- Email: [email protected]
Platelet glycoprotein inhibitors in patients with medically managed acute coronary syndrome: Does the enthusiasm exceed the science?
SPECIAL CONTRIBUTION
Platelet Glycoprotein Inhibitors in Patients With Medically Managed Acute Coronary Syndrome: Does the Enthusiasm Exceed the Science?
From the UCLA School of Medicine,* UCLA Emergency Medicine Center,‡ UCLA School of Nursing,§ and the Department of Emergency Medicine, Olive View-UCLA Medical Center,II Los Angeles, CA.
David L. Schriger, MD, MPH*‡ Mel E. Herbert, MD, MBBS, BMedSci*§II
Dr. Schriger has received an unrestricted gift from the Pfizer Corporation for the evaluation of computerized psychiatric screening in the emergency department. He has also received unrestricted gifts from the MedAmerica Corporation for support of his health services research. Reprints not available from the authors. Address for correspondence: David L. Schriger, MD, MPH, UCLA Emergency Medicine Center, UCLA School of Medicine, 924 Westwood Blvd, Suite 300, Los Angeles, CA 90024-2924; 310-794-0593; E-mail [email protected]. Copyright © 2001 by the American College of Emergency Physicians. 0196-0644/2001/$35.00 + 0 47/1/117881 doi:10.1067/mem.2001.117881
SEPTEMBER 2001
38:3
ANNALS OF EMERGENCY MEDICINE
See related articles, p. 229 and p. 241. Guidelines from the American Heart Association and the American College of Cardiology, as well as numerous review articles, have strongly and enthusiastically recommended that platelet glycoprotein IIb/IIIa inhibitors be used in patients with medically managed unstable angina or non–ST-segment myocardial infarction (UA/NSTEMI). We explore whether there is sufficient experimental evidence to justify these recommendations. We review the 4 large randomized trials of US Food and Drug Administration–approved platelet glycoprotein IIb/IIIa inhibitors that included medically managed UA/NSTEMI patients, first taking each trial’s results at face value and then in the context of likely biases. The risk differences, unadjusted for potential biases, are 2.5% (0.6%, 4.4%) for the Platelet Receptor Inhibition in Ischemic Syndrome Management (PRISM) study, 2.3% (–1.9%, 6.5%) for the Platelet Receptor Inhibition in Ischemic Syndrome Management in Patients Limited by Unstable Signs and Symptoms (PRISM-PLUS) study (tirofiban plus heparin), 0.9% (–0.9%, 2.8%) for the Platelet Glycoprotein IIb/IIIa in Unstable Angina Receptor Suppression Using Integrilin Therapy (PURSUIT) trial, –0.2% (–1.7%, 1.3%) for the least harmful treatment arm of the Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries (GUSTO-IV) trial, and –1.9% (–6.8%, 3.0%) for the PRISM-PLUS study (tirofiban alone) (positive numbers indicate benefit). The 95% confidence interval produced by combining the studies using a random effects model is –1.3% to 3.2% (mean 0.9%); this is consistent with drugs providing a small benefit, no benefit, or causing harm. Confounding caused by the nonrandom selection of patients for percutaneous transluminal coronary angioplasty and coronary artery bypass grafting in all trials except GUSTO-IV and problems arising from the fact that enrolled patients were much sicker than typical UA/ NSTEMI patients are likely to have biased the studies away
2 4 9
PLATELET GLYCOPROTEIN INHIBITORS Schriger & Herbert
from the null and make an assertion of benefit even more tenuous. Given the equivocal results, it would appear that the authors are relying on opinion rather than evidence to formulate their conclusions. Clinicians should understand that opinion and factors other than medical evidence may influence the content of the recommendations. [Schriger DL, Herbert ME. Platelet glycoprotein inhibitors in patients with medically managed acute coronary syndrome: does the enthusiasm exceed the science? Ann Emerg Med. September 2001;38:249-255.] INTRODUCTION
One article in this issue of Annals summarizes the 2000 American Heart Association (AHA) and American College of Cardiology (ACC) guidelines for the management of patients with unstable angina and non–ST-segment elevation myocardial infarction (UA/NSTEMI); another explains how emergency physicians should implement the recommendations found in the guidelines.1,2 Both articles enthusiastically endorse the use of “super-aspirin” platelet glycoprotein IIb/IIIa inhibitors for patients with UA/ NSTEMI. The article that summarizes the guidelines states, “Once the decision is made to admit a high-risk acute coronary syndrome (ACS) patient who does not have ST-segment elevation, standard medical therapy should include aspirin, a β-blocker, antithrombin therapy, and a platelet glycoprotein IIb/IIIa inhibitor [our emphasis].”1 A statement such as this, presented without caveat, implies that physicians who do not use platelet glycoprotein IIb/IIIa agents are practicing bad medicine. Because these guidelines are reportedly “evidence-based” and carry the supposed imprimatur of the AHA and ACC, readers are likely to assume that such strident endorsement is reflective of compelling, unequivocal supporting evidence.3 For, if the evidence was not compelling, why would the guidelines’ authors and the authors of these review articles endorse the agents so unequivocally? We demonstrate that there is insufficient evidence to support unconditional recommendation of the platelet glycoprotein IIb/IIIa inhibitors. We begin by examining the conclusions reached when available evidence is assumed to be free of bias. We then consider how bias at the level of individual trials, trial synthesis, and guideline creation might affect conclusions. Because of space concerns, our critique is restricted to the use of platelet glycoprotein IIb/IIIa inhibitors in patients with medically managed UA/NSTEMI. This is not to say that we do not have
2 5 0
similar concerns about other aspects of the recommendations made in the aforementioned guidelines and articles. C H A R A C T E R I S T I C S O F T H E D ATA
Four large randomized trials evaluated US Food and Drug Administration (FDA)-approved glycoprotein inhibitors and included at least some ACS patients who are managed without percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG) (Table 1). The Platelet Receptor Inhibition in Ischemic Syndrome Management (PRISM) and Platelet Receptor Inhibition in Ischemic Syndrome Management in Patients Limited by Unstable Signs and Symptoms (PRISM-PLUS) studies involved tirofiban (Aggrastat); the Platelet Glycoprotein IIb/IIIa in Unstable Angina Receptor Suppression Using Integrilin Therapy (PURSUIT) trial involved eptifibatide (Integrelin); and the Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries (GUSTO-IV) trial involved abciximab (Reopro).4-7 The first 3 studies included some patients who received medical management only, as well as some patients who underwent PCI or CABG. In the GUSTO-IV trial, all patients received medical management only. We begin by considering features of the trials that are relevant to the interpretation of a face value analysis, that is, an analysis that assumes that the trials are free of bias and error. We consider the metric used to describe results, the use of composite outcome measures, the timing and choice of outcome measures, and the potential for spectrum bias (eg, the enrolled patients were considerably sicker than patients encountered in typical emergency departments). Relative risk (RR; the percentage of patients with the outcome in the treatment group divided by the percentage with the outcome in the control group) is the primary metric for all 4 trials. When an outcome is rare, RR can be misleading because a value of 0.33 (a 66% reduction in the outcome) seems far more impressive than the same change described as a risk difference (RD; also known as the absolute risk reduction) of 2%, from 3% to 1%. Another advantage of reporting RD is that its reciprocal, the “number needed to treat” (the number of patients who need to be treated to achieve one additional good or bad outcome), can be easily gleaned. In this article, we represent the data as RDs so that readers have access to a metric that has practical clinical relevance. Transformation from RR to RD is a mechanical process that required no assumptions or interpretation on our part.
