Smoking prevents dementia? Smoking causes dementia? Over the past decade a succession of research findings has produced apparently conflicting evidence on this question. In the early 1990s results from case-control and family studies suggested a protective effect.1,2 The findings were widely reported,3 including in the mass media, and some scientists stated publicly that they would consider taking up smoking if they had a family history of dementia.
Tobacco companies began to sponsor conferences on dementia, perhaps because it seemed to offer them a lifeline in an otherwise relentless sequence of findings about the deleterious effects of smoking. If smoking reduced life expectancy and also reduced the likelihood of survivors developing dementia then, from a policy perspective, there might be a role for the habit in later life.
The potential protective effects have some biological plausibility. Alzheimer's disease affects neurotransmitter systems, particularly the cholinergic system. Nicotine is a cholinergic agonist. The effect of nicotine on cognition, attention, and reaction time has been studied in non-cognitively impaired individuals, and nicotine is under investigation as a therapeutic agent in several disorders.4 A drug that acts on nicotinic receptors in the brain has just received regulatory approval in Sweden, the first such drug to be approved in the European Union. These effects are likely to be short term, with no obvious mechanism for long term effects. Moreover, smoking's effect on risk for vascular disease, including cerebrovascular disease,5 makes it a likely risk factor for vascular dementia over the longer term.6
The effects of smoking on dementia clearly need investigating in relatively unbiased populations and in longitudinal studies of reasonable duration in which the risk factors are examined before the onset of any dementia. In reporting one such study of British doctors in this week's BMJ (p1097), Doll et al also discuss the earlier case-control studies of risk for Alzheimer's disease, highlighting the deficiencies of this approach for dementia.7 In order to take part in a case-control study, patients need to have survived. For Alzheimer's disease the diagnostic criteria demand exclusion of those with vascular disorders, thus excluding those more likely to have smoked in earlier life. These are among the many reasons why the early case-control studies might have found a spurious “protective effect.” Although these limitations were discussed in most of the papers reporting an apparent protective effect and in commentaries,8 the reservations did not appear in all the media reports.
Currently a series of longitudinal studies of cognitive decline and dementia is under way in Europe. The EURODEM group has reported the first combined results of a group of European incidence studies in which people aged 65 and over were followed for dementia over two to three years, having answered questions on smoking at baseline.8 Doll et al review the findings from a selection of these studies,7 which provide no evidence of a protective effect and, if anything, suggest an increased risk.9 In a further longitudinal study from Sweden cross sectional analysis showed the apparent protective effect of smoking, but continued follow up revealed an increased risk of subsequent dementia in smokers.10 These provide evidence about risk over a short period of observation. The strength of the study reported in this week's issue is its much longer time frame.
Over the years the study of British doctors has provided crucial evidence on the wide ranging effects of smoking. The cohort has now reached the age at which the investigators can look at the impact of smoking in mid-life on mortality from dementia. The study was not designed to measure dementia as an outcome, and its measurement of dementia is relatively weak, as documentation of dementia on death certificates is known to be incomplete and usually mentioned only in moderate to severe cases. Diagnosis of subtypes of dementia is notoriously difficult during life. These limitations are unlikely, however, to have introduced systematic bias. The findings reported by Doll et al provide no evidence of a significant protective effect in men.
This study cannot provide the detail necessary to answer questions about the short term therapeutic effects of nicotine, its effects on minimal dementia and cognitive decline, gene-environment interaction, or gender specific effects. However, its findings are an important counterbalance to the potentially biased earlier studies. As Doll et al point out, these findings will be followed by many other longitudinal studies that should clarify the relation between smoking at various stages of life and subsequent dementia as most cohort studies now include some measures of cognition.
In the meantime, taking both the European cohort findings and the British doctors study together, the public health message is clear: at the population level there is no protective effect of smoking in dementia.