Early mortality rates: a tool for phase III trials or for changing standard practice?

Defining and describing (dose-related) side-effects of chemotherapy are among the aims of phase I and II oncology trials. Phase I studies are intended to assess the safety of a drug or regimen, phase II to confirm the feasibility and assess long-term safety. They are usually performed in a single or just a few institutions. Due to the novelty of the regimen and the fact that safety is considered a major concern, it is sometimes the case that patient selection in such trials is rather conservative. If phase I and II trials are performed appropriately, the safety aspect should become less of a concern in phase III trials since investigators will know what to expect, how to manage the observed side-effects, and how to avoid entering patients that should not be entered in view of organ dysfunction, poor clinical condition and other factors. Nevertheless, since phase III studies have a different aim, and are mostly multi-centre, it is possible that the population becomes less well selected, whilst remaining within the limits of the protocol criteria. Most of the time, phase I and II studies have indeed appropriately predicted side-effects and designed ways to manage them. Fortunately, exceptions are truly rare. However, that does not mean that such exceptions do not occur. Whether this means that ‘‘Early mortality detected in a clinical trial is one of the most important parameters for assessing the safety of chemotherapy regimes’’, as indicated by Katopodis and colleagues [1], in their paper in the current issue of European Journal of Cancer entitled ‘‘60-Day all-cause mortality rates in patients treated for gastrointestinal cancers, in randomised