Clustering of continuous and binary outcomes at the general practice level in individually randomised studies in primary care - a review of 10 years of primary care trials

Abstract

Background: In randomised controlled trials, the assumption of independence of individual observations is fundamental to the design, analysis and interpretation of studies. However, in individually randomised trials in primary care, this assumption may be violated because patients are naturally clustered within primary care practices. Ignoring clustering may lead to a loss of power or, in some cases, type I error. Methods: Clustering can be quantified by intra-cluster correlation (ICC), a measure of the similarity between individuals within a cluster with respect to a particular outcome. We reviewed 17 trials undertaken by the Department of Primary Care at the University of Southampton over the last ten years. We calculated the ICC for the primary and secondary outcomes in each trial at the practice level and determined whether ignoring practice-level clustering still gave valid inferences. Where multiple studies collected the same outcome measure, the median ICC was calculated for that outcome. Results: The median intra-cluster correlation (ICC) for all outcomes was 0.016, with interquartile range 0.00-0.03. The median ICC for symptom severity was 0.02 (interquartile range (IQR) 0.01 to 0.07) and for reconsultation with new or worsening symptoms was 0.01 (IQR 0.00, 0.07). For HADS anxiety the ICC was 0.04 (IQR 0.02, 0.05) and for HADS depression was 0.02 (IQR 0.00, 0.05). The median ICC for EQ. 5D-3 L was 0.01 (IQR 0.01, 0.04). Conclusions: There is evidence of clustering in individually randomised trials primary care. The non-zero ICC suggests that, depending on study design, clustering may not be ignorable. It is important that this is fully considered at the study design phase.

Effectiveness and implementation success of a co-produced physical activity referral scheme in Germany: study protocol of a pragmatic cluster randomised trial

“…With an assumed drop-out rate of 10%, a sample size of 567 participants from 24 medical practices (α = 0.0125, power = 80%) is required to detect differences between both intervention groups of low to moderate effect size ( f = 0.175). To account for the effects of clustering, an intraclass correlation (ICC) of 0.02 is considered, estimated based on previous studies for cRCTs in primary care [ 26 , 27 ]. We consider this ICC to be realistic in our trial as the participating practices are similarly structured (professional background, qualification level, treating patient group size and region) and were trained by the research team.…”

Clinical Observation, Management and Function Of low back pain Relief Therapies (COMFORT): A cluster randomised controlled trial protocol

“…We have assumed an ICC of 0.045 based on 145 ICCs from cluster randomised trials in primary care. 55 This sample size will also provide at least 80% power to detect changes in secondary outcomes, including a minimum difference of 10% for pain (1 point on the 0-10 Numerical Rating Scale; assuming SD of 1.9), 50 and 10% on the 24-item RMDQ (assuming SD of 5.4). 50 This sample size will also provide >80% power to detect a difference of 20% in the proportion of people who continue to use opioids at 1 year and 20% reduction in the proportion of dispensations for strong, long-acting opioids.…”

Methodological aspects of a randomized within-patient concurrent controlled design for clinical trials in spine surgery

“…43,44 As a consequence, studies might be underpowered or estimated treatment effects invalid. 12,45–47 The report of methodological and statistical aspects of the within-patient RCT design, especially regarding the correlation of outcomes within subjects, can further demonstrate its value and assist the set-up of future studies. Another limitation of several previous within-patient controlled trials is the lack of randomization, which unnecessarily limited their level of evidence.…”