D RUG TH ER A PY

Review Article

Drug Therapy A L A S T A I R J . J . W O O D , M. D . , Editor

A SPIRIN , H EPARIN , AND F IBRINOLYTIC T HERAPY IN S USPECTED A CUTE M YOCARDIAL I NFARCTION RORY COLLINS, M.B., B.S., RICHARD PETO, F.R.S., COLIN BAIGENT, B.M., B.CH., AND PETER SLEIGHT, D.M.

I

N this article, we review the randomized clinical trials of aspirin, of heparin, and of fibrinolytic therapy in patients with suspected acute myocardial infarction to determine which of these treatments have been shown to improve survival and other major clinical outcomes. ASPIRIN Benefits of Aspirin during and after Suspected Myocardial Infarction

The Second International Study of Infarct Survival (ISIS-2) demonstrated conclusively the substantial value of aspirin therapy in patients with suspected acute myocardial infarction1 (and other studies have shown the value of aspirin in patients with unstable angina2). In the ISIS-2 trial, assignment to one month of treatment with 162.5 mg of enteric-coated aspirin per day (with the first tablet crushed or chewed for a rapid antiplatelet effect) produced a significant reduction of about one fifth in mortality (P0.001) (Fig. 1). This reduction represents the avoidance of about 25 early deaths for every 1000 patients with suspected acute myocardial infarction who are treated with aspirin for one month (Table 1), and further follow-up in the ISIS-2 trial found that these early benefits of short-term treatment persist for at least several years.6 Assignment to one month of aspirin therapy in the ISIS-2 trial also approximately halved patients’ risk of reinfarction and of stroke, typically preventing an additional 10 nonfatal reinfarctions and 3 nonfatal

From the Clinical Trial Service Unit and Epidemiological Studies Unit, Nuffield Department of Clinical Medicine, University of Oxford, Oxford, United Kingdom. Address reprint requests to Professor Collins at the Clinical Trial Service Unit and Epidemiological Studies Unit, Radcliffe Infirmary, Oxford OX2 6HE, United Kingdom. ©1997, Massachusetts Medical Society.

strokes per 1000 treated patients.1 Moreover, in other studies, continuation of aspirin therapy beyond the first month approximately doubled these early benefits, with about 40 further deaths, reinfarctions, or strokes prevented per 1000 patients during the first few years of additional treatment.2 Benefits in Different Types of Patients

There is no good evidence that the proportional reduction in mortality produced by aspirin therapy is larger or smaller in any identifiable category of patient.1 In particular, the effects on overall mortality are similar whether or not patients receive fibrinolytic therapy (Fig. 1) or heparin (Table 1). Likewise, the effects of aspirin in the ISIS-2 trial were similar irrespective of the delay from the onset of symptoms to the start of treatment (the proportional reduction in mortality for those who underwent randomization within 4 hours was 257 percent; for a delay of 5 to 12 hours, it was 217 percent; and for a delay of 13 to 24 hours, it was 2112 percent). This argues for routine use of aspirin therapy in all patients with suspected acute myocardial infarction, even if they present late after the onset of symptoms. The apparent lack of a survival advantage in the first day or two after myocardial infarction, which is seen both with aspirin and with fibrinolytic therapy (Fig. 1), does not imply7 that aspirin therapy can be delayed without loss. Platelet activation and aggregation occur in the acute phase of myocardial infarction and are further increased by fibrinolytic therapy,8-10 with an increased risk of early coronary-artery reocclusion and reinfarction. Aspirin inhibits this acute platelet activation.10 Even in the absence of fibrinolytic therapy, aspirin reduced the rate of inhospital reinfarction in the ISIS-2 trial by about one third (from 2.9 percent to 1.9 percent), whereas among patients who had been given fibrinolytic therapy, the rate of in-hospital reinfarction was halved by aspirin therapy (from 3.8 percent to 1.8 percent). Indeed, the excess of reinfarction that normally follows fibrinolytic therapy appeared to be abolished completely by aspirin therapy.1 Value of Wider Use of Aspirin

Inhibition of cyclooxygenase-dependent platelet activation may well be responsible for most or all of the protective effects of aspirin.11 If so, then patients with suspected acute myocardial infarction or unstable angina should immediately be given a few hundred milligrams (e.g., at least 160 mg) of aspirin to produce virtually complete inhibition of cyclooxygenase in less than one hour.12,13 Daily doses Vol ume 336

Numbe r 12



847

The New England Journal of Medi ci ne

Routine hospital care only, 13% cumulative mortality (568 of 4300 patients)

Aspirin only

Mortality (%)

10

Streptokinase only

Streptokinase and aspirin, 8% cumulative mortality (343 of 4292 patients)

5

0 0

1

2

3

4

5

Weeks after Starting Treatment Figure 1. Cumulative Mortality from Vascular Causes up to Day 35 in the ISIS-2 Trial. A total of 17,187 patients were randomly assigned within 24 hours after the onset of suspected acute myocardial infarction to one of four regimens: placebo infusion and placebo tablets (i.e., routine hospital care); placebo infusion and 162.5 mg of aspirin daily for one month (aspirin only); an infusion of 1.5 million units of streptokinase over a one-hour period and placebo tablets (streptokinase only); or both streptokinase and aspirin.1

substantially lower than the 160-mg dose that was used in the ISIS-2 trial may take more than 24 hours to produce full inhibition.11,14,15 Aspirin therapy should then continue for at least several years, perhaps at a somewhat lower daily dose2; virtually complete inhibition of cyclooxygenase can be maintained indefinitely with a dose of about 75 mg per day,15 but not with much lower doses. In clinical trials, doses of 75 to 325 mg per day were protective, with, as expected pharmacologically, no clinical evidence of greater benefit from higher doses2 (which may cause adverse gastric effects). Hence, the appropriate dose for long-term aspirin therapy appears to be the lowest dose in the range of 75 to 325 mg per day that is convenient to prescribe; if available, enteric-coated aspirin may further help to limit gastric toxicity. Despite the strength of the evidence of a substantial benefit, many patients who present with suspected acute myocardial infarction still do not receive immediate aspirin therapy. Indeed, in a survey of the recent care of Medicare patients with acute myocardial infarction, only about half received aspirin on admission.16 Likewise, in the years after a myocardial infarction, long-term aspirin therapy is not used as 848 

Ma r ch 2 0 , 1 9 9 7

widely as it should be. In the Scandinavian Simvastatin Survival Study,17 for example, among patients who had had myocardial infarction only about half were taking aspirin in 1994, and in the North American Survival and Ventricular Enlargement Study,18 only about two thirds were receiving aspirin during 1992. Some of this lack of long-term aspirin use may be due to the preference of some physicians for anticoagulation with warfarin and related drugs.16 However, although oral anticoagulant therapy may be useful in patients who are not receiving aspirin,19,20 such therapy is difficult to control and is associated with a substantial risk of bleeding. Moreover, in a small trial of aspirin versus warfarin versus placebo in patients after myocardial infarction, there were significantly fewer vascular events in the patients given aspirin than in those given warfarin or placebo.21 And, as yet, there is no clear evidence of an additional benefit when warfarin is added to aspirin therapy (although some such studies are still in progress).22 Recommendations for Routine Aspirin Therapy

The avoidance of aspirin therapy during and after suspected myocardial infarction may sometimes be justified by very clear contraindications (for example, definite evidence of a serious allergic reaction to aspirin, recent severe gastrointestinal bleeding, or recent intracranial hemorrhage), but in many circumstances it does not appear to be. Given the definite and substantial benefits of aspirin therapy during and after myocardial infarction (and in unstable angina), it should not be withheld from patients with only mild contraindications (such as a vague history of allergy or a history of gastrointestinal bleeding or ulcer that is not recent), for whom the benefits are likely to outweigh any risks. In the rare patients who have a definite allergy to aspirin, some other antiplatelet drug (e.g., ticlopidine) should be considered.2 HEPARIN Addition of Standard Heparin Regimens to Aspirin

Most trials of heparin therapy in patients with acute myocardial infarction have been too small on their own to detect the moderate effects on the rates of major vascular events (e.g., death, reinfarction, stroke, or pulmonary embolism) that might realistically be expected.23 Taken together, however, the studies of heparin therapy in patients who did not routinely receive aspirin (most of whom also did not receive fibrinolytic therapy) do indicate some net benefit from heparin, despite the occurrence of major bleeding in about 10 patients per 1000 treated (Table 2).3 However, it is no longer relevant to assess heparin in the absence of aspirin, because the ISIS-2 trial found aspirin to be of sub-

D R UG TH ER A PY

TABLE 1. EFFECTS

OF

ASPIRIN

WITH AND WITHOUT PLANNED INTRAVENOUS HEPARIN THERAPY IN ACUTE MYOCARDIAL INFARCTION IN ISIS-2.*

PATIENTS

WITH

SUSPECTED

EFFECT/1000 PATIENTS ASSIGNED ASPIRIN NO IV HEPARIN PLANNED†

EVENT

EFFECT/1000

(N  6539)

PLACEBO

ASSIGNED

(N  6547)

ASPIRIN

PATIENTS ASPIRIN

P

VALUE

no. (%)

Death

626 (9.6)

778 (11.9)

Reinfarction

104 (1.6)

188 (2.9)

42 (0.6)

65 (1.0)

710 (11.0)

940 (14.6)

17 (0.3)

17 (0.3)

Stroke Stroke, reinfarction, or death Major bleeding

P VALUE

EFFECT/1000

PATIENTS ASPIRIN

(REGARDLESS OF HEPARIN USE)

