Mega-trials and equivalence trials: experience from the INJECT study

Transcription

1 European Heart Journal (1996) 17 {Supplement ), Mega-trials and equivalence trials: experience from the INJECT study J. R. Hampton Department of Cardiovascular Medicine, University Hospital, Nottingham, United Kingdom As treatments for acute myocardial infarction have grown in number and effectiveness, the post-infarction mortality rate has fallen, and new therapies can provide only a small additional advantage in extending survival. To prove such an advantage of a new drug over its predecessors in the same drug class requires a trial of at least patients. Proving 'equivalence' rather than superiority requires only 6000 patients. The INJECT trial was designed with the modest goal of determining whether the novel agent reteplase (recombinant plasminogen activator) has an effect on mortality equivalent to that of streptokinase. Between August 1993 and September 1994, 6010 patients were randomized to receive either reteplase (n = 3004) or streptokinase (n = 3OO6). Both treatment groups had similar rates of bleeding events, extension or recurrence of myocardial infarction, and in-hospital stroke followed by 6 months of disability. recipients had a lower incidence of Introduction The number of patients needed to give a trial acceptable power depends essentially on the event rate in untreated patients and on the size of benefit that a new treatment can be expected to confer. The number of patients thought to be necessary for a convincing clinical trial has increased dramatically in the past two decades. In the 1970s, when the role of /?-blockade after myocardial infarction was the main focus of interest in cardiology, a trial including 1000 patients was considered large 1 ' 1. In the 1980s, when thrombolytic agents were being compared with placebo in the management of acute myocardial infarction, a few thousand patients was the norm' 21. In the 1990s, individual trials comparing thrombolytic agents have involved more than patients' 3 ' 4 '. As each new therapy becomes established, the mortality rate associated with the condition being investigated falls. Thus, new drug treatments can be expected Correspondence: J. R. Hampton, DM, MA, DPhil, FRCP, FFPM, FESC, Department of Cardiovascular Medicine, University Hospital. Nottingham NG7 2UH, U.K. cardiac events in hospital and fewer allergic reactions. Although the number of diagnosed haemorrhagic strokes was higher in reteplase recipients, more patients receiving streptokinase had strokes of uncertain aetiology. At 35 days, the mortality rate associated with reteplase use was approximately 0-5% lower than that associated with streptokinase administration, and the upper limit of the 90% confidence interval for the difference between these rates was a superiority of streptokinase of 0-73%. Because the INJECT trial provided a probability of 0-95 that the mortality rate associated with reteplase use would be either lower than that with streptokinase administration or at most 0-73% worse, reteplase and streptokinase were proved equivalent according to the trial definition and limits. (Eur Heart J 1996; 17 (Suppl E): 28-34) Key Words: Acute myocardial infarction, reteplase, streptokinase, equivalence, mortality, INJECT trial. to provide only a small additional benefit, and trials must be large to have the power to detect these small differences. When a disorder is associated with a high mortality rate and a totally new form of therapy appears, such as angiotensin converting enzyme (ACE) inhibitors for the treatment of heart failure, trials can still be small. However, even with heart failure, trials enlarged from a few hundred patients in the first ACE inhibitor survival study 15 ' to a few thousand 161 as less severely ill patients were treated. Totally new types of therapy such as /?-blockers, thrombolytics, and ACE inhibitors appear only rarely. Even when they do, large trials may be required to justify adding the new agents to existing therapies of proven value. For example, mortality end-point trials of ^-blockers in heart failure must be conducted on a background of ACE inhibitor therapy. Since only a small additional survival benefit can be anticipated, such trials will have to include several thousand patients. But the problem is potentially much greater when two drugs in the same drug class are being compared: it was the minute difference in outcome anticipated between 0I95-668X/96/0E $18.00/ The European Society of Cardiology

