An evidence-based score to detect prevalent peripheral artery disease (PAD)

Abstract

Detection of peripheral artery disease (PAD) typically entails collection of medical history, physical examination, and noninvasive imaging, but whether a risk factor-based model has clinical utility in population screening is unclear. Our objective was to derive and validate a new score for estimating PAD probability in individuals or populations. PAD presence was determined by a history of previous or current intermittent claudication associated with an ankle-brachial index (ABI) of < 0.9 or previous lower extremity arterial intervention. Multivariable stepwise logistic regression identified cross-sectional correlates of PAD from demographic, clinical, and laboratory variables. Analyses were derived from 18,049 US REACH (REduction of Atherothrombosis for Continued Health) Registry outpatients with a complete baseline risk factor profile (enrolled from December 2003 to June 2004). Model performance was assessed internally using 10-fold cross validation, and effect estimates were used to generate the score. The model was externally validated using the Framingham Offspring Study. Age, sex, smoking, diabetes mellitus, body mass index, hypertension stage, and history of heart failure, coronary artery disease, and cerebrovascular disease were predictive of PAD prevalence. The model had reasonable discrimination on derivation and internal validation (c-statistic = 0.61 and 0.60, respectively) and external validation (c-statistic = 0.63 [ABI < 0.9] or 0.64 [clinical PAD]). The model-estimated PAD prevalence varied more than threefold from lowest to highest decile (range, 4.5-16.7) and corresponded closely with actual PAD prevalence in each population. In conclusion, this new tool uses clinical variables to estimate PAD prevalence. While predictive power may be limited, it may improve PAD detection in vulnerable, at-risk populations.