Accepted Manuscript

Pathophysiology, screening and treatment of ROP: A multi-disciplinary perspective

Tailoi Chan-Ling, Glen A. Gole, Graham E. Quinn, Samuel J. Adamson, Brian A. Darlow

PII: S1350-9462(16)30077-5 DOI: 10.1016/j.preteyeres.2017.09.002 Reference: JPRR 685

To appear in: Progress in Retinal and Eye Research

Received Date: 3 February 2017 Revised Date: 18 September 2017 Accepted Date: 20 September 2017

Please cite this article as: Chan-Ling, T., Gole, G.A., Quinn, G.E., Adamson, S.J., Darlow, B.A., Pathophysiology, screening and treatment of ROP: A multi-disciplinary perspective, Progress in Retinal and Eye Research (2017), doi: 10.1016/j.preteyeres.2017.09.002.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. ACCEPTED MANUSCRIPT Pathophysiology, Screening and Treatment of ROP: a multi-disciplinary perspective Tailoi Chan-Ling 1 Glen A. Gole 2 Graham E. Quinn 3 Samuel J. Adamson 1 Brian A. Darlow 4

Affiliations: 1 Department of Anatomy, School of Medical Sciences and Bosch Institute, , NSW 2006, [email protected] 2 Discipline of Paediatrics and Child Health, University of Queensland, Qld Children’s Hospital, Sth Brisbane, Qld 4101, Australia [email protected] 3. Division of , The Children’s Hospital of Philadelphia and Department of Ophthalmology, Perelman School of , University of Pennsylvania, Philadelphia, PA, USA [email protected] 4. Department of Paediatrics, University of Otago, Christchurch, New Zealand. [email protected]. MANUSCRIPT * TC-L, GAG, GEQ and BAD contributed equally to this review.

Corresponding author : Tailoi Chan-Ling Department of Anatomy School of Medical Sciences and Bosch Institute University of Sydney, NSW 2006 Australia [email protected] T: +61 2 9351 2596ACCEPTED F: +61 2 9351 6556

Acknowledgement: This work has been supported by: the National Health and Medical Research Council of Australia (#571100, #1005730), the Brian M Kirby Foundation - Gift of Sight Initiative, the Bonnie Babes Foundation (Lucy Turnbull AO and The Hon. Prime Minister of Australia ), the Baxter Charitable Foundation, the Sydney Foundation, the Alma Hazel Eddy Trust, The Rebecca L Cooper Medical Research Foundation and the Geoffrey Arnott Foundation to TC-L, the NWG Macintosh Memorial Fund to SJA and Cure Kids New Zealand to BAD. 1 Table of Contents ACCEPTED MANUSCRIPT 2 Abstract ...... 3 3 List of Abbreviations ...... 4 4 Introduction - Epidemiology of ROP ...... 7 5 Purpose ...... 7 6 Cellular and Molecular Mechanisms of Human Retinal Vascularization ...... 8 7 Development of the human retinal vasculature ...... 8 8 ‘Physiological hypoxia’ model of formation of human retinal vasculature ...... 10 9 Cellular & molecular mechanisms of formation of the foveal avascular zone (FAZ) ...... 11 10 Glial, vascular and neuronal interactions in the formation of the foveal vasculature ...... 11 11 Neurosensory Retina in ROP ...... 13 12 Effects of premature birth on retinal development and function: Impact on extremely premature infants 13 ...... 14 14 Choroidal involution as a component of ROP ...... 15 15 Optical Coherence Tomography (OCT) and ROP ...... 16 16 Unique susceptibility of posterior ROP: Basis in vasculogenic origin of posterior vessels...... 16 17 ROP develops in two phases ...... 18 18 Initiating event in pathogenesis of ROP - delayed vascularisation relative to neuronal maturation (Phase 1 19 ROP) ...... 19 20 Phase 2 ROP: Hypoxia-induced pathological neovascularization ...... 19 21 Limitations of current management strategies for ROP, including anti-VEGF: ...... 20 22 The Rationale for Dark-Rearing (DR), a non-invasive for the treatment of ROP ...... 20 23 Scientific rationale for distinct pathogenetic mechanisms underlying zone 1 & zone 2 ROP ...... 22 24 VEGF165 subserves multiple functions: angiogenic growth factor, vascular permeability factor and 25 neuronal survival factor...... 23 26 Background ...... 23 27 VEGF Biology ...... MANUSCRIPT...... 24 28 Control and expression of VEGF and associated genes via HIF-1 ...... 24 29 VEGF Signaling ...... 25 30 VEGF and pathogenesis of ROP ...... 26 31 VEGF and neuroprotection in the retina ...... 26 32 Role of VEGF in maintaining vessel stability in early postnatal period...... 27 33 Role of VEGF in vascular permeability ...... 28 34 Screening for ROP ...... 28 35 Telemedicine and ROP screening ...... 30 36 Fluorescein angiography before and after treatment for ROP ...... 34 37 Treatment of ROP ...... 36 38 CRYO-ROP Study ...... 37 39 Laser versus Cryotherapy ...... 37 40 Laser for ROP...... 38 41 Timing of intervention for severe ROP ...... 40 42 Side effects of laser ...... ACCEPTED...... 40 43 Long term outcomes of laser treatment ...... 41 44 Recommendations for long term follow up...... 42 45 Outcomes after laser for AP-ROP ...... 42 46 Anti-VEGF therapy in ROP: The pros and cons ...... 43 47 VEGF-based therapeutic options ...... 43 48 BEAT-ROP Study ...... 45 49 Anti-VEGF therapy in combination with laser ...... 46 50 Dosage considerations ...... 48 51 Long term systemic effects of anti-VEGF treatment ...... 49 52 Ocular complications of anti-VEGF treatment ...... 51 53 Long-term ophthalmic outcomes of anti-VEGF treatment for ROP...... 51 1 54 Vitrectomy – structural and visual outcomes ...... 51 ACCEPTED MANUSCRIPT 55 Effects of premature birth on retinal development/function: Impact on larger premature infants .52 56 ROP and Neurodevelopment ...... 52 57 Recent trials of oxygen saturation targeting ...... 54 58 Gradual oxygen withdrawal as strategy to minimize Phase 2 ROP – Evidence from the 59 kitten model of ROP ...... 56 60 Low serum levels of Insulin-like Growth Factor-1 (IGF-1) correlate with ROP severity ...... 58 61 Omega 3 supplementation ...... 59 62 Human Milk ...... 59 63 Novel molecular targets for ROP over the horizon ...... 60 64 Propranolol ...... 60 65 EPO ...... 61 66 Anti-inflammatory therapy and ROP ...... 61 67 Recommendations for current best practice for extremely premature infants ...... 63 68 Non-ophthalmic strategies for prevention of, and treatment for ROP ...... 64 69 Recommendation of best practice for larger infants ...... 65 70 Future Directions ...... 65 71 Overview ...... 66 72 73 MANUSCRIPT

ACCEPTED

2 74 Abstract ACCEPTED MANUSCRIPT 75 The population of infants at risk for retinopathy of prematurity (ROP) varies by world region; in countries 76 with well developed neonatal intensive care services, the highest risk infants are those born at less than 28 77 weeks gestational age (GA) and less than 1 kg at birth, while, in regions where many aspects of neonatal 78 intensive and ophthalmological care are not routinely available, more mature infants up to 2000 g at birth 79 and 37 weeks GA are also at risk for severe ROP. Treatment options for both groups of patients include 80 standard retinal laser photocoagulation or, more recently, intravitreal anti-VEGF drugs. In addition to 81 detection and treatment of ROP, this review highlights new opportunities created by telemedicine, where 82 screening and diagnosis of ROP in remote locations can be undertaken by non-ophthalmologists using 83 digital fundus cameras. The ophthalmological care of the ROP infant is undertaken in the wider context 84 of neonatal care and general wellbeing of the infant. Because of this context, this review takes a multi- 85 disciplinary perspective with contributions from retinal vascular biologists, pediatric ophthalmologists, an 86 epidemiologist and a neonatologist. This review highlights the latest insights regarding cellular and 87 molecular mechanisms in the formation of the retinal vasculature in the human infant, pathogenesis of 88 ROP, detection and treatment of severe ROP, the risks and benefits of anti-VEGF therapy, the 89 identification of new over the horizon, and the optimal neonatal care regimen for best ROP 90 outcomes, and the benefits and pitfalls of telemedicine in the remote screening and diagnosis of ROP, all 91 of which have the potential to improve ROP outcomes 92 MANUSCRIPT

ACCEPTED

3 93 List of Abbreviations ACCEPTED MANUSCRIPT 94 ACTRN - Australian Clinical Trial Registration Number 95 ADPase - Adenosine Diphosphatase 96 AP-ROP - Aggressive Posterior Retinopathy of Prematurity 97 APC - Astrocyte Precursor Cell 98 AVIOX - Achieved versus intended pulse oximeter saturation 99 BBB - Blood Brain Barrier 100 BEAT-ROP - Bevacizumab Eliminates the Angiogenic Threat of Retinopathy of Prematurity Study 101 BOOST - Benefits of Oxygen Saturation Targeting Trial 102 BOOST-NZ - Benefits of Oxygen Saturation Targeting Trial - New Zealand 103 BPD - bronchopulmonary dysplasia 104 BRB - Blood Retinal Barrier 105 BW - Birth Weight 106 CAP - Caffeine for Apnoea of Prematurity 107 CD39 - Cluster of Differentiation 39 108 CHOPROP - Children's Hospital of Philadelphia Model 109 CI - Confidence Interval 110 CLD - chronic neonatal lung disease 111 CME - cystoid macular edema 112 CNN - Candian Neonatal Network MANUSCRIPT 113 COT - Canadian Oxygen Trial 114 CPAP - Constant Positive Airway Pressure 115 CRYO-ROP - Multicenter Trial of Cryotherapy for Retinopathy of Prematurity Study 116 d - diopter 117 DHA - docosahexaenoic acid 118 DNA - Deoxyribonucleic Acid 119 e-ROP - Telemedicine Approaches to Evaluating Acute-phase Retinopathy of Prematurity Study (e-ROP) 120 ELBW - Extremely Low Birth Weight 121 EPA - eicosapentaenoic acid 122 EphA6 - Ephrin A6 ACCEPTED 123 Epo - Erythropoietin 124 ERG - Electroretinogram 125 ETROP - The Early Treatment for Retinopathy of Prematurity Study 126 FA - Fluorescein Angiography 127 FAZ - Foveal Avascular Zone 128 fiO2 - fraction of inspired oxygen

4 129 FIRS - fetal inflammatory response syndrome ACCEPTED MANUSCRIPT 130 GA - Gestational Age 131 GCL - Ganglion Cell Layer 132 GFAP - Glial Fibrillary Acidic Protein 133 HIF1a - Hypoxia Inducible Factor 1 alpha 134 HRQL - Health Related Quality of Life 135 IGF-1 - Insulin-like Growth Factor 1 136 IGFBP3 - Insulin-like Growth Factor Binding Protein 3 137 IgG - Immunoglobulin G 138 IH - infantile hemangioma 139 INL - Inner Nuclear Layer 140 IOP - Intraocular Pressure 141 IPD - individual patient data 142 IQ - Intelligence Quotient 143 IVB - Intravitreal Bevacizumab 144 KIDROP - Karnataka Internet Assisted Diagnosis of Retinopathy of Prematurity 145 LCPUA - long chain polyunsaturated fatty acid 146 ml - millilitre 147 NDD - Neurodevelopmental Delay MANUSCRIPT 148 ng - nanogram 149 NICU - Neonatal Intensive Care Unit 150 OCT - Optical Coherence Tomography 151 OIR - oxygen induced retinopathy, term applied to the animal models of retinopathy of prematurity 152 ONH - Optic Nerve Head 153 ONL - Outer Nuclear Layer 154 OR - odds ratio 155 Pax2 - Paired box gene 2 156 PEDF - Pigment Epithelium-Derived Factor 157 pg - picogram 158 PlGF - Placental GrowthACCEPTED Factor 159 PMA - Post-Menstrual Age 160 PVL - Periventricular leukomalacia 161 RCT - Randomized Controlled Trial 162 rhEpo - recombinant Erythropoietin 163 rhIGFBP3 - recombinant Insulin-like Growth Factor Binding Protein 3 164 RLF - Retrolental Fibroplasia

5 165 ROP - Retinopathy of Prematurity ACCEPTED MANUSCRIPT 166 RPC - Radial Peri-papillary Capillary 167 RPE - Retinal Pigment Epithelium 168 RW-ROP - Referral Warranted Retinopathy of Prematurity 169 siRNA - Small Interfering Ribonucleic Acid 170 SOT - supplemental oxygen therapy 171 spO2 - peripheral capillary oxygen saturation 172 SUNDROP - Stanford University Network for Diagnosis of Retinopathy of Prematurity 173 SUPPORT - Surfactant, Positive Pressure, and Oxygenation Randomized Trial 174 TW-ROP - Treatment-Warranted Retinopathy of Prematurity 175 VEGF - Vascular Endothelial Growth Factor 176 VEGFR2 - Vascular Endothelial Growth Factor Receptor 2 177 VHL - Von Hippel Lindau protein 178 VPC - Vascular Precursor Cell 179 WINROP - weight, insulin-like growth factor I, neonatal, retinopathy of prematurity algorithim 180 WMI - White Matter Injury 181

MANUSCRIPT

ACCEPTED

6 182 Introduction - Epidemiology ofACCEPTED ROP MANUSCRIPT 183 Globally, it has been estimated that more than 20,000 infants are blinded annually from retinopathy of 184 prematurity (ROP), and an additional 12,300 have mild to moderate visual impairment (Blencowe et al., 185 2013) (Figure 1). This estimate, which updates previous figures based largely on reviews of the causes of 186 blindness amongst children attending schools for the blind (Gilbert, 2008) is derived from published data 187 on the incidence of preterm birth, mortality rates and the proportion of ROP requiring treatment, and 188 suggests the greatest burden of disease is now in the rapidly developing economies of India, China, South 189 East Asia and South America. In addition to vision loss, ROP can cause widespread deficits in visual 190 function (Fielder et al., 2015) including reduced contrast sensitivity, visual field defects, colour vision 191 deficits, strabismus and refractive errors. In countries with well-developed neonatal care, data from large 192 neonatal networks show a recent decline in the incidence of severe ROP (defined either as stage 3 or 193 more, or as receiving treatment) in infants of 28 weeks gestational age (GA) or less (Stoll et al., 2015) or 194 with birthweight 5001-1500g (Darlow et al., 2017, Ellsbury et al., 2016).A recent study based on data on 195 infants of 24 0 to 27 6 weeks’ gestation prospectively collected by 10 population-based networks from 10 196 countries and one region from 2007-13 (involving 48,087 infants), found the overall trend in the 197 incidence ROP needing treatment declined from 2007-13 but was stable in the last six years, and with 198 only the UK experiencing a rising trend (Darlow 2017). However, with increasing numbers and 199 increasing survival of extremely premature infants (<26 weeks GA), there is likely to be an increasing 200 number of infants affected by ROP (Austeng et al., 2009,MANUSCRIPT Costeloe et al., 2012, Painter et al., 2015). In 201 middle- and low-income countries where there is often less expert neonatal care, a lack of oxygen 202 saturation monitors and little or no advanced training for neonatologists, ophthalmologists and neonatal 203 nurses, both any and severe ROP are seen not only in the smallest infants if they do survive but also in 204 infants who are more mature than the very preterm infants seen in the developed world, who have average 205 gestation of 25 weeks and a birth weight of approximately 750 g (Gilbert et al., 2005). Premature infants 206 upwards to 34 weeks GA and sometimes >1500g birthweight develop ROP in countries with still 207 developing neonatal care services.(Gilbert et al., 2005). These differences in care vary both by region and 208 sometimes within the region according to the resources and levels of care provided (Zin et al., 2010). 209 210 Purpose ACCEPTED 211 The increase in the incidence of ROP in middle income and newly industrialized countries, which is 212 occurring at the same time as the increasing survival of very immature infants, has been referred to as the 213 “third epidemic of ROP ” (Gilbert, 2008, Quinn et al., 2010), making a review of the pathogenesis and 214 treatment options for ROP timely. The two patient groups present different challenges in terms of ROP 215 detection and care, so there is a pressing need for improved strategies to manage ROP globally. Many 216 new developments in retinal biology, ROP pathogenesis, and treatment of ROP have arisen since our last 217 review on ROP in Progress in Retinal and Eye Research (Chan-Ling and Stone, 1993). This critical 7 218 review brings a collected perspective from the viewpoints of retinal/VEGF biologists, paediatric ACCEPTED MANUSCRIPT 219 ophthalmologists, epidemiologists and neonatologists with specialist interest in ROP. 220 221 Cellular and Molecular Mechanisms of Human Retinal Vascularization 222 Development of the human retinal vasculature

223 The human retina is vascularized by two distinct mechanisms: vasculogenesis and angiogenesis (Hughes 224 et al., 2000). Vasculogenesis is the de novo formation of vessels by the aggregation of endothelial 225 precursors. Vessels develop from vascular precursor cells (VPCs) that aggregate into solid vascular cords, 226 which then become patent and differentiate to form primitive endothelial tubes. On the other hand, 227 formation of vessels by angiogenesis occurs through budding from existing vessels; this process takes 228 place through the migration and proliferation of vascular endothelial cells (Chan-Ling et al., 1990).

229 The first event in the development of the retinal vasculature is the de novo formation of vessels by 230 vasculogenesis, detectable in the human retina before 12 weeks GA (Hughes et al., 2000). The spindle- 231 shaped VPCs can be Nissl stained (Hughes et al., 2000) and also express CD39 (Hasegawa et al., 2008), 232 vascular endothelial growth factor receptor 2 (VEGFR2), and ADPase (Chan-Ling et al., 2004a). The 233 VPCs migrate from the optic disk where vascular chords begin to coalesce on the inner surface of the 234 retina, behind the wave of individual spindle cells, beginning in the region proximal to the optic nerve 235 head and progressing outward towards the peripheral retina, as early as 14–15 weeks GA (Figure 2A). 236 This maturation of the retinal vasculature takes place with MANUSCRIPT a central to peripheral topography centered 237 around the optic nerve head. These chords become patent to form a primordial vascular bed centered 238 around the optic nerve head and expand in a four-lobed pattern to become the future 4 major artery-vein 239 pairs seen in the adult human retina. These first primordial vessels, formed through vasculogenesis, are 240 typically radial, have uniform diameter and have low capillary density (Figure 2B). The formation of 241 primitive vessels lags behind the leading edge of VPCs by a distance of at least 1 mm (Chan-Ling et al., 242 2004a). The formation of retinal vessels via the vasculogenic process appears to be complete by 21 243 weeks. The formation of solid vascular cords and transformation of these cords to form the first 244 primordial patent vessels centered around the optic nerve head expand in a four-lobed pattern, as seen in 245 Figure 2A,B. The lobes extend the farthest in the temporal and superior directions, and they correspond to 246 the future location of theACCEPTED four major artery–vein pairs in the adult retina. The lobes curve around the 247 region of the incipient fovea, leaving this area free of spindle cells and vascular cords. The fovea and 248 perifoveal region remains avascular through 25 weeks GA. This avascularity at the incipient fovea is also 249 observed in the macaque retina in development, where neither retinal vessels, nor their accompanying 250 astrocytes grow into a circumscribed region which, at a later stage, develops into the foveal depression 251 (Provis, 2001).

252 8 253 The primordial vessels formed by vasculogenesis are insufficient to meet the metabolic demands of the ACCEPTED MANUSCRIPT 254 posterior retina as the capillaries show minimal branching. Capillary density is expanded in the posterior 255 retina via angiogenic budding from the primordial vessels. In contrast to the vasculogenic process, no 256 spindle-shaped mesenchymal cells are seen in the peripheral two-thirds of the retina; as a consequence, 257 the formation of the blood vessels in the peripheral retina, as well as the formation of the outer, deep 258 plexus and the radial RPCs near the ONH occurs by angiogenesis. Concomitant with the process of 259 retinal vascular formation, retinal cells of the astrocyte (Chu et al., 2001) and pericyte (Hughes and Chan- 260 Ling, 2004) lineages migrate from the optic nerve and enter the human retina between 12-26 weeks. 261 Pax2 +/GFAP - astrocyte precursor cells and vascular precursor cells invade the retina from the optic nerve 262 head (Figure 2A-C) where astrocytes serve as a template for migration of the vascular lineage cells 263 towards the avascular peripheral retina (Gerhardt et al., 2003). Further, cells of the astrocytic lineage

264 express VEGF 165 to regulate angiogenic vascular formation in the inner retinal plexus (Stone et al., 1995, 265 Provis et al., 1997).

266 267 Angiogenic growth is driven by neuronal maturation. Beginning in the region surrounding the fovea 268 around 25–26 weeks GA, the superficial vessels start to bud and grow down from the superficial plexus 269 of the retina (Hughes et al., 2000) towards the inner nuclear layer. The vessels ramify at the junction of 270 the inner nuclear and the outer plexiform layers. This stage coincides with the period of eye opening at 271 25-26 weeks gestational age in the human infant when theMANUSCRIPT visually evoked potential, indicative of a 272 functional visual pathway and photoreceptor activity, is first detectable (Purpura, 1975). The growth of 273 capillaries that descend from the superficial plexus to form the deep vascular bed is not preceded by

274 VPCs or astrocyte precursor cells (APCs). Rather, VEGF 165 expression in this region is associated with

275 the soma of Muller cells, suggesting that these cells are the source of the VEGF 165 that stimulates and 276 guides endothelial cell growth in the outer vascular plexus (Stone et al., 1995). Unlike the inner plexus, 277 the formation of the outer plexus begins at and is centered around the fovea (Hughes et al., 2000), whilst 278 the fovea itself remains avascular. Formation of the outer plexus corresponds with the maturation pattern 279 of the neuronal retina, suggesting that increased metabolic demands from active neurons produce local, 280 ‘physiological hypoxia’ to induce the growth of vessels to these tissues (Chan-Ling et al., 1995a). It is 281 important to note that ACCEPTEDthe ‘physiologic’ levels of hypoxia do not damage the surrounding tissue, but is 282 sufficient to signal the need for increased oxygen and other metabolic needs. Figure 2D demonstrates this 283 progression of the patent vasculature in both the superficial and deeper vascular plexuses in fetal 284 development (25 weeks GA) to full-term (40 weeks GA). 285

9 286 ‘Physiological hypoxia’ model ofA CCEPTEDformation ofMANUSCRIPT human retinal vasculature 287 It had long been suspected that the oxygen needs of the tissue drives the developing vasculature, and that 288 a hypoxia-induced vasoformative growth factor mediates this process (Michaelson et al., 1954). The 289 central role of oxygen in the control of retinal vessel development was demonstrated by Patz and Ashton 290 (Patz et al., 1953, Ashton et al., 1954, Ashton, 1966, Ashton, 1970) with subsequent studies providing 291 evidence that oxygen controls the microarchitecture of the retinal vessels (Glaser et al., 1980, D'amore et 292 al., 1981, Chen and Chen, 1984, Chen and Chen, 1987, Chen and Chen, 1980, Phelps, 1990) and that the 293 pattern of vessel growth mirrors that of the developing retinal neurons (Chan-Ling et al., 1990, Hughes et 294 al., 2000) These observations led us to propose that ‘physiological hypoxia’ is the stimulus for normal 295 retinal angiogenesis (Chan-Ling, 1994). The “physiological hypoxia” model of retinal angiogenesis 296 demonstrated localized angiogenesis is coincident with unmet metabolic demand (Chan-Ling et al.,

297 1995a) with the subsequent identification of vascular endothelial growth factor (VEGF 165 ) as the 298 angiogenic, hypoxia-induced growth factor responsible for vascular development in the retina of the rat 299 and cat (Stone et al., 1995, Provis et al., 1997) and human retina (Provis et al., 1997). This level of tissue 300 hypoxia is sufficient to induce hypoxia inducible factor-1 alpha (HIF-1α) (Semenza and Wang, 1992),

301 which in turn stimulates the expression of VEGF 165 by astrocytes and Müller cells (Stone et al., 1995) to 302 direct endothelial growth where it is needed. Once the new vessels become patent, providing blood flow 303 to the developing retinal tissues the process of vascular remodeling and retraction continues for a number 304 of weeks and is still ongoing in a full-term neonate (GerhardtMANUSCRIPT et al., 2003, Hughes et al., 2000). The 305 balance between ‘physiological hypoxia’ and hyperoxia, vessel growth and vessel remodeling (Hughes 306 and Chan-Ling, 2000) leads to the formation of a vasculature that precisely meets the metabolic demands 307 of the human retina.

308 Figure 3A-B displays VEGF expression during normal human fetal development, at 12, 14 and 16 weeks 309 GA (Provis et al., 1997). VEGF expression in the RPE at 12 weeks GA precedes that of the inner retina, 310 where VEGF is not readily detectable (Provis et al., 1997). At 14-16 weeks GA, expression of VEGF is 311 still not apparent in the inner retina, and there is no evidence of a vascular bed internal to the ganglion 312 cell layer. By 20 weeks GA (Figure 3B+C), VEGF expression co-localized with vessels formed by 313 angiogenesis as determined in our earlier studies (Hughes et al., 2000)(Figure 2C). Figure 3D 314 demonstrates the highlyACCEPTED conserved process of temporal and topographical expression of VEGF in 315 mammalian retinal development, where the rat has an RPE focused expression of VEGF early in 316 development (Postnatal Day 3), that gradually transforms to an astrocyte and Muller cell driven 317 expression of VEGF at both the ganglion cell layer and the mid-level of the INL (coincident with location 318 of Müller cell somas) at Postnatal Day 10, similar to that also demonstrated in cat,and human.

10 319 Cellular & molecular mechanismsACCEPTED of formation MANUSCRIPT of the foveal avascular zone (FAZ) 320 The human and primate retina is characterized by the presence of an avascular zone at the fovea 321 (Reviewed in (Provis and Madigan, 2013, Provis, 2001) This avascular zone has been denoted as the 322 “foveal avascular zone” (FAZ) (Figure 3E-H). Similar avascular zones can be seen in the kitten (Figure 323 3K) and primate retina (Figure 3L-M). There is an interdependent relationship between retinal vascular 324 development, foveal development and photoreceptor maturation, and modulation of this relationship in 325 ROP can have significant outcome on visual acuity in later life. The FAZ forms in the ganglion cell layer 326 (GCL), between 25 and 27 weeks (Hughes et al., 2000) and overlies the foveal cone mosaic, which is 327 responsible for central vision and is specialized to provide maximum visual acuity. Because of the 328 thinness of this region the FAZ is not reliant on the inner and outer retinal vascular plexuses for 329 oxygenation; rather it is supplied by diffusion from the choriocapillaris, allowing the fovea to remain 330 vessel free and permitting the tight packing of cone photoreceptors required for high visual acuity. 331 Similarly, the topography of astrocyte/vascular invasion in the human (Chu et al., 2001, Hughes et al., 332 2000) kitten (Ling and Stone, 1988, Chan-Ling et al., 1990); and primate retina (Provis et al., 1997) is 333 characterized by the avoidance of the temporal raphe and incipient fovea (Figure 3E,H,K).

334 Whilst the absence of astrocytes and blood vessels from both the incipient and adult fovea has been noted 335 by earlier workers, Kozulin and colleagues have added key molecular insights regarding the mechanisms 336 by which the fovea maintains its avascularity (Kozulin et al., 2010, Kozulin et al., 2009). In human, the 337 FAZ is specified at 20 weeks GA by ganglion cell and RPEMANUSCRIPT expression of the anti-angiogenic and anti- 338 migratory factor, pigment epithelium–derived factor (PEDF) (Dawson et al., 1999) and similar 339 mechanisms have also been demonstrated in the primate retina (Kozulin et al., 2010, Kozulin et al., 2009) 340 (Figure 3L). The avascular zone in the incipient fovea is further defined by the expression of EphA6 (a 341 receptor for both guidance factors ephrin-A1 and ephrin-A4) in the GCL. Studies have shown that 342 astrocytes express the ephrin-A1 and ephrin-A4 ligands (Kozulin et al., 2009). As EphA6 interacts 343 directly with ephrin-A1 and ephrin-A4 to mediate a repellent response, the developing vasculature is, in 344 effect, guided around the areas of high EphA6 expression at the FAZ (Figure 3M). Thus, this evidence 345 indicates that the FAZ is defined by the combined effects of both EphA6 and PEDF; a gradient of EphA6 346 expression regulates the rate (and possibly the path) of migrating astrocytes and a gradient of PEDF 347 expression suppressesACCEPTED endothelial proliferation (Provis and Madigan, 2013). 348 349 Glial, vascular and neuronal interactions in the formation of the foveal vasculature

350 The first retinal neurons to exit the mitotic cycle are the foveal cone photoreceptors (Provis et al., 1985), 351 which largely remain in the central retina from 10-12 weeks GA, where they slowly migrate (in 352 comparison to the rods) in a centrifugal fashion towards the periphery to create low levels of cone density 353 - though the formation of the foveal pit does not take place until much later – approximately 26-28 weeks

11 354 GA (Provis et al., 1985, Yuodelis and Hendrickson, 1986). Further, although the foveal cone ACCEPTED MANUSCRIPT 355 photoreceptors are the first cells to differentiate, they are the last to achieve adult-like characteristics 356 (Narayanan and Wadhwa, 1998, Yuodelis and Hendrickson, 1986), so that the formation of the fovea is 357 incomplete at term.

358 At full term, the outer nuclear layer (ONL) is comprised of immature cone photoreceptors and the fovea 359 still includes cells from the GCL and inner nuclear layer (INL) (Hendrickson et al., 2012). 360 Morphologically the cone photoreceptors in the incipient foveal pit remain cuboidal, little changed from 361 their appearance at mid-gestation, whereas those at the rim of the fovea are elongated, with inner and 362 outer segments beginning to develop along with axonal processes (Yuodelis and Hendrickson, 1986) 363 (Figure 4A-E). The high density of cone photoreceptors needed for best visual acuity is achieved by the 364 mass displacement of cones from peripheral areas of the retina to the fovea. This displacement is 365 evidenced by the fact there is no detectable cell death during this migratory period, and no increase in 366 photoreceptor numbers across the mosaic (Provis and Madigan, 2013, Cornish et al., 2004). The 367 centripetal migration of cone photoreceptors from the periphery towards the fovea occurs from 13 to 24 368 weeks GA (Diaz-Araya and Provis, 1992), the beginnings of cone maturation at the foveal pit and the 369 onset of the ERG (Dreher and Robinson, 1988) coincide with the appearance of the first descending 370 vessels to form the outer vascular plexus at the fovea between 25-26 weeks. The increase in metabolic 371 demand of the maturing neurons leads to the development of ‘physiological hypoxia’, which as noted 372 above, is the stimulus for the formation of the deeper vascularMANUSCRIPT bed (Chan-Ling, 1994, Chan-Ling et al., 373 1995a, Chan-Ling et al., 1995b, Provis et al., 1997). Figure 4F-I shows in a series of scanning electron 374 micrographs, the marked changes in cone packing and morphology through the gestational period. J-K 375 demonstrates the changes in morphology observable from 25 weeks gestation age (J), where the cones 376 appear short and rounded, to 5 months post-partum (K), where the cones and rods have elongated to a 377 point where they are comparable to the appearance of adult photoreceptors.

378 At 24 weeks GA, only the ganglion cells at the fovea and surrounding posterior retina, extending between 379 the raphe, fovea and para-optic nerve head regions have exited the mitotic cycle, representing 380 approximately half of the posterior retinal area (Provis et al., 1985, Sandercoe et al., 1999). At 26 weeks, 381 the superficial vascular plexus in the posterior pole is fully formed with the exception of the foveal 382 avascular zone. The deeperACCEPTED vascular plexus has not yet formed. A small margin at the very edge of retina 383 is avascular at this stage because it is thin enough to be adequately nourished by diffusion from the 384 choriocapillaris (Hughes et al., 2000). The FAZ is defined by the superficial vascular plexus at the 385 GCL/inner plexiform layer interface. As the inner retinal vasculature further develops, at 32 weeks GA, 386 perifoveal cones and those on the foveal rim begin to elongate; this is in contrast to the developmental lag 387 seen in the cones of the central fovea, where they are cuboidal in shape. The foveal depression continues 388 to deepen until 15 months after birth, due primarily to the migration of inner retinal cells, including

12 389 ganglion, amacrine and bipolar cells, towards the parafoveal region. Prior to birth, the FAZ is more than ACCEPTED MANUSCRIPT 390 1000 um in diameter, but it progressively narrows after birth due to migration of cones from the 391 periphery. During maturation, the foveolar cones develop both outer segment and basal axon processes, 392 and reduce in diameter and elongate to become tightly packed in the foveal region, enabling high acuity 393 vision. The arrangement of foveal cones is such that the basal axons of these cells form the structure 394 known as ‘Henle's fibre layer’ with the dendrites of the bipolar and ganglion cells.

395 The basal axons of the foveal cones are obliquely arranged in Henle's fibre layer (almost parallel to the 396 retinal surface, as compared to the perpendicular arrangement in extramacular retina) where they are 397 connected with their respective bipolar and ganglion cells. The ratio between foveal cones and their 398 bipolar and ganglion cells varies between cone types, with each foveal L and M cone presynaptic to one 399 inner and one outer midget bipolar cell, while S cones are presynaptic to one outer, but no inner midget 400 bipolar cell (Ahmad et al., 2003). The work of Ahmad and colleagues has demonstrated that for each 401 foveal cone pedicle, there is close to one outer diffuse bipolar cell (1:0.88), but only 0.40 inner diffuse 402 bipolar cells to each cone. The Muller cell and amacrine cell ratios are close to 1 (1.01 for each cell type 403 respectively), however the horizontal cell ratio is 1:0.83 for each foveal cone pedicle. The previously held 404 assumption that foveal cones are connected in a 1:1 fashion to both the bipolar and ganglion cells is not 405 necessarily the case, with Ahmad et.al demonstrating that there are 3.4 cone bipolar cells and 2.6 ganglion 406 cells per cone. 407 In light of the space limitations in the foveal region, where MANUSCRIPT only cones appear, the bipolar and ganglion 408 cells are located relatively distant from their corresponding photoreceptors, necessitating the unique and 409 delicate arrangement of this layer to interconnect them (Drasdo et al., 2007). By 45 months after birth, the 410 cones have decreased in diameter from 7.5 um to 2 um, allowing for an increased density of cones from 411 18 cones/100 um 2 at P7, to 45 cones/100 um 2 by adulthood (Yuodelis and Hendrickson, 1986).

412 Interactions between the glial, vascular and neuronal lineages are critical for normal foveal and vessel 413 development (Chan-Ling, 1994) . Because of this, conditions such as retinopathy of prematurity, where 414 normal retinal vascular development is altered can also have serious implications for foveal development 415 as shown by Maldonado and colleagues (Maldonado et al., 2011). In contrast to the adult fovea, they 416 demonstrate that several signs of immaturity are evident in premature neonates: a shallow foveal pit, a 417 persistence of inner retinalACCEPTED layers, and a thin photoreceptor layer that is thinnest at the foveal center as 418 well as the presence of cystoid macula edema in 58% of premature neonates, that appeared to affect inner 419 foveal maturation.

420 421 Neurosensory Retina in ROP

422 ROP onset occurs when the neurosensory retina is quite immature. The photoreceptors are the last cells to 423 diffentiate from the retinal neuroepithelium during human embryological development. While ROP is 13 424 characterized by vascular abnormalities, there is compelling evidence that the neurosensory retina and ACCEPTED MANUSCRIPT 425 also cortical deficiencies contribute to the visual outcome in this devastating disease (Dutton, 2013). ROP 426 onset in infants is coincident with rapid elongation of the rod photoreceptor outer segments. Utilizing a 427 combination of electroretinogram (ERG), psycophysical measures and retinal imaging we now know that 428 rod photoreceptor and post-receptor responses are significantly altered years after active ROP disease 429 have subsided (as reviewed by Hansen et al PIRER 2017). These alterations include persistent rod 430 dysfunction, evidence of compensatory remodeling of the inner retina neuronal circuitry as evidenced by 431 both full-field ERG and multifocal ERG. Further, both hypoxia and hyperoxia have been demonstrated to 432 be injurious to rods in young adult rats, (Wellard et al., 2005), thus both hypoxia and hyperoxia are likely 433 to also be injurious to rods during neonatal development with consequences on ROP and visual function.

434 435 Effects of premature birth on retinal development and function: Impact on extremely 436 premature infants

437 As the FAZ is demarcated by 28 weeks GA in normal development, differences in FAZ size and foveal 438 morphology are less pronounced in infants born after this time point. This is supported by studies that 439 demonstrate that foveal thickness is inversely correlated with gestational age (Wu et al., 2012b, Wagner- 440 Schuman et al., 2011, Chui et al., 2012, Akerblom et al., 2011) and that 28 weeks is a “critical gestational 441 age” in terms of fine foveal structure because preterm birth before 28 weeks of age is associated with a 442 failure of the inner retinal layers to migrate away from theMANUSCRIPT fovea, resulting in increased foveal thickness 443 compared to full-term controls (Wu et al., 2012b).

444 With the increased survival of extremely premature (EP; <28 weeks GA; <1000g birth weight) infants 445 and particularly those <26 weeks GA and <750g birth weight, the consequences of premature birth on 446 foveal vascular and neural development have increased significance. It has been shown that many 447 premature infants have significantly decreased visual function despite apparently normal retinal 448 vasculature (Dobson et al., 1994, CRYO-ROP, 2001, Fulton et al., 2001, Fulton et al., 2005, Barnaby et 449 al., 2007, Hammer et al., 2008). The most immature group of infants are 3 times more likely than term 450 infants to have poor visual acuity and other impairments such as strabismus, amblyopia and colour vision 451 defects indicative of damage to the macula and visual pathways. (Msall et al., 2004, O'Connor et al., 452 2007, Hou et al., 2011).ACCEPTED Recent follow-up data from the Extremely Preterm Infants in Sweden Study 453 (EXPRESS; infants <27 weeks’ gestation) at 6.5 years of age showed that 37.9% had ophthalmologic 454 abnormalities compared with 6.2% of matched controls, and 4.8% had visual impairment (<20/60 vision) 455 and 2.1% were blind (Hellgren 2016). OCT undertaken in a subset of the cohort at the same age showed 456 central macula thickness to be significantly increased in the extremely premature group (Molnar 2017).

457

14 458 The delayed formation of the vasculature around the fovea in preterm infants causes a later formation of ACCEPTED MANUSCRIPT 459 the FAZ (i.e. at 30 weeks GA rather than at 25-26 weeks). This lag coincides with the time when anti- 460 angiogenic and guidance factors that mediate formation of the FAZ are in decline, leading to capillary 461 overgrowth of the fovea. Numerous studies have shown that prematurity is directly related to a shallower 462 foveal depression and a thicker central fovea. The persistence of the GCL and INL cells in the central 463 fovea suggests that the mismatch between vascular to neuronal development that occurs in prematurity 464 lead to a reduction in the migration of inner retinal neurons out of the fovea. In addition, published studies 465 suggest that premature birth is linked to changes in both rod and cone photoreceptor response, and other 466 ocular anomalies including impaired ocular growth; increased incidence and magnitude of refractive 467 error, particularly myopia; acuity deficits; field defects; and strabismus. (Reviewed in (Fulton et al., 468 2009). School age children who were premature have been shown to have reduced photoreceptor function

469 (Akerblom et al., 2014). Moreover, as VEGF 165 and PEDF expression as well as the Eph receptor 470 pathways are directly modulated by oxygen, exposure of the neonate to higher levels of oxygen than in 471 utero would be predicted to affect the development of the fovea, with significant impact on their visual 472 acuity.

473 Choroidal involution as a component of ROP

474 ROP has always been considered a disease of the inner retinal vasculature but there is now both animal 475 and human evidence that the choroid is actively involved in ROP, contributing to photoreceptor damage. 476 Choroidal involution, confined to the central areas of the MANUSCRIPT retina has been demonstrated as a key 477 component of 50/10 oxygen-induced retinopathy in the rat (rat model of ROP), leading to degeneration of 478 photoreceptors in the central area of the retina, and corresponding deficits in retinal function as measured 479 by multifocal electroretinography, which measures changes in discrete areas of the retina (Shao et al., 480 2011). Shao and colleagues demonstrated that in a hyperoxic state (as seen in animal models of OIR and

481 ROP), the high levels of prostaglandin D 2 (PGD 2) which control autoregulation of the choroid in

482 immature animals are converted into the electrophile 15d-PGJ 2, which is cytotoxic to endothelial cells of 483 the developing choroid. The complex mechanisms involved in choroidal damage in ROP have recently 484 been reviewed by Rivera and colleagues (Rivera et al., 2016). ERG research studies on infants with ROP 485 have demonstrated the compromise of photoreceptor function (Akerblom et al., 2014, Fulton et al., 2009). 486 Infants with ROP mayACCEPTED have anomalies in colour discrimination and degeneration of the RPE; while older 487 subjects (5-23 years) with severe ROP can show deficits in rod responses not seen in infants with mild or 488 no ROP raising the possibility of progressive compromise of rod function, related to dark adaptation, over 489 time (Fulton et al., 2009). Outer retinal damage could explain some of those cases of regressed ROP with 490 subnormal visual acuity, despite apparently normal fundi (Molnar et al 2017).

491

15 492 Optical Coherence TomographyA (OCT)CCEPTED and ROPMANUSCRIPT 493 OCT scanning is now a standard tool for the management of adult retina cases particularly those 494 involving leaky vasculature such as diabetic retinopathy and age-related . Pediatric 495 scanning has been severely limited because, until recently, commercially available OCT scanners have 496 been table mounted. With the development of hand held OCT scanners, progress has been rapid (see 497 excellent reviews by Maldonado & Toth (Maldonado and Toth, 2013) and Lee and colleagues (Lee et al., 498 2016). There are significant differences in the retinal layers observed in OCT scans between infants and 499 adults. The dynamic processes involved in the development of the fovea which include centrifugal inner 500 layer migration (to form the foveal pit) and centripetal outer retinal growth involving thicker inner and 501 outer segment growth as well as an increase in the cone density have now been documented in vivo in 502 infants from 31 weeks PMA upwards (Maldonado et al., 2010) and the changes correlated with histology 503 (Vajzovic et al., 2012).

504 New findings, unsuspected on clinical exam, have been reported in premature infants using OCT. 505 Epiretinal membranes, otherwise rare in childhood, have been reported in a third of premature infants and 506 may contribute to abnormal foveal contour formation (Lee et al., 2011a). The most significant OCT 507 finding has been that cystoid macular edema (involving both intracellular and extracellular components) 508 occurs very commonly in premature infants with ROP, over 50% in one series (Maldonado et al., 2012); 509 21-29% in another series (Gursoy et al., 2016) and has several different morphologic characteristics 510 (Maldonado et al., 2012, Vinekar et al., 2011). Some full-term MANUSCRIPT infants show subfoveal fluid (Cabrera et 511 al., 2012) but we are not aware that CME has been reported in full-term infants. Increased severity of 512 CME was correlated with an increased likelihood for laser treatment (Maldonado et al., 2012) and may 513 correlate with ultimate visual outcome (Vinekar et al., 2015b, Rothman et al., 2015b) and is also 514 associated with poorer language and motor skills at 18-24 months (Rothman et al., 2015a). This clinical 515 outcome correlates well with experimental models of ROP (kitten model in this case), where a transient 516 breakdown of the blood retinal barrier is a consistent feature of the vaso-proliferative phase of ROP 517 (Chan-Ling and Stone, 1992). For this aspect of infant ROP pathogenesisis, therapeutic options based on 518 anti-VEGF agents could provide dual benefits due to the fact that in addition to blocking vascular 519 endothelial proliferation, VEGF has previously been identified as ‘vascular permeability factor’(Senger et 520 al., 1983). As a consequence,ACCEPTED anti-VEGF agents may reduce pathological neovascularization and vascular 521 leakage in ROP, however the effects of anti-VEGF agents on vascular leakage remains to be studied 522 clinically.