2 The INJECT study 29 different thrombolytics that led to the huge Third International Study of Infarct Survival (ISIS-3) and Global Utilization of Streptokinase and t-pa for Occluded Coronary Arteries (GUSTO) study Comparative studies of agents within a drug class will always be necessary; it will never be appropriate to adopt a new drug purely on the basis of its pharmacological properties. Furthermore, it can never be safely assumed that the effect on survival resulting from treatment with one thrombolytic will be seen with use of all drugs of that type. Not only may each drug have a different safety profile and, therefore, a different balance between risk and benefit, but unless direct comparison is made, doubt will exist about the equivalence of dosage. There may be little point in having 10 or more different thrombolytic agents, /?-blockers, or ACE inhibitors available in the therapeutic armamentarium. However, a few examples of each comprising different ancillary properties, adverse effects, and potential beneficial effects are necessary. It is usually easier to establish differences between drugs within a group in terms of adverse effects, symptomatic relief, and cost, than to determine whether one particular drug has an advantage over another in prolonging patient survival. Once the effect on mortality of one or two drugs of each class has been established by comparison with placebo (e.g. streptokinase and alteplase [t-pa] in acute myocardial infarction), symptom relief, adverse effects, and cost comparisons become more important than trivial differences new drugs from that class might contribute to survival. Thus, if a new drug within an established class has a definite advantage over its predecessors in terms of adverse effects or cost, before using it a clinician will need to know only that its effect on survival is equivalent to that of the other drugs in its class. The problem, then, is: what is meant by 'equivalent'? The new thrombolytic agent reteplase (r-pa) exemplifies all these points. is a non-glycosylated deletion mutant of wild type t-pa. It consists of the kringle-2 and the proteinase domains, but lacks the kringle-1, finger, and growth factor domains of t-pa. The modification results in less high-affinity fibrin binding, a longer halflife, and a greater thrombolytic potency than seen with t-pa. In an animal model, reteplase was superior to anistreplase, streptokinase, and urokinase, achieving more rapid, complete, and sustained thrombolysis' 71. Clinically, this agent has the considerable advantage of administration by bolus injection rather than by infusion. is not antigenic and can be used repeatedly. Production costs of reteplase are potentially lower than those of alteplase. In an angiographic study 181, reteplase given as a double bolus of 10MU+10MU achieved more rapid, complete, and sustained thrombolysis in an infarctrelated artery than did standard dose alteplase. The findings of this study, which included more than 600 patients, suggest that reteplase treatment was associated with improved left ventricular function at the time of hospital discharge. The complication rate also appeared to be no higher with reteplase than with alteplase. therefore seems to be an attractive thrombolytic agent. On the basis of the angiographic study, reteplase administration might be expected to reduce mortality in acute myocardial infarction % more than administration of alteplase. Unfortunately, a trial with the power to demonstrate such a difference would have to include more than patients. Such a trial would be time consuming and expensive, and it would not necessarily justify the efforts of the research community. The International Joint Efficacy Comparison of Thrombolytics (INJECT) trial was therefore designed to determine not whether reteplase use was associated with a better survival rate than another thrombolytic, but whether its effect on survival was equivalent. The concept of equivalence Equivalence is a practical, clinical, and philosophical concept. Although equivalence must be described in statistical terms, it does not have a statistical foundation. The clinician must establish what can be accepted as 'equivalent'; the statistician will then calculate the size of the trial necessary. Figure 1 illustrates the possible results of some hypothetical studies of two drugs named A and B. In the first trial, drug A led to a risk reduction compared with drug B. In this small trial, the confidence intervals (CIs) were relatively wide; however, since they were clear of the line of zero effect, A was significantly superior to B. In the second trial, B was significantly better than A. The third trial was large, and the CIs around the result were small. Drugs A and B produced the same effect (i.e. the relative difference between them was zero); thus, no clinician would doubt that A and B were equivalent in terms of mortality. In the fourth trial, A and B again produced the same effect (zero observed difference), but the trial was smaller and the CI around the result was wider. Can A and B now be considered equivalent? This is not a statistical problem: the clinician must decide whether the degree of uncertainty indicated by the CI is acceptable. In the fifth trial, although the effect of A was superior to that of B, the CI around the result included the possibility that there was no difference between the drugs. A clinician might decide that it would be inappropriate to consider B equivalent to A, but the statistician would tell him that within the limits of the trial (limits pre-determined by the trial size and thus the width of the CI), it would be appropriate to consider A and B equivalent. Although in the fifth trial A was not demonstrably superior to B, a term such as 'at least equivalent' to B could justifiably be used.

