Should we target higher or lower oxygen saturation targets in the preterm infant?

Citation: Askie LM, Darlow BA, Finer N, Schmidt B, Stenson B, Tarnow-Mordi W, 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.

Reviewed by:
Sarah Drennan, MD; Edgardo Szyld, MD, MSc
Institution: University of Oklahoma Health Sciences Center at Oklahoma City, Oklahoma, USA.

Corresponding Author:
Sarah Drennan, MD
The University of Oklahoma
Department of Pediatrics
Division of Neonatal-Perinatal Medicine
PO Box 26901, Oklahoma City, OK 73126-0901
1200 Everett Drive, 7^th Floor North Pavilion,
OKC, OK 73104
Phone: (405) 271-5215
Email
Conflict of interest: none 

Keywords: Oxygen saturations, Extremely preterm infants, Neurodevelopmental Impairments, Mortality     

Type of Investigation: prognosis 

Question:

In infants born before 28 weeks gestation, does lower oxygen saturation targets (85-89%) versus higher oxygen saturation targets (91%-95%) affect death or major disability at a corrected age of 18-24 months?

Methods:

Design: Prospectively designed, multi-center, meta-analysis of 5 prospective randomized control trials 

Allocation: Subjects were randomized to one of two groups: lower oxygen saturation targeting (85-89%) or higher oxygen saturation targeting (91-95%) while in the NICU 

Blinding: Intervention groups were blinded to parents, caregivers, and outcome assessors by using a modified pulse oximeter, adjusted to display saturations 88-92% ±3%. True values were only displayed if actual SpO₂ <84% or >96%. 

Follow-up period: Infants were followed out to 18-24 months corrected age.  

Setting: 5 large multicenter trials taking place in the United States, Canada, Argentina, Finland, Germany, Israel, United Kingdom, Australia, and New Zealand, orchestrated by the Neonatal Oxygenation Prospective Meta-analysis (NeOProM) Collaboration.

Patients: Study eligibility includes randomized trials with an adequate level of allocation concealment and registration on publicly accessible trials registry. 

• Inclusion Criteria: infants born before 28 weeks' gestation and enrolled within 24 hours of life. 
• Exclusion Criteria: no exclusion criteria. 

Intervention: Supplemental oxygen to maintain oxygen saturations via pulse oximeter in the range of 85-89% in one group compared to 91-95% in the second group. Intervention was continued until 36 weeks post-menstrual age with slight variations per study (see Supplement 3).

Outcomes:

Primary: composite of death or major disability at a corrected age of 18-24 months (major disability defined as any of the following: Bayley-III cognitive or language score <85, severe visual loss defined as cannot fixate or legally blind with visual acuity <6/60 in both eyes, CP with GMFCS level 2 or higher, or deafness requiring hearing aids)

Secondary: Total of 16 secondary outcomes including the components of the primary outcome and other major morbidities were studied.

Analysis and sample size: An estimated sample size of approximately 5000 infants was required to detect an hypothesized difference of 4% in the primary outcome of death or major disability between lower and higher Spo₂ target ranges. The preplanned total sample size was 5230 infants; however, due to two trials (BOOST II in the United Kingdom and BOOST II in Australia) stopping early due to interim analysis of higher mortality rates, the total number of participants was slightly lower at 4965 infants. This number provides 80% power (using a 2-sided p value of 0.05) to detect an absolute risk difference of 4% between groups. This correlates to a number needed to treat of 25 infants to prevent 1 major adverse outcome¹.

Patient follow-up: Follow-up was performed out to 18-24 months corrected age with available data for 90% of infants for the primary outcome.

Main Results:

There was no significant difference in the primary composite outcome of death or major disability at 18 to 24 months corrected age between the lower SpO₂ (85-89%) target range group as compared to the higher SpO₂ (91-95%) target range group (53.3% in lower vs 51.6% in higher group; risk difference 1.7% [95% CI -1.3%, 4.6%]; RR 1.04 [95% CI 0.98, 1.09] P=.21). When alternate measures of disability were used, there was again no statistically significant between group differences in the rate of death or major disability (51.2% in lower vs 49.3% in higher group; risk difference 1.7% [95% CI -1.2%, 4.5%]; RR 1.04 [95% CI 0.98, 1.09] P=.20). Eleven of the sixteen secondary outcomes showed no statistically significant difference. This included primary major disability, cerebral palsy, deafness, and others.

