Smoking as a risk factor for Alzheimer's disease: contrasting evidence from a systematic review of case-control and cohort studies.

To investigate the risk of Alzheimer's disease (AD) associated with smoking. Meta-analyses of case-control and cohort studies. Index Medicus-Medline (1966-April 2000) and PsycINFO (1984-April 2000) databases were systematically consulted for the retrieval of references. This search was supplemented by manual search of relevant references quoted by other studies and reviews. Irrelevant abstracts and articles were identified by one of the authors. These papers were retrieved and examined by at least two of the authors, who initially assessed them for the relevance of the exposure (smoking), outcome (AD) and study-design (case-control or cohort study). Two reviewers rated independently the quality of selected papers. Whenever possible, raw data were extracted and the crude odds ratio (OR) calculated using the Cornfield method. The pooled risk ratios were estimated using a fixed-effects model. Twenty-one case-control studies reported data on 5323 subjects. The estimated pooled odds ratio (OR) was 0.74 [95% confidence interval (CI) = 0.66-0.84]. In another analysis incorporating ORs adjusted for confounding variables (such as age, sex, schooling and alcohol use), the pooled odds ratio was 0.82 (95% CI = 0.70-0.97). Finally, in a analysis that included only the four case-control studies that used matched design the pooled odds ratio was 0.82 (95% CI = 0.53-1.27). Eight cohort studies reported data on 43 885 people at risk-the overall relative risk (RR) of AD among ever smokers was 1.10 (95% CI = 0.94-1.29). Restricting the analysis to the two cohort studies that described the number of subjects who were smokers at baseline and later developed AD produced a RR of 1.99 (95% CI = 1.33-2.98). Case-control and cohort studies produce conflicting results as to the direction of the association between smoking and AD. Survival bias and other methodological problems associated with case-control studies may partly explain this difference. Access to information collected by ongoing follow-up studies may contribute to clarify the role of smoking in AD. If new results confirm that smoking is associated with increased risk of AD, then smoking prevention and cessation should become public health priorities in the fight against dementia.