Introduction
Retinopathy of prematurity (ROP) was first associated with high oxygen at birth in the 1950s. An early theory was that, initially, high oxygen caused areas of avascular retina that became hypoxic after the infant was removed from high supplemental oxygen into ambient air. Then, secondly, the hypoxic, avascular retina produced angiogenic factors that caused blood vessels to grow abnormally into the vitreous.1 Since this original two-phased hypothesis was proposed, technology to regulate and monitor oxygen has been developed, and an important angiogenic factor in proliferative retinopathies was identified: vascular endothelial growth factor (VEGF). The goal of this article is to review major clinical trials that address treatment of severe ROP, especially recent anti-VEGF trials. However, questions still remain, particularly regarding outcomes following anti-VEGF treatments. Experimental evidence will be presented to provide insight into the mechanisms of action of anti-VEGF treatments on retina, causes for recurrences of severe ROP, and effects on the developing infant. Potential directions for future clinical and research studies will also be discussed.
Early Clinical Trials for ROP: Oxygen Studies
Early clinical trials provided evidence that high, uncontrolled oxygen at birth caused ROP at a time when premature babies were, on average, only about 2 months premature and born larger than infants who develop ROP today.2 With technology to regulate and monitor oxygen, ROP virtually disappeared. However, further advances in neonatal care led to the survival of extremely premature infants, and ROP reemerged.3 Later studies were performed that tested high or low oxygen saturation targets, but none has led to universally accepted standards among neonatologists. The Supplemental Therapeutic Oxygen for Prethreshold ROP (STOP-ROP) study tested high supplemental oxygen in infants with pre-threshold ROP to prevent threshold ROP,4 whereas the Surfactant, Positive Pressure and Pulse Oximetry Trial (SUPPORT),5 Benefits of Oxygen Saturation Targeting (BOOSTII),6,7 and Canadian Oxygen Trial (COT)8 compared lower oxygen saturation targets on ROP (85% to 89% vs. 91% to 95% SaO2). SUPPORT and BOOSTII found reduced ROP but increased mortality in the 85% to 89% SaO2 group, but COT found no difference between oxygen saturation targets on ROP or mortality. An analysis of published randomized trials did not find differences.9 Based on oxygen studies to date, high oxygen at birth is avoided, and neonatal nurseries tend to choose oxygen saturation targets based on individual data and outcomes. Some investigators found intermittent episodes of oxygen desaturation associated with severe ROP, which suggests a role for fluctuations in oxygenation in severe ROP.10–12
Experimental Evidence Supporting Study Outcomes
Fluctuations in oxygenation in newborn rat pups recreated arterial oxygen levels similar to transcutaneous values in premature infants with severe ROP.12,13 A rat model of fluctuating oxygen-induced retinopathy (OIR) was developed with features similar to infants with severe ROP: oxygen fluctuations recreating extremes similar to preterm infants at risk of severe ROP, poor postnatal growth, and appearance of early compromise in physiologic vascularity and delayed physiologic retinal vascular development with later development of intravitreal neovascularization. Using the model, it was found that oxygen fluctuations increased the expression of VEGF14 and of reactive oxygen species that triggered signaling through VEGF receptor 2 and the transcription factor, STAT3, in endothelial cells to lead to intravitreal neovascularization.15,16
Clinical Trials Testing Ablation of Avascular Retina
Given the finding of delayed physiologic retinal vascular development causing broad areas of avascular retina and the association with severe ROP, studies were developed to test cryotherapy and later laser when indirect ophthalmoscopic delivery of laser became available to ablate the peripheral retina (Cryotherapy for ROP17 [Cryo-ROP] and Early Treatment for ROP18 [ETROP]). Both multicenter trials reduced severe ROP compared to controls. Subsequent analyses of Cryo-ROP also identified eyes with zone I severe (“threshold”) ROP as having worse outcomes than eyes with zone II threshold ROP. The ETROP study redefined severe ROP to encompass a less severe level than threshold ROP, called type 1 ROP, and found laser treatment beneficial to reduce poor outcomes from severe ROP. Laser to the peripheral avascular retina has since been recommended as treatment of severe, now defined as type 1 ROP.19
Current Anti-VEGF Clinical Trials
VEGF was recognized as important in the pathogenesis of proliferative diabetic retinopathy and neovascular age-related macular degeneration.20–22 Preclinical studies were done in OIR models using animals that vascularized their retinas after birth and shared similarities to features of ROP. Therefore, it was logical to predict that VEGF would be involved in the pathology of severe ROP. There is now strong evidence that many methods to inhibit VEGF reduce plus disease and intravitreal neovascularization using models representative of human ROP.23–26 Furthermore, inhibiting signaling through VEGF receptor 2 (VEGFR2) specifically in endothelial cells not only reduces intravitreal neovascularization (stage 3 ROP), but also extends physiologic retinal vascular development.16 Although this seems counter intuitive, regulation of VEGFR2 signaling orders the division of proliferating endothelial cells so they extend into the retina instead of growing aberrantly into the vitreous, and this experimental finding also has been seen in human ROP following treatment with anti-VEGF agents.27
Bevacizumab Eliminated the Angiogenic Threat of Retinopathy of Prematurity (BEAT-ROP)
Because of the success of anti-VEGF agents in adult diseases, the Bevacizumab Eliminated the Angiogenic Threat of Retinopathy of Prematurity (BEAT-ROP) study tested bevacizumab (Avastin; Genentech, South San Francisco, CA) at half the adult dose (0.625 mg in 0.025mL) given as an intravitreal injection against laser in eyes with zone I or posterior zone II ROP, stage 3 and plus disease.27 The study reported that bevacizumab reduced recurrent severe ROP to 4% compared to laser in 22% of enrolled infants by 54 weeks postmenstrual age (PMA).
