This study was undertaken to test the hypothesis that a reduction in midthigh muscle cross-sectional area obtained by CT scan (MTCSA(CT)) is a better predictor of mortality in chronic obstructive pulmonary disease (COPD) than low body mass index (BMI). We also wished to evaluate whether anthropometric measurements could be used to estimate MTCSA(CT). One hundred forty-two patients with COPD (age = 65 +/- 9 years, mean +/- SD, 26 F, BMI = 26 +/- 6 kg/m(2), FEV(1) = 42 +/- 16% predicted) were recruited from September 1995 to April 2000 with a mean follow-up of 41 +/- 18 months. The primary end-point was all-cause mortality during the study period. A Cox proportional hazards regression model was used to predict mortality using the following independent variables: age, sex, daily use of corticosteroid, FEV(1), DL(CO), BMI, thigh circumference, MTCSA(CT), peak exercise workrate, Pa(O2), and Pa(CO2). Only MTCSA(CT) and FEV(1) were found to be significant predictors of mortality (p = 0.0008 and p = 0.01, respectively). A second analysis was also performed with FEV(1) and MTCSA(CT) dichotomized. Patients were divided into four subgroups based on FEV(1) (< or >or= 50% predicted) and MTCSA(CT) (< or >or= 70 cm(2)). Compared with patients with an FEV(1) >or= 50% predicted and a MTCSA(CT) >or= 70 cm(2), those with an FEV(1) < 50% predicted and a MTCSA(CT) >or= 70 cm(2) had a mortality odds ratio of 3.37 (95% confidence interval, 0.41-28.00), whereas patients with an FEV(1) < 50% predicted and a MTCSA(CT) < 70 cm(2) had a mortality odds ratio of 13.16 (95% confidence interval, 1.74-99.20). MTCSA(CT) could not be estimated with sufficient accuracy from anthropometric measurements. In summary, we found in this cohort of patients with COPD that (1) MTCSA(CT) was a better predictor of mortality than BMI, and (2) MTCSA had a strong impact on mortality in patients with an FEV(1) < 50% predicted. These findings suggest that the assessment of body composition may be useful in the clinical evaluation of these patients.