A Cohort Study of Incident Myopia in Singaporean Children

Abstract

purpose. To determine the risk factors of incident myopia in a school-based cohort study in Singaporean children.

methods. A 3-year prospective cohort study was conducted in Singaporean school children aged 7 to 9 years in three schools at entry. Chinese children without myopia at baseline (n = 994) were included in the analysis. The main outcome was incident myopia, defined as spherical equivalent (SE) at least −0.75 D based on cycloplegic autorefraction. Other definitions of incident myopia, at least −0.5 D and at least −1.0 D, were also assessed.

results. After controlling for school, age, gender, income, reading in books per week and intelligence quotient (IQ) test scores, we found the relative risk (RR) of incident myopia defined as −0.75 D to be 1.55 (95% confidence interval [CI] 1.18–2.04) for two versus no myopic parents. The multivariate RR of myopia for IQ in the third versus first tertile was 1.50 (95% CI, 1.19–1.89). However, the RR of incident myopia was 1.01 (95% CI, 0.97–1.05) for every unit increase in books read per week. Similar results were obtained with definitions of −0.5 and −1.0 D for incident myopia.

conclusions. These data provide new prospective evidence of essential links between parental myopia, IQ scores and subsequent myopia development. However, reading in books per week was not associated with incident myopia.

Myopia affects a significant proportion of school children and is particularly common among East Asians. In countries such as Singapore and Taiwan, prevalence rates of up to 80% have been reported in individuals at the time of graduation from high school. ^{1 2} Exact and relative contributions of specific risk factors for incident myopia have not been evaluated in detail, despite well-documented rates of prevalent myopia in growing children. ²  

In previous studies, largely cross-sectional, risk factors of myopia have been investigated in school children, but there are few prospective cohort studies that have delineated the associations of major risk factors with incident myopia. Parental history of myopia has been linked to prevalence in several prior human studies, but whether parental myopia denotes a common family environment or genetic susceptibility is generally difficult to resolve. ³ In cross-sectional analyses of the Orinda Longitudinal Study of 366 eighth grade children, the risks of myopia were 1.02 times higher for each diopter-hour per week increase in near work, but not all cross-sectional studies have substantiated a myopia risk from visual activity at near. ^{4 5} Baseline cross-sectional data from the Singapore Cohort Study of the Risk Factors for Myopia (SCORM), a population-based study of school children in Singapore, showed that those who read more than two books per week had 3.05 times higher risks of higher myopia (spherical equivalent [SE] at least −3.0 D) than those who read less than this amount. ⁶ Intelligence quotient (IQ) test scores and height were also positively correlated with myopia in cross-sectional studies in Singapore. ^{7 8} However, definite temporal relationships for exposure (e.g., near work) and disease (myopia) cannot be established from cross-sectional data. For instance, a positive association between near work and myopia may occur in a cross-sectional study, because myopic children may be more likely to engage in near work rather than because near work is a risk factor for myopia onset. 

In the current cohort study, we examined and compared the risk factors for incident myopia in a 3-year follow-up study of Singaporean children aged 7 to 9 years who were not myopic at baseline. This cohort study assessed, in children without myopia, the relationship between baseline risk factors and the probability of development of myopia during the follow-up period. 

Methods

The SCORM cohort study was initiated in 1999 in two schools located in the northeastern and southeastern parts of Singapore, and in 2001 in one school located in the west. The methodology has been described previously. ^{7 8 9 10} In brief, all eligible children in grades 1 to 3 were invited to participate in the study. Eighty-one children who had serious medical or ocular conditions, such as congenital cataract were excluded from the study. Of 2186 eligible Chinese children, 1478 (67.6%) agreed to participate in the initial cross-sectional survey. ⁸ In the assessments of incident myopia, 994 Chinese children without prevalent myopia defined as −0.75 D at baseline were included in this cohort analysis. The total duration of follow-up was 3 years. Secondary analyses were conducted for 293 Malays and 97 Indians who did not have prevalent myopia at baseline. The tenets of the Declaration of Helsinki were observed, and the data collection procedures were reviewed and approved by the Singapore Eye Research Institute Ethics Committee. Informed written consent was obtained after the nature of the study was explained to the parents. 

