Introduction

Several small retrospective studies have pointed to a negative association between smoking and the occurrence of Alzheimer disease or dementia in general. This might suggest that smoking could have a "protective" influence against the development of dementias - as it appears to have against Parkinson's disease.¹ To address this possibility, Doll and Peto (the authors of the article on smoking and the occurrence of Parkinson's disease) have reported their analysis of a prospective study they conducted in physicians in the UK.

Method

In connection with their findings on lung cancer and smoking, in 1951 Doll began collecting data on smoking habits from 34,439 male physicians on the British medical register and resident in the UK. Changes in such habits were sought at 6-11 year intervals up to 1998, and additional personal information was collected in the most recent 20 years.

By December 1998, 24,133 deaths had been recorded in the cohort. Dementia was mentioned in the death certificate in 473 of these subjects who died between 1962 and 1998; earlier deaths were not analyzed as they were within 10 years of first recording smoking habits.

As disease may itself influence smoking habits, the smoking data used were those from records made 10 or more years before death. The primary comparison was between subjects who were continuing smokers at that time, and those who had never smoked or were ex-smokers (the ex-smokers who died with dementia had stopped smoking on average 34 years before).

The two chief types of dementia were Alzheimer disease and cerebral vascular dementia, but an accurate distinction could not be made without histopathologic examination of postmortem tissue. Such results were available from 81 subjects with dementia. Comparison of the death certificate diagnoses and the diagnoses after full review of data from these subjects, including histopathology, showed reasonably good correspondence.

Each death caused by dementia was matched with 4 live age-matched controls, and each subject who died with dementia mentioned as an associated condition was matched with 4 controls with the same year of birth who had died in the same year as the subject. Standard Mantel-Haenszel analyses of contingency tables yielded relative risk values with 95% confidence limits.

Results

As the distinction between different types of dementia was uncertain, the 473 subjects with any dementia were analyzed. Their mean age at death was 81 years - only 2% of deaths with dementia occurred before 65. The relative risk of any dementia occurring in smokers was 0.96 (95%CI, 0.78-1.18).

Dementia probably or definitely due to Alzheimer disease was also unrelated to smoking; the relative risk was 0.99 (95%CI, 0.78-1.25), with the mean age at death being 82. There was no evidence of a significant association between continued smoking and the aggregate of all other types of dementia (relative risk 0.86, 95%CI 0.55-1.34). Most of these 103 cases were vascular dementia; there were 6 Lewy body dementias.

Comment

The results of this prospective study do not suggest that prolonged smoking either increases or decreases the likelihood that men will develop dementia of any type, including Alzheimer disease. Unlike the earlier retrospective studies reported, this study was long (over 45 years) and had complete follow-up, with the subjects classified according to their smoking habits about 15 years before their deaths. Although differentiation into types of dementia was not perfect, substantial misclassification would not invalidate the main conclusions.

There are two important implications of these findings. First, simple but well-conceived prospective studies can provide conclusive results that may refute the somewhat hasty associations uncovered by retrospective analyses. The authors review briefly other published prospective studies; in aggregate, these suggest a non-significant adverse effect of smoking in the short term (2-3 years).

Second, the results should remove one more excuse smokers have used to continue their habit. The slight apparent protective effect that smoking may have on the occurrence of Parkinson's disease and some inflammatory bowel diseases is clearly outweighed by the detrimental effects it has on other organs.