The Enigma of Drug-Eluting Stents: Title and subTitle BreakHope, Hype, Humility, and Advancing Patient Care

Interventional cardiology is an iterative science. Advances come in spurts as interventionalists gradually become increasingly familiar with its technologies and limitations. This iterative learning process in the use of percutaneous coronary intervention (PCI) has been tremendous over the last 3 decades. While adverse events were frequent and restenosis was common in the first decade, PCI procedures did relieve symptoms for patients with coronary disease. In the second decade of PCI use, new techniques were developed to scrape, burn, or excise plaque, in an attempt to achieve better outcomes. While there was much hope and hype with these approaches, they all had fundamental problems that prevented their broader use.

Intracoronary bare-metal stents humbled clinicians because warfarin was initially required to prevent in-stent thrombosis and resulted in bleeding complications in many patients. Advances in delivery of bare-metal stents, application of high-pressure balloon inflation, and combination therapy of aspirin and clopidogrel set the stage for broader and safer use of stents, although restenosis remained a problem. The third decade of PCI use was ushered in with great hope and hype, as drug-eluting stents became approved and used regularly. The irrational exuberance was partially driven by the hope that cardiologists could offer safe technical procedures with a low rate of restenosis. However, humility soon set in as the first reports of unexpected late stent thrombosis, myocardial infarction, and even death occurred more than 1 year after the initial procedures involving drug-eluting stents.¹ Could these problems have been anticipated? And what can be expected now, as PCI moves into a more mature stage of development?

While iterative and exploratory science is not suitable for development of clinical guidelines, mature science always is. The highest-quality evidence is attributed to multiple randomized controlled trials (RCTs)—or a single large RCT—that have sufficient size (ie, provide adequate power) to observe or exclude meaningful treatment effects, are conducted in representative patient populations, have relevant clinical end points, and are analyzed appropriately. Such trials are best performed when the science is mature, so that the intervention is stable enough to be tested in large cohorts to detect events of rare frequency (usually less than 10%—as now applies for all end points in PCI). By their very design, RCTs are limited because they provide only a snapshot of a disease and a population and because they frequently exclude the sickest patients. Observational studies complement RCTs and are especially valuable for examining the uptake of a new therapy into broader community practice and including groups that have not been adequately studied in trials. These types of studies can provide valuable practice information regarding quality of care, unusual adverse effects, uptake beyond the populations studied, and comparative effectiveness.²

In the best situations, the approval of a device and subsequent guideline recommendations for its use follow a similar path. Challenges arise when uptake of a device into clinical practice moves faster than the evidence accumulates regarding its benefits and harms or when unexpected events occur after device implantation. The approval of drug-eluting stents was rapid, and the major trials leading to approval were conducted in the setting of low-risk clinically stable patients with low lesion complexity. However, the uptake of drug-eluting stents into clinical practice quickly moved beyond the evidence on which approval had been based and included high-risk groups of patients that had not been studied in adequately sized clinical trials.^{3 - 4} For instance, in the CRUSADE registry,³ the average use of drug-eluting stents was 50% within 6 months of approval and increased to 82% within 1 year; however, this expanded use was mostly in high-risk patients. Thus, much of what was learned in the last few years regarding late complications of drug-eluting stent use has been in a population for which there is a paucity of RCT data. Moreover, most of the assessment of efficacy and risk is derived from observational data, which, while important, have inherent flaws that allow the introduction of bias that cannot easily be accounted for.

Within 6 months of US approval and 2 years after European Union approval of drug-eluting stents, McFadden et al¹ reported that 4 patients experienced late unexpected stent thrombosis. This report highlighted the ability of an experienced interventional cardiologist to discern a pattern of rare events that now is suspected to occur in less than 1% of all drug-eluting stent procedures.⁵ Following subsequent reports of stent thrombosis, the US Food and Drug Administration (FDA) issued a warning,⁶ and investigators around the world began reporting both on the clinical follow-up of a spectrum of patients as well as on some of the pathobiology behind this potential problem.^{7 - 8} Because of safety concerns, the FDA convened an expert advisory panel in December 2006, resulting in an updated statement.^{9 - 10} Whereas the FDA statement noted that drug-eluting stents are associated with a small increase in stent thrombosis compared with bare-metal stents, “the concerns about thrombosis do not outweigh the benefits of [drug-eluting stents].”⁹ In addition, the panel agreed that due to safety concerns the drug-eluting stent labels should state that when such stents are used off-label, patient outcomes may not be the same as the results observed in the RCTs used for approval.

Recently, investigators assessing the use of drug-eluting stents have focused on the long-term experience from the original approval-based RCTs^{11 - 15} or from a variety of both industry- and independently-supported observational databases.^{16 - 18} A key observation has been that the late stent thrombosis issue might be attenuated through the use of long-term dual antiplatelet therapy.^{19 - 21}

In this issue of JAMA, the reports by Win et al²² and by Beohar et al²³ represent the 2 latest additions to this ongoing controversy. Both groups, using observational registries, report on the indications and outcomes associated with use of drug-eluting stents in US clinical practice between 2004 and 2005, when use of such stents was high. Both databases allow classification of drug-eluting stent use as “on-label” vs “off-label,” relying on patient and angiographic indications that had been included in the approval-based pivotal RCTs, and both studies report 1-year clinical follow-up and in-hospital outcomes.

