The Development of Retinal Function and Refractive Error in Children With Retinopathy of Prematurity

Abstract

Purpose: The purpose of this study was to test the hypothesis that retinopathy of prematurity (ROP) prolongs development of rod-mediated thresholds for detection of stimuli at 10 degrees but not 30 degrees eccentricity. In addition, to evaluate the thresholds at each site for an association with visual acuity (VA) and spherical equivalent (SE).

Methods: We estimated rod-mediated dark-adapted thresholds (DATs) for the detection of 2 degree diameter, 50 ms, blue (λ < 510 nm) flashes at 10 degrees and 30 degrees eccentric in former preterm subjects (n = 111), stratified by ROP severity: None (n = 32), Mild (n = 66), and Severe (n = 13). We also tested Term-born (n = 28) controls. To determine the age at half-maximal sensitivity (Age_half) for each group and eccentricity, we fit DATs to logistic growth curves. We obtained VA and SE for Preterm subjects and evaluated the course of threshold development at 10 degrees and 30 degrees for significant association with VA and SE predicted at age 10 years.

Results: DAT development at 10 degrees was significantly delayed in ROP (Mild and Severe); ROP did not significantly alter DAT development at 30 degrees. At age 10 years, among Preterm subjects, both VA and SE were significantly associated with group (None, Mild, and Severe). SE was predicted by the course of DAT development at 30 degrees. VA was not associated with the course of DAT development at 10 degrees.

Conclusions: At 10 degrees, ROP—whether mild or severe—is associated with significant delays in DAT development, evidence that the late-maturing central retina is vulnerable to ROP. The association of 30 degree threshold and myopia are evidence that more peripheral retina is important to refractive development.

In the developing primate retina, the photoreceptors in the posterior retina mature more slowly than do those in the peripheral retina.¹ The protracted course of foveal development has been much studied.^2,3 The rods surrounding the fovea and within the rod ring mature more slowly than the peripheral rods.⁴ The rod ring, defined anatomically as the region with a high number of rods per unit area, is concentric with the fovea and passes just nasal to the optic nerve head and temporally 20 degrees eccentric to the fovea. In the healthy term-born infant, the dark-adapted visual threshold (DAT) for detection of a small test spot within the rod ring at 10 degrees eccentric matures more slowly than the threshold at 30 degrees eccentric.^5,6 By age 6 months, the normal DAT at both 10-degree and 30-degree eccentricities have become equal and equivalent to those in adults.⁶ The difference in the course of normal threshold development at 10 degrees and 30 degrees is well accounted for by the difference in quantum catch by the maturing rods.^7,8 

In prematurely born infants with a history of mild retinopathy of prematurity (ROP) that resolved spontaneously without any treatment, DAT development is slower, especially at 10 degrees, than in healthy term-born infants.^5,6,9 In infants with mild ROP, DATs do not reach mature values until 12 months or more post-term, a significantly older age than in typically developing term-born infants.⁹ Herein, we characterized DAT development in a larger sample of prematurely born subjects, designated as preterm, including those with a history of ROP so severe as to have required treatment. Logistic growth curves have provided a reasonable summary of the course of development for a number of retinal parameters.^7,8,10,11 We used logistic growth curves to summarize the course of DAT development at each eccentricity, 10 degrees and 30 degrees. We also summarized visual acuity (VA) and spherical equivalent (SE) development for the preterm group and evaluated DATs as predictors of VA and SE in childhood. We evaluated the course of threshold development at 10 degrees as a predictor at age 10 years of VA, putatively mediated by the anatomically nearby fovea that is known to be impacted by ROP.^12–14 Acknowledging that the peripheral retina is a controller of eye growth and refractive development^15,16 and that high refractive errors are frequent in those with a history of ROP,^17–20 we evaluated the course of threshold development at 30 degrees as a predictor of SE at age 10 years. 

Methods

Subjects

We divided our Preterm subjects (Table 1) born at or before 32 weeks postmenstrual age (PMA), into 3 groups. Term is at 40 weeks PMA. The first group, None, included those who never developed ROP. The Mild group included those who had ROP that resolved without treatment. The Severe group included those whose ROP required treatment, which was ablation of the peripheral avascular retina either by cryotherapy (n = 1) or laser (n = 12). In all Preterm subjects, ROP severity and management were symmetric, right and left eye. We excluded any subject with an eye disease other than ROP, including strabismus, or any systemic disease or syndrome. We also studied healthy term-born (Term) subjects (n = 28). We previously reported some of these data.^5,9,10,21 

Table 1.

