March 15th, 2011
GRAVITAS Editorial: Why Invite the Fox into the Henhouse?
Richard A. Lange, MD, MBA and L. David Hillis, MD
Results from the GRAVITAS randomized trial, presented at the 2010 American Heart Association conference, have just been published in JAMA. They show that high-dose clopidogrel did not improve outcomes after percutaneous coronary intervention (PCI) in patients with “high on-treatment platelet reactivity” (i.e., during clopidogrel therapy), compared with clopidogrel given at conventional doses.
Our interpretation at CardioExchange, reported at the time of the AHA conference, was that assessing platelet reactivity doesn’t effectively identify individuals who are at high risk for a cardiovascular event after PCI. Although 41% of patients were reported to have had high on-treatment platelet reactivity (according to the VerifyNow P2Y12 test), only 2.3% of them had a cardiovascular event in the 6 months after PCI.
In a comment on our post, Dr. Sanjay Kaul bolstered our explanation by noting that 25 of 1080 patients (2.3%) with and 8 of 586 (1.4%) without high on-treatment platelet reactivity had adverse cardiac events. He calculated these statistics:
- sensitivity: 76%
- specificity: 35%
- positive predictive value: 2%
- negative predictive value: 99%
- positive likelihood ratio: 1.17
- negative likelihood ratio: 0.69
- area under the curve: 0.59
In this week’s JAMA, the GRAVITAS article is accompanied by an editorial suggesting that personalized therapy with high-dose clopidogrel was not effective because the cutpoint for defining high on-treatment platelet reactivity (P2Y12 reaction units ≥230) may have been too high. On the contrary, multiple studies have shown that this cutpoint provides the best predictive value of post-PCI ischemic complications, according to receiver-operating-characteristic analysis, and identifies the top tertile of post-treatment reactivity.
Notably, the lead author of the editorial receives grant support from Accumetrics, manufacturer of the VerifyNow P2Y12 test, and is an advocate for its use. We commend the author for disclosing his potential conflict of interest but have difficulty understanding why JAMA invited a major proponent of this platelet-reactivity test to write the editorial. Rather than acknowledging that the test has a negligible impact on risk prediction, the editorial simply calls for different platelet-function cutpoints to be used.
Do you consider this to be a real or merely a perceived conflict of interest? Do you routinely look at the COI disclosures that accompany published studies and editorials?
11 Responses to “GRAVITAS Editorial: Why Invite the Fox into the Henhouse?”
NEJM Journal Watch — Recent Cardiology Articles- Transcatheter Valve Replacement for Severe Tricuspid Regurgitation
- Time in Blood Pressure Target Range and Outcomes in Hypertension: Real-World Data
- When to Start Anticoagulation After Ischemic Stroke in Patients with Atrial Fibrillation
- Aortic-Valve Replacement vs. Clinical Surveillance for Asymptomatic Severe Aortic Stenosis
- Secondary Hypertension in Young Adults
I fail to see how lower cut points for platelet function test will improve the poor to modest prognostic utility of VerifyNow assay for ischemic endpoints post-stenting. The tiny impact on MACE risk prediction is underscored by the statistics shown above. Using a lower cut point will reduce the already poor specificity further resulting in a lower discriminant capacity (lower AUC). Furthermore, how does lowering the cut point overcome the possibility of HPRA being a nonmodifiable variable?
The important lesson to learn is that HPRA is a correlate of poor outcomes post-stenting, not a “major cardiovascular risk factor” as claimed by the editorialists. Association should not be confused with causation!
There are 4 desirable attributes required for a screening test to be clinically useful:
1. The test should be practical in the clinical setting.
2. The test should help characterize low- and high-risk patients.
3. The test should identify at-risk individuals that should lead to a treatment that improves outcome.
4. The test should be cost-effective.
VerifyNow platelet function “point of care” assay fulfills only the first of these attributes.
Thus, the available evidence base is insufficient to recommend platelet function measurements in the care of patients with cardiovascular disease at the present time.
I agree with all above, except for AUC issue. This remains the same and represents the ability to discriminate between 2 random patients, one from the MACE group and one not. The” optimal” cutoff will vary depending on sensitivity and specificity trade-offs but is often defined as the test value corresponding to the point on the ROC curve furthest from the identity line. I think.
AS have said in a previous comment and published in an editorial in detail in JACC several years ago, The “Verify Now” test by Accumetrics is not a valid test because it detects only MacRO-platelet aggrerates and not MICRO- platelet aggregates,. This is easily shown buy doing a dose response curve to either Integrelin, Aggrestat or Repro.
Such a test will show a flat response above the critical level of 80% inhibition of ADP stimulation.Doing a simple eletronic platelet count with a coulter counter and calculating inhibition as the % of the count after ADP addition / the count before addition of ADP will show an accurate linear res;osnse. This is similar to the “Platelet Works” test. It is cheap, uses readily available equipment present in nearly every lab, and accurate.Routine use of such a test would reduce both bleeding and stent thrombosis. If the public ever realizes how simple the solution is to this problem, they will question the credibility of ACC and cardiologists for not using it before waiting for the arrival of a properly designed clinical trial. Some things are just common sense and don’t required clinical trials.
