Disparities in Environmental Exposures to Endocrine-Disrupting Chemicals and Diabetes Risk in Vulnerable Populations

Abstract

Burgeoning epidemiological, animal, and cellular data link environmental endocrine-disrupting chemicals (EDCs) to metabolic dysfunction. Disproportionate exposure to diabetes-associated EDCs may be an underappreciated contributor to disparities in metabolic disease risk. The burden of diabetes is not uniformly borne by American society; rather, this disease disproportionately affects certain populations, including African Americans, Latinos, and low-income individuals. The purpose of this study was to review the evidence linking unequal exposures to EDCs with racial, ethnic, and socioeconomic diabetes disparities in the U.S.; discuss social forces promoting these disparities; and explore potential interventions. Articles examining the links between chemical exposures and metabolic disease were extracted from the U.S. National Library of Medicine for the period of 1966 to 3 December 2016. EDCs associated with diabetes in the literature were then searched for evidence of racial, ethnic, and socioeconomic exposure disparities. Among Latinos, African Americans, and low-income individuals, numerous studies have reported significantly higher exposures to diabetogenic EDCs, including polychlorinated biphenyls, organochlorine pesticides, multiple chemical constituents of air pollution, bisphenol A, and phthalates. This review reveals that unequal exposure to EDCs may be a novel contributor to diabetes disparities. Efforts to reduce the individual and societal burden of diabetes should include educating clinicians on environmental exposures that may increase disease risk, strategies to reduce those exposures, and social policies to address environmental inequality as a novel source of diabetes disparities.

Introduction

Diabetes is a complex and devastating metabolic disease that arises from impairments in insulin production and/or action with consequential derangements in global energy metabolism. In the U.S., diabetes is the leading cause of adult blindness, kidney failure, and nontraumatic amputations; moreover, it is a central driver of cardiovascular disease, the leading cause of death among people with diabetes. Diabetes disproportionately afflicts African Americans, Latinos, and low-income individuals. Compared with non-Hispanic whites, the risk of developing diabetes is estimated to be 66% higher for Hispanics and 77% higher for African Americans (1). Indeed, 17.9% of African Americans and 20.5% of Mexican Americans have diabetes compared with only 9.1% of non-Hispanic whites, and these disparities in diabetes prevalence have been amplified over the past decade (2). Furthermore, age-adjusted diabetes mortality rates are significantly higher among Hispanics and non-Hispanic blacks than non-Hispanic whites (3). Understanding the complete array of factors that contribute to racial and ethnic differences in the pathogenesis of metabolic disease is critical for addressing the disproportionate burden of diabetes in communities of color.

Although physical inactivity and caloric excess undoubtedly are risk factors, emerging evidence implicates environmental endocrine-disrupting chemicals (EDCs) as contributors to the diabetes epidemic. The Endocrine Society defines an EDC as an exogenous chemical or mixture of chemicals that interferes with any aspect of hormone action (4). Of note, the dramatic rise in U.S. diabetes rates correlates closely with synthetic chemical production (5), and these associations are now supported by epidemiological, animal, and cellular data that demonstrate that EDCs can interfere with insulin secretion and action as well as with other pathways that regulate glucose homeostasis. Despite a history of environmental pollution disproportionately affecting communities of color in the U.S. (6), the potential contribution of environmental toxicants to racial and ethnic differences in diabetes risk is underappreciated.

The issue of environmental injustice first entered widespread consciousness in 1982 when residents of the predominantly African American community of Warren County, North Carolina, made national news by laying themselves across a rural road to prevent encroaching trucks from dumping dirt laden with polychlorinated biphenyls (PCBs) in their community (6). This media attention prompted empirical examination of the community’s claim that toxic waste facilities were being disproportionately sited in low-income communities and communities of color. This research has grown substantially since the 1980s, with the majority of evidence showing racial and socioeconomic disparities in exposures to myriad environmental hazards (6). In addition to higher exposure to air pollution nationwide (7), unequal exposures among people of color are also rooted in patterns of occupation, housing conditions, and neighborhood infrastructure (8,9). This article reviews the state of the evidence linking ethnic, racial, and socioeconomic disparities in pollutant exposure in the U.S. to EDCs linked to diabetes.

Unequal Environmental Exposures and Diabetes Risk

Scientific evidence linking EDCs with the development of diabetes and other metabolic disorders continues to grow. Of note, exposures to several toxicants have been prospectively linked to diabetes risk, including PCBs, organochlorine (OC) pesticides, various chemical constituents of air pollution, bisphenol A (BPA), and phthalates (Table 1); moreover, exposure to these EDCs is higher among African Americans, Latinos, and low-income individuals (Supplementary Table 1). These unequal exposures raise the possibility that EDCs are underappreciated contributors to diabetes disparities.

Table 1.

