University of Southern Denmark Increasing Prevalence of Myopia in Europe and the Impact of Education

Williams, Katie M; Bertelsen, Geir; Cumberland, Phillippa; Wolfram, Christian; Verhoeven, Virginie J M; Anastasopoulos, Eleftherios; Buitendijk, Gabriëlle H S; Cougnard-Grégoire, Audrey; Creuzot-Garcher, Catherine; Erke, Maja Gran; Hogg, Ruth; Höhn, René; Hysi, Pirro; Khawaja, Anthony P; Korobelnik, Jean-François; Ried, Janina; Vingerling, Johannes R; Bron, Alain; Dartigues, Jean-François; Fletcher, Astrid; Hofman, Albert; Kuijpers, Robert W A M; Luben, Robert N; Oxele, Konrad; Topouzis, Fotis; von Hanno, Therese; Mirshahi, Alireza; Foster, Paul J; van Duijn, Cornelia M; Pfeiffer, Norbert; Delcourt, Cécile; Klaver, Caroline C W; Rahi, Jugnoo; Hammond, Christopher J; European Eye Epidemiology (E(3)) Consortium; Grauslund, Jakob

prevalence of myopia seems to be lower 4,5 ; however, there is evidence of an increasing prevalence in the United States and elsewhere, 6e8 particularly among young adults. 9 This is of concern because myopia, even when appropriately corrected, is associated with an increased risk of sightthreatening diseases, such as myopic maculopathy, retinal detachment, glaucoma, and cataract. 10 Myopic maculopathy is currently untreatable and already contributes to visual impairment in working-age adults. 11 Increasing myopia levels in Europe carry implications for public health policy in both the provision of clinical services and the economic sequelae from the resulting visual impairment among the working population.
Myopia is a highly heritable trait, 12,13 and to date a number of genetic polymorphisms have been associated with refractive error, albeit explaining only a small proportion of this heritability. 14,15 Environmental factors play a key role in myopia development and must explain the recent changes in prevalence. 16 Myopia has been associated with education, near work, urbanization, prenatal factors, socioeconomic status, cognitive ability, season of birth, light, and time spent outdoors. 2,16e25 One of the strongest and most replicated risk factors is educational attainment, 16,26 and there is some evidence of interaction between genetic factors and education influencing the risk of myopia. 27 The increased levels of higher education over the 20th century 28 might be a causative factor, or marker of a causative factor, for increasing myopia prevalence.
The aims of this study are to identify whether myopia is becoming more common across Europe and to examine whether increasing levels of education explain any temporal trend, using data from more than 60 000 participants from the European Eye Epidemiology (E 3 ) Consortium.

Study Population
The E 3 consortium is a collaborative initiative to share and metaanalyze epidemiologic data on common eye diseases across Europe. Thirty-three studies are currently part of the consortium, and a range of ophthalmic data are available on approximately 124 000 individuals from population-based and case-control cohorts. All studies adhere to the tenets of the Declaration of Helsinki, and relevant local ethical committee approvals with specific study consent were obtained.
Refractive error measurements from 68 350 adults within the 15 E 3 population-based studies that had data on refractive error were included. These included population-based cross-sectional or cohort studies, with 2 studies recruiting participants nationally and 13 studies recruiting from a local population. Further details on each study are provided in Table 1 and the Supplementary information (available at www.aaojournal.org). Exclusion criteria included subjects who had cataract or refractive surgery, retinal detachment, or other conditions, such as keratoconus, which might influence refraction (n ¼ 6404). Data on age at refraction and birth year were available for 61 946 individuals, with information on education level for 60 125 subjects. Participants were mainly middle to late age; 98% were of European descent (where ethnicity was known), predominantly from Northern and Western Europe; and refractive examinations were performed from 1990 to 2013 (Table 1).

Study Variables
Noncycloplegic refractions were performed on all individuals using subjective refraction, autorefraction, or a combination of focimetry with subjective refraction. Spherical equivalent was calculated using the standard formula (spherical equivalent ¼ sphere þ [cylinder/2]). Myopia was defined as À0.75 diopters. Myopia prevalence by age was calculated, using 5-and 10-year age bands from !15 years to !90 years. To study the impact of education on myopia, given the variation in educational systems across Europe, we established a simplified 3-tier level of education across all cohorts. Primary education was defined as those leaving school before 16 years of age, secondary education was defined as those leaving education up to the age of 19 years, and higher education was defined as those leaving education at or after the age of 20 years. Those aged younger than 20 years at the time of refraction (and therefore unable to have reached the highest education tier) were excluded from this analysis to avoid misclassification bias.
We investigated the evidence for a cohort effect on increasing myopia prevalence by observing variations in myopia prevalence within defined age bands. These analyses are focused on the age range constituting the majority of our cohort (40e80 years of age, birth year 1910e1979, n ¼ 56 088), meaning the youngest and oldest participants, for whom we had no comparative birth cohort, were not considered. Prevalence between different birth cohorts was examined, initially using decade bins (1910e1970) and subsequently in 2 birth cohort groups divided by the median birth decade (1940e1949). Finally we examined the influence of education by examining the myopia prevalence between birth cohorts with the additional stratification of educational status.