ANNALS OF EMERGENCY MEDICINE
38:3 SEPTEMBER 2001
PLATELET GLYCOPROTEIN INHIBITORS Schriger & Herbert
The 4 trials use composite end points (typically death or myocardial infarction [MI], or death, MI, or recurrent ischemia) as their primary outcome measure. Readers should understand that composite end points can be helpful when each component is measuring a unique aspect of the outcome. However, in many studies in cardiology, the components of the composite end points are not independent (eg, if a drug decreases the number of MIs it should also decrease mortality). Because death is likely to be the outcome of greatest concern to patients, it may be more straightforward to measure only mortality. This is because a drug could improve the composite end point by decreasing a less important outcome but have no effect (or a paradoxical effect) on the more important one. Most composite end points weigh each of the components equally, yet patients may have very different beliefs about the importance of each component. A meta-analysis that separately analyzed each component of the composite outcome of these studies failed to demonstrate evidence of effect for any component.8 Our analysis focuses on death or nonfatal MI at 30 days because this was the most clinically important outcome available in all studies. Only the PRISM-PLUS trial reported 6-month outcomes. This is unfortunate, be-
cause long-term outcomes are especially important when an intervention is expensive, and typical charges per patient for a usual course of therapy with these drugs is US$1,625 for eptifibatide, US$1,260 for tirofiban, and US$2,160 for abciximab.9,10 If death is deferred and the patient lives for several years with a good quality of life, the expense of these agents may be justified. If the patient lives only 6 weeks or has poor quality of life, the expense may not be worthwhile. The inclusion of longer-term outcomes and measures of functional status, general health, or quality of life would have been advantageous to those trying to determine the utility of these drugs. It is regrettable that such measures are omitted from many trials. Emergency physicians should understand that the patients enrolled in these trials were much sicker than the cohort of patients with chest pain (or the cohort with unstable angina) presenting to the typical ED. Each trial required some combination of transient ST or T-wave changes, persistent (not relieved by nitroglycerin) or recurrent chest pain, or positive cardiac enzyme or marker levels. ECG changes, which are a marker for increased short-term risk of death or MI, were observed in more than 75% of patients enrolled in these studies
Table 1.
Summary of the platelet glycoprotein IIb/IIIa randomized trials of FDA-approved agents that included medically managed patients.* RD for Death or MI at 30 D
Study
Limbs (No.)
Medically Managed Patients, No. (%)
% of Patients With ECG Changes
Medically Managed Patients
Patients With PTCA or CABG
1,999 (62)
75
2.5 (0.6, 4.4)
0.3 (–4.5, 5.2)
719 (46) —†
>90 >90
PRISM
(a) Tirofiban (1,616) (b) Placebo (1,616)
PRISM-PLUS
(a) Tirofiban and heparin (773) (b) Placebo (797) (c) Tirofiban (345)
PURSUIT
(a) High-dose eptifibatide and heparin‡ (4,722) (b) Placebo and heparin (4,739) (c) Low-dose eptifibatide and heparin (1,787)II
5,866 (62)§
92
0.9 (–0.9, 2.8)
2.4 (0.1, 4.7)
GUSTO-IV
(a) High-dose abciximab and heparin (2,590) (b) Placebo and heparin (2,598) (c) Low-dose abciximab and heparin (2,612)II
5,188 (100)
—¶
–0.2 (–1.7 to 1.3)
NA
2.3 (–1.9, 6.5) 4.2 (–0.1, 8.4) –1.9% (–6.8%, 3.0%)†
*Patients
in all limbs received aspirin unless contraindicated. does not report data stratified on the method of management. The numbers represent all patients. ‡Ninety percent of patients received heparin in the PURSUIT trial. §This number is extrapolated (see text footnote). IIThese limbs performed worse than the other limbs and are not considered in the calculations because we wanted to consider each therapy at its most beneficial dosing regimen. ¶This number has not been published. All patients had 0.5-mm ST depression, positive troponin levels, or both indicators. †PRISM-PLUS
SEPTEMBER 2001
38:3
ANNALS OF EMERGENCY MEDICINE
2 5 1
PLATELET GLYCOPROTEIN INHIBITORS Schriger & Herbert
(Table 1).11 Approximately 40% of enrolled patients had had a previous MI, 13% had had angioplasty, and 15% had had CABG. In short, the patients studied in these trials were at much higher risk for bad outcomes than a cohort of ED patients being evaluated for chest pain thought to be of cardiac origin. Because a drug designed to prevent MI and death can only help patients who are likely to have such outcomes but can harm anyone who receives it, the net benefit will increase when the drug is given to sicker patients. Thus, it is reasonable to expect that the magnitude of any benefit found in patients in these studies will be greater than that found in a more typical population of ED UA/NSTEMI patients. FA C E VA L U E A N A LY S I S
If we assume that the 4 cited studies are free from all forms of error and bias, the 30-day death or nonfatal MI RD for medically managed patients is 2.5% (0.6%, 4.4%) in the PRISM study, 2.3% (–1.9%, 6.5%) in the tirofiban plus heparin limb of the PRISM-PLUS study, –1.90%* (–6.8%, 3.0%) in the tirofiban-alone limb of the PRISM-PLUS study, 0.9%† (–0.9%, 2.8%) in the PURSUIT trial, and *This
is the value for both medically and invasively managed patients. Stratified data were not reported. Because these agents are known to benefit invasively managed patients, this deviation biases our analysis in favor of the drug’s efficacy.
†The
value reflects patients who did not undergo PCI in the first 72 hours. Unlike the other trials, the value does include patients who underwent PCI after 72 hours or who underwent CABG at any time. No data were provided on patients who were solely managed medically. The N (and, therefore, the confidence interval [CI]) is extrapolated from the N provided in the paper.
–0.2% (–1.7%, 1.3%) for the least harmful treatment limb of the GUSTO-IV trial (Table 1, Figure). Although these data are consistent with a small (≅1%) net benefit in this cohort of very ill UA/NSTEMI patients, they are also equally consistent with the drugs having a more substantial effect, having no effect, or causing harm. No amount of subgroup, Bayesian analysis, meta-analysis, or evidence-based analysis can eliminate the uncertainty inherent in these data. Nevertheless, it is instructive to examine whether these studies conclusively show that the platelet glycoprotein IIb/IIIa inhibitors are beneficial, even with the assumption that there are no biases in the design or execution of the studies. We can combine the results of the 5 limbs using a random effects meta-analytic technique. “Random effects” means that our mathematic model for combining the studies is based on the assumption that the results of each study should not be identical and should vary randomly about an average value.12,13 This is reasonable because the studies were conducted with different inclusion criteria, different agents, and different protocols. The combined result has a 95% confidence interval from –1.3% to 3.2% (mean 0.9%). Twenty-one percent of the distribution lies below zero and 79% lies above zero. This means that, although the exact effect of these drugs is unknown, our current knowledge suggests that there is a 79% possibility that they provide some benefit and a 21% possibility that they produce net harm. Even fixed effect meta-analysis, which makes the unreasonable assumption that the studies are identical, produces a 95% confidence interval that, although narrower, still
Figure.
