Methods
In this retrospective analysis and validation study, we developed a histopathological whole-slide image (WSI)-based score using deep learning allied to digital scanning of conventional haematoxylin and eosin-stained tumour tissue sections, to predict tumour recurrence in a development dataset of 651 patients with distinctly good or poor disease outcome. The six single nucleotide polymorphism-based score, which was detected in paraffin-embedded tumour tissue samples, and the Leibovich score, which was established using clinicopathological risk factors, were combined with the WSI-based score to construct a multimodal recurrence score in the training dataset of 1125 patients. The multimodal recurrence score was validated in 1625 patients from the independent validation dataset and 418 patients from The Cancer Genome Atlas set. The primary outcome measured was the recurrence-free interval (RFI).
Findings
The multimodal recurrence score had significantly higher predictive accuracy than the three single-modal scores and clinicopathological risk factors, and it precisely predicted the RFI of patients in the training and two validation datasets (areas under the curve at 5 years: 0·825–0·876 vs 0·608–0·793; p<0·05). The RFI of patients with low stage or grade is usually better than that of patients with high stage or grade; however, the RFI in the multimodal recurrence score-defined high-risk stage I and II group was shorter than in the low-risk stage III group (hazard ratio [HR] 4·57, 95% CI 2·49–8·40; p<0·0001), and the RFI of the high-risk grade 1 and 2 group was shorter than in the low-risk grade 3 and 4 group (HR 4·58, 3·19–6·59; p<0·0001).
Interpretation
Our multimodal recurrence score is a practical and reliable predictor that can add value to the current staging system for predicting localised renal cell carcinoma recurrence after surgery, and this combined approach more precisely informs treatment decisions about adjuvant therapy.
Funding
National Natural Science Foundation of China, and National Key Research and Development Program of China.
The prediction of the risk of recurrence after surgical treatment for localised clear cell renal cell carcinoma (ccRCC) usually relies on the TNM staging system and three established single-modality models that use clinical, genomic, or histopathological information. The study by Gui et al. evaluated a novel recurrence scoring system integrating the three modalities—clinical, genomic, and histopathological to predict such recurrence.
Study design included retrospective analysis and validation with 2 external cohorts. The authors used histopathological whole-slide image (WSI)-based score to predict tumour recurrence with distinctly good (recurrence >7 years) or poor disease (recurrence within 3 years) outcome, and the six single nucleotide polymorphism-based score, which was detected in paraffin-embedded tumour tissue samples, and the Leibovich score. The multimodal recurrence score was validated in 1,625 patients from the independent validation dataset and 418 patients from The Cancer Genome Atlas set. The primary endpoint was recurrence free interval.
For the multimodal scoring system, the AUC at 5 years for predicting tumour recurrence was 0.825, which was significantly higher than that of any single-modality risk model alone, and also significantly higher compared with the clinicopathological risk factors.
The authors stratified KEYNOTE-564 (1) dataset with their multimodal system and demonstrated that the survival of patients with high-stage and high-grade disease was not always inferior to that of patients with low-stage and low-grade disease.
The authors conclude that their recurrence scoring system is a practical and reliable prognostic tool which can complement the available staging system to predict recurrence after surgery and guide clinicians about decisions on adjuvant therapy.
Better stratification of patients to high and low risk of recurrence after (partial)nephrectomy is required not only for adjuvant therapy but also for surveillance strategies. Currently, there is no validated biomarkers available. Would this be the best recurrence scoring system available today? I would address that the authors used the 2003 published Leibovich score in their model. Nevertheless the updated Leibovich score includes features such as: symptoms, grade, coagulative necrosis, sarcomatoid differentiation, tumour size, presence of fat invasion, tumour thrombus level, extension beyond the kidney, and presence of regional lymph node involvement. With a c-index of 0.83 (95% CI: 0.82–0.84) (2), it remains high in validated cohorts (3). Before being adopted, the model needs to be compared to all current prognostic ccRCC models (Leibovich 2003 and updated 2018, UISS) with regards to discrimination, calibration and net benefit. Yet, it demonstrated promising results.