Two secondary outcomes favored the lower SpO₂ target and three favored the higher SpO₂ target group.

1. Those outcomes favoring the lower SpO2 target group included treatment of ROP and need for supplemental oxygen at 36 weeks corrected age. Infants in the lower target group showed a lower rate of ROP treatment (10.9% in lower vs 14.9% in higher group; risk difference -4% [95% CI -6.1%, 2%]; RR 0.74 [95% CI 0.63, 0.86], P=<.003) and oxygen treatment at a postmenstrual age of 36 weeks. Of note, there was no significant difference between the groups in severe visual impairment.

2. Those outcomes favoring the higher SpO2 target group included death before PMA of 36 wk, severe necrotizing enterocolitis, and PDA treatment with surgical ligation. The lower SpO2 target range group was associated with an increased incidence of death at a corrected age of 18 to 24 months (19.9% in lower group vs 17.1% in higher group; risk difference 2.8% [95% CI 0.6%, 5%]; RR 1.17 [95% CI 1.04, 1.31] P=.01), but not major disability. The lower target group showed an increase in death also at postmenstrual age of 36 weeks and at hospital discharge. Infants in the lower target group had an increased incidence of severe necrotizing enterocolitis leading to surgery or death (9.2% in lower group vs 6.9% in higher group; risk difference 2.3% [95% CI 0.8%, 3.8%]; RR, 1.33 [95% CI 1.1, 1.61]; P=.003) and PDA treated with surgical ligation (P=.046).

Study Conclusions:

In this prospective, multicenter, meta-analysis comprised of five randomized controlled trials, the authors concluded there was no difference for death or disability as a composite outcome between extremely preterm infants in a lower targeted SpO₂ group (85-89%) than infants in a higher targeted SpO₂ group (91-95%). There were however differences in secondary outcomes that favored those in a lower or higher target group.

Commentary:

 This meta-analysis stands out as unique due to the significant international concerted prospective effort by NeOProM to answer one specific clinical question which allowed for a large sample size and sound methodological approach. The main strength of this meta-analysis is the prospective design of five randomized controlled trials with a standardized, pre-planned protocol^2,3 and investigator agreement to submit individual participant data for pooled analysis (see EBM lesson).

By using a common protocol, this allows the meta-analysis to overcome a common flaw which is that many meta-analyses fail to perform a convincing synthesis of studies due to heterogeneity between dissimilar studies. A common protocol allows for comparison amongst studies using the same inclusion criteria, study measures, and outcomes. This therefore limits heterogeneity and the risk of altered validity.

Together, the results of these five randomized controlled trials highlight the significant difficulty in pinpointing the ideal oxygen saturation target range for extremely preterm infants. The primary outcome in this study is a composite of death or major disability. Composite outcomes are advantageous in allowing investigators to use a smaller sample size to reach statistical power; however, composite outcomes may be problematic when the individual outcomes pull in opposite directions, leading to an overall less precise result. Lower compared to higher oxygen saturation targets may lead to a trade-off in risks and benefits for this fragile population. Investigators concluded that the higher SpO₂ target range group was associated with less mortality and cases of severe NEC with a trade-off of higher rates of ROP and need for supplemental oxygen at 36 weeks corrected age.  However, as the authors point out there was "no adjustment for multiple comparisons. Therefore, because of the potential for type I error, the prespecified secondary outcomes and the subgroup analyses should be considered exploratory".  Outcomes within a clinical trial sample represent averages and true benefits and harms may differ from those in these analyses. There are limitations in studies of this nature compared to real life scenarios which include setting alarm limits and staff response to such alarms to maintain goal saturations within this range. As we know it can be difficult to maintain SpO₂ in a defined range. As was shown in this study, despite being controlled for, study participants in each group still had overlap in their oxygen saturation ranges. For this reason, additional studies are needed to look at the effect of alarm limits and targeting compliance⁴, and studies are already underway to evaluate automated methods to achieve narrow targeted saturation ranges. These may improve compliance in targeting oxygen saturations in extremely preterm infants.