Demanding a clinical trial is just a convenient dodge and represents ” passing the buck of responsibility”. The p value is only one type of evidence for “evidence based medicine”. It doesn’t replace common sense.
B.A. Kottke MD,PhD. FACC
Emeritus Professor of Medicine
Mayo College of Medicine
I used to disregard VerifyNow until I had a few patients who presented either with stent thrombosis or TIAs while on dual antiplatelet therapy and the test showed 0% P2Y12 inhibition. These highly resistant patients do not respond to simply doubling the maintenance dose of clopidogrel. So, the result of GRAVITAS is no surprise. However, if I place a stent at a critical location I would like to know that my therapy is working and my patient is not on a 75mg placebo. The positive predictive value is low because in many cases a well expanded stent in a brisk flow environment does fine on aspirin alone. Most patients with a LDL of 140 also have a low positive predictive value for CV events but we still measure it and treat it. In this discussion I see a philosophical difference: the minimalists do not want to measure or do anything that does not have 100% accuracy or efficacy or predictive value. We went through this for years with the coronary calcium score.
Competing interests pertaining specifically to this post, comment, or both:
None
The is quite different from statins and CAC testing (statin risks are well defined). Routinely checking platelet assays and guessing how to react could be a risky business, particularly in light of bleeding issues and prasugrel. “Common sense” is not that common (Voltaire, not me) but may be not to check these tests at this point.
But this 0 percent result…also noted in another case on this Exchange…how common? Is it real..repeated?
Regarding the question of conflict of interest vs potential conflict of interest. How does it matter whether the writer is trying to persuade the reader because the writer is convinced or because the writer has economic motives for wanting to convince the reader whether or not the writer is convinced? We need to know the bias of the writer only to the extent that we are unable to evaluate the authors arguments on our own. When we cannot adequately evaluate the merits of a study on our own, we have to use our faith in the author to help decide how much weight to give the study results and the study conclusion until we find some more trusted authority to help. I am annoyed and offended if I think that the author is trying to dupe me. I am only disappointed if I feel that the author is duped by enthusiasm for a preferred result. Both sources of bias leave me doubtful of the reliability of what I read. The reader should assume that all authors are biased and try to evaluate the merits of a study based on their understanding of the limitations of the study design, setting, execution and analysis. Unfortunately, I don’t consider myself capable of adequately critiquing the methods, randomization and statistical analysis in particular, of a study. I can, however, follow simple logic going from results to conclusions. This step seems to be beyond many authors, since too often they make the leap from “is in this situation” to “is always” or from “is” to “ought” . This blog line was provoked by an editorial in JAMA March, 16,2011. A commentary in the same issue,a page or two later, “Reexamining Metrics for Glucose Control” gave just as good an example of bias. The author informs the reader that there are some populations in which the HgA1c may not be as predictive, accurate or consistent. Some alternatives to HgA1c might be of some value for patients with “persistent inconsistencies between blood glucose monitoring data and HgA1c levels”. His conclusion is, “As of now, however, for the clinician, no single index should be used in isolation;…” That was quite a jump. What has happened to Catherine de Angelis? Has someone threatened her life? Two foxes in the hen houses. What is going on?
Competing interests pertaining specifically to this post, comment, or both:
A strongly pro single payer and proponent of increased funding to NIH coupled with strict requirements for researchers to be truly independent from commercial influence, and open access to all published medical research, no $35 per article via internet access, thanks to Elsevier. How can we have evidence based medicine when we can’t evaluate the evidence?
In GRAVITA, platelet inhibition testing was done post-PCI. Is it possible that a study of on-medication platelet inhibition level done pre-procedure might, especially in ACS/high risk patients, be more likely to show such testing to be clinically useful?
Competing interests pertaining specifically to this post, comment, or both:
None
To Sanjay…..nice summary
To Powell…cut point will alter AUC
To Bruce…” Some things are just common sense and don’t required clinical trials” Except for parachutes, not many things fall in this category. Certainly not most treatments.
To Alain….VerifyNow vs cholesterol for risk prediction? The former is association and the latter is causative.
To David….Good comments (0% inhibition “fishy” since values are reported as arbitrary P2Y12 reaction units (PRU), not % inhibition)
To Carol….the foxes are everywhere?
To Hisham….provocative question. Do you think it matters?
To Richard.
Not sure, but if there is potential benefit to higher on treatment paltelet inhibition , it is more likely to be seen in a high MACE group and when the level of inhibition studied is the one pre-intervention rather than the one 6 to 12 hours following iatrogenic arterial injury.
Can’t say that I’ve previously considered the question Hisham has posed. Namely, does PCI (i.e.,contrast agent, arterial injury, or some other factor) alter platelet inhibition assay results. Anybody know the answer to this?
Correction…mia culpa! David Powell is correct that ROC curve shows trade-offs of sensitivity and specificity for all potential (i.e., “cut points”_. Every cutpoint will have a uniquely associated point on the curve. The AUC should not change with any cut point.
Thanks, David!!!