Are phase III trials still important for FDA drug approval?

Abstract

The pace of new drug approvals by the United States Food and Drug Administration (FDA) over the past decade has been rapid, particularly in the realm of oncology. The availability of new drugs has expanded treatment possibilities and improved outcomes in many patients. Traditionally, drug approval requires randomized phase III studies, better yet placebo-controlled and double-blind when feasible. Such studies need large investments in terms of financial capital and patient participation, but they are considered the final arbiter of drug efficacy. However, as Drs. Al Hadidi and Ramos describe in this issue, many FDA drug approvals are based exclusively on early phase clinical trial data (phase I/II) [1]. Specifically, early phase trials led to 40% of the 49 approvals in the field of hematologic malignancies over a nearly five-year period from January 2016 to May 2020. The traditional goal of early phase clinical trials is to establish dose and schedule, evaluate toxicity, and, in preliminary fashion, assess efficacy; not until a phase III study is performed can a more definitive statement be made about efficacy. With so many drugs approved based solely on phase I/II data, are phase III trials still important for FDA drug approval? The authors report that the percentage of approvals supported by phase III/IV trials in their dataset increased over time, up to 69% in 2019. However, it is worth noting that even late phase clinical trials can be fallible. Clinical trial populations may not be representative of the population who will receive a drug (at best) or intentionally chosen to produce a “favorable” outcome (at worst). In one retrospective analysis, outcomes were worse for patients with newly diagnosed acute myeloid leukemia (AML) treated “off-study” with regimens also under investigation as a clinical trial at an academic center [2]. Further, many phase III trials are designed to reach an arbitrary significance threshold of p1⁄4 0.05, ignoring the frequent lack of a relationship between statistically and clinically significant improvement. The “fragility index” measures the ease with which a significant p value can become non-significant in a randomized trial. One recent analysis found that a large percentage of phase III clinical trials for anti-cancer medications approved by the FDA have a low fragility index, indicating that small changes in the classification of events could easily shift the interpretation of the outcome [3]. The objectivity of the whole p-value-based approach has itself been questioned [4]. Further, outcomes from several recent phase III trials in AML raise questions about earlier FDA approvals. For example, the FLT3 inhibitor gilteritinib was approved in December 2018 for patients with relapsed/refractory (R/ R) FLT3-mutated AML based on results from the randomized phase III trial ADMIRAL [5]. However, the LACEWING study, which randomized R/R FLT3-mutated AML patients to receive azacitidine±gilteritinib failed to meet its primary endpoint of overall survival [6]. A similar failure was seen in the phase III IDHENTIFY study, which randomized R/R IDH2-mutated AML patients to receive azacitidine±enasidenib; the study did not meet its primary endpoint [7]. The approvals of gilteritinib and enasidenib were not conditional based on the referenced trials, so both drugs will continue to be commercially available. It will be important to analyze the forthcoming results from the LACEWING and IDHENTIFY trials to inform clinicians’ understanding about the differences between previous and current data, since negative findings can also change clinical practice [8]. A related issue is the use of inappropriate control arms in phase III trials [9]. Nonetheless we agree with the authors’ emphasis on the importance of confirming clinical activity in phase III studies when feasible for hematologic