Time to full vascularization stratified by gestational age (GA) and era of birth and censored by death and treatment for retinopathy of prematurity for the full study cohort (N = 596), as estimated by the Kaplan-Meier method.
eTable 1. Clinical Characteristics and of Study Cohort
eTable 2. Time-to-Event Analysis
eFigure. Rates of Treated ROP by Gestational Age Group and Era
Customize your JAMA Network experience by selecting one or more topics from the list below.
Shukla A, Sonnie C, Worley S, et al. Comparison of Biphasic vs Static Oxygen Saturation Targets Among Infants With Retinopathy of Prematurity. JAMA Ophthalmol. 2019;137(4):417–423. doi:10.1001/jamaophthalmol.2018.7021
Does the incidence and severity of retinopathy of prematurity (ROP) decrease with biphasic oxygen targets compared with static standards of the Surfactant, Positive Pressure, and Pulse Oximetry Trial (SUPPORT) without affecting mortality?
In a cohort study comparing incidence of type 1 ROP using a biphasic oxygen standard with static SUPPORT standards, type 1 ROP increased in the postintervention cohort (2% pre-SUPPORT vs 6% post-SUPPORT). There was an increase in any ROP overall (20% pre-SUPPORT vs 28% post-SUPPORT), and mortality was unchanged (5% pre-SUPPORT and 6% post-SUPPORT).
Biphasic oxygen targets are associated with decreased incidence and severity of ROP without increasing mortality.
The Surfactant, Positive Pressure, and Pulse Oximetry Randomized Trial (SUPPORT) demonstrated that static low oxygen saturation decreased retinopathy of prematurity (ROP) but increased mortality compared with static high oxygen saturation cohorts.
To compare outcomes of a biphasic oxygen protocol with static targets recommended by SUPPORT.
Design, Setting, and Participants
Retrospective cohort study comparing biphasic vs static standards 41 months prior to and 42 months after a change from biphasic to static SUPPORT standards at a level III neonatal intensive care unit (Fairview Hospital, Cleveland, Ohio). The study included infants born at a corrected gestational age (CGA) of 31 weeks or younger or birth weight 1500 g or less. Data were analyzed between August 2010 and July 2017.
The pre-SUPPORT group underwent biphasic protocol target saturations of 85% to 92% at younger than 34 weeks’ CGA and greater than 95% at 34 weeks’ CGA or older. The post-SUPPORT group underwent a constant 91% to 95% target.
Main Outcomes and Measures
Primary outcome was incidence of type 1 ROP. Secondary outcomes were incidence of any ROP, time to full vascularization, and mortality.
Of 596 eligible infants, 562 were included in ophthalmic analysis. Three hundred three patients were boys (54%); 399 were white (71%), 87 were black (15%), and 76 were of other or unknown race/ethnicity (14%). Mean (SD) CGA and birth weight were 29 (2) weeks and 1151 (346) g, respectively. Any ROP overall increased (53 [20%] pre-SUPPORT vs n = 86 [28%] post-SUPPORT; absolute difference, 8%; 95% CI, 1%-15%; odds ratio, 1.6; 95% CI, 1.05-2.3; P = .03). Type 1 ROP increased in the post-SUPPORT era (n = 6 [2%] pre-SUPPORT vs n = 18 [6%] post-SUPPORT; absolute difference, 4%; 95% CI, 0.4%-7%; odds ratio, 2.7; 95% CI, 1.05-6.9; P = .03). There was a delay in vascularization in the post-SUPPORT group (n = 6 [2%] pre-SUPPORT vs n = 18 [6%] post-SUPPORT; absolute difference, 4%; 95% CI, 0.4%-7%; P = .03).
Conclusions and Relevance
Compared with static oxygen standards, biphasic oxygen targets are associated with decreased incidence and severity of ROP without increasing mortality.