IV HEPARIN PLANNED‡

(N  2048)

PLACEBO

ASSIGNED

(N  2053)

ASPIRIN

P

VALUE

no. (%)

235 fewer 133 fewer 42 fewer 367 fewer No apparent excess

0.001

178 (8.7)

238 (11.6)

0.001

52 (2.6)

96 (4.7)

0.03

9 (0.4)

23 (1.1)

0.001

208 (10.2)

302 (14.9)

14 (0.7)

16 (0.8)

299 fewer 215 fewer 73 fewer 4711 fewer No apparent excess

0.003 0.001 0.02 0.001

0.001

255 fewer 152 fewer 41 fewer 385 fewer 01

0.001 0.002 0.001 0.8

*Data are from ISIS-21 and are for patients who were randomly assigned to receive aspirin or placebo for one month. This table should not be used to assess the effects of heparin, because heparin treatment was not randomly assigned. Just before randomly assigning each patient to aspirin or placebo, doctors were asked whether they “planned” to use intravenous heparin for that patient, and on the discharge form they reported whether they had actually done so. Deaths were recorded through the first 35 days, whereas reinfarctions, strokes, and major bleeding episodes were recorded only if they occurred before hospital discharge. Mortality percentages are based on all randomized patients (numbers given at top of columns), whereas percentages for all other events are based only on the 99 percent of patients with discharge forms. Data on heparin use include only patients discharged alive for whom discharge forms were available. IV denotes intravenous. Plus–minus values are SD. The methods used in all tables and figures to analyze the results of individual trials and to combine the results from different trials, with appropriate weight given to each trial, are described in detail elsewhere.1-5 †Of this group, 6 percent were given intravenous heparin, and a further 50 percent were given subcutaneous heparin. ‡Of this group, 77 percent were given intravenous heparin, and a further 14 percent were given subcutaneous heparin.

stantial value even when given with heparin (Table 1). The key question now is whether heparin should be added to aspirin. Thus far, a total of about 68,000 patients have been randomly assigned in trials to receive either aspirin plus heparin or aspirin alone; 93 percent of these patients also received fibrinolytic therapy (Table 2).24-30 Any further reductions in the rate of death or of other major vascular events with the heparin regimens in these trials were small (52 deaths prevented per 1000 patients), and there was a small increase in episodes of major bleeding.3 Most of this evidence about adding heparin to aspirin therapy comes from the 62,067 patients who received fibrinolytic therapy in the Second Gruppo Italiano per lo Studio della Sopravvivenza nell’Infarto Miocardico (GISSI-2) trial24,25 and the Third International Study of Infarct Survival (ISIS-3)26 (Table 3). This raises the question whether particular aspects of these two trials led to an underestimate of the possible effects of heparin. In both studies, for example, heparin therapy was begun several hours (12 hours in GISSI-2 and 4 hours in ISIS-3) after the start of any fibrinolytic therapy (streptokinase, tissue plasminogen activator [t-PA], or anistreplase), and the heparin was given subcutaneously (12,500 IU twice daily for

about one week), which caused further delay. During the period of scheduled heparin treatment in these two trials, there was some evidence of a reduction in mortality (2071 deaths [6.8 percent] with heparin, aspirin, and fibrinolytic therapy, as compared with 2239 deaths [7.3 percent] with aspirin and fibrinolytic therapy alone), suggesting the prevention of about 5 deaths per 1000 patients assigned heparin. However, there was no significant effect of heparin on mortality at 35 days (22 fewer deaths per 1000 patients assigned heparin) (Table 3) or at 6 months (13 fewer deaths per 1000 patients).26 Addition of More Intensive Heparin Regimens to Aspirin

The GISSI-2 and ISIS-3 trials studied high-dose subcutaneous heparin regimens, and an intravenous regimen could have produced more intensive anticoagulation. A total of only about 1300 patients with suspected acute myocardial infarction have been studied in trials that compared intravenous heparin (24,000 to 25,000 IU per day for two to three days) plus aspirin with aspirin given alone.27-30 Hence, even in aggregate, the unpromising results of these small trials are inconclusive33,34: there were 32 deaths (4.7 percent) among 676 patients assigned intraveVol ume 336

Numbe r 12



849

The New England Journal of Medi ci ne

TABLE 2. EFFECTS

OF

HEPARIN

WITH OR WITHOUT ASPIRIN IN PATIENTS WITH SUSPECTED ACUTE IN A SYSTEMATIC OVERVIEW OF 26 RANDOMIZED TRIALS.*

TRIALS

EVENT

WITH

NO ROUTINE ASPIRIN†

NO

TRIALS

WITH

MYOCARDIAL INFARCTION

ROUTINE ASPIRIN‡

EFFECT/1000

EFFECT/1000

HEPARIN

ANTITHROMBOTIC

PATIENTS

ONLY

AGENT

ASSIGNED

(N2684)

(N 2775)

P

HEPARIN

VALUE

ASPIRIN PLUS

no. (%)

PATIENTS

HEPARIN

ASPIRIN ONLY

ASSIGNED

(N34,035)

(N 34,055)

P

HEPARIN

VALUE

52 fewer 31 fewer 11 more 10.4 fewer 31 more

0.03

no. (%)

Death

284 (11.4)

378 (14.9)

Reinfarction

142 (6.7)

176 (8.2)

Stroke

23 (1.1)

44 (2.1)

Pulmonary embolism Major bleeding

46 (2.0)

91 (3.8)

31 (1.9)

14 (0.9)

3511 fewer 158 fewer 104 fewer 195 fewer 104 more

0.002

2932 (8.6)

3092 (9.1)

0.08

1009 (3.0)

1103 (3.3)

0.01

397 (1.2)

375 (1.1)

0.001

82 (0.3)

117 (0.4)

0.01

342 (1.0)

234 (0.7)

0.04 0.4 0.01 0.001

*Data are from Collins et al.3 This table should not be used to assess the effects of aspirin or fibrinolytic therapy, because these treatments were not randomly assigned. Deaths were generally recorded only before hospital discharge, and reinfarctions, strokes, pulmonary embolisms, and major bleeding episodes were recorded only if they occurred before hospital discharge. For the six trials (or trial strata) with routine aspirin therapy, data on mortality and nonfatal events were available for 98 to 100 percent of the randomized patients (numbers given at top of columns), except for data on pulmonary embolism, which were available for only about 85 percent. For the 21 trials (or trial strata) with no routine aspirin therapy, data on mortality were available for about 90 percent of all randomized patients; data on reinfarction, stroke, and pulmonary embolism for 75 to 85 percent; and data on major bleeding episodes for only about 60 percent. A few of the trials without routine aspirin therapy had 2:1 randomization, and the control group was counted twice in the present analyses to maintain balance.21 Plus–minus values are SD. †Of the patients in these trials, 14 percent were given fibrinolytic therapy. A total of about 100 additional patients were randomly assigned to treatment in several of these trials, but their treatment-group allocation and outcome are unknown. ‡Of the patients in these trials, 93 percent were given fibrinolytic therapy.

nous heparin plus aspirin, as compared with 31 deaths (4.7 percent) among 659 patients assigned aspirin alone; 23 (3.4 percent) as compared with 22 (3.3 percent) reinfarctions; and 12 (1.8 percent) as compared with 4 (0.6 percent) strokes.3 Among patients with unstable angina, there is also very little evidence of improvement in major clinical outcomes with the addition of intravenous heparin to aspirin, either in the individual trials35-40 or (despite its title) in a formal meta-analysis of those trials.41 Promising results with the addition of subcutaneous lowmolecular-weight heparin to aspirin were recently reported, however.42 A more reliable assessment of intravenous heparin therapy is provided by the large Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries trial (GUSTO-1),31,32 in which — among more than 20,000 patients with suspected acute myocardial infarction who were assigned to therapy with streptokinase — treatment with aspirin plus the subcutaneous heparin regimen used in the ISIS-3 trial was compared with aspirin plus at least 48 hours of intravenous heparin (Table 3). For the patients assigned to receive intravenous heparin, an initial bolus dose of 5000 IU was to be followed by an infusion of 1000 IU per hour, adjusted to aim for an activated partial-thromboplastin time 850 

Mar ch 2 0 , 1 9 9 7

of 60 to 85 seconds.43 In the GUSTO-1 trial, intravenous heparin therapy was not associated with any reduction in mortality or in stroke, and perhaps by chance, there was a conventionally significant excess of reinfarctions associated with intravenous, as compared with subcutaneous, heparin (Table 3). Also, there was no good evidence of any difference in clinical outcome in any subgroup of patients, such as those with anterior infarction, who might have been at particular risk of thrombotic complications. Therefore, despite the small improvements in coronary-artery patency that are noted when intravenous heparin is added to adequate doses of aspirin after treatment with t-PA28 or streptokinase,44 the intravenous heparin regimen studied in the GUSTO-1 trial did not appear to confer any clinical advantage over high-dose subcutaneous heparin (even if allowance is made for some noncompliance with the assigned heparin regimen32) or, indirectly, over aspirin alone (Table 3). Intravenous heparin is, however, associated with a small increase in major bleeding. Bleeding with More Intensive Heparin Regimens

Higher activated partial-thromboplastin times during heparin therapy are associated with higher rates of coronary-artery patency a few days after acute my-

D R UG TH ER A PY

TABLE 3. EFFECTS

OF

ADDING HEPARIN TO ASPIRIN AND FIBRINOLYTIC THERAPY IN PATIENTS WITH SUSPECTED ACUTE MYOCARDIAL INFARCTION IN THE GISSI-2, ISIS-3, AND GUSTO-1 TRIALS.*

