The evidence for an inverse relationship between smoking and dementia is inconclusive and based on small, frequently flawed retrospective studies. Doll and colleagues examined data from a large-scale epidemiologic study of British male physicians to investigate the relationship between smoking and dementia.

Data on the 34,439 physicians in the study were initially collected in 1951 during one of the first studies of smoking and lung cancer. The survey of smoking habits was repeated in 1957, 1966, 1972, 1978, 1990 and 1998. Additional information was collected in 1978, 1990 and 1998. Follow-up was discontinued in 10.1 percent of the survivors, mainly because they moved overseas.

The current analysis concerns 24,133 deaths that occurred from 1962 to December 31, 1998. Of these deaths, 483 included a mention of dementia on the death certificate. Eight deaths in which dementia was mentioned were excluded because they occurred less than 10 years after data on smoking habits were first obtained. Two additional deaths were excluded because the dementia was attributed to head injury and glioma.

Of the remaining 473 persons with dementia, data analysis did not indicate a relationship between dementia and smoking. The mean age at the time of death was 81 years. No significant difference was found between continuing smokers and long-term nonsmokers in the incidence of dementia in general or in the incidence of specific types of dementia.

The authors conclude that smoking does not increase or decrease the likelihood of dementia. In comparing their findings with those of four other prospective studies, the authors believe that the previously reported inverse relationship between smoking and dementia is an artifact. Although the effect is small, persistent smoking might increase rather than decrease the rate of dementia.