ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’,...

37
Supplementary Materials for Ambient air pollution and markers of fetal growth: a retrospective population-based cohort study of 2.57 million term singleton births in China Pi Guo 1, † , Yuliang Chen 1, † , Haisheng Wu 1 , Jing Zeng 2 , Zhisheng Zeng 2 , Weiping Li 3 , Qingying Zhang 1 , Xia Huo 4 , Wenru Feng 5 , Jiumin Lin 6 , Huazhang Miao 2, * , Yingxian Zhu 2, * 1 Department of Preventive Medicine, Shantou University Medical College, Shantou 515041, China 2 Guangdong Women and Children Hospital, No.521 Xingnan Road, Guangzhou 511442, China 3 Clinical Cohort Research CenterFirst Affiliated Hospital of Shantou University Medical College, Shantou 515041, China 4 Laboratory of Environmental Medicine and Developmental Toxicology, Guangzhou and Guangdong Key Laboratory of Environmental Pollution and Health, School of Environment, Jinan University, Guangzhou 510632, China 5 Guangzhou Center for Disease Control and Prevention, Guangzhou 510440, China 6 Department of Hepatology and Infectious Diseases, the Second Affiliated Hospital, Shantou University Medical College, Shantou 515041, China These two authors contributed equally in this work. * Corresponding authors Huazhang Miao, Department of Healthcare, Guangdong Women and Children Hospital, Guangzhou 511442, China (Email: [email protected]) 1

Transcript of ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’,...

Page 1: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Materials for

Ambient air pollution and markers of fetal growth: a retrospective

population-based cohort study of 2.57 million term singleton births in

ChinaPi Guo 1, †, Yuliang Chen 1, †, Haisheng Wu 1, Jing Zeng 2, Zhisheng Zeng 2, Weiping Li 3,

Qingying Zhang 1, Xia Huo 4, Wenru Feng 5, Jiumin Lin 6, Huazhang Miao 2, *, Yingxian Zhu 2, *

1 Department of Preventive Medicine, Shantou University Medical College, Shantou 515041,

China2 Guangdong Women and Children Hospital, No.521 Xingnan Road, Guangzhou 511442, China3 Clinical Cohort Research Center,First Affiliated Hospital of Shantou University Medical

College, Shantou 515041, China4 Laboratory of Environmental Medicine and Developmental Toxicology, Guangzhou and

Guangdong Key Laboratory of Environmental Pollution and Health, School of Environment,

Jinan University, Guangzhou 510632, China5 Guangzhou Center for Disease Control and Prevention, Guangzhou 510440, China6 Department of Hepatology and Infectious Diseases, the Second Affiliated Hospital, Shantou

University Medical College, Shantou 515041, China

† These two authors contributed equally in this work.

* Corresponding authorsHuazhang Miao, Department of Healthcare, Guangdong Women and Children Hospital, Guangzhou 511442, China (Email: [email protected])Yingxian Zhu, Department of Healthcare, Guangdong Women and Children Hospital, Guangzhou 511442, China (Email: [email protected])

1

Page 2: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

ContentsSupplementary Figure 1: Exclusion process of birth data used in analysis. Supplementary Figure 2: Parameter tuning in inverse distance weighting (IDW) algorithm. Two main parameters including the inverse distance weighting power M and the nearest observations numbers N were assessed. The coefficient of determination R2 was used to identify the optimal combination of the values of two parameters. Parameters tuning for air pollutants and meteorological factors: CO (A), NO2 (B), O3 (C), PM10 (D), PM2.5

