trial design • drug development • drug registration • patient benefit • risks of bias Pivotal clinical trials: biased? A recent study by Naci et al. examined the design characteristics, risks of bias and reporting adequacies of pivotal randomized controlled trials of cancer drugs approved by the EMA in the period 2014-2016 [1]. During this period, 32 new cancer drugs were approved by the EMA on the basis of 54 pivotal trials-of these, 41 (76%) were randomized controlled trials and 13 (24%) were non-randomized or single arm trials. The study reported that 49% of trials were judged to be at high risk of bias based on aspects of their design, conduct or analysis. Furthermore, only 10 (26%) randomized controlled trials measured overall survival as a primary end point, with the majority of trials evaluating surrogate metrics such as progression free survival or response rates. Furthermore, there were also discrepancies between scientific publications and regulatory documentation, thereby raising concerns of reporting inadequacies and risks of bias in both sources of information. These findings raise serious questions regarding clinical trial design and the drug registration process, particularly with regards to patient benefits. The problem is not restricted to Europe alone; previous studies have also raised concerns with US FDA approvals. For example, a study that assessed all FDA approvals of novel drugs, across all areas of medicine in the period 2005-2012, reported wide disparities in pivotal trial characteristics across different therapeutic areas [2]. While overall 219 of 448 (48.9%) pivotal trials included a surrogate end point, for cancer it was 46 of 55 (83.6%). Pivotal trials in cancer were also less likely to be randomized, double-blinded or to not include a placebo or active comparator. Another notable finding was that most cancer approvals were on the basis of a single pivotal trial, whereas for other therapeutic areas, approvals were based on two to four pivotal trials. A different study examining FDA cancer drug approvals based on response rate showed that many cancer drugs are approved based on low or modest response rates, typically tested in single-arm studies [3]. It could be argued that the surrogate end points used in such pivotal trials are acceptable as they correlate to overall survival and/or quality of life in the long term [4]. However, the data to support such an argument is lacking [5]. A recent systematic review of the association between surrogate end points and overall survival in oncology showed that the vast majority (82%) are low or moderate in strength [6]. A study of FDA approvals of cancer drugs during 2008-2012, a period in which 67% of approvals were made on the basis of a surrogate end point, assessed whether subsequent data emerged to show an increase in overall survival [7]. With a median follow-up of 4.4 years, the study reported that of 36 drugs approved with surrogate end points, only five subsequently reported improved overall survival in a randomized controlled trial, 18...