Results of randomized trials of FDA-approved platelet glycoprotein IIb/IIIa inhibitors that included at least some medically managed patients. Each study result is represented by a line that demarcates the 95% CI for the absolute RD. The central tick is the mean. Positive values indicate that the drug is better than the control. The combined interval was created using a random effects hierarchic Bayesian model.12,13
PRISM-PLUS (tirofiban; n=695) GUSTO-IV (abciximab and heparin; n=5,188) PURSUIT (eptifibatide and heparin; n=5,866) PRISM-PLUS (tirofiban and heparin; n=719) PRISM (tirofiban; n=1,999) Combined (random effects model; n=14,467) –8%
–4% Favors control
0%
4% Favors drug
8%
Risk difference (RD)
2 5 2
ANNALS OF EMERGENCY MEDICINE
38:3 SEPTEMBER 2001
PLATELET GLYCOPROTEIN INHIBITORS Schriger & Herbert
includes negative values. Thus, even if we assume that each study was free from bias, the data fail to provide unequivocal evidence that the platelet glycoprotein IIb/IIIa inhibitors are beneficial. Furthermore, we should remember that the patients in these studies were much sicker than typical UA/NSTEMI patients and that any benefit is likely to be reduced or nullified if these agents are given to more representative patients. TRIAL BIAS
The aforementioned analysis assumed that the individual trials were unbiased. The point estimates and CIs reported in the trials are accurate if (1) each trial was faithfully and honestly undertaken and reported; (2) randomization was successful; (3) each trial was free of selection bias, workup bias, differential subject loss, and bias in outcome assessment; (4) the statistical models accurately reflect the underlying phenomena; (5) the nonrandom allocation to interventions (eg, percutaneous transluminal coronary angioplasty [PTCA], CABG) was not associated with both treatment and outcome, that is, the interventions did not confound the results; and (6) the results produced in the study hospitals are representative of the community and have external validity. Violations of any of these assumptions can lead to biased point estimates or inappropriately narrow CIs for the drug’s effect, particularly for its effect when used in community practice. The results of these trials are sufficiently close to the null that even small violations of the aforementioned assumptions could shift the reported point estimates from “helpful” to “harmful.” These studies were all sponsored by pharmaceutical manufacturers. The potential for bias in such trials has been considered in detail. 14 Learned scientists and epidemiologists could legitimately argue the validity of each assumption in each trial. Of particular concern is the fact that in each trial except GUSTO-IV, the decision to perform angiography, PTCA, or CABG was not randomized. If there is an association between the study limb and the likelihood of intervention and an association between having an invasive procedure and outcome, then the study is confounded. For example, in the PURSUIT trial, the frequency of cardiac catheterization was 79% in North America, 58% in western Europe, 46% in Latin America, and 20% in eastern Europe. The approximate RDs for these regions were 4%, 1%, –0.5%, and –1%, respectively. Roughly two thirds of the patients were enrolled at sites in North America and western Europe. The marked
SEPTEMBER 2001
38:3
ANNALS OF EMERGENCY MEDICINE
variation in effect by geographic region and the correlation of results with the frequency of cardiac catheterization raise questions regarding the effectiveness of the glycoprotein inhibitors in medically managed patients. Furthermore, if more patients had been enrolled in the regions with lower catheterization rates, the overall RD could have shifted to favor the control group. As mentioned previously, another crucial concern is that the 2 trials of tirofiban without heparin versus heparin alone (PRISM, RD 2.5% [0.6%, 4.4%] and PRISM-PLUS, RD –1.9% [–6.9%, 3.0%]) produced divergent results. In response to these legitimate concerns regarding bias, we could widen the reported CIs to reflect the additional uncertainty.* When we do so, there is even less support for the argument that these agents are unequivocally beneficial. In summary, we believe that these data are inconclusive. It is unquestionable that the agents have biological activity, but there are innumerable instances in medicine where biological activity does not translate into medical benefit. We simply do not know yet which, if any, medically managed patients will experience sufficient benefit from the platelet glycoprotein IIb/IIIa inhibitors to justify the harms and costs. The demographics and baseline risk of a patient, the presence of ECG changes, the release of cardiac enzymes, the specifics of the hospital, and the treatment strategy (eg, are interventions planned?) are all likely to affect the outcome of glycoprotein therapy in ways that we do not yet understand. CONSIDERING COSTS
In most situations, considerations of cost-effectiveness begin after the health benefit of a medication has been unequivocally established. Thus, we might consider a discussion of the cost-effectiveness of the platelet glycoprotein IIb/III inhibitors in medically managed ACS patients premature. Nevertheless, we briefly examine these issues. The best-known platelet inhibitor is aspirin. One MI or death is prevented for every 12 to 43 patients treated, at a total cost of less than US$5 (Table 2).15 Even if the platelet glycoprotein IIb/IIIa inhibitors provide benefit, the number needed to treat to benefit (NNTB) is considerably more than for aspirin. Estimates of the NNTB range from 18 to infinity, with estimates of cost per *Mathematic
methods can be used to produce credible intervals that reflect likely biases. We do not present such techniques here because the face value analysis established that these agents are not unequivocally beneficial, and the consideration of likely biases would only increase uncertainty.
2 5 3
PLATELET GLYCOPROTEIN INHIBITORS Schriger & Herbert
MI prevented or life saved at 30 days of US$50,000 to infinity. Thus, even if subsequent research establishes the efficacy and effectiveness of these agents, legitimate concerns regarding their cost-effectiveness remain. P O T E N T I A L B I A S I N C R E AT I O N O F R E C O M M E N D AT I O N S
Given the equivocal and conflicting data for medically managed patients found in these trials, a reasonable con-
Table 2.
