A Population-Based Study of the Refractive Outcome in 10-Year-Old Preterm and Full-Term Children FREE

PREMATURELY BORN children have a higher risk of ophthalmological^{1 - 4} and neurodevelopmental^{5 - 8} problems than those born at term. Severe changes are usually diagnosed early in life. Minor lesions, however, may be detected later, in preschool- or school-aged children. Long-term follow-up studies are, therefore, important. Many investigations^{9 - 14} on the ophthalmological outcome in prematurely born children have been performed, but few have been strictly population based. In Stockholm County, children with a birth weight (BW) of 1500 g or less, born between September 1, 1988, and October 31, 1990, were included in a prospective population-based study¹⁵ on the incidence of retinopathy of prematurity (ROP). The follow-up at 3½ years has been reported.^{16 - 17} Recently, we performed a 10-year ophthalmological follow-up study of various visual functions in the same cohort. A control group of children born at term was also included. The present study reports the refractive outcome in these preterm and full-term children at the age of 10 years, and evaluates the effects of prematurity per se, of ROP, and of cryotreatment on refraction. Longitudinal data with analysis of the development of refractive errors of the prematurely born children will be provided later.

In the previous prospective population-based study¹⁵ on the incidence of ROP in Stockholm County, 260 children, with a BW of 1500 g or less and a survival of 8 weeks or more, were included. Of these children, 40% had ROP and 11% underwent cryotreatment. The criterion for treatment was ROP stage 3 in at least 4 contiguous clock hours in zone II, even in the absence of "plus disease." In the present follow-up study, the prematurely born children and their caregivers were asked by letter whether they wanted to participate in a 10-year ophthalmological follow-up study. The children were located with the help of a 10-digit identification number used in Sweden. From the Swedish National Board of Health and Social Welfare, we obtained a similar number of children born at term (39-41 weeks) with normal BWs (3000-4000 g) who were randomly selected to provide a control group. The full-term children were born in exactly the same period and in the same geographical area as the prematurely born children. These children and their families were also contacted and asked to participate.

Retinoscopy during cycloplegia was performed by 2 of us (E.K.L. and G.E.H.) 45 minutes after the instillation of a mixture of 0.85% cyclopentolate hydrochloride and 1.5% phenylephrine hydrochloride. Spherical equivalents were calculated. Myopia was defined as less than 0 diopters (D), and moderate or high myopia as less than −3 D. Clinically significant myopia was defined as −1 D or less; hypermetropia was significant when greater than 3 D. Anisometropia was defined as significant when the difference between the eyes was 1 D or more, and high if 2 D or more. Astigmatism was recorded as a negative cylinder, and was defined as significant when 1 D or more and as high when 2 D or more. The axes of astigmatism (≥1 D) were divided as follows: with the rule (0°-15° and 165°-180°), against the rule (75°-105°), and oblique (16°-74° and 106°-164°). A significant refractive error of any kind was defined as hypermetropia of more than 3 D, myopia of −1 D or less, astigmatism of 1 D or more in 1 or both eyes, and/or anisometropia of 1 D or more.

Retinopathy of prematurity was divided into mild ROP (stages 1-2) and severe ROP (stages 3-5). Severe ROP was further divided into severe untreated ROP (ROP stage 3 not fulfilling the criterion for treatment) and cryotreated ROP. Right and left eyes were evaluated separately. When anisometropia and significant refractive outcome were analyzed, the premature group was divided into no, mild, untreated severe, and cryotreated ROP, according to the most severely affected eye.

The study was approved by the local ethics committee at Karolinska Institutet.

An independence test of contingency tables was used to compare nominal or ordinal data, and an unpaired t test was used for continuous data. A Wilcoxon matched-pair signed rank test was used to compare the right and left eyes regarding spherical equivalent and astigmatism. A 1-way analysis of variance was used for analyses of the spherical equivalents. Because the sample sizes were unbalanced and the variances were not homogeneous, an analysis of variance model with separate variance estimates was used (Proc Mixed in SAS statistical software).¹⁸ To evaluate the spherical equivalents, gestational age at birth, BW, stage of ROP, and cryotreatment were included in stepwise regression analyses.

Astigmatism and anisometropia were negatively skewed, and data were analyzed using the Kruskal-Wallis analysis of variance by ranks, followed by multiple comparisons between the groups based on mean ranks.¹⁹ The P values were then corrected with the Bonferroni procedure. A stepwise logistic regression analysis was also performed to determine the effects of gestational age at birth, BW, stage of ROP, and cryotreatment on astigmatism of 1 D or more (yes or no).

