We calculated 1–5 years’ individual time-weighted cumulative exposure to PM2.5 and PM10 based on daily estimation of pollutants from LUR model, geocoded information of both residential and work addresses, and characteristics of personal activity pattern. PM2.5 and PM10 were identified as risk factors for carotid atherosclerosis. The maximum 32.2% and 21.3% increased risk of carotid atherosclerosis were found with two-year exposure to PM2.5 and PM10, respectively. The consistent results were obtained treating PM as categorical variables. The linear relationship between PM2.5 exposure and the risk of carotid atherosclerosis was observed. However, such significant linear or nonlinear relationship was not shown with one, four or five-year exposure to PM10. Robust results were demonstrated by a variety of sensitivity analyses. The overall adverse effect of time-weighted exposure to mixed ambient pollutants on the risk of carotid atherosclerosis was proved, and PM2.5 accounted for a large part of the whole positive effect. Our results suggest that higher levels of PM2.5 and PM10 exposure would increase the risk of carotid atherosclerosis. The study specifically demonstrated the need to reduce individual exposure to ambient pollutants.
The detailed mechanisms of PM-induced atherosclerotic lesion were not clearly and fully illustrated, but several processes would potentially describe the progression according to some mechanistic researches and experiments. Exposure to PM2.5 and PM10 would promote the release of inflammatory cytokines and reactive oxygen species, induce peroxidative stress and inflammatory process, and further cause endothelial malfunction and vascular remodeling [47, 48]. Tonne had investigated the association between oxidative potential of PM10 and CIMT, but it was not statistically significant [49]. More original studies should be performed to further investigate the relationship between oxidation and inflammation derived from PM and carotid atherosclerosis. We observed larger effect sizes for PM2.5 than PM10. It can be explained by greater pulmonary deposition and differences in sources and constituents. Ambient PM2.5 primarily emits from fossil combustion, and PM10 consists of higher fractions of particles, such as suspended dust. Larger particles had relatively less infiltration efficiency into indoor environment. Our study also demonstrated the significant association between multiple pollutant exposures and the risk of carotid atherosclerosis based on quantile g-computation. This method was presented to obtain the pooled estimation of multiple exposures and avoid collinearity. It allows distinct effect directions for the exposures on study outcomes, and can compare their relative importance. The essential role of ambient pollutants on carotid atherosclerosis was highlighted, and evidence about the adverse effect of ambient pollutants was enriched.
It was evidenced that increased atherosclerotic burden was observed in areas with higher levels of air pollutants [50]. PM2.5 estimated by AOD adjusting for visibility and relative humidity was proved to be associated with atherosclerosis among healthy Chinese Han population [51]. The significant association of atherosclerosis with interpolated PM2.5 was demonstrated among treatment groups of four clinical trials, but not among the whole population or other subgroups [33]. Study derived from Whitehall II study reported the significant association between the 52 weeks’ average exposure to PM10 estimated by a geostatistical spatiotemporal model with the percent change of CIMT before ultrasound scan [49] without adjustment for important blood biochemical indicators. Study among randomized participants of two rigorous clinical trials also found the significantly higher percent change of CIMT with PM2.5 exposure estimated by interpolation model, but some variables, such as blood lipids, were not controlled [52]. Long-term association between PM2.5 and PM10 derived from gaussian dispersion model and the prevalence of carotid plaques was significant for an exposed time of lag 15–18 years at three-year follow-up [23]. Among healthy participants of Multicultural Community Health Assessment Trial (M-CHAT), the association between PM2.5 predicted based on LUR model and the progression of carotid artery atherosclerosis was not significant [27]. A study of midlife population from a prospective cohort found the significant association between estimated individual PM2.5 exposure (industry and household emissions) and carotid atherosclerosis, but such relationship was not shown with PM2.5 and PM10 derived from traffic and heating [24]. Once in a study, several well-performed cohorts were combined and the overall effect was pooled [53]. But the significant associations were not observed. Those results were not totally consistent. Although our study identified ambient PM as risk factor for carotid atherosclerosis, potential heterogeneity existed, which had an impact on the results.
