Risk variances of refractive error and axial length explained by polygenic risk scores of spherical equivalents

Abstract

Purpose : Genome-wide association studies identified 134 genetic variants of spherical equivalent in a meta-analysis conducted by the Consortium for Refractive Error and Myopia (CREAM) and 23andMe. We calculated the risk variances of refractive error and axial length explained by the carriership of these genetic variants of spherical equivalent, using polygenic risk scores, in the Generation R (children) and Rotterdam Study I, II and III (adults) cohorts.

Methods : Our analyses were performed in the population-based child cohort Generation R (follow up at 6 years, n= 3937; follow up at 9 years, n= 3414) and the adult cohort Rotterdam Study I, II and III (RS I-III; n= 10,792). Axial length was measured using the IOLmaster 700 in Generation R and the Lensstar LS900 in RS I-III. Cycloplegic refraction was performed at the follow-up at 9 years in Generation R using the Retinomax 3, while non-cycloplegic refraction was performed in RS I-III using a Topcon RM-A2000 auto refractor. Pearson correlation coefficient was calculated in SPSS version 21. Polygenic risk scores (PRS) were calculated using the summary statistics from the meta-analysis of CREAM and 23andMe (n=160,402). SNPs were selected on imputation quality score > 0.8 and minor allele frequency > 0.01. We performed P-value-informed clumping using PLINK. For each individual, PRS for spherical equivalent were calculated based on a range of P-value thresholds (from p<5x10^-8 up to p<1). We calculated the variance for spherical equivalent and axial length explained by each PRS model using regression analyses.

Results : In Generation R, mean axial length was 22.37±0.75 mm at 6 years, and 23.12±0.84 mm at 9 years. In RS I-III, mean axial length was 23.52±1.16 mm at 60 years. Mean spherical equivalent was 0.74±1.32 D in Generation R at 9 years and 0.27±2.42 D in RS I-III. The Pearson correlation coefficient between axial length and spherical equivalent was -0.62 at the age of 9 in Generation R and -0.73 in RS I-III. The maximum variance explained by PRS for axial length was 14.8% at 6 years and, 11.7% at 9 years in Generation R; and 7.4% in RS I-III. The maximum variance for spherical equivalent explained by PRS was 9.2% in Generation R and 7.8% in RS I-III.

Conclusions : PRS derived from a GWAS in adults is as applicable, if not better, to children to estimate the risk variances of spherical equivalent and axial length explained by genetics.

This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.