A number of studies subsequently reported recurrent severe ROP following bevacizumab treatment.28 A follow-up study including infants from the original BEAT-ROP study reported recurrences in 8.3% of infants, on average, 16 weeks after injection.29 This contrasted to a very low risk in eyes that did not receive anti-VEGF treatment but had persistent avascular retina in zone III without the development of pre-threshold ROP or worse by 45 weeks PMA.30
Dose De-Escalation of Bevacizumab: ROP1
One way to reduce the risk of anti-VEGF in the developing retina already compromised by oxygen stresses of prematurity would be adjusting the dose given to infant eyes. The dose of bevacizumab in the BEAT-ROP study was chosen based on available adult doses. Infant eye vitreous is approximately 1 mL compared to about 4 mL in the adult eye. Low intravitreal volumes reduce the risk of increased intraocular pressure, but administering small volumes can be associated with error and variability among individual injections. In addition to the risk from anti-VEGF to the retina, there is risk to the developing infant. Although the discrepancy in eye volumes between infant and adult is about 1:4, the difference in blood volumes is about 120 mL to approximately 6,000 mL or 1:20. Therefore, drug that enters the blood stream would be diluted into a lower blood volume in the infant compared to the adult and be a higher concentration. In fact, several studies found that bevacizumab was present in the blood stream of infants after single intravitreal injections and reduced serum VEGF for up to 2 months.31 Based on these concerns, the Pediatric Eye Disease Investigator Group (PEDIG) tested de-escalating doses of bevacizumab in type 1 ROP.32 This study included all type 1 severe ROP and not only posterior zone II or zone I with stage 3 ROP and plus disease. Also, doses were prepared by pharmacies and used 0.05 mL volume for higher and 0.01 mL volume for lower concentrations tested.
The study reported that severe ROP was reduced at 1 month with one-twentieth the dose used in BEAT-ROP, or 0.031 mg, in 55 of 58 infants of the 61 enrolled. At 6 months, 25 of 61 study eyes received another form of treatment, either another injection or laser; three had early failure within 4 weeks, 11 had recurrent vascular activity after 4 weeks, and 11 had persistent avascular retina. All 10 eyes given the 0.031 mg dose had success with attached retinas and no retreatment for early failure or late recurrence at 6 months, but three eyes had additional treatment for persistent peripheral avascular retina. At 6 months, stage 4A ROP developed in one eye, stage 4B developed in one eye, and stage 5 developed in two eyes from different infants. Six infants died from preexisting conditions prior to enrollment.32
Comparing Alternative Ranibizumab Dosages for Safety and Efficacy in ROP (CARE ROP)
Ranibizumab (Eylea; Regeneron, Tarrytown, NY) is a humanized recombinant antibody Fab fragment that is structurally derived from the light chains of bevacizumab and has a higher affinity for VEGF with a shorter half-life.33,34 Unlike bevacizumab, ranibizumab has been reported to have less of an effect on systemic VEGF.31,35
The Comparing Alternative Ranibizumab Dosages for Safety and Efficacy in Retinopathy of Prematurity (CARE-ROP) study compared 0.12 mg of intravitreous ranibizumab to 0.2 mg in Type 1 ROP in 19 infants and found equal success rates between the two groups in controlling acute ROP (94.4% in 0.12 mg group and 92.9% in 0.2 mg group), which was defined as not requiring rescue therapy at 24 weeks. However, infants were allowed repeated doses of ranibizumab, if needed, to replenish the drug. The VEGF plasma levels were not altered systemically in either group.35
Experimental Evidence Supporting anti-VEGF Study Outcomes
There are several potential reasons why severe ROP recurs after anti-VEGF treatment. Inhibition of VEGF with an intravitreal neutralizing anti-rat antibody, analogous to bevacizumab in human, caused recurrent IVNV and increased retinal VEGF in an experimental rat OIR model of ROP, suggesting compensatory increases in retinal VEGF signaling being one mechanism for recurrent intravitreal neovascularization.36 This would support the idea of why additional anti-VEGF reduces severity in some cases of recurrent severe ROP. However, additional studies show that other signaling pathways, including the mitogen-activated protein kinase (MAP kinase) pathways, are involved in pathologic angiogenesis in the rat model of ROP.37 VEGF is transcribed under conditions of hypoxia, which stabilizes hypoxia-inducible factors that initiate transcription of not only VEGF, but also of other angiogenic factors, including the angiopoietins and erythropoietin. These compounds have also been found to be involved in pathologic angiogenesis.38,39 Therefore, when VEGF is inhibited, hypoxia-inducible factors still transcribe other angiogenic factors. Finally, stimuli besides hypoxia, such as reactive oxygen species and inflammation, lead to angiogenesis through VEGF and other signaling mechanisms.