 

Characteristics of the 111 Preterm Subjects

 

All adult subjects and parents of infants and children provided written informed consent after explanation of the nature and possible consequences of the study; assent was obtained from children who were capable of providing it. This study conformed to the tenets of the Declaration of Helsinki and was approved by the Boston Children's Hospital Institutional Review Board. 

Dark-Adapted Thresholds

Following our previously reported methods, we estimated DATs for detection of blue (λ < 510 nm), 2-degree diameter, 50-ms duration test spots presented on a rear projection screen to the right or left of a 30-minute arc, red, center fixation light flickered at 1 hertz.^5,9,21 We used a two-alternative, spatial forced-choice, transformed up-down staircase to determine the DAT at each eccentricity, 10 degrees or 30 degrees.²² The number of trials for the staircase ranged from 19 to 51 (median 37). We obtained three to five alternations to determine the threshold. The subjects were tested using binocular viewing. We required that a subject complete the DAT test at both eccentricities within a single session. 

Preterm subjects completed 296 sessions (Table 2). Of the 111 Preterm subjects, 49 contributed longitudinal (range = 2–10, median = 4) sessions, comprising 79% of all Preterm sessions. In the longitudinally studied Preterm subjects, median PMA was 0.98 (range = 0.9–11.7) years at first session and 2.5 (range = 1.1–22.4) years at last session. The median interval between first and last session was 0.86 (range = 0.04–10.7) years. Of the Preterm subjects who contributed only one DAT session (None = 17, Mild = 39, and Severe = 6), the median was 1.4 (range = 0.36–18.5) years. 

Table 2.

 

Tests in Preterm Subjects

 

Additionally, we report 46 DAT sessions from Term subjects. Nine of the Term subjects were infants and participated in a previous longitudinal study⁹ using the same apparatus and procedure; each completed sessions at ages 10-, 18-, and 26-weeks post-term, that is 50-, 58-, and 66-weeks PMA. We tested the remaining 19 Term subjects once as adults. 

We modeled DAT development as logistic growth:  

\begin{eqnarray} T\left( {Age} \right) = \frac{{Ag{e^b}}}{{Ag{e^b} + Ag{e_{{\rm{half}}}}^b}}{\rm{\ }}\, \cdot \,{T_{{\rm{max}}}}\end{eqnarray}

(1)

 

In Equation 1, T(Age) is the threshold predicted at postmenstrual year Age, T_max is DAT in adulthood, Age_half is the postmenstrual year at which threshold reaches half of the adult value (T_max), and the exponent, b, is related to the slope of the curve. 

Initially, we fit the Term DATs to Equation 1 to obtain T_max, the adult value. We then expressed all thresholds as percent T_max. Next, we fit each group (Term, None, Mild, and Severe) at the two eccentricities, 10 degrees and 30 degrees, to Equation 1. The exponent b was shared among the eight fits, Age_half was free to vary, and T_max was fixed at 100%. In other words, variability among groups and between the two eccentricities is captured by Age_half. We tested, by all 28 pairwise comparisons²³ (Table 3), whether each developmental growth curve differed significantly from each other following Bonferroni adjusted α = 0.05/28 = 0.00179 per test. Table 3 displays the P value for each pairwise comparison. 

Table 3.

 

Pairwise Comparison of DAT Development, P Values

 

Finally, with exponent b specified, we created a nomogram applicable to each individual's data (Fig. 1). By the fit of the nomogram, we determined Age_half for each individual at each eccentricity, 10 degrees and 30 degrees. To be eligible for this analysis we required a subject to have at least one DAT between the 5th and 95th percentiles before age 5 years. Data from 63 Preterm subjects met this criterion. 

Figure 1.

