Previous studies have shown that Alzheimer's disease (AD) is associated with a reduced risk of cancer. However, most studies exclude those with undiagnosed dementia. The only way to overcome this methodological issue is to examine all the participants or to screen the population for symptoms of dementia with a validated instrument and confirm any suspected dementia patients with a clinical examination (i.e., a two-phase investigation method). We used this methodology to estimate whether cancer-specific mortality is associated with AD and other types of dementia in a prospective population-based study (NEDICES) involving 5,278 elderly people. Community-dwelling subjects with and without dementia were identified and followed for a median of 12.5 years, after which the death certificates of those who deceased were examined. A total of 1,976 (47.1%) died, including 277 who had possible or probable AD and 126 with non-AD dementia. Cancer was reported significantly less often in those with possible or probable AD (5.8%) or non-AD dementia (6.3%) than in those without dementia (26.5%). In an unadjusted Cox model, hazard ratio (HR) of cancer-specific mortality in participants with AD = 0.45 (p = 0.002) and HR in participants with non-AD dementia = 0.62 (p = 0.179) when compared to the non-demented group. In a Cox model that adjusted for a variety of demographic factors and co-morbidities, HRs of cancer-specific mortality in participants with AD = 0.50 (p = 0.028) and 0.97 (p = 0.942) in non-AD dementia. This study provides further evidence of an inverse association between cancer and AD.