Role of recruitment bias in stepped-wedge cluster randomised controlled trials: a systematic review

Increased popularity of stepped-wedge cluster randomised trials (SW-CRT) highlights the importance of understanding and appropriate mitigation of sources of bias within this trial design. While current evidence suggests that 'conventional' cluster randomised controlled trials (RCTs) are at a higher risk of recruitment bias than individually randomised trials, this review aims to estimate the risk of recruitment bias in SW-CRTs.

Systematic review with search conducted on four databases. Risk of bias (RoB) was assessed using subdomain 1a (randomisation process) and 1b (timing of identification or recruitment of participants) of the Cochrane RoB tool 2.0 (extension for cluster RCTs).

MEDLINE, Embase, CINAHL, Cochrane Library were searched on 9 February 2024.

SW-CRTs published in 2023 were included.

Two independent reviewers screened and extracted all eligible papers. RoB was assessed with the Cochrane RoB tool.

Overall, 808 papers were screened, and 64 studies were included in the review. Most studies were deemed to have a high RoB (n=35, 55%), some concerns were noticed in 20 studies (31%), and 9 (14%) were considered to have a low RoB. The description of the randomisation process in the included papers was sometimes poorly reported (in 15 studies (23%) problems with the randomisation process were identified), and 21 studies (33%) had issues with sampling strategy (recruiting participants after randomisation by unmasked staff).

The review revealed that SW-CRTs are prone to recruitment bias, but the risks are comparable to cluster RCTs. When SW-CRTs are unable to recruit prior to randomisation, mitigation strategies could be implemented to reduce bias. A separate tool for RoB assessment in SW-CRTs is required to address the complexities of this trial design.