Highlights
- A prospective BabySTEPS cohort study published in JAMA Ophthalmology found that greater retinal nerve fiber layer (RNFL) thickness at 36 weeks’ postmenstrual age (PMA) was associated with better 2-year motor and cognitive outcomes in very preterm infants.
- RNFL thickness added predictive value beyond standard neonatal covariates, increasing explained variance for Bayley-III motor scores from 0.36 to 0.53.
- Choroidal thickness also showed association with motor outcome, but its incremental predictive value was weaker than that of RNFL thickness.
- The study strengthens the concept that the preterm retina is a noninvasive biomarker of central nervous system development, while also underscoring the need for multicenter validation and standardization of neonatal OCT metrics.
Background
Very preterm (VPT) birth, generally defined as delivery before 32 weeks’ gestation, remains strongly associated with later motor, cognitive, behavioral, and visual morbidity. Even as survival has improved, clinicians still lack robust bedside biomarkers that can identify which infants are at highest risk for later neurodevelopmental impairment while they are still in the neonatal intensive care unit (NICU). Current risk stratification relies on gestational age, birth weight, growth restriction, cranial ultrasonography, brain MRI in selected cases, and evolving clinical history. These tools are useful but imperfect, and many are either nonspecific or not easily repeated.
The retina is developmentally part of the central nervous system. Its neuronal layers, ganglion cell axons, and vascular support structures may reflect systemic and cerebral maturation, hypoxic-ischemic exposure, inflammation, nutrition, and growth. This has made retinal optical coherence tomography (OCT), especially handheld OCT adapted for premature infants, an attractive candidate biomarker platform. Unlike brain MRI, OCT can be performed serially at the bedside, without ionizing radiation and often without sedation.
The central question has been whether retinal microanatomy in the NICU carries clinically meaningful prognostic information beyond ophthalmic disease alone. The new BabySTEPS analysis by Gustafson and colleagues directly addresses that question by linking OCT measurements obtained at 36 +/- 2 weeks’ PMA to neurodevelopmental outcomes at 2 years of age. The findings are clinically notable because they move neonatal OCT from descriptive anatomy and retinopathy of prematurity (ROP) phenotyping toward developmental prognostication.
Key Content
The BabySTEPS Study: Design and Main Findings
Gustafson et al. reported a prospective longitudinal cohort analysis nested within BabySTEPS (Study of Eye Imaging in Preterm Infants), a single-center NICU cohort of VPT infants at risk for ROP. Infants underwent OCT imaging of both eyes during routine ROP examinations, and outcomes were assessed at age 2 years using the Bayley Scales of Infant and Toddler Development, Third Edition (BSID-III), the Child Behavior Checklist, and the Modified Checklist for Autism in Toddlers-Revised (M-CHAT-R).
Among 72 infants with both OCT data and neurodevelopmental follow-up, mean gestational age was 27.6 weeks and mean birth weight was 945.5 g. In multivariable models adjusted for gestational age, small-for-gestational-age status, maternal education, and PMA at imaging, greater RNFL thickness at 36 weeks’ PMA was associated with:
- Higher Bayley-III motor scores: +7.50 points per 10-um increase in RNFL thickness (95% CI, 4.38 to 10.62; P<.001)
- Higher Bayley-III cognitive scores: +3.71 points per 10-um increase (95% CI, 0.73 to 6.69; P=.02)
- Lower autism risk screening scores on M-CHAT-R: -0.64 per 10-um increase (95% CI, -1.19 to -0.09; P=.03)
- Lower Child Behavior Checklist internalizing problems scores: -2.25 per 10-um increase (95% CI, -4.41 to -0.09; P=.04)
Greater choroidal thickness was associated with higher motor scores (+4.84 points per 100-um increase; 95% CI, 0.45 to 9.23; P=.03), but the choroid contributed less predictive information than RNFL. Importantly, adding RNFL thickness to infant factors improved model performance for motor outcome substantially (R2 from 0.36 to 0.53) and modestly for cognition (R2 from 0.29 to 0.35).
These effect sizes are clinically meaningful. A Bayley score difference of several points at the individual level may influence referral intensity, early intervention planning, and parental counseling, especially when integrated with other risk markers.
