For the genome-wide survival analysis, we used 14,406 AD case and 25,849 control samples from the IGAP consortium (Table 1a). 8,253,925 SNPs passed quality control and were included for meta-analysis across all cohorts (Supplementary Table 1), which showed little evidence of genomic inflation (λ = 1.026). Four loci showed genome-wide significant associations (P < 5×10−8) with AAOS: BIN1 (P=7.6×10−13), MS4A (P=5.1×10−11), PICALM (P=4.3×10−14), and APOE (P=1.2×10−67) (Supplementary Fig. 1). While SNPs within BIN16, PICALM6,10, and APOE6,8–10,22 loci have previously been shown to be associated with AAO, this is the first time that the MS4A locus is reported to be associated with an AAO-related phenotype. The minor allele of rs7930318 near MS4A4A is associated with delayed AAO. Four other AD risk loci previously reported in the IGAP GWAS1 showed associations that reached suggestive significance (P < 1.0×10−5): CR1 (P=1.2×10−6), SPI1/CELF1 (P=5.4×10−6), SORL1 (P=1.8×10−7), and FERMT2 (P=1.0×10−5). The direction of effects were concordant with the previous IGAP GWAS logistic regression analysis for AD risk1 at all suggestive loci: AD risk-increasing alleles were all associated with a hazard ratio above 1 and earlier
