Retinopathy of prematurity and neurodevelopmental outcome and quality of life at 10 years of age

Objective In a cohort of 10-year-old children born extremely preterm, we evaluated the hypothesis that increasing severity of retinopathy of prematurity (ROP) is associated with increasing frequency of unfavorable neurodevelopmental and quality of life outcomes. Study Design Study participants were classified according to the severity of ROP. At 10 years of age, their neurocognitive abilities, academic achievement, and gross motor function were assessed, and they were evaluated for autism spectrum disorder, anxiety, depression, and quality of life. Results After adjustment for sample attrition and confounders, only the association with lower quality of life persisted. Increasing severity of visual impairment was associated with worse neurodevelopmental outcomes and lower quality of life. Conclusion Among extremely preterm children, severity of visual impairment, but not severity of ROP, was associated with adverse neurodevelopmental outcomes at 10 years of age. Both severe ROP and more severe visual impairment were associated with lower quality of life.


What's known on this subject
Retinopathy of prematurity, a highly prevalent condition among extremely preterm neonates, has been associated with suboptimal neurodevelopmental outcomes, but studies of outcomes after infancy are limited.
What this study adds: This study suggests that after adjustment for confounding factors, increasing severity of retinopathy of prematurity is not associated with worse neurodevelopmental outcomes in middle childhood but is associated with lower quality of life.

Introduction
2][3][4] Survivors are at high risk for retinopathy of prematurity (ROP), with a prevalence of 10-33%. 2,5In most cases, ROP resolves without serious visual sequelae, however, severe ROP is associated with worse cognitive and motor abilities in early childhood, [5][6][7] even after adjusting for neonatal brain injury.Further, since the brain and eye share embryonic origins, risk factors associated to ROP such as oxidative stress and in ammation might have harmful effects on the development of the central nervous system. 8,9Nonetheless, few studies of neurodevelopmental outcomes among children with a history of ROP have followed individuals into older childhood. 10,11e follow up study of Cryotherapy for Retinopathy of Prematurity (CRYO-ROP) demonstrated that very low birth weight infants who develop threshold ROP are at increased risk of unfavorable visual outcomes and functional limitations in self-care, continency, mobility, communication, and social cognition skills at ve years of age. 6At 8 years of age, CRYO-ROP study participants with unfavorable visual status had worse developmental, educational, and social skills than those who had favorable visual status. 12Furthermore, in a cohort of extremely low birth weight infants born in Victoria, Australia 1991-1992 (n = 180), study participants who previously had severe ROP exhibited more impairments at 17-18 years of age than peers without severe ROP on measures of visual processing, visual-motor integration, visual learning, IQ, and academic achievement tests in reading, spelling, and math; after controlling for perinatal risk factors, only visual acuity scores remained signi cant between the groups 10 These studies describing the relationship between severe ROP and neurodevelopmental outcomes in school-aged children did not include children born within the past two decades.Recently, a secondary analysis of data collected from the South Korean National Health Insurance Service (KNHIS) data base throughout the 10-year follow-up found a higher prevalence of developmental delay in children with severe ROP compared with those mild ROP, although this study was based on registry data unadjusted for neonatal morbidities (mechanical ventilation, necrotizing enterocolitis, bronchopulmonary dysplasia, periventricular leukomalacia, and severe intraventricular hemorrhage) which may contribute to neurodevelopmental impairments. 11 increase the understanding of associations between severe ROP and neurocognitive, neurobehavioral and quality of life outcomes at school age, we analyzed the data from the ELGAN Study of neonates born before 28 weeks of gestation and prospectively followed to 10 years of age.We hypothesized that extremely preterm infants with severe ROP would have less favorable neurocognitive, neurobehavioral, and quality of life outcomes in middle childhood as compared to ELGAN participants without severe ROP.P

Participants
The ELGAN study is a multicenter, prospective, observational study of the risk of structural and functional neurologic disorders in extremely preterm infants. 13,14A total of 1506 infants born before the 28th week of gestation were enrolled during the years 2002-2004; 1222 were discharged alive from neonatal intensive care units (NICUs) and 1198 survived to 10 years.A subset of survivors from whom data were collected on neonatal systemic in ammation (n = 966; 81% of survivors) was recruited for evaluation at 10 years of age.Of those recruited, 889 (92%) returned for follow-up.After excluding 14 children who did not have retinal examination data to classify their level of ROP, our nal sample was 875 children (73% of surviving members of the cohort).

