Study Participants and Design
From November 2016 to January 2018, we conducted a longitudinal study in order to investigate the health impacts of air pollution and its various chemical constituents on adults in Beijing. Based on the spatial distribution of the annual average PM2.5 levels in 2015, five communities were selected as sampling sites to represent different pollutant levels in Beijing. Two samplers were set up at each sampling site [28]. We included participants who have lived in those communities for more than five years and will still be there in the next few years. Those who were unable to accomplish the follow-up studies were excluded. Participants who had severe cardiovascular diseases (stroke, congestive heart failure and myocardial infarction) and cancers were also excluded. Four repeated measurements were conducted. Visit 1 was conducted from November 2016 to December 2016, then followed by 3 follow-ups: visit 2 in May 2017, visit 3 in November 2017, and visit 4 in January 2018. Health assessment questionnaires, physical examination and biological sample collection were performed during each visit. We included 98 participants who met the prespecified criteria. Among them, 97 completed all 4 visits, and 1 participant completed 3 visits. Thus, data from a total of 391 person-times was compiled and applied in this analysis. The ethics was approved by Institutional Review Board of Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences (IBMS, CAMS). Each participant completed a written informed consent.
Ambient PM2.5 sampling in the communities
During the field survey, we measured concentrations of PM2.5, the US Environmental Protection Agency’s (USEPA) 16 priority PAHs, organic carbon (OC) and element carbon (EC) at fixed samplers in each community. A total of ten fixed samplers were set up in five communities, including Fangshan, Dongcheng, Chaoyang, Liu Hegou and Qian Nantai, with no major roads or factories within a radius of 400, 150, 150, 150 and 160 meters, respectively. Medium flow samplers (TH-150C, Wuhan Tianhong, China) were used to measure PM2.5 mass concentration. Each sampler was equipped with quartz-fiber filters to determine the inorganic and organic constituents. Five daily successive samples were obtained prior to each visit.
Laboratory analysis for PAHs and other constituents
The PM2.5 mass concentrations were determined following the standard operation procedures. Further details can be found elsewhere [28]. Quartz filters were analyzed by a thermal-optical carbon analyzer (Model 2001A, Atmoslytic Inc., USA) to determine organic (OC) and elemental carbon (EC) [28].
In order to measure PAHs concentration, 1/2 of each filter was used for analysis. NAP-D8, ACP-D10, PHE-D10, CHR-D12 and BghiP-D12 (AccuStandard Inc.) were used as recovery rate indicator. The samples of filter aliquots were extracted by ultrasonication for 60 min, with 90 ml of dichloromethane and acetone (1:1) solvent mixture. The solvent extracts were concentrated under reduced pressure on a rotary evaporator. The solvent was replaced by adding 10 ml n-hexane and continuing the rotary evaporation. Purification was performed using an aluminum peroxide/silica gel (1:2) column. PAHs were extracted with 70 ml dichloromethane/hexane (3:7) solvent mixture. The extract was rotary evaporated to 1 ml and then blown with nitrogen until dryness. Internal standards (m-terphenyl-d14) was added prior to Gas chromatography-mass spectrometry (GC-MS) analysis (Agilent 7890A GC system, USA). Selected Ion Monitoring (SIM) was employed to quantify the 16 PAHs designated as high priority pollutants by the USEPA (naphthalene (NAP), acenaphthylene (ACY), acenaphthene (ACP), fluorene (FLU), phenanthrene (PHE), anthracene (ANT), fluoranthene (FLT), pyrene (PYR), benzo(a)anthracene (BaA), chrysene (CHR), dibenz(a,h)anthracene (DahA), benzo(b)fluoranthene (BbF), benzo(k)fluoranthene (BkF), benzo(a)pyrene (BaP), indeno(1,2,3-cd)pyrene (IcdP) and benzo(g,h,i)perylene (BghiP)) [29]. The chromatographic condition was as follows: DB-5MS capillary column (60m×250μm×0.25μm) was used. The inlet temperature was 280 °C. Helium was used as the carrier gas with a flow rate of 1.5 mL/min. Split-flow sample injection was used with a split ratio was 4.5:1.
Given the cumulative lag effects of air pollutants [30], 1- to 5-days moving average (MA) concentrations of PAHs and other pollutants were examined. Specifically, 1-day MA was defined as the pollutant concentrations from 8 a.m. the day before clinic visit to 8 a.m. the day of the clinic visit, and 2-days MA was defined as the average pollutant concentrations from 8 a.m. two days before clinic visit to the day of clinic visit, etc.
Quality control and quality assurance
Recoveries were calculated by adding known concentrations of recovery indicators to each sample, where NAP-D8 was the recovery indicator for NAP; ACP-D10 as the recovery indicator for ACY, ACP, and FLU; PHE-D10 as the recovery indicator for PHE and ANT; CHR-D12 as the recovery indicator for FLT, PYR, BaA, and CHR; BghiP-D12 as the recovery indicator for BbF, BkF, BaP, DahA, IcdP and BghiP. Blank values were deducted from the experimental data and the actual concentrations were recovery-corrected. The method limit of detection ranged from 0.4 to 0.9 ng/m3 and the instrument detection limit ranged from 0.05 to 0.94 μg/L. The concentration of PAHs below the detection limits was calculated as 1/2 of the detection limits. The average recoveries for each indicator ranged from 63.4% to 91.4%.
