Many protocols have been used in clinical and research settings for collecting systolic blood pressure (SBP) measurements to calculate the ankle-brachial index (ABI); however, it is not known how useful it is to replicate measurements and which measures best reflect cardiovascular risk. Standardized measurements of ankle and arm SBP from 5140 overweight or obese individuals with type 2 diabetes were used to estimate sources of variation. Measurement characteristics of leg-specific ABI, as calculated using a standard algorithm based on the highest SBP of the dorsalis pedis or posterior tibial arteries, were projected using simulations. Coefficients of variability ranged from 2% to 3% when single SBP measurements were used and ABI was overestimated by 2-3%. Taking two SBP measurements at each site reduced standard errors and bias each by 30-40%. The sensitivity of detecting low ABI ranges exceeded 90% for ABI within 0.05 of the 0.90 clinical cut-point. The average and the minimum of the two (i.e. right and left) leg-specific ABI values had similar U-shaped relationships with Framingham risk scores; however, the average leg ABI had slightly greater precision. Replicating SBP measurements reduces the error and bias of ABI. Averaging leg-specific values may increase power for characterizing cardiovascular disease risk.