Time Spent Outdoors Partly Accounts for the Effect of Education on Myopia

Abstract

Purpose: The purpose of this study was to investigate if education contributes to the risk of myopia because educational activities typically occur indoors or because of other factors, such as prolonged near viewing.

Methods: This was a two-sample Mendelian randomization study. Participants were from the UK Biobank, Avon Longitudinal Study of Parents and Children, and Generation R. Genetic variants associated with years spent in education or time spent outdoors were used as instrumental variables. The main outcome measures were: (1) spherical equivalent refractive error attained by adulthood, and (2) risk of an early age-of-onset of spectacle wear (EAOSW), defined as an age-of-onset of 15 years or below.

Results: Time spent outdoors was found to have a small genetic component (heritability 9.8%) that tracked from childhood to adulthood. A polygenic score for time outdoors was associated with children's time outdoors; a polygenic score for years spent in education was inversely associated with children's time outdoors. Accounting for the relationship between time spent outdoors and myopia in a multivariable Mendelian randomization analysis reduced the size of the causal effect of more years in education on myopia to -0.17 diopters (D) per additional year of formal education (95% confidence interval [CI] = -0.32 to -0.01) compared with the estimate from a univariable Mendelian randomization analysis of -0.27 D per year (95% CI = -0.41 to -0.13). Comparable results were obtained for the outcome EAOSW.

Conclusions: Accounting for the effects of time outdoors reduced the estimated causal effect of education on myopia by 40%. These results suggest about half of the relationship between education and myopia may be mediated by children not being outdoors during schooling.

Figure 1.

Outline of the study design. Panel (A) The study made use of three sets of GWAS summary statistics: a GWAS for time spent outdoors, a GWAS for spherical equivalent refractive error and a GWAS for years spent in education (EduYears). The three GWAS analyses were performed in nonoverlapping samples. The GWAS for spherical equivalent refractive error and the GWAS for time outdoors were newly performed analyses carried out in the current work, using samples from the UK Biobank. The GWAS for EduYears was performed in earlier work reported by the Within Family GWAS Consortium; the GWAS for EduYears included participants from UK Biobank and other cohorts (HUNT, Minnesota Twins, MoBa, Netherlands Twin Register, ORCADES, QIMR Berghofer Medical Research Institute, Swedish Twin Registry I, II, and YATSS, TEDS, TwinsUK, and Viking). Panel (B) Two child cohorts, ALSPAC and Generation R, were used to validate that (i) the GWAS for time outdoors identified genetic variants associated with the time children spent outdoors, and (ii) the GWAS for EduYears identified genetic variants associated with the time children spent reading. The Mendelian randomization analyses were performed using only the GWAS summary statistics for Time Outdoors, EduYears, and spherical equivalent refractive error. (Note that a GWAS for an early age-of-onset of spectacle wear [EAOSW] was also performed; this was done using the same sample as the GWAS for spherical equivalent refractive error. The summary statistics from the GWAS for EAOSW were used in place of those from the GWAS for spherical equivalent refractive error in a separate set of Mendelian randomization analyses). The sample size for the ALSPAC and Generation R analyses varied depending on the number of questionnaire respondents at each age, as detailed in Supplementary Tables S2 and S3.

Figure 2.

Association of polygenic scores for time outdoors or years spent in education with the time children spend outdoors or reading. Error bars show 95% CI. These results are presented in tabular form in Supplementary Tables S2 and S3.

Figure 3.

Principles of multivariable Mendelian randomization to estimate the causal effect of education on refractive error mediated by time spent outdoors. SNPs associated with years spent in education are used as instrumental variables to assess the causal effect of education on refractive error (SNP_Education → Education → Refractive error). SNPs associated with time outdoors are used as instrumental variables to assess the causal effect of time outdoors on refractive error (SNP_{Time outdoors} → Time outdoors → Refractive error). The causal pathway Education → Time outdoors → Refractive error is the pathway through which the causal effect of education on refractive error is mediated by time outdoors. The causal pathway Time outdoors → Education → Refractive error is the pathway through which the causal effect of time outdoors on refractive error is mediated by education. In a univariable Mendelian randomization analysis to assess the effect of education on refractive error, IVs for years spent in education acting through the pathway SNP_Education → Time outdoors → Refractive error would yield a biased estimate. In a multivariable Mendelian randomization analysis, the use of IVs for both years spent in education and time outdoors would avoid this source of bias.

Figure 4.

Univariable Mendelian randomization analysis of exposure to years spent in education (EduYears) or time spent outdoors on the outcomes of spherical equivalent refractive error or an early age-of-onset of spectacle wear (EAOSW). Each data point represents an independent genetic variant chosen for its strong association with either EduYears or time outdoors. Error bars show 95% CI. The slope of the solid line corresponds to the inverse-variance weighted Mendelian randomization (MR-IVW) causal effect estimate.

Figure 5.

Comparison of univariable vs. multivariable Mendelian randomization analysis to estimate the causal effects of exposure to years spent in education (EduYears) and time spent outdoors on the outcomes of spherical equivalent refractive error, or an early age-of-onset of spectacle wear (EAOSW). Error bars show 95% CI.