Difficulties with cluster randomized trials in primary care

Cluster randomized trials (cRCT) are a popular and common design option for trials based in primary care. Because the number of eligible patients are limited within each primary care facility, to collect a large enough sample it becomes necessary to involve multiple health care providers from multiple health care clinics. This raises the question of confounders due to location and intervention.

In 2016 we published a cRCT on a cognitive based education program for patients with subacute or chronic back pain recruited from primary care facilities in Norway (the COPE study) [1]. 16 GPs and 20 PTs participated and recruited 216 patients to the study. In the study, an “Explain Pain” model was adapted to fit into Norwegian practices. The results provided evidence of no benefit from the intervention compared to usual care.

This lack of effect motivated us to search for explanations. In order to understand more about how each provider delivered the intervention, we conducted a qualitative study involving the participating health care providers (open access, free to download) [2]. All 36 health care providers from the intervention and the control group were invited to take part in a focus group discussion that investigated their experiences from the study. Six clinicians from the intervention group and four from the control group accepted the invitation.

The study identified a number of factors that could have influenced the results of the COPE study. Firstly, we found several potential selection biases. The providers reported not including patients when they had a gut feeling that the patient would not accept the pain education messages of the study. Logistical factors included patients being inadvertently overlooked for the study if the secretary or the provider forgot to ask for participation. Further, if the patient came in late in the evening, it was reported as bothersome to add the extra load the study demanded to their appointment. Some providers also admitted that they needed good chemistry with the patient to be willing to spend so much time with him or her.

Further and perhaps factors more influential to a poor outcome included difficulties with adhering to the methods of the study accurately. Some providers were unsure or did not actually believe in parts of the pain education messages and tended to skip the difficult parts of the pain education manual. There was also a possibility of double communication when the providers offered specific treatment, like manipulation, in addition to the message of back pain as a self-recovering condition.

In the control group we found that most of the providers were highly skilled and very interested in back pain and probably gave pain education messages very much in line with the “Explain Pain” model.  Because both the intervention and control providers reported providing similar interventions, this raises the possibility that the differences between the groups is likely to have been small..

In conclusion, we believe cRCT is the most appropriate methodology to use for trials in a primary care environment  but care must be taken in the design of trials to minimise potential biases and confounders. We would encourage researchers planning a cRCT make use of  tools like the Cochrane risk of bias tool, the PRECIS-2 tool and the Ottawa statement on the ethics of cRCT.

About Erik L. Werner

Erik L. Werner is a professor in general practice at the University of Oslo, Norway. With special interest in low back pain, he has conducted several studies on low back pain for 15 years. His ph.d. dissertation was based on a media campaign on coping with back pain, largely inspired by the Australian mass media campaign in 1997 “Back Pain: Don’t take it lying down”.

References

[1] Werner EL, Storheim K, Løchting I, Wisløff T, Grotle M. Cognitive patient education for low back pain in primary care: a cluster randomized controlled trial and cost-effectiveness analysis. Spine; 2016; 41 (6): 455-62

[2] Werner EL, Løchting I, Storheim K, Grotle M. A focus group study to understand biases and confounders in a cluster randomized controlled trial on low back pain in primary care in Norway. BMC Fam Pract 2018; 19:71