Of the 260 prematurely born children in the prospective population-based incidence study, ¹⁵ 12 had been excluded in the previous follow-up studies^{16 - 17} (7 had died, 4 had ophthalmological or general diseases not related to prematurity, and 1 had moved abroad). The remaining study population at that time consisted of 248 prematurely born children. Of those children, 1 had died since the previous follow-up and 1 who had moved abroad during the previous follow-up immigrated at the age of 10 years and could be located again. Moreover, 32 children dropped out during the present study (6 had emigrated, 1 had a protected identity, and 25 denied participation). The remaining study population in the 10-year ophthalmological follow-up, thus, included 216 prematurely born children. Two of these children refused eyedrops and could not undergo refraction during cycloplegia (1 in the group with no ROP and 1 in the group with severe ROP that was not treated). They were excluded from the present study on refraction. Therefore, the total dropout frequency was 13.7% (34/248). Finally, 1 child had bilateral retinal detachments and could not undergo refraction, and was excluded from this analysis. The final population for refraction consisted of 213 prematurely born children. The control group consisted of 217 children born at term.

All children were examined at 10 years ±3 months, except 1 prematurely born child, who was examined at the age of 10½ years. Data on demographic features and ROP stage are given in Table 1.

There was no statistically significant difference between the right and left eyes in analyses of spherical equivalence (P =.12) or astigmatism (P = .63). We, therefore, chose to provide detailed results of the right eyes only.

The mean spherical equivalent of the right eyes was 0.52 D (range, −20.0 to 5.62 D), and the median value of astigmatism was −0.5 D. The distributions of refractive errors and the axes of astigmatism are given in Table 2, Table 3, and Table 4. Clinically significant myopia(≤−1 D) was detected in 15 (7.0%) of 213 premature children. Regarding anisometropia, the median value was 0.25 D and the distribution is provided in Table 5. The overall refractive outcome is given in Figure 1.

The mean spherical equivalent of the right eyes was 0.64 D (range, −2.25 to 4.00 D), and the median value of astigmatism was 0 D. The distributions of refractive errors and the axes of astigmatism are illustrated in Table 2, Table 3, and Table 4.Clinically significant myopia (≤−1 D) was found in 5 (2.3%) of 217 control children. The median value of anisometropia was 0.12 D. The distribution of anisometropia is illustrated in Table 5. The overall refractive outcome is given in Figure 1.

The mean spherical equivalents in the control and premature groups were similar. In the group of prematurely born children, cryotreated eyes differed from the other subgroups and had a lower mean value (−1.27 D). A difference in the distribution was found between the 2 groups (P =.008) (Table 2). There was also a significant difference when comparing the controls with the children without ROP (ie, the prematurity per se group) (P = .02). In a multiple regression analysis of the prematurely born children, gestational age at birth, BW, stage of ROP, and cryotreatment were included as independent risk factors. Only cryotreatment was significantly associated with a reduction in spherical equivalent (P<.001).

The prevalence of hypermetropia (>3 D) was significantly higher in the premature group than in the control group (P = .03). Within the premature group, we found no difference in the prevalence of hypermetropia.

Clinically significant myopia (≤−1 D) was more common in the premature group than in the controls (P = .02) (Figure 2). In the premature group, the cryotreated children had the highest prevalence of myopia of less than 0 D, clinically significant myopia (≤−1 D), and moderate or high myopia (<−3 D) (Table 2 and Figure 2).

The control group had significantly less total astigmatism (median value) than the premature group (P<.001). The distribution of astigmatism is given in Table 3. The prevalence of astigmatism of 1 D or more was significantly higher in the premature group than in the control group (P<.001). In the premature group, the cryotreated eyes had the highest prevalence of astigmatism of 1 D or more and 2 D or more.

An oblique axis of astigmatism was more common in the premature group than in the control group (Table 4).No other obvious difference for the axis of astigmatism was found between these groups or in the preterm group.

In a stepwise logistic multiple regression analysis of astigmatism (≥1 D) in the prematurely born children, gestational age at birth, BW, stage of ROP, and cryotreatment were included as independent risk factors. Cryotreatment was the only significant risk factor for astigmatism (P<.001).

Anisometropia (median value) was significantly greater in the premature group than in the control group (P<.001). Moreover, anisometropia (≥1 D) was more common in the premature group (P<.001) (Table 5). In the premature group, the cryotreated children had the highest frequency of anisometropia (≥1 D), especially high anisometropia (≥2 D).

Significant refractive errors, defined as hypermetropia of greater than 3 D, myopia of −1 D or less, astigmatism of 1 D or more in 1 or both eyes, and/or anisometropia of 1 D or more, were more common in prematurely born children (63/213) than in children born at term (17/217) (P<.001) (Figure 1). Cryotreated children had the highest risk; those with untreated ROP or those without ROP were also prone to develop significant refractive errors.

