OBJECTIVES: We evaluated whether treatment initiation of clopidogrel among symptomatic peripheral artery disease (SPAD) was impacted by post-stenting prevention of thromboembolic complications in patients with comorbid coronary artery disease (CAD), in spite of primary use as secondary prevention in SPAD.  METHODS: Patient records from Jan 1, 2006 through June 30, 2010 were extracted from the MarketScan Commercial Claims and Encounters database. Patients met inclusion criteria if they had a record of SPAD but were excluded if they had stroke/TIA or contraindications to anti-platelet therapy. Claims identified 1 year prior to and 3 years post-earliest record of SPAD were included in the analysis. Clopidogrel initiators ± 90 days from the index date were identified. A classification tree model was created with the outcome being the likelihood of treatment initiation with clopidogrel controlling for select baseline covariates. The model used the misclassification criterion, required a minimum of 50 observations/leaf and a maximum of 2 branches/ node.  As a comparison, a stepwise logistic regression model was also constructed. RESULTS: Of 16,377 SPAD patients, 5,164 (32%)  initiated clopidogrel treatment. The output model had 18 leaves, ranging from 0.3%-53.4% of the population. The most important variables were PAD outpatient payment (importance score (IS) = 1), CAD [IS=0.89], pre-index percutaneous coronary intervention (PCI) [IS=0.46] and pre-index statin use (IS=0.33). Similar results were obtained with the logistic regression model. The highest probability leaf for clopidogrel initiation (74.6%; 2.5% of the population) and included patients with low PAD outpatient payment (<$145), comorbid CAD, pre-index PCI and ≥1 pre-index all-cause inpatient admission.   CONCLUSIONS: Our results show clopidogrel use in SPAD patients may be impacted by the presence of comorbid CAD and/or prior PCI use. When evaluating the effectiveness of clopidogrel for SPAD-related outcomes, comorbid CAD and/or PCI use should be considered as confounding variables.