Retinopathy of prematurity (ROP) is the leading cause of childhood blindness, increasing as more children are successfully resuscitated at lower birth weight and gestational age (GA).1 Worldwide, 14.9 million children are born prematurely, with an estimated 184 700 cases of new blindness each year attributable to ROP.2 In 2010, in high-income countries, an estimated 6300 infants required treatment for ROP, with 1700 of them becoming blind or severely visually impaired.3 Similar to other neonatal diseases, ROP is considered to be multifactorial, involving oxygen supplementation, nutrition, prematurity, low birth weight and extrauterine growth retardation, sepsis, glucose imbalance, and blood product transfusion, among several others.4
The impetus for this study lies in the pathogenesis of ROP. Oxygen supplementation is linked to hyperoxia-induced retinovascular growth attenuation caused by diffusion of oxygen from the choriocapillaris, a highly vascularized layer of unfenestrated blood vessels beneath the retina, to peripheral areas of undeveloped retina. Termed phase 1, this leads to retinovascular growth attenuation and vaso-obliteration. This ischemic hyperoxia creates wide expanses of subsequent hypoxic retina, termed phase 2, that induces pathologic angiogenesis in response to acute overexpression of cytokines from these avascular areas.5-7
The 2-step pathogenesis of ROP stimulated the idea that early physiologic hypoxia might decrease ROP by inducing the normal sequential development of the retina.8 Randomized clinical trials, such as the NeoProm Collaboration, have previously compared target ranges of oxygen saturations between 85% and 92% until 36 weeks’ corrected GA (CGA) to continuous 91% to 95%. The results were summarized in a 2018 meta-analysis9 that concluded that lower (85%-89%) oxygen saturation vs higher (91%-95%) oxygen saturation increased the risk of death and necrotizing enterocolitis by at least 3.0% without a significant association with the composite outcome of death or major disability including blindness.9-11 Prior to the SUPPORT trial, multiple nurseries adopted a biphasic approach that adjusted oxygen concentration using saturation targets of 85% to 92% until either 32 or 34 weeks’ CGA and then increasing targets to greater than 95% beyond 34 weeks’ CGA.8,12,13 However, biphasic trials are unlike static approaches described in this meta-analysis9 because low oxygen saturations were maintained until at least 36 weeks, 2 to 4 weeks longer than in a biphasic approach.
Quiz Ref IDIn light of the increased risk of mortality at lower oxygen saturations reported in multiple studies, most neonatologists now recommend static and higher saturations. The purpose of this study was to compare ophthalmic outcomes and mortality within this new oxygen protocol (91%-95%) with a biphasic oxygen protocol (85%-92% at less than 34 weeks’ CGA and more than 95% at least 34 weeks’ CGA). The primary outcome measure was type 1 ROP. Secondary outcomes compared infant mortality and incidence of any ROP.
This is a retrospective study conducted in a single, level III neonatal intensive care unit (NICU) at the Cleveland Clinic Children’s Hospital, Cleveland, Ohio. The Cleveland Clinic internal review board considered this a minimal-risk study (16-1595) and approved it. Informed consent was not obtained because the study was retrospective, noninterventional, nonrandomized, and patients were deidentified. This unit is 1 of 3 cared for by the Ophthalmology Department, Cole Eye Institute, Cleveland Clinic, in which a biphasic oxygen protocol was designed and previously tested.8
Inclusion criteria were infants born earlier than 31 weeks and/or weighing less than 1500 g who received at least 1 dilated eye examination. The data collection was done 41 months prior to (pre-SUPPORT) and 42 months after (post-SUPPORT) the change in institutional oxygen saturation targets. Preintervention infants were maintained at 85% to 92% for younger than 34 weeks’ CGA and greater than 95% for 34 weeks’ CGA or older. All postintervention infants were kept at 90% to 95% constantly, regardless of GA. Exclusion criterion was death prior to first eye examination, which occurred 4 weeks after birth. Primary outcome measure was presence of type 1 ROP (treated with laser surgery, Avastin, or lens-sparing vitrectomy); secondary outcome measures included presence of any ROP and death. Multiple demographic and clinical characteristics of the NICU course were collected for both groups of patients.
