Article Text
Abstract
Aim: Evidence suggests that reading may be an important risk factor for myopia, but recent reports find that performance in non-verbal intelligence tests may be more important or that near-work is not associated with myopia.
Methods: Non-cycloplegic autorefraction data were available at the ages of 7 and 10 years from a birth cohort study. Children whose right eye spherical equivalent autorefraction was ⩽−1.50 D were categorised as “likely to be myopic.” The authors tested associations between school-based Standardised Assesment Tests (SATS) for reading and mathematics, maternal report of child liking reading, the Wescher Objective Reading Dimension (WORD) test results, verbal and non-verbal IQ, and the child being in the “likely to be myopic” group.
Results: 6871 children (59.7% of remaining cohort) had refractive and risk factor data at 7, of whom 1.5% were in the “likely to be myopic” group. Predictors (odds ratios, OR: 95% CI) of concurrent (at 7) risk for myopia were good performance in the SATS reading (2.60:1.61, 4.19; p<0.001), SATS maths (1.90: 1.19, 3.05; p = 0.008), the WORD (2.72:1.60, 4.64; p = 0.001) and verbal IQ tests (1.99, 1.13, 3.52; p = 0.055) after adjustment for the number of myopic parents (p = 0.014) and ethnicity (p = 0.129). However, the strongest predictor of incident myopia developing between 7 and 10 years was the parental report of whether the child liked reading: (4.05:1.27, 12.89; p = 0.031), adjusted for parental myopia (p = 0.033) and ethnicity (p = 0.008).
Conclusions: Factors associated with reading may play a part in myopia development. Further comparisons of different measures of reading-related activity or verbal ability may help clarify which of the related behavioural characteristics are causally related to myopia prevalence.
Statistics from Altmetric.com
Increases in the prevalence of myopia have caused concern and stimulated much research. Studies have demonstrated high rates of myopia in Eastern and Far Eastern countries,1 2 but there is also evidence of increases in other parts of the world such as Israel.3 While twin studies suggest that myopia is 80–90% heritable,4 5 familial aggregation studies suggest lower heritability,6 and the observed large changes in prevalence over very few generations have illustrated the importance of environmental factors.7 Reading or closework have been associated with concurrent myopia in two major cohorts,8 9 while the Sydney Myopia study recently reported that near-work was not associated with concurrent myopia.10 The SCORM study has reported that performance in non-verbal IQ tests predicted concurrent and future myopia11 12 and discussed that performance in these tests might be a surrogate for reading exposure. We aimed to compare the usefulness of reading and IQ as risk factors for the development of myopia in an ongoing UK birth cohort study.
METHODS
Avon Longitudinal Study of Parents and Children
The Avon Longitudinal Study of Parents and Children (ALSPAC) study was open to all pregnant women living in the area, which was Avon, Southwest UK, with an expected date of delivery between 1 April 2001 and 31 December 2002.13 A total of 14 541 women joined the study, which has been estimated as approximately 85% of those eligible; 13 988 children were alive at 1 year. The characteristics of the ALSPAC cohort are generally representative of the 1991 UK census data for the same geographic area (details available on the website13). Approximately 11 500 children and their families are still participating in the study. There is under-representation of very deprived families as well as of children from non-white ethnic backgrounds. The children are now aged 15–16 years. Ethical approval for the study was obtained from the ALSPAC Law and Ethics Committee and the Local Research Ethics Committees. Written consent was obtained from the mothers at recruitment.
Refractive data
Cycloplegia could not be used because the ALSPAC clinics are multidisciplinary, and the variable effects of cycloplegic drops on different children may have interfered with other investigators’ data collection (eg, tests of attention or reaction time). Non-cycloplegic autorefraction using a Canon R50 (Clement Clarke, Haag Streit UK, Harlow, UK) was used at assessments held when the children were aged 7 and then 10 years. This device estimates refractive error for each eye separately and provides a numerical printout of three consecutive readings plus the average of these. In a nested validation study, receiver operating characteristic (ROC) curves were derived to estimate the effectiveness of the non-cycloplegic autorefractor as a screening device for myopia. Full details are presented separately.14 The largest area under the curve for myopia was obtained when the target was set as –1.50 D (or worse). Using a cut-point of –1.50 D mean spherical equivalent on the autorefractor gave 0.96 specificity and 0.65 sensitivity for “true” myopia; other cut-points were either less sensitive or less specific.
