Associations Between Elemental Constituents of Fine Particulate Matter and Subclinical Atherosclerosis in Adolescents and Young Adults


 BackgroundExisting studies have demonstrated the relationships between particulate matter (PM) exposure and subclinical atherosclerosis; however, whether PM and its elemental constituents predispose to atherosclerosis remains unclear in adolescents and young adults. This cross-sectional study included 789 subjects between the ages of 12 to 30 years who lived in Taipei metropolis since childhood. Health examination and carotid intima-media thickness (CIMT) measurements were performed between 2006 and 2008. Land use regression (LUR) model was used to estimate participants’ one-year exposure to fine particulate matter (PM2.5) and eight elemental constituents, i.e., silicon (Si), sulfur (S), titanium (Ti), manganese (Mn), iron (Fe), nickel (Ni), copper (Cu), and zinc (Zn). The associations between percent differences in CIMT at common carotid artery (CCA) segments and air pollutants were analyzed.ResultsAn interquartile range increment of PM2.5 (4.5 μg/m3), Fe (34.7 ng/m3), and Zn (20.7 ng/m3) are associated with 0.77% (95% confidence interval; 95%CI: 0.05 to 1.50), 0.83% (0.01 to 1.65), and 1.22% (0.35 to 2.10) higher for combined CIMT, respectively; while Mn (2.0 ng/m3) exposure is associated with 0.31% (0.01 to 0.60) higher for right CIMT. Stratified analyses show PM2.5 and elemental constituents, especially Zn, are associated with CIMT among subjects who are 18 years or older, females, lower household income, non-smokers, normal weight, non-hypertensive, non-hyperglycemic, or non-hypercholesterolemic. ConclusionsLong-term exposures to PM2.5 and elemental constituents mainly originating from traffic and industry operations are associated with subclinical atherosclerosis in young population. Individual characteristics, health behaviors, and biometric measures, may modify air pollution-related subclinical atherosclerosis.

or CIMT [16,17], except higher augmentation index observed with exposed to PM 2.5 among young females or non-smokers by subgroup analyses [17]. An epidemiological study conducted in Netherland children demonstrated PM 2.5 , NO 2 , and nitrogen oxides (NO X ) exposures are associated with decreased carotid artery distensibility, but not CIMT [18]. Children residing <100 meters from heavily tra cked road were reported to have higher CIMT measurements compared with those who living ≥ 200 meters away, suggesting urban tra c exposures promotes atherosclerotic process in children [19]. The heterogeneity as abovementioned may indicate more surveys on concentrations and source-speci c compositions of air pollutants and the mediation of individual demographics are necessary while evaluating atherosclerotic effects of air pollution in young population. Therefore, we designed a cross-sectional study consisted of adolescents and young adults living in Taipei metropolis and applied land use regression model (LUR) to estimate individual exposures to ne particulate matter (PM 2.5 ) and eight elemental constituents. The associations of one-year exposures to PM 2.5 and elemental constituents with CIMT values, and the potential mediation of individual characteristics were also examined.

Study Subjects
In this cross-sectional study, we included 789 subjects who aged 12-30 years and lived in Taipei metropolis. The study subjects were selected from the YOung TAiwanese (YOTA) Cohort Study, a nationwide urine screening program which was conducted among 103,756 school-age children between the ages of 6 to 18 from 1992 to 2000. Detailed information on this program is provided in previous study [20]. After excluding 38,118 subjects with unreliable or missing data and 96,659 subjects who did not live in Taipei metropolis, 7160 subjects were invited for the follow-up health examination via telephone or mail during 2006 2008. A total of 789 subjects completed the follow-up health examination. The participants in this study contained higher proportion of females, the majority of whom were older in age and possess higher systolic and diastolic blood pressures (BP), serum cholesterol levels compared to the 6,371 subjects who were lost to follow-up (Supplemental Table 1, see additional le 1). All of the participants and their parents signed informed consent documents upon enrollment in the study. The study was approved by the Ethics Committee of National Taiwan University Hospital (NTUH).

Health Data
The follow-up health examination was conducted in NTUH from 2006 to 2008 and consisted of a clinician interview, a structured and self-reported questionnaire, venous blood biochemical analysis, as well as BP and CIMT measurements. The clinician interview and questionnaire provided information of individual characteristics, such as age, sex, household income [New Taiwan Dollar (NTD) per month], smoking, and alcohol consumption. Body mass index (BMI) was calculated as body weight (kg) divided by the square of body height (in meters). Here, overweight is de ned by BMI ≥25 kg/m 2 .

