Background
It is unknown whether lymph node metastases (LNM) and their localization negatively affect clinical outcome in metastatic renal cell carcinoma (mRCC) patients.
Objective
To evaluate the clinicopathological features, survival outcome, and treatment response in mRCC patients with LNM versus those without LNM after treatment with targeted therapies (TT).
Design, setting, and participants
Patients (n = 2996) were first analyzed without consideration of lymph node (LN) localization or histologic subtype. Additional analyses (n = 1536) were performed in subgroups of patients with supradiaphragmatic (SPD) LNM, subdiaphragmatic (SBD) LNM, and patients with LNM in both locations (SPD+/SBD+) without histologic considerations, and then separately in clear cell RCC (ccRCC) and non–clear cell RCC (nccRCC) patients, respectively.
Outcome measurements and statistical analysis
The primary outcome was overall survival (OS) and the secondary outcome was progression-free survival (PFS).
Results and limitations
All patients with LNM had worse PFS (p = 0.001) and OS (p < 0.001) compared to those without LNM. Compared to patients without LNM (PFS 8.8 mo; OS 25.1 mo), any SBD LNM involvement was associated with worse PFS (SBD, 6.8 mo; p = 0.003; SPD+/SBD+, 5.5 mo; p < 0.001) and OS (SBD, 16.2 mo; p < 0.001; SPD+/SBD+, 11.5 mo; p < 0.001). Both SBD and SPD+/SBD+ LNM were retained as independent prognostic factors in multivariate analyses (MVA) for PFS (p = 0.006 and p = 0.022, respectively) and OS (both p < 0.001), while SPD LNM was not an independent risk factor. Likewise, in ccRCC, SBD LNM (19.8 mo) and SPD+/SBD+ LNM (12.85 mo) patients had the worst OS. SPD+/SBD+ LNM (p = 0.006) and SBD LNM (p = 0.028) were independent prognostic factors for OS in MVA, while SPD LNM was not significant (p = 0.301). The study is limited by its retrospective design and the lack of pathologic evaluation of LNM in all cases.
Conclusions
The metastatic spread of RCC to SBD lymph nodes is associated with poor prognosis in mRCC patients treated with TT.
Patient summary
The presence of lymph node metastases below the diaphragm is associated with shorter survival outcome when metastatic renal cell carcinoma (mRCC) patients are treated with targeted therapies. Clinical trials should evaluate whether surgical removal of regional lymph nodes at the time of nephrectomy may improve outcomes in high-risk RCC patients.
Renal cell carcinoma (RCC) is a heterogeneous disease consisting of several tumor types that have their own genetic, molecular, and clinical characteristics. Historical studies in the cytokine era by the University of California Los Angeles (UCLA) and the National Institutes of Health (NIH) described poor prognosis and worse treatment response to interleukin-2 (IL-2) in patients with metastatic involvement of the retroperitoneal lymph nodes [1] and [2]. Patients with and without retroperitoneal lymph node involvement had median overall survival (OS) times of 8.5 mo and 14.7 mo (p = 0.0004), respectively [2] . Pantuck et al [1] reported a better objective response rate in N0M1 compared to N+M1 patients (p = 0.01), with survival times of 10.5 mo and 20.4 mo, respectively. The adverse prognostic impact of retroperitoneal lymph node metastases (LNM) on survival outcome in patients with metastatic RCC (mRCC) was confirmed by several other institutions during the cytokine era [3], [4], and [5]. However, in the era of targeted therapies (TT), the survival outcome for patients with LNM has not yet been well characterized.
RCC preferentially metastasizes via hematogenous routes. Bianchi et al [6] recently described hematogenous metastatic sites in 80% of 11 157 patients with mRCC, while extension of the disease into lymph nodes was described only in 20% of patients. However, the authors did not provide details about the distribution of lymph node localization. The dominant regional LNM sites in RCC are retro- and paracaval, pre- and paraaortic, and interaortocaval lymph nodes in an anatomically intact retroperitoneum [7] . LNM can be unpredictable, and can also occur via direct extension to the thorax, supraclavicular lymph nodes, and iliac lymph nodes [7] . It has not been documented if particular sites of positive lymph node localization have a negative influence on the survival outcome of mRCC patients treated with TT.
Whether retroperitoneal lymph node dissection (LND) improves survival outcome is the subject of ongoing scientific debate. A phase 3 clinical trial (EORTC 30881) demonstrated positive lymph nodes in only 4% of investigated cases having no apparent involvement on computed tomography (CT) imaging. There was no advantage in OS, recurrence-free survival, or progression-free survival (PFS) for RCC patients who underwent LND [8] . The study has been criticized, however, because the majority of patients included had low-stage disease. Moreover, uniform surgical templates were not used for LND [9] . However, retrospective studies have demonstrated that LNM are more prevalent in high-risk patients with high Fuhrman grades, sarcomatoid features, locally advanced tumor stage, tumor size >10 cm, and tumor necrosis [10] . It has been suggested that LND is particularly beneficial in these high-risk patients [11] and [12]. Trends in the surgical management of RCC that have moved away from open radical surgery with extended LND to laparoscopic and robotic surgery with minimal hilar LND, combined with the negative results of phase 3 studies showing no additional benefit of LND over removal of the primary tumor alone, have decreased enthusiasm for performing extended LND. Therefore, it is common clinical practice to perform LND only in high-risk patients, and with a diagnostic rather than a curative intention [7] . Retrospective studies may be unable to add new insights into the current debate because only a prospective randomized trial with clear inclusion and exclusion criteria would be able to overcome this selection bias. Before planning such trials, however, it is prudent to better understand the impact of LNM and different lymph node localizations on the survival outcome for mRCC patients treated with current state-of-the-art therapies using data from the International mRCC Database Consortium (IMDC).
The IMDC database includes centers from North America (Canada, USA), Europe (Denmark, Greece, Belgium), Asia (Singapore, Japan, South Korea), and New Zealand. Data were collected from August 15, 2008 until December 31, 2013. At the time of analysis, the database contained data on 3405 patients who had received first-line targeted therapies between 2003 and 2013. The final study cohort comprised 2996 patients who were treated with first-line vascular endothelial growth factor (VEGF) inhibitors (n = 2823), mammalian target of rapamycin (mTOR) inhibitors (n = 165), or a combination of both (n = 8) therapies between April 2003 and August 2013. Patients were excluded from the analyses if they had received experimental therapies in first-line treatment (n = 7) or if no information was available on their lymph node status (n = 402; Fig. 1 ). Lymph node status was determined according to standard pathologic and CT criteria.
Information about LND was available for 890 patients. Full LND was performed in only 51 patients, and a limited (hilar) dissection was performed in 173 patients. These numbers were too low to allow any analyses and conclusions regarding the influence of LND on the survival outcome for mRCC patients treated with TT. Whether LND could be beneficial for mRCC patients was therefore not considered in the present study.
All centers obtained local Institutional Review Board approval before including data in the IMDC database. Baseline patient characteristics included demographic, clinicopathologic, and laboratory data, as previously described [13] . Survival data were retrospectively and prospectively collected from medical chart reviews and electronic records. Uniform data templates were used to ensure consistent data collection at each institution. The majority of patients were treated using standard care, but a subset received therapy as part of clinical trials. All patients were considered in consecutive series to avoid selection bias.
The primary hypothesis of the current study was that mRCC patients with LNM have worse survival outcome than patients without LNM when treated with TT. The secondary hypothesis was that worse survival outcome for mRCC patients with LNM is dictated by subdiaphragmatic (SPD) LNM rather than supradiaphragmatic (SBD) LNM.
Therefore, treatment response to current standard-of-care TT agents, OS, and PFS were compared between patients with LNM and those without LNM. We did not compare outcomes between the drugs used for treatment because these comparisons are frequently biased outside a clinical trial and there may not be sufficient power for such an analysis. Additional analyses were performed for subgroups of patients with exclusively SPD lymph node involvement, exclusively SBD lymph node involvement, or with lymph node involvement in both sites (SPD+/SBD+). Previous studies have demonstrated that tumor biology has an important impact on the survival outcome for mRCC patients [14] . Thus, the different biology of RCC was considered by separate analyses for clear cell RCC (ccRCC) and non-ccRCC (nccRCC) patients. Patients with SPD primarily had mediastinal lymph node involvement, while patients with SBD primarily had retroperitoneal lymph node involvement. Lymph node status was determined before initiation of TT. The selection process for patient groups analyzed in the study is shown in Figure 1 .
OS was defined as the time period between TT initiation and the date of death, with censoring on the day of the last follow-up visit. PFS was defined as the period between treatment initiation and progression, drug cessation, or death, with censoring at the last follow-up visit. Kaplan-Meier plots were used to estimate median OS and PFS, and univariate comparisons were performed using the log rank test. Uni- and multivariate (MVA) Cox regression analyses were performed to test the association of predefined prognostic factors and LNM with survival outcome. The analyses were performed with backward stepwise selection criteria, and significance was tested using the Wald statistic [13] . In MVA analyses, the IMDC (Heng) risk criteria were applied for adjustment [15] . The IMDC risk criteria include a time period from diagnosis to treatment of <1 yr, Karnofsky performance score <80%, anemia, hypercalcemia, thrombophilia, and neutrophilia. Upper and lower limits of normal for the laboratory parameters were based on institutional limits. In addition, adjustment was performed for bone and liver metastases because interactions of LNM with other metastatic sites should be considered. The prognostic relevance of bone and liver metastases has previously been demonstrated by our group [16] .
A secondary aim was to analyze baseline characteristics. Patient and tumor characteristics were compared using the Student t test for continuous variables, or the chi-square test and the Fisher exact test for categorical variables.
Statistical analyses were performed using SPSS version 22 (IBM, Chicago, IL, USA), and a two-sided value of p < 0.05 was considered statistically significant.
At the time of data analysis, 2525 (84.3%) patients had stopped taking their TT; the median time on TT was 6.0 mo (25th percentile, 2.7 mo; 75th percentile, 13.0 mo). Best overall response data to first-line therapy were available for 2559 patients. There were no significant differences in response to first-line TT (p = 0.059) when patients were compared without consideration of RCC subtype or LNM localization ( Table 1 ).
Table 1 Treatment response to targeted therapies in patients with lymph node metastases (LNM) compared to those without LNM
| Best response | |||||
|---|---|---|---|---|---|
| CR | PR | SD | PD | ||
| All RCC subtypes | |||||
| No LMN | n | 14 | 304 | 556 | 290 |
| % | 1.2 | 26.1 | 47.8 | 24.9 | |
| LMN | n | 9 | 314 | 690 | 382 |
| % | 0.6 | 22.5 | 49.5 | 27.4 | |
| Clear cell RCC | |||||
| No LNM | n | 3 | 108 | 261 | 95 |
| % | 0.6 | 23.1 | 55.9 | 20.3 | |
| SPD LNM | n | 2 | 73 | 149 | 55 |
| % | 0.7 | 26.2 | 53.4 | 19.7 | |
| SBD LNM | n | 2 | 50 | 92 | 42 |
| % | 1.1 | 26.9 | 49.5 | 22.6 | |
| SPD+/SBD+ LNM | n | 0 | 28 | 58 | 36 |
| % | 0.0 | 23.0 | 47.5 | 29.5 | |
| Non–clear cell RCC | |||||
| No LNM | n | 0 | 3 | 21 | 21 |
| % | 0.0 | 6.7 | 46.7 | 46.7 | |
| SPD LNM | n | 0 | 3 | 12 | 3 |
| % | 0.0 | 16.7 | 66.7 | 16.7 | |
| SBD LNM | n | 0 | 7 | 24 | 19 |
| % | 0.0 | 14.0 | 48.0 | 38.0 | |
| SPD+/SBD+ LNM | n | 1 | 4 | 10 | 6 |
| % | 4.8 | 19.0 | 47.6 | 28.6 | |
CR = complete remission; PR = partial response; SD = stable disease; PD = progressive disease; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm.
Both response data and LNM localization data were available for 1054 ccRCC patients and 134 nccRCC patients. Comparison of data by LNM localization revealed no difference in best response to first-line therapies for either ccRCC (p = 0.457) or nccRCC (p = 0.192; Table 1 ).
Patients with LNM generally had worse PFS compared to those with no LNM without considering LNM localization or RCC subtype (median PFS [95% confidence interval] 7.8 mo [7.1–8.5 mo] vs 6.1 [5.6–6.6 mo], log rank p = 0.001; Fig. 2 A).
Fig. 2 (A) Comparison of progression-free survival (PFS) in patients with positive lymph nodes versus patients without positive lymph nodes (median 7.8 mo vs. 6.1 mo) without consideration of histologic subtype. Patients with lymph node metastases (LNM) had worse PFS (hazard ratio [HR] 1.14, 95% confidence interval [CI] 1.06–1.24; p = 0.001). (B) Comparison of PFS according to lymph node localization. Patients without LNM had the best PFS (8.8 mo), and patients with both supradiaphragmatic (SPD) and subdiaphragmatic (SBD) LNM localizations (SPD+/SBD+) had the worst PFS (5.5 mo). Patients with SBD LNM had significantly worse PFS compared to patients with no LNM (6.8 vs 8.8 mo; p = 0.003) and those with SPD LNM (6.8 vs 8.3 mo; p = 0.022). Any metastatic involvement of retroperitoneal lymph nodes, either alone (HR 1.25, 95% CI 1.08–1.45) or in combination with SPD+/SBD+LNM (HR 1.50, 95% CI 1.24–1.74; p < 0.001) was associated with higher risk of progression. By contrast, SPD LNM alone was not a prognostic factor for progression (HR 1.03, 95% 0.895–1.18; p = 0.695). (C) PFS by lymph node localization in clear cell renal cell carcinoma (ccRCC; median PFS: no nodes, 9.9 mo; SPD LNM, 8.7 mo; SBD LNM, 8.0 mo; and both, 5.6 mo). Patients with SPD LNM (HR 1.09, 95% CI 0.94–1.25) and SBD LNM (HR 1.04, 95% CI 0.96–1.14) had similar PFS, and this was not significantly worse than the PFS of patients without LNM (p = 0.265 and p = 0.333 for SPD and SBD LNM, respectively).
Comparison of PFS rates according to LNM sites revealed that patients with SBD LNM had significantly shorter PFS (6.8 mo [5.7–8.0 mo]) compared to those without LNM (8.8 mo [7.6–10.0 mo]; p = 0.003) or SPD LNM (8.3 mo [7.2–9.6 mo]; p = 0.022). The worst PFS was observed in patients with SPD+/SBD+ LNM (5.4 mo [4.6–6.4 mo]), but this did not significantly differ from PFS in patients with SBD LNM (p = 0.093; Fig. 2 B).
In ccRCC, however, comparison according to LNM localization revealed no significant differences in PFS between patients without LNM (9.9 mo [8.6–11.1 mo]) and those with SPD LNM (8.7 mo [7.3–10.1 mo]) or SBD LNM (8.0 mo [6.9–9.2 mo]) (p = 0.312; Fig. 2 C). SPD+/SBD+ LNM patients had the worst survival outcome (5.6 mo [4.0–7.2 mo]) in comparison to patients without LNM (p < 0.001), with SPD LNM (p = 0.006), and with SBD LNM (p = 0.035; Fig. 2 C). In nccRCC, there was no difference in PFS between patients with no LNM (3.9 mo [2.9–5.0 mo]), SPD LNM (4.6 mo [0.0–9.9 mo]), SBD LNM (5.4 mo [4.7–6.0 mo]), and SPD+/SBD+ LNM (5.4 mo [3.2–7.3 mo]).
In MVA analyses, LNM remained an independent prognostic factor (HR 1.13 [1.03–1.24]; p = 0.013) when adjusted for IMDC risk criteria, as well as liver and bone metastases ( Table 2 ) [16] . In subgroup analysis, SBD LNM (HR 1.31 [1.08–1.60]; p = 0.006) and SPD+/SBD LNM (HR 1.29 [1.04–1.61]; p = 0.022) were independent prognostic factors for PFS, while SPD LNM was not independently associated with PFS (p = 1.00; Table 2 ). In ccRCC, only SPD+/SBD+ LNM (p = 0.025) was independently associated with PFS ( Table 2 ).
Table 2 Multivariate analyses for association between lymph node metastases (LNM) and progression-free survival (PFS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time <1 yr | 1.33 | 1.21–1.47 | <0.001 |
| Low hemoglobin | 1.37 | 1.24–1.52 | <0.001 |
| Neutrophilia | 1.43 | 1.25–1.63 | <0.001 |
| Thrombocytosis | 1.16 | 1.03–1.32 | 0.019 |
| Karnofsy PS <80% | 1.68 | 1.50–1.88 | <0.001 |
| LNM | 1.13 | 1.03–1.24 | 0.013 |
| Liver metastases | 1.26 | 1.12–1.41 | <0.001 |
| Bone metastases | 1.27 | 1.15–1.40 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.19–1.61 | <0.001 |
| Low hemoglobin | 1.42 | 1.21–1.67 | <0.001 |
| Neutrophilia | 1.46 | 1.19–1.80 | <0.001 |
| Thrombocytosis | 1.22 | 1.01–1.47 | 0.045 |
| Karnofsy PS <80% | 1.94 | 1.65–2.28 | <0.001 |
| Liver metastases | 1.39 | 1.18–1.65 | <0.001 |
| Bone metastases | 1.45 | 1.25–1.68 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.83–1.20 | 0.990 |
| SBD LNM | 1.32 | 1.08–1.60 | 0.006 |
| SPD+/SBD+ LNM | 1.29 | 1.04–1.61 | 0.022 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.17–1.64 | <0.001 |
| Low hemoglobin | 1.43 | 1.21–1.69 | <0.001 |
| Neutrophilia | 1.38 | 1.09–1.76 | 0.009 |
| Thrombocytosis | 1.97 | 1.64–2.38 | <0.001 |
| Karnofsy PS <80% | 1.51 | 1.24–1.85 | <0.001 |
| Bone metastases | 1.64 | 1.38–1.94 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.82–1.22 | 0.998 |
| SBD LNM | 1.23 | 0.98–1.55 | 0.077 |
| SPD+/SBD+ LNM | 1.33 | 1.04–1.71 | 0.025 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm; ccRCC = clear cell RCC.
At the time of the analysis, 1960 patients had died; the median OS time for the total patient cohort was 20.0 mo (25th percentile, 42.3 mo; 75th percentile, 7.9 mo). Patients with metastatic spread to lymph nodes had a worse survival outcome than patients without LNM (24.0 mo [21.6–26.3 mo] vs 16.0 mo [14.9–17.1 mo], p < 0.001; Fig. 3 A).
Fig. 3 (A) Patients with lymph node metastases (LNM) had a higher risk of death (median overall survival [OS] 23.9 mo, 95% confidence interval [CI] 21.6–26.3 mo) than those without LNM (median OS 16.0 mo, 95% CI 14.9–17.1; hazard ratio [HR] 1.41, 95% CI 1.28–1.54; p < 0.001). (B) A comparison without consideration of the histologic subtypes demonstrates that any lymph node involvement was associated with worse survival outcome compared to patients without LNM. Median survival times were 25.2 mo, 20.3 mo, 16.2 mo, and 11.5 mo in patients with no LNM and lymph node involvement in supradiaphragmatic (SPD), subdiaphragmatic (SBD), and both localizations (SPD+/SBD+), respectively. The HRs (95% CIs) for patients with SPD, SBD, and SPD+/SBD+ LNM were 1.24 (1.06–1.44), 1.45 (1.23–1.71), and 1.93 (1.60–2.33), respectively. (C) In clear cell renal cell carcinoma (ccRCC), patients with SPD and SBD LNM had similar median OS (SPD, 21.6 mo; SBD, 19.8 mo). However, both groups had a significantly higher risk of death (SPD: HR 1.24, 95% CI 1.05–1.47; SBD: HR 1.28, 95% CI 1.05–1.56) in comparison to those without lymph nodes. Patients with SPD+/SBD+ metastases had the worst OS (12.85 mo) and a twofold higher risk of death than patients without lymph node involvement (HR 2.04, 95% CI 1.68–2.48). (D) In non–clear cell RCC (nccRCC), median OS times according to lymph node localization were 15.8 mo for no lymph nodes, 16.1 mo for SPD LNM, 11.2 mo for SBD LNM, and 12.5 mo for SPD+/SBD+ LNM. There were no significant differences between the groups (log rank p = 0.444).
In subgroup analyses, patients without LNM had a median OS of 25.2 mo (21.9–28.5 mo) compared to 20.3 mo (17.1–23.5 mo) for SPD LNM (p = 0.093), 16.2 mo (13.5–20.0 mo) for SBD LNM (p < 0.001), and 11.5 mo (9.8–13.3 mo) (p < 0.001) for SPD+/SBD+ LNM ( Fig. 3 B).
In ccRCC, further analysis considering LNM localization revealed similar OS times for patients with SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13]), and were shorter than for patients without LNM (26.97 mo [23.32– 30.62], p = 0.01; Fig. 3 C). Patients with SPD+/SBD+ LNM had significantly (p = 0.001) worse OS (12.85 mo [9.45–16.24 mo]) compared to SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13 mo]; Fig. 3 C). In nccRCC, analyses according to LNM revealed no significant differences in OS (p = 0.444; Fig. 3 C).
For all histological subtypes, LNM were an independent prognostic factor for adverse OS in MVA analysis (p < 0.001; Table 3 ). Additional MVA according to LNM sites revealed that SBD LNM (p < 0.001) and SPD+/SBD+ LNM (p < 0.001) were independent prognostic factors for OS. SPD LNM were not independently associated with OS (p = 0.999; Table 3 ). In ccRCC, MVA according to LNM localization confirmed an independent association between SBD LNM (p = 0.028) and SPD+/SBD+ LNM (p = 0.006) and OS ( Table 3 ).
Table 3 Multivariate Cox regression analysis of association between lymph node metastases (LNM) and overall survival (OS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time < 1 yr | 1.34 | 1.19–1.49 | <0.001 |
| Low hemoglobin | 1.51 | 1.31–1.70 | <0.001 |
| High Ca | 1.47 | 1.25–1.73 | <0.001 |
| Neutrophilia | 1.67 | 1.45–1.92 | <0.001 |
| Thrombophilia | 1.31 | 1.14–1.51 | <0.001 |
| Karnofsky PS <80% | 1.88 | 1.67–2.13 | <0.001 |
| Liver metastases | 1.50 | 1.32–1.70 | <0.001 |
| Bone metastases | 1.50 | 1.34–1.67 | <0.001 |
| LNM | 1.33 | 1.19–1.48 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time < 1 yr | 1.32 | 1.11–1.56 | 0.001 |
| Low hemoglobin | 1.45 | 1.21–1.73 | <0.001 |
| High Ca | 1.38 | 1.08–1.77 | 0.010 |
| Neutrophilia | 1.78 | 1.43–2.22 | <0.001 |
| Thrombophilia | 1.49 | 1.21–1.83 | <0.001 |
| Karnofsky PS <80% | 2.19 | 1.82–2.62 | <0.001 |
| Liver metastases | 1.84 | 1.53–2.20 | <0.001 |
| Bone metastases | 1.61 | 1.36–1.90 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.57 | 0.98–1.48 | 0.075 |
| SBD LNM | 1.59 | 1.26–1.96 | <0.001 |
| SPD+/SBD+ LNM | 1.32 | 1.25–2.02 | <0.001 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time < 1 yr | 1.36 | 1.12–1.64 | 0.002 |
| Low hemoglobin | 1.36 | 1.11–1.67 | 0.003 |
| High Ca | 1.44 | 1.10–1.89 | 0.009 |
| Neutrophilia | 1.65 | 1.27–2.13 | <0.001 |
| Thrombophilia | 1.52 | 1.19–1.91 | 0.001 |
| Karnofsky PS <80% | 2.12 | 1.72–2.61 | <0.001 |
| Liver metastases | 1.94 | 1.57–2.41 | <0.001 |
| Bone metastases | 1.84 | 1.53–2.22 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.13 | 0.90–1.42 | 0.301 |
| SBD LNM | 1.34 | 1.03–1.74 | 0.028 |
| SPD+/SBD+ LNM | 1.47 | 1.11–1.93 | 0.006 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm; ccRCC = clear cell RCC.
An overall comparison of patients with and without LNM is shown in Table 4 . Patients with LNM presented more often with nccRCC (p < 0.001) and higher Fuhrman grades (p = 0.007), and had less often undergone prior nephrectomy (p < 0.001). The IMDC risk factors of high calcium (p = 0.001), thrombocytosis (p < 0.001), neutrophilia (p < 0.001), and Karnofsky performance status <80% (p = 0.009) were all more frequently seen in patients with LNM. Comparison of additional metastatic sites revealed that LNM are related to a higher overall metastatic burden (one or more additional metastatic site p < 0.001).
Table 4 General comparison of clinicopathologic features of patients with lymph node metastases (LNM) compared to those without LNM a
| No LNM (n = 1380) |
LNM (n = 1616) |
p value | |
|---|---|---|---|
| Mean age, yr (SD) | 58.65 (11.36) | 58. 50 (11.30) | 0.716 |
| Male, n (%) | 1003 (72.7) | 1182 (73.1) | 0.805 |
| Nephrectomy, n (%) | 1134 (82.4) | 1218 (75.4) | <0.001 |
| Fuhrman grade n (%) | 0.007 | ||
| 1 | 44 (4.2) | 33 (2.8) | |
| 2 | 302 (28.6) | 269 (23.1) | |
| 3 | 426 (40.3) | 501 (43.1) | |
| 4 | 284 (26.9) | 359 (30.9) | |
| Non–clear cell RCC, n (%) | 127 (9.8) | 253 (17.4) | <0.001 |
| Sarcomatoid features, n (%) | 143 (12.0) | 170 (12.0) | 0.586 |
| Therapy class, n (%) | |||
| AntiVEGF | 1316 (95.4) | 1507 (93.3) | |
| mTOR | 60 (4.3) | 105 (6.5) | |
| Combination | 4 (0.3) | 4 (0.2) | |
| DTT <1 yr, n (%) | 736 (53.4) | 905 (56.0) | 0.151 |
| Low HB (<LLN), n (%) | 691 (56.0) | 877 (58.1) | 0.261 |
| High calcium, n (%) | 98 (8.6) | 179 (12.6) | 0.001 |
| Neutrophilia, n (%) | 154 (13.1) | 286 (19.3) | <0.001 |
| Thrombocytosis, n (%) | 172 (15.4) | 299 (22.4) | <0.001 |
| Karnofsky PS <80%, n (%) | 261 (21.5) | 372 (25.9) | 0.009 |
| Lung metastases, n (%) | 939 (68.0) | 1149 (72.5) | 0.016 |
| Brain metastases, n (%) | 100 (7.2) | 141 (9.4) | 0.009 |
| Liver metastases, n (%) | 278 (20.2) | 328 (21.7) | 0.337 |
| Bone metastases, n (%) | 481 (34.9) | 551 (35.9) | 0.462 |
| Other metastic sites, n (%) | 470 (34.1) | 643 (43.1) | <0.001 |
| >1 metastatic site, n (%) | 805 (58.3%) | 1446 (89.5%) | <0.001 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; mTOR = mammalian target of rapamycin; PS = performance status; RCC = renal cell carcinoma; SD = standard deviation; VEGF = vascular endothelial growth factor.
The LNM localization distribution and clinicopathologic features in ccRCC and nccRCC are shown in Table 5 . In ccRCC, IMDC risk factors of neutrophilia (p = 0.049) and thrombophilia (p = 0.016) were more common. Patients with LNM had undergone nephrectomy less frequently (p = 0.001). In patients with ccRCC, those with LNM had concomitant lung (p < 0.001) and other metastases (eg, adrenal glands, other soft tissues; p = 0.009) more often than those without LNM; there was no difference in the frequency of bone (p = 0.276), liver (p = 0.943), or brain metastases (p = 0.187).
Table 5 Distribution of metastatic lymphatic sites in relation to the diaphragm in clear cell and non–clear cell renal carcinoma (RCC) a
| Clear cell RCC (n = 1218) |
Non–clear cell RCC (n = 159) |
|
|---|---|---|
| No LNM, n (%) | 552 (45.6) | 52 (32.7) |
| SPD LNM, n (%) | 321 (26.6) | 22 (13.8) |
| SBD LNM, n (%) | 204 (16.9) | 54 (34.0) |
| SPD+/SBD+ LNM, n (%) | 132 (10.9) | 31 (19.5) |
| LMN | No LMN | p value | LMN | No LMN | p value | |
|---|---|---|---|---|---|---|
| Nephrectomy, n (%) | 542 (81.2) | 492 (88.2) | 0.001 | 82 (76.6) | 43 (82.7) | 0.418 |
| Male, n (%) | 515 (77.7) | 403 (72.6) | 0.045 | 75 (70.1) | 37 (71.2) | 1.00 |
| DTT <1 yr, n (%) | 348 (52.5) | 265 (47.9) | 0.120 | 65 (60.7) | 29 (55.8) | 0.607 |
| Low HB (<LLN) | 331 (54.0) | 233 (48.7) | 0.088 | 49 (51.0) | 19 (45.2) | 0.582 |
| High calcium, n (%) | 70 (12.3) | 40 (9.5) | 0.184 | 5 (5.7) | 4 (12.1) | 0.254 |
| Neutrophilia, n (%) | 95 (15.9) | 53 (11.6) | 0.049 | 24 (25.3) | 7 (16.7) | 0.376 |
| Thrombophilia, n (%) | 106 (22.0) | 63 (15.6) | 0.016 | 17 (21.0) | 10 (25.6) | 0.642 |
| Karnofsky PS <80%, n (%) | 145 (24.7) | 95 (20.0) | 0.076 | 24 (24.7) | 10 (20.4) | 0.680 |
| Sarcomatoid features, n (%) | 52 (9.1) | 62 (12.6) | 0.074 | 14 (15.4) | 7 (15.6) | 1.00 |
| Lung metastases, n (%) | 507 (76.8) | 372 (67.0) | <0.001 | 52 (49.1) | 20 (38.5) | 0.237 |
| Brain metastases, n (%) | 65 (9.8) | 42 (7.6) | 0.187 | 4 (3.8) | 2 (3.8) | 1.00 |
| Liver metastases, n (%) | 132 (20.0) | 112 (20.2) | 0.943 | 32 (30.2) | 28 (53.8) | 0.005 |
| Bone metastases, n (%) | 220 (33.3) | 202 (36.4) | 0.276 | 35 (33.0) | 20 (38.5) | 0.594 |
| Other metastatic sites, n (%) | 287 (45.7) | 211 (38.2) | 0.009 | 40 (40.8) | 20 (38.5) | 0.862 |
| >1 metastatic site, n (%) | 602 (90.8) | 320 (57.7) | <0.001 | 87 (81.3) | 30 (7.7) | 0.002 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; PS = performance status; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm.
In nccRCC, liver metastases were found more often in patients without LNM (p = 0.005). A comparison of other clinicopathologic features between nccRCC patients with and without LNM revealed no significant differences ( Table 5 ).
It has previously been shown that only a minority of patients with kidney cancer undergo LND at the time of nephrectomy. This is the case even for high-risk patients [17] , even though retrospective studies have suggested a therapeutic benefit of LND [11] and [18]. It remains uncertain whether extended LND can improve the survival outcome for mRCC patients. However, if it were shown that SBD LNM affect the survival outcome for mRCC patients, this could form the basis for a rationale in favor of performing LND, or at least justify the need for prospective clinical trials to address this question. In this regard, an understanding of the influence of LNM on treatment response and survival outcome for mRCC patients undergoing TT could make an important contribution to the debate on surgical management of RCC.
Our study has demonstrated that LNM are associated with advanced IMDC risk factors, are an adverse prognostic factor for PFS and OS, and are associated with a higher metastatic burden. Furthermore, subanalyses showed that these differences depend on LNM localization.
Different cellular clones of RCC appear to differ in their propensity to spread to different organ sites [19] ; however, whether these differences impact on the survival outcome of mRCC patients is not completely understood. A previous IMDC study highlighted the prognostic importance of bone and liver metastases for survival outcome in mRCC patients, and suggested that aggressive RCC subclones tend to metastasize to these organs [16] . In addition, another study by our group showed that metastatic spread to other organs such as the pancreas or soft tissue tends to develop as a late clinical symptom, and has favorable treatment response and survival outcomes [20] . While these tumors may represent a favorable RCC subtype, our current results suggest that RCCs that spread to regional retroperitoneal lymph nodes represent another aggressive RCC subtype that develops bone metastases.
It has been suggested that LNM in ccRCC occur independently of von Hippel Lindau (VHL) gene inactivation [21] . Current agents used for mRCC treatment modulate pathways that are dysregulated because of VHL inactivation. This would imply that LNM might be less amenable to treatment with VEGF and mTOR inhibitors. However, our findings demonstrated no significant differences in best response to first-line therapies in patients with LNM versus those without LNM in both ccRCC and nccRCC patient groups. Nonetheless, patients with LNM had significantly shorter PFS than patients without LNM. While there was only a trend for an independent association between SBD LNM and short PFS, SPD+/SBD+ LNM were independently related to shorter PFS. Patients with SPD+/SBD+ LNM had additional hematogenous metastatic sites in 97% of cases. Thus, the high metastatic burden may explain the shorter response duration of first-line agents in this group of patients. However, a shorter duration for treatment success could also support the hypothesis that LNM are caused by RCC clones that do not have an inactive VHL gene function.
In the cytokine era, LNM were associated with shorter OS [1] and [2]. Our study demonstrates similar findings in the TT era. Interestingly, a worse survival outcome was observed for SBD LNM, both when occurring exclusively and in combination with thoracic lymph node involvement. Moreover, SBD LNM remained an independent prognostic factor for PFS and OS when analyzed without consideration of histologic subtype, and were an independent prognostic factor for OS in ccRCC. Conversely, SPD LNM were not an independent prognostic factor. This is notable and underscores the unique nature of RCCs that spread via retroperitoneal lymph nodes. Because the best response to TT did not different between SBD and SPD LNM, it appears that survival outcome is influenced by numerous factors. Patients with SBD LNM had faster disease progression before treatment initiation than those with SPD LNM (diagnosis to treatment <1 yr; 63% vs 46%; p < 0.001; data not shown). Despite a best response that was broadly similar, it seems that the disease progresses faster in SBD LNM patients during the later clinical course than in SPD LNM patients. In addition, SBD LNM were more often associated with anemia compared to SPD LNM (59% vs 48%; p = 0.008; data not shown). Anemia has multifactorial causes in cancer patients. However, one of the most important contributing factors is immune system dysregulation [22] . Dysregulation of the immune system may also be the reason for metastatic spread to SBD sites. Collectively, while risk factors that are currently unknown appear to dictate the clinical course in patients with SPD metastases, SBD metastases are independently associated with PFS and OS.
The majority of patients analyzed in this study did not undergo formal LND. Therefore, the effect of LND cannot be determined retrospectively. However, the current results indicate that SBD lymph nodes may have substantial negative effects on survival outcome and it is worth speculating whether extended removal of these metastatic sites could improve the clinical outcome for mRCC patients. The IMDC has recently shown that cytoreductive nephrectomy may provide therapeutic benefit in a large number of patients [23] . With regard to our current findings, it would be worth determining the role of LND in these patients.
The current study has several limitations that should be considered when interpreting the results. The IMDC database comprises retrospectively collected data for consecutive series of mRCC patients. These individual series, moreover, do not have standardized follow-up protocols, but this limitation may more accurately reflect real world outcomes. In addition, there was no centralized pathologic or radiologic review. Finally, lymph node status was determined clinically according to CT images in a number of cases, and it has been demonstrated that clinical and pathologic lymph node status is not necessarily always correlated.
LNM are associated with poor prognosis and adverse prognostic IMDC risk factors in mRCC patients treated with TT. Much of this increased risk associated with LNM appears to be dictated by SBD localization rather than SPD localization. Whether LND could improve the outcome for this group of patients remains unknown, but it is an important clinical question requiring prospective trials.
Author contributions: Daniel Y. Heng had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kroeger, Heng.
Acquisition of data: all authors.
Analysis and interpretation of data: Kroeger, Heng, Choueiri.
Drafting of the manuscript: Kroeger, Heng, Choueiri.
Critical revision of the manuscript for important intellectual content: all authors.
Statistical analysis: Kroeger, Heng.
Obtaining funding: None.
Administrative, technical, or material support: all authors.
Supervision: Choueiri, Heng.
Other (specify): None.
Financial disclosures: Daniel Y. Heng certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Benoit Beuselinck is an investigator of the EudraCT: 2011-006085-40/MetaSun trial supported by Pfizer. Frede Donskov has received research support from Novartis and GlaxoSmithKline. Toni K. Choueiri has received research funding from Pfizer and has an advisory role at Pfizer, GSK, Novartis, and Bayer. All other authors declare no conflicts of interest to the current study.
Funding/Support and role of the sponsor: None.
Acknowledgments: Dr. Choueiri is supported in part by the Trust family and the Loker Pinard and Michael Brigham Funds for Kidney Cancer Research at the Dana-Farber Cancer Institute. Dr. Kroeger thanks Dr. Klatte, Department of Urology, Medical University of Vienna, for statistical advice.
Renal cell carcinoma (RCC) is a heterogeneous disease consisting of several tumor types that have their own genetic, molecular, and clinical characteristics. Historical studies in the cytokine era by the University of California Los Angeles (UCLA) and the National Institutes of Health (NIH) described poor prognosis and worse treatment response to interleukin-2 (IL-2) in patients with metastatic involvement of the retroperitoneal lymph nodes [1] and [2]. Patients with and without retroperitoneal lymph node involvement had median overall survival (OS) times of 8.5 mo and 14.7 mo (p = 0.0004), respectively [2] . Pantuck et al [1] reported a better objective response rate in N0M1 compared to N+M1 patients (p = 0.01), with survival times of 10.5 mo and 20.4 mo, respectively. The adverse prognostic impact of retroperitoneal lymph node metastases (LNM) on survival outcome in patients with metastatic RCC (mRCC) was confirmed by several other institutions during the cytokine era [3], [4], and [5]. However, in the era of targeted therapies (TT), the survival outcome for patients with LNM has not yet been well characterized.
RCC preferentially metastasizes via hematogenous routes. Bianchi et al [6] recently described hematogenous metastatic sites in 80% of 11 157 patients with mRCC, while extension of the disease into lymph nodes was described only in 20% of patients. However, the authors did not provide details about the distribution of lymph node localization. The dominant regional LNM sites in RCC are retro- and paracaval, pre- and paraaortic, and interaortocaval lymph nodes in an anatomically intact retroperitoneum [7] . LNM can be unpredictable, and can also occur via direct extension to the thorax, supraclavicular lymph nodes, and iliac lymph nodes [7] . It has not been documented if particular sites of positive lymph node localization have a negative influence on the survival outcome of mRCC patients treated with TT.