 

Determination of Age_half as described in text. DATs obtained at seven sessions for a Mild subject are shown: 10 degrees (closed symbols) and 30 degrees (open symbols). Blue curves are nomograms derived from Equation 1, fit through the 10-degree (solid) and 30-degree (dashed) data. Gray shaded regions represent the upper and lower 5% intervals. Note the developmental delay at 10 degrees relative to 30 degrees.

Visual Acuity and Spherical Equivalent

At the start of each DAT session, we measured VA using age-appropriate tests (Teller Acuity Cards, Allen, HOTV, or Early Treatment Diabetic Retinopathy Study Snellen). We extracted additional VAs and results of cycloplegic refraction from the medical records of the Preterm subjects. We expressed all VAs as logMAR and refractions as SE (diopters [D]). VA had been measured between 0.3 and 34.7 years PMA and refraction between 0.2 and 34.7 years PMA (see Table 2). On the advice of an anonymous reviewer, we summarized the development of VA and SE by organizing our Preterm data into 12 bins, evenly spaced across log age; we computed the predication limits for each bin. To be included in this analysis, we required a given bin to have a minimum of 12 observations. We then used regression analysis to describe the course of development of VA and SE. 

We applied linear mixed-effects models (fitlme; MATLAB R2019b; MathWorks, Natick, MA, USA) with fixed effects for Preterm group (None, Mild, and Severe), age (log PMA), eccentricity (10 degrees and 30 degrees) and random effect for eye, stratified by subject. We optimized the model using restricted maximum likelihood (REML) estimates of the parameters, with intercept at term (40 weeks PMA). We then used the model to predict (predict; MATLAB) VA and SE at age 10 years (10.8 years PMA) for the Preterm subjects. 

In the 58 Preterm subjects who were fit to the nomogram and also had VA and SE measured, we used forward and backward stepwise linear regression (stepwiselm; MATLAB), limited by Akaike information criterion (AIC), to determine whether DAT development predicts VA, having variables Preterm group (None, Mild, and Severe), Age_half at 10 degrees, and Age_half at 30 degrees. Then, to detect significant predictors of SE, we used a second stepwise linear regression with variables Preterm group (None, Mild, and Severe), Age_half at 10 degrees, Age_half at 30 degrees, and VA. 

Results

In Figure 2A, we present DAT development at the two eccentricities (10 degrees and 30 degrees) for each group (Term, None, Mild, and Severe). In Figure 2B, we show the developmental curves from panel A along with confidence intervals of Age_half. Table 3 outlines results of the pairwise comparisons. In all groups, DAT development at 10 degrees lagged that at 30 degrees. The courses of DAT development in those without ROP (Term and None) were statistically indistinguishable. Between the groups with ROP (Mild and Severe) the courses of DAT development did not differ. DAT development in the two ROP groups was significantly delayed at 10 degrees but not 30 degrees. DAT development at 30 degrees did not differ significantly among any of the groups. 

Figure 2.

 

Development of dark-adapted threshold (DAT) stratified by group. (A) Thresholds in Term (black, circles), None (green, triangles), Mild (blue, diamonds), and Severe (red, squares) subjects fitted to Equation 1 (exponent, b = 6.38) at 10 degrees (solid lines, closed symbols) and 30 degrees (dashed lines, open symbols) eccentric. Adult Term subjects are plotted at age 18 years. (B) Curves from A replotted at increased scale (top) and Age_half with ±95% confidence intervals (bottom). Results of pairwise comparisons among Age_half are in Table 2.

We present all VAs and all SEs for the subjects in Figure 3. VA in the Preterm subjects improved with age (P = 7.05e-310). The rate of improvement in VA varied significantly with group (P = 1.52e-42)—fastest in the None group (−0.894 logMAR/log age), slowest in the Severe group (−0.416 logMAR/log age), and intermediately in the Mild group (−0.653 logMAR/log age). Concurrently, SE progressed toward myopia (P = 9.28e-104). The rate of progression varied by group (P = 3.07e-52)—fastest in the Severe group (−7.47 D/log age), slowest in the None group (−1.27 D/log age), and intermediately in the Mild group (−2.89 D/log age). Early high myopia was common in those with Severe ROP, but also frequent in the Mild group. For clarity, we show in Figure 4 the VA and SE data binned by age, along with prediction limits. Each bin contained 12 to 159 (median 48) observations. 

Figure 3.