GISSI-2

EVENT

AND

ISIS-3†

INDIRECT ASSESSMENT: EFFECT/1000 PATIENTS ASSIGNED INTRAVENOUS HEPARIN IN ADDITION TO ASPIRIN§

GUSTO-1‡

P VALUE

EFFECT/1000 PATIENTS EFFECT/1000

ASPIRIN PLUS HIGH-DOSE

ASSIGNED INTRAVENOUS

PATIENTS

ASPIRIN PLUS

HIGH-DOSE

SUBCUTANEOUS

ASPIRIN

ASSIGNED

INTRAVENOUS

SUBCUTANEOUS

INSTEAD OF

HEPARIN

ONLY

SUBCUTANEOUS

HEPARIN

HEPARIN

SUBCUTANEOUS

(N31,017)

(N 31,050)

P VALUE

(N 10,410)

(N 9841)

P

HEPARIN

HEPARIN

VALUE

0.93.8 more 7.22.7 more 2.01.6 more 1.11.0 more 2.61.6 more

0.8

no. (%)

Death

ASPIRIN PLUS

3100 (10.0) 3172 (10.2)

Reinfarction

927 (3.0)

1010 (3.3)

Any stroke

376 (1.2)

359 (1.2)

Hemorrhagic stroke Major or severe bleeding¶

150 (0.5)

120 (0.4)

312 (1.0)

213 (0.7)

no. (%)

2.22.4 fewer 2.71.4 fewer 0.60.9 more 1.00.5 more 3.20.7 more

0.4

763 (7.4)

712 (7.3)

0.06

438 (4.2)

343 (3.5)

0.5

144 (1.4)

117 (1.2)

0.07

59 (0.6)

45 (0.5)

0.001

151 (1.5)

117 (1.2)

0.01 0.2 0.3 0.1

14 fewer 43 more 32 more 21 more 62 more

0.8 0.2 0.2 0.08 0.001

*Data are from the GISSI-224 and ISIS-326 direct randomized comparisons of aspirin plus high-dose subcutaneous heparin versus aspirin alone and from the GUSTO-131 randomized comparison of aspirin plus intravenous heparin versus aspirin plus high-dose subcutaneous heparin. This table should not be used to make direct nonrandomized comparisons between the absolute event rates in different trials, because the patient populations may have differed substantially in age and other characteristics. By contrast with Table 2, which involved only in-hospital mortality in the GISSI-2 and ISIS-3 trials (and most of the other trials), this table includes deaths recorded throughout the first 35 days for the GISSI-2 and ISIS-3 trials and throughout the first 30 days for the GUSTO-1 trial. Numbers randomized (given at the top of columns) and numbers with follow-up are from the ISIS-3 report26 and GUSTO-131 (supplemented with revised GUSTO-1 data from the National Auxiliary Publications Service), and numbers with events and the percentages (based on patients with follow-up) are from the ISIS-3 report26 and Van de Werf et al.32 Plus–minus values are SD. †Only the patients in this trial who were randomly assigned to fibrinolytic therapy with streptokinase, t-PA, or anistreplase are included. ‡Only the patients in this trial who were randomly assigned to receive streptokinase are included. §Effects were assessed indirectly by combining the results of the directly randomized comparison within GISSI-2 and ISIS-3 with those of the directly randomized comparison within GUSTO-1, with appropriate weight based on the sizes of the studies given to each comparison.5 ¶Data from ISIS-3 and GISSI-2 involved “major” bleeding episodes, and data from GUSTO-1 involved “severe” bleeding episodes.

ocardial infarction.45-47 Hence, a somewhat more intensive intravenous heparin regimen was initially studied in one arm of the Global Use of Strategies to Open Occluded Coronary Arteries trial (GUSTO-II) — 1300 IU per hour for patients weighing 80 kg or more, with the upper boundary of the target range for the activated partial-thromboplastin time increased to 90 seconds48 — and in two other studies.49,50 Although an average of only about 20 percent more heparin was given, all three trials were stopped prematurely because of intracerebral hemorrhage and other major bleeding with the high-dose heparin regimen (and with the direct thrombin inhibitors) in patients given fibrinolytic therapy.48-50 This suggests that despite increased rates of coronary-artery patency, more intensive anticoagulation with heparin (or, perhaps, with other antithrombotic

drugs) may not be an appropriate strategy, at least in conjunction with fibrinolytic therapy. Avoidance of Routine Heparin Therapy in Acute Myocardial Infarction

In the relevant comparison of aspirin plus heparin with aspirin alone, there is little evidence of any additional clinical advantage with subcutaneous or with intravenous heparin (Tables 2 and 3), at least among patients with suspected acute myocardial infarction receiving streptokinase, t-PA, or other fibrinolytic agents. Moreover, if heparin therapy is made much more intensive then the incidence of cerebral hemorrhage and other major bleeding rises appreciably. Further large trials of the effects of adding heparin or newer antithrombotic agents to aspirin, especially among patients not receiving fibrinolytic Vol ume 336

Numbe r 12



851

Lives Saved/1000 Patients

The New England Journal of Medi ci ne

40

3000

Loss of benefit per hour of delay: 1.6  0.6 lives per 1000 patients

14,000

30

12,000 9000

20

10 7000

0 0

6

12

18

24

Hours from Onset of Symptoms to Randomization Figure 2. Absolute Number of Lives Saved at One Month per 1000 Patients Treated with Fibrinolytic Therapy, Plotted against the Time from the Onset of Symptoms to Randomization among 45,000 Patients with ST-Segment Elevation or Bundle-Branch Block. This figure from the FTT overview 4 includes data from all nine randomized trials with more than 1000 patients that compared patients given fibrinolytic therapy and controls (with all patients from the Anglo-Scandinavian Study of Early Thrombolysis55 and Late Assessment of Thrombolytic Efficacy 58 trials included, irrespective of ECG changes). For patients whose delay times were recorded as 0 to 1, 2 to 3, 4 to 6, 7 to 12, and 13 to 24 hours, the absolute benefit (SD) in lives saved per 1000 patients treated is plotted against the mean recorded delay time for each category (1.0, 2.5, 4.8, 9.1, and 17.5 hours, respectively). The area of each black square and the extent to which it influences the line drawn through the five points are approximately proportional to the number of patients in the category on which it is based (with these numbers shown in the figure).

apy Trialists’ Collaborative Group (FTT) found that fibrinolytic therapy is beneficial for a wide range of patients with suspected acute myocardial infarction. In particular, fibrinolytic therapy reduced mortality in patients presenting with either ST-segment elevation or bundle-branch block up to at least 12 hours after the onset of symptoms (see below). For patients with ST-segment elevation, the proportional reductions in mortality were similar regardless of whether the location of the infarct was anterior, inferior, or other.4 Thus, although the absolute benefits appear to be greatest among patients with anterior ST-segment elevation, there is good evidence that the patients with inferior ST-segment elevation, who are at somewhat lower risk, also benefit. It is not yet clear whether fibrinolytic therapy produces worthwhile benefits among certain other types of patients at appreciable risk for early death from cardiac causes who present with suspected acute myocardial infarction but with other abnormalities on their electrocardiograms (ECGs), such as ST-segment depression alone.4 Among these patients at medium or high risk, further trials of fibrinolytic therapy versus control that are large enough to assess directly the effects on mortality are therefore still needed. On the other hand, patients with suspected acute myocardial infarction who present with fairly normal ECG findings are usually at such low cardiac risk that any reduction in deaths from cardiac causes may not outweigh the hazards of fibrinolytic therapy (notably, early hemorrhagic stroke). Such low-risk patients might best be treated in the hospital by monitoring their ECGs frequently and giving them fibrinolytic therapy only if ST-segment elevation or bundle-branch block develops. Delay from Onset of Symptoms to Fibrinolytic Therapy

therapy, could provide valuable information about the balance of risks and benefits.51 The currently available evidence with respect to heparin is, however, already extensive (Tables 2 and 3), and it does not provide any clear justification for routine heparin therapy in patients with suspected acute myocardial infarction (or unstable angina). FIBRINOLYTIC THERAPY Benefits of Fibrinolytic Therapy in Patients with ST-Segment Elevation or Bundle-Branch Block

Several randomized trials comparing fibrinolytic therapy with no fibrinolytic therapy in patients with suspected acute myocardial infarction have been large enough on their own to demonstrate that mortality during the first month or so can be reduced substantially by such treatment (see, for example, Fig. 1) and that the survival benefit persists for at least several years.6 The recent overview 4 of five-week mortality in the nine largest trials1,26,52-58 by the Fibrinolytic Ther852 

Mar ch 2 0 , 1 9 9 7

The earlier that fibrinolytic therapy is started after the onset of symptoms the greater the benefit. However, in the FTT overview of the nine largest trials, involving 46,000 patients, the slope of absolute benefit plotted against increasing delay is fairly gradual and is not significantly steeper in the first few hours after the onset of symptoms than in subsequent hours4 (Fig. 2). A retrospective analysis of the results of the Gruppo Italiano per lo Studio della Streptochinasi nell’Infarto Miocardico (GISSI-1) trial suggested that fibrinolytic therapy might be especially effective if started within one hour after the onset of symptoms52 — the so-called golden hour.59 However, this suggestion (and similar claims based largely on data-dependent selective emphasis on particular small trials59) is not supported by the results among patients with ST-segment elevation or bundle-branch block in a meta-analysis of all the large trials.4 Indeed, if the patients in the “hypothesisgenerating” GISSI-1 trial are excluded from the analysis, then the apparent benefit in the patients

