Article Text

Download PDFPDF

Environmental factors explain socioeconomic prevalence differences in myopia in 6-year-old children
  1. J Willem L Tideman1,2,
  2. Jan Roelof Polling1,3,
  3. Albert Hofman2,
  4. Vincent WV Jaddoe2,4,
  5. Johan P Mackenbach5,
  6. Caroline CW Klaver1,2
  1. 1 Department of Ophthalmology, Erasmus Medical Center, Rotterdam, The Netherlands
  2. 2 Department of Epidemiology, Erasmus Medical Center, Rotterdam, The Netherlands
  3. 3 Optometry & Orthoptics, University of Applied Science, Utrecht, The Netherlands
  4. 4 Department of Paediatrics, Erasmus Medical Center, Rotterdam, The Netherlands
  5. 5 Department of Public Health, Erasmus Medical Center, Rotterdam, The Netherlands
  1. Correspondence to Caroline CW Klaver, Erasmus Medical Center NA2808; PO Box 5201, 3008 AE Rotterdam, The Netherlands; c.c.w.klaver{at}


Purpose High myopia (≤−6 D) usually has its onset before 10 years of age and can lead to blinding complications later in life. We examined whether differences in myopia prevalences in socioeconomic risk groups could be explained by differences in lifestyle factors.

Methods A total of 5711 six-year-old children participating in the prospective population-based birth cohort study Generation R underwent a stepwise ophthalmic examination, which included visual acuity and objective cycloplegic refraction to identify children with myopia (≤−0.5D). Daily activities, ethnicity, factors representing family socioeconomic status and housing were ascertained by questionnaire. Risk assessments of myopia and mediation analyses were performed using logistic regression; attenuation of risks was calculated by bootstrapping.

Results Prevalence of myopia was 2.4% (n=137). Myopic children spent more time indoors and less outdoors than non-myopic children (p<0.01), had lower vitamin D (p=0.01), had a higher body mass index and participated less in sports (p=0.03). Children of non-European descent (OR 2.60; 95% CI 1.84 to 3.68), low maternal education (OR 2.27; 95% CI 1.57 to 3.28) and low family income (OR 2.62; 95% CI 1.8 to 3.74) were more often myopic. Lifestyle factors explained the majority of the increased risk for ethnicity (82%; 95% CI 55 to 120), maternal education (69%; 95% CI 45 to 109) and family socioeconomic status (71%; 95% CI 46 to 104).

Conclusion This study found environmental factors to be strong risk factors for myopia already at the age of 6 years. The myopia prevalence differences in socioeconomic groups were greatly determined by differences in distribution of these environmental risk factors, highlighting the importance of lifestyle adjustments in young children developing myopia.

  • myopia
  • children
  • lifestyle
  • mediation

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Myopia (nearsightedness) is an eye disorder with increasing prevalence,1 burden2 and corresponding economic costs3 during the past two decades. Most challenging for public health are the visual consequences, in particular of pathological myopia,1 due to myopic macular degeneration, glaucoma and retinal detachment.2 These changes often lead to irreversible visual impairment, emphasising the need to unravel the underlying causes.

The classical risk profile for myopia includes education,4 5 other socioeconomic factors6 and ethnicity.7–9 Higher education coincides with an almost three times increased risk.5 In East Asia prevalences reach 85% among school leavers,8 whereas in Europe prevalence rates are now approaching 50% in 25-year-olds.9 The biological basis for these risk factors is unclear, as is the consistency across age groups and countries. This lack of insight in the causal relationship hinders the development of effective clinical and public health campaigns. Recent research focus has shifted to the study of behavioural factors; of these, time spent outdoors,10 11 reading and indoor activities have been launched as the most prominent determinants.10 Whether these behavioural factors are consistent and whether they mediate in the association between classical risk factors and myopia have not been settled.

This study addresses the consistency of currently known environmental and socioeconomic risk factors in a cohort of young multiethnic children. We conducted mediation analysis to decipher the relevant components underlying the associations, and estimated to what extent these mediators explain the differential occurrence of myopia.

Population and methods

Study population

This study was embedded in the Generation R Study, a population-based prospective cohort study of pregnant women and their children in Rotterdam, The Netherlands. The complete methodology has been described elsewhere.12 Briefly, a total of 9778 pregnant women were included in the study, and children were born between April 2002 and January 2006. The children were invited at age 6 years for examination by trained nurses at the research centre. Of the initial cohort, 6690 (68.4%) children participated in the physical examination. The study protocol was approved by the Medical Ethical Committee of the Erasmus Medical Center, Rotterdam (MEC 217.595/2002/20), and written informed consent was obtained from all participants.