Retinopathy of Prematurity
Retinal examinations were performed by ophthalmologists experienced in ROP screening at each NICU where participants received care.Participating ophthalmologists helped to prepare a manual and data collection form, and then participated in efforts to minimize observer variability.6][17] In keeping with these guidelines, the rst ophthalmologic examination was performed between the 31st and 33rd post-menstrual week.Follow-up exams were performed as clinically indicated until normal vascularization began in zone III.For the current analyses, severe ROP was de ned as ROP stage 3,4, 5 and was strati ed into those not requiring and those requiring surgery; less-severe ROP was de ned as ROP stage 1, 2 and not requiring surgery.ELGAN participants without ROP served as the comparison group.

Ten-Year Follow-Up Visit
Assessments were selected to provide the most comprehensive information about cognitive functions and selected academic skills.While the child was tested, the parent or caregiver completed a general health questionnaire (medical diagnoses and receipts of medications), as well as questionnaires about functional communication, social-emotional and adaptive behaviors, and quality of life.

Eye disorders at ten years
At the time of the ten-year follow-up visit, a parent or guardian reported whether or not the child had these eye disorders: requirement for corrective lenses, blindness in one eye, blindness in both eyes, strabismus not requiring surgery, or strabismus requiring surgery.Based on these data, study participants are classi ed according to the most severe vision or eye problem using the following hierarchy: none, corrective lenses, strabismus not requiring surgery, strabismus requiring surgery, blind in one eye, or blind in both eyes.

Cognitive function
To evaluate cognitive functioning at 10 years of age, we used the school-age Differential Ability Scales-II (DAS-II) 18 and the "Developmental NEuroPSYchological Assessment" (NEPSY-II). 19Participants who were not able to obtain basal score on any given test because of severe motor impairment, were assigned the lowest score for the given test.We employed latent pro le analysis to identify subgroups of extremely preterm (EP) children with similar pro les on 9 measures of verbal and nonverbal IQ (DAS-II Verbal and Nonverbal Reasoning scales), working memory (DAS-II Recall of Digits Backwards, Recall Sequential Order), concept generation and mental exibility (NEPSY-II Animal Sorting), auditory attention and set switching (NEPSY-II Auditory Attention, Response Set), and simple inhibition and inhibition shifting (NEPSY-II Inhibition and Inhibition Switching).Using this approach, four neurocognitive pro les were identi ed : 1) "normal" -mean IQ and executive function scores within the normal range (< 0.5SD) on all measures; 2) "low-normal"mean IQ and executive function scores ranged from 0.5-1 standard deviations below the mean in the normative sample; 3) "moderately impaired" -mean IQ and executive function scores between 1.5 and 2.5 standard deviations below the mean in the normative sample; and 4) "severely impaired" -mean IQ and executive function scores between 3 to 4 standard deviations below the mean in the normative sample. 20cademic achievement was evaluated using the Wechsler Individual Achievement Test-III (WIAT-III) Word Reading and Numerical Operations subtests. 21

Gross motor function
Gross motor function was assessed with Gross Motor Function Classi cation System (GMFCS), an ordinal scale that ranges from I to V, with higher levels indicative of greater functional limitations. 22ilepsy Epilepsy was identi ed using a two-stage screening process that included administration of brief screening questionnaire to the parents, followed by an open-ended interview conducted by an epileptologist. 23,24A diagnosis of epilepsy required the consensus of 2 epileptologists and was de ned as two or more seizures after the neonatal period that were not provoked by fever, trauma, or infection of the central nervous system.

Autism assessment and Social Responsiveness Scales
Parents completed the Social Communication Questionnaire (SCQ) and if the score was greater than 11, they were asked to complete the Autism Diagnostic Interview, Revised (ADIR).Children who met the ADIR criteria for autism spectrum disorder (ASD) were evaluated with the Autism Diagnostic Observation Schedule 2 (ADOS2).All children who met standardized criteria for ASD on both ADIR and ADOS2 were classi ed as having ASD. 25 We used the Social Responsive Scale (SRS), completed by the parent, to assess a constellation of social challenges in social awareness, social cognition, social communication, social motivation and autistic mannerisms (stereotypies and restricted interests), behaviors observed in children with autism (35).As previously described, SRS was modeled as a binary outcome (≤ 65 or > 65). 26Scores > 65 indicate more impaired social responsiveness.

