The intrinsic prognostic value of the ankle–brachial index is independent from its mode of calculation
Résumé
The object of this study was to compare the prognostic value of different methods of ankle-brachial index (ABI) calculation. From April 1998 to September 2008, we calculated the ABI in 1223 patients before coronary artery bypass grafting. The ABI was calculated according to five different calculation modes of the numerator. The patients were classified into three groups: clinical peripheral artery disease (PAD), subclinical PAD if no clinical history but abnormal ABI (< 0.90 or > 1.40), and no PAD. The primary outcome was total mortality. During a follow-up of 7.6 years (0.1-15.9), 406 patients (33%) died. The prevalence of the subclinical PAD varied from 22% to 29% according to the different modes of ABI calculation. Areas under the ROC curve to predict mortality according to different calculation modes varied from 0.608 ± 0.020 to 0.625 ± 0.020 without significant differences. The optimal ABI threshold to predict mortality varied for every method, ranging from 0.87 to 0.95. In multivariate models, ABI was significantly and independently associated with total mortality (hazard ratio (HR) = 1.46, 95% CI: 1.15-1.85, p = 0.002); however, this association was not significantly different between the various methods (HRs varying from 1.46 to 1.67). The use of the optimal ABI threshold for each calculation mode (rather than standard 0.90) allowed a slight improvement of the model. In conclusion, the ABI prognostic value to predict mortality is independent from its method of calculation. The use of different optimal thresholds for each method enables a comparable prognosis value.