3 30 J. R Hampton -* Drug A better DrugB better»- riais 1 2 a. 1 3 u o HH 53 4 C o Ea. 5 o Difference in mortality Figure 1 Comparison of mortality rates associated with drugs A and B in five hypothetical trials. Bullets represent observed trial results; bars represent confidence intervals (CIs) around the observed results. Although CIs are wide in trial 1, they avoid the line of zero effect, so that A is clearly superior to B. Similarly, B is superior to A in trial 2. In trial 3, the zero relative difference between A and B describes their effects as equal; the small CI reinforces this conclusion. The clinician must decide whether A and B are equivalent in trial 4, which was small and therefore associated with a wide CI, and whether A can be considered superior in trial 5, in which the CI included the possibility of zero effect. When designing a trial to study equivalence, it is thus necessary to determine the acceptable width of the CIs around the results. It is also necessary to predict the likely point estimate (i.e. the observed result), for its position will determine whether the CI around it will extend beyond the line of zero effect. Design of the INJECT trial From the outset, the purpose of the INJECT trial was to establish the equivalence of reteplase with another thrombolytic agent in the reduction of mortality following acute myocardial infarction. It would have been unethical to include a placebo-treated group; however, considerable debate arose over whether to compare the use of reteplase with that of streptokinase or alteplase. Because provision of alteplase would have been prohibitively expensive, and because streptokinase is the standard thrombolytic used throughout most of Europe, streptokinase was chosen as the comparator. The next problem was to decide on an acceptable CI within which reteplase would be considered equivalent to streptokinase. On pragmatic clinical grounds, it was decided that reteplase would be acceptable as equivalent to streptokinase if the CI around the trial results excluded the possibility that the mortality associated with reteplase use was more than 1% worse (in absolute terms) than the mortality associated with streptokinase administration. If one end of the CI had to be less than the streptokinase-associated mortality plus 1%, then the size of the trial would be controlled by the length of the CI and what the actual result (the point estimate comparison) was expected to be. On the basis of the angiographic studies with reteplase, the reteplaseassociated mortality rate was expected to be 1% less than that with streptokinase. If this proved true, a trial of 6000 patients would have 90% power to achieve the trial objective. Possible results that might have been observed in the INJECT trial are illustrated in Fig. 2. If the observed result proved to be equal mortality rates in the groups given streptokinase or reteplase (absolute difference 0%), then the CI around that result would include the possibility that the real effect of reteplase was more than 1% worse than that with streptokinase; by the definition of the trial, equivalence would not have been proven. If the mortality rate with reteplase were between 0-5% and 1% less than that seen with streptokinase, the CI would include zero, so it would not be possible to conclude that reteplase is superior to streptokinase. However, because the CI would not include the possibility that the effect of reteplase is more than 1% worse than that of streptokinase, equivalence could be claimed. Only if the difference between treatments approached 2% in favour of reteplase would it be possible (with a trial of 6000 patients) to conclude that mortality associated with reteplase use is superior to that with streptokinase administration. Eur Heart J, Vol 17, Suppl E 1996

4 The INJECT study 31 >> "p 1 C 1 oe a Retepl ase better Streptokin "5 2 E 2 a Absolute difference (%) i i i i i i Figure 2 Hypothetical 35-day mortality results used in planning the INJECT trial. -associated mortality would be accepted as equivalent to that of streptokinase if the confidence interval (CI) around the results excluded the possibility that the effect of reteplase is >1% worse than that of streptokinase (dashed line). Thus, in example 1, equivalence could not be proved. In example 2, the reteplase-associated mortality rate is <1% greater that the mortality rate with streptokinase (i.e. <1% greater than zero difference), consistent with equivalence; however, since the CI included no difference between treatments, superiority cannot be claimed. The third estimate shows a difference approaching 2% in favour of reteplase, sufficient to signify superiority. INJECT=International Joint Efficacy Comparison of Thrombolytics. better Streptokinase better Difference in 35 day mortality (%) Figure 3 Actual 35-day mortality rates in the INJECT trial. was associated with a mortality 0-51% lower (bullet) than that of streptokinase; the confidence interval (CI) was % to 0-73%. Since the CI includes zero effect, superiority could not be claimed. However, by the trial definition, the two drugs could be considered equivalent. Dashed line=limit of potential equivalence.

