Analysis Slams Use of Clopidogrel Loss-of-Function Gene Test

December 29, 2011

December 29, 2011 (London, United Kingdom) — A new meta-analysis has concluded that although there is an association between clopidogrel loss-of-function genotypes (CYP2C19 alleles) and responsiveness to the drug, there is no significant association of the genotype with CV events [1]. Dr Michael V Holmes (University College London, UK) and colleagues report their findings in the December 28, 2011 issue of JAMA.

For example, in the clopidogrel-based studies included in their meta-analysis, "there was strong evidence of small-study bias. We also identified evidence for selective outcome reporting bias," the group writes. "Taken together, these observations cast doubt on the association of CYP2C19 genotype with clinical cardiovascular end points."

The study provides no evidence to support CYP2C19 genotype testing, and therefore, physicians should use such tests "rarely, if ever, and interpret the results with caution," says Dr Steven E Nissen (Cleveland Clinic, OH) in an accompanying editorial [2].

But other doctors approached by heartwire are up in arms about the meta-analysis, saying it is misleading and contains numerous flaws. At the heart of their argument is the fact that they maintain that clopidogrel is most efficacious in preventing events in patients receiving stents. However, this review looks at clopidogrel use in many different settings and uses CV outcomes that are too broad, they argue.

Moreover, when the authors drill down to individual end points, they do find a significant increase in risk of myocardial infarction and stent thrombosis in patients with the loss-of-function genotype, the critics state.

While Holmes and a coauthor, Dr Aroon D Hingorani (University College London, UK), told heartwire --and clearly state in their report--that stent thrombosis is "a possible exception," to their overall conclusion, "we don't have enough data to be able to conclusively say that the association with stent thrombosis is robust," they maintain.

Broad vs Narrow Analysis Debated

Dr Marc Sabatine (Brigham and Women's Hospital, Boston, MA), who has authored a previous meta-analysis on the subject, told heartwire that Holmes et al, "started with a laudable goal, to update a meta-analysis, but sadly they have done the field a bit of a disservice by not more carefully selecting which studies, in which patients, and which outcomes, to put together."

Genotype is important information to use to make decisions, but only in stented patients.

They included studies "where patients were managed conservatively--where there hasn't really been any debate about the importance of the CYP2C19 variant. Moreover, they include other studies of clopidogrel, such as those [in patients] with stable coronary disease, or patients with atrial fibrillation who happen to get clopidogrel, where there is a modest, if any, benefit of clopidogrel. So in that sense, it's really somewhat silly to say that they've found that genetic variants that alter clopidogrel metabolism weren't important in those settings. Well of course, because clopidogrel isn't important in those settings," he adds.

Dr Gilles Montalescot (Hôpital Pitié-Salpêtrière, Paris, France) agrees: "It does not make sense. Some of these studies were not even performed in coronary patients, and some were performed in coronary patients but without stenting, and this is the real miscue, because we know from all the data we have now that the genotype thing is important in coronary patients when they get stented. Elsewhere, it's probably of little value. I feel uneasy with the conclusion, and I feel uneasy with the editorial."

"If you look at all meta-analyses of stented patients, they are clearly in favor of a genetic impact. Genotype is important information to use to make decisions, but only in stented patients."

Pharmacogenetic analyses require a strong understanding of statistics, genetics, and clinical cardiology. If one doesn't have all those elements, it can be easy to fall into traps.

Sabatine adds: "These pharmacogenetic analyses require a strong understanding not only of statistics, but of genetics and clinical cardiology. If one doesn't have all those elements, it can be easy to fall into traps and make mistakes."

But Holmes and Hingorani, who are part of a genetics epidemiology group, but are also both physicians--with Hingorani holding qualifications in clinical pharmacology and internal medicine--say they stand by their review.

"We are interested in how genomic advances might be translated into clinical practice. We want to ensure that genetic tests are [held] to similar standards as other healthcare technologies, so it was our aim to conduct a critical appraisal of this particular test because it seems to be one that is closest to translation," Hingorani notes.

It was our aim to conduct a critical appraisal of this particular test, because it seems to be one that is closest to translation.

"We wanted to be as broad as possible, and the licensed indications for clopidogrel are wide-ranging," he added, in defense of the decision to include all of the studies that were incorporated. The range of trials is also "similar"--although more have been added--to those in another review, published in BMJ this summer [3], which came to an analogous conclusion: that the current evidence does not support the use of individualized antiplatelet regimens guided by CYP2C19 genotype, he argues.

