A 16-gene assay to predict recurrence after surgery in localised renal cell carcinoma: development and validation studies

Summary

Background

The likelihood of tumour recurrence after nephrectomy in localised clear cell renal cell carcinoma is well characterised by clinical and pathological parameters. However, these assessments can be improved and personalised by the addition of molecular characteristics of each patient's tumour. We aimed to develop and validate a prognostic multigene signature to improve prediction of recurrence risk in clear cell renal cell carcinoma.

Methods

In the development stage, we investigated the association between expression of 732 genes, measured by reverse-transcription PCR, and clinical outcome in 942 patients with stage I–III clear cell renal cell carcinoma who had undergone a nephrectomy at the Cleveland Clinic (OH, USA). 516 genes were associated with recurrence-free interval. 11 of these genes were selected by further statistical analyses, and were combined with five reference genes (ie, 16 genes in total), from which a recurrence score algorithm was developed. The recurrence score was then validated in an independent cohort of 626 patients from France with stage I–III clear cell renal cell carcinoma who had also undergone nephrectomy. The association between the recurrence score and the risk of recurrence and cancer-specific survival in the first 5 years after surgery was assessed using Cox proportional hazard regression, stratified by tumour stage (stage I vs stage II vs III).

Findings

In our primary univariate analysis, the continuous recurrence score (median 37, IQR 31–45) was significantly associated with recurrence-free interval (hazard ratio 3·91 [95% CI 2·63–5·79] for a 25-unit increase in score, p<0·0001). In multivariable analyses, the recurrence score was significantly associated with the risk of tumour recurrence (hazard ratio per 25-unit increase in the score 3·37 [95% CI 2·23–5·08], p<0·0001) after stratification by stage and adjustment for tumour size, grade, or Leibovich score. The recurrence score was able to identify a clinically significant number of both high-risk stage I and low-risk stage II–III patients. A heterogeneity study on separate samples showed little to no intratumoural variability among the 16 genes.

Interpretation

Our findings validate the recurrence score as a predictor of clinical outcome in patients with stage I–III clear cell renal cell carcinoma, providing a more accurate and individualised risk assessment beyond existing clinical and pathological parameters.

Funding

Genomic Health Inc and Pfizer Inc.

Introduction

Clear cell renal cell carcinoma is the most common subtype of kidney cancer and its incidence is increasing in high-income countries.¹ More than 80% of sporadic clear cell renal cell carcinomas are characterised by misregulation of the von Hippel-Lindau pathway and related hypoxia-inducible factor-responsive genes.² Although clear cell renal cell carcinoma is more resistant to cytotoxic chemotherapy than other tumour types,³ anti-angiogenic targeted therapies have shown substantial clinical activity in the metastatic setting.⁴ Several of these agents are now being assessed in the adjuvant setting after initial diagnosis of localised disease.⁵

The only curative therapy for patients with stage I–III clear cell renal cell carcinoma is surgery or ablation, and around 30% of patients treated for localised disease will relapse.⁶ Assessment of recurrence risk for patients with this disease is based on clinical and pathological features. The tumour–node–metastasis (TNM) staging system for clear cell renal cell carcinoma enables stratification of patients by pathological stage.⁷ Several systems combine TNM stage with clinical and pathological features to increase their prognostic capability, including the Mayo clinic score developed by Leibovich and others,8, 9 the University of California Los Angeles (UCLA) Integrated Staging System,¹⁰ and the Memorial Sloan Kettering postoperative nomogram.¹¹ Although useful, these scoring systems rely on parameters such as tumour grade, tumour necrosis, and patient performance status, which can be subject to inter-observer variability and might not entirely account for individual tumour biology.12, 13

Multigene assays have been shown to provide prognostic information beyond traditional parameters in several tumour types,14, 15, 16 and some are now included in treatment guidelines for these tumours.¹⁷ We sought to identify and validate a novel prognostic gene signature for clear cell renal cell carcinoma.