5 32 J. R Hampton Table 1 Comparability zation Variable n Age (years) Mean (SD) >75 years (%) Male (%) Smokers (%) Current Former Never Previous morbidity (%) Myocardial infarction Angina Heart failure Hypertension Diabetes Qualifying ECG (%) ST elevation: anterior* ST elevation: inferiorf Bundle branch block Undefined Time to treatment (h) Mean (SD) Sill (yo) >6 h (%) of treatment groups at randomi (11-5) (2-8) / / 201 Including lateral and antero-lateral. ( Including infero-lateral and infero-posterior. ECG=electrocardiogram; SD = standard deviation. The INJECT trial Streptokinase (11-7) (2-4) 18-6 The full results of the INJECT trial have been described elsewhere' 9 '. A total of 6010 patients were recruited between August 1993 and September 1994; 3004 patients were randomized to reteplase and 3006 were randomized to streptokinase. Nine countries took part in the study: United Kingdom (n = 2280), Germany (n=1909), Poland (n=1098), Sweden (n = 292), Hungary (n=175), Finland (n=129), Spain (n = 53), Lithuania (n=46), and Austria (n = 28). The two treatment groups were well matched (Table 1). Patients older than 75 years comprised 11-9% of participants, and patients with a history of previous myocardial infarction comprised 14-6%. Aspirin was prescribed for 93% of patients, and intravenous heparin for 99%. Myocardial infarction, either definite or possible, was the confirmed admission diagnosis for 95% of patients in both groups. The distribution of infarct sites was similar in the two treatment groups. Mortality and stroke incidence Of all randomized patients, 270 patients in the reteplase group died within 35 days after myocardial infarction compared to 285 patients in the streptokinase gtoup. Mortality rates at 35 days were thus 90% for reteplase and 9-5% for streptokinase, a difference of - 0-5% with Table 2 Incidence and outcome of in-hospital strokes All strokes* 37 (1-2) Haemorrhagict 23 (0-8) Embolic 10(0-3) Uncertain 4(01) Outcome Death <, 35 days 19 (0 6) Death days 2(01) Disability 4(01) Recovery 6 (0-2) Living (status unknown) 6 (0-2) Streptokinase 30 (10) 11 (0-4) 9 (0-3) 10 (0-3) 11 (0-4) 1 (<01) 6 (0-2) 7 (0-2) 5 (0-2) Includes one patient in the reteplase group who suffered an embolic stroke before treatment initiation and did not receive study medication. -( Includes two cases in the reteplase group assessed by the investigator as haemorrhagic, but no computed tomography scan or post mortem evidence available. a 90% Cl of % to 0-73%. Because the confidence interval includes zero effect (Fig. 3), this study did not demonstrate superiority of reteplase to streptokinase. However, since the possibility that the effect of reteplase is 1% worse than that of streptokinase was excluded, the two drugs can be considered equivalent according to the trial definition. The reteplase group had a small, non-significant excess of total in-hospital strokes compared to the streptokinase group (37 compared with 30 patients). Of all the patients with stroke, 18 in the reteplase group and 19 in the streptokinase group survived to day 35. Two strokes in the reteplase group were described as haemorrhagic on the basis of clinical evidence alone; however, in all other cases of stroke classified as either haemorrhagic or embolic, the cause was confirmed by computed tomography scan or post mortem examination. Although a greater proportion of strokes in the reteplase group appeared to result from cerebral haemorrhage, more strokes of uncertain aetiology occurred in the streptokinase group (10 vs 4 in the reteplase group); this may have over-emphasized the difference in incidence of haemorrhagic stroke between groups (Table 2). Functional assessment at 6 months of all patients who suffered an in-hospital stroke shows comparable degrees of residual disability. When the end-points of 35-day mortality and continuing disability from an in-hospital stroke were combined, the rates were 9-2% in the reteplase group and 9-8% in the streptokinase group, a difference of - 0-6% (95% CI, to 0-88). Extended or recurrent myocardial infarction The incidence of recurrent myocardial infarction or extension of infarction was similar in both treatment groups. In the reteplase group, 149 patients (50%) experienced a new myocardial infarction or extension in