(E),SO2 (F), mean temperature (G), and relative humidity (H). The largest 10-fold cross validation R2 is marked with a red triangle, which corresponds to the optimal combination of parameters. Supplementary Figure 3: Spatial variation of air pollutant concentration. (A) Spatial variation of PM2.5 concentration (Unit: μg/m3). (B) Spatial variation of NO2 concentration (Unit: μg/m3). (C) Spatial variation of O3 concentration (Unit: μg/m3). (D) Spatial variation of PM10 concentration (Unit: μg/m3). (E) Spatial variation of SO2 concentration (Unit: μg/m3).Supplementary Figure 4: Spatial variation of meteorological factors. (A) Spatial variation of mean ambient temperature (Unit: ). (B℃ ) Spatial variation of relative humidity.Supplementary Figure 5: Spatial variation of number of term low birth weight (LBW) (<2500 g and ≥37 weeks gestation) neonates and mean birth weight of term infants. (A) Spatial variation of number of term LBW neonates. (B) Spatial variation of mean birth weight of term infants.Supplementary Figure 6: Odds of term small for gestational age (SGA), associated with interquartile range (IQR) increases in air pollutant concentration in the first trimester, in single and two air pollutant models. The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated. Supplementary Figure 7: Odds of term small for gestational age (SGA), associated with interquartile range (IQR) increases in air pollutant concentration in the second trimester, in single and two air pollutant models. The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated. Supplementary Figure 8: Odds of term small for gestational age (SGA), associated with interquartile range (IQR) increases in air pollutant concentration in the third trimester, in single and two air pollutant models. The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above.

2

Page 3: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

PM2.5 and PM10 were not entered into the same model together as they were too highly correlated.Supplementary Figure 9: Odds of term low birth weight (LBW), associated with interquartile range (IQR) increases in air pollutant concentration in the first trimester, in single and two air pollutant models. The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated. Supplementary Figure 10: Odds of term low birth weight (LBW), associated with interquartile range (IQR) increases in air pollutant concentration in the second trimester, in single and two air pollutant models. The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10

were not entered into the same model together as they were too highly correlated. Supplementary Figure 11: Odds of term low birth weight (LBW), associated with interquartile range (IQR) increases in air pollutant concentration in the third trimester, in single and two air pollutant models. The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated. Supplementary Figure 12: The associations of term birth weight with interquartile range (IQR) increases in air pollutant concentration in the first trimester, in single and two air pollutant models. The model regression coefficients rescaled to IQR increments specify to pollutants were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated. Supplementary Figure 13: The associations of term birth weight with interquartile range (IQR) increases in air pollutant concentration in the second trimester, in single and two air pollutant models. The model regression coefficients rescaled to IQR increments specify to pollutants were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term

3

Page 4: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

(pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated.Supplementary Figure 14: The associations of term birth weight with interquartile range (IQR) increases in air pollutant concentration in the third trimester, in single and two air pollutant models. The model regression coefficients rescaled to IQR increments specify to pollutants were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated.Supplementary Figure 15: The associations of term small for gestational age (SGA) with interquartile range (IQR) increases in air pollutant concentration over the entire pregnancy, stratified by potential modifiers (maternal age or neonate’s sex). The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during entire pregnancy in each subgroup. P-values for heterogeneity in the effects in the subgroups were estimated by using the DerSimonian-Laird method. Generalized estimating equation models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, meteorological factors and an interaction term (pollutant exposures × month of conception). Supplementary Figure 16: The associations of term low birth weight (LBW) with interquartile range (IQR) increases in air pollutant concentration over the entire pregnancy, stratified by potential modifiers (maternal age or neonate’s sex). The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during entire pregnancy in each subgroup. P-values for heterogeneity in the effects in the subgroups were estimated by using the DerSimonian-Laird method. Generalized estimating equation models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception). Supplementary Figure 17: Sensitivity analyses for association of air pollutant exposure across the entire pregnancy with the risk of term small for gestational age (SGA) in Guangdong, China. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. The ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’, ‘Sensitivity analysis 6’ and ‘Sensitivity analysis 7’ represents the analysis of variation of parameter value in inverse distance weighting algorithm, the impact of potential misclassification of exposure, the Han nationality, the ambient temperature adjusted as a nonlinear term in the model, the use of 10th percentile as the cutting point to determine term SGA, and the restricted subgroup of births with gestational age < 40 weeks, respectively. Supplementary Figure 18: Sensitivity analyses for association of air pollutant exposure across the entire pregnancy with the risk of term low birth weight (LBW) in Guangdong, China. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette

4

Page 5: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. The ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’ represents the analysis of variation of parameter value in inverse distance weighting algorithm, the impact of potential misclassification of exposure, the Han nationality, the ambient temperature adjusted as a nonlinear term in the model, and the restricted subgroup of births with gestational age < 40 weeks respectively. Supplementary Figure 19: Sensitivity analyses for association of air pollutant exposure across the entire pregnancy with the risk of term birth weight in Guangdong, China. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. The ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’ represents the analysis of variation of parameter value in inverse distance weighting algorithm, the impact of potential misclassification of exposure, the Han nationality, the ambient temperature adjusted as a nonlinear term in the model, and the restricted subgroup of births with gestational age < 40 weeks, respectively.Table S1. Characteristics of the study population and distribution of pregnancy outcomes by city.