Why the RNFL Might Track Neurodevelopment
The biologic plausibility of the BabySTEPS findings is strong. The RNFL consists of retinal ganglion cell axons, which form the optic nerve and project centrally to the lateral geniculate nucleus and beyond. In preterm infants, adverse exposures such as white matter injury, chronic inflammation, impaired growth, intermittent hypoxemia, sepsis, and nutritional deficits may influence both retinal neuronal maturation and cerebral development. Thus, a thinner RNFL may represent a peripheral signature of a more generalized disturbance in neuroaxonal development.
This concept aligns with broader neonatal neuroscience. Premature brain injury is no longer viewed only as focal hemorrhage or cystic periventricular leukomalacia; it is increasingly understood as a diffuse maturational disturbance affecting axons, oligodendrocytes, subplate neurons, cortical folding, and long-range connectivity. Volpe’s framework of the “encephalopathy of prematurity” remains particularly relevant here: the same systemic and developmental pressures that alter brain microstructure may also perturb retinal neurodevelopment. The BabySTEPS results fit well within that model.
Where This Study Fits in the Evolution of Preterm OCT Research
Neonatal OCT literature has evolved in roughly three stages.
Stage 1: Feasibility and descriptive imaging. Early handheld spectral-domain OCT studies established that high-resolution retinal imaging was technically feasible in awake neonates and premature infants. Investigators including Maldonado, Toth, and colleagues demonstrated bedside acquisition of retinal images capable of revealing foveal maturation, subclinical macular edema, vitreoretinal interface changes, and optic nerve features not visible on routine indirect ophthalmoscopy. This methodological advance was foundational; without it, the current biomarker work would not be possible.
Stage 2: Association with ocular disease and systemic prematurity. Subsequent studies described delayed foveal development, persistence of inner retinal layers, cystoid macular edema, and altered choroidal development in preterm infants, particularly those with severe ROP or major systemic illness. These observations suggested that retinal microanatomy was shaped by more than ocular disease alone. Investigators also began to report associations between retinal findings and factors such as gestational age, growth, oxygen exposure, and postnatal morbidity.
Stage 3: Biomarker and outcome prediction. The newest phase asks whether NICU OCT can predict later visual or neurodevelopmental outcomes. The BabySTEPS report is one of the clearest demonstrations to date that a quantitative retinal measure, especially RNFL thickness, may provide independent prognostic information for nonvisual developmental domains.
Context from Prior Literature on Preterm Retinal Development
Several themes from prior PubMed-indexed studies help contextualize the current findings.
First, preterm birth alters normal foveal and retinal development. Compared with term infants, preterm infants frequently show delayed centrifugal migration of inner retinal layers from the fovea and other markers of immature macular architecture on OCT. These structural signatures likely reflect interruption of in utero retinal maturation during the third trimester.
Second, retinal abnormalities in prematurity often extend beyond classic ROP staging. OCT has revealed subclinical cystoid macular edema, abnormal foveal contour, variable photoreceptor maturation, and choroidal thinning. Some of these abnormalities have been linked to poorer visual function or more severe neonatal illness. This is important because it means the retina may capture global developmental stress not fully represented by ophthalmoscopic ROP stage alone.
Third, optic nerve and peripapillary measurements may be especially relevant to neurodevelopment. In older children and adults, RNFL thinning is established as a correlate of neuroaxonal injury in several neurologic diseases. Translating that concept to prematurity is biologically sensible, though neonatal norms and measurement reproducibility remain active areas of work.
Interpreting the Choroid: Signal, but Less Robust than RNFL
The association between thicker choroid and better motor scores is also noteworthy. The choroid supports the outer retina through vascular supply and may be sensitive to systemic hemodynamics, oxygen regulation, growth restriction, and vascular development. Preterm birth can disrupt angiogenic signaling, especially through insulin-like growth factor 1 and vascular endothelial growth factor pathways that are also central to ROP pathobiology.
However, the BabySTEPS analysis suggests that choroidal thickness is a weaker predictor than RNFL thickness. There are several possible explanations. Choroidal measurements may be more vulnerable to technical variability, segmentation difficulty, or physiologic fluctuation. The choroid may also reflect a broader vascular or metabolic state, whereas RNFL thickness may more directly index neuroaxonal integrity. Clinically, this points toward RNFL as the more promising first-line candidate biomarker for future validation studies.
