Automated versus manual oxygen control in preterm infants on respiratory support: a randomised crossover trial

In the delicate world of neonatal care, where tiny lungs struggle to breathe on their own, oxygen is both a lifeline and a potential hazard. For premature infants, the body's ability to regulate oxygen levels is often immature, making them vulnerable to swings that are too low or too high. Too little oxygen can starve the brain and other vital organs, while too much can damage the developing eyes and lungs. Because of this narrow margin for error, doctors and nurses constantly adjust the amount of oxygen mixed into the air these babies breathe, watching a monitor that displays their blood oxygen levels in real time. This task requires intense focus and split-second decisions, yet human attention naturally wanders, and the sheer volume of data can be overwhelming. The question facing modern medicine is whether a machine, programmed to react instantly and precisely, can manage this balance better than a human caregiver, especially in hospitals where staff are stretched thin.

A team of researchers at the Aga Khan University Hospital in Karachi, Pakistan, set out to answer this question with a direct comparison. They enrolled twenty-five premature infants who were already receiving breathing support, either through a tube in the windpipe or through masks and nasal prongs. Each baby served as their own control in a carefully designed experiment. For twelve hours, a nurse would manually adjust the oxygen levels based on the monitor readings, just as they would in a normal shift. Then, for the next twelve hours, an automated system took over. This system, known as OxyGenie, is a closed-loop controller that reads the baby's oxygen levels every second and makes tiny, continuous adjustments to the oxygen mixture without human intervention. The order of these two periods was randomized, meaning some babies started with the machine and others with the nurse, ensuring that the results were not skewed by the time of day or the baby's natural recovery.

The results showed a clear advantage for the automated system. When the machine was in charge, the babies spent significantly more time within the ideal target range for blood oxygen, which the researchers set between 90 and 94 percent. During the manual periods, the babies were in this safe zone about 38 percent of the time. When the automated controller took over, that figure rose to nearly 56 percent. This improvement was not just a matter of staying in the middle; the machine also kept the babies safer at the extremes. The time spent with dangerously low oxygen levels dropped from about 12 percent down to 7 percent, and the time spent with dangerously high levels fell even more dramatically, from 16 percent down to just 3 percent, but only when the babies were actually receiving supplemental oxygen.

Crucially, the researchers found that this better control did not come at the cost of exposing the babies to more oxygen overall. The average amount of oxygen the infants received remained the same whether a nurse or the machine was in charge. The difference lay in how that oxygen was delivered. Human adjustments tend to be coarse, often changing the oxygen level in large, round steps, which can cause the baby's oxygen levels to overshoot the target and then swing back. The machine, by contrast, made thousands of tiny, frequent adjustments, keeping the levels steady and centered. This precision meant the babies experienced fewer severe drops in oxygen, which are the events most likely to cause long-term harm. In fact, the number of times a baby's oxygen level fell below 80 percent for more than ten seconds was cut in half when the machine was managing the care.

The study also highlighted a subtle but important detail in how oxygen safety is measured. When a baby is breathing normal room air and their oxygen level is slightly above the target range, it is not considered dangerous because they are not receiving extra oxygen. However, standard measurements often count this as "too high" regardless of the source. The researchers adjusted their analysis to account for this, and when they did, the benefits of the automated system became even more pronounced. The machine was able to wean babies off supplemental oxygen more effectively, returning them to room air sooner and keeping them there, whereas human caregivers tended to leave the oxygen on a bit longer. This suggests that the machine's ability to make fine-tuned decisions allows it to recognize when a baby no longer needs help, a nuance that is difficult for a busy nurse to maintain constantly.

This research took place in a setting that differs significantly from the wealthy, well-staffed hospitals where most similar studies have been conducted. The unit in Karachi operates without dedicated respiratory therapists and faces high staff turnover, conditions that make the task of manual oxygen titration even more challenging. The fact that the automated system performed so well in this environment suggests that such technology could be a vital tool for improving care in lower-resource regions where specialist staff are scarce. While the study did not measure long-term health outcomes like eye or lung disease directly, it demonstrated that the machine can achieve a level of physiological stability that human hands struggle to match. The findings indicate that automated control offers a reliable way to keep premature infants safe from the dangers of both too little and too much oxygen, reducing the burden on staff and potentially saving lives in the most vulnerable moments of a newborn's life.