523 524 Unique susceptibility of posterior ROP: Basis in vasculogenic origin of posterior vessels

525 Earlier work has shown that the human retina is vascularized by a combination of vasculogenesis and 526 angiogenesis (Hughes et al., 2000). Between 12-15 weeks GA large numbers of vascular precursor cells 16 527 stream in superficially from the optic nerve head where they form solid cords of vascular precursor cells ACCEPTED MANUSCRIPT 528 before becoming patent (Chan-Ling et al., 2004a). At 15 weeks GA, patent radial vessels with few 529 interconnecting segments emerge from the optic disc in the GCL to form a primordial vascular bed, and 530 this primordial vascular bed formed by vasculogenesis only extends to the central one third of the 531 posterior human retina (Hughes et al., 2000, Flynn and Chan-Ling, 2006). In contrast, angiogenesis is 532 responsible for vascular formation for the remaining superficial vasculature. In its most severe form, ROP 533 is located in the posterior retina and affects the smallest, most premature infants and posterior aggressive 534 ROP shows a similar topography and distribution to the primordial vessels formed by vasculogenesis. We 535 hypothesized that the clinical presentation and therapeutic outcome observed in Zone 1 vs. Zone 2 ROP 536 reflected the difference in topography, distribution and age of appearance between retinal vessels formed 537 by the vasculogenic versus the angiogenic process (Flynn and Chan-Ling, 2006, Hughes et al., 2000).

538 Zone 1 ROP, which is less sensitive to laser/cryotherapy and demonstrates poorer anatomic and visual 539 outcomes (Good and Early Treatment for Retinopathy of Prematurity Cooperative, 2004), (Wani et al., 540 2013, Spandau et al., 2013, Gunn et al., 2014, Gole et al., 2011), is correlated with vessel development by 541 both vasculogenesis and angiogenesis, but vasculogenesis is only responsible for the formation of the 542 primordial vessels in a small region surrounding the optic nerve head (Hughes et al., 2000) . Even within 543 the central one third of the retina surrounding the optic disc, angiogenesis contributes to increasing 544 vascular density once the primordial vessels are formed. In contrast, the remaining peripheral and deep 545 retinal vessels are formed totally by angiogenesis via HIF-1 MANUSCRIPTα induced VEGF expression, so they are 546 expected to be responsive to anti-VEGF 165 treatment options. This is supported by the fact that the 547 peripheral ROP responds well to laser ablation (ETROP, 2003) which works by destroying retinal tissue, 548 reducing need for oxygen and also increasing diffusion from the choroid. When the current international 549 classification was developed, knowledge of the processes of human retinal vascular development was 550 incomplete.

551 Vasculogenesis is seen in the human retina at 14 to 15 weeks GA, despite an absence of detectable VEGF 552 mRNA in the retina at this time (Provis et al., 1997) so vasculogenesis appears to take place largely

553 independent of VEGF 165 in the retina. Further, vasculogenesis has been reported to be insensitive to 554 modulation of VEGF in knockout mice. In these animals, in which not only paracrine but also autocrine 555 VEGF production is lost,ACCEPTED vessels still form by vasculogenesis although they are highly abnormal 556 (Carmeliet et al., 1996) so there may be a certain amount of overlap between vasculogenesis and 557 angiogenesis in the course of development. An additional possible driving force for Zone 1 ROP could be 558 the later maturation of the parafoveal rod photoreceptors, where elongation of these rod outer segments 559 increase metabolic activity and oxygen consumption (Hendrickson and Drucker, 1992) this in turn 560 upregulates VEGF expression mediated by increased expression of HIF-1α.

17 561 As extremely premature infants are now surviving as young as 22 weeks GA, at a time when ACCEPTED MANUSCRIPT 562 vasculogenesis may still be taking place, and due to our incomplete understanding of the cellular and 563 molecular mechanisms of vasculogenesis, other approaches to therapy for ROP need to be considered and 564 thoroughly investigated especially for Zone 1 posterior ROP. 565 566 An additional factor that could contribute to the unique susceptibility of the posterior pole to ROP 567 damage is the concurrent maturation of the posterior neurosensory retina with birth of the extreme 568 preterm infant (less than 26 weeks GA), since both cone and rod systems are still in development between 569 26 to 36 weeks GA (Hendrickson and Drucker, 1992). At 26 weeks GA, in the parafoveal area, 570 photoreceptors have a rudimentary inner segment, while by 36 weeks GA, short outer segments are 571 present on both rods and cones (Hendrickson and Drucker, 1992). Importantly, the studies of 572 Hendrickson and Drucker demonstrate that the rods in the peripheral retina develop earlier than those in 573 central retina, while the parafoveal photoreceptors develop well in advance of the foveal cones thus 574 contributing to the ‘physiological hypoxia’ required for driving HIF1-α/VEGF induced angiogensesis 575 even very early in human embryological development. Thus, premature birth could have significant 576 impact on their normal differentiation both structurally and as reflected in their function via ERG 577 determination (Hansen et al., 2017, Fulton et al., 2009, Fulton et al., 2005, Fulton et al., 2001). 578 579 ROP develops in two phases MANUSCRIPT 580 Human ROP develops in two phases (Smith, 2008, Chan -Ling et al., 1992). The first phase involves 581 cessation of vessel growth and some loss of already formed vessels driven by higher than physiologic 582 intrauterine blood oxygen levels and loss of maternal and placental growth factors consequent on 583 premature birth. This higher than normal intrauterine arterial oxygen tension results in the flooding of the 584 retina with oxygen from a poorly auto-regulated choroidal vasculature, thus negating the ‘physiological 585 hypoxia’ stimulus required for normal retinal vascularization. The second phase occurs when the 586 avascular retina becomes more metabolically active and the relative hypoxia upregulates VEGF; 587 pathologic neovascularization with the potential for retinal detachment then follows. In animal models 588 (oxygen-induced retinopathy) the first (hyperoxic) phase results in widespread vaso-obliteration followed 589 on return to room air by vasoproliferation. Human ROP and oxygen-induced retinopathy are therefore 590 qualitatively different ACCEPTEDdisorders in the hyperoxic phase and retinal detachment does not occur in most 591 animal models.

18 592 ACCEPTED MANUSCRIPT 593 Initiating event in pathogenesis of ROP - delayed vascularisation relative to neuronal 594 maturation (Phase 1 ROP) 595 Premature infants are exposed to higher oxygen tension after birth than that experienced in utero where 596 oxygen tension is around 30-35mmHg (Rudolph, 1970). In the face of this increased oxygen tension, the 597 choroidal vasculature, with its limited ability to auto-regulate (for review, see (Kur et al., 2012), floods 598 the retina with oxygen. At this point the developing retina is only partially vascularized and the hyperoxia

599 leads to a down-regulation of VEGF 165 expression by astrocytes, Müller cells (Stone et al., 1995, Stone et 600 al., 1996), pericytes (Darland et al., 2003) and RPE (Shima et al., 1995) via hypoxia induced factor-alpha 601 (HIF-1α). The loss of this vascular endothelial proliferation and survival factor results in apoptosis and 602 vascular regression and is combined with cessation of angiogenesis (Chan-Ling et al., 1992, Alon et al., 603 1995). As a consequence, during the period when the neonate is exposed to high oxygen, the rate of 604 vascular formation is delayed relative to the maturation of the neuronal elements, which proceeds 605 normally as it is unimpeded by oxygen tension in the retina (Chan-Ling et al., 1995b, Hartnett and Penn, 606 2012). Because of this disconnect, the infant will have varying degrees of non-vascularized but 607 metabolically active retina, thus providing a massive hypoxic stimulus for neovascularization seen in 608 ROP (Chan-Ling et al., 1995b). Further contributing to this injury is the damaging effect of hyperoxia 609 during phase 1 of ROP and hypoxia in Phase 2 to the neurosensory retina (Wellard et al., 2005). Figure 5 610 is a schematic representation of the underlying pathogenicMANUSCRIPT mechanisms underlying the various stages of 611 ROP (Reprinted from (Hartnett and Penn, 2012)

612 The clinical fundus picture of a neonate recently removed from a high oxygen environment shows limited 613 retinal vascular extent and density. This predisposes to significant retinal hypoxia, because the attenuated 614 retinal vascularization is accompanied by degeneration of astrocytes and the subsequent failure of the 615 blood retinal barrier, seen in the neovascular phase (Phase 2) of ROP (Chan-Ling and Stone, 1992, Smith, 616 2008, Hartnett and Penn, 2012). 617 618 Phase 2 ROP: Hypoxia-induced pathological neovascularization 619 The later, sight-threatening stages of ROP develop due to a mismatch between the extent of retinal 620 vascularization and theACCEPTED metabolic demands of the maturing neurons during the period of hyperoxic 621 exposure. The metabolic needs of the developing retina are no longer being met by the exogenously 622 supplied oxygen and the retina is now relatively hypoxic due to delayed vascularization relative to

623 neuronal maturation, which results in upregulated VEGF 165 expression. Vascular growth resumes; 624 however, the blood vessels are abnormal and leaky and can grow into the vitreous, leading to sight- 625 threatening tractional retinal detachment. The pathological vessels that arise during this second phase of

626 ROP that grow under a marked upregulation of VEGF 165 (Roberto et al., 1996, Stone et al., 1996), lack

19 627 astrocytic ensheathment and an intact blood retinal barrier (BRB) (Chan-Ling and Stone, 1992). If normal ACCEPTED MANUSCRIPT 628 formation of the retinal vasculature can take place during the hyperoxic phase of the disease, then there 629 will be no peripheral retinal ischemia to drive hypoxia-induced vaso-proliferation during the second 630 proliferative phase of ROP, thus the neovascular stage of the disease can theoretically be prevented. 631 (Chan-Ling et al., 1995b, Hartnett and Penn, 2012). Our current studies utilizing dark rearing to protect 632 the forming retinal vasculature during the hyperoxic phase of disease were undertaken to provide 633 evidence in support of this hypothesis. (Chan-Ling et al., 2016, Chan-Ling et al., 2014a, Chan-Ling et al., 634 2014b). 635 636 Limitations of current management strategies for ROP, including anti-VEGF: 637 Current treatment strategies for ROP focus on advanced disease when retinal vessels are already 638 proliferating uncontrollably (Phase 2), and include cryotherapy, laser therapy and most recently 639 intraocular injection of anti-VEGF constructs. These treatments are expensive, invasive, and can be 640 difficult to access in the various at-risk populations. Further, VEGF is a neurotrophic factor (Falk et al., 641 2010, Genetos et al., 2010) as well as an important pro-angiogenic factor (Nguyen et al., 2012, Reinders 642 et al., 2003). Several studies show that following injection into the eye, anti-VEGF agents enter the 643 systemic circulation (Darlow et al., 2013a, Hard and Hellstrom, 2011, Hoerster et al., 2013, Miyake et al., 644 2010) and there are reports of long term ROP recurrence with anti-VEGF therapies. Thus, there is a risk 645 that anti-VEGF agents injected into the neonatal eye canMANUSCRIPT compromise development of other CNS 646 structures and/or organs. Furthermore, newborn rats injected with a single systemic dose of a VEGF 647 receptor inhibitor develop pulmonary hypertension & show maldevelopment of the lungs, which persists 648 into adulthood (Le Cras et al., 2002). Hence it is possible that early neonatal use of anti-VEGF agents, 649 when the lungs are underdeveloped, might exacerbate developmental disorders associated with 650 prematurity. 651 At present, there are no therapies in clinical use that address the key pathogenic events in the 652 hyperoxic/vaso-obliterative Phase 1 of ROP . We proposed and showed that moderation of retinal 653 hyperoxia in Phase 1 reduced oxygen-induced vaso-obliteration of the retinal vessels, resulting in a

654 smaller hypoxic stimulus ( PO2) in the transition to room air in the 50/10 oscillating model (Penn et al., 655 1994) of rat OIR (Chan-Ling et al., 2016, Chan-Ling et al., 2014a, Chan-Ling et al., 2014b). Moderation 656 of retinal hyperoxia viaACCEPTED ‘Dark Rearing’ in Phase 1 of ROP is the mechanism underlying this non-invasive 657 treatment. 658 659 The Rationale for Dark-Rearing (DR), a non-invasive therapy for the treatment of ROP

660 Limiting O 2 administration to preterm infants has long been known to significantly increase the incidence 661 of brain damage (Low et al., 1993, Pryds, 1994) and more recently lower oxygen saturation targets have 662 been demonstrated to increase mortality (see section on Recent trials of oxygen saturation targeting). So 20 663 while attempts to limit O administration to preterm infants may benefit the retina, it has severe 2 ACCEPTED MANUSCRIPT 664 consequences for lung function and brain development. An alternative approach is to lower O 2 levels in

665 the retina by increasing local O 2 consumption of the photoreceptors while infants are in the high oxygen 666 environment. Rod and cone photoreceptors are considered by many to be the most metabolically active 667 cells in the body. Photoreceptors have the highest level of oxidative metabolism in the retina, as 668 demonstrated through the use of oxygen micro-electrodes (Alder et al., 1990, Alder et al., 1983, 669 Linsenmeier, 1986, Yu et al., 1994) and by cytochrome oxidase histochemistry (Kageyama and Wong- 670 Riley, 1984). 671 672 Photoreceptors are more metabolically active in the dark than in the light. Activation of rhodopsin by 673 light activates a G-protein-coupled receptor that triggers a second messenger cascade, causing cGMP- 674 gated ion channels to close, resulting in hyperpolarization of the cell. In the dark, the cGMP-gated 675 channels are open, allowing the passage of ions into the cell (the “dark current”) and depolarization. 676 Restoration of membrane ion balance by active ion pumps means the energy demands of the

677 photoreceptors in the dark are much higher than in light. In cat retina, O 2 consumption by PRs is 50% 678 higher in the dark as compared to light (Linsenmeier, 1986) 679 680 The earliest published report of detectable rod outer segments in the rat eye is postnatal day 8 (P8) (Ratto 681 et al., 1991), however, as this process is a continuum spanningMANUSCRIPT a number of days, it is likely that the 682 earliest primordial outer segments are present prior than P8. We have shown that dark rearing rats in 683 normal oxygen conditions resulted in higher vascular density in neonatal rat retina from birth to postnatal 684 day 7 (Chan-Ling et al., 2016, Chan-Ling et al., 2014a, Chan-Ling et al., 2014b), during the first postnatal 685 week while vessels are actively forming (Stone et al., 1995). Our findings of increased retinal vascular 686 density in early retinal development in dark-reared animals support those of Rao and collegaues (Rao et 687 al., 2013), who demonstrated an increased vascular density in mice that were dark-reared in normal 688 oxygen conditions before birth to P8. Both our work, and the work of Rao has demonstrated that vascular 689 density can be increased in darkness, even when light-sensitive rod outer segments are not fully formed. 690 Newborn mice are light-sensitive at a time where rods and cones are not fully formed (Hattar et al., 691 2002), a phenomenon that has been attributed to the melanopsin-containing retinal ganglion cells, which 692 are intrinsically photosensitive.ACCEPTED Rao and colleagues have demonstrated that mice deficient in the Opn4 693 gene which encodes melanopsin, display increases in retinal vascular density at P8 that mimic those seen 694 in dark-reared wild-type animals, and that the melanopsin photo transduction pathway also contribute to 695 regulation of retinal vascular formation and vascular pruning/remodeling. 696 697 Taken together, we suggest that the melanopsin dependent increases in vascular density that can be 698 effected by dark-rearing in early postnatal development, coupled with the increases in local retinal

21 699 hypoxia caused by the continuous depolarization of the light-sensitive photroreceptors in the dark, can be ACCEPTED MANUSCRIPT 700 combined to reduce the burden of vascular damage caused by hyperoxia in the oxygen-induced 701 retinopathy model. 702 703 Further, we were able to demonstrate that dark rearing from birth to postnatal day 14 reduced 704 significantly the frequency of pre-retinal neovascularisation at postnatal day 18, as well as the severity of 705 retinal neovascularisation during the hypoxic Phase 2 (postnatal day 14-18) of rat OIR. Further, we

706 showed that dark rearing resulted in changes in expression of key angiogenic genes HIF1 α, VEGF 164 , as 707 well as the early response transcription factor AP1/Jun, an indicator of oxidative and retinal stress (Chan- 708 Ling et al., 2016, Chan-Ling et al., 2014a, Chan-Ling et al., 2014b). A recent review by Hansen and 709 colleagues (Hansen et al., 2017) has provided a systematic review of the neural retina of ROP and arrived 710 at a similar proposal. These authors extended our rationale for dark rearing during the hyperoxic Phase 1 711 of ROP to also include a recommendation for continuous light in Phase 2 ROP, at relatively low levels to 712 dampen the dark current, diminishing the angiogenic signal and protecting the retina from both hypoxic 713 damage and neovascularization . However, empirical evidence in support of this hypothesis for 714 continuous light in Phase 2 ROP is still lacking. 715 716 Scientific rationale for distinct pathogenetic mechanisms underlying zone 1 & zone 2 ROP 717 Flynn and Chan-Ling (Flynn and Chan-Ling, 2006) presentedMANUSCRIPT a hypothesis that ROP may be the result of 718 two distinct clinical-pathologic processes which resemble each other in certain respects but differ 719 substantially in their location in the retina, their morphology, and their need for different treatment. Table 720 1 summarizes the differences between Zone 1 and Zone 2 ROP, as observed with indirect 721 ophthalmoscopy. 722 723 Flynn and Chan-Ling (Flynn and Chan-Ling, 2006) provided both anatomic and clinical evidence in 724 support of the hypothesis that the more severe ROP seen in posterior disease can be attributed to two 725 distinct pathologic pathways of disease. Our hypothesis is that in Zone 1, retinal vascularization occurs 726 through vasculogenesis while in Zone 2 (and Zone 3) the retina is vascularized through angiogenesis. Our 727 studies of normal human fetal retinal vascularization have shown that vasculogenesis, which involves 728 previous invasion of vascularACCEPTED precursor cells (sometimes also known as angioblasts), is responsible for 729 the formation of the primordial vessels that are seen at the central posterior region of the human retina, 730 whereas angiogenesis formation through budding from existing blood vessels are responsible for the 731 formation of the remaining retinal blood vessels in the human retina. As a consequence, the course of 732 ROP in these locations might also follow different cellular and molecular cues. In support of this 733 conclusion, we have argued that Zone 1 ROP is clinically less responsive to ablative therapy than is Zone 734 2 ROP (Flynn and Chan-Ling, 2006). According to the ‘physiologic hypoxia’ model of retinal

22 735 vascularization (Chan-Ling et al., 1995a) the reason that ablative therapy is effective is because it results ACCEPTED MANUSCRIPT 736 in the destruction of the retinal tissue that would normally act as a metabolic sump - thus taking away the

737 hypoxia-induced stimulus to produce VEGF 165 . Our hypothesis in 2006, argued that Zone 1 ROP is less 738 responsive to laser therapy because the yet to be identified molecular cue that induces vasculogenesis in

739 the posterior pole appears to be VEGF 165 -independent, however, it predated studies of anti-VEGF therapy

740 in infants. An alternative interpretation is that a second source of VEGF 165 , such as vitreal macrophages 741 as reported by Naug and associates (Naug et al., 2000) is responsible for the reduced effect of laser 742 therapy in Zone 1 ROP. 743 744

Clinical Presentation of Zone 1 vs Zone 2 Retinopathy of Prematurity

Zone 1 ROP (Vasculogenic) ROP Characteristic Zone 2 ROP (Angiogenic)

Posterior Pole Location Outside Posterior Pole

Always 12 o'clock hours Extent Variable extent

Normal arcades absent Vascular Pattern Normal arcades preserved

Absent Ridge Present

Never present "Wedge" Often present Precapillary tangles Microvascular Pattern MANUSCRIPT Layered Capillaries Never? Regression Common Response to Laser Poor (55.2% unfavourable) 50% reduction in unfavourable outcomes Therapy 745 Table 1: Modified from Flynn & Chan-Ling, Am J Path 2006 746 747 VEGF165 subserves multiple functions: angiogenic growth factor, vascular 748 permeability factor and neuronal survival factor. 749 Background 750 Almost from the earliestACCEPTED clinical description of ROP, numerous researchers had postulated the existence 751 of an inducible factor that could stimulate angiogenesis in various developmental and disease states 752 (Michaelson et al., 1954, Henkind and De Oliveira, 1967, Shakib et al., 1968, Folkman, 1971). In 1983, 753 the work of Senger and colleagues (Senger et al., 1983) identified a “vascular permeability factor” in 754 tumor ascites fluid from various experimental animals, which when applied to cell cultures in vitro 755 demonstrated similar increases in microvascular permeability observed in vivo . Subsequently, Ferrara and 756 Henzel demonstrated the role of this vascular permeability factor in angiogenesis (Ferrara and Henzel,

23 757 1989) and suggested, due to its specific target effects on endothelial cells, the factor be named “vascular ACCEPTED MANUSCRIPT 758 endothelial growth factor”. 759 760 The work of Shweiki, Keshet and colleagues (Shweiki et al., 1992), Chan-Ling (Chan-Ling et al., 1995a), 761 Stone (Stone et al., 1995), and Adamis, Miller, Shima & D’Amore (Adamis et al., 1993, Miller et al.,

762 1994, Shima et al., 1996) and associated laboratories in the late 1980s and early 1990s led to VEGF 165 763 being identified as the prime molecular mechanism for the angiogenic growth and development of the 764 retinal vasculature. With the key molecular cue in angiogenesis now identified, specific antagonists and

765 molecular blockers of VEGF 165 were synthesised, and eventually taken to market for the treatment of 766 various angiogenic dependent diseases, ranging from colon cancer to age-related macular degeneration, 767 proliferative diabetic retinopathy and ROP (Reviewed in (Amadio et al., 2016)). 768 769 VEGF Biology 770 The vascular endothelial growth factor (VEGF) family currently consists of VEGF-A, -B, -C, -D, -E and 771 placental growth factor (PlGF) . (Kowanetz and Ferrara, 2006). Among them, VEGF-A is considered the

772 key angiogenic factor with various splicing variants such as VEGF-A125 , -A145 , -A165 , -A183 , -A189 , and -

773 A206 (Robinson and Stringer, 2001). The generic use of the term “VEGF” most often refers to VEGF-A. 774 VEGF-A is indispensable during embryonic vascular development, and the loss of a single VEGF-A 775 allele results in embryonic lethality with defective vasculatureMANUSCRIPT in a mouse model (Carmeliet et al., 1996). 776 Hypoxia strongly up-regulates VEGF-A expression via increased levels of hypoxia inducible factor (HIF- 777 1α) (Safran and Kaelin, 2003). Under normal conditions, HIF is ubiquitinated and degenerated by binding 778 to von Hippel-Lindau (VHL) protein, but, in hypoxic conditions, HIF cannot bind to VHL, resulting in 779 increased levels of HIF. Increased HIF binds to the DNA chain at the HIF-1 binding site, the hypoxia 780 response element (HRE), and enhances many pro-angiogenic gene transcriptional activities, including the 781 VEGF-A gene (Safran and Kaelin, 2003). 782

783 Beyond the vasculature, VEGF 165 is essential for maintenance and development of the blood-brain barrier 784 (BBB), vascular endothelial cell proliferation, survival & migration, the proliferation, survival and 785 migration of neurons andACCEPTED neural stem cells, proliferation of astrocytes, and proliferation and migration of 786 microglia. (Zachary, 2005). VEGF 165 is also important for the development of the kidney, (Simon et al., 787 1995) lung (Voelkel et al., 2006) and brain (Sentilhes et al., 2010). 788 789 Control and expression of VEGF and associated genes via HIF-1 790 The expression of VEGF is controlled via hypoxia-inducible factor 1 (HIF-1). HIF-1 has been identified 791 in, and is highly conserved across all metazoan life (Kaelin and Ratcliffe, 2008) from C.elegans 792 (nematode worm) (Jiang et al., 2001) to Homo sapiens (Semenza and Wang, 1992), highlighting the key 24 793 role that HIF-1 has played in adaptation to the conditions in which life exists on Earth. HIF-1 is a ACCEPTED MANUSCRIPT 794 heterodimeric protein consisting of two subunits, HIF-1α and HIF-1β. HIF-1β is constitutively expressed; 795 however, the expression of HIF-1α is regulated by the concentration of oxygen within the cell. HIF-1α 796 rapidly accumulates under hypoxic conditions, and is then rapidly degraded via ubiquitination by von 797 Hippel-Lindau (VHL) protein on return to normal oxygen conditions. This process has been demonstrated 798 to take less than one minute, with no other protein having a shorter cellular half-life (Yu et al., 1998). The 799 rapid expression and degradation of HIF-1α make it key to all cellular responses to changes in oxygen, 800 and core to the expression of genes and subsequent transcription of proteins involved in vascular 801 development making HIF-1α innately involved in the pathogenesis of ROP. 802 803 HIF-1α is the primary transcriptional regulator of hypoxia related genes, and activates more than 70 genes 804 that are crucial in vascular development, angiogenesis, proliferation, and cell survival (Semenza, 2003, 805 Semenza, 2004). Of these, in terms of retinal vascular development, the most important is VEGF. HIF- 806 1αis also the master transcriptional regulator at a cellular level for numerous other genes that have a 807 role in ROP, including erythropoitein (EPO) (Smith, 2008), angiopoietin 1 & 2 (Sato et al., 2011) and 808 platelet-derived growth factor subunit B (PDGFB) (Wilkinson-Berka et al., 2004). 809 810 Conversely, the presence of the mainly placental derived growth factor, insulin-like growth factor 1 (IGF- 811 1), is essential for the regulation of HIF-1α protein synthesis;MANUSCRIPT removing the neonate from the intrauterine 812 environment can have potentially serious effects on HIF-1 α expression, fetal development as a whole, and 813 ultimately, normal retinal vascular development (Smith, 2008). Early removal of the neonate from normal 814 intrauterine conditions, leading to a drop in the serum levels of IGF-1 has been observed in humans, as 815 well as in the mouse model of ROP (Hellstrom et al., 2001, Lofqvist et al., 2006a, Smith, 2004). 816 817 VEGF Signaling 818 The VEGF signalling pathway is well understood, with three major tyrosine kinase receptors expressed 819 by vascular endothelial cells, these being VEGFR2, VEGFR1 and VEGFR3 (Ferrara et al., 2003). 820 VEGFR2 is the major signalling receptor for VEGF-A and is the most critical receptor in terms of 821 angiogenesis, hence approaches which target VEGF-A constitute the majority of anti-VEGF therapeutics. 822 VEGF-A and PlacentalACCEPTED Growth Factor (PlGF) also bind to VEGFR1. VEGFC and D bind to VEGFR3, 823 which is the key regulator of lymphangiogenic activity. 824 825 When bound to VEGFR2, VEGF-A triggers a molecular cascade of events, the majority of which effect 826 the vascular endothelium. Some of these processes are effected immediately (in a period of seconds or 827 minutes), while others that are delayed in their onset (in the region of hours and days) (reviewed in (Nagy 828 et al., 2007)). Immediate in vivo effects of VEGF-A signalling include increases in vascular dilatation and 25 829 vascular permeability, followed by a delayed onset of downstream gene expression, angiogenesis and the ACCEPTED MANUSCRIPT 830 activation and mobilization of endothelial precursor cells. 831 832 VEGF and pathogenesis of ROP 833 Normal retinal vascular development relies on localized hypoxia at the developing vascular front (Chan- 834 Ling et al., 1995a). This localized ‘physiological hypoxia’ (Chan-Ling, 1994) induces the expression of

835 VEGF 165 by astrocytes and Muller glia and the consequent growth of blood vessels, which then relieves

836 the hypoxia, and reduces the local expression of VEGF 165 . This localised ‘wave-like’ hypoxic gradient 837 drives the formation of retinal vessels towards the avascular peripheral retina until the retina is fully 838 vascularized (Chan-Ling et al., 1995a (Smith, 2004, Gerhardt et al., 2003, Stone et al., 1995). 839 840 The delivery of supplemental oxygen to premature infants (Phase 1 ROP) supresses the normal levels of 841 hypoxia, leading to a delay in retinal vessel growth, and can obliterate the nascent vasculature through 842 apoptosis of existing vascular endothelial cells (Chan-Ling et al., 1992, Hughes et al., 2000) highlighting

843 the essential role of VEGF 165 as a vasotrophic factor in conjunction with its angiogenic role (Alon et al., 844 1995, Pierce et al., 1996). 845

846 Fluctuating fraction of inspired oxygen (FiO 2) has been shown to be a risk factor in human ROP 847 (McColm et al., 2004), further accentuating the mismatchMANUSCRIPT between the developing central vessels and the 848 metabolic needs of the avascular peripheral neural retina. 849 850 Hellgren and colleagues (Hellgren et al., 2016) measured local VEGF concentrations in the vitreous and 851 subretinal fluid of infants undergoing treatment for ROP, and found that these were elevated compared to 852 infants undergoing for cataract or acute retinal detachment. Further, they also observed that 853 infants requiring treatment for ROP have significantly higher serum levels of VEGF at 34, 35 and 36 854 weeks GA (the typical time of onset of pathological neovascularization) than infants without ROP. Serum 855 levels of VEGF are reduced after both laser and intravitreal anti-VEGF treatment, however laser 856 treatment leads to a less pronounced reduction in circulating VEGF as compared to intravitreal anti- 857 VEGF treatment (Kong et al., 2015a). These findings support the need for caution when considering anti- 858 VEGF therapies for ROP,ACCEPTED as they demonstrate the distinction between laser treatment with side effects 859 largely restricted to the eye, whereas anti-VEGF treatment could have undesirable systemic side effects. 860 861 VEGF and neuroprotection in the retina 862 VEGF has neurotrophic and neuroprotective effects on neuronal and glial cells in culture and in vivo, and 863 can stimulate the proliferation and survival of neural stem cells. Saint-Geniez and colleagues (Saint- 864 Geniez et al., 2008) demonstrated that after 14 days of VEGF neutralization, a significant increase of 26 865 apoptotic cells in the inner & outer nuclear layers was seen in the mouse retina. Further, addition of ACCEPTED MANUSCRIPT 866 exogenous VEGF to freshly isolated photoreceptor cells and outer-nuclear-layer explants demonstrated a 867 neuroprotective role for VEGF, while siRNA-based suppression of VEGF expression in a Müller cell line 868 supported an autocrine role in Müller cell survival. These findings indicate that intravitreal anti-VEGF 869 therapies should be used with caution in the treatment of ROP in premature infants until long-term studies 870 have been completed on their ocular and systemic side effects. 871 872 Ford and colleagues also have demonstrated that RPE derived VEGF-A plays a role in the maintenance of 873 the choriocapillaris (Ford et al., 2011), and that VEGF plays a vital role in ciliary body homeostasis (Ford 874 et al., 2012), where systemic neutralisation of VEGF led to thinning of non-pigmented epithelium, 875 impaired ciliary body function, and a decrease in intra-ocular pressure (IOP). Considering that the non- 876 pigmented epithelium produces aqueous humor, it is possible that intravitreal anti-VEGF therapy may 877 have off target effects on lens transparency as well as IOP. It should be noted however, that these studies 878 were undertaken using a complete VEGF blockade on a continuing basis, and that the blockade effected 879 by current anti-VEGF therapy, for example bevacizumab or ranibizumab, is not total. 880 881 Role of VEGF in maintaining vessel stability in early postnatal period 882 The role of VEGF as a survival factor for the nascent endothelium has been demonstrated by the work of 883 Benjamin and colleagues in the rat (Benjamin et al., 1998),MANUSCRIPT where they elucidated a VEGF dependent 884 “plasticity window” in the developing retinal vasculature. Hyperoxia downregulates VEGF expression in 885 Phase 1 ROP, leading to reduced angiogenesis and excessive capillary regression in the retina. In a rat 886 model of ROP, Benjamin and colleagues (Benjamin et al., 1998) demonstrated that at an early stage of 887 postnatal development (postnatal days 4-6), hyperoxic exposure caused marked regression of retinal 888 capillaries, but that hyperoxia had no effect on capillaries at later stages of development (postnatal days 889 10-13), when they have been ensheathed by pericytes. Chan-Ling and colleagues showed that vessel 890 stability is attained during normal neonatal development and also in the recovery phase of kitten ROP, 891 when a desmin-ensheathment ratio (DER) of 0.9 is reached (Chan-Ling et al., 2004b). 892 893 Further, Benjamin and colleagues showed that hyperoxia induced vaso-obliteration at postnatal days 4-6 894 could be prevented byACCEPTED intravitreal administration of VEGF prior to hyperoxic exposure, thus acting as an 895 endothelial survival factor, and that the administration of VEGF in one eye can increase the rate of 896 pericyte coating of developing capillaries compared to the other, untreated eye under hyperoxia. These 897 findings suggest that VEGF is not only a vital survival factor for the developing retinal vasculature, but 898 also an integral factor for the stabilization and normalization of the retinal vascular network in 899 development, especially in hyperoxia related diseases such as ROP. 900

27 901 Role of VEGF in vascular permeabilityACCEPTED MANUSCRIPT 902 The initial discovery of VEGF was intrinsically related to a search for factors that caused increases in 903 microvascular permeability in tumors (Reviewed in (Schlingemann and van Hinsbergh, 1997)). The work 904 of Senger and colleagues (Senger et al., 1983) in discovering “vascular permeability factor” (VPF) laid 905 the basis for the subsequent studies which identified the angiogenic role of VEGF, and ultimately 906 recognized that VPF and VEGF were identical (Senger et al., 1990, Plouet and Moukadiri, 1990). The 907 majority of research in recent years has focused on the angiogenic role of VEGF, however its role as a 908 potent permeability factor is poorly understood in terms of ROP pathogenesis. 909 910 Stone, Chan-Ling and colleagues demonstrated that overexpression of VEGF (Stone et al., 1995) and 911 astrocyte degeneration (Chan-Ling and Stone, 1992) in Phase 2 ROP has a direct effect on the blood- 912 retinal-barrier (BRB), causing a transient breakdown in the integrity of the BRB, extravasation of plasma 913 proteins and tissue edema especially where VEGF is over-expressed. Further, with this increase in edema, 914 there is a likely increase in inflammation and the expression of inflammatory cytokines. Lee and 915 Dammann suggest that early anti-inflammatory treatment may be beneficial by reducing the release of 916 VEGF regulated pro-inflammatory cytokines in Phase 2 ROP (Lee and Dammann, 2012) and their 917 subsequent role in pathological neovascularization. 918 919 Screening for ROP MANUSCRIPT 920 Based on the large datasets accumulated by the Early Treatment For Retinopathy Of Prematurity 921 Cooperative Group, (ETROP) (ETROP, 2003, Good et al., 2005) and CRYO-ROP studies (Palmer et al., 922 1991), the timing of development and progression of ROP was found to be based largely on the 923 postmenstrual age (PMA) of the infant and not the time from birth. Onset of disease in most infants is 924 between 31 and 33 weeks PMA with the greatest severity noted between 35-38 weeks PMA (Reynolds et 925 al., 2002) Chen et al 2015. The results of the examination should be recorded in terms of the International 926 Classification of ROP - Revised (ICROP, 2005). In brief, the examination for acute phase ROP should 927 record at least these three components: the extent of retinal vascularization (Zone I is most posterior and 928 Zone III most peripheral), the extent of abnormality at the junction between the vascularized and 929 avascular retina (no ROP or stage 1-5 ROP with worrisome disease usually greater than stage 2 ROP) and 930 the presence of plus diseaseACCEPTED (markedly increased dilation and tortuosity of the posterior pole vessels,) or 931 pre-plus disease (ICROP, 2005). 932 933 The standard for these examinations is a retinal examination by an ophthalmologist experienced in ROP 934 using a binocular indirect ophthalmoscope (Fierson et al., 2013) although screening for potentially serious 935 disease is likely to be supplanted partially or completely by wide field digital imaging in the future (see 936 section below on telemedicine). The greater availability of wide field retinal cameras would also facilitate 28 937 a new screening paradigm, namely a led programme as suggested by Gilbert and colleagues ACCEPTED MANUSCRIPT 938 (Gilbert et al., 2016). In any setting this could be combined with remote image interpretation, which 939 might be most valuable in middle-income regions by both increasing coverage of at risk infants and 940 reducing the ophthalmology workload. 941 942 Careful timing of the initial examination of at-risk infants and the results of the examination are used to 943 determine the need for and timing of subsequent examination or treatment. Most infants who require 944 treatment are still in hospital and readily accessible for screening. However, it is essential that all 945 stakeholders and care providers are aware of the scheduled examinations (Fierson et al., 2013) so that at 946 risk patients are identified in time to allow treatment and are not lost to follow up. 947 948 At present, the recommendations for detection of ROP in premature infants in the US (Fierson et al., 949 2013) are based largely on the birth weight (BW) and gestational age (GA) of the infant. That is, retinal 950 examinations are recommended for infants with BW <1500 g or with a GA of 30 weeks or less. An 951 examination is also recommended for those slightly larger (1500-<2000g BW) or more mature (>30 952 weeks GA) infants who have an “unstable clinical course” in the opinion of the attending neonatologist or 953 pediatrician.(Fierson et al., 2013) These guidelines have been confirmed to satisfactorily identify 954 extremely low birth weight infants at risk in the US (Kennedy et al., 2014). The timing of the initial and 955 subsequent retinal examinations is critical for the detectionMANUSCRIPT of acute phase disease as intervention in 956 serious disease is time-sensitive. The present US recommendation for initiating ROP examinations is 957 based on the gestational age of the infant and, up to 27 weeks GA the initial examination should be at 31- 958 32 weeks PMA (recognizing that the most immature infants will be 8 or 9 weeks after birth). For the 959 infants with gestations less than 25 weeks, an earlier examination should be considered if serious 960 comorbidities are noted. For infants greater than 27 weeks GA, the recommendation is to start retinal 961 examinations at 4 weeks after birth (Fierson et al., 2013). 962 963 The timing of subsequent examinations is critically important to avoid missing treatable disease not noted 964 at the initial exam. In the current system, the follow up examination is determined by the examining 965 ophthalmologist based on the findings of the initial examination. Based on the ETROP findings (ETROP, 966 2003) for the highest riskACCEPTED eyes (eyes that have retinal vessels extending only into Zone I, eyes that have 967 low stage ROP in Zone I, or eyes with any stage 3 ROP), an examination is recommended in one week or 968 less. Less frequent examinations, every 1-2 weeks or 2 weeks, are recommended for eyes with mild ROP 969 in Zone II or eyes with no ROP but vessels extending into Zone II, with 2-3 week follow up for eyes with 970 Zone III ROP or retinal vascularization well into Zone III. 971

29 972 The decision of when to stop surveillance for acute phase ROP is difficult though there are suggested ACCEPTED MANUSCRIPT 973 guidelines developed in the United States. These include Zone III vascularization with no history of ROP 974 and a PMA of >35 weeks, full retinal vascularization to or near the ora serrata (this is of particular 975 importance if bevacizumab has been used), an examination at 50 weeks in eyes with a history of stage 3 976 ROP or Zone I ROP, or regression of ROP well into Zones II or III (Fierson et al., 2013). If 977 vascularization has not reached the ora serrata at later times, no specific recommendation was made and 978 clinical circumstance would determine further follow up. The UK Guidelines recommend that screening 979 in babies without ROP can cease when normal vascularisation has extended into Zone III usually after 37 980 weeks postmenstrual age or there are clear signs that the active progression of ROP has halted and 981 regression has commenced (Wilkinson et al., 2008). 982 983 As reviewed by Wilson (Wilson et al., 2013), the answers to the basic questions of which infants should 984 be screened, and when, will differ depending on the setting and population and based upon the local 985 epidemiology and resources available. In common with many middle-income countries, in India, larger 986 (up to 2750 gm) and more mature babies (up to 35 weeks PMA) at birth can develop severe ROP 987 requiring treatment (Vinekar et al., 2007). Current Indian guidelines are under revision (personal 988 communication Anand Vinekar MD, Dec 2016) and will likely suggest commencing ROP screening for 989 all preterm neonates who are born < 34 weeks gestation and/or < 1750 grams birth weight; as well as in 990 infants 34-36+6/7 weeks gestation or 1750-2000 grams MANUSCRIPTbirth weight if they have risk factors for ROP. 991 The first retinal examination is recommended not later than 4 weeks of age or 30 days after birth in 992 infants born ≥ 28 weeks of gestational age. Infants born < 28 weeks or < 1200 grams birth weight should 993 be screened early, by 2-3 weeks of age, to enable early identification of AP-ROP. In China, where large 994 infants also can develop ROP, often within a few weeks after birth, screening is recommended at 4-10 995 weeks after birth but if birth weight is >2,000g, screening should take place 3 weeks after birth (Chen et 996 al., 2015). 997 998 Telemedicine and ROP screening 999 Binocular indirect ophthalmoscopy by an ophthalmologist experienced in ROP is the current standard for 1000 detection of eyes that require treatment, but fewer than 10% of those infants undergoing serial 1001 examinations actually ACCEPTEDrequire treatment (Reynolds et al., 2002). This low rate of treatment, coupled with 1002 several factors including the limited availability of ophthalmologists with sufficient ROP experience and 1003 the increasing survival of premature infants worldwide, (Gilbert et al., 2005) has required the 1004 development of alternative screening approaches. The potential role of telemedicine as a screening tool to 1005 select eyes that need evaluation by an ophthalmologist for possible treatment has been bolstered by the 1006 extensive 25 year experience in the successful implementation of retinal imaging in the evaluation of 1007 diabetic retinopathy (Lee, 1999). Systematic evaluation of the efficacy of telemedicine in ROP is 30 1008 necessary because most studies evaluating ROP telemedicine undertaken in the last 15-20 years are ACCEPTED MANUSCRIPT 1009 relatively small. In a 2015 review of the current status of telemedicine in ROP, Fierson and Capone, on 1010 behalf of the American Academy of Section on Ophthalmology, the American Academy of 1011 Ophthalmology and the American Association of Orthoptists (Fierson et al., 2015) contended that the 1012 development of imaging in premature infants coupled with the ability to review those images remotely 1013 provided a useful adjunct to but should not serve as a replacement for repeated binocular indirect 1014 ophthalmoscopy by an on-site ophthalmologist. This review highlighted a number of studies conducted in 1015 NICUs in high-income countries in which the image findings were compared to results of an examination 1016 by an ophthalmologist (defined variably as the “reference standard” or “criterion standard”). Studies with 1017 samples ranging from 43 to 1257 infants had sensitivities of 57% to 100% with wide 95% confidence 1018 intervals for documenting image findings consistent with the reference standard. Outcome measures 1019 chosen for the various studies, varying from the presence of plus disease to stage 3 ROP on zone I or II, 1020 make comparison of outcomes difficult and limit conclusions about the overall validity of the approach. 1021 There is, unfortunately, no consensus on which measures best determine the validity of a telemedicine 1022 approach to screening for serious ROP to identify those infants who need to be evaluated by an 1023 ophthalmologist for potential treatment. 1024 1025 Ells and colleagues developed the term “referral-warranted ROP” to include specific retinal findings that 1026 indicated an eye should be evaluated by an ophthalmologistMANUSCRIPT for possible treatment (Ells et al., 2003). 1027 These findings included the presence of ROP in Zone I, stage 3 ROP or worse, or plus disease. In this 1028 small series of 44 patients, there was 100% sensitivity (95% CI 85-100) and 98% specificity (95% CI 86- 1029 100). Subsequent studies in the US, Canada, and UK that compared image grading with the results of the 1030 clinical examination used various other outcome measures including type 1 ROP (treatment severity), 1031 type 2 and type 1 ROP and “clinically significant ROP.” The varying endpoints selected for the studies 1032 makes comparison of results across studies difficult (Fierson et al., 2015). Most of these studies had fewer 1033 than 100 patients. 1034 1035 The largest study to date that compared retinal image grading to the clinical examination is the National 1036 Eye Institute supported “Telemedicine approaches to evaluating acute-phase ROP” e-ROP Study which 1037 was conducted in 13 NICUsACCEPTED in North America from May 2011 to October 2013 and which enrolled more 1038 than 1200 infants with birth weights of <1251g. (Quinn et al., 2014) In e-ROP, in addition to a clinical 1039 examination by a study ophthalmologist, non- including neonatal nurses, ophthalmic 1040 photographers and technicians obtained 6 wide-field digital images using the RetCam Shuttle (Clarity 1041 Medical System, Pleasantville, CA) from each eye of an infant. The individuals obtaining the images had 1042 undergone an extensive training course and certification process and had been selected for a high comfort 1043 level in dealing with infants in the NICU setting (Karp et al., 2015). Image sets obtained included 5

