Long-term anticoagulation after acute thromboembolic limb ischemia.

[Dr. Forbes replies] In their paper,1 Forbes and associates evaluate the benefit of long-term anticoagulation after thromboembolectomy in patients without either atrial fibrillation or a cardiac thrombus. In this observational study, with retrospective and prospective components, 3 aspects of study design and data analysis deserve to be addressed. First, not all patients in the study used anticoagulation on a long-term basis. At the time of follow-up, only 79% of patients with atrial fibrillation or a cardiac thrombus (group 1) and 39% of patients without these conditions (group 2) were still taking anticoagulants. Therefore, the groups included a mix of long-term and short- term users of warfarin, and this mix was different in the 2 groups. Such a mix could have introduced a statistical bias in the comparisons by causing a dilution of the effect being studied. Second, in observational studies like this one, it is imperative that the 2 groups be comparable, except for the risk factor under study. In this study, group 2 includes 10 patients (out of 31) with malignant disease, whereas group 1 includes none. Cancer patients are at higher risk of death and thrombotic disorders.2 They should therefore be excluded to make group 2 more comparable to group 1. Finally, the outcome of amputation described in Table 3 occurred in 4 patients who “underwent lower extremity amputation during the initial hospitalization for acute ischemia.” Therefore, this outcome did not occur during follow-up but rather before the exposure being studied (warfarin treatment at the time of discharge). The epidemiologic principle of directionality requires that the outcome be observed after the exposure, so these subjects should not have been included in the analysis. To reduce these sources of bias, it would be helpful if the data could be shown for the 2 groups after removing from group 2 the 10 patients with malignant disease and the 4 patients who underwent an amputation during the initial hospitalization. Showing the results only for the long-term users of anticoagulation treatment would be also useful. Daniel Suissa, BSc Medical student Montreal, Que.