Surrogate endpoints for overall survival in locally advanced head and neck cancer: meta-analyses of individual patient data

Abstract

The gold standard endpoint in randomised trials of locally advanced head and neck squamous-cell carcinoma (HNSCC) is overall survival. Our objective was to study whether duration of locoregional control or event-free survival (EFS) could be considered as surrogate endpoints to estimate the effect of radiotherapy and chemotherapy on overall survival. This would allow a reduction in the duration and cost of the development of new treatments. Individual patient data from 104 trials (22 744 patients), with 116 treatment-control comparisons, from four meta-analyses on hyperfractionated or accelerated radiotherapy and concomitant, induction, or adjuvant chemotherapy were analysed. Duration of locoregional control was defined as the time from randomisation to the first locoregional event and EFS as the time to any first event (ie, locoregional relapse, distant recurrence, or death). At the individual level, a rank correlation coefficient between the surrogate endpoint and overall survival was used to assess surrogacy; at the trial level, a correlation coefficient R between treatment effects was used. At the individual level, overall survival was more strongly correlated with EFS (range of correlations 0.82-0.90) than with locoregional control (0.65-0.76). For radiotherapy, treatment effects on both locoregional control and EFS were strongly correlated with those on overall survival (R=0.94 and 0.98, respectively). For chemotherapy, the correlations between treatment effects on EFS and overall survival were stronger than those between locoregional control and overall survival (range of R 0.79-0.93 vs 0.53-0.84, respectively). EFS is a better correlate with overall survival than locoregional control and could be used as a surrogate for overall survival to assess the treatment effect of radiotherapy and chemotherapy in randomised trials of locally advanced HNSCC.