Objectives: To compare bone mineral densities (BMDs) of the affected and unaffected limbs in stroke patients at multiple sites; to study longitudinal changes during a 3-month rehabilitation program; and to relate BMDs to demographic, impairment, and disability variables. Design: Descriptive study. Setting: Tertiary rehabilitation center. Patients: One hundred four consecutive hemiplegic inpatients, 69 men, age 56.5 +/- 13.2yrs, 47 with left-sided brain lesion. Median days from onset to admission and median length of stay days were 83 and 105.5, respectively Main Outcome Measures: BMDs of proximal humerus, distal radius, femoral neck, and calcaneus bilaterally, and third lumbar vertebra, measured with dual-energy x-ray absorptiometry (DXA), were compared between affected and unaffected sides at admission and discharge. Results: Stroke Impairment Assessment Set (SIAS) motor scores, Functional Independence Measure (FIM) scores, grip strength, and awake/sleep heart rate counts (activity index) improved significantly at discharge. Affected/unaffected BMD ratios were 88.3% to 98.4% at admission and 79.6% to 98.8% at discharge, lowest for the humerus. Discharge/admission ratios were 89.1% to 97.8% for the affected and 97.4% to 100% for the unaffected side. All BMDs were intercorrelated (R = .438 to .873). They correlated significantly with age, body weight, grip strength, FIM scores, and activity index. Factors selected to explain BMD with multiple regression analysis differed according to the site and timing of the measurement. Conclusions: BMDs of the affected side were lower and most marked in the humerus. Longitudinally, not only the affected but the unaffected BMDs decreased. Age, sex, duration of stroke, anthropometric measurements, motor paralysis, muscle strength, and activity level contributed differently to bone loss according to the site and timing of the measurement. (C) 1999 by the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation.