Supplemental material

Personal exposure to particulate air pollution and vascular damage in peri-urban South India

Otavio T. Ranzani, Carles Milà, Margaux Sanchez, Santhi Bhogadi, Bharati Kulkarni, Kalpana Balakrishnan, Sankar Sambandam, Jordi Sunyer, Julian D Marshall, Sanjay Kinra, Cathryn Tonne

1

Supplementary Information Text

Definitions of cardiometabolic comorbidities

a

Alberti K.G.M.M., Eckel R.H., Grundy S.M., et al. Harmonizing the metabolic syndrome: A joint interim statement

of the international diabetes federation task force on epidemiology and prevention; National heart, lung, and

blood institute; American heart association; World heart federation; International. Circulation.

2009;120(16):1640-1645. doi:10.1161/CIRCULATIONAHA.109.192644.b Misra A, Chowbey P, Makkar BM, et al. Consensus statement for diagnosis of obesity, abdominal obesity and the

metabolic syndrome for Asian Indians and recommendations for physical activity, medical and surgical

management. J Assoc Physicians India. 2009;57:163–170.

2

Variable DefinitionAbdominal obesity Waist circumference ≥ 80 cm for women and ≥90 cm for men.a

Hypertension

Systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90mmHg or anti-hypertensive intake.Anti-hypertensive intake was from the question: Are you on regular medication for your high blood pressure? Options: a) Yes, b) No.

Impaired fasting glucose Fasting plasma glucose level ≥ 100 mg/dL

Diabetes

Fasting plasma glucose level ≥ 126 mg/dL or diabetes diagnosed.Diagnosed diabetes was from the question: Have you been diagnosed with diabetes? Options: a) Yes, b) No. Medications for diabetes were from the question: Are you on a regular tablets for your diabetes? Options: a) Yes, b) No.

Metabolic syndrome

At least 3 of the following criteria:a a) abdominal obesity, b) systolic blood pressure ≥130 mmHG or diastolic blood pressure ≥85mmHg or anti-hypertensive intake, c) impaired fasting glucose, d) HDL cholesterol <50 mg/dL for women and <40 mg/dL for men, e) Triglycerides ≥150 md/dL

ObesityBody-mass index (kg/m2), classified as:Underweight (<18.5); Normal weight (18.5-22.9); Overweight (23.0-24.9); Obese (25.0 -)b

Further description of statistical analysis Of the 6944 participants enrolled in the third follow-up of Andhra Pradesh Children and Parent

Study (APCAPS) cohort (first clinic visit established within the villages), we included adults (age ≥ 18

years), men and non-pregnant women, with available and reliable cardiovascular outcomes

measurement. Among those eligible for inclusion (n=6229), 3445 (50%) attended the second clinic visit at

the National Institute of Nutrition (NIN) located in Hyderabad, for the cardiovascular risk profile measures.

To adjust for potential selection bias, we used inverse probability weighting (IPW) technique to account

for the population representativeness (Robins et al., 2000; Seaman and White, 2013). The IPW were

applied in two steps a) deriving the probability of attending the NIN for each individual and b) estimating

the weight for each participant (1/probability of attending), which was subsequently used in the linear

mixed model. Therefore, each participant contribution to the model is weighted, to achieve estimates

that aim to be representative of the source population (i.e., participants with a profile that are more likely

to attend receive lower weights compared with participants with a profile not attending). Thus, the final

model was fit in a pseudopopulation, generated with the derived weights (Robins et al., 2000; Seaman

and White, 2013).

We followed the published recommendations for the IPW generation (Seaman et al., 2012;

Seaman and White, 2013). The model to derive the attending probability to the NIN was build using a

logistic regression model, where the outcome was binary Yes/No (Yes - participants who attended NIN,

NO – participants who did not attend NIN). We used as covariates variables related to the fact of

attending or not, but also related to our main outcomes, as recommended by the literature (Seaman et

al., 2012; Seaman and White, 2013). The covariates were all those included in the full adjustment model

(model 4), with education and occupation as originally collected (education = 6 levels; occupation = 10

levels) and the auxiliary variables systolic blood pressure, diastolic blood pressure, impaired fasting

glucose/diabetes, abdominal circumference, and metabolic syndrome criteria. The model also included an

interaction term between village-ID, age, and gender, to account for the underlying potential process of

not attending the NIN, which we hypothesized to be closely related to each village characteristics (e.g.,

distance to the NIN), and the participant age and gender accounted within villages. Sensitivity analysis

using different auxiliary variables (e.g. distance to the NIN instead of village-ID) showed similar results

(data not shown).

We assessed the IPW building process and met all suggested requirements: a) We observed a

high degree of overlap between weights among participants who attended or not the clinic, b) We did not

observe problems with large weights (Seaman et al., 2012; Seaman and White, 2013), and c) we used the

recommended additional tests to evaluate the model (Hosmer-Lemeshow, p=0.885; Hinkley´s method,

p=0.791) (Seaman et al., 2012; Seaman and White, 2013).

