I read with interest the study by Garrodet al on the use of supplemental oxygen during pulmonary rehabilitation in patients with COPD with exercise hypoxaemia.1 Most studies on supplemental oxygen are difficult to interpret because they are either not randomised or placebo controlled. It is therefore important that they have dealt with these problems. Nevertheless, I have some difficulty in interpreting their results. The patients were randomised to either training with air (AT) or oxygen (OT). These two groups were tested without additional gas (test 1), with gas (air or oxygen) before training (test 2), and with gas after six weeks of training (test 3). I have reproduced their data on the shuttle walk test (SWT) and post SWT Borg score (Borg) in table1.

When they compared AT and OT at test 2 they found that oxygen had a significant effect on the SWT (mean difference 27.3 m) and on the Borg score (mean difference –0.68). Surprisingly, although the difference between OT and AT at test 1 was larger, this difference was reported as insignificant.

If changes in the mean outcome measures (test 2 minus test 1) are compared (as they did after six weeks of training), the differences between OT and AT are –1.7 m for SWT and 0.02 for the Borg score. The results therefore depend on the methods used to compare the two groups. The authors need to explain why they compared absolute values when they evaluated the acute effect of oxygen and differences when they evaluated the effect of oxygen during training.