Effect of exposure to ambient Particulate Matter (PM2.5) onAdverse Infant Outcomes: An ecological study.

Introduction

Air pollution (including ambient) is a serious rising global concern that claims almost 3.3 million deaths which are caused annually, with an estimated increase of 50% by 2050.^1,2 Majority of the cities in the world have average levels of particulate matters 10 (PM10) and3 2.5 (PM_2.5) above the recommended concentrations. In fact the burden of diseases due to ambient PM_2.5 outnumbers the combined burden of global health threats such as Malaria and Human Immunodeficiency Virus-Acquired Immunodeficiency Syndrome (HIV-AIDS).⁴ With the increasingly urban population of the developing world, the environmental air pollutants have also increased exponentially.^{5, 6} With emissions from the motor vehicles being the main source of ambient air pollution, the vast majority of the urban population breathe polluted air with particulate matter.⁷ The respiratory consequences are the effect of acute exposure to air pollutants; mortality follows over the long term exposure. Children, elderly, and pregnant women are most susceptible population to air pollution.⁸ Exposure to ambient air pollutants during pregnancy can have an adverse effect on neonatal birth outcomes and respiratory health.9 The pollutants inhaled by the pregnant mother during pregnancy can cross the alveoli and placental barrier and act directly on developing fetus; can also affect indirectly through systemic changes in the mother leading to a decreased supply of fetal nutrients and oxygen.⁹ The suggestive mechanism of actions for these changes are oxidative stress,¹⁰ interrupted oxygen supply,^11,12 decreased DNA methylation, and disturbed hemodynamic response.^13-15 Therefore it is possible that exposure to air pollutants perhaps can modulate the immune and inflammatory system leading to epigenetic changes.^16-18 Longer exposure to these pollutants during pregnancy may lead to changes in organ development, resulting in complications in later life.¹⁹ The maternal and child health outcomes in low and middle-income countries (LMICs) are poorer,²⁰ as 98% of the cities in LMICs do not meet ambient air quality guidelines compared to High-income countries (HICs) where 56% of the cities meet ambient air quality.³ Although several studies have explored adverse birth outcomes and ambient air pollution, there exists a relative dearth of studies that draw firm conclusions about causality on the effect of ambient air pollutants in terms of neonatal outcomes.²¹ Thus through our study we aim to provide evidence by assessing the magnitude of ambient pollutants concentration and its impact on low birth weight (LBW), mortality, and respiratory morbidities in children. To achieve this, we adopted an interdisciplinary framework to examine our objectives encompassing maternal and child health, environmental health, and big data analytics.

Hence the objectives of our study were, to estimate the level of PM_2.5 across selected LMICs and Highincome countries (HICs) and to establish/ascertain the predictors of ambient air pollution correlated with adverse infant outcomes. We hypothesized that countries with high air pollution levels have higher LBW, infant mortality, and the incidence of pneumonia - than countries with low air pollution levels.

Materials and Methods

Study design and study population

An ecological study using aggregate data from population-level rather that individuals was conducted using data for the year 2017 from 10 countries: six LMICsand four HICs as seen in Table 1. Countries were classified based on the World Bank Atlas method22 of Gross National Income(GNI) per capita - LMICs: between $1,006 to$3955, and as HICs: $12,235 or more.23 Based on relatively low amounts of data regarding exposure to ambient air pollutants and its effect on neonatal outcome the following countries were finalized for the study: LMICs - India, Bangladesh, El Salvador, Ethiopia, Indonesia, Nepal, Philippines, Uganda and Vietnam; and HICs - United States of America (USA), United Kingdom(UK), United Arab Emirates(UAE), and China. HICs were chosen to compare the pollution data with LMIC countries.

