Study Links 76 Environmental and Lifestyle Factors to Dementia Risk in Nearly 500,000 Adults

A broader picture of dementia risk

A study of 499,992 UK Biobank participants has linked 76 environmental, lifestyle and social factors to dementia, drawing together associations involving air pollution, sleep, employment, housing and physical activity. Over a median follow-up of 15.3 years, 8,881 participants developed dementia, including 4,000 who developed Alzheimer's disease.

The researchers examined 193 exposures rather than concentrating on a small set of familiar health habits. Among the 76 factors associated with Alzheimer's disease and related dementias, 45 were linked to higher risk and 31 to lower risk. A separate analysis identified 34 factors associated specifically with Alzheimer's disease.

The team also combined exposure information into an Exposomic Risk Score. Participants in the highest quarter of that score had nearly three times the relative dementia risk of those in the lowest quarter. That finding describes a pattern across multiple factors, not the effect of one food, habit or living arrangement.

The central caution is that association does not establish cause and effect. The study cannot show that eating cheese prevents dementia, renting a home causes it, or changing any single factor would produce the difference observed between groups.

What the researchers measured

The research, led by scientists at Harvard T.H. Chan School of Public Health, Harvard Medical School and collaborating institutions, was published in Alzheimer's & Dementia on September 29. Harvard Medical School introduced the findings on October 8.

Chenshuang Li, a postdoctoral fellow at Harvard Chan School, was the first author. Peng Gao, an assistant professor of environmental health and exposomics, and Harvard Medical School's Chirag Patel were the corresponding authors. Harvard Chan School's study announcement describes the work as an extensive systematic evaluation of environmental exposures and dementia.

The researchers conducted an exposome-wide association study using UK Biobank health, genetic and lifestyle data collected over time. The announcement identifies Patel and Gao as co-senior authors.

The approach examines the exposome, meaning the environmental influences people encounter throughout their lives. Here, the term includes social and behavioral circumstances as well as substances in the air or physical surroundings. The researchers organized the exposures into five domains:

• Chemical exposures, including air pollutants and medication use.
• Physical exposures, such as noise.
• Lifestyle factors, including diet, exercise and sleep.
• Social conditions, including education and economic circumstances.
• Ecosystems, including access to green space.

These categories describe what the study examined. They do not mean that every example within them was associated with dementia.

Economic circumstances stood out

Unemployment was associated with a 72% higher dementia risk than employment, while financial hardship was associated with a 39% higher risk. Lower educational attainment and housing circumstances, including renting rather than owning a home, were also linked to higher risk.

In the other direction, higher household income and living with a spouse or partner were associated with lower risk. The association for living with a partner was a 22% reduction. These are relative comparisons between groups, not percentage point changes in an individual's probability of developing dementia.

The housing result needs particular care. Renting can be connected with income, housing quality and other circumstances that may themselves be relevant to health. The analysis does not establish that ownership protects the brain or that purchasing a home would reduce dementia risk.

The same difficulty applies across the social findings. Education, employment, income and health are connected. Studying them together expands the picture of risk, but does not fully separate the contribution of each one.

Pollution and medications require different interpretations

Greater exposure to nitrogen dioxide was associated with higher dementia risk. Nitrogen dioxide is an air pollutant produced by vehicle exhaust and fuel combustion, making this finding relevant to the environments where people live as well as their personal behavior.

Taking more medications was also associated with higher risk, but the researchers cautioned against interpreting that as a direct effect of the drugs. People taking several medicines may have multiple chronic conditions. Their medication count can therefore reflect underlying illness and more complex health needs.

These findings illustrate why apparently similar statistical results can call for different explanations. Pollution is an environmental exposure, while medication use may be a marker of existing disease. The study does not justify stopping prescribed treatment to reduce dementia risk.

Sleep and activity showed contrasting patterns

Current smoking, inappropriate alcohol use, frequent daytime dozing and unusually short or long sleep were associated with higher dementia risk. The sleep findings included durations outside roughly seven to eight hours a night.

Physical activities such as walking for pleasure, other exercise, gardening and demanding home or yard work were associated with lower risk. The examples extend beyond formal workouts to movement that can be part of everyday life.

Sleep and activity are also areas where the direction of an association can be difficult to determine. Early cognitive changes may affect how much someone sleeps, whether they doze during the day and how physically active they are, before dementia is diagnosed.

That possibility is called reverse causation: an emerging illness changes the measured behavior, rather than the behavior being the cause of the illness. The researchers identified it as one reason further work is needed before the associations can be treated as prevention targets.