Methodological Strengths of the BabySTEPS Analysis
The study has several strengths that enhance credibility.
- Prospective longitudinal design: OCT was obtained in the NICU before outcome ascertainment at 2 years.
- Standardized developmental assessment: Bayley-III remains widely used in preterm follow-up programs, and behavioral screening instruments broadened the outcome scope.
- Adjustment for key confounders: Gestational age, growth restriction, maternal education, and PMA at imaging are relevant developmental covariates.
- Quantitative retinal layer analysis: Rather than relying on a qualitative OCT impression, the investigators tested anatomically specific measures with biologic plausibility.
- Incremental prediction analysis: Reporting change in R2 made the practical value of RNFL thickness clearer than association statistics alone.
Limitations and Cautions
Despite the excitement, this is not yet a practice-changing standalone prognostic test.
Sample size and center effects. The cohort included 72 infants with complete imaging and follow-up, which is respectable for a neonatal imaging study but still modest for predictive modeling. A single-center design also raises concern about local practice patterns, imaging protocols, and follow-up selection.
Risk of residual confounding. Although adjustment was thoughtful, other factors could influence both retinal structure and development, including severe bronchopulmonary dysplasia, necrotizing enterocolitis, sepsis, exposure to corticosteroids, severe brain injury, nutritional trajectories, and postnatal growth. Some may lie on the same causal pathway, complicating interpretation.
Outcome timing. Two-year Bayley testing is important but imperfect. Early developmental scores are clinically useful, yet they do not fully capture later executive, language, visual-perceptual, school, or behavioral outcomes. Longer-term follow-up is essential to understand whether RNFL predicts persistent disability or only early developmental variation.
Standardization challenges. Neonatal OCT acquisition, segmentation, image quality thresholds, and normative reference ranges are not yet harmonized across centers and devices. Before implementation, clinicians need reproducible protocols and validated cut points.
Generalizability beyond ROP-screened populations. The study focused on VPT infants at risk for ROP within a specialized tertiary NICU. Whether similar relationships hold for broader preterm populations or different healthcare settings remains unknown.
Clinical Implications for Neonatology, Ophthalmology, and Follow-up Programs
The immediate implication is not that OCT should replace neurologic assessment or MRI, but that it may become a useful adjunctive biomarker. A plausible future workflow would incorporate NICU handheld OCT into routine ophthalmic screening, extracting automated measurements such as RNFL thickness and integrating them with gestational age, growth, cranial imaging, and cardiorespiratory morbidity to refine developmental risk estimates.
Potential benefits include:
- Earlier referral to physical, occupational, and developmental therapies for infants with thin RNFL or other adverse OCT signatures
- Targeted counseling for families regarding follow-up intensity
- Better enrichment of high-risk cohorts for neuroprotective or developmental intervention trials
- A scalable, repeatable biomarker for multicenter neonatal research
However, implementation should be cautious. A biomarker that changes monitoring intensity must demonstrate robust external validity, calibration across populations, and added clinical utility beyond existing risk tools. It should also avoid unintended harms such as over-labeling infants based on uncertain thresholds.
Expert Commentary
The BabySTEPS report is one of the strongest pieces of evidence so far supporting the idea that the retina can function as a “window to the developing brain” in very preterm infants. Its most compelling contribution is not merely the statistical significance of the associations, but the incremental predictive value of RNFL thickness after adjustment for well-known clinical risk factors.
From a translational standpoint, the findings are appealing because they leverage an examination many preterm infants already undergo for ROP surveillance. If retinal OCT can supply both ophthalmic and neurodevelopmental information, its value proposition in the NICU changes substantially. This could encourage wider deployment of handheld OCT platforms and spur development of automated analytic pipelines.
At the same time, the field should avoid premature determinism. Neonatal development is plastic, and structural biomarkers reflect risk, not destiny. A thinner RNFL should not be interpreted as a fixed forecast of poor outcome, nor should a thicker RNFL reassure clinicians that follow-up can be reduced. Rather, these measures may help identify infants who would benefit most from close developmental surveillance and early intervention.
