This thesis focuses on the pivotal, yet frequently underappreciated aspect of randomized controlled trials (RCTs) – the control arm. Our research highlights the critical importance of rigorously designing and implementing this component within RCTs.
The rise of the randomized controlled trial as one of the most robust methods for estimating the effect of a medical intervention is born out of a long history of ideas, particularly in statistical and medical sciences. Randomization has three main advantages: balancing potential confounding factors (known and unknown), avoiding or at least limiting the issue of analytic flexibility, and finally determining a time-zero - the time of randomization - which avoids numerous biases which can affect observational studies. The principle of equipoise, and uncertainty about the result of an experiment, is crucial as it justifies conducting a randomized trial. Estimates suggest that the frequency of suboptimal control arms, within modern clinical trials in oncology, varies from 11% to 17%.
However, based on our works, this proportion may be underestimated, as no systematic framework has been applied to assess for control arm’s quality. Beyond factors classically assessed for, we described specific features which can unfavorably affect the control arm. For instance, when the control arm is left to the “physician's discretion”, this choice may seem optimal, yet we show that it was restricted in 85% of cases, and such restriction can deprive the patients of essential treatment options. We also demonstrate that the treatment duration in the control arm, when compared to the duration in previous clinical trials in identical settings, can decrease. This raises the question of whether investigators are less incentivized to promote the control treatment, particularly in open-label trials. Lastly, we show that a majority (55%) of head- to-head registration clinical trials in oncology have dose modification rules or supportive care usage that penalize the control arm.
In the conclusion, we higlight that the qualitative analysis of the control arm is not taken into consideration by professional society evaluation scores (such as ASCO and ESMO) and other systematic evaluation methodologies from the evidence-based medicine (EBM) movement. We propose an evaluation framework that could prove beneficial to clinicians, patients, Health Technology Assessment (HTA) agencies, and regulators.