D R UG TH ER A PY

who underwent randomization within one hour after the onset of symptoms is slightly less than that shown in Figure 2. Overall, among patients with ST-segment elevation or bundle-branch block, fibrinolytic therapy produces a definite reduction in mortality not only among those presenting within 6 hours after the onset of symptoms (saving about 30 lives per 1000 patients treated) but also among those presenting from 7 to 12 hours after onset (saving about 20 lives per 1000 patients treated). Thus, each hour of delay recorded was associated with a mean (SD) reduction in the number of lives saved of about 1.60.6 per 1000 patients treated. This estimate may have been somewhat reduced by inaccuracies in assessing the delays,60 so the real effect of each additional hour of delay may be slightly greater, involving perhaps 2 (or even 3) extra deaths per 1000 patients treated. In principle, the trials comparing prehospital with in-hospital fibrinolytic therapy 61-65 might provide direct evidence of the importance of an extra hour of delay, but even in aggregate these trials are far too small66 to measure reliably differences of only a few deaths per 1000 patients treated. Too few patients presenting more than 12 hours after the onset of symptoms have been studied to determine whether the benefits of fibrinolytic therapy in such patients outweigh the risks. However, the general pattern shown in Figure 2 strongly suggests that at least among patients with ST-segment elevation or bundle-branch block who are treated 12 to 18 hours after onset, there may be some net benefit (perhaps about 10 lives saved per 1000 patients treated). The FTT overview indicates that this gradual diminution in the size of the benefit may be due to the association of the lateness of starting fibrinolytic treatment with the size of the excess of deaths on the day treatment is given and on the following day, whereas the subsequent reduction in mortality with fibrinolytic therapy appears to be little affected by the time when patients are treated.4 The causes of this so-called early hazard are unclear.4,67,68 But, if it could be reduced by some adjuvant therapy, then the benefits of fibrinolytic therapy might be increased substantially, not just among patients treated late after the onset of symptoms, but also among those treated early. Benefits in Elderly and Other High-Risk Patients

The FTT overview of the large trials of fibrinolytic therapy provides no evidence for withholding such treatment on the basis of age from patients with suspected acute myocardial infarction.4 The absolute survival advantages at five weeks with the use of fibrinolytic therapy among patients with ST-segment elevation or bundle-branch block were similar in young and old patients (and persisted over the long term6): 154 lives saved per 1000 treated patients

less than 55 years of age, 215 lives saved per 1000 patients 55 to 64 years of age, 376 lives saved per 1000 patients 65 to 74 years of age, and 1314 lives saved per 1000 patients 75 years of age or older. Although the absolute excess of deaths on the day of treatment and the following day with fibrinolytic therapy increased with age,4 so too did the subsequent absolute reduction in deaths. Hence, age alone should not be considered a contraindication to fibrinolytic treatment. It has been suggested that fibrinolytic therapy may be relatively ineffective in high-risk patients, such as those with cardiogenic shock, and that primary coronary angioplasty should be considered instead.69-71 However, the available evidence indicates that fibrinolytic therapy is especially effective among high-risk patients such as those presenting with hypotension or tachycardia, many of whom have heart failure or cardiogenic shock.4 In the FTT overview, among patients with ST-segment elevation or bundle-branch block whose initial systolic blood pressure was below 100 mm Hg, 6421 deaths were averted per 1000 patients given fibrinolytic therapy (29.7 percent dead with fibrinolytic therapy, as compared with 36.1 percent dead in the control group; P0.002), and among those whose heart rate was above 100 beats per minute, 4912 deaths per 1000 were averted (18.9 percent dead, as compared with 23.8 percent dead; P0.001). In the GISSI-1 trial, which was the only large trial to record cardiogenic shock and heart failure at entry, the proportional reduction in mortality with streptokinase among patients with either heart failure or shock appeared to be similar to that found in the overall study — again suggesting substantial benefit.52 Hence, fibrinolytic therapy is of demonstrated value, whereas the evidence on primary angioplasty is limited to inconclusive results from small trials among mostly low-risk patients72 (including a small substudy in the GUSTO-II trial73), and no appropriately largescale studies are currently being conducted. Recommendations for Routine Fibrinolytic Therapy

Despite the strength of the available evidence, many eligible patients who present with suspected acute myocardial infarction still do not receive fibrinolytic therapy.16,74,75 For example, in a recent survey about four fifths of Medicare patients presenting with acute myocardial infarction did not receive fibrinolytic therapy, and even among the patients considered by the survey’s authors to be ideal candidates for fibrinolytic therapy, about one third did not receive it (moreover, the definition of ideal candidates inappropriately excluded those presenting 7 to 12 hours after the onset of symptoms).16 Similarly, an extensive survey conducted throughout Europe in 1993 and 1994 found that more than one third of patients in whom fibrinolytic therapy was clearly indicated did Vol ume 336

Numbe r 12



853

The New England Journal of Medi ci ne

TABLE 4. DIRECT RANDOMIZED COMPARISONS OF THE STANDARD STREPTOKINASE REGIMEN WITH VARIOUS t-PA–BASED FIBRINOLYTIC REGIMENS IN PATIENTS WITH SUSPECTED ACUTE MYOCARDIAL INFARCTION IN THE GISSI-2, ISIS-3, AND GUSTO-1 TRIALS.*

TRIAL

AND

TREATMENT

NO. OF PATIENTS

ANY

RANDOMIZED

STROKE

ANY DEATH

DEATH NOT RELATED TO STROKE†

STROKE OR DEATH

number (percent)

GISSI-2‡ Streptokinase t-PA Effect/1000 patients treated with t-PA instead of streptokinase ISIS-3§ Streptokinase t-PA Effect/1000 patients treated with t-PA instead of streptokinase GUSTO-1¶ Streptokinase (subcutaneous heparin) Streptokinase (intravenous heparin) t-PA alone t-PA plus streptokinase Effect/1000 patients treated with t-PA–based regimens instead of streptokinase x 22 heterogeneity of effects between 3 trials P value Weighted average of all 3 trials Effect/1000 patients treated with t-PA–based regimens instead of streptokinase P value

10,396 10,372

98 (0.9) 136 (1.3) 3.71.5 more

958 (9.2) 993 (9.6) 3.64.0 more

13,780 13,746

141 (1.0) 188 (1.4) 3.51.3 more

1455 (10.6) 1418 (10.3) 2.43.7 fewer

9,841 10,410 10,396 10,374

117 (1.2) 144 (1.4) 161 (1.6) 170 (1.6) 3.01.2 more

712 (7.3) 763 (7.4) 653 (6.3) 723 (7.0) 6.62.5 fewer

0.7 0.3 3.30.8 more 0.001

5.6 0.06 2.91.9 fewer 0.1

916 (8.8) 931 (9.0) 1.74.0 more

1014 (9.8) 1067 (10.3) 5.34.2 more

1389 (10.1) 1530 (11.1) 1325 (9.6) 1513 (11.0) 4.43.6 1.03.8 fewer fewer 666 (6.8) 709 (6.8) 585 (5.6) 647 (6.2) 8.62.4 fewer

783 (8.0) 853 (8.2) 746 (7.2) 817 (7.9) 5.52.6 fewer

7.0 0.03

5.4 0.07

4.91.8 fewer 0.01

1.61.9 fewer 0.4

*This table should not be used to make direct nonrandomized comparisons between the absolute event rates in different trials, because the patient populations may have differed substantially in age and other characteristics. Deaths recorded throughout the first 35 days are included for GISSI-2 and ISIS-3 and throughout the first 30 days for GUSTO-1. Numbers randomized and numbers with follow-up are from the ISIS-3 report26 and GUSTO-131 (supplemented with revised GUSTO-1 data from the National Auxiliary Publications Service), and numbers with events and the percentages (based on patients with follow-up) are from the ISIS-3 report26 and Van de Werf et al.32 Plus–minus values are SD. In all three trials, streptokinase was given in an intravenous infusion of 1.5 million units over a period of one hour. †Death not related to stroke was defined as death without recorded stroke. ‡In the GISSI-2 trial, the t-PA regimen involved an initial bolus of 10 mg, followed by 50 mg in the first hour and 20 mg in each of the second and third hours. §In the ISIS-3 trial, the t-PA regimen involved 40,000 clot-lysis units per kilogram of body weight as an initial bolus, followed by 360,000 units per kilogram in the first hour and 67,000 units per kilogram in each of the next three hours. ¶In the GUSTO-1 trial, the t-PA alone regimen involved an initial bolus of 15 mg, followed by 0.75 mg per kilogram (up to 50 mg) in the first 30 minutes and 0.5 mg per kilogram (up to 35 mg) in the next hour; in the GUSTO-1 trial the other t-PA–based regimen involved 0.1 mg of t-PA per kilogram (up to 9 mg) as an initial bolus and 0.9 mg per kilogram (up to 81 mg) in the remainder of the first hour, plus 1 million units of streptokinase in the first hour. The weights are proportional to the sample sizes of the trials, so this average gives most weight to the GUSTO-1 trial and least to the GISSI-2 trial.5

not receive it (with the underuse of therapy most marked in women and older patients).74 Remaining uncertainties about fibrinolytic therapy should not obscure the definite evidence of substantial benefit — irrespective of the patient’s age or sex, the site of infarction, blood pressure, heart rate, the presence of heart failure, or a history of myocardial infarction or diabetes — in patients who present with ST-segment elevation or bundle-branch block up to at least 12 hours after the onset of symptoms.4 Among these patients, fibrinolytic therapy typically prevents at least 20 to 30 deaths per 1000 patients 854 

Mar ch 2 0 , 1 9 9 7

treated, with the greatest benefits in patients treated earlier and, generally, in those at higher risk. By contrast, fibrinolytic therapy was associated in the large trials with only about 4 extra strokes per 1000 patients treated, of which 2 strokes were associated with early death (and so were already accounted for in the reduction of overall mortality), 1 was moderately or severely disabling, and 1 was not. Hence, all such patients should be routinely considered for fibrinolytic therapy, and as with aspirin, it should be withheld only if there is some extremely clear contraindication (such as a very high risk of intracerebral bleeding).

