How Much Oxygen Is Right for a Baby Born Very Early?

What a 2026 computer model from the United States (King et al., JAMA Network Open) and a survey of newborn units in 13 countries (Darlow et al., Neonatology) tell us about balancing survival against an eye disease of prematurity

Babies born very early often need extra oxygen, and their care team chooses a target range for how much oxygen their blood should carry. A slightly higher range saves some lives, but a few more babies then need treatment for an eye disease of prematurity. Two 2026 studies, a US computer model and a survey of newborn units in 13 countries, show that this balance differs from hospital to hospital and that units still set their targets quite differently.

Why oxygen is both life-saving and risky for very premature babies

This article explains two new studies: a computer model by King and colleagues in the United States [1] and an international survey by Darlow and colleagues [2]. To understand them, it helps to know how oxygen is checked in a newborn intensive care unit (NICU). Babies wear a small clip-on light sensor called a pulse oximeter. It shows the "oxygen saturation", the percentage of the blood's oxygen carriers that are loaded with oxygen. Nurses and machines adjust the oxygen the baby breathes to keep this number within a chosen range. An expert group from the American Academy of Pediatrics has explained why this matters: too much oxygen can harm the lungs and the eyes, but too little may also cause harm [3].

The eye problem is called retinopathy of prematurity, or ROP. In babies born very early, the blood vessels at the back of the eye (the retina) have not finished growing, and they can grow abnormally. Babies born very early have regular eye checks so that serious ROP can be found and treated to protect sight.

Keeping a tiny baby exactly in range is hard. In one study of 84 babies born before 28 weeks of pregnancy, at 14 hospitals in 3 countries, called the AVIOx study (achieved versus intended pulse oximeter saturation), babies spent only between 16% and 64% of their time inside the intended range, and were more often above it than below it [4].

How doctors searched for the right range

For years nobody knew whether a lower or a higher range was safer. So, between 2005 and 2014, five large trials randomly assigned babies born before 28 weeks to a lower range (85–89%) or a higher one (91–95%) [5]. One of them, the Surfactant, Positive Pressure, and Oxygenation Randomized Trial (SUPPORT), was carried out in a group of US hospitals called the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) Neonatal Research Network. It found that more babies died before going home with the lower range, but fewer developed severe ROP [6]. The Canadian Oxygen Trial (COT), run at 25 hospitals in Canada, the United States, Argentina, Finland, Germany and Israel, found no clear difference in death or disability at 18 months [7].

Researchers then combined the five trials in one large analysis called the Neonatal Oxygenation Prospective Meta-analysis (NeOProM), which included 4965 babies. Put simply, for every 100 babies, about 20 died with the lower range compared with about 17 with the higher range. Meanwhile, about 15 in 100 needed eye treatment with the higher range compared with about 11 in 100 with the lower range [5]. Overall rates of death or disability by age 2 were similar. Our earlier article on these trials explains them in more detail.

That left hospitals with a hard choice rather than a simple answer. Experts suggested that each hospital should weigh the trade-off using its own situation, such as how often its babies die or need eye treatment [8][9]. Many units did change. In a 2015 survey of units belonging to the International Network for Evaluating Outcomes in Neonates (iNeo), 60% had recently changed their targets, mostly upwards [10]. In a 2021 survey of 170 US units, 78% had changed, yet there was still a lot of variation [11].

Two new studies looking at the same question from different sides

The two 2026 studies did not refer to each other, but they fit together well. The first, by King and colleagues, asks: if the trial results are applied to real American hospitals, how many lives would a higher range save, and how many more babies would need eye treatment? The second, by Darlow and colleagues, repeats the 2015 iNeo survey to ask what units actually do today. One study is about what the evidence implies; the other is about what happens in practice. Putting them side by side shows where the two match and where they do not.

What the computer model found

King and colleagues did not treat any babies. They built a computer model of one year of US births between 23 and 27 weeks of pregnancy, 19 050 babies in all, using real outcome data from hospitals in the Vermont Oxford Network, a large group of newborn units [1]. Their key measure was simple: how many extra babies need eye treatment for each baby's life saved by the higher range?

For the whole country, using the higher range everywhere was linked to about 638 fewer deaths each year and about 551 more babies treated for ROP [1]. Put another way, for every 1000 babies born this early, about 33 more would survive, and among every 1000 survivors about 29 more would need eye treatment. On average that is slightly less than one extra treated baby for each life saved. The higher range also meant fewer cases of a serious bowel illness needing surgery, but more babies still needing oxygen at what would have been 36 weeks of pregnancy.

