Proposed Section Structure
This article is organized into the following sections: clinical background and unmet need; study design and methods; key results; mechanistic and clinical interpretation; strengths and limitations; implications for neonatal practice and future research; funding and trial information; and references.
Highlights
Among very preterm infants, greater retinal nerve fiber layer thickness measured by optical coherence tomography at 36 ± 2 weeks’ postmenstrual age was associated with better 2-year motor and cognitive performance.
Thicker retinal nerve fiber layer was also associated with lower autism screening risk scores and fewer internalizing behavior problems at age 2 years.
Choroidal thickness showed a weaker signal than retinal nerve fiber layer thickness, with a significant association only for motor outcomes.
Adding retinal nerve fiber layer thickness to a model containing key infant factors improved prediction of Bayley motor outcomes substantially and cognitive outcomes modestly.
Background and Clinical Need
Very preterm birth, typically defined as birth before 32 weeks’ gestation, remains a major determinant of long-term neurodevelopmental morbidity. Even in the absence of severe intraventricular hemorrhage or cystic white matter injury, survivors are at increased risk of motor delay, cognitive impairment, behavioral symptoms, and later school difficulties. Clinicians in neonatal intensive care units therefore need practical biomarkers that identify infants at highest risk early enough to guide monitoring, parent counseling, and developmental intervention.
Optical coherence tomography, or OCT, offers an appealing possibility. This noninvasive imaging modality provides high-resolution cross-sectional views of retinal layers. Because the retina is embryologically part of the central nervous system, retinal microstructure may reflect broader neurodevelopmental processes. In adults and older children, retinal nerve fiber layer thinning has been linked to neurodegenerative and neuroinflammatory conditions. In preterm infants, bedside or handheld OCT has already expanded understanding of foveal development, macular edema, and retinopathy of prematurity. The unanswered question is whether retinal microanatomy can function not only as an ocular biomarker, but also as an early marker of brain and developmental health.
The BabySTEPS study addresses this translational gap by testing whether retinal findings at approximately 36 weeks’ postmenstrual age are associated with standardized neurodevelopmental outcomes at 2 years. If confirmed, such a marker could complement conventional risk factors such as gestational age, fetal growth restriction, and social determinants.
Study Design and Methods
Design and Setting
This was a prospective, longitudinal cohort study from BabySTEPS, the Study of Eye Imaging in Preterm Infants. Infants were enrolled at a single-center neonatal intensive care unit from August 12, 2016, through November 12, 2019, and were followed through their outpatient neurodevelopmental visit at age 2 years. Data analysis was conducted from January 5, 2023, to December 1, 2025.
Participants
The cohort included very preterm infants, defined as less than 32 weeks’ gestational age, who were at risk for retinopathy of prematurity and underwent retinal OCT imaging. The analyzed sample consisted of 72 infants with both imaging and 2-year neurodevelopmental follow-up. Of these, 33 infants, or 45.8%, were female. Mean gestational age was 27.6 weeks with a standard deviation of 2.6 weeks, and mean birth weight was 945.5 g with a standard deviation of 287.8 g.
Exposure and Imaging Measures
The exposure of interest was OCT imaging of both eyes obtained concurrently with routine examinations for retinopathy of prematurity. The principal retinal microanatomic variables examined at 36 ± 2 weeks’ postmenstrual age included retinal nerve fiber layer thickness and choroidal thickness. These structures have biologic plausibility as markers of central nervous system and vascular development, respectively. The retinal nerve fiber layer contains unmyelinated axons of retinal ganglion cells, while the choroid is a vascular layer important for outer retinal metabolism.
Outcomes
Neurodevelopmental outcomes at age 2 years were assessed using the Bayley Scales of Infant and Toddler Development, Third Edition, including motor and cognitive composite scores. Behavioral and autism-related screening outcomes were measured using the Child Behavior Checklist and the Modified Checklist for Autism in Toddlers-Revised.
Statistical Analysis
The investigators used multivariable linear regression to test associations between OCT measures and 2-year outcomes. Models were adjusted for gestational age, small for gestational age status at birth, mother’s highest education level, and postmenstrual age at imaging. These covariates were appropriate choices, capturing biological maturity, fetal growth, social context, and timing of imaging. The study also evaluated whether OCT measures improved predictive performance beyond infant factors alone, using changes in R2.
Key Findings
Retinal Nerve Fiber Layer Thickness and Motor Development
The strongest and most clinically compelling signal in the study was the association between retinal nerve fiber layer thickness and motor outcome. For every 10-µm increase in retinal nerve fiber layer thickness, the Bayley motor score increased by a mean of 7.50 points, with a 95% confidence interval of 4.38 to 10.62 and P < .001. That effect size is notable because Bayley composite scores are scaled to a mean of 100 with a standard deviation of 15. An approximately 7.5-point difference per 10 µm suggests a potentially meaningful shift in developmental trajectory, not merely a statistically detectable one.
