In the 70 years since the first description, retinopathy of prematurity (ROP) has been the focus of intensive basic and clinical research. Over time, worldwide, there have been several phenotypes of ROP described. Here, we explore whether these are part of a single spectrum or are separate and distinct entities.

First described in 1942 by Terry, clinical1 and experimental studies2 3 provided compelling evidence that ‘retrolental fibroplasia’—as ROP was then known—was related to uncontrolled oxygen exposure, although some acknowledged that the mechanism was likely complex. This led clinicians to the restriction of supplemental oxygen concentrations to less than 40%, with predictions that this would eliminate an important cause of infant blindness. Sadly, such forecasts were not fulfilled, and ROP-induced blindness still occurred. Subsequently with increasing preterm survival, the population developing ROP had ever decreasing gestational age. Other risk factors for ROP were recognised, leading to the notion that ROP ‘cannot always be prevented’4—a view which prevailed for several decades.

The classification of retrolental fibroplasia published in 19535 did not describe features of early acute ROP, likely due to limitations from the use of the direct ophthalmoscope. The three iterations of the International Classification of Retinopathy of Prematurity (ICROP) of 1984, 2005 and 2021 (for all, see Chiang et al)6 describe ROP in far more detail and have changed clinical practice worldwide. ICROP facilitated a number of clinical studies and trials documenting which forms of ROP will regress spontaneously and which will likely require intervention, together with indications for treatment. Treatment options have evolved from peripheral retinal ablation using cryotherapy to laser therapy, and since the early 2010s, injection of antivascular endothelial factor (VEGF) agents.

The fundamental prerequisite for the development of ROP is incomplete peripheral retinal vascularisation, for which gestational age is the major determinant. Elevated blood oxygen concentrations (compared with normal intrauterine levels) lead to interruption of retinal vascular growth and retinal ischaemia, with neovascularisation at the advancing margin of vascular growth during the recovery phase. Thus, among the most immature infants in whom vascularisation has yet to advance into the periphery the disease is located posteriorly—closer to the optic disc. This contrasts with more mature infants for whom vascularisation has already progressed more peripherally. In addition to gestational age and oxygen exposure, several factors increase the risk of ROP, including fetal growth restriction, poor postnatal growth and causes of oxidative stress including neonatal sepsis and systemic inflammation.7 8

ROP affects different populations across the world. In high-income countries, the risk of sight-threatening ROP is mostly confined to posterior disease among extremely preterm (<28 weeks of gestation) or extremely low birth weight (<1000 g birth weight) infants, who rarely survived in the 1950s. In 2005, aggressive-posterior ROP (APROP) was incorporated into ICROP as an acute severe phenotype. This primarily affects extremely premature infants and is a consequence of research-driven advances in neonatal care, leading to increasing survival. It remains difficult to control blood oxygen concentrations with absolute precision, meaning that primary prevention is not always possible. The Neonatal Oxygen Prospective Meta-analysis collaboration of five international trials of two different oxygen saturation (SpO2) targets enrolled 4695 infants of <28 weeks of gestation. Compared with an SpO2 target of 91%–95%, aiming for 85%–89% was associated with increased mortality and prevalence of necrotising enterocolitis, but a lower risk of sight-threatening ROP.9 Thus, it is not a question of simply reducing exposure to supplemental oxygen, but rather one of fine tuning its administration to optimise quality survival while decreasing the risk of sight-threatening ROP—in essence the dilemma posed by Cross in 1973.10 Given this, services need to monitor the occurrence of ROP closely and provide timely therapeutic intervention; the disease might not be preventable but vision loss is largely avoidable through timely screening and treatment.

In regions where resources for medical care are more limited, survival at low gestations is less common, and sight-threatening disease frequently affects more mature infants (33–35 weeks of gestation, around 2000 g birth weight) who have received unblended/unregulated supplemental oxygen.11 Recently, a particularly aggressive phenotype has been described, mainly from India, which occurs in more mature infants given unblended oxygen.12 This novel type and APROP have been categorised as ‘aggressive retinopathy of prematurity’ (AROP).6 It is instructive to reflect that the RLF of 19535 strikingly resembles this novel phenotype.

Retinal vascular development commences in utero but can be perturbed ex utero when the baby is exposed to higher oxygen levels than those experienced during fetal life. Experimentally this induces cessation of retinovascular growth with some loss/regression of pre-existing vessels2 3; the vascular changes being related both to the absolute level and wide fluctuations in oxygen tension. As the eye develops, the mismatch between vascular supply and the increasing oxygen requirements of the developing retina, renders the retina relatively hypoxic, triggering vascular endothelial growth factor-induced neovascularisation, which is the first visually detected sign of ROP. Importantly, in animal models of oxygen-induced retinopathy, all other conditions are kept constant and only the oxygen exposure varied, a situation challenging to replicate in clinical studies. Vessel die-back to more posterior retinal zones, ‘reversing’ the direction of vascular development, is a feature of these studies.

The reports of Shah et al,12 Padhi et al 13 and others have provided the first objective evidence of vessel die- back in the human infant (figure 1), which had not previously been observed in the human infant, both because of late referral and the challenge of repeatedly and objectively measuring the peripheral extent of retinal vascularisation during the neonatal period. Thus, the retinal lesion appears to commence peripherally, but progressively retreats to posterior retinal regions which are more likely to threaten sight. We contend that these different manifestations of ROP are expressions of a similar aetiopathogenesis differing only in the oxygen management.

ROP development. Male infant born at 30 weeks gestation and 1,380 g birthweight.A shows a featureless peripheral retina at 31 weeks 3 days postmenstrual age with early retinopathy and vessels visible in anterior Zone II. B at 33 weeks and 3 days) the vessels had regressed to posterior Zone II. C at 34 weeks PMA regression with vessel die back to Zones I and II. This infant received oxygen at 100% for around 7 days. NICU has no oxygen blenders. Clinical sepsis was diagnosed and treated with antibiotics.

Where the management of supplemental oxygen is not optimal, sight-threatening ROP currently occurring in larger preterm infants should be preventable. Avoidance of unblended 100% oxygen during postdelivery stabilisation and neonatal care, with careful titration when supplemental oxygen is required, demands careful education and investment in air/oxygen blenders and monitoring equipment. There is an urgent need to educate and train all who care for the newborn in the use of supplemental oxygen, as well as other key elements of care such as the prevention of sepsis,14 and to increase awareness of this disease in the ophthalmologic community in developing economies.15 To fail in this effort increases the likelihood of blindness from a largely preventable condition and increases the burden of ROP screening in these settings where fewer resources are directed to the care of preterm infants and ROP detection and treatment. To fail sends us back to the 1950s, to before we understood the condition and its aetiology. The circle is very nearly complete with one important difference—we now have better understanding what we need to do!

Consent obtained from parent(s)/guardian(s)

Not applicable.