We followed the published recommendations for the multiple imputation (MI) followed the IPW

generation (Seaman et al., 2012; Seaman and White, 2013; Sterne et al., 2009). We conducted multiple

3

imputation for covariates using multivariate chained equation methods using the mice package in R

(Buuren and Groothuis-Oudshoorn, 2011; Sterne et al., 2009). We investigated the missingness pattern

and assumed a Missing at Random (MAR) mechanism. We used all covariates used in the full adjustment

model, the outcomes, and auxiliary variables. We generated 10 imputed datasets, with 50 iterations, and

the estimates were pooled following the Rubin´s rule. We checked the variables distribution and

convergence. To allow for the hierarchical structure of the data, we used the predictive mean matching

method for all covariates and entered the village-ID as dummy variables (Little, 1988; Vink et al., 2015).

All analyses were conducted with R-3.4.2 (R Core Team, 2017), with the packages tidyverse (Wickham, 2016), mice (Buuren and Groothuis-Oudshoorn, 2011), miceadds (Robitzsch et al., 2017), lme4 (Bates et al., 2015; Bates, 2010), mgcv(Wood, 2011), and ggplot2(Wickham, 2011).

Bates, D., Machler, M., Bolker, B.M., Walker, S.C., 2015. Fitting Linear Mixed-Effects Models using lme4. J. Stat. Softw. 67, 1–48. https://doi.org/10.18637/jss.v067.i01

Bates, D.M., 2010. lme4: Mixed-effects modeling with R. Springer.

Buuren, S. van, Groothuis-Oudshoorn, K., 2011. mice : Multivariate Imputation by Chained Equations in R. J. Stat. Softw. 45. https://doi.org/10.18637/jss.v045.i03

Little, R.J.A., 1988. Missing-data adjustments in large surveys. J. Bus. Econ. Stat. 6, 287–296. https://doi.org/10.1080/07350015.1988.10509663

R Core Team, 2017. R. R Core Team. https://doi.org/3-900051-14-3

Robins, J.M., Hernan, M.A., Brumback, B., 2000. Marginal structural models and causal inference in epidemiology. Epidemiology 11, 550–560.

Robitzsch, A., Grund, S., Henke, T., 2017. miceadds: Some additional multiple imputation functions, especially for mice.

Seaman, S.R., White, I.R., 2013. Review of inverse probability weighting for dealing with missing data. Stat. Methods Med. Res. 22, 278–295. https://doi.org/10.1177/0962280210395740

Seaman, S.R., White, I.R., Copas, A.J., Li, L., 2012. Combining Multiple Imputation and Inverse-Probability Weighting. Biometrics 68, 129–137. https://doi.org/10.1111/j.1541-0420.2011.01666.x

Sterne, J.A., White, I.R., Carlin, J.B., Spratt, M., Royston, P., Kenward, M.G., Wood, A.M., Carpenter, J.R., 2009. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ 338, b2393. https://doi.org/10.1136/bmj.b2393

Vink, G., Lazendic, G., Buuren, S. Van, 2015. Partitioned predictive mean matching as a multilevel imputation technique. Psychol. Test Assess. Model. 5, 1–16.

Wickham, H., 2016. tidyverse: Easily Install and Load “Tidyverse” Packages., R package version 1.0.0.Wickham, H., 2011. ggplot2. Wiley Interdiscip. Rev. Comput. Stat. 3, 180–185. https://doi.org/10.1002/wics.147

Wood, S.N., 2011. Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. J. R. Stat. Soc. Ser. B Stat. Methodol. https://doi.org/10.1111/j.1467-9868.2010.00749.x

4

Fig. S1 Direct-acyclic graph (DAG) for the association between personal prediction of particulate matter and cardiovascular markers.

5

Fig S2. Linear correlation between the three vascular damage markers stratified by gender

r = 0.54 r = 0.39

Men Women

4 8 12 4 8 12

0.5

1.0

1.5

2.0

cf-PWV - m/s

CIM

T -

mm

r = 0.52 r = 0.44

Men Women

0 20 40 60 0 20 40 60

0.5

1.0

1.5

2.0

AIx - %

CIM

T -

mm

r = 0.53 r = 0.50

Men Women

0 20 40 60 0 20 40 60

4

8

12

AIx - %

cf-P

WV

- m

/s

AIx = augmentation index; cfPWV = carotid-femoral pulse wave velocity; CI = confidence interval; CIMT = carotid intima-media thickness; r = pearson correlation coefficient.