5

Page 6: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 1: Exclusion process of birth data used in analysis.

6

Page 7: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 2: Parameter tuning in inverse distance weighting (IDW) algorithm. Two main parameters including the inverse distance weighting power M and the nearest observations numbers N were assessed. The coefficient of determination R2 was used to identify the optimal combination of the values of two parameters. Parameters tuning for air pollutants and meteorological factors: CO (A), NO2 (B), O3 (C), PM10 (D), PM2.5 (E), SO2 (F), mean temperature (G), and relative humidity (H). The largest 10-fold cross validation R2 is marked with a red triangle, which corresponds to the optimal combination of parameters.

7

Page 8: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 3: Spatial variation of air pollutant concentration. (A) Spatial variation of PM2.5 concentration (Unit: μg/m3). (B) Spatial variation of NO2 concentration (Unit: μg/m3). (C) Spatial variation of O3 concentration (Unit: μg/m3). (D) Spatial variation of PM10 concentration (Unit: μg/m3). (E) Spatial variation of SO2 concentration (Unit: μg/m3).

8

Page 9: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 4: Spatial variation of meteorological factors. (A) Spatial variation of mean ambient temperature (Unit: ). (B℃ ) Spatial variation of relative humidity.

9

Page 10: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 5: Spatial variation of number of term LBW (<2500 g and ≥37 weeks gestation) neonates and mean birth weight of term infants. (A) Spatial variation of number of term LBW neonates. (B) Spatial variation of mean birth weight of term infants.

10

Page 11: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 6: Odds of term small for gestational age (SGA), associated with interquartile range (IQR) increases in air pollutant concentration in the first trimester, in single and two air pollutant models. The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated.

11

Page 12: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 7: Odds of term small for gestational age (SGA), associated with interquartile range (IQR) increases in air pollutant concentration in the second trimester, in single and two air pollutant models. The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated.

12

Page 13: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 8: Odds of term small for gestational age (SGA), associated with interquartile range (IQR) increases in air pollutant concentration in the third trimester, in single and two air pollutant models. The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated.

13

Page 14: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 9: Odds of term low birth weight (LBW), associated with interquartile range (IQR) increases in air pollutant concentration in the first trimester, in single and two air pollutant models. The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated.

14

Page 15: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 10: Odds of term low birth weight (LBW), associated with interquartile range (IQR) increases in air pollutant concentration in the second trimester, in single and two air pollutant models. The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10

were not entered into the same model together as they were too highly correlated.

15

Page 16: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 11: Odds of term low birth weight (LBW), associated with interquartile range (IQR) increases in air pollutant concentration in the third trimester, in single and two air pollutant models. The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated.

16

Page 17: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 12: The associations of term birth weight with interquartile range (IQR) increases in air pollutant concentration in the first trimester, in single and two air pollutant models. The model regression coefficients rescaled to IQR increments specify to pollutants were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated.

17

Page 18: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 13: The associations of term birth weight with interquartile range (IQR) increases in air pollutant concentration in the second trimester, in single and two air pollutant models. The model regression coefficients rescaled to IQR increments specify to pollutants were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated.

18

Page 19: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 14: The associations of term birth weight with interquartile range (IQR) increases in air pollutant concentration in the third trimester, in single and two air pollutant models. The model regression coefficients rescaled to IQR increments specify to pollutants were used to assess the overall effect of exposure during the period. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. PM2.5 and PM10 were not entered into the same model together as they were too highly correlated.

19

Page 20: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 15: The associations of term small for gestational age (SGA) with interquartile range (IQR) increases in air pollutant concentration over the entire pregnancy, stratified by potential modifiers (maternal age or neonate’s sex). The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during entire pregnancy in each subgroup. P-values for heterogeneity in the effects in the subgroups were estimated by using the DerSimonian-Laird method. Generalized estimating equation models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, meteorological factors and an interaction term (pollutant exposures × month of conception).

