Background:
The biological pathways by which long term exposure to air pollution may promote cardiovascular disease (CVD) are not yet fully understood. There are important research gaps in terms of effects of air pollutants on blood pressure, inflammation, atherosclerosis and cardiac hypertrophy. While adverse effects of air pollutants on CVD mortality are firmly established, further work is needed to explain mixed results of studies with recent follow-up periods.

Air pollution contains a mix of gaseous and particulate compounds. From a statistical viewpoint, most studies so far have taken a uni-pollutant approach, looking at the effect of one air pollutant at a time. Traditional frequentist models are not suited for multi-pollutant research because their estimates tend to be unstable due to high correlations between air pollutants and because they do not provide estimation of the combined health effect of multiple air pollutants.

Objectives:
The aim of this PhD was to investigate associations of long term exposure to air pollution with cardiovascular events in the UK, and explore the underlying biological pathways.

Objectives were firstly to investigate if long term exposure to air pollutants is associated with 1) CVD mortality and morbidity longitudinally, 2) systolic blood pressure (SBP), inflammation, atherosclerosis and cardiac hypertrophy cross-sectionally, and secondly to explore use of alternative statistical methods using Bayesian approaches to overcome the limitations of frequentist models in order to perform multi-pollutant research.


Methods:
Two cohort studies were used: Southall And BRent REvisited (SABRE) and the National Survey of Health and Development (NSHD). SABRE is a tri-ethnic cohort of Europeans, South Asians and Afro-Caribbeans made up of 4857 participants from North and West London, recruited in 1989 and followed up in 2008-11. NSHD is a national cohort of 5362 participants born in March 1946 and followed up at multiple occasions, including 1989 and 2006-10, which is broadly representative of the general population but is all European.

Estimates of air pollution concentrations were available at the postcode level from previous studies. Concentration estimates of black smoke (BS), nitrogen dioxide (NO2) and sulfur dioxide (SO2) were available for 1991 and assigned to 1989, and of NO2, nitrogen oxides (NOx), particular matter (PM) of different sizes (PM10, PMcoarse and PM2.5) for 2006-10.

The two cohorts were analysed separately, and were combined at the individual level. CVD mortality and the combined endpoint of fatal and non-fatal CVD events were investigated longitudinally with respect to the air pollutants. Cross-sectional associations, in 1989 and 2006-10, were then investigated between the air pollutants and intermediate markers of CVD. These were SBP (with adjustment for those on anti-hypertensive medication), C-Reactive Protein (CRP) and Interleukin 6 (IL-6) as markers of inflammation, carotid intima-media thickness (CIMT) as marker of atherosclerosis, and relative wall thickness (RWT) and left ventricular mass (LVM) as markers of cardiac hypertrophy. Models were adjusted for age, gender, diabetes status, smoking history, socio-economic status (SES) and ethnicity. Interaction between air pollution and ethnicity was investigated using interaction terms and stratification.

Profile regression (PR) and Bayesian kernel machine regression (BKMR) were used on SBP in 1989 cross-sectional analyses in NSHD to assess the combined effect of NO2, SO2 and BS. The combined effect of these three air pollutants on CVD mortality between 1989 and 2015 in SABRE was estimated using PR.

Results:
For events longitudinally, elevated levels of SO2 and BS increased risk of fatal and non-fatal CVD events in SABRE. Hazard ratios (HR) were 1.10 (95% CI: 0.99 - 1.23, p=0.07) and 1.27 (95% CI: 0.90 - 1.80, p=0.17) per 10 microgram/m3 increase in SO2 and BS respectively. Associations with CVD mortality were small in magnitude and non statistically significant.

Cross-sectionally, SBP was greater in those exposed to higher levels of SO2 and BS. In NSHD, SBP was increased by 2.4 mmHg (95% CI: 1.5 - 3.3, p<0.01) and 4.0 mmHg (95% CI: 2.7 - 5.2, p<0.01) per increase of 10 microgram/m3 in SO2 and BS respectively. Detrimental associations between NOx and PM2.5 and inflammatory markers were found. In the combined analysis, CRP was increased by 0.04 mg/L (95% CI: 0.01 - 0.07, p=0.02) per 10 microgram/m3 increase in NOx and IL-6 by 0.36 mg/L (95% CI: -0.05 - 0.78, p=0.09) per 10 microgram/m3 increase in PM2.5. Some evidence of an association between NO2, NOx and PM2.5 and LVM was observed in NSHD. Associations of air pollutants with SBP and CIMT were stronger in South Asians than in Europeans.

Using PR and BKMR, a simultaneous increase from the first to the fourth quartiles of NO2, SO2 and BS was estimated to increase SBP by approximately 5 mmHg (95% CrI: 3 - 7, P(>0) = 100%) in NSHD in 1989 and to double the odds (95% CrI: 1.5 - 2.7, P(>1) = 100%) of CVD mortality between 1989 and 2015 in SABRE.

Discussion:
This PhD investigated the effect of long term exposure to air pollutants on CVD events and intermediate markers of CVD. There is evidence supporting two potential biological pathways from the results: 1) the inhalation of air pollutants may lead to chronic inflammation, which could contribute to the development and/or progression of atherosclerosis and to left ventricular hypertrophy; 2) long term exposure to air pollutants may raise blood pressure. Further cohort studies are needed to confirm these hypotheses in different populations.

PR and BKMR are among novel Bayesian methods that allow the assessment of the health effect of multiple air pollutants simultaneously. Future researchers are encouraged to employ these new methods, as humans are exposed in real-life to a mixture of pollutants. Moreover, these methods can identify which pollutant(s) has the largest effect among the mixture, which is particularly relevant for future public health policies.