6

Fig S3. Linear correlation between the mean arterial blood pressure and pulse wave velocity and between heart rate and augmentation index stratified by gender.

r = 0.65 r = 0.60

Men Women

75 100 125 150 75 100 125 150

4

8

12

Blood pressure - mmHg

cf-P

WV

- m

/s

r = -0.11 r = -0.33

Men Women

50 75 100 50 75 1000

20

40

60

Heart rate - beats per minute

AIx

- %

AIx = augmentation index; cfPWV = carotid-femoral pulse wave velocity; CI = confidence interval; CIMT = carotid intima-media thickness; r = pearson correlation coefficient.

7

Table S1. Association of personal predicted exposure to PM2.5 with cardiovascular markers stratified by gender.

Model Exposure Men(n = 1564)

Women(n = 1453)

Outcome: CIMT (mm) Mean (SD) 0.802 (0.26) 0.868 (0.24)Difference in CIMT

(95% CI)Difference in CIMT

(95% CI)Model 1 (crude)a PM 2.5 (10 µg/m3) 0.115 (0.102, 0.128) 0.026 (0.016, 0.037)Model 2 (Model1 +Age)b PM 2.5 (10 µg/m3) 0.023 (0.012, 0.034) 0.004 (-0.005, 0.013)Model 3 (Model 2 + Demo)c PM 2.5 (10 µg/m3) 0.027 (0.015, 0.039) 0.006 (-0.004, 0.015)Model 4 (Model 3 + Education)d PM 2.5 (10 µg/m3) 0.026 (0.014, 0.037) 0.003 (-0.006, 0.013)

Model S (Sensitivity analysis: (Model 3 + Education + SLI)f PM 2.5 (10 µg/m3) 0.025 (0.013, 0.037) -0.001 (-0.011, 0.010)

Outcome: PWV (m/s) Mean (SD) 7.02 (1.4) 6.99 (1.3)Difference in cfPWV

(95% CI)Difference in cfPWV

(95% CI)Model 1 (crude)a PM 2.5 (10 µg/m3) 0.530 (0.459, 0.601) 0.063 (-0.001, 0.128)Model 2 (Model1 +Age)b PM 2.5 (10 µg/m3) 0.040 (-0.020, 0.100) -0.048 (-0.096, 0.001)Model 3 (Model 2 + Demo)c PM 2.5 (10 µg/m3) 0.070 (0.010, 0.131) -0.013 (-0.062, 0.036)Model 4 (Model 3 + SES)d PM 2.5 (10 µg/m3) 0.069 (0.008, 0.131) -0.015 (-0.064, 0.035)Model 5 (Model 4 + BP)e PM 2.5 (10 µg/m3) 0.041 (-0.01, 0.092) -0.028 (-0.071, 0.014)

Model S (Sensitivity analysis: (Model 3 + Education + SLI)f PM 2.5 (10 µg/m3) 0.069 (0.007, 0.130) 0 (-0.056, 0.056)

Outcome: AIx (%) Mean (SD) 21.9 (11) 24.6 (12)Difference in AIx

(95% CI)Difference in AIx

(95% CI)Model 1 (crude)a PM 2.5 (10 µg/m3) 4.061 (3.505, 4.616) 1.227 (0.741, 1.713)Model 2 (Model1 +Age)b PM 2.5 (10 µg/m3) 0.527 (0.029, 1.025) 0.405 (0.006, 0.804)Model 3 (Model 2 + Demo)c PM 2.5 (10 µg/m3) 0.833 (0.334, 1.331) 0.667 (0.281, 1.053)Model 4 (Model 3 + SES)d PM 2.5 (10 µg/m3) 0.783 (0.282, 1.284) 0.627 (0.237, 1.017)Model 5 (Model 4 + Height)e PM 2.5 (10 µg/m3) 0.750 (0.253, 1.247) 0.532 (0.138, 0.926)

Model S (Sensitivity analysis: Model 3 + Education + SLI)f PM 2.5 (10 µg/m3) 0.786 (0.280, 1.292) 0.596 (0.160, 1.032)

Analysis conducted in 10 multiple imputed datasets, using a linear mixed model (random intercepts: households nested to villages), with correction for selection bias through inverse probability weighting. a Model 1 has no adjustment for CIMT and cfPWV, and it is adjusted for heart rate for AIx.b Model 2: Model 1 + age (natural spline, df=3 for CIMT, linear for cfPWV and AIx). c Model 3: Model 2 + body-mass index, fruits and vegetables consumption, alcohol consumption and physical activity.d Model 4: Model 3 + education.e Model 5: Model 4 + mean blood pressure for cfPWV; and + height for AIx.f Model S: Model 3 + education + standard living index

AIx = augmentation index; cfPWV = carotid-femoral pulse wave velocity; CI = confidence interval; CIMT = carotid intima-

8

media thickness; PM 2.5 = particulate matter with an aerodynamic diameter of 2.5 micrometers or less; SLI = standard living index.