20

Page 21: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 16: The associations of term low birth weight (LBW) with interquartile range (IQR) increases in air pollutant concentration over the entire pregnancy, stratified by potential modifiers (maternal age or neonate’s sex). The odds ratios and corresponding 95% confidence intervals (CIs), rescaled to IQR increments specify to pollutants, were used to assess the overall effect of exposure during entire pregnancy in each subgroup. P-values for heterogeneity in the effects in the subgroups were estimated by using the DerSimonian-Laird method. Generalized estimating equation models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception).

21

Page 22: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 17: Sensitivity analyses for association of air pollutant exposure across the entire pregnancy with the risk of term small for gestational age (SGA) in Guangdong, China. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. The ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’, ‘Sensitivity analysis 6’ and ‘Sensitivity analysis 7’ represents the analysis of variation of parameter value in inverse distance weighting algorithm, the impact of potential misclassification of exposure, the Han nationality, the ambient temperature adjusted as a nonlinear term in the model, the use of 10th percentile as the cutting point to determine term SGA, and the restricted subgroup of births with gestational age < 40 weeks, respectively.

22

Page 23: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 18: Sensitivity analyses for association of air pollutant exposure across the entire pregnancy with the risk of term low birth weight (LBW) in Guangdong, China. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. The ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’ represents the analysis of variation of parameter value in inverse distance weighting algorithm, the impact of potential misclassification of exposure, the Han nationality, the ambient temperature adjusted as a nonlinear term in the model, and the restricted subgroup of births with gestational age < 40 weeks respectively.

23

Page 24: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Supplementary Figure 19: Sensitivity analyses for association of air pollutant exposure across the entire pregnancy with the risk of term birth weight in Guangdong, China. Adjusted models were adjusted for maternal age, maternal ethnicity, maternal cigarette smoking state, mode of delivery, gravidity, year of conception, month of conception, neonate’s sex, gestational age as linear and quadratic terms, meteorological factors and an interaction term (pollutant exposures × month of conception), in addition to including the air pollutant shown above. The ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’ represents the analysis of variation of parameter value in inverse distance weighting algorithm, the impact of potential misclassification of exposure, the Han nationality, the ambient temperature adjusted as a nonlinear term in the model, and the restricted subgroup of births with gestational age < 40 weeks, respectively.

24

Page 25: ars.els-cdn.com · Web viewThe ‘Sensitivity analysis 2’, ‘Sensitivity analysis 3’, ‘Sensitivity analysis 4’, ‘Sensitivity analysis 5’ and ‘Sensitivity analysis 6’

Table S1. Characteristics of the study population and distribution of pregnancy outcomes by city.

City Number Percentage of total birth (%)

Term SGA (%)

Term LBW (%)

Mean term birth weight (g) Girls (%)

Guangzhou 422799 16.47 10.63 2.09 3203.68 47.00Shenzhen 27370 1.07 10.91 1.78 3217.34 46.42Zhuhai 20021 0.78 10.79 2.07 3205.56 47.54Shantou 139202 5.42 16.96 2.37 3152.71 48.62Foshan 12227 0.48 11.20 2.41 3196.71 46.44Shaoguan 85776 3.34 11.54 2.10 3206.09 47.10Zhanjiang 230441 8.98 15.34 2.32 3179.51 46.73Zhaoqing 115796 4.51 15.10 2.64 3144.63 45.57Jiangmen 99124 3.86 13.00 2.42 3167.90 48.16Maoming 210346 8.19 15.28 2.55 3154.01 46.45Huizhou 104035 4.05 13.94 2.23 3161.03 47.04Meizhou 142178 5.54 13.20 2.04 3171.18 47.38Shanwei 90731 3.53 16.14 2.30 3153.05 47.53Heyuan 94476 3.68 14.17 2.40 3157.49 47.14Yangjiang 90285 3.52 13.62 2.11 3173.01 46.80Qingyuan 131516 5.12 13.74 2.42 3162.77 46.23Dongguan 119761 4.66 11.26 2.10 3198.98 46.46Zhongshan 78086 3.04 11.93 2.19 3193.73 47.19Chaozhou 45202 1.76 14.85 2.18 3165.65 47.91Jieyang 219335 8.54 15.50 2.15 3169.48 47.93Yunfu 88750 3.46 15.83 2.86 3132.43 46.79

SGA=small for gestational age; LBW=low birth weight (<2,500 g).

25