FDA-approved antiplatelet drugs, number needed to treat to benefit or harm, and cost per event prevented. NNTB or (NNTH) 1 Death or MI at 30 D* Agent
Mean
Aspirin Tirofiban
19 63‡
Eptifibatide
43
Abciximab
(500)
Cost per Life Saved (US$) or MI Prevented at 30 D†
95% CI
Mean
95% CI
NNTB 12, 43 NNTB 30, ∞ NNTH 500, ∞ NNTB 15, ∞ NNTH 111, ∞ NNTB 77, ∞ NNTH 59, ∞
Pennies $79,000
Pennies, $5 $38,000, ∞
$70,000
$24,000, ∞
Infinite
Infinite
NNTB, number need to benefit; NNTH, number needed to harm. *NNTH is relevant when the control patients do better than those in the active limb. NNTB and NNTH are calculated as 1/RD. †Costs per life saved or MI prevented = NNTB × the drug’s average wholesale cost.9,10 ‡Calculations include medically managed patients in the PRISM study and the tirofiban and heparin limb of the PRISM-PLUS study and all patients in the tirofiban-only limb of the PRISMPLUS study.
clusion would be to continue enrolling patients in randomized trials of these drugs. It is difficult to accept the conclusion of the ACC/AHA guidelines that there is “A” level evidence that these are Class 1 agents. Class 1 is defined as “…evidence or general agreement that a specific procedure is useful and effective.”3 Because the evidence is inconclusive, it must be the “general agreement” that is driving these recommendations. That being the case, we can conclude that their content is primarily based on the opinions and biases of the authors. The guideline summary article in this issue of Annals states “both eptifibatide and tirofiban have been shown to have benefits in UA/NSTEMI patients who do not undergo PCI.”1 This statement obscures the uncertainties that we have detailed and fails to acknowledge that the patients in these trials are much sicker than typical UA/NSTEMI patients. Table 3 shows factors that affect the potential for interpretive bias in the development of recommendations and shows that the amount, quality, and consistency of evidence regarding the effects of platelet glycoprotein IIb/IIIa inhibitors in medically managed UA/NSTEMI patients is such that there is a high potential for biased interpretation. In the absence of compelling data, relying on the opinions of experts has been shown to be a process fraught with pitfalls.16,17 In summary, we are troubled that so much effort has been expended in the development and dissemination of guidelines and review articles for a treatment that has not been convincingly established as beneficial. We find in the evidence neither sufficient reason to support the widespread adoption of this therapy nor reason to bar its use. Although we choose to treat our medically managed UA/NSTEMI patients without these drugs, we respect
Table 3.
Potential for biased interpretation of data, modifying characteristics, and state of platelet glycoprotein IIb/IIIa data. Data Characteristic Is Associated With: Low Potential for Interpretive Bias
High Potential for Interpretive Bias
Large magnitude of effect Large number of observations Consistent results among studies Consistent results among subgroups Good match between studies’ target populations and population of interest Studies’ outcomes are valid and clinically meaningful
Small magnitude of effect Small number of observations Inconsistent results among studies Inconsistent results among subgroups Poor match between studies’ target populations and population of interest Outcomes have not been validated or are not patient-centered
2 5 4
State of the Platelet Glycoprotein IIa/IIIb Data Small magnitude of effect Moderately large number of observations There are some inconsistencies There are inconsistencies, particularly in the PURSUIT trial Poor match because patients in studies are far sicker than typical ED chest pain patients 30-day death and nonfatal MI are reasonable outcomes, but longer follow-up and functional status and quality-of-life measures would be desirable
ANNALS OF EMERGENCY MEDICINE
38:3 SEPTEMBER 2001
PLATELET GLYCOPROTEIN INHIBITORS Schriger & Herbert
physicians who choose to use them, so long as they have made an educated decision based on their own evaluation of the data. We thank Dr. Richelle Cooper, Dr. Jerome Hoffman, and especially Dr. Michael Callaham for their helpful comments on drafts of this manuscript.
REFERENCES 1. Pollack CV Jr, Gibler WB. 2000 ACC/AHA guidelines for the management of patients with unstable angina and non–ST-segment elevation myocardial infarction: a practical summary for emergency physicians. Ann Emerg Med. 2001;38:229-240. 2. Pollack CV Jr, Gibler WB. Advances create opportunities: implementing the major tenets of the new unstable angina guidelines in the emergency department. Ann Emerg Med. 2001;38:241-248. 3. Braunwald E, Antman EM, Beasley JW, et al. ACC/AHA guidelines for the management of patients with unstable angina and non-ST-segment elevation myocardial infarction. A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee on the Management of Patients With Unstable Angina). J Am Coll Cardiol. 2000;36:970-1062. Available at: http://www.acc.org/clinical/guidelines/unstable/unstable.pdf. Accessed July 17, 2001. 4. The PRISM-PLUS Study Investigators. Inhibition of the platelet glycoprotein IIb/IIIa receptor with tirofiban in unstable angina and non–Q-wave myocardial infarction. Platelet Receptor Inhibition in Ischemic Syndrome Management in Patients Limited by Unstable Signs and Symptoms (PRISM-PLUS) Study Investigators. N Engl J Med. 1998;338:1488-1497. 5. The PRISM Study Investigators. A comparison of aspirin plus tirofiban with aspirin plus heparin for unstable angina. Platelet Receptor Inhibition in Ischemic Syndrome Management (PRISM) Study Investigators. N Engl J Med. 1998;338:1498-1505. 6. Inhibition of platelet glycoprotein IIb/IIIa with eptifibatide in patients with acute coronary syndromes. The PURSUIT Trial Investigators. Platelet Glycoprotein IIb/IIIa in Unstable Angina: Receptor Suppression Using Integrilin Therapy. N Engl J Med. 1998;339:436-443. 7. Califf RM. Glycoprotein IIb/IIIa blockade and thrombolytics: early lessons from the SPEED and GUSTO IV trials. Am Heart J. 1999;138:S12-S15. 8. Brown DL, Fann CS, Chang CJ. Effect of glycoprotein IIb/IIIa inhibitors on the individual components of composite endpoints used in clinical trials of unstable angina and non–Q-wave myocardial infarction. Cardiovasc Drugs Ther. 2000;14:253-258. 9. Two new antiplatelet drugs for angioplasty and acute coronary syndromes. Med Lett Drugs Ther. 1998;40:89-92. 10. Medical Economics Staff. Drug Topics Red Book, 2000. Oradell, NJ: Medical Economics Company; 2000. 11. Goldman L, Cook EF, Johnson PA, et al. Prediction of the need for intensive care in patients who come to the emergency department with acute chest pain. N Engl J Med. 1996;334:14981504. 12. Eddy DM, Hasselblad V, Shacter R. Meta-analysis by the Confidence Profile Method: The Statistical Synthesis of Evidence. San Diego, CA: Academic Press; 1991. 13. Eddy DM, Hasselblad V. FAST*PRO: Software for Meta-analysis by the Confidence Profile Method. San Diego, CA: Academic Press; 1992:93-98. 14. Bero LA, Rennie D. Influences on the quality of published drug studies. Int J Technol Assess Health Care. 1996;12:209-237. 15. Lewis HD, Davis JW, Archibald DG, et al. Protective effects against acute myocardial infarction and death in men with unstable angina. Results of a Veterans Administration Cooperative Study. N Engl J Med. 1983;309:396-403. 16. Antman EM, Lau J, Kupelnick B, et al. A comparison of results of meta-analyses of randomized control trials and recommendations of clinical experts: treatments for myocardial infarction. JAMA. 1992;268:240-248. 17. Eddy DM. Clinical decision making: from theory to practice: the challenge. JAMA. 1990;263:287-290.