Prospective studies documenting associations between EDC exposure and diabetes risk

Reference	Population	Outcome and comparison	Effect estimate (95% CI)
PCBs
Vasiliu et al., 2006 (21)	1,384 subjects without diabetes in the Michigan polybrominated biphenyls cohort followed for 25 years	Incident diabetes in women with the highest vs. lowest serum PCB levels	IDR: 2.33 (1.25–4.34)*
Wang et al., 2008 (22)	378 subjects and 370 matched referents from the Yucheng poisoning in Taiwan in the 1970s	Incident diabetes in women who consumed rice bran oil laced with PCBs as well as a subgroup who developed chloracne, a manifestation of dioxin-like PCB exposure	OR: 2.1 (1.1–4.5)*; Chloracne OR: 5.5 (2.1–13.4)*
Turyk et al., 2009 (27)	471 Great Lakes sport fish consumers without diabetes followed from 1994/1995 to 2005	Incident diabetes among the highest vs. lowest tertile of PCB levels	Total PCBs IRR: 1.8 (0.6–5.0); PCB 118 IRR: 1.3 (0.5–3.0)
Wu et al., 2013 (23)	Two case-control studies of women without diabetes from the NHS and a meta-analysis of pooled data with six additional prospective studies	Incident diabetes after pooling of data and comparing highest PCB exposure group with the referent	Pooled OR: 1.70 (1.28–2.27)*
Rignell-Hydbom et al., 2009 (26)	Case-control study of women age 50–59 years in southern Sweden	Incident diabetes in 39 patients and matched control subjects after ≥6 years of follow-up comparing the highest quartile of PCB levels with the referent	OR: 1.6 (0.61–4.0)
Lee et al., 2010 (25)	90 patients and control subjects in a nested case-control study followed for ∼18 years	Incident diabetes comparing second sextile or quartile with the referent for a summary measure of 16 POPs, including 12 PCBs as well as individual PCBs	PCB sum OR: 5.3*; PCB 187 OR: 2.8 (1.1–7.4)*
Lee et al., 2011 (24)	725 participants from the PIVUS study	Incident diabetes comparing a summary measure of 14 PCBs across quintiles with the referent	Quintile 2 OR: 4.5 (0.9–23.5); Quintile 3 OR: 5.1 (1.0–26.0); Quintile 4 OR: 8.8 (1.8–42.7)*; Quintile 5 OR: 7.5 (1.4–38.8)*; Ptrend < 0.01
Song et al., 2016 (29)	Meta-analysis of 13 cross-sectional and 8 prospective studies published before 8 March 2014 examining links between PCBs and diabetes risk	Pooled diabetes risk in the highest vs. lowest exposure groups for PCBs	RR: 2.39 (1.86–3.08)*
OC pesticides
Wu et al., 2013 (23)	Two case-control studies of women without diabetes from the NHS and a meta-analysis of pooled data with six additional prospective studies	Incident diabetes comparing highest tertile of plasma HCB levels in NHS and highest to lowest exposure group in pooled prospective studies	NHS OR: 3.14 (1.28–7.67)*; Pooled OR: 2.00 (1.13–3.53)*
Turyk et al., 2009 (27)	471 Great Lakes sport fish consumers without diabetes followed from 1994/1995 to 2005	Incident diabetes comparing tertiles of serum DDE levels with the referent	Tertile 2 IRR: 5.5 (1.2–25.1)*; Tertile 3 IRR: 7.1 (1.6–31.9)*
Rignell-Hydbom et al., 2009 (26)	Case-control study of women age 50–59 years in southern Sweden	Incident diabetes in 39 patients and matched control subjects after ≥6 years of follow-up comparing the highest quartile of DDE levels with the referent	OR: 5.5 (1.2–25)*
Lee et al., 2010 (25)	90 patients and control subjects in a nested case-control study followed for ∼18 years	Incident diabetes comparing second sextile with the referent for summary measure of 16 POPs (including 3 OC pesticides) or second quartile with the referent for trans-nonachlor	Sum OR: 5.4 (1.6–18.4)*; Trans-nonachlor OR: 4.3 (1.5–12.6)*
Lee et al., 2011 (24)	725 participants from the PIVUS study	Incident diabetes comparing quintiles of OC pesticides or summary measure of three OC pesticides with the referent	Quintile 5 sum OC pesticides OR: 3.4 (1.0–11.7); Ptrend = 0.03; Quintile 4 trans-nonachlor OR: 4.2 (1.3–13.3)*; Ptrend = 0.03
Van Larebeke et al., 2015 (28)	973 participants of the Flemish Environment and Health Survey	Risk of incident diabetes calculated for a doubling of serum or comparing 90th percentile with 10th percentile of levels for HCB (men and women) or DDE (men only)	Doubled HCB OR: 1.61 (1.07–2.42)*; 90th vs. 10th percentile HCB OR: 6.27*; Doubled DDE OR: 1.66 (1.09–2.53)*; 90th vs. 10th percentile DDE OR: 5.39*
Starling et al., 2014 (35)	13,637 women from the Agricultural Health Study	Incident diabetes for ever use of the OC pesticide dieldrin	HR: 1.99 (1.12–3.54)*
Song et al., 2016 (29)	Meta-analysis of 11 cross-sectional and 6 prospective studies published before 8 March 2014 examining links among various pesticides and diabetes risk	Pooled diabetes risk in the highest vs. lowest exposure groups for pesticides	RR: 2.30 (1.81–2.93)*
Chemical constituents of air pollution
Brook et al., 2016 (41)	65 adults with metabolic syndrome and insulin resistance from the Air Pollution and Cardiometabolic Diseases China Study	Change in HOMA-IR per SD increase in personal-level black carbon or PM2.5 exposure during the fourth and fifth days of assessment	Day 4 black carbon: 0.18 (0.01–0.36)*; Day 5 black carbon: 0.22 (0.04–0.39)*; Day 4 PM2.5: 0.18 (0.02–0.34)*; Day 5 pm2.5: 0.22 (0.08–0.36)*