Whether retroperitoneal lymph node dissection (LND) improves survival outcome is the subject of ongoing scientific debate. A phase 3 clinical trial (EORTC 30881) demonstrated positive lymph nodes in only 4% of investigated cases having no apparent involvement on computed tomography (CT) imaging. There was no advantage in OS, recurrence-free survival, or progression-free survival (PFS) for RCC patients who underwent LND [8] . The study has been criticized, however, because the majority of patients included had low-stage disease. Moreover, uniform surgical templates were not used for LND [9] . However, retrospective studies have demonstrated that LNM are more prevalent in high-risk patients with high Fuhrman grades, sarcomatoid features, locally advanced tumor stage, tumor size >10 cm, and tumor necrosis [10] . It has been suggested that LND is particularly beneficial in these high-risk patients [11] and [12]. Trends in the surgical management of RCC that have moved away from open radical surgery with extended LND to laparoscopic and robotic surgery with minimal hilar LND, combined with the negative results of phase 3 studies showing no additional benefit of LND over removal of the primary tumor alone, have decreased enthusiasm for performing extended LND. Therefore, it is common clinical practice to perform LND only in high-risk patients, and with a diagnostic rather than a curative intention [7] . Retrospective studies may be unable to add new insights into the current debate because only a prospective randomized trial with clear inclusion and exclusion criteria would be able to overcome this selection bias. Before planning such trials, however, it is prudent to better understand the impact of LNM and different lymph node localizations on the survival outcome for mRCC patients treated with current state-of-the-art therapies using data from the International mRCC Database Consortium (IMDC).
The IMDC database includes centers from North America (Canada, USA), Europe (Denmark, Greece, Belgium), Asia (Singapore, Japan, South Korea), and New Zealand. Data were collected from August 15, 2008 until December 31, 2013. At the time of analysis, the database contained data on 3405 patients who had received first-line targeted therapies between 2003 and 2013. The final study cohort comprised 2996 patients who were treated with first-line vascular endothelial growth factor (VEGF) inhibitors (n = 2823), mammalian target of rapamycin (mTOR) inhibitors (n = 165), or a combination of both (n = 8) therapies between April 2003 and August 2013. Patients were excluded from the analyses if they had received experimental therapies in first-line treatment (n = 7) or if no information was available on their lymph node status (n = 402; Fig. 1 ). Lymph node status was determined according to standard pathologic and CT criteria.
Information about LND was available for 890 patients. Full LND was performed in only 51 patients, and a limited (hilar) dissection was performed in 173 patients. These numbers were too low to allow any analyses and conclusions regarding the influence of LND on the survival outcome for mRCC patients treated with TT. Whether LND could be beneficial for mRCC patients was therefore not considered in the present study.
All centers obtained local Institutional Review Board approval before including data in the IMDC database. Baseline patient characteristics included demographic, clinicopathologic, and laboratory data, as previously described [13] . Survival data were retrospectively and prospectively collected from medical chart reviews and electronic records. Uniform data templates were used to ensure consistent data collection at each institution. The majority of patients were treated using standard care, but a subset received therapy as part of clinical trials. All patients were considered in consecutive series to avoid selection bias.
The primary hypothesis of the current study was that mRCC patients with LNM have worse survival outcome than patients without LNM when treated with TT. The secondary hypothesis was that worse survival outcome for mRCC patients with LNM is dictated by subdiaphragmatic (SPD) LNM rather than supradiaphragmatic (SBD) LNM.
Therefore, treatment response to current standard-of-care TT agents, OS, and PFS were compared between patients with LNM and those without LNM. We did not compare outcomes between the drugs used for treatment because these comparisons are frequently biased outside a clinical trial and there may not be sufficient power for such an analysis. Additional analyses were performed for subgroups of patients with exclusively SPD lymph node involvement, exclusively SBD lymph node involvement, or with lymph node involvement in both sites (SPD+/SBD+). Previous studies have demonstrated that tumor biology has an important impact on the survival outcome for mRCC patients [14] . Thus, the different biology of RCC was considered by separate analyses for clear cell RCC (ccRCC) and non-ccRCC (nccRCC) patients. Patients with SPD primarily had mediastinal lymph node involvement, while patients with SBD primarily had retroperitoneal lymph node involvement. Lymph node status was determined before initiation of TT. The selection process for patient groups analyzed in the study is shown in Figure 1 .
OS was defined as the time period between TT initiation and the date of death, with censoring on the day of the last follow-up visit. PFS was defined as the period between treatment initiation and progression, drug cessation, or death, with censoring at the last follow-up visit. Kaplan-Meier plots were used to estimate median OS and PFS, and univariate comparisons were performed using the log rank test. Uni- and multivariate (MVA) Cox regression analyses were performed to test the association of predefined prognostic factors and LNM with survival outcome. The analyses were performed with backward stepwise selection criteria, and significance was tested using the Wald statistic [13] . In MVA analyses, the IMDC (Heng) risk criteria were applied for adjustment [15] . The IMDC risk criteria include a time period from diagnosis to treatment of <1 yr, Karnofsky performance score <80%, anemia, hypercalcemia, thrombophilia, and neutrophilia. Upper and lower limits of normal for the laboratory parameters were based on institutional limits. In addition, adjustment was performed for bone and liver metastases because interactions of LNM with other metastatic sites should be considered. The prognostic relevance of bone and liver metastases has previously been demonstrated by our group [16] .
A secondary aim was to analyze baseline characteristics. Patient and tumor characteristics were compared using the Student t test for continuous variables, or the chi-square test and the Fisher exact test for categorical variables.
Statistical analyses were performed using SPSS version 22 (IBM, Chicago, IL, USA), and a two-sided value of p < 0.05 was considered statistically significant.
At the time of data analysis, 2525 (84.3%) patients had stopped taking their TT; the median time on TT was 6.0 mo (25th percentile, 2.7 mo; 75th percentile, 13.0 mo). Best overall response data to first-line therapy were available for 2559 patients. There were no significant differences in response to first-line TT (p = 0.059) when patients were compared without consideration of RCC subtype or LNM localization ( Table 1 ).
Table 1 Treatment response to targeted therapies in patients with lymph node metastases (LNM) compared to those without LNM
| Best response | |||||
|---|---|---|---|---|---|
| CR | PR | SD | PD | ||
| All RCC subtypes | |||||
| No LMN | n | 14 | 304 | 556 | 290 |
| % | 1.2 | 26.1 | 47.8 | 24.9 | |
| LMN | n | 9 | 314 | 690 | 382 |
| % | 0.6 | 22.5 | 49.5 | 27.4 | |
| Clear cell RCC | |||||
| No LNM | n | 3 | 108 | 261 | 95 |
| % | 0.6 | 23.1 | 55.9 | 20.3 | |
| SPD LNM | n | 2 | 73 | 149 | 55 |
| % | 0.7 | 26.2 | 53.4 | 19.7 | |
| SBD LNM | n | 2 | 50 | 92 | 42 |
| % | 1.1 | 26.9 | 49.5 | 22.6 | |
| SPD+/SBD+ LNM | n | 0 | 28 | 58 | 36 |
| % | 0.0 | 23.0 | 47.5 | 29.5 | |
| Non–clear cell RCC | |||||
| No LNM | n | 0 | 3 | 21 | 21 |
| % | 0.0 | 6.7 | 46.7 | 46.7 | |
| SPD LNM | n | 0 | 3 | 12 | 3 |
| % | 0.0 | 16.7 | 66.7 | 16.7 | |
| SBD LNM | n | 0 | 7 | 24 | 19 |
| % | 0.0 | 14.0 | 48.0 | 38.0 | |
| SPD+/SBD+ LNM | n | 1 | 4 | 10 | 6 |
| % | 4.8 | 19.0 | 47.6 | 28.6 | |
CR = complete remission; PR = partial response; SD = stable disease; PD = progressive disease; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm.
Both response data and LNM localization data were available for 1054 ccRCC patients and 134 nccRCC patients. Comparison of data by LNM localization revealed no difference in best response to first-line therapies for either ccRCC (p = 0.457) or nccRCC (p = 0.192; Table 1 ).
Patients with LNM generally had worse PFS compared to those with no LNM without considering LNM localization or RCC subtype (median PFS [95% confidence interval] 7.8 mo [7.1–8.5 mo] vs 6.1 [5.6–6.6 mo], log rank p = 0.001; Fig. 2 A).
Fig. 2 (A) Comparison of progression-free survival (PFS) in patients with positive lymph nodes versus patients without positive lymph nodes (median 7.8 mo vs. 6.1 mo) without consideration of histologic subtype. Patients with lymph node metastases (LNM) had worse PFS (hazard ratio [HR] 1.14, 95% confidence interval [CI] 1.06–1.24; p = 0.001). (B) Comparison of PFS according to lymph node localization. Patients without LNM had the best PFS (8.8 mo), and patients with both supradiaphragmatic (SPD) and subdiaphragmatic (SBD) LNM localizations (SPD+/SBD+) had the worst PFS (5.5 mo). Patients with SBD LNM had significantly worse PFS compared to patients with no LNM (6.8 vs 8.8 mo; p = 0.003) and those with SPD LNM (6.8 vs 8.3 mo; p = 0.022). Any metastatic involvement of retroperitoneal lymph nodes, either alone (HR 1.25, 95% CI 1.08–1.45) or in combination with SPD+/SBD+LNM (HR 1.50, 95% CI 1.24–1.74; p < 0.001) was associated with higher risk of progression. By contrast, SPD LNM alone was not a prognostic factor for progression (HR 1.03, 95% 0.895–1.18; p = 0.695). (C) PFS by lymph node localization in clear cell renal cell carcinoma (ccRCC; median PFS: no nodes, 9.9 mo; SPD LNM, 8.7 mo; SBD LNM, 8.0 mo; and both, 5.6 mo). Patients with SPD LNM (HR 1.09, 95% CI 0.94–1.25) and SBD LNM (HR 1.04, 95% CI 0.96–1.14) had similar PFS, and this was not significantly worse than the PFS of patients without LNM (p = 0.265 and p = 0.333 for SPD and SBD LNM, respectively).
Comparison of PFS rates according to LNM sites revealed that patients with SBD LNM had significantly shorter PFS (6.8 mo [5.7–8.0 mo]) compared to those without LNM (8.8 mo [7.6–10.0 mo]; p = 0.003) or SPD LNM (8.3 mo [7.2–9.6 mo]; p = 0.022). The worst PFS was observed in patients with SPD+/SBD+ LNM (5.4 mo [4.6–6.4 mo]), but this did not significantly differ from PFS in patients with SBD LNM (p = 0.093; Fig. 2 B).
In ccRCC, however, comparison according to LNM localization revealed no significant differences in PFS between patients without LNM (9.9 mo [8.6–11.1 mo]) and those with SPD LNM (8.7 mo [7.3–10.1 mo]) or SBD LNM (8.0 mo [6.9–9.2 mo]) (p = 0.312; Fig. 2 C). SPD+/SBD+ LNM patients had the worst survival outcome (5.6 mo [4.0–7.2 mo]) in comparison to patients without LNM (p < 0.001), with SPD LNM (p = 0.006), and with SBD LNM (p = 0.035; Fig. 2 C). In nccRCC, there was no difference in PFS between patients with no LNM (3.9 mo [2.9–5.0 mo]), SPD LNM (4.6 mo [0.0–9.9 mo]), SBD LNM (5.4 mo [4.7–6.0 mo]), and SPD+/SBD+ LNM (5.4 mo [3.2–7.3 mo]).
In MVA analyses, LNM remained an independent prognostic factor (HR 1.13 [1.03–1.24]; p = 0.013) when adjusted for IMDC risk criteria, as well as liver and bone metastases ( Table 2 ) [16] . In subgroup analysis, SBD LNM (HR 1.31 [1.08–1.60]; p = 0.006) and SPD+/SBD LNM (HR 1.29 [1.04–1.61]; p = 0.022) were independent prognostic factors for PFS, while SPD LNM was not independently associated with PFS (p = 1.00; Table 2 ). In ccRCC, only SPD+/SBD+ LNM (p = 0.025) was independently associated with PFS ( Table 2 ).
Table 2 Multivariate analyses for association between lymph node metastases (LNM) and progression-free survival (PFS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time <1 yr | 1.33 | 1.21–1.47 | <0.001 |
| Low hemoglobin | 1.37 | 1.24–1.52 | <0.001 |
| Neutrophilia | 1.43 | 1.25–1.63 | <0.001 |
| Thrombocytosis | 1.16 | 1.03–1.32 | 0.019 |
| Karnofsy PS <80% | 1.68 | 1.50–1.88 | <0.001 |
| LNM | 1.13 | 1.03–1.24 | 0.013 |
| Liver metastases | 1.26 | 1.12–1.41 | <0.001 |
| Bone metastases | 1.27 | 1.15–1.40 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.19–1.61 | <0.001 |
| Low hemoglobin | 1.42 | 1.21–1.67 | <0.001 |
| Neutrophilia | 1.46 | 1.19–1.80 | <0.001 |
| Thrombocytosis | 1.22 | 1.01–1.47 | 0.045 |
| Karnofsy PS <80% | 1.94 | 1.65–2.28 | <0.001 |
| Liver metastases | 1.39 | 1.18–1.65 | <0.001 |
| Bone metastases | 1.45 | 1.25–1.68 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.83–1.20 | 0.990 |
| SBD LNM | 1.32 | 1.08–1.60 | 0.006 |
| SPD+/SBD+ LNM | 1.29 | 1.04–1.61 | 0.022 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.17–1.64 | <0.001 |
| Low hemoglobin | 1.43 | 1.21–1.69 | <0.001 |
| Neutrophilia | 1.38 | 1.09–1.76 | 0.009 |
| Thrombocytosis | 1.97 | 1.64–2.38 | <0.001 |
| Karnofsy PS <80% | 1.51 | 1.24–1.85 | <0.001 |
| Bone metastases | 1.64 | 1.38–1.94 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.82–1.22 | 0.998 |
| SBD LNM | 1.23 | 0.98–1.55 | 0.077 |
| SPD+/SBD+ LNM | 1.33 | 1.04–1.71 | 0.025 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm; ccRCC = clear cell RCC.
At the time of the analysis, 1960 patients had died; the median OS time for the total patient cohort was 20.0 mo (25th percentile, 42.3 mo; 75th percentile, 7.9 mo). Patients with metastatic spread to lymph nodes had a worse survival outcome than patients without LNM (24.0 mo [21.6–26.3 mo] vs 16.0 mo [14.9–17.1 mo], p < 0.001; Fig. 3 A).
Fig. 3 (A) Patients with lymph node metastases (LNM) had a higher risk of death (median overall survival [OS] 23.9 mo, 95% confidence interval [CI] 21.6–26.3 mo) than those without LNM (median OS 16.0 mo, 95% CI 14.9–17.1; hazard ratio [HR] 1.41, 95% CI 1.28–1.54; p < 0.001). (B) A comparison without consideration of the histologic subtypes demonstrates that any lymph node involvement was associated with worse survival outcome compared to patients without LNM. Median survival times were 25.2 mo, 20.3 mo, 16.2 mo, and 11.5 mo in patients with no LNM and lymph node involvement in supradiaphragmatic (SPD), subdiaphragmatic (SBD), and both localizations (SPD+/SBD+), respectively. The HRs (95% CIs) for patients with SPD, SBD, and SPD+/SBD+ LNM were 1.24 (1.06–1.44), 1.45 (1.23–1.71), and 1.93 (1.60–2.33), respectively. (C) In clear cell renal cell carcinoma (ccRCC), patients with SPD and SBD LNM had similar median OS (SPD, 21.6 mo; SBD, 19.8 mo). However, both groups had a significantly higher risk of death (SPD: HR 1.24, 95% CI 1.05–1.47; SBD: HR 1.28, 95% CI 1.05–1.56) in comparison to those without lymph nodes. Patients with SPD+/SBD+ metastases had the worst OS (12.85 mo) and a twofold higher risk of death than patients without lymph node involvement (HR 2.04, 95% CI 1.68–2.48). (D) In non–clear cell RCC (nccRCC), median OS times according to lymph node localization were 15.8 mo for no lymph nodes, 16.1 mo for SPD LNM, 11.2 mo for SBD LNM, and 12.5 mo for SPD+/SBD+ LNM. There were no significant differences between the groups (log rank p = 0.444).
In subgroup analyses, patients without LNM had a median OS of 25.2 mo (21.9–28.5 mo) compared to 20.3 mo (17.1–23.5 mo) for SPD LNM (p = 0.093), 16.2 mo (13.5–20.0 mo) for SBD LNM (p < 0.001), and 11.5 mo (9.8–13.3 mo) (p < 0.001) for SPD+/SBD+ LNM ( Fig. 3 B).
In ccRCC, further analysis considering LNM localization revealed similar OS times for patients with SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13]), and were shorter than for patients without LNM (26.97 mo [23.32– 30.62], p = 0.01; Fig. 3 C). Patients with SPD+/SBD+ LNM had significantly (p = 0.001) worse OS (12.85 mo [9.45–16.24 mo]) compared to SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13 mo]; Fig. 3 C). In nccRCC, analyses according to LNM revealed no significant differences in OS (p = 0.444; Fig. 3 C).
For all histological subtypes, LNM were an independent prognostic factor for adverse OS in MVA analysis (p < 0.001; Table 3 ). Additional MVA according to LNM sites revealed that SBD LNM (p < 0.001) and SPD+/SBD+ LNM (p < 0.001) were independent prognostic factors for OS. SPD LNM were not independently associated with OS (p = 0.999; Table 3 ). In ccRCC, MVA according to LNM localization confirmed an independent association between SBD LNM (p = 0.028) and SPD+/SBD+ LNM (p = 0.006) and OS ( Table 3 ).
Table 3 Multivariate Cox regression analysis of association between lymph node metastases (LNM) and overall survival (OS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time < 1 yr | 1.34 | 1.19–1.49 | <0.001 |
| Low hemoglobin | 1.51 | 1.31–1.70 | <0.001 |
| High Ca | 1.47 | 1.25–1.73 | <0.001 |
| Neutrophilia | 1.67 | 1.45–1.92 | <0.001 |
| Thrombophilia | 1.31 | 1.14–1.51 | <0.001 |
| Karnofsky PS <80% | 1.88 | 1.67–2.13 | <0.001 |
| Liver metastases | 1.50 | 1.32–1.70 | <0.001 |
| Bone metastases | 1.50 | 1.34–1.67 | <0.001 |
| LNM | 1.33 | 1.19–1.48 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time < 1 yr | 1.32 | 1.11–1.56 | 0.001 |
| Low hemoglobin | 1.45 | 1.21–1.73 | <0.001 |
| High Ca | 1.38 | 1.08–1.77 | 0.010 |
| Neutrophilia | 1.78 | 1.43–2.22 | <0.001 |
| Thrombophilia | 1.49 | 1.21–1.83 | <0.001 |
| Karnofsky PS <80% | 2.19 | 1.82–2.62 | <0.001 |
| Liver metastases | 1.84 | 1.53–2.20 | <0.001 |
| Bone metastases | 1.61 | 1.36–1.90 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.57 | 0.98–1.48 | 0.075 |
| SBD LNM | 1.59 | 1.26–1.96 | <0.001 |
| SPD+/SBD+ LNM | 1.32 | 1.25–2.02 | <0.001 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time < 1 yr | 1.36 | 1.12–1.64 | 0.002 |
| Low hemoglobin | 1.36 | 1.11–1.67 | 0.003 |
| High Ca | 1.44 | 1.10–1.89 | 0.009 |
| Neutrophilia | 1.65 | 1.27–2.13 | <0.001 |
| Thrombophilia | 1.52 | 1.19–1.91 | 0.001 |
| Karnofsky PS <80% | 2.12 | 1.72–2.61 | <0.001 |
| Liver metastases | 1.94 | 1.57–2.41 | <0.001 |
| Bone metastases | 1.84 | 1.53–2.22 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.13 | 0.90–1.42 | 0.301 |
| SBD LNM | 1.34 | 1.03–1.74 | 0.028 |
| SPD+/SBD+ LNM | 1.47 | 1.11–1.93 | 0.006 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm; ccRCC = clear cell RCC.
An overall comparison of patients with and without LNM is shown in Table 4 . Patients with LNM presented more often with nccRCC (p < 0.001) and higher Fuhrman grades (p = 0.007), and had less often undergone prior nephrectomy (p < 0.001). The IMDC risk factors of high calcium (p = 0.001), thrombocytosis (p < 0.001), neutrophilia (p < 0.001), and Karnofsky performance status <80% (p = 0.009) were all more frequently seen in patients with LNM. Comparison of additional metastatic sites revealed that LNM are related to a higher overall metastatic burden (one or more additional metastatic site p < 0.001).
Table 4 General comparison of clinicopathologic features of patients with lymph node metastases (LNM) compared to those without LNM a
| No LNM (n = 1380) |
LNM (n = 1616) |
p value | |
|---|---|---|---|
| Mean age, yr (SD) | 58.65 (11.36) | 58. 50 (11.30) | 0.716 |
| Male, n (%) | 1003 (72.7) | 1182 (73.1) | 0.805 |
| Nephrectomy, n (%) | 1134 (82.4) | 1218 (75.4) | <0.001 |
| Fuhrman grade n (%) | 0.007 | ||
| 1 | 44 (4.2) | 33 (2.8) | |
| 2 | 302 (28.6) | 269 (23.1) | |
| 3 | 426 (40.3) | 501 (43.1) | |
| 4 | 284 (26.9) | 359 (30.9) | |
| Non–clear cell RCC, n (%) | 127 (9.8) | 253 (17.4) | <0.001 |
| Sarcomatoid features, n (%) | 143 (12.0) | 170 (12.0) | 0.586 |
| Therapy class, n (%) | |||
| AntiVEGF | 1316 (95.4) | 1507 (93.3) | |
| mTOR | 60 (4.3) | 105 (6.5) | |
| Combination | 4 (0.3) | 4 (0.2) | |
| DTT <1 yr, n (%) | 736 (53.4) | 905 (56.0) | 0.151 |
| Low HB (<LLN), n (%) | 691 (56.0) | 877 (58.1) | 0.261 |
| High calcium, n (%) | 98 (8.6) | 179 (12.6) | 0.001 |
| Neutrophilia, n (%) | 154 (13.1) | 286 (19.3) | <0.001 |
| Thrombocytosis, n (%) | 172 (15.4) | 299 (22.4) | <0.001 |
| Karnofsky PS <80%, n (%) | 261 (21.5) | 372 (25.9) | 0.009 |
| Lung metastases, n (%) | 939 (68.0) | 1149 (72.5) | 0.016 |
| Brain metastases, n (%) | 100 (7.2) | 141 (9.4) | 0.009 |
| Liver metastases, n (%) | 278 (20.2) | 328 (21.7) | 0.337 |
| Bone metastases, n (%) | 481 (34.9) | 551 (35.9) | 0.462 |
| Other metastic sites, n (%) | 470 (34.1) | 643 (43.1) | <0.001 |
| >1 metastatic site, n (%) | 805 (58.3%) | 1446 (89.5%) | <0.001 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; mTOR = mammalian target of rapamycin; PS = performance status; RCC = renal cell carcinoma; SD = standard deviation; VEGF = vascular endothelial growth factor.
The LNM localization distribution and clinicopathologic features in ccRCC and nccRCC are shown in Table 5 . In ccRCC, IMDC risk factors of neutrophilia (p = 0.049) and thrombophilia (p = 0.016) were more common. Patients with LNM had undergone nephrectomy less frequently (p = 0.001). In patients with ccRCC, those with LNM had concomitant lung (p < 0.001) and other metastases (eg, adrenal glands, other soft tissues; p = 0.009) more often than those without LNM; there was no difference in the frequency of bone (p = 0.276), liver (p = 0.943), or brain metastases (p = 0.187).
Table 5 Distribution of metastatic lymphatic sites in relation to the diaphragm in clear cell and non–clear cell renal carcinoma (RCC) a
| Clear cell RCC (n = 1218) |
Non–clear cell RCC (n = 159) |
|
|---|---|---|
| No LNM, n (%) | 552 (45.6) | 52 (32.7) |
| SPD LNM, n (%) | 321 (26.6) | 22 (13.8) |
| SBD LNM, n (%) | 204 (16.9) | 54 (34.0) |
| SPD+/SBD+ LNM, n (%) | 132 (10.9) | 31 (19.5) |
| LMN | No LMN | p value | LMN | No LMN | p value | |
|---|---|---|---|---|---|---|
| Nephrectomy, n (%) | 542 (81.2) | 492 (88.2) | 0.001 | 82 (76.6) | 43 (82.7) | 0.418 |
| Male, n (%) | 515 (77.7) | 403 (72.6) | 0.045 | 75 (70.1) | 37 (71.2) | 1.00 |
| DTT <1 yr, n (%) | 348 (52.5) | 265 (47.9) | 0.120 | 65 (60.7) | 29 (55.8) | 0.607 |
| Low HB (<LLN) | 331 (54.0) | 233 (48.7) | 0.088 | 49 (51.0) | 19 (45.2) | 0.582 |
| High calcium, n (%) | 70 (12.3) | 40 (9.5) | 0.184 | 5 (5.7) | 4 (12.1) | 0.254 |
| Neutrophilia, n (%) | 95 (15.9) | 53 (11.6) | 0.049 | 24 (25.3) | 7 (16.7) | 0.376 |
| Thrombophilia, n (%) | 106 (22.0) | 63 (15.6) | 0.016 | 17 (21.0) | 10 (25.6) | 0.642 |
| Karnofsky PS <80%, n (%) | 145 (24.7) | 95 (20.0) | 0.076 | 24 (24.7) | 10 (20.4) | 0.680 |
| Sarcomatoid features, n (%) | 52 (9.1) | 62 (12.6) | 0.074 | 14 (15.4) | 7 (15.6) | 1.00 |
| Lung metastases, n (%) | 507 (76.8) | 372 (67.0) | <0.001 | 52 (49.1) | 20 (38.5) | 0.237 |
| Brain metastases, n (%) | 65 (9.8) | 42 (7.6) | 0.187 | 4 (3.8) | 2 (3.8) | 1.00 |
| Liver metastases, n (%) | 132 (20.0) | 112 (20.2) | 0.943 | 32 (30.2) | 28 (53.8) | 0.005 |
| Bone metastases, n (%) | 220 (33.3) | 202 (36.4) | 0.276 | 35 (33.0) | 20 (38.5) | 0.594 |
| Other metastatic sites, n (%) | 287 (45.7) | 211 (38.2) | 0.009 | 40 (40.8) | 20 (38.5) | 0.862 |
| >1 metastatic site, n (%) | 602 (90.8) | 320 (57.7) | <0.001 | 87 (81.3) | 30 (7.7) | 0.002 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; PS = performance status; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm.
In nccRCC, liver metastases were found more often in patients without LNM (p = 0.005). A comparison of other clinicopathologic features between nccRCC patients with and without LNM revealed no significant differences ( Table 5 ).
It has previously been shown that only a minority of patients with kidney cancer undergo LND at the time of nephrectomy. This is the case even for high-risk patients [17] , even though retrospective studies have suggested a therapeutic benefit of LND [11] and [18]. It remains uncertain whether extended LND can improve the survival outcome for mRCC patients. However, if it were shown that SBD LNM affect the survival outcome for mRCC patients, this could form the basis for a rationale in favor of performing LND, or at least justify the need for prospective clinical trials to address this question. In this regard, an understanding of the influence of LNM on treatment response and survival outcome for mRCC patients undergoing TT could make an important contribution to the debate on surgical management of RCC.
Our study has demonstrated that LNM are associated with advanced IMDC risk factors, are an adverse prognostic factor for PFS and OS, and are associated with a higher metastatic burden. Furthermore, subanalyses showed that these differences depend on LNM localization.
Different cellular clones of RCC appear to differ in their propensity to spread to different organ sites [19] ; however, whether these differences impact on the survival outcome of mRCC patients is not completely understood. A previous IMDC study highlighted the prognostic importance of bone and liver metastases for survival outcome in mRCC patients, and suggested that aggressive RCC subclones tend to metastasize to these organs [16] . In addition, another study by our group showed that metastatic spread to other organs such as the pancreas or soft tissue tends to develop as a late clinical symptom, and has favorable treatment response and survival outcomes [20] . While these tumors may represent a favorable RCC subtype, our current results suggest that RCCs that spread to regional retroperitoneal lymph nodes represent another aggressive RCC subtype that develops bone metastases.
It has been suggested that LNM in ccRCC occur independently of von Hippel Lindau (VHL) gene inactivation [21] . Current agents used for mRCC treatment modulate pathways that are dysregulated because of VHL inactivation. This would imply that LNM might be less amenable to treatment with VEGF and mTOR inhibitors. However, our findings demonstrated no significant differences in best response to first-line therapies in patients with LNM versus those without LNM in both ccRCC and nccRCC patient groups. Nonetheless, patients with LNM had significantly shorter PFS than patients without LNM. While there was only a trend for an independent association between SBD LNM and short PFS, SPD+/SBD+ LNM were independently related to shorter PFS. Patients with SPD+/SBD+ LNM had additional hematogenous metastatic sites in 97% of cases. Thus, the high metastatic burden may explain the shorter response duration of first-line agents in this group of patients. However, a shorter duration for treatment success could also support the hypothesis that LNM are caused by RCC clones that do not have an inactive VHL gene function.
In the cytokine era, LNM were associated with shorter OS [1] and [2]. Our study demonstrates similar findings in the TT era. Interestingly, a worse survival outcome was observed for SBD LNM, both when occurring exclusively and in combination with thoracic lymph node involvement. Moreover, SBD LNM remained an independent prognostic factor for PFS and OS when analyzed without consideration of histologic subtype, and were an independent prognostic factor for OS in ccRCC. Conversely, SPD LNM were not an independent prognostic factor. This is notable and underscores the unique nature of RCCs that spread via retroperitoneal lymph nodes. Because the best response to TT did not different between SBD and SPD LNM, it appears that survival outcome is influenced by numerous factors. Patients with SBD LNM had faster disease progression before treatment initiation than those with SPD LNM (diagnosis to treatment <1 yr; 63% vs 46%; p < 0.001; data not shown). Despite a best response that was broadly similar, it seems that the disease progresses faster in SBD LNM patients during the later clinical course than in SPD LNM patients. In addition, SBD LNM were more often associated with anemia compared to SPD LNM (59% vs 48%; p = 0.008; data not shown). Anemia has multifactorial causes in cancer patients. However, one of the most important contributing factors is immune system dysregulation [22] . Dysregulation of the immune system may also be the reason for metastatic spread to SBD sites. Collectively, while risk factors that are currently unknown appear to dictate the clinical course in patients with SPD metastases, SBD metastases are independently associated with PFS and OS.
The majority of patients analyzed in this study did not undergo formal LND. Therefore, the effect of LND cannot be determined retrospectively. However, the current results indicate that SBD lymph nodes may have substantial negative effects on survival outcome and it is worth speculating whether extended removal of these metastatic sites could improve the clinical outcome for mRCC patients. The IMDC has recently shown that cytoreductive nephrectomy may provide therapeutic benefit in a large number of patients [23] . With regard to our current findings, it would be worth determining the role of LND in these patients.
The current study has several limitations that should be considered when interpreting the results. The IMDC database comprises retrospectively collected data for consecutive series of mRCC patients. These individual series, moreover, do not have standardized follow-up protocols, but this limitation may more accurately reflect real world outcomes. In addition, there was no centralized pathologic or radiologic review. Finally, lymph node status was determined clinically according to CT images in a number of cases, and it has been demonstrated that clinical and pathologic lymph node status is not necessarily always correlated.
LNM are associated with poor prognosis and adverse prognostic IMDC risk factors in mRCC patients treated with TT. Much of this increased risk associated with LNM appears to be dictated by SBD localization rather than SPD localization. Whether LND could improve the outcome for this group of patients remains unknown, but it is an important clinical question requiring prospective trials.
Author contributions: Daniel Y. Heng had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kroeger, Heng.
Acquisition of data: all authors.
Analysis and interpretation of data: Kroeger, Heng, Choueiri.
Drafting of the manuscript: Kroeger, Heng, Choueiri.
Critical revision of the manuscript for important intellectual content: all authors.
Statistical analysis: Kroeger, Heng.
Obtaining funding: None.
Administrative, technical, or material support: all authors.
Supervision: Choueiri, Heng.
Other (specify): None.
Financial disclosures: Daniel Y. Heng certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Benoit Beuselinck is an investigator of the EudraCT: 2011-006085-40/MetaSun trial supported by Pfizer. Frede Donskov has received research support from Novartis and GlaxoSmithKline. Toni K. Choueiri has received research funding from Pfizer and has an advisory role at Pfizer, GSK, Novartis, and Bayer. All other authors declare no conflicts of interest to the current study.
Funding/Support and role of the sponsor: None.
Acknowledgments: Dr. Choueiri is supported in part by the Trust family and the Loker Pinard and Michael Brigham Funds for Kidney Cancer Research at the Dana-Farber Cancer Institute. Dr. Kroeger thanks Dr. Klatte, Department of Urology, Medical University of Vienna, for statistical advice.
Renal cell carcinoma (RCC) is a heterogeneous disease consisting of several tumor types that have their own genetic, molecular, and clinical characteristics. Historical studies in the cytokine era by the University of California Los Angeles (UCLA) and the National Institutes of Health (NIH) described poor prognosis and worse treatment response to interleukin-2 (IL-2) in patients with metastatic involvement of the retroperitoneal lymph nodes [1] and [2]. Patients with and without retroperitoneal lymph node involvement had median overall survival (OS) times of 8.5 mo and 14.7 mo (p = 0.0004), respectively [2] . Pantuck et al [1] reported a better objective response rate in N0M1 compared to N+M1 patients (p = 0.01), with survival times of 10.5 mo and 20.4 mo, respectively. The adverse prognostic impact of retroperitoneal lymph node metastases (LNM) on survival outcome in patients with metastatic RCC (mRCC) was confirmed by several other institutions during the cytokine era [3], [4], and [5]. However, in the era of targeted therapies (TT), the survival outcome for patients with LNM has not yet been well characterized.
RCC preferentially metastasizes via hematogenous routes. Bianchi et al [6] recently described hematogenous metastatic sites in 80% of 11 157 patients with mRCC, while extension of the disease into lymph nodes was described only in 20% of patients. However, the authors did not provide details about the distribution of lymph node localization. The dominant regional LNM sites in RCC are retro- and paracaval, pre- and paraaortic, and interaortocaval lymph nodes in an anatomically intact retroperitoneum [7] . LNM can be unpredictable, and can also occur via direct extension to the thorax, supraclavicular lymph nodes, and iliac lymph nodes [7] . It has not been documented if particular sites of positive lymph node localization have a negative influence on the survival outcome of mRCC patients treated with TT.
Whether retroperitoneal lymph node dissection (LND) improves survival outcome is the subject of ongoing scientific debate. A phase 3 clinical trial (EORTC 30881) demonstrated positive lymph nodes in only 4% of investigated cases having no apparent involvement on computed tomography (CT) imaging. There was no advantage in OS, recurrence-free survival, or progression-free survival (PFS) for RCC patients who underwent LND [8] . The study has been criticized, however, because the majority of patients included had low-stage disease. Moreover, uniform surgical templates were not used for LND [9] . However, retrospective studies have demonstrated that LNM are more prevalent in high-risk patients with high Fuhrman grades, sarcomatoid features, locally advanced tumor stage, tumor size >10 cm, and tumor necrosis [10] . It has been suggested that LND is particularly beneficial in these high-risk patients [11] and [12]. Trends in the surgical management of RCC that have moved away from open radical surgery with extended LND to laparoscopic and robotic surgery with minimal hilar LND, combined with the negative results of phase 3 studies showing no additional benefit of LND over removal of the primary tumor alone, have decreased enthusiasm for performing extended LND. Therefore, it is common clinical practice to perform LND only in high-risk patients, and with a diagnostic rather than a curative intention [7] . Retrospective studies may be unable to add new insights into the current debate because only a prospective randomized trial with clear inclusion and exclusion criteria would be able to overcome this selection bias. Before planning such trials, however, it is prudent to better understand the impact of LNM and different lymph node localizations on the survival outcome for mRCC patients treated with current state-of-the-art therapies using data from the International mRCC Database Consortium (IMDC).
The IMDC database includes centers from North America (Canada, USA), Europe (Denmark, Greece, Belgium), Asia (Singapore, Japan, South Korea), and New Zealand. Data were collected from August 15, 2008 until December 31, 2013. At the time of analysis, the database contained data on 3405 patients who had received first-line targeted therapies between 2003 and 2013. The final study cohort comprised 2996 patients who were treated with first-line vascular endothelial growth factor (VEGF) inhibitors (n = 2823), mammalian target of rapamycin (mTOR) inhibitors (n = 165), or a combination of both (n = 8) therapies between April 2003 and August 2013. Patients were excluded from the analyses if they had received experimental therapies in first-line treatment (n = 7) or if no information was available on their lymph node status (n = 402; Fig. 1 ). Lymph node status was determined according to standard pathologic and CT criteria.
Information about LND was available for 890 patients. Full LND was performed in only 51 patients, and a limited (hilar) dissection was performed in 173 patients. These numbers were too low to allow any analyses and conclusions regarding the influence of LND on the survival outcome for mRCC patients treated with TT. Whether LND could be beneficial for mRCC patients was therefore not considered in the present study.
All centers obtained local Institutional Review Board approval before including data in the IMDC database. Baseline patient characteristics included demographic, clinicopathologic, and laboratory data, as previously described [13] . Survival data were retrospectively and prospectively collected from medical chart reviews and electronic records. Uniform data templates were used to ensure consistent data collection at each institution. The majority of patients were treated using standard care, but a subset received therapy as part of clinical trials. All patients were considered in consecutive series to avoid selection bias.
The primary hypothesis of the current study was that mRCC patients with LNM have worse survival outcome than patients without LNM when treated with TT. The secondary hypothesis was that worse survival outcome for mRCC patients with LNM is dictated by subdiaphragmatic (SPD) LNM rather than supradiaphragmatic (SBD) LNM.
Therefore, treatment response to current standard-of-care TT agents, OS, and PFS were compared between patients with LNM and those without LNM. We did not compare outcomes between the drugs used for treatment because these comparisons are frequently biased outside a clinical trial and there may not be sufficient power for such an analysis. Additional analyses were performed for subgroups of patients with exclusively SPD lymph node involvement, exclusively SBD lymph node involvement, or with lymph node involvement in both sites (SPD+/SBD+). Previous studies have demonstrated that tumor biology has an important impact on the survival outcome for mRCC patients [14] . Thus, the different biology of RCC was considered by separate analyses for clear cell RCC (ccRCC) and non-ccRCC (nccRCC) patients. Patients with SPD primarily had mediastinal lymph node involvement, while patients with SBD primarily had retroperitoneal lymph node involvement. Lymph node status was determined before initiation of TT. The selection process for patient groups analyzed in the study is shown in Figure 1 .
OS was defined as the time period between TT initiation and the date of death, with censoring on the day of the last follow-up visit. PFS was defined as the period between treatment initiation and progression, drug cessation, or death, with censoring at the last follow-up visit. Kaplan-Meier plots were used to estimate median OS and PFS, and univariate comparisons were performed using the log rank test. Uni- and multivariate (MVA) Cox regression analyses were performed to test the association of predefined prognostic factors and LNM with survival outcome. The analyses were performed with backward stepwise selection criteria, and significance was tested using the Wald statistic [13] . In MVA analyses, the IMDC (Heng) risk criteria were applied for adjustment [15] . The IMDC risk criteria include a time period from diagnosis to treatment of <1 yr, Karnofsky performance score <80%, anemia, hypercalcemia, thrombophilia, and neutrophilia. Upper and lower limits of normal for the laboratory parameters were based on institutional limits. In addition, adjustment was performed for bone and liver metastases because interactions of LNM with other metastatic sites should be considered. The prognostic relevance of bone and liver metastases has previously been demonstrated by our group [16] .
A secondary aim was to analyze baseline characteristics. Patient and tumor characteristics were compared using the Student t test for continuous variables, or the chi-square test and the Fisher exact test for categorical variables.
Statistical analyses were performed using SPSS version 22 (IBM, Chicago, IL, USA), and a two-sided value of p < 0.05 was considered statistically significant.
