Bias, confounders and limitations
From a methodological point of view, a cluster randomized design would be the most sound design for the study. The rationale for performing an individually randomized trial is that we observed in a pilot study [44], that physiotherapists were not very keen to deliver a treatment without a proper training programme, although hard evidence of efficacy of a physical exercise training programme COPD was still lacking. Therefore, it would be impossible to recruit sufficient physiotherapy practices with treating only control patients. We have found a solution to this problem by allowing the physiotherapist offering the possibility for patients in the control group to participate in the physical exercise training programme after the study-period, in case the intervention has proven to be efficacious. To tackle contamination, we will train and instruct the physiotherapists thoroughly in advance of the study and monitor and instruct them throughout the
intervention period. Physiotherapists can only participate if they are willing to deliver both the intervention treatment
and sham-treatment. Another strategy to minimize contamination is that patients of the intervention and control group will not be in the same physiotherapy setting at the same time. So, the physiotherapists can focus their mind on just one treatment at the time.
Both the physiotherapists and the patients are not blinded during this study, since they are aware of the treatment procedures. Physiotherapists will conduct the measurements as well as the treatment in patients. Due to practical considerations it is not feasible to perform all measurements in many different practice settings by a single researcher.
To assure a high quality and univocal treatment, the participating physiotherapists will be trained and instructed extensively before the start of the training. Also, throughout the intervention period the physiotherapists will be monitored continuously.
The researcher will visit the participating physiotherapy practices frequently and will have regular contact by telephone and email in order to check the compliance with the treatment protocols. As the population in Limburg is the least physically active population of the Netherlands, this might influence the external validity of the study. Another limitation is
that the six-minute walk tests are performed on different tracks, which will influence the variability. An advantage of the randomization on patient level instead of physiotherapy practice level is that patients are assigned to smaller and longer passages in a non-differential manner and an equal distribution of patients from the intervention group and control group can be expected per physiotherapy practice. Since we are interested in the difference scores (4 or 6 months minus baseline measurement) and participants are assessed in the same passage on all occasions, we think that the variability is acceptable.