OBJECTIVE To assess the outcome of a clinical judgement-based approach to weaning and extubation and to adduce the predictive accuracy of various mechanical respiratory indices measured in parallel. METHODS Prospective study. METHODS Multidisciplinary intensive care unit at a university teaching hospital. METHODS 163 consecutive mechanically ventilated patients, excluding tracheotomy, for weaning trial and extubation. METHODS Using bedside clinical assessment, aided by arterial gas analysis, patients were weaned from mechanical ventilation to spontaneous ventilation via to continuous positive airway pressure (CPAP) circuit (with pressure support) of a microprocessor-controlled ventilator. Extubation occurred from the CPAP circuit at 7 cmH2O pressure support, fractional inspired oxygen (FIO2) < or = 0.5 and CPAP level of < or = 5 cmH2O, such that the partial pressure of oxygen in arterial blood (PaO2) was > or = 65 mmHg. Before extubation, observation for a 1-h (T0 and T60) trial period allowed measurement of vital capacity (VC), expired minute volume (VE), respiratory rate/tidal volume (f/VT) and maximal inspiratory pressure (MIP) using a one-way valve technique over 25 s. RESULTS Over 7 months, 163 patients (62 females and 101 males; mean (SD) age 64(15) years) were considered. There were 91 surgical (18 with chronic obstructive pulmonary disease; COPD) and 72 medical (28 with COPD) patients. Ventilation was for > or = 1 day (median 5 days, range 1-31) in 115 [group I; APACHE II score 23(8)] and < or = 1 day in 48 [Group II; APACHE II score 17(6)]. Three patients (all Group I: 2 surgical, 1 medical) were reintubated within 24 h, an overall extubation failure rate of 1.8%. In group I, at T0, PaO2/FIO2 was 238(65), f/VT 50(26), MIP 44(21) cmH2O, VE 10.6(3.7) l/min, VC 13(5) ml/kg. Cardiorespiratory variables did not change significantly in either group, T0 to T60. For prediction of reintubation (n = 163), only VE (threshold > 10 l/min) and f/VT (threshold > 100) demonstrated moderate sensitivity and specificity at T60: 67 and 52% and 33 and 94%, respectively. CONCLUSIONS Bedside clinical judgement of weaning and extubation produces satisfactory outcomes. As a routine, mechanical predictive indices have limited utility.