At the time of data analysis, 2525 (84.3%) patients had stopped taking their TT; the median time on TT was 6.0 mo (25th percentile, 2.7 mo; 75th percentile, 13.0 mo). Best overall response data to first-line therapy were available for 2559 patients. There were no significant differences in response to first-line TT (p = 0.059) when patients were compared without consideration of RCC subtype or LNM localization ( Table 1 ).
Table 1 Treatment response to targeted therapies in patients with lymph node metastases (LNM) compared to those without LNM
| Best response | |||||
|---|---|---|---|---|---|
| CR | PR | SD | PD | ||
| All RCC subtypes | |||||
| No LMN | n | 14 | 304 | 556 | 290 |
| % | 1.2 | 26.1 | 47.8 | 24.9 | |
| LMN | n | 9 | 314 | 690 | 382 |
| % | 0.6 | 22.5 | 49.5 | 27.4 | |
| Clear cell RCC | |||||
| No LNM | n | 3 | 108 | 261 | 95 |
| % | 0.6 | 23.1 | 55.9 | 20.3 | |
| SPD LNM | n | 2 | 73 | 149 | 55 |
| % | 0.7 | 26.2 | 53.4 | 19.7 | |
| SBD LNM | n | 2 | 50 | 92 | 42 |
| % | 1.1 | 26.9 | 49.5 | 22.6 | |
| SPD+/SBD+ LNM | n | 0 | 28 | 58 | 36 |
| % | 0.0 | 23.0 | 47.5 | 29.5 | |
| Non–clear cell RCC | |||||
| No LNM | n | 0 | 3 | 21 | 21 |
| % | 0.0 | 6.7 | 46.7 | 46.7 | |
| SPD LNM | n | 0 | 3 | 12 | 3 |
| % | 0.0 | 16.7 | 66.7 | 16.7 | |
| SBD LNM | n | 0 | 7 | 24 | 19 |
| % | 0.0 | 14.0 | 48.0 | 38.0 | |
| SPD+/SBD+ LNM | n | 1 | 4 | 10 | 6 |
| % | 4.8 | 19.0 | 47.6 | 28.6 | |
CR = complete remission; PR = partial response; SD = stable disease; PD = progressive disease; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm.
Both response data and LNM localization data were available for 1054 ccRCC patients and 134 nccRCC patients. Comparison of data by LNM localization revealed no difference in best response to first-line therapies for either ccRCC (p = 0.457) or nccRCC (p = 0.192; Table 1 ).
Patients with LNM generally had worse PFS compared to those with no LNM without considering LNM localization or RCC subtype (median PFS [95% confidence interval] 7.8 mo [7.1–8.5 mo] vs 6.1 [5.6–6.6 mo], log rank p = 0.001; Fig. 2 A).
Fig. 2 (A) Comparison of progression-free survival (PFS) in patients with positive lymph nodes versus patients without positive lymph nodes (median 7.8 mo vs. 6.1 mo) without consideration of histologic subtype. Patients with lymph node metastases (LNM) had worse PFS (hazard ratio [HR] 1.14, 95% confidence interval [CI] 1.06–1.24; p = 0.001). (B) Comparison of PFS according to lymph node localization. Patients without LNM had the best PFS (8.8 mo), and patients with both supradiaphragmatic (SPD) and subdiaphragmatic (SBD) LNM localizations (SPD+/SBD+) had the worst PFS (5.5 mo). Patients with SBD LNM had significantly worse PFS compared to patients with no LNM (6.8 vs 8.8 mo; p = 0.003) and those with SPD LNM (6.8 vs 8.3 mo; p = 0.022). Any metastatic involvement of retroperitoneal lymph nodes, either alone (HR 1.25, 95% CI 1.08–1.45) or in combination with SPD+/SBD+LNM (HR 1.50, 95% CI 1.24–1.74; p < 0.001) was associated with higher risk of progression. By contrast, SPD LNM alone was not a prognostic factor for progression (HR 1.03, 95% 0.895–1.18; p = 0.695). (C) PFS by lymph node localization in clear cell renal cell carcinoma (ccRCC; median PFS: no nodes, 9.9 mo; SPD LNM, 8.7 mo; SBD LNM, 8.0 mo; and both, 5.6 mo). Patients with SPD LNM (HR 1.09, 95% CI 0.94–1.25) and SBD LNM (HR 1.04, 95% CI 0.96–1.14) had similar PFS, and this was not significantly worse than the PFS of patients without LNM (p = 0.265 and p = 0.333 for SPD and SBD LNM, respectively).
Comparison of PFS rates according to LNM sites revealed that patients with SBD LNM had significantly shorter PFS (6.8 mo [5.7–8.0 mo]) compared to those without LNM (8.8 mo [7.6–10.0 mo]; p = 0.003) or SPD LNM (8.3 mo [7.2–9.6 mo]; p = 0.022). The worst PFS was observed in patients with SPD+/SBD+ LNM (5.4 mo [4.6–6.4 mo]), but this did not significantly differ from PFS in patients with SBD LNM (p = 0.093; Fig. 2 B).
In ccRCC, however, comparison according to LNM localization revealed no significant differences in PFS between patients without LNM (9.9 mo [8.6–11.1 mo]) and those with SPD LNM (8.7 mo [7.3–10.1 mo]) or SBD LNM (8.0 mo [6.9–9.2 mo]) (p = 0.312; Fig. 2 C). SPD+/SBD+ LNM patients had the worst survival outcome (5.6 mo [4.0–7.2 mo]) in comparison to patients without LNM (p < 0.001), with SPD LNM (p = 0.006), and with SBD LNM (p = 0.035; Fig. 2 C). In nccRCC, there was no difference in PFS between patients with no LNM (3.9 mo [2.9–5.0 mo]), SPD LNM (4.6 mo [0.0–9.9 mo]), SBD LNM (5.4 mo [4.7–6.0 mo]), and SPD+/SBD+ LNM (5.4 mo [3.2–7.3 mo]).
In MVA analyses, LNM remained an independent prognostic factor (HR 1.13 [1.03–1.24]; p = 0.013) when adjusted for IMDC risk criteria, as well as liver and bone metastases ( Table 2 ) [16] . In subgroup analysis, SBD LNM (HR 1.31 [1.08–1.60]; p = 0.006) and SPD+/SBD LNM (HR 1.29 [1.04–1.61]; p = 0.022) were independent prognostic factors for PFS, while SPD LNM was not independently associated with PFS (p = 1.00; Table 2 ). In ccRCC, only SPD+/SBD+ LNM (p = 0.025) was independently associated with PFS ( Table 2 ).
Table 2 Multivariate analyses for association between lymph node metastases (LNM) and progression-free survival (PFS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time <1 yr | 1.33 | 1.21–1.47 | <0.001 |
| Low hemoglobin | 1.37 | 1.24–1.52 | <0.001 |
| Neutrophilia | 1.43 | 1.25–1.63 | <0.001 |
| Thrombocytosis | 1.16 | 1.03–1.32 | 0.019 |
| Karnofsy PS <80% | 1.68 | 1.50–1.88 | <0.001 |
| LNM | 1.13 | 1.03–1.24 | 0.013 |
| Liver metastases | 1.26 | 1.12–1.41 | <0.001 |
| Bone metastases | 1.27 | 1.15–1.40 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.19–1.61 | <0.001 |
| Low hemoglobin | 1.42 | 1.21–1.67 | <0.001 |
| Neutrophilia | 1.46 | 1.19–1.80 | <0.001 |
| Thrombocytosis | 1.22 | 1.01–1.47 | 0.045 |
| Karnofsy PS <80% | 1.94 | 1.65–2.28 | <0.001 |
| Liver metastases | 1.39 | 1.18–1.65 | <0.001 |
| Bone metastases | 1.45 | 1.25–1.68 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.83–1.20 | 0.990 |
| SBD LNM | 1.32 | 1.08–1.60 | 0.006 |
| SPD+/SBD+ LNM | 1.29 | 1.04–1.61 | 0.022 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.17–1.64 | <0.001 |
| Low hemoglobin | 1.43 | 1.21–1.69 | <0.001 |
| Neutrophilia | 1.38 | 1.09–1.76 | 0.009 |
| Thrombocytosis | 1.97 | 1.64–2.38 | <0.001 |
| Karnofsy PS <80% | 1.51 | 1.24–1.85 | <0.001 |
| Bone metastases | 1.64 | 1.38–1.94 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.82–1.22 | 0.998 |
| SBD LNM | 1.23 | 0.98–1.55 | 0.077 |
| SPD+/SBD+ LNM | 1.33 | 1.04–1.71 | 0.025 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm; ccRCC = clear cell RCC.
At the time of the analysis, 1960 patients had died; the median OS time for the total patient cohort was 20.0 mo (25th percentile, 42.3 mo; 75th percentile, 7.9 mo). Patients with metastatic spread to lymph nodes had a worse survival outcome than patients without LNM (24.0 mo [21.6–26.3 mo] vs 16.0 mo [14.9–17.1 mo], p < 0.001; Fig. 3 A).
Fig. 3 (A) Patients with lymph node metastases (LNM) had a higher risk of death (median overall survival [OS] 23.9 mo, 95% confidence interval [CI] 21.6–26.3 mo) than those without LNM (median OS 16.0 mo, 95% CI 14.9–17.1; hazard ratio [HR] 1.41, 95% CI 1.28–1.54; p < 0.001). (B) A comparison without consideration of the histologic subtypes demonstrates that any lymph node involvement was associated with worse survival outcome compared to patients without LNM. Median survival times were 25.2 mo, 20.3 mo, 16.2 mo, and 11.5 mo in patients with no LNM and lymph node involvement in supradiaphragmatic (SPD), subdiaphragmatic (SBD), and both localizations (SPD+/SBD+), respectively. The HRs (95% CIs) for patients with SPD, SBD, and SPD+/SBD+ LNM were 1.24 (1.06–1.44), 1.45 (1.23–1.71), and 1.93 (1.60–2.33), respectively. (C) In clear cell renal cell carcinoma (ccRCC), patients with SPD and SBD LNM had similar median OS (SPD, 21.6 mo; SBD, 19.8 mo). However, both groups had a significantly higher risk of death (SPD: HR 1.24, 95% CI 1.05–1.47; SBD: HR 1.28, 95% CI 1.05–1.56) in comparison to those without lymph nodes. Patients with SPD+/SBD+ metastases had the worst OS (12.85 mo) and a twofold higher risk of death than patients without lymph node involvement (HR 2.04, 95% CI 1.68–2.48). (D) In non–clear cell RCC (nccRCC), median OS times according to lymph node localization were 15.8 mo for no lymph nodes, 16.1 mo for SPD LNM, 11.2 mo for SBD LNM, and 12.5 mo for SPD+/SBD+ LNM. There were no significant differences between the groups (log rank p = 0.444).
In subgroup analyses, patients without LNM had a median OS of 25.2 mo (21.9–28.5 mo) compared to 20.3 mo (17.1–23.5 mo) for SPD LNM (p = 0.093), 16.2 mo (13.5–20.0 mo) for SBD LNM (p < 0.001), and 11.5 mo (9.8–13.3 mo) (p < 0.001) for SPD+/SBD+ LNM ( Fig. 3 B).
In ccRCC, further analysis considering LNM localization revealed similar OS times for patients with SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13]), and were shorter than for patients without LNM (26.97 mo [23.32– 30.62], p = 0.01; Fig. 3 C). Patients with SPD+/SBD+ LNM had significantly (p = 0.001) worse OS (12.85 mo [9.45–16.24 mo]) compared to SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13 mo]; Fig. 3 C). In nccRCC, analyses according to LNM revealed no significant differences in OS (p = 0.444; Fig. 3 C).
For all histological subtypes, LNM were an independent prognostic factor for adverse OS in MVA analysis (p < 0.001; Table 3 ). Additional MVA according to LNM sites revealed that SBD LNM (p < 0.001) and SPD+/SBD+ LNM (p < 0.001) were independent prognostic factors for OS. SPD LNM were not independently associated with OS (p = 0.999; Table 3 ). In ccRCC, MVA according to LNM localization confirmed an independent association between SBD LNM (p = 0.028) and SPD+/SBD+ LNM (p = 0.006) and OS ( Table 3 ).
Table 3 Multivariate Cox regression analysis of association between lymph node metastases (LNM) and overall survival (OS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time < 1 yr | 1.34 | 1.19–1.49 | <0.001 |
| Low hemoglobin | 1.51 | 1.31–1.70 | <0.001 |
| High Ca | 1.47 | 1.25–1.73 | <0.001 |
| Neutrophilia | 1.67 | 1.45–1.92 | <0.001 |
| Thrombophilia | 1.31 | 1.14–1.51 | <0.001 |
| Karnofsky PS <80% | 1.88 | 1.67–2.13 | <0.001 |
| Liver metastases | 1.50 | 1.32–1.70 | <0.001 |
| Bone metastases | 1.50 | 1.34–1.67 | <0.001 |
| LNM | 1.33 | 1.19–1.48 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time < 1 yr | 1.32 | 1.11–1.56 | 0.001 |
| Low hemoglobin | 1.45 | 1.21–1.73 | <0.001 |
| High Ca | 1.38 | 1.08–1.77 | 0.010 |
| Neutrophilia | 1.78 | 1.43–2.22 | <0.001 |
| Thrombophilia | 1.49 | 1.21–1.83 | <0.001 |
| Karnofsky PS <80% | 2.19 | 1.82–2.62 | <0.001 |
| Liver metastases | 1.84 | 1.53–2.20 | <0.001 |
| Bone metastases | 1.61 | 1.36–1.90 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.57 | 0.98–1.48 | 0.075 |
| SBD LNM | 1.59 | 1.26–1.96 | <0.001 |
| SPD+/SBD+ LNM | 1.32 | 1.25–2.02 | <0.001 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time < 1 yr | 1.36 | 1.12–1.64 | 0.002 |
| Low hemoglobin | 1.36 | 1.11–1.67 | 0.003 |
| High Ca | 1.44 | 1.10–1.89 | 0.009 |
| Neutrophilia | 1.65 | 1.27–2.13 | <0.001 |
| Thrombophilia | 1.52 | 1.19–1.91 | 0.001 |
| Karnofsky PS <80% | 2.12 | 1.72–2.61 | <0.001 |
| Liver metastases | 1.94 | 1.57–2.41 | <0.001 |
| Bone metastases | 1.84 | 1.53–2.22 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.13 | 0.90–1.42 | 0.301 |
| SBD LNM | 1.34 | 1.03–1.74 | 0.028 |
| SPD+/SBD+ LNM | 1.47 | 1.11–1.93 | 0.006 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm; ccRCC = clear cell RCC.
An overall comparison of patients with and without LNM is shown in Table 4 . Patients with LNM presented more often with nccRCC (p < 0.001) and higher Fuhrman grades (p = 0.007), and had less often undergone prior nephrectomy (p < 0.001). The IMDC risk factors of high calcium (p = 0.001), thrombocytosis (p < 0.001), neutrophilia (p < 0.001), and Karnofsky performance status <80% (p = 0.009) were all more frequently seen in patients with LNM. Comparison of additional metastatic sites revealed that LNM are related to a higher overall metastatic burden (one or more additional metastatic site p < 0.001).
Table 4 General comparison of clinicopathologic features of patients with lymph node metastases (LNM) compared to those without LNM a
| No LNM (n = 1380) |
LNM (n = 1616) |
p value | |
|---|---|---|---|
| Mean age, yr (SD) | 58.65 (11.36) | 58. 50 (11.30) | 0.716 |
| Male, n (%) | 1003 (72.7) | 1182 (73.1) | 0.805 |
| Nephrectomy, n (%) | 1134 (82.4) | 1218 (75.4) | <0.001 |
| Fuhrman grade n (%) | 0.007 | ||
| 1 | 44 (4.2) | 33 (2.8) | |
| 2 | 302 (28.6) | 269 (23.1) | |
| 3 | 426 (40.3) | 501 (43.1) | |
| 4 | 284 (26.9) | 359 (30.9) | |
| Non–clear cell RCC, n (%) | 127 (9.8) | 253 (17.4) | <0.001 |
| Sarcomatoid features, n (%) | 143 (12.0) | 170 (12.0) | 0.586 |
| Therapy class, n (%) | |||
| AntiVEGF | 1316 (95.4) | 1507 (93.3) | |
| mTOR | 60 (4.3) | 105 (6.5) | |
| Combination | 4 (0.3) | 4 (0.2) | |
| DTT <1 yr, n (%) | 736 (53.4) | 905 (56.0) | 0.151 |
| Low HB (<LLN), n (%) | 691 (56.0) | 877 (58.1) | 0.261 |
| High calcium, n (%) | 98 (8.6) | 179 (12.6) | 0.001 |
| Neutrophilia, n (%) | 154 (13.1) | 286 (19.3) | <0.001 |
| Thrombocytosis, n (%) | 172 (15.4) | 299 (22.4) | <0.001 |
| Karnofsky PS <80%, n (%) | 261 (21.5) | 372 (25.9) | 0.009 |
| Lung metastases, n (%) | 939 (68.0) | 1149 (72.5) | 0.016 |
| Brain metastases, n (%) | 100 (7.2) | 141 (9.4) | 0.009 |
| Liver metastases, n (%) | 278 (20.2) | 328 (21.7) | 0.337 |
| Bone metastases, n (%) | 481 (34.9) | 551 (35.9) | 0.462 |
| Other metastic sites, n (%) | 470 (34.1) | 643 (43.1) | <0.001 |
| >1 metastatic site, n (%) | 805 (58.3%) | 1446 (89.5%) | <0.001 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; mTOR = mammalian target of rapamycin; PS = performance status; RCC = renal cell carcinoma; SD = standard deviation; VEGF = vascular endothelial growth factor.
The LNM localization distribution and clinicopathologic features in ccRCC and nccRCC are shown in Table 5 . In ccRCC, IMDC risk factors of neutrophilia (p = 0.049) and thrombophilia (p = 0.016) were more common. Patients with LNM had undergone nephrectomy less frequently (p = 0.001). In patients with ccRCC, those with LNM had concomitant lung (p < 0.001) and other metastases (eg, adrenal glands, other soft tissues; p = 0.009) more often than those without LNM; there was no difference in the frequency of bone (p = 0.276), liver (p = 0.943), or brain metastases (p = 0.187).
Table 5 Distribution of metastatic lymphatic sites in relation to the diaphragm in clear cell and non–clear cell renal carcinoma (RCC) a
| Clear cell RCC (n = 1218) |
Non–clear cell RCC (n = 159) |
|
|---|---|---|
| No LNM, n (%) | 552 (45.6) | 52 (32.7) |
| SPD LNM, n (%) | 321 (26.6) | 22 (13.8) |
| SBD LNM, n (%) | 204 (16.9) | 54 (34.0) |
| SPD+/SBD+ LNM, n (%) | 132 (10.9) | 31 (19.5) |
| LMN | No LMN | p value | LMN | No LMN | p value | |
|---|---|---|---|---|---|---|
| Nephrectomy, n (%) | 542 (81.2) | 492 (88.2) | 0.001 | 82 (76.6) | 43 (82.7) | 0.418 |
| Male, n (%) | 515 (77.7) | 403 (72.6) | 0.045 | 75 (70.1) | 37 (71.2) | 1.00 |
| DTT <1 yr, n (%) | 348 (52.5) | 265 (47.9) | 0.120 | 65 (60.7) | 29 (55.8) | 0.607 |
| Low HB (<LLN) | 331 (54.0) | 233 (48.7) | 0.088 | 49 (51.0) | 19 (45.2) | 0.582 |
| High calcium, n (%) | 70 (12.3) | 40 (9.5) | 0.184 | 5 (5.7) | 4 (12.1) | 0.254 |
| Neutrophilia, n (%) | 95 (15.9) | 53 (11.6) | 0.049 | 24 (25.3) | 7 (16.7) | 0.376 |
| Thrombophilia, n (%) | 106 (22.0) | 63 (15.6) | 0.016 | 17 (21.0) | 10 (25.6) | 0.642 |
| Karnofsky PS <80%, n (%) | 145 (24.7) | 95 (20.0) | 0.076 | 24 (24.7) | 10 (20.4) | 0.680 |
| Sarcomatoid features, n (%) | 52 (9.1) | 62 (12.6) | 0.074 | 14 (15.4) | 7 (15.6) | 1.00 |
| Lung metastases, n (%) | 507 (76.8) | 372 (67.0) | <0.001 | 52 (49.1) | 20 (38.5) | 0.237 |
| Brain metastases, n (%) | 65 (9.8) | 42 (7.6) | 0.187 | 4 (3.8) | 2 (3.8) | 1.00 |
| Liver metastases, n (%) | 132 (20.0) | 112 (20.2) | 0.943 | 32 (30.2) | 28 (53.8) | 0.005 |
| Bone metastases, n (%) | 220 (33.3) | 202 (36.4) | 0.276 | 35 (33.0) | 20 (38.5) | 0.594 |
| Other metastatic sites, n (%) | 287 (45.7) | 211 (38.2) | 0.009 | 40 (40.8) | 20 (38.5) | 0.862 |
| >1 metastatic site, n (%) | 602 (90.8) | 320 (57.7) | <0.001 | 87 (81.3) | 30 (7.7) | 0.002 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; PS = performance status; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm.
In nccRCC, liver metastases were found more often in patients without LNM (p = 0.005). A comparison of other clinicopathologic features between nccRCC patients with and without LNM revealed no significant differences ( Table 5 ).
It has previously been shown that only a minority of patients with kidney cancer undergo LND at the time of nephrectomy. This is the case even for high-risk patients [17] , even though retrospective studies have suggested a therapeutic benefit of LND [11] and [18]. It remains uncertain whether extended LND can improve the survival outcome for mRCC patients. However, if it were shown that SBD LNM affect the survival outcome for mRCC patients, this could form the basis for a rationale in favor of performing LND, or at least justify the need for prospective clinical trials to address this question. In this regard, an understanding of the influence of LNM on treatment response and survival outcome for mRCC patients undergoing TT could make an important contribution to the debate on surgical management of RCC.
Our study has demonstrated that LNM are associated with advanced IMDC risk factors, are an adverse prognostic factor for PFS and OS, and are associated with a higher metastatic burden. Furthermore, subanalyses showed that these differences depend on LNM localization.
Different cellular clones of RCC appear to differ in their propensity to spread to different organ sites [19] ; however, whether these differences impact on the survival outcome of mRCC patients is not completely understood. A previous IMDC study highlighted the prognostic importance of bone and liver metastases for survival outcome in mRCC patients, and suggested that aggressive RCC subclones tend to metastasize to these organs [16] . In addition, another study by our group showed that metastatic spread to other organs such as the pancreas or soft tissue tends to develop as a late clinical symptom, and has favorable treatment response and survival outcomes [20] . While these tumors may represent a favorable RCC subtype, our current results suggest that RCCs that spread to regional retroperitoneal lymph nodes represent another aggressive RCC subtype that develops bone metastases.
It has been suggested that LNM in ccRCC occur independently of von Hippel Lindau (VHL) gene inactivation [21] . Current agents used for mRCC treatment modulate pathways that are dysregulated because of VHL inactivation. This would imply that LNM might be less amenable to treatment with VEGF and mTOR inhibitors. However, our findings demonstrated no significant differences in best response to first-line therapies in patients with LNM versus those without LNM in both ccRCC and nccRCC patient groups. Nonetheless, patients with LNM had significantly shorter PFS than patients without LNM. While there was only a trend for an independent association between SBD LNM and short PFS, SPD+/SBD+ LNM were independently related to shorter PFS. Patients with SPD+/SBD+ LNM had additional hematogenous metastatic sites in 97% of cases. Thus, the high metastatic burden may explain the shorter response duration of first-line agents in this group of patients. However, a shorter duration for treatment success could also support the hypothesis that LNM are caused by RCC clones that do not have an inactive VHL gene function.
In the cytokine era, LNM were associated with shorter OS [1] and [2]. Our study demonstrates similar findings in the TT era. Interestingly, a worse survival outcome was observed for SBD LNM, both when occurring exclusively and in combination with thoracic lymph node involvement. Moreover, SBD LNM remained an independent prognostic factor for PFS and OS when analyzed without consideration of histologic subtype, and were an independent prognostic factor for OS in ccRCC. Conversely, SPD LNM were not an independent prognostic factor. This is notable and underscores the unique nature of RCCs that spread via retroperitoneal lymph nodes. Because the best response to TT did not different between SBD and SPD LNM, it appears that survival outcome is influenced by numerous factors. Patients with SBD LNM had faster disease progression before treatment initiation than those with SPD LNM (diagnosis to treatment <1 yr; 63% vs 46%; p < 0.001; data not shown). Despite a best response that was broadly similar, it seems that the disease progresses faster in SBD LNM patients during the later clinical course than in SPD LNM patients. In addition, SBD LNM were more often associated with anemia compared to SPD LNM (59% vs 48%; p = 0.008; data not shown). Anemia has multifactorial causes in cancer patients. However, one of the most important contributing factors is immune system dysregulation [22] . Dysregulation of the immune system may also be the reason for metastatic spread to SBD sites. Collectively, while risk factors that are currently unknown appear to dictate the clinical course in patients with SPD metastases, SBD metastases are independently associated with PFS and OS.
The majority of patients analyzed in this study did not undergo formal LND. Therefore, the effect of LND cannot be determined retrospectively. However, the current results indicate that SBD lymph nodes may have substantial negative effects on survival outcome and it is worth speculating whether extended removal of these metastatic sites could improve the clinical outcome for mRCC patients. The IMDC has recently shown that cytoreductive nephrectomy may provide therapeutic benefit in a large number of patients [23] . With regard to our current findings, it would be worth determining the role of LND in these patients.
The current study has several limitations that should be considered when interpreting the results. The IMDC database comprises retrospectively collected data for consecutive series of mRCC patients. These individual series, moreover, do not have standardized follow-up protocols, but this limitation may more accurately reflect real world outcomes. In addition, there was no centralized pathologic or radiologic review. Finally, lymph node status was determined clinically according to CT images in a number of cases, and it has been demonstrated that clinical and pathologic lymph node status is not necessarily always correlated.
LNM are associated with poor prognosis and adverse prognostic IMDC risk factors in mRCC patients treated with TT. Much of this increased risk associated with LNM appears to be dictated by SBD localization rather than SPD localization. Whether LND could improve the outcome for this group of patients remains unknown, but it is an important clinical question requiring prospective trials.
Author contributions: Daniel Y. Heng had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kroeger, Heng.
Acquisition of data: all authors.
Analysis and interpretation of data: Kroeger, Heng, Choueiri.
Drafting of the manuscript: Kroeger, Heng, Choueiri.
Critical revision of the manuscript for important intellectual content: all authors.
Statistical analysis: Kroeger, Heng.
Obtaining funding: None.
Administrative, technical, or material support: all authors.
Supervision: Choueiri, Heng.
Other (specify): None.
Financial disclosures: Daniel Y. Heng certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Benoit Beuselinck is an investigator of the EudraCT: 2011-006085-40/MetaSun trial supported by Pfizer. Frede Donskov has received research support from Novartis and GlaxoSmithKline. Toni K. Choueiri has received research funding from Pfizer and has an advisory role at Pfizer, GSK, Novartis, and Bayer. All other authors declare no conflicts of interest to the current study.
Funding/Support and role of the sponsor: None.
Acknowledgments: Dr. Choueiri is supported in part by the Trust family and the Loker Pinard and Michael Brigham Funds for Kidney Cancer Research at the Dana-Farber Cancer Institute. Dr. Kroeger thanks Dr. Klatte, Department of Urology, Medical University of Vienna, for statistical advice.
Renal cell carcinoma (RCC) is a heterogeneous disease consisting of several tumor types that have their own genetic, molecular, and clinical characteristics. Historical studies in the cytokine era by the University of California Los Angeles (UCLA) and the National Institutes of Health (NIH) described poor prognosis and worse treatment response to interleukin-2 (IL-2) in patients with metastatic involvement of the retroperitoneal lymph nodes [1] and [2]. Patients with and without retroperitoneal lymph node involvement had median overall survival (OS) times of 8.5 mo and 14.7 mo (p = 0.0004), respectively [2] . Pantuck et al [1] reported a better objective response rate in N0M1 compared to N+M1 patients (p = 0.01), with survival times of 10.5 mo and 20.4 mo, respectively. The adverse prognostic impact of retroperitoneal lymph node metastases (LNM) on survival outcome in patients with metastatic RCC (mRCC) was confirmed by several other institutions during the cytokine era [3], [4], and [5]. However, in the era of targeted therapies (TT), the survival outcome for patients with LNM has not yet been well characterized.
RCC preferentially metastasizes via hematogenous routes. Bianchi et al [6] recently described hematogenous metastatic sites in 80% of 11 157 patients with mRCC, while extension of the disease into lymph nodes was described only in 20% of patients. However, the authors did not provide details about the distribution of lymph node localization. The dominant regional LNM sites in RCC are retro- and paracaval, pre- and paraaortic, and interaortocaval lymph nodes in an anatomically intact retroperitoneum [7] . LNM can be unpredictable, and can also occur via direct extension to the thorax, supraclavicular lymph nodes, and iliac lymph nodes [7] . It has not been documented if particular sites of positive lymph node localization have a negative influence on the survival outcome of mRCC patients treated with TT.
Whether retroperitoneal lymph node dissection (LND) improves survival outcome is the subject of ongoing scientific debate. A phase 3 clinical trial (EORTC 30881) demonstrated positive lymph nodes in only 4% of investigated cases having no apparent involvement on computed tomography (CT) imaging. There was no advantage in OS, recurrence-free survival, or progression-free survival (PFS) for RCC patients who underwent LND [8] . The study has been criticized, however, because the majority of patients included had low-stage disease. Moreover, uniform surgical templates were not used for LND [9] . However, retrospective studies have demonstrated that LNM are more prevalent in high-risk patients with high Fuhrman grades, sarcomatoid features, locally advanced tumor stage, tumor size >10 cm, and tumor necrosis [10] . It has been suggested that LND is particularly beneficial in these high-risk patients [11] and [12]. Trends in the surgical management of RCC that have moved away from open radical surgery with extended LND to laparoscopic and robotic surgery with minimal hilar LND, combined with the negative results of phase 3 studies showing no additional benefit of LND over removal of the primary tumor alone, have decreased enthusiasm for performing extended LND. Therefore, it is common clinical practice to perform LND only in high-risk patients, and with a diagnostic rather than a curative intention [7] . Retrospective studies may be unable to add new insights into the current debate because only a prospective randomized trial with clear inclusion and exclusion criteria would be able to overcome this selection bias. Before planning such trials, however, it is prudent to better understand the impact of LNM and different lymph node localizations on the survival outcome for mRCC patients treated with current state-of-the-art therapies using data from the International mRCC Database Consortium (IMDC).
The IMDC database includes centers from North America (Canada, USA), Europe (Denmark, Greece, Belgium), Asia (Singapore, Japan, South Korea), and New Zealand. Data were collected from August 15, 2008 until December 31, 2013. At the time of analysis, the database contained data on 3405 patients who had received first-line targeted therapies between 2003 and 2013. The final study cohort comprised 2996 patients who were treated with first-line vascular endothelial growth factor (VEGF) inhibitors (n = 2823), mammalian target of rapamycin (mTOR) inhibitors (n = 165), or a combination of both (n = 8) therapies between April 2003 and August 2013. Patients were excluded from the analyses if they had received experimental therapies in first-line treatment (n = 7) or if no information was available on their lymph node status (n = 402; Fig. 1 ). Lymph node status was determined according to standard pathologic and CT criteria.
Information about LND was available for 890 patients. Full LND was performed in only 51 patients, and a limited (hilar) dissection was performed in 173 patients. These numbers were too low to allow any analyses and conclusions regarding the influence of LND on the survival outcome for mRCC patients treated with TT. Whether LND could be beneficial for mRCC patients was therefore not considered in the present study.
All centers obtained local Institutional Review Board approval before including data in the IMDC database. Baseline patient characteristics included demographic, clinicopathologic, and laboratory data, as previously described [13] . Survival data were retrospectively and prospectively collected from medical chart reviews and electronic records. Uniform data templates were used to ensure consistent data collection at each institution. The majority of patients were treated using standard care, but a subset received therapy as part of clinical trials. All patients were considered in consecutive series to avoid selection bias.
The primary hypothesis of the current study was that mRCC patients with LNM have worse survival outcome than patients without LNM when treated with TT. The secondary hypothesis was that worse survival outcome for mRCC patients with LNM is dictated by subdiaphragmatic (SPD) LNM rather than supradiaphragmatic (SBD) LNM.
Therefore, treatment response to current standard-of-care TT agents, OS, and PFS were compared between patients with LNM and those without LNM. We did not compare outcomes between the drugs used for treatment because these comparisons are frequently biased outside a clinical trial and there may not be sufficient power for such an analysis. Additional analyses were performed for subgroups of patients with exclusively SPD lymph node involvement, exclusively SBD lymph node involvement, or with lymph node involvement in both sites (SPD+/SBD+). Previous studies have demonstrated that tumor biology has an important impact on the survival outcome for mRCC patients [14] . Thus, the different biology of RCC was considered by separate analyses for clear cell RCC (ccRCC) and non-ccRCC (nccRCC) patients. Patients with SPD primarily had mediastinal lymph node involvement, while patients with SBD primarily had retroperitoneal lymph node involvement. Lymph node status was determined before initiation of TT. The selection process for patient groups analyzed in the study is shown in Figure 1 .
OS was defined as the time period between TT initiation and the date of death, with censoring on the day of the last follow-up visit. PFS was defined as the period between treatment initiation and progression, drug cessation, or death, with censoring at the last follow-up visit. Kaplan-Meier plots were used to estimate median OS and PFS, and univariate comparisons were performed using the log rank test. Uni- and multivariate (MVA) Cox regression analyses were performed to test the association of predefined prognostic factors and LNM with survival outcome. The analyses were performed with backward stepwise selection criteria, and significance was tested using the Wald statistic [13] . In MVA analyses, the IMDC (Heng) risk criteria were applied for adjustment [15] . The IMDC risk criteria include a time period from diagnosis to treatment of <1 yr, Karnofsky performance score <80%, anemia, hypercalcemia, thrombophilia, and neutrophilia. Upper and lower limits of normal for the laboratory parameters were based on institutional limits. In addition, adjustment was performed for bone and liver metastases because interactions of LNM with other metastatic sites should be considered. The prognostic relevance of bone and liver metastases has previously been demonstrated by our group [16] .
A secondary aim was to analyze baseline characteristics. Patient and tumor characteristics were compared using the Student t test for continuous variables, or the chi-square test and the Fisher exact test for categorical variables.
Statistical analyses were performed using SPSS version 22 (IBM, Chicago, IL, USA), and a two-sided value of p < 0.05 was considered statistically significant.
At the time of data analysis, 2525 (84.3%) patients had stopped taking their TT; the median time on TT was 6.0 mo (25th percentile, 2.7 mo; 75th percentile, 13.0 mo). Best overall response data to first-line therapy were available for 2559 patients. There were no significant differences in response to first-line TT (p = 0.059) when patients were compared without consideration of RCC subtype or LNM localization ( Table 1 ).
Table 1 Treatment response to targeted therapies in patients with lymph node metastases (LNM) compared to those without LNM
| Best response | |||||
|---|---|---|---|---|---|
| CR | PR | SD | PD | ||
| All RCC subtypes | |||||
| No LMN | n | 14 | 304 | 556 | 290 |
| % | 1.2 | 26.1 | 47.8 | 24.9 | |
| LMN | n | 9 | 314 | 690 | 382 |
| % | 0.6 | 22.5 | 49.5 | 27.4 | |
| Clear cell RCC | |||||
| No LNM | n | 3 | 108 | 261 | 95 |
| % | 0.6 | 23.1 | 55.9 | 20.3 | |
| SPD LNM | n | 2 | 73 | 149 | 55 |
| % | 0.7 | 26.2 | 53.4 | 19.7 | |
| SBD LNM | n | 2 | 50 | 92 | 42 |
| % | 1.1 | 26.9 | 49.5 | 22.6 | |
| SPD+/SBD+ LNM | n | 0 | 28 | 58 | 36 |
| % | 0.0 | 23.0 | 47.5 | 29.5 | |
| Non–clear cell RCC | |||||
| No LNM | n | 0 | 3 | 21 | 21 |
| % | 0.0 | 6.7 | 46.7 | 46.7 | |
| SPD LNM | n | 0 | 3 | 12 | 3 |
| % | 0.0 | 16.7 | 66.7 | 16.7 | |
| SBD LNM | n | 0 | 7 | 24 | 19 |
| % | 0.0 | 14.0 | 48.0 | 38.0 | |
| SPD+/SBD+ LNM | n | 1 | 4 | 10 | 6 |
| % | 4.8 | 19.0 | 47.6 | 28.6 | |
CR = complete remission; PR = partial response; SD = stable disease; PD = progressive disease; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm.
Both response data and LNM localization data were available for 1054 ccRCC patients and 134 nccRCC patients. Comparison of data by LNM localization revealed no difference in best response to first-line therapies for either ccRCC (p = 0.457) or nccRCC (p = 0.192; Table 1 ).
Patients with LNM generally had worse PFS compared to those with no LNM without considering LNM localization or RCC subtype (median PFS [95% confidence interval] 7.8 mo [7.1–8.5 mo] vs 6.1 [5.6–6.6 mo], log rank p = 0.001; Fig. 2 A).
Fig. 2 (A) Comparison of progression-free survival (PFS) in patients with positive lymph nodes versus patients without positive lymph nodes (median 7.8 mo vs. 6.1 mo) without consideration of histologic subtype. Patients with lymph node metastases (LNM) had worse PFS (hazard ratio [HR] 1.14, 95% confidence interval [CI] 1.06–1.24; p = 0.001). (B) Comparison of PFS according to lymph node localization. Patients without LNM had the best PFS (8.8 mo), and patients with both supradiaphragmatic (SPD) and subdiaphragmatic (SBD) LNM localizations (SPD+/SBD+) had the worst PFS (5.5 mo). Patients with SBD LNM had significantly worse PFS compared to patients with no LNM (6.8 vs 8.8 mo; p = 0.003) and those with SPD LNM (6.8 vs 8.3 mo; p = 0.022). Any metastatic involvement of retroperitoneal lymph nodes, either alone (HR 1.25, 95% CI 1.08–1.45) or in combination with SPD+/SBD+LNM (HR 1.50, 95% CI 1.24–1.74; p < 0.001) was associated with higher risk of progression. By contrast, SPD LNM alone was not a prognostic factor for progression (HR 1.03, 95% 0.895–1.18; p = 0.695). (C) PFS by lymph node localization in clear cell renal cell carcinoma (ccRCC; median PFS: no nodes, 9.9 mo; SPD LNM, 8.7 mo; SBD LNM, 8.0 mo; and both, 5.6 mo). Patients with SPD LNM (HR 1.09, 95% CI 0.94–1.25) and SBD LNM (HR 1.04, 95% CI 0.96–1.14) had similar PFS, and this was not significantly worse than the PFS of patients without LNM (p = 0.265 and p = 0.333 for SPD and SBD LNM, respectively).
Comparison of PFS rates according to LNM sites revealed that patients with SBD LNM had significantly shorter PFS (6.8 mo [5.7–8.0 mo]) compared to those without LNM (8.8 mo [7.6–10.0 mo]; p = 0.003) or SPD LNM (8.3 mo [7.2–9.6 mo]; p = 0.022). The worst PFS was observed in patients with SPD+/SBD+ LNM (5.4 mo [4.6–6.4 mo]), but this did not significantly differ from PFS in patients with SBD LNM (p = 0.093; Fig. 2 B).