D R UG TH ER A PY

DIFFERENT FIBRINOLYTIC REGIMENS Coronary-Artery Patency versus Cerebral Hemorrhage

The first large trials of fibrinolytic therapy involved the administration of 1.5 million units of streptokinase in one hour. This regimen takes a median of about 90 minutes from the start of treatment to open the occluded coronary arteries.26 More intensive fibrinolytic regimens, generally based on the administration of t-PA, do not significantly increase the proportion of arteries opened within the first few hours, but their action is faster.26,44,76 The exact reduction in the length of time to patency is not known, but the median figure of 90 minutes with streptokinase cannot be reduced by more than about 60 minutes. These transient differences in early patency have not been shown to produce differences in ventricular function.44,76 Other things being equal, however, opening the arteries about 30 to 60 minutes earlier should be

TRIAL AND TREATMENT

STROKE OR DEATH

a good thing; Figure 2 indicates that it might prevent 1 or 2 deaths from cardiac causes per 1000 patients treated. Other things are not equal, however, because more intensive fibrinolytic regimens are associated with a few extra strokes per 1000 patients. The fundamental question is whether any cardiovascular advantages of more intensive fibrinolytic regimens outweigh any cerebrovascular disadvantages. Because the advantages and disadvantages are both likely to be small (probably involving at most only a few events per 1000 patients treated), any difference between them is likely to be even smaller and hence difficult to measure reliably. Evidence from Large, Randomized Trials and Avoidance of Selective Emphasis

Two main requirements must be met if differences of only a few adverse events per 1000 treated pa-

EVENTS/1000 PATIENTS TREATED (AND 99% CI) RELATIVE TO WITHIN-TRIAL AVERAGE

EFFECT/1000 PATIENTS TREATED WITH t-PA INSTEAD OF STREPTOKINASE (SD)

no. of events/ no. of patients (%)

GISSI-2 SK t-PA

1014/10,396 (9.8) 1067/10,372 (10.3)

5.3  4.2 more (P  0.2)

ISIS-3 SK t-PA

1530/13,780 (11.1) 1513/13,746 (11.0)

1.0  3.8 fewer (P  0.8)

GUSTO-1 SK (SC heparin) SK (IV heparin) t-PA alone t-PA  SK

783/9841 (8.0) 853/10,410 (8.2) 746/10,396 (7.2) 817/10,374 (7.9) All 3 trials (weighted averages) (9.4) SK-only regimens (9.2) t-PA– based regimens Heterogeneity tests between: All 4 groups in GUSTO-1: x32  8.3 (P  0.04) 2 t-PA groups in GUSTO-1: x12 3.7 (P not significant) GISSI-2, ISIS-3, and GUSTO-1: x22 5.4 (P not significant)

5.5  2.6 fewer

(P  0.04)

1.6  1.9 fewer

(P  0.4)

10 5 Fewer events per 1000 than within-trial average

0

5 10 More events per 1000 than within-trial average

Figure 3. Stroke or Death in the Three Large, Directly Randomized Comparisons (GISSI-2, ISIS-3, and GUSTO-1) of the Standard Streptokinase Regimen with More Intensive t-PA–Based Fibrinolytic Regimens. For each treatment group the number of events per 1000 patients treated (with its 99 percent confidence interval [CI]) is plotted, after subtraction of the overall average number of events per 1000 patients treated in that trial. Solid circles denote streptokinase-only regimens; open symbols denote t-PA–based regimens. (Subtraction of the overall within-trial risk does not affect the differences between different groups in one trial, but merely centers the results for each trial on the same vertical line.) The weighted average of the results from the three trials has weights proportional to the number of patients in those trials.5 SK denotes streptokinase, SC subcutaneous, and IV intravenous.

Vol ume 336

Numbe r 12



855

The New England Journal of Medi ci ne

tients are to be reliably demonstrated or disproved. First, there need to be large-scale, randomized trials, and second, the results need to be interpreted as a whole, without unduly selective emphasis on parts of the evidence that suit particular conclusions.5,23,77,78 There have been three large trials comparing the standard one-hour streptokinase regimen of 1.5 million units with t-PA–based regimens (the GISSI2,24,25,79 ISIS-3,26 and GUSTO-131,32,80 trials); stroke and death were the two main clinical end points (Table 4 and Fig. 3). The patients were randomized an average of about two to three hours after the onset of symptoms in the GISSI-2 trial, four hours after onset in the ISIS-3 trial, and two hours after onset in the GUSTO-1 trial. In each trial, the t-PA–based regimens were designed to ensure appreciably better coronary-artery patency 90 minutes after the start of treatment than the standard streptokinase regimen with which they were compared, and the accompanying dose of aspirin was sufficiently large to contribute substantially to the maintenance of early patency. Moreover, although one of the two t-PAbased regimens in the GUSTO-1 trial is often referred to as “front-loaded” or “accelerated,” the two t-PA–based regimens studied in that trial were actually very similar to each other in terms of the total dose of t-PA given in the first hour. Hence, it is appropriate to consider the results from all of the regimens in all three of these trials together. Such an emphasis on all the relevant results is important and is the chief way in which this review differs from most other commentaries on the same evidence. Risks of Stroke with Different Fibrinolytic Regimens

An increased risk of stroke was found with the t-PA–based fibrinolytic regimens in GISSI-2, ISIS-3, and GUSTO-1 (Table 4). Taking all three trials together, there was a significant excess of 3.30.8 strokes per 1000 patients treated with t-PA (P0.001), with an excess of similar size in each trial. Most of this excess risk occurred in the first day or so after the administration of t-PA and was attributed to an even more definite excess of 2.90.5 cerebral hemorrhages per 1000 patients treated with t-PA (95 percent confidence interval, 2 to 4; P0.001). The excess risk associated with these t-PA–based regimens increased substantially with older age and with higher blood pressure.26,79,80 A similar excess of 4 cerebral hemorrhages per 1000 patients treated, but no significant survival advantage, was observed with reteplase, another recombinant plasminogen activator that produces high rates of early patency, in a 6000-patient comparison with streptokinase.81 Deaths Not Related to Stroke

For mortality from all causes in the GISSI-2, ISIS-3, and GUSTO-1 trials, the evidence of heter856 

Mar ch 2 0 , 1 9 9 7

ogeneity between the treatment effects was not statistically significant, and the weighted average of the difference in effect (2.91.9 fewer deaths per 1000 patients treated with t-PA as compared with streptokinase) in all three trials was not significantly different from zero (P0.1) (Table 4). Mortality from all causes includes the excess of early deaths after stroke that occurred with the t-PA regimens in each of these three large, directly randomized comparisons (an average excess of about 2 deaths after stroke per 1000 patients treated with t-PA). A more sensitive assessment of any cardiac benefits of the t-PA regimens may be obtained by considering deaths not related to stroke (i.e., deaths of patients in whom no stroke was reported) separately from deaths after stroke (most of which were probably due to the stroke). There is some indication of heterogeneity of the results for deaths not related to stroke in the three different trials (P0.03) (Table 4). Overall, the t-PA–based regimens were associated with 4.91.8 fewer deaths not related to stroke per 1000 patients treated, but the 95 percent confidence interval for this estimate spans a wide range, from about 1 to about 9 fewer such deaths per 1000 patients treated. (This overall result remains about the same when the data from ISIS-3 are restricted to those from patients randomized within six hours after the onset of symptoms,24 as in the GISSI-2 and GUSTO-1 trials.) Each of the t-PA–based regimens studied in these three trials probably opened occluded coronary arteries about 30 to 60 minutes earlier than did the standard streptokinase regimen.44,76 This would change a delay of about three hours between the onset of symptoms and patency into a delay of about two and a half to two hours. Figure 2 indicates that this change is likely to reduce mortality from cardiac causes by only about 1 or 2 deaths per 1000 patients treated (irrespective of what the exact relation between benefit and delay until treatment may be during the “golden hour” after the onset of symptoms). Hence, the true cardiac benefit of these t-PA–based regimens may well be near the lower end of the 95 percent confidence interval for the overall result presented in Table 4 — that is, only a few less deaths not related to stroke per 1000 patients treated. Lack of Difference in Net Clinical Outcome with Different Fibrinolytic Regimens

The analyses of stroke or death in Figure 3 and Table 4 avoid any double counting of patients who had a stroke and also died. These directly randomized comparisons of streptokinase and t-PA again indicate no significant heterogeneity between the results in the different trials and no significant difference in net clinical outcome (defined as the occurrence of stroke or death). Taken together, these results suggest that t-PA–based regimens might con-

