The oxygen cost diagram (OCD) is a simple scale for quantifying a patient's evaluation of his tolerance to exercise frequently used in clinical trials; it has been shown to be well correlated with objective measures of capacity of ambulation such as the 6 min walk test (6' W). This study aimed to determine whether the OCD accurately depicts changes in capacity of ambulation either quantitatively or qualitatively. OCD ratings were analysed at baseline and after a 1 yr follow-up, in patients treated by non-invasive home mechanical ventilation, as well as objective measurements of pulmonary function [forced expiratory volume in 1 sec (FEV1), forced vital capacity (FVC), arterial blood gases], physical autonomy (6' W), resting dyspnoea (Borg scale) and scores for anxiety or depressive disorders (HAD). Forty-five patients (24 male, 21 female, aged 62 +/- 16 years, mean FEV1: 38 +/- 17% of predicted) were evaluated at baseline. OCD ratings were significantly correlated with 6 min walking distance (P < 0.0001)--although with a large variability around the regression line--but not with resting dyspnoea (Borg). Patients were re-evaluated after 352 +/- 90 days. Changes in OCD ratings were not significantly correlated with changes in FFV1 FVC, PaO2, PaCO2, 6' W, HAD scores or resting dyspnoea; furthermore--albeit for Borg scores--changes in OCD did not reflect the trend of changes in these parameters. These results show that although OCD ratings are well correlated with results of a 6' W test, they cannot be used to extrapolate individual performances, because of a large variability around the regression line, furthermore, changes in the OCD over 1 yr did not depict objective changes in 6' W test results, either quantitatively or qualitatively. The use of the OCD in clinical trials should be limited to the description of the patient's perception of exercise tolerance, as a component of health-related quality of life, with the awareness of possible discrepancies between changes in objective performances and changes in OCD ratings.