31 1044 fundus views of the optic disc and temporal, nasal, superior, and inferior views that included the optic ACCEPTED MANUSCRIPT 1045 disc at the edge of the image. In addition, an image of the anterior segment was obtained to assess 1046 pupillary dilation. Image sets from both eyes were then uploaded to a central server for masked grading 1047 by non- readers who had undergone extensive training by ROP experts (Daniel et al., 2015) 1048 1049 In the e-ROP study population, 19.4% of the patients developed referral-warranted ROP (RW-ROP) 1050 based on the clinical examination and 10% were determined to have type 1 ROP. The sensitivity of 1051 detecting RW-ROP on image grading in an infant was 90% (95% CI 85-94), specificity 87% (95% CI 84- 1052 90) and negative predictive value 97.3% at the RW-ROP prevalence of 19.4% in the e-ROP study. When 1053 only those infants who underwent treatment in one or both eyes are considered, the sensitivity for 1054 detecting RW-ROP in images of these patients was 98.2% (95% CI 94.4-99.4), specificity 80.2% (95% 1055 CI 77.0-83.0%) and negative predictive value 99.6% demonstrating the feasibility and accuracy of this 1056 protocol for the identification of ROP infants requiring referral to an ophthalmologist for further 1057 evaluation and treatment. In a further study, the e-ROP investigators documented that remote grading of 1058 images could be returned within 24 hours in more than 95% of cases (Quinn et al., 2016). 1059 1060 The e-ROP investigators have subsequently reported that no serious adverse events (death or prolonged 1061 hospitalization) were found related to e-ROP study visits and adverse events (primarily apnea, 1062 bradycardia, emesis, or reintubation shortly after e-ROPMANUSCRIPT procedures) were noted in 4.9% of infants and 1063 0.8% of study procedures (65 events/8311 procedures). One infant had a new retinal hemorrhage directly 1064 attributed to imaging. Adverse events were more likely to occur when the infant was on mechanical 1065 ventilation, regardless of whether the procedure was an eye examination or retinal imaging. Noteworthy 1066 clinical findings such as apnea greater than 10 seconds and severe bradycardia or tachypnea were more 1067 common than adverse events and the proportion was consistent with other smaller studies on fragile 1068 extremely premature infants (Wade et al., 2015) 1069 1070 Based on the findings of the e-ROP study, e-ROP investigators (Kemper et al., 2016) compared digital 1071 imaging strategies with bedside serial examinations. Using a microsimulation approach of a cohort of 1072 infants with gestational ages from 23-30 weeks GA (microsimulation being a technique to evaluate the 1073 outcomes from competingACCEPTED policies by statistically modeling the experience of individuals within a 1074 hypothetical cohort), they reported that all cases of type 1 ROP were detected using digital imaging in 1075 combination with an ROP examination at discharge. In this theoretical cohort of 650 infants, Kemper and 1076 colleagues (Kemper et al., 2016) demonstrated that examination alone required 1746 procedures while 1077 imaging plus a discharge examination required 1066 ROP examinations and 1786 digital imaging 1078 sessions, i.e. more total interventions for the infants, but fewer examinations required. These findings

32 1079 highlight a number of complex issues including availability of ophthalmologists for routine screening ACCEPTED MANUSCRIPT 1080 examinations, availability of well-trained imagers, and cost of these interventions (Quinn et al., 2016). 1081 1082 In addition to the studies that compare the results of eye examinations with the grading of images, there 1083 have been several studies reporting the implementation of ROP telemedicine in various settings 1084 throughout the world. The KIDROP (Karnataka Internet Assisted Diagnosis of Retinopathy of 1085 Prematurity) model provides ROP screening in more than 80 units in the Indian state of Karnataka, using 1086 technicians who obtain and grade the images at the bedside and report the need for an evaluation to an 1087 ophthalmologist who can then provide treatment if indicated (Vinekar et al., 2014). In more than 1600 at 1088 risk infants, the grading by the highest level of trained technician agreed with the grading of an ROP 1089 expert more than 90% of the time with a sensitivity of 95.7%, specificity 93.2% and importantly, a 1090 negative predictive value of 98.6%. This large and important model demonstrates that ROP screening 1091 services can be provided in regions with limited access to clinical examinations, but it is however, 1092 dependent on the availability of a sufficient number of ophthalmologists to provide treatment and the 1093 expertise of the trained technician. 1094 1095 In the US, ROP telemedicine is becoming increasingly available. The Stanford University for diagnosis of 1096 ROP (SUNDROP) study has reported a series of evaluations of infants meeting ROP screening criteria 1097 with images taken in the NICU by trained nurses and remotelyMANUSCRIPT evaluated by an ROP-experienced 1098 ophthalmologist. All infants were followed with an examination after discharge as clinically indicated. In 1099 the recent report of 608 infants over a 6 year period, images from 22 infants were graded as having type 1 1100 ROP (treatment-warranted TW-ROP in these studies) and were evaluated by an ophthalmologist who 1101 determined that all met type 1 ROP criteria except for one infant with only “stage 3 ROP insufficient to 1102 warrant intervention.” Sensitivity was calculated at 100%, specificity of 99.8% and a negative predictive 1103 value of 100%. Mean birth weight in this entire cohort was 1261g (range 420-3744g) (Wang et al., 2015) 1104 while only 1 infant requiring ROP treatment had a birth weight of more than 1000g. 1105 1106 Weaver and Murdock (Weaver and Murdock, 2012) reported experience in Montana with remote imaging 1107 by trained nurses in a level III NICU when there was no ophthalmologist within 200 miles of the center. 1108 In this study of 137 infants,ACCEPTED images were sent to ROP-experienced ophthalmologists and they 1109 recommended transfer of 13 infants to another institution for examination with 9 eventually requiring 1110 laser treatment. Such an approach allowed the majority of the infants to be cared for nearer to family and 1111 fewer than 10% were transferred to the remote facility for an examination. 1112 1113 The Fierson report, (Fierson et al., 2015) noted above, also concluded there is a “need for systematic 1114 studies” of ROP telemedicine highlighting cost versus benefit, and tradeoffs needed from the perspective

33 1115 of various stakeholders and provided details about important components of such a system. However, this ACCEPTED MANUSCRIPT 1116 report addresses largely NICU settings in high-income countries with good availability of 1117 ophthalmologists to evaluate infants at risk. Gilbert and colleagues (Gilbert et al., 2016) suggest that 1118 screening using retinal imaging could be led by neonatology (see section above on Screening for ROP). 1119 In low and middle income countries, if coupled with remote reading of images and ophthalmologists only 1120 coming to visit NICUs for a relatively small number of infants who appear to have severe ROP, such an 1121 approach has the potential to greatly increase assessment of at risk infants Clearly, this would require 1122 explicit definitions of roles and responsibilities with excellent training and quality control for imagers and 1123 image evaluators, and also the continued availability of ophthalmologists able to review the images as 1124 needed and to provide examinations with treatment if indicated. Such a shift would require “close 1125 collaboration between the team on the neonatal unit and ophthalmologists” with an important decision to 1126 be made about who would be responsible for evaluating the images. Such an approach could expand 1127 screening dramatically to the ‘at risk population’ as increasingly small birth weight and low gestational 1128 age infants survive worldwide. 1129 1130 Fluorescein angiography before and after treatment for ROP 1131 The findings on fluorescein angiography (FA) in ROP were first reported almost 40 years ago by Flynn 1132 and colleagues, a time when the term retrolental fibroplasia (RLF) was still being used to designate those 1133 changes in the eye of premature infants that came to be knownMANUSCRIPT as ROP (Flynn et al., 1977). These 1134 investigators noted a series of findings, largely on the retinal vasculature, which were present in eyes with 1135 RLF. In eyes with RLF, they proposed that the structure between the vascularized and avascular retina 1136 was a “functioning shunt comprised structurally of mesenchyme within which are primitive vascular 1137 channels connecting arteries to veins directly.” This structure could be circumferential and of variable 1138 thickness with the “retinal vessels feeding this shunt…dilated and tortuous.” On FA, the shunts “fill 1139 rapidly and leak profusely” with little color difference between the venules and arterioles. They also 1140 noted areas of capillary non-perfusion and “more posteriorly, capillary architecture…abnormal” with 1141 areas of non-perfusion posterior to the shunt. 1142 1143 As technology for imaging the very premature infant improved significantly, interest in using FA as a tool 1144 for understanding the ACCEPTEDmorphologic changes that occur in ROP was revived by a number of investigators 1145 in the early 2000s (Schulenburg and Tsanaktsidis, 2004, Yokoi et al., 2009, Ng et al., 2006). Clinician 1146 investigators, both neonatologists and ophthalmologists, at the Catholic University in Rome undertook a 1147 series of studies that further detailed the FA findings in eyes at risk for ROP.(Purcaro et al., 2012, Lepore 1148 et al., 2014, Lepore et al., 2011). In a 2012 report, they outlined several findings that detailed the 1149 “variable features of retinal and choroidal circulation in preterm infants with a high risk of developing 1150 ROP.”(Purcaro et al., 2012) They noted variability in choroidal filling in the posterior pole with linear

34 1151 and lobular patterns that varied over time. They also found that the abnormal vascular structure between ACCEPTED MANUSCRIPT 1152 the vascularized and avascular retina were easier to identify using FA than with indirect ophthalmoscopy. 1153 Abnormal branching and leakage were easier to identify with FA, with loss of the “normal dichotomous 1154 branching” as ROP develops in the immature eye. Of particular note was the finding of “hypofluorescent 1155 areas with peri-arteriolar loss of capillary bed in the posterior retina” behind the ROP structural 1156 abnormality at the junction of the vascularized and avascular retina. 1157 1158 These observations have been extended in the diagnosis and determination of the need for treatment of 1159 ROP in its most serious forms. Klufas and colleagues (Klufas et al., 2015) suggest that FA would 1160 improve sensitivity for the need for treatment over fundus imaging alone, largely by increasing the ability 1161 to identify stage 3 ROP more accurately. With the use of anti-VEGF drugs in the treatment for ROP 1162 (Mintz-Hittner et al., 2011) (See section: Anti-VEGF therapy in ROP: The Pros and Cons) FA has been 1163 useful for determining the extent of the abnormal retinal findings both before and after treatment. Lepore 1164 and colleagues (Lepore et al., 2014) conducted a small prospective study on infants with type 1, Zone I 1165 ROP. For treatment, 1 eye was randomly selected (using a random number series) to undergo 1166 conventional laser photoablation of the peripheral avascular retina; the fellow eye received an intravitreal 1167 injection of 0.5 mg bevacizumab in a 0.02 ml balanced salt solution.” At 9 months after treatment, they 1168 performed FA on 12 of the 13 enrolled patients. Among the majority of the bevacizumab-treated eyes, 1169 they documented large avascular areas with abnormal branchingMANUSCRIPT and shunt persistence in the retinal 1170 periphery, along with hyperfluorescent lesions, foveal avascular zone absence, and capillary dropout in 1171 the posterior pole (see Figure 6 (Lepore et al., 2014). These persistent abnormalities were rarely observed 1172 in the laser treated eyes. They concluded that there should be continued surveillance and concerns about 1173 the long-term visual function of eyes treated with anti-VEGF drugs based on FA monitoring of ROP 1174 progression. There is, as yet, no consensus, as to whether these avascular areas should be treated or 1175 simply observed long term (see findings reported by Mintz-Hittner et al 2016 below). 1176 1177 Other groups have noted abnormalities after anti-VEGF treatment of eyes with type 1 ROP. Tahija and 1178 colleagues (Tahija et al., 2014) examined FAs in 20 eyes of 10 patients and concluded that, “although 1179 bevacizumab appears effective in bringing resolution, complete normal retinal peripheral retinal 1180 vascularization was notACCEPTED achieved in” 11 of the 20 eyes. Henaine-Berra and colleagues (Henaine-Berra et 1181 al., 2014) documented, in a series of 47 eyes of 26 patients who underwent FA at least one month after 1182 treatment with bevacizumab, that there was “improvement of abnormal vascular findings.” They noted a 1183 foveal avascular zone in more than half of the eyes, “subsequent growth of vessels to the capillary-free 1184 zones in all eyes,” but persistence of vascular loops in most of the eyes. In 83% of the eyes, perivascular 1185 leakage was noted “through the end of the follow-up period.” Toy and colleagues (Toy et al., 2016) 1186 reported a specific ‘scalloped’ pattern of vascular regression after intravitreal bevacizumab. This was

35 1187 associated with chronic vascular arrest, which if further than two disc diameters from the ora serrata, was ACCEPTED MANUSCRIPT 1188 associated with an increased, long-term risk of disease recurrence. All of these studies and other 1189 investigators (Klufas and Chan, 2015, Mintz-Hittner et al., 2016) contend that longer follow-up is needed 1190 to determine the safety of anti-VEGF in ROP. 1191 1192 Treatment of ROP 1193 Treatment of ROP can be carried out for both the proliferative and cicatricial (i.e. scar forming) phases of 1194 ROP. Treatment before vitroretinal traction occurs can produce good visual results but once retinal 1195 detachment occurs, even peripherally, the visual results are often quite poor (Reynolds et al., 1995, 1196 Gilbert et al., 1992, Dobson et al., 1993, Reynolds et al., 1993, Gilbert et al., 1996). From this it follows 1197 that timely screening, based on natural history data (largely determined by findings of the CRYO-ROP 1198 study) (Palmer et al., 1991) will detect those children who require treatment well before they suffer 1199 blinding complications. Timing of treatment is therefore crucial. Treatment of the proliferative phase is 1200 directed towards reducing the VEGF driven angiogenic drive, either by ablating the peripheral retina or 1201 inactivating VEGF already released from areas of ischemic peripheral retina. 1202 1203 Treatment of acute ROP began in the 1970s, initially with photocoagulation which was pioneered in 1204 Japan (Reviewed in (Palmer et al., 1985)). Treatment was originally directed at the new vessels forming 1205 the ridge and only later at the more anterior avascular retina.MANUSCRIPT The xenon arc and argon laser equipment 1206 then available were not practicable for neonatal use and cryotherapy became the favored treatment option 1207 because it could be used with the binocular indirect ophthalmoscope. The early reports were difficult to 1208 interpret because there was no uniform grading system of reporting the changes of acute retinopathy and, 1209 as Palmer (Palmer et al., 1985) noted, most studies included only small numbers of patients and lacked 1210 adequate randomization and controls. 1211 1212 A major step forward in combating blindness from ROP was the publication in 1984 of the International 1213 Classification of Retinopathy of Prematurity (ICROP, 1984). This classification, which described ROP as 1214 seen with an indirect ophthalmoscope described the disease according to the location, extent and severity 1215 of the acute changes, became universally accepted. Location was described in terms of “zone” with the 1216 most posterior (Zone I)ACCEPTED defined by a circle centered on the disc with a radius of twice the disc macula 1217 distance. Zone II is doughnut-shaped and extended to the nasal ora serrate and Zone III is peripheral to 1218 that. Severity is indicated by “stage” with stage 1 (demarcation line) the least severe and stage 5 (total 1219 retinal detachment) the most severe acute phase disease. Extent was designated in clock hours along the 1220 peripheral retinopathy. Severity of disease was also indicated by the presence of “plus disease” defined as 1221 abnormal posterior pole vessel dilation and/or tortuosity. A description of cicatricial disease was added in

36 1222 1987 (ICROP, 1987). ICROP was revised and updated in 2005, when a new category of aggressive ACCEPTED MANUSCRIPT 1223 posterior ROP (AP-ROP) was added and “pre-plus” changes were defined (ICROP, 2005). 1224 1225 CRYO-ROP Study 1226 Following the publication of the international classification, randomized controlled trials (RCT) became 1227 possible. The most important early RCT was the CRYO-ROP study conducted in 23 centers in the US 1228 from 1986-8 (CRYO-ROP, 1988). This trial allocated treatment with cryotherapy to one eye with the 1229 fellow eye as control. The trial was concluded prematurely due to the demonstration of an overwhelming 1230 benefit to ablative treatment. Cryotherapy under indirect ophthalmoscopy became the standard of care for 1231 ROP treatment but was largely supplanted in western countries by laser in the 1990s because laser can be 1232 delivered with more precision and less morbidity (Simpson et al., 2012). 1233 1234 The ICROP classification allowed the multicenter randomized controlled trial of cryotherapy for 1235 threshold disease (CRYO-ROP) to be carried out (Palmer et al., 1985). “Threshold” disease (defined as 5 1236 contiguous or 8 total clock hours of stage 3 plus disease,) had been predicted to lead to blindness in 50% 1237 of cases. Fifteen-year outcomes showed a long-term benefit of treatment with unfavourable structural 1238 outcomes in 48% of untreated eyes compared with 27% in treated eyes, and unfavourable visual acuity 1239 (worse than 20/200) in 62% of untreated eyes versus 44% in treated eyes (Palmer et al., 2005). 1240 MANUSCRIPT 1241 The immediate benefit of the CRYO-ROP study was tha t there was now a randomized, controlled-trial- 1242 proven treatment for ROP. While cryotherapy has now been supplanted by laser, the lasting benefits of 1243 the CRYO-ROP study, now viewed from a distance of nearly 30 years, have been; (1) that the beneficial 1244 effects of intervention were maintained long term (Palmer et al., 2005); (2) the natural history of the 1245 development of sight threatening ROP was described in a large neonatal population, (Palmer et al., 1991); 1246 (3) that cryotherapy had much less effect on the visual field than would be thought from an examination 1247 of the fundus (5-10% loss) (Quinn and Dobson, 1996, CRYO-ROP, 2001); and (4) the finding that there 1248 was ongoing ocular morbidity (including late retinal detachment)(Quinn and Dobson, 1996) from ROP 1249 for years after apparently successful cryotherapy treatment (Palmer et al., 2005). Kremer and colleagues 1250 (Kremer et al., 1995) reported on 10 children who had Humphrey visual field examination 10 to 14 years 1251 after neonatal cryotherapyACCEPTED for ROP; field constriction was 10-20 degrees temporally and 10-30 degrees 1252 nasally. Affected children did not have any subjective field defects nor did they notice reduced night 1253 vision. 1254 1255 Laser versus Cryotherapy 1256 Modern, usually diode, trans-pupillary laser treatment has considerable advantages over cryotherapy in 1257 that it can be delivered precisely to the area requiring treatment and is less traumatic and painful. 37 1258 However, retinal laser photocoagulation is technically demanding to perform, usually requires some form ACCEPTED MANUSCRIPT 1259 of sedation, takes about an hour to treat two eyes and preferably should be carried out in “laser safe’ 1260 rooms to minimize risk to nearby staff and infants. The diode laser is readily portable which is a big 1261 advantage in developing countries where laser treatment can be carried out in remote locations. A number 1262 of studies have shown results with laser to be at least as good as cryotherapy (Simpson et al., 2012, 1263 Tsitsis et al., 1997) or better (Sahni et al., 2005) with stable long term outcomes (McLoone et al., 2006, 1264 Al-Otaibi et al., 2012). Sanghi and colleagues directly compared outcomes between diode (810nm) with 1265 green (532nm) laser and found equal outcomes with both lasers (Sanghi et al., 2010). Diode laser has the 1266 theoretical advantage that it is not absorbed by haemoglobin so it can be used in the presence of retinal or 1267 vitreous haemorrhages and still achieve retinal burns. Diode laser is less likely to damage the lens. 1268 Because of these advantages, in developed economies, diode laser is currently the mainstay of treatment. 1269 1270 However, diode laser burns are absorbed by melanin so they produce a deeper burn than green laser; 1271 potentially leading to worse visual outcomes due to more extensive photoreceptor damage. However, 1272 green laser has some reported advantages over diode; less pain due to less penetration (shorter 1273 wavelength) which may be an advantage for outpatient treatment, it is easier to see spots on retina (good 1274 for beginners- green laser spots are whiter) so there is less chance of over treatment, less power is 1275 required and the scars are less dense and OCT shows better preservation of photoreceptors (Anand 1276 Vinekar MD- personal communication). In addition, mostMANUSCRIPT retina specialists have a green laser for diabetic 1277 retinopathy so there is no need to acquire a diode laser especially for ROP; this is a particular advantage 1278 in middle-income countries such as India . 1279 1280 Laser for ROP 1281 The Early Treatment for Retinopathy of Prematurity Study (ETROP) enrolled infants with BW <1251 gm 1282 from 2000-2 and screened approximately 7,000 infants at 26 centers in the US to achieve a cohort of 401 1283 infants who were at high risk of an adverse outcome (Hardy et al., 2004). They were then entered into a 1284 RCT of earlier vs conventionally timed treatment of laser for ROP. The study concluded that laser applied 1285 earlier than threshold disease resulted in better structural and functional outcomes. From then on, laser 1286 became the standard of treatment in the developed world. 1287 ACCEPTED 1288 The ETROP study compared laser treatment of eyes reaching threshold (as above) with “pre-threshold” 1289 (Zone I ROP of any stage short of threshold; Zone II, with stage 2 in the presence of plus disease, or stage 1290 3 without plus; Zone II, with stage 3 in the presence of plus disease but not reaching threshold criteria). 1291 At 9 months post-treatment unfavourable structural outcome occurred in 15.9% eyes treated at threshold 1292 compared with 9.1% with earlier treatment, and unfavourable visual acuity was found in 19.5% and 1293 14.5% respectively (ETROP, 2003).

38 1294 ACCEPTED MANUSCRIPT 1295 The ETROP trial recommended laser for “type 1 disease”: Zone I ROP of any stage with plus disease, 1296 Zone I stage 3 ROP with or without plus disease, Zone II of stage 2 or 3 ROP with plus disease (ETROP, 1297 2003). There was a clear structural benefit for treating eyes with type 1 disease but not for type 2 disease 1298 (Good et al., 2005) so a “wait and see approach” was recommended for type 2 disease (Zone I, stage 1 or 1299 2 without plus; Zone 2, stage 3 without plus). This benefit for treating type 1 disease, and not type 2 1300 disease was confirmed for visual acuity as well (ETROP, 2011). A small study from Thailand also 1301 reported better structural and functional outcomes for pre-threshold vs threshold treatment of ROP 1302 (Warrasak et al., 2012). 1303 1304 Table 2: Indications for ROP treatment. Note-Type 1 ROP does encompass what was defined as AP- 1305 ROP* in ICROP II CRYO -ROP (CRYO -ROP, 1988 ) ET -ROP (Good and Early Treatment for Retinopathy of Prematurity Cooperative, 2004) ROP requiring Threshold Disease Type 1 ROP (Severe) treatment • 5 continuous or 8 cumulative • Zone I, any stage with Plus* clock hours of Stage 3 in Zone • Zone I, Stage 3 without Plus I or II with Plus Disease • Zone II, Stage 2 or 3 with Plus

Wait and watch - MANUSCRIPTType 2 ROP (Less Severe) may require • Zone I, Stage I or 2 without Plus treatment in the • Zone II, Stage 3 without Plus future 1306 1307 Strabismus is commonly seen in severe ROP; in the ETROP study, the cumulative incidence of 1308 strabismus after six years was close to 60%. Even in those children with good acuity scores and 1309 favourable anatomic outcomes, the prevalence of strabismus was around 30% but in children with 1310 impaired vision either from ocular or cerebral causes, the prevalence was 80% at age 6 years 1311 (VanderVeen et al., 2011) 1312 ACCEPTED 1313 There are many reports of favourable outcomes from laser treatment in the majority of patients (Tsitsis et 1314 al., 1997) but there is still appreciable visual morbidity related to induced refractive errors, especially 1315 myopia and anisometropia as well as strabismus (Yang et al., 2010). However, many of the patients with 1316 treated ROP also have periventricular leukomalacia which may be associated with cerebral vision 1317 impairment and other neurological co-morbidities such as nystagmus, making the contribution of ROP to 1318 the overall visual status difficult to assess accurately (Ospina et al., 2005, Siatkowski et al., 2013). In 1319 children with favorable structural outcomes in the ETROP study but with vision loss glaucoma has been reported to occur (Lee et al., 1998, Trigler et al., 2005) The acute 40 1356 pressure rise occurs frequently enough that most patients are routinely put on anti-glaucoma medications ACCEPTED MANUSCRIPT 1357 for a few days after laser treatment. Exudative detachments occur infrequently after laser treatment; they 1358 usually settle spontaneously, but bevacizumab has been used to treat this with good effect (Ehmann and 1359 Greve, 2014). Visual field constriction from laser is less than predicted from an examination of the 1360 fundus, particularly from treatment in Zone II, but laser treatment in Zone I disease must produce 1361 constriction of the visual field although this is difficult to assess in the setting of the frequent poor visual 1362 acuity in such children. 1363 1364 Cataract is a rare complication of laser for ROP occurring in about 2% of cases in the ETROP study, 1365 although it can occur without any treatment (Davitt et al., 2013). The BEAT-ROP Study had a 3/73 1366 (4.1%) prevalence of cataract and one case of corneal opacity requiring corneal transplantation (Mintz- 1367 Hittner et al., 2011), following laser which is higher than previously reported, so in the BEAT-ROP 1368 study, these effects may have been operator dependent. 1369 1370 Long term outcomes of laser treatment 1371 Myopia, and high myopia is a major long-term complication of laser and appears to be related to the 1372 number of burns applied (Katoch et al., 2011) which is in effect a proxy for disease severity. Earlier 1373 treatment does not appear to affect the prevalence of myopia (Quinn et al., 2013). The increase in myopia 1374 is due to a significantly thicker lens, shallower anterior chamberMANUSCRIPT and a steeper corneal curvature than 1375 control eyes (Yang et al., 2013). There is also increased myopic astigmatism due to a steeper vertical 1376 corneal curvature (Yang et al., 2010). Myopia

41 ≥ 1392 In a Swedish cohort, examined at 10 AyearsCCEPTED of age, 2 5%MANUSCRIPT had visual acuity 0.1 logMAR and/or strabismus 1393 and/or reduced contrast sensitivity with the main risk factors being neurological dysfunction, cryo-treated 1394 ROP and astigmatism, anisometropia or strabismus at 2.5 years of age (Holmstrom and Larsson, 2008). A 1395 Danish cohort study of 4 year-old-children, contemporaneous with the ETROP study reported better 1396 visual and refractive outcomes than the ETROP despite similar treatment protocols, possibly due to 1397 differences in delivery of care between Scandinavian and US health care systems (Fledelius et al., 2015) 1398 Long term survivors (decades) with ROP, especially from the pre-treatment era, can have ongoing 1399 problems in both the anterior and posterior segments occasionally leading to enucleation for blind painful 1400 eyes and insidious vision loss for which there is no obvious cause (Fledelius and Jensen, 2011). 1401 1402 Recommendations for long term follow up. 1403 Preterm infants as a group have a higher prevalence of amblyopia, strabismus and refractive errors than 1404 full-term infants (Holmstrom et al., 1999, O'Connor et al., 2007). While this is generally agreed upon, 1405 there are no standard recommendations for follow up after nursery discharge and practices vary widely 1406 (O'Connor et al., 2006). Most children with severe ROP or significant neurological deficits will be 1407 followed up anyway, but children with no or only mild ROP can still have refractive errors that might 1408 require intervention so if only one screening is carried out, it should be at age 24-30 months, if two 1409 screenings are carried out, the second screening should be carried out at 42-48 months of age (Holmstrom 1410 et al., 1999), which is typically co-incident with their preschoolMANUSCRIPT vision screening. 1411 1412 Outcomes after laser for AP-ROP 1413 In developed countries, ELBW babies are now the largest group still developing sight-threatening ROP. 1414 The worst visual outcomes occur in those infants with aggressive posterior ROP (AP-ROP) which 1415 typically occurs in Zone I. The visual outcomes for laser treatment of AP-ROP are generally poor (Wani 1416 et al., 2013, Spandau et al., 2013, Gunn et al., 2014) despite good anatomical success rates (Sanghi et al., 1417 2014). Early vitreous surgery has been advocated but the results remain poor especially if fibrovascular 1418 growth has already reached the vitreous base when surgery is performed (Azuma et al., 2013). 1419 After laser for AP-ROP, there may be early decrease in posterior vascular abnormality, but the 1420 retinopathy can reappear and progress rapidly to intractable retinal detachment (Vinekar et al., 2008). A 1421 second session of laserACCEPTED to fill in areas of regression of flat neovascularization produced by a first session 1422 will produce fewer haemorrhages and less fibrosis post treatment than a single session of heavier laser 1423 (Vinekar et al., 2015a). AP-ROP occurs in larger and more mature Indian infants than in high-income 1424 countries (Sanghi et al., 2010). Indeed the severe ROP seen in larger Indian infants may be a ‘hybrid’ of 1425 both typical AP-ROP and the classic ROP stages because it may also have a ridge at the junction of 1426 vascular and avascular retina (Sanghi et al., 2012). AP-ROP in India appears to have better outcomes 1427 from treatment if early screening, aggressive laser treatment followed by close follow-up and retreatment 42 1428 if necessary takes place (Jalali et al., 2011). Better outcomes in Indian infants with AP-ROP may also be ACCEPTED MANUSCRIPT 1429 due to genetic and racial differences including darker fundus pigmentation allowing a better uptake of 1430 laser burns (Sanghi et al., 2009). 1431 1432 Anti-VEGF therapy in ROP: The pros and cons 1433 Because vasoproliferation in ROP is largely VEGF driven and anti-VEGF drugs can be easily 1434 administered by intravitreal injection, ROP is an attractive target for such treatment. Anti-VEGF 1435 treatment is much more problematic in the premature newborn than adults, where its use is widespread 1436 for retinal vascular disease, because VEGF has a key role in normal development of many organs, 1437 including the brain, kidneys and lungs. The potential advantages of anti-VEGF therapy (Klufas and Chan, 1438 2015) include the fact that the injections can be administered quickly (2-3 minutes per eye) compared to 1439 laser photocoagulation (30-40 minutes per eye); require less specialized equipment, and appears to have 1440 fewer ocular side effects such as induced myopia and astigmatism. Intravitreal injection of anti-VEGF 1441 agents may allow vascularization to extend further into the retinal periphery than laser. Anti-VEGF 1442 treatment is therefore attractive in situations where laser treatment is difficult, such as in infants with 1443 opaque media, or who are very sick and unstable and unable to tolerate the longer (laser) procedure. In 1444 developing countries, the lack of availability of lasers and trained ophthalmologists who can operate them 1445 makes anti-VEGF treatment potentially useful, though it requires careful and sustained follow up to 1446 monitor for recurrence of retinopathy and long term systemicMANUSCRIPT complications (Hu et al., 2012, Lien et al., 1447 2016, Morin et al., 2015). 1448 1449 Anti-VEGF therapy is increasingly being employed as a first-line or rescue treatment for retinopathy of 1450 prematurity in many middle-income and emerging economies because it is widely available and easily 1451 accessible. A circumspect and detailed analysis of its benefits, and pitfalls, especially in relation to ROP 1452 recurrence (Chan et al., 2016, Mintz-Hittner et al., 2016), is required because the risks of long term ocular 1453 and systemic side effects are, as yet, not fully determined. Angiogenesis is still active in many organs of 1454 these infants at the time of treatment, and the potential risks and safety profiles of anti-VEGF agents 1455 (bevacizumab, ranibizumab and others) especially long term are not fully known. The growing body of 1456 literature which demonstrates that VEGF has a major neuroprotective role, the potential systemic toxicity 1457 of intravitreal injectionsACCEPTED of anti-VEGF in organs other than the eye; the long half-life of anti-VEGF 1458 agents in the body; and evidence of late reactivation of ROP post anti-VEGF treatment are all issues 1459 which must be examined before anti-VEGF therapy takes its place as a routine treatment for ROP. 1460 1461 VEGF-based therapeutic options 1462 There are currently four candidate anti-VEGF drugs for treatment of acute ROP (Stewart, 2012).

43 1463 Bevacizumab (trade name: Avastin) is a full-length IgG antibody developed and licensed for the systemic ACCEPTED MANUSCRIPT 1464 treatment of certain cancers in 2004 but, in part because it is much cheaper than ranibizumab, it has been 1465 used extensively off-label as intravitreal treatment of AMD and other proliferative retinopathies from 1466 2005 (Fung et al., 2006). Ranibizumab (trade name: Lucentis) is a humanized monoclonal antibody 1467 fragment active against all isoforms of VEGF, which was developed for the intravitreal treatment of some 1468 forms of age-related macular degeneration (AMD), and was licensed for this indication in 2006. 1469 Pegaptanib (trade name: Macugen) is a small molecule attached to polyethylene glycol to prevent rapid

1470 breakdown and which binds tightly to the VEGF 165 isoform: it is licensed for intravitreal use in adults. 1471 The most recently marketed drug, aflibercept (trade name: Eylea, also known as VEGF-Trap), is a fusion 1472 protein based on VEGF receptor domains that binds VEGF isoforms and is licensed for treatment of some 1473 forms of AMD but there are no published reports of its use for ROP. Whilst much cheaper than the other 1474 drugs, bevacizumab has a longer half-life in the vitreous and intravitreal injection results in much higher 1475 systemic concentrations, again with a much longer half-life (Table 3). 1476 1477 Table 3: Vitreous Peak Systemic Agent Mol. Wt Dose Vitreous half-life plasma half-life (KD) (mg) conc. (μM) (days) conc. (days) MANUSCRIPT(ng/ml)

Bevacizumab 150 1.25 2.1 10 20-687 20

Ranibizumab 50 0.5 2.5 9 0.79-2.9 0.09

Pegaptanib 50 0.3 1.5 3.9 6-7 10 1478 Adapted From: Table 1 - Tolentino M Surv Ophthalmol 2011 (Tolentino, 2011) 1479 1480 An increasing number of case series of either bevacizumab or ranibizumab treatment, usually in 1481 combination with laserACCEPTED but sometimes as monotherapy, have been reported (Micieli et al., 2009, Wallace 1482 and Wu, 2013, Mititelu et al., 2012). There have now been (Autrata et al., 2012, Karkhaneh et al., 2016, 1483 Kong et al., 2015b, Mintz-Hittner et al., 2011, Moran et al., 2014) five randomized trials of anti-VEGF 1484 agents as treatment for ROP published (Autrata et al., 2012, Karkhaneh et al., 2016, Kong et al., 2015b, 1485 Mintz-Hittner et al., 2011, Moran et al., 2014).These trials have small numbers of patients and several 1486 have design flaws. They cannot at this point be viewed as providing conclusive data on the ocular and 1487 systemic outcomes of the premature infants treated.

44 1488 ACCEPTED MANUSCRIPT 1489 BEAT-ROP Study 1490 The Bevacizumab Eliminates the Angiogenic Threat of Retinopathy of Prematurity (BEAT-ROP) trial 1491 (Mintz-Hittner et al., 2011), investigated the efficacy of anti-VEGF therapy for Zone I or posterior Zone 1492 II stage 3 + ROP. The treatment was 0.625mg intravitreal bevacizumab, with laser to the avascular retinal 1493 periphery as the control arm with a primary outcome measure of need for retreatment by 54 weeks PMA. 1494 There were 150 infants (300 eyes) enrolled with 143 infants surviving to 54 weeks. ROP requiring 1495 retreatment occurred in 4% of the intravitreal bevacizumab eyes and in 22% of laser treated eyes with a 1496 significant effect for Zone I ROP (P=0.003) but not for Zone II eyes (P=0.27). 1497 1498 The timing of recurrences needing treatment was later following bevacizumab (19.2±8.6 weeks) 1499 compared with laser (6.4±6.7 weeks). The trial was not sufficiently powered to fully assess the safety 1500 profile and long-term effects of the systemic administration of intravitreal bevacizumab in ROP 1501 treatment. Five infants died in the bevacizumab group (four of respiratory causes) and two infants died in 1502 the laser group; although not reaching statistical significance many would consider this of clinical 1503 significance. A follow up of refractive outcomes on 109 of 131 eligible patients showed a highly 1504 significant increase in myopia for laser treatment in both Zone I and Zone II eyes (Geloneck et al., 2014). 1505 Very high myopia ( ≥ -8.00D) occurred in 3/110 (2.7%) of eyes that received bevacizumab and 42/124 1506 (33.9%) eyes that were treated with laser (P<0.001for bothMANUSCRIPT Zone I and Zone II eyes). The authors 1507 attributed this difference to the negative effect laser has on anterior segment development. The myopia 1508 appeared to worsen with increasing numbers of laser burns, particularly if the eyes were retreated with 1509 laser after recurrence of ROP. 1510 1511 There were a number of design and other issues with the BEAT-ROP trial, as noted in subsequent 1512 critiques (Darlow et al., 2013a, Hard and Hellstrom, 2011, Moshfeghi and Berrocal, 2011) and 1513 correspondence in the publishing journal of the BEAT-ROP trial ( N Engl J Med 2011; 364:2359-2362), 1514 including the fact that treatment criteria were not standard, no details of the laser treatment were given 1515 and the results in the laser treated group were much poorer than in other series, the outcome was not 1516 masked and follow-up was only to 54 weeks PMA. Importantly, the trial was not powered to assess 1517 mortality or systemic toxicity.ACCEPTED 1518 1519 Five other published trials included four that compared an anti-VEGF agent with laser and one (Autrata 1520 et al., 2012) that investigated combination therapy; all were of limited patient numbers and varied 1521 considerably both in design and outcome measures. In 13 infants with type 1, zone I disease, Lepore and 1522 colleagues (Lepore et al., 2014) randomized one eye to treatment with bevacizumab and the fellow eye to 1523 laser, the primary outcome being retinal and choroidal vessel abnormalities at 9 months on fluorescein 45 1524 angiography. Persistent retinal abnormalities occurred more frequently after bevacizumab treatment. (See ACCEPTED MANUSCRIPT 1525 section above: Fluorescein Angiography before and after Treatment for ROP). In 14 infants with zone I or 1526 II, stage 3 plus disease, Moran and colleagues randomized eyes to either bevacizumab or laser treatment 1527 (Moran et al., 2014). There were three recurrences after bevacizumab and only one after laser treatment. 1528 One and two year follow-up was reported, including cranial MRI scans, which were said to demonstrate 1529 “no abnormality that could be attributed to bevacizumab” due to systemic exposure to VEGF during 1530 critical period (Moran et al., 2014). Kong (2015) studied 24 patients with type 1 disease who were quazi- 1531 randomized to bevacizumab 0.625 mg/dose, 0.25 mg/dose or laser treatment (Kong et al., 2015a). The 1532 primary outcome was serum free VEGF and IGF-1 concentrations to 90 days. Karkhaneh and colleagues 1533 (2016), in a study from Iran, enrolled 79 infants with zone II, stage 2 or 3 plus ROP (Karkhaneh et al., 1534 2016). The details of randomisation are uncertain and the primary outcome was treatment failure by 90 1535 weeks PMA; this occurring in 9/43 treated with bevacizumab and 1/36 treated with laser (Karkhaneh et 1536 al., 2016). The heterogeneous nature of these studies, generally small sample size and lack of details of 1537 the study design mean they add little to our understanding of relative advantages of bevacizumab 1538 compared to laser with the clear exception of Lepore (2014) (Lepore et al., 2014). 1539 1540 Following on from the BEAT-ROP study, Hittner and colleagues (Mintz-Hittner et al., 2016) recently 1541 reported on their management of recurrent ROP after intravitreal bevacizumab therapy (0.625mg) in a 1542 larger cohort of premature infants (241 infants, 471 eyes).MANUSCRIPT The recurrence rate was 7.2% for eyes. The 1543 risk factors were AP-ROP, prolonged hospital stay and lower birth weight. Most recurrences occurred 1544 between 45 and 55 weeks adjusted age (PMA) with recurrence up to 64.9 weeks. Recurrence was more 1545 likely if there was slow progression of retinal vascularization after the first treatment, and the signs of 1546 recurrence were return of plus disease and neovascularization, usually at the advancing edge of the retinal 1547 vessels. Wu and colleagues (Wu et al., 2013) reported on their experience with 162 eyes from 85 patients 1548 given intravitreal bevacizumab 0.625mg; they reported an 88% success rate for one injection with 1% of 1549 cases showing either intraocular haemorrhage, cataract and exotropia. 1550 1551 Anti-VEGF therapy in combination with laser 1552 The use of anti-VEGF therapy in combination with laser has been widely reported in small case series. 1553 Anti-VEGF can be usedACCEPTED as primary therapy followed by laser if complete vascularization fails to occur, as 1554 a combination treatment with laser or as a rescue treatment following failed laser treatment. In a 1555 retrospective non-randomized case series, Hwang and colleagues (Hwang et al., 2015) reported on the 1556 outcomes of 54 infants treated with either IVB or laser; the IVB treated eyes showed less myopia than the 1557 eyes treated with laser alone. A prospective randomized trial of pegaptanib, 0.3 mg in 0.02 mL, combined 1558 with laser therapy compared with laser therapy alone (although in one place the report states “laser 1559 therapy combined with cryotherapy”) for stage 3 plus in Zone I or posterior Zone II in 74 infants, has

46 1560 been published (Autrata et al., 2012). Important details of the methodology are missing, including the ACCEPTED MANUSCRIPT 1561 method of randomization. The primary outcome was absence of recurrence of stage 3 plus ROP in either 1562 eye by 55 weeks PMA. No recurrence of neovascularization was observed in 85.4% infants treated with 1563 pegaptanib and laser compared with 50% treated with laser alone (p=0.0197). A “final” unfavourable 1564 structural outcome (stage 4A or 4B) was seen in 10.3% eyes treated with pegaptanib plus laser compared 1565 with 39.3% eyes treated with laser alone (p=0.0149). This latter figure is in stark contrast to the 9% 1566 unfavourable structural outcomes at 9 months in the ET-ROP trial laser early treatment group (Good 1567 2004). The idea of using a low dose of ranibizumab followed by laser therapy after an interval to allow 1568 for growth of the eye has yet to be subject to trials. 1569 1570 Since publication of the BEAT-ROP trial there have been several reports highlighting longer term 1571 concerns following treatment with intravitreal bevacizumab. Bevacizumab in the higher dose seems to 1572 suppress all angiogenic activity in some eyes so that the periphery remains avascular. This has resulted in 1573 reported late recurrences of ROP, which have progressed to retinal detachment, occurring from 4 months 1574 (Jang et al., 2010, Hu et al., 2012) to a year after injection (Ittiara et al., 2013). It is, therefore, imperative 1575 that infants who receive such therapy have careful ophthalmological review for probably many months. 1576 Mintz-Hittner (Mintz-Hittner et al., 2016) recommended follow up until 65 weeks PMA. We endorse an 1577 unpublished recommendation of weekly follow up until 50 weeks PMA, then every two weeks till 60 1578 weeks and then every 3 weeks till 65-70 weeks PMA. (PersonalMANUSCRIPT communication, Tom Lee; (Mintz- 1579 Hittner et al., 2016). In addition, Isaac and colleagues (Isaac et al., 2016) reported that about 3 times more 1580 follow up visits were required in the first year for IVB-treated infants compared to laser-treated infants. 1581 1582 The timing of intravitreal injection is also important as emphasized by Mintz-Hittner (Mintz-Hittner, 1583 2009). Zepeda-Romero and colleagues (Zepeda-Romero et al., 2010) reported a “paradoxical response” to 1584 bevacizumab injection for severe ROP with subsequent fibrosis and need for vitrectomy. This response 1585 may be delayed up to four months after treatment (Lee et al., 2012). A number of cases have been 1586 reported where the process of fibrosis and retinal detachment appears to have been accelerated after 1587 bevacizumab injection (Honda et al., 2008), colloquially known as ‘Avastin crunch’ in diabetes (Arevalo 1588 et al., 2008). In proliferative membranes removed by vitrectomy after intravitreal bevacizumab for 1589 proliferative retinopathy,ACCEPTED fibrosis was accelerated by upregulation of transforming growth factor Beta 2 1590 and connective tissue growth factor. (Zhang et al., 2016). Bevacizumab is reported to stimulate fibroblast 1591 activity in vivo (Zhang et al., 2015), therefore its use is contraindicated once vitreoretinal traction 1592 commences. Recurrence of ROP followed by total retinal detachment has also been reported after 1593 ranibizumab injection (Jang et al., 2010). 1594

47 1595 Dosage considerations ACCEPTED MANUSCRIPT 1596 The dose of bevacizumab chosen for the BEAT-ROP study, half the adult dose, was arbitrary and 1597 probably excessive (Darlow et al., 2013a, Hard and Hellstrom, 2011, Spandau, 2013)Avery 2012), given 1598 that the infant’s body weight is 1/50th that of an adult, the vitreous volume is less than half, and retinal 1599 surface area around one-third (author’s calculations- Darlow et al 2013) . Sears (Sears, 2008) has 1600 estimated that 0.5-1mg intravitreal bevacizumab is 10,000 times that needed to neutralize VEGF in the 1601 vitreous in cases of ROP, and will lead to systemic concentrations 1,000 fold higher, on a molar basis, 1602 than serum VEGF-A. Case studies reporting good results using a lower dose of bevacizumab are 1603 beginning to appear (Harder et al., 2011, Spandau, 2013, Han et al., 2016, Khodabande et al., 2016). Two 1604 separate small case series comparing eyes injected with low dose (0.25mg) vs ‘conventional’ dose 1605 (0.625mg) bevacizumab found no difference in outcomes (Han et al., 2016, Kong et al., 2015b). 1606 1607 The advantage of the lower dose appears to be that the pathological neovascularization is still suppressed 1608 but normal peripheral angiogenesis can still occur. In a single case report, a dose of 0.16mg of 1609 bevacizumab was shown to be effective (Connor et al., 2015). Anecdotally, many ophthalmologists are 1610 now using the lower dose of 0.25mg in 0.01ml of bevacizumab with good results. A phase 1 dosing study 1611 of bevacizumab showed doses as low as 0.031mg were effective in treating ROP (Wallace et al 2017). 1612 A human study of the pharmacokinetics of lower doses of bevacizumab is underway (ROP1: 1613 NCT02390531) as is a small clinical study of ranibizumabMANUSCRIPT treatment (40 patients) and which will have 1614 five year follow-up (CARE-ROP:NCT02134447). An industry funded and organized international, 1615 multicenter trial (RAINBOW: Ranibizumab compared with laser therapy for the treatment of Infants 1616 Born With retinopathy of prematurity; NCT02375971) will enroll 300 infants <1500g and compare two 1617 different doses of ranibizumab (0.2mg, 0.1mg) with laser therapy. A five-year follow-up is planned 1618 (RAINBOW extension; NCT02640664). 1619 1620 Intravitreal bevacizumab escapes into the systemic circulation ((Hard and Hellstrom, 2011, Wu et al., 1621 2015); reviewed in (Darlow et al., 2013a)) (see Table 1 above). Miyake (Miyake et al., 2010) measured 1622 serum concentrations after a single injection of 1.25 mg bevacizumab in adult macaques. A peak 1623 concentration of 1430±186 ng/ml was reached at one week but concentrations were still half of this at 8 1624 weeks. Bevacizumab wasACCEPTED also detected in the fellow eye for the first week. Sato (Sato et al., 2012) 1625 measured both bevacizumab and VEGF concentrations in the system circulation in 11 infants who 1626 received 0.25 mg or 0.5 mg bevacizumab to one or both eyes having previously been treated with laser 1627 therapy. Serum bevacizumab reached a concentration of 1241 ng/ml at two weeks and there was a 1628 significantly correlated fall in the concentrations of serum VEGF concentrations to 269 pg/ml at this time. 1629 (Note that 1ng = 1000 pg). Lee and colleagues (Lee et al., 2011b) measured serum VEGF concentrations 1630 up to 8 weeks after treatment with only intravitreal bevacizumab, 0.62 mg per eye, in 11 infants. Pre-

48 1631 injection serum VEGF concentrations (SD) were 2050 (3010) pg/ml before injection, falling to 170 (100) ACCEPTED MANUSCRIPT 1632 pg/ml at 1 week and 50 (10) pg/ml at 7 weeks post-injection. By comparison, Pieh (Pieh et al., 2008) 1633 reported systemic median concentrations of VEGF at 32 weeks PMA in infants with untreated ROP and 1634 with no ROP to be 904 pg/ml and 658 pg/ml respectively. To add to reports from adults of clinical effects 1635 in the fellow eye following intravitreal bevacizumab injection (Avery et al., 2006), Karaca and colleagues 1636 (Karaca et al., 2013) reported three cases of regression of ROP in the fellow eye after unilateral 1637 monotherapy with bevacizumab. It is not only bevacizumab that will escape into the systemic circulation 1638 after intravitreal injection. Hoerster (2012) injected 0.2 mg ranibizumab into each eye of an infant at 33 1639 weeks PMA and reported serum VEGF levels fell to below the assay limit (2 pg/ml) at 3 weeks before 1640 recovering to pre-operative concentrations by 4 weeks. 1641 1642 Long term systemic effects of anti-VEGF treatment 1643 The editorial accompanying the BEAT-ROP study (Reynolds, 2011) expressed the opinion that “it seems 1644 reasonable to assume that intravitreal bevacizumab is safe” but this view is not shared by a majority of 1645 commentators. To date there have been few studies which have systematically examined long term 1646 developmental outcomes of very preterm infants treated with anti-VEGF agents. One report of 13 1647 consecutive patients treated with 1.25 mg of intravitreal bevacizumab suggested no evidence of 1648 neurodevelopmental problems up to five years of age (Martinez-Castellanos et al., 2013). Araz-Ersan 1649 (Araz-Ersan et al., 2015) reported no difference in neurodevelopmentalMANUSCRIPT outcomes between matched IVB 1650 and laser treated patients. 1651 1652 By contrast a study from a single unit in Taiwan compared two year outcomes for infants treated with 1653 laser alone (33), bevacizumab alone (12) a combination of laser and bevacizumab (16) (Lien et al., 2016); 1654 Compared with laser treatment alone, infants treated with intravitreal bevacizumab had lower mean 1655 mental developmental and performance developmental index scores (Bayley-II), and compared with the 1656 group receiving a combination of bevacizumab and laser the difference was significant. Most recently, a 1657 retrospective comparison of neurodevelopmental outcome at 18-22 months of age in infants of <29 weeks 1658 gestation in the Canadian Neonatal Network (CNN) dataset born in 2010-11, who had acute ROP treated 1659 with either standard laser therapy or intravitreal bevacizumab injection has been reported (Morin et al., 1660 2015). Ninety-eight infantsACCEPTED received laser treatment and 27 bevacizumab, in most cases as monotherapy. 1661 Bevacizumab treated infants had significantly lower Bayley-III composite motor scores and more 1662 frequently had neurodevelopmental impairment. After adjustment for confounding factors, bevacizumab 1663 treated infants still had severe neurodevelopmental impairment significantly more frequently than laser 1664 treated infants. One difficulty in drawing inferences from this observational study is the fact bevacizumab 1665 treated infants scored higher on a measure of severity of illness at admission and were more likely to have 1666 Zone I disease, both of which may be independently associated with adverse outcomes.

