Importance: Keratoconus can be a debilitating corneal ectasia in which the cornea thins, bulges, and steepens into a conical shape. Early features of keratoconus include myopia and irregular astigmatism, which affect vision and can be treated with contact lenses, collagen cross-linking, or, in advanced cases, corneal transplant. Recent estimates of the prevalence of keratoconus based on results of Scheimpflug imaging in young adults are as high as 1.2%. However, obtaining very large keratoconus data sets for a genome-wide association study (GWAS) is problematic because few population studies include Scheimpflug imaging and because severe keratoconus is relatively rare.
Objective: To identify novel keratoconus loci using corneal resistance factor (CRF) and central corneal thickness (CCT).
Design, setting, and participants: This multitrait GWAS used European ancestry CRF data from UK Biobank (UKB) (n = 105 427) and the Canadian Longitudinal Study on Aging (CLSA) (n = 18 307) and European ancestry CCT data from the International Glaucoma Genetics Consortium (IGGC) (n = 17 803). The CRF and CCT variants in published keratoconus data sets (4669 cases and 116 547 controls) were compared. The data set from UKB was compiled March 24, 2020; data were released from the CLSA in July 2020; and IGGC data were available from May 1, 2018.
Main outcomes and measures: Association of CRF and CCT variants with keratoconus risk.
Results: The GWAS included 4 cohorts: 105 427 UKB European ancestry (56 134 women [53.2%] and 49 293 men [46.7%]; mean [SD] age, 57  years), 5029 UKB South Asian ancestry (2368 women [47.1%] and 2661 men [52.9%]; mean [SD] age, 54  years), 902 UKB East Asian ancestry (622 women [68.9%] and 280 men [31.0%]; mean [SD] age, 53  years), and 18 307 CLSA European ancestry (9260 women [50.6%] and 9047 men [49.4%]; mean [SD] age, 63  years) participants. A total of 369 CRF and 233 CCT loci were identified, including 36 novel CRF loci and 114 novel CCT loci. Twenty-nine CRF loci and 24 CCT loci were associated with keratoconus. Polygenic risk scores (PRS) were constructed using CRF- and CCT-associated variants and published keratoconus variants. The PRS result showed that adding a CRF- or CCT-based PRS to the keratoconus PRS from previously published variants improved the prediction area under the receiver operating characteristic curve (from 0.705 to 0.756 for CRF and from 0.715 to 0.755 for CCT).
Conclusions and relevance: These findings support the use of multitrait modeling of corneal parameters in a relatively large data set to identify new keratoconus risk loci and enhance polygenic risk score models.