Holmes also notes that the US FDA guidance related to this subject, issued 18 months ago, is not limited to only patients who undergo stenting, "so in the setting of the guidance, the broader analysis is more applicable."

Too Many Soft End Points?

In the meta-analysis, Holmes et al examined 32 studies that evaluated CYP2C19 genotyping as a predictor of response to clopidogrel therapy, including 42 016 patients reporting 3545 CVD events, 579 stent thromboses, and 1413 bleeding events. The analysis is unique and differs from previous ones in that studies were considered in two separate groups, Hingorani explains.

Firstly, they took all those studies in which all patients were exposed to clopidogrel--26 in total--termed "treatment only" trials, and looked at the association of genotype with outcome. Individuals with one or more CYP2C19 alleles had lower levels of active clopidogrel metabolites, less platelet inhibition, lower risk of bleeding, and higher risk of CVD events (relative risk 1.18).

But there was evidence of publication bias among these studies, the authors say, whereby larger associations were reported among smaller studies (those with <99 events had a point estimate for hazard ratio for CV morbidity and mortality of 1.83). When analyses were restricted to studies with 200 or more events, the point estimate was 0.97, indicating no evidence of a less favorable outcome in slower metabolizers.

Nissen observes that "their careful attention to the issue of publication bias allowed the authors to more accurately assess the clinical relevance of slow metabolism of clopidogrel. The results are compelling."

The other part of the analysis centers on six randomized clinical trials in which subgroup analyses were used; patients had received clopidogrel or placebo, and had at least one CYP2C19 loss-of-function variant or not. This is "the most robust approach," noted Hingorani, and in these trials, the CYP2C19 genotype was not associated with the modification of the effect of clopidogrel on CVD end points or bleeding.

Sabatine points out that, while the subject of publication bias is well established, it is not the key issue here, and that a slightly biased result in the small studies would, by definition, have a small impact on the pooled results.

As well as the choice of trials to include, another vital component is what was included in the CV end points reviewed, he notes, with "soft" outcomes, such as rehospitalization for acute coronary syndrome and target vessel revascularization, incorporated.

"You have to look at outcomes that clopidogrel is known to reduce," says Sabatine. "It's been shown to reduce CV death, myocardial infarction, stroke, and particularly, stent thrombosis. In this meta-analysis, they lump together a far broader range of events, and include what we would describe as softer CV events; but clopidogrel doesn't affect those events, so obviously, genetic variants affecting clopidogrel will have no impact."

They include what we would describe as softer CV events; but clopidogrel doesn't affect those events, so obviously, genetic variants affecting clopidogrel will have no impact.

Holmes responds: "We tried to report the totality of evidence, and some studies only reported composite end points, which included these individual components--some of which are 'soft'--so when we pool the data together, this is what we are limited to." But he adds that they did a sensitivity analysis, where they excluded each study individually to see if this had an overall bearing on the summary estimates, and they did not see any evidence that suggested this was the case for any of the studies.

"If you pick on one outcome of many, you don't get a full picture of the effect of the genotype on all relevant outcomes, and obviously all cardiovascular events are relevant in this context," Hingorani adds.

"The Data Don't Fit With Their Conclusions"

Hingorani notes that they also, "where possible, dissect out individual components and end points, and report those separately."

Sabatine seizes on this analysis to illustrate his point, noting that "when they start to drill down to more appropriate end points, we see close to a 40% reduction in fatal and non-fatal MI in those who have a loss-of-function variant, and a 1.5- to 2-fold increased risk of stent thrombosis in the same patients, and these are highly statistically significant, and thus, their data don't fit with their conclusions."

When they start to drill down to more appropriate end points, their data don't fit with their conclusions.

He also points out that three studies that only reported stent thrombosis as a primary outcome were excluded from the overall analysis. Holmes says this is true, because when they looked at the totality of evidence, they couldn't include studies that reported a single end point. But these three studies were included in the analysis of stent thrombosis as a separate outcome, he notes.

Both Holmes and Hingorani reiterate that stent thrombosis could be the one exception to their overall conclusion, but maintain that the data are not conclusive enough to state this unequivocally.

When pressed by heartwire as to whether they would agree that clopidogrel was more efficacious in the context of stenting than in all other clinical settings, Hingorani said, "I'd have to look more closely at the evidence, I couldn't tell you offhand."