D R UG TH ER A PY

fer a nonsignificant improvement in net clinical outcome of only 1 or 2 events per 1000 patients. However, the extra hazard with these t-PA regimens is definite (about 3 additional cerebral hemorrhages per 1000 patients treated, P0.001), whereas any excess of benefit over hazard is uncertain (Fig. 3). Indeed, when the definite hazard of 3 extra strokes per 1000 patients treated with t-PA is taken together with the indirect evidence (Fig. 2) of what slightly earlier fibrinolytic treatment can achieve, this suggests a nonsignificant excess of hazard over benefit. Hence, consideration of all the evidence does not demonstrate any clear differences in net clinical outcome between these different fibrinolytic regimens. GENERAL PROBLEMS OF UNDULY SELECTIVE EMPHASIS

Even if the overall evidence on a particular therapeutic question is significant, unduly selective emphasis on particular parts of the evidence may well generate false negative results.5,77,78 In ISIS-2, for example, the overall results indicated that aspirin substantially reduces mortality (Fig. 1), but selective emphasis on just those patients born under the astrological sign of Libra or Gemini misleadingly suggests that for those patients, aspirin is of no benefit.1 Likewise, in the trials comparing fibrinolytic therapy with control, the overall results strongly favor fibrinolytic therapy, so the apparent lack of benefit in subgroups with certain findings on their ECGs (e.g., ST-segment depression) might well be misleading.4 Conversely, if the overall evidence is not clearly significant, then selective emphasis on particular results could easily generate false positive results.5,77,78 In the three main trials comparing one fibrinolytic regimen with another, the overall results did not clearly favor any particular treatment (Fig. 3). So, although the results of GISSI-2 appear to favor streptokinase and those of GUSTO-1 do not, it may be misleading to place much emphasis on either one of these two observations — or, still more so, on the observations in just one of the two t-PA–based treatment groups in GUSTO-1. Once an unexpected result has been noticed in a particular trial or subgroup, retrospective explanations for it may be proposed. However, such post hoc explanations can be proposed for almost any apparent finding and so should be viewed fairly skeptically. For example, when trials comparing different fibrinolytic regimens were first being planned, it was widely expected that therapy with fibrin-specific agents, such as t-PA, would result in fewer strokes and substantially fewer deaths from cardiac causes than would streptokinase therapy. The results of GISSI-2, however, indicated that these expectations of a much better net clinical outcome with the t-PA regimen approved for use at that time were not correct, and this finding was reinforced by the re-

sults of ISIS-3. It was then proposed that the failure of t-PA in these two trials might be due to the lack of use of intravenous heparin.82-87 However, these proposed explanations involved inappropriate nonrandomized comparisons between the rates of mortality and stroke associated with t-PA in the GISSI-2 and ISIS-3 trials and the very much lower rates of these events in selected small trials of t-PA in which event rates in the placebo group were also low.88 By contrast, the results from the large, randomized trials shown in Tables 3 and 4 reveal that stroke is more common with all the t-PA regimens studied than with the standard streptokinase regimen, and that there is little difference between the clinical effects of intravenous heparin and those of high-dose subcutaneous heparin (irrespective of whether streptokinase or t-PA is given). Before the results of GUSTO-1 were available, quantitative calculations had suggested that the regimen involving t-PA alone might not result in a measurable improvement in outcome89 and that the addition to such a regimen of an agent that was not fibrin-specific, such as streptokinase, would provide a more intensive fibrinolytic regimen, and should therefore result in a better outcome (at least, if strokes were ignored).90,91 Since the results of GUSTO-1 became available, however, only the findings with the regimen involving t-PA alone have been emphasized,31,92-94 and the less promising findings with the regimen involving t-PA plus streptokinase (Fig. 3) have been ignored. The justification often given for this selective emphasis is that in a substudy of infarct-related coronary-artery patency among just a small percentage of the patients in GUSTO-1, the regimen involving t-PA alone appeared to produce somewhat better patency at 90 minutes after the start of treatment than did the regimen involving t-PA plus streptokinase.44 Within the first hour after the start of treatment, however, the scheduled average doses of t-PA in these two regimens were actually very similar: 82 mg in the regimen involving t-PA alone and 78 mg in the other t-PA–based regimen.31 If, as the trial protocol originally envisaged, emphasis had been placed not just on grade 3 patency (i.e., an open artery with normal flow) but on grade 2 or 3 patency (i.e., an open artery with delayed or normal flow) then the apparent difference in patency at 90 minutes would have been smaller and less convincing.44 Moreover, irrespective of which definition of patency is emphasized, the measurements at 180 minutes show a nonsignificantly opposite pattern, indicating slightly worse patency with the regimen involving t-PA alone.44 Thus, neither these measurements of patency nor the first 60 minutes of drug dosage can justify selective emphasis on the regimen involving t-PA alone in GUSTO-1. The view that therapy with t-PA, if used in some Vol ume 336

Numbe r 12



857

The New England Journal of Medi ci ne

special way, might produce as many as 10 more survivors per 1000 patients treated than streptokinase31,92-94 involves selective emphasis on just one of the three large, directly randomized comparisons between the two agents and, within that trial, on just one of the two t-PA–based regimens studied (despite the lack of significant difference between these two regimens in net clinical outcome) (Fig. 3). Moreover, this view ignores the biologic implausibility of so slight a difference in the length of time to patency causing such a difference in outcome; Figure 2 indicates that it would take a delay of several hours for the cardiac benefits from fibrinolytic therapy to be reduced by as much as 10 deaths per 1000 patients treated, whereas the difference between t-PA and streptokinase in the time to patency is typically less than one hour. On the basis of all the available evidence from large trials (Fig. 2 and 3 and Tables 3 and 4), the true difference in stroke-free survival between one fibrinolytic regimen and another is likely to be, at most, only a very few events more or less per 1000 patients treated. Moreover, the excess risk of stroke if streptokinase is replaced by t-PA is definite, whereas any excess of benefit over risk is not definite and is likely to be close to zero. (The fact that there is no significant difference in net clinical outcome between streptokinase and t-PA if all the relevant trial results are considered together invalidates health economic calculations of whether the extra costs of t-PA are justified.95) Hence, on the principle of doing no harm unless one can be reasonably sure of doing more good than harm,96 the standard one-hour regimen of 1.5 million units of streptokinase would generally be the fibrinolytic treatment of choice (except for patients who have had streptokinase before, in whom persisting antibodies may make it less effective97). IMPLICATIONS FOR CLINICAL PRACTICE

Aspirin should be started immediately in all patients with suspected acute myocardial infarction or unstable angina, unless some particularly strong contraindication exists (in which case a different antiplatelet drug should be considered), and should generally be continued not only throughout the hospital stay but also for some years after discharge.1,2 In addition, patients who present with ST-segment elevation or bundle-branch block up to at least 12 hours after the onset of the symptoms of acute myocardial infarction (regardless of age, site of infarction, or blood pressure) should generally be given fibrinolytic therapy as soon as practicable, unless there is a very definite contraindication.4 Patients who present with no changes, or only minor ones, on their ECGs can be treated with aspirin and monitored, with fibrinolytic therapy given only if major changes develop. The choice of fibrinolytic regimen appears to 858 

Mar ch 2 0 , 1 9 9 7

make little difference to the overall probability of stroke-free survival, because the regimens that dissolve coronary thrombi more rapidly produce greater risks of cerebral hemorrhage.24-26,31 Nor is there good evidence that, among patients who are given aspirin and fibrinolytic therapy, the routine addition of either intravenous or subcutaneous heparin produces any worthwhile improvement in outcome.3 We have discussed at some length the comparison of one fibrinolytic regimen with another. This is not because the choice of fibrinolytic regimen is medically important, but because the perception that it is important might discourage the use of fibrinolytic therapy by those who fear the increase in cerebral hemorrhage (or other costs) of the more intensive regimens. It is, however, important that any uncertainties about which fibrinolytic regimen or dose of aspirin to use do not engender uncertainty about whether to use fibrinolytic and antiplatelet therapies routinely. REFERENCES 1. ISIS-2 (Second International Study of Infarct Survival) Collaborative Group. Randomised trial of intravenous streptokinase, oral aspirin, both, or neither among 17 187 cases of suspected acute myocardial infarction: ISIS-2. Lancet 1988;2:349-60. 2. Antiplatelet Trialists’ Collaboration. Collaborative overview of randomised trials of antiplatelet therapy. I. Prevention of death, myocardial infarction, and stroke by prolonged antiplatelet therapy in various categories of patients. BMJ 1994;308:81-106. [Erratum, BMJ 1994;308:1540.] 3. Collins R, MacMahon S, Flather M, et al. Clinical effects of anticoagulant therapy in suspected acute myocardial infarction: systematic overview of randomised trials. BMJ 1996;313:652-9. 4. Fibrinolytic Therapy Trialists’ (FTT) Collaborative Group. Indications for fibrinolytic therapy in suspected acute myocardial infarction: collaborative overview of early mortality and major morbidity results from all randomised trials of more than 1000 patients. Lancet 1994;343:311-22. [Erratum, Lancet 1994;343:742.] 5. Peto R, Pike MC, Armitage P, et al. Design and analysis of randomized clinical trials requiring prolonged observation of each patient. II. Analysis and examples. Br J Cancer 1977;35:1-39. 6. Baigent C, Collins R. ISIS-2: 4-year mortality follow-up of 17,187 patients after fibrinolytic and antiplatelet therapy in suspected acute myocardial infarction. Circulation 1993;88:Suppl I:I-291. abstract. 7. Anderson HV, Willerson JT. Thrombolysis in acute myocardial infarction. N Engl J Med 1993;329:703-9. 8. Fitzgerald DJ, Catella F, Roy L, FitzGerald GA. Marked platelet activation in vivo after intravenous streptokinase in patients with acute myocardial infarction. Circulation 1988;77:142-50. 9. Kerins DM, Roy L, FitzGerald GA, Fitzgerald DJ. Platelet and vascular function during coronary thrombolysis with tissue-type plasminogen activator. Circulation 1989;80:1718-25. 10. Fitzgerald DJ, Wright F, FitzGerald GA. Increased thromboxane biosynthesis during coronary thrombolysis: evidence that platelet activation and thromboxane A2 modulate the response to tissue-type plasminogen activator in vivo. Circ Res 1989;65:83-94. 11. Patrono C. Aspirin as an antiplatelet drug. N Engl J Med 1994;330: 1287-94. 12. Patrignani P, Filabozzi P, Patrono C. Selective cumulative inhibition of platelet thromboxane production by low-dose aspirin in healthy subjects. J Clin Invest 1982;69:1366-72. 13. Reilly IAG, FitzGerald GA. Inhibition of thromboxane formation in vivo and ex vivo: implications for therapy with platelet inhibitory drugs. Blood 1987;69:180-6. 14. Clarke RJ, Mayo G, Price P, FitzGerald GA. Suppression of thromboxane A2 but not of systemic prostacyclin by controlled-release aspirin. N Engl J Med 1991;325:1137-41. 15. Berglund U, Wallentin L. Persistent inhibition of platelet function during long-term treatment with 75 mg acetylsalicylic acid daily in men with unstable coronary artery disease. Eur Heart J 1991;12:428-33. 16. Ellerbeck EF, Jencks SF, Radford MJ, et al. Quality of care for