Exposure assessment;

The exposure to air pollutants was collected using an advanced software system and program, retrieved live on an hourly-basis automatically from the Plume Labs Application Program Interface (API). Plume Lab that tracks air pollution and captures metrics on absolute concentration of major harmful atmospheric pollutants over 60 countries across the global. It provides real time information and projections for the next 24 hours as well. However, only 10 countries had data available for PM_2.5 levels. For our study, we built a technical software system that has three subsystems: i) data capturing - to obtain data scheduled from the API; ii) input processing - to process stored data, filters and stores in the Microsoft Azure database, iii) reporting - that aggregates the stored data and generates reports in Microsoft excel format and transmits the reports to the research team through periodic emails. The data of hourly intervals throughout 24 hours in a day for all locations were stored in the Microsoft Azure database (The statistical analysis and visual analytics were made using these reports. Data extraction process began on August 10th, 2017 and continued until October 15th, 2017 - the last day of UN Data Challenge for Climate Change. The time stamp for online data collection was uniform across all countries. Therefore, the analysis of air pollutants included the measurement and analysis of the air pollutants over 56 days.

Outcome assessment;

The primary outcome of interest was low birth weight (LBW) while the secondary outcomes were infant mortality rate (IMR) and the incidence of pneumonia. The recently published reports on the same were used to extract. These included the World Bank data from 2015²⁴ for LBW, and the Global Burden of Diseases (GBD) 2015 data for secondary outcomes.²⁵ The GBD uses different sources to populate the information. Most information comes from the data providers like International Household survey network, Integrated Public Use Micro data Series International, (IPUMS-I), IPUMSUSA, Simple Online Data Archive for Population Studies (SodaPop), Synapse (Sage Bionetwork repository) and Health Data. Gov. The data on infant mortality and incidence of pneumonia have been included from the Global Health Data exchange that includes GDB data.²⁵

Confounders;

Based on available evidence, we included the proportion of households using solid fuels; as this is both related to air pollution levels and adverse health outcome^26-31, smoking history in women was included,^32-37 female literacy was chosen as a confounder as literacy has shown to reduce in LBWincidence and also has impact on bringing down household air pollution,^38-43 proportion of children exclusively breastfed for 6 months,^44,45,46,47 as confounders. The data were extracted from the Health, Nutrition and Population Statistics⁴⁸ for all the countries. The confounders were adjusted during analysis and tested for their effect of interaction on PM_2.5. We also included temperature^49-56 and humidity^57-60 as confounders Weather Data was collected manually from a single source for all locations using a website (www. timeanddate.com).

Statistical analysis;

The data were analyzed using SPSS software version 23. The dependent and independent variables have checked for normality. The descriptive statistics were reported. The association between PM_2.5 and LBW, IMR and incidence of pneumonia were quantified using univariate and Multiple Linear Regression. Before proceeding to the regression analysis the assumptions of linear regression were tested.

Results

The mean PM_2.5 levels of the respective countries are shown in table-1 for the 56 days of the data extraction period. As seen in the table, the mean level of PM_2.5 was highest in Indonesia and Bangladesh followed by Ethiopia and India among the LMICs and in China among HICs. El Salvador (LMIC) and USA (HIC) had minimum levels of PM_2.5

We further conducted univariate and multivariate linear regression analyses to understand the relationship between PM_2.5 and selected adverse infant outcomes. Univariate analysisin Table 2 showed no significant association between PM_2.5 level with LBW and IMR however there was significant association with incidence of pneumonia.Thefactors as female literacy,smoking habits of women, use of solid fuels, and exclusive breastfeeding rate have been adjusted in the multivariate model. Beta coefficient, 95% confidence interval and p-value are reported in table 3. The multivariate linear regression analysis results show no relationship between PM_2.5 levels with LBW and IMR. There was significant association between mean PM_2.5 levels and incidence of pneumonia even after adjusting for confounders.

Discussion

The results of our study that included data from Lower and middle-income countries (LMIC) and high-income countries (HICs) suggests that exposure to the ambient level of PM_2.5 is not associated with low birth weight, infant mortality,with the exception of mean level of PM_2.5 that has shown significant correlation with incidence of pneumonia.