Strict oxygen targeting was maintained as much as possible, with alarms set at 1% greater than and less than the target oxygen saturation range on Massimo pulse oximeters. All infants were examined and classified according to the Early Treatment Retinopathy of Prematurity Trial using binocular indirect ophthalmoscopy.14 Type 1 ROP was defined as retinopathy in zone 1 (with stage 3 or any stage with plus disease); or zone 2 retinopathy (with stage 2-3 and plus disease).
Race/ethnicity was determined by medical record review; per hospital policy, patients (in this case, mothers of the study infants) filled out their race/ethnicity during admission to the hospital.
Data were described using medians and quartiles or means and standard deviations for continuous variables and counts and percentages for categorical variables. Pre-SUPPORT and post-SUPPORT groups were compared on demographic and clinical characteristics using analysis of variance or Kruskal-Wallis tests for continuous and ordinal characteristics and χ2 or Fisher exact tests for categorical characteristics. Absolute risk differences, odds ratios, and the 95% confidence intervals of these were estimated for outcome measures. To adjust for the clinically relevant risk factors of GA, birth weight, sepsis, and intrafamily correlation among siblings, the preintervention and postintervention groups were compared on ROP outcomes using logistic models with generalized estimating equations. The pre-SUPPORT and post-SUPPORT groups were compared on time to full vascularization with a Cox proportional hazards model THAT adjusted for GA group, intervention-GA group interaction, and the intrafamily correlation, with patients censored at death or treatment. Patients who died or were lost to follow-up were excluded from the analyses of ROP incidence, severity, and treatment. All tests were 2-tailed. SAS, version 9.4 software (SAS Institute) was used for all analyses. The P value level of significance was .05, and all P values were 2-sided.
A total of 596 patients were included based on the GA and/or birth weight criteria; eTable 1 in the Supplement describes the clinical characteristics of the entire cohort eligible for ophthalmic screening. There were 275 patients in the pre-SUPPORT and 321 in the post-SUPPORT cohort. Of the 596 infants, 562 were eligible for ROP screening based on our inclusion criteria of at least 1 eye examination; 34 were excluded from analysis owing to death that occurred prior to ROP screening (15 from the pre-SUPPORT and 19 from the post-SUPPORT group). Among those screened for ROP, mean (SD) GA and birth weight were 29 (2) weeks and 1151 (346) g, respectively. Three hundred three patients (54%) were male; 399 (71%) were white, 87 (15%) were black, and 76 (14%) were of other or unknown race/ethnicity.
The demographics of the patients included in the ROP analysis are shown in Table 1 by SUPPORT era. Pre-SUPPORT and post-SUPPORT groups did not differ on GA or birth weight. The post-SUPPORT cohort had a lower rate of cesarean delivery (n = 210 [81%] pre-SUPPORT vs n = 220 [73%] post-SUPPORT; P = .03), had more exposure to antenatal steroids (n = 204 [78%] vs n = 272 [90%]; P <.001), and was more often resuscitated with continuous positive airway pressure (n = 92 [35%] vs n = 168 [56%]; P < .001) and room air (n = 11 [4%] vs n = 26 [9%]; P = .04) after delivery than the pre-SUPPORT cohort. The post-SUPPORT group also had a lower incidence of respiratory distress syndrome (n = 186 [72%] vs n = 192 [64%]; P = .05) and patent ductus arteriosus (n = 62 [24%] vs n = 40 (13%); P < .001). The rates of supplemental oxygen at 36 weeks (n = 50 [19%] vs n = 73 [24%]) and at discharge (n = 40 [15%] vs n = 55 [18%]) did not differ.
Table 2 presents the ophthalmic outcomes of the entire study cohort and the cohort eligible for ROP screening, respectively. Morality rates were similar within study periods: 15 of 275 patients (5%) died in the pre-SUPPORT era and 19 of 321 patients (6%) died in the post-SUPPORT era (absolute difference, 0.4%; 95% CI, −3% to 4%; P = .81). There was an increase in any ROP overall in the post-SUPPORT compared with the pre-SUPPORT cohort (n = 86 [27%] pre-SUPPORT vs n = 54 [20%] post-SUPPORT; absolute difference, 7%; 95% CI, 0.04%-14%; P = .04).