D R UG TH ER A PY

Medicare patients with acute myocardial infarction: a four-state pilot study from the Cooperative Cardiovascular Project. JAMA 1995;273: 1509-14. 17. Scandinavian Simvastatin Survival Study Group. Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S). Lancet 1994;344:13839. 18. Lamas GA, Pfeffer MA, Hamm P, Wertheimer J, Rouleau J-L, Braunwald E. Do the results of randomized clinical trials of cardiovascular drugs influence medical practice? N Engl J Med 1992;327:241-7. 19. Smith P, Arnesen H, Holme I. The effect of warfarin on mortality and reinfarction after myocardial infarction. N Engl J Med 1990;323:14752. 20. Anticoagulants in the Secondary Prevention of Events in Coronary Thrombosis (ASPECT) Research Group. Effect of long-term oral anticoagulant treatment on mortality and cardiovascular morbidity after myocardial infarction. Lancet 1994;343:499-503. 21. Meijer A, Verheugt FWA, Werter CPJ, Lie KI, van der Pol JMJ, van Eenige MJ. Aspirin versus coumadin in the prevention of reocclusion and recurrent ischemia after successful thrombolysis: a prospective placebo-controlled angiographic study: results of the APRICOT Study. Circulation 1993;87:1524-30. 22. Cairns JA. Oral anticoagulants or aspirin after myocardial infarction? Lancet 1994;343:497-8. 23. Yusuf S, Collins R, Peto R. Why do we need some large, simple, randomized trials? Stat Med 1984;3:409-20. 24. Gruppo Italiano per lo Studio della Sopravvivenza nell’Infarto Miocardico. GISSI-2: a factorial randomised trial of alteplase versus streptokinase and heparin versus no heparin among 12 490 patients with acute myocardial infarction. Lancet 1990;336:65-71. 25. The International Study Group. In-hospital mortality and clinical course of 20 891 patients with suspected acute myocardial infarction randomised between alteplase and streptokinase with or without heparin. Lancet 1990;336:71-5. 26. ISIS-3 (Third International Study of Infarct Survival) Collaborative Group. ISIS-3: a randomised comparison of streptokinase vs tissue plasminogen activator vs anistreplase and of aspirin plus heparin vs aspirin alone among 41 299 cases of suspected acute myocardial infarction. Lancet 1992;339:753-70. 27. ISIS (International Studies of Infarct Survival) Pilot Study Investigators. Randomized factorial trial of high-dose intravenous streptokinase, of oral aspirin and of intravenous heparin in acute myocardial infarction. Eur Heart J 1987;8:634-42. 28. de Bono DP, Simoons ML, Tijssen J, et al. Effect of early intravenous heparin on coronary patency, infarct size, and bleeding complications after alteplase thrombolysis: results of a randomised double blind European Cooperative Study Group trial. Br Heart J 1992;67:122-8. 29. Col J, Decoster O, Hanique G, et al. Infusion of heparin conjunct to streptokinase accelerates reperfusion of acute myocardial infarction: results of a double blind randomized study (OSIRIS). Circulation 1992;86:Suppl I:I-259. abstract. 30. O’Connor CM, Meese R, Carney R, et al. A randomized trial of intravenous heparin in conjunction with anistreplase (anisoylated plasminogen streptokinase activator complex) in acute myocardial infarction: the Duke University Clinical Cardiology Study (DUCCS) 1. J Am Coll Cardiol 1994;23:11-8. 31. The GUSTO Investigators. An international randomized trial comparing four thrombolytic strategies for acute myocardial infarction. N Engl J Med 1993;329:673-82. 32. Van de Werf F, Topol EJ, Lee KL, et al. Variations in patient management and outcomes for acute myocardial infarction in the United States and other countries: results from the GUSTO trial. JAMA 1995;273:158691. 33. Ridker PM, Hebert PR, Fuster V, Hennekens CH. Are both aspirin and heparin justified as adjuncts to thrombolytic therapy for acute myocardial infarction? Lancet 1993;341:1574-7. 34. Mahaffey KW, Granger CB, Collins R, et al. Overview of randomized trials of intravenous heparin in patients with acute myocardial infarction treated with thrombolytic therapy. Am J Cardiol 1996;77:551-6. 35. Théroux P, Ouimet H, McCans J, et al. Aspirin, heparin, or both to treat acute unstable angina. N Engl J Med 1988;319:1105-11. 36. The RISC Group. Risk of myocardial infarction and death during treatment with low dose aspirin and intravenous heparin in men with unstable coronary artery disease. Lancet 1990;336:827-30. 37. Cohen M, Adams PC, Hawkins L, Bach M, Fuster V. Usefulness of antithrombotic therapy in resting angina pectoris or non-Q-wave myocardial infarction in preventing death and myocardial infarction (a pilot study from the Antithrombotic Therapy in Acute Coronary Syndromes Study Group). Am J Cardiol 1990;66:1287-92. 38. Holdright D, Patel D, Cunningham D, et al. Comparison of the effect

of heparin and aspirin versus aspirin alone on transient myocardial ischemia and in-hospital prognosis in patients with unstable angina. J Am Coll Cardiol 1994;24:39-45. 39. Cohen M, Adams PC, Parry G, et al. Combination antithrombotic therapy in unstable rest angina and non-Q-wave infarction in nonprior aspirin users: primary end points analysis from the ATACS trial. Circulation 1994;89:81-8. 40. Gurfinkel EP, Manos EJ, Mejaíl RI, et al. Low molecular weight heparin versus regular heparin or aspirin in the treatment of unstable angina and silent ischemia. J Am Coll Cardiol 1995;26:313-8. 41. Oler A, Whooley MA, Oler J, Grady D. Adding heparin to aspirin reduces the incidence of myocardial infarction and death in patients with unstable angina: a meta-analysis. JAMA 1996;276:811-5. 42. Fragmin during Instability in Coronary Artery Disease (FRISC) study group. Low-molecular-weight heparin during instability in coronary artery disease. Lancet 1996;347:561-8. 43. Cruickshank MK, Levine MN, Hirsh J, Roberts R, Siguenza M. A standard heparin nomogram for the management of heparin therapy. Arch Intern Med 1991;151:333-7. 44. The GUSTO Angiographic Investigators. The effects of tissue plasminogen activator, streptokinase, or both on coronary-artery patency, ventricular function, and survival after acute myocardial infarction. N Engl J Med 1993;329:1615-22. [Erratum, N Engl J Med 1994;330:516.] 45. Granger CB, Califf RM, Hirsh J, Woodlief LH, Topol EJ. APTTs after thrombolysis and standard intravenous heparin are often low and correlate with body weight, age and sex: experience from the GUSTO Trial. Circulation 1992;86:Suppl I:I-258. abstract. 46. Arnout J, Simoons M, de Bono D, Rapold HJ, Collen D, Verstraete M. Correlation between level of heparinization and patency of the infarctrelated coronary artery after treatment of acute myocardial infarction with alteplase (rt-PA). J Am Coll Cardiol 1992;20:513-9. 47. Hsia J, Kleiman N, Aguirre F, Chaitman BR, Roberts R, Ross AM. Heparin-induced prolongation of partial thromboplastin time after thrombolysis: relation to coronary artery patency. J Am Coll Cardiol 1992;20: 31-5. 48. The Global Use of Strategies to Open Occluded Coronary Arteries (GUSTO) IIa Investigators. Randomized trial of intravenous heparin versus recombinant hirudin for acute coronary syndromes. Circulation 1994; 90:1631-7. 49. Antman EM. Hirudin in acute myocardial infarction: safety report from the Thrombolysis and Thrombin Inhibition in Myocardial Infarction (TIMI) 9A Trial. Circulation 1994;90:1624-30. 50. Neuhaus K-L, von Essen R, Tebbe U, et al. Safety observations from the pilot phase of the randomized r-Hirudin for Improvement of Thrombolysis (HIT-III) study. Circulation 1994;90:1638-42. 51. Hennekens CH, O’Donnell CJ, Ridker PM. Current and future perspectives on antithrombotic therapy of acute myocardial infarction. Eur Heart J 1995;16:Suppl D:2-9. 52. Gruppo Italiano per lo Studio della Streptochinasi nell’Infarto Miocardico (GISSI). Effectiveness of intravenous thrombolytic treatment in acute myocardial infarction. Lancet 1986;1:397-402. 53. The I.S.A.M. Study Group. A prospective trial of intravenous streptokinase in acute myocardial infarction (I.S.A.M.): mortality, morbidity, and infarct size at 21 days. N Engl J Med 1986;314:1465-71. 54. AIMS (APSAC Intervention Mortality Study) Trial Study Group. Effect of intravenous APSAC on mortality after acute myocardial infarction: preliminary report of a placebo-controlled clinical trial. Lancet 1988;1:545-9. 55. Wilcox RG, von der Lippe G, Olsson CG, Jensen G, Skene AM, Hampton JR. Trial of tissue plasminogen activator for mortality reduction in acute myocardial infarction: Anglo-Scandinavian Study of Early Thrombolysis (ASSET). Lancet 1988;2:525-30. 56. Rossi P, Bolognese L. Comparison of intravenous urokinase plus heparin versus heparin alone in acute myocardial infarction. Am J Cardiol 1991;68:585-92. 57. EMERAS (Estudio Multicéntrico Estreptoquinasa Repúblicas de América del Sur) Collaborative Group. Randomised trial of late thrombolysis in patients with suspected acute myocardial infarction. Lancet 1993; 342:767-72. 58. LATE Study Group. Late Assessment of Thrombolytic Efficacy (LATE) study with alteplase 6-24 hours after onset of acute myocardial infarction. Lancet 1993;342:759-66. 59. Boersma E, Maas ACP, Deckers JW, Simoons ML. Early thrombolytic treatment in acute myocardial infarction: reappraisal of the golden hour. Lancet 1996;348:771-5. 60. Peto R. Two properties of multiple regression analysis; and regression to the mean (and regression from the mean). In: Fletcher C, Peto R, Tinker C, Speizer FE, eds. The natural history of chronic bronchitis and emphysema: an eight-year study of early chronic obstructive lung disease in working men in London. Oxford, England: Oxford University Press, 1976: 218-23.

