Participant and trial characteristics reported in predictive analyses of trial attrition: An umbrella review of systematic reviews of randomised controlled trials across multiple conditions

Abstract

Background Trial attrition poses several risks for the validity of randomised controlled trials (RCTs). To better understand attrition, studies have explored and identified predictors among participant and trial characteristics. Reviews of these have so far been limited to single conditions. We performed an umbrella review to explore which participant and trial characteristics are reported in predictive analyses of trial attrition in systematic reviews of RCTs across multiple conditions. Methods We searched MEDLINE, Embase, Web of Science and the Online Resource for Research in Clinical TriAls for systematic reviews of RCTs that evaluated associations between participant/trial characteristics and attrition. We included quantitative systematic reviews of adult populations that evaluated any participant/trial characteristic and any attrition outcome. Review quality was appraised using R-AMSTAR. A review-level narrative synthesis was conducted. Results We identified 88 reviews of RCTs evaluating characteristics associated with attrition. Included reviews encompassed 33 different conditions. Over half (50/88, 56.8%) were of RCTs for psychological conditions. All but one examined trial characteristics (87/88, 98.9%) and fewer than half (42/88, 47.7%) evaluated participant characteristics. Reviews typically reported on participant age (33/42, 78.6%), sex (29/42, 69.1%) and the type (13/42, 31%) or severity (10/42, 23.8%) of an index condition. Trial characteristics typically reported on were intervention type (56/87, 64.4%), intervention frequency/intensity (29/87, 33.3%), intervention delivery/format (26/87, 29.9%), trial duration (16/87, 18.4%), publication/reporting year (15/87, 17.2%) and sample size (15/87, 31.9%). Retention strategies were rarely reported (2/87, 2.3%). No characteristic was examined for every condition. Some reviews of certain conditions found that age (12/33, 36.4%), intervention type (29/56, 51.8%) and trial duration (9/16, 56.3%) were associated with attrition, but no characteristic was reportedly associated across multiple conditions. Conclusions Across conditions, reviews conducting predictive analyses of attrition in RCTs typically report on several characteristics. These are participant age, sex and the type or severity of index condition, as well as the type, frequency or intensity and delivery or format of a trial intervention, trial duration, publication/reporting year and sample size. Future studies should consider exploring these characteristics as a core set when evaluating predictive factors of attrition in RCTs across multiple conditions. Registration PROSPERO: CRD42023398276