What do these new studies add? First, they represent a major shift in clinical research over the last few years, from a few well-established observational databases to multiple registry-based studies that take advantage of improvements in information technology that now allows easier collection of clinical information, aggregation and analysis of those data, and rapid reporting. However, these studies also highlight a major flaw in the US system of device safety surveillance—ie, there is no reliable system that allows systematic follow-up and subsequent public reporting for all implanted devices. The current system depends on isolated registries, frequently supported by industry, and on voluntary reporting by vigilant experienced individual clinicians.

These 2 studies report a very high frequency of off-label use of drug-eluting stents. In the study by Win et al,²² 55% of patients had at least 1 off-label characteristic, whereas in the study by Beohar et al,²³ 47% of patients were categorized as receiving a drug-eluting stent for either an off-label or untested indication. This is consistent with reports on the uptake of drug-eluting stent use, indicating that the vast majority of patients undergoing PCI in the United States during this period received a drug-eluting stent, despite estimates that the RCTs studied a very select minority of low-risk patients who undergo coronary stenting. Patients with complex angiographic disease (including chronic total occlusions, vein graft disease, and bifurcation lesions) were not included in the RCTs, nor were those with multivessel coronary artery disease or urgent acute coronary syndromes. These patients are all represented in the 2 registries. Both studies indicate higher acute and late ischemic complication rates among patients receiving drug-eluting stents for off-label compared with on-label indications.

The rates of early complications reported by Win et al,²² especially myocardial infarction, among the patients receiving drug-eluting stents for off-label indications are much (9-10 times) higher than those reported by Beohar et al.²³ This finding may be due to ascertainment bias; Win et al²² described a careful plan to obtain the periprocedural biomarkers, while Beohar et al²³ did not. This also may reflect a sponsor bias; in the study by Win et al,²² the registry was funded by the manufacturers of antiplatelet agents, whereas the study by Beohar et al²³ was funded by a device manufacturer. This difference in event rates for registries conducted in essentially the same types of hospitals at a similar time point highlights the need for large national registry programs that use common data standards and agreed-on collection variables. Otherwise, data cannot be compared across registries, and the power of having larger sample sizes available for analysis is diminished.

Additionally, the marked difference in event rates likely is one reason the 2 groups of investigators arrived at different final conclusions, despite both studies showing similar findings. Win et al²² note that ischemic event rates are very high among the patients receiving drug-eluting stents for off-label indications and that clinicians should be cautious in extrapolating results of the on-label RCTs to higher-risk patients. In contrast, Beohar et al²³ acknowledge that there is higher risk associated with off-label stent use but that the overall low absolute event rates should be comforting. Event rates, especially those for myocardial infarction, will be lower if biomarkers are not systematically obtained in these studies.

The studies by Win et al²² and by Beohar et al²³ provide 1-year follow-up information that is of great value. Stent thrombosis occurred more commonly among the patients receiving drug-eluting stents for off-label indications, and this finding remains after adjustment for dual antiplatelet therapy, suggesting an independent risk contribution of receiving a drug-eluting stent for an off-label indication. Clinicians need to acknowledge that few data are available to guide the duration of clopidogrel use when the studies of drug-eluting stents were performed. The 2001 PCI guidelines²⁴ also were not entirely clear on the optimal duration of clopidogrel use. However, for the highest-risk patients with acute coronary syndromes (those not included in the RCTs assessing drug-eluting stents), 1 year of clopidogrel use was recommended, but this recommendation generally was not followed by the clinical community. Having complete long-term follow-up, including information on medication compliance, is critical to being able to complete these analyses and look for clinically significant associations. There needs to be a nationally agreed-on mechanism for linking long-term outcomes to these existing databases to allow better and more complete device safety surveillance. A system such as the one in use in Sweden¹⁷ provides long-term tracking of cardiovascular outcomes among PCI-treated patients and allows clinicians, investigators, and policy makers to adjust practice recommendations rapidly as more data accrue in large patient samples.

Drug-eluting stenting is clearly a breakthrough technology that provides hope that coronary artery restenosis can be reduced if not eliminated. But clinicians need to be cautious in too rapidly adopting a technology until more evidence accumulates and must be careful to not succumb to the hope and hype of a product based on limited approval data. The community of clinicians, investigators, regulators, manufacturers, journalists, and policy makers must remain humbled by the notion that despite the belief that it is possible to truly understand the complex biology of coronary disease and that data can be extrapolated easily across patient subsets, high-quality evidence, particularly robust RCT data, is needed to guide practice decisions. On the other hand, the public needs to recognize that medicine remains an iterative and adaptive science and that physicians continue to learn as technology matures.

Clinicians must take advantage of all the information sources available (RCTs, observational studies, and clinical vigilance) in making clinical decisions. Moreover, it is essential to harness the power of this disparate information by working toward common data standards, efficient methods of data aggregation, and systems that allow rapid performance of clinical trials that can be integrated into clinical practice and that can provide a true reflection of the patients receiving these sometimes enigmatically evolving interventions.