Safe Oxygen Targets for Preemies Unclear, Doctors to Choose

Neil Canavan

May 10, 2013

WASHINGTON, DC — Are lower oxygen saturation targets safe for extremely preterm infants? Two major new studies come to different conclusions.

"For years, we clinicians have searched for the right balance between the competing risks caused by oxygen excess and oxygen deprivation," said Barbara Schmidt, MD, chair in neonatology at the University of Pennsylvania in Philadelphia.

Dr. Schmidt presented 18-month outcomes from the Canadian Oxygen Trial here at the Pediatric Academic Societies 2013 Annual Meeting. The results were published online May 5 in JAMA to coincide with the presentation.

"We performed this double-blind randomized trial to help resolve the uncertainty," she explained.

In the Canadian trial, 2 levels of oxygen supplementation — target oxygen saturation measured by pulse oximetry of 85% to 89% and of 91% to 95% — started within 24 hours of birth, were compared in preterm infants born at 23 to 27 weeks. Treatment was to continue until the postmenstrual age of 40 weeks or hospital discharge.

The primary end point was the composite of infant death, survival with severe disability (such as bilateral blindness), and gross motor function impairment.

A total of 1201 infants were randomized in a 1:1 fashion to the higher or lower saturation target. To blind investigators to the treatment group, oximeters displayed saturation levels either 3% higher or 3% lower than the actual measures. Researchers adjusted the fraction of inspired oxygen to achieve a displayed saturation of 88% to 92%; to ensure overall safety, alarms were set at 86% and 94%.

At baseline, both groups were balanced for weight and sex ratio, singleton births, and exposure to antenatal corticosteroids.

For years, we clinicians have searched for the right balance between the competing risks caused by oxygen excess and oxygen deprivation.

Of note, the software for the oximeter was upgraded during the investigation. When measurements taken using old and new oximeter software were compared, the statistical differences were consistent.

After 18 months, there was no statistical difference between the lower and higher saturation groups in death or disability (51.5% vs 49.7%; = .52).

In fact, "we found little evidence for differences in any of the components that make up the composite end point," Dr. Schmidt reported.

"Targeting the lower saturation range did reduce postmenstrual age at last use of oxygen therapy by almost 1 week (= .03), but no other significant differences were seen in secondary outcomes, such as severe retinopathy or brain injury as seen on cranial ultrasound," she added.

After her presentation, Dr. Schmidt was asked by an audience member to consider her results in the light of the similar SUPPORT trial, which found greater mortality at lower saturation targets.

Dr. Schmidt replied that, although the trials do have similarities, there are also critical differences, particularly in the very highest and very lowest values actually observed for the 2 target ranges. It is possible that "we did not find excess mortality in the low target or increased retinopathy in the high target because of tighter control."

In practice, for levels outside the target range, "we believe that these alarm limits are critically important and need to be enforced at all times. We know it's difficult, but it needs to be done," she emphasized.

Results from a second study, known as the Benefits of Oxygen Saturation Targeting (BOOST II) trial, conflict with the findings of this Canadian study.

BOOST II Trial Halted

In the BOOST II trial, mortality differences between the 2 groups are significant, reported Benjamin Stenson, MD, from the Royal Infirmary of Edinburgh, in the United Kingdom, who presented the results here.

In fact, recruitment to the British and Australian BOOST II cohorts was stopped in December 2010 after an interim analysis of data showed an increased mortality risk in infants randomized to the lower saturation target.

The results were published online May 5 in the New England Journal of Medicine to coincide with their presentation.

When we revised oximeters, we saw a large and highly statistically difference in mortality.

The 5230 infants enrolled in the BOOST II trial were randomized in the 24 hours after birth to the same saturation target levels as the Canadian trial. As in the that study, oximeters were offset and investigators were blinded. Alarm limits for the oximeters varied by study site: in Australia, they were set at 86% and 94% (n = 1135); in New Zealand, they were set at 87% and 93% (n = 340); and in the United Kingdom, they were set at 94% and unit preference (n = 1350).

As in the Canadian trial, oximeters were upgraded midstudy, except in New Zealand, where the study had already been completed.

"When we were using the original oximeters, we saw no statistically significant differences in mortality between treatment groups," said Dr. Stenson. "However, when we revised oximeters, we saw a large and highly statistically difference in mortality." Survival was greater in infants in the higher saturation group than in the lower saturation group (23.1% vs 15.9%; risk ratio, 1.45; = .002).

Table. BOOST II Mortality Outcomes By Oximeter Calibration

Software 85%–89% Saturation Target 91%–95% Saturation Target Risk Ratio (95% Confidence Interval)
Old 15.6% 17.3% 0.90 (0.70–1.16)
New 23.2% 15.8% 1.47 (1.16–1.86)


Other differences were observed for secondary outcomes. There was less retinopathy in the lower group than in the higher group (10.6% vs 13.5%; = .045), but more necrotizing enterocolitis (10.4% vs 8.0%; P = .04).

"Slightly more infants in the high-target group were oxygen-dependent at 36 weeks (RR, 0.88), but this was not observed in the British cohort," Dr. Stenson noted.

In an interview with Medscape Medical News, Lawrence Rhein, MD, director of the Center for Healthy Infant Lung Development at the Children's Hospital in Boston, Massachusetts, noted that "the take-home message of these studies is that close attention to the saturation limits is important."

He pointed out that the Canadian trial shows that paying close attention to the saturation targets results in no difference in outcome; there were no negative consequences in either treatment group.

Close attention to the saturation limits is important.

"But I think there's a huge asterisk in that," Dr. Rhein said. "No negative consequences with this practice assumes that people have paid, and will continue to pay, close attention to those saturation limits. Yet, in real-world practice, I'm not sure that always happens. So if you are going to practice that way, you have to make sure that you adhere very strictly within that low range."

Dr. Rhein pointed to a number of outliers in the BOOST trial — data points outside the target ranges where real dangers to the patients reside. "It was that issue that people were talking about, these excursions, and how you can avoid the greatest risk to patients at extremes of both ranges." Some hypothesize that these outliers explain the difference between the opposing findings of the 2 trials, Dr. Rhein said. If that is the case, that would tend to frame the controversy as one more of accuracy than of appropriate targeting.

"I would say that if you are going to change practice, err on the side of raising your lower target saturation limit, and then make sure that on the high end, you have the nurses maintain strict alarm limits," he said.

Raise the Lower Saturation Target

As for the protocols used in his unit, "our set points are 88% to 92%," Dr. Rhein said. "That way, when we do a good job of keeping within range, we won't have too many outliers that are less than 85%, like in the SUPPORT trial, and we shouldn't have too many that are above 95%." But again, this very much depends on the diligence of the staff.

Now that these much-anticipated data have been released, Dr. Rhein suggests that the next hot topic in neonatology is going to be long-term outcomes. "When big studies like this come out with preliminary findings, these short-term outcomes, the next thing people look for is what are these kids going to be like when they're 5."

Dr. Rhein said he would also like to see data on how long a given range of saturation targets should be sustained. "In all of these studies, they had a target range that stayed the same on day-of-life 1 and discharge. There are many units, including ours, that change the saturation parameters once a child reaches a gestational age of 33 to 36 weeks. That aspect wasn't addressed in these studies."

Dr. Schmidt, Dr. Stenson, and Dr. Rhein have disclosed no relevant financial relationships.

JAMA. Published online May 5, 2013. Abstract

N Engl J Med. Published online May 5, 2013. Abstract

Pediatric Academic Societies (PAS) 2013 Annual Meeting: Abstracts 2180.6 and 2180.8. Presented May 5, 2013.