When the Machine Adjusts the Oxygen: What Research Says About Automatic Oxygen Control for Premature Babies

A review of sixteen studies comparing computer-controlled oxygen with nurse-controlled oxygen in premature newborns, and what it means for families

If your baby is in a neonatal intensive care unit and needs help with breathing, one of the most constant tasks at the bedside is adjusting how much oxygen your baby receives. A review of sixteen randomised studies involving 431 premature babies found that when a computer made those adjustments automatically, babies spent roughly 12 percentage points more of the day within their intended oxygen range — about three extra hours out of every twenty-four — and nurses had to make far fewer manual changes. What the research has not yet shown is whether that steadier oxygen leads to healthier babies in the long run.

Why Oxygen Is Such a Delicate Balance

Babies born very early have lungs that are not finished developing, and many need extra oxygen to survive. But oxygen is a medicine, and like any medicine the dose matters. Too little oxygen — called hypoxaemia — starves the brain and other organs. Too much oxygen — hyperoxaemia — damages the delicate blood vessels growing in a premature baby's eyes and can contribute to a serious eye condition called retinopathy of prematurity, or ROP, which in its severe forms can threaten sight.

This is not a new discovery, but it took a long time to pin down. In the 1990s, Flynn and colleagues followed a group of premature babies and measured their oxygen levels continuously, showing that the more time a baby spent with high oxygen, the more likely and more severe the eye disease became [1]. That finding pushed neonatal units toward giving less oxygen — but a large review of the evidence at the time found that nobody actually knew where the safe middle ground was [2].

The question was finally answered by an international project called NeOProM, which combined the individual records of babies from five separate large trials. It found that aiming for a lower oxygen saturation range of 85 to 89 percent led to more deaths and disability than aiming for 91 to 95 percent [3]. Most units around the world now use a target somewhere in the low-to-mid nineties. (Oxygen saturation, usually written SpO₂, is the percentage of the blood's oxygen-carrying capacity that is actually being used. It is measured painlessly by a small light-sensor taped to a hand or foot.)

The Problem Nobody Could Solve

Knowing the right target turned out to be much easier than actually hitting it. A study called AVIOx looked at what oxygen levels babies born before 28 weeks were actually achieving, compared with what their units intended, and found a considerable gap [4]. Another group of researchers asked the question directly in their paper's title — can oxygen saturation be maintained in the desired range? — and the answer was largely no [5].

This is not because anyone is careless. A premature baby's oxygen level can swing within seconds, often because the baby pauses in their breathing, and a nurse who is also caring for one or two other babies simply cannot respond to every fluctuation. Researchers confirmed this by showing that how well a unit hits its oxygen targets depends partly on how many babies each nurse is looking after [6]. If you have sat at a bedside and watched the numbers rise and fall while a nurse reaches over to turn a dial, you have seen this problem in action.

The idea of letting a computer do the adjusting is surprisingly old. Claure and colleagues published one of the first controlled tests of a closed-loop oxygen controller back in 2001, in babies who were having frequent drops in oxygen [7], and Urschitz and colleagues ran the first randomised trial of such a device in 2004 [8]. "Closed-loop" simply means the machine watches the oxygen sensor and adjusts the oxygen supply on its own, second by second, without waiting for a person. Over the following two decades a number of small trials followed, including a randomised trial of a system called CLAC [9] and a multi-hospital study of automated control in ventilated premature babies [10]. But each study was small, each used a different machine, and no single one gave a clear answer.

What the Researchers Did

A team from the Children's Hospital of Chongqing Medical University in China — Yihan Zhang, Yuxuan Du and Yuan Shi — set out to combine all these small studies into one clearer picture [11]. This kind of study is called a systematic review and meta-analysis: researchers search medical databases thoroughly, decide in advance what kinds of study they will include, and then pool the numerical results mathematically so that many small studies can speak with one larger voice.

They searched eight databases in May 2022 and registered their plan publicly beforehand so their methods could not be changed to suit the results. They accepted only randomised trials — studies in which chance, not a doctor's preference, decides which treatment a baby receives, which is the strongest design for a fair comparison. Sixteen trials met their standards, involving 431 premature babies in total. Most of these babies were born extremely early, at around 25 to 28 weeks of pregnancy, and were studied when they were a few weeks old.

Almost all of the trials used a "crossover" design, meaning each baby spent some hours on automatic oxygen control and some hours on nurse-controlled oxygen, so each baby served as their own comparison. The study periods were short — anywhere from 2 to 24 hours. The machines used varied considerably from study to study, which matters for how the results should be read.

What They Found

The headline result is that automatic control kept babies in their intended oxygen range for about 12 percentage points more of the time than nurse-controlled oxygen. Across a full day, that is close to three extra hours in the safe zone. When researchers looked only at the studies using a wider target of 85 to 96 percent, the benefit was about 10.7 percentage points, and the studies agreed with each other closely.

Most of that improvement came from cutting down time spent with oxygen levels too high — a reduction of about 7 percentage points. Time spent too low fell by about 3 percentage points overall, and time spent seriously low (below 80 percent saturation) fell by about 1 percentage point.

Two other findings are worth knowing. First, nurses had to make dramatically fewer manual oxygen changes when the machine was running — a real reduction in a repetitive, interrupting task. Second, and reassuringly, the babies' average oxygen level and average oxygen dose were essentially unchanged. The machine was not giving more oxygen or less oxygen overall; it was giving it more steadily. The researchers also found that automatic control did not reduce episodes of a slow heart rate, called bradycardia.

