Determination of ABI allows the non-invasive and reliable detection and quantification of PAD. It is one of the most widely used methods in the epidemiology of PAD since it can be easily performed in large-scale studies, with low variability of measurements between different observers [12, 17, 24]. At the same time, it is increasingly used as screening measure in primary care. However, different modes of ABI calculation are used in the literature. In some reports, the highest arterial pressure in each leg is used, whilst in others the lowest arterial pressure serves to determine the ABI. Alternatively, only the tibialis posterior or dorsalis pedis pressure is used, or the pressures of one leg are averaged [18, 24–27]. However, the optimal method for ABI calculation for estimating the prevalence of PAD and predicting mortality and other outcomes in PAD has to our knowledge not been determined.
This is the first large-scale epidemiological study to screen an unselected sample of patients in primary care for the prevalence of PAD using different methods for the calculation of the ABI. The data confirm the high prevalence of PAD in primary care [28, 29], indicating that – at least – about every fifth subject aged ≥65 years, is a PAD patient.
In our study, and as recommended by the American Heart Association , taking the higher of the two systolic pressures (method #1, tibial posterior and anterior artery) resulted in the most conservative estimate of the prevalence of PAD. However, at ABI values > 1.5, the presence of poorly compressible ankle arteries due to calcified vessels should be taken into account. Incompressible arteries are frequently observed in diabetics and haemodialysis patients and may result in incorrectly high ABI values .
Modes of ABI calculations in which the dorsalis pedis (which is the distal extension of the anterior tibial artery) pressure (method #3) serves as the numerator may be misleading and not generally suitable in primary care, since an absent dorsalis pedis signal due to hypoplasia is described in 8–12% of healthy subjects without PAD [11, 32]. The smaller diameter of the dorsalis pedis artery relative to the posterior tibial artery may be more difficult for general physicians to locate with the Doppler probe (leading to an underestimate of the true pressure), and in addition, artery pressures due to pulse wave reflection in smaller vessels may be different in the two vessels. 
Thus, if the numerator were defined by the dorsalis pedis pressure (method #3) or the lowest ankle pressure (method #2) (which would be 0 if the dorsalis pedis pressure is not detectable), the value 0 would be attributed to these legs although PAD would be absent. Likewise, in a smaller study in healthy subjects, Aboyans et al.  disqualified methods involving the lower arterial pressure at the ankle, since this method cannot distinguish between hypoplastic or obstructed arteries. As in our approach, that study also recommended the use of the higher arterial pressure at the ankle for calculation of the ABI, which yielded a PAD prevalence estimate of about 5% at the cut-off point of 0.9. This lower value can be explained by the fact that patients with history of PAD were excluded in the study by Aboyans et al. Moreover, the cohort was significantly younger, with only 26% of the patients being > 65 years .
In another approach, McDermott et al. correlated three modes of ABI determination to leg functioning parameters in PAD patients . As in the present observations, using the lower of the two arterial pressures at the ankle (method #2) resulted in a significantly higher PAD prevalence. However, their results suggested that the ABI, determined by averaging the dorsalis pedis and posterior tibial arterial pressures in each leg, may be most predictive of walking endurance and walking speed in patients with peripheral arterial disease. In contrast to that study, which was performed with regard to leg functioning parameters on selected patients with high prevalence of PAD, the present investigation was performed on a large-scale general population aged 65 years or older. Consequently, the results may not be directly comparable.
Another finding is of considerable importance for the practical use of the ABI for screening purposes. While the different modes of ABI calculation profoundly affected the estimation of PAD prevalence, there was no major effect on the association between other concomitant manifestations of atherothrombotic disease, i.e. cardiovascular events. ROC analyses were performed to describe the relationship between sensitivity and specificity of different ABI values (according to different modes of calculation) with respect to history of cardiovascular events. Interestingly, the respective ROC curves differed only marginally between the diverse modes of calculation, indicating that the strength of association between the ABI and cardiovascular events is largely independent of the mode of ABI determination. Thus, an abnormal ABI, irrespective of the mode of calculation, is associated with the same prevalence of cardiovascular events at a "mode-specific" cut-off point.
For the assessment of exercise ABI, in contrast to a previous methodological study  that used the posterior tibial artery or the dorsalis pedis artery (which is the distal extension of the anterior tibial artery), due to the reasons given above [14, 32] we chose to focus on measurements on one vessel, the posterior tibial artery. Of note, even the determination of the ABI after exercise did not increase the strength of association with concurrent (or history of) cardiovascular events. Since ABI calculation after exercise is a demanding procedure in terms of time and manpower, and since there is no obvious advantage over other methods, this approach may not be suitable for routine determination of the ABI in general practice.