Eye Examinations

The same procedures were performed by trained ophthalmologists, optometrists, and optometry students in the schools annually. Cycloplegia was induced in each eye by the instillation of 3 drops of 1% cyclopentolate hydrochloride (cyclogyl: Alcon-Couvreur, Puurs, Belgium) 5 minutes apart. At least 30 minutes after the last drop, five consecutive refraction and keratometry readings were obtained by using one of two calibrated autokeratorefractometers (model RK5; Canon, Inc., Ltd., Tochigiken, Japan). Contact ultrasound biometry measures were performed using one of two biometry machines (probe frequency of 10 mHz; Echoscan model US-800; Nidek Co., Ltd., Tokyo, Japan), after 1 drop of 0.5% proparacaine hydrochloride (Alcaine: Alcon-Couvreur) was administered. The average of six ultrasound values was taken if the standard deviation of the six measurements was less than 0.12 mm. If the standard deviation of the six measurements was 0.12 mm or greater, the data were not included, and the measurements were repeated until SD < 0.12 mm. 

Incident Myopia

SE was defined as sphere plus half cylinder. Three definitions of myopia were used for analysis: −0.5, −0.75, and −1.0 D. That is, incident myopia in Chinese children was defined as an SE of at least −0.5 D among children without prevalent myopia, defined as at least −0.5 D at baseline (n = 888); an SE of at least −0.75 D among children without prevalent myopia, defined as at least −0.75 D at baseline (n = 994); or an SE of at least −1.0 D among children with no myopia, defined as at least −1.0 D at baseline (n = 1057). An SE of −0.75 D was used for the primary analyses, but data are reported for the other definitions as well. 

Questionnaire Data

A parent-administered questionnaire was completed during the baseline visit (1999 and 2001). Both English and Chinese translated versions were used. Ethnicity was assessed by asking the father to classify his ethnicity in the following groups: Chinese (refers to persons of Chinese origin such as Fujian, Guangdong, and Hajia), Malay (refers to persons of Malay or Indonesian origin such as Javanese), Indian (refers to persons with ancestry originating on the Indian subcontinent), and others (comprising all persons other than the first three categories such as European, Eurasian, Middle Eastern, and Japanese). Persons of mixed heritage were classified under the ethnic group of their fathers. ¹¹ The ethnicity of the child was determined using the father’s reported ethnicity according to the definition adopted by the Singapore Population Census 2000. ¹² We asked about total combined family monthly income, as well as paternal and maternal educational level. Also included were questions about the number of books read per week (a book was classified as a novel or school text book), the number of hours per day spent reading, using the computer, watching television, and playing video games on weekdays and weekends. ¹⁰ The average weighted number of hours of each activity per day was calculated (5/7 · hours per weekday + 2/7 · hours per weekend). The use of diopter-hours in this study was adopted from Mutti et al. ⁴ as follows: (3 × hours spent reading) + (2 × hours spent on computer or playing videos games) + (1 × hours spent watching television). Other data described the number of hours spent on outdoor games and activities per week, and in night lighting before 2 years of age. Parents were deemed myopic if they reported that they were currently wearing spectacles or contact lenses to see clearly at far distances. If short-sighted, the parents reported the spherical component of the prescription for their spectacles or contact lenses in three categories: −0.5 to −2.99, −3.0 to −5.99, and −6.0 D or less. The parents were asked to recall the spherical component from the lens prescription obtained from the attending eye care professional. The Raven Standard Progressive Matrices Test, a nonverbal IQ test that does not require reading or linguistic ability was administered by psychologists and trained research assistants in the school. ¹³ Age-specific IQ tertiles were created for age 7 years (Raven scores for tertile 1 were 11–37; tertile 2, 38–44; tertile 3, 45–55), 8 years (tertile 1, 11–39; tertile 2, 40–46; tertile 3, 47–57), and 9 years (tertile 1, 10–41; tertile 2, 42–47; tertile 3, 48–56). The study personnel performing the eye examinations were masked to information from the questionnaire. 

