Targeted Oxygen in the Resuscitation of Preterm Infants, a Randomized Clinical Trial
BACKGROUND AND OBJECTIVES: Lower concentrations of oxygen (O2) (≤30%) are recommended for preterm resuscitation to avoid oxidative injury and cerebral ischemia. Effects on long-term outcomes are uncertain. We aimed to determine the effects of using room air (RA) or 100% O2 on the combined risk of death and disability at 2 years in infants <32 weeks’ gestation.
METHODS: A randomized, unmasked study designed to determine major disability and death at 2 years in infants <32 weeks’ gestation after delivery room resuscitation was initiated with either RA or 100% O2 and which were adjusted to target pulse oximetry of 65% to 95% at 5 minutes and 85% to 95% until NICU admission.
RESULTS: Of 6291 eligible patients, 292 were recruited and 287 (mean gestation: 28.9 weeks) were included in the analysis (RA: n = 144; 100% O2: n = 143). Recruitment ceased in June 2014, per the recommendations of the Data and Safety Monitoring Committee owing to loss of equipoise for the use of 100% O2. In non-prespecified analyses, infants <28 weeks who received RA resuscitation had higher hospital mortality (RA: 10 of 46 [22%]; than those given 100% O2: 3 of 54 [6%]; risk ratio: 3.9 [95% confidence interval: 1.1–13.4]; P = .01). Respiratory failure was the most common cause of death (n = 13).
CONCLUSIONS: Using RA to initiate resuscitation was associated with an increased risk of death in infants <28 weeks’ gestation. This study was not a prespecified analysis, and it was underpowered to address this post hoc hypothesis reliably. Additional data are needed.
- CI —
- confidence interval
- DSMC —
- Data and Safety Monitoring Committee
- Fio2 —
- fractional inspired oxygen
- O2 —
- PDA —
- patent ductus arteriosus
- Spo2 —
- pulse oximetry
- RA —
- room air
- ROP —
- retinopathy of prematurity
What’s Known on This Subject:
To reduce oxidative stress, clinicians use lower amounts of oxygen, including room air, to resuscitate newborn infants, but the effects of this method on the short and long-term outcomes of preterm infants are unknown.
What This Study Adds:
This study was underpowered for the primary outcome. There are important limitations, but results show that initiating resuscitation of extremely premature infants with lower oxygen levels may increase mortality compared with using 100% oxygen.
One hundred percent oxygen (O2) has been used for delivery room resuscitation of newborn infants for more than a century,1 but animal2 and human3–6 data illustrate that this approach leads to oxidative stress5 and organ injury.6 Saugstad et al first showed that ventilation with room air (RA; 21% O2) did not worsen outcomes compared with using 100% O2 when used to resuscitate mature, hypoxic infants.3,4 A meta-analysis of randomized controlled trials of >1300 mostly term/near-term, hypoxic infants then showed that RA significantly reduced the risk of death when used instead of 100% O2 for resuscitation (relative risk: 0.71 [95% confidence interval (CI): 0.54–0.94]; P = .015)7 and reduced severe encephalopathy.8 Most infants, however, were recruited from developing countries,3,4 and because RA resuscitation did not make a difference to 2-year neurodevelopmental outcomes,9 the applicability for using RA in developed countries with different resuscitation resources is therefore unclear.10
Even less clear is the evidence for using less O2 to stabilize preterm infants. Preterm infants, in contrast to term infants, may need supplemental O211 because of lung immaturity, but antioxidant defenses are also suboptimal until the third trimester.12 This scenario leads to a precarious balance between O2 need and toxicity. Nevertheless, growing evidence shows that preterm infants can be supported in the short term with <100% O2. Although this approach has not been associated with major adverse consequences,11,13–21 including death,22 cerebral vasoconstriction (the implications of which are uncertain) has been noted.13
Over the last decade, recommendations for the use of O2 in the delivery suite have changed substantially. Previously, 100% O2 was standard of care.23,24 After 2006, RA25 or low levels of fractional inspired oxygen (Fio2), such as 30%,26,27 became the standard of care for preterm infant resuscitation. Furthermore, novel preductal oximetry (Spo2) data from healthy, spontaneously breathing term28 and preterm29 infants showed that Spo2 increased only gradually after birth. In 2010, Spo2 targeting was also recommended as standard of care. This approach has led to distinct change in clinical practice. In 2008, 50% of Australian and New Zealand perinatal centers used 100% O2 to resuscitate preterm infants.30 However, in 2015, a survey of 630 clinicians from 25 countries found that only 4 used 100% O2 and >70% used Fio2 ≤40% O2.31
Despite these findings, the long-term implications of using anything other than 100% O2 to resuscitate preterm infants are unknown. In 2007, the To2rpido (Targeted Oxygen in the Resuscitation of Preterm Infants and their Developmental Outcomes) study was designed to examine 2-year outcomes at 2 years of age for preterm infants <32 weeks’ gestation after resuscitation with either RA or 100% O2. At that time, only 1 study had been published11 to suggest that preterm resuscitation was possible with <100% O2. We hypothesized that using RA would decrease risk of death and major disability at 2 years of age by 20% compared with 100% O2, but recruitment difficulties arose due to loss of equipoise for using 100% O2. As a consequence, recruitment had to be stopped in June 2014 on the advice of the Data and Safety Monitoring Committee (DSMC), which nevertheless recommended that short-term outcomes be reported before availability of 2-year primary outcomes.