"As clinicians, we all have to make decisions in conditions of uncertainty with the best available evidence," he added. "It was our aim to bring together the current best available evidence, in order that informed decisions can be made, and to update previous systematic reviews in this area, but also to compare and contrast the findings from treatment-only studies and trials."

"We really need to come to a consensus on what level of evidence we require to translate pharmacogenetic testing to clinical practice. It's a pretty fast-moving area, there are discoveries being made all the time, and we need to now prepare ourselves in relation to study design, analysis, reporting, and evaluation, so we can utilize that information most effectively."

Where Do We Go From Here?

The issue of identifying poor responders to clopidogrel--either through genetic testing or platelet function tests--has been a controversial topic since the US FDA put a boxed warning on the drug 18 months ago. This stated that pinpointing poor responders to clopidogrel could allow doctors to implement "alternative treatment strategies." At the time, the American Heart Association and American College of Cardiology issued a consensus statement that conflicted with the FDA's position, stating that the evidence base was "insufficient to recommend either routine genetic or platelet function testing at the present time."

In his editorial, Nissen says "it now appears the FDA warning reflected a case of 'irrational exuberance,'" and he calls for a large randomized controlled trial.

But Sabatine points out that the FDA warning "came out at a time when there wasn't easy access to genotyping, and they didn't give specific guidance as to what alternative therapy to give." The field has moved on rapidly since then, he argues, and while genotyping may be a "moot point" if people are using newer antiplatelet drugs for stenting, clopidogrel still remains the dominant agent used. And 25–30% of patients (depending on ethnicity) will harbor a genetic variant that affects clopidogrel function. When these patients receive a stent, the risk of CV events--most notably stent thrombosis--"is quite clear," he maintains.

He acknowledges, however, that many of those in favor of testing for CYP2C19 genotypes are "still grappling with whom to test and what to do about it. I think [the FDA] have appropriately highlighted an issue with this medication, and now it's up to us as clinicians to figure out what the reasonable options would be and what is the most effective strategy. We need to put the information together to develop a sensible algorithm."

Montalescot says a large randomized, controlled trial is impractical and "impossible to imagine." He notes that a recent risk model developed in Paris that incorporates two genetic risk factors--the CYP2C19 genotypes and ABCB1 genotypes--provides greater discriminatory power in defining stent-thrombosis risk compared with a model that incorporates clinical risk factors only.

He says around a quarter of patients receiving stents are being tested for clopidogrel loss-of-function variants in his hospital, using quick turnaround, point-of-care tests when requested by either the treating doctor or patient.

And although he acknowledges that stent thrombosis is less of an issue with newer drug-eluting stents than older ones, he says "it is still a concern for the patients with the wrong genotype."

If I would have to get a stent implanted, I would certainly like this information.

Those who are identified as carrying either the CYP2C19 genotypes or the ABCB1 genotypes are offered prasugrel, as ticagrelor is not yet available in France, "or we may double or triple the dose of clopidogrel," he notes.

Genetic testing of stented patients is also occurring at other institutions, such as Scripps Research Institute, La Jolla, CA, and Vanderbilt University, Nashville, TN, as highlighted in a videoblog on earlier this week.

"It's easy and it's inexpensive," Montalescot says, noting that in France, it costs around $90 per patient, and the patient does not pay. "If I would have to get a stent implanted, I would certainly like this information," he concluded.

Hingorani has provided free advice to colleagues at University College London on the design and analysis of pharmacogenetic studies, including an analysis funded, in part, by Celera. All other authors report no conflicts of interest. Nissen also reports no conflicts of interest. Sabatine has received research grant support from AstraZeneca, Bristol-Myers Squibb/Sanofi-Aventis, Daiichi Sankyo, Merck, and Nanosphere, and has provided consulting for Bristol-Myers Squibb and Sanofi-Aventis. Montalescot has received research grants to his institution or consulting/lecture fees from Abbott Vascular, AstraZeneca, Bayer, Biotronik, Boehringer-Ingelheim, Boston Scientific, Cleveland Clinic Foundation, Cardiovascular Research Foundation, Cordis, Daiichi Sankyo, Duke Cardiovascular Research Institute, Eli Lilly, Europa, Fédération Française de Cardiologie, Fondation de France, GlaxoSmithKline, ICM, INSERM, Lead-up, Medtronic, Menarini, Nanospheres, Novartis, Pfizer, Sanofi-Aventis, Servier, Société Française de Cardiologie, Stentys, The Medicines Company, and the Thrombolysis in Myocardial Infarction (TIMI) group.