Back

Mediation of the relationship between air pollution and dementia: A UK Biobank study.

Taylor, K.; Harris, M.; Hui, E. K.; Anderson, E.; Mukadam, N.

2026-02-22 epidemiology
10.64898/2026.02.20.26346698 medRxiv
Show abstract

BackgroundAir pollution is a potentially modifiable risk factor for dementia with a population attributable risk fraction of 3%. Little is known about the causal mechanisms behind the association, so we aimed to investigate this. MethodsData from the UK Biobank were used to investigate the association between six measures of air pollution (NO2, NOx, PM2{middle dot}5-10, PM2{middle dot}5, PM2{middle dot}5 absorbance and PM10) and dementia incidence. Indirect pathways through four mediators (cardiovascular conditions, mental health treatment, insufficient exercise and social isolation) were explored. Logistic regression was used to model the associations between air pollution, mediators and dementia. Casual mediation analysis implemented using the g-formula was used to investigate the joint indirect effect through the mediators. FindingsExposure to the highest quintile of PM2{middle dot}5 (Rte:1{middle dot}14, 95% CI:1{middle dot}06-1{middle dot}23), NOx (Rte:1{middle dot}11, 95% CI:1{middle dot}03-1{middle dot}20) or NO2 (Rte:1{middle dot}08, 95% CI:0{middle dot}99-1{middle dot}16), compared to the lowest quintile, was associated with higher dementia risk. Most of the observed association resulted from the direct effect of air pollution, consisting of pathways not captured through considered mediators. Amongst those in the highest PM2{middle dot}5 quintile, jointly intervening on the four mediators would result in a 1% reduction in risk of dementia (Rpnie:1{middle dot}01, 95% CI: 1{middle dot}01-1{middle dot}02). The randomised pure natural indirect effect was similar for NO2 (Rpnie:1{middle dot}01, 95% CI: 1{middle dot}00-1{middle dot}01) and NOx (Rpnie:1{middle dot}01, 95% CI: 1{middle dot}01-1{middle dot}02). InterpretationMost of the association between dementia and PM2{middle dot}5, NO2 and NOx occurs through the direct effect of air pollution, or other unmeasured mediators, and not pathways through these four mediators. FundingMedical Research Council (Grant MR/W006774/1).

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

1
Environmental Research
46 papers in training set
Top 0.1%
22.8%
2
Environmental Pollution
35 papers in training set
Top 0.2%
10.6%
3
Environment International
42 papers in training set
Top 0.1%
10.2%
4
Environmental Health Perspectives
17 papers in training set
Top 0.1%
6.4%
50% of probability mass above
5
Science of The Total Environment
179 papers in training set
Top 1%
6.4%
6
BMJ Open
554 papers in training set
Top 7%
2.6%
7
Scientific Reports
3102 papers in training set
Top 47%
2.4%
8
Thorax
32 papers in training set
Top 0.4%
1.8%
9
GeoHealth
10 papers in training set
Top 0.3%
1.7%
10
Eurosurveillance
80 papers in training set
Top 0.7%
1.7%
11
International Journal of Epidemiology
74 papers in training set
Top 2%
1.5%
12
PLOS ONE
4510 papers in training set
Top 58%
1.3%
13
The Innovation
12 papers in training set
Top 0.5%
1.3%
14
Clinical Infectious Diseases
231 papers in training set
Top 3%
1.2%
15
BMC Public Health
147 papers in training set
Top 4%
1.2%
16
Nature Communications
4913 papers in training set
Top 56%
1.2%
17
PLOS Global Public Health
293 papers in training set
Top 5%
1.0%
18
ERJ Open Research
44 papers in training set
Top 0.6%
1.0%
19
International Journal of Environmental Research and Public Health
124 papers in training set
Top 6%
1.0%
20
Spatial and Spatio-temporal Epidemiology
10 papers in training set
Top 0.1%
0.9%
21
eLife
5422 papers in training set
Top 53%
0.9%
22
Journal of Alzheimer’s Disease
39 papers in training set
Top 0.9%
0.9%
23
American Journal of Epidemiology
57 papers in training set
Top 1%
0.9%
24
BMC Medicine
163 papers in training set
Top 6%
0.8%
25
JNCI: Journal of the National Cancer Institute
16 papers in training set
Top 0.7%
0.8%
26
Environmental Science & Technology
64 papers in training set
Top 2%
0.8%
27
Indoor Air
10 papers in training set
Top 0.3%
0.8%
28
Brain, Behavior, and Immunity
105 papers in training set
Top 3%
0.5%
29
PLOS Medicine
98 papers in training set
Top 6%
0.5%
30
Journal of Infection
71 papers in training set
Top 4%
0.5%