ADHD
The parent or caregiver completed the Child Symptom Inventory (CSI-4) Parent Check List, 27 and the child's current teacher completed the teacher CSI-4 checklist.In order to parallel clinical decision making, three contexts in which ADHD symptoms could manifest were considered.Two of the contexts were taken from the CSI-4 norm-based cutoffs for the parent and teacher reports of ADHD symptoms, and the third context was based on the parent's indication at interview of the child being diagnosed previously to have ADHD by a clinician.Participants were included in the ADHD symptom group if they met any 2 or 3 criteria. 28,29

Anxiety and Depression
For identi cation of anxiety disorders and depressive disorders, parents completed the CSI-4 parent check list and teachers completed CSI-4 teachers check list. 27The forms asked for details about generalized anxiety, social phobia, depressive disorder or dysthymic disorder.For this analysis, we de ned parent-reported anxiety as an average T-score > 60 across three CSI-4 scales: general anxiety, separation anxiety, social phobia, and parent-reported depression as an average T score > 60 across two CSI-4 scales: dysthymia and major depression.Teacher-reported anxiety and depression were de ned analogously except that only general and anxiety and social phobia were included in the de nition of anxiety, because the teacher-reported CSI-4 scale does not include items on which the T-score for separation anxiety is based.

Pediatric Quality of Life Inventory
For the quality of life (QoL) measure, a parent or a caregiver completed the PedsQL 4.0 generic core scales at 10 year-follow-up. 30,31The 23-item PedsQL was designed to measure core dimensions of QoL in areas of physical (8 items), emotional (5 items), social (5 items) and school (5 items) with each item scored on a vepoint Likert scale.Each dimension was summed and linearly transformed to a 0-100 scale with higher scales indicating better QoL.In the analyses reported here we dichotomized the total PedsQL using a cutoff of < 70, which is two standard deviations below the mean in the reference sample.

Data Analysis
To analyze associations between severity of ROP and outcomes at 10 years of age we used chi square tests for univariate analysis, and logistic regression models for multivariable analyses.The strength of association was estimated as odds ratios and adjusted odds ratios (aOR) and 95% con dence intervals (CI), with individuals without ROP as the referent group.
We created directed acyclic graphs 32 to identify the minimally su cient adjustment sets of variables to include in multivariable regression models to close non-causal, biasing pathways (Figure S1).Multivariable analyses adjusted for maternal age, eligibility for government-subsidized health insurance, number of years of formal education completed by the mother at the time of the study participant's birth, marital status (married versus unmarried), duration of mechanical ventilation, gestational age, sepsis, sex, and surgical necrotizing enterocolitis.In a sensitivity analysis we adjusted for the set of variables just listed plus maternal fever during pregnancy.We expressed results as odds ratios with 95% con dence intervals.In sensitivity analyses we applied parametric multiple imputation to evaluate the potential for bias resulting from informative missingness due to sample attrition.Missing values were imputed using the adjustment covariates, and ten iterations of imputed datasets were analyzed; nal odds ratios and p-values represent the mean odds ratios and p-values of the ten iterations.Statistical signi cance was de ned as p < 0.05.