49 1667 ACCEPTED MANUSCRIPT 1668 An additional factor is that the few studies reported to date on neurodevelopmental outcomes often do not 1669 state the dosage of bevacizumab used. Many centers using such therapy are now favouring a lower dose 1670 (often 0.3mg) than that used in the BEAT-ROP study (0.625 mg) 1671 1672 But absence of evidence of harm does not mean that no harm occurs. Le Cras (Le Cras et al., 2002) 1673 treated newborn rats with a single systemic dose of the VEGF receptor inhibitor Su-5416. The lungs of 1674 treated pups developed reduced septation and larger air spaces, together with pulmonary hypertension 1675 (Figure 7). What is more, these changes persisted into adulthood and the histopathological features are 1676 hard to distinguish from those seen in bronchopulmonary dysplasia. In the BEAT-ROP study (Mintz- 1677 Hittner et al., 2011) four of the five deaths in the bevacizumab treated group and one of the two in the 1678 laser treated group were from respiratory causes and it is clearly not possible to know whether or not 1679 bevacizumab contributed to this difference. 1680 1681 VEGF has multiple actions in many organs beyond vasculogenesis and angiogenesis (Ferrara, 2004, 1682 Tolentino, 2011, Simon et al., 1995, Zachary, 2005, Voelkel et al., 2006, Sentilhes et al., 2010). Very 1683 preterm infants are born before organogenesis is complete, indeed many organs such as the eye are still 1684 undergoing major growth and development for many months to years. Many of these infants have a 1685 complicated medical course and the associated early morbiditiesMANUSCRIPT (bronchopulmonary dysplasia, 1686 necrotizing enterocolitis, intra-ventricular haemorrhage and periventricular leukomalacia) are well 1687 described, as are problems of developmental delay and learning difficulties in middle-childhood. The 1688 potential systemic adverse effects of anti-VEGF therapy may contribute to, or mimic, all of these 1689 disorders and so may not be easy to prove or refute without large multicenter randomized trials. We have 1690 previously estimated that 742 infants would be required to have 90% power to detect an absolute 10% 1691 increase or decrease in the composite outcome of death or moderate/severe disability at 2 years of age 1692 (Darlow et al., 2013a). 1693 1694 Several commentators have called for further animal studies and RCTs (Fleck and Stenson, 2013, 1695 Wallace and Wu, 2013), pharmacokinetic studies in human infants, and a registry of cases treated with 1696 anti-VEGF agents. A nationalACCEPTED registry of ROP was initiated in Sweden in 2006 (SWEDROP) and has 1697 reported details of infants treated with anti-VEGF agents since the first use of this therapy in 2008 1698 (Holmstrom et al., 2016).At present SWEDROP does not include long-term follow-up information, 1699 however many of the large neonatal networks do include mode of ROP treatment and some also collect at 1700 least a minimum follow-up dataset at 24-36 months of age (Shah et al., 2014). If the findings from the 1701 Canadian Neonatal Network reported above (Morin et al., 2015) are confirmed in other networks or other 1702 reports it should lead to much greater caution in the use of this therapy (Quinn and Darlow, 2016).

50 1703 ACCEPTED MANUSCRIPT 1704 Ocular complications of anti-VEGF treatment 1705 Endophthalmitis is rare after intravitreal anti-VEGF injections. In a meta-analysis (McCannel, 2011), 1706 McCannel reported a rate of 0.05% in an adult population, with an increased percentage of Strep species 1707 compared with post cataract endophthalmitis suggesting oropharyngeal droplet transfer during injection; 1708 he recommended routine sterile precautions including the wearing of surgical masks to minimize the risk 1709 of oropharyngeal transmission. In patients given intravitreal injections for retinal disease, post-operative 1710 antibiotic drops did not change the endophthalmitis rate (Geloneck et al., 2014). Reported ocular 1711 complications of injection for ROP include retinal breaks causing rhegmatogenous retinal detachment, 1712 vascular attenuation, optic atrophy and choroidal rupture (Jalali et al., 2013). 1713 1714 Long-term ophthalmic outcomes of anti-VEGF treatment for ROP 1715 There appears to much less induced myopia with IVB than laser; Harder and colleagues (Harder et al., 1716 2013) and Chen and colleagues (Chen et al., 2014) reported less myopia in their IVB treated group than 1717 the laser treated groups. In the BEAT-ROP study, more myopia and very high myopia (>-8.00D) 1718 occurred in the laser treated group (Geloneck et al., 2014). 1719 1720 Vitrectomy – structural and visual outcomes 1721 Once retinal detachment occurs, the visual outcomes areMANUSCRIPT generally poor (Quinn et al., 1991, Seaber et al., 1722 1995, Trese and Droste, 1998, Shah et al., 2009b, Singh et al., 2012), with the best anatomical success 1723 rates at around 50% for stage 5 disease (Gadkari et al., 2015), but frequently as low as 15-25% (Shah et 1724 al., 2009b, Yu et al., 2006, Choi et al., 2011). Poor results and severe complications have been reported, 1725 even with lens-sparing vitrectomy (Yu et al., 2006). There may be a role for intravitreal bevacizumab (Xu 1726 et al., 2013) prior to surgery to reduce vascular activity, which allows surgery to be carried out earlier 1727 with shorter operative times. Lens-sparing vitrectomy may have a role to play for Stage 4a ROP with 1728 some encouraging vision results initially reported with grating acuity (Repka et al., 2006) but long-term 1729 vision results remained poor (Repka et al., 2011). Late retinal detachment can occur years later in 1730 children with cicatricial ROP with generally poor visual outcomes (Park et al., 2004). Good form vision 1731 was reported after surgeryACCEPTED for Stage 5 disease (Mintz-Hittner et al., 1997) but the functional benefit to 1732 such children, often in the setting of neurological disease affecting vision, is debatable. Untreated adults 1733 with “marked posterior segment abnormalities secondary to ROP” have been shown to have good 1734 corrected visual acuities (Ferrone et al., 1998) so the lack of controls in most anecdotal case series of 1735 retinal detachment surgery makes the benefits difficult to put into a proper perspective. 1736

51 1737 Quinn and colleagues (Quinn et al., 1991) concluded that the poor visual outcomes from retinal ACCEPTED MANUSCRIPT 1738 detachment surgery “suggest that efforts are well-spent in attempting to prevent retinal detachment in 1739 ROP”. Little has changed in the intervening years. 1740 1741 Effects of premature birth on retinal development/function: Impact on larger 1742 premature infants 1743 There are a number of factors leading to ROP blindness in less mature infants in high income countries 1744 compared to more mature infants in low and middle income countries. In the latter regions, both primary 1745 and secondary prevention pathways may not be employed robustly, resulting in failures in terms of 1746 optimal neonatal intensive care procedures. In addition, as Blencowe et al (2013) document the mean 1747 incidence of any ROP for infants <32 weeks (from 42 unit based studies 2000-10 in low and middle 1748 income countries) is 36.5% or nearly twice that from 13 population based high-income country studies at 1749 21.8%. In the more mature infant, exposure to 100% oxygen for hours will likely lead to dieback of the 1750 retinal vessels with subsequent development of stage 4 or 5 ROP (Shah et al., 2009a). Further, our own 1751 work (Hughes et al., 2000) showed that vascularization of the human retina is far from complete even at 1752 32WG, and the work of Benjamin and colleagues (Benjamin et al., 1998), and Chan-Ling and colleagues 1753 (Chan-Ling et al., 2004b) showed that VEGF is required for endothelial cell survival until immature 1754 retinal capillaries are protected from VEGF withdrawal, by close ensheathment of pericytes. This implies 1755 that any over administration of oxygen to these larger AsianMANUSCRIPT infants will result in ‘delayed 1756 vascularization’ - the initiating event in the pathogenesis of ROP. 1757 1758 However, when there are good detection and treatment programs available, the overall outcome may be 1759 excellent because many of these more mature infants will be less likely to have white matter injury and 1760 other morbidities seen in the micro-preemies cared for in high-income countries. Fielder and colleagues 1761 (Fielder et al., 2015) extend the data from Blencowe (Blencowe et al., 2013) to look at specific features of 1762 visual morbidity in the middle-income setting where 80% of children with severe visual morbidity from 1763 ROP are found. 1764 1765 ROP and NeurodevelopmentACCEPTED 1766 This is a complex area and a full treatment of the topic is beyond the scope of this review. There has been 1767 a recent review of key studies that has addressed some of the issues involved (Beligere et al., 2015). 1768 Discussion of neurodevelopmental outcome for infants with ROP are complicated by the fact that the 1769 infants at highest risk of ROP are also at risk of other morbidities of prematurity, including white-matter 1770 injury (WMI) and neurodevelopmental delay (NDD). Cerebral damage, associated with WMI may indeed 1771 be the commonest cause of visual impairment in very preterm children in high income countries (Rahi et 1772 al., 2003); this was confirmed in a more recent cohort (Slidsborg et al., 2012). Schmidt and colleagues 52 1773 (Schmidt et al., 2014) undertook a secondary analysis of data from the Caffeine for Apnea of Prematurity ACCEPTED MANUSCRIPT 1774 (CAP) trial, where 1582 children (95 with severe ROP) were assessed at 5 years of age. Of children with 1775 severe ROP, 39.5% had at least 1 nonvisual disability and 14.9% two, compared with 15.5% and 2.4% 1776 other children (P< .001 in both cases). ROP and NDD are more common with increasing immaturity, and 1777 oxygen toxicity and inflammation are understood to be contributing factors in the pathogenesis of both 1778 these and other morbidities of prematurity (Chen et al., 2011). 1779 1780 Several studies have reported that VLBW is associated with increased rates of myopia, squint and 1781 amblyopia as well as cortical (cerebral) visual impairment in early (O'Connor et al., 2002) and middle 1782 childhood (Darlow et al., 1997, Holmstrom and Larsson, 2008, Fledelius, 1996b, Fledelius, 1996a) 1783 reviewed by Holmstrom & Larsson (Holmstrom and Larsson, 2013). Two reports from a prospective 1784 study of ROP (infants born in 1985-7) assessed visual outcomes at 10-14 years (O'Connor et al., 2002, 1785 Stephenson et al., 2007) and noted that 50% had adverse visual outcomes, principally reduced acuity, 1786 myopia or strabismus, but children with regressed mild ROP had no increased risk of ophthalmic 1787 problems. In addition, children with adverse visual outcomes also had worse neuropsychological 1788 outcomes and there is likely to be a significant association between visual and neurological development. 1789 1790 Ex-preterm infants also have abnormal higher visual processing that result in difficulties with visual 1791 scenes (for example, finding a toy against a patterned background),MANUSCRIPT visual guidance (for example, 1792 walking over uneven ground) and recognition and orientation, all manifestations of cerebral visual 1793 impairment (Dutton, 2009, Ortibus et al., 2011) 1794 1795 A recent study suggests that much of the demonstrable association between severe ROP and NND is the 1796 result of common shared risk factors. In a cohort of 17-18 year old former extremely low birth weight 1797 (ELBW: birth weight <1000g) infants from Victoria, Australia, Molloy and colleagues (Molloy et al., 1798 2015) found those with a history of severe ROP had poorer test performance including on visual 1799 processing and visual-motor integration as well as reading, arithmetic and IQ compared with those 1800 without severe ROP. They also reported that after controlling for neonatal factors (severe WMI, post- 1801 natal corticosteroids and neonatal surgery) the differences between the groups diminished with only 1802 poorer visual acuity remainingACCEPTED significant. 1803 1804 Arguably one of the most important developmental outcomes of prematurity is overall quality of life. 1805 Quinn and colleagues (Quinn et al., 2004) reported health related quality of life scores (HRQL – a perfect 1806 score being 1) as assessed by parents at 10 years of age in 244 infants randomized in the Cryotherapy for 1807 Retinopathy of Prematurity (CRYO-ROP) trial compared with 102 CRYO-ROP participants who did not 1808 develop ROP. Children in the randomized cohort who were bilaterally blind or had low vision rarely had

53 1809 scores above 0.65 (possible scores 0 to 1.00 ranging from death to perfect health). However, randomized ACCEPTED MANUSCRIPT 1810 children who were sighted in at least one eye had most scores above 0.65 with a distribution very similar 1811 to the scores for children without ROP. 1812 1813 Recent trials of oxygen saturation targeting 1814 One of the most enduring questions in neonatology has been how much oxygen should preterm infants be 1815 exposed to in order to maximise morbidity-free survival and avoid the competing adverse outcomes of 1816 death, neurodevelopmental impairment, bronchopulmonary dysplasia (BPD) and retinopathy of 1817 prematurity (ROP) (Tin and Wariyar, 2002, Silverman, 2004). From the 1990s onwards, most 1818 neonatologists have used the rather indirect measurement of oxygen saturation as measured by continuous 1819 pulse oximetry (SpO2) as the preferred means of monitoring oxygen exposure because it is both simple to 1820 undertake and non-invasive (Hay et al., 1991, Darlow and Morley, 2015) and this question has 1821 crystallised to “what is the most appropriate SpO2 target for very preterm infants”. 1822 1823 In 2003 an international group of neonatologists came together to discuss this issue (Cole et al., 2003). As 1824 a result of these discussions, five randomized controlled trials, all of which planned to compare a SpO2 1825 target of 85-89% with 91-95% (being the lower and higher ends of the commonly adopted range) in 1826 infants of <29 weeks gestation, were funded. From the outset there was agreement to pool data in a 1827 prospective individual patient data (IPD) meta-analysis that would include 5000 infants (Askie et al., 1828 2011). The primary outcome for these trials was combined MANUSCRIPT death or neurodevelopmental impairment at 1829 18-24 months but both severe ROP and longer-term visual impairment were important secondary 1830 outcomes. 1831 1832 The early results from these trials (collectively involving 4965 infants) have been reviewed (Askie, 2013, 1833 Fleck and Stenson, 2013, Saugstad and Aune, 2014). All five trials (SUPPORT; COT; BOOST-NZ; 1834 BOOST II Australia; BOOST UK) have now reported outcomes at 18-24 months with no significant 1835 differences between the target groups in the combined outcome of death or neurodevelopmental 1836 impairment, or in severe visual impairment (Vaucher et al., 2012, Schmidt et al., 2013, Darlow et al., 1837 2014, Groups, 2016). Even though the higher target was significantly associated with an increased risk of 1838 severe ROP on meta-analysisACCEPTED of all five studies (Saugstad and Aune, 2014, Askie et al., 2017), severe 1839 visual impairment from retinal causes at 18-24 months was low in all groups at 1% or less (Vaucher et al., 1840 2012, Schmidt et al., 2013, Darlow et al., 2014), which suggests that in these high-income countries 1841 screening and treatment was largely successful. 1842 1843 The United States based SUPPORT study, which also randomized infants to early continuous positive 1844 airways pressure (CPAP) or intubation and surfactant using a 2-by-2 factorial design, had earlier reported

54 1845 outcomes at hospital discharge (Network et al., 2010). Mortality was significantly increased in the lower ACCEPTED MANUSCRIPT 1846 target group (19.9% vs 16.2%: RR 1.27; 95% CI 1.01-1.60) (P<0.05). As a result of this finding, 1847 recruitment to two other trials (BOOST II Australia and BOOST UK) was terminated prematurely 1848 (Johnston et al., 2011). These two trials, together with the COT trial, had also upgraded the pulse 1849 oximeter software to eliminate a small (2%) over-estimation of SpO2 in the 87% to 90% range 1850 approximately half-way through recruitment (Johnston et al., 2011, Schmidt et al., 2013). Combined data 1851 from the three BOOST II trials at hospital discharge showed increased mortality in the lower target group 1852 in the 1117 infants cared for with the upgraded monitors (23.2% vs 15.8%: RR 1.47; 95% CI 1.16-1.86) 1853 (P<.001) (Schmidt et al., 2013). Subsequent interim meta-analysis of then published data (4911 infants; 1854 including hospital mortality for BOOST II Australia and BOOST UK and 18 – 24 month data for 1855 SUPPORT, COT and BOOST-NZ) confirmed significantly increased mortality in infants randomized to 1856 the lower target group (RR 1.18; 95% CI 1.04-1.34) (Saugstad and Aune, 2014). The recently published 1857 two-year outcome from BOOST II Australia and BOOST UK also showed significantly increased 1858 mortality in infants randomized to the lower target group (pooled data including original and upgraded 1859 monitors, RR 1.20; 95%CI 1.01-1.43) (P=.04) (Groups, 2016) 1860 1861 The just completed Cochrane review (Askie et al., 2017) confirms that increased mortality is associated 1862 with the lower target (RR 1.16 [95%CI 1.03, 1.31], number needed to treat to result in one additional 1863 death being 31 [95%CI 16, 168]), but a reduced risk of MANUSCRIPTROP treatment (RR 0.72 [95%CI 0.61, 0.85], 1864 number needed to treat to prevent one additional case being 34 [95%CI 21, 63]) although there was 1865 moderate heterogeneity (I 2 statistic = 69%) for this outcome largely because of a much greater difference 1866 between the groups in one trial (SUPPORT). Whilst further information may become available from the 1867 awaited individual patient data meta-analysis (Askie et al., 2011) these trials have already changed 1868 practice such that the drift to use ever lower saturation targets, principally to avoid ROP and BPD, has 1869 ceased and we now know that a SpO2 target range of 85-89% is too low in these vulnerable infants. What 1870 is less certain still is what target to adopt with most commentators recommending the higher target used 1871 in these RCTs of 91-95%, at least in these very premature infants (Polin and Bateman, 2013, Saugstad 1872 and Aune, 2014, Bancalari and Claure, 2013). There are concerns, however, that adoption of this 1873 recommendation may result in another epidemic of hyperoxia, particularly in lesser resourced areas where 1874 careful monitoring is oftenACCEPTED hard to achieve (Darlow and Morley, 2015, Sola et al., 2014). Perhaps 1875 reflecting these concerns the American Academy of Pediatrics (AAP) has stated “Recent RCTs suggest 1876 that a targeted oxygen saturation range of 90% to 95% may be safer than 85% to 89%, at least for some 1877 infants – however, the ideal oxygen saturation range for extremely low birth weight infants remains 1878 unknown.” (Cummings Pediatrics 2016). 1879

55 1880 It is important to recognise that recommendations about alarm limits are expert opinion only because ACCEPTED MANUSCRIPT 1881 these have not been studied in randomized controlled trials. However, prior to these recent oxygen 1882 saturation targeting studies the AVIOX study found target compliance was better when the high alarm 1883 was 1% above the target and low alarm 1-2% below (Hagadorn et al. Pediatrics 118:1574, 2006). The 1884 AAP follow the BOOST-NZ (Darlow 2014) suggestion that the high alarm be set at 95%. In high 1885 resource settings for very premature infants we recommend an oxygen saturation target of 90-94% with 1886 alarms set at 95% and 89%. In poorer resource settings where more mature infants predominate, the high 1887 alarm should still be set at 95% but it may be reasonable for the lower margin of the saturation target to 1888 be slightly wider. (WHO 2016) 1889 1890 Gradual oxygen withdrawal as strategy to minimize Phase 2 ROP pathology – 1891 Evidence from the kitten model of ROP 1892 Our earlier study (Chan-Ling et al., 1995b) showed that the proliferative vasculopathy in Phase 2 of the 1893 kitten model of ROP can be significantly reduced by a regimen of supplemental oxygen therapy (SOT - 1894 i.e. delayed O2 withdrawal), where the aim is to attempt to mimic physiological levels of hypoxia in the 1895 inner retinal layers during the vascularization of the retina in Phase 2 of ROP pathogenesis. Our 1896 experimental studies led us to conclude that optimal revascularization requires a balance between the rate 1897 and quality of revascularization. We were able to demonstrate that SOT allowed astrocyte survival and 1898 prevented overexpression of VEGF. Empirical evidence was provided in our study detailed in (Chan-Ling 1899 et al., 1995b)and is further supported by VEGF in-situ hybridization.MANUSCRIPT 1900 1901 VEGF in-situ hybridization studies support the pathogenetic mechanism of ROP. Figure 8A-B (from 1902 (Stone et al., 1996) shows the normal expression of VEGF during normal development of the cat retina 1903 via in-situ hybridization at postnatal day (P) 4, where VEGF expression can be determined in dark-field 1904 illumination by a white hybridization product detectable close to the inner surface of the retina, internal to 1905 the inner plexiform layer. Figure 4B shows a similar region of retina to 8A in bright-field illumination at 1906 P3, to demonstrate the location of VEGF expression in the inner retina. Figure 8C shows a kitten raised 1907 for 3 days in room air, then placed in a high oxygen (70-80%) environment for 24 hours. Suppression of 1908 the normally observed VEGF signal at the inner retinal surface via hyperoxia is evident. Figure 8D shows 1909 a kitten raised in roomACCEPTED air for 24 hours, then placed in 70-80% oxygen for 3 days. The VEGF signal at the 1910 inner retina is completely suppressed via hyperoxia. Figure 8E-G shows the marked increase in VEGF 1911 expression in kittens raised in hyperoxia (70-80% oxygen) for 4 days and returned to room air (21% 1912 oxygen). At 3, 11 and 27 days post-hyperoxic exposure, the VEGF signal is clearly detectable and 1913 remains strong throughout the time course – confirming the 2-phase hypothesis of ROP pathogenesis. 1914 1915 Figure 8H-M shows the localization of VEGF expression via in-situ hybridization, and the corresponding 1916 expression of glial fibrillary acid protein (GFAP) in kittens raised in normoxia (Figure 8H-I), in 11 days 56 1917 post-hyperoxia for 4 days (Figure 8J-K) and 27 days post-hyperoxia for 4 days (Figure 8L). Figure 8H ACCEPTED MANUSCRIPT 1918 shows normal VEGF expression during ‘physiological hypoxia’ while Fig 8J, shows the marked 1919 upregulation of VEGF expression during vaso-proliferative phase (2) of the kitten model of ROP and it’s 1920 gradual resolution by 27 days post hyperoxia exposure. Figure 8M shows GFAP staining in a hyperoxic 1921 retina at 3 days post exposure, again confirming the marked down-regulation of VEGF expression during 1922 the period of hyperoxic retina. 1923 1924 Evidence in support of our conclusion that a regimen of supplemental oxygen therapy (SOT - i.e. delayed 1925 O2 withdrawal), where the aim is to attempt to mimic physiological levels of hypoxia in the inner retinal 1926 layers during the vascularization of the retina in Phase 2 of ROP (Chan-Ling et al., 1995b) comes from 1927 VEGF in-situ hybridization studies. Figure 8N-P (from (Stone et al., 1996)), shows a kitten raised for 4 1928 days in 70% to 80% oxygen and then for 3 days in 45% oxygen, the level shown by Chan-Ling and 1929 colleagues to normalize the formation of vessels for several days after exposure to hyperoxia. At the edge 1930 of the reforming vessels, astrocytes are normal in morphology and in their location in the axon layer, 1931 internal to the ganglion cell layer (Figure 8N), confirming our earlier work (Chan-Ling and Stone, 1992). 1932 Further, the expression of VEGF is normal in its location in the axon layer, superficial to the neurones of 1933 the ganglion cell layer (Figure 8O-P). Finally, the expression of VEGF during SOT is relatively normal. 1934 The pattern of VEGF expression across the edge of the developing vessels is compared with the normal 1935 pattern in Figure 8A. Peak VEGF expression during SOTMANUSCRIPT was found lateral to the edge of the developing 1936 vessels, as in the normally developing retina, and was not elevated. We suggest that SOT prevents 1937 proliferative vasculopathy by preventing overexpression of VEGF, which occurs if the animal is returned 1938 to room air abruptly. However, differences are apparent in VEGF expression between SOT and normal 1939 development. The peak signal appears reduced in SOT, VEGF expression central to the edge of vessels is 1940 relatively low, and expression peripheral to the vessels is relatively high. Since our findings demonstrated 1941 that excessive hypoxia induces the vasculopathy characterised as the initiating event in ROP, while 1942 hyperoxia inhibits the rate of retinal vascularization, the successful implementation of SOT requires 1943 negotiation of a fine line between hyperoxia and hypoxia. The penalty of excessive hypoxia is the 1944 vasculopathy of ROP, with resultant formation of a proliferative, leaky vasculature associated with the 1945 degeneration of astrocytes. The penalty of even a slight level of hyperoxia is that vessel formation is 1946 slowed (though otherwiseACCEPTED free of pathology) delaying revasculaization of the retina. 1947 1948 Translating this (SOT) approach to the clinical situation would not be straightforward. Observational 1949 studies suggest that on average compliance with saturation targets can be achieved on average only 50% 1950 of the time (range 16-64%) (Hagadorn et al., 2006) and that both hpoxemic and hyperoxemic events are 1951 common. Automated systems of adjustment of inspired oxygen are being investigated and in general do 1952 result in better compliance with oxygen saturation targets than manual adjustments but this is still not

57 1953 more than 70-80% of the time and clinical benefits have so far not been demonstrated (Poets and Franz, ACCEPTED MANUSCRIPT 1954 2017). 1955 1956 Low serum levels of Insulin-like Growth Factor-1 (IGF-1) correlate with ROP 1957 severity 1958 IGF-1 is a polypeptide protein hormone essential for fetal development at all stages of pregnancy. The 1959 biological effects of IGF signaling in the CNS involve many aspects of normal brain development (Dore 1960 et al., 1997). IGF-1 and its receptors are abundantly expressed in neuroepithelial cells in the early 1961 developing brain (Bondy et al., 1992), where IGF-1 acts as an autocrine proliferation (Hodge et al., 2004) 1962 and pro-survival factor (Drago et al., 1991). IGF-1 promotes neuronal migration (Hurtado-Chong et al., 1963 2009), neurogenesis and synaptogenesis (O'Kusky et al., 2000). IGF-1 enhances axon outgrowth, thus 1964 contributing to the formation of appropriate circuitries in the brain (Ozdinler and Macklis, 2006). IGF-1 1965 enhances neuron metabolism (Bondy and Cheng, 2002), and modulates neuronal excitability (Carro et al., 1966 2000). These properties, together with IGF-1 anti-apoptotic actions (Carro et al., 2003) and the effects of 1967 anti-oxidant and anti-inflammation (Higashi et al., 2010), are crucial for the ability of IGF-1 to protect 1968 neurons against insults. IGF-1 also influences the formation of astrocytes (Torres-Aleman et al., 1994), 1969 and regulates oligodendrocyte differentiation (Barres et al., 1992). IGF-1 deficiency reduces glucose 1970 utilization by 30-60% in developing mouse brain with great decreasing in structures (Cheng et al., 2000). 1971 1972 Serum levels of IGF-1 rise rapidly during the third trimesterMANUSCRIPT of pregnancy (Gluckman and Harding, 1973 1997). Consequently, premature birth is associated with a rapid fall in serum IGF-1 levels, as maternal 1974 sources of IGF-1 are lost (Lineham et al., 1986). Hellstrom and colleagues have shown that low serum 1975 levels of insulin like growth factor-1 (IGF-1) correlate with ROP severity, and hypothesised that the 1976 duration of low serum IGF-1 levels correlate with the severity of ROP because low IGF-1 prevents 1977 normal vascular development, causing an increase in tissue hypoxia (Hellstrom et al., 2003). A lack of 1978 IGF-1 has been demonstrated to prevent normal retinal vascular growth in mice, despite the presence of 1979 VEGF, as a minimum level of IGF-1 is required for VEGF signalling (Hellstrom et al., 2001, Smith et al., 1980 1999). In very premature infants low IGF-1 concentrations have also been associated with poor post-natal 1981 weight gain (reviewed in (Hellstrom et al., 2013)) and head growth (Lofqvist et al., 2006b). Algorithms 1982 that combine serial IGF-1ACCEPTED concentrations and post-natal weight gain have been used to predict infants at 1983 risk of severe ROP (Hellstrom et al., 2013) although serial weight measurements alone might be 1984 sufficient (Hellstrom et al., 2009, Binenbaum, 2013) 1985 1986 Whilst the correlation between serum IGF-1 levels and ROP severity is supportive of IGF-1 1987 supplementation in premature infants, one problem with such an approach is its short half-life. Lofqvist 1988 and colleagues (Lofqvist et al., 2009) conducted a limited trial of an intravenously administered 3 hour 1989 infusion of an equimolar preparation of recombinant IGF-I and its binding protein, insulin-like growth 58 1990 factor binding protein-3 (rhIGFBP-3) (mecasemin rinfabate; trade name: iPlex) to premature infants at ACCEPTED MANUSCRIPT 1991 postnatal day 3, followed by serial measurements of IGF-I and IGFBP-3 over a 14 day time period 1992 (Hellstrom et al., 2016a). The rationale behind the IGFBP-3 inclusion is that in the perinatal period, 1993 IGFBP-3 extends the IGF-1 half-life by protecting against degradation and prevents passage into the 1994 extravascular compartment. The investigators found that the administration of this preparation was 1995 effective in increasing serum IGF-1 levels to that found in utero at the same age but administered IGF-1 1996 had a mean half-life of only 0.86 hours. 1997 1998 The same investigators (Ley et al., 2013) in a phase II study, established the feasibility of longer term (7 1999 day) administration of IGF-1 to achieve corresponding normal intrauterine IGF levels in infants at 27 2000 WG. There were no side effects of the infusion noted, including no hypoglycaemia. However, the study 2001 also showed that post-infusion, serum IGF-1 levels fell below corresponding intra-uterine levels, 2002 suggesting that a longer infusion period than 7 days is necessary to maintain physiologic IGF-1 2003 concentrations in premature infants. Subsequently a full efficacy trial (NCT01096784) has been carried 2004 out in 121 infants of 23-27 WG who were given mecasemin rinfabate, 250µg/kg, by 24 hour infusion 2005 until 30 weeks PMA. Preliminary results (quoted by (Hellstrom et al., 2016b)) showed no effect on the 2006 primary endpoint of severe ROP but that there was a reduction in bronchopulmonary dysplasia and 2007 intraventricular haemorrhage. The full report must we be awaited before drawing any conclusions. 2008 MANUSCRIPT 2009 Omega 3 supplementation 2010 Connor and colleagues (Connor et al., 2007) studied the effect of the omega-3 fatty acids 2011 eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), derived from fish, and the omega-6 fatty 2012 acid arachidonic acid on the loss of blood vessels, the re-growth of healthy vessels, and the growth of 2013 destructive abnormal vessels in a mouse model of oxygen-induced retinopathy. Increasing omega-3 fatty 2014 acids and decreasing omega-6 fatty acids in the diet reduced the area of vessel loss that ultimately causes 2015 the growth of the abnormal vessels and blindness. Omega-6 fatty acid contributes to the growth of 2016 abnormal blood vessels in the retina. This study results suggests that increasing omega-3 fatty acid intake 2017 in premature infants may significantly decrease the occurrence of ROP. One large trial of increased dose 2018 of the omega-3 DHA from 2 to 40 weeks PMA in 657 infants of <33 weeks gestation (Makrides et al., 2019 2009) in which the primaryACCEPTED outcome was neurodevelopment at 18 months, showed no effect on ROP. A 2020 further RCT of omega-3 long chain polyunsaturated fatty acid (LCPUFA) for the prevention of ROP in 2021 very low birth-weight infants by Hellstrom, Lofqvist and Smith (NCT02486042) is currently underway. 2022 2023 Human Milk 2024 There are many reasons to support early and continued feeding of very preterm infants with mother’s own 2025 breast milk. Several observational studies have suggested human milk may reduce the risk of any 2026 (Hylander et al., 2001) or severe ROP (Okamoto et al., 2007) but not all studies found a positive 59 2027 association (Heller et al., 2007). Manzoni (Manzoni et al., 2013b), in a secondary analysis of two ACCEPTED MANUSCRIPT 2028 multicenter RCTs (aimed at assessing fluconazole prophylaxis and lactoferrin supplementation to reduce 2029 sepsis), reported that exclusive human milk feeding was associated with a significant reduction in the 2030 incidence of any and threshold ROP. A recent systematic review of nine observational studies (Bharwani 2031 et al., 2016) found receipt of any human milk intake was associated with a significant reduction in both 2032 any and severe ROP (Hylander et al., 2001, Okamoto et al., 2007, Heller et al., 2007, Manzoni et al., 2033 2013a). 2034 2035 Novel molecular targets for ROP over the horizon 2036 Propranolol 2037 A chance finding in two children treated with oral propranolol for cardiac conditions that such treatment 2038 led to regression of infantile haemangiomas (IH) (Leaute-Labreze et al., 2008) resulted in rapid adoption 2039 of this treatment worldwide; propranolol is now considered the standard of care for severe and high-risk 2040 IHs (Pride et al., 2013), although such use remains off-label (Drolet et al., 2013). The exact mechanism 2041 of action of propranolol in causing IH regression remains unknown (Storch and Hoeger, 2010) and 2042 despite a favourable safety profile in children there is a long list of potentially serious side effects 2043 including bradycardia, hypotension, cold extremities, hypoglycaemia, and pulmonary symptoms (Drolet 2044 et al., 2013). 2045 MANUSCRIPT 2046 There are several reports of the presence of IHs being associated with ROP (Praveen 2009; Hyland 2013). 2047 and propranolol has been proposed as a potential treatment for the latter (Filippi 2010; Bührer 2015). A 2048 small pilot study of 52 infants with stage 2 ROP without plus in Zone II explored the safety and effects of 2049 oral administration of propranolol (0.25 or 0.5 mg/kg/ 6 hourly)(Filippi et al., 2013). Propranolol 2050 treatment was associated with fewer infants progressing to treatment criteria but 19% of propranolol 2051 treated infants had serious side effects including hypotension and bradycardia. Two other small trials 2052 have also reported fewer infants progressing to treatment in the propranolol versus control group 2053 (Makhoul 2013; Bancalari 2016). Animal work has so far produced conflicting results (Ristori et al., 2054 2011, Chen et al., 2012). Hard and Hellstrom (Hard and Hellstrom, 2011) have noted key roles for 2055 adrenergic receptors and noradrenalin as a neurotransmitter in the brains of animal models and cautioned 2056 that there are many unknownsACCEPTED about beta-adrenergic blockage in the premature infant’s brain. It is worth 2057 remembering that an earlier promising treatment for IH, -interferon, was subsequently shown to cause 2058 spastic diplegia (Michaud et al., 2004). The risk-benefit profile for propranolol treatment of ROP would 2059 need to be carefully established in rigorously conducted animal studies and clinical trials before any 2060 adoption into practice. 2061

60 2062 EPO ACCEPTED MANUSCRIPT 2063 Erythropoeitin (Epo) is another oxygen-regulated growth factor with some similar actions on 2064 angiogenesis to VEGF (Chen and Smith, 2008). Human recombinant erythropoietin (rhEpo) has been 2065 used in preterm infants to prevent anaemia and there is current interest in its possible role in 2066 neuroprotection (Fauchere et al., 2015). Fauchere and colleagues (Fauchere 2015) undertook a large 2067 phase II RCT (450 infants) of 3 doses of rhEpo given in the first 48 hours after birth. No excess mortality 2068 or morbidity was reported and, in particular, concerns that early high dose rhEpo might increase the risk 2069 of ROP were not born out (Ohlsson and Aher, 2012). A further, recent systematic review and meta- 2070 analysis also found no effect of rhEpo administration on ROP (Fang et al., 2016). 2071 2072 Anti-inflammatory therapy and ROP 2073 Sepsis has been associated with an increased risk of ROP in a number of studies. A report from the Israel 2074 Neonatal Network found an odds ratio (OR) of 2.04 (95% CI 1.32 to 3.16) for risk of severe ROP 2075 following early-onset sepsis (Klinger et al., 2010). A nested case control study of 622 infants born at <30 2076 weeks’ gestation and screened for ROP in one institution found any sepsis increased the risk of any ROP 2077 (OR 2.1, 95%CI 1.4 to 3.2) and that the risk was higher in infants of 28-29 weeks’ gestation than more 2078 immature infants (Chen et al., 2011). Candida infection has also been strongly linked with both any and 2079 severe ROP (reviewed in (Bharwani and Dhanireddy, 2008)). 2080 MANUSCRIPT 2081 More generally there is strong evidence for a cascade of events, mediated by various cytokine and 2082 chemokine responses, from maternal infection/inflammation (chorioamnionitis), leading to premature 2083 birth and a fetal inflammatory response syndrome (FIRS) and subsequently to morbidities including peri- 2084 ventricular white matter injury (Yoon et al., 1997, Gomez et al., 1998). Dammann and co-investigators 2085 have proposed similar mechanisms may contribute to ROP (Dammann et al., 2009, Lee and Dammann, 2086 2012). In mouse pups systemic inflammatory distress was induced by an intraperitoneal injection of 2087 lipopolysaccharide, which was demonstrated to induce a marked increase in retinal activated microglia 2088 and to lead to aberrant vascular development with similarities to the pattern seen in ROP (Tremblay et al., 2089 2013). A large observational study of over 800 extremely low birth weight infants (B Wt <1000g) found 2090 the pattern of systemic cytokine and chemokine responses from birth to 21 days differed in infants who 2091 developed ROP comparedACCEPTED with controls (Sood et al., 2010). Sato and colleagues (Sato et al., 2009) 2092 measured vitreous concentrations of 27 cytokines in the vitreous of eyes with stage 4 ROP and found 2093 several cytokines to have significantly higher concentrations compared with control eyes. Notably 2094 concentrations of VEGF were significantly higher in the presence of active vasoproliferative ROP. 2095 2096 Given this background it is not surprising that there have been increasing reports linking treatments that 2097 have anti-inflammatory actions with reduced risks of ROP, although in the main this has not been the 61 2098 primary study outcome. Here we briefly review trials of lactoferrin, pentoxifylline, dexamethasone and ACCEPTED MANUSCRIPT 2099 caffeine, although there is also limited information on a variety of other agents with anti-inflammatory 2100 actions, including cyclooxygenase inhibitors, polyunsaturated fatty acids (see later in this review) and 2101 granulocyte colony stimulating factor (Mataftsi et al., 2011, Mutlu and Sarici, 2013, Beharry et al., 2016, 2102 Shah et al., 2016). 2103 2104 Lactoferrin is an iron binding glycoprotein found in both human and cow’s milk and which has 2105 antimicrobial, antioxidant and anti-inflammatory properties(Lonnerdal, 2009) . In a three-arm RCT of 2106 oral supplements to breastmilk (bovine lactoferrin, bovine lactoferrin plus probiotic, controls) aimed at 2107 prevention of late-onset sepsis, Manzoni and colleagues found that as well as a significant reduction in 2108 the primary outcome, lactoferrin was associated with a significant reduction in ROP requiring treatment 2109 from 11.3% to 3.9% (p=0.02) (Manzoni et al., 2009). Further trials are underway, including a large 2110 multicenter trial in Australia, New Zealand, Europe and North America, which will enroll 1,500 very low 2111 birth weight infants and which has a primary outcome of a composite of all-cause mortality or any of five 2112 morbidities including treated ROP [ACTRN12611000247976].

