Individual patient data meta-analysis of high-frequency oscillatory ventilation trials in preterm infants: addressing questions left unanswered by aggregate data meta-analysis

Article type
Authors
Cools F, Askie L, Offringa M
Abstract
Background: The interpretation of aggregate data meta-analysis of trials comparing high-frequency oscillatory ventilation (HFOV) with conventional ventilation (CV) in preterm infants with respiratory distress syndrome has been confounded by heterogeneity in study design and outcomes. Also, it does not allow exploration of how patient-level characteristics might affect the treatment effect. Objectives: To determine if HFOV offers benefit over CV for neonatal outcomes, to identify subgroups of infants who might benefit more or less from HFOV, and to explore effect-modifying factors. Methods: A collaborative study group (PreVILIG) provided all available original patient level data from elective HFOV trials. New variables and new composite primary outcomes were created using pre-specified common definitions. A two-stage method was used for the meta-analyses. Results: Data from 3229 preterm infants (89% of all eligible infants) were obtained. We found no difference in effect between HFOV and CV for the outcomes ‘death or bronchopulmonary dysplasia at 36 weeks postmenstrual age’ and ‘death or severe adverse neurological event’. There was no evidence that either type of ventilator or ventilation strategy, or patient characteristics such as gestational age at birth, intrauterine growth retardation or initial lung disease severity modified the overall treatment effect. Compared with infants randomised more than 4 hours after intubation, infants randomised within 4 hours benefited significantly more from HFOV. Compared with infants whose mother did, infants whose mothers did not receive antenatal corticosteroids benefited more from HFOV in terms of duration of respiratory support. Conclusion: This first neonatal IPD meta-analysis addressed some important limitations of the existing aggregate data meta-analysis, hence improving the quality of the evidence. In addition, it provides new information regarding effect modification by several patient-and trial-level factors. Additional one-stage analyses are planned to elucidate further the complex effect modifications.