SEPTEMBER 2001
38:3
ANNALS OF EMERGENCY MEDICINE
2 5 5
Platelet Glycoprotein Inhibitors in Patients With Medically Managed Acute Coronary Syndrome: Does the Enthusiasm Exceed the Science?
From the UCLA School of Medicine,* UCLA Emergency Medicine Center,‡ UCLA School of Nursing,§ and the Department of Emergency Medicine, Olive View-UCLA Medical Center,II Los Angeles, CA.
David L. Schriger, MD, MPH*‡ Mel E. Herbert, MD, MBBS, BMedSci*§II
Dr. Schriger has received an unrestricted gift from the Pfizer Corporation for the evaluation of computerized psychiatric screening in the emergency department. He has also received unrestricted gifts from the MedAmerica Corporation for support of his health services research. Reprints not available from the authors. Address for correspondence: David L. Schriger, MD, MPH, UCLA Emergency Medicine Center, UCLA School of Medicine, 924 Westwood Blvd, Suite 300, Los Angeles, CA 90024-2924; 310-794-0593; E-mail [email protected]. Copyright © 2001 by the American College of Emergency Physicians. 0196-0644/2001/$35.00 + 0 47/1/117881 doi:10.1067/mem.2001.117881
SEPTEMBER 2001
38:3
ANNALS OF EMERGENCY MEDICINE
See related articles, p. 229 and p. 241. Guidelines from the American Heart Association and the American College of Cardiology, as well as numerous review articles, have strongly and enthusiastically recommended that platelet glycoprotein IIb/IIIa inhibitors be used in patients with medically managed unstable angina or non–ST-segment myocardial infarction (UA/NSTEMI). We explore whether there is sufficient experimental evidence to justify these recommendations. We review the 4 large randomized trials of US Food and Drug Administration–approved platelet glycoprotein IIb/IIIa inhibitors that included medically managed UA/NSTEMI patients, first taking each trial’s results at face value and then in the context of likely biases. The risk differences, unadjusted for potential biases, are 2.5% (0.6%, 4.4%) for the Platelet Receptor Inhibition in Ischemic Syndrome Management (PRISM) study, 2.3% (–1.9%, 6.5%) for the Platelet Receptor Inhibition in Ischemic Syndrome Management in Patients Limited by Unstable Signs and Symptoms (PRISM-PLUS) study (tirofiban plus heparin), 0.9% (–0.9%, 2.8%) for the Platelet Glycoprotein IIb/IIIa in Unstable Angina Receptor Suppression Using Integrilin Therapy (PURSUIT) trial, –0.2% (–1.7%, 1.3%) for the least harmful treatment arm of the Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries (GUSTO-IV) trial, and –1.9% (–6.8%, 3.0%) for the PRISM-PLUS study (tirofiban alone) (positive numbers indicate benefit). The 95% confidence interval produced by combining the studies using a random effects model is –1.3% to 3.2% (mean 0.9%); this is consistent with drugs providing a small benefit, no benefit, or causing harm. Confounding caused by the nonrandom selection of patients for percutaneous transluminal coronary angioplasty and coronary artery bypass grafting in all trials except GUSTO-IV and problems arising from the fact that enrolled patients were much sicker than typical UA/ NSTEMI patients are likely to have biased the studies away
2 4 9
PLATELET GLYCOPROTEIN INHIBITORS Schriger & Herbert
from the null and make an assertion of benefit even more tenuous. Given the equivocal results, it would appear that the authors are relying on opinion rather than evidence to formulate their conclusions. Clinicians should understand that opinion and factors other than medical evidence may influence the content of the recommendations. [Schriger DL, Herbert ME. Platelet glycoprotein inhibitors in patients with medically managed acute coronary syndrome: does the enthusiasm exceed the science? Ann Emerg Med. September 2001;38:249-255.] INTRODUCTION
One article in this issue of Annals summarizes the 2000 American Heart Association (AHA) and American College of Cardiology (ACC) guidelines for the management of patients with unstable angina and non–ST-segment elevation myocardial infarction (UA/NSTEMI); another explains how emergency physicians should implement the recommendations found in the guidelines.1,2 Both articles enthusiastically endorse the use of “super-aspirin” platelet glycoprotein IIb/IIIa inhibitors for patients with UA/ NSTEMI. The article that summarizes the guidelines states, “Once the decision is made to admit a high-risk acute coronary syndrome (ACS) patient who does not have ST-segment elevation, standard medical therapy should include aspirin, a β-blocker, antithrombin therapy, and a platelet glycoprotein IIb/IIIa inhibitor [our emphasis].”1 A statement such as this, presented without caveat, implies that physicians who do not use platelet glycoprotein IIb/IIIa agents are practicing bad medicine. Because these guidelines are reportedly “evidence-based” and carry the supposed imprimatur of the AHA and ACC, readers are likely to assume that such strident endorsement is reflective of compelling, unequivocal supporting evidence.3 For, if the evidence was not compelling, why would the guidelines’ authors and the authors of these review articles endorse the agents so unequivocally? We demonstrate that there is insufficient evidence to support unconditional recommendation of the platelet glycoprotein IIb/IIIa inhibitors. We begin by examining the conclusions reached when available evidence is assumed to be free of bias. We then consider how bias at the level of individual trials, trial synthesis, and guideline creation might affect conclusions. Because of space concerns, our critique is restricted to the use of platelet glycoprotein IIb/IIIa inhibitors in patients with medically managed UA/NSTEMI. This is not to say that we do not have
2 5 0
similar concerns about other aspects of the recommendations made in the aforementioned guidelines and articles. C H A R A C T E R I S T I C S O F T H E D ATA
Four large randomized trials evaluated US Food and Drug Administration (FDA)-approved glycoprotein inhibitors and included at least some ACS patients who are managed without percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG) (Table 1). The Platelet Receptor Inhibition in Ischemic Syndrome Management (PRISM) and Platelet Receptor Inhibition in Ischemic Syndrome Management in Patients Limited by Unstable Signs and Symptoms (PRISM-PLUS) studies involved tirofiban (Aggrastat); the Platelet Glycoprotein IIb/IIIa in Unstable Angina Receptor Suppression Using Integrilin Therapy (PURSUIT) trial involved eptifibatide (Integrelin); and the Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries (GUSTO-IV) trial involved abciximab (Reopro).4-7 The first 3 studies included some patients who received medical management only, as well as some patients who underwent PCI or CABG. In the GUSTO-IV trial, all patients received medical management only. We begin by considering features of the trials that are relevant to the interpretation of a face value analysis, that is, an analysis that assumes that the trials are free of bias and error. We consider the metric used to describe results, the use of composite outcome measures, the timing and choice of outcome measures, and the potential for spectrum bias (eg, the enrolled patients were considerably sicker than patients encountered in typical emergency departments). Relative risk (RR; the percentage of patients with the outcome in the treatment group divided by the percentage with the outcome in the control group) is the primary metric for all 4 trials. When an outcome is rare, RR can be misleading because a value of 0.33 (a 66% reduction in the outcome) seems far more impressive than the same change described as a risk difference (RD; also known as the absolute risk reduction) of 2%, from 3% to 1%. Another advantage of reporting RD is that its reciprocal, the “number needed to treat” (the number of patients who need to be treated to achieve one additional good or bad outcome), can be easily gleaned. In this article, we represent the data as RDs so that readers have access to a metric that has practical clinical relevance. Transformation from RR to RD is a mechanical process that required no assumptions or interpretation on our part.