The To2rpido study was an international multicenter, nonblinded, randomized controlled trial involving 2 centers in Australia, 3 in Malaysia, and 1 in Qatar. There were no changes to the trial design during recruitment.
Live-born infants <32 weeks’ gestation or ≤1250 g birth weight (if gestation was uncertain) were eligible if their mothers presented at least 6 hours before delivery and if informed, written parental or guardian consent was obtained. Infants were excluded at any time if they were diagnosed with cardiorespiratory or other abnormalities that had the potential to affect oxygenation or mortality.
Randomization and Masking
Infants were randomized by computer-generated sequence once birth was considered imminent. They were grouped into blocks of 10 stratified according to gestation (<28 weeks’ gestation and 28–31 weeks’ gestation) to ensure even gestational distribution. The study was not masked because ethics committees considered blinding to be difficult to achieve, considering that infants given 100% O2 would theoretically become pinker more rapidly than infants started on RA.
Each center used a dedicated Rad-7 oximeter (Masimo Corporation, Irvine, CA) preloaded with the Trendcom software (Masimo Corporation) to enable downloading of Spo2 and heart rates onto an Excel file (Microsoft Corporation, Bellvue, WA) for later verification of hard copy data. Rad-7 oximeters were used because they were equipped with HI-FI trauma sensors that were able to provide 2-second averaging for maximum sensitivity to rapid oximetry changes. Signal Extraction Technology in the Rad-7 oximeters also used techniques to optimize Spo2 measurement in challenging clinical situations (eg, poor perfusion).32 Oximeter probes were placed on the infant’s right wrist for preductal Spo2 measurement. The Neopuff Infant Resuscitator (Fisher and Paykel Healthcare, Auckland, New Zealand) was used in all participating centers to deliver positive-end expiratory pressure–controlled intermittent breaths at delivery. Ventilation strategies were not proscribed as part of the protocol. O2 blenders were used in all delivery suites for Fio2 blending.
Definition of the Time of Life
Because delayed cord clamping was practiced in some centers, time of birth (or time 0) was defined as time of cord clamping.
Respiratory support was initiated with RA or 100% O2 depending on randomization. Other aspects of resuscitation were provided as per expert committee guidelines at the time of trial design.24,25 Clinicians were advised by protocol to increase Fio2 by 10% every minute (or less) if preductal saturations were <65% before 5 minutes or <80% after 5 minutes and to decrease Fio2 by 10% every minute if Spo2 was >95% at any time. Fio2 could be increased immediately to 100% at any time if the infant’s heart rate remained persistently <100 beats/min despite adequate ventilation (duration of bradycardia left to clinical discretion), if Spo2 was <65% at 5 minutes, or if external cardiac massage or resuscitation medications (eg, adrenaline) were required.
Hard copy records of heart rates, Fio2, Spo2, and resuscitation events were viewed visually and recorded by a team member who was not involved in the resuscitation process. Ventilatory parameters from admission into the NICU and until 12 hours of life were also noted.
The primary outcome was death and major disability at 2 years corrected age (gestational age plus chronological age minus 40 weeks). Major disability was defined as ≥1 of the following: (1) composite cognitive score <85 and/or language score <85 on the Bayley Scale of Infant Development III; (2) severe visual loss (<6/60 vision); (3) cerebral palsy with Gross Motor Function Classification System level ≥2; or (4) deafness requiring hearing aids.
