Order of patient entry and outcomes in phase II clinical trials: A meta-analysis of individual patient data.

Background: Prior meta-analysis of stem-cell transplantation trials for renal-cell carcinoma observed that clinical outcomes vary by subjects' order of entry, specifically their quartile of accrual. We test this hypothesis using meta-analysis of individual patient data from diverse Phase II trials conducted by an oncology consortium.

Methods: Eligible were all Phase II trials in hematologic or solid tumors opened and closed by California Cancer Consortium during 2005-2020. Excluded were trials closed in first quartile or currently embargoed pending publication and subjects ineligible per protocol or untreated on study. The primary risk factor was entry by quartile of planned sample size. As a cross-protocol endpoint, primary outcome was time to discontinuation of intervention. One-stage meta-analysis used a shared frailty model with trial as random effect. As covariates, stepwise selection retained tumor type, obesity, their interaction, calendar year, entry at least 3 years post-diagnosis, and performance status but rejected age, sex, randomized design, and class of drug.

Results: Twenty trials (including 8 terminated early, 2 not published) included n = 923 subjects. Most (90.6%) subjects discontinued intervention, usually for disease progression or toxicity. Independently of covariates, risk of discontinuation increased (p < 0.0001) with each quartile of entry (Hazards Ratio 1.13, 95% CI 1.06-1.22), culminating at Quartile 4 (HR 1.46, 1.36-1.57). The 95% prediction interval for the Hazards Ratio in future trials was (1.04-1.24). Progression-free survival similarly worsened by quartile of entry.

Conclusion: In Phase II trials, clinical outcome worsens with quartile of entry. This finding merits independent replication, and the cause of this phenomenon merits investigation.