ANNALS OF EMERGENCY MEDICINE
38:3 SEPTEMBER 2001
PLATELET GLYCOPROTEIN INHIBITORS Schriger & Herbert
The 4 trials use composite end points (typically death or myocardial infarction [MI], or death, MI, or recurrent ischemia) as their primary outcome measure. Readers should understand that composite end points can be helpful when each component is measuring a unique aspect of the outcome. However, in many studies in cardiology, the components of the composite end points are not independent (eg, if a drug decreases the number of MIs it should also decrease mortality). Because death is likely to be the outcome of greatest concern to patients, it may be more straightforward to measure only mortality. This is because a drug could improve the composite end point by decreasing a less important outcome but have no effect (or a paradoxical effect) on the more important one. Most composite end points weigh each of the components equally, yet patients may have very different beliefs about the importance of each component. A meta-analysis that separately analyzed each component of the composite outcome of these studies failed to demonstrate evidence of effect for any component.8 Our analysis focuses on death or nonfatal MI at 30 days because this was the most clinically important outcome available in all studies. Only the PRISM-PLUS trial reported 6-month outcomes. This is unfortunate, be-
cause long-term outcomes are especially important when an intervention is expensive, and typical charges per patient for a usual course of therapy with these drugs is US$1,625 for eptifibatide, US$1,260 for tirofiban, and US$2,160 for abciximab.9,10 If death is deferred and the patient lives for several years with a good quality of life, the expense of these agents may be justified. If the patient lives only 6 weeks or has poor quality of life, the expense may not be worthwhile. The inclusion of longer-term outcomes and measures of functional status, general health, or quality of life would have been advantageous to those trying to determine the utility of these drugs. It is regrettable that such measures are omitted from many trials. Emergency physicians should understand that the patients enrolled in these trials were much sicker than the cohort of patients with chest pain (or the cohort with unstable angina) presenting to the typical ED. Each trial required some combination of transient ST or T-wave changes, persistent (not relieved by nitroglycerin) or recurrent chest pain, or positive cardiac enzyme or marker levels. ECG changes, which are a marker for increased short-term risk of death or MI, were observed in more than 75% of patients enrolled in these studies
Table 1.
Summary of the platelet glycoprotein IIb/IIIa randomized trials of FDA-approved agents that included medically managed patients.* RD for Death or MI at 30 D
Study
Limbs (No.)
Medically Managed Patients, No. (%)
% of Patients With ECG Changes
Medically Managed Patients
Patients With PTCA or CABG
1,999 (62)
75
2.5 (0.6, 4.4)
0.3 (–4.5, 5.2)
719 (46) —†
>90 >90
PRISM
(a) Tirofiban (1,616) (b) Placebo (1,616)
PRISM-PLUS
(a) Tirofiban and heparin (773) (b) Placebo (797) (c) Tirofiban (345)
PURSUIT
(a) High-dose eptifibatide and heparin‡ (4,722) (b) Placebo and heparin (4,739) (c) Low-dose eptifibatide and heparin (1,787)II
5,866 (62)§
92
0.9 (–0.9, 2.8)
2.4 (0.1, 4.7)
GUSTO-IV
(a) High-dose abciximab and heparin (2,590) (b) Placebo and heparin (2,598) (c) Low-dose abciximab and heparin (2,612)II
5,188 (100)
—¶
–0.2 (–1.7 to 1.3)
NA
2.3 (–1.9, 6.5) 4.2 (–0.1, 8.4) –1.9% (–6.8%, 3.0%)†
*Patients
in all limbs received aspirin unless contraindicated. does not report data stratified on the method of management. The numbers represent all patients. ‡Ninety percent of patients received heparin in the PURSUIT trial. §This number is extrapolated (see text footnote). IIThese limbs performed worse than the other limbs and are not considered in the calculations because we wanted to consider each therapy at its most beneficial dosing regimen. ¶This number has not been published. All patients had 0.5-mm ST depression, positive troponin levels, or both indicators. †PRISM-PLUS
SEPTEMBER 2001
38:3
ANNALS OF EMERGENCY MEDICINE
2 5 1
PLATELET GLYCOPROTEIN INHIBITORS Schriger & Herbert
(Table 1).11 Approximately 40% of enrolled patients had had a previous MI, 13% had had angioplasty, and 15% had had CABG. In short, the patients studied in these trials were at much higher risk for bad outcomes than a cohort of ED patients being evaluated for chest pain thought to be of cardiac origin. Because a drug designed to prevent MI and death can only help patients who are likely to have such outcomes but can harm anyone who receives it, the net benefit will increase when the drug is given to sicker patients. Thus, it is reasonable to expect that the magnitude of any benefit found in patients in these studies will be greater than that found in a more typical population of ED UA/NSTEMI patients. FA C E VA L U E A N A LY S I S
If we assume that the 4 cited studies are free from all forms of error and bias, the 30-day death or nonfatal MI RD for medically managed patients is 2.5% (0.6%, 4.4%) in the PRISM study, 2.3% (–1.9%, 6.5%) in the tirofiban plus heparin limb of the PRISM-PLUS study, –1.90%* (–6.8%, 3.0%) in the tirofiban-alone limb of the PRISM-PLUS study, 0.9%† (–0.9%, 2.8%) in the PURSUIT trial, and *This
is the value for both medically and invasively managed patients. Stratified data were not reported. Because these agents are known to benefit invasively managed patients, this deviation biases our analysis in favor of the drug’s efficacy.
†The
value reflects patients who did not undergo PCI in the first 72 hours. Unlike the other trials, the value does include patients who underwent PCI after 72 hours or who underwent CABG at any time. No data were provided on patients who were solely managed medically. The N (and, therefore, the confidence interval [CI]) is extrapolated from the N provided in the paper.
–0.2% (–1.7%, 1.3%) for the least harmful treatment limb of the GUSTO-IV trial (Table 1, Figure). Although these data are consistent with a small (≅1%) net benefit in this cohort of very ill UA/NSTEMI patients, they are also equally consistent with the drugs having a more substantial effect, having no effect, or causing harm. No amount of subgroup, Bayesian analysis, meta-analysis, or evidence-based analysis can eliminate the uncertainty inherent in these data. Nevertheless, it is instructive to examine whether these studies conclusively show that the platelet glycoprotein IIb/IIIa inhibitors are beneficial, even with the assumption that there are no biases in the design or execution of the studies. We can combine the results of the 5 limbs using a random effects meta-analytic technique. “Random effects” means that our mathematic model for combining the studies is based on the assumption that the results of each study should not be identical and should vary randomly about an average value.12,13 This is reasonable because the studies were conducted with different inclusion criteria, different agents, and different protocols. The combined result has a 95% confidence interval from –1.3% to 3.2% (mean 0.9%). Twenty-one percent of the distribution lies below zero and 79% lies above zero. This means that, although the exact effect of these drugs is unknown, our current knowledge suggests that there is a 79% possibility that they provide some benefit and a 21% possibility that they produce net harm. Even fixed effect meta-analysis, which makes the unreasonable assumption that the studies are identical, produces a 95% confidence interval that, although narrower, still
Figure.