6 The INJECT study 33 Table 3 In-hospital bleeding events Total patients affected Significant bleeding* Required transfusion Location Puncture site Gastrointestinal tract Genito-urinary tract Haemoglobin decrease >3 g. dl~ ' 450(150) 138 (4-6) 21 (0-7) 135 (4-5) 75 (2-5) 48 (1-6) 75 (2-5) Streptokinase 460(15-3) 141 (4-7) 30 (10) 150(50) 84 (2-8) 54(1-8) 96 (3-2) 'Excluding gum or injection site bleeding, subcutaneous bleeding, haematuria, and any decrease in haemoglobin level described as non-serious. hospital compared to 162 patients (5-4%) in the streptokinase group. In general, the reteplase group had a lower incidence of cardiac events in hospital. There were statistically significant differences between the groups regarding atrial fibrillation, asystole, cardiac shock, heart failure, and hypotension. Other adverse effects Bleeding events in the two treatment groups were generally minor. Such events severe enough to require transfusion affected 07% of patients in the reteplase group and 1 0% of patients in the streptokinase group (Table 3). Patients in the reteplase group had fewer allergic reactions than those in the streptokinase group. Three in-hospital allergic events were reported as 'serious' in the reteplase group (two hypotension, one unspecified) and 15 were reported in the streptokinase group (three anaphylactic shock, two angioneurotic oedema, one bronchospasm, one pyrexia, eight unspecified). No antibodies were detected in the second blood sample taken from reteplase-treated patients either at hospital discharge or 35-day follow up. INJECT and the mega-trials Studies such as ISIS-3 [3] and GUSTO [4] exemplify the study size that must be considered to identify a difference in mortality of approximately 1%. Such studies require very large resources and if a further reduction in mortality must be proven before a new thrombolytic agent can be introduced, it is unlikely that many alternative therapies will become available even if they offer other advantages such as ease of administration or a more acceptable adverse event profile. More modest alternatives to large trials must be explored; equivalence trials such as INJECT may offer one such possibility. Although INJECT is an equivalence trial, its statistical basis differs from that of a trial designed to show that two drugs are likely to have the same mortality rates. In a conventional equivalence trial, the starting point is the belief that treatments are truly identical. The objective is to confirm this belief within pre-defined limits. For INJECT, the starting point was the belief that the new agent does offer a small mortality benefit over the comparator. Rather than trying to prove that this small benefit exists, a more conservative target was chosen: to show that the new agent is associated with a mortality rate not more than 1% worse than that associated with streptokinase. Since the mortality rate associated with reteplase use was anticipated to be 1% less than that with streptokinase, a trial of 6000 patients provided appropriate power to achieve this objective. The trial resulted in a 35-day mortality rate for reteplase approximately 0-5% lower than that for streptokinase. The upper limit of the 90% CI (two sided) for the difference in mortality rates between randomized treatment groups was 0-71%. The INJECT trial, therefore, gives a probability of 0-95 (one sided) that the true mortality associated with reteplase