Secondary outcomes were as follows: (1) neonatal morbidities such as retinopathy of prematurity (ROP) grade ≥333; (2) bronchopulmonary dysplasia34; (3) necrotizing enterocolitis requiring surgery or resulting in death; and (4) duration of ventilatory support and respiratory status on NICU admission.
Statistical Analysis, Including Additional Assessments
Sample Size Calculation
A total of 1976 infants (988 infants in each arm) were required to demonstrate decreased incidence of death and neurodevelopmental disability by 20%, from 30% to 24%, with a 2-sided, type I error rate of 5% and 80% power. A correction factor of 1.15 was applied to account for 15% attrition to follow-up to 2 years of age.
Analysis was conducted on an intention-to-treat basis, after excluding infants with life-threatening congenital anomalies (see Inclusion Criteria section). Summary statistics (n, mean, SD, median, minimum, and maximum) were used for the analysis of continuous variables. Counts and percentages were used for categorical variables. Student’s t test and Mann-Whitney tests were used for continuous variables according to their distributions. The χ2 test was used for categorical variables. Factors known to influence mortality were entered into a regression analysis of hospital mortality, including: (1) RA or 100% O2; (2) male sex; (3) gestation <28 weeks; (4) heart rate <100 beats/min at 5 minutes; and (5) Spo2 <80% at 5 minutes. Analyses were considered significant at P < .05; no adjustments were made for multiple comparisons, and missing data were not imputed.
Interim Analysis and Stopping Guidelines
A difference of at least 3 SDs in a major endpoint (or in a combination of major end points) suggesting net clinical benefit or harm was considered by the DSMC as justification to stop the study prematurely or to increase the frequency of trial monitoring or to instigate modification of trial design. The DSMC planned to review interim data and emerging evidence from other studies at 25% (n = 500), 50% (n = 1000), and 75% (n = 1500) of enrollment and to advise the steering committee if premature termination of the trial was required.
Ethics Approval and Trial Registration
The study was approved by the following: the South Easy Sydney Human Research Ethics Committee (06/065), the Hunter New England Ethics Committee (09/08/18/5.07) as the principal Australian site; the University of Malaya Ethics Committee (601.17); the Ministry of Health, Malaysia; Universiti Kebangsaan Malaysia Committee (220.127.116.11/244/SPP3); and the Hamad Medical Corporation, Qatar.
After review by the DSMC, new recruitment was ceased in June 2014 due to difficulties in meeting target recruitment rates. Of 6291 eligible patients, 292 infants were recruited and randomized to treatment between first January 2008 and June 14, 2014. The most common reasons for failed recruitment were: precipitous delivery, clinician preference, and inability of the research team to attend. Two infants were withdrawn after consent was given, resulting in randomization of 290 infants (RA: n = 145; 100% O2: n = 145). A female infant (gestational age: 26.3 weeks) randomized to the 100% O2 arm was delivered precipitously. She was included in the 100% O2 group on an intention-to-treat basis, but her resuscitation data were not included. Two infants from the 100% O2 group (1 with hypertrophic cardiomyopathy, 1 with a congenital diaphragmatic hernia) and 1 from the RA group (lung hypoplasia and dysplasia) were excluded after birth (see Exclusion Criteria and Methods), leaving 144 infants in the RA group and 143 infants in the 100% O2 group to be included in the final analysis (Fig 1).
Patient Demographic Characteristics
Maternal and infant characteristics are shown in Table 1. Most mothers were multiparous and gave birth via cesarean delivery. Almost all mothers received at least 1 dose of antenatal steroids, and there were no differences between the groups for significant antenatal problems or infant parameters. Forty-six (32%) of 144 RA infants and 54 (38%) of 143 100% O2 infants were <28 weeks’ gestation (OR: 0.8 [95% CI: 0.5–1.3]; P = .32).
One infant in the RA group was given adrenaline and cardiac compressions. No infant died during resuscitation or was asystolic. Most were given continuous positive airway pressure, and approximately one-third were intubated. Age at admission to the NICU, admission Fio2, and temperature did not differ (Table 2).
As shown in Fig 2, heart rates were significantly lower in the RA group until the third to fourth minute of life. Spo2 was also significantly lower in infants in the RA group for up to minute 8 of life but were not significantly different thereafter. Infants commenced on 100% O2 had Spo2 ranges mostly within recommended targets,26 but RA infants did not meet these targets until minute 8 of life (Fig 3). Infants started on 100% O2 received significantly more O2 until minute 8 of life. Fio2 was not different after that until nursery admission (Fig 4 and Table 2). These differences were not significantly different between the groups when the infants were examined according to gestational age (ie, below or above 28 weeks’ gestation) (Figs 2–4).
