Nevin Manimala Statistics

Intracluster correlation coefficients in osteoarthritis cluster randomized trials: a systematic review

Osteoarthritis Cartilage. 2023 Sep 15:S1063-4584(23)00919-6. doi: 10.1016/j.joca.2023.09.006. Online ahead of print.


OBJECTIVES: The design, analysis and interpretation of cluster randomized clinical trials (RCTs) requires accounting for potential correlation of observations on individuals within the same cluster. Reporting of observed intracluster correlation coefficients (ICCs) in cluster RCTs, as recommended by CONSORT, facilitates sample size calculation of future cluster RCTs and understanding of the trial statistical power. Our objective was to summarize observed ICCs in osteoarthritis (OA) cluster RCTs.

DESIGN: Systematic review of knee/hip OA cluster RCTs. We searched CENTRAL for trials published from 2012, when CONSORT cluster RCTs extension was published, to September 2022. We calculated the proportion of cluster RCTs that reported observed ICCs. Of those that did, we extracted observed ICCs.

PROSPERO: CRD42022365660.

RESULTS: We screened 1121 references, and included 20 cluster RCTs. Only 5 trials (25%) reported the observed ICC for at least one outcome variable. ICC values for pain outcomes were: 0, 0.01, 0.18; for physical function outcomes were: 0, 0.06, 0.13 (knee)/0.27 (hip); WOMAC total: 0.02, 0.02; symptoms of anxiety/depression: 0.22; disability: 0; and global change: 0. One out of four (25%) trials reported an ICC that was larger than the ICC used for sample size calculation and thus was underpowered.

CONCLUSIONS: Despite CONSORT statement recommendations for reporting of cluster RCTs, few OA trials reported the observed ICC. Given the importance of the ICC to interpretation of trial results and future trial design, this reporting gap warrants attention.

PMID:37717903 | DOI:10.1016/j.joca.2023.09.006

By Nevin Manimala

Portfolio Website for Nevin Manimala