The striking finding was how much this varied between hospitals. Outcomes already differ a great deal from one US unit to another [12]. In hospitals where babies were more likely to die but less likely to need eye treatment, the higher range cost only about 1 extra treated baby for every 6 lives saved. In hospitals with low death rates but high rates of eye treatment, it meant about 2.5 extra treated babies for each life saved [1]. If people judge that 4 or more extra treated babies per life saved is an acceptable price, using the higher range in every hospital came out best in more than 9 out of 10 model runs.

A model is only as good as the information put into it, and the authors are open about its limits. It used outcome data from US hospitals only, so the numbers may not apply directly elsewhere. It assumed that the trial results apply in the same way at every hospital, and it counted only deaths and eye treatment, not every other illness a premature baby can face [1].

What newborn units around the world actually do

Darlow and colleagues sent a questionnaire in July 2023 to the directors of 608 units in 12 networks across 13 countries, including Australia, New Zealand, Canada, Japan, several European countries and Brazil. In all, 382 units (63%) replied [2]. For babies born before 29 weeks, the most usual top of the range was 95%. The bottom of the range was most often 90%; about 1 unit in 4 used 88%, and about 1 unit in 8 still used 85% or lower. Targets inside each network had changed little since 2015.

About a quarter of units used machines that adjust oxygen automatically for some babies, from 7 in 10 units in Israel to none in Finland. Most units decided which babies needed eye checks using both birth weight and length of pregnancy, but the exact rules differed. Eye injections, a newer treatment for ROP, were used in 43% of units in 2015 and 76% in 2023 [2].

What this means for families

If your baby is on oxygen, the target range, and the alarm limits set around it, are a deliberate choice, made to balance survival against the risk of eye disease. The evidence suggests that, for most hospitals, a slightly higher range saves more lives than it costs in eye treatment. It is reasonable to ask your baby's team what range they use and why.

ROP treatment is improving. Laser treatment was standard for many years, and eye injections are increasingly used instead [13]. In the RAINBOW trial, which compared an injected medicine called ranibizumab with laser in 225 babies at 87 hospitals in 26 countries, treatment worked in 80 of 100 babies given the higher dose of ranibizumab and 66 of 100 given laser, although the difference was not quite large enough to be certain [14]. Five years later, severe short-sightedness was less common in children treated with the higher dose of ranibizumab (8 in 100) than with laser (24 in 100) [15]. Doctors still take ROP seriously, because babies who need treatment can have other developmental needs as they grow [16][17]. That is one reason follow-up visits after the NICU matter.

What researchers are working on next

Many units, especially in Canada, use "in-between" ranges such as 88–93%. A large Canadian study found that more babies survived without major complications with that range than with 90–95%, but it was not a randomised trial, so it cannot prove cause and effect [18]. The computer model did not test these ranges [1]. Researchers are also studying machines that adjust oxygen automatically. These probably keep babies in range for more of the time [19], but a trial of 1082 babies at 32 units in China, Germany, the Netherlands and the United Kingdom, called FiO2-C (automated control of the oxygen a baby breathes), found no clear difference in serious outcomes [20]. Darlow and colleagues now plan to link each unit's practices to how its babies do [2]. Together, these studies help each hospital make a careful, well-explained choice about the oxygen its smallest patients receive.