In ccRCC, however, comparison according to LNM localization revealed no significant differences in PFS between patients without LNM (9.9 mo [8.6–11.1 mo]) and those with SPD LNM (8.7 mo [7.3–10.1 mo]) or SBD LNM (8.0 mo [6.9–9.2 mo]) (p = 0.312; Fig. 2 C). SPD+/SBD+ LNM patients had the worst survival outcome (5.6 mo [4.0–7.2 mo]) in comparison to patients without LNM (p < 0.001), with SPD LNM (p = 0.006), and with SBD LNM (p = 0.035; Fig. 2 C). In nccRCC, there was no difference in PFS between patients with no LNM (3.9 mo [2.9–5.0 mo]), SPD LNM (4.6 mo [0.0–9.9 mo]), SBD LNM (5.4 mo [4.7–6.0 mo]), and SPD+/SBD+ LNM (5.4 mo [3.2–7.3 mo]).
In MVA analyses, LNM remained an independent prognostic factor (HR 1.13 [1.03–1.24]; p = 0.013) when adjusted for IMDC risk criteria, as well as liver and bone metastases ( Table 2 ) [16] . In subgroup analysis, SBD LNM (HR 1.31 [1.08–1.60]; p = 0.006) and SPD+/SBD LNM (HR 1.29 [1.04–1.61]; p = 0.022) were independent prognostic factors for PFS, while SPD LNM was not independently associated with PFS (p = 1.00; Table 2 ). In ccRCC, only SPD+/SBD+ LNM (p = 0.025) was independently associated with PFS ( Table 2 ).
Table 2 Multivariate analyses for association between lymph node metastases (LNM) and progression-free survival (PFS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time <1 yr | 1.33 | 1.21–1.47 | <0.001 |
| Low hemoglobin | 1.37 | 1.24–1.52 | <0.001 |
| Neutrophilia | 1.43 | 1.25–1.63 | <0.001 |
| Thrombocytosis | 1.16 | 1.03–1.32 | 0.019 |
| Karnofsy PS <80% | 1.68 | 1.50–1.88 | <0.001 |
| LNM | 1.13 | 1.03–1.24 | 0.013 |
| Liver metastases | 1.26 | 1.12–1.41 | <0.001 |
| Bone metastases | 1.27 | 1.15–1.40 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.19–1.61 | <0.001 |
| Low hemoglobin | 1.42 | 1.21–1.67 | <0.001 |
| Neutrophilia | 1.46 | 1.19–1.80 | <0.001 |
| Thrombocytosis | 1.22 | 1.01–1.47 | 0.045 |
| Karnofsy PS <80% | 1.94 | 1.65–2.28 | <0.001 |
| Liver metastases | 1.39 | 1.18–1.65 | <0.001 |
| Bone metastases | 1.45 | 1.25–1.68 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.83–1.20 | 0.990 |
| SBD LNM | 1.32 | 1.08–1.60 | 0.006 |
| SPD+/SBD+ LNM | 1.29 | 1.04–1.61 | 0.022 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.17–1.64 | <0.001 |
| Low hemoglobin | 1.43 | 1.21–1.69 | <0.001 |
| Neutrophilia | 1.38 | 1.09–1.76 | 0.009 |
| Thrombocytosis | 1.97 | 1.64–2.38 | <0.001 |
| Karnofsy PS <80% | 1.51 | 1.24–1.85 | <0.001 |
| Bone metastases | 1.64 | 1.38–1.94 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.82–1.22 | 0.998 |
| SBD LNM | 1.23 | 0.98–1.55 | 0.077 |
| SPD+/SBD+ LNM | 1.33 | 1.04–1.71 | 0.025 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm; ccRCC = clear cell RCC.
At the time of the analysis, 1960 patients had died; the median OS time for the total patient cohort was 20.0 mo (25th percentile, 42.3 mo; 75th percentile, 7.9 mo). Patients with metastatic spread to lymph nodes had a worse survival outcome than patients without LNM (24.0 mo [21.6–26.3 mo] vs 16.0 mo [14.9–17.1 mo], p < 0.001; Fig. 3 A).
Fig. 3 (A) Patients with lymph node metastases (LNM) had a higher risk of death (median overall survival [OS] 23.9 mo, 95% confidence interval [CI] 21.6–26.3 mo) than those without LNM (median OS 16.0 mo, 95% CI 14.9–17.1; hazard ratio [HR] 1.41, 95% CI 1.28–1.54; p < 0.001). (B) A comparison without consideration of the histologic subtypes demonstrates that any lymph node involvement was associated with worse survival outcome compared to patients without LNM. Median survival times were 25.2 mo, 20.3 mo, 16.2 mo, and 11.5 mo in patients with no LNM and lymph node involvement in supradiaphragmatic (SPD), subdiaphragmatic (SBD), and both localizations (SPD+/SBD+), respectively. The HRs (95% CIs) for patients with SPD, SBD, and SPD+/SBD+ LNM were 1.24 (1.06–1.44), 1.45 (1.23–1.71), and 1.93 (1.60–2.33), respectively. (C) In clear cell renal cell carcinoma (ccRCC), patients with SPD and SBD LNM had similar median OS (SPD, 21.6 mo; SBD, 19.8 mo). However, both groups had a significantly higher risk of death (SPD: HR 1.24, 95% CI 1.05–1.47; SBD: HR 1.28, 95% CI 1.05–1.56) in comparison to those without lymph nodes. Patients with SPD+/SBD+ metastases had the worst OS (12.85 mo) and a twofold higher risk of death than patients without lymph node involvement (HR 2.04, 95% CI 1.68–2.48). (D) In non–clear cell RCC (nccRCC), median OS times according to lymph node localization were 15.8 mo for no lymph nodes, 16.1 mo for SPD LNM, 11.2 mo for SBD LNM, and 12.5 mo for SPD+/SBD+ LNM. There were no significant differences between the groups (log rank p = 0.444).
In subgroup analyses, patients without LNM had a median OS of 25.2 mo (21.9–28.5 mo) compared to 20.3 mo (17.1–23.5 mo) for SPD LNM (p = 0.093), 16.2 mo (13.5–20.0 mo) for SBD LNM (p < 0.001), and 11.5 mo (9.8–13.3 mo) (p < 0.001) for SPD+/SBD+ LNM ( Fig. 3 B).
In ccRCC, further analysis considering LNM localization revealed similar OS times for patients with SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13]), and were shorter than for patients without LNM (26.97 mo [23.32– 30.62], p = 0.01; Fig. 3 C). Patients with SPD+/SBD+ LNM had significantly (p = 0.001) worse OS (12.85 mo [9.45–16.24 mo]) compared to SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13 mo]; Fig. 3 C). In nccRCC, analyses according to LNM revealed no significant differences in OS (p = 0.444; Fig. 3 C).
For all histological subtypes, LNM were an independent prognostic factor for adverse OS in MVA analysis (p < 0.001; Table 3 ). Additional MVA according to LNM sites revealed that SBD LNM (p < 0.001) and SPD+/SBD+ LNM (p < 0.001) were independent prognostic factors for OS. SPD LNM were not independently associated with OS (p = 0.999; Table 3 ). In ccRCC, MVA according to LNM localization confirmed an independent association between SBD LNM (p = 0.028) and SPD+/SBD+ LNM (p = 0.006) and OS ( Table 3 ).
Table 3 Multivariate Cox regression analysis of association between lymph node metastases (LNM) and overall survival (OS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time < 1 yr | 1.34 | 1.19–1.49 | <0.001 |
| Low hemoglobin | 1.51 | 1.31–1.70 | <0.001 |
| High Ca | 1.47 | 1.25–1.73 | <0.001 |
| Neutrophilia | 1.67 | 1.45–1.92 | <0.001 |
| Thrombophilia | 1.31 | 1.14–1.51 | <0.001 |
| Karnofsky PS <80% | 1.88 | 1.67–2.13 | <0.001 |
| Liver metastases | 1.50 | 1.32–1.70 | <0.001 |
| Bone metastases | 1.50 | 1.34–1.67 | <0.001 |
| LNM | 1.33 | 1.19–1.48 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time < 1 yr | 1.32 | 1.11–1.56 | 0.001 |
| Low hemoglobin | 1.45 | 1.21–1.73 | <0.001 |
| High Ca | 1.38 | 1.08–1.77 | 0.010 |
| Neutrophilia | 1.78 | 1.43–2.22 | <0.001 |
| Thrombophilia | 1.49 | 1.21–1.83 | <0.001 |
| Karnofsky PS <80% | 2.19 | 1.82–2.62 | <0.001 |
| Liver metastases | 1.84 | 1.53–2.20 | <0.001 |
| Bone metastases | 1.61 | 1.36–1.90 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.57 | 0.98–1.48 | 0.075 |
| SBD LNM | 1.59 | 1.26–1.96 | <0.001 |
| SPD+/SBD+ LNM | 1.32 | 1.25–2.02 | <0.001 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time < 1 yr | 1.36 | 1.12–1.64 | 0.002 |
| Low hemoglobin | 1.36 | 1.11–1.67 | 0.003 |
| High Ca | 1.44 | 1.10–1.89 | 0.009 |
| Neutrophilia | 1.65 | 1.27–2.13 | <0.001 |
| Thrombophilia | 1.52 | 1.19–1.91 | 0.001 |
| Karnofsky PS <80% | 2.12 | 1.72–2.61 | <0.001 |
| Liver metastases | 1.94 | 1.57–2.41 | <0.001 |
| Bone metastases | 1.84 | 1.53–2.22 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.13 | 0.90–1.42 | 0.301 |
| SBD LNM | 1.34 | 1.03–1.74 | 0.028 |
| SPD+/SBD+ LNM | 1.47 | 1.11–1.93 | 0.006 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm; ccRCC = clear cell RCC.
An overall comparison of patients with and without LNM is shown in Table 4 . Patients with LNM presented more often with nccRCC (p < 0.001) and higher Fuhrman grades (p = 0.007), and had less often undergone prior nephrectomy (p < 0.001). The IMDC risk factors of high calcium (p = 0.001), thrombocytosis (p < 0.001), neutrophilia (p < 0.001), and Karnofsky performance status <80% (p = 0.009) were all more frequently seen in patients with LNM. Comparison of additional metastatic sites revealed that LNM are related to a higher overall metastatic burden (one or more additional metastatic site p < 0.001).
Table 4 General comparison of clinicopathologic features of patients with lymph node metastases (LNM) compared to those without LNM a
| No LNM (n = 1380) |
LNM (n = 1616) |
p value | |
|---|---|---|---|
| Mean age, yr (SD) | 58.65 (11.36) | 58. 50 (11.30) | 0.716 |
| Male, n (%) | 1003 (72.7) | 1182 (73.1) | 0.805 |
| Nephrectomy, n (%) | 1134 (82.4) | 1218 (75.4) | <0.001 |
| Fuhrman grade n (%) | 0.007 | ||
| 1 | 44 (4.2) | 33 (2.8) | |
| 2 | 302 (28.6) | 269 (23.1) | |
| 3 | 426 (40.3) | 501 (43.1) | |
| 4 | 284 (26.9) | 359 (30.9) | |
| Non–clear cell RCC, n (%) | 127 (9.8) | 253 (17.4) | <0.001 |
| Sarcomatoid features, n (%) | 143 (12.0) | 170 (12.0) | 0.586 |
| Therapy class, n (%) | |||
| AntiVEGF | 1316 (95.4) | 1507 (93.3) | |
| mTOR | 60 (4.3) | 105 (6.5) | |
| Combination | 4 (0.3) | 4 (0.2) | |
| DTT <1 yr, n (%) | 736 (53.4) | 905 (56.0) | 0.151 |
| Low HB (<LLN), n (%) | 691 (56.0) | 877 (58.1) | 0.261 |
| High calcium, n (%) | 98 (8.6) | 179 (12.6) | 0.001 |
| Neutrophilia, n (%) | 154 (13.1) | 286 (19.3) | <0.001 |
| Thrombocytosis, n (%) | 172 (15.4) | 299 (22.4) | <0.001 |
| Karnofsky PS <80%, n (%) | 261 (21.5) | 372 (25.9) | 0.009 |
| Lung metastases, n (%) | 939 (68.0) | 1149 (72.5) | 0.016 |
| Brain metastases, n (%) | 100 (7.2) | 141 (9.4) | 0.009 |
| Liver metastases, n (%) | 278 (20.2) | 328 (21.7) | 0.337 |
| Bone metastases, n (%) | 481 (34.9) | 551 (35.9) | 0.462 |
| Other metastic sites, n (%) | 470 (34.1) | 643 (43.1) | <0.001 |
| >1 metastatic site, n (%) | 805 (58.3%) | 1446 (89.5%) | <0.001 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; mTOR = mammalian target of rapamycin; PS = performance status; RCC = renal cell carcinoma; SD = standard deviation; VEGF = vascular endothelial growth factor.
The LNM localization distribution and clinicopathologic features in ccRCC and nccRCC are shown in Table 5 . In ccRCC, IMDC risk factors of neutrophilia (p = 0.049) and thrombophilia (p = 0.016) were more common. Patients with LNM had undergone nephrectomy less frequently (p = 0.001). In patients with ccRCC, those with LNM had concomitant lung (p < 0.001) and other metastases (eg, adrenal glands, other soft tissues; p = 0.009) more often than those without LNM; there was no difference in the frequency of bone (p = 0.276), liver (p = 0.943), or brain metastases (p = 0.187).
Table 5 Distribution of metastatic lymphatic sites in relation to the diaphragm in clear cell and non–clear cell renal carcinoma (RCC) a
| Clear cell RCC (n = 1218) |
Non–clear cell RCC (n = 159) |
|
|---|---|---|
| No LNM, n (%) | 552 (45.6) | 52 (32.7) |
| SPD LNM, n (%) | 321 (26.6) | 22 (13.8) |
| SBD LNM, n (%) | 204 (16.9) | 54 (34.0) |
| SPD+/SBD+ LNM, n (%) | 132 (10.9) | 31 (19.5) |
| LMN | No LMN | p value | LMN | No LMN | p value | |
|---|---|---|---|---|---|---|
| Nephrectomy, n (%) | 542 (81.2) | 492 (88.2) | 0.001 | 82 (76.6) | 43 (82.7) | 0.418 |
| Male, n (%) | 515 (77.7) | 403 (72.6) | 0.045 | 75 (70.1) | 37 (71.2) | 1.00 |
| DTT <1 yr, n (%) | 348 (52.5) | 265 (47.9) | 0.120 | 65 (60.7) | 29 (55.8) | 0.607 |
| Low HB (<LLN) | 331 (54.0) | 233 (48.7) | 0.088 | 49 (51.0) | 19 (45.2) | 0.582 |
| High calcium, n (%) | 70 (12.3) | 40 (9.5) | 0.184 | 5 (5.7) | 4 (12.1) | 0.254 |
| Neutrophilia, n (%) | 95 (15.9) | 53 (11.6) | 0.049 | 24 (25.3) | 7 (16.7) | 0.376 |
| Thrombophilia, n (%) | 106 (22.0) | 63 (15.6) | 0.016 | 17 (21.0) | 10 (25.6) | 0.642 |
| Karnofsky PS <80%, n (%) | 145 (24.7) | 95 (20.0) | 0.076 | 24 (24.7) | 10 (20.4) | 0.680 |
| Sarcomatoid features, n (%) | 52 (9.1) | 62 (12.6) | 0.074 | 14 (15.4) | 7 (15.6) | 1.00 |
| Lung metastases, n (%) | 507 (76.8) | 372 (67.0) | <0.001 | 52 (49.1) | 20 (38.5) | 0.237 |
| Brain metastases, n (%) | 65 (9.8) | 42 (7.6) | 0.187 | 4 (3.8) | 2 (3.8) | 1.00 |
| Liver metastases, n (%) | 132 (20.0) | 112 (20.2) | 0.943 | 32 (30.2) | 28 (53.8) | 0.005 |
| Bone metastases, n (%) | 220 (33.3) | 202 (36.4) | 0.276 | 35 (33.0) | 20 (38.5) | 0.594 |
| Other metastatic sites, n (%) | 287 (45.7) | 211 (38.2) | 0.009 | 40 (40.8) | 20 (38.5) | 0.862 |
| >1 metastatic site, n (%) | 602 (90.8) | 320 (57.7) | <0.001 | 87 (81.3) | 30 (7.7) | 0.002 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; PS = performance status; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm.
In nccRCC, liver metastases were found more often in patients without LNM (p = 0.005). A comparison of other clinicopathologic features between nccRCC patients with and without LNM revealed no significant differences ( Table 5 ).
It has previously been shown that only a minority of patients with kidney cancer undergo LND at the time of nephrectomy. This is the case even for high-risk patients [17] , even though retrospective studies have suggested a therapeutic benefit of LND [11] and [18]. It remains uncertain whether extended LND can improve the survival outcome for mRCC patients. However, if it were shown that SBD LNM affect the survival outcome for mRCC patients, this could form the basis for a rationale in favor of performing LND, or at least justify the need for prospective clinical trials to address this question. In this regard, an understanding of the influence of LNM on treatment response and survival outcome for mRCC patients undergoing TT could make an important contribution to the debate on surgical management of RCC.
Our study has demonstrated that LNM are associated with advanced IMDC risk factors, are an adverse prognostic factor for PFS and OS, and are associated with a higher metastatic burden. Furthermore, subanalyses showed that these differences depend on LNM localization.
Different cellular clones of RCC appear to differ in their propensity to spread to different organ sites [19] ; however, whether these differences impact on the survival outcome of mRCC patients is not completely understood. A previous IMDC study highlighted the prognostic importance of bone and liver metastases for survival outcome in mRCC patients, and suggested that aggressive RCC subclones tend to metastasize to these organs [16] . In addition, another study by our group showed that metastatic spread to other organs such as the pancreas or soft tissue tends to develop as a late clinical symptom, and has favorable treatment response and survival outcomes [20] . While these tumors may represent a favorable RCC subtype, our current results suggest that RCCs that spread to regional retroperitoneal lymph nodes represent another aggressive RCC subtype that develops bone metastases.
It has been suggested that LNM in ccRCC occur independently of von Hippel Lindau (VHL) gene inactivation [21] . Current agents used for mRCC treatment modulate pathways that are dysregulated because of VHL inactivation. This would imply that LNM might be less amenable to treatment with VEGF and mTOR inhibitors. However, our findings demonstrated no significant differences in best response to first-line therapies in patients with LNM versus those without LNM in both ccRCC and nccRCC patient groups. Nonetheless, patients with LNM had significantly shorter PFS than patients without LNM. While there was only a trend for an independent association between SBD LNM and short PFS, SPD+/SBD+ LNM were independently related to shorter PFS. Patients with SPD+/SBD+ LNM had additional hematogenous metastatic sites in 97% of cases. Thus, the high metastatic burden may explain the shorter response duration of first-line agents in this group of patients. However, a shorter duration for treatment success could also support the hypothesis that LNM are caused by RCC clones that do not have an inactive VHL gene function.
In the cytokine era, LNM were associated with shorter OS [1] and [2]. Our study demonstrates similar findings in the TT era. Interestingly, a worse survival outcome was observed for SBD LNM, both when occurring exclusively and in combination with thoracic lymph node involvement. Moreover, SBD LNM remained an independent prognostic factor for PFS and OS when analyzed without consideration of histologic subtype, and were an independent prognostic factor for OS in ccRCC. Conversely, SPD LNM were not an independent prognostic factor. This is notable and underscores the unique nature of RCCs that spread via retroperitoneal lymph nodes. Because the best response to TT did not different between SBD and SPD LNM, it appears that survival outcome is influenced by numerous factors. Patients with SBD LNM had faster disease progression before treatment initiation than those with SPD LNM (diagnosis to treatment <1 yr; 63% vs 46%; p < 0.001; data not shown). Despite a best response that was broadly similar, it seems that the disease progresses faster in SBD LNM patients during the later clinical course than in SPD LNM patients. In addition, SBD LNM were more often associated with anemia compared to SPD LNM (59% vs 48%; p = 0.008; data not shown). Anemia has multifactorial causes in cancer patients. However, one of the most important contributing factors is immune system dysregulation [22] . Dysregulation of the immune system may also be the reason for metastatic spread to SBD sites. Collectively, while risk factors that are currently unknown appear to dictate the clinical course in patients with SPD metastases, SBD metastases are independently associated with PFS and OS.
The majority of patients analyzed in this study did not undergo formal LND. Therefore, the effect of LND cannot be determined retrospectively. However, the current results indicate that SBD lymph nodes may have substantial negative effects on survival outcome and it is worth speculating whether extended removal of these metastatic sites could improve the clinical outcome for mRCC patients. The IMDC has recently shown that cytoreductive nephrectomy may provide therapeutic benefit in a large number of patients [23] . With regard to our current findings, it would be worth determining the role of LND in these patients.
The current study has several limitations that should be considered when interpreting the results. The IMDC database comprises retrospectively collected data for consecutive series of mRCC patients. These individual series, moreover, do not have standardized follow-up protocols, but this limitation may more accurately reflect real world outcomes. In addition, there was no centralized pathologic or radiologic review. Finally, lymph node status was determined clinically according to CT images in a number of cases, and it has been demonstrated that clinical and pathologic lymph node status is not necessarily always correlated.
LNM are associated with poor prognosis and adverse prognostic IMDC risk factors in mRCC patients treated with TT. Much of this increased risk associated with LNM appears to be dictated by SBD localization rather than SPD localization. Whether LND could improve the outcome for this group of patients remains unknown, but it is an important clinical question requiring prospective trials.
Author contributions: Daniel Y. Heng had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kroeger, Heng.
Acquisition of data: all authors.
Analysis and interpretation of data: Kroeger, Heng, Choueiri.
Drafting of the manuscript: Kroeger, Heng, Choueiri.
Critical revision of the manuscript for important intellectual content: all authors.
Statistical analysis: Kroeger, Heng.
Obtaining funding: None.
Administrative, technical, or material support: all authors.
Supervision: Choueiri, Heng.
Other (specify): None.
Financial disclosures: Daniel Y. Heng certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Benoit Beuselinck is an investigator of the EudraCT: 2011-006085-40/MetaSun trial supported by Pfizer. Frede Donskov has received research support from Novartis and GlaxoSmithKline. Toni K. Choueiri has received research funding from Pfizer and has an advisory role at Pfizer, GSK, Novartis, and Bayer. All other authors declare no conflicts of interest to the current study.
Funding/Support and role of the sponsor: None.
Acknowledgments: Dr. Choueiri is supported in part by the Trust family and the Loker Pinard and Michael Brigham Funds for Kidney Cancer Research at the Dana-Farber Cancer Institute. Dr. Kroeger thanks Dr. Klatte, Department of Urology, Medical University of Vienna, for statistical advice.
Renal cell carcinoma (RCC) is a heterogeneous disease consisting of several tumor types that have their own genetic, molecular, and clinical characteristics. Historical studies in the cytokine era by the University of California Los Angeles (UCLA) and the National Institutes of Health (NIH) described poor prognosis and worse treatment response to interleukin-2 (IL-2) in patients with metastatic involvement of the retroperitoneal lymph nodes [1] and [2]. Patients with and without retroperitoneal lymph node involvement had median overall survival (OS) times of 8.5 mo and 14.7 mo (p = 0.0004), respectively [2] . Pantuck et al [1] reported a better objective response rate in N0M1 compared to N+M1 patients (p = 0.01), with survival times of 10.5 mo and 20.4 mo, respectively. The adverse prognostic impact of retroperitoneal lymph node metastases (LNM) on survival outcome in patients with metastatic RCC (mRCC) was confirmed by several other institutions during the cytokine era [3], [4], and [5]. However, in the era of targeted therapies (TT), the survival outcome for patients with LNM has not yet been well characterized.
RCC preferentially metastasizes via hematogenous routes. Bianchi et al [6] recently described hematogenous metastatic sites in 80% of 11 157 patients with mRCC, while extension of the disease into lymph nodes was described only in 20% of patients. However, the authors did not provide details about the distribution of lymph node localization. The dominant regional LNM sites in RCC are retro- and paracaval, pre- and paraaortic, and interaortocaval lymph nodes in an anatomically intact retroperitoneum [7] . LNM can be unpredictable, and can also occur via direct extension to the thorax, supraclavicular lymph nodes, and iliac lymph nodes [7] . It has not been documented if particular sites of positive lymph node localization have a negative influence on the survival outcome of mRCC patients treated with TT.
Whether retroperitoneal lymph node dissection (LND) improves survival outcome is the subject of ongoing scientific debate. A phase 3 clinical trial (EORTC 30881) demonstrated positive lymph nodes in only 4% of investigated cases having no apparent involvement on computed tomography (CT) imaging. There was no advantage in OS, recurrence-free survival, or progression-free survival (PFS) for RCC patients who underwent LND [8] . The study has been criticized, however, because the majority of patients included had low-stage disease. Moreover, uniform surgical templates were not used for LND [9] . However, retrospective studies have demonstrated that LNM are more prevalent in high-risk patients with high Fuhrman grades, sarcomatoid features, locally advanced tumor stage, tumor size >10 cm, and tumor necrosis [10] . It has been suggested that LND is particularly beneficial in these high-risk patients [11] and [12]. Trends in the surgical management of RCC that have moved away from open radical surgery with extended LND to laparoscopic and robotic surgery with minimal hilar LND, combined with the negative results of phase 3 studies showing no additional benefit of LND over removal of the primary tumor alone, have decreased enthusiasm for performing extended LND. Therefore, it is common clinical practice to perform LND only in high-risk patients, and with a diagnostic rather than a curative intention [7] . Retrospective studies may be unable to add new insights into the current debate because only a prospective randomized trial with clear inclusion and exclusion criteria would be able to overcome this selection bias. Before planning such trials, however, it is prudent to better understand the impact of LNM and different lymph node localizations on the survival outcome for mRCC patients treated with current state-of-the-art therapies using data from the International mRCC Database Consortium (IMDC).
The IMDC database includes centers from North America (Canada, USA), Europe (Denmark, Greece, Belgium), Asia (Singapore, Japan, South Korea), and New Zealand. Data were collected from August 15, 2008 until December 31, 2013. At the time of analysis, the database contained data on 3405 patients who had received first-line targeted therapies between 2003 and 2013. The final study cohort comprised 2996 patients who were treated with first-line vascular endothelial growth factor (VEGF) inhibitors (n = 2823), mammalian target of rapamycin (mTOR) inhibitors (n = 165), or a combination of both (n = 8) therapies between April 2003 and August 2013. Patients were excluded from the analyses if they had received experimental therapies in first-line treatment (n = 7) or if no information was available on their lymph node status (n = 402; Fig. 1 ). Lymph node status was determined according to standard pathologic and CT criteria.
Information about LND was available for 890 patients. Full LND was performed in only 51 patients, and a limited (hilar) dissection was performed in 173 patients. These numbers were too low to allow any analyses and conclusions regarding the influence of LND on the survival outcome for mRCC patients treated with TT. Whether LND could be beneficial for mRCC patients was therefore not considered in the present study.
All centers obtained local Institutional Review Board approval before including data in the IMDC database. Baseline patient characteristics included demographic, clinicopathologic, and laboratory data, as previously described [13] . Survival data were retrospectively and prospectively collected from medical chart reviews and electronic records. Uniform data templates were used to ensure consistent data collection at each institution. The majority of patients were treated using standard care, but a subset received therapy as part of clinical trials. All patients were considered in consecutive series to avoid selection bias.
The primary hypothesis of the current study was that mRCC patients with LNM have worse survival outcome than patients without LNM when treated with TT. The secondary hypothesis was that worse survival outcome for mRCC patients with LNM is dictated by subdiaphragmatic (SPD) LNM rather than supradiaphragmatic (SBD) LNM.
Therefore, treatment response to current standard-of-care TT agents, OS, and PFS were compared between patients with LNM and those without LNM. We did not compare outcomes between the drugs used for treatment because these comparisons are frequently biased outside a clinical trial and there may not be sufficient power for such an analysis. Additional analyses were performed for subgroups of patients with exclusively SPD lymph node involvement, exclusively SBD lymph node involvement, or with lymph node involvement in both sites (SPD+/SBD+). Previous studies have demonstrated that tumor biology has an important impact on the survival outcome for mRCC patients [14] . Thus, the different biology of RCC was considered by separate analyses for clear cell RCC (ccRCC) and non-ccRCC (nccRCC) patients. Patients with SPD primarily had mediastinal lymph node involvement, while patients with SBD primarily had retroperitoneal lymph node involvement. Lymph node status was determined before initiation of TT. The selection process for patient groups analyzed in the study is shown in Figure 1 .
OS was defined as the time period between TT initiation and the date of death, with censoring on the day of the last follow-up visit. PFS was defined as the period between treatment initiation and progression, drug cessation, or death, with censoring at the last follow-up visit. Kaplan-Meier plots were used to estimate median OS and PFS, and univariate comparisons were performed using the log rank test. Uni- and multivariate (MVA) Cox regression analyses were performed to test the association of predefined prognostic factors and LNM with survival outcome. The analyses were performed with backward stepwise selection criteria, and significance was tested using the Wald statistic [13] . In MVA analyses, the IMDC (Heng) risk criteria were applied for adjustment [15] . The IMDC risk criteria include a time period from diagnosis to treatment of <1 yr, Karnofsky performance score <80%, anemia, hypercalcemia, thrombophilia, and neutrophilia. Upper and lower limits of normal for the laboratory parameters were based on institutional limits. In addition, adjustment was performed for bone and liver metastases because interactions of LNM with other metastatic sites should be considered. The prognostic relevance of bone and liver metastases has previously been demonstrated by our group [16] .
A secondary aim was to analyze baseline characteristics. Patient and tumor characteristics were compared using the Student t test for continuous variables, or the chi-square test and the Fisher exact test for categorical variables.
Statistical analyses were performed using SPSS version 22 (IBM, Chicago, IL, USA), and a two-sided value of p < 0.05 was considered statistically significant.
At the time of data analysis, 2525 (84.3%) patients had stopped taking their TT; the median time on TT was 6.0 mo (25th percentile, 2.7 mo; 75th percentile, 13.0 mo). Best overall response data to first-line therapy were available for 2559 patients. There were no significant differences in response to first-line TT (p = 0.059) when patients were compared without consideration of RCC subtype or LNM localization ( Table 1 ).
Table 1 Treatment response to targeted therapies in patients with lymph node metastases (LNM) compared to those without LNM
| Best response | |||||
|---|---|---|---|---|---|
| CR | PR | SD | PD | ||
| All RCC subtypes | |||||
| No LMN | n | 14 | 304 | 556 | 290 |
| % | 1.2 | 26.1 | 47.8 | 24.9 | |
| LMN | n | 9 | 314 | 690 | 382 |
| % | 0.6 | 22.5 | 49.5 | 27.4 | |
| Clear cell RCC | |||||
| No LNM | n | 3 | 108 | 261 | 95 |
| % | 0.6 | 23.1 | 55.9 | 20.3 | |
| SPD LNM | n | 2 | 73 | 149 | 55 |
| % | 0.7 | 26.2 | 53.4 | 19.7 | |
| SBD LNM | n | 2 | 50 | 92 | 42 |
| % | 1.1 | 26.9 | 49.5 | 22.6 | |
| SPD+/SBD+ LNM | n | 0 | 28 | 58 | 36 |
| % | 0.0 | 23.0 | 47.5 | 29.5 | |
| Non–clear cell RCC | |||||
| No LNM | n | 0 | 3 | 21 | 21 |
| % | 0.0 | 6.7 | 46.7 | 46.7 | |
| SPD LNM | n | 0 | 3 | 12 | 3 |
| % | 0.0 | 16.7 | 66.7 | 16.7 | |
| SBD LNM | n | 0 | 7 | 24 | 19 |
| % | 0.0 | 14.0 | 48.0 | 38.0 | |
| SPD+/SBD+ LNM | n | 1 | 4 | 10 | 6 |
| % | 4.8 | 19.0 | 47.6 | 28.6 | |
CR = complete remission; PR = partial response; SD = stable disease; PD = progressive disease; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm.
Both response data and LNM localization data were available for 1054 ccRCC patients and 134 nccRCC patients. Comparison of data by LNM localization revealed no difference in best response to first-line therapies for either ccRCC (p = 0.457) or nccRCC (p = 0.192; Table 1 ).
Patients with LNM generally had worse PFS compared to those with no LNM without considering LNM localization or RCC subtype (median PFS [95% confidence interval] 7.8 mo [7.1–8.5 mo] vs 6.1 [5.6–6.6 mo], log rank p = 0.001; Fig. 2 A).
Fig. 2 (A) Comparison of progression-free survival (PFS) in patients with positive lymph nodes versus patients without positive lymph nodes (median 7.8 mo vs. 6.1 mo) without consideration of histologic subtype. Patients with lymph node metastases (LNM) had worse PFS (hazard ratio [HR] 1.14, 95% confidence interval [CI] 1.06–1.24; p = 0.001). (B) Comparison of PFS according to lymph node localization. Patients without LNM had the best PFS (8.8 mo), and patients with both supradiaphragmatic (SPD) and subdiaphragmatic (SBD) LNM localizations (SPD+/SBD+) had the worst PFS (5.5 mo). Patients with SBD LNM had significantly worse PFS compared to patients with no LNM (6.8 vs 8.8 mo; p = 0.003) and those with SPD LNM (6.8 vs 8.3 mo; p = 0.022). Any metastatic involvement of retroperitoneal lymph nodes, either alone (HR 1.25, 95% CI 1.08–1.45) or in combination with SPD+/SBD+LNM (HR 1.50, 95% CI 1.24–1.74; p < 0.001) was associated with higher risk of progression. By contrast, SPD LNM alone was not a prognostic factor for progression (HR 1.03, 95% 0.895–1.18; p = 0.695). (C) PFS by lymph node localization in clear cell renal cell carcinoma (ccRCC; median PFS: no nodes, 9.9 mo; SPD LNM, 8.7 mo; SBD LNM, 8.0 mo; and both, 5.6 mo). Patients with SPD LNM (HR 1.09, 95% CI 0.94–1.25) and SBD LNM (HR 1.04, 95% CI 0.96–1.14) had similar PFS, and this was not significantly worse than the PFS of patients without LNM (p = 0.265 and p = 0.333 for SPD and SBD LNM, respectively).
Comparison of PFS rates according to LNM sites revealed that patients with SBD LNM had significantly shorter PFS (6.8 mo [5.7–8.0 mo]) compared to those without LNM (8.8 mo [7.6–10.0 mo]; p = 0.003) or SPD LNM (8.3 mo [7.2–9.6 mo]; p = 0.022). The worst PFS was observed in patients with SPD+/SBD+ LNM (5.4 mo [4.6–6.4 mo]), but this did not significantly differ from PFS in patients with SBD LNM (p = 0.093; Fig. 2 B).
In ccRCC, however, comparison according to LNM localization revealed no significant differences in PFS between patients without LNM (9.9 mo [8.6–11.1 mo]) and those with SPD LNM (8.7 mo [7.3–10.1 mo]) or SBD LNM (8.0 mo [6.9–9.2 mo]) (p = 0.312; Fig. 2 C). SPD+/SBD+ LNM patients had the worst survival outcome (5.6 mo [4.0–7.2 mo]) in comparison to patients without LNM (p < 0.001), with SPD LNM (p = 0.006), and with SBD LNM (p = 0.035; Fig. 2 C). In nccRCC, there was no difference in PFS between patients with no LNM (3.9 mo [2.9–5.0 mo]), SPD LNM (4.6 mo [0.0–9.9 mo]), SBD LNM (5.4 mo [4.7–6.0 mo]), and SPD+/SBD+ LNM (5.4 mo [3.2–7.3 mo]).
In MVA analyses, LNM remained an independent prognostic factor (HR 1.13 [1.03–1.24]; p = 0.013) when adjusted for IMDC risk criteria, as well as liver and bone metastases ( Table 2 ) [16] . In subgroup analysis, SBD LNM (HR 1.31 [1.08–1.60]; p = 0.006) and SPD+/SBD LNM (HR 1.29 [1.04–1.61]; p = 0.022) were independent prognostic factors for PFS, while SPD LNM was not independently associated with PFS (p = 1.00; Table 2 ). In ccRCC, only SPD+/SBD+ LNM (p = 0.025) was independently associated with PFS ( Table 2 ).
Table 2 Multivariate analyses for association between lymph node metastases (LNM) and progression-free survival (PFS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time <1 yr | 1.33 | 1.21–1.47 | <0.001 |
| Low hemoglobin | 1.37 | 1.24–1.52 | <0.001 |
| Neutrophilia | 1.43 | 1.25–1.63 | <0.001 |
| Thrombocytosis | 1.16 | 1.03–1.32 | 0.019 |
| Karnofsy PS <80% | 1.68 | 1.50–1.88 | <0.001 |
| LNM | 1.13 | 1.03–1.24 | 0.013 |
| Liver metastases | 1.26 | 1.12–1.41 | <0.001 |
| Bone metastases | 1.27 | 1.15–1.40 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.19–1.61 | <0.001 |
| Low hemoglobin | 1.42 | 1.21–1.67 | <0.001 |
| Neutrophilia | 1.46 | 1.19–1.80 | <0.001 |
| Thrombocytosis | 1.22 | 1.01–1.47 | 0.045 |
| Karnofsy PS <80% | 1.94 | 1.65–2.28 | <0.001 |
| Liver metastases | 1.39 | 1.18–1.65 | <0.001 |
| Bone metastases | 1.45 | 1.25–1.68 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.83–1.20 | 0.990 |
| SBD LNM | 1.32 | 1.08–1.60 | 0.006 |
| SPD+/SBD+ LNM | 1.29 | 1.04–1.61 | 0.022 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.17–1.64 | <0.001 |
| Low hemoglobin | 1.43 | 1.21–1.69 | <0.001 |
| Neutrophilia | 1.38 | 1.09–1.76 | 0.009 |
| Thrombocytosis | 1.97 | 1.64–2.38 | <0.001 |
| Karnofsy PS <80% | 1.51 | 1.24–1.85 | <0.001 |
| Bone metastases | 1.64 | 1.38–1.94 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.82–1.22 | 0.998 |
| SBD LNM | 1.23 | 0.98–1.55 | 0.077 |
| SPD+/SBD+ LNM | 1.33 | 1.04–1.71 | 0.025 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm; ccRCC = clear cell RCC.
At the time of the analysis, 1960 patients had died; the median OS time for the total patient cohort was 20.0 mo (25th percentile, 42.3 mo; 75th percentile, 7.9 mo). Patients with metastatic spread to lymph nodes had a worse survival outcome than patients without LNM (24.0 mo [21.6–26.3 mo] vs 16.0 mo [14.9–17.1 mo], p < 0.001; Fig. 3 A).