Page 4/19
The venous blood sample was collected from an antecubital vein after a fasting period of 10-14 hours.
The blood glucose, serum total cholesterol, triglyceride, high-and low-density lipoprotein cholesterol were analyzed using an auto-analyzer (Hitachi 7250 Special; Hitachi, Tokyo, Japan) in central lab at NTUH.
Hyperglycemia is de ned as fasting blood glucose level ≥100 mg/dL, and hypercholesterolemia is de ned as total cholesterol level ≥200 mg/dL. BP values were measured using a mercury sphygmomanometer in a standardized fashion, with the cuffsize adjusted to the circumference of the arm. The mean of two measurements obtained after 5-10 minutes of rest in the seated position with the legs uncrossed in a quiet room was used as the BP measurement. If the difference in the two BP measurements was greater than 10 mm Hg, a third BP measurement was obtained, and the average of the lowest two measured BP values was selected as the subject's BP. Hypertensive status was determined by participant either had self-reported physiciandiagnosed hypertension and used anti-hypertensive medication or had measured BP values ≥140 mmHg for systolic BP or ≥90 mmHg for diastolic BP.
CIMT Measurement CIMT, de ned as the distance from the front edge of the rst echogenic line (lumen-intima interface) to the front edge of the second echogenic line (media-adventitia interface) in the far wall of the vessel, was measured by an experienced technician using high-resolution B-mode ultrasonography (GE Vivid ultrasound system, Horten, Norway) equipped with a 3.5-to 10-MHz real-time B-mode scanner. The values of the CIMT at the common carotid artery (CCA) proximal to the carotid bifurcation were measured bilaterally. All scans were recorded on a digitalized memory system in Digital Imaging and Communications in Medicine (DICOM) format for subsequent o ine analysis. The digitized M-mode was later analyzed off-line using a computer program, in which each image was recalled with magni cation and the CIMT between two successive R waves was measured by automated analyzing software provided by the manufacturer. The details in the protocol for CIMT measurements has been described in previous studies [20]. In this study, we used averaging measurements of CIMT values at left CCA (LCCA), right CCA (RCCA), and means of bilateral CCA (combined CCA) as health outcomes. To ensure the reliability of repeat measurements, a technician conducted a second reading for randomly selected 30 subjects two weeks later. The reliability of CCA measurement had excellent intra-observer correlation coe cients of 98.8% for RCCA, and 98.5% for LCCA [21].
Air Pollution Exposure Assessment.
We used land use regression (LUR) model developed by Ho et al. (2015) to estimate individual's annual average exposure concentrations of PM 2.5 and eight elemental constituents, i.e., silicon (Si), sulfur (S), titanium (Ti), manganese (Mn), iron (Fe), nickel (Ni), copper (Cu), and zinc (Zn) [22]. This modeling approach was derived from projects of ESCAPE (European Study of Cohorts for Air Pollution Effects) [23]. In brief, PM 2.5 samples were collected at 20 low-level sites ( rst to third oors), ve mid-level sites (fourth to sixth oors), and ve high-level sites (seventh to ninth oors) from Taipei metropolis, Taiwan, between January and October of 2010. Each PM 2.5 sampling site was measured twice with a ve-month interval, and each measurement was collected for two-week period with a Harvard impactor (Air Diagnostics and Engineering Inc., Harrison, ME, USA). The eight elemental constituents were identi ed and quantitatively determined using energy-dispersive X-ray uorescence spectrometry at the National Taiwan University [24].
ArcGIS (version 10.1; ESRI) was used to obtain geographic information system information for land use data (residential land, industry, port, urban green, and natural space) and tra c information (total length of major roads and road segments, the distance to the nearest major road and the nearest road).
Predictor variables with multiple buffer sizes (100, 300, 500, 1000, and 5000 m for land use data; 25, 50, 100, 300, 500, and 1000 m for tra c data) were applied to estimate the in uence of spatial variability on PM 2.5 elemental constituent exposures. Supervised forward stepwise multiple regression to derive the nal LUR models. We summarized the equations and parameters in Supplemental Table 2 (Additional le 2). Potassium was excluded from this study because its associated predictor variables all presented nonsigni cant effects in the LUR model. Individual's annual exposure estimates which vary by more than three standard deviations (SD) were removed as outliers.

Statistical analyses.
Multiple linear regressions were applied to assess the associations of LCCA, RCCA, and combined CCA with an interquartile range (IQR) increase in annual averages of PM 2.5 and eight elemental constituents.
The following covariates were adjusted in the main model: age, sex, household income, smoking status, BMI, systolic BP, fasting blood glucose, and total cholesterol. In the extended model, we further adjusted for individual's urinary cotinine levels in addition to the selected covariates in the main model to consider the possible effect of environmental smoke. Strati ed analyses were performed to examine whether associations between CIMT values and air pollutants are modi ed by individual characteristics, including age (<18 years vs. ≥18 years), sex, smoking status (non-smoking vs. current smoking), household income (<NTD 50,000/month vs. ≥ NTD 50,000/month), overweightness (BMI <25 kg/m 2 vs. BMI ≥25 kg/m 2 ), hypertension, hyperglycemia (fasting glucose <100 mg/dL vs. fasting glucose ≥100 mg/dL), and hypercholesterolemia (total cholesterol <200 mg/dL vs. total cholesterol ≥200 mg/dL). The estimates are presented by percent differences and 95% con dence intervals (CIs) of CIMT values for an IQR increment in each of air pollutant. The analyses were performed using SAS software (version 9.1.3; SAS Institute Inc., Cary, NC).