Quiz Ref IDAll 562 patients screened for ROP were followed up until ROP resolution (Table 3). The rate of treatment-requiring (type 1) ROP was higher in the post-SUPPORT era (n = 6 [2%)] pre-SUPPORT vs n = 18 [6%] post-SUPPORT, absolute difference, 4%; 95% CI, 0.4%-7%; P = .03; Table 3; eFigure in the Supplement), as well as an increase in any ROP overall in the post-SUPPORT compared with the pre-SUPPORT cohort (n = 53 [20%] pre-SUPPORT vs n = 86 [28%] post-SUPPORT; absolute difference, 8%; 95% CI, 1%-15%; P = .03). The increased risk of overall ROP in the post-SUPPORT group remained after adjusting for the known risk factors of GA, birth weight, sepsis, and meningitis lasting greater than 3 days and for intrafamily correlation among siblings, with an adjusted odds ratio of 1.8 (95% CI, 1.1-3.1; P = .03); interactions between risk factors were not at the .05 level and removed from the final model. Severity of ROP was also worse in the post-SUPPORT cohort, as seen by the higher stages and lower zones of ROP (Table 3).
We next stratified the full cohort by GA (≤28 weeks and >28 weeks, Table 4). Among all infants born at 28 weeks’ GA or earlier, incidence of any ROP and treated ROP were higher in infants in the static oxygen cohort (any ROP, n = 45 [34%] pre-SUPPORT vs n = 72 [49%] post-SUPPORT; absolute difference, 14%; 95% CI, 3%-26%; P = .02; treated ROP, n = 7 [5%] pre-SUPPORT vs n = 18 [12%] post-SUPPORT; absolute difference, 7%; 95% CI, 0.2%-13%; P = .047). Among infants born after 28 weeks’ GA, the pre-SUPPORT and post-SUPPORT cohorts did not differ on overall ROP or treated ROP. The median days required to fully vascularize the retina were longer in the post-SUPPORT cohort for the younger GA group. There was no difference in mortality between the pre-SUPPORT and post-SUPPORT cohorts in either GA group (Table 4).
We performed a time-to-event analysis for time to full vascularization, with patients censored at time of death or treatment of ROP using a Cox proportional hazards model. The model was adjusted for the GA group and the intrafamily correlation in sibling groups; an interaction between GA and SUPPORT era was included in the model. We found that among patients in the 22-week to 25-week and 26-week to 28-week GA groups, those in the post-SUPPORT cohort were less likely to be fully vascularized at any day of life compared with patients in the pre-SUPPORT group, with the smallest infants being worst affected. (Figure; eTable 2 in the Supplement). When calculating the hazard ratio, the likelihood of infants 28 weeks or younger with avascular retinas was almost 40% more at any given day in post-SUPPORT vs pre-SUPPORT cohorts (eTable 2 in the Supplement).
Quiz Ref IDIn our study, we observed that after switching to the SUPPORT guidelines from a biphasic oxygen protocol, type 1 ROP incidence was increased. There was also an increased incidence of any ROP and a delay of retinal vascularization, associated with an increase in the overall number of retinal examinations after the change in practice. We did not observe any difference in mortality between the 2 cohorts. We hypothesize that biphasic oxygen saturation targets might be different than static oxygen targets. These findings do not contradict the outcomes of SUPPORT nor do they contradict the outcomes of Supplemental Therapeutic Oxygen for Prethreshold Retinopathy of Prematurity (STOP-ROP).15 A biphasic approach is different than these important studies because it reciprocally mirrors the 2-step hypothesis of neovascularization first postulated by Ashton et al16 and proven by Alon et al6 and Pierce et al.5 The timing of the switch to higher oxygen settings follows the natural history of ROP: early, physiologic hypoxia (oxygen targets 85%-92% saturation) coincides with phase 1 but induces retinal growth to remove the substrate for disease, whereas late hyperoxia coincides with phase 2 to reduce abnormal neovascularization, much as was found in STOP-ROP. This hypothesis is confirmed by pharmaceutical hypoxic preconditioning, in which small molecules are shown to eradicate experimental oxygen-induced retinopathy by stabilizing hypoxia inducible factor during phase 1.17
We believe that the difference in mortality in the 2 groups found in studies, such as SUPPORT, might not apply to a biphasic oxygen standard, in which we have determined that mortality is the same when comparing static SUPPORT parameters with our biphasic targets. An important question regarding mortality is not addressed by our study, which is whether early low oxygen increases the risk of death early in GA. The fact that there were equal percentages of infants who died in each group prior to any eye examination (within 4 to 6 weeks of birth) suggests but does not prove that early low oxygen targets are not associated with early death.
