Anticoagulant therapy in acute myocardial infarction: Demonstration of a selection bias in a retrospective study

Records of 851 patients with first acute myocardial infarction were analyzed in order to identify the effect of anticoagulant therapy on hospital mortality. A coronary prognostic index was applied and the patients were allocated into subgroups with similar predicted prognosis. When the mortality rates of treated versus non-treated patients were compared within each subgroup, no effect of anticoagulants on mortality could be demonstrated. However, when the patients were divided only according to anticoagulant therapy, a significantly lower mortality rate was found in the treated group: 12.3% versus 22.3% (p< 0.001). The disparity of results obtained via different methods of patient allocation may be explained by the existence of a selection bias expressed by the choice of better risk patients for anticoagulant treatment. Indeed, analysis of the records showed that the ratio of treated versus non-treated patients decreased in parallel with the worsening of prognosis.