Abstract
BACKGROUND: To examine if the significantly associated SNPs derived from the genome wide allelic association study on the AREDS cohort at the NEI (dbGAP) specifically confer risk for neovascular age-related macular degeneration (AMD). We ascertained 134 unrelated patients with AMD who had one sibling with an AREDS classification 1 or less and was past the age at which the affected sibling was diagnosed (268 subjects). Genotyping was performed by both direct sequencing and Sequenom iPLEX system technology. Single SNP analyses were conducted with McNemar's Test (both 2 x 2 and 3 x 3 tests) and likelihood ratio tests (LRT). Conditional logistic regression was used to determine significant gene-gene interactions. LRT was used to determine the best fit for each genotypic model tested (additive, dominant or recessive). RESULTS: Before release of individual data, p-value information was obtained directly from the AREDS dbGAP website. Of the 35 variants with P < 10-6 examined, 23 significantly modified risk of neovascular AMD. Many variants located in tandem on 1q32-q22 including those in CFH, CFHR4, CFHR2, CFHR5, F13B, ASPM and ZBTB were significantly associated with AMD risk. Of these variants, single SNP analysis revealed that CFH rs572515 was the most significantly associated with AMD risk (P < 10-6). Haplotype analysis supported our findings of single SNP association, demonstrating that the most significant haplotype, GATAGTTCTC, spanning CFH, CFHR4, and CFHR2 was associated with the greatest risk of developing neovascular AMD (P < 10-6). Other than variants on 1q32-q22, only two SNPs, rs9288410 (MAP2) on 2q34-q35 and rs2014307 (PLEKHA1/HTRA1) on 10q26 were significantly associated with AMD status (P = .03 and P < 10-6 respectively). After controlling for smoking history, gender and age, the most significant gene-gene interaction appears to be between rs10801575 (CFH) and rs2014307 (PLEKHA1/HTRA1) (P < 10-11). The best genotypic fit for rs10801575 and rs2014307 was an additive model based on LRT. After applying a Bonferonni correction, no other significant interactions were identified between any other SNPs. CONCLUSION: This is the first replication study on the NEI dbGAP SNPs, demonstrating that alleles on 1q, 2q and 10q may predispose an individual to AMD.