CpG-related single nucleotide polymorphisms (CGS) have the potential to perturb DNA methylation; however, their effects on Alzheimer disease (AD) risk have not been evaluated systematically. We conducted a genome-wide association study using a sliding-window approach to measure the combined effects of CGSes on AD risk in a discovery sample of 24 European ancestry cohorts (12,181 cases, 12,601 controls) from the Alzheimer's Disease Genetics Consortium (ADGC) and replication sample of seven European ancestry cohorts (7,554 cases, 27,382 controls) from the International Genomics of Alzheimer's Project (IGAP). The potential functional relevance of significant associations was evaluated by analysis of methylation and expression levels in brain tissue of the Religious Orders Study and the Rush Memory and Aging Project (ROSMAP), and in whole blood of Framingham Heart Study participants (FHS). Genome-wide significant (p < 5 × 10-8 ) associations were identified with 171 1.0 kb-length windows spanning 932 kb in the APOE region (top p < 2.2 × 10-308 ), five windows at BIN1 (top p = 1.3 × 10-13 ), two windows at MS4A6A (top p = 2.7 × 10-10 ), two windows near MS4A4A (top p = 6.4 × 10-10 ), and one window at PICALM (p = 6.3 × 10-9 ). The total number of CGS-derived CpG dinucleotides in the window near MS4A4A was associated with AD risk (p = 2.67 × 10-10 ), brain DNA methylation (p = 2.15 × 10-10 ), and gene expression in brain (p = 0.03) and blood (p = 2.53 × 10-4 ). Pathway analysis of the genes responsive to changes in the methylation quantitative trait locus signal at MS4A4A (cg14750746) showed an enrichment of methyltransferase functions. We confirm the importance of CGS in AD and the potential for creating a functional CpG dosage-derived genetic score to predict AD risk.