Study limitations and implications for future research
While a major strength of our study was its execution in “real world” clinical practice, utilizing existing resources, this also imposed some limitations. Firstly, due to ethical considerations, we were unable to include a second control group who did not receive any rehabilitation-type intervention. Secondly, it would have been informative to have a group in a twice-weekly or three-times-weekly supervised exercise schedule to determine whether this more exacting approach to exercise would add to the effects of the CDSMP. While this more intense exercise has its advocates,6 such an approach is highly resource-intensive; most other similar centers would have been unable to achieve this. Therefore, the effect of adding more than one supervised exercise session to the CDSMP is unknown, or indeed whether there are optimal numbers of weekly exercise sessions. This may be another area worthy of future research. However, even if more sessions are better, resource limitation will always be a major factor for generalizability within many health centers. Thirdly, participants were recruited from referrals to a hospital-based program that may differ from those who might self-refer to community-based CDSMPs. Nevertheless, our study reflects the usual practice for Australian PR, thus enhancing local generalizability. Fourthly, due to resource limitations, we were unable to offer separate sessions for CDSMP-exercise and CDSMP-only groups. While participants were requested not to discuss the exercise experience, vicarious “contamination” of the control group by the active intervention cannot be excluded. Fifthly, the leaders in this study were health professionals rather than peer leaders as is typical for other CDSMPs. Nevertheless, a recently published systematic review concluded that there were few differences between peerled or health professional-led self-management programs,27 suggesting this was unlikely to be a source of bias. Sixthly, we did not stratify randomization according to COPD severity. Although this did not vary a great deal, it may have yielded information as to a differential effect of the intervention and would be a consideration for future research. Finally, the CDSMP does not include a structured home exercise program, since under the license agreement, we were precluded from doing so.