RESUMEN
BACKGROUND: The safety and efficacy of bubble continuous positive airway pressure (bCPAP) for treatment of childhood severe pneumonia outside tertiary care hospitals is uncertain. We did a cluster-randomised effectiveness trial of locally made bCPAP compared with WHO-recommended low-flow oxygen therapy in children with severe pneumonia and hypoxaemia in general hospitals in Ethiopia. METHODS: This open, cluster-randomised trial was done in 12 general (secondary) hospitals in Ethiopia. We randomly assigned six hospitals to bCPAP as first-line respiratory support for children aged 1-59 months who presented with severe pneumonia and hypoxaemia and six hospitals to standard low-flow oxygen therapy. Cluster (hospital) randomisation was stratified by availability of mechanical ventilation. All children received treatment in paediatric wards (in a dedicated corner in front of a nursing station) with a similar level of facilities (equipment for oxygen therapy and medications) and staffing (overall, one nurse per six patients and one general practitioner per 18 patients) in all hospitals. All children received additional care according to WHO guidelines, supervised by paediatricians and general practitioners. The primary outcome was treatment failure (defined as any of the following: peripheral oxygen saturation <85% at any time after at least 1 h of intervention plus signs of respiratory distress; indication for mechanical ventilation; death during hospital stay or within 72 h of leaving hospital against medical advice; or leaving hospital against medical advice during intervention). The analysis included all children enrolled in the trial. We performed both unadjusted and adjusted analyses of the primary outcome, with the latter adjusted for the stratification variable and for the design effect of cluster randomisation, as well as selected potentially confounding variables, including age. We calculated effectiveness as the relative risk (RR) of the outcomes in the bCPAP group versus low-flow oxygen group. This trial is registered with ClinicalTrial.gov, NCT03870243, and is completed. FINDINGS: From June 8, 2021, to July 27, 2022, 1240 children were enrolled (620 in hospitals allocated to bCPAP and 620 in hospitals allocated to low-flow oxygen). Cluster sizes ranged from 103 to 104 children. Five (0·8%) of 620 children in the bCPAP group had treatment failure compared with 21 (3·4%) of 620 children in the low-flow oxygen group (unadjusted RR 0·24, 95% CI 0·09-0·63, p=0·0015; adjusted RR 0·24, 0·07-0·87, p=0·030). Six children died during hospital stay, all of whom were in the low-flow oxygen group (p=0·031). No serious adverse events were attributable to bCPAP. INTERPRETATION: In Ethiopian general hospitals, introduction of locally made bCPAP, supervised by general practitioners and paediatricians, was associated with reduced risk of treatment failure and in-hospital mortality in children with severe pneumonia and hypoxaemia compared with use of standard low-flow oxygen therapy. Implementation research is required in higher mortality settings to consolidate our findings. FUNDING: SIDA Sweden and Grand Challenges Ethiopia.