Author + information
- Daniel B. Mark, MD, MPH, FACC⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Daniel B. Mark, Director, Outcomes Research, Duke Clinical Research Institute, P.O. Box 17969, Durham, North Carolina 27715.
When drug-eluting stents (DES) were introduced clinically in 2003 as the long-sought cure for restenosis, the Center for Medicare and Medicaid Services (CMS) was so persuaded that these devices were a breakthrough technology that it took the almost unprecedented step of increasing reimbursement to account for their substantially greater cost. Debating the appropriate use of this technology, interventional cardiologists in the U.S. worried that they might be judged guilty of medical malpractice if they failed to provide DES wherever possible in patients undergoing percutaneous intervention (PCI). Within a few years, DES were being routinely used in over 80% of all PCIs across the U.S. and in several European countries.
For a brief halcyon period, it seemed that if cost were no object, all PCI patients should be given a drug-eluting rather than a bare-metal stent (BMS) wherever technically feasible. Then, in September 2006 at the World Congress of Cardiology meeting in Barcelona, the winds began to shift. Several groups presented data suggesting that DES were associated with a small post–1-year excess risk of stent thrombosis relative to BMS. Particularly concerning was the suggestion that relative to other types of acute coronary syndromes, DES thrombosis events seemed to be associated with exceptionally high rates of mortality and morbidity. The press sounded the alarm, concerned patients called their doctors asking whether they should have their stents removed, and the Food and Drug Administration convened a panel of experts to review the matter (1). Over 2 days in December 2006, this panel reviewed evidence from a wide variety of sources and heard the views of many experts. In the end, the panel reached the not-surprising conclusions that there was some reason for concern but that the evidence was far from clear and (of course) more evidence was needed (2). Cardiologists, who in the meantime were faced with the pragmatic problem of caring for patients with coronary artery disease, were left wondering how to navigate the choppy waters in which they now found themselves.
In this issue of the Journal, Garg et al. (3) present a decision-model–based quantitative analysis to help clarify the health outcome trade-offs between BMS with higher risk of target vessel repeat revascularization and DES with possible increased risk of very late stent thrombosis. Of the elements required for this modeling exercise on the BMS side, the least certain is the negative health outcome weight that should be assigned to repeat revascularization procedures performed for restenosis after stenting. To compare 2 treatment strategies in terms of net expected quality-adjusted life years (QALYs), the impact of restenosis must be expressed in terms either of life expectancy or quality of life lost. The standard utility theory approach requires determining how much of remaining life expectancy patients would be willing to forgo to avoid restenosis with repeat revascularization. For this parameter, Garg et al. (3) referred to previous work in the Stent-PAMI (Stent-Primary Angioplasty in Myocardial Infarction) trial where they found that patients without repeat revascularization had a 0.06 absolute advantage in quality adjusted life expectancy over patients with repeat revascularization (4). Using these data in the current model, Garg et al. (3) are in essence assuming that patients would be willing to trade 22 days of life in excellent health to avoid a repeat revascularization procedure. Although this estimate might be reasonable, it is important to recognize that it does not reflect the judgments of patients directly grappling with this difficult question.
The most critical and elusive aspect of this decision problem is the true risk of very late stent thrombosis with current generation DES (5). Detecting a presumed tiny absolute difference of 0.1% to 0.5% annually with precision would require over 10,000 patients followed for 3 or more years. Detecting this difference with shorter follow-up would raise the sample size requirements. Compounding these signal detection difficulties are problems in determining whether a late cardiac event in a coronary artery disease patient with prior PCI is due to stent thrombosis. If the patient dies before invasive study and no postmortem examination is performed or is not referred for invasive study, there is no way to be certain that the DES was involved in the acute ischemic event.
For these and other reasons, we cannot now determine whether there is a small but important excess risk of very late stent thrombosis with DES. To assess how this possibility can and should affect our decisions, we must use modeling as Garg et al. (3) have done. They begin by considering the choice between DES and BMS assuming that there is no late excess risk of stent thrombosis. In this situation, DES improves clinical outcomes through fewer repeat revascularization procedures, thereby providing a modest improvement in quality-adjusted survival over BMS. Disregarding cost, if lowering the probability of repeat revascularization is of some value to the well being of patients, the decision to routinely use DES is not controversial.
In the second pivotal case considered by Garg et al. (3), DES are assumed to have an excess risk of stent thrombosis after the first year of 0.13%/year for at least the next 3 years. In that scenario, the QALY benefits of the DES noted in the preceding text are cancelled out by the morbidity and mortality of these late adverse events so that the net difference in favor of the DES is 0.001 QALYs or approximately 9 extra hours of life in excellent health. If the excess risk of stent thrombosis is increased over 0.14% from 0.13%/year, the BMS strategy now provides the higher net yield of QALYs and becomes the preferred treatment.