Fig. 3 (A) Patients with lymph node metastases (LNM) had a higher risk of death (median overall survival [OS] 23.9 mo, 95% confidence interval [CI] 21.6–26.3 mo) than those without LNM (median OS 16.0 mo, 95% CI 14.9–17.1; hazard ratio [HR] 1.41, 95% CI 1.28–1.54; p < 0.001). (B) A comparison without consideration of the histologic subtypes demonstrates that any lymph node involvement was associated with worse survival outcome compared to patients without LNM. Median survival times were 25.2 mo, 20.3 mo, 16.2 mo, and 11.5 mo in patients with no LNM and lymph node involvement in supradiaphragmatic (SPD), subdiaphragmatic (SBD), and both localizations (SPD+/SBD+), respectively. The HRs (95% CIs) for patients with SPD, SBD, and SPD+/SBD+ LNM were 1.24 (1.06–1.44), 1.45 (1.23–1.71), and 1.93 (1.60–2.33), respectively. (C) In clear cell renal cell carcinoma (ccRCC), patients with SPD and SBD LNM had similar median OS (SPD, 21.6 mo; SBD, 19.8 mo). However, both groups had a significantly higher risk of death (SPD: HR 1.24, 95% CI 1.05–1.47; SBD: HR 1.28, 95% CI 1.05–1.56) in comparison to those without lymph nodes. Patients with SPD+/SBD+ metastases had the worst OS (12.85 mo) and a twofold higher risk of death than patients without lymph node involvement (HR 2.04, 95% CI 1.68–2.48). (D) In non–clear cell RCC (nccRCC), median OS times according to lymph node localization were 15.8 mo for no lymph nodes, 16.1 mo for SPD LNM, 11.2 mo for SBD LNM, and 12.5 mo for SPD+/SBD+ LNM. There were no significant differences between the groups (log rank p = 0.444).
In subgroup analyses, patients without LNM had a median OS of 25.2 mo (21.9–28.5 mo) compared to 20.3 mo (17.1–23.5 mo) for SPD LNM (p = 0.093), 16.2 mo (13.5–20.0 mo) for SBD LNM (p < 0.001), and 11.5 mo (9.8–13.3 mo) (p < 0.001) for SPD+/SBD+ LNM ( Fig. 3 B).
In ccRCC, further analysis considering LNM localization revealed similar OS times for patients with SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13]), and were shorter than for patients without LNM (26.97 mo [23.32– 30.62], p = 0.01; Fig. 3 C). Patients with SPD+/SBD+ LNM had significantly (p = 0.001) worse OS (12.85 mo [9.45–16.24 mo]) compared to SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13 mo]; Fig. 3 C). In nccRCC, analyses according to LNM revealed no significant differences in OS (p = 0.444; Fig. 3 C).
For all histological subtypes, LNM were an independent prognostic factor for adverse OS in MVA analysis (p < 0.001; Table 3 ). Additional MVA according to LNM sites revealed that SBD LNM (p < 0.001) and SPD+/SBD+ LNM (p < 0.001) were independent prognostic factors for OS. SPD LNM were not independently associated with OS (p = 0.999; Table 3 ). In ccRCC, MVA according to LNM localization confirmed an independent association between SBD LNM (p = 0.028) and SPD+/SBD+ LNM (p = 0.006) and OS ( Table 3 ).
Table 3 Multivariate Cox regression analysis of association between lymph node metastases (LNM) and overall survival (OS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time < 1 yr | 1.34 | 1.19–1.49 | <0.001 |
| Low hemoglobin | 1.51 | 1.31–1.70 | <0.001 |
| High Ca | 1.47 | 1.25–1.73 | <0.001 |
| Neutrophilia | 1.67 | 1.45–1.92 | <0.001 |
| Thrombophilia | 1.31 | 1.14–1.51 | <0.001 |
| Karnofsky PS <80% | 1.88 | 1.67–2.13 | <0.001 |
| Liver metastases | 1.50 | 1.32–1.70 | <0.001 |
| Bone metastases | 1.50 | 1.34–1.67 | <0.001 |
| LNM | 1.33 | 1.19–1.48 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time < 1 yr | 1.32 | 1.11–1.56 | 0.001 |
| Low hemoglobin | 1.45 | 1.21–1.73 | <0.001 |
| High Ca | 1.38 | 1.08–1.77 | 0.010 |
| Neutrophilia | 1.78 | 1.43–2.22 | <0.001 |
| Thrombophilia | 1.49 | 1.21–1.83 | <0.001 |
| Karnofsky PS <80% | 2.19 | 1.82–2.62 | <0.001 |
| Liver metastases | 1.84 | 1.53–2.20 | <0.001 |
| Bone metastases | 1.61 | 1.36–1.90 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.57 | 0.98–1.48 | 0.075 |
| SBD LNM | 1.59 | 1.26–1.96 | <0.001 |
| SPD+/SBD+ LNM | 1.32 | 1.25–2.02 | <0.001 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time < 1 yr | 1.36 | 1.12–1.64 | 0.002 |
| Low hemoglobin | 1.36 | 1.11–1.67 | 0.003 |
| High Ca | 1.44 | 1.10–1.89 | 0.009 |
| Neutrophilia | 1.65 | 1.27–2.13 | <0.001 |
| Thrombophilia | 1.52 | 1.19–1.91 | 0.001 |
| Karnofsky PS <80% | 2.12 | 1.72–2.61 | <0.001 |
| Liver metastases | 1.94 | 1.57–2.41 | <0.001 |
| Bone metastases | 1.84 | 1.53–2.22 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.13 | 0.90–1.42 | 0.301 |
| SBD LNM | 1.34 | 1.03–1.74 | 0.028 |
| SPD+/SBD+ LNM | 1.47 | 1.11–1.93 | 0.006 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm; ccRCC = clear cell RCC.
An overall comparison of patients with and without LNM is shown in Table 4 . Patients with LNM presented more often with nccRCC (p < 0.001) and higher Fuhrman grades (p = 0.007), and had less often undergone prior nephrectomy (p < 0.001). The IMDC risk factors of high calcium (p = 0.001), thrombocytosis (p < 0.001), neutrophilia (p < 0.001), and Karnofsky performance status <80% (p = 0.009) were all more frequently seen in patients with LNM. Comparison of additional metastatic sites revealed that LNM are related to a higher overall metastatic burden (one or more additional metastatic site p < 0.001).
Table 4 General comparison of clinicopathologic features of patients with lymph node metastases (LNM) compared to those without LNM a
| No LNM (n = 1380) |
LNM (n = 1616) |
p value | |
|---|---|---|---|
| Mean age, yr (SD) | 58.65 (11.36) | 58. 50 (11.30) | 0.716 |
| Male, n (%) | 1003 (72.7) | 1182 (73.1) | 0.805 |
| Nephrectomy, n (%) | 1134 (82.4) | 1218 (75.4) | <0.001 |
| Fuhrman grade n (%) | 0.007 | ||
| 1 | 44 (4.2) | 33 (2.8) | |
| 2 | 302 (28.6) | 269 (23.1) | |
| 3 | 426 (40.3) | 501 (43.1) | |
| 4 | 284 (26.9) | 359 (30.9) | |
| Non–clear cell RCC, n (%) | 127 (9.8) | 253 (17.4) | <0.001 |
| Sarcomatoid features, n (%) | 143 (12.0) | 170 (12.0) | 0.586 |
| Therapy class, n (%) | |||
| AntiVEGF | 1316 (95.4) | 1507 (93.3) | |
| mTOR | 60 (4.3) | 105 (6.5) | |
| Combination | 4 (0.3) | 4 (0.2) | |
| DTT <1 yr, n (%) | 736 (53.4) | 905 (56.0) | 0.151 |
| Low HB (<LLN), n (%) | 691 (56.0) | 877 (58.1) | 0.261 |
| High calcium, n (%) | 98 (8.6) | 179 (12.6) | 0.001 |
| Neutrophilia, n (%) | 154 (13.1) | 286 (19.3) | <0.001 |
| Thrombocytosis, n (%) | 172 (15.4) | 299 (22.4) | <0.001 |
| Karnofsky PS <80%, n (%) | 261 (21.5) | 372 (25.9) | 0.009 |
| Lung metastases, n (%) | 939 (68.0) | 1149 (72.5) | 0.016 |
| Brain metastases, n (%) | 100 (7.2) | 141 (9.4) | 0.009 |
| Liver metastases, n (%) | 278 (20.2) | 328 (21.7) | 0.337 |
| Bone metastases, n (%) | 481 (34.9) | 551 (35.9) | 0.462 |
| Other metastic sites, n (%) | 470 (34.1) | 643 (43.1) | <0.001 |
| >1 metastatic site, n (%) | 805 (58.3%) | 1446 (89.5%) | <0.001 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; mTOR = mammalian target of rapamycin; PS = performance status; RCC = renal cell carcinoma; SD = standard deviation; VEGF = vascular endothelial growth factor.
The LNM localization distribution and clinicopathologic features in ccRCC and nccRCC are shown in Table 5 . In ccRCC, IMDC risk factors of neutrophilia (p = 0.049) and thrombophilia (p = 0.016) were more common. Patients with LNM had undergone nephrectomy less frequently (p = 0.001). In patients with ccRCC, those with LNM had concomitant lung (p < 0.001) and other metastases (eg, adrenal glands, other soft tissues; p = 0.009) more often than those without LNM; there was no difference in the frequency of bone (p = 0.276), liver (p = 0.943), or brain metastases (p = 0.187).
Table 5 Distribution of metastatic lymphatic sites in relation to the diaphragm in clear cell and non–clear cell renal carcinoma (RCC) a
| Clear cell RCC (n = 1218) |
Non–clear cell RCC (n = 159) |
|
|---|---|---|
| No LNM, n (%) | 552 (45.6) | 52 (32.7) |
| SPD LNM, n (%) | 321 (26.6) | 22 (13.8) |
| SBD LNM, n (%) | 204 (16.9) | 54 (34.0) |
| SPD+/SBD+ LNM, n (%) | 132 (10.9) | 31 (19.5) |
| LMN | No LMN | p value | LMN | No LMN | p value | |
|---|---|---|---|---|---|---|
| Nephrectomy, n (%) | 542 (81.2) | 492 (88.2) | 0.001 | 82 (76.6) | 43 (82.7) | 0.418 |
| Male, n (%) | 515 (77.7) | 403 (72.6) | 0.045 | 75 (70.1) | 37 (71.2) | 1.00 |
| DTT <1 yr, n (%) | 348 (52.5) | 265 (47.9) | 0.120 | 65 (60.7) | 29 (55.8) | 0.607 |
| Low HB (<LLN) | 331 (54.0) | 233 (48.7) | 0.088 | 49 (51.0) | 19 (45.2) | 0.582 |
| High calcium, n (%) | 70 (12.3) | 40 (9.5) | 0.184 | 5 (5.7) | 4 (12.1) | 0.254 |
| Neutrophilia, n (%) | 95 (15.9) | 53 (11.6) | 0.049 | 24 (25.3) | 7 (16.7) | 0.376 |
| Thrombophilia, n (%) | 106 (22.0) | 63 (15.6) | 0.016 | 17 (21.0) | 10 (25.6) | 0.642 |
| Karnofsky PS <80%, n (%) | 145 (24.7) | 95 (20.0) | 0.076 | 24 (24.7) | 10 (20.4) | 0.680 |
| Sarcomatoid features, n (%) | 52 (9.1) | 62 (12.6) | 0.074 | 14 (15.4) | 7 (15.6) | 1.00 |
| Lung metastases, n (%) | 507 (76.8) | 372 (67.0) | <0.001 | 52 (49.1) | 20 (38.5) | 0.237 |
| Brain metastases, n (%) | 65 (9.8) | 42 (7.6) | 0.187 | 4 (3.8) | 2 (3.8) | 1.00 |
| Liver metastases, n (%) | 132 (20.0) | 112 (20.2) | 0.943 | 32 (30.2) | 28 (53.8) | 0.005 |
| Bone metastases, n (%) | 220 (33.3) | 202 (36.4) | 0.276 | 35 (33.0) | 20 (38.5) | 0.594 |
| Other metastatic sites, n (%) | 287 (45.7) | 211 (38.2) | 0.009 | 40 (40.8) | 20 (38.5) | 0.862 |
| >1 metastatic site, n (%) | 602 (90.8) | 320 (57.7) | <0.001 | 87 (81.3) | 30 (7.7) | 0.002 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; PS = performance status; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm.
In nccRCC, liver metastases were found more often in patients without LNM (p = 0.005). A comparison of other clinicopathologic features between nccRCC patients with and without LNM revealed no significant differences ( Table 5 ).
It has previously been shown that only a minority of patients with kidney cancer undergo LND at the time of nephrectomy. This is the case even for high-risk patients [17] , even though retrospective studies have suggested a therapeutic benefit of LND [11] and [18]. It remains uncertain whether extended LND can improve the survival outcome for mRCC patients. However, if it were shown that SBD LNM affect the survival outcome for mRCC patients, this could form the basis for a rationale in favor of performing LND, or at least justify the need for prospective clinical trials to address this question. In this regard, an understanding of the influence of LNM on treatment response and survival outcome for mRCC patients undergoing TT could make an important contribution to the debate on surgical management of RCC.
Our study has demonstrated that LNM are associated with advanced IMDC risk factors, are an adverse prognostic factor for PFS and OS, and are associated with a higher metastatic burden. Furthermore, subanalyses showed that these differences depend on LNM localization.
Different cellular clones of RCC appear to differ in their propensity to spread to different organ sites [19] ; however, whether these differences impact on the survival outcome of mRCC patients is not completely understood. A previous IMDC study highlighted the prognostic importance of bone and liver metastases for survival outcome in mRCC patients, and suggested that aggressive RCC subclones tend to metastasize to these organs [16] . In addition, another study by our group showed that metastatic spread to other organs such as the pancreas or soft tissue tends to develop as a late clinical symptom, and has favorable treatment response and survival outcomes [20] . While these tumors may represent a favorable RCC subtype, our current results suggest that RCCs that spread to regional retroperitoneal lymph nodes represent another aggressive RCC subtype that develops bone metastases.
It has been suggested that LNM in ccRCC occur independently of von Hippel Lindau (VHL) gene inactivation [21] . Current agents used for mRCC treatment modulate pathways that are dysregulated because of VHL inactivation. This would imply that LNM might be less amenable to treatment with VEGF and mTOR inhibitors. However, our findings demonstrated no significant differences in best response to first-line therapies in patients with LNM versus those without LNM in both ccRCC and nccRCC patient groups. Nonetheless, patients with LNM had significantly shorter PFS than patients without LNM. While there was only a trend for an independent association between SBD LNM and short PFS, SPD+/SBD+ LNM were independently related to shorter PFS. Patients with SPD+/SBD+ LNM had additional hematogenous metastatic sites in 97% of cases. Thus, the high metastatic burden may explain the shorter response duration of first-line agents in this group of patients. However, a shorter duration for treatment success could also support the hypothesis that LNM are caused by RCC clones that do not have an inactive VHL gene function.
In the cytokine era, LNM were associated with shorter OS [1] and [2]. Our study demonstrates similar findings in the TT era. Interestingly, a worse survival outcome was observed for SBD LNM, both when occurring exclusively and in combination with thoracic lymph node involvement. Moreover, SBD LNM remained an independent prognostic factor for PFS and OS when analyzed without consideration of histologic subtype, and were an independent prognostic factor for OS in ccRCC. Conversely, SPD LNM were not an independent prognostic factor. This is notable and underscores the unique nature of RCCs that spread via retroperitoneal lymph nodes. Because the best response to TT did not different between SBD and SPD LNM, it appears that survival outcome is influenced by numerous factors. Patients with SBD LNM had faster disease progression before treatment initiation than those with SPD LNM (diagnosis to treatment <1 yr; 63% vs 46%; p < 0.001; data not shown). Despite a best response that was broadly similar, it seems that the disease progresses faster in SBD LNM patients during the later clinical course than in SPD LNM patients. In addition, SBD LNM were more often associated with anemia compared to SPD LNM (59% vs 48%; p = 0.008; data not shown). Anemia has multifactorial causes in cancer patients. However, one of the most important contributing factors is immune system dysregulation [22] . Dysregulation of the immune system may also be the reason for metastatic spread to SBD sites. Collectively, while risk factors that are currently unknown appear to dictate the clinical course in patients with SPD metastases, SBD metastases are independently associated with PFS and OS.
The majority of patients analyzed in this study did not undergo formal LND. Therefore, the effect of LND cannot be determined retrospectively. However, the current results indicate that SBD lymph nodes may have substantial negative effects on survival outcome and it is worth speculating whether extended removal of these metastatic sites could improve the clinical outcome for mRCC patients. The IMDC has recently shown that cytoreductive nephrectomy may provide therapeutic benefit in a large number of patients [23] . With regard to our current findings, it would be worth determining the role of LND in these patients.
The current study has several limitations that should be considered when interpreting the results. The IMDC database comprises retrospectively collected data for consecutive series of mRCC patients. These individual series, moreover, do not have standardized follow-up protocols, but this limitation may more accurately reflect real world outcomes. In addition, there was no centralized pathologic or radiologic review. Finally, lymph node status was determined clinically according to CT images in a number of cases, and it has been demonstrated that clinical and pathologic lymph node status is not necessarily always correlated.
LNM are associated with poor prognosis and adverse prognostic IMDC risk factors in mRCC patients treated with TT. Much of this increased risk associated with LNM appears to be dictated by SBD localization rather than SPD localization. Whether LND could improve the outcome for this group of patients remains unknown, but it is an important clinical question requiring prospective trials.
Author contributions: Daniel Y. Heng had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kroeger, Heng.
Acquisition of data: all authors.
Analysis and interpretation of data: Kroeger, Heng, Choueiri.
Drafting of the manuscript: Kroeger, Heng, Choueiri.
Critical revision of the manuscript for important intellectual content: all authors.
Statistical analysis: Kroeger, Heng.
Obtaining funding: None.
Administrative, technical, or material support: all authors.
Supervision: Choueiri, Heng.
Other (specify): None.
Financial disclosures: Daniel Y. Heng certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Benoit Beuselinck is an investigator of the EudraCT: 2011-006085-40/MetaSun trial supported by Pfizer. Frede Donskov has received research support from Novartis and GlaxoSmithKline. Toni K. Choueiri has received research funding from Pfizer and has an advisory role at Pfizer, GSK, Novartis, and Bayer. All other authors declare no conflicts of interest to the current study.
Funding/Support and role of the sponsor: None.
Acknowledgments: Dr. Choueiri is supported in part by the Trust family and the Loker Pinard and Michael Brigham Funds for Kidney Cancer Research at the Dana-Farber Cancer Institute. Dr. Kroeger thanks Dr. Klatte, Department of Urology, Medical University of Vienna, for statistical advice.
Renal cell carcinoma (RCC) is a heterogeneous disease consisting of several tumor types that have their own genetic, molecular, and clinical characteristics. Historical studies in the cytokine era by the University of California Los Angeles (UCLA) and the National Institutes of Health (NIH) described poor prognosis and worse treatment response to interleukin-2 (IL-2) in patients with metastatic involvement of the retroperitoneal lymph nodes [1] and [2]. Patients with and without retroperitoneal lymph node involvement had median overall survival (OS) times of 8.5 mo and 14.7 mo (p = 0.0004), respectively [2] . Pantuck et al [1] reported a better objective response rate in N0M1 compared to N+M1 patients (p = 0.01), with survival times of 10.5 mo and 20.4 mo, respectively. The adverse prognostic impact of retroperitoneal lymph node metastases (LNM) on survival outcome in patients with metastatic RCC (mRCC) was confirmed by several other institutions during the cytokine era [3], [4], and [5]. However, in the era of targeted therapies (TT), the survival outcome for patients with LNM has not yet been well characterized.
RCC preferentially metastasizes via hematogenous routes. Bianchi et al [6] recently described hematogenous metastatic sites in 80% of 11 157 patients with mRCC, while extension of the disease into lymph nodes was described only in 20% of patients. However, the authors did not provide details about the distribution of lymph node localization. The dominant regional LNM sites in RCC are retro- and paracaval, pre- and paraaortic, and interaortocaval lymph nodes in an anatomically intact retroperitoneum [7] . LNM can be unpredictable, and can also occur via direct extension to the thorax, supraclavicular lymph nodes, and iliac lymph nodes [7] . It has not been documented if particular sites of positive lymph node localization have a negative influence on the survival outcome of mRCC patients treated with TT.
Whether retroperitoneal lymph node dissection (LND) improves survival outcome is the subject of ongoing scientific debate. A phase 3 clinical trial (EORTC 30881) demonstrated positive lymph nodes in only 4% of investigated cases having no apparent involvement on computed tomography (CT) imaging. There was no advantage in OS, recurrence-free survival, or progression-free survival (PFS) for RCC patients who underwent LND [8] . The study has been criticized, however, because the majority of patients included had low-stage disease. Moreover, uniform surgical templates were not used for LND [9] . However, retrospective studies have demonstrated that LNM are more prevalent in high-risk patients with high Fuhrman grades, sarcomatoid features, locally advanced tumor stage, tumor size >10 cm, and tumor necrosis [10] . It has been suggested that LND is particularly beneficial in these high-risk patients [11] and [12]. Trends in the surgical management of RCC that have moved away from open radical surgery with extended LND to laparoscopic and robotic surgery with minimal hilar LND, combined with the negative results of phase 3 studies showing no additional benefit of LND over removal of the primary tumor alone, have decreased enthusiasm for performing extended LND. Therefore, it is common clinical practice to perform LND only in high-risk patients, and with a diagnostic rather than a curative intention [7] . Retrospective studies may be unable to add new insights into the current debate because only a prospective randomized trial with clear inclusion and exclusion criteria would be able to overcome this selection bias. Before planning such trials, however, it is prudent to better understand the impact of LNM and different lymph node localizations on the survival outcome for mRCC patients treated with current state-of-the-art therapies using data from the International mRCC Database Consortium (IMDC).
The IMDC database includes centers from North America (Canada, USA), Europe (Denmark, Greece, Belgium), Asia (Singapore, Japan, South Korea), and New Zealand. Data were collected from August 15, 2008 until December 31, 2013. At the time of analysis, the database contained data on 3405 patients who had received first-line targeted therapies between 2003 and 2013. The final study cohort comprised 2996 patients who were treated with first-line vascular endothelial growth factor (VEGF) inhibitors (n = 2823), mammalian target of rapamycin (mTOR) inhibitors (n = 165), or a combination of both (n = 8) therapies between April 2003 and August 2013. Patients were excluded from the analyses if they had received experimental therapies in first-line treatment (n = 7) or if no information was available on their lymph node status (n = 402; Fig. 1 ). Lymph node status was determined according to standard pathologic and CT criteria.
Information about LND was available for 890 patients. Full LND was performed in only 51 patients, and a limited (hilar) dissection was performed in 173 patients. These numbers were too low to allow any analyses and conclusions regarding the influence of LND on the survival outcome for mRCC patients treated with TT. Whether LND could be beneficial for mRCC patients was therefore not considered in the present study.
All centers obtained local Institutional Review Board approval before including data in the IMDC database. Baseline patient characteristics included demographic, clinicopathologic, and laboratory data, as previously described [13] . Survival data were retrospectively and prospectively collected from medical chart reviews and electronic records. Uniform data templates were used to ensure consistent data collection at each institution. The majority of patients were treated using standard care, but a subset received therapy as part of clinical trials. All patients were considered in consecutive series to avoid selection bias.
The primary hypothesis of the current study was that mRCC patients with LNM have worse survival outcome than patients without LNM when treated with TT. The secondary hypothesis was that worse survival outcome for mRCC patients with LNM is dictated by subdiaphragmatic (SPD) LNM rather than supradiaphragmatic (SBD) LNM.
Therefore, treatment response to current standard-of-care TT agents, OS, and PFS were compared between patients with LNM and those without LNM. We did not compare outcomes between the drugs used for treatment because these comparisons are frequently biased outside a clinical trial and there may not be sufficient power for such an analysis. Additional analyses were performed for subgroups of patients with exclusively SPD lymph node involvement, exclusively SBD lymph node involvement, or with lymph node involvement in both sites (SPD+/SBD+). Previous studies have demonstrated that tumor biology has an important impact on the survival outcome for mRCC patients [14] . Thus, the different biology of RCC was considered by separate analyses for clear cell RCC (ccRCC) and non-ccRCC (nccRCC) patients. Patients with SPD primarily had mediastinal lymph node involvement, while patients with SBD primarily had retroperitoneal lymph node involvement. Lymph node status was determined before initiation of TT. The selection process for patient groups analyzed in the study is shown in Figure 1 .
OS was defined as the time period between TT initiation and the date of death, with censoring on the day of the last follow-up visit. PFS was defined as the period between treatment initiation and progression, drug cessation, or death, with censoring at the last follow-up visit. Kaplan-Meier plots were used to estimate median OS and PFS, and univariate comparisons were performed using the log rank test. Uni- and multivariate (MVA) Cox regression analyses were performed to test the association of predefined prognostic factors and LNM with survival outcome. The analyses were performed with backward stepwise selection criteria, and significance was tested using the Wald statistic [13] . In MVA analyses, the IMDC (Heng) risk criteria were applied for adjustment [15] . The IMDC risk criteria include a time period from diagnosis to treatment of <1 yr, Karnofsky performance score <80%, anemia, hypercalcemia, thrombophilia, and neutrophilia. Upper and lower limits of normal for the laboratory parameters were based on institutional limits. In addition, adjustment was performed for bone and liver metastases because interactions of LNM with other metastatic sites should be considered. The prognostic relevance of bone and liver metastases has previously been demonstrated by our group [16] .
A secondary aim was to analyze baseline characteristics. Patient and tumor characteristics were compared using the Student t test for continuous variables, or the chi-square test and the Fisher exact test for categorical variables.
Statistical analyses were performed using SPSS version 22 (IBM, Chicago, IL, USA), and a two-sided value of p < 0.05 was considered statistically significant.
At the time of data analysis, 2525 (84.3%) patients had stopped taking their TT; the median time on TT was 6.0 mo (25th percentile, 2.7 mo; 75th percentile, 13.0 mo). Best overall response data to first-line therapy were available for 2559 patients. There were no significant differences in response to first-line TT (p = 0.059) when patients were compared without consideration of RCC subtype or LNM localization ( Table 1 ).
Table 1 Treatment response to targeted therapies in patients with lymph node metastases (LNM) compared to those without LNM
| Best response | |||||
|---|---|---|---|---|---|
| CR | PR | SD | PD | ||
| All RCC subtypes | |||||
| No LMN | n | 14 | 304 | 556 | 290 |
| % | 1.2 | 26.1 | 47.8 | 24.9 | |
| LMN | n | 9 | 314 | 690 | 382 |
| % | 0.6 | 22.5 | 49.5 | 27.4 | |
| Clear cell RCC | |||||
| No LNM | n | 3 | 108 | 261 | 95 |
| % | 0.6 | 23.1 | 55.9 | 20.3 | |
| SPD LNM | n | 2 | 73 | 149 | 55 |
| % | 0.7 | 26.2 | 53.4 | 19.7 | |
| SBD LNM | n | 2 | 50 | 92 | 42 |
| % | 1.1 | 26.9 | 49.5 | 22.6 | |
| SPD+/SBD+ LNM | n | 0 | 28 | 58 | 36 |
| % | 0.0 | 23.0 | 47.5 | 29.5 | |
| Non–clear cell RCC | |||||
| No LNM | n | 0 | 3 | 21 | 21 |
| % | 0.0 | 6.7 | 46.7 | 46.7 | |
| SPD LNM | n | 0 | 3 | 12 | 3 |
| % | 0.0 | 16.7 | 66.7 | 16.7 | |
| SBD LNM | n | 0 | 7 | 24 | 19 |
| % | 0.0 | 14.0 | 48.0 | 38.0 | |
| SPD+/SBD+ LNM | n | 1 | 4 | 10 | 6 |
| % | 4.8 | 19.0 | 47.6 | 28.6 | |
CR = complete remission; PR = partial response; SD = stable disease; PD = progressive disease; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm.
Both response data and LNM localization data were available for 1054 ccRCC patients and 134 nccRCC patients. Comparison of data by LNM localization revealed no difference in best response to first-line therapies for either ccRCC (p = 0.457) or nccRCC (p = 0.192; Table 1 ).
Patients with LNM generally had worse PFS compared to those with no LNM without considering LNM localization or RCC subtype (median PFS [95% confidence interval] 7.8 mo [7.1–8.5 mo] vs 6.1 [5.6–6.6 mo], log rank p = 0.001; Fig. 2 A).
Fig. 2 (A) Comparison of progression-free survival (PFS) in patients with positive lymph nodes versus patients without positive lymph nodes (median 7.8 mo vs. 6.1 mo) without consideration of histologic subtype. Patients with lymph node metastases (LNM) had worse PFS (hazard ratio [HR] 1.14, 95% confidence interval [CI] 1.06–1.24; p = 0.001). (B) Comparison of PFS according to lymph node localization. Patients without LNM had the best PFS (8.8 mo), and patients with both supradiaphragmatic (SPD) and subdiaphragmatic (SBD) LNM localizations (SPD+/SBD+) had the worst PFS (5.5 mo). Patients with SBD LNM had significantly worse PFS compared to patients with no LNM (6.8 vs 8.8 mo; p = 0.003) and those with SPD LNM (6.8 vs 8.3 mo; p = 0.022). Any metastatic involvement of retroperitoneal lymph nodes, either alone (HR 1.25, 95% CI 1.08–1.45) or in combination with SPD+/SBD+LNM (HR 1.50, 95% CI 1.24–1.74; p < 0.001) was associated with higher risk of progression. By contrast, SPD LNM alone was not a prognostic factor for progression (HR 1.03, 95% 0.895–1.18; p = 0.695). (C) PFS by lymph node localization in clear cell renal cell carcinoma (ccRCC; median PFS: no nodes, 9.9 mo; SPD LNM, 8.7 mo; SBD LNM, 8.0 mo; and both, 5.6 mo). Patients with SPD LNM (HR 1.09, 95% CI 0.94–1.25) and SBD LNM (HR 1.04, 95% CI 0.96–1.14) had similar PFS, and this was not significantly worse than the PFS of patients without LNM (p = 0.265 and p = 0.333 for SPD and SBD LNM, respectively).
Comparison of PFS rates according to LNM sites revealed that patients with SBD LNM had significantly shorter PFS (6.8 mo [5.7–8.0 mo]) compared to those without LNM (8.8 mo [7.6–10.0 mo]; p = 0.003) or SPD LNM (8.3 mo [7.2–9.6 mo]; p = 0.022). The worst PFS was observed in patients with SPD+/SBD+ LNM (5.4 mo [4.6–6.4 mo]), but this did not significantly differ from PFS in patients with SBD LNM (p = 0.093; Fig. 2 B).
In ccRCC, however, comparison according to LNM localization revealed no significant differences in PFS between patients without LNM (9.9 mo [8.6–11.1 mo]) and those with SPD LNM (8.7 mo [7.3–10.1 mo]) or SBD LNM (8.0 mo [6.9–9.2 mo]) (p = 0.312; Fig. 2 C). SPD+/SBD+ LNM patients had the worst survival outcome (5.6 mo [4.0–7.2 mo]) in comparison to patients without LNM (p < 0.001), with SPD LNM (p = 0.006), and with SBD LNM (p = 0.035; Fig. 2 C). In nccRCC, there was no difference in PFS between patients with no LNM (3.9 mo [2.9–5.0 mo]), SPD LNM (4.6 mo [0.0–9.9 mo]), SBD LNM (5.4 mo [4.7–6.0 mo]), and SPD+/SBD+ LNM (5.4 mo [3.2–7.3 mo]).
In MVA analyses, LNM remained an independent prognostic factor (HR 1.13 [1.03–1.24]; p = 0.013) when adjusted for IMDC risk criteria, as well as liver and bone metastases ( Table 2 ) [16] . In subgroup analysis, SBD LNM (HR 1.31 [1.08–1.60]; p = 0.006) and SPD+/SBD LNM (HR 1.29 [1.04–1.61]; p = 0.022) were independent prognostic factors for PFS, while SPD LNM was not independently associated with PFS (p = 1.00; Table 2 ). In ccRCC, only SPD+/SBD+ LNM (p = 0.025) was independently associated with PFS ( Table 2 ).
Table 2 Multivariate analyses for association between lymph node metastases (LNM) and progression-free survival (PFS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time <1 yr | 1.33 | 1.21–1.47 | <0.001 |
| Low hemoglobin | 1.37 | 1.24–1.52 | <0.001 |
| Neutrophilia | 1.43 | 1.25–1.63 | <0.001 |
| Thrombocytosis | 1.16 | 1.03–1.32 | 0.019 |
| Karnofsy PS <80% | 1.68 | 1.50–1.88 | <0.001 |
| LNM | 1.13 | 1.03–1.24 | 0.013 |
| Liver metastases | 1.26 | 1.12–1.41 | <0.001 |
| Bone metastases | 1.27 | 1.15–1.40 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.19–1.61 | <0.001 |
| Low hemoglobin | 1.42 | 1.21–1.67 | <0.001 |
| Neutrophilia | 1.46 | 1.19–1.80 | <0.001 |
| Thrombocytosis | 1.22 | 1.01–1.47 | 0.045 |
| Karnofsy PS <80% | 1.94 | 1.65–2.28 | <0.001 |
| Liver metastases | 1.39 | 1.18–1.65 | <0.001 |
| Bone metastases | 1.45 | 1.25–1.68 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.83–1.20 | 0.990 |
| SBD LNM | 1.32 | 1.08–1.60 | 0.006 |
| SPD+/SBD+ LNM | 1.29 | 1.04–1.61 | 0.022 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.17–1.64 | <0.001 |
| Low hemoglobin | 1.43 | 1.21–1.69 | <0.001 |
| Neutrophilia | 1.38 | 1.09–1.76 | 0.009 |
| Thrombocytosis | 1.97 | 1.64–2.38 | <0.001 |
| Karnofsy PS <80% | 1.51 | 1.24–1.85 | <0.001 |
| Bone metastases | 1.64 | 1.38–1.94 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.82–1.22 | 0.998 |
| SBD LNM | 1.23 | 0.98–1.55 | 0.077 |
| SPD+/SBD+ LNM | 1.33 | 1.04–1.71 | 0.025 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm; ccRCC = clear cell RCC.
At the time of the analysis, 1960 patients had died; the median OS time for the total patient cohort was 20.0 mo (25th percentile, 42.3 mo; 75th percentile, 7.9 mo). Patients with metastatic spread to lymph nodes had a worse survival outcome than patients without LNM (24.0 mo [21.6–26.3 mo] vs 16.0 mo [14.9–17.1 mo], p < 0.001; Fig. 3 A).
Fig. 3 (A) Patients with lymph node metastases (LNM) had a higher risk of death (median overall survival [OS] 23.9 mo, 95% confidence interval [CI] 21.6–26.3 mo) than those without LNM (median OS 16.0 mo, 95% CI 14.9–17.1; hazard ratio [HR] 1.41, 95% CI 1.28–1.54; p < 0.001). (B) A comparison without consideration of the histologic subtypes demonstrates that any lymph node involvement was associated with worse survival outcome compared to patients without LNM. Median survival times were 25.2 mo, 20.3 mo, 16.2 mo, and 11.5 mo in patients with no LNM and lymph node involvement in supradiaphragmatic (SPD), subdiaphragmatic (SBD), and both localizations (SPD+/SBD+), respectively. The HRs (95% CIs) for patients with SPD, SBD, and SPD+/SBD+ LNM were 1.24 (1.06–1.44), 1.45 (1.23–1.71), and 1.93 (1.60–2.33), respectively. (C) In clear cell renal cell carcinoma (ccRCC), patients with SPD and SBD LNM had similar median OS (SPD, 21.6 mo; SBD, 19.8 mo). However, both groups had a significantly higher risk of death (SPD: HR 1.24, 95% CI 1.05–1.47; SBD: HR 1.28, 95% CI 1.05–1.56) in comparison to those without lymph nodes. Patients with SPD+/SBD+ metastases had the worst OS (12.85 mo) and a twofold higher risk of death than patients without lymph node involvement (HR 2.04, 95% CI 1.68–2.48). (D) In non–clear cell RCC (nccRCC), median OS times according to lymph node localization were 15.8 mo for no lymph nodes, 16.1 mo for SPD LNM, 11.2 mo for SBD LNM, and 12.5 mo for SPD+/SBD+ LNM. There were no significant differences between the groups (log rank p = 0.444).
In subgroup analyses, patients without LNM had a median OS of 25.2 mo (21.9–28.5 mo) compared to 20.3 mo (17.1–23.5 mo) for SPD LNM (p = 0.093), 16.2 mo (13.5–20.0 mo) for SBD LNM (p < 0.001), and 11.5 mo (9.8–13.3 mo) (p < 0.001) for SPD+/SBD+ LNM ( Fig. 3 B).
In ccRCC, further analysis considering LNM localization revealed similar OS times for patients with SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13]), and were shorter than for patients without LNM (26.97 mo [23.32– 30.62], p = 0.01; Fig. 3 C). Patients with SPD+/SBD+ LNM had significantly (p = 0.001) worse OS (12.85 mo [9.45–16.24 mo]) compared to SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13 mo]; Fig. 3 C). In nccRCC, analyses according to LNM revealed no significant differences in OS (p = 0.444; Fig. 3 C).
For all histological subtypes, LNM were an independent prognostic factor for adverse OS in MVA analysis (p < 0.001; Table 3 ). Additional MVA according to LNM sites revealed that SBD LNM (p < 0.001) and SPD+/SBD+ LNM (p < 0.001) were independent prognostic factors for OS. SPD LNM were not independently associated with OS (p = 0.999; Table 3 ). In ccRCC, MVA according to LNM localization confirmed an independent association between SBD LNM (p = 0.028) and SPD+/SBD+ LNM (p = 0.006) and OS ( Table 3 ).
Table 3 Multivariate Cox regression analysis of association between lymph node metastases (LNM) and overall survival (OS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time < 1 yr | 1.34 | 1.19–1.49 | <0.001 |
| Low hemoglobin | 1.51 | 1.31–1.70 | <0.001 |
| High Ca | 1.47 | 1.25–1.73 | <0.001 |
| Neutrophilia | 1.67 | 1.45–1.92 | <0.001 |
| Thrombophilia | 1.31 | 1.14–1.51 | <0.001 |
| Karnofsky PS <80% | 1.88 | 1.67–2.13 | <0.001 |
| Liver metastases | 1.50 | 1.32–1.70 | <0.001 |
| Bone metastases | 1.50 | 1.34–1.67 | <0.001 |
| LNM | 1.33 | 1.19–1.48 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time < 1 yr | 1.32 | 1.11–1.56 | 0.001 |
| Low hemoglobin | 1.45 | 1.21–1.73 | <0.001 |
| High Ca | 1.38 | 1.08–1.77 | 0.010 |
| Neutrophilia | 1.78 | 1.43–2.22 | <0.001 |
| Thrombophilia | 1.49 | 1.21–1.83 | <0.001 |
| Karnofsky PS <80% | 2.19 | 1.82–2.62 | <0.001 |
| Liver metastases | 1.84 | 1.53–2.20 | <0.001 |
| Bone metastases | 1.61 | 1.36–1.90 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.57 | 0.98–1.48 | 0.075 |
| SBD LNM | 1.59 | 1.26–1.96 | <0.001 |
| SPD+/SBD+ LNM | 1.32 | 1.25–2.02 | <0.001 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time < 1 yr | 1.36 | 1.12–1.64 | 0.002 |
| Low hemoglobin | 1.36 | 1.11–1.67 | 0.003 |
| High Ca | 1.44 | 1.10–1.89 | 0.009 |
| Neutrophilia | 1.65 | 1.27–2.13 | <0.001 |
| Thrombophilia | 1.52 | 1.19–1.91 | 0.001 |
| Karnofsky PS <80% | 2.12 | 1.72–2.61 | <0.001 |
| Liver metastases | 1.94 | 1.57–2.41 | <0.001 |
| Bone metastases | 1.84 | 1.53–2.22 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.13 | 0.90–1.42 | 0.301 |
| SBD LNM | 1.34 | 1.03–1.74 | 0.028 |
| SPD+/SBD+ LNM | 1.47 | 1.11–1.93 | 0.006 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm; ccRCC = clear cell RCC.