Weinmayr et al., 2015 (43)	3,607 individuals from the Heinz Nixdorf Recall Study in Germany followed for an average of 5.1 years	Incident diabetes relative IQR increase in PM10, PM2.5, traffic-specific PM10, and traffic-specific PM2.5 as well as comparison of living <100 m vs. 200 m from a busy road	PM10 RR: 1.20 (1.01–1.42)*; PM2.5 RR: 1.08 (0.89–1.29); Traffic PM10 RR: 1.11 (0.99–1.23); Traffic PM2.5 RR: 1.10 (0.99–1.23); <100 m RR: 1.37 (1.04–1.81)*
To et al., 2015 (93)	29,549 women from the Canadian National Breast Screening Study	Change in diabetes prevalence per 10 μg/m3 increase in PM2.5 exposure	PRR: 1.28 (1.16–1.41)*
Park et al., 2015 (46)	5,839 subjects in the Multi-Ethnic Study of Atherosclerosis cohort	Prevalent and incident diabetes risk per IQR increase in residential concentrations of PM2.5 or NOx	Prevalent DM PM2.5 OR: 1.09 (1.00–1.17)*; Prevalent DM NOx OR: 1.18 (1.01–1.38)*; Incident DM PM2.5 HR: 1.02 (0.95–1.10); Incident DM NOx HR: 1.00 (0.86–1.16)
Pope et al., 2015 (48)	669,046 participants from the American Cancer Society Cancer Prevention Study II	Risk of diabetes-associated death on death certificates per 10 μg/m3 increase in PM2.5	HR: 1.13 (1.02–1.26)*
Brook et al., 2013 (49)	2.1 million adults from the 1991 Canadian Census Mortality Follow-up Study	Diabetes-associated mortality per 10 μ/m3 increase in PM2.5	HR: 1.49 (1.37–1.62)*
Thiering et al., 2013 (42)	Fasting blood from 397 10-year-old children in two prospective German birth cohort studies	Change in HOMA-IR per 2-SD increase in ambient NO2 and PM10 and for every 500 m to nearest major road	NO2: 17.0% (5.0–30.3%)*; PM10: 18.7% (2.9–36.9%)*; Road proximity: 7.2% (0.8–14.0%)*
Coogan et al., 2012 (44)	3,992 black women living in Los Angeles followed for 10 years	Incident diabetes per IQR increase in NOx or 10 μg/m3 increase in PM2.5	NOx IRR: 1.25 (1.07–1.46)*; PM2.5 IRR: 1.63 (0.78–3.44)
Coogan et al., 2016 (47)	43,003 participants in the Black Women’s Health Study followed from 1995 to 2011	Incident diabetes per IQR increase in NO2 by using both land use repression and dispersion models	Land use HR: 0.96 (0.88–1.06); Dispersion HR: 0.94 (0.80–1.10)
Krämer et al., 2010 (45)	1,775 women without diabetes age 54–55 followed for 16 years in West Germany	Incident diabetes per IQR increase in exposure on the basis of data from monitoring stations, emission inventories, or land use regression models as well as distance from busy road (<100 m) relative to education status	Monitored PM10 HR: 1.16 (0.81–1.65); Monitored NO2 HR: 1.34 (1.02–1.76)*; Inventory traffic PM HR: 1.15 (1.04–1.27)*; Inventory traffic NO2 HR: 1.15 (1.04–1.27)*; Land use soot HR: 1.27 (1.09–1.48)*; Land use NO2 HR: 1.42 (1.16–1.73)*; <100 m/low education HR: 2.54 (1.31–4.91)*; <100 m/high education HR: 0.92 (0.58–1.47)
Schneider et al., 2008 (50)	22 people with type 2 diabetes living in North Carolina	Changes in vascular parameters per 10 μ/m3 increase in PM2.5 accounting for lag period (in days)	Lag 0 FMD: −17.3 (−34.6 to 0.0)*; Lag 1 SAEI: −17.0 (−27.5 to −6.4)*; Lag 3 SAEI: −15.1 (−29.3 to −0.9)*
O’Donnell et al., 2011 (51)	9,202 patients hospitalized with ischemic stroke	Risk of ischemic stroke among patients with diabetes per 10 μg/m3 increase in PM2.5	11% (1–22%)*
Brook et al., 2013 (40)	25 healthy adults from rural Michigan brought to an urban location for 5 consecutive days	Change in HOMA-IR per 10 μg/m3 increase in PM2.5	0.7 (0.1–1.3)*
BPA
Sun et al., 2014 (60)	971 incident type 2 diabetes case-control pairs from the NHS and NHS II	Incident diabetes after adjusting for BMI comparing highest with the referent quartile of urinary BPA levels	NHS OR: 0.81 (0.48–1.38); NHS II OR: 2.08 (1.17–3.69)*
Bi et al., 2016 (61)	2,209 middle-aged and elderly subjects without diabetes followed for 4 years	Incident diabetes risk in highest quartile vs. lowest quartile of urinary BPA level for each 10-point increase in a diabetes genetic risk score	OR: 1.89 (1.31–2.72)*
Hu et al., 2015 (62)	121 patients with type 2 diabetes followed for 6 years	Incident chronic kidney disease in patients with diabetes comparing highest with referent tertile of urinary BPA level	OR: 6.65 (1.47–30.04)*
Song et al., 2016 (29)	Meta-analysis of five cross-sectional and prospective studies published before 8 March 2014 examining links between BPA and diabetes risk	Pooled diabetes risk in the highest vs. lowest exposure groups for BPA	RR: 1.45 (1.13–1.87)*
Phthalates
Sun et al., 2014 (60)	971 incident type 2 diabetes case-control pairs from the NHS and NHS II	Incident diabetes after adjusting for BMI comparing highest with the referent quartile of urinary phthalate levels	NHS DEHP OR: 1.34 (0.77–2.30); NHS butyl phthalates OR: 0.91 (0.50–1.68); NHS total phthalates OR: 0.87 (0.49–1.53); NHS II DEHP OR: 1.91 (1.04–3.49)*; NHS II butyl phthalates OR: 3.16 (1.68–5.95)*; NHS II total phthalates OR: 2.14 (1.19–3.85)*
Watkins et al., 2016 (70)	250 children of women enrolled in the Early Life Exposure in Mexico to Environmental Toxicants cohort	Change in insulin secretion as assessed by a C-peptide index per IQR increase in either in utero MEP levels for pubertal boys or peripubertal DEHP for prepubertal girls	Pubertal boys: −17% (−29 to −3.3%)*; Prepubertal girls: 20% (2.5–41%)*
Song et al., 2016 (29)	Meta-analysis of four cross-sectional and prospective studies published before 8 March 2014 examining links between phthalates and diabetes risk	Pooled diabetes risk in the highest vs. lowest exposure groups for BPA	RR: 1.48 (0.98–2.25)