Assessment of myopia 

A two-step approach was performed to identify children with myopia. The first step included an ophthalmological examination consisting of visual acuity according to logMAR using LEA charts at a 3 m distance by means of the ETDRS method. Step 2 was carried out in children with a logMAR (Logarithm of the Minimum Angle of Resolution) visual acuity of >0.1 in at least one eye or in children with an ophthalmological history (visit to eye care practitioner), and included performance of automated subjective cycloplegic refraction (Topcon autorefractor KR8900 (Topcon, Tokyo, Japan)) in both eyes. Two drops (three in case of dark irises), with 5 min time interval, of cyclopentolate (1%) were administered ≥30 min before refractive error measurement. Pupil diameter was ≥6 mm at the time of the measurement. Spherical equivalent (SE) was calculated as sphere + ½ cylinder, and myopia was defined as SE ≤−0.5 D in at least one eye. Children with logMAR visual acuity ≤0.1 in both eyes, no glasses or ophthalmic history were classified as non-myopic.13

Ethnicity, education and income 

As a proxy for ethnicity, country of birth of the parents was obtained and determined by questionnaire using the method developed by Statistic Netherlands.14 Country of parental origin was grouped into Morocco, Turkey, Dutch Antilles and Surinam, and ‘other’ for risk estimation, and for final analysis grouped into European and non-European (see online supplementary table 1). Educational level of the mother and household income at age 6 years of the children were used to estimate social economic status. The highest educational level accomplished by mother and net household income were obtained by questionnaire and categorised into high (university or bachelor’s degree) or low (>3 years general secondary school or lower), and <€2400/month was categorised as low income (lowest tertile).

Supplementary Material

Supplementary material 1

Potential mediators

Type of house was categorised into a rental house or private property. Marital status was stratified into a single parent or living with a partner. Outdoor and indoor activities were time spent playing outdoors, biking and walking to school, time spent watching television and playing (handheld) computer games. Total time spent on activities was calculated in average hours/day. All outdoor activities (hour/day) were combined, and all indoor activities (hour/day) were subtracted to make a daily activity score of outdoor time relatively to indoor time per day, to avoid overfitting of the model.15 Sport participation was obtained using a questionnaire (‘Does your child participate in a sport?’). The measurements of 25(OH)D (25-hydroxy vitamin D, nmol/L) were conducted on blood samples collected at the research centre at 6 years of age, using the gold standard liquid chromatography/tandem mass spectrometry method at the Endocrine Laboratory of the Vrije Universiteit Medical Centre, Amsterdam, The Netherlands, between July 2013 and January 2014. Height and weight were measured at the research centre.

Statistical analysis

Differences in the European and non-European groups were calculated using χ2and Student’s t-test or Mann-Whitney U test. Four models were performed for testing associations between ethnicity, low income and low educational level versus myopia with logistic regression analysis. Model 1 included only adjustment for age and sex. Model 2 added inclusion for low household income versus myopia, and low maternal educational level versus myopia social economic factors such as marital status of the parents and rental house, and for ethnicity versus myopia marital status of the parents, rental house, and also family income and educational level of the mother. Model 3 included model 2 with additional inclusion of activity factors such as outdoor time relatively to indoor time per day and participation in sports. Model 4 included model 3 with additional inclusion of 25(OH)D and body mass index (BMI). Selected mediators were ordered and added to the model based on an hierarchical approach in which more distal mediators to the trait were first added to the model (see online supplementary figure 1).16 Multiple imputation procedures were used to replace missing covariates for the most likely values to avoid potential bias that may result from missing data,17 using fully conditional specification, an iterative of the Markov chain Monte Carlo approach.18 Data on playing outdoors (24.1%), data about housing (21.4%) and 25(OH)D (36.3%) were missing; all other covariates had missing values <20%. Mediation analyses was performed using the Baron and Kenny method,19 which requires mediators to fulfil the following criteria: (1) only factors associated with myopia independent of ethnicity or income/education were included in the model (table 1); and (b) mediators were unequally distributed over the ethnic groups (see online supplementary table 2), or between income/education groups (see online supplementary table 3). Differences in distributions were tested using logistic regression models. To calculate the attenuation of the effect estimate after adjusting for the mediator(s), the following formula was used: (100 × (B model 1 – B model 1 with explanatory factor) / (B model 1)). The bootstrap method was used to calculate a 95% CI around the percentage of attenuation with 1000 resamplings per imputed data set using the statistical program R. All other analysis were performed in SPSS V.

