ROP develops through a biphasic model driven by oxygen-mediated disruption of retinal angiogenic signaling [3]. In utero, the developing retina typically exists in a relatively hypoxic environment that maintains hypoxia-inducible factor (HIF) and vascular endothelial growth factor (VEGF) expression at normal levels to support organized vascularization. Following premature birth, exposure to higher oxygen tension suppresses HIF signaling and reduces VEGF expression, resulting in slowed vascular growth (phase I) [22]. Low insulin-like growth factor 1 (IGF-1) levels further disrupts physiologic vessel growth, due to limited nutritional and hepatic production [23].

As the peripheral retina matures, metabolic demand increases despite incomplete vascularization. The resulting tissue hypoxia drives further upregulation of VEGF, erythropoietin (EPO), and other pro-angiogenic mediations, leading to disorganized vessel formation (phase II). As the infant matures, IGF-1 levels rapidly rise to stimulate abnormal neovascularization in Phase II [24]. Because these angiogenic pathways are highly-oxygen sensitive, even modest differences in oxygen exposure may alter disease severity.

Landmark clinical trials have demonstrated this relationship (Table 1). The Supplemental Therapeutic Oxygen for Prethreshold ROP (STOP-ROP) study found that higher oxygen supplementation did not reduce progression to ROP and revealed increased adverse pulmonary complications [25]. Notably, a post hoc subgroup analysis found that increasing supplemental oxygen in infants without plus disease during Phase II could decrease the progression to threshold disease. Similarly, large multicenter trials such as Surfactant, Positive Airway Pressure, and Pulse Oximetry Trial (SUPPORT), BOOST, and the Canadian Oxygen Trial (COT) investigated different oxygen-saturation targets in extremely preterm infants [26–28]. Across trials, restriction of oxygen within the target range of 85–89% reduced treatment-warranted ROP, but were associated with increased morbidity and mortality [29]. In contrast, higher targets within the 91–96% oxygen-saturation range were associated with increased odds of treatment-warranted ROP [30]. The Neonatal Oxygenation Prospective Meta-analysis Collaboration (NeOProM) analysis from 5 randomized clinical trials reported that lower oxygen saturation targets reduced the likelihood of requiring treatment for ROP, but also increased NEC and mortality (with one additional death for every 2 cases of treatment-warranted ROP avoided) [31–33]. These studies demonstrate complex trade-offs between survival and ROP risk, reinforcing that oxygen exposure remains a key modifiable determinant of pathogenesis.

Observational studies spanning several decades suggest that race may influence treatment-warranted ROP risk, although the mechanisms underlying this association remain unclear. The initial CRYO-ROP study reported lower treatment-warranted ROP incidence in low-birth-weight Black infants despite similar overall incidence of ROP compared to White infants [4]. Subsequent studies including the Early Treatment for Retinopathy of Prematurity (ETROP) and the Evaluating Acute-Phase Retinopathy of Prematurity (e-ROP) demonstrated similar findings, with Black race associated with lower treatment-warranted disease compared to non-Black infants despite overall comparable ROP incidence [15, 16].

Additional cohort studies by Yang et al. and Port et al. further supported these findings by identifying Black race to be an independent protective factor for treatment-requiring ROP in multivariable analyses [17, 18]. In the Postnatal Growth and ROP (G-ROP-1 and G-ROP-2) studies, Black infants had significantly lower BW and GAs compared with White and Asian infants, yet also demonstrated lower adjusted risks of treatment-warranted ROP, any stage ROP, and plus disease [11]. Importantly, mean daily weight gain were similar during the first 30 days of life, and there were no differences in the timing of ROP onset or progression across racial groups. At days 31–40, however, postnatal daily-weight-gain was significantly lower among Black infants compared to White infants. Additionally, they found that Black infants were less likely to develop Zone 1 disease, which contradicted with a secondary analysis of the ET-ROP database that showed the opposite [34]. These findings suggest that differences in postnatal growth or early disease kinetics may not be able to fully explain the observed differences.

Although these epidemiological observations have been replicated across multiple cohorts (Table 1), they must be interpreted cautiously. A Los Angeles cohort study reported that apparent race/ethnicity differences in ROP outcomes were mainly driven by GA and neighborhood income rather than race alone [35]. Black infants were more likely to be born at lower GAs than non-Hispanic White neonates, but this association was attenuated after adjustment for household income and insurance status. Lower household income remained significantly associated with earlier gestational age, highlighting the important role of socioeconomic determinants in shaping prematurity risk, and consequently, ROP susceptibility.

Although structural factors provide important context, researchers have also explored whether biological mechanisms could contribute to racial differences in ROP severity. An early hypothesis by Saunders et al. proposed that retinal pigmentation might confer protection against oxidative injury, whereas Good et al. suggested that β-blocker receptor polymorphisms may play a role in their immunity to severe disease [4, 15]. Alternatively, Monos et al. speculated that greater amounts of melanin in the retinal pigment epithelium (RPE) or choroid would be a protective factor against developing ROP, independent of BW, GA, and length of oxygen therapy [36]. A subsequent study by Fan et al. found that light fundus pigmentation was consistently associated with higher risk features compared to medium or dark fundus pigmentation [37]. Melanin functions as a free radical scavenger to protect the retina from oxidative stress, providing a plausible mechanism for reducing ROP progression in high-risk infants [38]. In other retinal pathologies such as age-related macular degeneration, the loss of melanin in RPE was associated with increased inflammation and lipid peroxidization [39]. However, the reliability of evidence is limited, and the role of pigmentation in neonatal retinal physiology has not been clearly established.

