Correspondence
Thrombolytic Therapy for Myocardial Infarction
N Engl J Med 1998; 338:545-548February 19, 1998
- Article
To the Editor:
The investigators in the Global Use of Strategies to Open Occluded Coronary Arteries (GUSTO III) trial (Oct. 16 issue)1 make an important comparison between reteplase and alteplase, but their investigation has a serious flaw — namely, the lack of blinding. This may have led to serious biases. Might patients have had more anxiety (and more tachycardia and catecholamine release) if they knew they were receiving the experimental treatment instead of the standard treatment? Might a doctor subconsciously treat a patient differently if chest pain continued for half an hour after the completion of treatment (with reteplase) than if pain continued only in the middle of treatment (with an infusion of alteplase), even though the period of pain was the same?
Such possible biases could easily affect mortality by as much as 0.5 percent in one treatment group, totally changing the results. Is this not why double blinding has become the standard for randomized trials? Why was such an important trial performed without blinding, especially after the Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries (GUSTO I) trial had been criticized for the same fault?
1 ReferencesSteven Frei, M.D.
Sacred Heart Hospital, Allentown, PA 18102-34901
The Global Use of Strategies to Open Occluded Coronary Arteries (GUSTO III) Investigators
Full Text | Web of Science | Medline
To the Editor:
On the basis of the lower 95 percent confidence limit of the 1 percent absolute difference in 30-day mortality between an accelerated infusion and streptokinase (0.4 percent) reported in the GUSTO I trial,1 the Continuous Infusion versus Double-Bolus Administration of Alteplase (COBALT) investigators (Oct. 16 issue)2 conclude that a double bolus of alteplase was not equivalent to the accelerated protocol. As Ware and Antman mention in the accompanying editorial,3 if the GUSTO III investigators had applied this same criterion, they would also have concluded that reteplase was not equivalent to the accelerated infusion of alteplase.
If the gold standard for the nonequivalence of thrombolytic agents is to be the difference between alteplase and streptokinase reported in GUSTO I, one should compare the combined end point of mortality and stroke, since this latter event has been consistently shown to be increased with alteplase. Thus, the appropriate criterion should be the lower 95 percent confidence limit of the 0.9 percent absolute difference in the net clinical benefit. This suggests that equivalence can be claimed only if any excess events do not exceed a 0.27 percent limit, not a 0.4 percent limit. This lower limit would be even closer to zero if the statistically more appropriate random-effects model were used. These modifications to the gold standard would only make an equivalence between alteplase and other fibrinolytic regimens even more difficult to establish.
One side advantage of equivalence studies is that they force our focus away from a single point estimate of the difference between agents toward confidence intervals and eventually probability distributions. Such an approach has shown that there is considerable uncertainty about the clinical superiority of accelerated alteplase over streptokinase.4 This supports the editorialists' opinion that the proposed equivalence criterion is too stringent and should be relaxed. Of course, the consequence of such an action would be to admit that there is no clinically significant difference between streptokinase and alteplase. The GUSTO investigators have been unwilling to make this admission, and therefore all future equivalence trials of fibrinolytic agents seem destined to fail when measured against the streptokinase–alteplase yardstick. The adage, “He who lives by the sword dies by the sword,” seems particularly apt.