Characteristics of the study population is the main source of heterogeneity. BHMC study includes most of individuals taking health examination every year in Beijing. Participants in the same examination group hold the similar socioeconomical characteristics and activity patterns at work. We controlled important metabolic and behavior indicators to explore the independent effect of PM. Our results about the association between PM2.5 and PM10 and carotid atherosclerosis are convincing.
Accurate estimation of individual exposure to pollutants would help quantify the actual relationship between PM and carotid atherosclerosis. Compared with previous studies [14, 19, 23], we referred to critical considerations of both the primary predictive model and personal activity patterns. LUR model would reflect the characteristics of long-term pollutant distribution and put influencing factors of pollutants into consideration [54]. It represents improvement over studies that estimated pollutant exposures based on monitoring stations or interpolation. We used both residential and work addresses. And detailed personal activity mode was also considered, including commuting mode and indoor and outdoor activities at residential and work addresses and on commuting. The duration, location, frequency and type of physical activity were also considered. The method can better explain individual activity behaviors and characteristics of personal exposures [22, 28]. Additionally, cumulative exposure can reflect the overall cumulated effect of pollutant exposure compared with average level. In our analysis, the highest risk of carotid atherosclerosis was found with time-weighted two-year exposure to PM2.5 and PM10. Cumulated effects of PM should be further investigated in larger cohort studies.
Our results demonstrated that women and those with higher level of TG, LDL-C, BMI, UA and HDL-C had higher risk of carotid atherosclerosis when exposing to PM10. While men and individuals with lower level of TG were elusively sensitive to PM2.5 with higher risk of carotid atherosclerosis. All subgroups tested were identified a priori. It was once reported that a greater susceptibility for atherosclerotic lesion of ambient PM were interpreted among individuals with dyslipidemia [14], overweight [16] and hyperuricemia. One possible explanation for this finding is the weaker effects of PM compared with cardiometabolic risk factors on carotid atherosclerosis. Among individuals with abnormal level of HDL-C, BMI and TG, effect of PM on carotid atherosclerosis was inhibited. Pre-existing health conditions would also have an impact. Additionally, such results should be explained with caution, because they were kinds of secondary analysis in a smaller group, and the differences were likely due to chance. The observed vulnerability among men may be described as more time spent outdoors, larger airways and enhanced deposition of PM.
Our study has several advantages. First, individual exposure to PM was calculated based on prediction of ambient pollutants by LUR model at both residential and work addresses and personal activity patterns, which can better represent the actual personal exposure. Ambient and indoor PM are highly correlated, but infiltrated ambient PM should not be ignored. Our results may owe in part to individual time-weighted estimation of PM. Second, quantile g-computation was applied to control collinearity of the pollutants, and the overall adverse effects of individual exposure to multiple pollutants were obtained in our study. Double-pollutant model was also performed, and the robust results were not wholly obtained. It can be illustrated that the pollutants would mutually influence each other. And the interaction should not be ignored. Our study enriched the evidences of the pooled estimation about the total effects of multiple pollutant exposures, which was less explored before. High collinearity was avoided, and the results were convincing.
Some limitations would affect our findings. Household PM is also an important source of individual pollution, such as fuel use and tobacco smoking, but it was not considered in our study. Infiltration factors of different vehicles for PM10 were not clearly investigated. We applied infiltration factor of bus for PM2.5 in this study, which would induce misclassification. We collected individual activity information only at the baseline questionnaire. But it is less likely that participants in the same examination group would change their work or move frequently. Thus, our results are reliable. Additionally, small samples in subgroup analysis and participants enrolled only in Beijing inhibit the generalization. Further researches are required to explore the relationship between accurate estimation of time-weighted individual exposure to PM and the prevalence and progression of carotid atherosclerosis among larger sample of population in multicenter studies.