This is an analysis of OCT images captured between November 2016 and March 2021 under BabySTEPS (NCT02887157), a prospective observational study of retinal microanatomy in ROP. BabySTEPS was approved by the Duke University Health System Institutional Review Board and adhered to the tenets of the Declaration of Helsinki, Good Clinical Practice, and the Health Insurance Portability and Accountability Act, and each participant’s parent or guardian provided written informed consent.
As part of BabySTEPS, we imaged both eyes of each infant on the same day as standard-of-care clinical ROP examinations using investigational, bedside, swept-source OCT. Two ultra-compact handheld OCT probes were used for imaging: (1) 100-kHz OCT system centered at 1047 nm, with a handpiece that captured 512 A-scans per B-scan and 112 B-scans per 6 × 6-mm volume (UC2; November 2016–December 2018)
28; and (2) a 1060-nm, 200-kHz probe that captured 950 A-scans per B-scan and 256 B-scans per 10 × 10-mm volume (UC3; December 2018–March 2021).
29 OCT imagers measured pupil size prior to imaging in ambient light using black dots for scale and prioritized capture of the foveal center.
Custom infant-specific software, the Duke OCT Retinal Analysis Program Marking Code (DOCTRAP, MATLAB R2017b; MathWorks, Natick, MA), was used to autosegment retinal layers on foveal OCT volumes (
Fig. 1). As part of the primary analysis for BabySTEPS, an experienced OCT grader assessed the following: (1) visibility of all age-appropriate retinal layers (including retinal nerve fiber layer, ganglion cell layer, inner plexiform layer, inner nuclear layer, outer plexiform layer, outer nuclear layer, and retinal pigment epithelium); and (2) visibility of the CSJ. We decided not to analyze the visibility of the ellipsoid zone as an OCT outcome in this study, as the ellipsoid zone is absent in a significant number of preterm infants.
30 OCT imagers and graders were masked to demographic and clinical features, including fundus pigmentation. During standard-of-care ophthalmoscopic examinations, a fellowship-trained pediatric ophthalmologist graded fundus pigmentation as blond, medium, or dark using previously published fundus photographs as reference.
31 For infants who had two different fundus pigmentations recorded during the nursery stay (
n = 4), the most frequently recorded fundus pigmentation was used for analysis. Clinicians were masked to OCT findings.
We included all infants that had fundus pigmentation recorded as part of BabySTEPS and did not have ocular pathology that precluded OCT imaging (e.g., persistent tunica vasculosa lentis, vitreous hemorrhage, poor corneal clarity, cataract). For each included infant, we analyzed all OCT imaging sessions of both eyes that occurred between 30 and 42 weeks’ PMA, that captured the foveal center, and that occurred prior to any ROP treatment. We extracted the following variables from the clinical chart: birth weight, gestational age, sex, and PMA at OCT imaging. The maternal identification of race (American Indian or Alaska Native, Asian, Black or African American, Native Hawaiian or Other Pacific Islander, White, or Multiple) and maternal ethnicity (Hispanic or non-Hispanic) were collected for this National Institute of Health–funded project following the racial and ethnic standards set by the Office of Management and Budget.
32,33 We classified infants by the maternal identification of race and binarized maternal race as “White” versus “non-White” for analysis because only four infants were categorized as Asian, four as Multiple, and none as American Indian or Alaska Native or Native Hawaiian or Other Pacific Islander.
We performed all statistical analyses using R 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria). We performed Fisher’s exact test for comparison of categorical variables between the fundus pigmentation groups and performed ANOVA for comparison of continuous variables among the fundus pigmentation groups. The proportion of OCT scans with all retinal layers at each 2-week PMA window (i.e., 30–32 weeks, 32–34 weeks, 34–36 weeks, 36–38 weeks, 38–40 weeks, and 40–42 weeks) was calculated, both overall and for each pigmentation group. These proportions were also calculated with regard to CSJ visibility. We performed multivariable logistic regression to evaluate the association between fundus pigmentation (blonde, medium, or dark) and the following OCT outcomes: (1) all age-appropriate retinal layers visible (yes vs. no), and (2) CSJ visible (yes vs. no). We adjusted for birth weight, gestational age, PMA at OCT imaging, pupil size at OCT imaging, and OCT system (UC2 or UC3) used. We used generalized estimating equations to account for repeated-measures correlation and inter-eye correlation. Each feature was first evaluated in a univariable logistic regression model; those that were statistically significantly associated with OCT parameters were then included in a multivariable logistic regression model. Fundus pigmentation, the primary variable of interest, was included into each multivariable regression model, no matter whether it was associated with OCT parameters or not in the univariable logistic regression model. We performed a test of interaction between fundus pigmentation and age at OCT imaging, and when there was significant interaction we performed analysis of association between age at OCT imaging stratified by fundus pigmentation with regard to OCT outcomes. For these analyses, we again adjusted for clinical factors that were statistically significant on univariable regression analysis. A two-sided P ≤ 0.05 was considered as statistical significance for all analyses.