When retinal nerve fiber layer thickness was added to a model already containing core infant factors, the explained variance for the motor score increased from R2 = 0.36 to R2 = 0.53, an absolute increase of 0.17 with a 95% confidence interval of 0.04 to 0.26. In practical terms, this means the retinal measure materially improved prediction of later motor status beyond baseline clinical information.
Retinal Nerve Fiber Layer Thickness and Cognitive Development
Retinal nerve fiber layer thickness was also associated with better cognitive performance, although the magnitude was smaller than for motor outcomes. Each 10-µm increase in thickness corresponded to a mean Bayley cognitive score increase of 3.71 points, with a 95% confidence interval of 0.73 to 6.69 and P = .02. This association remained significant after adjustment for the prespecified covariates.
The predictive improvement for cognition was more modest than for motor function. Adding retinal nerve fiber layer thickness increased model R2 from 0.29 to 0.35, an absolute increase of 0.06 with a 95% confidence interval of 0.00 to 0.16. The lower bound touching 0.00 suggests that this gain should be interpreted more cautiously than the motor result, but it remains directionally supportive of added predictive value.
Behavioral and Autism-Screening Outcomes
Greater retinal nerve fiber layer thickness was associated with lower autism screening risk on the Modified Checklist for Autism in Toddlers-Revised. The mean score change was -0.64 per 10-µm increase, with a 95% confidence interval of -1.19 to -0.09 and P = .03. The study also found an association with fewer internalizing problems on the Child Behavior Checklist, with a mean change of -2.25 per 10-µm increase, 95% confidence interval -4.41 to -0.09, P = .04.
These findings are intriguing but should be interpreted with restraint. Screening tools at 2 years are useful for risk stratification, yet they are not definitive diagnostic instruments for autism spectrum disorder or psychiatric disease. Still, the direction of effect is consistent with the broader idea that retinal microstructure may encode information about early brain development and vulnerability.
Choroidal Thickness
Choroidal thickness showed a more limited association profile. A 100-µm increase in choroidal thickness was associated with a 4.84-point higher Bayley motor score, with a 95% confidence interval of 0.45 to 9.23 and P = .03. However, choroidal thickness improved prediction less than retinal nerve fiber layer thickness and did not add further predictive value once retinal nerve fiber layer thickness was included.
This pattern suggests that while the choroid may reflect aspects of vascular or systemic maturation, retinal nerve fiber layer thickness appears to be the more robust neurodevelopmental signal in this cohort.
Clinical and Biological Interpretation
The study’s central concept is biologically plausible. The retina and brain share developmental origins, and retinal ganglion cell axons form the nerve fiber layer. In very preterm infants, disturbances in axonal development, neuronal maturation, inflammation, oxygen instability, and impaired growth could plausibly affect both retinal architecture and later neurodevelopment. A thicker retinal nerve fiber layer in this context may indicate more preserved neuronal integrity or maturational advancement.
The stronger association with motor outcomes is also interesting. Motor impairment in preterm populations often reflects diffuse developmental vulnerability involving white matter, corticospinal tracts, cerebellar development, and sensorimotor network maturation. If retinal nerve fiber layer thickness tracks global neuronal development, one might expect a prominent relationship with motor function. The cognitive association, although smaller, points in the same direction.
At the bedside, the appeal of OCT lies in its noninvasive nature and increasing feasibility in neonatal units. A retinal biomarker could help identify infants who might benefit from intensified developmental surveillance, earlier referral to physical and occupational therapy, or more structured parent support after discharge. That said, the study does not justify using OCT as a stand-alone prognostic test. It is better viewed as a candidate component of a multimodal risk model that could eventually include neuroimaging, neurologic examination, growth parameters, and social risk factors.
Strengths of the Study
This investigation has several notable strengths. First, the prospective longitudinal design reduces some of the bias that can affect retrospective studies. Second, the use of standardized developmental measures at 2 years provides clinically relevant endpoints. Third, the authors adjusted for important baseline covariates, including gestational age, small for gestational age status, maternal education, and timing of imaging. Fourth, the study moved beyond simple association testing by examining incremental predictive value, which is essential if a biomarker is to become clinically useful.
Another strength is the focus on a precisely timed neonatal window, 36 ± 2 weeks’ postmenstrual age, which could be practical for future implementation. This timing often coincides with ongoing retinopathy surveillance and therefore may fit naturally into neonatal workflows.