Results of randomized trials of FDA-approved platelet glycoprotein IIb/IIIa inhibitors that included at least some medically managed patients. Each study result is represented by a line that demarcates the 95% CI for the absolute RD. The central tick is the mean. Positive values indicate that the drug is better than the control. The combined interval was created using a random effects hierarchic Bayesian model.12,13
PRISM-PLUS (tirofiban; n=695) GUSTO-IV (abciximab and heparin; n=5,188) PURSUIT (eptifibatide and heparin; n=5,866) PRISM-PLUS (tirofiban and heparin; n=719) PRISM (tirofiban; n=1,999) Combined (random effects model; n=14,467) –8%
–4% Favors control
0%
4% Favors drug
8%
Risk difference (RD)
2 5 2
ANNALS OF EMERGENCY MEDICINE
38:3 SEPTEMBER 2001
PLATELET GLYCOPROTEIN INHIBITORS Schriger & Herbert
includes negative values. Thus, even if we assume that each study was free from bias, the data fail to provide unequivocal evidence that the platelet glycoprotein IIb/IIIa inhibitors are beneficial. Furthermore, we should remember that the patients in these studies were much sicker than typical UA/NSTEMI patients and that any benefit is likely to be reduced or nullified if these agents are given to more representative patients. TRIAL BIAS
The aforementioned analysis assumed that the individual trials were unbiased. The point estimates and CIs reported in the trials are accurate if (1) each trial was faithfully and honestly undertaken and reported; (2) randomization was successful; (3) each trial was free of selection bias, workup bias, differential subject loss, and bias in outcome assessment; (4) the statistical models accurately reflect the underlying phenomena; (5) the nonrandom allocation to interventions (eg, percutaneous transluminal coronary angioplasty [PTCA], CABG) was not associated with both treatment and outcome, that is, the interventions did not confound the results; and (6) the results produced in the study hospitals are representative of the community and have external validity. Violations of any of these assumptions can lead to biased point estimates or inappropriately narrow CIs for the drug’s effect, particularly for its effect when used in community practice. The results of these trials are sufficiently close to the null that even small violations of the aforementioned assumptions could shift the reported point estimates from “helpful” to “harmful.” These studies were all sponsored by pharmaceutical manufacturers. The potential for bias in such trials has been considered in detail. 14 Learned scientists and epidemiologists could legitimately argue the validity of each assumption in each trial. Of particular concern is the fact that in each trial except GUSTO-IV, the decision to perform angiography, PTCA, or CABG was not randomized. If there is an association between the study limb and the likelihood of intervention and an association between having an invasive procedure and outcome, then the study is confounded. For example, in the PURSUIT trial, the frequency of cardiac catheterization was 79% in North America, 58% in western Europe, 46% in Latin America, and 20% in eastern Europe. The approximate RDs for these regions were 4%, 1%, –0.5%, and –1%, respectively. Roughly two thirds of the patients were enrolled at sites in North America and western Europe. The marked
SEPTEMBER 2001
38:3
ANNALS OF EMERGENCY MEDICINE
variation in effect by geographic region and the correlation of results with the frequency of cardiac catheterization raise questions regarding the effectiveness of the glycoprotein inhibitors in medically managed patients. Furthermore, if more patients had been enrolled in the regions with lower catheterization rates, the overall RD could have shifted to favor the control group. As mentioned previously, another crucial concern is that the 2 trials of tirofiban without heparin versus heparin alone (PRISM, RD 2.5% [0.6%, 4.4%] and PRISM-PLUS, RD –1.9% [–6.9%, 3.0%]) produced divergent results. In response to these legitimate concerns regarding bias, we could widen the reported CIs to reflect the additional uncertainty.* When we do so, there is even less support for the argument that these agents are unequivocally beneficial. In summary, we believe that these data are inconclusive. It is unquestionable that the agents have biological activity, but there are innumerable instances in medicine where biological activity does not translate into medical benefit. We simply do not know yet which, if any, medically managed patients will experience sufficient benefit from the platelet glycoprotein IIb/IIIa inhibitors to justify the harms and costs. The demographics and baseline risk of a patient, the presence of ECG changes, the release of cardiac enzymes, the specifics of the hospital, and the treatment strategy (eg, are interventions planned?) are all likely to affect the outcome of glycoprotein therapy in ways that we do not yet understand. CONSIDERING COSTS
In most situations, considerations of cost-effectiveness begin after the health benefit of a medication has been unequivocally established. Thus, we might consider a discussion of the cost-effectiveness of the platelet glycoprotein IIb/III inhibitors in medically managed ACS patients premature. Nevertheless, we briefly examine these issues. The best-known platelet inhibitor is aspirin. One MI or death is prevented for every 12 to 43 patients treated, at a total cost of less than US$5 (Table 2).15 Even if the platelet glycoprotein IIb/IIIa inhibitors provide benefit, the number needed to treat to benefit (NNTB) is considerably more than for aspirin. Estimates of the NNTB range from 18 to infinity, with estimates of cost per *Mathematic
methods can be used to produce credible intervals that reflect likely biases. We do not present such techniques here because the face value analysis established that these agents are not unequivocally beneficial, and the consideration of likely biases would only increase uncertainty.
2 5 3
PLATELET GLYCOPROTEIN INHIBITORS Schriger & Herbert
MI prevented or life saved at 30 days of US$50,000 to infinity. Thus, even if subsequent research establishes the efficacy and effectiveness of these agents, legitimate concerns regarding their cost-effectiveness remain. P O T E N T I A L B I A S I N C R E AT I O N O F R E C O M M E N D AT I O N S
Given the equivocal and conflicting data for medically managed patients found in these trials, a reasonable con-
Table 2.