9

Fig S4. Exposure-response function between personal predicted exposure to PM2.5 and Augmentation index (AIx) among men and exposure distribution according to predictors

Additive mixed effects model with nested random intercepts (household within village) and a smooth term (thin-plate spline basis) on PM2.5. Adjustment as Model 4: age, body-mass index, alcohol consumption, fruit and vegetable consumption, physical activity, education and heart rate. Grey shade areas represent 95% credible intervals. At the bottom, the probability densities were estimated using kernel smoothing methods. The men exposed to manual job, active tobacco smoking and biomass fuel use and those exposed to manual job and active tobacco smoking correspond to 506 (32.4%) participants.

10

Table S2. Association of personal predicted exposure to BC with cardiovascular markers stratified by gender.

Model Exposure Men(n = 1564)

Women (n = 1453)

Outcome: CIMT (mm) Mean (SD) 0.802 (0.26) 0.868 (0.24)Percent change in CIMT

(95% CI)Percent change in CIMT

(95% CI)Model 1 (crude)a BC (2 µg/m3) -0.052 (-0.090, -0.014) 0.023 (0.012, 0.034)Model 2 (Model1 +Age)b BC (2 µg/m3) -0.014 (-0.042, 0.014) 0.005 (-0.003, 0.013)Model 3 (Model 2 + Demo)c BC (2 µg/m3) -0.014 (-0.042, 0.013) 0.006 (-0.002, 0.014)Model 4 (Model 3 + SES)d BC (2 µg/m3) -0.015 (-0.042, 0.013) 0.004 (-0.004, 0.013)

Model S (Sensitivity analysis: (Model 3 + Education + SLI)f BC (2 µg/m3) -0.018 (-0.046, 0.010) -0.001 (-0.012, 0.011)

Outcome: PWV (m/s) Mean (SD) 7.02 (1.4) 6.99 (1.3)Difference in cfPWV

(95% CI)Difference in cfPWV

(95% CI)Model 1 (crude)a BC (2 µg/m3) -0.126 (-0.318, 0.066) 0.037 (-0.023, 0.097)Model 2 (Model1 +Age)b BC (2 µg/m3) -0.001 (-0.142, 0.139) -0.049 (-0.095, -0.003)Model 3 (Model 2 + Demo)c BC (2 µg/m3) -0.008 (-0.143, 0.127) -0.023 (-0.066, 0.019)Model 4 (Model 3 + SES)d BC (2 µg/m3) -0.008 (-0.144, 0.127) -0.025 (-0.068, 0.018)Model 5 (Model 4 + BP)e BC (2 µg/m3) -0.032 (-0.141, 0.078) -0.022 (-0.060, 0.017)

Model S (Sensitivity analysis: (Model 3 + Education + SLI)f BC (2 µg/m3) -0.015 (-0.153, 0.124) -0.012 (-0.065, 0.041)

Outcome: AIx (%) Mean (SD) 21.9 (11) 24.6 (12)Difference in AIx

(95% CI)Difference in AIx

(95% CI)Model 1 (crude)a BC (2 µg/m3) -1.066 (-2.622, 0.49) 0.968 (0.426, 1.511)Model 2 (Model1 +Age)b BC (2 µg/m3) 0.290 (-0.982, 1.562) 0.321 (-0.021, 0.663)Model 3 (Model 2 + Demo)c BC (2 µg/m3) 0.166 (-1.061, 1.393) 0.494 (0.138, 0.851)Model 4 (Model 3 + SES)d BC (2 µg/m3) 0.166 (-1.060, 1.391) 0.457 (0.099, 0.814)Model 5 (Model 4 + Height)e BC (2 µg/m3) -0.095 (-1.310, 1.120) 0.378 (0.026, 0.729)

Model S (Sensitivity analysis: (Model 3 + Education + SLI)f BC (2 µg/m3) 0.135 (-1.153, 1.422) 0.443 (-0.010, 0.896)

Analysis conducted in 10 multiple imputed datasets, using a linear mixed model (random intercepts: households nested to villages), with correction for selection bias through inverse probability weighting. a Model 1 has no adjustment for CIMT and cfPWV, and it is adjusted for heart rate for AIx.b Model 2: Model 1 + age (natural spline, df=3 for CIMT, linear for cfPWV and Aix). c Model 3: Model 2 + body-mass index, fruits and vegetables consumption, alcohol consumption and physical activity.d Model 4: Model 3 + education for men.e Model 5: Model 4 + mean blood pressure for cfPWV; and + height for AIx.f Model S: Model 3 + education + standard living index.AIx = augmentation index; BC = black carbon; cfPWV = carotid-femoral pulse wave velocity; CI = confidence interval; CIMT = carotid intima-media thickness; SLI = standard living index.

11

12

Figure S5. Crude and adjusted associations between personal predicted BC and three cardiovascular markers stratified by gender.