Results
The average age of study participants in the follow-up study was 21.3±3.3 years, and females account for 60.3% of all 789 individuals. The means (SD) of CIMT values at LCCA, RCCA, and combined CCA are all 0.45±0.06mm. Table 1 details the distribution of study subjects and CIMT measurements strati ed by individual characteristics. The CIMT values at LCCA RCCA, and combined CCA are higher in males, current smokers, or subjects who are overweight, hypertensive, hyperglycemic, or hypercholesterolemic. The CIMT values are not different between age and household income stratum. Table 2 shows the one-year average exposure concentrations of PM 2.5 and eight elemental constituents for study subjects. After removing 10 subjects whose estimated annual average concentrations of PM 2.5 exceed more than three SD, the mean (SD) values of annual average concentrations of PM 2.  Table 3, see additional le 3). Table 3

Discussion
This is the rst study to demonstrate chronic exposures to PM 2.5 and transition metal components, including Mn, Fe, and Zn, are associated with subclinical atherosclerosis in young population. Although previous studies reported on the associations between tra c proximity, O 3 , or NO X exposures and CIMT in young population, they failed to demonstrate PM are associated with CIMT [16,17]. There are several strengths in ours study ndings. First, our exposure assessment of air pollutants and elemental constituents were conducted with LUR model which accounted for the effects of vertical distribution for high-rise buildings in urban area, thus minimizing exposure measurement error. Second, our study design largely ruled out the exposure misclassi cation from relocation, since our participants were invited for Though several studies have demonstrated associations between PM and CIMT in middle-or old-age population [3,4,13,14,25], our study rst demonstrates the positive associations between PM 2.5 and CIMT in young population. The individuals in this study exposed high annual exposure concentrations of PM 2.5 (24.9 μg/m 3 ), which is almost twice as high compared to those reported in published studies in western countries, may contribute to this particular nding. Nevertheless, the point estimate of percent difference in CIMT with PM 2.5 exposure (0.77% with exposed to 4.5 μg/m 3 of PM 2.5 ) in this study is lower than a previous meta-analysis that showed a 5 μg/m 3 increment of long-term exposure to PM 2.5 is associated with 1.66% change in CIMT [26]. The heterogenous ndings between differernt age group may be attributable to that CIMT is chronic process in structural change and take longer time to demonstrate the measureable differences. Subroup analysis of this stuty showed signi cant associations between CIMT and PM 2.5 only among young adults but not adolescents also support abovementioned hypothesis. In addition to PM 2.5 , we further demonstrate long-term exposure to certain transition metals of PM 2.5 , namely Mn, Fe, and Zn, are associated with higher CIMT values. Epidemiological studies have demonstrated speci c human activities, such as residing tra c proximity and cooking fuels, are associated with increased CIMT measures [11,12,27,28]. Certain source-speci c components of PM, including organic carbon, elemental carbon, black carbon, and S were reported to associate with increased CIMT values in elderly population [12,[28][29][30][31]. PM 2.5absorbance exposure was associated to decreased carotid artery distensibility in young children [18]. Several studies also demonstrated exposures to elemental components of PM 2.5 contribute to other adverse cardiovascular effects related to atherosclerosis. Bilenko et al. (2015) demonstrated that PM 10 elements, including Fe, K, and Si associated with higher diastolic BP values in children [32]. PM 2.5 metal composition, including Ni, Fe, and vanadium, were also reported to associate with higher BP values in young adults or elderly persons [33,34]. Long-term exposures to Cu, Fe, and Zn, the transition metals of PM, may be associated with higher in ammatory biomarkers [35]. A panel study of 17 mail carriers showed that metal compositions of PM 2.5-1.0 , including sodium, magnesium, calcium, strontium, manganese, and cadmium, signi cantly increased the cardio-ankle vascular index, a surrogate marker of arterial stiffness [36]. These elements are emitted from multiple sources, such as Mn, Cu, Fe, and Zn from brake linings and tires; Si and Ti from road dust suspended by automobiles and wind; Ni and S from industrial or fossil fuel combustion [37,38]. According to the regression coe cients, the strongest in uence in LUR models for Fe and Zn are mainly attributed to tra c-or industry-related covariates, while Zn may also be partially emitted from population variables, such as cooking activities. Mn is primarily generated from the port activities, and partially from tra c, and industry operations. Because ship decommissioning activity does not exist in the buffers around sampling sites, Ho et al. (2015) speculated Mn possibly emitted by abrasive wear activities from ship movement in wharfs of Taipei metropolis [22]. Our study results suggest sourcespeci c PM 2.5 , primary from vehicular or industrial emissions, contribute to extra risk of subclinical atherosclerosis in young population.
