Early stopping of randomized controlled trials in pediatric critical care

Duffett M, Xu G, Cook DJ

Aims & Objectives

Early stopping of randomized controlled trials (RCTs) occurs for ethical, scientific, and resource reasons. Our objective was to describe the prevalence and characteristics of RCTs stopped early in pediatric critical care.

Methods

We included RCTs in pediatric critical care published in English from 1986–2017. We considered a trial to be stopped early if less than 90% of the intended sample size was randomized.

Results

157 (44%) of 354 published RCTs reported their planned sample size. Of these, 39 (25%) were stopped early: 15 (38%) for futility, 8 (21%) for poor recruitment, 6 (15%) for benefit, 3 (8%) for harm, 3 (8%) for other reasons and for 4 (10%) the reason was unclear. RCTs stopped early recruited a median (IQR) of 50% (35–65%) of their planned sample size. When compared to RCTs that were fully completed, RCTs stopped early had a larger planned sample size (200 [109, 320] vs. 73 [43, 144], p<0.001), randomized more children (82 [45–125] vs 78 [44–147], p<0.001), recruited participants for longer (28 [20–39] vs. 25 [15–38] months, p<0.001), were more often multicentred (51% vs. 26%, p=0.007), and their publications were cited more often (3.7 [2.0–8.4] vs. 2.6 [1.0–5.4] citations/year, p<0.001). Using logistic regression, multicentre design, larger planned sample size, a pharmaceutical intervention, and industry funding were not independently associated early stopping.

Conclusions

Early stopping occurred in one quarter of pediatric critical care trials, most often because of futility or poor recruitment. Reducing the incidence of early stopping remains a challenge for the pediatric critical care research community.

This abstract was presented orally at the World Congress on Pediatric Intensive and Critical Care (June 2018; Singapore).