Statistical Analysis
Study-specific summary data for myopia prevalence were obtained and combined in a random-effect meta-analysis stratified by age. A random-effects model was chosen over a fixed-effects model to allow for expected heterogeneity between studies as a result of varying study design. Age was standardized with demographic distribution adjustments to age-specific estimates according to the European Standard Population 2010. 29 Evidence for the presence of a cohort effect was investigated using random-effect meta-analyses of myopia prevalence stratified by age and birth year, and subsequently age, birth year, and educational level. Differences between estimates of myopia prevalence were evaluated using the analysis of variance test, proportion z tests, and prevalence ratios (relative difference in prevalence against a defined baseline). Differences were considered significant at P < 0.05.

Cohort Effect for Increasing Myopia Prevalence
There was a trend of higher myopia prevalence with more recent birth decade across all age groups (Fig 1), although sample sizes for some point estimates were low, resulting in wide CIs ( Table 2, available at www.aaojournal.org).

Influence of Education on Myopia Risk and the Cohort Effect
The association between education and myopia was investigated in the 13 studies from which these data were available (n ¼ 60 125 participants). Educational level was significantly associated with myopia prevalence across all age strata (P < 0.0001). Overall, the age-standardized myopia prevalence for those completing primary, secondary, and higher education was 25.4% (95% CI, 25.0e25.8), 29.1% (95% CI, 28.8e29.5), and 36.6% (95% CI, 36.1e37.2), respectively. In those aged 35 to 84 years, the majority of study subjects, myopia prevalence in participants with higher education was approximately double those with primary education (Fig 3). For example, in subjects aged 45 to 49 years when tested, the myopia prevalence was 26.3% (95% CI, 20.1e32.5) compared  with 51.4% (95% CI, 46.7e56.0) for those with primary and higher education, respectively, and in those aged 60 to 64 years, myopia prevalence was 14.0% (95% CI, 12.3e15.8) compared with 28.7% (95% CI, 25.4e32.0) for those with primary and higher education, respectively. The trends observed are less clear in younger subjects (<35 years) in Figure 3, most likely because of small sample sizes (n ¼ 216 aged 20e25 years, n ¼ 336 aged 25e30 years), which are further stratified by education level with corresponding wide CIs.
Levels of education throughout Europe have increased in the past 90 years (Fig 4). The proportion of individuals progressing to higher education increased from 4% of those born in the 1900s to 16% in the 1920s, 20% in the 1940s, 33% in the 1960s, and approximately 61% in the 1980s.
However, although those born more recently were more likely to have achieved a higher educational level, this alone did not explain the cohort effect of increasing myopia. As shown in Figure 5, for individuals aged 45 to 65 years (age range selected for minimal age-related myopia variance and large available sample size), the increase in myopia prevalence with a more recent birth decade was observed across all educational groups. This was most pronounced for participants achieving only a primary education, in whom myopia prevalence increased from 10.7% (95% CI, 7.6e13.8) to 28.1% (95% CI, 18.1e38.0) between birth decades 1920 to 1929 and 1960 to 1969 (P ¼ 0.001). The corresponding increase in myopia in those with higher education was from 26.0% (95% CI, 17.4e34.6) to 40.2% (95% CI, 30.5e50.0) (P ¼ 0.03). Compared with the reference risk of participants with primary education and born in the 1920s, the myopia prevalence ratio for those achieving a higher education was 2.43 (95% CI, 1.26e4.17) and for those born in the 1960s was 2.62 (95% CI, 1.31e5.00). Individuals born in the 1960s and completing higher education had approximately 4 times the baseline risk, with a prevalence ratio of 3.76 (95% CI, 2.21e6.57). Thus, the individual associations of educational level and birth cohort had an additive effect on myopia prevalence.  . Prevalence of myopia (spherical equivalent À0.75 diopters) with 95% confidence interval stratified by highest educational level achieved: primary education, leaving education at age <16 years; secondary education, leaving school at age 19 years; higher education, leaving school at age !20 years.