Analysis conducted in 10 multiple imputed datasets, using linear mixed models with households nested to villages as random intercepts and correction for selection bias through inverse probability weighting. Exposure BC was modelled as 2 μg/m3 increase. Model 1: BC and random effects for households nested to villages; Model 2: Model 1 + age (natural spline, df=3 for CIMT and linear term for cfPWV and AIx); Model 3: Model 2 + body-mass index, alcohol consumption, fruit and vegetable consumption, and physical activity; Model 4: Model 3 + education (main model); Model 5: Model 4 + blood pressure (for cfPWV) and height (for AIx). All models for AIx were adjusted for heart rate. The point estimates are represented by boxes and their 95% confidence intervals as whiskers. AIx: aortic augmentation index; BC: black carbon; cfPWV: carotid-femoral pulse wave velocity; CIMT = carotid intima-media thickness.

13

Fig. S6. Sensitivity analysis for the crude and adjusted associations between personal predicted PM2.5 and carotid intima-media thickness (CIMT) stratified by gender.

M1 (Crude) M2 (M1+Age) M3 (M2+Demo) M4 (M3+Educ)

Men

Wom

en

CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW

0.00

0.05

0.10

0.00

0.05

0.10Diff

eren

ce in

CIM

T -

mm

Exposure: PM2.5 (10 g m3) increase

Exposure to PM2.5 was modelled as 10 μg/m3 increase. We show the results from the models in four approaches: 1) Complete case data without inverse probability weighting (CC), 2) Complete case data with inverse probability weighting (CC-IPW), 3) Multiple imputed data (m=10) without inverse probability weighting (MI), 4) Multiple imputed data (m=10) with inverse probability weighting (our main analysis).Model 1: PM2.5 and random intercepts for households nested to villages; Model 2: Model 1 + age (natural spline, df=3); Model 3: Model 2 + body-mass index, alcohol consumption, fruit and vegetable consumption, and physical activity; Model 4: Model 3 + education (main model).

The point estimates are represented by boxes and their 95% confidence intervals as horizontal lines. CIMT = carotid intima-media thickness; PM 2.5 = particulate matter with an aerodynamic diameter of 2.5 micrometers or less.

14

Fig S7. Sensitivity analysis for the crude and adjusted associations between personal predicted PM2.5 and carotid-femoral pulse wave velocity (cf-PWV) stratified by gender.

M1 (Crude) M2 (M1+Age) M3 (M2+Demo) M4 (M3+Educ) M5 (M4+BP)

Men

Wom

en

CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW

0.0

0.2

0.4

0.6

0.0

0.2

0.4

0.6

Diff

eren

ce in

cfP

WV

- m

/s

Exposure: PM2.5 (10 g m3) increase

Exposure to PM2.5 was modelled as 10 μg/m3 increase. We show the results from the models in four approaches:

1) Complete case data without inverse probability weighting (CC), 2) Complete case data with inverse probability weighting (CC-IPW), 3) Multiple imputed data (m=10) without inverse probability weighting (MI), 4) Multiple imputed data (m=10) with inverse probability weighting (our main analysis).Model 1: PM2.5 and random intercepts for households nested to villages; Model 2: Model 1 + age (linear term); Model 3: Model 2 + body-mass index, alcohol consumption, fruit and vegetable consumption, and physical activity; Model 4: Model 3 + education (main model).The point estimates are represented by boxes and their 95% confidence intervals as horizontal lines. cfPWV: carotid-femoral pulse wave velocity; PM 2.5 = particulate matter with an aerodynamic diameter of 2.5 micrometers or less.

15

Fig S8. Sensitivity analysis for the crude and adjusted associations between personal predicted PM2.5 and aortic augmentation index (AIx) stratified by gender.

M1 (Crude) M2 (M1+Age) M3 (M2+Demo) M4 (M3+Educ) M5 (M4+Height)

Men

Wom

en

CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW

0

1

2

3

4

5

0

1

2

3

4

5

Diff

eren

ce in

AIx

- %

Exposure: PM2.5 (10 g m3) increase

Exposure to PM2.5 was modelled as 10 μg/m3 increase. We show the results from the models in four approaches:

1) Complete case data without inverse probability weighting (CC), 2) Complete case data with inverse probability weighting (CC-IPW), 3) Multiple imputed data (m=10) without inverse probability weighting (MI), 4) Multiple imputed data (m=10) with inverse probability weighting (our main analysis).Model 1: PM2.5 and random intercepts for households nested to villages (+ heart rate); Model 2: Model 1 + age (linear term); Model 3: Model 2 + body-mass index, alcohol consumption, fruit and vegetable consumption, and physical activity; Model 4: Model 3 + education (main model).

The point estimates are represented by boxes and their 95% confidence intervals as horizontal lines. AIx: aortic augmentation index; PM 2.5 = particulate matter with an aerodynamic diameter of 2.5 micrometers or less.

16

Fig S9. Sensitivity analysis for the crude and adjusted associations between personal predicted BC and carotid intima-media thickness (CIMT) stratified by gender.

