IMPORTANCE Family studies of centenarians and long-lived persons have found substantial familial aggregation of survival to extreme ages; however, the extent to which such familial longevity is characterized by cognitively intact survival is not established. OBJECTIVE To determine whether families with exceptional longevity are protected against cognitive impairment consistent with Alzheimer disease. DESIGN Cross-sectional analysis. SETTING Multisite study in New York, Massachusetts, Pennsylvania, and Denmark. PARTICIPANTS A total of 1870 individuals (1510 family members and 360 spouse controls) recruited through the Long Life Family Study. MAIN OUTCOME AND MEASURE Prevalence of cognitive impairment based on a diagnostic algorithm validated using the National Alzheimer's Coordinating Center data set. RESULTS The cognitive algorithm classified 546 individuals (38.5%) as having cognitive impairment consistent with Alzheimer disease. Long Life Family Study probands had a slightly but not statistically significant reduced risk of cognitive impairment compared with spouse controls (121 of 232 for probands vs 45 of 103 for spouse controls; odds ratio = 0.7; 95% CI, 0.4-1.4), whereas Long Life Family Study sons and daughters had a clearly reduced risk of cognitive impairment (11 of 213 for sons and daughters vs 28 of 216 for spouse controls; odds ratio = 0.4; 95% CI, 0.2-0.9). Restriction to nieces and nephews in the offspring generation attenuated this effect (37 of 328 for nieces and nephews vs 28 of 216 for spouse controls; odds ratio = 0.8; 95% CI, 0.4-1.4). CONCLUSIONS AND RELEVANCE Rates of cognitive impairment characteristic of Alzheimer disease were relatively high. In the proband generation, rates were comparable across family members and spouse controls, whereas sons and daughters of probands had significantly lower rates than spouse controls. Results suggest a delayed onset of cognitive impairment in families with exceptional longevity, but assessment of age-specific incidence rates is required to confirm this hypothesis.