Statistical Analyses

Because the correlations between the right and left eyes were high for refractive error (Pearson correlation coefficient = 0.96), only right eye data were analyzed. The differences in proportions were evaluated with χ² tests, and differences in means evaluated using analysis of variance for participants and nonparticipants. We examined the 3-year cumulative incidence of myopia (defined as 100 × [1 − Kaplan-Meier estimate]) according to the presence or absence of various exposures. The Cox proportional hazard models were then constructed with myopia as the outcome variable and the univariate risk factors or exposures of interest based on previous cross-sectional analysis of the cohort as covariates. Models with incident myopia defined as −0.75 D were constructed, followed by models with incident myopia defined as −0.5 and −1.0 D. A separate analysis of data from children of all ethnicities (Chinese, Malay, Indian, and others), using the definition of incident myopia of −0.75 D was also performed. For each of the risk factors, we calculated risk ratios (RR) or hazards ratios from two models: the age-, sex-, income-, and schooling-adjusted model and the multivariable-adjusted model with additional adjustment for all univariate risk factors for incident myopia using the three definitions (−0.5, −0.75, and −1.0 D). We performed stratified analyses to examine the effect of reading, IQ, and parental myopia within categories of selected variables. Effect modification was formally tested by including cross-product interaction terms, including reading X parental myopia, reading X IQ, IQ X parental myopia, age X reading, age X IQ, gender X reading, and gender X IQ in the corresponding multivariate models; interactions with P < 0.10 are presented. Statistical analyses were conducted on computer (SAS ver. 9.0; SAS, Cary, NC). 

Results

Children who were included in the cohort analysis for incident myopia defined as −0.75 D (n = 994) were similar to children who did not have a follow-up visit (n = 24) by age (P = 0.64), gender (P = 0.37), history of parental myopia (P = 0.26), IQ score (P = 0.25), parental income (P = 0.47), books read per week (P = 0.15), and baseline refraction (P = 0.24). 

In multivariate analysis, the RR for incident myopia defined as −0.75 D was 1.42 for children aged 7 years compared with 9 years (P = 0.02), and the risk of myopia was 1.21 times greater in females than in males (P = 0.04; Table 1 ). Children with two myopic parents had a 1.55 times higher risk of myopia (P = 0.002) and children with one myopic parent had a 1.56 times higher risk of myopia (P = 0.002) compared with children with no myopic parents, after adjustment for risk factors. In a similar multivariate analysis, the RR of myopia for any parent with myopia compared with no parent with myopia was 1.57 (95% CI, 1.25–1.96). After adjustment for risk factors, children with IQ scores in tertiles 3 and 2 had 1.50 (P < 0.001) and 1.37 times (P = 0.006) higher risks of incident myopia, respectively, compared with children with IQ in tertile 1. The RR of incident myopia was 1.03 (95% CI, 1.01–1.04) for every unit increase in Raven IQ score (P = 0.001). There was a link between reading and IQ scores. Children who read more than two books per week were more likely to have IQ scores in the highest tertile (36.6%), compared with children who read two or fewer books per week (26.5%). There was no evidence of associations between reading in books per week and total family income with incident myopia. There were no interactions between parental myopia and child IQ; reading and IQ; parental myopia and reading; age and IQ; or age and parental myopia, gender, and IQ; or gender and parental myopia, with incident myopia. 

Other factors, such as outdoor activity (RR = 1.01; 95% CI, 0.98–1.04), number of hours of reading per day (RR = 0.99; 95% CI, 0.92–1.07), computer use (RR = 0.94; 95% CI, 0.77–1.15), number of hours playing video games per day (RR = 0.94; 95% CI, 0.80–1.09), number of hours watching television per day (RR = 0.96; 95% CI, 0.87–1.06); diopter-hours (RR = 0.99; 95% CI, 0.97–1.01), total number of near work hours per day (RR = 0.97; 95% CI, = 0.93–1.01), night lighting before 2 years, and height (RR = 1.16; 95% CI, 0.96–1.40) were not associated with incident myopia in multivariate analyses. 

Multivariate analyses were also conducted for incident myopia defined as −0.5 D and incident myopia defined as −1.0 D (Table 2) . In multivariate analysis using a definition of −0.5 D, younger age (P = 0.002), but not female gender (P = 0.21) was associated with incident myopia defined as −0.5 D. Children with two myopic parents had a 1.39 times higher risk of myopia and children with one myopic parent had a 1.28 times higher risk of myopia than did children with no myopic parents, after adjustment for risk factors. Children with IQ in the highest tertile had a 1.45 times higher risk of incident myopia than did children with IQ in the lowest tertile. Parental myopia, IQ score in tertiles, age, and gender predicted incident myopia defined as −1.0 D in the same fashion as similar multivariate analyses for incident myopia defined as −0.75 D. 

