Impact of varying the definition of myopia on estimates of prevalence and associations with risk factors: time for an approach that serves research, practice and policy

Background Refractive error is an increasing global public health concern that requires robust and reliable research to identify modifiable risk factors and provide accurate estimates of population burden. We investigated the impact of reclassification of individuals when using different threshold values of spherical equivalent (SE) to define myopia, on estimates of frequency, distribution and associations with risk factors, to inform current international initiatives to standardise definitions. Methods A random sample of 1985 individuals from the 1958 British birth cohort, at age 44, had autorefraction and self-reported on educational attainment and social class. Refraction status assigned in three different models using SE: (A) moderate to high myopia -3 diopters (D) or more extreme (<=-3.00D), (B) hypermetropia +1.00D or more extreme (>=+1.00D) and (C) mild myopia using three different thresholds: -1.00D, -0.75D or -0.50D, hence reciprocal changes in definition of emmetropia. Results Frequency estimates and associations with risk factors altered significantly as the threshold value for myopia moved towards SE 0.0D: prevalence of mild myopia increased from 28% to 47%, the association with highest educational attainment attenuated and with higher social class strengthened, with changes in risk ratios of approximately 20%. Conclusion Even small changes in the threshold definition of myopia (+/- 0.25D) can significantly affect the conclusions of epidemiological studies, creating both false-positive and false-negative associations for specific risk factors. An international classification for refractive error, empirically evidenced and cognisant of the question(s) being addressed and the population(s) being studied, is needed to serve better translational research, practice and policy.