Background Low muscle oxidative capacity is an extrapulmonary manifestation of chronic obstructive pulmonary disease (COPD) with unclear aetiology. We sought to characterize locomotor muscle oxidative capacity in never smokers and ever smokers with and without COPD and determine clinical and behavioural features associated with low muscle oxidative capacity. Methods Two hundred forty-three adults enrolled in the Muscle Health Study, an observational study ancillary to COPDGene. Gastrocnemius oxidative capacity was measured by near-infrared spectroscopy from the muscle oxygen consumption recovery rate constant (k). Physical activity was measured by accelerometry (vector magnitude units [VMU]/min). Pulmonary assessments included spirometry (FEV1%predicted), diffusing capacity (DLCO) and quantitative chest computed tomography (CT). Eighty-seven variables related to COPD features were considered. Variables selected by univariate analysis of log-transformed k with p ≤ 0.20 and filtered by machine learning were entered into multivariable linear regression to determine association with k. Results Two hundred forty-one (53.1% female; 45.6% African American; 64 ± 10 years old) participants were allocated to analysis. FEV1%predicted, DLCO, CT, pack-years, age and VMU/min were among 24 variables selected by univariate analysis. After machine learning filtering on 162 (67%) cases with complete data, 11 variables were included in multivariable analysis. Only FEV1%predicted, age and race were significantly associated with k (R2 = 0.26). Model coefficients equate a 10% lower FEV1%predicted to a 4.4% lower k or 10 years of aging to a 9.7% lower k. In 118 cases with CT available, FEV1%predicted and age remained associated with k (R2 = 0.24). Physical activity was not retained in any model. Conclusions Physical activity or radiographic COPD manifestations were not significantly associated with muscle oxidative impairment. Across never smokers and ever smokers with and without COPD, locomotor muscle oxidative capacity was positively associated with FEV1%predicted and negatively associated with age.