No significant associations were detected with similar multivariate analyses of incident myopia conducted in Malays (n = 293) and Indians (n = 97), because of small sample sizes (Table 3) . Only the association of incident myopia for children with two myopic parents versus no myopic parents in Malay children was of borderline significance: RR = 2.29 (95% CI, 1.05–4.97; P = 0.04). The analysis of the influence of ethnicity on the development of myopia in all children in the cohort was also evaluated. After adjustment for school, age, gender, income, IQ scores, reading in books per week, and number of parents with myopia, the RR of myopia for Malays versus Chinese was 0.63 (95% CI, 0.48–0.83; P < 0.001), Indians versus Chinese was 0.96 (95% CI, 0.68–1.37; P = 0.84), and other ethnicities versus Chinese was 0.68 (95% CI, 0.30–1.54; P = 0.36). 

Discussion

This is one of the first prospective cohort investigations of the major possible risk factors for myopia. The risks of development of myopia in Chinese children were evaluated and found to be higher in children aged 7 to 9 years without myopia at baseline who were younger, female, had higher IQ scores, and had parents with myopia. In Chinese children, the multivariate analyses for incident myopia were robust across the different definitions of incident myopia of −0.5, −0.75, and −1.0 D. The consistency of the results from the three approaches using different definitions of myopia (−0.5, −0.75, and −1.0 D) supports possible associations between these factors and the development of myopia. In addition, the presentation of data using these three common definitions facilitates comparability with other studies. Secondary analyses were performed in Malays and Indians, but these findings are not conclusive because of the small number of Malays and Indians in our cohort. 

The risks of myopia were higher in Chinese children who are aged 7 years than in those aged 9 years at baseline. In this cohort, the prevalence rates of myopia in all 1478 Chinese children at baseline were 24.7% in 7-year-olds, 31.3% in 8-year-olds, and 49.7% in 9-year-olds. Thus, there may be an age effect. Twice as many of the 9-year-olds compared with 7-year-olds are already myopic. There are fewer older children at risk of development of myopia, since half of the 9-year-olds, but only one fourth of the 7-year-olds, were already myopic at baseline. 

Although prior cross-sectional studies have found that higher IQ scores and parental myopia are associated with myopia, the relative strengths of each association and temporal nature of the relationships have not been well established. ^{6 7} Our cohort study delineates in a single study population, the potential risk factors for incident myopia in a clear, temporal fashion. In multivariate analyses, there were positive relationships between IQ and parental myopia at baseline with the subsequent development of myopia in Chinese children within the next 3 years. Recall bias is less likely to occur in the present study compared with previous cross-sectional studies, because only children without myopia are enrolled in this cohort and the parents’ risk factor questionnaire data were collected at baseline before the children converted to myopia. 

Parental myopia was a significant risk factor for myopia in our population and could denote a hereditary origin, similar parent–child environment, or a combination of both. This relationship remained significant in Chinese children, regardless of whether incident myopia was defined as −0.5, −0.75, or −1.0 D. Other factors that may contribute to increased risks of myopia such as IQ were accounted for, and we found an independent association of parental myopia and the development of myopia, at least in linear models. Parental myopia is a complex measure, because the extent of similarities between the parent and child’s lifestyles may differ between families. Other available evidence of parental history of myopia as a risk factor has been studied primarily in cross-sectional association studies. There are no reports of prospective cohort studies of parental myopia and the development of new cases of myopia in Asian children. Children with a parental history of myopia had longer eyes and more myopic refractive error in a cross-sectional sample of the Orinda Longitudinal Study of Myopia of 716 children aged 6 to 14 years. ⁴ However, a report of 514 younger preschool Hong Kong children aged 2 to 6 years did not show any association between premyopic eye size and parental history of myopia. ¹⁴ Perhaps this association only becomes evident in older children who have already been exposed to early environmental factors. A cross-sectional study of Taiwanese adults aged 17 to 45 years found that the odds ratios of mild, moderate, and high myopia were 2.5, 3.7, and more than 5.5, respectively, if there was one or more highly myopic parent. ¹⁵ In an investigation of family history and adult-onset myopia in 66 myopic and 116 nonmyopic adults aged 25 to 35 years, myopia was associated with family history of myopia, even after adjustment for daily hours of reading (P < 0.005). ¹⁶ Thus, the link between family history and myopia has been observed across different age and ethnic groups. In our cohort, we demonstrated this association in a prospective fashion. The advantage of this analysis is that the effect of parental myopia, including any common environmental factor, is determined at the time the child was not myopic and before the onset of any myopia. 

