We designed a prospective multicenter randomized controlled study in three long-term weaning units (LWU) to evaluate which protocol, inspiratory pressure support ventilation (PSV) or spontaneous breathing trials (SB), is more effective in weaning patients with chronic obstructive pulmonary disease (COPD) requiring mechanical ventilation for more than 15 d. Fifty-two of 75 patients, failing an initial T-piece trial at admission, were randomly assigned to PSV or SB (26 in both groups). No significant difference was found in weaning success rate (73% versus 77% in the PSV and SB group, respectively), mortality rate (11.5% versus 7.6%), duration of ventilatory assistance (181 +/- 161 versus 130 +/- 106 h), LWU (33 +/- 12 versus 35 +/- 19 d), or total hospital stay. The results of these defined protocols were retrospectively compared with an "uncontrolled clinical practice" in weaning historical control patients. The overall 30-d weaning success rate was significantly greater (87% versus 70%) and the time spent under mechanical ventilation by survived and weaned patients was shorter in the patients in the study than in historical control patients (103 +/- 144 versus 170 +/- 127 h). The LWU and hospital stays were also significantly shorter (27 +/- 12 versus 38 +/- 18 and 38 +/- 17 versus 47 +/- 18 d). Spontaneous breathing trials and decreasing levels of PSV are equally effective in difficult-to-wean patients with COPD. The application of a well-defined protocol, independent of the mode used, may result in better outcomes than uncontrolled clinical practice.