Data for gestational subgroups (below or above 28 weeks’ gestation) are included in the Supplemental Materials.
Infants given 100% O2 required a significantly longer duration of respiratory support (defined as use of mandatory ventilation, continuous positive airway pressure, or high- and low-flow nasal cannula) compared with the RA group. There was no difference in prespecified morbidities, including bronchopulmonary dysplasia, ROP, patent ductus arteriosus (PDA), intraventricular hemorrhage, or necrotizing enterocolitis (Table 3). Infants in the 100% O2 group were more likely to require surgery, predominantly for conditions occurring after the first few days of life (eg, hernia repair and for PDA ligation).
Death before 28 days of age (neonatal death) or death before hospital discharge was not significantly different for all infants. However, 10 (22%) of the infants <28 weeks’ gestation in the RA group died before hospital discharge, compared with 3 (6%) infants given 100% O2 (risk ratio: 3.9 [95% CI: 1.1–13.4]; P = .01). The most common cause of death was respiratory failure (n = 13). Infants in the RA group died slightly later (median: 12 days; range: 2–95 days) than infants on 100% O2 (median: 5 days; range: 1–83 days). One infant <28 weeks’ gestation from the 100% O2 group died of sudden infant death syndrome 3 months after hospital discharge. No other child has died since hospital discharge (Table 4).
Regression analysis showed that both failure to meet target Spo2 by 5 minutes and gestation <28 weeks were significantly associated with the risk of death (adjusted odds ratios: 4.1 and 4.8, respectively) (Table 5); 100% O2, male sex, or low (<100 beats/min) 5-minute heart rate were not significant associations for death (model: B, 2.53; SE, 0.23; df = 1; Exp [B] = 0.08).
This study was the largest single randomized trial to examine outcomes of very preterm infants after delivery room resuscitation was initiated with either lower RA or higher (100%) O2. We emphasize that this study was underpowered due to recruitment difficulties associated with the lack of equipoise for using 100% O2.24–26 On the advice of the DSMC, we stopped recruitment in June 2014 but published short-term outcomes because the results would represent the single biggest contribution to any meta-analysis, the highest level of current evidence. The contribution of To2rpido data to a meta-analysis of 504 infants <28 weeks’ gestation shows that lower (≤30%) or higher (≥60%) O2 (risk ratio: 0.99 [95% CI: 0.52–1.91]) resuscitation makes no difference to hospital mortality35 despite current clinical preference to use lower Fio2.31
We emphasize that recruitment for this study was not stopped because of the unexpected finding of a 3-fold increase in deaths in infants <28 weeks’ gestation who were initially given RA. These infants were already at a high risk of death compared with more mature infants, and the precise etiology of death cannot be ascertained from this small group. The results were not prespecified and due to the small sample size, a change in just 1 death in either group would have resulted in very different outcomes. The results should be considered only as hypothesis-generating.
Our secondary results suggest possible differences in outcomes for survivors. RA infants needed less time on supplemental O2 and respiratory support. This scenario was noted previously by Vento et al15 and may be due to decreased oxidative injury.5,6,16 Infants on 100% O2 needed more surgery, but the cause for this requirement is speculative due to the heterogeneous nature of the surgeries. Some surgeries (eg, hernia repairs) can only be performed if the infant survives, although others (eg, ROP and PDA ligation) have been linked to high O2 exposure. High O2 levels, for example, increase free radical production and prostaglandin E2 expression, a ductal relaxant.36 The associations between lower O2 resuscitation and a decreased rate of PDA were also noted in a retrospective population study by Rabi et al.37
The present study was designed before publication of Spo2 target recommendations in expert guidelines.26,27 Despite widespread acceptance of these recommendations,31 the implications of these trajectories for preterm infants are unclear. The International Consensus on Cardiopulmonary Resuscitation of 2015 makes no Spo2 target recommendations in its resuscitation algorithm.27 Low Spo2 may not result in immediate apparent harm, but hypoxia might lead to delayed cellular damage. In animal studies, hypoxia is associated with delayed death, apoptosis, and irreversible cellular degeneration.38 In our study, death occurred slightly later in infants given RA (12 vs 4 days) but again, this outcome needs to be verified in larger studies.