The study also raises mechanistic questions worth pursuing. Is RNFL thickness primarily a marker of global maturational delay, diffuse white matter dysmaturation, altered axonal growth, nutritional insufficiency, or shared vascular-neurodevelopmental injury? How do OCT markers correlate with term-equivalent brain MRI, diffusion metrics, or electroencephalographic maturation? Can serial retinal trajectories outperform a single 36-week measurement? These are the key translational next steps.
Guideline-level implications remain preliminary. Current ROP screening statements from the American Academy of Pediatrics, American Academy of Ophthalmology, and related societies focus on detection and treatment of retinopathy, not developmental prediction. Neonatal OCT is not standard of care for neurodevelopmental risk assessment. Yet studies like BabySTEPS make it increasingly reasonable for tertiary centers and research networks to incorporate OCT into multimodal preterm phenotyping protocols.
Conclusion
The 2026 BabySTEPS cohort study adds an important new dimension to neonatal retinal imaging. In very preterm infants, greater RNFL thickness at 36 weeks’ PMA was associated with better motor and cognitive outcomes at 2 years and improved predictive performance beyond standard infant factors. Choroidal thickness also showed signal, particularly for motor outcome, but appeared less informative than RNFL.
Taken together with the broader preterm OCT literature, these findings support a coherent biologic model in which retinal microanatomy reflects systemic and cerebral maturation in the NICU. The retina is not just an end-organ affected by prematurity and ROP; it may also serve as an accessible biomarker of neurodevelopmental vulnerability.
The next phase for the field is clear: multicenter replication, technical standardization, correlation with brain imaging and longer-term outcomes, and testing of whether OCT-guided risk stratification actually improves care. If those steps succeed, neonatal retinal OCT could become a practical precision-medicine tool linking ophthalmology, neonatology, and developmental follow-up.
References
- Gustafson KE, Sarin N, He J, Ying GS, Tai V, Winter KP, Cotten CM, Toth CA, BabySTEPS Group. Very Preterm Infant Retinal Microanatomy at 36 Weeks’ Postmenstrual Age and 2-Year Neurodevelopment. JAMA Ophthalmology. 2026. PMID: 42240991.
- Fierson WM; American Academy of Pediatrics Section on Ophthalmology; American Academy of Ophthalmology; American Association for Pediatric Ophthalmology and Strabismus; American Association of Certified Orthoptists. Screening Examination of Premature Infants for Retinopathy of Prematurity. Pediatrics. 2018;142(6):e20183061.
- Hellstrom A, Smith LEH, Dammann O. Retinopathy of prematurity. Lancet. 2013;382(9902):1445-1457.
- Volpe JJ. Brain injury in premature infants: a complex amalgam of destructive and developmental disturbances. Lancet Neurology. 2009;8(1):110-124.
- Maldonado RS, O’Connell RV, Sarin N, Freedman SF, Wallace DK, Cotten CM, Toth CA. Dynamics of human foveal development after premature birth. Ophthalmology. 2011;118(12):2315-2325.
- Maldonado RS, O’Connell R, Ascher SB, Sarin N, Freedman SF, Wallace DK, Toth CA. Optimizing handheld spectral-domain optical coherence tomography imaging for neonates, infants, and children. Investigative Ophthalmology & Visual Science. 2010;51(5):2678-2685.
- Vinekar A, Avadhani K, Sivakumar M, et al. Understanding clinically undetected macular changes in early retinopathy of prematurity on spectral-domain optical coherence tomography. Investigative Ophthalmology & Visual Science. 2011;52(8):5183-5188.
- Twilhaar ES, de Kieviet JF, Aarnoudse-Moens CSH, van Elburg RM, Oosterlaan J. Academic trajectories of very preterm born children at school age: a systematic review and meta-analysis. Archives of Disease in Childhood Fetal and Neonatal Edition. 2018;103(4):F322-F330.
- Twilhaar ES, Wade RM, de Kieviet JF, van Goudoever JB, van Elburg RM, Oosterlaan J. Cognitive outcomes of children born extremely or very preterm since the 1990s and associated risk factors: a meta-analysis and meta-regression. JAMA Pediatrics. 2018;172(4):361-367.
