FDA and Medical Devices

Abstract & Commentary

By John P. DiMarco, MD, PhD

Source: Dhruva SS, et al. Strength of study evidence examined by the FDA in premarket approval of cardiovascular devices. JAMA. 2009;302:2679-2685

In this paper, Dhruva et al examine the types of studies that were involved in premarket approval (PMA) of cardiovascular devices by the FDA over a seven-year period. PMA applications are required for novel or high-risk medical devices, a relatively small fraction of the new medical devices marketed each year. In the PMA application, the sponsor submits data to the FDA to document the safety and effectiveness of the proposed device. After a PMA is granted, the FDA makes available a summary of the safety and effectiveness data, the SSED report, which it evaluated before granting the PMA. In this study, Dhruva et al reviewed the SSED reports from cardiovascular PMAs issued between 2000 and 2007. The types and numbers of devices covered were coronary stents (12), vascular stents (21), ablation catheters or systems (8), pacemaker or defibrillator systems or components (14), ventricular assist devices (2), cardiac valves (4), and occlusion devices or sealants (17). Supporting these 78 PMAs were 123 studies. Sixty-five percent of the PMAs were supported by a single study; only eight PMAs were supported by more than two studies.

Of the 123 studies included, only 33 (27%) of the SSEDs that were reviewed were randomized and just 17 were blinded. Except for studies on cardiac stents, most device studies were neither randomized nor blinded. Follow-up time varied by type of device, ranging from 365 days for intracardiac devices and endovascular grafts to one day for femoral artery hemostasis devices. Definition of a primary study endpoint was inconsistent. In fact, 17 of 123 studies (14%) did not state a specific primary endpoint. Many other studies reported multiple endpoints but did not explicitly specify one as the primary endpoint. Frequently, endpoint data were not compared with data from concurrent controls. Even when a control group was reported, in 34 instances, the controls were retrospective. Most primary endpoints were surrogate endpoints; for example, target lesion revascularization for a coronary stent, primary patency for an endoprosthesis, or lead implant success for a pacemaker or defibrillator lead and longer-term clinical outcomes were not reported.

Other findings of interest were also described in the review. Many studies did not include data from training, lead-in, or roll-in patients, and some studies had only a post-hoc analysis of a primary endpoint. Twenty-seven percent of the studies were performed completely outside the United States.

Dhruva et al conclude that their data suggest that there are significant limitations in publicly available data for the evidence leading to PMA of cardiovascular devices. They encourage the FDA to require higher quality studies, as well as make the data from these studies available for public analysis and comment.


The FDA has well-established protocols for approval of new drug applications. In most instances, the new drug, or new indication for an approved drug, must be supported by two randomized, controlled studies, each of which must achieve statistical significance. Rarely, a single large study may be powerful enough to justify approval. The compound must prove itself to be either superior to placebo for the proposed indication or noninferior to an accepted therapy for the condition being treated. The primary endpoint must be stated clearly. Applications based only on superiority of secondary endpoints are often not approved.

Such a rigorous approach has not been taken with PMA applications for medical devices, even class III devices that have significant risk potential. The approval of medical devices has some features notably different from that for drugs. Devices change rapidly as technology advances, and the life span of an individual model may only be quite short. Device failures may not appear for several years after implantation and often occur with a frequency that would be undetectable in the usual PMA application trial sample. These and other dissimilarities have led to a very individualized approach to the approval process for many devices. This paper by Dhruva et al emphasizes our need to improve the requirements for device approval. More robust initial trials and better post-market approval follow-up studies should be routinely required, and the FDA should be provided with adequate resources for analyzing all the data. As devices become more commonly used, we must continue to improve their safety and efficacy. Although it is unlikely that the same criteria now used for new drug applications will always be applicable to device trials, it is clear that more rigorous criteria need to be established for these high-cost, high-risk, but potentially life-saving technologies.