Balancing precision and affordability in assessing infant development in large-scale mortality trials: secondary analysis of a randomised controlled trial.

Objective: Large-scale mortality trials require reliable secondary assessments of impairment. We compared the Ages and Stages Questionnaire (ASQ-3), a screening tool self-administered by parents, in classifying impairment using the 'gold standard' Bayley Scales of Infant Development (Bayley-III), a diagnostic tool administered by trained assessors.

Design: Analysis of 405 children around 2 years corrected age from the Australian Placental Transfusion Study, a trial conducted over 8 years.

Setting: Secondary analysis of international, open-label, multicentre randomised trial.

Patients: Children born <30 weeks gestation.

Interventions: Immediate (<10 s) versus delayed (60 s+) cord clamping.

Main outcomes: ASQ-3 and Bayley-III assessments around 2 years corrected age. Impairment (or developmental delay) was defined as <2 SD below the mean (<70) for Bayley-III domains.

Results: The area under the receiver operating curve for ASQ-3 domains predicting delay was 0.75-0.99. Sensitivity for predicting delay was 57%-100%, while specificity was 88%-90%.We modelled the cost and sample size using a less expensive, though less precise, screening assessment for impairment compared with a more costly diagnostic assessment. For detecting a 25% reduction in the relative risk of delay, using ASQ-3 rather than Bayley-III could require double the sample size (15 000 to 30 000), but outcome assessment cost savings would be US$13M (EUR$12M). However, assessment cost savings may be outweighed by upscaling.

Conclusions: When measuring developmental outcomes in a large-scale clinical trial, using a more precise diagnostic tool may be financially prohibitive, so increasing the sample size and using a less precise but appropriately calibrated tool may be more affordable.

Trial registration number: ACTRN12610000633088.