Data are from studies from around the world (Supplementary Fig. 1). DM, diabetes; FMD, flow-mediated dilatation; IDR, incidence density ratio; PRR, prevalence rate ratio; SAEI, small-artery elasticity index.

*P < 0.05.

PCBs

Introduced in the U.S. in the 1930s for a variety of industrial purposes, PCBs are a class of synthetic compounds where various combinations of hydrogen atoms on the biphenyl (C₁₂H₁₀) structure are substituted with chlorine, resulting in 209 congeners designated by a unique number reflecting the extent and position of their chlorination (e.g., PCB 153). Although banned by the U.S. Environmental Protection Agency in 1977, PCBs remain detectable in human tissues as a result of their environmental and biological persistence (10). Higher PCB exposures among African Americans have been documented since the 1960s (11) (Supplementary Table 1). Ongoing human exposure to PCBs is due to the legacy of contamination in food, including certain fish (12); however, additional exposure sources include leaching from contaminated industrial sites and indoor construction materials (13,14). PCB waste is found in Superfund and toxic waste sites that are concentrated in neighborhoods of color (6). Although catfish consumption has been suggested as the main contributor to increased PCB levels in African Americans (15), the historical siting of PCB production and disposal sites in predominantly black communities is likely a significant additional contributor to increased contamination of locally sourced foods. One example of this phenomenon is Anniston, Alabama, a PCB manufacturing city from 1929 to 1971. African Americans not only lived closer to a former Monsanto PCB manufacturing plant but also had PCB levels three times higher than whites living in Anniston (16). Consumption of local fish and livestock were the strongest predictors of higher serum PCB levels among African Americans (17), whereas consumption of local dairy products and dredging near another PCB-contaminated Superfund site also predicted higher cord blood PCB levels in infants (18).

