PCN360 SURROGATE ENDPOINT VALIDATION STUDIES SUBMITTED TO THE GERMAN HEALTH TECHNOLOGY ASSESSMENT AUTHORITY, THE FEDERAL JOINT COMMITTEE (G-BA). CASE STUDIES IN ONCOLOGY

Abstract

Oncology trials using overall survival (OS) as a primary endpoint need substantial sample sizes and extensive follow-up. An alternative surrogate endpoint for OS is progression-free survival (PFS). The Institute for Quality and Efficiency in Health Care (IQWiG) has issued a guidance document for surrogate endpoint validation. We aimed to review submissions made to G-BA, where an endpoint validation study was submitted. Oncology submissions published between 2011 and 2019 with endpoint validation studies were included. Information on interventions, indication, pivotal study design, methodology, endpoints, criticism, and final recommendations were extracted. Out of 153 reviewed submissions, five submissions were included. Submissions were found in two indications: breast cancer (ribociclib, pertuzumab & palbociclib) and melanoma (dabrafenib and pembrolizumab). All submissions were in the advanced/metastatic setting except for pertuzumab which was in early-stage breast cancer. In all submissions, efficacy/safety evidence came from RCTs. The defined primary endpoint in these RCTs was PFS, except for pertuzumab which was in disease-free survival (DFS). Trial-level validation methodologies were applied in all submissions. For pertuzumab, additional analyses were also conducted using patient-level data. Due to the observed medium correlation between PFS or DFS and OS, it was concluded that the validity of PFS or DFS as surrogate endpoints was unclear. Consequently, the surrogate threshold effect (STE) was presented. Only in two submissions, ribociclib and pembrolizumab, the 95% confidence interval of the treatment effect on PFS lied below the estimated STE. Nevertheless, final G-BA conclusion for all submissions was that PFS/DFS is not a surrogate endpoint for OS. Key criticisms included inappropriate eligibility criteria for included studies regarding patient population, interventions and endpoint definition. Analyses proposed by IQWiG were done and presented in very few submissions to G-BA, and these were not perceived positively. This might be explained by the stringent threshold recommended by IQWiG.