By means of our ecological study that provides a cursory view through country-level data, our results suggests a greater burden of ill health related to environmental pollution. Therefore, we strongly recommend the exploration further through more studies to fill the knowledge gaps regarding the consequences of adverse infant outcomes and exposure to PM_2.5. We advocate the need for prospective cohort studies as a key research approach with good precision and sample size to re-confirm the evidence at individual-level.

For our primary outcome, the results of our study contradicted previous studies across the globe that reported a positive association between PM_2.5 and low birth weight.^61-63 Further studies by Muhammad et al found that exposure to particulate matter is associated with the decreased birth weight,⁶⁴ whileConnieng et al also examined the increased risk for low birth weight with an increase in exposure to PM_2.5. A retrospective cohort done in China showed that exposure to an incremental dose of 10 μg/m3 was associated with the decrease in birth weight by 4.94gm.⁶⁵ The difference in inferences may be due to the inclusion of only a few selected countries (which had the data available on PM_2.5 in the regression analysis and might be due to the population level data and not individual level PM_2.5 data that we have considered to evaluate the relationship with the adverse infant outcome.

Our results with respect to infant mortality show ambient level of PM_2.5 was not associated with Infant mortality. However, there is not enough of available studies to support our evidence on the association between PM_2.5 andneonatal mortality rate (NMR). The majority of the studies are based on post-neonatal mortality and under five mortality. We also found inconsistency with the available, limited sources of evidence. The study done in the United States on estimating the exposure of particulate matter with respect to the death of a cohort of infants observed that particulate matter is associated with increased risk of post-neonatal mortality.₆₆ while two other studies in California showed an increase in 5% risk for respiratory-related deaths among infants due to exposure to the particulate matter;₆₇ and reported the relationship between long-term exposure to PM_2.5 air pollution and post-neonatal infant death in all areas of California.⁶⁸

The role of PM_2.5 with respect to the incidence of pneumonia has been found significant. A study done in China shows increased risk of childhood pneumonia associated with particulate matter;^69-70 while developing countries showed positive association between air pollution and incidence of upper- and lower-respiratory infections in children.⁷¹ and association with hospitalization due to Pneumonia.⁷²

Strength and limitations:

In light of our findings our study had certain several strengths and limitations. One of the major strength of our study is that we have adjusted the possible covariates including health indicators in our analysis. Using the data made available by United Nations climate action team, we were able to offer inputs for further studies and policy support. Through our study, we show that it is possible to address modifiable risk factor such as air pollution using the big data analytics, and thereby, contribute towards at alleviating the ill health in infants and children.

Limitations include: use of country-level data since an ecological study was undertaken for a shorter duration. The data on exposure is the real-time data that we obtained for the year of 2017 but the outcome is extracted from the recently published Global Burden of Diseases data of 2015. We have analyzed the data with an assumption thatthe current level of ambient PM_2.5 will be similar in the year 2015. Hence, the causal inference could not be made. Other than this, the size of the sample we have chosen is small hence study results cannot be generalized. We have not adjusted for smoking status of men within the household and this is a limitation of this study.

Conclusion: The overall Infant health indicators seemed to be very poor in LMICs compared to HICs. In spite of this, the level of pollution is also becoming a major concern. As per the data collected from different capital cities of LMIC countries, it seems that the level of ambient PM_2.5 is higher in lower and middleincome countries. If we glance at the child health indicators we can observe that there are still many lower and middle-income countries with poor infant outcomes.

We have not observed a significant association between ambient PM_2.5 and any of the outcome indicators after controlling for the covariates. Hence, we recommend a future prospective study to confirm the association by assessing the individual exposure during pregnancy and its impact on infant health outcome.

Funding: No funding support received. Dr. Giridhara Babu was accepted to participate in the Data for Climate Action Challenge and was provided with access to data on air pollutants in real time.