A large body of evidence, including prospective epidemiological studies, supports the hypothesis that PCBs are metabolic disease risk factors. For example, residential proximity to PCB-contaminated waste sites is associated with higher diabetes hospitalization rates (19). Among female residents of Anniston, serum PCB levels were significantly associated with diabetes (20), whereas in a separate study with 25 years of follow-up, women with higher PCB levels exhibited increased diabetes incidence (incidence density ratio 2.33 [95% CI 1.25–4.34]) (21). Similarly, women exposed to PCB-laced rice bran oil during the Yucheng poisoning event in Taiwan also had an increased risk of developing diabetes (odds ratio [OR] 2.1 [95% CI 1.1–4.5]), with markedly higher risk among those who developed chloracne, a cutaneous manifestation of dioxin-like PCB exposure (OR 5.5 [95% CI 2.1–13.4]) (22). A meta-analysis that pooled data from the Nurses’ Health Study (NHS) with six prospective studies showed that total PCBs were associated with incident diabetes (OR 1.70 [95% CI 1.28–2.27]) (23). Further supporting these prospective links between PCB exposure and diabetes are data from cohort studies, including the Prospective Investigation of the Vasculature in Uppsala Seniors (PIVUS) (24) and a group followed for nearly 20 years (25). Finally, although not reaching statistical significance, a study of Swedish women suggested that higher levels of PCB 153 were similarly associated with increased rates of type 2 diabetes diagnosed after >6 years of follow-up (OR 1.6 [95% CI 0.61–4.0]) (26). Collectively, these data suggest an association between PCBs and diabetes risk, especially among women; however, some discrepant findings exist in the literature. In a study of Great Lakes sport fish consumers, PCB 118 and total PCBs were not associated with diabetes (27), and in a Flemish study that adjusted for correlated exposures, PCBs showed a negative association with self-reported diabetes (28). Despite these discrepancies, a meta-analysis of both cross-sectional and prospective studies published before March 2014 showed that in aggregate, total PCBs are associated with increased diabetes risk (relative risk [RR] 2.39 [95% CI 1.86–3.08]) (29). Taken within the context of animal and cellular data demonstrating that PCBs alter metabolic function (Supplementary Table 2), this evidence collectively suggests that PCBs contribute to diabetes risk and disparities.

OC Pesticides

OC pesticides were extensively used in the U.S. until the 1970s when most were banned because of their environmental persistence and toxicity to humans and wildlife; however, several OC pesticides and their metabolites are still measurable in the U.S. population. Levels of these compounds are greater in Mexican Americans and African Americans compared with whites (Supplementary Table 1). The prolonged use of OC pesticides outside of the U.S. for agricultural purposes or vector control is believed to contribute to higher levels in Latino populations (10,30,31). Indeed, on the basis of National Health and Nutrition Examination Survey data, people born outside the U.S. are more likely to be exposed to OC pesticides (32). However, the overrepresentation of Mexican Americans in U.S. agriculture may also play a role in exposure disparities (33). In addition, direct exposures to OC pesticides before their phaseout may have been passed down to offspring through breast milk and cord blood (34), likely resulting in higher body burdens at the start of life that persist into adulthood.

In concordance with animal and cellular data demonstrating the capacity of OC pesticides to disrupt multiple aspects of cellular and systemic glucose regulation (Supplementary Table 2), epidemiological studies from various regions of the world have associated OC pesticide exposure with diabetes and the metabolic syndrome (Table 1). For example, plasma hexachlorobenzene (HCB) was positively associated with incident type 2 diabetes, an effect confirmed in an accompanying meta-analysis (OR 2.00 [95% CI 1.13–3.53]) (23). Among Great Lakes sport fish consumers, levels of dichlorodiphenyldichloroethylene (DDE), a metabolite of dichlordiphenyltrichloroethane (DDT), were associated with incident diabetes (27), whereas a study in Swedish women showed that being in the highest quartile of DDE levels relative to the lowest quartile was associated with incident diabetes (OR 5.5 [95% CI 1.2–25]) (26). In a nested case-control cohort of individuals followed for nearly 20 years, the OC pesticides trans-nonachlor, oxychlordane, and mirex were nonlinearly associated with new-onset diabetes (25). In the PIVUS study, trans-nonachlor and a summary index of three OC pesticides also were positively associated with diabetes at age 75 years (24). In a Flemish biomonitoring program, OC pesticide levels measured in 2004–2005 were associated with self-reported diabetes in 2011; this included HCB as well as DDE in men (28). Finally, in the Agricultural Health Study, a large prospective cohort of pesticide applicators and their spouses, the OC pesticide dieldrin was associated with incident diabetes (hazard ratio [HR] 1.99 [95% CI 1.12–3.54]) (35). By compiling data across studies, a meta-analysis comparing the highest to lowest exposure groups demonstrated a strong positive correlation between OC pesticide exposure and diabetes rates (RR 2.30 [95% CI 1.81–2.93]) (29).

Traffic-Related Air Pollution and Particulate Matter

Traffic-related air pollution comprises various chemical components, including nitric oxides (NO_x), ozone, and particulate matter (PM), which is a mixture of particles and liquids classified by their diameter (e.g., <10 μm [PM₁₀] or <2.5 μm [PM_2.5]). Nationwide studies have shown that African Americans and Latinos are exposed to significantly more PM_2.5 (7,36), and ethnic and racial disparities in exposure to traffic-related air pollution exceed those between income groups (37) (Supplementary Table 1). NO₂ levels correlate closely with PM_2.5, ultrafine particles, and black carbon and thus serve as a proxy for traffic-related air pollution (38). Exposure to NO₂ is 38% higher for people of color than for non-Hispanic whites and 10% higher for people below the poverty line (37). Among nonwhite individuals living in poverty, children age <5 years are exposed to 23% higher NO₂ concentrations than the rest of the population (37). Of note, racial differences in NO₂ exposure are greater in large metropolitan centers compared with small-to-medium urban areas, likely reflecting racial and ethnic segregation around traffic corridors in major U.S. cities.