use is either better than that associated with streptokinase or at most 0-71% worse. The safety profile observed with reteplase administration was generally satisfactory. Although the haemorrhagic stroke rate associated with reteplase use was higher than that with streptokinase, most strokes were fatal and, therefore, included in the mortality figures. Both treatment groups had identical, small proportions of patients who experienced in-hospital stroke and remained disabled for 6 months. The overall mortality rate in the INJECT trial was somewhat higher than that in the GUSTO trial despite similar entry requirements and similar demographic profiles. Although the wider time window for INJECT provides a partial explanation for the mortality difference, the principal reason for this disparity may be that the INJECT trial was conducted outside the United States, where in clinical trials mortality rates after myocardial infarction seem to be lower. Conclusion Because it is now unethical to compare a new thrombolytic agent with placebo, comparisons with existing agents are necessary. To demonstrate the superiority of a new agent in terms of mortality would require a vast, expensive trial. However, when a new drug such as reteplase has definite advantages over existing thrombolytic agents, all that is clinically necessary is to prove that its effect on mortality following myocardial infarction is equivalent to that of other thrombolytics. Proving equivalence is not the same as failing to show a difference between treatments, and INJECT was a trial designed specifically with equivalence as an end-point. INJECT showed that reteplase is equivalent to streptokinase in its effect on mortality following myocardial infarction.

7 34 J. R. Hampton References [1] Hampton JR. Should every survivor of a heart attack be given a beta blocker? Evidence from clinical trials. Br Med J 1982; 285 (Part 1): [2] Fibrinolytics Therapy Tnalists' (h II) Collaborative Group. Indications for fibrinolytic therapy in suspected acute myocardial infarction: collaborative overview of mortality and major morbidity results from all randomised trials of more than 1000 patients. Lancet 1994; 343: [3] ISIS-3 (Third International Study of Infarct Survival) Collaborative Group. ISIS-3: a randomised trial of streptokinase vs tissue plasminogen activator vs anistreplase and of aspirin plus heparin vs aspirin alone among cases of suspected acute myocardial infarction. Lancet 1992; 339: [4] The GUSTO Investigators. An international randomized trial companng four thrombolytic strategies for acute myocardial infarction. N EngI J Med 1993; 329: [5] The CONSENSUS Trial Study Group. Effects of enalapril on mortality in severe congestive heart failure. Results of the Cooperative North Scandinavian Enalapril Survival Study (CONSENSUS). N Engl J Med 1987; 316: [6] Pfeffer MA and SAVE Investigators. Effect of captopnl on mortality and morbidity in patients with left ventricular dysfunction after myocardial infarction. Results of the Survival and Ventricular Enlargement Trial. N Engl J Med 1992; 327: [7] Martin U, Sponer G, Strein K. Evaluation of thrombolytic and systemic effects of the novel recombinant plasminogen activator BM compared with alteplase, anistreplase, streptokinase and urokinase in a canine model of coronary artery thrombosis. J Am Coll Cardiol 1992; 19: 433-^0. [8] Smalling RW, Bode C, Kalbfleisch J et al. More rapid, complete, and stable coronary thrombolysis with bolus administration of reteplase compared with alteplase infusion in acute myocardial infarction. Circulation 1995; 91: [9] INJECT Investigators. A randomised double-blind comparison of reteplase double-bolus administration with streptokinase in patients with acute myocardial infarction (INJECT): a trial to investigate equivalence. Lancet 1995; 346: Eur Heart J. Vol. 17. Suppl E 1996