Study Participants
Children who were evaluated at 10 years of age (n = 889), as compared to members of the ELGAN cohort who survived to 10 years of age (1198) were similar in maternal variables except that the study sample had slightly lower proportions of participants born to mothers with high school education or less, Medicaid eligibility, and Black race.Infant variables were similar for those evaluated at ten years as compared to all surviving cohort members (Table S1).Of the 875 children included in the present analysis, 16% (n = 138) had severe ROP (stages 3, 4, 5) requiring surgery; 15% (n = 129) had severe ROP not requiring surgery; 44% (n = ) † white matter injury was de ned as one or more of the following ndings, with agreement by two readers, on cranial ultrasound: persistent ventricular enlargement beyond the rst month of life, echodensity in the periventricular white matter, echolucency in the periventricular white matter ¶ chronic lung disease was de ned as receipt of supplemental oxygen at 36 weeks post-menstrual age Abbreviations: NICU -neonatal intensive care unit; IQ -intelligence quotient; NICU-neonatal intensive care unit; SIP -spontaneous intestinal perforation; PMA -postmenstrual age; SNAP -Score for Neonatal Acute Physiology ROP Severity and Outcomes at 10 years of age Table 2 summarizes the frequencies of eye or vision problems and frequencies of cognitive, behavioral, neurological, and quality of life outcomes as a function of severity of ROP.In general, as compared to individuals without ROP, those with less severe ROP had similar frequencies of eye problems; those who required surgery for ROP had the highest frequencies of problems; and those with more severe ROP who did not require surgery had frequencies intermediate between those with no ROP and those who required surgery.3 presents unadjusted and adjusted odds ratios for associations between ROP severity and adverse outcomes at 10 years.As compared to no ROP, less severe forms of ROP were not associated with outcomes at 10 years.In contrast, more severe ROP requiring surgery was associated with lower scores for numeric operations, higher likelihood of parent-reported anxiety, and lower scores for quality of life.More severe ROP that did not require surgery was associated with lower scores for quality of life but not with adverse neurodevelopmental outcomes.When maternal fever was added as a covariate in the sensitivity analysis (Table S2), the results remained the same.Relationship of eye disorders/vision problems with outcomes at 10 years Table 4 highlights the associations between any vision or eye and outcomes at 10 years.Study participants are classi ed according to the most severe vision or eye problem using the following hierarchy: none, corrective lenses, strabismus not requiring surgery, strabismus requiring surgery, blind in one eye, or blind in both eyes.Overall, we found that the frequencies of adverse cognitive, behavioral, and neurological outcomes were similar in groups with current eye or vision problems as compared to groups without current eye or vision problems, unless the individual was blind in one or both eyes (Table 4).
Table 5 presents unadjusted odds ratios for associations between eye problems and adverse outcomes at 10 years.Children who required corrective lenses but did not have strabismus or blindness (25.4% of participants), were more likely to have cognitive impairment and had lower scores for quality of life.Children who had strabismus but did not require surgery and were not blind (7.8% of participants), had higher odds of autism spectrum disorder.Children who had strabismus and required surgery (6.5% of participants) for strabismus were more likely to have lower scores for numeric operations, epilepsy, and lower scores for quality of life.Lastly, children with blindness in one or both eyes (7% of participants) were more likely to have cognitive impairment, low academic achievement test scores for numeric operations and word reading, low social responsiveness, epilepsy, and low quality of life.

Discussion
In this cohort of children born before 28 weeks of gestation, we found that the severity of visual impairment, was associated with the risk of adverse neurodevelopmental outcomes and with quality of life at 10 years of age.In contrast, severity of ROP during the neonatal period was minimally predictive of cognitive, academic, or behavioral outcomes once controlling for confounding medical and social variables.On the other hand, severity of ROP was associated with overall quality of life as reported by the parent.Thus, while associations were not found between ROP severity and outcomes evaluated with objective assessments associations were found for the more subjective outcome of quality of life, an important outcome from the perspective of the individual and family. 33ong participants in the large CRYO-ROP study, poor vision after threshold ROP was associated with an increased risk for disability 6 as well as worse educational outcomes and social skills at age 8 years. 12At age 10 years, parental perspectives on health status and health-related quality of life were described in 244 participants with severe (i.e., threshold) ROP and 102 participants who did not develop ROP.Children with severe ROP were more likely to have functional limitations in vision, hearing, speech, ambulation, dexterity, emotion, cognition, and pain and reduction in health-related quality of life (HRQL).In agreement with our ndings, among children who developed threshold ROP, those with poor visual outcomes had lower HRQL scores than those with better visual outcomes. 340][41] In contrast, and consistent with our analyses, Altendahl et al. found no association between severe ROP and neurodevelopmental outcome at 12, 24 and 36 months of age, after adjusting for birth weight, male sex, intraventricular hemorrhage, and eligibility for public health insurance. 42In another recent study of extremely preterm children, Brumbaugh et al. found that those with ROP stage ≤ 3 that regressed without intervention were no more likely to have adverse neurodevelopmental at 2 years corrected age than children born extremely preterm without a history of ROP.By contrast, children who underwent an ophthalmologic intervention for ROP had worse neurodevelopmental outcomes at 2 years corrected age than did children born extremely preterm who did not undergo intervention for ROP. 43Sources of variation across studies of the relationship of developmental outcomes to ROP include differences in years and locations of births, neonatal treatments and co-morbidities, outcome assessments, and control of confounding.In addition, educational services and neuroplasticity, manifesting as improvement in neurodevelopmental functions, 44 may attenuate associations between ROP, related visual de cits, and neurodevelopmental outcomes in middle childhood.
Our study has several strengths, including its large multi-center sample, making it unlikely that we missed important associations because of lack of statistical power.Second, a broad range of outcomes, including neuromotor, neurocognitive, neurobehavioral, and quality of life, were collected prospectively by examiners who were unaware of study participant's ROP status.Third, we selected infants based on gestational age, not birth weight, to minimize confounding due to factors related growth restriction.A limitation of our study is the sample attrition between the time of NICU discharge and follow-up.An additional limitation is the lack of information about visual acuity, which precluded our ability to investigate whether de cits in visual acuity mediate the association that we observed between ROP severity and quality of life in middle childhood.
In summary, ten years after birth, extremely low gestational age neonates who had severe ROP, as compared to neonates without this condition, were more likely to have low scores on overall pediatric quality of life.It is reassuring that in general, cognitive, neuromotor, and behavioral outcomes did not differ signi cantly as a function of ROP severity.Worse neurocognitive outcomes were associated with severe visual impairment (strabismus requiring surgery or blindness in one of both eyes).These ndings could serve as the basis for cautious optimism when counselling families of extremely preterm infants who have experienced severe ROP without severe visual impairment dysfunction.