M1 (Crude) M2 (M1+Age) M3 (M2+Demo) M4 (M3+Educ)

Men

Wom

en

CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW

-0.05

0.00

-0.05

0.00

Diff

eren

ce in

CIM

T -

mm

Exposure: Black carbon (2 g m3) increase

Exposure to BC was modelled as 2 μg/m3 increase. We show the results from the models in four approaches:

1) Complete case data without inverse probability weighting (CC), 2) Complete case data with inverse probability weighting (CC-IPW), 3) Multiple imputed data (m=10) without inverse probability weighting (MI), 4) Multiple imputed data (m=10) with inverse probability weighting (our main analysis).Model 1: BC and random intercepts for households nested to villages; Model 2: Model 1 + age (natural spline, df=3); Model 3: Model 2 + body-mass index, alcohol consumption, fruit and vegetable consumption, and physical activity; Model 4: Model 3 + education (main model).

The point estimates are represented by boxes and their 95% confidence intervals as horizontal lines. BC = black carbon; CIMT = carotid intima-media thickness.

17

Fig S10. Sensitivity analysis for the crude and adjusted associations between personal predicted BC and carotid-femoral pulse wave velocity (cf-PWV) stratified by gender.

M1 (Crude) M2 (M1+Age) M3 (M2+Demo) M4 (M3+Educ) M5 (M4+BP)

Men

Wom

en

CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW

-0.3

-0.2

-0.1

0.0

0.1

-0.3

-0.2

-0.1

0.0

0.1

Diff

eren

ce in

cfP

WV

- m

/s

Exposure: Black carbon (2 g m3) increase

Exposure to BC was modelled as 2 μg/m3 increase. We show the results from the models in four approaches:

1) Complete case data without inverse probability weighting (CC), 2) Complete case data with inverse probability weighting (CC-IPW), 3) Multiple imputed data (m=10) without inverse probability weighting (MI), 4) Multiple imputed data (m=10) with inverse probability weighting (our main analysis).Model 1: PM2.5 and random intercepts for households nested to villages; Model 2: Model 1 + age (linear term); Model 3: Model 2 + body-mass index, alcohol consumption, fruit and vegetable consumption, and physical activity; Model 4: Model 3 + education (main model).

The point estimates are represented by boxes and their 95% confidence intervals as horizontal lines. BC = black carbon; cfPWV: carotid-femoral pulse wave velocity.

18

Fig S11. Sensitivity analysis for the crude and adjusted associations between personal predicted BC and aortic augmentation index (AIx) stratified by gender.

M1 (Crude) M2 (M1+Age) M3 (M2+Demo) M4 (M3+Educ) M5 (M4+Height)

Men

Wom

en

CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW CC CC-IPW MI MI-IPW

-3

-2

-1

0

1

-3

-2

-1

0

1

Diff

eren

ce in

AIx

- %

Exposure: Black carbon (2 g m3) increase

Exposure to BC was modelled as 2 μg/m3 increase. We show the results from the models in four approaches:

1) Complete case data without inverse probability weighting (CC), 2) Complete case data with inverse probability weighting (CC-IPW), 3) Multiple imputed data (m=10) without inverse probability weighting (MI), 4) Multiple imputed data (m=10) with inverse probability weighting (our main analysis).Model 1: PM2.5 and random intercepts for households nested to villages (+ heart rate); Model 2: Model 1 + age (linear term); Model 3: Model 2 + body-mass index, alcohol consumption, fruit and vegetable consumption, and physical activity; Model 4: Model 3 + education (main model).

The point estimates are represented by boxes and their 95% confidence intervals as horizontal lines. AIx: aortic augmentation index; BC = black carbon.

19

Fig.S12. Adjusted associations (main model) between personal predicted PM2.5 and three cardiovascular markers in the whole population and subgroups by age and cardiometabolic risk factors.

personal predicted PM2.5 and CIMT

n=1564 n=860 n=704 n=1282 n=282 n=1166 n=398 n=1468 n=96 n=1217 n=347 n=1453 n=638 n=815 n=1156 n=297 n=1199 n=254 n=1387 n=66 n=1018 n=435

Men Women

M4

(Who

le p

opul

atio

n)

M4

(Age

<40

)

M4

(Age

>=4

0)

M4

(with

out M

etS

d)

M4

(with

Met

SD

)

M4

(with

out H

TN)

M4

(with

HTN

)

M4

(with

out D

M)

M4

(with

DM

)

M4

(with

out O

besi

ty)

M4

(with

Obe

sity

)

M4

(Who

le p

opul

atio

n)

M4

(Age

<40

)

M4

(Age

>=4

0)

M4

(with

out M

etS

d)

M4

(with

Met

SD

)

M4

(with

out H

TN)

M4

(with

HTN

)

M4

(with

out D

M)

M4

(with

DM

)

M4

(with

out O

besi

ty)

M4

(with

Obe

sity

)