In the present population-based study, significant refractive errors were almost 4 times more common in 10-year-old prematurely born children (29.6%) than in those born at term (7.8%). Cryotreated children had the highest prevalence(16 [64%] of 25 children), but prematurity per se was also an important risk factor for refractive errors (34 [26.2%] of 130 children). The prematurely born children had a higher prevalence of hypermetropia (>3 D) and clinically significant myopia (≤−1 D) than those born at term, and moderate or high myopia (<−3 D) was found only in the preterm group. Astigmatism and anisometropia were also more frequent and more marked in the preterm group. The cryotreated children had the highest prevalences of myopia (<0, ≤−1, and <−3 D), astigmatism, and anisometropia. Regarding hypermetropia, however, the findings in the cryotreated children were similar to those in the other preterm children. The refractive errors in the cryotreated children did not explain all the differences between the control and premature groups. We also found differences in the distribution of refractive errors between the control group and the prematurely born children without ROP (ie, prematurity per se). The latter group had a higher frequency of hypermetropia of greater than 3 D, myopia of less than −3 D, astigmatism, and anisometropia than children born at term.

Several researchers^{1 ,10 ,12 - 13 ,20 - 23} have evaluated refraction in prematurely born children of similar ages as in the present study. However, the findings are difficult to compare because of differences in epidemiological features and methods. Some studies have been hospital based, ^{1 ,21 - 23} while others have been population based.^{10 ,12 - 13 ,20} To our knowledge, only 3 population-based studies, ^{10 ,12 - 13} apart from this one, have been performed on 10-year-old children who also had been screened for ROP in the neonatal period (Table 6). The overall results of refraction in our prematurely born children seem to compare best with the findings of Fledelius.¹⁰

The present study was strictly population based regarding children born preterm and at term. The groups were not masked to the examiner, but there was no reason to expect this as a source of bias. The 2 study groups were born in the same period and in the same geographical area and were examined in exactly the same way, which ensured an accurate comparison between the groups. Such a comparison was more difficult in the other long-term population-based studies^{10 ,12 - 13} for the following reasons: one¹² had no control group, a cycloplegic refraction of the control group was not performed in one, ¹³ and one¹⁰ had a control group from the previous decade. A comparison of the refractive data of our control group with those of other studies^{13 ,20 - 21 ,24 - 28} is also difficult because of differences in epidemiological features and methods. By taking various inclusion criteria into account, we found that our results were comparable to those of other Scandinavian studies.^{24 - 26} Our group of controls, therefore, seems to be representative of the refractive status of full-term 10-year-old children in our population.

Several researchers^{2 ,29 - 32} have reported a high prevalence of myopia in prematurely born children. This was confirmed by our study, in which clinically significant myopia was more common in prematurely born children than in full-term children. When our results were reanalyzed according to the criteria of other studies, they resembled those of Fledelius, ¹⁰ who reported similar mean spherical equivalents and a similar prevalence of myopia of −0.25 D or less. Darlow et al¹² found a higher prevalence of myopia of −0.5 D or less. Their refraction, however, was not measured during cycloplegia, which could explain the finding. O'Connor et al¹³ found a higher prevalence of myopia. Their children were about 1 year older at examination than those in the present study. Considering the continuous annual change in refraction, ³³ it seems likely that the prevalences of myopia would have been similar if the examinations had been performed at the same age. However, it could not explain the great differences in myopia in children with severe ROP in the present study and in the study by O'Connor et al, although the latter study only included a few children with previous severe ROP.

Three types of myopia associated with premature birth have been reported(physiological myopia, myopia without ROP, and myopia induced by severe ROP).³⁴ The results of the present study confirm the 2 latter types. The cryotreated children had a high prevalence of myopia, in accord with other reports.^{22 - 23 ,35 - 37} Quinn et al²² found more severe myopia in their cryotreated eyes than found in our study. Moreover, they observed similar distributions of myopia in cryotreated and untreated eyes with threshold disease. In their study, cryotreatment was given at a more advanced stage of ROP (threshold disease)³⁸ than in our study. At that stage of the disease, the treatment and the disease per se may affect ocular growth to a greater extent and, therefore, explain the high prevalence of myopia in both groups. Whether the high prevalence of myopia in the cryotreated children in the present population was due to the treatment or the severe ROP per se cannot be established because all eyes fulfilling the criteria for treatment in the neonatal period were treated.

A higher prevalence of myopia in prematurely born children without ROP than in those born at term has also been reported.^{10 ,13 ,39} In the present study, prematurity per se seemed to have the greatest effect on myopia of less than −3 D, which could be due to a disturbance in ocular growth and refractive development, as previously reported.^{11 ,13 ,32 ,37 ,40 - 43}

In the present study, significant hypermetropia (>3 D) was more common in prematurely born children than in full-term children, unlike in some previous reports.^{24 ,44 - 45} In a comparison with the 3 population-based studies, O'Connor et al¹⁴ found a similar prevalence of hypermetropia of greater than 3 D. Darlow et al, ¹² however, reported a lower frequency of hypermetropia of 1 D or more than in the present study. This could be because of their photorefraction technique, which may underestimate refractive errors, and measurements of refraction without cycloplegia. Fledelius¹⁰ reported the prevalence of myopia, but did not comment on hypermetropia.