Table 1
Maternal and neonatal characteristics as a function of severity of retinopathy of prematurity.All percentages are Column totals, based on denominators that include participants with missing values.white matter injury was de ned as one or more of the following ndings, with agreement by two readers, on cranial ultrasound: persistent ventricular enlargement beyond the rst month of life, echodensity in the periventricular white matter, echolucency in the periventricular white matter ¶ chronic lung disease was de ned as receipt of supplemental oxygen at 36 weeks post-menstrual age † white matter injury was de ned as one or more of the following ndings, with agreement by two readers, on cranial ultrasound: persistent ventricular enlargement beyond the rst month of life, echodensity in the periventricular white matter, echolucency in the periventricular white matter ¶ chronic lung disease was de ned as receipt of supplemental oxygen at 36 weeks post-menstrual age Abbreviations: NICU -neonatal intensive care unit; IQ -intelligence quotient; NICU-neonatal intensive care unit; SIP -spontaneous intestinal perforation; PMA -postmenstrual age; SNAP -Score for Neonatal Acute Physiology † white matter injury was de ned as one or more of the following ndings, with agreement by two readers, on cranial ultrasound: persistent ventricular enlargement beyond the rst month of life, echodensity in the periventricular white matter, echolucency in the periventricular white matter ¶ chronic lung disease was de ned as receipt of supplemental oxygen at 36 weeks post-menstrual age Abbreviations: NICU -neonatal intensive care unit; IQ -intelligence quotient; NICU-neonatal intensive care unit; SIP -spontaneous intestinal perforation; PMA -postmenstrual age; SNAP -Score for Abbreviations: NICU -neonatal intensive care unit; IQ -intelligence quotient; NICU-neonatal intensive care unit; SIP -spontaneous intestinal perforation; PMA -postmenstrual age; SNAP -Score for

Table 2
Outcomes at 10 years of age as a function of ROP severity.

Table 3
Associations between severity of retinopathy of prematurity and outcomes at 10 years of age.Referent group is infants without ROP (n = 223).Data are unadjusted odds ratios (95% con dence intervals in parentheses).Bold indicates p < 0.05.Gross Motor Function Classi cation System † Multivariable analyses adjusted for maternal age, eligibility for government-subsidized health insurance, maternal years of formal education, status (married versus unmarried), duration of mechanical ventilation, gestational age, any early or late sepsis, and surgical NEC(Penrose drain, Exploratory Laparotomy, or both) disorder; IQ -intelligence quotient; GMFCS -Gross Motor Function Classi cation System † Multivariable analyses adjusted for maternal age, eligibility for government-subsidized health insurance, maternal years of formal education, marital status (married versus unmarried), duration of mechanical ventilation, gestational age, any early or late sepsis, and surgical NEC(Penrose drain, Exploratory Laparotomy, or both)

Table 4
Associations between any vision/eye problem and outcomes at 10 years.Study participants are classi ed according to the most severe vision/eye problem based on the following hierarchy: none/corrective lenses/strabismus not requiring surgery/strabismus requiring surgery/blind in one eye/blind in both eyes.P values are for chi square test for difference in proportions within rows.

Table 5
Odds ratios for associations between any vision/eye problem and outcomes at 10 years.Referent group is infants with no vision or eye problems (n = 474).Data are unadjusted odds ratios (95% con dence intervals in parentheses).Bold indicates p < 0.05.
Abbreviations: SRS -Social Responsiveness Scale (higher scores indicate more social challenges); ASD -autism spectrum disorder; IQ -intelligence quotient; ADHD -attention de cit hyperactivity disorder; GMFCS -Gross Motor Function Classi cation System