4 ReferencesJames Brophy, M.D.
Centre Hospitalier Angrignon, Verdun, QC H4G 2A3, Canada1
The GUSTO Investigators
Full Text | Web of Science | Medline2
The Continuous Infusion versus Double-Bolus Administration of Alteplase (COBALT) Investigators
Full Text | Web of Science | Medline3
Ware JH ,Antman EM . Equivalence trials. N Engl J Med 1997;337:1159-1161
Full Text | Web of Science | Medline4
Brophy JM ,Joseph L . Placing trials in context using Bayesian analysis: GUSTO revisited by Reverend Bayes. JAMA 1995;273:871-875
CrossRef | Web of Science | Medline
To the Editor:
The COBALT investigators randomly assigned 7169 patients with acute myocardial infarction to receive intravenous tissue plasminogen activator either as two 50-mg boluses given 30 minutes apart or as an accelerated infusion of 100 mg given over a period of 90 minutes. The incidence of the combined end point of death or nonfatal stroke at 30 days was 8.8 percent in the double-bolus group and 8.3 percent in the accelerated-infusion group (P = 0.45); the median activated partial-thromboplastin times at 6 hours were 101 seconds and 84 seconds, respectively (P<0.001). Of note, 30-day rates of death or nonfatal stroke in the two groups were identical for patients 75 years old or younger (6.3 percent) but not for the 934 patients who were older than 75 years (25.3 percent in the double-bolus group vs. 21.4 percent in the accelerated-infusion group). The authors suggest that the worse clinical outcome with the double-bolus regimen may be related to less frequent early coronary reperfusion.
In a previous study, Purvis et al., using the same double-bolus regimen of tissue plasminogen activator, reported a high percentage of grade 3 flow rates (according to the Thrombolysis in Myocardial Infarction classification) at 90 minutes (88 percent) in infarct-related arteries.1 Patients older than 75 years were not included in the study by Purvis et al., and no strokes occurred. The incidence of stroke after thrombolytic therapy increases significantly with age,2 whereas the rate of coronary reperfusion after thrombolysis has not been found to be age-dependent. It is known that bolus administration of tissue plasminogen activator, as compared with an infusion, is followed by less thrombin generation3 and that longer activated partial-thromboplastin times after thrombolytic treatment are associated with higher incidences of death and intracranial hemorrhage and with older age.4
These considerations indicate that double-bolus administration of tissue plasminogen activator may be associated with excessive bleeding in certain high-risk groups, such as the elderly, and this may account for the worse clinical outcome, as compared with a continuous infusion. In the elderly, frequent measurement of the activated partial-thromboplastin time may be of value in monitoring risk, adjusting antithrombotic treatment, and improving the outcome after thrombolytic therapy.
4 ReferencesFelicita Andreotti, M.D., Ph.D.
Università Cattolica del Sacro Cuore, 00168 Rome, Italy1
Purvis JA ,McNeill AJ ,Siddiqui RA , et al. Efficacy of 100 mg of double-bolus alteplase in achieving complete perfusion in the treatment of acute myocardial infarction. J Am Coll Cardiol 1994;23:6-10
CrossRef | Web of Science | Medline2
Maggioni AP ,Franzosi MG ,Santoro E , et al. The risk of stroke in patients with acute myocardial infarction after thrombolytic and antithrombotic treatment. N Engl J Med 1992;327:1-6
Full Text | Web of Science | Medline3
Andreotti F ,Kluft C ,Hackett DR ,Davies GJ ,Maseri A . Thrombin generation after fast or prolonged regimens of tissue-type plasminogen activator. Lancet 1993;342:937-938
CrossRef | Web of Science | Medline4
Granger CB ,Hirsch J ,Califf RM , et al. Activated partial thromboplastin time and outcome after thrombolytic therapy for acute myocardial infarction: results from the GUSTO-I Trial. Circulation 1996;93:870-878
Web of Science | Medline
To the Editor:
It seems to me that the matter of equivalence trials1 is being made inordinately complicated. A trial shows that the treatment difference lies between the confidence limits2 or below a one-sided confidence limit.3 It proves no more than this. And since there will always be some distance between upper and lower confidence limits, it is unwarranted to claim that the experiment shows equivalence unless any difference within the confidence limits can be considered equivalence. Whether this is so is a matter of judgment and opinion, which it seems best to separate from the statistical analysis. There seems no need to get entangled in null and alternative hypotheses.