Increasing evidence implicates air pollution in glucose dysregulation, including insulin resistance (39) (Table 1). In a small, but elegant study of residents of rural Michigan, exposure to urban air for only 4–5 h daily for 5 consecutive days increased HOMA insulin resistance (HOMA-IR) for each 10 μg/m³ increase in PM_2.5 (40). Similarly, in adults with the metabolic syndrome living in the Beijing metropolitan area, variations in black carbon and PM_2.5 have been associated with worsening insulin resistance (41). In Germany, long-term exposure to PM₁₀ and NO₂ was associated with greater insulin resistance in 10-year-old children (42). In addition, several studies have linked poor air quality with progression to diabetes. In one study of individuals without diabetes followed for 5.1 years, each interquartile range (IQR) increase in total PM₁₀ was associated with a 20% increased risk of developing type 2 diabetes (RR 1.20 [95% CI 1.01–1.42]) (43). Living <100 m (relative to >200 m) from a busy road was associated with a 37% increased risk of developing diabetes (RR 1.37 [95% CI 1.04–1.81]). Furthermore, higher levels of PM_2.5, traffic-specific PM₁₀, and traffic-specific PM_2.5 were associated with increased diabetes risk; however, these measures failed to reach statistical significance. In a study of black women living in Los Angeles, California, followed for 10 years, incident diabetes rates were increased for each IQR increase in NO_x (incidence rate ratio [IRR] 1.25 [95% CI 1.07–1.46]), whereas PM_2.5 was associated with a nonsignificant increase in incident diabetes (IRR 1.63 [95% CI 0.78–3.44]) (44). Among woman without diabetes from the Study of the Influence of Air Pollution on Lung, Inflammation, and Aging cohort followed for 16 years, incident diabetes increased by 15–42% per IQR of PM₁₀ or traffic-related air pollution (45). The data from prospective studies are not, however, uniform. In the Multi-Ethnic Study of Atherosclerosis, NO_x was associated with prevalent diabetes, and PM_2.5 trended toward an association (OR 1.09 [95% CI 1.00–1.17]), but no air pollution measure was associated with incident diabetes over 9 years of follow-up (46). In a long-term analysis of the Black Women’s Health Study with adjustment for multiple metabolic stressors, NO₂ was not associated with diabetes incidence (47). Despite this heterogeneity, epidemiological studies linking various chemical constituents of air pollution to diabetes risk coupled with animal studies demonstrating that exposures to air pollutants such as PM_2.5 and polyaromatic hydrocarbons disrupt metabolism and promote inflammation (Supplementary Table 2) suggest that differential exposure to air pollution may augment diabetes risk in low-income communities of color.

In addition to effects on diabetes development per se, air pollutants may also promote adverse outcomes in those with the disease. For example, PM_2.5 levels modeled for home addresses were linked to diabetes on death certificates (48), whereas a prospective analysis of >2 million adults revealed that a 10 μg/m³ increase in PM_2.5 was associated with increased diabetes-related mortality (49). These findings may be related to adverse vascular effects in individuals with diabetes. In 22 patients with type 2 diabetes living in North Carolina, daily measures of flow-mediated vasodilatation were decreased in association with PM_2.5 levels (50). The clinical significance of this finding may be reflected in data showing that each 10 μg/m³ increase in PM_2.5 was associated with an 11% increased risk of ischemic stroke in individuals with diabetes (51).

BPA

BPA is a ubiquitous synthetic chemical used in the manufacturing of polycarbonate and other plastics commonly used in consumer products; moreover, BPA is a component of sales receipts and epoxy resins lining food and beverage cans as well as water pipes. BPA exposure in the U.S. population is nearly universal (52). Although BPA is rapidly cleared from the body and single measurements may not reflect cumulative exposure (53), African Americans and people with lower incomes have higher BPA levels than the population at large (Supplementary Table 1). The reasons for these disparities are not clear, but reduced access to fresh food and consequential consumption of processed foods may partly explain these associations (54) because consuming foods packaged in plastic or cans increases BPA exposure (55). Moreover, among individuals with low food security, BPA levels are higher if they receive emergency food assistance, which includes canned foods (56). For example, 6–11-year-old children receiving emergency food assistance had BPA levels that were 54% higher than age-matched children from more affluent families (56).

