Purpose: To compare the performance of a novel method for identifying candidate gene-environment interaction loci, Scalable Cauchy Aggregate test using Multiple Phenotypes to test Interactions (SCAMPI) (doi:10.1101/2024.09.10.612314v1), to widely used existing methods, Levene’s test and conditional quantile regression (CQR). The 3 methods were evaluated for refractive error, a trait known to exhibit widespread gene-environment interaction effects. Design: A cohort study; a genome-wide association study to investigate variance heterogeneity of refractive error. Participants: A discovery sample of 77 880 UK Biobank participants with records of both refractive error and age of onset of spectacle wear (AOSW) and a validation sample of 257 265 UK Biobank participants with known AOSW. Methods: SCAMPI, Levene’s tests, and CQR were applied in the discovery sample. Variance quantitative trait loci (vQTLs) identified using SCAMPI were assessed in the independent validation sample. The SCAMPI vQTLs were further assessed for evidence of genotype-by-education interaction or gene–gene interaction, using linear regression. Main Outcome Measures: Genetic variance heterogeneity associated with refractive error (phenotypic variance differences across genotypes). Results: SCAMPI identified 15 independent vQTLs with P <5.0e-08 while 3 and 11 were found using Levene’s test and CQR, respectively. Of the 15 SCAMPI vQTLs, 12 (80%) were supported in the validation data set after accounting for multiple testing. The lead SCAMPI variants included known vQTL associated with refractive error, such as rs12193556 (LAMA2) and rs685352 (GJD2), as well as 3 novel candidate gene-environment interaction loci. Among these novel candidates was rs7077247, an intronic variant in the TCF7L2 gene, which was associated with a –0.085 D greater shift toward myopia in participants with a university degree (P = 9.43e-04). This variant is known to be associated with the risk of diabetes and TCF7L2 is a component of the Wnt signaling pathway. The SCAMPI vQTLs demonstrated minimal evidence for gene–gene interactions. Conclusions: The SCAMPI outperformed Levene’s test and CQR in identifying candidate gene-environment interaction loci associated with refractive error. Our results suggest that a variant in the TCF7L2 gene may confer susceptibility to myopia to a greater extent in those with higher vs. lower educational attainment and suggests a molecular link to Wnt signaling and the risk of myopia associated with intensive education. Financial Disclosure(s): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.