ctDNA Clearance as a Trial-Level Early Efficacy Signal in Solid Tumors: A Systematic Review and Meta-analysis of Randomized Trials

Background: Circulating tumor DNA (ctDNA) is a sensitive marker of residual disease and treatment activity, but its validity as a trial-level surrogate endpoint remains uncertain. Given scarce randomized evidence and clinical and methodological heterogeneity, we performed an exploratory, hypothesis-generating evaluation of trial-level associations between ctDNA clearance and clinical outcomes across solid tumors.

Methods: Randomized controlled trials reporting ctDNA clearance and overall survival (OS), progression-free survival (PFS), or recurrence-related outcomes were identified through April 2026. Analyses were exploratory and conducted at trial level. Weighted regression, leave-one-out cross-validation, copula correlation, Bayesian hierarchical modeling, and sensitivity analyses were performed, with cross-validated R² (R²cv) prioritized over apparent R² for out-of-sample interpretation.

Results: Eighteen trials were included. Out-of-sample performance was limited: R²cv was 0.41 for PFS, 0.26 for OS, and –1.66 for recurrence-related outcomes. Corresponding apparent R² estimates were 0.77 (95% CI 0.03–0.96), 0.56 (0.12–0.84), and 0.04 (0.00–0.92), respectively. The estimable PFS signal was derived from eight palliative/metastatic trials. Recurrence-related findings were inconclusive; after excluding the ctDNA-guided DYNAMIC-III trial, apparent R² increased to 0.67 but R²cv remained negative (–0.18). Post-treatment ctDNA status remained prognostic across endpoints.

Conclusion: These exploratory, hypothesis-generating findings identify an endpoint- and setting-dependent trial-level signal, with the largest observed association for PFS, but do not validate ctDNA clearance as a surrogate endpoint. Wide uncertainty intervals, few contributing trials, and limited cross-validated performance require prospective validation before use in trial qualification or individual treatment decisions.