Data on risk factors for myopia
Highest level of parental education
Data regarding their partner’s highest educational achievement were obtained from questionnaires to the mother during pregnancy and then were dichotomised to “secondary” (11 years or less) or “tertiary” (>11 years). Cases where no data were given were excluded.
Parental refractive error
Parental refractive error was estimated (during the pregnancy) by using the mother’s and her partner’s responses to a questionnaire item in which they were asked to describe their sight without glasses, in each eye. The options were, for each eye separately, “always very good,” “can’t see clearly at a distance,” “can’t see clearly close up” and “can’t see much at all.” The number of myopic parents was taken as the number of parents responding with “can’t see clearly at a distance” for either eye.
School attainment
We used the results of the nationally administered school-based tests of attainment (“SATS”) in reading and in mathematics. These tests measure the extent to which a child has achieved prespecified tasks in each subject. The assessments were made when the children were in year 2 at school (aged 6–7 years). The available grades are “W” (working towards an acceptable standard), “1,” “2c,” “2b,” “2a,” “3” and “4.” The target level for children at this age is at least 2. We therefore dichotomised the scores as “good” (scores 3 or 4) versus “the rest” (scores below 3). Age in months at testing was included in all analyses using the SATS results.
Exposure to reading
There was direct assessment of the child’s skill at reading made at the 7-year clinic, when members of the speech and language team administered the single word section of the Weschler Objective Reading Dimension (WORD) test.15 A score of up to 50 for single words of increasing difficulty read correctly was obtained. The scores were categorised into tertiles (three groups containing approximately equal numbers of children) for analysis. The mother was also asked questions about whether her child liked reading (can’t; does not like; quite likes; likes a lot) in a self-completion questionnaire posted to her when the child was aged 7 years.
Verbal and non-verbal intelligence quotient
The children’s intelligence quotient (IQ) was assessed at a clinic when they were aged 8 years, using the Weschler Intelligence Scales for Children (WISC-IIIUK). A shortened, alternate question version of the test was used, which took approximately 30 min, administered by graduate psychologists. This test has both verbal and non-verbal components, the scores of which can be derived separately. The scores were adjusted for the child’s age in months, according to the manual.16
Ethnicity
Data were obtained from the mother and her partner during the pregnancy, using the format asked in the 1991 UK Census. This format asked the person to categorise themself as one of: “white, black/Caribbean, Black/African, black/other, Indian, Pakistani, Bangladeshi, Chinese, other specified.” We conducted analyses investigating the potential effect of ethnicity in the cohort, by considering children as either “white versus non-white,” or “Asian vs non-Asian.” The child’s ethnicity was “non-white” if either parent was non-white.
Analysis
All singleton and twin children were included, but children from higher multiple births were excluded (n = 13). The autorefractor identified children (−1.50 D or more severe in right eye) who were “likely to be myopic.” We used these data as a model (“likely to be myopic” versus the rest) to test the strength of association between the suggested risk factors and being in the “likely to be myopic group.” Logistic regression was used to calculate unadjusted odds ratios (ORs, 95% CI) for each exposure or characteristic (reading, IQ) indicating the association between that factor and the chance of being in the “likely to be myopic” group at 7 years and to adjust these ORs for the other factors (sex, parental education, myopic parents, ethnicity). All children who were myopic at 7 (using ⩽–1.50 D) were removed for a cohort analysis of the likelihood of entering the “likely to be myopic” group between 7 and 10 years of age. Analyses were with SPSS v12.0.1 (SPSS, Chicago).