2113 Pentoxifylline is a phosphodiesterase inhibitor that suppresses TNF-α production and has other anti- 2114 inflammatory actions (Haque and Pammi, 2011). A meta-analysis of six RCTs investigating the effect of 2115 intravenous pentoxfylline as an adjunct to antibiotic therapy to prevent mortality or morbidity from 2116 suspected or proven sepsis found pentoxifylline was associated MANUSCRIPT with a significant decrease in all-cause 2117 mortality (Pammi and Haque, 2015) but only 416 infants have been studied and the evidence is low- 2118 quality. Only one trial assessed risk of ROP as a secondary outcome with no benefit seen from 2119 pentoxifylline compared with antibiotics alone (Pammi and Haque, 2015).

2120 Corticosteroids are widely used for their anti-inflammatory properties in all age groups (Barnes, 2006). In 2121 the 1980-90s corticosteroids, particularly dexamethasone, were frequently prescribed for preterm infants 2122 who were ventilated to facilitate extubation and reduce the risk of chronic neonatal lung disease (CLD) 2123 (Halliday et al., 2009). Meta-analysis of 28 trials of postnatal corticosteroids commenced within the first 2124 week of life also showed such therapy reduced the risk of any and severe ROP (Halliday et al., 2009). 2125 However, following a report linking early short course of postnatal corticosteroids with a significant 2126 increased risk of cerebralACCEPTED palsy their use was more restricted (Shinwell et al., 2000). Currently such 2127 therapy is more likely to be commenced after the first week of life when an increase in the risk of ROP, 2128 but not blindness, has been reported (Doyle et al., 2014). A recent analysis of 20 studies demonstrated 2129 that there is a trade-off between the risk of death or cerebral palsy and CLD that may aid clinicians in 2130 choosing whether to use such therapy, but the risks of ROP were not considered (Doyle et al., 2014).

2131 Caffeine is a methylxanthine, which acts as a respiratory stimulant and has been used since the 1970s to 2132 treat apnoea of prematurity (Beharry et al., 2016) . Amongst other actions caffeine may have some anti- 62 2133 inflammatory effects (Beharry et al., 2016). In the landmark Caffeine for Apnoea of Prematurity (CAP) ACCEPTED MANUSCRIPT 2134 trial involving over 2,000 very low birthweight infants, Schmidt and colleagues showed that caffeine 2135 given to prevent apnoea not only reduced the rate of death or disability at 18-21 months of age but also 2136 significantly reduced the risk of severe ROP (OR 0.61; 95% CI 0.42 to 0.89) (Schmidt et al., 2007). The 2137 mechanisms behind this reduction in ROP are unclear although caffeine treated infants were less likely to 2138 suffer chronic lung disease and were possibly more stable overall.

2139 Other possibilities for treatment of ROP in the future might include tyrosine kinase inhibitors and

2140 targeting splicing variants of VEGF so that VEGF 165 is inhibited but not VEGF 165b , which is both anti - 2141 angiogenic and neuroprotective (Wang, 2016), but these and other therapies still require extensive 2142 investigation in animal models for abnormalities in both the eye and other organs as a first step. 2143 2144 Recommendations for current best practice for extremely premature infants 2145 At our present state of knowledge, peripheral retinal ablation using laser photocoagulation remains the 2146 current treatment standard for severe acute phase ROP. The evidence is strong for Type 1 ROP in zone II, 2147 though this treatment which is confined to the eye is less effective in zone I disease. 2148 2149 Until more research into the use of anti-VEGF is available on both ocular and systemic effects of such 2150 treatments, use of these drugs should be considered in the setting of clinical investigations (preferably 2151 trials) or as a rescue therapy when conventional treatment MANUSCRIPT has failed or where child is so unwell 2152 systemically that the time taken and the manipulation necessary for laser becomes a risk to the child. This 2153 latter situation occurs relatively frequently in children with AP-ROP or with Type 1 ROP in zone I who 2154 are often very sick from respiratory disease when treatment for ROP becomes necessary. In this 2155 circumstance, treatment with anti-VEGF may well be the preferable option. Treatment with anti-VEGF 2156 drugs must be preceded by full and frank informed consent with the parents. The evidence is now strong 2157 that low dose bevacizumab (eg 0.3mg or less) is effective so the systemic risks should theoretically be 2158 lowered. In regions of the world with increasing survival of premature infants and limited access to 2159 ophthalmologic care, emphasis should be placed on both developing ROP expertise among 2160 ophthalmologists and providing the best neonatal care possible given the needs of the infant and resource 2161 needs and availability.ACCEPTED 2162 2163 An unknown issue is whether or not to laser the remaining avascular area in the retinal periphery after 2164 successful primary treatment with bevacizumab and whether laser and intravitreal anti-VEGF drugs 2165 should be used simultaneously as the primary treatment in some cases. What is also undetermined at the 2166 present time is how to decide on second line treatment after either failed laser or anti-VEGF treatment. In 2167 this circumstance, careful consideration needs to be given to avoiding the risk of accelerating fibrosis

63 2168 after bevacizumab by preferably using laser; timing of treatment is still crucial here with treatment ACCEPTED MANUSCRIPT 2169 necessary within a short time frame, probably no longer than 72 hours (authors recommendation).. 2170 2171 Telemedicine offers the possibility of more widespread and standardized ROP screening of infants at risk. 2172 Standard protocols for imaging and grading, training and accreditation of screeners, quality control by 2173 way of regular audit are all essential as this technology is adopted. Telemedicine should be considered an 2174 adjunct rather than a replacement for the examination of the fundus by ophthalmologists with specialist 2175 ROP expertise when resources both financial and personnel permit. Use of telemedicine as a standalone 2176 tool to identify those eyes that need to be seen and treated by an ophthalmologist requires further study. 2177 2178 Non-ophthalmic strategies for prevention of, and treatment for ROP 2179 The current epidemic of ROP in middle-income and developing countries is a mix of inadequate or no 2180 monitoring of oxygen and developing neonatal care. The landmark publication by Gilbert and colleagues 2181 (Gilbert et al., 2005) showed that in this setting many infants larger than 1500g and with gestations 32 2182 weeks or more develop severe ROP, a situation not experienced in higher income countries for several 2183 decades. Primary prevention of ROP remains key and improvements in perinatal care have the greatest 2184 potential to reduce the incidence of both any ROP and ROP progressing to an advanced stage and 2185 requiring treatment. As reviewed by Darlow and colleagues (Darlow et al., 2013b) many evidence-based 2186 interventions that decrease mortality and morbidity are notMANUSCRIPT expensive and are easily adopted, including 2187 wider use of ante-natal corticosteroids prior to preterm delivery; and in the labour ward delaying cord 2188 clamping by 30-60 seconds (which increases the haematocrit and decreases the need for transfusions) 2189 (Brocato et al., 2016), using plastic occlusive wraps for the infant to prevent hypothermia and avoidance 2190 of 100% oxygen at resuscitation. In the neonatal unit hand washing (or using alcohol based hand 2191 sanitizers) by all on entering and before handling each infant, together with avoidance of unfocussed use 2192 of wide-spectrum antibiotics, have the potential to greatly reduce nosocomial infection rates. Breast milk 2193 feeding must be supported, and pain prevented and adequately treated. Oxygen monitoring is essential 2194 and does require equipment but also a clear protocol on how to respond to alarms. Lastly, there are more 2195 costly changes that will bring improvement in the longer term including regionalisation of perinatal care, 2196 adequate neonatal transport systems and provision of sufficient trained nursing staff. 2197 ACCEPTED 2198 In highly developed countries, neonatal networks report marked center variations in major morbidities, 2199 including ROP, and demonstrate that it is possible to further reduce the rate of severe ROP to that of the 2200 best performing units (Darlow et al., 2005, Thomas et al., 2015, Lee et al., 2009). Quality improvement 2201 projects focusing on education about ROP as well as compliance with oxygen saturation targets have 2202 been associated with a reduced rate of ROP (Ellsbury and Ursprung, 2010) and comprehensive 2203 multifaceted programmes have produced further improvements in the rates of severe ROP and other 64 2204 morbidities (Ellsbury et al., 2016). Some interventions, including human recombinant IGF-1/IGF-B3 ACCEPTED MANUSCRIPT 2205 complex and omega-3 fatty acids, which have been beneficial in animal models of OIR, have not so far 2206 been shown to have similar effects in the newborn but further trials are to be reported or are underway. A 2207 recent phase III multicenter trial of inositol (NCT1954082), a carbohydrate and essential nutrient shown 2208 to reduce neonatal morbidity including ROP in the 1980s-90s (Hallman et al., 1992) has been terminated 2209 early because of safety concerns but no other details are currently available. Caffeine, which is widely 2210 used to prevent apnoea, is one of very few treatments shown to reduce the incidence of severe ROP, 2211 whilst lactoferrin is currently being investigated in trials where ROP is a secondary outcome. To date the 2212 results from small trials of propranolol to prevent ROP suggest that, although there might be slight 2213 benefit, the safety profile of the drug would prevent its use. Finally, as neonatologists review their oxygen 2214 saturation targets in the light of the recent randomized controlled trials, which have shown the lower 2215 target of 85%-89% in very preterm infants is associated with significantly increased mortality, there is 2216 increased emphasis on investigations of ways to achieve better compliance with targets and avoid both 2217 hyperoxaemia and repeat hypoxaemic episodes. 2218 2219 Recommendation of best practice for larger infants 2220 Sadly, severe ROP in infants larger than 1500g and with gestations of 32 weeks or more is a common 2221 occurrence in middle-income and developing countries (Gilbert et al., 2005). The best approach to deal 2222 with this situation must be through improved primary prevention,MANUSCRIPT achievable through organizational 2223 changes including regionalization of care, organization of neonatal transport systems and adequate 2224 nursing numbers, as well as wider adoption of often simple evidence-based practices in perinatal and 2225 neonatal care such as greater use of antenatal steroids and oxygen saturation targeting protocols. 2226 Regarding secondary prevention for these larger infants, who usually develop ROP in zone II, laser 2227 treatment remains the gold standard. Intravitreal anti-VEGF, despite its allure as a more simple therapy is 2228 not just “an inject and walk away” option but demands careful and frequent follow-up eye examinations 2229 as well as paediatric follow-up, both of which are often not easily undertaken in countries with 2230 developing neonatal intensive care. 2231 2232 Future DirectionsACCEPTED 2233 While much has been learnt in recent years key questions remain. 2234 Our earlier work argued that different pathogenic processes may drive Posterior Zone 1 ROP as distinct 2235 from neovascularization in other areas of the retina where hypoxia-driven VEGF is responsible, so it is 2236 important for biologists to identify the molecular basis of retinal vasculogenesis. Any insights into 2237 molecular cues that regulate vasculogenesis could have therapeutic potential for aggressive posterior 2238 (Zone 1) ROP. 2239 65 2240 Another major unresolved issue is the source of the spindle cells/myofibroblasts which cause tractional ACCEPTED MANUSCRIPT 2241 detachment seen on EM in cicatricial ROP (Kretzer et al., 1984) and how to ameliorate their effects. We 2242 have also observed fibroblasts in the vitreous during the pre-retinal neovascularization phase in the Penn 2243 (50/10) model of ROP (unpublished observations: Koina, Adamson and Chan-Ling). While we reported 2244 the vascular precursors as mesenchymal precursor cells (which means they would have the potential to 2245 become myofibroblasts in disease states) (Hughes et al., 2000) our evidence was only based on Nissl 2246 staining with no lineage specific studies to demonstrate the ability of the first mesenchymal vascular 2247 precursor cells to differentiate into fibroblasts. Studies to determine the role of inflammation in tractional 2248 ROP are also essential. Future studies need to address these questions because traction retinal detachment, 2249 is the end stage of ROP and is what causes blindness from ROP as current surgical interventions at this 2250 late stage have poor visual outcomes. 2251 2252 The potential harm from systemic exposure to anti-VEGF agents needs further study, given that VEGF 2253 has key roles in the development of many organs and also acts as a neuronal survival factor during 2254 embryonic development. We do not yet know the lowest dosage that is effective because there has not 2255 been adequate investigation of lower dosages. Further basic science animal studies at different stages of 2256 development, studies of the pharmacokinetics of these agents in human infants and adequately powered 2257 randomized clinical trials are required. We suggest that a multidisciplinary consensus conference should 2258 discuss the next best trial given the difficulties in recruitingMANUSCRIPT adequate numbers of infants and the problems 2259 of interpretation that result from small trials of poor design. One possibility to be considered might be a 2260 trial of an initial low dose anti-VEGF agent coupled with subsequent laser therapy. 2261 2262 Overview 2263 This review has detailed the current understanding of the cellular and molecular mechanisms of human 2264 retinal vascular formation, including the mechanism underlying the creation of the foveal avascular zone. 2265 Armed with this understanding of normal human fetal retinal vascularization, we have detailed the 2266 mechanism underlying the pathogenesis of the two phases of acute ROP. VEGF plays a fundamental role 2267 in inducing the normal growth of angiogenic blood vessels in the human fetal retina, and is of particular 2268 relevance as a therapeutic target for the treatment of ROP, since the dysregulation of VEGF expression in 2269 both the hyperoxic andACCEPTED hypoxic phases of ROP underlies the severity of the disease. We have reviewed 2270 the biology of VEGF and the studies comparing the relative efficacy of laser treatment with various anti- 2271 VEGF agents. Intravitreal VEGF does escape from the eye and suppresses systemic VEGF levels for up 2272 to two months and observational data are now emerging to suggest that IVB treated infants have a higher 2273 risk of severe neurodevelopmental impairment (Morin et al., 2016). Therefore we recommend that, with 2274 the exception of certain specific circumstances, use of anti-VEGF agents as a therapeutic intervention in 2275 ROP still needs to be approached with caution because of; 1) the risk of late recurrence of abnormal 66 2276 retinal vascularization; 2) the possible long term systemic side effects; and 3) the uncertainty about the ACCEPTED MANUSCRIPT 2277 optimum dosage for effectiveness in ROP treatment and minimal ocular and systemic effects. These 2278 concerns need to be systematically evaluated in larger trials of infants and with longer duration than have 2279 been carried out to date. 2280 2281 Since only a small proportion of infants at ROP risk warrants treatment and there is a general shortage of 2282 ROP skilled ophthalmologists, identifying those at highest risk using screening methods such as 2283 telemedicine (discussed above) or rate of weight gain algorithms (WINROP:(Hellstrom et al., 2009, Wu 2284 et al., 2012a); CHOPROP: (Binenbaum and Tomlinson, 2017, Binenbaum et al., 2012)) alone or in 2285 combination (Gurwin et al., 2017) may offer solutions to the maldistribution of ophthalmologists (located 2286 mainly in high income countries) and premature infants (most of whom are born in middle and low 2287 income countries). However, the conclusion of Fierson and colleagues (Fierson et al., 2015) that 2288 telemedicine should be considered an adjunct to and not a replacement for indirect ophthalmoscopy has to 2289 be taken into consideration before widespread adoption of telemedicine in high income countries such as 2290 the US. 2291 2292 We have also examined and detailed key proven neonatal interventions that should be adopted in the 2293 management of the ROP infant. With understanding of the various components of ROP pathogenesis and 2294 determining how various therapeutic options affect theseMANUSCRIPT pathways, clinicians can then make informed 2295 decisions regarding the best treatment for each infant. Whilst the role of oxygen and its effects on 2296 reducing the expression of VEGF as the initiating event in ROP is well understood, recent studies have 2297 increased our understanding of the various other pathways that also contribute to the pathogenesis of 2298 ROP. These other pathways offer the promise of new treatments for ROP in the future. 2299

ACCEPTED

67 2300 References: ACCEPTED MANUSCRIPT 2301 ADAMIS, A. P., SHIMA, D. T., YEO, K. T., YEO, T. K., BROWN, L. F., BERSE, B., D'AMORE, P. A. & 2302 FOLKMAN, J. 1993. Synthesis and secretion of vascular permeability factor/vascular 2303 endothelial growth factor by human retinal pigment epithelial cells. Biochem Biophys 2304 Res Commun, 193, 631-8. 2305 AHMAD, K. M., KLUG, K., HERR, S., STERLING, P. & SCHEIN, S. 2003. Cell density ratios in a 2306 foveal patch in macaque retina. Vis Neurosci, 20, 189-209. 2307 AKERBLOM, H., ANDREASSON, S., LARSSON, E. & HOLMSTROM, G. 2014. Photoreceptor 2308 Function in School-Aged Children is Affected by Preterm Birth. Transl Vis Sci Technol, 2309 3, 7. 2310 AKERBLOM, H., LARSSON, E., ERIKSSON, U. & HOLMSTROM, G. 2011. Central macular 2311 thickness is correlated with gestational age at birth in prematurely born children. Br J 2312 Ophthalmol, 95, 799-803. 2313 AL-OTAIBI, A. G., ALDREES, S. S. & MOUSA, A. A. 2012. Long term visual outcomes in laser 2314 treated threshold retinopathy of prematurity in Central Saudi Arabia. Saudi J 2315 Ophthalmol, 26, 299-303. 2316 ALDER, V. A., BEN-NUN, J. & CRINGLE, S. J. 1990. PO2 profiles and oxygen consumption in cat 2317 retina with an occluded retinal circulation. Invest. Ophth. Vis. Sci., 31, 31-36. 2318 ALDER, V. A., CRINGLE, S. J. & CONSTABLE, I. J. 1983. The retinal oxygen profile in cats. 2319 Research in Vision and Ophthalmology, 24, 30-36. 2320 ALON, T., HEMO, I., ITIN, A., PE'ER, J., STONE, J. & KESHET, E. 1995. Vascular endothelial 2321 growth factor acts as a survival factor for newly formed retinal vessels and has 2322 implications for retinopathy of prematurity. Nature Medicine., 1, 1024-8. 2323 AMADIO, M., GOVONI, S. & PASCALE, A. 2016. Targeting VEGF in eye neovascularization: 2324 What's new?: A comprehensive review on current therapies and oligonucleotide- 2325 based interventions under development. Pharmacol Res, 103, 253-69. 2326 ARAZ-ERSAN, B., KIR, N., TUNCER, S., AYDINOGLU-CANDAN, O., YILDIZ-INEC, D., AKDOGAN, B., 2327 EKICI, B., DEMIREL, A. & OZMEN, M. 2015. PreliminarMANUSCRIPTy anatomical and 2328 neurodevelopmental outcomes of intravitreal bevacizumab as adjunctive treatment 2329 for retinopathy of prematurity. Curr Eye Res, 40, 585-91. 2330 AREVALO, J. F., MAIA, M., FLYNN, H. W., JR., SARAVIA, M., AVERY, R. L., WU, L., EID FARAH, M., 2331 PIERAMICI, D. J., BERROCAL, M. H. & SANCHEZ, J. G. 2008. Tractional retinal 2332 detachment following intravitreal bevacizumab (Avastin) in patients with severe 2333 proliferative diabetic retinopathy. Br J Ophthalmol, 92, 213-6. 2334 ARVAS, S., SARICI, A. M. & AKAR, S. 2014. Diode laser photocoagulation posterior to the ridge 2335 in severe stage 3+ threshold retinopathy of prematurity. Cutan Ocul Toxicol, 33, 197- 2336 200. 2337 ASHTON, N. 1966. Oxygen and the growth and development of retinal vessels. In vivo and in 2338 vitro studies. The XX Francis I. Proctor Lecture. Am J Ophthalmol, 62, 412-35. 2339 ASHTON, N. 1970. Some aspects of the comparative pathology of oxygen toxicity in the 2340 retina. Ophthalmologica, 160, 54-71. 2341 ASHTON, N., WARD, B. & SERPELL, G. 1953. Role of oxygen in the genesis of retrolental 2342 fibroplasia; aACCEPTED preliminary report. Br J Ophthalmol, 37, 513-20. 2343 ASHTON, N., WARD, B. & SERPELL, G. 1954. Effect of oxygen on developing retinal vessels 2344 with particular reference to the problem of retrolental fibroplasia. Br J Ophthalmol, 2345 38, 397-432. 2346 ASKIE, L. M. 2013. Optimal oxygen saturations in preterm infants: a moving target. Curr Opin 2347 Pediatr, 25, 188-92. 2348 ASKIE, L. M., BROCKLEHURST, P., DARLOW, B. A., FINER, N., SCHMIDT, B., TARNOW-MORDI, 2349 W. & NE, O. C. G. 2011. NeOProM: Neonatal Oxygenation Prospective Meta-analysis 2350 Collaboration study protocol. BMC Pediatr, 11, 6.

68 2351 ASKIE, L. M., DARLOW, B. A., DAVIS,ACCEPTED P. G., FINER, MANUSCRIPT N., STENSON, B., VENTO, M. & WHYTE, R. 2352 2017. Effects of targeting lower versus higher arterial oxygen saturations on death or 2353 disability in preterm infants. Cochrane Database Syst Rev, 4, CD011190. 2354 AUSTENG, D., KALLEN, K. B., EWALD, U. W., JAKOBSSON, P. G. & HOLMSTROM, G. E. 2009. 2355 Incidence of retinopathy of prematurity in infants born before 27 weeks' gestation in 2356 Sweden. Arch Ophthalmol, 127, 1315-9. 2357 AUTRATA, R., SENKOVA, K., HOLOUSOVA, M., KREJCIROVA, I., DOLEZEL, Z. & BOREK, I. 2012. 2358 [Effects of intravitreal pegaptanib or bevacizumab and laser in treatment of threshold 2359 retinopathy of prematurity in zone I and posterior zone II--four years results]. Cesk 2360 Slov Oftalmol, 68, 29-36. 2361 AVERY, R. L., PEARLMAN, J., PIERAMICI, D. J., RABENA, M. D., CASTELLARIN, A. A., NASIR, M. A., 2362 GIUST, M. J., WENDEL, R. & PATEL, A. 2006. Intravitreal bevacizumab (Avastin) in the 2363 treatment of proliferative diabetic retinopathy. Ophthalmology, 113, 1695 e1-15. 2364 AXER-SIEGEL, R., BOURLA, D., FRILING, R., SHALEV, B., SIROTA, L., BENJAMINI, Y., 2365 WEINBERGER, D. & SNIR, M. 2004. Intraocular pressure variations after diode laser 2366 photocoagulation for threshold retinopathy of prematurity. Ophthalmology, 111, 2367 1734-8. 2368 AZUMA, N., ITO, M., YOKOI, T., NAKAYAMA, Y. & NISHINA, S. 2013. Visual outcomes after early 2369 vitreous surgery for aggressive posterior retinopathy of prematurity. JAMA 2370 Ophthalmol, 131, 1309-13. 2371 BANCALARI, E. & CLAURE, N. 2013. Oxygenation targets and outcomes in premature infants. 2372 JAMA, 309, 2161-2. 2373 BARNABY, A. M., HANSEN, R. M., MOSKOWITZ, A. & FULTON, A. B. 2007. Development of 2374 scotopic visual thresholds in retinopathy of prematurity. Invest Ophthalmol Vis Sci, 48, 2375 4854-60. 2376 BARNES, P. J. 2006. How corticosteroids control inflammation: Quintiles Prize Lecture 2005. 2377 Br J Pharmacol, 148, 245-54. 2378 BARRES, B. A., HART, I. K., COLES, H. S., BURNE, J. F.,MANUSCRIPT VOYVODIC, J. T., RICHARDSON, W. D. & 2379 RAFF, M. C. 1992. Cell death and control of cell survival in the oligodendrocyte lineage. 2380 Cell, 70, 31-46. 2381 BEHARRY, K. D., VALENCIA, G. B., LAZZARO, D. R. & ARANDA, J. V. 2016. Pharmacologic 2382 interventions for the prevention and treatment of retinopathy of prematurity. Semin 2383 Perinatol . 2384 BELIGERE, N., PERUMALSWAMY, V., TANDON, M., MITTAL, A., FLOORA, J., VIJAYAKUMAR, B. & 2385 MILLER, M. T. 2015. Retinopathy of prematurity and neurodevelopmental disabilities 2386 in premature infants. Semin Fetal Neonatal Med, 20, 346-53. 2387 BENJAMIN, L. E., HEMO, I. & KESHET, E. 1998. A plasticity window for blood vessel 2388 remodelling is defined by pericyte coverage of the preformed endothelial network 2389 and is regulated by PDGF-B and VEGF. Development, 125, 1591-8. 2390 BHARWANI, S. K. & DHANIREDDY, R. 2008. Systemic fungal infection is associated with the 2391 development of retinopathy of prematurity in very low birth weight infants: a meta- 2392 review. J Perinatol, 28, 61-6. 2393 BHARWANI, S. K., GREEN, B. F., PEZZULLO, J. C., BHARWANI, S. S., BHARWANI, S. S. & 2394 DHANIREDDY,ACCEPTED R. 2016. Systematic review and meta-analysis of human milk intake 2395 and retinopathy of prematurity: a significant update. J Perinatol, 36, 913-920. 2396 BINENBAUM, G. 2013. Algorithms for the prediction of retinopathy of prematurity based on 2397 postnatal weight gain. Clin Perinatol, 40, 261-70. 2398 BINENBAUM, G. & TOMLINSON, L. A. 2017. Postnatal Growth and Retinopathy of Prematurity 2399 Study: Rationale, Design, and Subject Characteristics. Ophthalmic Epidemiol, 24, 36-47. 2400 BINENBAUM, G., YING, G. S., QUINN, G. E., HUANG, J., DREISEITL, S., ANTIGUA, J., FOROUGHI, N. 2401 & ABBASI, S. 2012. The CHOP postnatal weight gain, birth weight, and gestational age 2402 retinopathy of prematurity risk model. Arch Ophthalmol, 130, 1560-5.

69 2403 BLENCOWE, H., LAWN, J. E., VAZQUEZ,ACCEPTED T., FIELDER, MANUSCRIPT A. & GILBERT, C. 2013. Preterm- 2404 associated visual impairment and estimates of retinopathy of prematurity at regional 2405 and global levels for 2010. Pediatric Research, 74, 35-49. 2406 BONDY, C., WERNER, H., ROBERTS, C. T., JR. & LEROITH, D. 1992. Cellular pattern of type-I 2407 insulin-like growth factor receptor gene expression during maturation of the rat 2408 brain: comparison with insulin-like growth factors I and II. Neuroscience, 46, 909-23. 2409 BONDY, C. A. & CHENG, C. M. 2002. Insulin-like growth factor-1 promotes neuronal glucose 2410 utilization during brain development and repair processes. Int Rev Neurobiol, 51, 189- 2411 217. 2412 BROCATO, B., HOLLIDAY, N., WHITEHURST, R. M., JR., LEWIS, D. & VARNER, S. 2016. Delayed 2413 Cord Clamping in Preterm Neonates: A Review of Benefits and Risks. Obstet Gynecol 2414 Surv, 71, 39-42. 2415 CABRERA, M. T., MALDONADO, R. S., TOTH, C. A., O'CONNELL, R. V., CHEN, B. B., CHIU, S. J., 2416 FARSIU, S., WALLACE, D. K., STINNETT, S. S., PANAYOTTI, G. M., SWAMY, G. K. & 2417 FREEDMAN, S. F. 2012. Subfoveal fluid in healthy full-term newborns observed by 2418 handheld spectral-domain optical coherence tomography. Am J Ophthalmol, 153, 167- 2419 75 e3. 2420 CARDEN, S. M., LUU, L. N., NGUYEN, T. X., HUYNH, T. & GOOD, W. V. 2008. Retinopathy of 2421 prematurity: postmenstrual age at threshold in a transitional economy is similar to 2422 that in developed countries. Clin Experiment Ophthalmol, 36, 159-61. 2423 CARMELIET, P., FERREIRA, V., BREIER, G., POLLEFEYT, S., KIECKENS, L., GERTSENSTEIN, M., 2424 FAHRIG, M., VANDENHOECK, A., HARPAL, K., EBERHARDT, C., DECLERCQ, C., PAWLING, 2425 J., MOONS, L., COLLEN, D., RISAU, W. & NAGY, A. 1996. Abnormal blood vessel 2426 development and lethality in embryos lacking a single VEGF allele. Nature, 380, 435-9. 2427 CARRO, E., NUNEZ, A., BUSIGUINA, S. & TORRES-ALEMAN, I. 2000. Circulating insulin-like 2428 growth factor I mediates effects of exercise on the brain. J Neurosci, 20, 2926-33. 2429 CARRO, E., TREJO, J. L., NUNEZ, A. & TORRES-ALEMAN, I. 2003. Brain repair and 2430 neuroprotection by serum insulin-like growthMANUSCRIPT factor I. Mol Neurobiol, 27, 153-62. 2431 CHAN, J. J., LAM, C. P., KWOK, M. K., WONG, R. L., LEE, G. K., LAU, W. W. & YAM, J. C. 2016. Risk 2432 of recurrence of retinopathy of prematurity after initial intravitreal ranibizumab 2433 therapy. Sci Rep, 6, 27082. 2434 CHAN-LING, T. 1994. Glial, neuronal and vascular interactions in the mammalian retina. In: 2435 OSBORNE, N. & CHADER, G. (eds.) Progress in Retinal Research. Oxford: Pergamon 2436 Press. 2437 CHAN-LING, T., ADAMSON, S., NATOLI, R., MACCARONE, R., KOINA, M., HU, P., LAU, J. C., 2438 PROVIS, J. M., LINSENMEIER, R. A. & BISTI, S. 2014a. Dark-Rearing (DR) Mitigates Pre- 2439 Retinal Neovascularization and OIR Pathology with no Detriment To Visual Function 2440 and Morphology: Compelling Evidence For DR as Non-Invasive Intervention in 2441 Retinopathy of Prematurity (ROP). Investigative Ophthalmology & Visual Science, 55, 2442 5902-5902. 2443 CHAN-LING, T., BARNETT, N. L., MACCARONE, R., PROVIS, J., KOINA, M., HU, P., NATOLI, R., 2444 BISTI, S., LINSENMEIER, R. A. & ADAMSON, S. 2016. Dark-Rearing (DR) precludes the 2445 initiating event in OIR and eliminates the pathology seen in the second phase of 2446 disease: RationaleACCEPTED for novel non-invasive treatment for ROP. Investigative 2447 Ophthalmology & Visual Science, 57. 2448 CHAN-LING, T., GOCK, B. & STONE, J. 1995a. The effect of oxygen on vasoformative cell 2449 division. Evidence that 'physiological hypoxia' is the stimulus for normal retinal 2450 vasculogenesis. Invest Ophthalmol Vis Sci, 36, 1201-14. 2451 CHAN-LING, T., GOCK, B. & STONE, J. 1995b. Supplemental oxygen therapy. Basis for 2452 Noninvasive treatment of retinopathy of prematurity. Invest. Opthalmol. Vis. Sci., 36, 2453 1215-1229. 2454 CHAN-LING, T., HALASZ, P. & STONE, J. 1990. Development of Retinal Vasculature in the Cat - 2455 Processes and Mechanisms. Current Eye Research, 9, 459-478. 70 2456 CHAN-LING, T., MACCARONE, R., AKOINA,CCEPTED M., LAU, MANUSCRIPT J. C., KOZULIN, P., NATOLI, R., PROVIS, J., 2457 BISTI, S., ADAMSON, S. & LINSENMEIER, R. A. 2014b. Dark rearing (DR) as a means of 2458 mimicking physiological hypoxia: a non-invasive intervention for retinopathy of 2459 prematurity (ROP). XXI Biennial Meeting of the International Society for Eye Research. 2460 San Francisco. 2461 CHAN-LING, T., MCLEOD, D. S., HUGHES, S., BAXTER, L., CHU, Y., HASEGAWA, T. & LUTTY, G. A. 2462 2004a. Astrocyte-endothelial cell relationships during human retinal vascular 2463 development. Investigative Ophthalmology & Visual Science, 45, 2020-2032. 2464 CHAN-LING, T., PAGE, M. P., GARDINER, T., BAXTER, L., ROSINOVA, E. & HUGHES, S. 2004b. 2465 Desmin ensheathment ratio as an indicator of vessel stability: evidence in normal 2466 development and in retinopathy of prematurity. Am J Pathol, 165, 1301-13. 2467 CHAN-LING, T. & STONE, J. 1992. Degeneration of astrocytes in feline retinopathy of 2468 prematurity causes failure of the blood-retinal barrier. Invest. Ophth. Vis. Sci., 33, 2469 2148-2159. 2470 CHAN-LING, T. & STONE, J. 1993. Retinopathy of prematurity: Origins in the architecture of 2471 the retina. Prog Retin Eye Res, 12, 155-178. 2472 CHAN-LING, T., TOUT, S., HOLLANDER, H. & STONE, J. 1992. Vascular changes and their 2473 mechanisms in the feline model of retinopathy of prematurity. Invest Ophthalmol Vis 2474 Sci, 33, 2128-47. 2475 CHEN, C. H. & CHEN, S. C. 1980. Angiogenic Activity of Vitreous and Retinal Extract. 2476 Investigative Ophthalmology & Visual Science, 19, 596-602. 2477 CHEN, C. H. & CHEN, S. C. 1984. Mediation of Vascular Endothelial Proliferation by Factors in 2478 Fetal Bovine Retinal Extracts and Serum. Exp Eye Res, 39, 469-478. 2479 CHEN, C. H. & CHEN, S. C. 1987. Evidence of the Presence of a Specific Vascular Endothelial 2480 Growth-Factor in Fetal Bovine Retina. Experimental Cell Research, 169, 287-295. 2481 CHEN, J., JOYAL, J. S., HATTON, C. J., JUAN, A. M., PEI, D. T., HURST, C. G., XU, D., STAHL, A., 2482 HELLSTROM, A. & SMITH, L. E. 2012. Propranolol inhibition of beta-adrenergic 2483 receptor does not suppress pathologic neovascularizMANUSCRIPTation in oxygen-induced 2484 retinopathy. Invest Ophthalmol Vis Sci, 53, 2968-77. 2485 CHEN, J. & SMITH, L. E. 2008. A double-edged sword: erythropoietin eyed in retinopathy of 2486 prematurity. J AAPOS, 12, 221-2. 2487 CHEN, M., CITIL, A., MCCABE, F., LEICHT, K. M., FIASCONE, J., DAMMANN, C. E. & DAMMANN, O. 2488 2011. Infection, oxygen, and immaturity: interacting risk factors for retinopathy of 2489 prematurity. Neonatology, 99, 125-32. 2490 CHEN, Y., FENG, J., GILBERT, C., YIN, H., LIANG, J. & LI, X. 2015. Time at treatment of severe 2491 retinopathy of prematurity in China: recommendations for guidelines in more mature 2492 infants. PLoS One, 10, e0116669. 2493 CHEN, Y. H., CHEN, S. N., LIEN, R. I., SHIH, C. P., CHAO, A. N., CHEN, K. J., HWANG, Y. S., WANG, N. 2494 K., CHEN, Y. P., LEE, K. H., CHUANG, C. C., CHEN, T. L., LAI, C. C. & WU, W. C. 2014. 2495 Refractive errors after the use of bevacizumab for the treatment of retinopathy of 2496 prematurity: 2-year outcomes. Eye (Lond), 28, 1080-6; quiz 1087. 2497 CHENG, C. M., REINHARDT, R. R., LEE, W. H., JONCAS, G., PATEL, S. C. & BONDY, C. A. 2000. 2498 Insulin-like growth factor 1 regulates developing brain glucose metabolism. Proc Natl 2499 Acad Sci U S A,ACCEPTED 97, 10236-41. 2500 CHOI, J., KIM, J. H., KIM, S. J. & YU, Y. S. 2011. Long-term results of lens-sparing vitrectomy for 2501 stages 4B and 5 retinopathy of prematurity. Korean J Ophthalmol, 25, 305-10. 2502 CHRISTIANSEN, S. P., DOBSON, V., QUINN, G. E., GOOD, W. V., TUNG, B., HARDY, R. J., BAKER, J. 2503 D., HOFFMAN, R. O., REYNOLDS, J. D., RYCHWALSKI, P. J., SHAPIRO, M. J. & EARLY 2504 TREATMENT FOR RETINOPATHY OF PREMATURITY COOPERATIVE, G. 2010. 2505 Progression of type 2 to type 1 retinopathy of prematurity in the Early Treatment for 2506 Retinopathy of Prematurity Study. Arch Ophthalmol, 128, 461-5.