Reading more books—an aggregate measure of the amount of near work—may promote aberrant eye growth and myopic refractive error. ^{17 18} The Orinda Longitudinal Study, found that the multivariate odds ratio for myopia was 1.02 (95% CI, 1.008–1.032) for every diopter-hour per week of near work. ⁴ On the contrary, in our cohort of young children aged 7 to 9 years who are not myopic at inception, reading in books per week did not predict the development of myopia in the next few years. We note that other measures of near work in our study, such as number of hours spent reading per day, using the computer, and diopter-hours are not associated with incident risks of myopia. These results are contradictory to a previous cross-sectional analysis of the SCORM data in which we found that children aged 7 to 9 years who read more than two books per week had a 3.05 times higher risk of moderate myopia (at least −3.0 D of myopia). ⁶ Perhaps children with moderate myopia read more after the development of myopia. The latest findings from our cohort study suggest that reading in books per week may not be causally linked to incident myopia. 

The lack of association for incident myopia in our longitudinal study may also occur because of small or undetectable influences on the development of myopia, and larger sample sizes may be needed to discern weak positive associations. The relationship between reading and myopia has not been well established, because this hypothesis has been addressed primarily in cross-sectional studies. ^{4 6} Further cohort analyses with objective measures of near work, such as 24-hour distance-sensing portable head-set devices and long follow-up of several years, are needed to determine the nature of the association between near work and incident myopia. 

Adjustment for better performance on IQ tests increased the risk of incident myopia defined in the three different ways (−0.5, −0.75, and −1.0 D), and this relationship remained significant even after controlling for the amount of reading in books per week. This finding had been observed in our cross-sectional baseline data and now is confirmed prospectively. ⁷ IQ performance is consistently associated with myopia in both cross-sectional and cohort studies, even after adjustment for other known factors. To our knowledge, this is the first longitudinal investigation of the association of IQ test scores and the subsequent development of myopia. In 157,748 Israeli males aged 17 to 19 years, there was a positive association between IQ and myopia, when educational level was controlled. ¹⁹ A similar study of 5942 Danish 18-year-old men showed that the prevalence of myopia rose from 8% to 31% for those with low to high IQ scores. ²⁰  

The mechanistic connection between IQ and myopia remains uncertain and will require further research to establish. Intelligence could be partly inherited and partly acquired (environmental intelligence). A child who spends considerable time reading and using the computer may assimilate larger volumes of information and thus develop “environmental” intelligence. Another consideration is that the nonverbal Raven test may not be entirely independent of reading and linguistic ability and may be influenced by previous experiences with similar test situations. There is evidence that training and practice improves test scores. For example, overall test scores have increased in the past few decades, perhaps because individuals are now more familiar with IQ tests. ²¹ Thus, nonverbal Raven test performance may be a composite measure of intelligence as well as the ability to perform well in tests. Our study also found that children who read more perform better on IQ tests. Although the independent effect of IQ scores on the development of myopia was assessed in multivariate analyses, residual confounding by reading may still occur despite statistical adjustment. In short, children from more academically oriented families (where the parents may be more myopic) may be introduced early to education and study harder and may become myopic, and in parallel may tend to perform better on both IQ and school tests. Apart from intelligence, the test may also measure how well subjects form perceptual relations and reason by analogy. Another explanation is that aspects of IQ test performance, such as personality, discipline, and intelligence, may be inherited, together with myopia, through a direct link or even a common and still yet unidentified genetic factor. 

Potentially modifiable risk factors including outdoor activity and night lighting are not associated with incident myopia in our dataset. Outdoor activity may not have a direct effect on development of myopia, or perhaps the relatively low amounts and slight variations in outdoor activity may lead to false-negative results. 