An overall comparison of patients with and without LNM is shown in Table 4 . Patients with LNM presented more often with nccRCC (p < 0.001) and higher Fuhrman grades (p = 0.007), and had less often undergone prior nephrectomy (p < 0.001). The IMDC risk factors of high calcium (p = 0.001), thrombocytosis (p < 0.001), neutrophilia (p < 0.001), and Karnofsky performance status <80% (p = 0.009) were all more frequently seen in patients with LNM. Comparison of additional metastatic sites revealed that LNM are related to a higher overall metastatic burden (one or more additional metastatic site p < 0.001).
Table 4 General comparison of clinicopathologic features of patients with lymph node metastases (LNM) compared to those without LNM a
| No LNM (n = 1380) |
LNM (n = 1616) |
p value | |
|---|---|---|---|
| Mean age, yr (SD) | 58.65 (11.36) | 58. 50 (11.30) | 0.716 |
| Male, n (%) | 1003 (72.7) | 1182 (73.1) | 0.805 |
| Nephrectomy, n (%) | 1134 (82.4) | 1218 (75.4) | <0.001 |
| Fuhrman grade n (%) | 0.007 | ||
| 1 | 44 (4.2) | 33 (2.8) | |
| 2 | 302 (28.6) | 269 (23.1) | |
| 3 | 426 (40.3) | 501 (43.1) | |
| 4 | 284 (26.9) | 359 (30.9) | |
| Non–clear cell RCC, n (%) | 127 (9.8) | 253 (17.4) | <0.001 |
| Sarcomatoid features, n (%) | 143 (12.0) | 170 (12.0) | 0.586 |
| Therapy class, n (%) | |||
| AntiVEGF | 1316 (95.4) | 1507 (93.3) | |
| mTOR | 60 (4.3) | 105 (6.5) | |
| Combination | 4 (0.3) | 4 (0.2) | |
| DTT <1 yr, n (%) | 736 (53.4) | 905 (56.0) | 0.151 |
| Low HB (<LLN), n (%) | 691 (56.0) | 877 (58.1) | 0.261 |
| High calcium, n (%) | 98 (8.6) | 179 (12.6) | 0.001 |
| Neutrophilia, n (%) | 154 (13.1) | 286 (19.3) | <0.001 |
| Thrombocytosis, n (%) | 172 (15.4) | 299 (22.4) | <0.001 |
| Karnofsky PS <80%, n (%) | 261 (21.5) | 372 (25.9) | 0.009 |
| Lung metastases, n (%) | 939 (68.0) | 1149 (72.5) | 0.016 |
| Brain metastases, n (%) | 100 (7.2) | 141 (9.4) | 0.009 |
| Liver metastases, n (%) | 278 (20.2) | 328 (21.7) | 0.337 |
| Bone metastases, n (%) | 481 (34.9) | 551 (35.9) | 0.462 |
| Other metastic sites, n (%) | 470 (34.1) | 643 (43.1) | <0.001 |
| >1 metastatic site, n (%) | 805 (58.3%) | 1446 (89.5%) | <0.001 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; mTOR = mammalian target of rapamycin; PS = performance status; RCC = renal cell carcinoma; SD = standard deviation; VEGF = vascular endothelial growth factor.
The LNM localization distribution and clinicopathologic features in ccRCC and nccRCC are shown in Table 5 . In ccRCC, IMDC risk factors of neutrophilia (p = 0.049) and thrombophilia (p = 0.016) were more common. Patients with LNM had undergone nephrectomy less frequently (p = 0.001). In patients with ccRCC, those with LNM had concomitant lung (p < 0.001) and other metastases (eg, adrenal glands, other soft tissues; p = 0.009) more often than those without LNM; there was no difference in the frequency of bone (p = 0.276), liver (p = 0.943), or brain metastases (p = 0.187).
Table 5 Distribution of metastatic lymphatic sites in relation to the diaphragm in clear cell and non–clear cell renal carcinoma (RCC) a
| Clear cell RCC (n = 1218) |
Non–clear cell RCC (n = 159) |
|
|---|---|---|
| No LNM, n (%) | 552 (45.6) | 52 (32.7) |
| SPD LNM, n (%) | 321 (26.6) | 22 (13.8) |
| SBD LNM, n (%) | 204 (16.9) | 54 (34.0) |
| SPD+/SBD+ LNM, n (%) | 132 (10.9) | 31 (19.5) |
| LMN | No LMN | p value | LMN | No LMN | p value | |
|---|---|---|---|---|---|---|
| Nephrectomy, n (%) | 542 (81.2) | 492 (88.2) | 0.001 | 82 (76.6) | 43 (82.7) | 0.418 |
| Male, n (%) | 515 (77.7) | 403 (72.6) | 0.045 | 75 (70.1) | 37 (71.2) | 1.00 |
| DTT <1 yr, n (%) | 348 (52.5) | 265 (47.9) | 0.120 | 65 (60.7) | 29 (55.8) | 0.607 |
| Low HB (<LLN) | 331 (54.0) | 233 (48.7) | 0.088 | 49 (51.0) | 19 (45.2) | 0.582 |
| High calcium, n (%) | 70 (12.3) | 40 (9.5) | 0.184 | 5 (5.7) | 4 (12.1) | 0.254 |
| Neutrophilia, n (%) | 95 (15.9) | 53 (11.6) | 0.049 | 24 (25.3) | 7 (16.7) | 0.376 |
| Thrombophilia, n (%) | 106 (22.0) | 63 (15.6) | 0.016 | 17 (21.0) | 10 (25.6) | 0.642 |
| Karnofsky PS <80%, n (%) | 145 (24.7) | 95 (20.0) | 0.076 | 24 (24.7) | 10 (20.4) | 0.680 |
| Sarcomatoid features, n (%) | 52 (9.1) | 62 (12.6) | 0.074 | 14 (15.4) | 7 (15.6) | 1.00 |
| Lung metastases, n (%) | 507 (76.8) | 372 (67.0) | <0.001 | 52 (49.1) | 20 (38.5) | 0.237 |
| Brain metastases, n (%) | 65 (9.8) | 42 (7.6) | 0.187 | 4 (3.8) | 2 (3.8) | 1.00 |
| Liver metastases, n (%) | 132 (20.0) | 112 (20.2) | 0.943 | 32 (30.2) | 28 (53.8) | 0.005 |
| Bone metastases, n (%) | 220 (33.3) | 202 (36.4) | 0.276 | 35 (33.0) | 20 (38.5) | 0.594 |
| Other metastatic sites, n (%) | 287 (45.7) | 211 (38.2) | 0.009 | 40 (40.8) | 20 (38.5) | 0.862 |
| >1 metastatic site, n (%) | 602 (90.8) | 320 (57.7) | <0.001 | 87 (81.3) | 30 (7.7) | 0.002 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; PS = performance status; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm.
In nccRCC, liver metastases were found more often in patients without LNM (p = 0.005). A comparison of other clinicopathologic features between nccRCC patients with and without LNM revealed no significant differences ( Table 5 ).
It has previously been shown that only a minority of patients with kidney cancer undergo LND at the time of nephrectomy. This is the case even for high-risk patients [17] , even though retrospective studies have suggested a therapeutic benefit of LND [11] and [18]. It remains uncertain whether extended LND can improve the survival outcome for mRCC patients. However, if it were shown that SBD LNM affect the survival outcome for mRCC patients, this could form the basis for a rationale in favor of performing LND, or at least justify the need for prospective clinical trials to address this question. In this regard, an understanding of the influence of LNM on treatment response and survival outcome for mRCC patients undergoing TT could make an important contribution to the debate on surgical management of RCC.
Our study has demonstrated that LNM are associated with advanced IMDC risk factors, are an adverse prognostic factor for PFS and OS, and are associated with a higher metastatic burden. Furthermore, subanalyses showed that these differences depend on LNM localization.
Different cellular clones of RCC appear to differ in their propensity to spread to different organ sites [19] ; however, whether these differences impact on the survival outcome of mRCC patients is not completely understood. A previous IMDC study highlighted the prognostic importance of bone and liver metastases for survival outcome in mRCC patients, and suggested that aggressive RCC subclones tend to metastasize to these organs [16] . In addition, another study by our group showed that metastatic spread to other organs such as the pancreas or soft tissue tends to develop as a late clinical symptom, and has favorable treatment response and survival outcomes [20] . While these tumors may represent a favorable RCC subtype, our current results suggest that RCCs that spread to regional retroperitoneal lymph nodes represent another aggressive RCC subtype that develops bone metastases.
It has been suggested that LNM in ccRCC occur independently of von Hippel Lindau (VHL) gene inactivation [21] . Current agents used for mRCC treatment modulate pathways that are dysregulated because of VHL inactivation. This would imply that LNM might be less amenable to treatment with VEGF and mTOR inhibitors. However, our findings demonstrated no significant differences in best response to first-line therapies in patients with LNM versus those without LNM in both ccRCC and nccRCC patient groups. Nonetheless, patients with LNM had significantly shorter PFS than patients without LNM. While there was only a trend for an independent association between SBD LNM and short PFS, SPD+/SBD+ LNM were independently related to shorter PFS. Patients with SPD+/SBD+ LNM had additional hematogenous metastatic sites in 97% of cases. Thus, the high metastatic burden may explain the shorter response duration of first-line agents in this group of patients. However, a shorter duration for treatment success could also support the hypothesis that LNM are caused by RCC clones that do not have an inactive VHL gene function.
In the cytokine era, LNM were associated with shorter OS [1] and [2]. Our study demonstrates similar findings in the TT era. Interestingly, a worse survival outcome was observed for SBD LNM, both when occurring exclusively and in combination with thoracic lymph node involvement. Moreover, SBD LNM remained an independent prognostic factor for PFS and OS when analyzed without consideration of histologic subtype, and were an independent prognostic factor for OS in ccRCC. Conversely, SPD LNM were not an independent prognostic factor. This is notable and underscores the unique nature of RCCs that spread via retroperitoneal lymph nodes. Because the best response to TT did not different between SBD and SPD LNM, it appears that survival outcome is influenced by numerous factors. Patients with SBD LNM had faster disease progression before treatment initiation than those with SPD LNM (diagnosis to treatment <1 yr; 63% vs 46%; p < 0.001; data not shown). Despite a best response that was broadly similar, it seems that the disease progresses faster in SBD LNM patients during the later clinical course than in SPD LNM patients. In addition, SBD LNM were more often associated with anemia compared to SPD LNM (59% vs 48%; p = 0.008; data not shown). Anemia has multifactorial causes in cancer patients. However, one of the most important contributing factors is immune system dysregulation [22] . Dysregulation of the immune system may also be the reason for metastatic spread to SBD sites. Collectively, while risk factors that are currently unknown appear to dictate the clinical course in patients with SPD metastases, SBD metastases are independently associated with PFS and OS.
The majority of patients analyzed in this study did not undergo formal LND. Therefore, the effect of LND cannot be determined retrospectively. However, the current results indicate that SBD lymph nodes may have substantial negative effects on survival outcome and it is worth speculating whether extended removal of these metastatic sites could improve the clinical outcome for mRCC patients. The IMDC has recently shown that cytoreductive nephrectomy may provide therapeutic benefit in a large number of patients [23] . With regard to our current findings, it would be worth determining the role of LND in these patients.
The current study has several limitations that should be considered when interpreting the results. The IMDC database comprises retrospectively collected data for consecutive series of mRCC patients. These individual series, moreover, do not have standardized follow-up protocols, but this limitation may more accurately reflect real world outcomes. In addition, there was no centralized pathologic or radiologic review. Finally, lymph node status was determined clinically according to CT images in a number of cases, and it has been demonstrated that clinical and pathologic lymph node status is not necessarily always correlated.
LNM are associated with poor prognosis and adverse prognostic IMDC risk factors in mRCC patients treated with TT. Much of this increased risk associated with LNM appears to be dictated by SBD localization rather than SPD localization. Whether LND could improve the outcome for this group of patients remains unknown, but it is an important clinical question requiring prospective trials.
Author contributions: Daniel Y. Heng had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kroeger, Heng.
Acquisition of data: all authors.
Analysis and interpretation of data: Kroeger, Heng, Choueiri.
Drafting of the manuscript: Kroeger, Heng, Choueiri.
Critical revision of the manuscript for important intellectual content: all authors.
Statistical analysis: Kroeger, Heng.
Obtaining funding: None.
Administrative, technical, or material support: all authors.
Supervision: Choueiri, Heng.
Other (specify): None.
Financial disclosures: Daniel Y. Heng certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Benoit Beuselinck is an investigator of the EudraCT: 2011-006085-40/MetaSun trial supported by Pfizer. Frede Donskov has received research support from Novartis and GlaxoSmithKline. Toni K. Choueiri has received research funding from Pfizer and has an advisory role at Pfizer, GSK, Novartis, and Bayer. All other authors declare no conflicts of interest to the current study.
Funding/Support and role of the sponsor: None.
Acknowledgments: Dr. Choueiri is supported in part by the Trust family and the Loker Pinard and Michael Brigham Funds for Kidney Cancer Research at the Dana-Farber Cancer Institute. Dr. Kroeger thanks Dr. Klatte, Department of Urology, Medical University of Vienna, for statistical advice.
Renal cell carcinoma (RCC) is a heterogeneous disease consisting of several tumor types that have their own genetic, molecular, and clinical characteristics. Historical studies in the cytokine era by the University of California Los Angeles (UCLA) and the National Institutes of Health (NIH) described poor prognosis and worse treatment response to interleukin-2 (IL-2) in patients with metastatic involvement of the retroperitoneal lymph nodes [1] and [2]. Patients with and without retroperitoneal lymph node involvement had median overall survival (OS) times of 8.5 mo and 14.7 mo (p = 0.0004), respectively [2] . Pantuck et al [1] reported a better objective response rate in N0M1 compared to N+M1 patients (p = 0.01), with survival times of 10.5 mo and 20.4 mo, respectively. The adverse prognostic impact of retroperitoneal lymph node metastases (LNM) on survival outcome in patients with metastatic RCC (mRCC) was confirmed by several other institutions during the cytokine era [3], [4], and [5]. However, in the era of targeted therapies (TT), the survival outcome for patients with LNM has not yet been well characterized.
RCC preferentially metastasizes via hematogenous routes. Bianchi et al [6] recently described hematogenous metastatic sites in 80% of 11 157 patients with mRCC, while extension of the disease into lymph nodes was described only in 20% of patients. However, the authors did not provide details about the distribution of lymph node localization. The dominant regional LNM sites in RCC are retro- and paracaval, pre- and paraaortic, and interaortocaval lymph nodes in an anatomically intact retroperitoneum [7] . LNM can be unpredictable, and can also occur via direct extension to the thorax, supraclavicular lymph nodes, and iliac lymph nodes [7] . It has not been documented if particular sites of positive lymph node localization have a negative influence on the survival outcome of mRCC patients treated with TT.
Whether retroperitoneal lymph node dissection (LND) improves survival outcome is the subject of ongoing scientific debate. A phase 3 clinical trial (EORTC 30881) demonstrated positive lymph nodes in only 4% of investigated cases having no apparent involvement on computed tomography (CT) imaging. There was no advantage in OS, recurrence-free survival, or progression-free survival (PFS) for RCC patients who underwent LND [8] . The study has been criticized, however, because the majority of patients included had low-stage disease. Moreover, uniform surgical templates were not used for LND [9] . However, retrospective studies have demonstrated that LNM are more prevalent in high-risk patients with high Fuhrman grades, sarcomatoid features, locally advanced tumor stage, tumor size >10 cm, and tumor necrosis [10] . It has been suggested that LND is particularly beneficial in these high-risk patients [11] and [12]. Trends in the surgical management of RCC that have moved away from open radical surgery with extended LND to laparoscopic and robotic surgery with minimal hilar LND, combined with the negative results of phase 3 studies showing no additional benefit of LND over removal of the primary tumor alone, have decreased enthusiasm for performing extended LND. Therefore, it is common clinical practice to perform LND only in high-risk patients, and with a diagnostic rather than a curative intention [7] . Retrospective studies may be unable to add new insights into the current debate because only a prospective randomized trial with clear inclusion and exclusion criteria would be able to overcome this selection bias. Before planning such trials, however, it is prudent to better understand the impact of LNM and different lymph node localizations on the survival outcome for mRCC patients treated with current state-of-the-art therapies using data from the International mRCC Database Consortium (IMDC).
The IMDC database includes centers from North America (Canada, USA), Europe (Denmark, Greece, Belgium), Asia (Singapore, Japan, South Korea), and New Zealand. Data were collected from August 15, 2008 until December 31, 2013. At the time of analysis, the database contained data on 3405 patients who had received first-line targeted therapies between 2003 and 2013. The final study cohort comprised 2996 patients who were treated with first-line vascular endothelial growth factor (VEGF) inhibitors (n = 2823), mammalian target of rapamycin (mTOR) inhibitors (n = 165), or a combination of both (n = 8) therapies between April 2003 and August 2013. Patients were excluded from the analyses if they had received experimental therapies in first-line treatment (n = 7) or if no information was available on their lymph node status (n = 402; Fig. 1 ). Lymph node status was determined according to standard pathologic and CT criteria.
Information about LND was available for 890 patients. Full LND was performed in only 51 patients, and a limited (hilar) dissection was performed in 173 patients. These numbers were too low to allow any analyses and conclusions regarding the influence of LND on the survival outcome for mRCC patients treated with TT. Whether LND could be beneficial for mRCC patients was therefore not considered in the present study.
All centers obtained local Institutional Review Board approval before including data in the IMDC database. Baseline patient characteristics included demographic, clinicopathologic, and laboratory data, as previously described [13] . Survival data were retrospectively and prospectively collected from medical chart reviews and electronic records. Uniform data templates were used to ensure consistent data collection at each institution. The majority of patients were treated using standard care, but a subset received therapy as part of clinical trials. All patients were considered in consecutive series to avoid selection bias.
The primary hypothesis of the current study was that mRCC patients with LNM have worse survival outcome than patients without LNM when treated with TT. The secondary hypothesis was that worse survival outcome for mRCC patients with LNM is dictated by subdiaphragmatic (SPD) LNM rather than supradiaphragmatic (SBD) LNM.
Therefore, treatment response to current standard-of-care TT agents, OS, and PFS were compared between patients with LNM and those without LNM. We did not compare outcomes between the drugs used for treatment because these comparisons are frequently biased outside a clinical trial and there may not be sufficient power for such an analysis. Additional analyses were performed for subgroups of patients with exclusively SPD lymph node involvement, exclusively SBD lymph node involvement, or with lymph node involvement in both sites (SPD+/SBD+). Previous studies have demonstrated that tumor biology has an important impact on the survival outcome for mRCC patients [14] . Thus, the different biology of RCC was considered by separate analyses for clear cell RCC (ccRCC) and non-ccRCC (nccRCC) patients. Patients with SPD primarily had mediastinal lymph node involvement, while patients with SBD primarily had retroperitoneal lymph node involvement. Lymph node status was determined before initiation of TT. The selection process for patient groups analyzed in the study is shown in Figure 1 .
OS was defined as the time period between TT initiation and the date of death, with censoring on the day of the last follow-up visit. PFS was defined as the period between treatment initiation and progression, drug cessation, or death, with censoring at the last follow-up visit. Kaplan-Meier plots were used to estimate median OS and PFS, and univariate comparisons were performed using the log rank test. Uni- and multivariate (MVA) Cox regression analyses were performed to test the association of predefined prognostic factors and LNM with survival outcome. The analyses were performed with backward stepwise selection criteria, and significance was tested using the Wald statistic [13] . In MVA analyses, the IMDC (Heng) risk criteria were applied for adjustment [15] . The IMDC risk criteria include a time period from diagnosis to treatment of <1 yr, Karnofsky performance score <80%, anemia, hypercalcemia, thrombophilia, and neutrophilia. Upper and lower limits of normal for the laboratory parameters were based on institutional limits. In addition, adjustment was performed for bone and liver metastases because interactions of LNM with other metastatic sites should be considered. The prognostic relevance of bone and liver metastases has previously been demonstrated by our group [16] .
A secondary aim was to analyze baseline characteristics. Patient and tumor characteristics were compared using the Student t test for continuous variables, or the chi-square test and the Fisher exact test for categorical variables.
Statistical analyses were performed using SPSS version 22 (IBM, Chicago, IL, USA), and a two-sided value of p < 0.05 was considered statistically significant.
At the time of data analysis, 2525 (84.3%) patients had stopped taking their TT; the median time on TT was 6.0 mo (25th percentile, 2.7 mo; 75th percentile, 13.0 mo). Best overall response data to first-line therapy were available for 2559 patients. There were no significant differences in response to first-line TT (p = 0.059) when patients were compared without consideration of RCC subtype or LNM localization ( Table 1 ).
Table 1 Treatment response to targeted therapies in patients with lymph node metastases (LNM) compared to those without LNM
| Best response | |||||
|---|---|---|---|---|---|
| CR | PR | SD | PD | ||
| All RCC subtypes | |||||
| No LMN | n | 14 | 304 | 556 | 290 |
| % | 1.2 | 26.1 | 47.8 | 24.9 | |
| LMN | n | 9 | 314 | 690 | 382 |
| % | 0.6 | 22.5 | 49.5 | 27.4 | |
| Clear cell RCC | |||||
| No LNM | n | 3 | 108 | 261 | 95 |
| % | 0.6 | 23.1 | 55.9 | 20.3 | |
| SPD LNM | n | 2 | 73 | 149 | 55 |
| % | 0.7 | 26.2 | 53.4 | 19.7 | |
| SBD LNM | n | 2 | 50 | 92 | 42 |
| % | 1.1 | 26.9 | 49.5 | 22.6 | |
| SPD+/SBD+ LNM | n | 0 | 28 | 58 | 36 |
| % | 0.0 | 23.0 | 47.5 | 29.5 | |
| Non–clear cell RCC | |||||
| No LNM | n | 0 | 3 | 21 | 21 |
| % | 0.0 | 6.7 | 46.7 | 46.7 | |
| SPD LNM | n | 0 | 3 | 12 | 3 |
| % | 0.0 | 16.7 | 66.7 | 16.7 | |
| SBD LNM | n | 0 | 7 | 24 | 19 |
| % | 0.0 | 14.0 | 48.0 | 38.0 | |
| SPD+/SBD+ LNM | n | 1 | 4 | 10 | 6 |
| % | 4.8 | 19.0 | 47.6 | 28.6 | |
CR = complete remission; PR = partial response; SD = stable disease; PD = progressive disease; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm.
Both response data and LNM localization data were available for 1054 ccRCC patients and 134 nccRCC patients. Comparison of data by LNM localization revealed no difference in best response to first-line therapies for either ccRCC (p = 0.457) or nccRCC (p = 0.192; Table 1 ).
Patients with LNM generally had worse PFS compared to those with no LNM without considering LNM localization or RCC subtype (median PFS [95% confidence interval] 7.8 mo [7.1–8.5 mo] vs 6.1 [5.6–6.6 mo], log rank p = 0.001; Fig. 2 A).
Fig. 2 (A) Comparison of progression-free survival (PFS) in patients with positive lymph nodes versus patients without positive lymph nodes (median 7.8 mo vs. 6.1 mo) without consideration of histologic subtype. Patients with lymph node metastases (LNM) had worse PFS (hazard ratio [HR] 1.14, 95% confidence interval [CI] 1.06–1.24; p = 0.001). (B) Comparison of PFS according to lymph node localization. Patients without LNM had the best PFS (8.8 mo), and patients with both supradiaphragmatic (SPD) and subdiaphragmatic (SBD) LNM localizations (SPD+/SBD+) had the worst PFS (5.5 mo). Patients with SBD LNM had significantly worse PFS compared to patients with no LNM (6.8 vs 8.8 mo; p = 0.003) and those with SPD LNM (6.8 vs 8.3 mo; p = 0.022). Any metastatic involvement of retroperitoneal lymph nodes, either alone (HR 1.25, 95% CI 1.08–1.45) or in combination with SPD+/SBD+LNM (HR 1.50, 95% CI 1.24–1.74; p < 0.001) was associated with higher risk of progression. By contrast, SPD LNM alone was not a prognostic factor for progression (HR 1.03, 95% 0.895–1.18; p = 0.695). (C) PFS by lymph node localization in clear cell renal cell carcinoma (ccRCC; median PFS: no nodes, 9.9 mo; SPD LNM, 8.7 mo; SBD LNM, 8.0 mo; and both, 5.6 mo). Patients with SPD LNM (HR 1.09, 95% CI 0.94–1.25) and SBD LNM (HR 1.04, 95% CI 0.96–1.14) had similar PFS, and this was not significantly worse than the PFS of patients without LNM (p = 0.265 and p = 0.333 for SPD and SBD LNM, respectively).
Comparison of PFS rates according to LNM sites revealed that patients with SBD LNM had significantly shorter PFS (6.8 mo [5.7–8.0 mo]) compared to those without LNM (8.8 mo [7.6–10.0 mo]; p = 0.003) or SPD LNM (8.3 mo [7.2–9.6 mo]; p = 0.022). The worst PFS was observed in patients with SPD+/SBD+ LNM (5.4 mo [4.6–6.4 mo]), but this did not significantly differ from PFS in patients with SBD LNM (p = 0.093; Fig. 2 B).
In ccRCC, however, comparison according to LNM localization revealed no significant differences in PFS between patients without LNM (9.9 mo [8.6–11.1 mo]) and those with SPD LNM (8.7 mo [7.3–10.1 mo]) or SBD LNM (8.0 mo [6.9–9.2 mo]) (p = 0.312; Fig. 2 C). SPD+/SBD+ LNM patients had the worst survival outcome (5.6 mo [4.0–7.2 mo]) in comparison to patients without LNM (p < 0.001), with SPD LNM (p = 0.006), and with SBD LNM (p = 0.035; Fig. 2 C). In nccRCC, there was no difference in PFS between patients with no LNM (3.9 mo [2.9–5.0 mo]), SPD LNM (4.6 mo [0.0–9.9 mo]), SBD LNM (5.4 mo [4.7–6.0 mo]), and SPD+/SBD+ LNM (5.4 mo [3.2–7.3 mo]).
In MVA analyses, LNM remained an independent prognostic factor (HR 1.13 [1.03–1.24]; p = 0.013) when adjusted for IMDC risk criteria, as well as liver and bone metastases ( Table 2 ) [16] . In subgroup analysis, SBD LNM (HR 1.31 [1.08–1.60]; p = 0.006) and SPD+/SBD LNM (HR 1.29 [1.04–1.61]; p = 0.022) were independent prognostic factors for PFS, while SPD LNM was not independently associated with PFS (p = 1.00; Table 2 ). In ccRCC, only SPD+/SBD+ LNM (p = 0.025) was independently associated with PFS ( Table 2 ).
Table 2 Multivariate analyses for association between lymph node metastases (LNM) and progression-free survival (PFS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time <1 yr | 1.33 | 1.21–1.47 | <0.001 |
| Low hemoglobin | 1.37 | 1.24–1.52 | <0.001 |
| Neutrophilia | 1.43 | 1.25–1.63 | <0.001 |
| Thrombocytosis | 1.16 | 1.03–1.32 | 0.019 |
| Karnofsy PS <80% | 1.68 | 1.50–1.88 | <0.001 |
| LNM | 1.13 | 1.03–1.24 | 0.013 |
| Liver metastases | 1.26 | 1.12–1.41 | <0.001 |
| Bone metastases | 1.27 | 1.15–1.40 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.19–1.61 | <0.001 |
| Low hemoglobin | 1.42 | 1.21–1.67 | <0.001 |
| Neutrophilia | 1.46 | 1.19–1.80 | <0.001 |
| Thrombocytosis | 1.22 | 1.01–1.47 | 0.045 |
| Karnofsy PS <80% | 1.94 | 1.65–2.28 | <0.001 |
| Liver metastases | 1.39 | 1.18–1.65 | <0.001 |
| Bone metastases | 1.45 | 1.25–1.68 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.83–1.20 | 0.990 |
| SBD LNM | 1.32 | 1.08–1.60 | 0.006 |
| SPD+/SBD+ LNM | 1.29 | 1.04–1.61 | 0.022 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.17–1.64 | <0.001 |
| Low hemoglobin | 1.43 | 1.21–1.69 | <0.001 |
| Neutrophilia | 1.38 | 1.09–1.76 | 0.009 |
| Thrombocytosis | 1.97 | 1.64–2.38 | <0.001 |
| Karnofsy PS <80% | 1.51 | 1.24–1.85 | <0.001 |
| Bone metastases | 1.64 | 1.38–1.94 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.82–1.22 | 0.998 |
| SBD LNM | 1.23 | 0.98–1.55 | 0.077 |
| SPD+/SBD+ LNM | 1.33 | 1.04–1.71 | 0.025 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm; ccRCC = clear cell RCC.
At the time of the analysis, 1960 patients had died; the median OS time for the total patient cohort was 20.0 mo (25th percentile, 42.3 mo; 75th percentile, 7.9 mo). Patients with metastatic spread to lymph nodes had a worse survival outcome than patients without LNM (24.0 mo [21.6–26.3 mo] vs 16.0 mo [14.9–17.1 mo], p < 0.001; Fig. 3 A).
Fig. 3 (A) Patients with lymph node metastases (LNM) had a higher risk of death (median overall survival [OS] 23.9 mo, 95% confidence interval [CI] 21.6–26.3 mo) than those without LNM (median OS 16.0 mo, 95% CI 14.9–17.1; hazard ratio [HR] 1.41, 95% CI 1.28–1.54; p < 0.001). (B) A comparison without consideration of the histologic subtypes demonstrates that any lymph node involvement was associated with worse survival outcome compared to patients without LNM. Median survival times were 25.2 mo, 20.3 mo, 16.2 mo, and 11.5 mo in patients with no LNM and lymph node involvement in supradiaphragmatic (SPD), subdiaphragmatic (SBD), and both localizations (SPD+/SBD+), respectively. The HRs (95% CIs) for patients with SPD, SBD, and SPD+/SBD+ LNM were 1.24 (1.06–1.44), 1.45 (1.23–1.71), and 1.93 (1.60–2.33), respectively. (C) In clear cell renal cell carcinoma (ccRCC), patients with SPD and SBD LNM had similar median OS (SPD, 21.6 mo; SBD, 19.8 mo). However, both groups had a significantly higher risk of death (SPD: HR 1.24, 95% CI 1.05–1.47; SBD: HR 1.28, 95% CI 1.05–1.56) in comparison to those without lymph nodes. Patients with SPD+/SBD+ metastases had the worst OS (12.85 mo) and a twofold higher risk of death than patients without lymph node involvement (HR 2.04, 95% CI 1.68–2.48). (D) In non–clear cell RCC (nccRCC), median OS times according to lymph node localization were 15.8 mo for no lymph nodes, 16.1 mo for SPD LNM, 11.2 mo for SBD LNM, and 12.5 mo for SPD+/SBD+ LNM. There were no significant differences between the groups (log rank p = 0.444).
In subgroup analyses, patients without LNM had a median OS of 25.2 mo (21.9–28.5 mo) compared to 20.3 mo (17.1–23.5 mo) for SPD LNM (p = 0.093), 16.2 mo (13.5–20.0 mo) for SBD LNM (p < 0.001), and 11.5 mo (9.8–13.3 mo) (p < 0.001) for SPD+/SBD+ LNM ( Fig. 3 B).
In ccRCC, further analysis considering LNM localization revealed similar OS times for patients with SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13]), and were shorter than for patients without LNM (26.97 mo [23.32– 30.62], p = 0.01; Fig. 3 C). Patients with SPD+/SBD+ LNM had significantly (p = 0.001) worse OS (12.85 mo [9.45–16.24 mo]) compared to SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13 mo]; Fig. 3 C). In nccRCC, analyses according to LNM revealed no significant differences in OS (p = 0.444; Fig. 3 C).
For all histological subtypes, LNM were an independent prognostic factor for adverse OS in MVA analysis (p < 0.001; Table 3 ). Additional MVA according to LNM sites revealed that SBD LNM (p < 0.001) and SPD+/SBD+ LNM (p < 0.001) were independent prognostic factors for OS. SPD LNM were not independently associated with OS (p = 0.999; Table 3 ). In ccRCC, MVA according to LNM localization confirmed an independent association between SBD LNM (p = 0.028) and SPD+/SBD+ LNM (p = 0.006) and OS ( Table 3 ).
Table 3 Multivariate Cox regression analysis of association between lymph node metastases (LNM) and overall survival (OS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time < 1 yr | 1.34 | 1.19–1.49 | <0.001 |
| Low hemoglobin | 1.51 | 1.31–1.70 | <0.001 |
| High Ca | 1.47 | 1.25–1.73 | <0.001 |
| Neutrophilia | 1.67 | 1.45–1.92 | <0.001 |
| Thrombophilia | 1.31 | 1.14–1.51 | <0.001 |
| Karnofsky PS <80% | 1.88 | 1.67–2.13 | <0.001 |
| Liver metastases | 1.50 | 1.32–1.70 | <0.001 |
| Bone metastases | 1.50 | 1.34–1.67 | <0.001 |
| LNM | 1.33 | 1.19–1.48 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time < 1 yr | 1.32 | 1.11–1.56 | 0.001 |
| Low hemoglobin | 1.45 | 1.21–1.73 | <0.001 |
| High Ca | 1.38 | 1.08–1.77 | 0.010 |
| Neutrophilia | 1.78 | 1.43–2.22 | <0.001 |
| Thrombophilia | 1.49 | 1.21–1.83 | <0.001 |
| Karnofsky PS <80% | 2.19 | 1.82–2.62 | <0.001 |
| Liver metastases | 1.84 | 1.53–2.20 | <0.001 |
| Bone metastases | 1.61 | 1.36–1.90 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.57 | 0.98–1.48 | 0.075 |
| SBD LNM | 1.59 | 1.26–1.96 | <0.001 |
| SPD+/SBD+ LNM | 1.32 | 1.25–2.02 | <0.001 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time < 1 yr | 1.36 | 1.12–1.64 | 0.002 |
| Low hemoglobin | 1.36 | 1.11–1.67 | 0.003 |
| High Ca | 1.44 | 1.10–1.89 | 0.009 |
| Neutrophilia | 1.65 | 1.27–2.13 | <0.001 |
| Thrombophilia | 1.52 | 1.19–1.91 | 0.001 |
| Karnofsky PS <80% | 2.12 | 1.72–2.61 | <0.001 |
| Liver metastases | 1.94 | 1.57–2.41 | <0.001 |
| Bone metastases | 1.84 | 1.53–2.22 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.13 | 0.90–1.42 | 0.301 |
| SBD LNM | 1.34 | 1.03–1.74 | 0.028 |
| SPD+/SBD+ LNM | 1.47 | 1.11–1.93 | 0.006 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm; ccRCC = clear cell RCC.
An overall comparison of patients with and without LNM is shown in Table 4 . Patients with LNM presented more often with nccRCC (p < 0.001) and higher Fuhrman grades (p = 0.007), and had less often undergone prior nephrectomy (p < 0.001). The IMDC risk factors of high calcium (p = 0.001), thrombocytosis (p < 0.001), neutrophilia (p < 0.001), and Karnofsky performance status <80% (p = 0.009) were all more frequently seen in patients with LNM. Comparison of additional metastatic sites revealed that LNM are related to a higher overall metastatic burden (one or more additional metastatic site p < 0.001).
Table 4 General comparison of clinicopathologic features of patients with lymph node metastases (LNM) compared to those without LNM a
| No LNM (n = 1380) |
LNM (n = 1616) |
p value | |
|---|---|---|---|
| Mean age, yr (SD) | 58.65 (11.36) | 58. 50 (11.30) | 0.716 |
| Male, n (%) | 1003 (72.7) | 1182 (73.1) | 0.805 |
| Nephrectomy, n (%) | 1134 (82.4) | 1218 (75.4) | <0.001 |
| Fuhrman grade n (%) | 0.007 | ||
| 1 | 44 (4.2) | 33 (2.8) | |
| 2 | 302 (28.6) | 269 (23.1) | |
| 3 | 426 (40.3) | 501 (43.1) | |
| 4 | 284 (26.9) | 359 (30.9) | |
| Non–clear cell RCC, n (%) | 127 (9.8) | 253 (17.4) | <0.001 |
| Sarcomatoid features, n (%) | 143 (12.0) | 170 (12.0) | 0.586 |
| Therapy class, n (%) | |||
| AntiVEGF | 1316 (95.4) | 1507 (93.3) | |
| mTOR | 60 (4.3) | 105 (6.5) | |
| Combination | 4 (0.3) | 4 (0.2) | |
| DTT <1 yr, n (%) | 736 (53.4) | 905 (56.0) | 0.151 |
| Low HB (<LLN), n (%) | 691 (56.0) | 877 (58.1) | 0.261 |
| High calcium, n (%) | 98 (8.6) | 179 (12.6) | 0.001 |
| Neutrophilia, n (%) | 154 (13.1) | 286 (19.3) | <0.001 |
| Thrombocytosis, n (%) | 172 (15.4) | 299 (22.4) | <0.001 |
| Karnofsky PS <80%, n (%) | 261 (21.5) | 372 (25.9) | 0.009 |
| Lung metastases, n (%) | 939 (68.0) | 1149 (72.5) | 0.016 |
| Brain metastases, n (%) | 100 (7.2) | 141 (9.4) | 0.009 |
| Liver metastases, n (%) | 278 (20.2) | 328 (21.7) | 0.337 |
| Bone metastases, n (%) | 481 (34.9) | 551 (35.9) | 0.462 |
| Other metastic sites, n (%) | 470 (34.1) | 643 (43.1) | <0.001 |
| >1 metastatic site, n (%) | 805 (58.3%) | 1446 (89.5%) | <0.001 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; mTOR = mammalian target of rapamycin; PS = performance status; RCC = renal cell carcinoma; SD = standard deviation; VEGF = vascular endothelial growth factor.
The LNM localization distribution and clinicopathologic features in ccRCC and nccRCC are shown in Table 5 . In ccRCC, IMDC risk factors of neutrophilia (p = 0.049) and thrombophilia (p = 0.016) were more common. Patients with LNM had undergone nephrectomy less frequently (p = 0.001). In patients with ccRCC, those with LNM had concomitant lung (p < 0.001) and other metastases (eg, adrenal glands, other soft tissues; p = 0.009) more often than those without LNM; there was no difference in the frequency of bone (p = 0.276), liver (p = 0.943), or brain metastases (p = 0.187).