71 2507 CHU, Y., HUGHES, S. & CHAN-LING,A CCEPTEDT. 2001. Different MANUSCRIPTiation and migration of astrocyte 2508 precursor cells and astrocytes in human fetal retina: relevance to optic nerve 2509 coloboma. FASEB J, 15, 2013-5. 2510 CHUI, T. Y., ZHONG, Z., SONG, H. & BURNS, S. A. 2012. Foveal avascular zone and its 2511 relationship to foveal pit shape. Optom Vis Sci, 89, 602-10. 2512 COLE, C. H., WRIGHT, K. W., TARNOW-MORDI, W., PHELPS, D. L. & PULSE OXIMETRY 2513 SATURATION TRIAL FOR PREVENTION OF RETINOPATHY OF PREMATURITY 2514 PLANNING STUDY, G. 2003. Resolving our uncertainty about oxygen therapy. 2515 Pediatrics, 112, 1415-9. 2516 CONNOR, A. J., PAPASTAVROU, V. T., HILLIER, R. J. & SHAFIQ, A. 2015. Ultra-low dose of 2517 intravitreal bevacizumab in the treatment of retinopathy of prematurity. J Pediatr 2518 Ophthalmol Strabismus, 52 Online, e20-1. 2519 CORNISH, E. E., XIAO, M., YANG, Z., PROVIS, J. M. & HENDRICKSON, A. E. 2004. The role of opsin 2520 expression and apoptosis in determination of cone types in human retina. Exp Eye Res, 2521 78, 1143-54. 2522 COSTELOE, K. L., HENNESSY, E. M., HAIDER, S., STACEY, F., MARLOW, N. & DRAPER, E. S. 2012. 2523 Short term outcomes after extreme preterm birth in England: comparison of two birth 2524 cohorts in 1995 and 2006 (the EPICure studies). BMJ, 345, e7976. 2525 CRYO-ROP 1988. Multicenter trial of cryotherapy for retinopathy of prematurity. 2526 Preliminary results. Cryotherapy for Retinopathy of Prematurity Cooperative Group. 2527 Arch Ophthalmol, 106, 471-9. 2528 CRYO-ROP 2001. Effect of retinal ablative therapy for threshold retinopathy of prematurity: 2529 results of Goldmann perimetry at the age of 10 years. Arch Ophthalmol, 119, 1120-5. 2530 D'AMORE, P. A., GLASER, B. M., BRUNSON, S. K. & FENSELAU, A. H. 1981. Angiogenic Activity 2531 from Bovine Retina - Partial-Purification and Characterization. Proceedings of the 2532 National Academy of Sciences of the United States of America-Biological Sciences, 78, 2533 3068-3072. 2534 DAMMANN, O., BRINKHAUS, M. J., BARTELS, D. B., MANUSCRIPTDORDELMANN, M., DRESSLER, F., KERK, J., 2535 DORK, T. & DAMMANN, C. E. 2009. Immaturity, perinatal inflammation, and 2536 retinopathy of prematurity: a multi-hit hypothesis. Early Hum Dev, 85, 325-9. 2537 DANIEL, E., QUINN, G. E., HILDEBRAND, P. L., ELLS, A., HUBBARD, G. B., 3RD, CAPONE, A., JR., 2538 MARTIN, E. R., OSTROFF, C. P., SMITH, E., PISTILLI, M., YING, G. S. & E, R. O. P. C. G. 2015. 2539 Validated System for Centralized Grading of Retinopathy of Prematurity: 2540 Telemedicine Approaches to Evaluating Acute-Phase Retinopathy of Prematurity (e- 2541 ROP) Study. JAMA Ophthalmol, 133, 675-82. 2542 DARLAND, D. C., MASSINGHAM, L. J., SMITH, S. R., PIEK, E., ST-GENIEZ, M. & D'AMORE, P. A. 2543 2003. Pericyte production of cell-associated VEGF is differentiation-dependent and is 2544 associated with endothelial survival. Developmental Biology, 264, 275-288. 2545 DARLOW, B. A., CLEMETT, R. S., HORWOOD, L. J. & MOGRIDGE, N. 1997. Prospective study of 2546 New Zealand infants with birth weight less than 1500 g and screened for retinopathy 2547 of prematurity: visual outcome at age 7-8 years. Br J Ophthalmol, 81, 935-40. 2548 DARLOW, B. A., ELLS, A. L., GILBERT, C. E., GOLE, G. A. & QUINN, G. E. 2013a. Are we there yet? 2549 Bevacizumab therapy for retinopathy of prematurity. Arch Dis Child Fetal Neonatal Ed, 2550 98, F170-4. ACCEPTED 2551 DARLOW, B. A., GILBERT, C. E. & QUIROGA, A. M. 2013b. Setting up and improving 2552 retinopathy of prematurity programs: interaction of neonatology, nursing, and 2553 ophthalmology. Clin Perinatol, 40, 215-27. 2554 DARLOW, B. A., HUTCHINSON, J. L., SIMPSON, J. M., HENDERSON-SMART, D. J., DONOGHUE, D. 2555 A. & EVANS, N. J. 2005. Variation in rates of severe retinopathy of prematurity among 2556 neonatal intensive care units in the Australian and New Zealand Neonatal Network. Br 2557 J Ophthalmol, 89, 1592-6. 2558 DARLOW, B. A., LUI, K., KUSUDA, S., REICHMAN, B., HAKANSSON, S., BASSLER, D., MODI, N., 2559 LEE, S. K., LEHTONEN, L., VENTO, M., ISAYAMA, T., SJORS, G., HELENIUS, K. K., ADAMS, 72 2560 M., RUSCONI, F., MORISAKI,A CCEPTEDN., SHAH, P. MANUSCRIPT S. & INTERNATIONAL NETWORK FOR 2561 EVALUATING OUTCOMES OF, N. 2017. International variations and trends in the 2562 treatment for retinopathy of prematurity. Br J Ophthalmol . 2563 DARLOW, B. A., MARSCHNER, S. L., DONOGHOE, M., BATTIN, M. R., BROADBENT, R. S., ELDER, 2564 M. J., HEWSON, M. P., MEYER, M. P., GHADGE, A., GRAHAM, P., MCNEILL, N. J., KUSCHEL, 2565 C. A., TARNOW-MORDI, W. O. & BENEFITS OF OXYGEN SATURATION TARGETING-NEW 2566 ZEALAND COLLABORATIVE, G. 2014. Randomized controlled trial of oxygen saturation 2567 targets in very preterm infants: two year outcomes. J Pediatr, 165, 30-35 e2. 2568 DARLOW, B. A. & MORLEY, C. J. 2015. Oxygen Saturation Targeting and Bronchopulmonary 2569 Dysplasia. Clin Perinatol, 42, 807-23. 2570 DAVITT, B. V., CHRISTIANSEN, S. P., HARDY, R. J., TUNG, B., GOOD, W. V. & EARLY TREATMENT 2571 FOR RETINOPATHY OF PREMATURITY COOPERATIVE, G. 2013. Incidence of cataract 2572 development by 6 months' corrected age in the Early Treatment for Retinopathy of 2573 Prematurity study. J AAPOS, 17, 49-53. 2574 DAWSON, D. W., VOLPERT, O. V., GILLIS, P., CRAWFORD, S. E., XU, H., BENEDICT, W. & BOUCK, 2575 N. P. 1999. Pigment epithelium-derived factor: a potent inhibitor of angiogenesis. 2576 Science, 285, 245-8. 2577 DIAZ-ARAYA, C. & PROVIS, J. M. 1992. Evidence of photoreceptor migration during early 2578 foveal development: a quantitative analysis of human fetal retinae. Vis Neurosci, 8, 2579 505-14. 2580 DOBSON, V., QUINN, G., FLYNN, J. T., GILBERT, W. S., HARDY, R. J., PHELPS, D. L., TUNG, B. & 2581 PALMER, E. A. 1993. Multicenter trial of cryotherapy for retinopathy of prematurity. 3 2582 1/2-year outcome--structure and function. Cryotherapy for Retinopathy of 2583 Prematurity Cooperative Group. Arch Ophthalmol, 111, 339-44. 2584 DOBSON, V., QUINN, G. E., SUMMERS, C. G., SAUNDERS, R. A., PHELPS, D. L., TUNG, B. & 2585 PALMER, E. A. 1994. Effect of acute-phase retinopathy of prematurity on grating acuity 2586 development in the very low birth weight infant. The Cryotherapy for Retinopathy of 2587 Prematurity Cooperative Group. Invest OphthalmolMANUSCRIPT Vis Sci, 35, 4236-44. 2588 DORE, S., KAR, S. & QUIRION, R. 1997. Rediscovering an old friend, IGF-I: potential use in the 2589 treatment of neurodegenerative diseases. Trends Neurosci, 20, 326-31. 2590 DOYLE, L. W., HALLIDAY, H. L., EHRENKRANZ, R. A., DAVIS, P. G. & SINCLAIR, J. C. 2014. An 2591 update on the impact of postnatal systemic corticosteroids on mortality and cerebral 2592 palsy in preterm infants: effect modification by risk of bronchopulmonary dysplasia. J 2593 Pediatr, 165, 1258-60. 2594 DRAGO, J., MURPHY, M., CARROLL, S. M., HARVEY, R. P. & BARTLETT, P. F. 1991. Fibroblast 2595 growth factor-mediated proliferation of central nervous system precursors depends 2596 on endogenous production of insulin-like growth factor I. Proc Natl Acad Sci U S A, 88, 2597 2199-203. 2598 DRASDO, N., MILLICAN, C. L., KATHOLI, C. R. & CURCIO, C. A. 2007. The length of Henle fibers 2599 in the human retina and a model of ganglion receptive field density in the visual field. 2600 Vision Res, 47, 2901-11. 2601 DREHER, B. & ROBINSON, S. R. 1988. Development of the retinofugal pathway in birds and 2602 mammals: evidence for a common 'timetable'. Brain Behav Evol, 31, 369-90. 2603 DROLET, B. A., FROMMELT,ACCEPTED P. C., CHAMLIN, S. L., HAGGSTROM, A., BAUMAN, N. M., CHIU, Y. E., 2604 CHUN, R. H., GARZON, M. C., HOLLAND, K. E., LIBERMAN, L., MACLELLAN-TOBERT, S., 2605 MANCINI, A. J., METRY, D., PUTTGEN, K. B., SEEFELDT, M., SIDBURY, R., WARD, K. M., 2606 BLEI, F., BASELGA, E., CASSIDY, L., DARROW, D. H., JOACHIM, S., KWON, E. K., MARTIN, 2607 K., PERKINS, J., SIEGEL, D. H., BOUCEK, R. J. & FRIEDEN, I. J. 2013. Initiation and use of 2608 propranolol for infantile hemangioma: report of a consensus conference. Pediatrics, 2609 131, 128-40. 2610 DUTTON, G. N. 2009. 'Dorsal stream dysfunction' and 'dorsal stream dysfunction plus': a 2611 potential classification for perceptual visual impairment in the context of cerebral 2612 visual impairment? Dev Med Child Neurol, 51, 170-2. 73 2613 DUTTON, G. N. 2013. The spectrumACCEPTED of cerebral MANUSCRIPT visual impairment as a sequel to premature 2614 birth: an overview. Doc Ophthalmol, 127, 69-78. 2615 EHMANN, D. & GREVE, M. 2014. Intravitreal bevacizumab for exudative retinal detachment 2616 post laser therapy for retinopathy of prematurity. Can J Ophthalmol, 49, 228-31. 2617 ELLS, A. L., GOLE, G. A., LLOYD HILDEBRAND, P., INGRAM, A., WILSON, C. M. & GEOFF 2618 WILLIAMS, R. 2013. Posterior to the ridge laser treatment for severe stage 3 2619 retinopathy of prematurity. Eye (Lond), 27, 525-30. 2620 ELLS, A. L., HOLMES, J. M., ASTLE, W. F., WILLIAMS, G., LESKE, D. A., FIELDEN, M., UPHILL, B., 2621 JENNETT, P. & HEBERT, M. 2003. Telemedicine approach to screening for severe 2622 retinopathy of prematurity: a pilot study. Ophthalmology, 110, 2113-7. 2623 ELLSBURY, D. L., CLARK, R. H., URSPRUNG, R., HANDLER, D. L., DODD, E. D. & SPITZER, A. R. 2624 2016. A Multifaceted Approach to Improving Outcomes in the NICU: The Pediatrix 100 2625 000 Babies Campaign. Pediatrics, 137. 2626 ELLSBURY, D. L. & URSPRUNG, R. 2010. Comprehensive Oxygen Management for the 2627 Prevention of Retinopathy of Prematurity: the pediatrix experience. Clin Perinatol, 37, 2628 203-15. 2629 ETROP 2003. Revised indications for the treatment of retinopathy of prematurity: results of 2630 the early treatment for retinopathy of prematurity randomized trial. Arch 2631 Ophthalmol, 121, 1684-94. 2632 ETROP 2011. Grating visual acuity results in the early treatment for retinopathy of 2633 prematurity study. Arch Ophthalmol, 129, 840-6. 2634 FALK, T., GONZALEZ, R. T. & SHERMAN, S. J. 2010. The Yin and Yang of VEGF and PEDF: 2635 Multifaceted Neurotrophic Factors and Their Potential in the Treatment of 2636 Parkinson's Disease. Int J Mol Sci, 11, 2875-900. 2637 FANG, J. L., SORITA, A., CAREY, W. A., COLBY, C. E., MURAD, M. H. & ALAHDAB, F. 2016. 2638 Interventions To Prevent Retinopathy of Prematurity: A Meta-analysis. Pediatrics, 2639 137. 2640 FAUCHERE, J. C., KOLLER, B. M., TSCHOPP, A., DAME,MANUSCRIPT C., RUEGGER, C., BUCHER, H. U. & SWISS 2641 ERYTHROPOIETIN NEUROPROTECTION TRIAL, G. 2015. Safety of Early High-Dose 2642 Recombinant Erythropoietin for Neuroprotection in Very Preterm Infants. J Pediatr, 2643 167, 52-7 e1-3. 2644 FERRARA, N. 2004. Vascular endothelial growth factor: basic science and clinical progress. 2645 Endocr Rev, 25, 581-611. 2646 FERRARA, N., GERBER, H. P. & LECOUTER, J. 2003. The biology of VEGF and its receptors. Nat 2647 Med, 9, 669-76. 2648 FERRARA, N. & HENZEL, W. J. 1989. Pituitary follicular cells secrete a novel heparin-binding 2649 growth factor specific for vascular endothelial cells. Biochem Biophys Res Commun, 2650 161, 851-8. 2651 FERRONE, P. J., TRESE, M. T., WILLIAMS, G. A. & COX, M. S. 1998. Good visual acuity in an adult 2652 population with marked posterior segment changes secondary to retinopathy of 2653 prematurity. Retina, 18, 335-8. 2654 FIELDER, A., BLENCOWE, H., O'CONNOR, A. & GILBERT, C. 2015. Impact of retinopathy of 2655 prematurity on ocular structures and visual functions. Arch Dis Child Fetal Neonatal 2656 Ed, 100, F179-84.ACCEPTED 2657 FIERSON, W. M., AMERICAN ACADEMY OF PEDIATRICS SECTION ON, O., AMERICAN ACADEMY 2658 OF, O., AMERICAN ASSOCIATION FOR PEDIATRIC, O., STRABISMUS & AMERICAN 2659 ASSOCIATION OF CERTIFIED, O. 2013. Screening examination of premature infants for 2660 retinopathy of prematurity. Pediatrics, 131, 189-95. 2661 FIERSON, W. M., CAPONE, A., JR., AMERICAN ACADEMY OF PEDIATRICS SECTION ON, O. & 2662 AMERICAN ACADEMY OF OPHTHALMOLOGY, A. A. O. C. O. 2015. Telemedicine for 2663 evaluation of retinopathy of prematurity. Pediatrics, 135, e238-54. 2664 FILIPPI, L., CAVALLARO, G., BAGNOLI, P., DAL MONTE, M., FIORINI, P., DONZELLI, G., TINELLI, 2665 F., ARAIMO, G., CRISTOFORI, G., LA MARCA, G., DELLA BONA, M. L., LA TORRE, A., 74 2666 FORTUNATO, P., FURLANETTO,ACCEPTED S., OSNAGHI, MANUSCRIPT S. & MOSCA, F. 2013. Oral propranolol for 2667 retinopathy of prematurity: risks, safety concerns, and perspectives. J Pediatr, 163, 2668 1570-1577 e6. 2669 FLECK, B. W. & STENSON, B. J. 2013. Retinopathy of prematurity and the oxygen conundrum: 2670 lessons learned from recent randomized trials. Clin Perinatol, 40, 229-40. 2671 FLEDELIUS, H. C. 1996a. Pre-term delivery and subsequent ocular development. A 7-10 year 2672 follow-up of children screened 1982-84 for ROP. 1) Visual function, slit-lamp findings, 2673 and fundus appearance. Acta Ophthalmol Scand, 74, 288-93. 2674 FLEDELIUS, H. C. 1996b. Pre-term delivery and subsequent ocular development. A 7-10 year 2675 follow-up of children screened 1982-84 for ROP. 3) Refraction. Myopia of prematurity. 2676 Acta Ophthalmol Scand, 74, 297-300. 2677 FLEDELIUS, H. C., BANGSGAARD, R., SLIDSBORG, C. & LACOUR, M. 2015. Refraction and visual 2678 acuity in a national Danish cohort of 4-year-old children of extremely preterm 2679 delivery. Acta Ophthalmol, 93, 330-8. 2680 FLEDELIUS, H. C. & JENSEN, H. 2011. Late subsequent ocular morbidity in retinopathy of 2681 prematurity patients, with emphasis on visual loss caused by insidious 'involutive' 2682 pathology: an observational series. Acta Ophthalmol, 89, 316-23. 2683 FLYNN, J. T., CASSADY, J., ESSNER, D., ZESKIND, J., MERRITT, J., FLYNN, R. & WILLIAMS, M. J. 2684 1979. Fluorescein angiography in retrolental fibroplasia: experience from 1969-1977. 2685 Ophthalmology, 86, 1700-23. 2686 FLYNN, J. T. & CHAN-LING, T. 2006. Retinopathy of Prematurity: Two Distinct Mechanisms 2687 That Underlie Zone 1 and Zone 2 Disease. American Journal of Ophthalmology, 142, 2688 46-59.e2. 2689 FLYNN, J. T., O'GRADY, G. E., HERRERA, J., KUSHNER, B. J., CANTOLINO, S. & MILAM, W. 1977. 2690 Retrolental fibroplasia: I. Clinical observations. Arch Ophthalmol, 95, 217-23. 2691 FOLKMAN, J. 1971. Tumor angiogenesis: therapeutic implications. N Engl J Med, 285, 1182-6. 2692 FORD, K. M., SAINT-GENIEZ, M., WALSHE, T., ZAHR, A. & D'AMORE, P. A. 2011. Expression and 2693 role of VEGF in the adult retinal pigment epitheliuMANUSCRIPTm. Invest Ophthalmol Vis Sci, 52, 2694 9478-87. 2695 FORD, K. M., SAINT-GENIEZ, M., WALSHE, T. E. & D'AMORE, P. A. 2012. Expression and Role of 2696 VEGF-A in the Ciliary Body. Invest Ophthalmol Vis Sci, 53, 7520-7. 2697 FULTON, A. B., HANSEN, R. M., MOSKOWITZ, A. & AKULA, J. D. 2009. The neurovascular retina 2698 in retinopathy of prematurity. Prog Retin Eye Res, 28, 452-82. 2699 FULTON, A. B., HANSEN, R. M., MOSKOWITZ, A. & BARNABY, A. M. 2005. Multifocal ERG in 2700 subjects with a history of retinopathy of prematurity. Doc Ophthalmol, 111, 7-13. 2701 FULTON, A. B., HANSEN, R. M., PETERSEN, R. A. & VANDERVEEN, D. K. 2001. The rod 2702 photoreceptors in retinopathy of prematurity: an electroretinographic study. Arch 2703 Ophthalmol, 119, 499-505. 2704 FUNG, A. E., ROSENFELD, P. J. & REICHEL, E. 2006. The International Intravitreal 2705 Bevacizumab Safety Survey: using the internet to assess drug safety worldwide. Br J 2706 Ophthalmol, 90, 1344-9. 2707 GADKARI, S. S., KULKARNI, S. R., KAMDAR, R. R. & DESHPANDE, M. 2015. Successful Surgical 2708 Management of Retinopathy of Prematurity Showing Rapid Progression despite 2709 Extensive RetinalACCEPTED Photocoagulation. Middle East Afr J Ophthalmol, 22, 393-5. 2710 GELONECK, M. M., CHUANG, A. Z., CLARK, W. L., HUNT, M. G., NORMAN, A. A., PACKWOOD, E. A., 2711 TAWANSY, K. A., MINTZ-HITTNER, H. A. & GROUP, B.-R. C. 2014. Refractive outcomes 2712 following bevacizumab monotherapy compared with conventional laser treatment: a 2713 randomized clinical trial. JAMA Ophthalmol, 132, 1327-33. 2714 GENETOS, D. C., CHEUNG, W. K., DECARIS, M. L. & LEACH, J. K. 2010. Oxygen tension modulates 2715 neurite outgrowth in PC12 cells through a mechanism involving HIF and VEGF. J Mol 2716 Neurosci, 40, 360-6.

75 2717 GERHARDT, H., GOLDING, M., FRUTTIGER,ACCEPTED M., MANUSCRIPTRUHRBERG, C., LUNDKVIST, A., ABRAMSSON, A., 2718 JELTSCH, M., MITCHELL, C., ALITALO, K., SHIMA, D. & BETSHOLTZ, C. 2003. VEGF guides 2719 angiogenic sprouting utilizing endothelial tip cell filopodia. J Cell Biol, 161, 1163-77. 2720 GILBERT, C. 2008. Retinopathy of prematurity: a global perspective of the epidemics, 2721 population of babies at risk and implications for control. Early Hum Dev, 84, 77-82. 2722 GILBERT, C., FIELDER, A., GORDILLO, L., QUINN, G., SEMIGLIA, R., VISINTIN, P. & ZIN, A. 2005. 2723 Characteristics of infants with severe retinopathy of prematurity in countries with 2724 low, moderate, and high levels of development: implications for screening programs. 2725 Pediatrics, 115, e518-25. 2726 GILBERT, C., WORMALD, R., FIELDER, A., DEORARI, A., ZEPEDA-ROMERO, L. C., QUINN, G., 2727 VINEKAR, A., ZIN, A. & DARLOW, B. 2016. Potential for a paradigm change in the 2728 detection of retinopathy of prematurity requiring treatment. Arch Dis Child Fetal 2729 Neonatal Ed, 101, F6-9. 2730 GILBERT, W. S., DOBSON, V., QUINN, G. E., REYNOLDS, J., TUNG, B. & FLYNN, J. T. 1992. The 2731 correlation of visual function with posterior retinal structure in severe retinopathy of 2732 prematurity. Cryotherapy for Retinopathy of Prematurity Cooperative Group. Arch 2733 Ophthalmol, 110, 625-31. 2734 GILBERT, W. S., QUINN, G. E., DOBSON, V., REYNOLDS, J., HARDY, R. J. & PALMER, E. A. 1996. 2735 Partial retinal detachment at 3 months after threshold retinopathy of prematurity. 2736 Long-term structural and functional outcome. Multicenter Trial of Cryotherapy for 2737 Retinopathy of Prematurity Cooperative Group. Arch Ophthalmol, 114, 1085-91. 2738 GLASER, B. M., DAMORE, P. A., MICHELS, R. G., PATZ, A. & FENSELAU, A. 1980. Demonstration 2739 of Vasoproliferative Activity from Mammalian Retina. Journal of Cell Biology, 84, 298- 2740 304. 2741 GLUCKMAN, P. D. & HARDING, J. E. 1997. The physiology and pathophysiology of intrauterine 2742 growth retardation. Hormone Research, 48, 11-16. 2743 GOLE, G. A. 1985. Animal Models of Retinopathy of Prematurity. In: SILVERMAN, W. A. & 2744 FLYNN, J. T. (eds.) Retinopathy of Prematurity.MANUSCRIPT London: Blackwell. 2745 GOLE, G. A., GUNN, D. J. & CARTWRIGHT, D. 2011. Outcome Of Laser Treatment Of AP-ROP In 2746 ELBW Babies. Investigative Ophthalmology & Visual Science, 52, 3147-3147. 2747 GOMEZ, R., ROMERO, R., GHEZZI, F., YOON, B. H., MAZOR, M. & BERRY, S. M. 1998. The fetal 2748 inflammatory response syndrome. Am J Obstet Gynecol, 179, 194-202. 2749 GOOD, W. V. & EARLY TREATMENT FOR RETINOPATHY OF PREMATURITY COOPERATIVE, G. 2750 2004. Final results of the Early Treatment for Retinopathy of Prematurity (ETROP) 2751 randomized trial. Trans Am Ophthalmol Soc, 102, 233-48; discussion 248-50. 2752 GOOD, W. V., HARDY, R. J., DOBSON, V., PALMER, E. A., PHELPS, D. L., QUINTOS, M. & TUNG, B. 2753 2005. The incidence and course of retinopathy of prematurity: findings from the early 2754 treatment for retinopathy of prematurity study. Pediatrics, 116, 15-23. 2755 GROUPS, T. B.-I. A. U. K. C. 2016. Outcomes of Two Trials of Oxygen-Saturation Targets in 2756 Preterm Infants. New England Journal of Medicine, 374, 749-760. 2757 GUNN, D. J., CARTWRIGHT, D. W. & GOLE, G. A. 2014. Prevalence and outcomes of laser 2758 treatment of aggressive posterior retinopathy of prematurity. Clin Experiment 2759 Ophthalmol, 42, 459-65. 2760 GURSOY, H., BILGEC,ACCEPTED M. D., EROL, N., BASMAK, H. & COLAK, E. 2016. The macular findings on 2761 spectral-domain optical coherence tomography in premature infants with or without 2762 retinopathy of prematurity. Int Ophthalmol, 36, 591-600. 2763 GURWIN, J., TOMLINSON, L. A., QUINN, G. E., YING, G. S., BAUMRITTER, A., BINENBAUM, G., 2764 POSTNATAL, G., RETINOPATHY OF PREMATURITY STUDY, G. & THE TELEMEDICINE 2765 APPROACHES TO EVALUATING ACUTE-PHASE RETINOPATHY OF PREMATURITY 2766 COOPERATIVE, G. 2017. A Tiered Approach to Retinopathy of Prematurity Screening 2767 (TARP) Using a Weight Gain Predictive Model and a Telemedicine System. JAMA 2768 Ophthalmol .

76 2769 HAGADORN, J. I., FUREY, A. M., NGHIEM,ACCEPTED T. H., SCHMIMANUSCRIPTD, C. H., PHELPS, D. L., PILLERS, D. A., 2770 COLE, C. H. & GROUP, A. V. S. 2006. Achieved versus intended pulse oximeter 2771 saturation in infants born less than 28 weeks' gestation: the AVIOx study. Pediatrics, 2772 118, 1574-82. 2773 HALLIDAY, H. L., EHRENKRANZ, R. A. & DOYLE, L. W. 2009. Early (< 8 days) postnatal 2774 corticosteroids for preventing chronic lung disease in preterm infants. Cochrane 2775 Database Syst Rev , CD001146. 2776 HALLMAN, M., BRY, K., HOPPU, K., LAPPI, M. & POHJAVUORI, M. 1992. Inositol 2777 supplementation in premature infants with respiratory distress syndrome. N Engl J 2778 Med, 326, 1233-9. 2779 HAMMER, D. X., IFTIMIA, N. V., FERGUSON, R. D., BIGELOW, C. E., USTUN, T. E., BARNABY, A. M. 2780 & FULTON, A. B. 2008. Foveal fine structure in retinopathy of prematurity: an adaptive 2781 optics Fourier domain optical coherence tomography study. Invest Ophthalmol Vis Sci, 2782 49, 2061-70. 2783 HAN, J., KIM, S. E., LEE, S. C. & LEE, C. S. 2016. Low dose versus conventional dose of 2784 intravitreal bevacizumab injection for retinopathy of prematurity: a case series with 2785 paired-eye comparison. Acta Ophthalmol . 2786 HANSEN, R. M., MOSKOWITZ, A., AKULA, J. D. & FULTON, A. B. 2017. The neural retina in 2787 retinopathy of prematurity. Prog Retin Eye Res, 56, 32-57. 2788 HAQUE, K. N. & PAMMI, M. 2011. Pentoxifylline for treatment of sepsis and necrotizing 2789 enterocolitis in neonates. Cochrane Database Syst Rev , CD004205. 2790 HARD, A. L. & HELLSTROM, A. 2011. On safety, pharmacokinetics and dosage of bevacizumab 2791 in ROP treatment - a review. Acta Paediatr, 100, 1523-7. 2792 HARDER, B. C., SCHLICHTENBREDE, F. C., VON BALTZ, S., JENDRITZA, W., JENDRITZA, B. & 2793 JONAS, J. B. 2013. Intravitreal bevacizumab for retinopathy of prematurity: refractive 2794 error results. Am J Ophthalmol, 155, 1119-1124 e1. 2795 HARDER, B. C., VON BALTZ, S., JONAS, J. B. & SCHLICHTENBREDE, F. C. 2011. Intravitreal 2796 bevacizumab for retinopathy of prematurity.MANUSCRIPT J Ocul Pharmacol Ther, 27, 623-7. 2797 HARDY, R. J., GOOD, W. V., DOBSON, V., PALMER, E. A., PHELPS, D. L., QUINTOS, M., TUNG, B. & 2798 EARLY TREATMENT FOR RETINOPATHY OF PREMATURITY COOPERATIVE, G. 2004. 2799 Multicenter trial of early treatment for retinopathy of prematurity: study design. 2800 Control Clin Trials, 25, 311-25. 2801 HARTNETT, M. E. & PENN, J. S. 2012. Mechanisms and management of retinopathy of 2802 prematurity. N Engl J Med, 367, 2515-26. 2803 HASEGAWA, T., MCLEOD, D. S., PROW, T., MERGES, C., GREBE, R. & LUTTY, G. A. 2008. Vascular 2804 precursors in developing human retina. Invest Ophthalmol Vis Sci, 49, 2178-92. 2805 HATTAR, S., LIAO, H. W., TAKAO, M., BERSON, D. M. & YAU, K. W. 2002. Melanopsin-containing 2806 retinal ganglion cells: architecture, projections, and intrinsic photosensitivity. Science, 2807 295, 1065-70. 2808 HAY, W. W., JR., THILO, E. & CURLANDER, J. B. 1991. Pulse oximetry in neonatal medicine. Clin 2809 Perinatol, 18, 441-72. 2810 HELLER, C. D., O'SHEA, M., YAO, Q., LANGER, J., EHRENKRANZ, R. A., PHELPS, D. L., POOLE, W. 2811 K., STOLL, B., DUARA, S., OH, W., LEMONS, J., POINDEXTER, B. & NETWORK, N. N. R. 2812 2007. HumanACCEPTED milk intake and retinopathy of prematurity in extremely low birth 2813 weight infants. Pediatrics, 120, 1-9. 2814 HELLGREN, G., LOFQVIST, C., HARD, A. L., HANSEN-PUPP, I., GRAM, M., LEY, D., SMITH, L. E. & 2815 HELLSTROM, A. 2016. Serum concentrations of vascular endothelial growth factor in 2816 relation to retinopathy of prematurity. Pediatr Res, 79, 70-5. 2817 HELLSTROM, A., ENGSTROM, E., HARD, A. L., ALBERTSSON-WIKLAND, K., CARLSSON, B., 2818 NIKLASSON, A., LOFQVIST, C., SVENSSON, E., HOLM, S., EWALD, U., HOLMSTROM, G. & 2819 SMITH, L. E. H. 2003. Postnatal serum insulin-like growth factor I deficiency is 2820 associated with retinopathy of prematurity and other complications of premature 2821 birth. Pediatrics, 112, 1016-1020. 77 2822 HELLSTROM, A., HARD, A. L., ENGSTROM,ACCEPTED E., NIKLASSON MANUSCRIPT, A., ANDERSSON, E., SMITH, L. & 2823 LOFQVIST, C. 2009. Early weight gain predicts retinopathy in preterm infants: new, 2824 simple, efficient approach to screening. Pediatrics, 123, e638-45. 2825 HELLSTROM, A., LEY, D., HANSEN-PUPP, I., HALLBERG, B., LOFQVIST, C., VAN MARTER, L., VAN 2826 WEISSENBRUCH, M., RAMENGHI, L. A., BEARDSALL, K., DUNGER, D., HARD, A. L. & 2827 SMITH, L. E. 2016a. Insulin-like growth factor 1 has multisystem effects on foetal and 2828 preterm infant development. Acta Paediatr, 105, 576-86. 2829 HELLSTROM, A., LEY, D., HANSEN-PUPP, I., HALLBERG, B., RAMENGHI, L. A., LOFQVIST, C., 2830 SMITH, L. E. & HARD, A. L. 2016b. IGF-I in the clinics: Use in retinopathy of 2831 prematurity. Growth Horm IGF Res, 30-31, 75-80. 2832 HELLSTROM, A., PERRUZZI, C., JU, M. H., ENGSTROM, E., HARD, A. L., LIU, J. L., ALBERTSSON- 2833 WIKLAND, K., CARLSSON, B., NIKLASSON, A., SJODELL, L., LEROITH, D., SENGER, D. R. & 2834 SMITH, L. E. H. 2001. Low IGF-I suppresses VEGF-survival signaling in retinal 2835 endothelial cells: Direct correlation with clinical retinopathy of prematurity. 2836 Proceedings of the National Academy of Sciences of the United States of America, 98, 2837 5804-5808. 2838 HELLSTROM, A., SMITH, L. E. & DAMMANN, O. 2013. Retinopathy of prematurity. Lancet, 382, 2839 1445-57. 2840 HENAINE-BERRA, A., GARCIA-AGUIRRE, G., QUIROZ-MERCADO, H. & MARTINEZ- 2841 CASTELLANOS, M. A. 2014. Retinal fluorescein angiographic changes following 2842 intravitreal anti-VEGF therapy. J AAPOS, 18, 120-3. 2843 HENDRICKSON, A. & DRUCKER, D. 1992. The development of parafoveal and mid-peripheral 2844 human retina. Behav Brain Res, 49, 21-31. 2845 HENDRICKSON, A., POSSIN, D., VAJZOVIC, L. & TOTH, C. A. 2012. Histologic development of 2846 the human fovea from midgestation to maturity. Am J Ophthalmol, 154, 767-778 e2. 2847 HENKIND, P. & DE OLIVEIRA, L. F. 1967. Development of retinal vessels in the rat. 2848 Investigative Ophthalmology & Visual Science, 6, 520-530. 2849 HIGASHI, Y., SUKHANOV, S., ANWAR, A., SHAI, S. Y.MANUSCRIPT & DELAFONTAINE, P. 2010. IGF-1, 2850 oxidative stress and atheroprotection. Trends Endocrinol Metab, 21, 245-54. 2851 HODGE, R. D., D'ERCOLE, A. J. & O'KUSKY, J. R. 2004. Insulin-like growth factor-I accelerates 2852 the cell cycle by decreasing G1 phase length and increases cell cycle reentry in the 2853 embryonic cerebral cortex. J Neurosci, 24, 10201-10. 2854 HOERSTER, R., MUETHER, P., DAHLKE, C., MEHLER, K., OBERTHUR, A., KIRCHHOF, B. & 2855 FAUSER, S. 2013. Serum concentrations of vascular endothelial growth factor in an 2856 infant treated with ranibizumab for retinopathy of prematurity. Acta Ophthalmol, 91, 2857 e74-5. 2858 HOLMSTROM, G., EL AZAZI, M. & KUGELBERG, U. 1999. Ophthalmological follow up of 2859 preterm infants: a population based, prospective study of visual acuity and 2860 strabismus. Br J Ophthalmol, 83, 143-50. 2861 HOLMSTROM, G., HELLSTROM, A., JAKOBSSON, P., LUNDGREN, P., TORNQVIST, K. & WALLIN, 2862 A. 2016. Five years of treatment for retinopathy of prematurity in Sweden: results 2863 from SWEDROP, a national quality register. Br J Ophthalmol, 100, 1656-1661. 2864 HOLMSTROM, G. & LARSSON, E. 2008. Long-term follow-up of visual functions in prematurely 2865 born children--aACCEPTED prospective population-based study up to 10 years of age. J AAPOS, 2866 12, 157-62. 2867 HOLMSTROM, G. & LARSSON, E. 2013. Outcome of retinopathy of prematurity. Clin Perinatol, 2868 40, 311-21. 2869 HONDA, S., HIRABAYASHI, H., TSUKAHARA, Y. & NEGI, A. 2008. Acute contraction of the 2870 proliferative membrane after an intravitreal injection of bevacizumab for advanced 2871 retinopathy of prematurity. Graefes Arch Clin Exp Ophthalmol, 246, 1061-3. 2872 HOU, C., NORCIA, A. M., MADAN, A., TITH, S., AGARWAL, R. & GOOD, W. V. 2011. Visual cortical 2873 function in very low birth weight infants without retinal or cerebral pathology. Invest 2874 Ophthalmol Vis Sci, 52, 9091-8. 78 2875 HU, J., BLAIR, M. P., SHAPIRO, M. AJ.,CCEPTED LICHTENSTEIN, MANUSCRIPT S. J., GALASSO, J. M. & KAPUR, R. 2012. 2876 Reactivation of retinopathy of prematurity after bevacizumab injection. Arch 2877 Ophthalmol, 130, 1000-6. 2878 HUGHES, S. & CHAN-LING, T. 2000. Roles of endothelial cell migration and apoptosis in 2879 vascular remodeling during development of the central nervous system. 2880 Microcirculation, 7, 317-33. 2881 HUGHES, S. & CHAN-LING, T. 2004. Characterization of smooth muscle cell and pericyte 2882 differentiation in the rat retina in vivo. Invest Ophthalmol Vis Sci, 45, 2795-806. 2883 HUGHES, S., YANG, H. & CHAN-LING, T. 2000. Vascularization of the human fetal retina: roles 2884 of vasculogenesis and angiogenesis. Invest Ophthalmol Vis Sci, 41, 1217-28. 2885 HURTADO-CHONG, A., YUSTA-BOYO, M. J., VERGANO-VERA, E., BULFONE, A., DE PABLO, F. & 2886 VICARIO-ABEJON, C. 2009. IGF-I promotes neuronal migration and positioning in the 2887 olfactory bulb and the exit of neuroblasts from the subventricular zone. Eur J 2888 Neurosci, 30, 742-55. 2889 HWANG, C. K., HUBBARD, G. B., HUTCHINSON, A. K. & LAMBERT, S. R. 2015. Outcomes after 2890 Intravitreal Bevacizumab versus Laser Photocoagulation for Retinopathy of 2891 Prematurity: A 5-Year Retrospective Analysis. Ophthalmology, 122, 1008-15. 2892 HYLANDER, M. A., STROBINO, D. M., PEZZULLO, J. C. & DHANIREDDY, R. 2001. Association of 2893 human milk feedings with a reduction in retinopathy of prematurity among very low 2894 birthweight infants. J Perinatol, 21, 356-62. 2895 ICROP 1984. An international classification of retinopathy of prematurity. The Committee 2896 for the Classification of Retinopathy of Prematurity. Arch Ophthalmol, 102, 1130-4. 2897 ICROP 1987. An international classification of retinopathy of prematurity. II. The 2898 classification of retinal detachment. The International Committee for the 2899 Classification of the Late Stages of Retinopathy of Prematurity. Arch Ophthalmol, 105, 2900 906-12. 2901 ICROP 2005. The International Classification of Retinopathy of Prematurity revisited. Arch 2902 Ophthalmol, 123, 991-9. MANUSCRIPT 2903 ISAAC, M., TEHRANI, N., MIRESKANDARI, K. & MEDSCAPE 2016. Involution patterns of 2904 retinopathy of prematurity after treatment with intravitreal bevacizumab: 2905 implications for follow-up. Eye (Lond), 30, 333-41. 2906 ITTIARA, S., BLAIR, M. P., SHAPIRO, M. J. & LICHTENSTEIN, S. J. 2013. Exudative retinopathy 2907 and detachment: a late reactivation of retinopathy of prematurity after intravitreal 2908 bevacizumab. J AAPOS, 17, 323-5. 2909 JALALI, S., BALAKRISHNAN, D., ZEYNALOVA, Z., PADHI, T. R. & RANI, P. K. 2013. Serious 2910 adverse events and visual outcomes of rescue therapy using adjunct bevacizumab to 2911 laser and surgery for retinopathy of prematurity. The Indian Twin Cities Retinopathy 2912 of Prematurity Screening database Report number 5. Arch Dis Child Fetal Neonatal Ed, 2913 98, F327-33. 2914 JALALI, S., KESARWANI, S. & HUSSAIN, A. 2011. Outcomes of a protocol-based management 2915 for zone 1 retinopathy of prematurity: the Indian Twin Cities ROP Screening Program 2916 report number 2. Am J Ophthalmol, 151, 719-724 e2. 2917 JANG, S. Y., CHOI, K. S. & LEE, S. J. 2010. Delayed-onset retinal detachment after an 2918 intravitreal injectionACCEPTED of ranibizumab for zone 1 plus retinopathy of prematurity. J 2919 AAPOS, 14, 457-9. 2920 JIANG, H., GUO, R. & POWELL-COFFMAN, J. A. 2001. The Caenorhabditis elegans hif-1 gene 2921 encodes a bHLH-PAS protein that is required for adaptation to hypoxia. Proc Natl Acad 2922 Sci U S A, 98, 7916-21. 2923 JOHNSTON, E. D., BOYLE, B., JUSZCZAK, E., KING, A., BROCKLEHURST, P. & STENSON, B. J. 2011. 2924 Oxygen targeting in preterm infants using the Masimo SET Radical pulse oximeter. 2925 Arch Dis Child Fetal Neonatal Ed, 96, F429-33. 2926 KAELIN, W. G., JR. & RATCLIFFE, P. J. 2008. Oxygen sensing by metazoans: the central role of 2927 the HIF hydroxylase pathway. Mol Cell, 30, 393-402. 79 2928 KAGEYAMA, G. H. & WONG-RILEY,A CCEPTEDM. T. T. 1984. MANUSCRIPT The histochemical localization of 2929 cytochrome oxidase in the retina and lateral geniculate nucleus of the ferret, cat, and 2930 monkey, with particular reference to retinal mosaics and on/off-center visual 2931 channels. J. Neurosci., 4, 2445-2459. 2932 KAISER, R. S. & TRESE, M. T. 2001. Iris atrophy, cataracts, and hypotony following peripheral 2933 ablation for threshold retinopathy of prematurity. Arch Ophthalmol, 119, 615-7. 2934 KARACA, C., ONER, A. O., MIRZA, E., POLAT, O. A. & SAHINER, M. 2013. Bilateral effect of 2935 unilateral bevacizumab injection in retinopathy of prematurity. JAMA Ophthalmol, 2936 131, 1099-101. 2937 KARKHANEH, R., KHODABANDE, A., RIAZI-EAFAHANI, M., ROOHIPOOR, R., GHASSEMI, F., 2938 IMANI, M., DASTJANI FARAHANI, A., EBRAHIMI ADIB, N. & TORABI, H. 2016. Efficacy of 2939 intravitreal bevacizumab for zone-II retinopathy of prematurity. Acta Ophthalmol, 94, 2940 e417-20. 2941 KARP, K. A., BAUMRITTER, A., PEARSON, D. J., PISTILLI, M., YING, G. S. & QUINN, G. E. 2015. 2942 Training retinal imagers for retinopathy of prematurity (ROP) screening. J AAPOS, 19, 2943 e26. 2944 KATOCH, D., SANGHI, G., DOGRA, M. R., BEKE, N. & GUPTA, A. 2011. Structural sequelae and 2945 refractive outcome 1 year after laser treatment for type 1 prethreshold retinopathy of 2946 prematurity in Asian Indian eyes. Indian J Ophthalmol, 59, 423-6. 2947 KEMPER, A. R., PROSSER, L. A., WADE, K. C., REPKA, M. X., YING, G. S., BAUMRITTER, A., QUINN, 2948 G. E. & E, R. O. P. S. C. G. 2016. A Comparison of Strategies for Retinopathy of 2949 Prematurity Detection. Pediatrics, 137. 2950 KENNEDY, K. A., WRAGE, L. A., HIGGINS, R. D., FINER, N. N., CARLO, W. A., WALSH, M. C., 2951 LAPTOOK, A. R., FAIX, R. G., YODER, B. A., SCHIBLER, K., GANTZ, M. G., DAS, A., 2952 NEWMAN, N. S., PHELPS, D. L., HEALTH, S. S. G. O. T. E. K. S. N. I. O. C. & HUMAN 2953 DEVELOPMENT NEONATAL RESEARCH, N. 2014. Evaluating retinopathy of 2954 prematurity screening guidelines for 24- to 27-week gestational age infants. J 2955 Perinatol, 34, 311-8. MANUSCRIPT 2956 KHODABANDE, A., NIYOUSHA, M. R. & ROOHIPOOR, R. 2016. A lower dose of intravitreal 2957 bevacizumab effectively treats retinopathy of prematurity. J AAPOS . 2958 KLINGER, G., LEVY, I., SIROTA, L., BOYKO, V., LERNER-GEVA, L., REICHMAN, B. & ISRAEL 2959 NEONATAL, N. 2010. Outcome of early-onset sepsis in a national cohort of very low 2960 birth weight infants. Pediatrics, 125, e736-40. 2961 KLUFAS, M. A. & CHAN, R. V. 2015. Intravitreal anti-VEGF therapy as a treatment for 2962 retinopathy of prematurity: what we know after 7 years. J Pediatr Ophthalmol 2963 Strabismus, 52, 77-84. 2964 KLUFAS, M. A., PATEL, S. N., RYAN, M. C., PATEL GUPTA, M., JONAS, K. E., OSTMO, S., 2965 MARTINEZ-CASTELLANOS, M. A., BERROCAL, A. M., CHIANG, M. F. & CHAN, R. V. 2015. 2966 Influence of Fluorescein Angiography on the Diagnosis and Management of 2967 Retinopathy of Prematurity. Ophthalmology, 122, 1601-8. 2968 KONG, L., BHATT, A. R., DEMNY, A. B., COATS, D. K., LI, A., RAHMAN, E. Z., SMITH, O. E. & 2969 STEINKULLER, P. G. 2015a. Pharmacokinetics of bevacizumab and its effects on serum 2970 VEGF and IGF-1 in infants with retinopathy of prematurity. Invest Ophthalmol Vis Sci, 2971 56, 956-61. ACCEPTED 2972 KONG, L., BHATT, A. R., DEMNY, A. B., COATS, D. K. & STEINKULLER, P. G. 2015b. Clinical 2973 efficacy and pharmacology of bevacizumab in treatment of retinopathy of prematurity 2974 (ROP): a comparison of two dosages. Journal of American Association for Pediatric 2975 Ophthalmology and Strabismus {JAAPOS}, 19, e27. 2976 KOWANETZ, M. & FERRARA, N. 2006. Vascular endothelial growth factor signaling pathways: 2977 therapeutic perspective. Clin Cancer Res, 12, 5018-22. 2978 KOZULIN, P., NATOLI, R., BUMSTED O'BRIEN, K. M., MADIGAN, M. C. & PROVIS, J. M. 2010. The 2979 cellular expression of antiangiogenic factors in fetal primate macula. Invest 2980 Ophthalmol Vis Sci, 51, 4298-306. 80 2981 KOZULIN, P., NATOLI, R., O'BRIEN,ACCEPTED K. M., MADIGAN, MANUSCRIPT M. C. & PROVIS, J. M. 2009. Differential 2982 expression of anti-angiogenic factors and guidance genes in the developing macula. 2983 Mol Vis, 15, 45-59. 2984 KREMER, I., NISSENKORN, I., LUSKY, M. & YASSUR, Y. 1995. Late visual field changes following 2985 cryotherapy for retinopathy of prematurity stage 3. Br J Ophthalmol, 79, 267-9. 2986 KRETZER, F. L., MEHTA, R. S., JOHNSON, A. T., HUNTER, D. G., BROWN, E. S. & HITTNER, H. M. 2987 1984. Vitamin E protects against retinopathy of prematurity through action on 2988 spindle cells. Nature, 309, 793-795. 2989 KUR, J., NEWMAN, E. A. & CHAN-LING, T. 2012. Cellular and physiological mechanisms 2990 underlying blood flow regulation in the retina and choroid in health and disease. 2991 Progress in Retinal and Eye Research, 31, 377-406. 2992 KUSHNER, B. J., ESSNER, D., COHEN, I. J. & FLYNN, J. T. 1977. Retrolental Fibroplasia. II. 2993 Pathologic correlation. Arch Ophthalmol, 95, 29-38. 2994 LAMBERT, S. R., CAPONE, A., JR., CINGLE, K. A. & DRACK, A. V. 2000. Cataract and phthisis 2995 bulbi after laser photoablation for threshold retinopathy of prematurity. Am J 2996 Ophthalmol, 129, 585-91. 2997 LE CRAS, T. D., MARKHAM, N. E., TUDER, R. M., VOELKEL, N. F. & ABMAN, S. H. 2002. 2998 Treatment of newborn rats with a VEGF receptor inhibitor causes pulmonary 2999 hypertension and abnormal lung structure. Am J Physiol Lung Cell Mol Physiol, 283, 3000 L555-62. 3001 LEAUTE-LABREZE, C., DUMAS DE LA ROQUE, E., HUBICHE, T., BORALEVI, F., THAMBO, J. B. & 3002 TAIEB, A. 2008. Propranolol for severe hemangiomas of infancy. N Engl J Med, 358, 3003 2649-51. 3004 LEE, A. C., MALDONADO, R. S., SARIN, N., O'CONNELL, R. V., WALLACE, D. K., FREEDMAN, S. F., 3005 COTTEN, M. & TOTH, C. A. 2011a. Macular features from spectral-domain optical 3006 coherence tomography as an adjunct to indirect ophthalmoscopy in retinopathy of 3007 prematurity. Retina, 31, 1470-82. 3008 LEE, B. J., KIM, J. H., HEO, H. & YU, Y. S. 2012. DelayedMANUSCRIPT onset atypical vitreoretinal traction 3009 band formation after an intravitreal injection of bevacizumab in stage 3 retinopathy 3010 of prematurity. Eye (Lond), 26, 903-9; quiz 910. 3011 LEE, G. A., HILFORD, D. J. & GOLE, G. A. 2004. Diode laser treatment of pre-threshold and 3012 threshold retinopathy of prematurity. Clin Experiment Ophthalmol, 32, 164-9. 3013 LEE, G. A., LEE, L. R. & GOLE, G. A. 1998. Angle-closure glaucoma after laser treatment for 3014 retinopathy of prematurity. J AAPOS, 2, 383-4. 3015 LEE, H., PROUDLOCK, F. A. & GOTTLOB, I. 2016. Pediatric Optical Coherence Tomography in 3016 Clinical Practice-Recent Progress. Invest Ophthalmol Vis Sci, 57, OCT69-79. 3017 LEE, J. & DAMMANN, O. 2012. Perinatal infection, inflammation, and retinopathy of 3018 prematurity. Semin Fetal Neonatal Med, 17, 26-9. 3019 LEE, P. 1999. Telemedicine: opportunities and challenges for the remote care of diabetic 3020 retinopathy. Arch Ophthalmol, 117, 1639-40. 3021 LEE, S.-J., KIM, S.-Y., YOO, B. C., KIM, H. W. & KIM, Y. H. 2011b. Plasma Level Of Vascular 3022 Endothelial Growth Factor In Retinopathy Of Prematurity After Intravitreal Injection 3023 Of Bevacizumab. Investigative Ophthalmology & Visual Science, 52, 3165-3165. 3024 LEE, S. K., AZIZ, K., SINGHAL,ACCEPTED N., CRONIN, C. M., JAMES, A., LEE, D. S., MATTHEW, D., OHLSSON, 3025 A., SANKARAN, K., SESHIA, M., SYNNES, A., WALKER, R., WHYTE, R., LANGLEY, J., 3026 MACNAB, Y. C., STEVENS, B. & VON DADELSZEN, P. 2009. Improving the quality of care 3027 for infants: a cluster randomized controlled trial. CMAJ, 181, 469-76. 3028 LEPORE, D., MOLLE, F., PAGLIARA, M. M., BALDASCINO, A., ANGORA, C., SAMMARTINO, M. & 3029 QUINN, G. E. 2011. Atlas of fluorescein angiographic findings in eyes undergoing laser 3030 for retinopathy of prematurity. Ophthalmology, 118, 168-75. 3031 LEPORE, D., QUINN, G. E., MOLLE, F., BALDASCINO, A., ORAZI, L., SAMMARTINO, M., PURCARO, 3032 V., GIANNANTONIO, C., PAPACCI, P. & ROMAGNOLI, C. 2014. Intravitreal bevacizumab