Ethnicity was a major factor linked to the possibility of development of myopia. The risks of myopia are lower in the Singapore Malay compared with the Singapore Chinese children, despite the high myopia prevalence rates among all major ethnic groups in Singapore. This observation concurs with previous published reports of interethnic differences in the prevalence of myopia in 15,095 Singaporean military conscripts aged 16 to 25 years in 82.2% of Chinese, 68.7% of Indians, and 65.0% of Malays were shown to be myopic, ¹ as well as a prior descriptive report of the 3-year cumulative incidence rates in SCORM. ⁹ Interethnic differences independently predicted myopia, after accounting for parental history of myopia and IQ test scores. Dissimilar genetic predisposition or cultural lifestyle may contribute to higher incidence rates of myopia in Chinese than in Malay children. Because of the small number of Indian and Malay children enrolled in our study, there is insufficient power to evaluate the risk factors for myopia in Malays and Indians. Future studies could examine the individual effects of specific risk factors on the development of myopia among children of distinct ethnicities. 

This prospective cohort design is novel because of the comparisons of the likelihood of development of new cases of myopia for each risk factor. Of note, children without myopia were followed up over time, to determine who had and who had not become myopic. Although inaccurate recall may still exist, the possibility of differential recall bias is reduced, because the risk-factor questionnaire was completed at baseline in nonmyopic children before the development of any myopia in the study. Thus, the recall of risk factors will not be apparently different for children with and without incident myopia. The availability of extensive risk factor information also allows us to explore the relative contributions of each factor simultaneously. Because there is no common consensus about the best definition for myopia, the risk factors in Chinese children were evaluated for incident myopia by using three common definitions: −0.5, −0.75, and −1.0 D. The potential causative factors remained significant across the three different definitions of myopia. Although major risk factors such as near work activity and parental myopia for incident myopia were evaluated in the present study, complete data describing other possible factors including personality and breastfeeding are not yet available from a study in three schools. ²² Other limitations of note include possible misclassifications of parent-reported near work and parental myopia. 

In summary, this prospective cohort study showed that both parental myopia and children’s IQ are important risk factors for incident myopia in young Singaporean children. 

 Supported by Grant NMRC/0695/2002 from the National Medical Research Council, Singapore. RAS was supported by the Paul and Evanina Bell Mackall Foundation Trust.

 Submitted for publication August 15, 2005; revised October 11 and December 22, 2005, and January 13 and 27, 2006; accepted March 16, 2006.

 Disclosure: S.-M. Saw, None; A. Shankar, None; S.-B. Tan, None; H. Taylor, None; D.T.H. Tan, None; R.A. Stone, None; T.-Y. Wong, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Factors Associated with Incident Myopia

	At Risk (n)	Cases (n)	Crude RR Myopia at least −0.75 D (95% CI)	Age, Sex, Income RR Myopia at least −0.75 D (95% CI)	Multivariable RR* Myopia at least −0.75 D (95% CI)
Age (y)
9	178	69	1 (Referent)	1 (Referent)	1 (Referent)
8	330	137	1.06 (0.79–1.42)	1.00 (0.74–1.35)	1.08 (0.80–1.45)
7	486	248	1.35 (1.03–1.76)	1.32 (1.00–1.73)	1.42 (1.07–1.87)
Gender
Male	490	209	1 (Referent)	1 (Referent)	1 (Referent)
Female	504	245	1.19 (0.99–1.43)	1.19 (0.99–1.43)	1.21 (1.01–1.46)
Parental income (Singapore dollars)
2000 or below	239	96	1 (Referent)	1 (Referent)	1 (Referent)
2001 to 5000	410	184	1.17 (0.92–1.49)	1.17 (0.92–1.49)	1.01 (0.76–1.29)
Above 5000	318	163	1.40 (1.10–1.79)	1.44 (1.12–1.84)	1.11 (0.84–1.45)
Parents with myopia
None	349	119	1 (Referent)	1 (Referent)	1 (Referent)
Either	416	214	1.63 (1.31–2.04)	1.57 (1.24–1.98)	1.56 (1.24–1.98)
Both	228	120	1.70 (1.32–2.19)	1.58 (1.20–2.08)	1.55 (1.18–2.04)
Books Read per Week (Continuous Variable)	994	454	1.02 (0.98–1.06)	1.00 (0.92–1.10)	1.01 (0.97–1.05)
IQ, tertiles of Raven score, †
Tertile 1	440	168	1 (Referent)	1 (Referent)	1 (Referent)
Tertile 2	256	130	1.40 (1.12–1.75)	1.39 (1.11–1.76)	1.37 (1.08–1.72)
Tertile 3	298	156	1.50 (1.20–1.88)	1.51 (1.20–1.90)	1.50 (1.19–1.89)

 Myopia was defined as spherical equivalent of at least −0.75 D in Chinese children.