In the present study, prematurity per se seemed to be of importance for significant hypermetropia. However, like Darlow et al, ¹² we found no difference in the prevalence within the premature group and, like Ricci, ⁴⁶ we found that cryotreated children had the same prevalence of hypermetropia as children without ROP. The normal process of emmetropization is thought to be disrupted in the development of myopia in prematurely born children.^{1 ,47 - 48} It remains to be determined whether this also accounts for the higher prevalence of significant hypermetropia, as found in the present study.

Astigmatism was significantly more common in prematurely born children than in full-term controls in the present study, in accord with the study by Fledelius.¹⁰ Darlow et al, ¹² however, found a lower prevalence, which may be explained by their photorefractive technique. O'Connor et al¹³ did not study astigmatism.

The degree of astigmatism in prematurely born children increases with the severity of ROP.⁴⁹ In the present study, the stage of ROP and cryotreatment per se were identified as risk factors for astigmatism of 1 D or more in univariate analyses. However, in the following stepwise logistic regression analysis, only cryotreatment remained an independent risk factor. This is in accord with the findings by Kent et al, ³⁷ who found that children with ROP stage 3 had significantly more severe astigmatism only when the cryotreated and laser-treated eyes were included. Quinn et al²² compared cryotreated eyes with untreated eyes with threshold ROP, and found a tendency for a higher frequency of astigmatism of 1 D or more in treated than untreated eyes, at the age of 10 years, indicating that cryotreatment per se may induce astigmatism. We are unable to evaluate this question because all children fulfilling our criteria for treatment underwent cryotreatment.

The prevalence of anisometropia (≥1 and ≥2 D) in the prematurely born children in the present study was significantly higher than in the full-term children, in accordance with the findings by Fledelius.¹⁰ Anisometropia is not mentioned in the 2 other population-based studies.^{12 - 13} Gallo and Lennerstrand²⁰ found a lower prevalence in 5- to 10-year-old prematurely born children, but their study was retrospective and based on medical records.

In the group of prematurely born children in the present study, the cryotreated children had the highest frequency of anisometropia, as noted by others.^{31 ,35} However, it remains uncertain whether the high prevalence of anisometropia in the cryotreated children was due to the treatment or the more severe ROP.

Theoretically, the dropout group can be a source of bias, but the low dropout frequency in the present study (13.7% [34/248]) minimized this risk. In comparing the 3 population-based studies, Darlow et al¹² had an even smaller dropout group of 9% of all survivors. Fledelius¹⁰ reported the refractive outcome for 48%, and O'Connor et al¹³ for 53%, of their original populations. In the present study, the significance of the results would not have changed if the estimated refraction of the dropout group had been included in the analyses.

In conclusion, refractive errors were more common in prematurely born children than in those born at term. Altogether, 29.6% of prematurely born and 7.8% of full-term children had some kind of significant refractive error. The mean spherical equivalents were similar in both groups, but the distribution of refractive errors differed. The prematurely born children had a higher prevalence of hypermetropia (>3 D) and clinically significant myopia (≤−1 D). Astigmatism and anisometropia were more common and more severe in premature children. Moreover, 26.2% of prematurely born children without ROP (ie, prematurity per se) had significant refractive errors. We, therefore, recommend that guidelines for follow-ups of prematurely born children include not only cryotreated or laser-treated children or children with previous severe ROP but also those without ROP in the neonatal period.

Corresponding author and reprints: Gerd E. Holmström, MD, PhD, Department of Ophthalmology, Uppsala University Hospital, 751 85 Uppsala, Sweden (e-mail: gerd.holmstrom@ogon.uu.se).

Submitted for publication February 11, 2003; final revision received May 15, 2003; accepted June 6, 2003.

This study was supported by Stiftelsen Synfrämjandets forskningsfond; Kronprinsessans Margarethas Arbetsnämnd för synskadade; Margit Thyselius fond; Stiftelsen Sigvard och Elsas Goljes minne; Sven Jerrings fond, 1: a majblomman; Svenska sällskapet för medicinsk forskning, Sweden; and Erik Rönnbergs donation for research on illnesses during the early childhood years.

We thank Elisabeth Berg, BSc, Department of Humanities, Informatics and Social Sciences, Karolinska Institute, Stockholm, Sweden, for her help with the statistical evaluation; and the Swedish National Board of Health and Social Welfare for its valuable help.