RESULTS
Participants
A total of 7834 children (68% of remaining cohort) attended the 7-year assessment, and autorefraction data for right eye were available for 7554. Failure to acquire data was usually due to staff sickness so that the vision session was not manned and rarely (<1% occasions) to machine error. Compared with the whole ALSPAC cohort, children who attended and for whom we had SATS and vision data were more likely to have achieved a good score (“3 or 4”) in their reading SATS than those who did not attend (39.2% attenders vs 19.6% non-attenders; p<0.001), and they were more likely to have myopic parents (46.1% vs 28.2%; p<0.001).
Distribution of risk factors
A total of 6871 children (59.7% of remaining cohort) had complete data for sex, mother’s partner’s education, parental myopia and ethnicity (table 1). Slightly fewer children had full data for the reading or IQ measures. Almost half of the children had at least one myopic parent, and half had a father who had completed tertiary education. Over 40% of the children were graded 3 or 4 in the reading SATS. Very few (2.6%) were “non-white,” and 61 were “Asian.” “Asian” comprises children categorised as Indian (n = 37), Pakistani (n = 6), Chinese (n = 17) and Mixed (n = 1). “Non-white” includes in addition a further 118 children categorised as black Caribbean (n = 78), black African (n = 11) and black/other (n = 23), and six had a combination of black parents.
Predictors of being in the “likely to be myopic” group at the 7-year assessment
Table 2 shows the ORs for six different measures estimating exposure to reading and/or intelligence, which are adjusted for gender, mother’s partner’s education, parental myopia and ethnicity. Children who had two myopic parents were more than twice as likely to be included in the “likely to be myopic” group than children with no myopic parents (p = 0.014). Gender and the highest level of the mother’s partner’s education were not predictive. The school-based SATS reading and maths tests and the clinic-based WORD test results were all predictive of being in the “likely to be myopic” group. The verbal IQ scores showed a similar pattern to the reading results, while the performance IQ scores were not predictive.
The parents’ assessment of whether the child liked reading suggested a positive association, but the confidence intervals were so wide that this could easily be a chance association. Ethnicity (Asian versus non-Asian) was a good predictor in unadjusted analyses, but the confidence intervals widened to include no evidence of a difference in the adjusted analyses.
Development of myopia between 7- and 10-cohort analysis
The proportion of children in the “likely to be myopic” group increased from 1.5% at the 7-year clinic to 3.6% at the 10-year clinic (n = 5527 with full data at 10). For the cohort analysis, all children who were in the “likely to be myopic group” at 7 were excluded, so that we could investigate the risks for developing myopia in the 3 years between the two clinics (table 3). All reading/IQ measures predicted the outcome. However, the mothers’ assessment of whether the child liked reading or not had the largest effect size with an adjusted OR (for “likes a lot” versus “does not like”) of 4.05 (1.27, 12.89; p = 0.031). Parental myopia remained a strong predictor in all adjusted analyses. Ethnicity (Asian versus non-Asian) was a much more robust predictor of incident myopia occurring between 7 and 10, than it had been for concurrent myopia at 7; adjusted OR 4.24 (1.45, 12.34; p = 0.008).
We repeated the cohort analysis excluding children whose parents said they “couldn’t see much at all” in either eye (n = 799) or who did not respond for either eye (n = 1850), as these parents may have had myopia. We had to dichotomise the “likes reading” data into “likes” versus “doesn’t like or quite likes” as there were no children who did not like reading who became myopic between 7 and 10. The results were then broadly similar to the main analysis (data not shown). Liking reading “a lot” (versus not liking or quite liking) was still associated with incident myopia (adjusted OR 1.75:1.09, 2.82; p = 0.022) after adjustment for number of myopic parents (adjusted OR for 2 vs 0 = 3.02: 1.72, 5.29; p<0.001) and for Asian vs non-Asian ethnicity (adjusted OR 7.22: 2.35, 22.23; p = 0.001).