References

  1. King BC, Zupancic JAF, Edwards EM, et al. Modeling Oxygen Saturation Targets, Mortality, and Retinopathy Among Extremely Preterm Infants. JAMA Netw Open. 2026;9(7):e2624727. doi:10.1001/jamanetworkopen.2026.24727 ↩
  2. Darlow BA, Hurrion EM, Battin M, et al. Oxygen Saturation Targeting and Retinopathy Management in Very Preterm Neonates: An International Survey. Neonatology. 2026:1-9. doi:10.1159/000552726 ↩
  3. Cummings JJ, Polin RA. Oxygen Targeting in Extremely Low Birth Weight Infants. Pediatrics. 2016;138(2):e20161576. doi:10.1542/peds.2016-1576 ↩
  4. Hagadorn JI, Furey AM, Nghiem TH, et al. Achieved versus intended pulse oximeter saturation in infants born less than 28 weeks' gestation: the AVIOx study. Pediatrics. 2006;118(4):1574-1582. doi:10.1542/peds.2005-0413 ↩
  5. Askie LM, Darlow BA, Finer N, et al. Association Between Oxygen Saturation Targeting and Death or Disability in Extremely Preterm Infants in the Neonatal Oxygenation Prospective Meta-analysis Collaboration. JAMA. 2018;319(21):2190-2201. doi:10.1001/jama.2018.5725 ↩
  6. Carlo WA, Finer NN, Walsh MC, et al. Target ranges of oxygen saturation in extremely preterm infants. N Engl J Med. 2010;362(21):1959-1969. doi:10.1056/NEJMoa0911781 ↩
  7. Schmidt B, Whyte RK, Asztalos EV, et al. Effects of targeting higher vs lower arterial oxygen saturations on death or disability in extremely preterm infants: a randomized clinical trial. JAMA. 2013;309(20):2111-2120. doi:10.1001/jama.2013.5555 ↩
  8. Askie LM, Darlow BA, Davis PG, et al. Effects of targeting lower versus higher arterial oxygen saturations on death or disability in preterm infants. Cochrane Database Syst Rev. 2017;4(4):CD011190. doi:10.1002/14651858.CD011190.pub2 ↩
  9. Schmidt B, Whyte RK. Oxygen saturation target ranges and alarm settings in the NICU: What have we learnt from the neonatal oxygenation prospective meta-analysis (NeOProM)?. Semin Fetal Neonatal Med. 2020;25(2):101080. doi:10.1016/j.siny.2020.101080 ↩
  10. Darlow BA, Vento M, Beltempo M, et al. Variations in Oxygen Saturation Targeting, and Retinopathy of Prematurity Screening and Treatment Criteria in Neonatal Intensive Care Units: An International Survey. Neonatology. 2018;114(4):323-331. doi:10.1159/000490372 ↩
  11. Parikh P, White RD, Tolia VN, Reyburn B, Guillory C, Ahmad KA. Variability in Practice and Implementation of Oxygen Target Saturation Policies in United States' Neonatal Intensive Care Units. Am J Perinatol. 2024;41(12):1728-1735. doi:10.1055/s-0044-1780523 ↩
  12. Horbar JD, Edwards EM, Greenberg LT, et al. Variation in Performance of Neonatal Intensive Care Units in the United States. JAMA Pediatr. 2017;171(3):e164396. doi:10.1001/jamapediatrics.2016.4396 ↩
  13. Hartnett ME, Stahl A. Laser versus Anti-VEGF: A Paradigm Shift for Treatment-Warranted Retinopathy of Prematurity. Ophthalmol Ther. 2023;12(5):2241-2252. doi:10.1007/s40123-023-00744-7 ↩
  14. Stahl A, Lepore D, Fielder A, et al. Ranibizumab versus laser therapy for the treatment of very low birthweight infants with retinopathy of prematurity (RAINBOW): an open-label randomised controlled trial. Lancet. 2019;394(10208):1551-1559. doi:10.1016/S0140-6736(19)31344-331344-3) ↩
  15. Marlow N, Reynolds JD, Lepore D, et al. Ranibizumab versus laser therapy for the treatment of very low birthweight infants with retinopathy of prematurity (RAINBOW): five-year outcomes of a randomised trial. EClinicalMedicine. 2024;71:102567. doi:10.1016/j.eclinm.2024.102567 ↩
  16. Schmidt B, Davis PG, Asztalos EV, Solimano A, Roberts RS. Association between severe retinopathy of prematurity and nonvisual disabilities at age 5 years. JAMA. 2014;311(5):523-525. doi:10.1001/jama.2013.282153 ↩
  17. Hellström A, Smith LEH, Hård AL. ROP: 80 Years after Its Detection - Where Do We Stand and How Long Will We Continue to Laser?. Neonatology. 2024;121(5):608-615. doi:10.1159/000538907 ↩
  18. Taylor RS, Singh B, Mukerji A, et al. Intermediate vs. High Oxygen Saturation Targets in Preterm Infants: A National Cohort Study. Neonatology. 2025;122(1):106-113. doi:10.1159/000540278 ↩
  19. Stafford IG, Lai NM, Tan K. Automated oxygen delivery for preterm infants with respiratory dysfunction. Cochrane Database Syst Rev. 2023;11(11):CD013294. doi:10.1002/14651858.CD013294.pub2 ↩
  20. Franz AR, Maiwald CA, Niemarkt HJ, et al. Automatic versus manual control of oxygen and neonatal clinical outcomes in extremely preterm infants: a multicentre, parallel-group, randomised, controlled, superiority trial. Lancet Child Adolesc Health. 2026;10(3):179-188. doi:10.1016/S2352-4642(25)00351-700351-7) ↩