3 ReferencesC. Dennis Thron, M.D.
5 Barrymore Rd., Hanover, NH 037551
Ware JH ,Antman EM . Equivalence trials. N Engl J Med 1997;337:1159-1161
Full Text | Web of Science | Medline2
The Global Use of Strategies to Open Occluded Coronary Arteries (GUSTO III) Investigators
Full Text | Web of Science | Medline3
The Continuous Infusion versus Double-Bolus Administration of Alteplase (COBALT) Investigators
Full Text | Web of Science | Medline
Author/Editor Response
The authors reply:
To the Editor: The issue of the equivalence of two alternative active therapies in clinical trials is a complex one and unfortunately is not elucidated by the letter by Brophy. In GUSTO I,1 we demonstrated the unequivocal clinical and statistical superiority of alteplase over streptokinase; treatment with alteplase prevented one of every seven deaths that would otherwise have occurred with the established (streptokinase) therapy. We maintain that the absolute difference of 1 percent in mortality (relative reduction of 15 percent in GUSTO I) should be regarded as the appropriate boundary for the definition of equivalence of comparative thrombolytic trials. As we stated in the GUSTO III report, “Our study was not designed to assess equivalence, nor did it have adequate power to do so.”2 However, as we duly acknowledged in the report, our data do not support the equivalence of alteplase and reteplase, using the definition of a 1 percent difference, since the 95 percent confidence intervals for the difference in 30-day mortality are -1.11 to 0.66 percent. To preempt the need for very large trials involving well in excess of 50,000 patients in the future, which might have adequate power to address this issue, we called for “careful scrutiny” of the boundaries for equivalence.2 Dr. Brophy's adage should thus be changed to “swallows his own sword.”
Regarding Dr. Frei's concern about the lack of a double-blind, double-dummy design, there are many considerations to keep in mind. First, several large-scale trials of mortality in the reperfusion era have been conducted on an open or double-blind basis. In each case, the findings of the open-design trial were fully validated by the subsequent double-blind trial — for example, the results of the first Gruppo Italiano per lo Studio della Sopravvivenza nell'Infarto Miocardico (GISSI) trial were validated by those of the Second International Study of Infarct Survival (ISIS-2), and the results of GISSI-2 were validated by those of ISIS-3.3 Second, mortality as an outcome measure is very unlikely to be affected by knowledge of the thrombolytic assignment. Third, the logistics of a double-blind and double-dummy design for active treatment comparisons are cumbersome, since the design necessitates a delay in the administration of one of the agents. For a bolus of a thrombolytic agent such as reteplase, this obligatory delay compromises the potential advantage of the drug. With the negative results of the GUSTO III trial (lack of superiority), at least the delay in the administration of reteplase cannot be invoked as the explanation. Fourth, for proper blinding, the cooperation of the sponsors is usually required to obtain a physically identical placebo with the regulatory approval to be administered in the trial. Unfortunately, having industry cooperate in providing a placebo for such comparative trials is not at all straightforward.
3 ReferencesEric J. Topol, M.D.
Cleveland Clinic Foundation, Cleveland, OH 44195Robert M. Califf, M.D.
Duke University Medical Center, Durham, NC 27710Robert Wilcox, M.D.
Queen's Medical Center University Hospital, Nottingham NG7 2UH, United Kingdomfor the GUSTO III Steering Committee
1
The GUSTO Investigators
Full Text | Web of Science | Medline2
The Global Use of Strategies to Open Occluded Coronary Arteries (GUSTO III) Investigators
Full Text | Web of Science | Medline3
Lee KL ,Califf RM ,Simes J ,Van de Werf F ,Topol EJ . Holding GUSTO up to the light. Ann Intern Med 1994;120:876-881
Web of Science | Medline
Author/Editor Response
Although the statistics involved are complex, defining the equivalence of two alternative therapies is a clinical issue. Together with the International Joint Efficacy Comparison of Thrombolytics (INJECT) investigators1 and the GUSTO III investigators, we believe that in the field of thrombolysis, a 1 percent difference in mortality, if still demonstrable on long-term follow-up, is clinically important. As shown in GUSTO I, this 1 percent difference results in the prevention of one of every seven deaths. For a disease with a high prevalence and mortality, this reduction is relevant at the population level.
