A pair of articles published in JAMA
examined the quality of studies used by the FDA to support its accelerated approval decisions and high-risk device modification approvals.
In one article
, researchers looked at the characteristics of clinical trials of drugs for 24 indications granted accelerated approval by the FDA from 2009 to 2013. These indications were supported by a total of 30 studies conducted prior to approval; 18 confirmatory studies were published after approval.
Of the preapproval studies, 40% were randomized and 20% were double blinded. As most of the drugs being studied were oncology drugs, the most common surrogate outcome used in the trials was disease response. For 14 indications, the FDA’s decision to grant accelerated approval was based solely on non-randomized, noncomparative single-group studies.
FDA acknowledged the limitations of these studies on its labels for the drugs in question, and required that 38 confirmatory trials be conducted to demonstrate the efficacy, long-term outcomes, and safety of these drugs. As of this April, 19 of these required studies had been completed.
Examining the findings of the 18 published confirmatory studies, the authors found that 56% were randomized but just 1 of the trials was double-blinded. In all but 1 of the studies, the primary endpoints were surrogate measures, most commonly disease response. For 2 of the indications granted accelerated approval, the confirmatory studies failed to demonstrate clinical benefit.
The researchers summarized that among the 24 indications studied, “42% had efficacy confirmed in postapproval trials a minimum of 3 years after approval, although both preapproval and postapproval trials had limitations in the study designs and end points used.”
The second article
also found limitations in study design and endpoints, but among the trials used to support FDA’s approval of modifications to high-risk medical devices, such as coronary stents or hip prostheses. In their analysis of 83 clinical studies used to support 78 approval supplements, the researchers identified several methodological shortcomings.
For instance, 45% of the studies were randomized and 30% were blinded. Eleven did not indicate the number of patients enrolled in the trial, while 40% did not include age data for all participants. The researchers identified 150 primary endpoints in the trials, but found that 57% specified the type of endpoint analysis and only 38% were compared with controls.
Of these primary endpoints, 81% were surrogate measures. The study authors also noted incomplete reporting in that some patients were excluded from endpoint analysis for two-thirds of the endpoints. In a trial of a modified radiology device for mammography, 91% of participants were not included in primary endpoint analysis, raising concerns about study bias.
According to the study authors, these findings indicate that “lower-quality data often supported changes in high-risk devices that are modifications to previously approved devices.” They recommended that the FDA should require more rigorous, high-quality studies and data before approving such modifications.