The fact that retinal vascularization is delayed is an important finding for several reasons. First, delayed vascularization in severely premature infants requires longer retinal follow-up after discharge. Second, there is, to our knowledge, no understanding of the association of incomplete retinal vascularization with vision or the chance of late ROP in adolescence. Finally, delayed vascularization in the higher-saturation range group confirms the role that oxygen plays in the pathogenesis of retinovascular growth attenuation.
Another important finding of this study is the increased incidence of ROP, and this may be relevant for future studies. The retinal examination reflects systemic hyperoxia and tracks with other oxygen-induced disease processes such as bronchopulmonary dysplasia. Theoretically, the perfect NICU course is one that provokes no ROP, that is, the retina vascularizes and is seen as immature (stage 0) during the entire course of NICU stay. In units with lower incidence of type 1 ROP, it may be prudent to use incidence of ROP as a primary outcome measure in studies that hope to define the optimal NICU practices that permit normal sequential growth of the severely premature infant ex utero.
Quiz Ref IDWe are aware that a single-center, retrospective study is a major limitation by nature, and thus, we are cautious to overstate the findings. The homogeneity of this NICU, its low overall severity of ROP to begin with, and the consistency of a single examiner make this a valuable comparative analysis. In addition, the SUPPORT trial showed that low-oxygen vs high-oxygen targets increased mortality; according to our study, we did not see any increase in mortality when comparing a biphasic oxygen target with static or constant high oxygen saturations. However, we are unsure whether there would have been an increase in mortality in our preintervention group, if after 34 weeks CGA the oxygen saturations were not increased (90%-95%). Our data only demonstrate that hypoxia in early gestation and increase in oxygen saturation later in gestation is associated with reduced ROP but not increased mortality risk.
In conclusion, biphasic oxygen standards have a different outcome than static standards. Previous oxygen standards have demonstrated important facts: STOP-ROP showed that hyperoxia in phase 2 can decrease neovascularization and vasodilation; SUPPORT demonstrated that hypoxia decreases ROP but increases mortality. A biphasic approach hopes to take the benefits of both static studies to integrate them into the 2-phase hypothesis of ROP.
Corresponding Author: Jonathan E. Sears, MD, Cole Eye Institute and Cellular and Molecular Medicine, 9500 Euclid Ave, Cleveland, OH 44195 (firstname.lastname@example.org).
Accepted for Publication: December 5, 2018.
Published Online: February 14, 2019. doi:10.1001/jamaophthalmol.2018.7021
Author Contributions: Drs Shukla and Sears had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Shukla, Moore, Rodriguez, Hoppe.
Study concept and design: Sears.
Acquisition, analysis, or interpretation of data: Shukla, Sonnie, Worley, Sharma, Howard, Sears.
Drafting of the manuscript: Shukla, Sonnie, Worley, Howard, Moore, Rodriguez, Sears.
Critical revision of the manuscript for important intellectual content: Shukla, Worley, Sharma, Moore, Rodriguez, Hoppe, Sears.
Statistical analysis: Worley, Sharma, Sears.
Administrative, technical, or material support: Shukla, Sonnie, Sharma, Howard, Moore, Rodriguez, Hoppe, Sears.
Study supervision: Sears.
Conflict of Interest Disclosures: None reported.
Funding/Support: Grant support was recived from the National Institutes of Health (R01 EY024972; Dr Sears); The Hartwell Foundation Biomedical Research Fellowship (Dr Sears); and the Research to Prevent Blindness Physician Scientist award (Dr Sears).
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.