Limitations and Cautions
Despite its promise, the study should not be overinterpreted. The sample size was modest at 72 infants, which limits precision and raises the possibility of model instability, especially when examining multiple outcomes. The study was conducted at a single center, so external validity remains uncertain. Preterm populations vary across institutions in illness severity, socioeconomic factors, oxygen management, nutrition, and developmental follow-up practices.
Residual confounding is also possible. Although the models included several key covariates, other important factors may influence both retinal development and neurodevelopmental outcome, such as major neonatal morbidities, sepsis, bronchopulmonary dysplasia, necrotizing enterocolitis, postnatal steroid exposure, nutrition, and brain MRI findings. The abstract does not indicate whether these were explored in sensitivity analyses.
Another limitation is outcome timing. Developmental testing at 2 years is clinically meaningful but not definitive for later cognition, executive function, academic skills, or confirmed autism diagnosis. Some early deficits resolve, while others emerge more clearly at school age. Replication with longer follow-up will be essential.
Finally, OCT acquisition and segmentation in preterm infants can be technically challenging. For a biomarker to be broadly adopted, reproducibility across devices, operators, image quality conditions, and analytic pipelines must be demonstrated.
Implications for Practice
At present, these findings are best understood as hypothesis-strengthening rather than practice-changing. Clinicians should not alter care solely on the basis of retinal nerve fiber layer thickness measured in a research context. However, the study supports continued integration of neonatal OCT into translational research and raises the possibility that retinal imaging could augment developmental risk assessment in the future.
For neonatologists and developmental specialists, the message is that the eye may offer an accessible window into the developing brain. For pediatric ophthalmologists, it highlights that retinal imaging in preterm infants may have prognostic significance beyond retinopathy of prematurity. For health systems, the eventual value proposition would depend on whether OCT-based risk stratification improves referral targeting, parental counseling, and long-term developmental outcomes.
Future Research Priorities
Independent validation is the immediate next step. Larger multicenter studies should test whether retinal nerve fiber layer thickness retains predictive value across diverse neonatal intensive care settings and patient populations. Future work should also examine whether combining OCT with cranial ultrasound, term-equivalent MRI, neurologic examination, and clinical risk factors yields better prediction than any single modality.
Additional priorities include defining normative retinal trajectories in preterm infants, standardizing image acquisition and analysis, and determining clinically meaningful thresholds. It will also be important to assess whether serial OCT measures outperform a single time point, and whether retinal biomarkers relate to later executive, visual-perceptual, school, and behavioral outcomes beyond age 2 years.
Ultimately, the field should move from association to actionability. A biomarker is most valuable when it changes care in a way that improves outcomes. Interventional studies may eventually ask whether OCT-informed developmental surveillance or early therapy allocation can improve motor and cognitive trajectories.
Conclusion
This BabySTEPS cohort study provides compelling early evidence that retinal nerve fiber layer thickness at 36 weeks’ postmenstrual age is associated with 2-year neurodevelopment in very preterm infants. Thicker retinal nerve fiber layer predicted better motor and cognitive scores and was linked to lower autism screening risk and fewer internalizing behavior symptoms. Choroidal thickness showed a weaker, primarily motor-related association.
The work is clinically relevant because OCT is noninvasive and increasingly feasible in neonatal care. Its importance lies not in immediate implementation as a prognostic test, but in establishing retinal microanatomy as a credible biomarker candidate for early neurodevelopmental risk stratification. Confirmation in larger, multicenter cohorts with longer-term follow-up is now essential.
Funding and ClinicalTrials.gov
The abstract provided does not report funding details or a ClinicalTrials.gov registration number. Readers should consult the full JAMA Ophthalmology publication for complete funding disclosures, conflicts of interest, and study registration information if applicable.
References
1. 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. Published online June 4, 2026. PMID: 42240991.
2. Volpe JJ. Brain injury in premature infants: a complex amalgam of destructive and developmental disturbances. Lancet Neurology. 2009;8(1):110-124.
3. Glass HC, Costarino AT, Stayer SA, Brett CM, Cladis F, Davis PJ. Outcomes for extremely premature infants. Anesthesia and Analgesia. 2015;120(6):1337-1351.
4. Tauqeer Z, Yonekawa Y, Cabrera MT, et al. Handheld optical coherence tomography in pediatric retinal disease. Journal of AAPOS. 2017;21(5):362-365.
5. Maldonado RS, O’Connell RV, Sarin N, et al. Dynamics of human foveal development after premature birth. Ophthalmology. 2011;118(12):2315-2325.
6. Fulton AB, Hansen RM, Moskowitz A, Akula JD. The neurovascular retina in retinopathy of prematurity. Progress in Retinal and Eye Research. 2009;28(6):452-482.