Table 1

Distribution of mediators for myopia, independent of ethnicity, maternal educational level or family income


A total of 5711 children were available for the analysis for ethnicity, maternal educational level and family income (figure 1). Of the total group, 31% (n=1764) were of a non-European descent. Cycloplegia was slower in children of non-European descent due to dark irises; however, this did not lead to a differential distribution of SE (Mann-Whitney U p=0.96). As shown in table 1, children with myopia were more likely to live with unmarried parents, to live in a rental home, to spend less time outdoors and more time indoors, have lower 25(OH)D levels, less participation in sports and a higher BMI.

Figure 1

Flow chart of participants eligible for analysis. SES, socioeconomic status.

Data on serum levels (such as 25(OH)D) had the highest proportion (36%) of missing values as result of refusal of blood withdrawal. Children with at least one predictor variable missing (n=3845) were less likely than children with complete data (n=1866) to participate in sports (p=0.04), to live in private property house (p<0.001), to have lower vitamin D (p<0.001) and to have lower BMI (p<0.001). Other mediators did not differ between the two groups. As this may cause selection bias, we imputed missing values and used data from all 5711 children for analysis.

Of the total group 2.4% (n=137) children were myopic. Children of Dutch Antilles, Surinamese (OR 3.29; 95% CI 2.13 to 5.10) and Moroccan descent (OR 2.35; 95% CI 1.34 to 4.13) were more likely to be myopic compared with their European peers. The total group of non-European children were more often myopic (table 2). When adjusting only for age and gender, low educational level of the mother and low family income were associated with a higher frequency of myopia (table 3).

Table 2

Association between ethnicity and risk of myopia with adjustment for mediators using various models

Table 3

Association between education and family income versus risk of myopia, and attenuation of the risk by mediators using various models

The association between family income or maternal education and risk for myopia was independent of ethnicity. In a sensitivity analysis performed for Europeans and non-Europeans, effect estimates showed similar effect in both groups: low education of the mother in Europeans (OR 1.79; 95% CI 1.07 to 2.99) and in non-Europeans (OR 1.66; 95% CI 0.93 to 2.96), and low family income in Europeans (OR 1.81; 95% CI 1.05 to 3.12) and in non-Europeans (OR 1.97; 95% CI 1.12 to 3.45).

All variables that remained significantly associated with myopia after adjustment for ethnicity, maternal educational level or family income entered the mediation analysis (table 1). As shown in table 2, 56% (95% CI 31 to 86) of the increased risk for non-European children was explained by differences in socioeconomic factors, and 82% (95% CI 55 to 120) could be explained by all mediators. For maternal education, 34% (95% CI 17 to 59) of the increased risk was explained by housing and marital status of the parents; an additional 29% was explained by daily activities and playing sports (table 3). The proportion of increased risk of myopia for low maternal education explained by all mediators was 69% (95% CI 45 to 109). For low family income, similar trends were observed, and the proportion of increased risk explained by all mediators was 71% (95% CI 46 to 104) (table 3).

As shown in tables 2 and 3, the differences in socioeconomic factors disappeared and became non-significant after adjusting for the more proximal lifestyle factors (ethnicity: OR 1.29 95% CI 0.83 to 1.99; maternal education: OR 1.40, 95% CI 0.92 to 2.12; and low family income: OR 1.47, 95% CI 0.96 to 2.25). The OR was slightly higher if in model 4 only 25(OH)D was added than with only BMI (low maternal education: OR 1.45, 95% CI 0.96 to 2.18 with BMI only vs OR 1.42, 95% CI 0.94 to 2.15 with 25(OH)D only; and low family income: OR 1.55, 95% CI 1.02 to 2.36 with BMI only vs OR 1.49, 95% CI 0.97 to 2.28 with 25(OH)D only). The mediating effect of 25(OH)D was higher for ethnicity, but additional adjustment for 25(OH)D or BMI did not alter significance of the association between ethnicity and myopia (with BMI only: OR 1.40, p=0.12; with 25(OH)D only: OR 1.31, p=0.23).


This study, which was performed in a multiethnic cohort in a densely populated area of the Netherlands, found environmental risk factors to be major determinants of myopia already occurring at the age of 6. We also found a higher frequency of myopia in children from families with low income, low maternal education and non-European ethnicity. Adjustment of these socioeconomic risk profiles for environmental factors caused the association to disappear, indicating a mediating effect of lifestyle.