Two important parameters were identified in sensitivity analysis as key determinants of the outcome of this model: the risks of restenosis with BMS and the length of time over which DES might continue to have an excess risk of stent thrombosis. According to this model, one should be willing to accept higher levels of risk of very late stent thrombosis for patients with higher risks of restenosis. Conversely, the longer we believe the elevated risk of very late stent thrombosis persists, the more we should favor the use of BMS. Importantly, the authors do not consider the possibility that extended use of clopidogrel, in addition to aspirin, beyond the first year after PCI might protect against this excess stent thrombosis risk at the cost of extra major bleeds (leaving aside the monetary cost of the therapy) (6).
The primary conclusion of this analysis can be restated as follows: the choice between a DES with a slight excess risk of stent thrombosis after the first year and a BMS with a much higher probability of restenosis requiring repeat revascularization during the first year is a toss-up (i.e., a situation in which the consequences of the 2 choices are so close as to be indistinguishable) (7). However, 1 further characteristic of a toss-up is that the major clinical outcomes are of equivalent value. Whether this condition is met in the present analysis depends on whether one accepts the trade of extra nonfatal revascularizations for a much smaller risk of DES thrombosis that could result in disability or death. The decision model by Garg et al. (3) assumes that patients are risk neutral with regard to this trade. In other words, they will choose the outcome with the higher net QALYs value regardless of how those QALYs are produced. Because very late stent thrombosis events have not translated into consistent elevations in overall mortality in DES patients relative to BMS patients, some interventionalists seem quite comfortable with this trade (5,8). As noted previously, however, studies reported to date have insufficient statistical power to deal with such small event rates. In such cases, the absence of evidence should not be confused with evidence of absence (9).
Although interventionalists might accept a trade between restenosis and very late stent thrombosis, patients are not so clearly risk neutral. They might be willing to accept some risk to avoid serious adverse health events such as death or disabling stroke, but things are less clear regarding the willingness to accept risk to avoid transient turbulence in their health that does not result in death or long-term disability. Indeed, Garg et al. (3) showed in sensitivity analysis that patients who are not willing to trade some of their remaining life expectancy to avoid repeat revascularization would have a much lower willingness to accept the late risks of DES.
Obviously, we need more data about the magnitude of late adverse events with DES. However, owing to the huge number of patients required and the difficulties of case finding in such studies, we cannot expect that this question will be easily or quickly settled. How then should these insights inform our clinical practice? Reserving DES for patients with the highest risk of restenosis is 1 reasonable strategy that improves the likely margin of benefit in QALYs for these devices. Since the reports at the World Congress of Cardiology in 2006, the use of DES has fallen significantly. Whether DES are in fact being selectively directed at the patients with the highest risk of restenosis requires empirical verification. Some clinicians have also chosen to assume that indefinite use of clopidogrel will protect against very late stent thrombosis. However, we lack clinical trial validation of this hypothesis as well as data on the long-term risk/benefit ratio and on the cost-effectiveness of such therapy (10).
There is also the issue of what patients want. Do they feel as passionately as interventional cardiologists about avoiding clinical restenosis? Are they really able to appreciate incremental risks on the order of magnitude of 2 to 5 of 1,000? How does the prospect of years of prophylactic clopidogrel therapy affect their preferences? Given the tremendous susceptibility of patients' decisions to framing bias, can we even provide a fair exposition of the problem and its attendant uncertainties and expect them to process the information and make a treatment selection on that basis? The analysis of Garg et al. (3) does reassure us that, for a risk-neutral individual who is willing to trade a small amount of life expectancy for a reduced risk of repeat target lesion revascularization, the 2 choices produce almost the same net clinical benefit, so either should be satisfactory. However, the issue of very long-term clopidogrel therapy in this decision problem does require further consideration.
Finally, the next generation drug–device combinations, already in use in Europe and now being introduced in the U.S., might provide different safety and efficacy profiles. Hopefully, the lessons of the last few years will teach us what data we require about these new devices that will allow us to shed the albatross of very late stent thrombosis and return to the halcyon days where cost was the only concern.
The author is grateful for the editorial assistance of Ms. Melanie Daniels and the critical suggestions of Drs. Manesh Patel and David Kong.
Dr. Mark has received >$10,000 in grant funding and <$10,000 in consulting fees from Medtronic.
↵⁎ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
- American College of Cardiology Foundation
- Food and Drug Administration
- Garg P.,
- Cohen D.J.,
- Gaziano T.,
- Mauri L.
- Cohen D.J.,
- Taira D.A.,
- Berezin R.H.,
- et al.,
- Stent-PAMI Investigators
- Kaul S.,
- Shah P.K.,
- Diamond G.A.
- Pfisterer M.,
- Brunner-La Rocca H.P.,
- Buser P.T.,
- et al.