PAHs grouping
The sixteen PAHs measured in the current study were divided into four categories according to their molecular weight or carcinogenicity: (1) low-molecular-weight PAHs (LMW-PAHs), including those with less than four rings, i.e., NAP, ACY, ACP, FLU, PHE and ANT; (2) high-molecular-weight PAHs (HMW-PAHs), including those with four or more rings [17]; (3) carcinogenic PAHs (c-PAHs), including BaA, CHR, BbF, BkF, BaP, IcdP, DahA and BghiP; (4) non-carcinogenic (nc-PAHs), including the remaining eight PAHs [29, 31, 32]. SPAHs represents the total concentration of 16 PAHs.
Meteorological measurements
We obtained hourly temperature and relative humidity from the China Meteorological Administration for the entire study period. Daily averages of those meteorological parameters were calculated. 5-days moving average of temperature and relative humidity were calculated for further analysis.
Questionnaire, physical examinations and biomarker measurements
Our study used a standardized field survey protocol, with a brief description as follows: Standardized questionnaires were designed to collect the following information: age, sex, education levels, hypertension, diabetes, alcohol consumption, smoking, and indoor smoking status; Physical examination was carried out for each resident, including the measurement of sitting blood pressure, height and weight, body mass index (BMI) was also estimated by the following equation: weight (kg) ÷ height2 (m2); Fasting venous blood of residents was collected at 8-9 a.m., and then blood glucose, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), total cholesterol (TC) and triglycerides (TG) were measured in the Department of Clinical Laboratory within the Peking Union Medical College Hospital.
According to the information provided by questionnaire, physical examinations and biomarker measurements, we redefined diabetes and hypertension patients. Participants who reported physician-diagnosed diabetes or fasting blood glucose ≥ 7.0 mmol/L were defined as diabetes. Participants who reported physician-diagnosed hypertension or SBP ≥ 140 mmHg and/or DBP ≥ 90 mmHg were defined as hypertension.
Additionally, in order to comprehensively evaluate the changes in lipid profiles and their potential effects, several lipoprotein ratios or “atherogenic indices” were calculated [33], which better predict cardiovascular diseases than the isolated lipid parameters [34-36]. To be specific, the following equations were used:
- Castelli risk indexes-I (CRI-I) = TC ¸ HDL-C
- Castelli risk indexes-II (CRI-II) = LDL-C ¸ HDL-C
- Non-HDL cholesterol (NHC) = TC - HDL-C
- Atherogenic coefficient (AC) = (TC - HDL-C) ¸ HDL-C
Statistical analysis
Ambient pollutants concentrations and population health data were merged according to the date of clinic visit. Hematological variables with heavily right-skewed distributions were log-transformed for further analyses. Spearman correlation coefficients between multiple categories of PAHs and the other pollutants were calculated.
Three linear mixed-effects models were fitted to evaluate the associations between multiple categories of PAHs and lipid profiles. In single-pollutant model (model 1), each category of PAHs with a specific days MA was incorporated as the fixed-effect term, and an unique identification (ID) for each participate was incorporated as the random-effect term. Several potential confounding factors were also included in this model: (1) individual characteristics, including age and BMI as continuous variables; sex, education, smoking and alcohol consumption as categorical variables; (2) meteorological factors were incorporated using natural splines with three degrees, including 5-days MA of mean temperature and relative humidity; (3) day of the week. We also fitted two-pollutant models (model 2) to control any potential confounding caused by other pollutants, each of which included adjustment for PM2.5, OC, and EC on the same specific days MA with PAHs, respectively. In order to control the collinearity among different pollutants, constituent-residual models (model 3) were fitted. Residuals were obtained in a linear model containing both PM2.5 and a specific category of PAHs and then added into the main model replacing pollutants.
We performed stratified analysis by cigarette smoking (yes, no), alcohol consumption, age (< 65y, ≥ 65y), BMI (< 25, ≥ 25), diabetes mellitus and hypertension. In addition, several sensitivity analyses were applied to test the robustness of the associations. (1) considering the potential associations between indoor smoking and indoor PAHs exposure [37], participants who were smokers or lived in smoker residency were excluded. (2) participates with diabetes or hypertension were excluded. (3) FLU was included in c-PAHs because of its relatively higher toxic equivalent factor compared to BghiP [38]. (4) the association between multiple categories of PAHs abundances per g of PM2.5 and lipid profiles were evaluated.
All statistical analyses were performed with R software, version 3.5.0 (R Foundation for Statistical Computing), using “lme4” and “splines” packages. For the association between PAHs concentrations and lipid profiles, model estimates were calculated as per 10 ng/m3 increase in PAHs exposure. For the association between PAHs abundances and lipid profiles, model estimates were calculated as per 1/1000 increase in PAHs abundances. A p-value less than 0.05 was considered statistically significant.