A retrospective population study by Rabi et al37 of 2326 infants ≤27 weeks’ gestation from Canadian NICU units supports the need for more evidence before lower O2 resuscitation can be universally recommended. After changes to the Canadian resuscitation guidelines from 100% O2 to RA or lower O2, infants after the change (n = 1244) were more likely to die or have severe neurologic injury (adjusted odds ratio: 1.36 [95% CI: 1.11–1.66]) than those born before the guideline change (n = 1082).39 The researchers noted that they did not have data regarding individual O2 exposure, and it would be misleading to attribute major outcomes to the initial levels of Fio2 used.
There are limitations to our study. Due to difficulties in recruiting, the study was stopped prematurely and is significantly underpowered for the primary outcome. Furthermore, ethics committees did not allow either blinding or consent waiver. Personnel who are not involved in clinical care may be able to respond to Spo2 changes more promptly, and the difference is demonstrated in a meta-analysis which showed that masked studies had lower mortality rates in the lower O2 arms.35 Parental consent may also take time and exclude the sickest infants.40 Finally, the study was conducted over a long period when clinical practice undoubtedly changed, with clinicians most likely becoming more adept at Spo2 targeting and Fio2 blending.31 Whether this factor has an effect on outcomes is uncertain.
The results of this study must be interpreted with caution due to its small sample size. However, it remains the largest randomized controlled trial to date. Our non-prespecified but potentially concerning finding of increased hospital death in infants <28 weeks’ gestation resuscitated with RA emphasizes the critical need to urgently examine this knowledge gap further with large and well-designed trials. Failure to do so could lead to irreversible harm to thousands of preterm infants around the world every year.
The authors thank the DSMC for their support and advice regarding this study (Prof Nicholas Evans [chair], Prof Val Gebski, and Dr Wendy Hague).
- Accepted October 14, 2016.
- Address correspondence to Ju Lee Oei, MBBS, FRACP, MD, Department of Newborn Care, Royal Hospital for Women, Barker St, Randwick, NSW, Australia, 2031. E-mail:
This trial has been registered with the Australian and New Zealand Clinical Trials Network Registry (www.anzctr.org.au/) (ACTRN 12610001059055) and the National Malaysian Research Registry (www.nmrr.gov.my) (NMRR-07-685-957).
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Supported by the Thrasher Research Fund for Children, United States, and the Leslie Stevens Fund for Newborn Research, Australia.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- O’Donnell CP,
- Gibson AT,
- Davis PG
- Vento M,
- Asensi M,
- Sastre J,
- García-Sala F,
- Pallardó FV,
- Viña J
- Wang CL,
- Anderson C,
- Leone TA,
- Rich W,
- Govindaswami B,
- Finer NN
- Lundstrøm KE,
- Pryds O,
- Greisen G
- Escrig R,
- Arruza L,
- Izquierdo I, et al
- Aguar M,
- Izquierdo M,
- Brugada M, et al
- Kumar VH,
- Wynn K,
- Ryan RM,
- Nielsen L,
- Reynolds AM,
- Carrion V
- International Liaison Committee on Resuscitation
- Perlman JM,
- Wyllie J,
- Kattwinkel J, et al; Neonatal Resuscitation Chapter Collaborators
- ↵Copublishing of the Pediatric and Neonatal Portions of the 2015 International Consensus on Cardiopulmonary Resuscitation and Emergency Cardiovascular Care Science With Treatment Recommendations and the 2015 American Heart Association Guidelines Update for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care. Pediatrics. 2015;136(suppl 2):S83–S87pmid:26471385
- Dawson JA,
- Kamlin CO,
- Vento M et al.
- Oei JL,
- Ghadge A,
- Coates E, et al; HOLISTIC (Higher Or Lower oxygen and Infant Saturation Targeting: International Collaboration) Study Group
- Ehrenkranz RA,
- Walsh MC,
- Vohr BR, et al; National Institutes of Child Health and Human Development Neonatal Research Network
- Oei JL,
- Vento M,
- Rabi Y, et al
- Clyman RI,
- Saugstad OD,
- Mauray F
- Wyckoff MH,
- Aziz K,
- Escobedo MB, et al
- Rich WD,
- Auten KJ,
- Gantz MG, et al; National Institute of Child Health and Human Development Neonatal Research Network
- Copyright © 2017 by the American Academy of Pediatrics