Vol ume 336

Numbe r 12



859

The New England Journal of Medi ci ne

61. Castaigne AD, Hervé C, Duval-Moulin A-M, et al. Prehospital use of APSAC: results of a placebo-controlled study. Am J Cardiol 1989;64:30A33A. 62. Schofer J, Büttner J, Geng G, et al. Prehospital thrombolysis in acute myocardial infarction. Am J Cardiol 1990;66:1429-33. 63. GREAT Group. Feasibility, safety, and efficacy of domiciliary thrombolysis by general practitioners: Grampian Region Early Anistreplase Trial. BMJ 1992;305:548-53. 64. The European Myocardial Infarction Project Group. Prehospital thrombolytic therapy in patients with suspected acute myocardial infarction. N Engl J Med 1993;329:383-9. 65. Weaver WD, Cerqueira M, Hallstrom AP, et al. Prehospital-initiated vs hospital-initiated thrombolytic therapy: the Myocardial Infarction Triage and Intervention trial. JAMA 1993;270:1211-6. 66. Leizorovicz A. Benefit from earlier thrombolytic therapy is certain, but what is the magnitude of benefit? BMJ 1996;312:215-6. 67. Braunwald E, Kloner RA. Myocardial reperfusion: a double-edged sword? J Clin Invest 1985;76:1713-9. 68. Honan MB, Harrell FE Jr, Reimer KA, et al. Cardiac rupture, mortality and the timing of thrombolytic therapy: a meta-analysis. J Am Coll Cardiol 1990;16:359-67. 69. Lange RA, Hillis LD. Immediate angioplasty for acute myocardial infarction. N Engl J Med 1993;328:726-8. 70. Bates ER, Topol EJ. Limitations of thrombolytic therapy for acute myocardial infarction complicated by congestive heart failure and cardiogenic shock. J Am Coll Cardiol 1991;18:1077-84. 71. Grines CL, O’Neill WW. Primary angioplasty: the optimal reperfusion strategy in the United States? Br Heart J 1995;73:405-6. 72. Michels KB, Yusuf S. Does PTCA in acute myocardial infarction affect mortality and reinfarction rates? A quantitative overview (meta-analysis) of the randomized clinical trials. Circulation 1995;91:476-85. 73. Ellis S. The GUSTO IIb angioplasty substudy. Presented at the American College of Cardiology Scientific Sessions, Orlando, Fla., March 27, 1996. 74. European Secondary Prevention Study Group. Translation of clinical trials into practice: a European population-based study of the use of thrombolysis for acute myocardial infarction. Lancet 1996;347:1203-7. 75. Pashos CL, Normand S-LT, Garfinkle JB, Newhouse JP, Epstein AM, McNeil BJ. Trends in the use of drug therapies in patients with acute myocardial infarction: 1988 to 1992. J Am Coll Cardiol 1994;23:1023-30. 76. Granger CB, White HD, Bates ER, Ohman EM, Califf RM. A pooled analysis of coronary arterial patency and left ventricular function after intravenous thrombolysis for acute myocardial infarction. Am J Cardiol 1994;74:1220-8. 77. Collins R, Gray R, Godwin J, Peto R. Avoidance of large biases and large random errors in the assessment of moderate treatment effects: the need for systematic overviews. Stat Med 1987;6:245-50. 78. Collins R, Peto R, Gray R, Parish S. Large-scale randomized evidence: trials and overviews. In: Weatherall DJ, Ledingham JGG, Warrell DA, eds. Oxford textbook of medicine. 3rd ed. Vol. 1. Oxford, England: Oxford University Press, 1996:21-32. 79. 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.

860 

Mar ch 2 0 , 1 9 9 7

80. Gore JM, Granger CB, Simoons ML, et al. Stroke after thrombolysis: mortality and functional outcomes in the GUSTO-1 trial. Circulation 1995;92:2811-8. 81. International Joint Efficacy Comparison of Thrombolytics. Randomised, double-blind comparison of reteplase double-bolus administration with streptokinase in acute myocardial infarction (INJECT): trial to investigate equivalence. Lancet 1995;346:329-36. [Erratum, Lancet 1995;346: 980.] 82. Prins MH, Hirsh J. Heparin as an adjunctive treatment after thrombolytic therapy for acute myocardial infarction. Am J Cardiol 1991;67:3A11A. 83. Sobel BE, Hirsh J. Principles and practice of coronary thrombolysis and conjunctive treatment. Am J Cardiol 1991;68:382-8. 84. Tiefenbrunn AJ, Sobel BE. Thrombolysis and myocardial infarction. Fibrinolysis 1991;5:1-15. 85. Scheidt S, Rogers W, Tiefenbrunn A, Bleich SD, Collen D. ISIS-3: implications for American practice: medical panel discussion. CVR & R. September 1991:37-61. 86. Roberts R, Aghababian R, Meister FL, Rogers WJ. ISIS-3 and other highlights of the ACC’s 40th Annual Scientific Session. Vol. 1. No. 6. Clinical Challenges in Acute Myocardial Infarction. June 1991:1-10. 87. Sobel BE, Collen D. Questions unresolved by the Third International Study of Infarct Survival. Am J Cardiol 1992;70:385-9. 88. Collins R, Peto R, Parish S, Sleight P. ISIS-3 and GISSI-2: no survival advantage with tissue plasminogen activator over streptokinase, but a significant excess of strokes with tissue plasminogen activator in both trials. Am J Cardiol 1993;71:1127-8. 89. Grines CL, Demaria AN. The Gusto trial: is it necessary? J Interventional Cardiol 1991;4:149-53. 90. Califf RM, Topol EJ, Stack RS, et al. Evaluation of combination thrombolytic therapy and timing of cardiac catheterization in acute myocardial infarction: results of thrombolysis and angioplasty in myocardial infarction — phase 5 randomized trial. Circulation 1991; 83:1543-56. 91. Grines CL, Nissen SE, Booth DC, et al. A prospective, randomized trial comparing combination half-dose tissue-type plasminogen activator and streptokinase with full-dose tissue-type plasminogen activator. Circulation 1991;84:540-9. 92. Ross AM, Califf RM, Topol EJ. GUSTO: the principal investigators respond to the criticism. J Myocardial Ischemia 1994;6:11-6. 93. Holmes DR Jr, Califf RM, Topol EJ. Lessons we have learned from the GUSTO trial. J Am Coll Cardiol 1995;25:Suppl:10S-17S. 94. Van de Werf F, Califf RM, Armstrong PW, et al. Progress culminating from ten years of clinical trials on thrombolysis for acute myocardial infarction. Eur Heart J 1995;16:1024-6. 95. Mark DB, Hlatky MA, Califf RM, et al. Cost effectiveness of thrombolytic therapy with tissue plasminogen activator as compared with streptokinase for acute myocardial infarction. N Engl J Med 1995;332:141824. [Erratum, N Engl J Med 1995;333:267.] 96. Warren KS, Mosteller F, eds. Doing more good than harm: the evaluation of health care interventions. Ann N Y Acad Sci 1993;703: 1-341. 97. Sleight P. Streptokinase is still the agent of choice for most patients with myocardial infarction. Am J Ther 1995;2:128-35.