The study included 100 preterm children without ROP (no-ROP), 50 preterm children with spontaneously resolved ROP documented with digital wide-field retinal imaging, and 30 children born at term. Of the 50 children, maximum stage 1 was detected in 21 children (zone 2:
n = 16; zone 3:
n = 5); another 21 children had stage 2 (zone 2:
n = 17; zone 3:
n = 4); and 8 children had stage 3 (zone 2:
n = 7; zone 3:
n = 1). All children from both groups were also further divided into sectors according to BW (<1000, 1000–1500, and >1500 g) and GA (<28, 28–32, and >32 weeks). Seven children (no-ROP:
n = 1; sr-ROP:
n = 2; term:
n = 4) were excluded because of poor scan image quality. The study groups' descriptive characteristics, including gestational age, birth weight, actual age, and sex have been published previously.
4 Table 1 summarizes the functional outcome in LIS and VA in the study groups at stimulus positions relevant for this study.
4 Statistically significant differences were observed for LIS and VA at the foveal center (at 0°) within all three groups, and for LIS at 2.8° between term and sr-ROP.
The mean values of foveal (0°), perifoveal (2.8°), and peripheral (8°) thicknesses of the whole retina are shown in
Figure 2A for all three groups. The mean values of the foveal minimum were highest in the sr-ROP group, intermediate in the no-ROP group, and lowest in the term-born group. These differences were statistically significant among all three groups in the foveal center (0°;
P = 0.011) and between sr-ROP and term in the perifoveal measuring point (2.8°;
P = 0.032). These data were confirmed by the ETDRS-based layer analysis, in which the C1 sector, representing the fovea, was significantly thicker in the no-ROP and sr-ROP groups compared with term. (
Supplementary Fig. S1).
The impact of the ROP stage on retinal thickness is shown in
Figure 2B. Retinal thickness was higher in all three subgroups (stage 1, 2, and 3) compared with term-born children, but no significant differences were observed among each other (
P = 0.012). In the prematurely born children, low GA and BW correlated significantly with a thicker central fovea (
Figs. 2C,
2D; GA:
P = 0.026; BW:
P = 0.037). Multiple regression analysis, including GA, BW, ROP (spontaneously regressed/without) showed that low GA was the only significant risk factor.
We observed significantly higher values of the mean thickness of the GCL+ layer for the sr-ROP group in relation to the no-ROP and term groups at 0° eccentricity (
Fig. 3A;
P = 0.014). The layer ONL+ was significantly thicker in the no-ROP and sr-ROP groups compared with term (
Fig. 3B;
P = 0.022). Again, these data were confirmed by the ETDRS grid–based layer analysis, in which C1, but no other sector, was significantly increased for both layers (Supplementary Figs.
S2,
S3;
P = 0.045). Similar to GCL+ and ONL+ thickness analysis, the INL+ thickness was significantly increased in C1 (
P = 0.048). All other sectors did not show any significantly changed layers within the groups (
Supplementary Fig. S4). The analysis of RNFL thickness demonstrated no significant differences in the ETDRS grid layer analysis (
Supplementary Fig. S5).
Measurements of cCOS, pCOS, and pROS did not reveal statistically significant changes between term and preterm children, independent of the presence or absence of ROP (cCOS:
P = 0.601; pCOS:
P = 0.979; pROS:
P = 0.573;
Fig. 4).
When correlating whole retina thickness, GCL+, ONL+, RNFL, or INL+ with LIS or VA data, no correlation was found (data not shown), indicating that a thickened retina (often along with a shallow foveal pit or its complete absence) alone is not the reason for reduced LIS and VA in preterm children.
When analyzing the individual thickness profiles of all children, a subset of preterm children differed in their distribution of individual layers above IS (
Fig. 5A). The retinal ratio analysis between different layers described in
Table 2 was initiated after the single layer analysis did not result in conclusive data, although it was particularly noticeable that several children had abnormal foveae. When looking at the ratio of the ONL+ to the entire retinal thickness at the foveal minimum (foveal center at 0°), 29.3% of the no-ROP group (29/99 children), compared with 39.6% of the sr-ROP (19/48 children) had a ratio of ONL+ to whole retinal thickness (“retina”) below 45%, which is below the limit (complete range of measurements) of what was found in term-born children (
Figs. 5A,
5B). The ratio of ONL+ to the sum of inner retinal layers (NFL + GCL + IPL + INL + OPL = IRL) and the ratio of IRL to whole retinal thickness differed significantly in the same subgroups (ANOVA
P = 0.043;
Fig. 5B,
Table 2). The impact of this morphologic anomaly, which we defined as macular developmental arrest (MDA), on LIS was significant (
Fig. 5C). Children with significantly differing ratios had significantly lower LIS values compared with term-born children or children with normal ratios, whether they belonged to the no-ROP or the sr-ROP group (
P = 0.009). No correlation was found between VA values and the calculated retinal ratios (
Supplementary Fig. S6).