Objective: We conducted a systematic evidence review on the diagnostic and prognostic value of the resting ankle-brachial index (ABI) in unselected populations. This review also examined the benefit and harms of treating generally asymptomatic persons with peripheral artery disease (PAD). We conducted this review to support the U.S. Preventive Services Task Force in updating its recommendation on screening for PAD.
Data Sources: We searched MEDLINE and the Cochrane Central Registry of Controlled Trials from 1996 through September 2012 to locate relevant English-language studies. We supplemented these searches with suggestions from experts' reference lists from 62 related systematic reviews.
Study Selection: Two investigators independently reviewed 4,434 abstracts and 418 articles against the specified inclusion criteria for diagnostic accuracy studies, prognostic studies, and treatment studies. Our review focused on the utility of using the ABI as a screening or prognostic tool in asymptomatic persons. We excluded populations with symptomatic PAD or known cardiovascular disease (CVD), diabetes, or severe chronic kidney disease. We included diagnostic accuracy studies that evaluated ABI against a reference standard. We included risk prediction studies that evaluated ABI's ability to predict future coronary artery disease (CAD) or CVD events in addition to the Framingham risk score (FRS). Treatment studies were limited to trials evaluating interventions to reduce CVD or maintain lower extremity function. We excluded interventions aimed primarily at management of lower extremity symptoms.
Data Extraction: We extracted all relevant study details (pertaining to population/setting, diagnostic test or intervention, reference standard or comparator, followup and outcomes), which varied by key question. Diagnostic accuracy studies had outcomes focused on measures of test performance (i.e., sensitivity and specificity). For risk prediction studies, outcomes focused on measures of risk reclassification (i.e., number reclassified, net reclassification index [NRI]), measures of discrimination (i.e., differences in the area under the curve [AUC]), or associations of risk adjusted for FRS predictors (i.e., hazard or risk ratios). We extracted any reported outcomes for treatment trials, including adverse effects. We independently appraised all articles for quality and excluded poor-quality studies.
Data Synthesis: Screening: In one fair-quality study (n=306) in older Swedish adults, the sensitivity of ABI (≤0.9) was low (15% to 20%) but specificity was near 100 percent, and the positive and negative predictive values for ABI were greater than 80 percent. Other diagnostic studies of ABI were primarily conducted in persons referred for vascular testing or with symptoms.
Risk prediction: From multiple population cohort studies (18 cohorts; n=52,510), low ABI (≤0.9) was generally associated with future CAD and CVD events, independent of FRS factors. The clinical relevance of the association of a low ABI (≤0.9) and the impact on risk reclassification for CAD and CVD events, however, was less certain. A well-conducted individual patient-level meta-analysis conducted by the ABI Collaboration demonstrated that ABI results could reclassify 10-year CAD risk for 19 percent of men and 36 percent of women when added to the FRS, across 13 population-based cohorts (n=43,919) representing a wide spectrum of persons. Five other studies (n=22,055) evaluated the additional prognostic value of ABI to the FRS using the AUC and/or NRI. In general, these studies suggest that the overall reclassification (among persons of any risk category) is low for CAD or CVD events, the NRI may be higher for older persons for total or hard CAD events (Health ABC; n=2,191), and the NRI is not significant for persons younger than age 65 years for total CVD events (ARIC; n=11,594).
Treatment: We excluded the majority of treatment trials because they focused on persons with intermittent claudication. In one good-quality trial (n=3,350), low-dose aspirin did not prevent CVD events in adults ages 50 to 75 years without known CVD who had a low ABI (≤0.9). In fact, there was a nonsignificant increase in major bleeding events. One smaller, fair-quality trial (n=355) showed that an intensive telephone counseling intervention aimed at adults with primarily asymptomatic PAD can decrease low-density lipoprotein cholesterol levels and achieve treatment goal levels (<100 mg/dL) compared with usual care.
Limitations: A general lack of evidence limited our understanding of the diagnostic test performance for screening ABI and treatment of asymptomatic PAD. The limitations in the understanding of the incremental prognostic value of ABI in CVD risk prediction are due to the differences in populations (e.g., age, sex, race/ethnicity), choice of referent group (i.e., definition of normal ABI), the definitions of composite CAD outcomes (i.e., hard vs. incident vs. total CAD) and risk categories (e.g., intermediate risk of 10% to 19%, 15% to 25%, or 5% to 20%), and measures of reclassification (e.g., number reclassified vs. NRI). These differences make comparisons between risk prediction studies difficult, which limited our ability to interpret findings.
Conclusions: There is very limited evidence examining the diagnostic accuracy of the ABI as a screening tool (one study) or examining the treatment of generally asymptomatic persons with PAD or a low ABI (two trials). However, there is a large body of evidence (18 population-based cohorts) suggesting that a low ABI is independently associated with increased CAD and CVD risk, after adjusting for FRS factors. Despite this association, the magnitude of risk reclassification of ABI in addition to FRS is still unclear and is likely small. The net reclassification may have the largest impact among persons age 65 years and older and persons at the thresholds of FRS risk categories.