Chronic heart failure (CHF) may impair lung gas diffusion, an effect that contributes to exercise limitation. We investigated whether diffusion improvement is a mechanism whereby physical training increases aerobic efficiency in CHF. Patients with CHF (n = 16) were trained (40 min of stationary cycling, 4 times/wk) for 8 wk; similar sedentary patients (n = 15) were used as controls. Training increased lung diffusion (DlCO, +25%), alveolar-capillary conductance (DM, +15%), pulmonary capillary blood volume (VC, +10%), peak exercise O2 uptake (peak VO2, +13%), and VO2 at anaerobic threshold (AT, +20%) and decreased the slope of exercise ventilation to CO2 output (VE/VCO2, -14%). It also improved the flow-mediated brachial artery dilation (BAD, from 4.8 +/- 0.4 to 8.2 +/- 0.4%). These changes were significant compared with baseline and controls. Hemodynamics were obtained in the last 10 patients in each group. Training did not affect hemodynamics at rest and enhanced the increase of cardiac output (+226 vs. +187%) and stroke volume (+59 vs. +49%) and the decrease of pulmonary arteriolar resistance (-28 vs. -13%) at peak exercise. Hemodynamics were unchanged in controls after 8 wk. Increases in DlCO and DM correlated with increases in peak VO2 (r = 0.58, P = 0.019 and r = 0.51, P = 0.04, respectively) and in BAD (r = 0.57, P < 0.021 and r = 0.50, P = 0.04, respectively). After detraining (8 wk), DlCO, DM, VC, peak VO2, VO2 at AT, VE/VCO2 slope, cardiac output, stroke volume, pulmonary arteriolar resistance at peak exercise, and BAD reverted to levels similar to baseline and to levels similar to controls. Results document, for the first time, that training improves DlCO in CHF, and this effect may contribute to enhancement of exercise performance.