What This Means for Families

If your baby's unit uses automatic oxygen control, this evidence suggests your baby will spend more time at the oxygen level their team is aiming for, and that this happens without the machine quietly overdosing or underdosing them. That is a genuinely good thing, and it frees your nurse's attention for the many other parts of your baby's care.

It is equally important to be clear about what the research does not show. Four hundred and thirty-one babies studied for a single day each cannot tell us whether automatic control leads to less eye disease, less lung disease, or better survival. Researchers strongly suspect steadier oxygen should help — there is good evidence linking patterns of repeated oxygen dips to eye disease [12] — but suspicion is not proof.

It is also worth knowing that automatic control does not replace your baby's nurse, and it should not. When a baby's oxygen need suddenly rises, the important question is why: a blocked breathing tube, an infection, a problem with the lungs. A machine can compensate for the change; only a clinician can find the cause. Ask your baby's team how they monitor oxygen trends over the day — it is a reasonable and welcome question.

What Researchers Are Working On Next

The biggest question — does steadier oxygen actually make babies healthier? — has now been tested directly. A large international trial called FiO₂-C was designed specifically to find out, planning to enrol more than two thousand babies born between 23 and 28 weeks and to follow them for death, bowel disease, lung disease and severe eye disease [13]. Its results, published in 2026 by Franz and colleagues from 32 hospitals in China, Germany, the Netherlands and the United Kingdom, found no significant difference in these outcomes between babies on automatic control and babies on nurse control [14]. The trial had to stop early because it could not recruit enough babies, which weakens its conclusions — but importantly, it also found no evidence that automatic control is harmful.

So the field sits in an honest middle ground: automatic oxygen control demonstrably does what it was designed to do, and it appears safe, but it has not yet been shown to change the outcomes that matter most to families. Researchers are now focusing on comparing specific machines against each other rather than treating them as interchangeable, on measuring whether automation genuinely reduces nursing workload and alarm fatigue, and on improving the algorithms — particularly so they do not increase oxygen during a pause in breathing, when the extra oxygen cannot reach the lungs anyway. An earlier pooled analysis reached broadly similar conclusions about the benefits of automation [15], and each new generation of research narrows the remaining uncertainty a little further.

References

  1. Flynn JT, Bancalari E, Snyder ES, et al. A cohort study of transcutaneous oxygen tension and the incidence and severity of retinopathy of prematurity. N Engl J Med. 1992;326(16):1050–1054. doi:10.1056/NEJM199204163261603 ↩
  2. Askie LM, Henderson-Smart DJ, Ko H. Cochrane review: restricted versus liberal oxygen exposure for preventing morbidity and mortality in preterm or low birth weight infants. Evid Based Child Health. 2010;5(1):371–413. doi:10.1002/14651858.CD001077.pub2 ↩
  3. 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 ↩
  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. Laptook AR, Salhab W, Allen J, Saha S, Walsh M. Pulse oximetry in very low birth weight infants: can oxygen saturation be maintained in the desired range? J Perinatol. 2006;26(6):337–341. doi:10.1038/sj.jp.7211500 ↩
  6. Sink DW, Hope SA, Hagadorn JI. Nurse:patient ratio and achievement of oxygen saturation goals in premature infants. Arch Dis Child Fetal Neonatal Ed. 2011;96(2):F93–F98. doi:10.1136/adc.2009.178616 ↩
  7. Claure N, Gerhardt T, Everett R, Musante G, Herrera C, Bancalari E. Closed-loop controlled inspired oxygen concentration for mechanically ventilated very low birth weight infants with frequent episodes of hypoxemia. Pediatrics. 2001;107(5):1120–1124. doi:10.1542/peds.107.5.1120 ↩
  8. Urschitz MS, Horn W, Seyfang A, et al. Automatic control of the inspired oxygen fraction in preterm infants: a randomized crossover trial. Am J Respir Crit Care Med. 2004;170(10):1095–1100. doi:10.1164/rccm.200407-929OC ↩
  9. Hallenberger A, Poets CF, Horn W, Seyfang A, Urschitz MS. Closed-loop automatic oxygen control (CLAC) in preterm infants: a randomized controlled trial. Pediatrics. 2014;133(2):e379–e385. doi:10.1542/peds.2013-1834 ↩
  10. Claure N, Bancalari E, D'Ugard C, et al. Multicenter crossover study of automated control of inspired oxygen in ventilated preterm infants. Pediatrics. 2011;127(1):e76–e83. doi:10.1542/peds.2010-0939 ↩
  11. Zhang Y, Du Y, Shi Y. The effectiveness of automated adjustment of inspired oxygen in preterm infants receiving respiratory support compared with manual: a systematic review and meta-analysis. Pediatr Discov. 2024;2(4):e57. doi:10.1002/pdi3.57 ↩
  12. Di Fiore JM, Kaffashi F, Loparo K, et al. The relationship between patterns of intermittent hypoxia and retinopathy of prematurity in preterm infants. Pediatr Res. 2012;72(6):606–612. doi:10.1038/pr.2012.132 ↩
  13. Maiwald CA, Niemarkt HJ, Poets CF, et al. Effects of closed-loop automatic control of the inspiratory fraction of oxygen (FiO₂-C) on outcome of extremely preterm infants — study protocol of a randomized controlled parallel group multicenter trial for safety and efficacy. BMC Pediatr. 2019;19(1):363. doi:10.1186/s12887-019-1735-9 ↩
  14. Franz AR, 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) ↩
  15. Mitra S, Singh B, El-Naggar W, McMillan DD. Automated versus manual control of inspired oxygen to target oxygen saturation in preterm infants: a systematic review and meta-analysis. J Perinatol. 2018;38(4):351–360. doi:10.1038/s41372-017-0037-z ↩