Recruitment characteristics of randomised trials in critical care: A systematic review.

Abstract

To summarise the temporal trends of recruitment and methodological characteristics of critical care randomised controlled trials with the primary outcome of mortality. PubMed was searched for articles meeting inclusion and exclusion criteria. Randomised controlled trials, with primary outcome of mortality, of adult and paediatric critical care patients treated in an intensive care unit, were included. Neonatal intensive care unit trials, non-English publications and conference proceedings or abstracts without full-length publications were excluded. Duplicate literature search, article selection and quality assessment were performed by two reviewers with disputes resolved through discussion. Data were extracted into a custom-designed Research Electronic Data Capture database. The search identified 67,199 records of which 230 were included. The annual number of critical care randomised controlled trials published increased gradually over a 30-year period from 0 in 1990 to 19 in 2014 with stabilisation at 8-11 between 2015 and 2020. Twenty-seven percent of randomised controlled trials were low risk in all categories using the Cochrane Risk of Bias tool. Methodological characteristics such as registration on clinical trials registries and data safety monitoring committee presence significantly (p < 0.001) increased over time. The median recruitment was 376 patients (interquartile range 125-895) with significant increase (p = 0.002) from 62 (interquartile range: 33-486) in 1991 to 725 (interquartile range: 537-2600) in 2020. This was accompanied by an increase in recruitment times. Thus overall, recruitment rates did not increase. Early cessation occurred in 23% (54/230) of randomised controlled trials with no temporal trend. The number, size and some methodological qualities of critical randomised controlled trials with primary outcome of mortality have increased over time, but rates of recruitment and early cessation have been unchanged.