Modeling Oxygen Saturation Targets, Mortality, and Retinopathy Among Extremely Preterm Infants

Importance: According to the Neonatal Oxygenation Prospective Meta-Analysis, a higher oxygen saturation as measured by pulse oximetry (Spo2) target range, compared with a lower Spo2 target, may reduce mortality but increase the risk of treated retinopathy of prematurity (ROP). Choosing a target range at an individual center may depend on the local baseline risks and the value placed on prioritizing survival over the risk of severe retinopathy. Objective: To estimate trade-offs between mortality and treated ROP associated with Spo2 targeting and the association between variation in baseline risks and these trade-offs. Design, Setting, and Participants: This decision analytical modeling study used US national birth data and the Vermont Oxford Network database (2019-2023), incorporating gestational age (GA) and 16 performance subgroups (based on quartiles of center-specific mortality and retinopathy risks), to simulate an annual birth cohort of infants born extremely preterm (GA 23-27 weeks). The data were analyzed between January and August 2025. Exposure: High Spo2 (91%-95%) vs low Spo2 (85%-89%) target ranges. Main Outcomes and Measures: The primary outcome was the risk trade-off ratio (RTOR) of additional cases of treated ROP per death averted. Estimates of outcomes comparing universal adoption of the high or low Spo2 targets were summarized using results of a probabilistic sensitivity analysis. Scenarios were also included that incorporated selective adoption of high or low targets based on performance quartiles. Results: Among 19 050 simulated infants (2372 12.5% born at GA 23 weeks, 3253 17.1% at GA 24 weeks, 3809 20.0% at GA 25 weeks, 4357 22.9% at GA 26 weeks, and 5259 27.6% at GA 27 weeks), which approximated the distribution of extremely preterm births in the US, universal adoption of the higher Spo2 target range was associated with a mean of 638 fewer deaths (95% uncertainty interval UI, -363 to -934 deaths) and a mean of 551 additional cases (95% UI, 435-659 additional cases) of treated ROP. The mean RTOR with the high Spo2 target range was 0.91 (95% UI, 0.61-1.43). Across performance subgroups, RTORs were 0.16 (95% UI, 0.00-0.40) in centers with the highest risk of mortality (4th quartile) and lowest risk of treated ROP (1st quartile) and 2.52 (95% UI, 1.39-4.72) for centers with the highest mortality (1st quartile) and lowest risk of treated ROP (4th quartile). Selective adoption of target ranges was favored in 50% of simulations if 2 cases of treated retinopathy per death averted were considered acceptable. Universal adoption of the high target range was favored in more than 90% of simulations if 4 or more cases of treated ROP were acceptable for each death averted. Conclusions and Relevance: This decision analytical modeling study of a US cohort of extremely preterm infants found that the RTOR for treated ROP and mortality during Spo2 targeting varied by GA and local baseline risks. This decision analysis may help inform the choice of Spo2 target ranges in individual clinical centers.