Aspirin versus heparin in the acute phase of unstable angina.

Aspirin Versus Heparin in the Acute Phase of Unstable AnginaAfter an initial four-armed, controlled trial demonstrated a 3.3% rate of infarction in the aspirin arm (that is, four infarcts in this group) and a 0.8% rate in the heparin arm (that is, one infarct), Theroux et all devised a second controlled trial between heparin and aspirin.The joint counting of events in the heparin and aspirin arms of the first trial and in the second trial shows nine events among the aspirin patients (four in the first trial, five in the second trial) and two events among the heparin patients (one in the first trial, one in the second trial).This study1 on aspirin versus heparin to prevent myocardial infarction during the acute phase of unstable angina raises the following issues.(1) What was the conditional probability to observe five events in the aspirin and one in the heparin group among the 245 additionally randomized patients, taking into account the four events and one event counted in the heparin and aspirin arms of the initial trial?Therefore, what is the true probability level of significance for the overall comparison of the heparin and aspirin groups in what is, despite the quotation of the report of O'Brien and Fleming2 among the references, not really a trial with interim analyses but an overall analysis of two successive trials?(2) The authors state that "A Cox regression logistic analysis was used to correct the baseline differences between the two groups" as a part of a legend of survival curves.As similar as the two Cox's models are (one is the linear logistic regression model3 and anoth- erusually quoted as "Cox's model" -is a nonlinear survival data regression model4), the reader may wonder which one has been used.Furthermore, the small number of analyzed events does not necessarily allow for performing those multivariate analyses.(3) The so-called factorial analysis, which is a comparison between the heparin and nonheparin groups obtained by the reunion of both trials, is incorrect because both groups are not balanced for aspirin.The same is true for the comparison between the aspirin and nonaspirin groups.It is not certain that the main conclusion of this study is affected by its poor statistical analysis, but it is unfortunate to read such statistical vagaries in Dr Theroux's report.