Food and device findings are not prescriptions

More frequent consumption of cheese and poultry was associated with lower dementia risk, while never eating eggs was associated with higher risk. Longer computer use and more frequent mobile phone use were also linked to lower risk.

These results are attention grabbing, but they do not demonstrate that adding cheese to a diet or spending more time on a phone protects against dementia. Food choices and technology use can be connected with education, income, other dietary habits and a person's existing health.

This problem is known as confounding. A separate factor can influence both the exposure being measured and the likelihood of disease, making an association difficult to interpret. For example, differences in socioeconomic circumstances could help explain a connection involving device use without the device itself providing protection.

The study therefore offers leads for further investigation, not a shopping list or a screen time recommendation. Its broader contribution is to place these everyday behaviors alongside social conditions and environmental exposures in the same analysis.

What the combined risk score shows

The Exposomic Risk Score brought together lifestyle, economic and environmental information to distinguish groups with different levels of dementia risk. Compared with the lowest quarter, risk was 1.43 times as high in the second quarter, 1.83 times as high in the third and 2.97 times as high in the highest quarter.

Expressed as relative increases, those comparisons are 43%, 83% and 197%. The progression suggests that the combined exposure profile contained useful information about differences in risk across the study population.

It does not establish an individual's absolute probability of developing dementia. Across the entire cohort, 8,881 diagnoses among 499,992 participants represent about 1.8% of participants during follow-up. That simple proportion is not the risk in any particular score group and should not be multiplied by the score ratios to estimate a person's chances.

The distinction matters because a relative comparison and an absolute probability answer different questions. Nearly three times the risk does not mean dementia was inevitable in the highest score group, nor does it identify one exposure responsible for the difference.

Prediction improved, but clinical use is not established

Adding environmental and lifestyle exposures improved a prediction model based on demographic information, including age and sex, and APOE genotype. APOE is a gene whose variants can influence susceptibility to Alzheimer's disease. Genetic susceptibility is not the same as a diagnosis or certainty that disease will develop.

The model was developed and tested using separate portions of the UK Biobank dataset. That is a useful check within the study, but both portions came from the same underlying population.

The researchers said the score still needs validation in other, more diverse populations before it can be used in clinical practice. The findings establish that exposure information improved prediction in this analysis, not that the score is ready for routine medical decisions.

That limitation also affects application outside the United Kingdom, including in Korea. Differences in population characteristics, exposure patterns and social circumstances mean that the same score cannot be assumed to perform equally well elsewhere.

What the findings support, and what remains uncertain

Most exposure information came from the beginning of the study. Changes in employment, housing, sleep, diet or pollution exposure over the following years may therefore not have been captured. Unmeasured factors and early undiagnosed illness could also have influenced the associations.

A separate Lancet commission estimate suggests that potentially modifiable factors may account for up to 45% of dementia risk. That estimate is not a result of this study and does not mean that 45% of its participants' diagnoses could have been prevented by changing the 76 associated exposures.

The Harvard study's contribution is more specific: it shows that behavior, social conditions, environmental exposures and genetic information can be examined together to improve understanding and prediction of dementia risk. It also supports research into interventions that address several exposures, rather than assuming a single change will reproduce the combined score's results.

In Harvard Chan School's announcement, the study's coauthors argued that prevention strategies targeting multiple modifiable exposures could help reduce dementia's burden. They described that effort in these terms:

which may prove essential for reducing the growing burden of dementia.

The Harvard Medical School announcement, adapted from a Harvard Chan School news story, gives the complete prevention statement:

The findings could support the development of multi-component intervention strategies targeting modifiable exposures for ADRD prevention — an effort, the authors wrote, “which may prove essential for reducing the growing burden of dementia.”

The reported publication sequence is short:

• September 29: The study was published in Alzheimer's & Dementia.
• October 8: Harvard Medical School introduced the findings.

The next scientific step is validation in other populations and further investigation of which associations reflect causes that can be changed. No date for completing that validation was announced.

Key Points

• The study examined 193 exposures in 499,992 UK Biobank participants over a median of 15.3 years.
• Researchers identified 76 factors associated with dementia and 34 associated specifically with Alzheimer's disease.
• Employment, finances, housing, pollution, sleep, physical activity and diet were among the associated factors.
• The highest quarter of the combined exposure score had 2.97 times the relative risk of the lowest quarter.
• The findings show associations, not proof that individual exposures cause or prevent dementia.
• The prediction model needs validation in other populations before clinical use.