Background No data exist comparing outcome prediction from arm exercise vs pharmacologic myocardial perfusion imaging (MPI) stress test variables in patients unable to perform treadmill exercise. Methods In this retrospective study, 2,173 consecutive lower extremity disabled veterans aged 65.4 ± 11.0 years (mean ± SD) underwent either pharmacologic MPI (1730 patients) or arm exercise stress tests (443 patients) with MPI (n = 253) or electrocardiography alone (n = 190) between 1997 and 2002. Cox multivariate regression models and reclassification analysis by integrated discrimination improvement (IDI) were used to characterize stress test and MPI predictors of cardiovascular mortality at >10-year follow-up after inclusion of significant demographic, clinical, and other variables. Results Cardiovascular death occurred in 561 pharmacologic MPI and 102 arm exercise participants. Multivariate-adjusted cardiovascular mortality was predicted by arm exercise resting metabolic equivalents (hazard ratio [HR] 0.52, 95% CI 0.39-0.69, P <.001), 1-minute heart rate recovery (HR 0.61, 95% CI 0.44-0.86, P <.001), and pharmacologic and arm exercise delta (peak-rest) heart rate (both P <.001). Only an abnormal arm exercise MPI prognosticated cardiovascular death by multivariate Cox analysis (HR 1.98, 95% CI 1.04-3.77, P <.05). Arm exercise MPI defect number, type, and size provided IDI over covariates for prediction of cardiovascular mortality (IDI = 0.074-0.097). Only pharmacologic defect size prognosticated cardiovascular mortality (IDI = 0.022). Conclusions Arm exercise capacity, heart rate recovery, and pharmacologic and arm exercise heart rate responses are robust predictors of cardiovascular mortality. Arm exercise MPI results are equivalent and possibly superior to pharmacologic MPI for cardiovascular mortality prediction in patients unable to perform treadmill exercise.