Hereditary hearing loss (HL) is a very heterogeneous trait, with 46 gene identifications for non-syndromic HL. Mutations in GJB2 cause up to half of all cases of severe-to-profound congenital autosomal recessive non-syndromic HL, with 35delG being the most frequent mutation in Caucasians. Although a genotype-phenotype correlation has been established for most GJB2 genotypes, the HL of 35delG homozygous patients is mild to profound. We hypothesise that this phenotypic variability is at least partly caused by the influence of modifier genes. By performing a whole-genome association (WGA) study on 35delG homozygotes, we sought to identify modifier genes. The association study was performed by comparing the genotypes of mild/moderate cases and profound cases. The first analysis included a pooling-based WGA study of a first set of 255 samples by using both the Illumina 550K and Affymetrix 500K chips. This analysis resulted in a ranking of all analysed single-nucleotide polymorphisms (SNPs) according to their P-values. The top 250 most significantly associated SNPs were genotyped individually in the same sample set. All 192 SNPs that still had significant P-values were genotyped in a second independent set of 297 samples for replication. The significant P-values were replicated in nine SNPs, with combined P-values between 3 x 10(-3) and 1 x 10(-4). This study suggests that the phenotypic variability in 35delG homozygous patients cannot be explained by the effect of one major modifier gene. Significantly associated SNPs may reflect a small modifying effect on the phenotype. Increasing the power of the study will be of greatest importance to confirm these results.