In the COBALT trial, which compared two ways of administering a marketed thrombolytic agent, the statistical criteria used to demonstrate equivalence with this (clinically determined) 1 percent boundary were very stringent. Indeed, the upper boundary of the difference had to be smaller than the lower 95 percent confidence limit of the difference in mortality between alteplase and streptokinase in GUSTO I. It was thought that to persuade clinicians to abandon accelerated infusion, indirect statistical evidence of superiority over streptokinase was needed as well. These stringent statistical criteria are not required for a new thrombolytic agent. For a new agent, it should suffice to demonstrate that the upper boundary of the difference in mortality with standard treatment does not exceed 1 percent and that therefore the two treatments have a similar efficacy.
The use of the combined end point of death and stroke, as suggested by Dr. Brophy, only increases the complexity of the problem. Indeed, neurologic symptoms in hemodynamically unstable patients are not always due to stroke, and the classification of the type of stroke and the degree of disability is sometimes impossible. The equivalence of thrombolytic agents should be established first on the basis of mortality alone, and other aspects, such as side effects, ease of use, and cost, must be left to the interpretation of the clinician.
With regard to Dr. Andreotti's comments, in the COBALT trial, the rates of hemorrhagic stroke increased with age, but the higher incidence with the double bolus as compared with infusion was confined to elderly patients. The activated partial-thromboplastin time was indeed longer after the double bolus and may be partly responsible for the higher risk of stroke. The patency rates after a double bolus in the Double-Bolus Lytic Efficacy (DouBLE) trial 2 suggest an equal efficacy for clot lysis but not superiority over infusion. On the basis of these observations, we suggested that a double bolus may be an alternative way of administering alteplase in selected populations (e.g., patients who are 75 years old or younger, especially those treated in an ambulance, because of the ease of use) and that a lower dose of heparin may be sufficient.
2 ReferencesFrans Van de Werf, M.D., Ph.D.
University Hospital Gasthuisberg, B-3000 Leuven, Belgiumfor the COBALT Investigators
1
International Joint Efficacy Comparison of Thrombolytics
CrossRef | Web of Science | Medline2
Bleich SD ,Adgey AAJ ,Pickering E , et al. An angiographic assessment of the efficacy and safety of front-loaded and bolus regimens of activase (alteplase, recombinant): the Double-Bolus Lytic Efficacy Trial (the DouBLE Trial). Circulation 1995;92:Suppl I:I-415 abstract.
Author/Editor Response
Dr. Thron is correct in his assertion that the confidence interval provides the statistical information needed to determine whether a clinical trial has met a specified criterion of equivalence. Rothman1 made an eloquent case for reliance on the confidence interval in the interpretation of clinical and epidemiologic studies. We also noted in our editorial that one can interpret an equivalence trial by forming the one-sided 100(1 - α) percent confidence interval for the difference, p T - pS, and concluding that the treatments are equivalent if the upper confidence limit is less than the equivalence criterion, δ. The principles of hypothesis testing are needed, however, in the planning of a trial. As the controversy over the design of AIDS trials in developing countries illustrates, the choice between an efficacy design and an equivalence design raises both scientific and ethical issues. Moreover, having chosen to conduct either a superiority trial or an equivalence trial, the investigator should specify the null and alternative hypotheses in advance and choose a sample size sufficient to provide good power when the alternative hypothesis is true. One cannot avoid the key questions of study design by choosing to rely on confidence intervals for interpretation of the study's results. Among other issues, the choice of study design determines whether a one-sided confidence interval is appropriate. We believe the hypothesis-testing formulation continues to be a useful framework for making decisions about study design.
1 ReferencesJames H. Ware, Ph.D.
Harvard School of Public HealthElliott M. Antman, M.D.
Brigham and Women's Hospital, Boston, MA 021151
Rothman KJ . A show of confidence. N Engl J Med 1978;299:1362-1363
Full Text | Web of Science | Medline
- Citing Articles (1)
Citing Articles
1
Sarah A. Spinler, Stephanie M. Inverso. (2001) Update on Strategies to Improve Thrombolysis for Acute Myocardial Infarction. Pharmacotherapy 21:6, 691-716
CrossRef
