Identifying effect modifiers of systemic hydrocortisone treatment initiated 7-14 days after birth in ventilated very preterm infants on long-term outcome: secondary analysis of a randomised controlled trial.

Objective: To explore clinical effect modifiers of systemic hydrocortisone in ventilated very preterm infants for survival and neurodevelopmental outcome at 2 years' corrected age (CA).

Design: Secondary analysis of a randomised placebo-controlled trial.

Setting: Dutch and Belgian neonatal intensive care units.

Patients: Infants born <30 weeks' gestational age (GA), ventilator-dependent in the second week of postnatal life.

Intervention: Infants were randomly assigned to systemic hydrocortisone (cumulative dose 72.5 mg/kg; n=182) or placebo (n=190).

Main Outcome Measures: The composite of death or neurodevelopmental impairment (NDI) at 2 years' CA and its components. Candidate effect modifiers (GA, small for GA, respiratory index, sex, multiple births, risk of moderate/severe bronchopulmonary dysplasia or death) were analysed using regression models with interaction terms and subpopulation treatment effect pattern plots.

Results: The composite outcome was available in 356 (96.0%) of 371 patients (one consent withdrawn). For this outcome, treatment effect heterogeneity was seen across GA subgroups (<27 weeks: hydrocortisone (n=141) vs placebo (n=156), 54.6% vs 66.2%; OR 0.61 (95% CI 0.38 to 0.98); ≥27 weeks: hydrocortisone (n=30) vs placebo (n=31), 66.7% vs 45.2%; OR 2.43 (95% CI 0.86 to 6.85); p=0.02 for interaction). This effect was also found for the component death (<27 weeks: 20.1% vs 32.1%; OR 0.53 (95% CI 0.32 to 0.90); ≥27 weeks: 28.1% vs 16.1%; OR 2.04 (95% CI 0.60 to 6.95); p=0.049 for interaction) but not for the component NDI. No differential treatment effects were observed across other subgroups.

Conclusion: This secondary analysis suggests that in infants <27 weeks' GA, systemic hydrocortisone may improve the outcome death or NDI, mainly driven by its component death. There was insufficient evidence for other selected candidate effect modifiers.