Table 5 Distribution of metastatic lymphatic sites in relation to the diaphragm in clear cell and non–clear cell renal carcinoma (RCC) a
| Clear cell RCC (n = 1218) |
Non–clear cell RCC (n = 159) |
|
|---|---|---|
| No LNM, n (%) | 552 (45.6) | 52 (32.7) |
| SPD LNM, n (%) | 321 (26.6) | 22 (13.8) |
| SBD LNM, n (%) | 204 (16.9) | 54 (34.0) |
| SPD+/SBD+ LNM, n (%) | 132 (10.9) | 31 (19.5) |
| LMN | No LMN | p value | LMN | No LMN | p value | |
|---|---|---|---|---|---|---|
| Nephrectomy, n (%) | 542 (81.2) | 492 (88.2) | 0.001 | 82 (76.6) | 43 (82.7) | 0.418 |
| Male, n (%) | 515 (77.7) | 403 (72.6) | 0.045 | 75 (70.1) | 37 (71.2) | 1.00 |
| DTT <1 yr, n (%) | 348 (52.5) | 265 (47.9) | 0.120 | 65 (60.7) | 29 (55.8) | 0.607 |
| Low HB (<LLN) | 331 (54.0) | 233 (48.7) | 0.088 | 49 (51.0) | 19 (45.2) | 0.582 |
| High calcium, n (%) | 70 (12.3) | 40 (9.5) | 0.184 | 5 (5.7) | 4 (12.1) | 0.254 |
| Neutrophilia, n (%) | 95 (15.9) | 53 (11.6) | 0.049 | 24 (25.3) | 7 (16.7) | 0.376 |
| Thrombophilia, n (%) | 106 (22.0) | 63 (15.6) | 0.016 | 17 (21.0) | 10 (25.6) | 0.642 |
| Karnofsky PS <80%, n (%) | 145 (24.7) | 95 (20.0) | 0.076 | 24 (24.7) | 10 (20.4) | 0.680 |
| Sarcomatoid features, n (%) | 52 (9.1) | 62 (12.6) | 0.074 | 14 (15.4) | 7 (15.6) | 1.00 |
| Lung metastases, n (%) | 507 (76.8) | 372 (67.0) | <0.001 | 52 (49.1) | 20 (38.5) | 0.237 |
| Brain metastases, n (%) | 65 (9.8) | 42 (7.6) | 0.187 | 4 (3.8) | 2 (3.8) | 1.00 |
| Liver metastases, n (%) | 132 (20.0) | 112 (20.2) | 0.943 | 32 (30.2) | 28 (53.8) | 0.005 |
| Bone metastases, n (%) | 220 (33.3) | 202 (36.4) | 0.276 | 35 (33.0) | 20 (38.5) | 0.594 |
| Other metastatic sites, n (%) | 287 (45.7) | 211 (38.2) | 0.009 | 40 (40.8) | 20 (38.5) | 0.862 |
| >1 metastatic site, n (%) | 602 (90.8) | 320 (57.7) | <0.001 | 87 (81.3) | 30 (7.7) | 0.002 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; PS = performance status; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm.
In nccRCC, liver metastases were found more often in patients without LNM (p = 0.005). A comparison of other clinicopathologic features between nccRCC patients with and without LNM revealed no significant differences ( Table 5 ).
It has previously been shown that only a minority of patients with kidney cancer undergo LND at the time of nephrectomy. This is the case even for high-risk patients [17] , even though retrospective studies have suggested a therapeutic benefit of LND [11] and [18]. It remains uncertain whether extended LND can improve the survival outcome for mRCC patients. However, if it were shown that SBD LNM affect the survival outcome for mRCC patients, this could form the basis for a rationale in favor of performing LND, or at least justify the need for prospective clinical trials to address this question. In this regard, an understanding of the influence of LNM on treatment response and survival outcome for mRCC patients undergoing TT could make an important contribution to the debate on surgical management of RCC.
Our study has demonstrated that LNM are associated with advanced IMDC risk factors, are an adverse prognostic factor for PFS and OS, and are associated with a higher metastatic burden. Furthermore, subanalyses showed that these differences depend on LNM localization.
Different cellular clones of RCC appear to differ in their propensity to spread to different organ sites [19] ; however, whether these differences impact on the survival outcome of mRCC patients is not completely understood. A previous IMDC study highlighted the prognostic importance of bone and liver metastases for survival outcome in mRCC patients, and suggested that aggressive RCC subclones tend to metastasize to these organs [16] . In addition, another study by our group showed that metastatic spread to other organs such as the pancreas or soft tissue tends to develop as a late clinical symptom, and has favorable treatment response and survival outcomes [20] . While these tumors may represent a favorable RCC subtype, our current results suggest that RCCs that spread to regional retroperitoneal lymph nodes represent another aggressive RCC subtype that develops bone metastases.
It has been suggested that LNM in ccRCC occur independently of von Hippel Lindau (VHL) gene inactivation [21] . Current agents used for mRCC treatment modulate pathways that are dysregulated because of VHL inactivation. This would imply that LNM might be less amenable to treatment with VEGF and mTOR inhibitors. However, our findings demonstrated no significant differences in best response to first-line therapies in patients with LNM versus those without LNM in both ccRCC and nccRCC patient groups. Nonetheless, patients with LNM had significantly shorter PFS than patients without LNM. While there was only a trend for an independent association between SBD LNM and short PFS, SPD+/SBD+ LNM were independently related to shorter PFS. Patients with SPD+/SBD+ LNM had additional hematogenous metastatic sites in 97% of cases. Thus, the high metastatic burden may explain the shorter response duration of first-line agents in this group of patients. However, a shorter duration for treatment success could also support the hypothesis that LNM are caused by RCC clones that do not have an inactive VHL gene function.
In the cytokine era, LNM were associated with shorter OS [1] and [2]. Our study demonstrates similar findings in the TT era. Interestingly, a worse survival outcome was observed for SBD LNM, both when occurring exclusively and in combination with thoracic lymph node involvement. Moreover, SBD LNM remained an independent prognostic factor for PFS and OS when analyzed without consideration of histologic subtype, and were an independent prognostic factor for OS in ccRCC. Conversely, SPD LNM were not an independent prognostic factor. This is notable and underscores the unique nature of RCCs that spread via retroperitoneal lymph nodes. Because the best response to TT did not different between SBD and SPD LNM, it appears that survival outcome is influenced by numerous factors. Patients with SBD LNM had faster disease progression before treatment initiation than those with SPD LNM (diagnosis to treatment <1 yr; 63% vs 46%; p < 0.001; data not shown). Despite a best response that was broadly similar, it seems that the disease progresses faster in SBD LNM patients during the later clinical course than in SPD LNM patients. In addition, SBD LNM were more often associated with anemia compared to SPD LNM (59% vs 48%; p = 0.008; data not shown). Anemia has multifactorial causes in cancer patients. However, one of the most important contributing factors is immune system dysregulation [22] . Dysregulation of the immune system may also be the reason for metastatic spread to SBD sites. Collectively, while risk factors that are currently unknown appear to dictate the clinical course in patients with SPD metastases, SBD metastases are independently associated with PFS and OS.
The majority of patients analyzed in this study did not undergo formal LND. Therefore, the effect of LND cannot be determined retrospectively. However, the current results indicate that SBD lymph nodes may have substantial negative effects on survival outcome and it is worth speculating whether extended removal of these metastatic sites could improve the clinical outcome for mRCC patients. The IMDC has recently shown that cytoreductive nephrectomy may provide therapeutic benefit in a large number of patients [23] . With regard to our current findings, it would be worth determining the role of LND in these patients.
The current study has several limitations that should be considered when interpreting the results. The IMDC database comprises retrospectively collected data for consecutive series of mRCC patients. These individual series, moreover, do not have standardized follow-up protocols, but this limitation may more accurately reflect real world outcomes. In addition, there was no centralized pathologic or radiologic review. Finally, lymph node status was determined clinically according to CT images in a number of cases, and it has been demonstrated that clinical and pathologic lymph node status is not necessarily always correlated.
LNM are associated with poor prognosis and adverse prognostic IMDC risk factors in mRCC patients treated with TT. Much of this increased risk associated with LNM appears to be dictated by SBD localization rather than SPD localization. Whether LND could improve the outcome for this group of patients remains unknown, but it is an important clinical question requiring prospective trials.
Author contributions: Daniel Y. Heng had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kroeger, Heng.
Acquisition of data: all authors.
Analysis and interpretation of data: Kroeger, Heng, Choueiri.
Drafting of the manuscript: Kroeger, Heng, Choueiri.
Critical revision of the manuscript for important intellectual content: all authors.
Statistical analysis: Kroeger, Heng.
Obtaining funding: None.
Administrative, technical, or material support: all authors.
Supervision: Choueiri, Heng.
Other (specify): None.
Financial disclosures: Daniel Y. Heng certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Benoit Beuselinck is an investigator of the EudraCT: 2011-006085-40/MetaSun trial supported by Pfizer. Frede Donskov has received research support from Novartis and GlaxoSmithKline. Toni K. Choueiri has received research funding from Pfizer and has an advisory role at Pfizer, GSK, Novartis, and Bayer. All other authors declare no conflicts of interest to the current study.
Funding/Support and role of the sponsor: None.
Acknowledgments: Dr. Choueiri is supported in part by the Trust family and the Loker Pinard and Michael Brigham Funds for Kidney Cancer Research at the Dana-Farber Cancer Institute. Dr. Kroeger thanks Dr. Klatte, Department of Urology, Medical University of Vienna, for statistical advice.
Renal cell carcinoma (RCC) is a heterogeneous disease consisting of several tumor types that have their own genetic, molecular, and clinical characteristics. Historical studies in the cytokine era by the University of California Los Angeles (UCLA) and the National Institutes of Health (NIH) described poor prognosis and worse treatment response to interleukin-2 (IL-2) in patients with metastatic involvement of the retroperitoneal lymph nodes [1] and [2]. Patients with and without retroperitoneal lymph node involvement had median overall survival (OS) times of 8.5 mo and 14.7 mo (p = 0.0004), respectively [2] . Pantuck et al [1] reported a better objective response rate in N0M1 compared to N+M1 patients (p = 0.01), with survival times of 10.5 mo and 20.4 mo, respectively. The adverse prognostic impact of retroperitoneal lymph node metastases (LNM) on survival outcome in patients with metastatic RCC (mRCC) was confirmed by several other institutions during the cytokine era [3], [4], and [5]. However, in the era of targeted therapies (TT), the survival outcome for patients with LNM has not yet been well characterized.
RCC preferentially metastasizes via hematogenous routes. Bianchi et al [6] recently described hematogenous metastatic sites in 80% of 11 157 patients with mRCC, while extension of the disease into lymph nodes was described only in 20% of patients. However, the authors did not provide details about the distribution of lymph node localization. The dominant regional LNM sites in RCC are retro- and paracaval, pre- and paraaortic, and interaortocaval lymph nodes in an anatomically intact retroperitoneum [7] . LNM can be unpredictable, and can also occur via direct extension to the thorax, supraclavicular lymph nodes, and iliac lymph nodes [7] . It has not been documented if particular sites of positive lymph node localization have a negative influence on the survival outcome of mRCC patients treated with TT.
Whether retroperitoneal lymph node dissection (LND) improves survival outcome is the subject of ongoing scientific debate. A phase 3 clinical trial (EORTC 30881) demonstrated positive lymph nodes in only 4% of investigated cases having no apparent involvement on computed tomography (CT) imaging. There was no advantage in OS, recurrence-free survival, or progression-free survival (PFS) for RCC patients who underwent LND [8] . The study has been criticized, however, because the majority of patients included had low-stage disease. Moreover, uniform surgical templates were not used for LND [9] . However, retrospective studies have demonstrated that LNM are more prevalent in high-risk patients with high Fuhrman grades, sarcomatoid features, locally advanced tumor stage, tumor size >10 cm, and tumor necrosis [10] . It has been suggested that LND is particularly beneficial in these high-risk patients [11] and [12]. Trends in the surgical management of RCC that have moved away from open radical surgery with extended LND to laparoscopic and robotic surgery with minimal hilar LND, combined with the negative results of phase 3 studies showing no additional benefit of LND over removal of the primary tumor alone, have decreased enthusiasm for performing extended LND. Therefore, it is common clinical practice to perform LND only in high-risk patients, and with a diagnostic rather than a curative intention [7] . Retrospective studies may be unable to add new insights into the current debate because only a prospective randomized trial with clear inclusion and exclusion criteria would be able to overcome this selection bias. Before planning such trials, however, it is prudent to better understand the impact of LNM and different lymph node localizations on the survival outcome for mRCC patients treated with current state-of-the-art therapies using data from the International mRCC Database Consortium (IMDC).
The IMDC database includes centers from North America (Canada, USA), Europe (Denmark, Greece, Belgium), Asia (Singapore, Japan, South Korea), and New Zealand. Data were collected from August 15, 2008 until December 31, 2013. At the time of analysis, the database contained data on 3405 patients who had received first-line targeted therapies between 2003 and 2013. The final study cohort comprised 2996 patients who were treated with first-line vascular endothelial growth factor (VEGF) inhibitors (n = 2823), mammalian target of rapamycin (mTOR) inhibitors (n = 165), or a combination of both (n = 8) therapies between April 2003 and August 2013. Patients were excluded from the analyses if they had received experimental therapies in first-line treatment (n = 7) or if no information was available on their lymph node status (n = 402; Fig. 1 ). Lymph node status was determined according to standard pathologic and CT criteria.
Information about LND was available for 890 patients. Full LND was performed in only 51 patients, and a limited (hilar) dissection was performed in 173 patients. These numbers were too low to allow any analyses and conclusions regarding the influence of LND on the survival outcome for mRCC patients treated with TT. Whether LND could be beneficial for mRCC patients was therefore not considered in the present study.
All centers obtained local Institutional Review Board approval before including data in the IMDC database. Baseline patient characteristics included demographic, clinicopathologic, and laboratory data, as previously described [13] . Survival data were retrospectively and prospectively collected from medical chart reviews and electronic records. Uniform data templates were used to ensure consistent data collection at each institution. The majority of patients were treated using standard care, but a subset received therapy as part of clinical trials. All patients were considered in consecutive series to avoid selection bias.
The primary hypothesis of the current study was that mRCC patients with LNM have worse survival outcome than patients without LNM when treated with TT. The secondary hypothesis was that worse survival outcome for mRCC patients with LNM is dictated by subdiaphragmatic (SPD) LNM rather than supradiaphragmatic (SBD) LNM.
Therefore, treatment response to current standard-of-care TT agents, OS, and PFS were compared between patients with LNM and those without LNM. We did not compare outcomes between the drugs used for treatment because these comparisons are frequently biased outside a clinical trial and there may not be sufficient power for such an analysis. Additional analyses were performed for subgroups of patients with exclusively SPD lymph node involvement, exclusively SBD lymph node involvement, or with lymph node involvement in both sites (SPD+/SBD+). Previous studies have demonstrated that tumor biology has an important impact on the survival outcome for mRCC patients [14] . Thus, the different biology of RCC was considered by separate analyses for clear cell RCC (ccRCC) and non-ccRCC (nccRCC) patients. Patients with SPD primarily had mediastinal lymph node involvement, while patients with SBD primarily had retroperitoneal lymph node involvement. Lymph node status was determined before initiation of TT. The selection process for patient groups analyzed in the study is shown in Figure 1 .
OS was defined as the time period between TT initiation and the date of death, with censoring on the day of the last follow-up visit. PFS was defined as the period between treatment initiation and progression, drug cessation, or death, with censoring at the last follow-up visit. Kaplan-Meier plots were used to estimate median OS and PFS, and univariate comparisons were performed using the log rank test. Uni- and multivariate (MVA) Cox regression analyses were performed to test the association of predefined prognostic factors and LNM with survival outcome. The analyses were performed with backward stepwise selection criteria, and significance was tested using the Wald statistic [13] . In MVA analyses, the IMDC (Heng) risk criteria were applied for adjustment [15] . The IMDC risk criteria include a time period from diagnosis to treatment of <1 yr, Karnofsky performance score <80%, anemia, hypercalcemia, thrombophilia, and neutrophilia. Upper and lower limits of normal for the laboratory parameters were based on institutional limits. In addition, adjustment was performed for bone and liver metastases because interactions of LNM with other metastatic sites should be considered. The prognostic relevance of bone and liver metastases has previously been demonstrated by our group [16] .
A secondary aim was to analyze baseline characteristics. Patient and tumor characteristics were compared using the Student t test for continuous variables, or the chi-square test and the Fisher exact test for categorical variables.
Statistical analyses were performed using SPSS version 22 (IBM, Chicago, IL, USA), and a two-sided value of p < 0.05 was considered statistically significant.
At the time of data analysis, 2525 (84.3%) patients had stopped taking their TT; the median time on TT was 6.0 mo (25th percentile, 2.7 mo; 75th percentile, 13.0 mo). Best overall response data to first-line therapy were available for 2559 patients. There were no significant differences in response to first-line TT (p = 0.059) when patients were compared without consideration of RCC subtype or LNM localization ( Table 1 ).
Table 1 Treatment response to targeted therapies in patients with lymph node metastases (LNM) compared to those without LNM
| Best response | |||||
|---|---|---|---|---|---|
| CR | PR | SD | PD | ||
| All RCC subtypes | |||||
| No LMN | n | 14 | 304 | 556 | 290 |
| % | 1.2 | 26.1 | 47.8 | 24.9 | |
| LMN | n | 9 | 314 | 690 | 382 |
| % | 0.6 | 22.5 | 49.5 | 27.4 | |
| Clear cell RCC | |||||
| No LNM | n | 3 | 108 | 261 | 95 |
| % | 0.6 | 23.1 | 55.9 | 20.3 | |
| SPD LNM | n | 2 | 73 | 149 | 55 |
| % | 0.7 | 26.2 | 53.4 | 19.7 | |
| SBD LNM | n | 2 | 50 | 92 | 42 |
| % | 1.1 | 26.9 | 49.5 | 22.6 | |
| SPD+/SBD+ LNM | n | 0 | 28 | 58 | 36 |
| % | 0.0 | 23.0 | 47.5 | 29.5 | |
| Non–clear cell RCC | |||||
| No LNM | n | 0 | 3 | 21 | 21 |
| % | 0.0 | 6.7 | 46.7 | 46.7 | |
| SPD LNM | n | 0 | 3 | 12 | 3 |
| % | 0.0 | 16.7 | 66.7 | 16.7 | |
| SBD LNM | n | 0 | 7 | 24 | 19 |
| % | 0.0 | 14.0 | 48.0 | 38.0 | |
| SPD+/SBD+ LNM | n | 1 | 4 | 10 | 6 |
| % | 4.8 | 19.0 | 47.6 | 28.6 | |
CR = complete remission; PR = partial response; SD = stable disease; PD = progressive disease; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm.
Both response data and LNM localization data were available for 1054 ccRCC patients and 134 nccRCC patients. Comparison of data by LNM localization revealed no difference in best response to first-line therapies for either ccRCC (p = 0.457) or nccRCC (p = 0.192; Table 1 ).
Patients with LNM generally had worse PFS compared to those with no LNM without considering LNM localization or RCC subtype (median PFS [95% confidence interval] 7.8 mo [7.1–8.5 mo] vs 6.1 [5.6–6.6 mo], log rank p = 0.001; Fig. 2 A).
Fig. 2 (A) Comparison of progression-free survival (PFS) in patients with positive lymph nodes versus patients without positive lymph nodes (median 7.8 mo vs. 6.1 mo) without consideration of histologic subtype. Patients with lymph node metastases (LNM) had worse PFS (hazard ratio [HR] 1.14, 95% confidence interval [CI] 1.06–1.24; p = 0.001). (B) Comparison of PFS according to lymph node localization. Patients without LNM had the best PFS (8.8 mo), and patients with both supradiaphragmatic (SPD) and subdiaphragmatic (SBD) LNM localizations (SPD+/SBD+) had the worst PFS (5.5 mo). Patients with SBD LNM had significantly worse PFS compared to patients with no LNM (6.8 vs 8.8 mo; p = 0.003) and those with SPD LNM (6.8 vs 8.3 mo; p = 0.022). Any metastatic involvement of retroperitoneal lymph nodes, either alone (HR 1.25, 95% CI 1.08–1.45) or in combination with SPD+/SBD+LNM (HR 1.50, 95% CI 1.24–1.74; p < 0.001) was associated with higher risk of progression. By contrast, SPD LNM alone was not a prognostic factor for progression (HR 1.03, 95% 0.895–1.18; p = 0.695). (C) PFS by lymph node localization in clear cell renal cell carcinoma (ccRCC; median PFS: no nodes, 9.9 mo; SPD LNM, 8.7 mo; SBD LNM, 8.0 mo; and both, 5.6 mo). Patients with SPD LNM (HR 1.09, 95% CI 0.94–1.25) and SBD LNM (HR 1.04, 95% CI 0.96–1.14) had similar PFS, and this was not significantly worse than the PFS of patients without LNM (p = 0.265 and p = 0.333 for SPD and SBD LNM, respectively).
Comparison of PFS rates according to LNM sites revealed that patients with SBD LNM had significantly shorter PFS (6.8 mo [5.7–8.0 mo]) compared to those without LNM (8.8 mo [7.6–10.0 mo]; p = 0.003) or SPD LNM (8.3 mo [7.2–9.6 mo]; p = 0.022). The worst PFS was observed in patients with SPD+/SBD+ LNM (5.4 mo [4.6–6.4 mo]), but this did not significantly differ from PFS in patients with SBD LNM (p = 0.093; Fig. 2 B).
In ccRCC, however, comparison according to LNM localization revealed no significant differences in PFS between patients without LNM (9.9 mo [8.6–11.1 mo]) and those with SPD LNM (8.7 mo [7.3–10.1 mo]) or SBD LNM (8.0 mo [6.9–9.2 mo]) (p = 0.312; Fig. 2 C). SPD+/SBD+ LNM patients had the worst survival outcome (5.6 mo [4.0–7.2 mo]) in comparison to patients without LNM (p < 0.001), with SPD LNM (p = 0.006), and with SBD LNM (p = 0.035; Fig. 2 C). In nccRCC, there was no difference in PFS between patients with no LNM (3.9 mo [2.9–5.0 mo]), SPD LNM (4.6 mo [0.0–9.9 mo]), SBD LNM (5.4 mo [4.7–6.0 mo]), and SPD+/SBD+ LNM (5.4 mo [3.2–7.3 mo]).
In MVA analyses, LNM remained an independent prognostic factor (HR 1.13 [1.03–1.24]; p = 0.013) when adjusted for IMDC risk criteria, as well as liver and bone metastases ( Table 2 ) [16] . In subgroup analysis, SBD LNM (HR 1.31 [1.08–1.60]; p = 0.006) and SPD+/SBD LNM (HR 1.29 [1.04–1.61]; p = 0.022) were independent prognostic factors for PFS, while SPD LNM was not independently associated with PFS (p = 1.00; Table 2 ). In ccRCC, only SPD+/SBD+ LNM (p = 0.025) was independently associated with PFS ( Table 2 ).
Table 2 Multivariate analyses for association between lymph node metastases (LNM) and progression-free survival (PFS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time <1 yr | 1.33 | 1.21–1.47 | <0.001 |
| Low hemoglobin | 1.37 | 1.24–1.52 | <0.001 |
| Neutrophilia | 1.43 | 1.25–1.63 | <0.001 |
| Thrombocytosis | 1.16 | 1.03–1.32 | 0.019 |
| Karnofsy PS <80% | 1.68 | 1.50–1.88 | <0.001 |
| LNM | 1.13 | 1.03–1.24 | 0.013 |
| Liver metastases | 1.26 | 1.12–1.41 | <0.001 |
| Bone metastases | 1.27 | 1.15–1.40 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.19–1.61 | <0.001 |
| Low hemoglobin | 1.42 | 1.21–1.67 | <0.001 |
| Neutrophilia | 1.46 | 1.19–1.80 | <0.001 |
| Thrombocytosis | 1.22 | 1.01–1.47 | 0.045 |
| Karnofsy PS <80% | 1.94 | 1.65–2.28 | <0.001 |
| Liver metastases | 1.39 | 1.18–1.65 | <0.001 |
| Bone metastases | 1.45 | 1.25–1.68 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.83–1.20 | 0.990 |
| SBD LNM | 1.32 | 1.08–1.60 | 0.006 |
| SPD+/SBD+ LNM | 1.29 | 1.04–1.61 | 0.022 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.17–1.64 | <0.001 |
| Low hemoglobin | 1.43 | 1.21–1.69 | <0.001 |
| Neutrophilia | 1.38 | 1.09–1.76 | 0.009 |
| Thrombocytosis | 1.97 | 1.64–2.38 | <0.001 |
| Karnofsy PS <80% | 1.51 | 1.24–1.85 | <0.001 |
| Bone metastases | 1.64 | 1.38–1.94 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.82–1.22 | 0.998 |
| SBD LNM | 1.23 | 0.98–1.55 | 0.077 |
| SPD+/SBD+ LNM | 1.33 | 1.04–1.71 | 0.025 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm; ccRCC = clear cell RCC.
At the time of the analysis, 1960 patients had died; the median OS time for the total patient cohort was 20.0 mo (25th percentile, 42.3 mo; 75th percentile, 7.9 mo). Patients with metastatic spread to lymph nodes had a worse survival outcome than patients without LNM (24.0 mo [21.6–26.3 mo] vs 16.0 mo [14.9–17.1 mo], p < 0.001; Fig. 3 A).
Fig. 3 (A) Patients with lymph node metastases (LNM) had a higher risk of death (median overall survival [OS] 23.9 mo, 95% confidence interval [CI] 21.6–26.3 mo) than those without LNM (median OS 16.0 mo, 95% CI 14.9–17.1; hazard ratio [HR] 1.41, 95% CI 1.28–1.54; p < 0.001). (B) A comparison without consideration of the histologic subtypes demonstrates that any lymph node involvement was associated with worse survival outcome compared to patients without LNM. Median survival times were 25.2 mo, 20.3 mo, 16.2 mo, and 11.5 mo in patients with no LNM and lymph node involvement in supradiaphragmatic (SPD), subdiaphragmatic (SBD), and both localizations (SPD+/SBD+), respectively. The HRs (95% CIs) for patients with SPD, SBD, and SPD+/SBD+ LNM were 1.24 (1.06–1.44), 1.45 (1.23–1.71), and 1.93 (1.60–2.33), respectively. (C) In clear cell renal cell carcinoma (ccRCC), patients with SPD and SBD LNM had similar median OS (SPD, 21.6 mo; SBD, 19.8 mo). However, both groups had a significantly higher risk of death (SPD: HR 1.24, 95% CI 1.05–1.47; SBD: HR 1.28, 95% CI 1.05–1.56) in comparison to those without lymph nodes. Patients with SPD+/SBD+ metastases had the worst OS (12.85 mo) and a twofold higher risk of death than patients without lymph node involvement (HR 2.04, 95% CI 1.68–2.48). (D) In non–clear cell RCC (nccRCC), median OS times according to lymph node localization were 15.8 mo for no lymph nodes, 16.1 mo for SPD LNM, 11.2 mo for SBD LNM, and 12.5 mo for SPD+/SBD+ LNM. There were no significant differences between the groups (log rank p = 0.444).
In subgroup analyses, patients without LNM had a median OS of 25.2 mo (21.9–28.5 mo) compared to 20.3 mo (17.1–23.5 mo) for SPD LNM (p = 0.093), 16.2 mo (13.5–20.0 mo) for SBD LNM (p < 0.001), and 11.5 mo (9.8–13.3 mo) (p < 0.001) for SPD+/SBD+ LNM ( Fig. 3 B).
In ccRCC, further analysis considering LNM localization revealed similar OS times for patients with SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13]), and were shorter than for patients without LNM (26.97 mo [23.32– 30.62], p = 0.01; Fig. 3 C). Patients with SPD+/SBD+ LNM had significantly (p = 0.001) worse OS (12.85 mo [9.45–16.24 mo]) compared to SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13 mo]; Fig. 3 C). In nccRCC, analyses according to LNM revealed no significant differences in OS (p = 0.444; Fig. 3 C).
For all histological subtypes, LNM were an independent prognostic factor for adverse OS in MVA analysis (p < 0.001; Table 3 ). Additional MVA according to LNM sites revealed that SBD LNM (p < 0.001) and SPD+/SBD+ LNM (p < 0.001) were independent prognostic factors for OS. SPD LNM were not independently associated with OS (p = 0.999; Table 3 ). In ccRCC, MVA according to LNM localization confirmed an independent association between SBD LNM (p = 0.028) and SPD+/SBD+ LNM (p = 0.006) and OS ( Table 3 ).
Table 3 Multivariate Cox regression analysis of association between lymph node metastases (LNM) and overall survival (OS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time < 1 yr | 1.34 | 1.19–1.49 | <0.001 |
| Low hemoglobin | 1.51 | 1.31–1.70 | <0.001 |
| High Ca | 1.47 | 1.25–1.73 | <0.001 |
| Neutrophilia | 1.67 | 1.45–1.92 | <0.001 |
| Thrombophilia | 1.31 | 1.14–1.51 | <0.001 |
| Karnofsky PS <80% | 1.88 | 1.67–2.13 | <0.001 |
| Liver metastases | 1.50 | 1.32–1.70 | <0.001 |
| Bone metastases | 1.50 | 1.34–1.67 | <0.001 |
| LNM | 1.33 | 1.19–1.48 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time < 1 yr | 1.32 | 1.11–1.56 | 0.001 |
| Low hemoglobin | 1.45 | 1.21–1.73 | <0.001 |
| High Ca | 1.38 | 1.08–1.77 | 0.010 |
| Neutrophilia | 1.78 | 1.43–2.22 | <0.001 |
| Thrombophilia | 1.49 | 1.21–1.83 | <0.001 |
| Karnofsky PS <80% | 2.19 | 1.82–2.62 | <0.001 |
| Liver metastases | 1.84 | 1.53–2.20 | <0.001 |
| Bone metastases | 1.61 | 1.36–1.90 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.57 | 0.98–1.48 | 0.075 |
| SBD LNM | 1.59 | 1.26–1.96 | <0.001 |
| SPD+/SBD+ LNM | 1.32 | 1.25–2.02 | <0.001 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time < 1 yr | 1.36 | 1.12–1.64 | 0.002 |
| Low hemoglobin | 1.36 | 1.11–1.67 | 0.003 |
| High Ca | 1.44 | 1.10–1.89 | 0.009 |
| Neutrophilia | 1.65 | 1.27–2.13 | <0.001 |
| Thrombophilia | 1.52 | 1.19–1.91 | 0.001 |
| Karnofsky PS <80% | 2.12 | 1.72–2.61 | <0.001 |
| Liver metastases | 1.94 | 1.57–2.41 | <0.001 |
| Bone metastases | 1.84 | 1.53–2.22 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.13 | 0.90–1.42 | 0.301 |
| SBD LNM | 1.34 | 1.03–1.74 | 0.028 |
| SPD+/SBD+ LNM | 1.47 | 1.11–1.93 | 0.006 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm; ccRCC = clear cell RCC.
An overall comparison of patients with and without LNM is shown in Table 4 . Patients with LNM presented more often with nccRCC (p < 0.001) and higher Fuhrman grades (p = 0.007), and had less often undergone prior nephrectomy (p < 0.001). The IMDC risk factors of high calcium (p = 0.001), thrombocytosis (p < 0.001), neutrophilia (p < 0.001), and Karnofsky performance status <80% (p = 0.009) were all more frequently seen in patients with LNM. Comparison of additional metastatic sites revealed that LNM are related to a higher overall metastatic burden (one or more additional metastatic site p < 0.001).
Table 4 General comparison of clinicopathologic features of patients with lymph node metastases (LNM) compared to those without LNM a
| No LNM (n = 1380) |
LNM (n = 1616) |
p value | |
|---|---|---|---|
| Mean age, yr (SD) | 58.65 (11.36) | 58. 50 (11.30) | 0.716 |
| Male, n (%) | 1003 (72.7) | 1182 (73.1) | 0.805 |
| Nephrectomy, n (%) | 1134 (82.4) | 1218 (75.4) | <0.001 |
| Fuhrman grade n (%) | 0.007 | ||
| 1 | 44 (4.2) | 33 (2.8) | |
| 2 | 302 (28.6) | 269 (23.1) | |
| 3 | 426 (40.3) | 501 (43.1) | |
| 4 | 284 (26.9) | 359 (30.9) | |
| Non–clear cell RCC, n (%) | 127 (9.8) | 253 (17.4) | <0.001 |
| Sarcomatoid features, n (%) | 143 (12.0) | 170 (12.0) | 0.586 |
| Therapy class, n (%) | |||
| AntiVEGF | 1316 (95.4) | 1507 (93.3) | |
| mTOR | 60 (4.3) | 105 (6.5) | |
| Combination | 4 (0.3) | 4 (0.2) | |
| DTT <1 yr, n (%) | 736 (53.4) | 905 (56.0) | 0.151 |
| Low HB (<LLN), n (%) | 691 (56.0) | 877 (58.1) | 0.261 |
| High calcium, n (%) | 98 (8.6) | 179 (12.6) | 0.001 |
| Neutrophilia, n (%) | 154 (13.1) | 286 (19.3) | <0.001 |
| Thrombocytosis, n (%) | 172 (15.4) | 299 (22.4) | <0.001 |
| Karnofsky PS <80%, n (%) | 261 (21.5) | 372 (25.9) | 0.009 |
| Lung metastases, n (%) | 939 (68.0) | 1149 (72.5) | 0.016 |
| Brain metastases, n (%) | 100 (7.2) | 141 (9.4) | 0.009 |
| Liver metastases, n (%) | 278 (20.2) | 328 (21.7) | 0.337 |
| Bone metastases, n (%) | 481 (34.9) | 551 (35.9) | 0.462 |
| Other metastic sites, n (%) | 470 (34.1) | 643 (43.1) | <0.001 |
| >1 metastatic site, n (%) | 805 (58.3%) | 1446 (89.5%) | <0.001 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; mTOR = mammalian target of rapamycin; PS = performance status; RCC = renal cell carcinoma; SD = standard deviation; VEGF = vascular endothelial growth factor.
The LNM localization distribution and clinicopathologic features in ccRCC and nccRCC are shown in Table 5 . In ccRCC, IMDC risk factors of neutrophilia (p = 0.049) and thrombophilia (p = 0.016) were more common. Patients with LNM had undergone nephrectomy less frequently (p = 0.001). In patients with ccRCC, those with LNM had concomitant lung (p < 0.001) and other metastases (eg, adrenal glands, other soft tissues; p = 0.009) more often than those without LNM; there was no difference in the frequency of bone (p = 0.276), liver (p = 0.943), or brain metastases (p = 0.187).
Table 5 Distribution of metastatic lymphatic sites in relation to the diaphragm in clear cell and non–clear cell renal carcinoma (RCC) a
| Clear cell RCC (n = 1218) |
Non–clear cell RCC (n = 159) |
|
|---|---|---|
| No LNM, n (%) | 552 (45.6) | 52 (32.7) |
| SPD LNM, n (%) | 321 (26.6) | 22 (13.8) |
| SBD LNM, n (%) | 204 (16.9) | 54 (34.0) |
| SPD+/SBD+ LNM, n (%) | 132 (10.9) | 31 (19.5) |
| LMN | No LMN | p value | LMN | No LMN | p value | |
|---|---|---|---|---|---|---|
| Nephrectomy, n (%) | 542 (81.2) | 492 (88.2) | 0.001 | 82 (76.6) | 43 (82.7) | 0.418 |
| Male, n (%) | 515 (77.7) | 403 (72.6) | 0.045 | 75 (70.1) | 37 (71.2) | 1.00 |
| DTT <1 yr, n (%) | 348 (52.5) | 265 (47.9) | 0.120 | 65 (60.7) | 29 (55.8) | 0.607 |
| Low HB (<LLN) | 331 (54.0) | 233 (48.7) | 0.088 | 49 (51.0) | 19 (45.2) | 0.582 |
| High calcium, n (%) | 70 (12.3) | 40 (9.5) | 0.184 | 5 (5.7) | 4 (12.1) | 0.254 |
| Neutrophilia, n (%) | 95 (15.9) | 53 (11.6) | 0.049 | 24 (25.3) | 7 (16.7) | 0.376 |
| Thrombophilia, n (%) | 106 (22.0) | 63 (15.6) | 0.016 | 17 (21.0) | 10 (25.6) | 0.642 |
| Karnofsky PS <80%, n (%) | 145 (24.7) | 95 (20.0) | 0.076 | 24 (24.7) | 10 (20.4) | 0.680 |
| Sarcomatoid features, n (%) | 52 (9.1) | 62 (12.6) | 0.074 | 14 (15.4) | 7 (15.6) | 1.00 |
| Lung metastases, n (%) | 507 (76.8) | 372 (67.0) | <0.001 | 52 (49.1) | 20 (38.5) | 0.237 |
| Brain metastases, n (%) | 65 (9.8) | 42 (7.6) | 0.187 | 4 (3.8) | 2 (3.8) | 1.00 |
| Liver metastases, n (%) | 132 (20.0) | 112 (20.2) | 0.943 | 32 (30.2) | 28 (53.8) | 0.005 |
| Bone metastases, n (%) | 220 (33.3) | 202 (36.4) | 0.276 | 35 (33.0) | 20 (38.5) | 0.594 |
| Other metastatic sites, n (%) | 287 (45.7) | 211 (38.2) | 0.009 | 40 (40.8) | 20 (38.5) | 0.862 |
| >1 metastatic site, n (%) | 602 (90.8) | 320 (57.7) | <0.001 | 87 (81.3) | 30 (7.7) | 0.002 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; PS = performance status; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm.
In nccRCC, liver metastases were found more often in patients without LNM (p = 0.005). A comparison of other clinicopathologic features between nccRCC patients with and without LNM revealed no significant differences ( Table 5 ).
It has previously been shown that only a minority of patients with kidney cancer undergo LND at the time of nephrectomy. This is the case even for high-risk patients [17] , even though retrospective studies have suggested a therapeutic benefit of LND [11] and [18]. It remains uncertain whether extended LND can improve the survival outcome for mRCC patients. However, if it were shown that SBD LNM affect the survival outcome for mRCC patients, this could form the basis for a rationale in favor of performing LND, or at least justify the need for prospective clinical trials to address this question. In this regard, an understanding of the influence of LNM on treatment response and survival outcome for mRCC patients undergoing TT could make an important contribution to the debate on surgical management of RCC.
Our study has demonstrated that LNM are associated with advanced IMDC risk factors, are an adverse prognostic factor for PFS and OS, and are associated with a higher metastatic burden. Furthermore, subanalyses showed that these differences depend on LNM localization.
Different cellular clones of RCC appear to differ in their propensity to spread to different organ sites [19] ; however, whether these differences impact on the survival outcome of mRCC patients is not completely understood. A previous IMDC study highlighted the prognostic importance of bone and liver metastases for survival outcome in mRCC patients, and suggested that aggressive RCC subclones tend to metastasize to these organs [16] . In addition, another study by our group showed that metastatic spread to other organs such as the pancreas or soft tissue tends to develop as a late clinical symptom, and has favorable treatment response and survival outcomes [20] . While these tumors may represent a favorable RCC subtype, our current results suggest that RCCs that spread to regional retroperitoneal lymph nodes represent another aggressive RCC subtype that develops bone metastases.
It has been suggested that LNM in ccRCC occur independently of von Hippel Lindau (VHL) gene inactivation [21] . Current agents used for mRCC treatment modulate pathways that are dysregulated because of VHL inactivation. This would imply that LNM might be less amenable to treatment with VEGF and mTOR inhibitors. However, our findings demonstrated no significant differences in best response to first-line therapies in patients with LNM versus those without LNM in both ccRCC and nccRCC patient groups. Nonetheless, patients with LNM had significantly shorter PFS than patients without LNM. While there was only a trend for an independent association between SBD LNM and short PFS, SPD+/SBD+ LNM were independently related to shorter PFS. Patients with SPD+/SBD+ LNM had additional hematogenous metastatic sites in 97% of cases. Thus, the high metastatic burden may explain the shorter response duration of first-line agents in this group of patients. However, a shorter duration for treatment success could also support the hypothesis that LNM are caused by RCC clones that do not have an inactive VHL gene function.