*  Multivariable risk ratio (RR) from Cox proportional hazard models were adjusted for school and all factors presented in the table.

†  See the Methods section for age-specific IQ tertiles.

Table 2.

 

Factors Associated with Incident Myopia, by Other Definitions in Chinese Children

	At Risk (n)	Cases (n)	Multivariate RR* Myopia at least −0.5 D (95% CI)	At Risk (n)	Cases (n)	Multivariable RR* Myopia at least −1.0 D (95% CI)
Age (y)
9	157	69	1 (Referent)	193	75	1 (Referent)
8	294	132	1.03 (0.76–1.39)	347	132	0.98 (0.73–1.32)
7	437	244	1.43 (1.08–1.90)	517	251	1.33 (1.01–1.75)
Gender
Male	434	214	1 (Referent)	523	201	1 (Referent)
Female	454	231	1.07 (0.89–1.29)	534	257	1.32 (1.10–1.59)
Parental income (Singapore dollars)
2000 or below	224	101	1 (Referent)	254	92	1 (Referent)
2001 to 5000	376	192	1.10 (0.86–1.41)	435	195	1.10 (0.86–1.41)
Above 5000	288	152	1.09 (0.83–1.44)	339	160	1.11 (0.84–1.46)
Parents with myopia
None	316	134	1 (Referent)	367	116	1 (Referent)
Either	373	197	1.28 (1.02–1.61)	443	216	1.58 (1.25–2.00)
Both	199	114	1.39 (1.06–1.82)	246	125	1.63 (1.24–2.14)
Books read per week (continuous variable)	888	445	1.01 (0.97–1.05)	1057	458	1.01 (0.97–1.06)
IQ, tertiles of Raven score, †
Tertile 1	375	160	1 (Referent)	465	167	1 (Referent)
Tertile 2	263	139	1.17 (0.92–1.48)	270	130	1.39 (1.10–1.75)
Tertile 3	250	146	1.45 (1.15–1.83)	322	161	1.52 (1.21–1.91)

 Myopia was Defined as spherical equivalent of at least −0.5 D and −1.0 D.

*  Multivariable risk ratio (RR) from Cox proportional hazard models were adjusted for school and all factors presented in the table.

†  See the Methods section for age-specific IQ tertiles.

Table 3.

 

Factors Associated with Incident Myopia in Malay and Indian Children

	At Risk (n)	Cases (n)	Multivariate RR* Malays (95% CI)	At Risk (n)	Cases (n)	Multivariable RR* Indians (95% CI)
Age (y)
9	71	20	1 (Referent)	26	11	1 (Referent)
8	91	19	0.82 (0.43–1.58)	29	13	1.39 (0.56–3.42)
7	131	33	1.10 (0.59–2.07)	42	10	0.71 (0.25–1.99)
Gender
Male	139	27	1 (Referent)	46	20	1 (Referent)
Female	154	45	1.49 (0.91–2.44)	51	14	0.67 (0.33–1.36)
Parental income (Singapore dollars)
2000 or below	205	47	1 (Referent)	35	11	1 (Referent)
2001 to 5000	73	21	1.02 (0.58–1.77)	47	19	1.16 (0.49–2.76)
Above 5000	8	1	0.42 (0.06–3.23)	11	3	0.68 (0.16–2.96)
Parents with myopia
None	194	41	1 (Referent)	60	20	1 (Referent)
Either	80	23	1.43 (0.85–2.42)	33	13	1.59 (0.76–3.32)
Both	18	8	2.29 (1.05–4.97)	3	1	1.35 (0.13–14.37)
Books read per week (continuous variable)	293	72	0.98 (0.89–1.07)	97	34	1.02 (0.93–1.12)
IQ, tertiles of Raven score, †
Tertile 1	195	44	1 (Referent)	67	19	1 (Referent)
Tertile 2	63	16	1.22 (0.66–2.25)	19	9	1.83 (0.79–4.26)
Tertile 3	35	12	1.75 (0.82–3.70)	11	6	2.06 (0.71–5.96)

 Myopia was defined as spherical equivalent of at least −0.75 D.

*  Multivariable risk ratio (RR) from Cox proportional hazard models were adjusted for school and all factors presented in the table.

†  See the Methods section for age-specific IQ tertiles.

The authors thank Angela Cheng and Liang Yu for their assistance in data collection and statistical analysis.