To determine whether smoking is associated with Alzheimer's disease (AD).
Analyses were conducted using three Canadian data sets: the University of Western Ontario Dementia Study (200 cases, 163 controls), the Canadian Study of Health and Aging (258 cases, 258 controls), and the patient database from the Clinic for Alzheimer Disease and Related Disorders at the Vancouver Hospital and Health Sciences Centre (566 cases, 277 controls). The association between smoking and AD was investigated using bivariate analyses and multiple logistic regression models adjusted for the potential confounders age, sex, educational level, family history of dementia, head injury, and hypertension.
The results of bivariate analyses were inconsistent across the three data sets, with smoking status a significant protective factor, a significant risk factor, or not associated with AD. The results of multiple logistic regression models, however, were consistent: any association between smoking status and AD disappeared in all three data sets after adjustment for confounders.
Smoking status was consistently not associated with AD across all three data sets after adjustment for confounders. Failure to adjust for relevant confounders may explain inconsistent reports of the influence of smoking on AD. Any protective effect of smoking may be limited to specific AD subtypes (e.g., early onset AD).