Disparities in BPA exposure may contribute to metabolic disease burden because increasing evidence associates BPA with diabetes. Analyses that explore the association between urinary BPA levels and metabolic disease are complicated by BPA’s rapid excretion (57); moreover, although no definitive evidence exists that urinary excretion of BPA is influenced by race/ethnicity, lack of adjustment for renal function can complicate urinary assessments in population studies (58). Despite these caveats, the National Toxicology Program concluded that BPA could exert effects on glucose homeostasis and insulin release on the basis of animal and in vitro studies (59). Although some heterogeneity exists across studies, the literature that supports this conclusion demonstrates myriad BPA-induced metabolic disruptions across multiple animal and cellular model systems, including alterations in body weight regulation, insulin action, and insulin secretion as well as specific disruptions in β-cell, α-cell, hepatocyte, and adipocyte function and development (Supplementary Table 2). This conclusion is further supported by limited prospective human studies (Table 1). In data from the NHS, extremes of BPA quartiles were associated with incident diabetes after adjusting for BMI (OR 2.08 [95% CI 1.17–3.69]) in NHS II but not NHS (60), suggesting that age modifies BPA-associated diabetes risk because the mean age in NHS II was 45.6 years versus 65.6 years in NHS. Alternatively, these differences may have arisen from period-cohort effects in which the extent, diversity, or timing of exposures may have been greater or more deleterious in NHS II. Furthermore, evidence that the BPA-diabetes association is modified by a diabetes genetic risk score (61) suggests that some populations are more sensitive to the diabetogenic effects of BPA. Of note, BPA may exacerbate diabetes complications because high levels of BPA have been associated with a markedly increased rate of developing chronic kidney disease (OR 6.65 [95% CI 1.47–30.04]) (62). In one meta-analysis that aggregated cross-sectional and prospective studies, a comparison of the highest to lowest exposure groups demonstrated a positive association between BPA and diabetes (RR 1.45 [95% CI 1.13–1.87]) (29), a finding similar to a second, more recent meta-analysis of prevalent diabetes in three cross-sectional studies (OR 1.47 [95% CI 1.21–1.80]) (63). Thus, on the basis of reasonable evidence, differential BPA exposure may promote diabetes disparities.

Phthalates

Phthalates are a diverse class of widely used synthetic compounds. High–molecular weight (HMW) phthalates are mainly used as plasticizers in food packaging, toys, and building materials, such as polyvinyl chloride (PVC); low–molecular weight phthalates are used in pharmaceuticals, personal care products, and solvents. Phthalates are not covalently bound within products and, thus, can volatilize or leach out, thereby facilitating absorption through dermal contact, ingestion, and inhalation. Higher phthalate exposures among people of color and people with low income have been documented in various studies (Supplementary Table 1), although the sources of these exposure differences are difficult to discern given the widespread commercial use of phthalates. Reduced access to fresh fruits and vegetables and increased consumption of fat-rich foods in low-income populations may augment exposure differences because certain high-fat foods are a major source of HMW phthalates (64). Weathering of older construction materials in low-income households may increase inhalational phthalate exposure (65). Furthermore, purchasing inexpensive products likely contributes to disproportionate phthalate exposures according to an evaluation of products at dollar stores that revealed that 32% of PVC-containing products exceed phthalate limits established for children’s products by the Consumer Product Safety Commission (66). Of note, personal care products and cosmetics also contribute to phthalate exposure (67), especially in women, who typically have the highest concentrations of phthalates (68). Indeed, certain feminine hygiene products were found to be at least partially responsible for higher levels of monoethyl phthalate (MEP) in African American women (69). These data provide provocative evidence of racial, ethnic, and socioeconomic disparities in phthalate exposure; however, additional studies are needed to further illuminate the sources of these differences.

Several epidemiological studies have linked higher phthalate exposure with diabetes (Table 1). In data from NHS II, total urinary phthalate metabolites were associated with diabetes (60). In this analysis, metabolites of butyl phthalates and diethylhexyl phthalate (DEHP) were associated with diabetes (OR 3.16 [95% CI 1.68–5.95] and 1.91 [95% CI 1.04–3.49], respectively). Similar to BPA, these associations may be age-related or a consequence of period-cohort effects because similar findings were not observed with the older, original NHS. In the Early Life Exposures in Mexico to Environmental Toxicants cohort, in utero levels of MEP were associated with reduced insulin secretion in pubertal boys (70). In the meta-analysis of Song et al. (29), urinary concentrations of phthalates were nearly significantly associated with diabetes (RR 1.48 [95% CI 0.98–2.25]). With supportive cellular and animal data demonstrating that various phthalates have the capacity to promote dysfunction in multiple metabolic tissues (Supplementary Table 2), further prospective studies are justified to define the relationship between phthalate exposures and diabetes risk, particularly among vulnerable populations.