0.00

0.05

0.10

Diff

eren

ce in

CIM

T - m

m

personal predicted PM2.5 and cfPWV

n=1564 n=860 n=704 n=1282 n=282 n=1166 n=398 n=1468 n=96 n=1217 n=347 n=1453 n=638 n=815 n=1156 n=297 n=1199 n=254 n=1387 n=66 n=1018 n=435

Men Women

M4

(Who

le p

opul

atio

n)

M4

(Age

<40

)

M4

(Age

>=4

0)

M4

(with

out M

etS

d)

M4

(with

Met

Sd)

M4

(with

out H

TN)

M4

(with

HTN

)

M4

(with

out D

M)

M4

(with

DM

)

M4

(with

out O

besi

ty)

M4

(with

Obe

sity

)

M4

(Who

le p

opul

atio

n)

M4

(Age

<40

)

M4

(Age

>=4

0)

M4

(with

out M

etS

d)

M4

(with

Met

Sd)

M4

(with

out H

TN)

M4

(with

HTN

)

M4

(with

out D

M)

M4

(with

DM

)

M4

(with

out O

besi

ty)

M4

(with

Obe

sity

)

-0.2

0.0

0.2

0.4

Diff

eren

ce in

cfP

WV

- m

/s

20

personal predicted PM2.5 and AIx

n=1564 n=860 n=704 n=1282 n=282 n=1166 n=398 n=1468 n=96 n=1217 n=347 n=1453 n=638 n=815 n=1156 n=297 n=1199 n=254 n=1387 n=66 n=1018 n=435

Men Women

M4

(Who

le p

opul

atio

n)

M4

(Age

<40

)

M4

(Age

>=4

0)

M4

(with

out M

etS

d)

M4

(with

Met

Sd)

M4

(with

out H

TN)

M4

(with

HTN

)

M4

(with

out D

M)

M4

(with

DM

)

M4

(with

out O

besi

ty)

M4

(with

Obe

sity

)

M4

(Who

le p

opul

atio

n)

M4

(Age

<40

)

M4

(Age

>=4

0)

M4

(with

out M

etS

d)

M4

(with

Met

Sd)

M4

(with

out H

TN)

M4

(with

HTN

)

M4

(with

out D

M)

M4

(with

DM

)

M4

(with

out O

besi

ty)

M4

(with

Obe

sity

)

-2

0

2

4

Diff

eren

ce in

AIx

- %

Analysis conducted in 10 multiple imputed datasets, using linear mixed models with households nested within villages as random intercepts and correction for selection bias through inverse probability weighting. Exposure PM2.5 was modeled as 10 μg/m3 increase. Model 4: PM2.5 + age + body-mass index, alcohol consumption, fruit and vegetable consumption, physical activity + education (main model); All models for AIx were adjusted for heart rate. The point estimates are represented by boxes and their 95% confidence intervals as whiskers. AIx: aortic augmentation index; BC: black carbon; cfPWV: carotid-femoral pulse wave velocity; CIMT = carotid intima-media thickness; DM = diabetes mellitus; HTN = hypertension; MetSd = metabolic syndrome.

21

Fig. S13. Adjusted associations (main model) between personal predicted BC and three cardiovascular markers in the whole population and subgroups by age and cardiometabolic risk factors.

personal predicted BC and CIMT

n=1564 n=860 n=704 n=1282 n=282 n=1166 n=398 n=1468 n=96 n=1217 n=347 n=1453 n=638 n=815 n=1156 n=297 n=1199 n=254 n=1387 n=66 n=1018 n=435

Men Women

M4

(Who

le p

opul

atio

n)

M4

(Age

<40

)

M4

(Age

>=4

0)

M4

(with

out M

etS

d)

M4

(with

Met

SD

)

M4

(with

out H

TN)

M4

(with

HTN

)

M4

(with

out D

M)

M4

(with

DM

)

M4

(with

out O

besi

ty)

M4

(with

Obe

sity

)

M4

(Who

le p

opul

atio

n)

M4

(Age

<40

)

M4

(Age

>=4

0)

M4

(with

out M

etS

d)

M4

(with

Met

SD

)

M4

(with

out H

TN)

M4

(with

HTN

)

M4

(with

out D

M)

M4

(with

DM

)

M4

(with

out O

besi

ty)

M4

(with

Obe

sity

)

-0.2

-0.1

0.0

0.1

Diff

eren

ce in

CIM

T - m

m

personal predicted BC and cfPWV

n=1564 n=860 n=704 n=1282 n=282 n=1166 n=398 n=1468 n=96 n=1217 n=347 n=1453 n=638 n=815 n=1156 n=297 n=1199 n=254 n=1387 n=66 n=1018 n=435

Men Women

M4

(Who

le p

opul

atio

n)

M4

(Age

<40

)

M4

(Age

>=4

0)

M4

(with

out M

etS

d)

