Investigating Time to Next Treatment As a Surrogate Endpoint for Overall Survival in Previously Untreated Intermediate- to Poor-Risk Advanced Renal Cell Carcinoma Patients: An Insight from the Phase III CheckMate-214
Author(s)
Branchoux S*1;Sofeu CL2;Kurt M3;Gaudin AF4;Italiano A5;Rondeau V2, Bellera C6
1Bristol Myers Squibb, RUEIL MALMAISON, France, 2Bordeaux Population Health Center, ISPED, Centre INSERM U1219, Bordeaux, France, 3Bristol Myers Squibb, Princeton, NJ, USA, 4Bristol-Myers Squibb, Rueil-Malmaison, France, 5Institut Bergonié Comprehensive Cancer Centre, Bordeaux, France, 6Epicene Team (Cancer & Environnement), Bordeaux Population Health Center, ISPED, Centre INSERM U1219, Bordeaux, France
OBJECTIVES: Time to next treatment (TNT) may be a patient-relevant endpoint by capturing durable response, post-progression benefit, and the time free of systemic anticancer therapy, which may be achieved with immune-checkpoint inhibitors (ICI). This study investigated TNT as a surrogate endpoint (SE) of overall survival (OS) in previously untreated intermediate- to poor-risk advanced renal cell carcinoma (aRCC) patients.
METHODS: Patient-level data from CheckMate-214 randomized clinical trial (RCT) (NCT02231749) with 42 months’ minimum follow-up were used. In this RCT, patients were randomly assigned to nivolumab combined with ipilimumab (NIVO+IPI) versus sunitinib. The SE one-step validation method based on a joint frailty model was used where the country of residence was applied to define synthetic clusters. Kendall’s 𝜏 and the coefficient of determination (R2trial) were estimated for measurement of the association at the individual and cluster levels, respectively. Surrogate threshold effect (STE), the maximum threshold hazard ratio (HR) for TNT that would translate into OS benefit, was estimated. A leave-one-out cross-validation analysis was performed to evaluate model robustness and predictive accuracy.
RESULTS: Twenty-two clusters were considered from 847 patients. The association between TNT and OS was deemed acceptable at the individual level (Kendall’s 𝜏 = 0.59; 95% confidence interval (CI) [0.56; 0.64]) and strong at the cluster level (R2trial close to 1; 95% CI [0.97; 1.03]). The estimated STE was 0.70 for NIVO+IPI versus sunitinib. Cross-validation results showed association measures were minimally influenced when the analyses were repeated with slightly smaller and balanced clusters. The predictive accuracy of the model was modest (45%), with optimistic OS HR predictions when TNT HR was below the STE.
CONCLUSIONS: TNT may be a valuable SE in previously untreated aRCC patients treated with ICI. Further surrogacy analyses considering multiple RCTs of ICI-treated aRCC patients are warranted for confirming these findings and for improving the accuracy of the predictions.
METHODS: Patient-level data from CheckMate-214 randomized clinical trial (RCT) (NCT02231749) with 42 months’ minimum follow-up were used. In this RCT, patients were randomly assigned to nivolumab combined with ipilimumab (NIVO+IPI) versus sunitinib. The SE one-step validation method based on a joint frailty model was used where the country of residence was applied to define synthetic clusters. Kendall’s 𝜏 and the coefficient of determination (R2trial) were estimated for measurement of the association at the individual and cluster levels, respectively. Surrogate threshold effect (STE), the maximum threshold hazard ratio (HR) for TNT that would translate into OS benefit, was estimated. A leave-one-out cross-validation analysis was performed to evaluate model robustness and predictive accuracy.
RESULTS: Twenty-two clusters were considered from 847 patients. The association between TNT and OS was deemed acceptable at the individual level (Kendall’s 𝜏 = 0.59; 95% confidence interval (CI) [0.56; 0.64]) and strong at the cluster level (R2trial close to 1; 95% CI [0.97; 1.03]). The estimated STE was 0.70 for NIVO+IPI versus sunitinib. Cross-validation results showed association measures were minimally influenced when the analyses were repeated with slightly smaller and balanced clusters. The predictive accuracy of the model was modest (45%), with optimistic OS HR predictions when TNT HR was below the STE.
CONCLUSIONS: TNT may be a valuable SE in previously untreated aRCC patients treated with ICI. Further surrogacy analyses considering multiple RCTs of ICI-treated aRCC patients are warranted for confirming these findings and for improving the accuracy of the predictions.
Conference/Value in Health Info
2020-11, ISPOR Europe 2020, Milan, Italy
Value in Health, Volume 23, Issue S2 (December 2020)
Code
EX3
Topic
Clinical Outcomes
Topic Subcategory
Relating Intermediate to Long-term Outcomes
Disease
Oncology