Surprisingly, young children from families with a non-European ethnic background and/or a low socioeconomic status appear to be more often myopic in Rotterdam. To demystify the underlying causal structure of the profiles, we conceptualised the potential mediating pathways and ranked them in a causal diagram (see online supplementary figure 1). We considered living circumstances such as housing as more distal mediators, and daily activities such as nearwork and outdoor exposure as more proximal mediators. Children from families with a non-European ethnic background and/or a low socioeconomic status appeared to spend more time performing indoor activities, and to have less compensation by outdoor exposure, participated less in sports, had more often lower vitamin D levels, higher BMI and were living more often in rental houses than children from more advantaged families. We enriched the model step by step with these factors in the mediation analysis, and the decomposition of the increased risk was most profound when all factors were taken into account. Consequently, this suggests that the risk profiles based on education and income4 6 20–23 do not cause myopia, but represent certain living conditions and habits that are more directly involved in the pathogenesis of myopia. Our findings add to the growing bulk of evidence that daily activities of children are an important cause of myopia,10 and specifically show that these activities also underlie associations with ethnic background and socioeconomic status.

The myopia profiles found in this study may be specific for young children growing up in a big city in Europe. As daily habits change with age, the profiles may be modified as the children grow older. They may also remain in the same direction, because awareness of lifestyle risks may become greater in the highly educated parents. This was also the case for the association between socioeconomic status and smoking after the discovery of its detrimental health effects.24 25

Strengths of this study were the prospective design, which decreased the chance of selection bias, the mix of ethnicities and the wide range of variables available for analysis. Limitations were the relatively low number of myopes due to the young age of the children and the limited number of covariates that could be added to the mediation analysis. Another limitation was the lack of data on parental myopia, a well-known myopia risk factor and incomplete data on some of the variables which were not randomly distributed. To avoid selection bias, we applied the fully conditional specification method to replace missing variables, a widely accepted method for imputation.17

The myopia prevalence in the 6-year-olds of our study was somewhat higher than in Australian children of comparable age (1.5%),10 but much lower than in 7-year-old Chinese children (6.7%).26 We found an increased risk of myopia in non-European, more specifically in Dutch-Antillean or Surinamese and Moroccan ethnicity. Our group of children with an Asian descent was small, which hampered direct comparison with other ethnic comparison studies. Multiethnic studies with large number of Asians usually estimate the highest prevalence of myopia in the Asians, even at very young ages.27

In summary, this study in a large cohort of young children living in Western Europe demonstrated an important role of lifestyle in the development of myopia at a young age. Risks for socioeconomic groups should be deconstructed and deciphered into living circumstances and daily activities. For clinicians and researchers in the field of myopia, it is important to bear in mind that socioeconomic risk groups may differ between populations, but that mediators proximal to the trait are likely to remain the same. The more proximal risk factors can be modified at an individual level by increasing the level of outdoor activity in children, or as a population intervention with more time dedicated to outdoor exposure at schools.


The Generation R Study is conducted by the Erasmus Medical Center in close collaboration with the School of Law and Faculty of Social Sciences of the Erasmus University Rotterdam, the Municipal Health Service Rotterdam area, Rotterdam, the Rotterdam Homecare Foundation, Rotterdam, and the Stichting Trombosedienst & Artsenlaboratorium Rijnmond (STAR-MDC), Rotterdam. We gratefully acknowledge the contribution of children and parents, general practitioners, hospitals, midwives and pharmacies in Rotterdam. The authors were supported by the following foundations: MaculaFonds, Novartis Fonds, ODAS, LSBS, Oogfonds and ANVVB that contributed through UitZicht (grant 2014–38). The funding organisations had no role in the design or conduct of this research. They provided unrestricted grants.


View Abstract


  • Contributors Design and conduct of the study: JT and CCWK. Collection and management of the data: JT and JRP. Analysis and interpretation of the data: JT, JRP and CCWK. Preparation, review and approval of the manuscript: JT, JRP, VWVJ, AH, JPM and CCWK.

  • Funding The authors were supported by the following foundations: MaculaFonds, Novartis Fonds, ODAS, LSBS, Oogfonds and ANVVB that contributed through UitZicht (grant 2014-38). The funding organisations had no role in the design or conduct of this research. They provided unrestricted grants.

  • Competing interests None declared.

  • Patient consent Obtained.

  • Ethics approval Medical Ethical Committee of the Erasmus Medical Center, Rotterdam.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement No additional unpublished data from the study are available.

  • Correction notice This paper has been amended since it was published Online First. Owing to a scripting error, some of the publisher names in the references were replaced with ’BMJ Publishing Group'. This only affected the full text version, not the PDF. We have since corrected these errors and the correct publishers have been inserted into the references.

Linked Articles

  • At a glance
    Keith Barton James Chodosh Jost B Jonas