Experimental studies provide evidence that it is not only persistent peripheral avascular retina that becomes hypoxic and stimulates hypoxia-induced expression of angiogenic factors, but also compromise of existing physiologic vascularity. Visualization of hypoxic tissue using a compound that formed conjugates in tissue having 1% oxygen showed that hypoxic retina occurred in between capillaries of vascularized retina in the rat model of fluctuating OIR. This contrasted with no hypoxic retina in the peripheral avascular retina in pups raised in room air40 (Figure 1). In the rat OIR model, an intravitreal neutralizing rat anti-VEGF antibody, analogous to bevacizumab in humans, significantly reduced capillary density in existing vascularized retina compared to control treated OIR animals.41 This supports the notion that oxygen stresses similar to those to which premature infants are exposed not only reduce physiologic retinal vascularization but also cause hypoxic retina within the vascularized retina, and that intravitreal anti-VEGF further compromises vascularized retina. The outcomes suggest anti-VEGF treatment at too high a dose or for less severe ROP than type 1 ROP may injure vascularized retina, stimulate additional angiogenic factor expression, and lead to recurrent severe ROP.
Still, the presence of persistent avascular retina is a concern in an infant treated with anti-VEGF, even if physiologic vascularization proceeds into zone III. Persistent avascular retina can be difficult to detect in infants. Careful indirect ophthalmoscopy is difficult as an infant becomes larger and matures. RetCam (Natus Medical, Pleasanton, CA) imaging is difficult and tends to under call zone III ROP and may miss peripheral avascular retina.42 Fluorescein angiography (FA) can be helpful. In eyes treated with the anti-VEGF, ranibizumab, for severe ROP, only 50% were vascularized into zone III when evaluated with fluorescein angiography.43
Refractive Error after anti-VEGF or Laser Treatment
Several studies have reported less myopia with anti-VEGF treatment compared to laser treatment for severe ROP,29,44,45 but the evidence to date is based on association and not causation. Some reports define a difference in the development of the anterior eye in anti-VEGF versus laser-treated infants. One study reported deeper anterior chambers in eyes treated with anti-VEGF.46 Other studies report refractive differences in premature infant eyes that have macular edema and the refractive errors changed when the edema resolved.47
Assessing Retinal Development, Evolution and Treatment of ROP by Retinal Imaging Modalities
Most trials in ROP have relied on clinician assessment and drawings that depict retinal development and features of ROP severity. Retinal images of infants taken with the RetCam system by trained experts were of adequate quality in 91% of ROP screening images,48 but agreement between image grading and clinical examination as to whether referral was warranted ranged from 46.5% to 70%, depending on which was considered the gold standard.49 Foveal vasculature and vascular perfusion are not detected in retinal images, but abnormalities in these were detected in small fluorescein angiographic studies of infants with ROP during early development and later in childhood, as well as after high oxygen exposure50–56 or after anti-VEGF treatment.55,57 FA has the drawback of intravenous dye injection, but retinal images alone do not provide details about retinal layers, vascular plexuses, the choroid, vitreoretinal traction or intra- or subretinal fluid — all of which are affected in prematurity, ROP, and by treatment for ROP.
Alternative imaging modalities such as spectral-domain optical coherence tomography (SD-OCT) and OCT angiography (OCTA) have provided objective documentation of unrecognized subclinical features that may be relevant to ROP, including after anti-VEGF treatment.58 Bedside SD-OCT imaging is obtained with barely visible near-infrared illumination and can be performed without pharmacological dilation.59 Delineation of retinal features across the vascular-avascular junction, retinal surface budding into the vitreous cavity, plus disease, inner retinal and photoreceptor maturation at the fovea,60 macular edema with cystoid spaces,61 choroidal thickness,62 retinal schisis, and detachment are only a few of the features that have been characterized with SD-OCT.56,59,61,63–67 OCTA, based on light reflectance from moving blood cells, enables imaging of retinal vascular flow without dye injection. The delineation of retinal vasculature through OCTA is impacting the assessment of children with a history of preterm birth and ROP68–72 and will likely to provide new insight into normal and abnormal vascular development including the effects of various treatments for ROP. These imaging modalities do not replace the need for clinical documentation, especially of the peripheral retina, which is poorly imaged with current OCT. More definitive, larger scale studies are needed. The RAnibizumab Compared With Laser Therapy for the Treatment of INfants BOrn Prematurely With Retinopathy of Prematurity (RAINBOW) study (NCT02375971) testing ranibizumab versus laser for severe ROP will provide additional information as will the study Analyzing Retinal Microanatomy in ROP (NCT02887157).