Linking early and late endpoints in adjuvant immunotherapy trials: an evaluation of multiple clinical outcomes

Overall survival (OS) in adjuvant and perioperative immunotherapy trials requires prolonged follow-up. We examined whether treatment effects on disease-free survival (DFS) and recurrence-free survival (RFS) were associated with OS at matched 2-, 3-, and 5-year windows.

The main analysis included 23 strict-primary randomized trials. DFS and RFS were modeled separately across 6 paired time windows using 2 co-primary trial-level approaches: a conventional OS-variance-weighted regression and a conditional errors-in-variables model accounting for uncertainty in both log-HRs. Because within-trial correlations were unavailable, ρ was varied from 0 to 0.75. Eligibility-sensitivity studies were supplementary; broad-taxonomy evidence was excluded. EFS was descriptive, PFS and PFS2 were analyzed separately, and QoL regression required at least 5 verified contrasts.

Nineteen trials contributed 2-year same-window pairs (12 DFS; 7 RFS), 8 contributed 3-year pairs, and 5 contributed 5-year pairs. For 2-year early endpoints vs 2-year OS, weighted slopes were 0.014 (95% CI, -0.286 to 0.314) for DFS and 0.630 (95% CI, -1.243 to 2.502) for RFS. Eight of 9 fitted weighted-regression intervals included zero; the only nominally positive result involved 4 RFS trials. In errors-in-variables analyses, slopes for the largest DFS and RFS comparisons ranged from 0.011 to 0.264 and from 0.758 to 0.885, respectively, and all profile CIs included zero. Six strict-primary QoL contrasts were verified, but no window reached n = 5.

Strict-primary analyses did not demonstrate a consistent or precise early-endpoint-OS relationship. Sparse later windows and uncertainty in both treatment-effect estimates preclude surrogate validation.