Control arm overperformance in phase 3 oncology clinical trials

Area of Science:

• Oncology Clinical Trials
• Biostatistics
• Meta-epidemiology

Background:

• Accurate prediction of control arm outcomes is crucial for power calculations and enrollment targets in randomized clinical trials (RCTs).
• Unexpected overperformance by the control arm can lead to underpowered studies, potentially affecting trial results and conclusions.

Purpose of the Study:

• To estimate the prevalence of control arm overperformance in phase 3 oncology RCTs.
• To assess the association between control arm overperformance and trial outcomes.

Main Methods:

• A meta-epidemiological study was conducted on two-arm, superiority-design, phase 3 oncology RCTs.
• Data on power calculations, justifications, and outcomes were extracted from publications and protocols.
• Control arm performance was quantified as the ratio of observed to pretrial estimated outcomes; overperformance was defined as a 10% improvement.

Main Results:

• Of 385 included RCTs, 43% exhibited overperformance.
• Overperformance was associated with lower odds of meeting the primary endpoint (aOR, 0.989; P = 0.015).
• RCTs using overall survival (OS) as the primary endpoint were more likely to show overperformance (56% vs. 33%; aOR, 2.43; P = 0.001).

Conclusions:

• Control arm overperformance is frequent in phase 3 oncology RCTs and impacts study power.
• The high prevalence of overperformance in OS-based trials may indicate underappreciation of postprogression therapies.
• Alternative trial design strategies are needed for more reliable estimation of treatment effects.