The number of people living with dementia worldwide is expected to triple in the next 30 years, with 135 million people predicted to have dementia by 2050 due to population ageing. This makes the dementia epidemic one of the greatest global challenges for health and social care in the 21st century. The failure of multi-domain interventions and drug trials has led to interest in measures that delay the onset of or prevent dementia. A healthy lifestyle is important for major chronic diseases, but its role in cognitive ageing and dementia remains unclear. The long preclinical phase of dementia, involving changes in various disease biomarkers over several years, highlights the importance of longitudinal studies to determine how health behaviours shape the risk of cognitive decline and dementia.
Analysing the role of health behaviours in ageing outcomes, such as dementia, is not straightforward: biases that arise from behaviours being self-reported and residual confounding can affect findings. However, the strength of longitudinal studies is their ability to assess critical windows of exposure and threshold of risk, and to carry out careful analysis, which takes selection (individuals with unhealthy behaviours are more likely to die before they can be included in studies which begin in old age) and reverse causation biases into account.
Using data from the Whitehall II cohort study that assessed health behaviours starting from age 35, we have shown that previous studies based on older people might have underestimated the role of smoking on cognitive ageing because of selection bias. And in a recent paper, we showed that reverse causation is a likely explanation of the protective effect of physical activity for dementia observed in previous studies based on older adults with a short follow-up.
Our new analysis of alcohol consumption was motivated by its absence from risk factors included in the most recent guidelines for dementia prevention. The effects of alcohol consumption on health are complex. The association with cognitive health is thought to be J-shaped, with both abstinence and high alcohol consumption associated with adverse cognitive outcomes. J-shape associations are counterintuitive in that the dose-response effect is part of the criteria used for causal inference in epidemiological studies. However, such associations exist for other exposures, such as body mass index, with detrimental health effects observed for both underweight and obesity, although the underlying biological mechanisms are likely to differ at the two ends of the distribution.
Most previous studies on alcohol and dementia have been in the elderly population. As people tend to drink less as they get older, alcohol consumption in these studies might not reflect previous behaviour. Furthermore, one-off measures to identify abstainers and heavy drinkers are prone to misclassification. This prompted us to use data from several time points, starting in midlife, to identify regular heavy drinkers and to differentiate long term abstainers from former and occasional drinkers. We found similar excess risks among long term abstainers, former drinkers, and occasional drinkers. Interestingly, comparable findings exist in relation to cardiometabolic outcomes. We thus examined the role of cardiometabolic health in the association between alcohol consumption and dementia. We found that some of the excess risk in abstainers was partly explained by cardiometabolic health, supporting the biological plausibility of the association. However, a previous report from the Whitehall II MRI substudy on 527 participants found a linear rather than a J-shaped association between alcohol consumption and hippocampal atrophy.
Taken together, it remains unclear if the inconsistency in results is due to the selected nature of the MRI substudy, compensatory mechanisms in moderate drinkers that protect them from the clinical expression of brain damage, or a real benign effect of moderate alcohol consumption. Future research, which does not rely on reported alcohol consumption, is needed to provide insight into the shape of the association between alcohol consumption and cognitive dysfunction.
Regarding the role of heavy alcohol consumption in dementia risk, we identified a threshold of 14 units per week above which the risk of dementia increases in a linear fashion. The harmful effect of high alcohol consumption was confirmed in analysis of the CAGE alcohol dependence scale and history of alcohol-related chronic disease hospitalization. Taken together, results from this paper and previous studies highlight the detrimental effect of heavy alcohol consumption for dementia. The definition of harmful alcohol consumption in several countries continues to be much higher than 14 units per week; our findings would encourage the downward revision of such guidelines in order to promote the cognitive health of future generations.
Given the long preclinical period of dementia and the continuing absence of curative solutions, it is imperative that we better understand the role of modifiable risk factors. For health behaviours, we need to pay particular attention to the age of exposure, the threshold of risk, and the impact of behavioural changes over time. This will help build an evidence base that will inform future interventions and prevention policies.
Séverine Sabia is an epidemiologist based at the National Institute of Health and Medical Research (Inserm) in France and associate researcher at University College London.
Competing interests: None declared.