Limitations specifically pointed out by the authors include this overlap in oxygen exposure between treatment groups, in addition to early stopping of two trials due to interim mortality analysis, potential for false positive results based on multiple comparisons from a large number of secondary and sub-group analyses, and finally, cautious interpretation of results in units without early ROP screening and skilled nursing care to respond to alarm limits. Again, this study specifically addressed oxygen targeting and not alarm limits.

There continues to be substantial practice variation and controversy concerning the optimal oxygen saturation targeting range in extremely preterm infants. There is also no information as to whether these ranges should be adjusted for gestational age. Future studies to definitely answer this question are unlikely as it would be difficult to improve upon the existing study by repeating a large international study with an increased sample size. To date, this meta-analysis provides the best available evidence which suggests that infants in a higher oxygen saturation targeted group have a lower incidence of mortality and severe NEC with a higher incidence of ROP treatment but not blindness.

EBM Lesson: Individual Participant Data versus Aggregate Data in Meta-analyses

Trial investigators in this meta-analysis pre-planned both individual patient data (IPD) analysis and aggregate data analysis as final goals. IPD analysis was performed following a Cochrane review⁵ of the same five randomized, controlled trials using aggregate data with similarly conclusive results. IPD reviews require raw data on each patient included in a trial by those responsible for the trial. The information can then be collected, data reviewed for quality, and analyzed as a larger dataset by those coordinating the review. Using IPD in a systematic review offers many benefits including: more complete data, use of unpublished data, more flexible analysis of patient subgroups and outcomes, analysis of time to event, double-checking and correction of data, and longer-term follow-up⁶. Potential disadvantages include the acquisition of data taking longer, requiring more costs, and inability to obtain IPD from researchers. In this meta-analysis, because two trials were stopped early by investigators, this unfortunately lowered the final sample size beneath that which was originally calculated for the study. Results of reviews have been compared when using IPD versus aggregate data with differing results^6,7. Sometimes an important difference is found between the two approaches. Often IPD reviews "find lower estimates of treatment effect and less statistically significant results", though the reverse can also occur⁶. When possible, an IPD review should be performed to compare results to aggregate data analysis. In this particular meta-analysis, while the results validated those of the Cochrane Review, IPD did allow for new insights into the subgroup analysis, particularly as it showed consistency of results across groups such as SGA and very early gestational ages. 

References:

1.      Askie LM, Brocklehurst P, Darlow BA, Finer N, Schmidt B, Tarnow-Mordi W. NeOProM: Neonatal Oxygenation Prospective Meta-analysis Collaboration study protocol. BMC Pediatrics. 2011;11:6.

2.      Simes RJ. Prospective meta-analysis of cholesterol-lowering studies: the Prospective Pravastatin Pooling (PPP) Project and the Cholesterol Treatment Trialists (CTT) Collaboration. The American journal of cardiology. 1995;76(9):122C-126C.

3.      Ioannidis J. Next-generation systematic reviews: prospective meta-analysis, individual-level data, networks and umbrella reviews. British Journal of Sports Medicine. 2017;51(20):1456-1458.

4.      Schmidt B, Whyte RK, Roberts RS. Trade-off between lower or higher oxygen saturations for oxygen saturation targeting (BOOST) II trial reports its primary outcome. The Journal of Pediatrics. 2014;165(1):6-8.

5.      Askie LM, Darlow BA, Davis PG, Finer N, Stenson B, Vento M, et al. Effects of targeting lower versus higher arterial oxygen saturations on death or disability in preterm infants. The Cochrane database of systematic reviews. 2017;4:Cd011190.

6.      Clarke MJ, Stewart LA. Obtaining individual patient data from randomised controlled trials. In: Egger M, Smith, GD, Altman DG, eds. Systematic reviews in health care: meta-analysis in context. 2^nd ed. London: BMJ Publishing Group;2009. p. 109-121.