FDA-approved antiplatelet drugs, number needed to treat to benefit or harm, and cost per event prevented. NNTB or (NNTH) 1 Death or MI at 30 D* Agent
Mean
Aspirin Tirofiban
19 63‡
Eptifibatide
43
Abciximab
(500)
Cost per Life Saved (US$) or MI Prevented at 30 D†
95% CI
Mean
95% CI
NNTB 12, 43 NNTB 30, ∞ NNTH 500, ∞ NNTB 15, ∞ NNTH 111, ∞ NNTB 77, ∞ NNTH 59, ∞
Pennies $79,000
Pennies, $5 $38,000, ∞
$70,000
$24,000, ∞
Infinite
Infinite
NNTB, number need to benefit; NNTH, number needed to harm. *NNTH is relevant when the control patients do better than those in the active limb. NNTB and NNTH are calculated as 1/RD. †Costs per life saved or MI prevented = NNTB × the drug’s average wholesale cost.9,10 ‡Calculations include medically managed patients in the PRISM study and the tirofiban and heparin limb of the PRISM-PLUS study and all patients in the tirofiban-only limb of the PRISMPLUS study.
clusion would be to continue enrolling patients in randomized trials of these drugs. It is difficult to accept the conclusion of the ACC/AHA guidelines that there is “A” level evidence that these are Class 1 agents. Class 1 is defined as “…evidence or general agreement that a specific procedure is useful and effective.”3 Because the evidence is inconclusive, it must be the “general agreement” that is driving these recommendations. That being the case, we can conclude that their content is primarily based on the opinions and biases of the authors. The guideline summary article in this issue of Annals states “both eptifibatide and tirofiban have been shown to have benefits in UA/NSTEMI patients who do not undergo PCI.”1 This statement obscures the uncertainties that we have detailed and fails to acknowledge that the patients in these trials are much sicker than typical UA/NSTEMI patients. Table 3 shows factors that affect the potential for interpretive bias in the development of recommendations and shows that the amount, quality, and consistency of evidence regarding the effects of platelet glycoprotein IIb/IIIa inhibitors in medically managed UA/NSTEMI patients is such that there is a high potential for biased interpretation. In the absence of compelling data, relying on the opinions of experts has been shown to be a process fraught with pitfalls.16,17 In summary, we are troubled that so much effort has been expended in the development and dissemination of guidelines and review articles for a treatment that has not been convincingly established as beneficial. We find in the evidence neither sufficient reason to support the widespread adoption of this therapy nor reason to bar its use. Although we choose to treat our medically managed UA/NSTEMI patients without these drugs, we respect
Table 3.
Potential for biased interpretation of data, modifying characteristics, and state of platelet glycoprotein IIb/IIIa data. Data Characteristic Is Associated With: Low Potential for Interpretive Bias
High Potential for Interpretive Bias
Large magnitude of effect Large number of observations Consistent results among studies Consistent results among subgroups Good match between studies’ target populations and population of interest Studies’ outcomes are valid and clinically meaningful
Small magnitude of effect Small number of observations Inconsistent results among studies Inconsistent results among subgroups Poor match between studies’ target populations and population of interest Outcomes have not been validated or are not patient-centered
2 5 4
State of the Platelet Glycoprotein IIa/IIIb Data Small magnitude of effect Moderately large number of observations There are some inconsistencies There are inconsistencies, particularly in the PURSUIT trial Poor match because patients in studies are far sicker than typical ED chest pain patients 30-day death and nonfatal MI are reasonable outcomes, but longer follow-up and functional status and quality-of-life measures would be desirable
ANNALS OF EMERGENCY MEDICINE
38:3 SEPTEMBER 2001
PLATELET GLYCOPROTEIN INHIBITORS Schriger & Herbert
physicians who choose to use them, so long as they have made an educated decision based on their own evaluation of the data. We thank Dr. Richelle Cooper, Dr. Jerome Hoffman, and especially Dr. Michael Callaham for their helpful comments on drafts of this manuscript.
REFERENCES 1. Pollack CV Jr, Gibler WB. 2000 ACC/AHA guidelines for the management of patients with unstable angina and non–ST-segment elevation myocardial infarction: a practical summary for emergency physicians. Ann Emerg Med. 2001;38:229-240. 2. Pollack CV Jr, Gibler WB. Advances create opportunities: implementing the major tenets of the new unstable angina guidelines in the emergency department. Ann Emerg Med. 2001;38:241-248. 3. Braunwald E, Antman EM, Beasley JW, et al. ACC/AHA guidelines for the management of patients with unstable angina and non-ST-segment elevation myocardial infarction. A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee on the Management of Patients With Unstable Angina). J Am Coll Cardiol. 2000;36:970-1062. Available at: http://www.acc.org/clinical/guidelines/unstable/unstable.pdf. Accessed July 17, 2001. 4. The PRISM-PLUS Study Investigators. Inhibition of the platelet glycoprotein IIb/IIIa receptor with tirofiban in unstable angina and non–Q-wave myocardial infarction. Platelet Receptor Inhibition in Ischemic Syndrome Management in Patients Limited by Unstable Signs and Symptoms (PRISM-PLUS) Study Investigators. N Engl J Med. 1998;338:1488-1497. 5. The PRISM Study Investigators. A comparison of aspirin plus tirofiban with aspirin plus heparin for unstable angina. Platelet Receptor Inhibition in Ischemic Syndrome Management (PRISM) Study Investigators. N Engl J Med. 1998;338:1498-1505. 6. Inhibition of platelet glycoprotein IIb/IIIa with eptifibatide in patients with acute coronary syndromes. The PURSUIT Trial Investigators. Platelet Glycoprotein IIb/IIIa in Unstable Angina: Receptor Suppression Using Integrilin Therapy. N Engl J Med. 1998;339:436-443. 7. Califf RM. Glycoprotein IIb/IIIa blockade and thrombolytics: early lessons from the SPEED and GUSTO IV trials. Am Heart J. 1999;138:S12-S15. 8. Brown DL, Fann CS, Chang CJ. Effect of glycoprotein IIb/IIIa inhibitors on the individual components of composite endpoints used in clinical trials of unstable angina and non–Q-wave myocardial infarction. Cardiovasc Drugs Ther. 2000;14:253-258. 9. Two new antiplatelet drugs for angioplasty and acute coronary syndromes. Med Lett Drugs Ther. 1998;40:89-92. 10. Medical Economics Staff. Drug Topics Red Book, 2000. Oradell, NJ: Medical Economics Company; 2000. 11. Goldman L, Cook EF, Johnson PA, et al. Prediction of the need for intensive care in patients who come to the emergency department with acute chest pain. N Engl J Med. 1996;334:14981504. 12. Eddy DM, Hasselblad V, Shacter R. Meta-analysis by the Confidence Profile Method: The Statistical Synthesis of Evidence. San Diego, CA: Academic Press; 1991. 13. Eddy DM, Hasselblad V. FAST*PRO: Software for Meta-analysis by the Confidence Profile Method. San Diego, CA: Academic Press; 1992:93-98. 14. Bero LA, Rennie D. Influences on the quality of published drug studies. Int J Technol Assess Health Care. 1996;12:209-237. 15. Lewis HD, Davis JW, Archibald DG, et al. Protective effects against acute myocardial infarction and death in men with unstable angina. Results of a Veterans Administration Cooperative Study. N Engl J Med. 1983;309:396-403. 16. Antman EM, Lau J, Kupelnick B, et al. A comparison of results of meta-analyses of randomized control trials and recommendations of clinical experts: treatments for myocardial infarction. JAMA. 1992;268:240-248. 17. Eddy DM. Clinical decision making: from theory to practice: the challenge. JAMA. 1990;263:287-290.
SEPTEMBER 2001
38:3
ANNALS OF EMERGENCY MEDICINE
2 5 5