M4

(with

Met

Sd)

M4

(with

out H

TN)

M4

(with

HTN

)

M4

(with

out D

M)

M4

(with

DM

)

M4

(with

out O

besi

ty)

M4

(with

Obe

sity

)

M4

(Who

le p

opul

atio

n)

M4

(Age

<40

)

M4

(Age

>=4

0)

M4

(with

out M

etS

d)

M4

(with

Met

Sd)

M4

(with

out H

TN)

M4

(with

HTN

)

M4

(with

out D

M)

M4

(with

DM

)

M4

(with

out O

besi

ty)

M4

(with

Obe

sity

)

-1.0

-0.5

0.0

0.5

Diff

eren

ce in

cfP

WV

- m

/s

22

personal predicted BC and AIx

n=1564 n=860 n=704 n=1282 n=282 n=1166 n=398 n=1468 n=96 n=1217 n=347 n=1453 n=638 n=815 n=1156 n=297 n=1199 n=254 n=1387 n=66 n=1018 n=435

Men WomenM

4 (W

hole

pop

ulat

ion)

M4

(Age

<40

)

M4

(Age

>=4

0)

M4

(with

out M

etS

d)

M4

(with

Met

Sd)

M4

(with

out H

TN)

M4

(with

HTN

)

M4

(with

out D

M)

M4

(with

DM

)

M4

(with

out O

besi

ty)

M4

(with

Obe

sity

)

M4

(Who

le p

opul

atio

n)

M4

(Age

<40

)

M4

(Age

>=4

0)

M4

(with

out M

etS

d)

M4

(with

Met

Sd)

M4

(with

out H

TN)

M4

(with

HTN

)

M4

(with

out D

M)

M4

(with

DM

)

M4

(with

out O

besi

ty)

M4

(with

Obe

sity

)

0

4

8

Diff

eren

ce in

AIx

- %

Analysis conducted in 10 multiple imputed datasets, using linear mixed models with households nested within villages as random intercepts and correction for selection bias through inverse probability weighting. Exposure BC was modelled as 2 μg/m3 increase. Model 4: BC + age + body-mass index, alcohol consumption, fruit and vegetable consumption, physical activity + education (main model); All models for AIx were adjusted for heart rate. The point estimates are represented by boxes and their 95% confidence intervals as whiskers. AIx: aortic augmentation index; BC: black carbon; cfPWV: carotid-femoral pulse wave velocity; CIMT = carotid intima-media thickness; DM = diabetes mellitus; HTN = hypertension; MetSd = metabolic syndrome.

23

Fig. S14. Association between vascular markers and measured personal PM2.5 (n=104 for men, n=83 for women)

Men Women

M1

(Cru

de)

M2

(M1+

Age

)

M3

(M2+

Dem

o)

M4

(M3+

Edu

c)

M1

(Cru

de)

M2

(M1+

Age

)

M3

(M2+

Dem

o)

M4

(M3+

Edu

c)

-0.005

0.000

0.005

0.010

Diff

eren

ce in

CIM

T -

mm

Men Women

M1

(Cru

de)

M2

(M1+

Age

)

M3

(M2+

Dem

o)

M4

(M3+

Edu

c)

M5

(M4+

BP

)

M1

(Cru

de)

M2

(M1+

Age

)

M3

(M2+

Dem

o)

M4

(M3+

Edu

c)

M5

(M4+

BP

)

-0.04

0.00

0.04

Diff

eren

ce in

cfP

WV

- m

/s

Men Women

M1

(Cru

de)

M2

(M1+

Age

)

M3

(M2+

Dem

o)

M4

(M3+

Edu

c)

M5

(M4+

BP

)

M1

(Cru

de)

M2

(M1+

Age

)

M3

(M2+

Dem

o)

M4

(M3+

Edu

c)

M5

(M4+

BP

)

-0.8

-0.4

0.0

Diff

eren

ce in

AIx

- %

Analysis conducted in 10 multiple imputed datasets, using linear mixed models with households nested within villages as random intercepts. PM2.5 exposure was modelled as 10 μg/m3 increase. Model 1: PM2.5 and random effects for households nested within villages; Model 2: Model 1 + age; Model 3: Model 2 + body-mass index, alcohol consumption, fruit and vegetable consumption, and physical activity; Model 4: Model 3 + education (main model); Model 5: Model 4 + blood pressure (for cfPWV) and height (for AIx). All models for AIx were adjusted for heart rate. The point estimates are represented by boxes and their 95% confidence intervals as whiskers. AIx: aortic augmentation index; cfPWV: carotid-femoral pulse wave velocity; CIMT = carotid intima-media thickness; PM 2.5 = particulate matter with an aerodynamic diameter of 2.5 micrometers or less. AIx: aortic augmentation index; BC: black carbon; cfPWV: carotid-femoral pulse wave velocity; CIMT = carotid intima-media thickness.

24