Several uncontrolled studies report improvement in lung function, gas exchange, and exercise capacity after bilateral lung volume reduction surgery (LVRS). We recruited 200 patients with severe chronic obstructive pulmonary disease (COPD) for a prospective randomized trial of pulmonary rehabilitation versus bilateral LVRS with stapling resection of 20 to 40% of each lung. Pulmonary function tests, gas exchange, 6-min walk distance, and symptom-limited maximal exercise testing were done in all patients at baseline and after 8 wk of rehabilitation. Patients were then randomized to either 3 additional months of rehabilitation or LVRS. Thirty-seven patients met study criteria and were enrolled into the trial. Eighteen patients were in the medical arm; 15 of 18 patients completed 3 mo of additional pulmonary rehabilitation. Thirty-two patients underwent LVRS (19 in the surgical arm, 13 crossover from the medical arm). After 8 wk of pulmonary rehabilitation, pulmonary function tests remained unchanged compared with baseline data. However, there was a trend toward a higher 6-min walk distance (285 +/- 96 versus 269 +/- 91 m, p = 0.14) and total exercise time on maximal exercise test was significantly longer compared with baseline values (7.4 +/- 2.1 versus 5.8 +/- 1.7 min, p < 0.001). In 15 patients who completed 3 mo of additional rehabilitation, there was a trend to a higher maximal oxygen consumption (V O(2)max) (13.3 +/- 3.0 versus 12.6 +/- 3.3, p < 0.08). In contrast, at 3 mo post-LVRS, FVC (2.79 +/- 0.59 versus 2.36 +/- 0.55 L, p < 0.001) and FEV(1) (0.85 +/- 0.3 versus 0.65 +/- 0.16 L, p < 0.005) increased whereas TLC (6.53 +/- 1.3 versus 7.65 +/- 2.1 L, p < 0.001) and residual volume (RV) (3.7 +/- 1.2 versus 4.9 +/- 1.1 L, p < 0.001) decreased when compared with 8 wk postrehabilitation data. In addition, Pa(CO(2)) decreased significantly 3 mo post-LVRS compared with 8 wk postrehabilitation. Six-minute walk distance (6MWD), total exercise time, and V O(2)max were higher after LVRS but did not reach statistical significance. However, when 13 patients who crossed over from the medical to the surgical arm were included in the analysis, the increases in 6MWD (337 +/- 99 versus 282 +/- 100 m, p < 0.001) and V O(2)max (13.8 +/- 4 versus 12.0 +/- 3 ml/kg/min, p < 0.01) 3 mo post-LVRS were highly significant when compared with postrehabilitation data. The Sickness Impact Profile (SIP), a generalized measure of quality of life (QOL), was significantly improved after 8 wk of rehabilitation and was maintained after 3 mo of additional rehabilitation. A further improvement in QOL was observed 3 mo after LVRS compared with the initial improvement gained after 8 wk of rehabilitation. There were 3 (9.4%) postoperative deaths, and one patient died before surgery (2.7%). We conclude that bilateral LVRS, in addition to pulmonary rehabilitation, improves static lung function, gas exchange, and QOL compared with pulmonary rehabilitation alone. Further studies need to evaluate the risks, benefits, and durability of LVRS over time.