Arsenic in drinking water may increase the risk of preterm birth

Preterm birth, when a baby is born before 37 weeks, is a pressing public health problem because babies born early have greater risks of mortality and health complications and later disease in child and adulthood. In 2017 alone, California had more than 400,000 preterm births an increased rate of 8.6%compared to the previous year.

Though we know that genetic factors, exposures to chemicals including air pollutants, and social factors such as and race and poverty can all raise preterm birth risk, there’s a lot we still don’t know about the causes of preterm birth—especially when it comes to environmental and social stressors. We do know that pregnant women in California are exposed to multiple environmental pollutants from air, food, water and consumer products. We also know that pregnant women can experience social stress due to factors such poverty, food insecurity and discrimination.

So in our new study, we set out to investigate the relationship between preterm birth and cumulative burdens of multiple environmental exposure and social stressors. We used a novel integrative big data approach to link two large datasets—1.8 million California birth records with environmental exposure information from CalEnviroScreen, a database that contains thousands of data points on environmental pollution and social factors for every census tract in California. By leveraging large datasets, we were able to reveal new, and surprising, patterns.

We found that arsenic pollution in drinking water is significantly associated with an increase in preterm birth of 1%, by comparing preterm birth in populations with different levels of arsenic contamination in drinking water. This may seem very small, but because there are so many pregnant women exposed to arsenic in their drinking water, it can translate into sizeable population impacts.. In our study of almost 2 million births, most of the pregnant women are living in area with arsenic contamination in the drinking water. For example, if the preterm birth rate for population without arsenic exposure is 7.0%, our study suggests that the same population with arsenic exposure in drinking water will have 1% increase in preterm birth rate (7.07%), which is around 1400 additional preterm births across a population of 2 million because of arsenic in drinking water. This shows how a small increase in risk spread across many people can translate to big impacts.

An additional finding is that there are many people living in area where arsenic concentration in drinking water is higher than regulatory standard. For example, the U.S. EPA’s maximum contaminant levels (MCL) is 10 parts per billion (ppb), which suggests that many people living in those areas are at higher risks of arsenic exposure (shown in figure below).

Source: Figure modified from Huang, et al., 2018.

U.S. EPA adopted this standard for arsenic in drinking water in 2001 with consideration for the various health effects associated with arsenic exposure, including “cancerous effects (skin, bladder, lung, kidney, nasal passages, liver and prostate) and non-cancerous effects: cardiovascular, pulmonary, immunological, neurological and endocrine (e.g., diabetes) effects”. While pregnant women exposed to arsenic in drinking water may have increased risk of preterm birth, the rest of the population may face different risks related to these other diseases.

In addition, we found that both environmental chemical exposures and social stressors such as PM2.5, nitrate in drinking water and neighborhood unemployment rate collectively are associated with increased risk of preterm birth. This study adds to the evidence that cumulative impacts of multiple environmental and social factors are associated with adverse health outcomes. Note that there are some limitations and assumptions made in this study. For example, given the design of the environmental database used, we assume constant pollution during the entire pregnancy, and therefore didn’t account for temporal variability.

The power of our study shows how data collected in California can be used to identify potential risk factors that deserve more attention in research, monitoring and efforts that prevent exposures to harmful pollutant levels in order to better improve prenatal health.

Other co-authors on this work include: Tracey J. Woodruff (UCSF), Rebecca J. Baer (UCSD), Komal Bangia (OEHHA), Laura M. August (OEHHA), Laura L. Jellife-Palowski (UCSF), Amy M. Padula (UCSF, senior author), Marina Sirota (UCSF, senior author).

1+1>2: Evaluating how risks of pollutants and stressors stack up

Constant exposure to environmental pollutants can sicken our bodies, and social stressors such as poverty and psychological burden can further aggravate the health effects. For example, traffic-related air pollution has been linked to the onset of childhood asthma. However, if children exposed to air pollutants also experience violence, their risks of developing asthma can be doubled or tripled. Researchers and Scientists consider ‘cumulative risks’ (details here) as the elevated risks from the combined effects of multiple environmental and social stressors or agents, in this case air pollution and exposure to violence.

Understanding cumulative risk is important because we know very little about how synergistic interactions of multiple chemical exposures and social factors can increase risk of severe diseases, even though we are exposed to numerous chemicals via air, water, soil and consumer products on a daily basis. Until recently, a systematic-review of human and animal evidence confirmed the cumulative adverse effects of prenatal-exposure to chemicals and psychosocial stress on fetal growth. Individuals and different health organizations can take more effective actions to address such health issues once we can gain better knowledge relevant to cumulative risks.

The number of cumulative risk and impact studies increased tremendously over the past decade, but the relevant modeling methods have been underdeveloped to evaluate the joint exposures. Proper selection and use of statistical modeling techniques will generate accurate scientific results, which will contribute to sound environmental and public health policies.

To summarize modeling methods utilized to quantify the cumulative effects of multiple stressors in previous studies, we performed the first ever review on statistical models used to evaluate cumulative risks. I conducted a systematic search to identify original peer-reviewed research articles published between Jan 1, 2012- June 21, 2017 that evaluated both environmental and social stressors, and analyzed their health effects. We focused on human subject studies that provided quantitative method information. Eventually, we identified 31 eligible articles– the majority used simple regression models and focused on air pollutants and socio-economic status (SES) with various health outcomes, as shown in the table below.

Chemical Stressors Measured Social Stressors Measured Health Effects Studied
Air pollutants, drinking water pollutants, climate indicators, metals, silica, BPA SES, race, neighborhood features, psychological factors, physical disorder, material hardship, education, employment, housing, urbanization, neighborhood features, access to health services, health prevention program, violent crime Mortality, morbidity, nutritional status, IQ, cancer risk, child behavior, heart defects, blood lead level, blood pressure, respiratory disease, autoimmune disease, diabetes hospitalization, pregnancy outcomes

We found that simple regression models, including multivariable and logistic regression models, are commonly used in cumulative risk studies, especially evaluating the combined effects of both chemical and non-chemical stressors. However, we wondered whether regression methods can properly cover all the combinations of different possible research settings such as research questions, study designs and data. Unfortunately, no. We found that regression methods, similar to other modeling techniques, has its own advantages and limitations (details here), and concluded that no single modeling technique can be applied universally to all cumulative risk studies.

With increasing knowledge in exposure science and the advent of more quantitative tools in the era of ‘big data’, we recommend that other data mining and machine learning techniques such as deep learning be considered in cumulative risk research, particularly when it comes to understanding the combined risks of multiple stressors. To provide public health protection for vulnerable populations such as pregnant women and children, risk assessment can take advantage of the established modeling methods while more approaches are being developed. Other authors on the study include Drs. Aolin Wang, Rachel Morello-Frosch (UC Berkeley), Juleen Lam, Marina Sirota, Amy Padula and Tracey Woodruff.