Linking Environmental Exposures to Diabetes Risk in Vulnerable Populations

Most studies examining links between EDCs and diabetes have done so without consideration of race, ethnicity, or socioeconomic status; however, recent reports have begun to interrogate these important interactions. In a cross-sectional study investigating the associations between phthalates and insulin resistance, an interaction with race demonstrated that Mexican American (P = 0.001) and non-Hispanic black adolescents (P = 0.002) had significant increments in HOMA-IR with higher levels of HMW phthalates or DEHP that were not observed in non-Hispanic whites (P ≥ 0.74) (71). Similarly, in stratified models, HMW phthalates and DEHP were more strongly associated with HOMA-IR in adolescents from households with lower income. In another cross-sectional study, phthalate levels were positively associated with fasting blood glucose, fasting insulin, or HOMA-IR; however, the dose-response relationship was stronger among African Americans and Mexican Americans than among whites (72). In the meta-analysis of Song et al. (29), the impact of PCBs on diabetes risk was higher in nonwhite populations (RR 2.91 [95% CI 1.60–5.30]) compared with their white counterparts (RR 1.94 [95% CI 1.42–2.62]); similarly, associations between OC pesticides and diabetes were stronger in nonwhites (RR 2.64 [95% CI 1.56–4.49]) than in whites (RR 1.95 [95% CI 1.40–2.71]). Although these associations are likely partially attributable to higher EDC exposures, these findings also suggest that African Americans and Latinos have heightened sensitivity to the diabetogenic effects of environmental contaminants because of potential synergy with other diabetes risk factors.

The current literature provides evidence that five classes of environmental toxicants are linked to diabetes risk in humans, and for each, vulnerable populations are disproportionately exposed. The strength of epidemiological evidence for these five classes varies, with the most consistent findings observed for the persistent pollutants (PCBs and OCs), likely reflecting their long biological half-lives, the stability of their quantitation, and the longer time they have been studied. Among all five classes, however, is provocative evidence of diabetes-promoting effects as well as disparities in exposure. Thus, although further study is required, the unequal burden of environmental risk factors in ethnically, racially, and economically segregated neighborhoods of color may contribute to interethnic differences in metabolic health.

Origins of Differential Environmental Exposures

Addressing disparities in environmental health necessitates understanding the sociological forces that shape society. Segregation profoundly influences individual socioeconomic status, reinforces unhealthy neighborhood environments, and modifies individual behaviors (73), all of which influence metabolic disease susceptibility. Reduced access to affordable healthy foods, as seen in many African American and Latino neighborhoods, promotes unhealthy eating habits (54), whereas lack of safety and reduced access to green space can limit physical activity (74). Thus, the built environment in many communities of color potentiates two key drivers of diabetes risk, namely diet and exercise.

In addition, historical economic and political racialization of residential areas and the labor force has promoted today’s racial segregation and the codecline of environmental health in these neighborhoods (75). Indeed, living in highly segregated metropolitan areas is associated with a greater health risk from industrial air pollution, with African Americans at enhanced risk relative to non-Hispanic whites (76). Despite improvements in air quality over time, African Americans remain exposed to significantly more air pollution than non-Hispanic whites (77). Accounts of the industrial division of labor by race in major U.S. cities document how people of color were restricted to low-wage, hazardous occupations while simultaneously being confined to low-income housing near these industries (75). Similar labor divisions also occurred in agriculture (33). Grandfathering clauses allow older industrial facilities, often located in America’s metropolitan centers, to opt out of the stricter environmental regulations required of newer facilities, thereby clustering industrial toxins within these urban cores (76). Suburbanization was accompanied by expansion and clustering of highways near and through neighborhoods of color (78), leading to higher traffic-related air pollution exposure among African Americans and Latinos (37). A shifted focus to suburban economic development with consequential disinvestment in inner city neighborhoods has perpetuated a legacy of environmental inequality (75). The cumulative effects of these cultural forces enhance exposure to environmental toxicants among African American, Latino, and low-income communities; addressing this history is essential to eliminating disparities in metabolic health.

Environmental Health in the Diabetes Clinic

With increasing evidence that pollutants promote metabolic dysfunction and likely contribute to diabetes disparities, environmental health will become an important component of clinical practice. As such, physicians need to be acquainted with these data to meaningfully address the concerns of their patients who are increasingly troubled about these links. In addition, as these data mature, policies to improve environmental quality should become components of comprehensive diabetes prevention and management strategies. Such efforts may have significant benefits. On the basis of recent intriguing analyses of the PIVUS study, 25% reductions in representative compounds from several chemical classes discussed herein (PCBs, OC pesticides, and phthalates) as well as perfluoroalkyl substances are predicted to reduce diabetes prevalence in Europe by 13% (95% CI 2–22%), with a projected cost savings of €4.51 billion/year (79). Thus, the identification of patient-specific exposures and implementation of exposure reduction strategies may reduce the burden of diabetes on both the individual and society at large.

Identifying Patients With Unique Diabetes Phenotypes

Astute clinicians revolutionized diabetes care by recognizing unique disease phenotypes that were subsequently linked to specific genetic variants and targeted therapeutics (i.e., maturity onset diabetes of the young). Similarly, comprehensive occupational and environmental histories in patients without the classical clinical features of type 2 diabetes and without a genetic explanation may identify unique chemical exposures that promote disease development. Similarly, patients whose medication needs are greater than anticipated may have background exposures that exacerbate metabolic dysfunction. