Development and validation of a questionnaire-based myopia proxy in adults: the LifeLines Cohort Study.

ABSTRACT

AIMS: To build a questionnaire-based myopia proxy and to validate the proxy by confirming its association with educational attainment and a Polygenic Risk Score (PRS) for myopia. METHODS: Data were collected between 2014 and 2017 from 88 646 Dutch adults from the LifeLines Cohort. First, we performed principal component analysis (PCA) to responses of five refraction-status questions. Second, we measured the refractive state in a subset of LifeLines participants (n=326) and performed logistic regression using myopia (mean spherical equivalent <-0.5 D) as a dependent variable and the principal components (PCs) as independent variables. We identified specificity, sensitivity and the classification threshold. Third, the classification equation was applied to the remaining LifeLines participants. The value of the proxy was then explored by calculating its association with educational attainment and a PRS of myopia. RESULTS: A total of 77 096 participants (58.1% women) were eligible for the PCA. The first two PCs had a specificity of 91.9% (95% CI 87.8% to 95.4%) and a sensitivity of 90.4% (95% CI 84.3% to 96.4%) for myopia. The area under the receiver operating characteristic curve was 95.0% (95% CI 92.2% to 97.8%). The age-standardised prevalence of proxy-inferred myopia was 33.8% (95% CI 33.4% to 34.3%). Compared with low education level, the ORs of proxy-inferred myopia were 1.66 (95% CI 1.58 to 1.74, p=5.94×10-90) and 2.54 (95% CI 2.41 to 2.68, p=4.04×10-271) for medium and high education levels, respectively. Similarly, individuals at the top 10% of PRS (vs lower 90%) had an OR of 2.18 (95% CI 1.98 to 2.41, p=6.57×10-56) for proxy-inferred myopia, whereas those at the highest decile had an OR of 4.51 (95% CI 3.9 to 5.21, p=1.74×10-89) when compared with the lowest decile. CONCLUSION: Self-administered refractive error-related questions could be used as an effective tool to capture proxy-inferred myopic cases in a population-based setting.