Medical Device Daily Washington Editor
SAN FRANCISCO — Device makers have a lot of audiences to keep track of, and one of yesterday's sessions at this year's Transcatheter Cardiovascular Therapeutics conference sheds light on a necessary evil of clinical trials for medical devices: that of the composite efficacy or safety endpoint. The problem for device makers is that one audience, namely FDA, pretty much mandates such endpoints, but another audience, physicians, are not always sure what to make of these aggregated measures.
Despite the headaches associated with interpreting composite endpoints, the panel agreed on at least one thing; they're not going away anytime soon. On the other hand, one unspoken message was that device makers can do doctors a big favor by publishing statements about their trials that highlight in great detail the relative contributions of the components to the composite score.
Panel moderator Stuart Pocock, PhD, of the London School of Hygiene and Tropical Medicine (London), opened the session by observing, "we think we need [composite endpoints] because individual outcomes lack power in their own right." This was a tacit acknowledgement that composites are popular partly because they help keep clinical trial enrollment numbers down. Whatever the utility of composite endpoints, though, "we soon learn to hate them because we know that their clinical importance varies across components" embedded in the study data, he observed.
Pocock did not seriously question whether physicians will pay attention to composites, but offered a cautionary statement: "My number one message is that you're going to do it, but explore the details."
Pocock referred to a survey appearing in the Annals of Internal Medicine of 304 trials for cardiovascular devices, 73% of which used composites. This is most likely a reference to the article by Eric Lim, MD, et al in the Nov. 4, 2008, edition of the Annals. According to Pocock, most of those composite measures included death (98% of the 304 studies), myocardial infarcts (92%) and reintervention (54%).
An example offered by Pocock was the Typhoon trial, which used cardiac death, infarct and target vessel revascularization (TVR) as a composite. "If you don't look closely, you might find" the P score of .004 reassuring as to the Cypher's superiority, "but all the action is in the TVR," he said. The Typhoon study compared the Cypher, made by Cordis (Miami Lakes, Florida) to a bare-metal stent.
Flipping the coin over, Pocock noted that the Syntax trial, which compared stenting to bypass grafting, was billed in the write-up in the New England Journal of Medicine as "failing to show non-inferiority" due to the use of a composite, but Pocock asserted that "this was not the greatest choice" of endpoints. Revascularization rates drove the higher rate of major adverse events in the stent arm of the study, but bypass incurred significantly higher rates of stroke. Of revascularization, Pocock remarked that the better perspective might be that "this is part of the package" that allows patients to avoid sternotomy, "and not really a clinical event" like death or stroke.
Pocock made several other arguments on this topic, but concluded with a suggestion that physicians "always pay attention to the individual components" of a composite. Thus, while regulatory authorities "force us to choose a single endpoint ... real life is not like that" and physicians would do well to approach the literature with the view that most trials "have multiple trials within them."
David Cohen, MD, of St. Luke's Mid-America Heart Institute (Kansas City, Missouri), started his talk by arguing that "there are advantages" to composites, including that they "improve statistical power" even as they "avoid analytical challenges for competing risks."
Composites in and of themselves offer no clinical advantages, though, Cohen asserted. "We usually interpret them to imply that all endpoints have equal clinical weight," he said, but "that's not usually the case," especially if those individual endpoints trend in utterly different directions.
Cohen discussed several examples as well as an analysis of the topic by Ignacio Ferriera-Gonzalez appearing in the British Medical Journal in 2007. In the abstract, Ferriera-Gonzalez wrote: "The use of composite end points in cardiovascular trials is frequently complicated by large gradients in importance to patients and in magnitude of the effect of treatment across component end points." He also made the argument that "higher event rates and larger treatment effects associated with less important components may result in misleading impressions of the impact of treatment."
Cohen said he sees a lot to be wary of where composites are concerned and confessed to a pet peeve. "I would really like to get rid of MACE (major adverse cardiac events) as an endpoint," he confessed.
Andrew Farb, MD, of the Office of Device Evaluation at FDA's Center for Devices and Radiological Health opened by promising, "I'm going to join the amen chorus about composite endpoints."
Farb noted that a growing contributor to the proliferation of composites is the add-on treatment, such as device-plus-medical management, which is more commonplace now. Composites are "best suited for well-characterized disease states," and "when there's strong consensus in the clinical community that the components run in the same direction" for the most part, he observed.
"There's an art to creating a composite endpoint," Farb offered, and the endpoints should obviously not be mutually exclusive or inclusive, and they should avoid inflation and deflation of the composite.
Hence, while composites can reduce sample size because of the greater number of events recorded, the idea that equal weight of components assumes equal clinical significance "can be controversial," Farb said.
This remark hints at the possibility of some sort of rubric for relative weighting of the endpoint components, but Farb acknowledged that there is as yet no consensus on how to weigh them. At the end of his talk, Farb offered a comment that summed up the intractability of the situation: "Composite endpoints should be based on sound clinical judgment" and "not [used] solely for statistical advantage." n
Mark McCarty, 703-966-3694;