Statistical analyses A power analysis prior to this trial revealed a total subject number of 50 (25 in each group) would provide an 80% chance of detecting a large effect size in the difference in the proportions of each sample (Cohen 1992), at the 0.05 level of significance using a two-sided test. The primary outcome measure in this study was the proportion of postoperative pulmonary complications. Earlier studies had demonstrated large physiotherapeutic treatment effects of deep breathing exercises in high risk subjects in physiotherapy treatment groups compared to non-treatment groups (Olsen et al 1997(. Therefore we anticipated a large treatment effect when designing this study. Fifty-eight subjects were considered to be sufficient in case of data loss or dropouts.