Article Text

Inconclusive messages from equivalence trials in thrombolysis
  1. V BERTELE,
  2. V TORRI,
  3. S GARATTINI
  1. “Mario Negri” Institute, Milan
  2. and Consozzio “Mario Negri Sud”
  3. S Mazia Imbaro, Italy
  4. email: garattini{at}irfmn.mnegri.it

    Statistics from Altmetric.com

    Glossary

    COMPASS,
    Comparison trial of saruplase and streptokinase
    GISSI,
    Gruppo Italiano per lo studio della sopravvivenza nell’infarto
    GUSTO,
    Global use of strategies to open occluded coronary arteries
    INJECT,
    International joint efficacy comparison of thrombolytics
    ISIS,
    International study of infarct survival

    Editor,—Since the mid-80s a series of large scale randomised clinical trials has progressively proved the effectiveness of thrombolytic treatment in acute myocardial infarction but, despite prodigious efforts, the superiority of intensive strategies based on tissue-type plasminogen activator (tPA) over the standard regimen with streptokinase has not been proved.1The certainty of the benefit has fuelled the search for new thrombolytic agents for a guaranteed market; however, the uncertainty of further benefits from new drugs has suggested testing their equivalence rather than superiority in impracticable trials. This approach has given poor results as demonstrated if one assesses the additional benefits (deaths avoided) and risks (excess of strokes) produced by single steps in the search for better or equivalent thrombolytic agents. We compared indirectly the efficacy and safety of tPA, reteplase, and saruplase by combining the results of studies assessing these agents and streptokinase, as well as streptokinase and no thrombolytic agent (fig 1).2

    Figure 1

    Absolute number of deaths avoided and absolute excess of strokes/1000 patients treated with different thrombolytic agents compared to prethrombolytic era on the basis of the result of the respective superiority or equivalence trials. Numerical data indicate point estimates and limits of 95% confidence intervals (horizontal bars). Mortality and stroke rate were assumed to be 10% and 1.5%, respectively, in patients not treated with thrombolytic agents.

    Based on the results of the equivalence trials INJECT3 and COMPASS,4 the minimum expected effect of reteplase and saruplase on mortality compared to prethrombolysis controls is smaller than the minimum effect attained with standard thrombolytic treatment (seven fewer deaths would be avoided by using reteplase or saruplase compared with tPA). Likewise, the maximum expected excess of strokes with reteplase or saruplase is greater than with current thrombolytic agents (13 more strokes would occur with reteplase and 11 with saruplase compared with streptokinase). This means that reteplase and saruplase may be worse than the worst expected effect of effective thrombolytic strategies, such as those based on streptokinase (in GISSI-I and ISIS-2) or on tPA (in GISSI-2, ISIS-3, and GUSTO-I).1

    The GUSTO-III trial,5 which assessed the superiority of reteplase over alteplase in a larger population than INJECT, provides a better—although not definitive—estimate of the risk–benefit profile of reteplase. As with streptokinase and tPA, further studies might have indicated at least whether reteplase did not differ substantially from standard thrombolytics.

    The place of reteplase and saruplase in the current thrombolytic treatment of acute myocardial infarction have not been established by INJECT and COMPASS. Tests of equivalence are far less precise than superiority trials, and the efficacy–safety profile of drugs assessed in either way cannot be reliably compared. This is mainly because of the small trial populations in several equivalence studies. These small samples are not the results of the equivalence hypothesis but rest on the assumptions that: the new drug is actually more effective or safer than the standard comparator; the real aim of the study is not so much to assess the equivalence as to prove that the new drug is not inferior to the standard comparator; and that the equivalence will be proved even if the confidence interval of the difference is wide. This is difficult to accept if the study concerns “hard” outcome events that are rare in high prevalence disease, such as death in acute myocardial infarction. For example, reteplase would have been considered equivalent to streptokinase even if—on the basis that confidence intervals should not reach +1%—10 more deaths/1000 patients treated with the new drug had occurred in the INJECT trial in addition to the 95 reported with the comparator. Saruplase would also have been considered equivalent, if an excess of 30–35 deaths/1000 patients (odds ratio saruplase:streptokinase < 1.5) had been reported in the COMPASS study in addition to the 67 with streptokinase.

    These devices ultimately imply that the confidence interval for estimated equivalence spans from a large benefit to a large risk. This offers no useful practical information to patients and physicians. Moreover, both the non-inferiority aim and the inconclusive results raise doubts about the ethics of randomisation in equivalence trials.

    References

    View Abstract

    Request permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.