Relapse-Free Survival (Rfs) As a Surrogate Endpoint for Overall Survival (Os) in Adjuvant Interferon Trials in Patients (Pts) with Resectable Cutaneous Melanoma: an Individual Patient Data (Ipd) Meta-Analysis

Abstract

ABSTRACT Aim: We aimed to assess whether RFS is a valid surrogate for OS for high risk stage I-III melanoma through a meta-analysis of randomized controlled trials. Methods: All trials with available IPD on RFS (time until recurrence/death) and OS were collected from high risk stage I-III pts enrolled in 11 randomized adjuvant trials (N = 5826) comparing interferon (IFN) to observation: EORTC (18871, 18952, 18991), ECOG (1684, 1690), MRC (AIMHIGH), WHO-16, NCCTG (83-7052), DKG 80-1, DeCOG and Scottish group. In addition IPD from ECOG 1694 trial, IFN vs GMK vaccine (N = 842), and ECOG 2696, IFN administered concomitantly/sequentially with GMK vaccine vs GMK vaccine (N = 107) were added. We applied a two-level modeling approach assessing Spearman's individual-level correlation of RFS and OS (rho) and the trial-level coefficient of determination (R2) of the treatment effect on RFS and on OS. Results: At a 7-yr median follow-up, a consistent advantage of IFN vs No IFN was observed for RFS and OS (see table). R2 could not reliably be estimated when all trials were considered. A proper R2 estimate was obtained after exclusion of ECOG 2696 trial, which was an outlier [HR(RFS) = 0.72 vs HR(OS) = 1.11]. As a sensitivity analysis, the model was also fitted after excluding the other vaccine comparator trial (ECOG 1694). In these two analyses, RFS and OS were strongly correlated and the surrogate threshold effect (STE) was estimated to be a HR of 0.774 and 0.786, respectively. The leave-one-out cross-validation showed consistently good model accuracy. All pts (N = 6815) All but ECOG2696(N = 6708) No vaccine trials (N = 5826) No IFN IFN HR * No IFN IFN HR * No IFN IFN HR * 5-yr RFS 35.2% 38.8% .87 35.2% 38.8% .88 34.0% 38.0% .88 5-yr OS 47.2% 50.3% .91 47.0% 50.2% .91 45.4% 49.0% .91 rho* * .887 .887 .887 R2* * - .906 .924 STE* * - .774 .786 * Forest plots; **Plackett copula. Conclusions: We found that both individual-level and trial-level correlations are close to 1, suggesting a valid surrogacy of RFS for OS. If, in a future trial, the treatment effect HR on RFS were less than 0.774, one would expect this trial to also show a treatment effect on OS. Disclosure: A.M.M. Eggermont: Member of advisory board: AMGEN, BMS, GSK, MedImmune, MSD; P. Lorigan: Advisory board member: BMS, Celgene, GSK, Merck, Novartis, Roche. Other supportive relationships: Support for travel BMS, Merck, Roche; J. Kirkwood: Consultancy: BMS, Celgene, MSD, Prometheus, Vical, Ziopharma; C. Garbe: Member of advisory board: BMS, GSK, MSD, Novartis, Philogen, Roche Corporate sponsore research: BMS, GSK, Roche; D. Cameron: Member of advisory board (compensated to employer): Roche Corporate sponsored research (paid to another institution): Roche; S. Kotapati: Stock ownership: BMS Corporate sponsored research: BMS; C. Konto: Stock ownership: BMS Corporate sponsored research: BMS Other: employee of BMS; T. Chen: Stock ownership: BMS Corporate sponsored research: BMS; M. Buyse: Stock ownership: IDDI. All other authors have declared no conflicts of interest.