abstract
- BACKGROUND AND AIMS: Patients with coronary artery disease (CAD) who also have peripheral artery disease (PAD) are at high risk of subsequent cardiovascular events and mortality. Despite this, PAD in patients with CAD often remains undiagnosed. The objective of this analysis was to assess clinical factors that predict the presence of PAD in patient with documented CAD who also have PAD. METHODS: In a post hoc analysis of patients with CAD in the COMPASS trial, we developed separate prediction models for symptomatic lower extremity PAD and documented carotid artery disease (Model 1), asymptomatic lower extremity PAD defined as ABI <0.9 (Model 2) and for any PAD (symptomatic or asymptomatic; Model 3). Using logistic regression models, candidate variables were chosen to predict the presence of PAD. Overall model performance was evaluated for discrimination and calibration using the concordance statistic and Hosmer and Lemeshow Goodness-of-fit chi-square, respectively. The final model was validated by bootstrapping. RESULTS: Of 23,402 participants, 3484 (14.9%) had a history of symptomatic PAD or carotid artery disease (Model 1), 1422 (5.7%) participants had asymptomatic PAD (Model 2) and 4906 (20.6%) had any PAD (Model 3). Model 1 demonstrated a C-statistic of 0.667 and goodness-of-fit p-value of 0.859. Model 2 demonstrated a C-statistic of 0.626 and goodness-of-fit p-value of 0.250. Model 3 demonstrated a C-statistic of 0.646 and goodness-of-fit p-value of 0.240. CONCLUSION: Routinely available clinical information is only marginally useful to identify patients with CAD and concomitant PAD.