Discussion
Reading, non-verbal IQ and myopia
Two reading assessments (WORD, SATS), verbal IQ and the school SATS maths tests all were associated with concurrent myopia at 7 years of age. This complements a recent report from SCORM that myopic children have higher school grades in Singapore.17 In contrast, the best indicator of incident (between 7 and 10) myopia was whether the child liked reading or not, although this had not predicted concurrent myopia status. Children who enjoy reading may do so for more prolonged periods between the ages of 7 and 10 than they do at 7, or they may read faster or with greater concentration. Alternatively, the results could be an indirect indicator of some activity that a “keen reader” is less likely to do, such as play outside or engage in sports, as these negative associations with myopia have also been described.9 10 18 The lack of association between non-verbal or performance IQ with concurrent myopia seen here, in contrast to the positive association reported recently from Singapore,11 may be due to differences in the IQ tests used, to the fact that the present analysis is likely to be less sensitive as we have used a model for “likely to be myopic” rather than actual refraction, or to population differences between the two studies. Alternatively, in Singapore some students may spend a very large proportion of their time studying and taking practice test assessments and thus may perform better at IQ tests, and the results may indicate a greater time spent studying.
Ethnicity and myopia
The data presented do not support the hypothesis that Asian children have similar risks of myopia to non-Asian children after adjusting for any of the available measures of reading or IQ. Two earlier studies which examined refraction in University students in the UK have reported either similar rates for white and British Asian students19 or a higher prevalence of myopia in students of Asian extraction as compared with white students.20 A multicentre American study designed to investigate refractive differences between ethnic groups within the US also reported that the Asian children had the highest prevalence of myopia.21 Asian children had the greatest rates of progression in one arm of the study, in The Correction of Myopia Evaluation Trial (COMET).22 However, in the US, the term “Asian” is likely to include children with a different range of ethnic origins as compared with the children in this UK study who were predominantly descended from the Indian subcontinent, so the ability to make transatlantic comparisons is limited.
Although definitive data regarding the risks for children of Asian ancestry living in the UK will need to come from studies with appropriate sampling, the present results suggest that factors other than school attainment need to be considered to evaluate their risks properly.
Methodological considerations—strengths and weaknesses
The simple model we have used (in which a group with a higher risk for myopia is compared with the rest of the sample) is likely to be less sensitive than using accurate refractive results. Modern dedicated eye studies now include biometry and sophisticated ocular imaging techniques that are not available in this study. The numbers of Asian or non-white children in the study are few, and the methods used to define ethnicity do not take account of recent advances in genetic epidemiology and taxonomy. Despite these limitations, the data reflect well-established patterns such as the risk of myopia increasing with age and with number of myopic parents.9 10 12
This is the largest study in children reporting on associations between behaviour and/or school attainment with current or incident myopia and is also the first contemporary report on myopia risk in children in a UK cohort for several decades. The data may therefore be useful for dedicated eye studies by confirming established trends in different populations and by indicating which characteristics or risk factors may be useful for further research. School-based measures of reading ability in these children at 7 were a useful predictor of present and future myopia but did not explain the excess risk for Asian children. Parental assessment of how much the child liked reading was the strongest predictor of incident myopia, but again this did not explain the excess risk for Asian children. A recent study using the UK 1958 cohort data indicates that three-quarters of individuals who were myopic by adulthood were not yet myopic at the age of 16.23 Therefore, later childhood and adolescent or early adult activities, including reading habits, are likely to be important in order to fully understand the development of myopia.
Acknowledgments
We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses. The UK Medical Research Council, the Wellcome Trust and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors, and CW will serve as guarantor for the contents of this paper.
REFERENCES
Footnotes
Funding: This research was specifically funded by the South West Regional Health Authority (Research and Development Directorate).
Competing interests: None.
Ethics approval: Ethical approval for the study was obtained from the ALSPAC Law and Ethics Committee and the Local Research Ethics Committees.
Patient consent: Written consent was obtained from the mothers at recruitment.