The First DES Meta-Analysis: What do 14 studies and 6,675 total patients tell us about stent thrombosis?
What prompted the meta-analysis? It was actually done a while ago, because one of my fellows, Anthony Bavry, actually presented these data at the American Heart Association meeting in 2005. At the time, it didn’t really generate any interest and I think was viewed with a fair amount of skepticism. Of course, now the climate has changed a bit and other reports have come out suggesting that there might be an increased rate of late stent thrombosis. In that context, our data is viewed as more believable. In fact, I think that the skepticism has pretty much died out regarding the core findings.
What prompted the study was the fact that at that point in time, we were concerned there might be an excess rate of late stent thrombosis with DES versus BMS. There was also the work of Dr. Renu Vermani, who reported that DES might have less full coverage of endothelial cells as compared to BMS, which in theory could predispose to clotting. In our own clinical practice, any number of my colleagues and I had seen some patients presenting with a thrombotic event what seemed to be rather a long time after their DES implantation, more so than we recalled seeing with BMS.
What we found was a small but real excess hazard, about 0.5%, or in more concrete terms, that about 1 patient in 200 who got a DES instead of a BMS may have suffered a late thrombotic event. Some of the more recent studies and analyses presented have also come up with a similar number as far as magnitude.
The Basel Stent Cost-effectiveness Trial—Late Thrombotic Events (BASKET-LATE) study was published in the December 2006 Journal of the American College of Cardiology and presented earlier that year at the ACC meeting. It actually showed that in the context of discontinuation of anti-platelet therapy, that there was a higher rate of death or myocardial infarction (MI) in DES patients. Even a large study doesn’t really address the mechanisms, but BASKET-LATE suggested that perhaps the reason for any sort of excess late thrombotic event is the discontinuation of anti-platelet therapy. A number of analyses, registry analyses and others, have shown that discontinuation of anti-platelet therapy is the most powerful predictor of stent thrombosis. Perhaps the solution is extending the duration of anti-platelet therapy, although that approach might raise individual bleeding risk. It certainly would increase cost, so in the absence of a large-scale randomized clinical trial supporting that approach, we don’t really know.
Data from the European Society of Cardiology meeting in early September 2006 suggested that DES increase death and MI, and with the sirolimus stent, data showed an increase in non-cardiac death. When all of this data was re-evaluated on a patient level analysis at TCT, however, none of these findings held true. As a result, do you think it is now time to start reporting meta-analysis data on an individual patient level rather than a study level? There is never any downside to a patient meta-analysis when you have access to the per-patient data, but we only had access to data that were in the public domain at the time. Full access to the data is not always available in contemporary clinical trials. The majority of the time the data is held by the companies themselves, so for us, it was impossible to go for a per-patient meta-analysis.
Just to clarify, our particular analysis wasn’t reporting an excess of death or heart attack, it was reporting an excess in late stent thrombosis. Indeed, in the randomized clinical trials such as the ones you mentioned that were presented at TCT, it’s not clear there’s an excess of death or MI in that clinical trial population. Where there has been an excess of death or MI in DES patients noted was in the BASKET-LATE study.
It seems intuitive to extend ASA + clopidogrel for longer durations. However, at TCT, Antonio Colombo presented data that suggests that there was no benefit to ASA + clopidogrel beyond 6 months. Since the last double-blind, placebo-controlled randomized trial to look specifically at this issue, Clopidogrel for the Reduction of Events During Observation (CREDO) trial, was only out to one year, what do we do? A year of clopidogrel use was proven with BMS, so there is no reason not to go at least a year with DES, in the absence of bleeding concerns. Obviously the combination of aspirin plus clopidogrel raises bleeding risk. In a patient that is at high risk of bleeding, the risk may outweigh the benefit with respect to reducing late stent thrombosis. Assuming that the patient is at low bleeding risk as best you can tell from their history, I would have recommended going a year just on the basis of the CREDO data. Now, based on all the reports of an excess rate of late stent thrombosis, such as our study and at least some reports showing a high rate of death or MI after clopidogrel discontinuation in DES patients, I’d be inclined to extend the duration of therapy to at least two years, though many times we say indefinitely, pending further data. Again, with the confirmation to the extent possible that the patient appears to be at low bleeding risk. If the patient has been on aspirin and clopidogrel for a year without a bleed, in a sense they’ve passed a bleeding stress test and I would feel more confident continuing them on with their aspirin and clopidogrel. On the other hand, if they didn’t do so well that year and had a number of bleeding problems, even if it was just frequent nuisance bleeding, then I might weigh the risk of that individual patient. If it were an unprotected left main, the nuisance bleeding might be acceptable, and I would tell the patient, if you have the occasional nosebleed, please try and put up with it. On the other hand, if it was a distal RCA or branch vessel with a DES, if the patient was having frequent nosebleeds, it might be best not to continue in the absence of data.
What are really needed though, are data. Short of doing a large, randomized clinical trial of various durations of clopidogrel after DES (one-year versus multi-year therapy), we won’t have an answer to the question you posed, other than different physicians’ gut feeling. My gut feeling would be to extend therapy if a patient was at low bleeding risk and could afford it.