 

Visual acuity (VA) and spherical equivalent (SE) by Preterm group. Curves plot the developmental course determined using linear mixed effects modeling for the three Preterm groups: None (green, triangles), Mild (blue, diamonds), and Severe (red, squares). Bar graphs plot the mean (±SD) predicted values at age 10 years. Gray shaded bands represent the 95% prediction intervals for normal at age 10 years.^24-27

Figure 4.

 

Development of visual acuity (top) and spherical equivalent (bottom) for the Preterm groups: None (green, triangles), Mild (blue, diamonds), and Severe (red, squares). Points are mean values for bins as detailed in the text. Solid lines are regressions through the points, logistic for visual acuity and log-linear for spherical equivalent. Dashed lines indicate the prediction limits (95, 75, 50, and 5%) as shown in the bottom right panel.

At age 10 years, subjects in the None group typically reached 20/20 (logMAR 0.00) and were hyperopic. For both VA and SE, subjects in the Severe group were typically outside the prediction limits for normal,^24–27 whereas most subjects in the None and Mild groups were within normal limits. Although VA was strongly associated with the Preterm group (None, Mild, or Severe), it was not associated with DAT development (that is, Age_half) at 10 degrees or 30 degrees. Notably, among the Preterm subjects, Age_half at 30 degrees was a predictor of SE at age 10 years. 

Discussion

In this first-ever investigation of DAT development in Severe ROP, we find the course significantly delayed, as in Mild ROP.⁹ Relative to the course in healthy controls, the delay in those with ROP, whether Mild or Severe, was at 10 degrees but not 30 degrees. This is consistent with the impact of ROP on the late maturing central retina, that includes the 10 degree site.^4,5 Severe ROP does not delay DAT development at the more eccentric (30 degree) site; this is evidence that ablation of the peripheral avascular retina does not impact sensitivity for the detection of test spots at 30 degrees. We are reminded that in children with a history of ablative treatment for ROP, in photopic conditions, their peripheral visual fields are quite intact compared to those in healthy children.^28–30 

We have explained the lag in normal DAT development at 10 degrees by low rhodopsin content and consequent decreased quantum catch by the immature rod outer segments.^5–7,10 In the ROP groups (Mild and Severe) DAT development at 10 degrees is even slower than normal. Re-organization of the post-receptor circuitry,^31,32 including the enlargement of receptive fields,³³ has been demonstrated in the ROP retina. We suspect that this is compensatory to persistent rod dysfunction,^9–11,32 possibly related to disorganized rod outer segment structure, as found in rat models of ROP.³⁴ It may take time for compensatory remodeling of the retinal circuitry to occur, resulting in a prolonged course of DAT development. 

Development of threshold at the near-central test spot, 10 degrees eccentric, was not associated with visual acuity. The course of normal foveal cone maturation is notoriously prolonged and anatomically distinct from the maturation of the rods and the rod-driven circuitry.^2,35 Therefore, it is plausible that, in the ROP retina, factors governing maturation of the fovea and rod circuitry at the 10 degree site, approximately 3 millimeters away from the fovea, follow different developmental programs. Given that the peripheral retina is a controller of eye growth,^{15,16,20,36–38} we have found, as hypothesized, that the course of rod-mediated DAT development at 30 degrees is associated with development of spherical equivalent. 

An alternative to the peripheral ablation used in our Severe ROP subjects is intravitreal injection of an antibody targeting vascular endothelial growth factor (anti-VEGF). Early high myopia is less frequent and less severe in ROP eyes treated using anti-VEGF than in eyes treated by peripheral retinal ablation.^17,19,39 Accordingly, it will be of considerable interest to study retinal development in eyes treated using anti-VEGF therapy. Will retinal development at 30 degrees in anti-VEGF treated eyes be associated with spherical equivalent? 

Acknowledgments

The authors thank Nebyou Mergia and Victoria Selian for their assistance in gathering and aggregating the medical record data used in this analysis. The authors thank their colleague, Jane Patrick, Science Writer, for her contributions to this report. 

Supported by National Institutes of Health (NIH) R01 EY010597. 

Disclosure: H. De Bruyn, None; R.M. Hansen, None; J.D. Akula, None; A.B. Fulton, None