Discussion
Our study provides the first evidence that myopia is becoming more common across Western and Northern Europe, with a clear trend of higher myopia prevalence in participants with a more recent birth year (Fig 1). This is similar to the increase reported in North America and, albeit to a lesser extent, Southeast Asian populations. 6,7,32,33 Evidence of increasing myopia prevalence carries clinical and economic implications. The increased requirement for detection and treatment of myopia, entailing glasses, contact lenses, or more recently laser refractive surgery, has significant implications for clinical optometric and ophthalmic service provision, and the health care system. Additional ophthalmic services will be needed for treatable sightthreatening complications, such as retinal detachment, glaucoma, and cataract. 10,34 The increasing prevalence of myopia also implies that untreatable complications, such as myopic maculopathy, most commonly seen in high myopia, will become more common. This will result in more visual impairment in middle-to older-aged individuals, including a proportion of the working-age population, with consequent economic implications.
Myopia has been strongly associated with education, 2,21,24,35 and we explored this using a simple 3-tier classification of educational level. Increasing educational level had a strong effect, with myopia twice as common in those achieving a higher education compared with participants leaving school before 16 years of age. There was a clear trend of increasing prevalence of myopia across the tiers of education level, suggesting a potential additive effect of years of education. This interesting association may reflect a number of factors: greater near work activities with more education and less time in outdoor light, shared genetic factors underlying myopia and intelligence, or factors related to educational opportunity, such as socioeconomic status or maternal nutrition. These associations have been explored in younger cohorts, 18e21,36,37 although causal pathways are yet to be fully understood.
Reasons for the observed cohort effect are clearly multifactorial, and education is an obvious possible explanation; in our data, only 12% of participants born in the 1920s went on to higher education, compared with 33% born in the 1960s. This educational expansion has been observed across Europe in both men and women, with a sharp trajectory toward mass higher education after World War II. 28,38 In addition to the disruption of education and economic consequences of World War II, adverse health outcomes have been reported in young people growing up at that time, notably diabetes, depression, and heart disease. 39 Although there is no known direct link between these health issues and myopia, the deprivation may have affected eye growth and resulting refraction. Certainly there was an increase in myopia in subjects born after 1950, but it is difficult to be certain what aspect of the seismic changes in Europe after the war might be responsible.
Although the younger generations were more educated, we found a clear increase in the prevalence of myopia across the birth cohorts within each educational stratum, as well as the additive effect of educational status. Therefore, increasing levels of myopia were not explained by education alone, and a more recent birth year and higher educational level had an additive effect on myopia risk. Our simple 3tier education stratification may be subject to residual confounding from variation in educational practices, and it may be these, rather than changes in education level, that are contributing to the observed cohort effect. In the latter half of the last century, there was increasing use of computers, increasing length of the educational day with increased after-school tuition, and less outdoor play as a result of reduced recess time. 35

Study Limitations
The E 3 consortium has provided a large data set to metaanalyses' temporal trends and educational associations for myopia prevalence across Europe. Limitations to this consortium meta-analysis include heterogeneity between studies. Contributing studies inherently differed in study design and cohort sampling. In acknowledgment of this heterogeneity, we performed a random-effect rather than a fixed-effect metaanalysis, assuming no fixed effect between studies. There are also differences between European countries in terms of urbanization, economy, social class, education, and lifestyle, which are known to influence myopia. Data on these variables at an individual or study-specific level were not uniformly available, and data often were collected from middle-aged and older participants, so retrospective collection of potential contributing factors such as outdoor exposure, amount of reading, and area of residence during the critical first 20 years of refractive error development would be impossible. In addition, potential multicollinearity of these likely highly correlated factors (e.g., reading and education) would make assessment of separate effects difficult. In an attempt to reduce heterogeneity arising from these associated Figure 5. Myopia prevalence (spherical equivalent À0.75 diopters) by birth cohort and educational level in individuals aged 45 to 65 years: primary education, leaving education at age <16 years; secondary education, leaving school at age 19 years; higher education, leaving school at age !20 years.
factors, we stratified the random-effects meta-analysis by age and educational level (both significantly associated with myopia). Applicability of our findings is greatest for middleto older-aged individuals and for those from Northern and Western European countries, given the sampled ages and the location of the E 3 studies (Table 1), although ultimately the degree to which these studies are representative of the underlying population is unknown.
Further limitations include the crude nature in which education was classed, which as previously acknowledged may result in residual confounding. In addition, education status was collected retrospectively and therefore prone to recall error, possibly heightened in older participants. Refractions were all noncycloplegic, although this is reasonable given the age of participants. 40,41 Finally, these data are not longitudinal, so we have not examined reasons for the lower prevalence with age within birth decades, although the cohort effect we identified may be part of this explanation. Other reasons include the well-known hyperopic shift with age and could include other factors, such as censoring with age if myopic subjects receive earlier cataract surgery.
In conclusion, the prevalence of myopia is increasing in Europe, a finding that is not fully explained by increasing education levels despite higher educational achievement being associated with myopia and becoming more widespread in Europe. The changes in prevalence are similar to those observed in North America, although they remain far less than those identified in Southeast Asia, possibly because of differing intensity of education from an early age. 1,6,35 High levels of myopia were detected in the younger adults with a more recent birth year, of whom approximately half were affected. This has significant implications for the future; increasing myopia prevalence, and specifically high levels in younger individuals, will potentially result in an increasing burden of associated visual impairment in the future. UK; UCLA Center for Eye Epidemiology, Los Angeles, CA;