In the cytokine era, LNM were associated with shorter OS [1] and [2]. Our study demonstrates similar findings in the TT era. Interestingly, a worse survival outcome was observed for SBD LNM, both when occurring exclusively and in combination with thoracic lymph node involvement. Moreover, SBD LNM remained an independent prognostic factor for PFS and OS when analyzed without consideration of histologic subtype, and were an independent prognostic factor for OS in ccRCC. Conversely, SPD LNM were not an independent prognostic factor. This is notable and underscores the unique nature of RCCs that spread via retroperitoneal lymph nodes. Because the best response to TT did not different between SBD and SPD LNM, it appears that survival outcome is influenced by numerous factors. Patients with SBD LNM had faster disease progression before treatment initiation than those with SPD LNM (diagnosis to treatment <1 yr; 63% vs 46%; p < 0.001; data not shown). Despite a best response that was broadly similar, it seems that the disease progresses faster in SBD LNM patients during the later clinical course than in SPD LNM patients. In addition, SBD LNM were more often associated with anemia compared to SPD LNM (59% vs 48%; p = 0.008; data not shown). Anemia has multifactorial causes in cancer patients. However, one of the most important contributing factors is immune system dysregulation [22] . Dysregulation of the immune system may also be the reason for metastatic spread to SBD sites. Collectively, while risk factors that are currently unknown appear to dictate the clinical course in patients with SPD metastases, SBD metastases are independently associated with PFS and OS.
The majority of patients analyzed in this study did not undergo formal LND. Therefore, the effect of LND cannot be determined retrospectively. However, the current results indicate that SBD lymph nodes may have substantial negative effects on survival outcome and it is worth speculating whether extended removal of these metastatic sites could improve the clinical outcome for mRCC patients. The IMDC has recently shown that cytoreductive nephrectomy may provide therapeutic benefit in a large number of patients [23] . With regard to our current findings, it would be worth determining the role of LND in these patients.
The current study has several limitations that should be considered when interpreting the results. The IMDC database comprises retrospectively collected data for consecutive series of mRCC patients. These individual series, moreover, do not have standardized follow-up protocols, but this limitation may more accurately reflect real world outcomes. In addition, there was no centralized pathologic or radiologic review. Finally, lymph node status was determined clinically according to CT images in a number of cases, and it has been demonstrated that clinical and pathologic lymph node status is not necessarily always correlated.
LNM are associated with poor prognosis and adverse prognostic IMDC risk factors in mRCC patients treated with TT. Much of this increased risk associated with LNM appears to be dictated by SBD localization rather than SPD localization. Whether LND could improve the outcome for this group of patients remains unknown, but it is an important clinical question requiring prospective trials.
Author contributions: Daniel Y. Heng had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kroeger, Heng.
Acquisition of data: all authors.
Analysis and interpretation of data: Kroeger, Heng, Choueiri.
Drafting of the manuscript: Kroeger, Heng, Choueiri.
Critical revision of the manuscript for important intellectual content: all authors.
Statistical analysis: Kroeger, Heng.
Obtaining funding: None.
Administrative, technical, or material support: all authors.
Supervision: Choueiri, Heng.
Other (specify): None.
Financial disclosures: Daniel Y. Heng certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Benoit Beuselinck is an investigator of the EudraCT: 2011-006085-40/MetaSun trial supported by Pfizer. Frede Donskov has received research support from Novartis and GlaxoSmithKline. Toni K. Choueiri has received research funding from Pfizer and has an advisory role at Pfizer, GSK, Novartis, and Bayer. All other authors declare no conflicts of interest to the current study.
Funding/Support and role of the sponsor: None.
Acknowledgments: Dr. Choueiri is supported in part by the Trust family and the Loker Pinard and Michael Brigham Funds for Kidney Cancer Research at the Dana-Farber Cancer Institute. Dr. Kroeger thanks Dr. Klatte, Department of Urology, Medical University of Vienna, for statistical advice.
Renal cell carcinoma (RCC) is a heterogeneous disease consisting of several tumor types that have their own genetic, molecular, and clinical characteristics. Historical studies in the cytokine era by the University of California Los Angeles (UCLA) and the National Institutes of Health (NIH) described poor prognosis and worse treatment response to interleukin-2 (IL-2) in patients with metastatic involvement of the retroperitoneal lymph nodes [1] and [2]. Patients with and without retroperitoneal lymph node involvement had median overall survival (OS) times of 8.5 mo and 14.7 mo (p = 0.0004), respectively [2] . Pantuck et al [1] reported a better objective response rate in N0M1 compared to N+M1 patients (p = 0.01), with survival times of 10.5 mo and 20.4 mo, respectively. The adverse prognostic impact of retroperitoneal lymph node metastases (LNM) on survival outcome in patients with metastatic RCC (mRCC) was confirmed by several other institutions during the cytokine era [3], [4], and [5]. However, in the era of targeted therapies (TT), the survival outcome for patients with LNM has not yet been well characterized.
RCC preferentially metastasizes via hematogenous routes. Bianchi et al [6] recently described hematogenous metastatic sites in 80% of 11 157 patients with mRCC, while extension of the disease into lymph nodes was described only in 20% of patients. However, the authors did not provide details about the distribution of lymph node localization. The dominant regional LNM sites in RCC are retro- and paracaval, pre- and paraaortic, and interaortocaval lymph nodes in an anatomically intact retroperitoneum [7] . LNM can be unpredictable, and can also occur via direct extension to the thorax, supraclavicular lymph nodes, and iliac lymph nodes [7] . It has not been documented if particular sites of positive lymph node localization have a negative influence on the survival outcome of mRCC patients treated with TT.
Whether retroperitoneal lymph node dissection (LND) improves survival outcome is the subject of ongoing scientific debate. A phase 3 clinical trial (EORTC 30881) demonstrated positive lymph nodes in only 4% of investigated cases having no apparent involvement on computed tomography (CT) imaging. There was no advantage in OS, recurrence-free survival, or progression-free survival (PFS) for RCC patients who underwent LND [8] . The study has been criticized, however, because the majority of patients included had low-stage disease. Moreover, uniform surgical templates were not used for LND [9] . However, retrospective studies have demonstrated that LNM are more prevalent in high-risk patients with high Fuhrman grades, sarcomatoid features, locally advanced tumor stage, tumor size >10 cm, and tumor necrosis [10] . It has been suggested that LND is particularly beneficial in these high-risk patients [11] and [12]. Trends in the surgical management of RCC that have moved away from open radical surgery with extended LND to laparoscopic and robotic surgery with minimal hilar LND, combined with the negative results of phase 3 studies showing no additional benefit of LND over removal of the primary tumor alone, have decreased enthusiasm for performing extended LND. Therefore, it is common clinical practice to perform LND only in high-risk patients, and with a diagnostic rather than a curative intention [7] . Retrospective studies may be unable to add new insights into the current debate because only a prospective randomized trial with clear inclusion and exclusion criteria would be able to overcome this selection bias. Before planning such trials, however, it is prudent to better understand the impact of LNM and different lymph node localizations on the survival outcome for mRCC patients treated with current state-of-the-art therapies using data from the International mRCC Database Consortium (IMDC).
The IMDC database includes centers from North America (Canada, USA), Europe (Denmark, Greece, Belgium), Asia (Singapore, Japan, South Korea), and New Zealand. Data were collected from August 15, 2008 until December 31, 2013. At the time of analysis, the database contained data on 3405 patients who had received first-line targeted therapies between 2003 and 2013. The final study cohort comprised 2996 patients who were treated with first-line vascular endothelial growth factor (VEGF) inhibitors (n = 2823), mammalian target of rapamycin (mTOR) inhibitors (n = 165), or a combination of both (n = 8) therapies between April 2003 and August 2013. Patients were excluded from the analyses if they had received experimental therapies in first-line treatment (n = 7) or if no information was available on their lymph node status (n = 402; Fig. 1 ). Lymph node status was determined according to standard pathologic and CT criteria.
Information about LND was available for 890 patients. Full LND was performed in only 51 patients, and a limited (hilar) dissection was performed in 173 patients. These numbers were too low to allow any analyses and conclusions regarding the influence of LND on the survival outcome for mRCC patients treated with TT. Whether LND could be beneficial for mRCC patients was therefore not considered in the present study.
All centers obtained local Institutional Review Board approval before including data in the IMDC database. Baseline patient characteristics included demographic, clinicopathologic, and laboratory data, as previously described [13] . Survival data were retrospectively and prospectively collected from medical chart reviews and electronic records. Uniform data templates were used to ensure consistent data collection at each institution. The majority of patients were treated using standard care, but a subset received therapy as part of clinical trials. All patients were considered in consecutive series to avoid selection bias.
The primary hypothesis of the current study was that mRCC patients with LNM have worse survival outcome than patients without LNM when treated with TT. The secondary hypothesis was that worse survival outcome for mRCC patients with LNM is dictated by subdiaphragmatic (SPD) LNM rather than supradiaphragmatic (SBD) LNM.
Therefore, treatment response to current standard-of-care TT agents, OS, and PFS were compared between patients with LNM and those without LNM. We did not compare outcomes between the drugs used for treatment because these comparisons are frequently biased outside a clinical trial and there may not be sufficient power for such an analysis. Additional analyses were performed for subgroups of patients with exclusively SPD lymph node involvement, exclusively SBD lymph node involvement, or with lymph node involvement in both sites (SPD+/SBD+). Previous studies have demonstrated that tumor biology has an important impact on the survival outcome for mRCC patients [14] . Thus, the different biology of RCC was considered by separate analyses for clear cell RCC (ccRCC) and non-ccRCC (nccRCC) patients. Patients with SPD primarily had mediastinal lymph node involvement, while patients with SBD primarily had retroperitoneal lymph node involvement. Lymph node status was determined before initiation of TT. The selection process for patient groups analyzed in the study is shown in Figure 1 .
OS was defined as the time period between TT initiation and the date of death, with censoring on the day of the last follow-up visit. PFS was defined as the period between treatment initiation and progression, drug cessation, or death, with censoring at the last follow-up visit. Kaplan-Meier plots were used to estimate median OS and PFS, and univariate comparisons were performed using the log rank test. Uni- and multivariate (MVA) Cox regression analyses were performed to test the association of predefined prognostic factors and LNM with survival outcome. The analyses were performed with backward stepwise selection criteria, and significance was tested using the Wald statistic [13] . In MVA analyses, the IMDC (Heng) risk criteria were applied for adjustment [15] . The IMDC risk criteria include a time period from diagnosis to treatment of <1 yr, Karnofsky performance score <80%, anemia, hypercalcemia, thrombophilia, and neutrophilia. Upper and lower limits of normal for the laboratory parameters were based on institutional limits. In addition, adjustment was performed for bone and liver metastases because interactions of LNM with other metastatic sites should be considered. The prognostic relevance of bone and liver metastases has previously been demonstrated by our group [16] .
A secondary aim was to analyze baseline characteristics. Patient and tumor characteristics were compared using the Student t test for continuous variables, or the chi-square test and the Fisher exact test for categorical variables.
Statistical analyses were performed using SPSS version 22 (IBM, Chicago, IL, USA), and a two-sided value of p < 0.05 was considered statistically significant.
At the time of data analysis, 2525 (84.3%) patients had stopped taking their TT; the median time on TT was 6.0 mo (25th percentile, 2.7 mo; 75th percentile, 13.0 mo). Best overall response data to first-line therapy were available for 2559 patients. There were no significant differences in response to first-line TT (p = 0.059) when patients were compared without consideration of RCC subtype or LNM localization ( Table 1 ).
Table 1 Treatment response to targeted therapies in patients with lymph node metastases (LNM) compared to those without LNM
| Best response | |||||
|---|---|---|---|---|---|
| CR | PR | SD | PD | ||
| All RCC subtypes | |||||
| No LMN | n | 14 | 304 | 556 | 290 |
| % | 1.2 | 26.1 | 47.8 | 24.9 | |
| LMN | n | 9 | 314 | 690 | 382 |
| % | 0.6 | 22.5 | 49.5 | 27.4 | |
| Clear cell RCC | |||||
| No LNM | n | 3 | 108 | 261 | 95 |
| % | 0.6 | 23.1 | 55.9 | 20.3 | |
| SPD LNM | n | 2 | 73 | 149 | 55 |
| % | 0.7 | 26.2 | 53.4 | 19.7 | |
| SBD LNM | n | 2 | 50 | 92 | 42 |
| % | 1.1 | 26.9 | 49.5 | 22.6 | |
| SPD+/SBD+ LNM | n | 0 | 28 | 58 | 36 |
| % | 0.0 | 23.0 | 47.5 | 29.5 | |
| Non–clear cell RCC | |||||
| No LNM | n | 0 | 3 | 21 | 21 |
| % | 0.0 | 6.7 | 46.7 | 46.7 | |
| SPD LNM | n | 0 | 3 | 12 | 3 |
| % | 0.0 | 16.7 | 66.7 | 16.7 | |
| SBD LNM | n | 0 | 7 | 24 | 19 |
| % | 0.0 | 14.0 | 48.0 | 38.0 | |
| SPD+/SBD+ LNM | n | 1 | 4 | 10 | 6 |
| % | 4.8 | 19.0 | 47.6 | 28.6 | |
CR = complete remission; PR = partial response; SD = stable disease; PD = progressive disease; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm.
Both response data and LNM localization data were available for 1054 ccRCC patients and 134 nccRCC patients. Comparison of data by LNM localization revealed no difference in best response to first-line therapies for either ccRCC (p = 0.457) or nccRCC (p = 0.192; Table 1 ).
Patients with LNM generally had worse PFS compared to those with no LNM without considering LNM localization or RCC subtype (median PFS [95% confidence interval] 7.8 mo [7.1–8.5 mo] vs 6.1 [5.6–6.6 mo], log rank p = 0.001; Fig. 2 A).
Fig. 2 (A) Comparison of progression-free survival (PFS) in patients with positive lymph nodes versus patients without positive lymph nodes (median 7.8 mo vs. 6.1 mo) without consideration of histologic subtype. Patients with lymph node metastases (LNM) had worse PFS (hazard ratio [HR] 1.14, 95% confidence interval [CI] 1.06–1.24; p = 0.001). (B) Comparison of PFS according to lymph node localization. Patients without LNM had the best PFS (8.8 mo), and patients with both supradiaphragmatic (SPD) and subdiaphragmatic (SBD) LNM localizations (SPD+/SBD+) had the worst PFS (5.5 mo). Patients with SBD LNM had significantly worse PFS compared to patients with no LNM (6.8 vs 8.8 mo; p = 0.003) and those with SPD LNM (6.8 vs 8.3 mo; p = 0.022). Any metastatic involvement of retroperitoneal lymph nodes, either alone (HR 1.25, 95% CI 1.08–1.45) or in combination with SPD+/SBD+LNM (HR 1.50, 95% CI 1.24–1.74; p < 0.001) was associated with higher risk of progression. By contrast, SPD LNM alone was not a prognostic factor for progression (HR 1.03, 95% 0.895–1.18; p = 0.695). (C) PFS by lymph node localization in clear cell renal cell carcinoma (ccRCC; median PFS: no nodes, 9.9 mo; SPD LNM, 8.7 mo; SBD LNM, 8.0 mo; and both, 5.6 mo). Patients with SPD LNM (HR 1.09, 95% CI 0.94–1.25) and SBD LNM (HR 1.04, 95% CI 0.96–1.14) had similar PFS, and this was not significantly worse than the PFS of patients without LNM (p = 0.265 and p = 0.333 for SPD and SBD LNM, respectively).
Comparison of PFS rates according to LNM sites revealed that patients with SBD LNM had significantly shorter PFS (6.8 mo [5.7–8.0 mo]) compared to those without LNM (8.8 mo [7.6–10.0 mo]; p = 0.003) or SPD LNM (8.3 mo [7.2–9.6 mo]; p = 0.022). The worst PFS was observed in patients with SPD+/SBD+ LNM (5.4 mo [4.6–6.4 mo]), but this did not significantly differ from PFS in patients with SBD LNM (p = 0.093; Fig. 2 B).
In ccRCC, however, comparison according to LNM localization revealed no significant differences in PFS between patients without LNM (9.9 mo [8.6–11.1 mo]) and those with SPD LNM (8.7 mo [7.3–10.1 mo]) or SBD LNM (8.0 mo [6.9–9.2 mo]) (p = 0.312; Fig. 2 C). SPD+/SBD+ LNM patients had the worst survival outcome (5.6 mo [4.0–7.2 mo]) in comparison to patients without LNM (p < 0.001), with SPD LNM (p = 0.006), and with SBD LNM (p = 0.035; Fig. 2 C). In nccRCC, there was no difference in PFS between patients with no LNM (3.9 mo [2.9–5.0 mo]), SPD LNM (4.6 mo [0.0–9.9 mo]), SBD LNM (5.4 mo [4.7–6.0 mo]), and SPD+/SBD+ LNM (5.4 mo [3.2–7.3 mo]).
In MVA analyses, LNM remained an independent prognostic factor (HR 1.13 [1.03–1.24]; p = 0.013) when adjusted for IMDC risk criteria, as well as liver and bone metastases ( Table 2 ) [16] . In subgroup analysis, SBD LNM (HR 1.31 [1.08–1.60]; p = 0.006) and SPD+/SBD LNM (HR 1.29 [1.04–1.61]; p = 0.022) were independent prognostic factors for PFS, while SPD LNM was not independently associated with PFS (p = 1.00; Table 2 ). In ccRCC, only SPD+/SBD+ LNM (p = 0.025) was independently associated with PFS ( Table 2 ).
Table 2 Multivariate analyses for association between lymph node metastases (LNM) and progression-free survival (PFS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time <1 yr | 1.33 | 1.21–1.47 | <0.001 |
| Low hemoglobin | 1.37 | 1.24–1.52 | <0.001 |
| Neutrophilia | 1.43 | 1.25–1.63 | <0.001 |
| Thrombocytosis | 1.16 | 1.03–1.32 | 0.019 |
| Karnofsy PS <80% | 1.68 | 1.50–1.88 | <0.001 |
| LNM | 1.13 | 1.03–1.24 | 0.013 |
| Liver metastases | 1.26 | 1.12–1.41 | <0.001 |
| Bone metastases | 1.27 | 1.15–1.40 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.19–1.61 | <0.001 |
| Low hemoglobin | 1.42 | 1.21–1.67 | <0.001 |
| Neutrophilia | 1.46 | 1.19–1.80 | <0.001 |
| Thrombocytosis | 1.22 | 1.01–1.47 | 0.045 |
| Karnofsy PS <80% | 1.94 | 1.65–2.28 | <0.001 |
| Liver metastases | 1.39 | 1.18–1.65 | <0.001 |
| Bone metastases | 1.45 | 1.25–1.68 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.83–1.20 | 0.990 |
| SBD LNM | 1.32 | 1.08–1.60 | 0.006 |
| SPD+/SBD+ LNM | 1.29 | 1.04–1.61 | 0.022 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time <1 yr | 1.38 | 1.17–1.64 | <0.001 |
| Low hemoglobin | 1.43 | 1.21–1.69 | <0.001 |
| Neutrophilia | 1.38 | 1.09–1.76 | 0.009 |
| Thrombocytosis | 1.97 | 1.64–2.38 | <0.001 |
| Karnofsy PS <80% | 1.51 | 1.24–1.85 | <0.001 |
| Bone metastases | 1.64 | 1.38–1.94 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.00 | 0.82–1.22 | 0.998 |
| SBD LNM | 1.23 | 0.98–1.55 | 0.077 |
| SPD+/SBD+ LNM | 1.33 | 1.04–1.71 | 0.025 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm; ccRCC = clear cell RCC.
At the time of the analysis, 1960 patients had died; the median OS time for the total patient cohort was 20.0 mo (25th percentile, 42.3 mo; 75th percentile, 7.9 mo). Patients with metastatic spread to lymph nodes had a worse survival outcome than patients without LNM (24.0 mo [21.6–26.3 mo] vs 16.0 mo [14.9–17.1 mo], p < 0.001; Fig. 3 A).
Fig. 3 (A) Patients with lymph node metastases (LNM) had a higher risk of death (median overall survival [OS] 23.9 mo, 95% confidence interval [CI] 21.6–26.3 mo) than those without LNM (median OS 16.0 mo, 95% CI 14.9–17.1; hazard ratio [HR] 1.41, 95% CI 1.28–1.54; p < 0.001). (B) A comparison without consideration of the histologic subtypes demonstrates that any lymph node involvement was associated with worse survival outcome compared to patients without LNM. Median survival times were 25.2 mo, 20.3 mo, 16.2 mo, and 11.5 mo in patients with no LNM and lymph node involvement in supradiaphragmatic (SPD), subdiaphragmatic (SBD), and both localizations (SPD+/SBD+), respectively. The HRs (95% CIs) for patients with SPD, SBD, and SPD+/SBD+ LNM were 1.24 (1.06–1.44), 1.45 (1.23–1.71), and 1.93 (1.60–2.33), respectively. (C) In clear cell renal cell carcinoma (ccRCC), patients with SPD and SBD LNM had similar median OS (SPD, 21.6 mo; SBD, 19.8 mo). However, both groups had a significantly higher risk of death (SPD: HR 1.24, 95% CI 1.05–1.47; SBD: HR 1.28, 95% CI 1.05–1.56) in comparison to those without lymph nodes. Patients with SPD+/SBD+ metastases had the worst OS (12.85 mo) and a twofold higher risk of death than patients without lymph node involvement (HR 2.04, 95% CI 1.68–2.48). (D) In non–clear cell RCC (nccRCC), median OS times according to lymph node localization were 15.8 mo for no lymph nodes, 16.1 mo for SPD LNM, 11.2 mo for SBD LNM, and 12.5 mo for SPD+/SBD+ LNM. There were no significant differences between the groups (log rank p = 0.444).
In subgroup analyses, patients without LNM had a median OS of 25.2 mo (21.9–28.5 mo) compared to 20.3 mo (17.1–23.5 mo) for SPD LNM (p = 0.093), 16.2 mo (13.5–20.0 mo) for SBD LNM (p < 0.001), and 11.5 mo (9.8–13.3 mo) (p < 0.001) for SPD+/SBD+ LNM ( Fig. 3 B).
In ccRCC, further analysis considering LNM localization revealed similar OS times for patients with SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13]), and were shorter than for patients without LNM (26.97 mo [23.32– 30.62], p = 0.01; Fig. 3 C). Patients with SPD+/SBD+ LNM had significantly (p = 0.001) worse OS (12.85 mo [9.45–16.24 mo]) compared to SPD LNM (21.55 mo [17.94–25.17 mo]) and SBD LNM (19.81 mo [15.50–24.13 mo]; Fig. 3 C). In nccRCC, analyses according to LNM revealed no significant differences in OS (p = 0.444; Fig. 3 C).
For all histological subtypes, LNM were an independent prognostic factor for adverse OS in MVA analysis (p < 0.001; Table 3 ). Additional MVA according to LNM sites revealed that SBD LNM (p < 0.001) and SPD+/SBD+ LNM (p < 0.001) were independent prognostic factors for OS. SPD LNM were not independently associated with OS (p = 0.999; Table 3 ). In ccRCC, MVA according to LNM localization confirmed an independent association between SBD LNM (p = 0.028) and SPD+/SBD+ LNM (p = 0.006) and OS ( Table 3 ).
Table 3 Multivariate Cox regression analysis of association between lymph node metastases (LNM) and overall survival (OS) in renal cell carcinoma
| Prognostic factor | HR | 95% CI | p value |
|---|---|---|---|
| Baseline regression | |||
| Diagnosis–treatment time < 1 yr | 1.34 | 1.19–1.49 | <0.001 |
| Low hemoglobin | 1.51 | 1.31–1.70 | <0.001 |
| High Ca | 1.47 | 1.25–1.73 | <0.001 |
| Neutrophilia | 1.67 | 1.45–1.92 | <0.001 |
| Thrombophilia | 1.31 | 1.14–1.51 | <0.001 |
| Karnofsky PS <80% | 1.88 | 1.67–2.13 | <0.001 |
| Liver metastases | 1.50 | 1.32–1.70 | <0.001 |
| Bone metastases | 1.50 | 1.34–1.67 | <0.001 |
| LNM | 1.33 | 1.19–1.48 | <0.001 |
| Inclusion of LNM localization in all RCC subtypes | |||
| Diagnosis–treatment time < 1 yr | 1.32 | 1.11–1.56 | 0.001 |
| Low hemoglobin | 1.45 | 1.21–1.73 | <0.001 |
| High Ca | 1.38 | 1.08–1.77 | 0.010 |
| Neutrophilia | 1.78 | 1.43–2.22 | <0.001 |
| Thrombophilia | 1.49 | 1.21–1.83 | <0.001 |
| Karnofsky PS <80% | 2.19 | 1.82–2.62 | <0.001 |
| Liver metastases | 1.84 | 1.53–2.20 | <0.001 |
| Bone metastases | 1.61 | 1.36–1.90 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.57 | 0.98–1.48 | 0.075 |
| SBD LNM | 1.59 | 1.26–1.96 | <0.001 |
| SPD+/SBD+ LNM | 1.32 | 1.25–2.02 | <0.001 |
| Inclusion of LNM localization in ccRCC | |||
| Diagnosis–treatment time < 1 yr | 1.36 | 1.12–1.64 | 0.002 |
| Low hemoglobin | 1.36 | 1.11–1.67 | 0.003 |
| High Ca | 1.44 | 1.10–1.89 | 0.009 |
| Neutrophilia | 1.65 | 1.27–2.13 | <0.001 |
| Thrombophilia | 1.52 | 1.19–1.91 | 0.001 |
| Karnofsky PS <80% | 2.12 | 1.72–2.61 | <0.001 |
| Liver metastases | 1.94 | 1.57–2.41 | <0.001 |
| Bone metastases | 1.84 | 1.53–2.22 | <0.001 |
| LNM | |||
| No LNM | Reference | ||
| SPD LNM | 1.13 | 0.90–1.42 | 0.301 |
| SBD LNM | 1.34 | 1.03–1.74 | 0.028 |
| SPD+/SBD+ LNM | 1.47 | 1.11–1.93 | 0.006 |
HR = hazard ratio; CI = confidence interval; PS = performance status; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sides of the diaphragm; ccRCC = clear cell RCC.
An overall comparison of patients with and without LNM is shown in Table 4 . Patients with LNM presented more often with nccRCC (p < 0.001) and higher Fuhrman grades (p = 0.007), and had less often undergone prior nephrectomy (p < 0.001). The IMDC risk factors of high calcium (p = 0.001), thrombocytosis (p < 0.001), neutrophilia (p < 0.001), and Karnofsky performance status <80% (p = 0.009) were all more frequently seen in patients with LNM. Comparison of additional metastatic sites revealed that LNM are related to a higher overall metastatic burden (one or more additional metastatic site p < 0.001).
Table 4 General comparison of clinicopathologic features of patients with lymph node metastases (LNM) compared to those without LNM a
| No LNM (n = 1380) |
LNM (n = 1616) |
p value | |
|---|---|---|---|
| Mean age, yr (SD) | 58.65 (11.36) | 58. 50 (11.30) | 0.716 |
| Male, n (%) | 1003 (72.7) | 1182 (73.1) | 0.805 |
| Nephrectomy, n (%) | 1134 (82.4) | 1218 (75.4) | <0.001 |
| Fuhrman grade n (%) | 0.007 | ||
| 1 | 44 (4.2) | 33 (2.8) | |
| 2 | 302 (28.6) | 269 (23.1) | |
| 3 | 426 (40.3) | 501 (43.1) | |
| 4 | 284 (26.9) | 359 (30.9) | |
| Non–clear cell RCC, n (%) | 127 (9.8) | 253 (17.4) | <0.001 |
| Sarcomatoid features, n (%) | 143 (12.0) | 170 (12.0) | 0.586 |
| Therapy class, n (%) | |||
| AntiVEGF | 1316 (95.4) | 1507 (93.3) | |
| mTOR | 60 (4.3) | 105 (6.5) | |
| Combination | 4 (0.3) | 4 (0.2) | |
| DTT <1 yr, n (%) | 736 (53.4) | 905 (56.0) | 0.151 |
| Low HB (<LLN), n (%) | 691 (56.0) | 877 (58.1) | 0.261 |
| High calcium, n (%) | 98 (8.6) | 179 (12.6) | 0.001 |
| Neutrophilia, n (%) | 154 (13.1) | 286 (19.3) | <0.001 |
| Thrombocytosis, n (%) | 172 (15.4) | 299 (22.4) | <0.001 |
| Karnofsky PS <80%, n (%) | 261 (21.5) | 372 (25.9) | 0.009 |
| Lung metastases, n (%) | 939 (68.0) | 1149 (72.5) | 0.016 |
| Brain metastases, n (%) | 100 (7.2) | 141 (9.4) | 0.009 |
| Liver metastases, n (%) | 278 (20.2) | 328 (21.7) | 0.337 |
| Bone metastases, n (%) | 481 (34.9) | 551 (35.9) | 0.462 |
| Other metastic sites, n (%) | 470 (34.1) | 643 (43.1) | <0.001 |
| >1 metastatic site, n (%) | 805 (58.3%) | 1446 (89.5%) | <0.001 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; mTOR = mammalian target of rapamycin; PS = performance status; RCC = renal cell carcinoma; SD = standard deviation; VEGF = vascular endothelial growth factor.
The LNM localization distribution and clinicopathologic features in ccRCC and nccRCC are shown in Table 5 . In ccRCC, IMDC risk factors of neutrophilia (p = 0.049) and thrombophilia (p = 0.016) were more common. Patients with LNM had undergone nephrectomy less frequently (p = 0.001). In patients with ccRCC, those with LNM had concomitant lung (p < 0.001) and other metastases (eg, adrenal glands, other soft tissues; p = 0.009) more often than those without LNM; there was no difference in the frequency of bone (p = 0.276), liver (p = 0.943), or brain metastases (p = 0.187).
Table 5 Distribution of metastatic lymphatic sites in relation to the diaphragm in clear cell and non–clear cell renal carcinoma (RCC) a
| Clear cell RCC (n = 1218) |
Non–clear cell RCC (n = 159) |
|
|---|---|---|
| No LNM, n (%) | 552 (45.6) | 52 (32.7) |
| SPD LNM, n (%) | 321 (26.6) | 22 (13.8) |
| SBD LNM, n (%) | 204 (16.9) | 54 (34.0) |
| SPD+/SBD+ LNM, n (%) | 132 (10.9) | 31 (19.5) |
| LMN | No LMN | p value | LMN | No LMN | p value | |
|---|---|---|---|---|---|---|
| Nephrectomy, n (%) | 542 (81.2) | 492 (88.2) | 0.001 | 82 (76.6) | 43 (82.7) | 0.418 |
| Male, n (%) | 515 (77.7) | 403 (72.6) | 0.045 | 75 (70.1) | 37 (71.2) | 1.00 |
| DTT <1 yr, n (%) | 348 (52.5) | 265 (47.9) | 0.120 | 65 (60.7) | 29 (55.8) | 0.607 |
| Low HB (<LLN) | 331 (54.0) | 233 (48.7) | 0.088 | 49 (51.0) | 19 (45.2) | 0.582 |
| High calcium, n (%) | 70 (12.3) | 40 (9.5) | 0.184 | 5 (5.7) | 4 (12.1) | 0.254 |
| Neutrophilia, n (%) | 95 (15.9) | 53 (11.6) | 0.049 | 24 (25.3) | 7 (16.7) | 0.376 |
| Thrombophilia, n (%) | 106 (22.0) | 63 (15.6) | 0.016 | 17 (21.0) | 10 (25.6) | 0.642 |
| Karnofsky PS <80%, n (%) | 145 (24.7) | 95 (20.0) | 0.076 | 24 (24.7) | 10 (20.4) | 0.680 |
| Sarcomatoid features, n (%) | 52 (9.1) | 62 (12.6) | 0.074 | 14 (15.4) | 7 (15.6) | 1.00 |
| Lung metastases, n (%) | 507 (76.8) | 372 (67.0) | <0.001 | 52 (49.1) | 20 (38.5) | 0.237 |
| Brain metastases, n (%) | 65 (9.8) | 42 (7.6) | 0.187 | 4 (3.8) | 2 (3.8) | 1.00 |
| Liver metastases, n (%) | 132 (20.0) | 112 (20.2) | 0.943 | 32 (30.2) | 28 (53.8) | 0.005 |
| Bone metastases, n (%) | 220 (33.3) | 202 (36.4) | 0.276 | 35 (33.0) | 20 (38.5) | 0.594 |
| Other metastatic sites, n (%) | 287 (45.7) | 211 (38.2) | 0.009 | 40 (40.8) | 20 (38.5) | 0.862 |
| >1 metastatic site, n (%) | 602 (90.8) | 320 (57.7) | <0.001 | 87 (81.3) | 30 (7.7) | 0.002 |
a High calcium, thrombocytosis, and neutrophilia were determined according to institutional upper limits of normal. Other metastatic sites include adrenal glands and soft tissues.
DTT = diagnosis–treatment time; HB = hemoglobin; LLN = lower limit of normal; PS = performance status; RCC = renal cell carcinoma; SBD = subdiaphragmatic; SPD = supradiaphragmatic; SPD+/SBD+ = lymph node involvement on both sites of the diaphragm.
In nccRCC, liver metastases were found more often in patients without LNM (p = 0.005). A comparison of other clinicopathologic features between nccRCC patients with and without LNM revealed no significant differences ( Table 5 ).
It has previously been shown that only a minority of patients with kidney cancer undergo LND at the time of nephrectomy. This is the case even for high-risk patients [17] , even though retrospective studies have suggested a therapeutic benefit of LND [11] and [18]. It remains uncertain whether extended LND can improve the survival outcome for mRCC patients. However, if it were shown that SBD LNM affect the survival outcome for mRCC patients, this could form the basis for a rationale in favor of performing LND, or at least justify the need for prospective clinical trials to address this question. In this regard, an understanding of the influence of LNM on treatment response and survival outcome for mRCC patients undergoing TT could make an important contribution to the debate on surgical management of RCC.
Our study has demonstrated that LNM are associated with advanced IMDC risk factors, are an adverse prognostic factor for PFS and OS, and are associated with a higher metastatic burden. Furthermore, subanalyses showed that these differences depend on LNM localization.
Different cellular clones of RCC appear to differ in their propensity to spread to different organ sites [19] ; however, whether these differences impact on the survival outcome of mRCC patients is not completely understood. A previous IMDC study highlighted the prognostic importance of bone and liver metastases for survival outcome in mRCC patients, and suggested that aggressive RCC subclones tend to metastasize to these organs [16] . In addition, another study by our group showed that metastatic spread to other organs such as the pancreas or soft tissue tends to develop as a late clinical symptom, and has favorable treatment response and survival outcomes [20] . While these tumors may represent a favorable RCC subtype, our current results suggest that RCCs that spread to regional retroperitoneal lymph nodes represent another aggressive RCC subtype that develops bone metastases.
It has been suggested that LNM in ccRCC occur independently of von Hippel Lindau (VHL) gene inactivation [21] . Current agents used for mRCC treatment modulate pathways that are dysregulated because of VHL inactivation. This would imply that LNM might be less amenable to treatment with VEGF and mTOR inhibitors. However, our findings demonstrated no significant differences in best response to first-line therapies in patients with LNM versus those without LNM in both ccRCC and nccRCC patient groups. Nonetheless, patients with LNM had significantly shorter PFS than patients without LNM. While there was only a trend for an independent association between SBD LNM and short PFS, SPD+/SBD+ LNM were independently related to shorter PFS. Patients with SPD+/SBD+ LNM had additional hematogenous metastatic sites in 97% of cases. Thus, the high metastatic burden may explain the shorter response duration of first-line agents in this group of patients. However, a shorter duration for treatment success could also support the hypothesis that LNM are caused by RCC clones that do not have an inactive VHL gene function.
In the cytokine era, LNM were associated with shorter OS [1] and [2]. Our study demonstrates similar findings in the TT era. Interestingly, a worse survival outcome was observed for SBD LNM, both when occurring exclusively and in combination with thoracic lymph node involvement. Moreover, SBD LNM remained an independent prognostic factor for PFS and OS when analyzed without consideration of histologic subtype, and were an independent prognostic factor for OS in ccRCC. Conversely, SPD LNM were not an independent prognostic factor. This is notable and underscores the unique nature of RCCs that spread via retroperitoneal lymph nodes. Because the best response to TT did not different between SBD and SPD LNM, it appears that survival outcome is influenced by numerous factors. Patients with SBD LNM had faster disease progression before treatment initiation than those with SPD LNM (diagnosis to treatment <1 yr; 63% vs 46%; p < 0.001; data not shown). Despite a best response that was broadly similar, it seems that the disease progresses faster in SBD LNM patients during the later clinical course than in SPD LNM patients. In addition, SBD LNM were more often associated with anemia compared to SPD LNM (59% vs 48%; p = 0.008; data not shown). Anemia has multifactorial causes in cancer patients. However, one of the most important contributing factors is immune system dysregulation [22] . Dysregulation of the immune system may also be the reason for metastatic spread to SBD sites. Collectively, while risk factors that are currently unknown appear to dictate the clinical course in patients with SPD metastases, SBD metastases are independently associated with PFS and OS.
The majority of patients analyzed in this study did not undergo formal LND. Therefore, the effect of LND cannot be determined retrospectively. However, the current results indicate that SBD lymph nodes may have substantial negative effects on survival outcome and it is worth speculating whether extended removal of these metastatic sites could improve the clinical outcome for mRCC patients. The IMDC has recently shown that cytoreductive nephrectomy may provide therapeutic benefit in a large number of patients [23] . With regard to our current findings, it would be worth determining the role of LND in these patients.
The current study has several limitations that should be considered when interpreting the results. The IMDC database comprises retrospectively collected data for consecutive series of mRCC patients. These individual series, moreover, do not have standardized follow-up protocols, but this limitation may more accurately reflect real world outcomes. In addition, there was no centralized pathologic or radiologic review. Finally, lymph node status was determined clinically according to CT images in a number of cases, and it has been demonstrated that clinical and pathologic lymph node status is not necessarily always correlated.
LNM are associated with poor prognosis and adverse prognostic IMDC risk factors in mRCC patients treated with TT. Much of this increased risk associated with LNM appears to be dictated by SBD localization rather than SPD localization. Whether LND could improve the outcome for this group of patients remains unknown, but it is an important clinical question requiring prospective trials.
Author contributions: Daniel Y. Heng had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kroeger, Heng.
Acquisition of data: all authors.
Analysis and interpretation of data: Kroeger, Heng, Choueiri.
Drafting of the manuscript: Kroeger, Heng, Choueiri.
Critical revision of the manuscript for important intellectual content: all authors.
Statistical analysis: Kroeger, Heng.
Obtaining funding: None.
Administrative, technical, or material support: all authors.
Supervision: Choueiri, Heng.
Other (specify): None.
Financial disclosures: Daniel Y. Heng certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: Benoit Beuselinck is an investigator of the EudraCT: 2011-006085-40/MetaSun trial supported by Pfizer. Frede Donskov has received research support from Novartis and GlaxoSmithKline. Toni K. Choueiri has received research funding from Pfizer and has an advisory role at Pfizer, GSK, Novartis, and Bayer. All other authors declare no conflicts of interest to the current study.
Funding/Support and role of the sponsor: None.
Acknowledgments: Dr. Choueiri is supported in part by the Trust family and the Loker Pinard and Michael Brigham Funds for Kidney Cancer Research at the Dana-Farber Cancer Institute. Dr. Kroeger thanks Dr. Klatte, Department of Urology, Medical University of Vienna, for statistical advice.
