(Stroke. 2004;35:537.)
© 2004 American Heart Association, Inc.
Original Contributions |
Editorial Comment—An Approach to the Estimation of the Risk of TTP During Clopidogrel Therapy
Section of Quantitative Investigation, Department of Neurology, New York University School of Medicine, New York, NY
In this issue of Stroke, Zakarija and 12 co-authors1 describe 37 cases of thrombotic thrombocytopenic purpura (TTP) that occurred in the United States in the years 1998 to 2002 in association with clopidogrel (Plavix) therapy. This is of interest to those of us who (as characterized by Donnan and Davis2) are "strokologists," since it raises questions about our understanding of the CAPRIE study.3 In CAPRIE there were 25 cases of platelet counts <100 000 per mm3 in 9553 patients receiving clopidogrel 75 mg/d, and 25 cases in 9546 patients receiving aspirin 325 mg/d (2.6 cases per thousand with each treatment); no cases of TTP were reported.
We pose 2 questions: (1) Why did Zakarija et al find cases of TTP while the CAPRIE investigators did not? (2) Can we estimate the true risk of TTP during clopidogrel therapy?
With regard to the first question, a possible explanation is that TTP occurs late in clopidogrel treatment, and was not seen in CAPRIE because treatment duration was too short. This explanation fails for 2 reasons: (1) Zakarija et al found that TTP occurred within the first 30 days in 87.5% of cases, and (2) mean treatment duration in CAPRIE was fairly long (1.63 years). Alternatively, we could conclude that very large numbers of instances of thrombocytopenia must occur before even a single instance of TTP will develop (all 37 patients had thrombocytopenia), and that the CAPRIE cohort was too small to produce these numbers. This seems a more plausible explanation, and presumably underlies the pharmaceutical companies’ postmarketing estimate of 4 cases per million patients, as cited by Zakarija et al.
This brings us to the second question: do we accept the pharmaceutical companies’ estimate of 4 cases per million patients, or can we examine the issue ourselves? Zakarija et al did not think that they could make an independent estimate; they commented: "An accurate estimate of incidence rate of clopidogrel-associated TTP cannot be determined from our data." We, however, see a way of using their data to make estimates of TTP occurrence; we present our methods and results below.
Zakarija et al used 3 case-finding approaches. Thirteen of their cases were identified through the active querying of physicians performing plasma exchange, and of hematologists, in 8 urban areas. According to the New York Times World Almanac 2001, these 8 areas accounted for 40 million (14.8%) of the 1998 US population of 270 million. On the assumptions of completeness of ascertainment in the 8 areas, and of uniformity of distribution of cases throughout the United States, the 5-year total for the United States should have been 13/0.148=approximately 88 cases. Obviously the assumptions can be questioned; nevertheless, if the second assumption is correct then the 37-13=24 cases identified through 2 passive reporting channels represent under-ascertainment compared with the active querying method.
We have reviewed financial statements estimating the sales of Plavix by the manufacturers for the years 1998 to 2002. The approximate values for these years were $134 million, $485 million, $801 million, $1.1 billion, and $1.6 billion, respectively. On the estimate of $50 wholesale price for a month’s supply, the number of "Plavix months" (PMs) of agent sold in 1998 would be $134 million/$50=approximately 2.7 million. By similar calculation the PM values in millions for the subsequent years would be approximately 9.7, 16.0, 22.0, and 32.0. The total PM for 1998 to 2002 would be 82.4 million. If each patient took 1 month of treatment and then received no more clopidogrel the total number of patients treated would have been 82.4 million and the rate of TTP would have been 88/82.4 million or approximately 1.1 per million; the order of magnitude is similar to the pharmaceutical companies’ estimate of 4 per million.
The 4 per million rate from 88 TTP cases in 5 years would have required a total of 22 million patients: 88 per 22 million=4 per 1 million; in that circumstance the duration of treatment would have averaged (82.4 million/22 million)=3.745 months.
What about longer durations of treatment? The maximal treatment duration assumption would be that all patients begun on treatment at any time in the 5-year period remained on treatment as of December 31, 2002. Let us start in 1998 with the assumption that all 2.7 million PMs where consumed by first-time users whose entry into treatment was distributed uniformly across the year. Therefore these first-time users averaged 6 PMs during the year; 2.7 million/6=450 000 new patients in 1998. In 1999 these 450 000 patients would have used 450 000x12=5.4 million PMs out of the 9.7 million PMs sold that year; 9.7 million-5.4 million=4.3 million PMs, which were consumed by 4.3 million/6=approximately 717 000 patients new to treatment in 1999. Application of the algorithm to the next 3 years leads to the estimation of 333 000 new patients in 2000, 667 000 new patients in 2001, and 999 000 new patients in 2002. From this algorithm the total of new patients (and therefore of all patients treated) for 1998 through 2002 is 3.166 million; 88 TTP cases in 3.166 million patients would give a rate of approximately 27.8 per million, an order of magnitude greater than the pharmaceutical companies’ estimate.
The range of our estimates—1.1 to 27.8 per million—reflects the assumptions we made. (For instance, the estimate of $65 rather than $50 per PM would result in the range of 1.4 to 36.2 per million.) It should be possible to replace some of these assumptions with data. The manufacturers can make a valuable contribution by publishing the actual net numbers of Plavix tablets sold for the years 1998 to 2002; this would obviate the need to use price estimates. Data giving ratios of new prescriptions to refills should be solicited from major retail pharmacy chains; these data would also contribute importantly to the refining of the estimation of risk of TTP during clopidogrel therapy.
 |
References |
|---|
 Top References |
|---|
1. Zakarija A et al. Clopidogrel-associated thrombotic thrombocytopenic purpura (TTP): An update of pharmacovigilance efforts conducted by independent researchers, the pharmaceutical suppliers, and the Food and Drug Administration. Stroke. 2004; 35: 533–538.
2. Donnan GA, Davis SM. Neurologist, internist, or strokologist? Stroke. 2003; 34: 2765.
3. CAPRIE Steering Committee. A randomized, blinded, trial of clopidogrel vs. aspirin in patients at risk of ischaemic events (CAPRIE). Lancet. 1996; 348: 1329–1339.[CrossRef][Medline] [Order article via Infotrieve]
This article has been cited by other articles:
 |
|
 |
|
  C. L. Bennett, B. Kim, A. Zakarija, N. Bandarenko, D. K. Pandey, C. G. Buffie, J. M. McKoy, A. D. Tevar, J. F. Cursio, P. R. Yarnold, et al. Two Mechanistic Pathways for Thienopyridine-Associated Thrombotic Thrombocytopenic Purpura: A Report From the SERF-TTP Research Group and the RADAR Project J. Am. Coll. Cardiol., September 18, 2007; 50(12): 1138 - 1143. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. L. Bennett, J. R. Nebeker, E. A. Lyons, M. H. Samore, M. D. Feldman, J. M. McKoy, K. R. Carson, S. M. Belknap, S. M. Trifilio, G. T. Schumock, et al. The Research on Adverse Drug Events and Reports (RADAR) Project JAMA, May 4, 2005; 293(17): 2131 - 2140. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||