81 3033 versus laser treatment in AtypeCCEPTED 1 retinopathy MANUSCRIPT of prematurity: report on fluorescein 3034 angiographic findings. Ophthalmology, 121, 2212-9. 3035 LEY, D., HANSEN-PUPP, I., NIKLASSON, A., DOMELLOF, M., FRIBERG, L. E., BORG, J., LOFQVIST, 3036 C., HELLGREN, G., SMITH, L. E., HARD, A. L. & HELLSTROM, A. 2013. Longitudinal 3037 infusion of a complex of insulin-like growth factor-I and IGF-binding protein-3 in five 3038 preterm infants: pharmacokinetics and short-term safety. Pediatr Res, 73, 68-74. 3039 LIEN, R., YU, M. H., HSU, K. H., LIAO, P. J., CHEN, Y. P., LAI, C. C. & WU, W. C. 2016. 3040 Neurodevelopmental Outcomes in Infants with Retinopathy of Prematurity and 3041 Bevacizumab Treatment. PLoS One, 11, e0148019. 3042 LINEHAM, J. D., SMITH, R. M., DAHLENBURG, G. W., KING, R. A., HASLAM, R. R., STUART, M. C. & 3043 FAULL, L. 1986. Circulating Insulin-Like Growth Factor-I Levels in Newborn 3044 Premature and Full-Term Infants Followed Longitudinally. Early Human Development, 3045 13, 37-46. 3046 LING, T. & STONE, J. 1988. The development of astrocytes in the cat retina: evidence of 3047 migration from the optic nerve. Developmental Brain Research, 44, 73-85. 3048 LINSENMEIER, R. A. 1986. Effects of light and darkness on oxygen distribution and 3049 consumption in the cat retina. J. Gen. Physiol., 88, 521-542. 3050 LOFQVIST, C., ANDERSSON, E., SIGURDSSON, J., ENGSTROM, E., HARD, A. L., NIKLASSON, A., 3051 SMITH, L. E. & HELLSTROM, A. 2006a. Longitudinal postnatal weight and insulin-like 3052 growth factor I measurements in the prediction of retinopathy of prematurity. Arch 3053 Ophthalmol, 124, 1711-8. 3054 LOFQVIST, C., ENGSTROM, E., SIGURDSSON, J., HARD, A. L., NIKLASSON, A., EWALD, U., 3055 HOLMSTROM, G., SMITH, L. E. & HELLSTROM, A. 2006b. Postnatal head growth deficit 3056 among premature infants parallels retinopathy of prematurity and insulin-like 3057 growth factor-1 deficit. Pediatrics, 117, 1930-8. 3058 LOFQVIST, C., NIKLASSON, A., ENGSTROM, E., FRIBERG, L. E., CAMACHO-HUBNER, C., LEY, D., 3059 BORG, J., SMITH, L. E. H. & HELLSTROM, A. 2009. A Pharmacokinetic and Dosing Study 3060 of Intravenous Insulin-Like Growth Factor-IMANUSCRIPT and IGF-Binding Protein-3 Complex to 3061 Preterm Infants. Pediatric Research, 65, 574-579. 3062 LONNERDAL, B. 2009. Nutritional roles of lactoferrin. Curr Opin Clin Nutr Metab Care, 12, 3063 293-7. 3064 MAKRIDES, M., GIBSON, R. A., MCPHEE, A. J., COLLINS, C. T., DAVIS, P. G., DOYLE, L. W., 3065 SIMMER, K., COLDITZ, P. B., MORRIS, S., SMITHERS, L. G., WILLSON, K. & RYAN, P. 2009. 3066 Neurodevelopmental outcomes of preterm infants fed high-dose docosahexaenoic 3067 acid: a randomized controlled trial. JAMA, 301, 175-82. 3068 MALDONADO, R. S., IZATT, J. A., SARIN, N., WALLACE, D. K., FREEDMAN, S., COTTEN, C. M. & 3069 TOTH, C. A. 2010. Optimizing hand-held spectral domain optical coherence 3070 tomography imaging for neonates, infants, and children. Invest Ophthalmol Vis Sci, 51, 3071 2678-85. 3072 MALDONADO, R. S., O'CONNELL, R., ASCHER, S. B., SARIN, N., FREEDMAN, S. F., WALLACE, D. K., 3073 CHIU, S. J., FARSIU, S., COTTEN, M. & TOTH, C. A. 2012. Spectral-domain optical 3074 coherence tomographic assessment of severity of cystoid macular edema in 3075 retinopathy of prematurity. Arch Ophthalmol, 130, 569-78. 3076 MALDONADO, R. S., ACCEPTEDO'CONNELL, R. V., SARIN, N., FREEDMAN, S. F., WALLACE, D. K., COTTEN, C. 3077 M., WINTER, K. P., STINNETT, S., CHIU, S. J., IZATT, J. A., FARSIU, S. & TOTH, C. A. 2011. 3078 Dynamics of human foveal development after premature birth. Ophthalmology, 118, 3079 2315-25. 3080 MALDONADO, R. S. & TOTH, C. A. 2013. Optical coherence tomography in retinopathy of 3081 prematurity: looking beyond the vessels. Clin Perinatol, 40, 271-96. 3082 MANZONI, P., GUARDIONE, R., BONETTI, P., PRIOLO, C., MAESTRI, A., MANSOLDO, C., 3083 MOSTERT, M., ANSELMETTI, G., SARDEI, D., BELLETTATO, M., BIBAN, P. & FARINA, D. 3084 2013a. Lutein and zeaxanthin supplementation in preterm very low-birth-weight

82 3085 neonates in neonatal intensiveACCEPTED care units: MANUSCRIPT a multicenter randomized controlled trial. 3086 Am J Perinatol, 30, 25-32. 3087 MANZONI, P., RINALDI, M., CATTANI, S., PUGNI, L., ROMEO, M. G., MESSNER, H., STOLFI, I., 3088 DECEMBRINO, L., LAFORGIA, N., VAGNARELLI, F., MEMO, L., BORDIGNON, L., SAIA, O. S., 3089 MAULE, M., GALLO, E., MOSTERT, M., MAGNANI, C., QUERCIA, M., BOLLANI, L., 3090 PEDICINO, R., RENZULLO, L., BETTA, P., MOSCA, F., FERRARI, F., MAGALDI, R., 3091 STRONATI, M., FARINA, D., ITALIAN TASK FORCE FOR THE, S. & PREVENTION OF 3092 NEONATAL FUNGAL INFECTIONS, I. S. O. N. 2009. Bovine lactoferrin supplementation 3093 for prevention of late-onset sepsis in very low-birth-weight neonates: a randomized 3094 trial. JAMA, 302, 1421-8. 3095 MANZONI, P., STOLFI, I., PEDICINO, R., VAGNARELLI, F., MOSCA, F., PUGNI, L., BOLLANI, L., 3096 POZZI, M., GOMEZ, K., TZIALLA, C., BORGHESI, A., DECEMBRINO, L., MOSTERT, M., 3097 LATINO, M. A., PRIOLO, C., GALLETTO, P., GALLO, E., RIZZOLLO, S., TAVELLA, E., 3098 LUPARIA, M., CORONA, G., BARBERI, I., TRIDAPALLI, E., FALDELLA, G., VETRANO, G., 3099 MEMO, L., SAIA, O. S., BORDIGNON, L., MESSNER, H., CATTANI, S., DELLA CASA, E., 3100 LAFORGIA, N., QUERCIA, M., ROMEO, M., BETTA, P. M., RINALDI, M., MAGALDI, R., 3101 MAULE, M., STRONATI, M., FARINA, D., ITALIAN TASK FORCE FOR THE, S. & 3102 PREVENTION OF NEONATAL FUNGAL INFECTIONS, I. S. O. N. 2013b. Human milk 3103 feeding prevents retinopathy of prematurity (ROP) in preterm VLBW neonates. Early 3104 Hum Dev, 89 Suppl 1, S64-8. 3105 MARTINEZ-CASTELLANOS, M. A., SCHWARTZ, S., HERNANDEZ-ROJAS, M. L., KON-JARA, V. A., 3106 GARCIA-AGUIRRE, G., GUERRERO-NARANJO, J. L., CHAN, R. V. & QUIROZ-MERCADO, H. 3107 2013. Long-term effect of antiangiogenic therapy for retinopathy of prematurity up to 3108 5 years of follow-up. Retina, 33, 329-38. 3109 MATAFTSI, A., DIMITRAKOS, S. A. & ADAMS, G. G. 2011. Mediators involved in retinopathy of 3110 prematurity and emerging therapeutic targets. Early Hum Dev, 87, 683-90. 3111 MCCANNEL, C. A. 2011. Meta-analysis of endophthalmitis after intravitreal injection of anti- 3112 vascular endothelial growth factor agents:MANUSCRIPT causative organisms and possible 3113 prevention strategies. Retina, 31, 654-61. 3114 MCCOLM, J. R., CUNNINGHAM, S., WADE, J., SEDOWOFIA, K., GELLEN, B., SHARMA, T., 3115 MCINTOSH, N. & FLECK, B. W. 2004. Hypoxic oxygen fluctuations produce less severe 3116 retinopathy than hyperoxic fluctuations in a rat model of retinopathy of prematurity. 3117 Pediatr Res, 55, 107-13. 3118 MCLOONE, E., O'KEEFE, M., MCLOONE, S. & LANIGAN, B. 2006. Long term functional and 3119 structural outcomes of laser therapy for retinopathy of prematurity. Br J Ophthalmol, 3120 90, 754-9. 3121 MICHAELSON, I. C., HERZ, N., LEWKOWITZ, E. & KERTESZ, D. 1954. Effect of Increased Oxygen 3122 on the Development of the Retinal Vessels - an Experimental Study. British Journal of 3123 Ophthalmology, 38, 577-587. 3124 MICHAUD, A. P., BAUMAN, N. M., BURKE, D. K., MANALIGOD, J. M. & SMITH, R. J. 2004. Spastic 3125 diplegia and other motor disturbances in infants receiving interferon-alpha. 3126 Laryngoscope, 114, 1231-6. 3127 MICIELI, J. A., SURKONT, M. & SMITH, A. F. 2009. A systematic analysis of the off-label use of 3128 bevacizumabACCEPTED for severe retinopathy of prematurity. Am J Ophthalmol, 148, 536-543 3129 e2. 3130 MILLER, J. W., ADAMIS, A. P., SHIMA, D. T., D'AMORE, P. A., MOULTON, R. S., O'REILLY, M. S., 3131 FOLKMAN, J., DVORAK, H. F., BROWN, L. F., BERSE, B. & ET AL. 1994. Vascular 3132 endothelial growth factor/vascular permeability factor is temporally and spatially 3133 correlated with ocular angiogenesis in a primate model. Am J Pathol, 145, 574-84. 3134 MINTZ-HITTNER, H. A. 2009. Reply: Intravitreal injection of bevacizumab (Avastin) for 3135 treatment of stage 3 retinopathy of prematurity in zone I or posterior zone II. Retina, 3136 29, 562-4.

83 3137 MINTZ-HITTNER, H. A., GELONECK,ACCEPTED M. M. & CHUANG, MANUSCRIPT A. Z. 2016. Clinical Management of 3138 Recurrent Retinopathy of Prematurity after Intravitreal Bevacizumab Monotherapy. 3139 Ophthalmology, 123, 1845-55. 3140 MINTZ-HITTNER, H. A., KENNEDY, K. A., CHUANG, A. Z. & GROUP, B.-R. C. 2011. Efficacy of 3141 intravitreal bevacizumab for stage 3+ retinopathy of prematurity. N Engl J Med, 364, 3142 603-15. 3143 MINTZ-HITTNER, H. A., O'MALLEY, R. E. & KRETZER, F. L. 1997. Long-term form identification 3144 vision after early, closed, lensectomy-vitrectomy for stage 5 retinopathy of 3145 prematurity. Ophthalmology, 104, 454-9. 3146 MITITELU, M., CHAUDHARY, K. M. & LIEBERMAN, R. M. 2012. An evidence-based meta- 3147 analysis of vascular endothelial growth factor inhibition in pediatric retinal diseases: 3148 part 1. Retinopathy of prematurity. J Pediatr Ophthalmol Strabismus, 49, 332-40. 3149 MIYAKE, T., SAWADA, O., KAKINOKI, M., SAWADA, T., KAWAMURA, H., OGASAWARA, K. & 3150 OHJI, M. 2010. Pharmacokinetics of bevacizumab and its effect on vascular endothelial 3151 growth factor after intravitreal injection of bevacizumab in macaque eyes. Invest 3152 Ophthalmol Vis Sci, 51, 1606-8. 3153 MOLLOY, C. S., ANDERSON, P. J., ANDERSON, V. A. & DOYLE, L. W. 2015. The long-term 3154 outcome of extremely preterm (<28 weeks' gestational age) infants with and without 3155 severe retinopathy of prematurity. J Neuropsychol . 3156 MORAN, S., O'KEEFE, M., HARTNETT, C., LANIGAN, B., MURPHY, J. & DONOGHUE, V. 2014. 3157 Bevacizumab versus diode laser in stage 3 posterior retinopathy of prematurity. Acta 3158 Ophthalmol, 92, e496-7. 3159 MORIN, J., LUU, T. M., SUPERSTEIN, R., OSPINA, L. H., LEFEBVRE, F., SIMARD, M. N., SHAH, V., 3160 SHAH, P. S., KELLY, E. N., CANADIAN NEONATAL, N. & THE CANADIAN NEONATAL 3161 FOLLOW-UP NETWORK, I. 2016. Neurodevelopmental Outcomes Following 3162 Bevacizumab Injections for Retinopathy of Prematurity. Pediatrics, 137. 3163 MORIN, J., SUPERSTEIN, R., LUU, T. M., OSPINA, L. H., LEFEBVRE, F., SIMARD, M., SHAH, V., 3164 SHAH, P. S. & KELLY, E. N. 2015. NeurodevelopmentalMANUSCRIPT outcomes of preterm infants 3165 treated with bevacizumab for severe retinopathy of prematurity. Invest Ophthalmol 3166 Vis Sci, 56, 4323. 3167 MOSHFEGHI, D. M. & BERROCAL, A. M. 2011. Retinopathy of prematurity in the time of 3168 bevacizumab: incorporating the BEAT-ROP results into clinical practice. 3169 Ophthalmology, 118, 1227-8. 3170 MSALL, M. E., PHELPS, D. L., HARDY, R. J., DOBSON, V., QUINN, G. E., SUMMERS, C. G., 3171 TREMONT, M. R. & CRYOTHERAPY FOR RETINOPATHY OF PREMATURITY 3172 COOPERATIVE, G. 2004. Educational and social competencies at 8 years in children 3173 with threshold retinopathy of prematurity in the CRYO-ROP multicenter study. 3174 Pediatrics, 113, 790-9. 3175 MUTLU, F. M. & SARICI, S. U. 2013. Treatment of retinopathy of prematurity: a review of 3176 conventional and promising new therapeutic options. Int J Ophthalmol, 6, 228-36. 3177 NAGY, J. A., DVORAK, A. M. & DVORAK, H. F. 2007. VEGF-A and the induction of pathological 3178 angiogenesis. Annu Rev Pathol, 2, 251-75. 3179 NARAYANAN, K. & WADHWA, S. 1998. Photoreceptor morphogenesis in the human retina: a 3180 scanning electronACCEPTED microscopic study. Anat Rec, 252, 133-9. 3181 NAUG, H. L., BROWNING, J., GOLE, G. A. & GOBE, G. 2000. Vitreal macrophages express 3182 vascular endothelial growth factor in oxygen-induced retinopathy. Clinical and 3183 Experimental Ophthalmology, 28, 48-52. 3184 NETWORK, S. S. G. O. T. E. K. S. N. N. R., CARLO, W. A., FINER, N. N., WALSH, M. C., RICH, W., 3185 GANTZ, M. G., LAPTOOK, A. R., YODER, B. A., FAIX, R. G., DAS, A., POOLE, W. K., 3186 SCHIBLER, K., NEWMAN, N. S., AMBALAVANAN, N., FRANTZ, I. D., 3RD, PIAZZA, A. J., 3187 SANCHEZ, P. J., MORRIS, B. H., LAROIA, N., PHELPS, D. L., POINDEXTER, B. B., COTTEN, C. 3188 M., VAN MEURS, K. P., DUARA, S., NARENDRAN, V., SOOD, B. G., O'SHEA, T. M., BELL, E. F.,

84 3189 EHRENKRANZ, R. A., WATTERBERG,ACCEPTED K. L.MANUSCRIPT & HIGGINS, R. D. 2010. Target ranges of 3190 oxygen saturation in extremely preterm infants. N Engl J Med, 362, 1959-69. 3191 NG, E. Y., LANIGAN, B. & O'KEEFE, M. 2006. Fundus fluorescein angiography in the screening 3192 for and management of retinopathy of prematurity. J Pediatr Ophthalmol Strabismus, 3193 43, 85-90. 3194 NGUYEN, B. T., MINKIEWICZ, V., MCCABE, E., CECILE, J. & MOWA, C. N. 2012. Vascular 3195 endothelial growth factor induces mRNA expression of pro-inflammatory factors in 3196 the uterine cervix of mice. Biomed Res, 33, 363-72. 3197 O'CONNOR, A. R., STEPHENSON, T., JOHNSON, A., TOBIN, M. J., MOSELEY, M. J., RATIB, S., NG, Y. 3198 & FIELDER, A. R. 2002. Long-term ophthalmic outcome of low birth weight children 3199 with and without retinopathy of prematurity. Pediatrics, 109, 12-8. 3200 O'CONNOR, A. R., STEWART, C. E., SINGH, J. & FIELDER, A. R. 2006. Do infants of birth weight 3201 less than 1500 g require additional long term ophthalmic follow up? Br J Ophthalmol, 3202 90, 451-5. 3203 O'CONNOR, A. R., WILSON, C. M. & FIELDER, A. R. 2007. Ophthalmological problems 3204 associated with preterm birth. Eye (Lond), 21, 1254-60. 3205 O'KEEFE, M., BURKE, J., ALGAWI, K. & GOGGIN, M. 1995. Diode laser photocoagulation to the 3206 vascular retina for progressively advancing retinopathy of prematurity. Br J 3207 Ophthalmol, 79, 1012-4. 3208 O'KUSKY, J. R., YE, P. & D'ERCOLE, A. J. 2000. Insulin-like growth factor-I promotes 3209 neurogenesis and synaptogenesis in the hippocampal dentate gyrus during postnatal 3210 development. J Neurosci, 20, 8435-42. 3211 OHLSSON, A. & AHER, S. M. 2012. Early erythropoietin for preventing red blood cell 3212 transfusion in preterm and/or low birth weight infants. Cochrane Database Syst Rev , 3213 CD004863. 3214 OKAMOTO, T., SHIRAI, M., KOKUBO, M., TAKAHASHI, S., KAJINO, M., TAKASE, M., SAKATA, H. & 3215 OKI, J. 2007. Human milk reduces the risk of retinal detachment in extremely low- 3216 birthweight infants. Pediatr Int, 49, 894-7. MANUSCRIPT 3217 ORTIBUS, E., LAENEN, A., VERHOEVEN, J., DE COCK, P., CASTEELS, I., SCHOOLMEESTERS, B., 3218 BUYCK, A. & LAGAE, L. 2011. Screening for cerebral visual impairment: value of a CVI 3219 questionnaire. Neuropediatrics, 42, 138-47. 3220 OSPINA, L. H., LYONS, C. J., MATSUBA, C., JAN, J. & MCCORMICK, A. Q. 2005. Argon laser 3221 photocoagulation for retinopathy of prematurity: long-term outcome. Eye (Lond), 19, 3222 1213-8. 3223 OZDINLER, P. H. & MACKLIS, J. D. 2006. IGF-I specifically enhances axon outgrowth of 3224 corticospinal motor neurons. Nat Neurosci, 9, 1371-81. 3225 PAINTER, S. L., WILKINSON, A. R., DESAI, P., GOLDACRE, M. J. & PATEL, C. K. 2015. Incidence 3226 and treatment of retinopathy of prematurity in England between 1990 and 2011: 3227 database study. Br J Ophthalmol, 99, 807-11. 3228 PALMER, E. A., BIGLAN, A. W. & HARDY, R. J. 1985. Retinal ablative therapy for active 3229 proliferative retinopathy of prematurity: history, current status and prospects. In: 3230 W.A., S. & FLYNN, J. T. (eds.) Contemporary Issues in Fetal and Neonatal Medicine 3231 (Volume 2 - Retinopathy of Prematurity). Boston: Blackwell Scientific. 3232 PALMER, E. A., FLYNN,ACCEPTED J. T., HARDY, R. J., PHELPS, D. L., PHILLIPS, C. L., SCHAFFER, D. B. & 3233 TUNG, B. 1991. Incidence and early course of retinopathy of prematurity. The 3234 Cryotherapy for Retinopathy of Prematurity Cooperative Group. Ophthalmology, 98, 3235 1628-40. 3236 PALMER, E. A., HARDY, R. J., DOBSON, V., PHELPS, D. L., QUINN, G. E., SUMMERS, C. G., KROM, C. 3237 P., TUNG, B. & CRYOTHERAPY FOR RETINOPATHY OF PREMATURITY COOPERATIVE, G. 3238 2005. 15-year outcomes following threshold retinopathy of prematurity: final results 3239 from the multicenter trial of cryotherapy for retinopathy of prematurity. Arch 3240 Ophthalmol, 123, 311-8.

85 3241 PAMMI, M. & HAQUE, K. N. 2015.A PentoxifyllineCCEPTED MANUSCRIPT for treatment of sepsis and necrotizing 3242 enterocolitis in neonates. Cochrane Database Syst Rev , CD004205. 3243 PARK, K. H., HWANG, J. M., CHOI, M. Y., YU, Y. S. & CHUNG, H. 2004. Retinal detachment of 3244 regressed retinopathy of prematurity in children aged 2 to 15 years. Retina, 24, 368- 3245 75. 3246 PATZ, A. 1954. Oxygen studies in retrolental fibroplasia. IV. Clinical and experimental 3247 observations. Am J Ophthalmol, 38, 291-308. 3248 PATZ, A., EASTHAM, A., HIGGINBOTHAM, D. H. & KLEH, T. 1953. Oxygen studies in retrolental 3249 fibroplasia. II. The production of the microscopic changes of retrolental fibroplasia in 3250 experimental animals. Am J Ophthalmol, 36, 1511-22. 3251 PENN, J. S., HENRY, M. M. & TOLMAN, B. L. 1994. Exposure to alternating hypoxia and 3252 hyperoxia causes severe proliferative retinopathy in the newborn rat. Pediatr Res, 36, 3253 724-31. 3254 PHELPS, D. L. 1990. Oxygen and developmental retinal capillary remodeling in the kitten. 3255 Invest Ophthalmol Vis Sci, 31, 2194-200. 3256 PIEH, C., AGOSTINI, H., BUSCHBECK, C., KRUGER, M., SCHULTE-MONTING, J., ZIRRGIEBEL, U., 3257 DREVS, J. & LAGREZE, W. A. 2008. VEGF-A, VEGFR-1, VEGFR-2 and Tie2 levels in plasma 3258 of premature infants: relationship to retinopathy of prematurity. Br J Ophthalmol, 92, 3259 689-93. 3260 PIERCE, E. A., FOLEY, E. D. & SMITH, L. E. 1996. Regulation of vascular endothelial growth 3261 factor by oxygen in a model of retinopathy of prematurity. Arch Ophthalmol, 114, 3262 1219-28. 3263 PLOUET, J. & MOUKADIRI, H. 1990. Characterization of the receptor to vasculotropin on 3264 bovine adrenal cortex-derived capillary endothelial cells. J Biol Chem, 265, 22071-4. 3265 POETS, C. F. & FRANZ, A. R. 2017. Automated FiO2 control: nice to have, or an essential 3266 addition to neonatal intensive care? Arch Dis Child Fetal Neonatal Ed, 102, F5-F6. 3267 POLIN, R. A. & BATEMAN, D. 2013. Oxygen-saturation targets in preterm infants. N Engl J 3268 Med, 368, 2141-2. MANUSCRIPT 3269 PRIDE, H. B., TOLLEFSON, M. & SILVERMAN, R. 2013. What's new in pediatric ?: 3270 part II. Treatment. J Am Acad Dermatol, 68, 899 e1-11; quiz 910-2. 3271 PROVIS, J. M. 2001. Development of the primate retinal vasculature. Prog Retin Eye Res, 20, 3272 799-821. 3273 PROVIS, J. M., LEECH, J., DIAZ, C. M., PENFOLD, P. L., STONE, J. & KESHET, E. 1997. 3274 Development of the human retinal vasculature: cellular relations and VEGF 3275 expression. Exp Eye Res, 65, 555-68. 3276 PROVIS, J. M. & MADIGAN, M. C. 2013. Foveal Development and Photoreceptor Development. 3277 In: HARTNETT, M. E. (ed.) Pediatric Retina. Philadelphia: Wolters-Kluwer. 3278 PROVIS, J. M., VAN DRIEL, D., BILLSON, F. A. & RUSSELL, P. 1985. Development of the human 3279 retina: patterns of cell distribution and redistribution in the ganglion cell layer. J 3280 Comp Neurol, 233, 429-51. 3281 PURCARO, V., BALDASCINO, A., PAPACCI, P., GIANNANTONIO, C., MOLISSO, A., MOLLE, F., 3282 LEPORE, D. & ROMAGNOLI, C. 2012. Fluorescein angiography and retinal vascular 3283 development in premature infants. J Matern Fetal Neonatal Med, 25 Suppl 3, 53-6. 3284 PURPURA, D. P. 1975.ACCEPTED Morphogenesis of visual cortex in the preterm infant. In: BRAZIER, M. 3285 A. B. (ed.) Growth and development of the brain : nutritional, genetic, and 3286 environmental factors. New York: Raven. 3287 QUINN, G. E. & DARLOW, B. A. 2016. Concerns for Development After Bevacizumab 3288 Treatment of ROP. Pediatrics, 137. 3289 QUINN, G. E. & DOBSON, V. 1996. Outcome of prematurity and retinopathy of prematurity. 3290 Curr Opin Ophthalmol, 7, 51-6. 3291 QUINN, G. E., DOBSON, V., BARR, C. C., DAVIS, B. R., FLYNN, J. T., PALMER, E. A., ROBERTSON, J. 3292 & TRESE, M. T. 1991. Visual acuity in infants after vitrectomy for severe retinopathy of 3293 prematurity. Ophthalmology, 98, 5-13. 86 3294 QUINN, G. E., DOBSON, V., DAVITT,ACCEPTED B. V., WALLACE, MANUSCRIPT D. K., HARDY, R. J., TUNG, B., LAI, D., GOOD, 3295 W. V. & EARLY TREATMENT FOR RETINOPATHY OF PREMATURITY COOPERATIVE, G. 3296 2013. Progression of myopia and high myopia in the Early Treatment for Retinopathy 3297 of Prematurity study: findings at 4 to 6 years of age. J AAPOS, 17, 124-8. 3298 QUINN, G. E., DOBSON, V., SAIGAL, S., PHELPS, D. L., HARDY, R. J., TUNG, B., SUMMERS, C. G., 3299 PALMER, E. A. & GROUP, C.-R. C. 2004. Health-related quality of life at age 10 years in 3300 very low-birth-weight children with and without threshold retinopathy of 3301 prematurity. Arch Ophthalmol, 122, 1659-66. 3302 QUINN, G. E., GILBERT, C., DARLOW, B. A. & ZIN, A. 2010. Retinopathy of prematurity: an 3303 epidemic in the making. Chin Med J (Engl), 123, 2929-37. 3304 QUINN, G. E., YING, G. S., DANIEL, E., HILDEBRAND, P. L., ELLS, A., BAUMRITTER, A., KEMPER, 3305 A. R., SCHRON, E. B., WADE, K. & E, R. O. P. C. G. 2014. Validity of a telemedicine system 3306 for the evaluation of acute-phase retinopathy of prematurity. JAMA Ophthalmol, 132, 3307 1178-84. 3308 QUINN, G. E., YING, G. S., REPKA, M. X., SIATKOWSKI, R. M., HOFFMAN, R., MILLS, M. D., 3309 MORRISON, D., DANIEL, E., BAUMRITTER, A., HILDEBRAND, P. L., SCHRON, E. B., ELLS, 3310 A. L., WADE, K. & KEMPER, A. R. 2016. Timely implementation of a retinopathy of 3311 prematurity telemedicine system. J AAPOS, 20, 425-430 e1. 3312 RAHI, J. S., CABLE, N. & BRITISH CHILDHOOD VISUAL IMPAIRMENT STUDY, G. 2003. Severe 3313 visual impairment and blindness in children in the UK. Lancet, 362, 1359-65. 3314 RAO, S., CHUN, C., FAN, J., KOFRON, J. M., YANG, M. B., HEGDE, R. S., FERRARA, N., 3315 COPENHAGEN, D. R. & LANG, R. A. 2013. A direct and melanopsin-dependent fetal light 3316 response regulates mouse eye development. Nature, 494, 243-6. 3317 RATTO, G. M., ROBINSON, D. W., YAN, B. & MCNAUGHTON, P. A. 1991. Development of the 3318 light response in neonatal mammalian rods. Nature, 351, 654-7. 3319 RECSAN, Z., SZAMOSI, A., KARKO, C., VAMOS, R., SEBESTYEN, M., FODOR, M. & SALACZ, G. 3320 2004. [Refraction and visual acuity after laser coagulation treatment for retinopathy 3321 of prematurity in stage 3+]. Ophthalmologe,MANUSCRIPT 101, 45-9. 3322 REINDERS, M. E., SHO, M., IZAWA, A., WANG, P., MUKH OPADHYAY, D., KOSS, K. E., GEEHAN, C. 3323 S., LUSTER, A. D., SAYEGH, M. H. & BRISCOE, D. M. 2003. Proinflammatory functions of 3324 vascular endothelial growth factor in alloimmunity. J Clin Invest, 112, 1655-65. 3325 REPKA, M. X., TUNG, B., GOOD, W. V., CAPONE, A., JR. & SHAPIRO, M. J. 2011. Outcome of eyes 3326 developing retinal detachment during the Early Treatment for Retinopathy of 3327 Prematurity study. Arch Ophthalmol, 129, 1175-9. 3328 REPKA, M. X., TUNG, B., GOOD, W. V., SHAPIRO, M., CAPONE, A., JR., BAKER, J. D., BARR, C. C., 3329 PHELPS, D. L. & VAN HEUVEN, W. A. 2006. Outcome of eyes developing retinal 3330 detachment during the Early Treatment for Retinopathy of Prematurity Study 3331 (ETROP). Arch Ophthalmol, 124, 24-30. 3332 REYNOLDS, J., DOBSON, V., QUINN, G. E., GILBERT, W. S., TUNG, B., ROBERTSON, J. & FLYNN, J. 3333 T. 1993. Prediction of visual function in eyes with mild to moderate posterior pole 3334 residua of retinopathy of prematurity. Cryotherapy for Retinopathy of Prematurity 3335 Cooperative Group. Arch Ophthalmol, 111, 1050-6. 3336 REYNOLDS, J. D. 2011. Bevacizumab for retinopathy of prematurity. N Engl J Med, 364, 677-8. 3337 REYNOLDS, J. D., DOBSON,ACCEPTED V., QUINN, G., GILBERT, W. S., TUNG, B. & FLYNN, J. T. 1995. Visual 3338 Function and Structure in ROP. In: SHAPIRO, M., BIGLAN, A. W. & MILLER, M. T. (eds.) 3339 PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON RETINOPATHY OF 3340 PREMATURITY. Amsterdam: Kugler. 3341 REYNOLDS, J. D., DOBSON, V., QUINN, G. E., FIELDER, A. R., PALMER, E. A., SAUNDERS, R. A., 3342 HARDY, R. J., PHELPS, D. L., BAKER, J. D., TRESE, M. T., SCHAFFER, D., TUNG, B., CRYO, R. 3343 O. P. & GROUPS, L.-R. C. S. 2002. Evidence-based screening criteria for retinopathy of 3344 prematurity: natural history data from the CRYO-ROP and LIGHT-ROP studies. Arch 3345 Ophthalmol, 120, 1470-6.

87 3346 RISTORI, C., FILIPPI, L., DAL MONTE,ACCEPTED M., MARTINI, MANUSCRIPT D., CAMMALLERI, M., FORTUNATO, P., LA 3347 MARCA, G., FIORINI, P. & BAGNOLI, P. 2011. Role of the adrenergic system in a mouse 3348 model of oxygen-induced retinopathy: antiangiogenic effects of beta-adrenoreceptor 3349 blockade. Invest Ophthalmol Vis Sci, 52, 155-70. 3350 RIVERA, J. C., MADAAN, A., ZHOU, T. E. & CHEMTOB, S. 2016. Review of the mechanisms and 3351 therapeutic avenues for retinal and choroidal vascular dysfunctions in retinopathy of 3352 prematurity. Acta Paediatr, 105, 1421-1433. 3353 ROBERTO, K. A., CONAWAY, K. J., TUDOR, S. D., KAUFMAN, C. E. & PENN, J. S. 1996. Systemic 3354 oxygen fluctuation contributes to development of retinopathy in newborn rats by 3355 augmenting the production of retinal VEGF. Investigative Ophthalmology & Visual 3356 Science, 37, 612-612. 3357 ROBINSON, C. J. & STRINGER, S. E. 2001. The splice variants of vascular endothelial growth 3358 factor (VEGF) and their receptors. J Cell Sci, 114, 853-65. 3359 ROTHMAN, A. L., TRAN-VIET, D., GUSTAFSON, K. E., GOLDSTEIN, R. F., MAGUIRE, M. G., TAI, V., 3360 SARIN, N., TONG, A. Y., HUANG, J., KUPPER, L., COTTEN, C. M., FREEDMAN, S. F. & TOTH, 3361 C. A. 2015a. Poorer neurodevelopmental outcomes associated with cystoid macular 3362 edema identified in preterm infants in the intensive care nursery. Ophthalmology, 3363 122, 610-9. 3364 ROTHMAN, A. L., TRAN-VIET, D., VAJZOVIC, L., TAI, V., SARIN, N., HOLGADO, S., GUSTAFSON, K. 3365 E., COTTEN, C. M., FREEDMAN, S. F. & TOTH, C. A. 2015b. Functional Outcomes of Young 3366 Infants with and without Macular Edema. Retina, 35, 2018-27. 3367 RUDOLPH, A. M. 1970. The changes in the circulation after birth. Their importance in 3368 congenital heart disease. Circulation, 41, 343-59. 3369 SAFRAN, M. & KAELIN, W. G., JR. 2003. HIF hydroxylation and the mammalian oxygen-sensing 3370 pathway. J Clin Invest, 111, 779-83. 3371 SAHNI, J., SUBHEDAR, N. V. & CLARK, D. 2005. Treated threshold stage 3 versus 3372 spontaneously regressed subthreshold stage 3 retinopathy of prematurity: a study of 3373 motility, refractive, and anatomical outcomesMANUSCRIPT at 6 months and 36 months. Br J 3374 Ophthalmol, 89, 154-9. 3375 SAINT-GENIEZ, M., MAHARAJ, A. S. R., WALSHE, T. E., TUCKER, B. A., SEKIYAMA, E., KURIHARA, 3376 T., DARLAND, D. C., YOUNG, M. J. & D'AMORE, P. A. 2008. Endogenous VEGF Is Required 3377 for Visual Function: Evidence for a Survival Role on Muller Cells and Photoreceptors. 3378 PLos One, 3. 3379 SANDERCOE, T. M., MADIGAN, M. C., BILLSON, F. A., PENFOLD, P. L. & PROVIS, J. M. 1999. 3380 Astrocyte proliferation during development of the human retinal vasculature. Exp Eye 3381 Res, 69, 511-23. 3382 SANGHI, G., DOGRA, M. R., DAS, P., VINEKAR, A., GUPTA, A. & DUTTA, S. 2009. Aggressive 3383 posterior retinopathy of prematurity in Asian Indian babies: spectrum of disease and 3384 outcome after laser treatment. Retina, 29, 1335-9. 3385 SANGHI, G., DOGRA, M. R., DOGRA, M., KATOCH, D. & GUPTA, A. 2012. A hybrid form of 3386 retinopathy of prematurity. Br J Ophthalmol, 96, 519-22. 3387 SANGHI, G., DOGRA, M. R., KATOCH, D. & GUPTA, A. 2014. Aggressive posterior retinopathy of 3388 prematurity in infants >/= 1500 g birth weight. Indian J Ophthalmol, 62, 254-7. 3389 SANGHI, G., DOGRA,ACCEPTED M. R., VINEKAR, A. & GUPTA, A. 2010. Frequency-doubled Nd:YAG (532 3390 nm green) versus diode laser (810 nm) in treatment of retinopathy of prematurity. Br 3391 J Ophthalmol, 94, 1264-5. 3392 SATO, T., KUSAKA, S., SHIMOJO, H. & FUJIKADO, T. 2009. Simultaneous analyses of vitreous 3393 levels of 27 cytokines in eyes with retinopathy of prematurity. Ophthalmology, 116, 3394 2165-9. 3395 SATO, T., SHIMA, C. & KUSAKA, S. 2011. Vitreous levels of angiopoietin-1 and angiopoietin-2 3396 in eyes with retinopathy of prematurity. Am J Ophthalmol, 151, 353-7 e1. 3397 SATO, T., WADA, K., ARAHORI, H., KUNO, N., IMOTO, K., IWAHASHI-SHIMA, C. & KUSAKA, S. 3398 2012. Serum concentrations of bevacizumab (avastin) and vascular endothelial 88 3399 growth factor in infants withACCEPTED retinopathy MANUSCRIPT of prematurity. Am J Ophthalmol, 153, 327- 3400 333 e1. 3401 SAUGSTAD, O. D. & AUNE, D. 2014. Optimal oxygenation of extremely low birth weight 3402 infants: a meta-analysis and systematic review of the oxygen saturation target studies. 3403 Neonatology, 105, 55-63. 3404 SCHLINGEMANN, R. O. & VAN HINSBERGH, V. W. 1997. Role of vascular permeability 3405 factor/vascular endothelial growth factor in eye disease. Br J Ophthalmol, 81, 501-12. 3406 SCHMIDT, B., DAVIS, P. G., ASZTALOS, E. V., SOLIMANO, A. & ROBERTS, R. S. 2014. Association 3407 between severe retinopathy of prematurity and nonvisual disabilities at age 5 years. 3408 JAMA, 311, 523-5. 3409 SCHMIDT, B., ROBERTS, R. S., DAVIS, P., DOYLE, L. W., BARRINGTON, K. J., OHLSSON, A., 3410 SOLIMANO, A., TIN, W. & CAFFEINE FOR APNEA OF PREMATURITY TRIAL, G. 2007. 3411 Long-term effects of caffeine therapy for apnea of prematurity. N Engl J Med, 357, 3412 1893-902. 3413 SCHMIDT, B., WHYTE, R. K., ASZTALOS, E. V., MODDEMANN, D., POETS, C., RABI, Y., SOLIMANO, 3414 A., ROBERTS, R. S. & CANADIAN OXYGEN TRIAL, G. 2013. Effects of targeting higher vs 3415 lower arterial oxygen saturations on death or disability in extremely preterm infants: 3416 a randomized clinical trial. JAMA, 309, 2111-20. 3417 SCHULENBURG, W. E. & TSANAKTSIDIS, G. 2004. Variations in the morphology of retinopathy 3418 of prematurity in extremely low birthweight infants. Br J Ophthalmol, 88, 1500-3. 3419 SEABER, J. H., MACHEMER, R., ELIOTT, D., BUCKLEY, E. G., DEJUAN, E. & MARTIN, D. F. 1995. 3420 Long-term visual results of children after initially successful vitrectomy for stage V 3421 retinopathy of prematurity. Ophthalmology, 102, 199-204. 3422 SEARS, J. E. 2008. Anti-vascular endothelial growth factor and retinopathy of prematurity. Br 3423 J Ophthalmol, 92, 1437-8. 3424 SEMENZA, G. L. 2003. Targeting HIF-1 for cancer therapy. Nat Rev Cancer, 3, 721-32. 3425 SEMENZA, G. L. 2004. Hydroxylation of HIF-1: oxygen sensing at the molecular level. 3426 Physiology (Bethesda), 19, 176-82. MANUSCRIPT 3427 SEMENZA, G. L. & WANG, G. L. 1992. A nuclear factor induced by hypoxia via de novo protein 3428 synthesis binds to the human erythropoietin gene enhancer at a site required for 3429 transcriptional activation. Mol Cell Biol, 12, 5447-54. 3430 SENGER, D. R., CONNOLLY, D. T., VAN DE WATER, L., FEDER, J. & DVORAK, H. F. 1990. 3431 Purification and NH2-terminal amino acid sequence of guinea pig tumor-secreted 3432 vascular permeability factor. Cancer Res, 50, 1774-8. 3433 SENGER, D. R., GALLI, S. J., DVORAK, A. M., PERRUZZI, C. A., HARVEY, V. S. & DVORAK, H. F. 3434 1983. Tumor cells secrete a vascular permeability factor that promotes accumulation 3435 of ascites fluid. Science, 219, 983-5. 3436 SENTILHES, L., MICHEL, C., LECOURTOIS, M., CATTEAU, J., BOURGEOIS, P., LAUDENBACH, V., 3437 MARRET, S. & LAQUERRIERE, A. 2010. Vascular endothelial growth factor and its high- 3438 affinity receptor (VEGFR-2) are highly expressed in the human forebrain and 3439 cerebellum during development. J Neuropathol Exp Neurol, 69, 111-28. 3440 SHAH, P. K., NARENDRAN, V., KALPANA, N. & GILBERT, C. 2009a. Severe retinopathy of 3441 prematurity in big babies in India: history repeating itself? Indian J Pediatr, 76, 801-4. 3442 SHAH, P. K., NARENDRAN,ACCEPTED V., KALPANA, N. & TAWANSY, K. A. 2009b. Anatomical and visual 3443 outcome of stages 4 and 5 retinopathy of prematurity. Eye (Lond), 23, 176-80. 3444 SHAH, P. K., PRABHU, V., KARANDIKAR, S. S., RANJAN, R., NARENDRAN, V. & KALPANA, N. 3445 2016. Retinopathy of prematurity: Past, present and future. World J Clin Pediatr, 5, 35- 3446 46. 3447 SHAH, P. S., LEE, S. K., LUI, K., SJORS, G., MORI, R., REICHMAN, B., HAKANSSON, S., FELICIANO, 3448 L. S., MODI, N., ADAMS, M., DARLOW, B., FUJIMURA, M., KUSUDA, S., HASLAM, R., MIREA, 3449 L. & INTERNATIONAL NETWORK FOR EVALUATING OUTCOMES OF, N. 2014. The 3450 International Network for Evaluating Outcomes of very low birth weight, very