The strati ed analyses show that associations of combined CCA IMT values with PM 2.5 , Mn, Fe, or Zn are stronger in subjects who are females or lower household incomes than those with contrary stratum, which may suggest females or subjects with lower socioeconomic (SES) status are more vulnerable to exposure of air pollution, resulting in the acceleration of subclinical atherosclerosis. Existing research support our nding that carotid arterial wall thickness is more pronounced in females exposed to PM 2.5 [8, 25,39]. This observed vulnerability among females could be due to females having smaller airways, resulting in enhanced deposition of ne particles. More frequent exposures to cooking fuel among women may further contribute to the stronger associations of PM 2.5 exposures with Taiwanese females.
The epidemiological study in India showing associations between use of unvented stove and higher CIMT values, especially in women [11], further support our ndings. The mediation effect of SES on relationships between air pollution and cardiovascular health still remains inconclusive [40]; however, several studies reported lower individual or neighborhood SES status may enhance the air pollutionrelated cardiovascular risk. Higher risk estimates of cardiovascular events with exposure to PM 2.5 were observed among participants living in low-SES neighborhoods [41]. Dragano et al. (2009) observed that women in the lowest income stratum have a signi cantly higher level of coronary artery calci cation (CAC) associated with pollution exposure compared to women in the highest income stratum [42]. Diez Roux et al. (2004) reported that low SES status also suffer from worse health outcomes resulting from psychosocial stress, which may mediate the effect of air pollution-related atherosclerosis [43]. The higher air pollution and noise exposures in subjects with lower SES due to their residency proximity to tra c-or industrial-area may also contribute to the mediation between air pollution and subclinical atherosclerosis [42].
Strati ed analyses of this study further demonstrate the vulnerability to air pollution in low cardiovascular risk subjects of non-smoking, normal weight, non-hypertensive, non-hyperglycemic, or nonhypercholesterolemic young population. Some other studies also agree with our ndings. Kauffman et al. (2016) reported that the progression of coronary artery calci cation with exposures to PM 2.5 and NO X might be greater in non-diabetic, non-obese, non-hypercholesterolemic subjects [8]. Epidemiological studies also reported that associations between PM and decreased renal function are stronger in nondiabetic subjects, which may share atherosclerotic change as the common pathophysiologic pathway [44,45]. Our previous study also observed that CIMT values are associated with per uorinated chemicals in healthy young subgroups [21]. One possible explanation to the stronger effect of air pollution-related atherosclerosis in healthier young subjects is that the effect of air pollution on CIMT is weaker than the traditional cardiovascular risk factors such as obesity, smoking, hypertension, hyperglycemia, or hypercholesterolemia, which results in ndings of insigni cant air pollution-related atherosclerosis on subjects with unhealthy lifestyle or comorbidities. In other words, healthy young population must be more aware of air pollution-related atherogenic effect.  Table 2), which could possibly in uence the accuracy of model prediction. Third, the results of strati ed analyses may be biased under multiple comparisons, and insigni cant ndings among comorbid groups may be attributed to small sample size and wide con dence intervals. More studies are necessary to elucidate the population susceptibility of subclinical atherosclerosis to air pollution. Other possible unmeasured confounders are ambient or tra c noise, and endocrine disrupting chemicals such as per uorinated compounds and phthalates, which have been shown to associate with atherosclerosis [47,48] The main models calculated by multiple liner models, adjusted for age, sex, household income, smoking status, body mass index, systolic blood pressure, fasting glucose, and cholesterol. The extended models were further adjusted for urinary cotinine levels in addition to covariates in the main models. Figure 1 Percent changes (95% CIs) of combined CCA values in association with an increment of (A) 4.5 μg/m3 for PM2.5; (B) 2.0 ng/m3 for Mn; (C) 34.7 ng/m3 for Fe; and (D) 20.7 ng/m3 for Zn strati ed by age (<18 vs. ≥18 years), sex, household income (<NTD 50,000/month vs. ≥ NTD 50,000/month), smoking, overweightness (BMI <25 vs. BMI ≥25 kg/m2), hypertension, hyperglycemia (fasting glucose <100 mg/dL vs. ≥100 mg/dL), or hypercholesterolemia (total cholesterol <200 mg/dL vs. ≥200 mg/dL). The estimates were calculated by linear regressions, adjusted for age, sex, household income, smoking, body mass index, and individual comorbid conditions other than analyzed stratum.