Another potential explanation involves genetic differences in pathways regulating angiogenesis and vascular development, including VEGF signaling, HIF, IGF-1, Wnt pathways [40–42]. Variations in these genes could potentially influence susceptibility to oxygen-induced vascular injury. However, few studies have examined genetic variation in ROP risk, with one genome-wide association study identifying certain variants of the Glioma-associated oncogene family zinc finger 3 (GLI3) gene to be significantly associated with stage 3 ROP or worse [43].

These studies are limited and have not identified alleles whose distribution and effect sizes account for the epidemiologic patterns seen in Black and White infants. Hypoxia-induced reactive oxygen species and angiogenic signaling are central to ROP, yet no study has linked race-specific oxidative-stress mechanisms to differential disease risk.

Beyond individual-level factors, structural determinants of health strongly influence prematurity, which is a well-established risk factor for ROP. Neighborhood deprivation indices such as the Area Deprivation Index (ADI) incorporate measures of poverty, education, employment, housing quality, and access to transportation. Higher ADI (indicating greater neighborhood disadvantage) has been associated with missed initial ROP follow-up (up to 30% lower exam completion) [44], which may lead to under-detection of progression to treatment-requiring disease. Similarly, a study using the Child Opportunity Index (COI) found that infants from very low-opportunity neighborhoods had higher odds of treatment-warranted ROP, although the association was attenuated after adjusting for lower GA and BW (indicators of prematurity) [45].

Racial differences often correlate with disparities in socioeconomic status, access to prenatal care, maternal health, and healthcare resources [46]. Black race is strongly associated with high ADI neighborhoods and greater socioeconomic disadvantage, where preterm birth rates increase two-fold due to maternal hypertension, infection, psychosocial stress, and limited prenatal care [47–49]. Furthermore, Black infants are more frequently treated in NICUs with limited resources and healthcare worker understaffing [46, 50–52]. In addition, Black and Hispanic families tend to be more affected by housing instability, food insecurity, and other adverse disparities in mortality and childhood health [53]. On a mechanistic basis, predominantly human breast milk feeding has been associated with improved IGF-1 signaling and lower ROP risk at any stage for infants under 1800 g [12]. Disparities in access to maternal leave, lactation support, and breastfeeding resources may plausibly contribute to variation in IGF-1 mediated vascular development [54, 55]. These additional mechanisms may confound the relationship between race and ROP, including differences in oxygen exposure and survival-related selection.

Given the central role of oxygen in ROP pathogenesis, differential oxygen exposure represents a potential mediator linking race and disease risk. Large clinical trials examining oxygen targets have demonstrated that higher oxygen saturation targets were strongly associated with increased rates of treatment-warranted ROP (Type 1) [27]. Oxygen saturation levels in the NICU fluctuate due to variations in clinical practice, respiratory support equipment, and infant physiology.

Recent work has therefore begun to explore alternative explanations involving differences in clinical monitoring and care delivery. For example, racial bias in pulse oximetry has been reported, with pulse oximeters tending to overestimate arterial oxygen saturation in individuals with darker skin pigmentation by 2.4 fold compared to White infants at equivalent true saturations [46, 56]. Survival-related analysis of hospitalized adults with COVID-19 similarly demonstrate that occult hypoxemia (SaO₂<88% with concurrent SpO₂ of 92–96%) due to pulse oximeter inaccuracy is more common in Black patients and is associated with reduced access to timely oxygen therapy [57]. Among those who did meet treatment eligibility recognition, Black patients had a median delay of 1 h longer than White patients. Oxygen-targeting trials including SUPPORT, BOOST, and COT demonstrate that relatively small differences in oxygen saturation ranges can alter the incidence of ROP, with lower targets reducing treatment-warranted disease [26–28]. Pulse oximeter overestimation of oxygen saturation in infants with darker skin pigmentation could result in lower true arterial oxygen exposure despite similar SpO₂ targets. Reduced hyperoxia may thus attenuate phase I vascular suppression and decrease treatment-warranted ROP [32], providing an explanation for lower observed rates among Black infants. However, occult hypoxemia could also disproportionately increase mortality and organ dysfunction, as seen in studies that analyzed clinical outcomes in adults with “hidden hypoxemia” [32, 58]. Death prior to ROP screening may produce a healthy survivor effect, creating the appearance of lower ROP incidence despite higher underlying risk. Thus, these competing factors suggest that clinical studies should consider composite outcomes of death and treatment-warranted ROP [59].

Survival bias, or competing risk bias, may contribute further to the observed racial differences. While screening for ROP begins at 31–34 weeks PMA, treatment-warranted ROP typically develops between 36 and 37 weeks PMA [60]. Infants who die earlier, particularly those born at the lowest gestational ages, may therefore not survive long enough to enter the period of highest risk for ROP. Black infants in the U.S. experience over twice the mortality rate of white infants, experiencing nearly 4-fold as many deaths related to low GA and BW [61]. Methodologic work on racial health disparities has shown that when groups experience different early mortality, racial comparisons are made among selectively healthier survivors rather than the original population at risk (“healthy survivor” selection) [62]. As a result, the remaining cohort of surviving infants may be more clinically stable, potentially lowering the observed incidence of treatment-warranted ROP. Of note, mortality rates were similar across racial groups in the G-ROP-1 cohort, but mainly included infants with incomplete ROP data (infants died before ROP outcome was known, or other reasons such as transfer of care) [11].

A similar phenomenon has been described in BPD risk prediction models, in which increasing respiratory illness severity paradoxically lowered the estimated risk of moderate to severe BPD because the sickest infants die before reaching the diagnostic time point [63]. This illustrates how early mortality can remove the highest-risk infants from the population at risk for later morbidity, thereby distorting observed disease progression. However, most ROP studies do not include racial differences in mortality rates and represents an area for further investigation.