89 3451 preterm neonates (iNeo):A aCCEPTED protocol for MANUSCRIPT collaborative comparisons of international 3452 health services for quality improvement in neonatal care. BMC Pediatr, 14, 110. 3453 SHAKIB, M., DE OLIVEIRA, L. F. & HENKIND, P. 1968. Development of retinal vessels. II. 3454 Earliest stages of vessel formation. Invest Ophthalmol, 7, 689-700. 3455 SHAO, Z., DORFMAN, A. L., SESHADRI, S., DJAVARI, M., KERMORVANT-DUCHEMIN, E., 3456 SENNLAUB, F., BLAIS, M., POLOSA, A., VARMA, D. R., JOYAL, J. S., LACHAPELLE, P., 3457 HARDY, P., SITARAS, N., PICARD, E., MANCINI, J., SAPIEHA, P. & CHEMTOB, S. 2011. 3458 Choroidal involution is a key component of oxygen-induced retinopathy. Invest 3459 Ophthalmol Vis Sci, 52, 6238-48. 3460 SHIMA, D. T., ADAMIS, A. P., FERRARA, N., YEO, K. T., YEO, T. K., ALLENDE, R., FOLKMAN, J. & 3461 D'AMORE, P. A. 1995. Hypoxic induction of endothelial cell growth factors in retinal 3462 cells: identification and characterization of vascular endothelial growth factor (VEGF) 3463 as the mitogen. Mol Med, 1, 182-93. 3464 SHIMA, D. T., GOUGOS, A., MILLER, J. W., TOLENTINO, M., ROBINSON, G., ADAMIS, A. P. & 3465 D'AMORE, P. A. 1996. Cloning and mRNA expression of vascular endothelial growth 3466 factor in ischemic retinas of Macaca fascicularis. Invest Ophthalmol Vis Sci, 37, 1334- 3467 40. 3468 SHINWELL, E. S., KARPLUS, M., REICH, D., WEINTRAUB, Z., BLAZER, S., BADER, D., YURMAN, S., 3469 DOLFIN, T., KOGAN, A., DOLLBERG, S., ARBEL, E., GOLDBERG, M., GUR, I., NAOR, N., 3470 SIROTA, L., MOGILNER, S., ZARITSKY, A., BARAK, M. & GOTTFRIED, E. 2000. Early 3471 postnatal dexamethasone treatment and increased incidence of cerebral palsy. Arch 3472 Dis Child Fetal Neonatal Ed, 83, F177-81. 3473 SHWEIKI, D., ITIN, A., SOFFER, D. & KESHET, E. 1992. Vascular endothelial growth factor 3474 induced by hypoxia may mediate hypoxia-initiated angiogenesis. Nature, 359, 843-5. 3475 SIATKOWSKI, R. M., GOOD, W. V., SUMMERS, C. G., QUINN, G. E. & TUNG, B. 2013. Clinical 3476 characteristics of children with severe visual impairment but favorable retinal 3477 structural outcomes from the Early Treatment for Retinopathy of Prematurity 3478 (ETROP) study. J AAPOS, 17, 129-34. MANUSCRIPT 3479 SILVERMAN, W. A. 2004. A cautionary tale about supplemental oxygen: the albatross of 3480 neonatal medicine. Pediatrics, 113, 394-6. 3481 SIMON, M., GRONE, H. J., JOHREN, O., KULLMER, J., PLATE, K. H., RISAU, W. & FUCHS, E. 1995. 3482 Expression of vascular endothelial growth factor and its receptors in human renal 3483 ontogenesis and in adult kidney. Am J Physiol, 268, F240-50. 3484 SIMONS, B. D., WILSON, M. C., HERTLE, R. W. & SCHAEFER, D. B. 1998. Bilateral hyphemas and 3485 cataracts after diode laser retinal photoablation for retinopathy of prematurity. J 3486 Pediatr Ophthalmol Strabismus, 35, 185-7. 3487 SIMPSON, J. L., MELIA, M., YANG, M. B., BUFFENN, A. N., CHIANG, M. F. & LAMBERT, S. R. 2012. 3488 Current role of cryotherapy in retinopathy of prematurity: a report by the American 3489 Academy of Ophthalmology. Ophthalmology, 119, 873-7. 3490 SINGH, R., REDDY, D. M., BARKMEIER, A. J., HOLZ, E. R., RAM, R. & CARVOUNIS, P. E. 2012. 3491 Long-term visual outcomes following lens-sparing vitrectomy for retinopathy of 3492 prematurity. Br J Ophthalmol, 96, 1395-8. 3493 SLIDSBORG, C., BANGSGAARD, R., FLEDELIUS, H., JENSEN, H., GREISEN, G. & LA COUR, M. 2012. 3494 Cerebral damageACCEPTED may be the primary risk factor for visual impairment in preschool 3495 children born extremely premature. Archives of Ophthalmology, 130, 1410-1417. 3496 SMITH, L. E. 2004. Pathogenesis of retinopathy of prematurity. Growth Horm IGF Res, 14 3497 Suppl A, S140-4. 3498 SMITH, L. E. 2008. Through the eyes of a child: understanding retinopathy through ROP the 3499 Friedenwald lecture. Invest Ophthalmol Vis Sci, 49, 5177-82. 3500 SMITH, L. E. H., SHEN, W., PERRUZZI, C., SOKER, S., KINOSE, F., XU, X. H., ROBINSON, G., 3501 DRIVER, S., BISCHOFF, J., ZHANG, B., SCHAEFFER, J. M. & SENGER, D. R. 1999. Regulation 3502 of vascular endothelial growth factor-dependent retinal neovascularization by 3503 insulin-like growth factor-1 receptor. Nature Medicine, 5, 1390-1395. 90 3504 SOLA, A., GOLOMBEK, S. G., MONTESACCEPTED BUENO, M. MANUSCRIPT T., LEMUS-VARELA, L., ZULUAGA, C., 3505 DOMINGUEZ, F., BAQUERO, H., YOUNG SARMIENTO, A. E., NATTA, D., RODRIGUEZ 3506 PEREZ, J. M., DEULOFEUT, R., QUIROGA, A., FLORES, G. L., MORGUES, M., PEREZ, A. G., 3507 VAN OVERMEIRE, B. & VAN BEL, F. 2014. Safe oxygen saturation targeting and 3508 monitoring in preterm infants: can we avoid hypoxia and hyperoxia? Acta Paediatr, 3509 103, 1009-18. 3510 SOOD, B. G., MADAN, A., SAHA, S., SCHENDEL, D., THORSEN, P., SKOGSTRAND, K., HOUGAARD, 3511 D., SHANKARAN, S., CARLO, W. & NETWORK, N. N. R. 2010. Perinatal systemic 3512 inflammatory response syndrome and retinopathy of prematurity. Pediatr Res, 67, 3513 394-400. 3514 SPANDAU, U. 2013. What is the optimal dosage for intravitreal bevacizumab for retinopathy 3515 of prematurity? Acta Ophthalmol, 91, e154. 3516 SPANDAU, U., TOMIC, Z., EWALD, U., LARSSON, E., AKERBLOM, H. & HOLMSTROM, G. 2013. 3517 Time to consider a new treatment protocol for aggressive posterior retinopathy of 3518 prematurity? Acta Ophthalmol, 91, 170-5. 3519 STEPHENSON, T., WRIGHT, S., O'CONNOR, A., FIELDER, A., JOHNSON, A., RATIB, S. & TOBIN, M. 3520 2007. Children born weighing less than 1701 g: visual and cognitive outcomes at 11- 3521 14 years. Arch Dis Child Fetal Neonatal Ed, 92, F265-70. 3522 STEWART, M. W. 2012. The expanding role of vascular endothelial growth factor inhibitors 3523 in ophthalmology. Mayo Clin Proc, 87, 77-88. 3524 STOLL, B. J., HANSEN, N. I., BELL, E. F., WALSH, M. C., CARLO, W. A., SHANKARAN, S., LAPTOOK, 3525 A. R., SANCHEZ, P. J., VAN MEURS, K. P., WYCKOFF, M., DAS, A., HALE, E. C., BALL, M. B., 3526 NEWMAN, N. S., SCHIBLER, K., POINDEXTER, B. B., KENNEDY, K. A., COTTEN, C. M., 3527 WATTERBERG, K. L., D'ANGIO, C. T., DEMAURO, S. B., TRUOG, W. E., DEVASKAR, U., 3528 HIGGINS, R. D., EUNICE KENNEDY SHRIVER NATIONAL INSTITUTE OF CHILD, H. & 3529 HUMAN DEVELOPMENT NEONATAL RESEARCH, N. 2015. Trends in Care Practices, 3530 Morbidity, and Mortality of Extremely Preterm Neonates, 1993-2012. JAMA, 314, 3531 1039-51. MANUSCRIPT 3532 STONE, J., CHAN-LING, T., PE'ER, J., ITIN, A., GNES SIN, H. & KESHET, E. 1996. Roles of vascular 3533 endothelial growth factor and astrocyte degeneration in the genesis of retinopathy of 3534 prematurity. Invest Ophthalmol Vis Sci, 37, 290-9. 3535 STONE, J., ITIN, A., ALON, T., PE'ER, J., GNESSIN, H., CHAN-LING, T. & KESHET, E. 1995. 3536 Development of retinal vasculature is mediated by hypoxia-induced vascular 3537 endothelial growth factor (VEGF) expression by neuroglia. J Neurosci, 15, 4738-47. 3538 STORCH, C. H. & HOEGER, P. H. 2010. Propranolol for infantile haemangiomas: insights into 3539 the molecular mechanisms of action. Br J Dermatol, 163, 269-74. 3540 TAHIJA, S. G., HERSETYATI, R., LAM, G. C., KUSAKA, S. & MCMENAMIN, P. G. 2014. Fluorescein 3541 angiographic observations of peripheral retinal vessel growth in infants after 3542 intravitreal injection of bevacizumab as sole therapy for zone I and posterior zone II 3543 retinopathy of prematurity. Br J Ophthalmol, 98, 507-12. 3544 THOMAS, K., SHAH, P. S., CANNING, R., HARRISON, A., LEE, S. K. & DOW, K. E. 2015. 3545 Retinopathy of prematurity: Risk factors and variability in Canadian neonatal 3546 intensive care units. J Neonatal Perinatal Med, 8, 207-14. 3547 TIN, W. & WARIYAR,ACCEPTED U. 2002. Giving small babies oxygen: 50 years of uncertainty. Semin 3548 Neonatol, 7, 361-7. 3549 TOLENTINO, M. 2011. Systemic and ocular safety of intravitreal anti-VEGF therapies for 3550 ocular neovascular disease. Surv Ophthalmol, 56, 95-113. 3551 TORRES-ALEMAN, I., PONS, S. & AREVALO, M. A. 1994. The insulin-like growth factor I system 3552 in the rat cerebellum: developmental regulation and role in neuronal survival and 3553 differentiation. J Neurosci Res, 39, 117-26. 3554 TOY, B. C., SCHACHAR, I. H., TAN, G. S. & MOSHFEGHI, D. M. 2016. Chronic Vascular Arrest as a 3555 Predictor of Bevacizumab Treatment Failure in Retinopathy of Prematurity. 3556 Ophthalmology, 123, 2166-75. 91 3557 TREMBLAY, S., MILOUDI, K., CHAYCHI,ACCEPTED S., FAVRET, MANUSCRIPT S., BINET, F., POLOSA, A., LACHAPELLE, P., 3558 CHEMTOB, S. & SAPIEHA, P. 2013. Systemic inflammation perturbs developmental 3559 retinal angiogenesis and neuroretinal function. Invest Ophthalmol Vis Sci, 54, 8125-39. 3560 TRESE, M. T. & DROSTE, P. J. 1998. Long-term postoperative results of a consecutive series of 3561 stages 4 and 5 retinopathy of prematurity. Ophthalmology, 105, 992-7. 3562 TRIGLER, L., WEAVER, R. G., JR., O'NEIL, J. W., BARONDES, M. J. & FREEDMAN, S. F. 2005. Case 3563 series of angle-closure glaucoma after laser treatment for retinopathy of prematurity. 3564 J AAPOS, 9, 17-21. 3565 TSITSIS, T., TASMAN, W., MCNAMARA, J. A., BROWN, G. & VANDER, J. 1997. Diode laser 3566 photocoagulation for retinopathy of prematurity. Trans Am Ophthalmol Soc, 95, 231-6; 3567 discussion 237-45. 3568 VAJZOVIC, L., HENDRICKSON, A. E., O'CONNELL, R. V., CLARK, L. A., TRAN-VIET, D., POSSIN, D., 3569 CHIU, S. J., FARSIU, S. & TOTH, C. A. 2012. Maturation of the human fovea: correlation 3570 of spectral-domain optical coherence tomography findings with histology. Am J 3571 Ophthalmol, 154, 779-789 e2. 3572 VANDERVEEN, D. K., BREMER, D. L., FELLOWS, R. R., HARDY, R. J., NEELY, D. E., PALMER, E. A., 3573 ROGERS, D. L., TUNG, B., GOOD, W. V. & EARLY TREATMENT FOR RETINOPATHY OF 3574 PREMATURITY COOPERATIVE, G. 2011. Prevalence and course of strabismus through 3575 age 6 years in participants of the Early Treatment for Retinopathy of Prematurity 3576 randomized trial. J AAPOS, 15, 536-40. 3577 VAUCHER, Y. E., PERALTA-CARCELEN, M., FINER, N. N., CARLO, W. A., GANTZ, M. G., WALSH, M. 3578 C., LAPTOOK, A. R., YODER, B. A., FAIX, R. G., DAS, A., SCHIBLER, K., RICH, W., NEWMAN, 3579 N. S., VOHR, B. R., YOLTON, K., HEYNE, R. J., WILSON-COSTELLO, D. E., EVANS, P. W., 3580 GOLDSTEIN, R. F., ACARREGUI, M. J., ADAMS-CHAPMAN, I., PAPPAS, A., HINTZ, S. R., 3581 POINDEXTER, B., DUSICK, A. M., MCGOWAN, E. C., EHRENKRANZ, R. A., BODNAR, A., 3582 BAUER, C. R., FULLER, J., O'SHEA, T. M., MYERS, G. J., HIGGINS, R. D. & NETWORK, S. S. G. 3583 O. T. E. K. S. N. N. R. 2012. Neurodevelopmental outcomes in the early CPAP and pulse 3584 oximetry trial. N Engl J Med, 367, 2495-504.MANUSCRIPT 3585 VINEKAR, A., AVADHANI, K., SIVAKUMAR, M., MAHENDRAD AS, P., KURIAN, M., BRAGANZA, S., 3586 SHETTY, R. & SHETTY, B. K. 2011. Understanding clinically undetected macular 3587 changes in early retinopathy of prematurity on spectral domain optical coherence 3588 tomography. Invest Ophthalmol Vis Sci, 52, 5183-8. 3589 VINEKAR, A., DOGRA, M. R., SANGTAM, T., NARANG, A. & GUPTA, A. 2007. Retinopathy of 3590 prematurity in Asian Indian babies weighing greater than 1250 grams at birth: ten 3591 year data from a tertiary care center in a developing country. Indian J Ophthalmol, 55, 3592 331-6. 3593 VINEKAR, A., GILBERT, C., DOGRA, M., KURIAN, M., SHAINESH, G., SHETTY, B. & BAUER, N. 3594 2014. The KIDROP model of combining strategies for providing retinopathy of 3595 prematurity screening in underserved areas in India using wide-field imaging, tele- 3596 medicine, non-physician graders and smart phone reporting. Indian J Ophthalmol, 62, 3597 41-9. 3598 VINEKAR, A., JAYADEV, C., MANGALESH, S., KUMAR, A. K., BAUER, N., CAPONE, A., JR., TRESE, 3599 M. & SHETTY, B. 2015a. COMPARING THE OUTCOME OF SINGLE VERSUS MULTIPLE 3600 SESSION LASERACCEPTED PHOTOABLATION OF FLAT NEOVASCULARIZATION IN ZONE 1 3601 AGGRESSIVE POSTERIOR RETINOPATHY OF PREMATURITY: A Prospective 3602 Randomized Study. Retina, 35, 2130-6. 3603 VINEKAR, A., MANGALESH, S., JAYADEV, C., MALDONADO, R. S., BAUER, N. & TOTH, C. A. 3604 2015b. Retinal Imaging of Infants on Spectral Domain Optical Coherence Tomography. 3605 BioMed Research International, 2015, 12. 3606 VINEKAR, A., TRESE, M. T., CAPONE, A., JR. & PHOTOGRAPHIC SCREENING FOR RETINOPATHY 3607 OF PREMATURITY COOPERATIVE, G. 2008. Evolution of retinal detachment in 3608 posterior retinopathy of prematurity: impact on treatment approach. Am J 3609 Ophthalmol, 145, 548-555. 92 3610 VOELKEL, N. F., VANDIVIER, R. W.A &CCEPTED TUDER, R. MANUSCRIPT M. 2006. Vascular endothelial growth factor in 3611 the lung. Am J Physiol Lung Cell Mol Physiol, 290, L209-21. 3612 WADE, K. C., PISTILLI, M., BAUMRITTER, A., KARP, K. A., GONG, A., KEMPER, A. R., YING, G. S. & 3613 QUINN, G. E. 2015. Safety of Retinopathy of Prematurity Examination and Imaging in 3614 Premature Infants. J AAPOS, 165, 994-1000 e.2. 3615 WAGNER-SCHUMAN, M., DUBIS, A. M., NORDGREN, R. N., LEI, Y., ODELL, D., CHIAO, H., WEH, E., 3616 FISCHER, W., SULAI, Y., DUBRA, A. & CARROLL, J. 2011. Race- and sex-related 3617 differences in retinal thickness and foveal pit morphology. Invest Ophthalmol Vis Sci, 3618 52, 625-34. 3619 WALLACE, D. K. & WU, K. Y. 2013. Current and future trends in treatment of severe 3620 retinopathy of prematurity. Clin Perinatol, 40, 297-310. 3621 WANG, H. 2016. Anti-VEGF therapy in the management of retinopathy of prematurity: what 3622 we learn from representative animal models of oxygen-induced retinopathy. Eye and 3623 Brain, 2016, 81-90. 3624 WANG, S. K., CALLAWAY, N. F., WALLENSTEIN, M. B., HENDERSON, M. T., LENG, T. & 3625 MOSHFEGHI, D. M. 2015. SUNDROP: six years of screening for retinopathy of 3626 prematurity with telemedicine. Can J Ophthalmol, 50, 101-6. 3627 WANI, V. B., SABTI, K. A., KUMAR, N., RAIZADA, S., KANDARI, J. A., HARBI, M. A., SAWAAN, R., 3628 RAJARAM, U., AL-NAQEEB, N. & SHUKKUR, M. 2013. Structural and functional results of 3629 indirect diode laser treatment for retinopathy of prematurity from 1999 to 2003 in 3630 Kuwait. Clin Ophthalmol, 7, 271-8. 3631 WARRASAK, S., NAWARUTKULCHAI, S. & SINSAWAT, P. 2012. Functional result and visual 3632 outcome in early versus conventional treatment of retinopathy of prematurity. J Med 3633 Assoc Thai, 95 Suppl 4, S107-15. 3634 WEAVER, D. T. & MURDOCK, T. J. 2012. Telemedicine detection of type 1 ROP in a distant 3635 neonatal intensive care unit. J AAPOS, 16, 229-33. 3636 WELLARD, J., LEE, D., VALTER, K. & STONE, J. 2005. Photoreceptors in the rat retina are 3637 specifically vulnerable to both hypoxia andMANUSCRIPT hyperoxia. Vis Neurosci, 22, 501-7. 3638 WILKINSON, A. R., HAINES, L., HEAD, K. & FIELDER, A. R. 2008. UK retinopathy of prematurity 3639 guideline. Early Hum Dev, 84, 71-4. 3640 WILKINSON-BERKA, J. L., BABIC, S., DE GOOYER, T., STITT, A. W., JAWORSKI, K., ONG, L. G., 3641 KELLY, D. J. & GILBERT, R. E. 2004. Inhibition of platelet-derived growth factor 3642 promotes pericyte loss and angiogenesis in ischemic retinopathy. Am J Pathol, 164, 3643 1263-73. 3644 WILSON, C. M., ELLS, A. L. & FIELDER, A. R. 2013. The challenge of screening for retinopathy 3645 of prematurity. Clin Perinatol, 40, 241-59. 3646 WU, C., LOFQVIST, C., SMITH, L. E., VANDERVEEN, D. K., HELLSTROM, A. & CONSORTIUM, W. 3647 2012a. Importance of early postnatal weight gain for normal retinal angiogenesis in 3648 very preterm infants: a multicenter study analyzing weight velocity deviations for the 3649 prediction of retinopathy of prematurity. Arch Ophthalmol, 130, 992-9. 3650 WU, W. C., KUO, H. K., YEH, P. T., YANG, C. M., LAI, C. C. & CHEN, S. N. 2013. An updated study of 3651 the use of bevacizumab in the treatment of patients with prethreshold retinopathy of 3652 prematurity in taiwan. Am J Ophthalmol, 155, 150-158 e1. 3653 WU, W. C., LIEN, R., LIAO,ACCEPTED P. J., WANG, N. K., CHEN, Y. P., CHAO, A. N., CHEN, K. J., CHEN, T. L., 3654 HWANG, Y. S. & LAI, C. C. 2015. Serum levels of vascular endothelial growth factor and 3655 related factors after intravitreous bevacizumab injection for retinopathy of 3656 prematurity. JAMA Ophthalmol, 133, 391-7. 3657 WU, W. C., LIN, R. I., SHIH, C. P., WANG, N. K., CHEN, Y. P., CHAO, A. N., CHEN, K. J., CHEN, T. L., 3658 HWANG, Y. S., LAI, C. C., HUANG, C. Y. & TSAI, S. 2012b. Visual acuity, optical 3659 components, and macular abnormalities in patients with a history of retinopathy of 3660 prematurity. Ophthalmology, 119, 1907-16.

93 3661 XU, Y., ZHANG, Q., KANG, X., ZHU,A Y.,CCEPTED LI, J., CHEN, MANUSCRIPT Y. & ZHAO, P. 2013. Early vitreoretinal 3662 surgery on vascularly active stage 4 retinopathy of prematurity through the 3663 preoperative intravitreal bevacizumab injection. Acta Ophthalmol, 91, e304-10. 3664 YANG, C. S., WANG, A. G., SHIH, Y. F. & HSU, W. M. 2013. Long-term biometric optic 3665 components of diode laser-treated threshold retinopathy of prematurity at 9 years of 3666 age. Acta Ophthalmol, 91, e276-82. 3667 YANG, C. S., WANG, A. G., SUNG, C. S., HSU, W. M., LEE, F. L. & LEE, S. M. 2010. Long-term visual 3668 outcomes of laser-treated threshold retinopathy of prematurity: a study of refractive 3669 status at 7 years. Eye (Lond), 24, 14-20. 3670 YOKOI, T., HIRAOKA, M., MIYAMOTO, M., YOKOI, T., KOBAYASHI, Y., NISHINA, S. & AZUMA, N. 3671 2009. Vascular abnormalities in aggressive posterior retinopathy of prematurity 3672 detected by fluorescein angiography. Ophthalmology, 116, 1377-82. 3673 YOON, B. H., ROMERO, R., KIM, C. J., KOO, J. N., CHOE, G., SYN, H. C. & CHI, J. G. 1997. High 3674 expression of tumor necrosis factor-alpha and interleukin-6 in periventricular 3675 leukomalacia. Am J Obstet Gynecol, 177, 406-11. 3676 YU, A. Y., FRID, M. G., SHIMODA, L. A., WIENER, C. M., STENMARK, K. & SEMENZA, G. L. 1998. 3677 Temporal, spatial, and oxygen-regulated expression of hypoxia-inducible factor-1 in 3678 the lung. Am J Physiol, 275, L818-26. 3679 YU, D.-Y., CRINGLE, S. J., ALDER, V. A. & SU, E.-N. 1994. Intraretinal oxygen distribution in rats 3680 as a function of systemic blood pressure. Am. J. Physiol., 267, H2498-H2507. 3681 YU, Y. S., KIM, S. J., KIM, S. Y., CHOUNG, H. K., PARK, G. H. & HEO, J. W. 2006. Lens-sparing 3682 vitrectomy for stage 4 and stage 5 retinopathy of prematurity. Korean J Ophthalmol, 3683 20, 113-7. 3684 YUODELIS, C. & HENDRICKSON, A. 1986. A qualitative and quantitative analysis of the human 3685 fovea during development. Vision Res, 26, 847-55. 3686 ZACHARY, I. 2005. Neuroprotective role of vascular endothelial growth factor: Signalling 3687 mechanisms, biological function, and therapeutic potential. Neurosignals, 14, 207- 3688 221. MANUSCRIPT 3689 ZEPEDA-ROMERO, L. C., LIERA-GARCIA, J. A., GUTIERRE Z-PADILLA, J. A., VALTIERRA- 3690 SANTIAGO, C. I. & AVILA-GOMEZ, C. D. 2010. Paradoxical vascular-fibrotic reaction 3691 after intravitreal bevacizumab for retinopathy of prematurity. Eye (Lond), 24, 931-3. 3692 ZHANG, M., CHU, S., ZENG, F. & XU, H. 2015. Bevacizumab modulates the process of fibrosis in 3693 vitro. Clin Exp Ophthalmol, 43, 173-9. 3694 ZHANG, Q., QI, Y., CHEN, L., SHI, X., BAI, Y., HUANG, L., YU, W., JIANG, Y., ZHAO, M. & LI, X. 2016. 3695 The relationship between anti-vascular endothelial growth factor and fibrosis in 3696 proliferative retinopathy: clinical and laboratory evidence. Br J Ophthalmol, 100, 3697 1443-50. 3698 ZIN, A. A., MOREIRA, M. E., BUNCE, C., DARLOW, B. A. & GILBERT, C. E. 2010. Retinopathy of 3699 prematurity in 7 neonatal units in Rio de Janeiro: screening criteria and workload 3700 implications. Pediatrics, 126, e410-7. 3701 3702 3703 ACCEPTED

94 3704 Figure 1 ACCEPTED MANUSCRIPT 3705 The Third Epidemic of ROP 3706 Figure 1 shows the distribution of the estimated number of children with severe visual impairment or 3707 blindness from ROP by geographic region or economic demographic for 2010 with the greatest burden of 3708 pre-term birth in East/Southeast Asia and Pacific, South Asia and Sub-Saharan Africa, accounting for 3709 78% of total in 2010. Visual impairment or blindness from ROP, however, is concentrated in the middle- 3710 income countries of Asia, Latin America and North Africa, the Middle East and Eastern Europe, where 3711 access to neonatal intensive care is increasing. The circle charts represent the total number of preterm 3712 infants surviving the neonatal period (<32 weeks GA in all regions, and 32-36 weeks GA in countries 3713 with a neonatal mortality rate (NMR) of greater than 5 per 1000 live births) at risk of developing any 3714 stage of ROP. [Figure 1 modified from (Blencowe et al., 2013)] 3715 3716

MANUSCRIPT

ACCEPTED

95 3717 Figure 2: ACCEPTED MANUSCRIPT 3718 Maps of Nissl-stained cells and vasculature. A: Map indicating the area occupied by all ADPase + cells 3719 (red stippling) in the 16-WG fetus. B: ADPase-incubated retina from a 12-week gestation (WG) fetus. 3720 The topography of the vasculature is butterfly-shaped and is limited to the peripapillary retina. 3721 (Reproduced with permission from (Chan-Ling et al., 2004a)) C: Vasculogenesis from 14-21 WG: The 3722 first event in retinal vasculature formation in the human retina is invasion of spindle shaped single 3723 mesenchymal vascular precursor cells (VPCs) rom the optic nerve head prior to 14 WG. The distribution 3724 of the spindle cells is shown at 14.5, 18, and 21 WG. The stippled regions show spindle cell distribution 3725 at each age; the white regions show areas with vascular cords. With maturity, the outer limit of the 3726 vascular cords expanded, where VPCs did not. At 21 WG, no spindle cells were evident in the retina. It is 3727 clear from these maps that the area formed by the vasculogenesis is not circular in the developing human 3728 retina. The X indicates the location of the incipient fovea. D: The outer limits of patent vessels in 3729 superficial and deep vascular plexuses, and Retinal Peripapillary Capillaries (RPCs) from 25-40WG. 3730 Formation of the inner vascular plexus had begun by 14 to 15 WG. These early vessels centered on the 3731 optic disc and showed a four-lobed topography. In following weeks, the inner vascular plexus extended 3732 peripherally, curving around the incipient fovea. By 32 WG, the inner plexus had reached its outer limits, 3733 leaving a narrow rim of avascular tissue at the periphery of the retina. In contrast, the formation of the 3734 outer vascular plexus began in the perifoveal region at about 25 to 26 WG and subsequently spread with 3735 an elongated topography along the horizontal meridian (HughesMANUSCRIPT et al., 2000). 3736

ACCEPTED

96 3737 Figure 3: ACCEPTED MANUSCRIPT 3738 A: Bright-and darkfield sections of retinae at 12, 14, 16 and 20 weeks gestation. At 12 WG the inner and 3739 outer nuclear layers are partially separated and the RPE is well differentiated. At the inner surface 3740 scattered neurones marked the level of the ganglion cell layer; there was no evidence of astrocytes or 3741 vessels internal to the ganglion cells. In darkfield illumination at 12WG the RPE was prominent but there 3742 was no evidence of a concentration of grains marking VEGF signal in the inner retina. At 14 and 16 WG 3743 the inner and outer nuclear layers are better differentiated and no evidence of a vascular layer internal to 3744 the GCL was evident on the hybridized sections. There was no evidence of a concentration of grains 3745 marking VEGF signal in the inner retina in these preparations. B: Section through the optic disc (od) 3746 region of a human fetal retina at 20 WG showing VEGF signal in inner retina in darkfield illumination 3747 and the corresponding brightfield image. Note the greater differentiation of the retinal layers in the 3748 section to the right (temporal) of the optic disc. The vascular layer lies superficially, VEGF mRNA 3749 expression being limited to the most distal portions indicated in the lightfield image by asterisks. A&B 3750 from (Provis et al., 1997). C: VEGF expression at the inner surface of the retina B, P3 rat retina. The 3751 horizontal arrow indicates the direction of vessel extension, and the vertical arrow indicates the extent of 3752 astrocyte spreading. Note the non-uniform distribution of VEGF-expressing cells in the astrocyte cell 3753 level, as compared with the uniform distribution of VEGF-expressing cells in the retinal pigmented 3754 epithelium. D: High magnification of P10 rat retina, showing VEGF-producing cells residing around the 3755 midline of the INL. Certain-positives are marked by arrowsMANUSCRIPT to emphasize the punctate pattern of VEGF- 3756 expressing cells (Stone et al., 1995). 3757 E: Fundus of the adult human retina. Shown are the optic disk (OD), the macula (broken circle), and the 3758 fovea centralis (asterisk). F: High magnification image of a flatmount of a human fovea. Note the vessels 3759 near the fovea (arrows); no vessels are seen on the surface of the fovea. G: Diagram of a human fetal 3760 retina flatmount at 19WG, showing the vascularized region (gray shading), the macula (circle), and the 3761 fovea (asterisk). Note how the region of vascularization curves around the fovea and does not enter. 3762 (Kozulin et al., 2009) 3763 H: CD34+ vessels in a human fetal retina at 25WG. Visible is the foveal region, showing the absence of 3764 blood vessels. I and J: CD34+ vessels in a human fetal retina at 25WG. J is distant from the fovea (I). 3765 Note the finer caliber and more regular meshwork of the vasculature in the perifoveal region (J). From 3766 (Hughes et al., 2000) ACCEPTED 3767 K: Superior-temporal area of a P35 cat retina showing the distribution of astrocytes in the area centralis. 3768 Astrocytes have not entered the area centralis (marked with an X) they are spread along the axon bundles 3769 and blood vessels that skirt around the area centralis to the raphe region (indicated with an arrow) From 3770 (Ling and Stone, 1988). 3771 L: In situ hybridization for PEDF mRNA (dark reaction product) in fetal macaque retina. Vessel profiles 3772 are indicated in by oblique arrows. In the Fd 150 fovea (lower), relatively high levels of PEDF mRNA

97 3773 are present in the GCL compared with levels in the GCL near the optic disc. Higher magnification inset ACCEPTED MANUSCRIPT 3774 of the transverse section at right shows increased PEDF expression in the RPE at Fd 150. (Kozulin et al., 3775 2010) 3776 M: In situ hybridization for Eph-A6 expression in a Fd 115 macaque retina: At the developing fovea and 3777 in adjacent temporal retina, peak Eph-A6 mRNA expression is detected in the inner ganglion cell layer 3778 (GCL) within the foveal avascular area (bracket). Large arrows mark vessels at the inner margin of the 3779 perifoveal plexus. The area between the two large arrows is devoid of vessels. Smaller arrows indicate 3780 vessel profiles deep in the GCL (GCL plexus), which is a characteristic of macular vessels.(Kozulin et al., 3781 2010) 3782 Abbreviations: Fd (Fetal Day); ganglion cell layer (GCL); inner nuclear layer (INL); inner plexiform 3783 layer (IPL); outer nuclear layer (ONL); outer plexiform layer (OPL); nerve fiber layer (NFL); retinal 3784 pigmented epithelium (RPE). 3785

MANUSCRIPT

ACCEPTED

98 3786 Figure 4: ACCEPTED MANUSCRIPT 3787 A-D: Development of human foveolar cones from 2pm plastic sections stained with azure II-methylene 3788 blue. One cone is outlined in each figure for clarity. The outer limiting membrane is marked by a small 3789 arrow. PE = pigment epithelium; OPL = outer plexiform layer: M = Muller glial cell processes; CP = 3790 cone synaptic pedicles; OS = outer segments. E:.Low power light micrograph of postnatal human fovea at 3791 birth. Section is from the center of the foveola. The black lines and arrows mark the width of the rod-free 3792 foveola; at birth the foveola is so wide that only half can be shown. P= photoreceptor nuclei; G = 3793 ganglion cell layer; 2 pm glycol methacrylate sections stained with azure II-methylene blue, 164 x. 3794 Reprinted with permission from (Yuodelis and Hendrickson, 1986). 3795 F-K: SEM of cone photoreceptors from 18WG to 5 months post-partum 3796 F: Mosaic pattern of photoreceptor inner segments at 18-19WG. Note larger cone inner segment (open 3797 arrow) surrounded by smaller cone inner segments (arrowheads). Both large and small cone inner 3798 segments are encircled by rod inner segments. x31,750. G: Scleral view of the sensory retina at 19–20 3799 weeks of gestation. Note the prominent cone inner segments (ci). In between cone inner segments 3800 connecting cilia of rods can be seen. x33,325. H: Photoreceptor inner segments at 22–23 weeks gestation. 3801 Both large (open arrow) and small (arrowheads) cone inner segments show elongated cilia from the apical 3802 ends. x33,325. I: Developing photoreceptors from posterior retina at 24-25WG. Note oval inner segments 3803 of both large (arrow) and small (arrowheads) cones. Cilia of the rods are distributed in the spaces between 3804 the cone inner segments. x31,750. J: Transversely fracturedMANUSCRIPT retinal tissue of 24-25WG human fetus 3805 showing both rod (ri) and cone (ci) inner segments and their cell bodies (cb). The ELM is located at the 3806 junctional level of the inner segments and cell bodies. At the level of the outer plexiform layer a few thick 3807 fibres are distributed. x33,325. K: Photoreceptors in posterior retina of a 5-month-old infant. Note the 3808 fully developed rods (r) and cone (c), comparable to adult photoreceptors. In the ONL the cone cell 3809 bodies (cb) are located at the uppermost row. x33,325. Reprinted with permission from (Narayanan and 3810 Wadhwa, 1998) 3811 L-O: Sections through the fovea at 40 WG using OCT and by histology. L: The layers of the retina are 3812 labeled using conventional terminology (left), and the hyperreflective bands 0 to 4 are indicated to the 3813 right. Band 0 corresponds to the layer of cone pedicles (the outer plexiform layer), Bands 1 and 2 are 3814 distinct about 1.5 mm or more from the central fovea; band 1 represents the ELM and band 2 the 3815 ellipsoid. Bands 3 andACCEPTED 4 represent components of the RPE; they are seen separately in the adult fovea, but 3816 not at 40 WG as photoreceptors are not fully elongated. M: Shows the foveal region of a 40 WG, stained 3817 with H&E and scaled to the same aspect ratio used for the OCT (1x: 4.3y). N: shows the same region at 3818 standard aspect ratio (1:1). GCL, ganglion cell layer; HFL, Henle fiber layer; INL, inner nuclear layer; 3819 IPL, inner plexiform layer; NFL, nerve fiber layer; ONL, outer nuclear layer; OPL, outer plexiform layer. 3820 Reprinted with permission from (Provis and Madigan, 2013) 3821

99 3822 ACCEPTED MANUSCRIPT 3823 Figure 5: 3824 In retinopathy of prematurity, there is initially delayed physiologic retinal vascular development, resulting 3825 in a peripheral avascular area of the retina (phase 1). Later, vasoproliferation in the form of intravitreal 3826 (or extraretinal) angiogenesis can occur at the junction of avascular and vascularized retina (phase 2). As 3827 shown in the lower panel, increased vascular endothelial growth factor (VEGF) induced by hypoxia 3828 delays physiologic retinal vascular development by interfering with ordered vascular development; 3829 decreased VEGF in high oxygen also delays physiologic retinal vascular development by reducing 3830 developmental angiogenesis. [abbreviations used: EPO erythropoietin, ERK extracellular signal-regulated 3831 kinase, HIF hypoxia-inducible factor, IGF-1 insulin-like growth factor 1, MEK mitogen-activated 3832 protein–ERK, O2 oxygen, pAKT phosphorylated protein kinase B, PI3 phosphatidylinositol 3, pJAK 3833 phosphorylated Janus kinase, pSTAT3 phosphorylated signal transducer and activator of transcription 3, 3834 ROS reactive oxygen species, and VEGFR vascular endothelial growth factor receptor.] (Reprinted with 3835 permission from (Hartnett and Penn, 2012)) 3836

MANUSCRIPT

ACCEPTED

100 3837 Figure 6: ACCEPTED MANUSCRIPT 3838 Fluorescein angiographic images before treatment and 9 months after treatment showing the junction 3839 between the vascularized and avascular retina in an infant born at 24 weeks' gestational age [A and B, 3840 bevacizumab (Avastin) injected; C and D, laser treated]. There is persistent leakage and irregular 3841 branching in both eyes at the time of treatment (A and C). In the eye injected with bevacizumab at 9 3842 months' follow-up (B), there is persistent avascular retina even posterior to the previous retinopathy 3843 (black circles) together with hypofluorescent areas (white circles), capillary tufts (white arrow), and a 3844 peripheral shunt at the junction between vascularized and avascular retina. In D, at9 months after 3845 conventional laser treatment, the previously avascular retina shows laser scars with no capillary tuft, 3846 shunts, or persistent avascular regions.. (Adapted from (Lepore et al., 2014) 3847

MANUSCRIPT

ACCEPTED

101 3848 Figure 7: ACCEPTED MANUSCRIPT 3849 Histology of newborn rat lungs at 3 weeks (“Infant”) and 3-4 months (“Adult”) both for non-treated 3850 controls and after treatment with one systemic dose of the VEGF receptor inhibitor, Su-5416 (see text). 3851 From (Le Cras et al., 2002)(Reproduced with permission). 3852 3853 3854

MANUSCRIPT

ACCEPTED

102 3855 Figure 8 ACCEPTED MANUSCRIPT 3856 Vascular endothelial growth factor (VEGF) expression in normoxia and hyperoxia in the kitten model of 3857 retinopathy of prematurity (ROP). All sections were hybridized with the probe for VEGF mRNA and are 3858 shown with the inner surface uppermost. Arrows in A, E, F, and G are positioned so that their stubs are 3859 located at the peripheral limits of blood vessels, and they point to the periphery of the retina. A: Region 3860 from the edge of vessels in a normally developing 3-day-old retina. The signal is found internal to the 3861 inner plexiform layer (i). B: The same region under bright-field illumination. The retina has separated 3862 during processing from the retinal pigment epithelium. C: Region from the edge of surviving vasculature 3863 from the retina of a kitten raised for 3 days in room air, then for 24 hours in 70% to 80% oxygen. The 3864 signal at the inner surface of the retina has been suppressed; signal is apparent at the outer surface, over 3865 the pigment epithelium. D: Region from the retina of a kitten kept in room air for 24 hours, then in 70% 3866 to 80% oxygen for 3 days. The signal at the inner surface of the retina has been suppressed; signal is 3867 apparent at the outer surface, over the pigment epithelium. E: Region from the edge of the developing 3868 vasculature in the retina of a kitten raised for 4 days in high (70% to 80%) oxygen, then in room air for 3 3869 days. Signal is strong at the inner surface and apparent over the pigment epithelium. F: Region from the 3870 edge of the developing vasculature in the retina of a kitten raised for 4 days in high (70% to 80%) 3871 oxygen, then in room air for 11 days. Signal is very strong at the inner surface. Signal at the pigment 3872 epithelium is not distinct from light reflected by the pigment granules in the pigment epithelium. G: 3873 Region from the edge of the developing vasculature in theMANUSCRIPT retina of a kitten raised for 4 days in high (70% 3874 to 80%) oxygen, then in room air for 27 days. Signal remains strong at the inner surface. A weak signal is 3875 apparent at the outer surface. From (Stone et al., 1996) 3876 3877 H-M: Localization of vascular endothelial growth factor (VEGF) expression and glial fibrillary acidic 3878 protein (GFAP) label in normoxia, retinopathy of prematurity, and hyperoxia. All sections are positioned 3879 with the inner limiting membrane uppermost. H & I: Bright-(H)( and dark-field (I) views of the labeling 3880 of VEGF mRNA in a P10 rat retina The VEGF signal lies superficial to the neurones of the GCL, marked 3881 with arrows. The arrows in H and I point to the same position in the section. J-K: Bright- and dark-field 3882 views of the labeling of VEGF mRNA in a P15 retina, post 4 days of oxygen exposure from P0-P4. The 3883 VEGF signal is distributed throughout the ganglion cell layer and is intense over many neurone-like 3884 somas. L: Bright-fieldACCEPTED view of the labeling of the VEGF mRNA probe hybridized in a P31 animal, 27 3885 days post 4 days of exposure from P0-P4. The VEGF signal is strong over the neurones of the GCL and 3886 weak over the more internal layer of retina, where GFAP+ labeling is strong (not shown). M: GFAP 3887 labeling in the retina of a 4-day-old animal exposed to high oxygen for 3 days. The labeling is normal in 3888 intensity and in its location internal to the GCL. ipl = inner plexiform layer. From (Stone et al., 1996) 3889

103 3890 In marked contrast, gradual withdrawal from hyperoxia (Supplemental oxygen therapy - SOT) resulted in ACCEPTED MANUSCRIPT 3891 a level of VEGF expression closer in extent to that seen during normal development in the cat retina (N- 3892 P) Adjacent sections from the retina of a kitten raised for 4 days in high (70% to 80%) oxygen and then in 3893 45% oxygen for 3 days. (N) The distribution of GFAP+ cells is normal, lying around vessels and along 3894 axons superficial to neurons of the ganglion cell layer (gel). (O+P) Bright- and dark-field views show 3895 that, as in the normally developing retina, the VEGF signal concentrates over structures that lie 3896 superficial to the neurones (marked by arrows) of the ganglion cell layer. From (Stone et al., 1996) 3897

MANUSCRIPT

ACCEPTED

104 ACCEPTED MANUSCRIPT

MANUSCRIPT

ACCEPTED ACCEPTED MANUSCRIPT

MANUSCRIPT

ACCEPTED

ACCEPTED MANUSCRIPT

MANUSCRIPT

ACCEPTED ACCEPTED MANUSCRIPT

MANUSCRIPT

ACCEPTED

ACCEPTED MANUSCRIPT

MANUSCRIPT

ACCEPTED ACCEPTED MANUSCRIPT

MANUSCRIPT

ACCEPTED ACCEPTED MANUSCRIPT

MANUSCRIPT

ACCEPTED ACCEPTED MANUSCRIPT

MANUSCRIPT

ACCEPTED ACCEPTED MANUSCRIPT Highlights

• The ophthalmological care of infants with retinopathy of prematurity (ROP) is undertaken in the wider context of neonatal care and general wellbeing of the infant.

• This review takes a multi-disciplinary perspective on ROP, bringing together contributions from retinal vascular biologists, pediatric ophthalmologists, epidemiologists and neonatologists.

• With two distinct populations of infants with ROP in the developed and developing world, different approaches to the disease must be undertaken.

• This review details the current understanding of the cellular and molecular mechanisms of human retinal vascular formation, detailing the mechanisms underlying the pathogenesis of the two phases of acute ROP.

• We highlight the risks and benefits of anti-VEGF therapy in ROP. The potential harm from systemic exposure to anti-VEGF agents needs further study, given that VEGF has key roles in the development of many organs and also acts as a neuronal survival factor during embryonic development.

• We recommend with the exception of certain key circumstances, use of anti- VEGF agents as a therapeutic intervention in ROP still needs to be approached with caution because of; 1) the risk of late recurrence of retinopathy due to incomplete peripheral retinal vascularization;MANUSCRIPT 2) possible long term systemic side effects; and 3) the uncertainty about the correct dosage.

• We identify new therapies over the horizon, and the optimal neonatal care regimen for best ROP outcomes, and the benefits and pitfalls of telemedicine in the remote screening and diagnosis of ROP, all of which have the potential to improve ROP outcomes.

ACCEPTED