Adaptive design increased success but does not significantly accelerate confirmatory oncology trials.

BACKGROUND

Adaptive designs are increasingly employed to improve trial efficiency; however, their benefits over fixed designs in late-phase oncology trials remain underexplored. We aimed to determine whether adaptive designs in phase 3 cancer trials are associated with primary completion time and primary endpoint success compared with fixed designs.

METHODS

This cross-sectional study analyzed phase 3 randomized lung and breast cancer trials initiated between 2000 and 2020 using Aggregate Analysis of ClinicalTrials.gov database (static version downloaded April 15, 2025). Eligible trials had time-to-event primary endpoints and were completed or terminated. Trials were classified as adaptive or nonadaptive based on preplanned design modifications. Associations between trial design and primary completion time were examined using Kaplan-Meier curves and multivariable cause-specific Cox regression, while associations between trial design and primary endpoint success were examined using multivariable logistic regression.

RESULTS

Of 421 eligible trials, 210 employed adaptive designs. Adoption of adaptive designs increased over time, with group-sequential designs being the most common adaptive approach. Adaptive designs showed earlier primary completion, but no statistically significant association was observed after multivariable adjustment. Adaptive trials had higher hazards of early stopping for efficacy or futility and lower hazards of administrative termination. Among 396 trials with available results, 163 achieved primary endpoint success. Adaptive trials had higher odds of success than nonadaptive trials (66.26% vs 33.74%; odds ratio [OR] = 2.363, 95% CI = 1.491 to 3.778; P < .001).

CONCLUSIONS

Adaptive designs were associated with higher primary endpoint success but not shorter primary completion time. Given high failure rates in late-phase oncology, adaptive designs may improve efficiency primarily by facilitating data-driven decisions rather than accelerating trial timelines.