Remarks on the Lancet Stent Study
A Letter to Colleagues from Dr. Donald S. Baim, M.D.
Executive Vice President, Chief Medical and Scientific Officer
Boston Scientific Corporation
September 14, 2007
Statement of Dr. Donald S. Baim, M.D.
September 14, 2007
The TAXUS® Paclitaxel-Eluting Stent and the Cypher® Sirolimus-Eluting Stent have each been implanted more than 4 million times worldwide since their respective approvals several years ago. A large number of individual studies comparing the devices have been published, some showing a benefit of one device over the other in some measure of efficacy (preventing restenosis) or safety (incidence of subsequent death, myocardial infarction, or stent thrombosis) – sometimes favoring the TAXUS Stent and sometimes favoring the Cypher Stent. Because the findings are so variable, looking at any single study (or the latest study) as an estimate of the “Truth” about relative performance of these excellent drug eluting stents can be very confusing.
As an example, the Bern-Rotterdam registry presented at ESC Vienna last week and just published in Lancet, followed 8,146 patients for two to four years, and found a small difference in cumulative stent thrombosis rates favoring the Cypher Stent over the TAXUS Stent. On the other hand, the New York State Registry just published in the American Heart Journal (Hannan et al), reported on 11,781 patients followed for 18 months and found small trends in death, death/MI, and repeat revascularization favoring the TAXUS Stent.
One way to make sense of such conflicting data is to try to assemble many separate studies into a single overall analysis (called a “meta-analysis”). Even that effort is challenged by the fact that individual studies differ in quality (some are randomized and others are non-randomized registries), use different definitions, and can be “selected” for inclusion so as to favor one stent or another in one or more performance parameters.
Thus, the recent Schomig meta-analysis of 16 trials including 5,562 patients (JACC October 2007) was strongly influenced by four older trials that had very strong outcomes favoring the Cypher Stent, helping it reach the conclusion that the TAXUS Stent and the Cypher Stent had the same risk of death or MI, but that the Cypher Stent had lower rates of stent thrombosis and repeat procedures. Yet the accompanying editorial by Kereiakes pointed out that when ALL of the available data were included (some 29 studies with more than 46,000 patients [including the four older trials mentioned above]) there was not even a trend favoring one stent over the other in either stent thrombosis or repeat revascularization. In essence, this broader view supports the impression that the two drug eluting stent platforms perform similarly in all important clinical performance measures in general patient populations (as they had in the more restrictive patient populations studied in the pivotal FDA trials).
An additional meta-analysis by Stettler (Lancet September 2007) tries to infer how the Cypher Stent and the TAXUS Stent compare, reasoning indirectly from published data. The indirect comparison tries to draw inferences by conjoining separate trials that compare the Cypher Stent to the bare metal (Bx Velocity) Stent and trials that compare the TAXUS Stent to a different bare metal (Express) Stent.
Unfortunately, the types of patients treated in the respective Cypher Stent and TAXUS Stent trials included in the Stettler meta-analysis were different (there is no Cypher Stent trial analogue of the small vessel and long lesion patients studied in TAXUS V and VI). In addition, using the separate trials to estimate how the Cypher Stent and the TAXUS Stent compare assumes that the two bare metal stents (BMS) used in the respective trials (Bx Velocity Stent and Express Stent) perform equally. In these trials, clinical performance of the two bare metal stents does not support equivalence based on the data. The Bx Velocity Stent had higher rates of repeat revascularization at four years [23.6 vs 20%] and higher rates of ARC stent thrombosis [1.8 vs 1.4% at four years] than the Express Stent used in the TAXUS trials. The attempt to force the two different bare metal stents to be "equal" in the network meta-analysis distorts the estimate of how the Cypher Stent and the TAXUS Stent compare, and leads to a misleading conclusion.
The inferred comparisons have problems since, according to the authors, "Criteria for consistency of the network were satisfied for all outcomes except stent thrombosis and target lesion revascularization," indicating that the indirect inferences did not even agree with the comparatively small number of direct comparisons included in creating the "network."
Most importantly, if the findings of either the Schomig or Stettler meta-analyses were actually valid estimates of the “Truth”, one would expect their strong findings to be echoed in the broad collection of trials and large registries (including the large and carefully conducted DEScover, STENT, and NY State registries). It should give pause for thought that the findings of the meta-analyses are not echoed there, nor in the Kereiakes overview of all the available trials and real-world registries that include more than 55,000 patients. One has to question the allegations of superiority of either stent from the recent meta-analyses with their inherent limitations.
Meta-analyses are well known as “statistical alchemy” for their claims to extract informational gold from the base metal of individual trials of variable quality and for the troublesome fact that meta-analyses commonly come up with answers that are then disproved by subsequent larger trial experiences. Although it may be difficult to see through the parade of different study results and different ways of selecting or combining groups of studies, we remain secure looking at ALL of the available data in totality, indicating that the TAXUS and Cypher drug-eluting stents have entirely comparable patterns of both efficacy (preventing restenosis) and safety (incidence of subsequent death, myocardial infarction, or stent thrombosis) in real world use.
|
