IMPORTANCE: Mutations in known causal Alzheimer disease (AD) genes account for only 1% to 3%of patients and almost all are dominantly inherited. Recessive inheritance of complex phenotypes can be linked to long (>1-megabase [Mb]) runs of homozygosity (ROHs) detectable by single-nucleotide polymorphism (SNP) arrays. OBJECTIVE: To evaluate the association between ROHs and AD in an African American population known to have a risk for AD up to 3 times higher than white individuals. DESIGN, SETTING, AND PARTICIPANTS: Case-control study of a large African American data set previously genotyped on different genome-wide SNP arrays conducted from December 2013 to January 2015. Global and locus-basedROHmeasurementswere analyzed using rawor imputed genotype data.We studied the rawgenotypes from 2 case-control subsets grouped based on SNP array: Alzheimer's Disease Genetics Consortium data set (871 cases and 1620control individuals) and Chicago Health and Aging Project-Indianapolis Ibadan Dementia Study data set (279 cases and 1367 control individuals).We then examined the entire data set using imputed genotypes from 1917 cases and 3858 control individuals. MAIN OUTCOMES AND MEASURES: The ROHs larger than 1Mb, 2Mb, or 3Mb were investigated separately for global burden evaluation, consensus regions, and gene-based analyses. RESULTS: The African American cohort had a lowdegree of inbreeding (F × 0.006). In the Alzheimer's Disease Genetics Consortium data set, we detected a significantly higher proportion of cases with ROHs greater than 2Mb (P =.004) or greater than 3Mb (P =.02), aswell as a significant 114-kilobase consensus region on chr4q31.3 (empirical P value 2 =.04; ROHs >2 Mb). In the Chicago Health and Aging Project-Indianapolis Ibadan Dementia Study data set, we identified a significant 202-kilobase consensus region on Chr15q24.1 (empirical P value 2 =.02; ROHs >1 Mb) and a cluster of 13 significant genes on Chr3p21.31 (empirical P value 2 =.03; ROHs >3 Mb). Atotal of 43 of 49 nominally significant genescommonfor both data sets also mapped to Chr3p21.31. Analyses of imputed SNP data from the entire data set confirmed the association of AD with global ROH measurements (12.38 ROHs >1Mb in cases vs 12.11 in controls; 2.986Mb average size of ROHs >2Mb in cases vs 2.889Mb in controls; and 22%of cases with ROHs >3Mb vs 19% of controls) and a gene-cluster on Chr3p21.31 (empirical P value 2 =.006-.04; ROHs >3 Mb). Also, we detected a significant association between AD and CLDN17 (empirical P value 2 =.01; ROHs >1 Mb), encoding a protein from the Claudin family, members of whichwere previously suggested as ADbiomarkers. CONCLUSIONS AND RELEVANCE: To our knowledge, we discovered the first evidence of increased burden of ROHs among patients with AD from an outbred African American population, which could reflect either the cumulative effect of multiple ROHs to AD or the contribution of specific loci harboring recessive mutations and risk haplotypes in a subset of patients. Sequencing is required to uncover AD variants in these individuals.