Has the time come for clopidogrel responsiveness testing? That’s a challenging question, again, because here we are not really talking about people that are on clopidogrel that are having problems with their stent thrombosing. This question would refer more to patients who aren’t on clopidogrel and in some cases, aren’t on clopidogrel or aspirin, and are having these events occur. It might be useful, but is probably a smaller part of the pie. Responsiveness testing is potentially useful for the occasional patient that will be on both medications and have a late thrombotic event. It is an unusual thing to happen after a year; there are reports of it happening, but I think that if variability of response to clopidogrel is an issue, it’s more likely to be an issue with an acute patient coming in with a high thrombotic load, and perhaps less of an issue a year and a half after a DES is placed. Still, responsiveness testing might be useful if a patient has persistently high platelet reactivity. One might consider then whether a DES is the best way to go. In theory, that information could be very useful. So I think there are a lot of good scientific questions, but I hesitate to recommend anything clinically, because now we are straying rather far from evidence-based medicine.
At the current rate of stent thrombosis, ranging from 0.4–0.6% year over year, do you feel this trend will continue indefinitely, or will it begin to plateau at some point? That is the million-dollar question. I don’t know the answer. Right now there is no reason to think that risk stops after a period of time, which is a bit disconcerting. Renu Vermani’s work, in patients that have undergone autopsy, showed that the rates of endothelialization were lower for DES than BMS. It seems to me that the risk persists, but we do not have enough long-term follow up of patients to know. I can say, with reasonable certainty, at least out to a couple of years — but for the rest of your life? Right now, I would say maybe, and if you forced me to commit to an answer, I would say probably.
Did data show any difference in vessel size? I’m thinking of the original ‘role’ of DES as being for those with 3.25 or less vessels (not a strict guideline, but perceived as such by many labs) before the larger commitment to use most vessels. It would be interesting to see if late thrombosis occurred in larger vessels. It seems the small vessel benefit is still significant over risk (especially in diabetic patients). We didn’t look at vessel size, but that’s a good question. There are technical factors that can predispose to stent thrombosis as well; for example, if a stent is under-deployed and not well-opposed, which can occur especially in arteries with bifurcation lesions, bifurcation stenting and long segments of overlapping stents. There are certainly technical aspects, either with the implantation or with patient characteristics such as renal failure, that other studies have shown predispose to stent thrombosis with bare-metal stents, but also in the contemporary era with DES. We simply didn’t have that in our analysis.
Could you discuss how the paclitaxel and sirolimus stents each fared in the study? We looked at both stents as a class. I realize it may not always be ideal to look at drugs or devices as a class, because individual products can have product-specific differences, either good or bad, but that was our intent. Anything beyond that is really more speculative or tenuous. Our findings and the excess hazard rate of about 1 in 200 would apply to both stents.
The other interesting thing about our analysis was the time course of the late stent thrombosis, which was about three to four months on average with a BMS, and on average, with a DES, about 18 months. That is given the limited period of follow-up of the fourteen studies, which was about a couple of years. The time course with DES is much more protracted, it seems, and is something of which physicians should be aware. It is not that the rates are so high, but rather it’s the time course that is dissimilar from BMS. If physicians know that, it might affect the way they manage a DES patient. If a year or year and a half after their procedure, the patient is going to need surgery, and there’s some need to discontinue anti-platelet therapy, at least the physician can make the decision to do so in a more informed way, understanding that there might be some additional risk to the DES patient.
The protracted time course was apparent for both the drug-eluting stents in our analysis, paclitaxel and sirolimus. I wouldn’t use our paper to try and differentiate the DES in terms of particular hazards. I happen to believe it exists with both. As far as the actual event rate, the p-value wasn’t significant with the sirolimus, but it was with the paclitaxel. Yet the lack of a significant p-value associated with the sirolimus-eluting stent does not mean that it doesn’t have this late risk and only paclitaxel does. I wouldn’t over-interpret the p-value. Again, it wasn’t our intent to compare the two stents, but to look at them as a class.
Behind this meta-analysis are issues in the individual trials about duration of clopidogrel therapy and premature discontinuation, and whether or not the stents were optimally deployed with higher pressure (20ATM) than labeled. We also have the issue of whether post deployment treatment with non-compliant balloons was utilized to optimize stent-vessel wall apposition, as well as the degree to which stented patients continued to smoke or work/live in high second-hand smoke environments. Finally, the role of ASA and clopidogrel resistance and degree of intrinsic inflammation in the individual patient suffering stent thrombosis remain unmeasured. All these issues cannot be corrected for in a meta-analysis of trials of patients with stents. I would agree that stent thrombosis is multi-factorial. All those things mentioned could in part have contributed. The fact still remains that the rates observed were higher with DES than with BMS. Those differences, in a study that was almost 7000 patients, would likely even themselves out. Yes, there were smokers, but the numbers were pretty similar between the two groups. Yes, there might have been issues with a particular stent having been underdeployed or not post-dilated, but again, in the almost 7000 patients we studied, one would expect all those factors to even out. We didn’t measure all those factors, but you would expect them to be almost randomly distributed, in a 50% form, between the two arms of the study. I don’t think considering these factors would invalidate our findings, but they could certainly predispose to stent thrombosis in an individual patient. When the numbers of actual events are low, such as with stent thrombosis, I would concede that even in a 7000-patient study that these factors may have affected events. Yet our purpose wasn’t to be alarmist or to say that DES are bad or shouldn’t be used. Rather it was just to raise a note of caution when DES are being used, because stent thrombosis is a hazard even in patients that are perhaps lower-risk than what is being used in real-life, since there is a lot of off-label use of DES in the U.S. This hazard exists and we felt reporting it was the right thing to do. The exact degree of hazard can be contested. I think at this point in time, there are other analyses that have come out or will be coming out that are supportive of this concept. You can argue the exact magnitude of the problem, but I think to dismiss it at this point would not be the correct approach.
Dr. Bhatt can be contacted at bhattd (at) ccf. org
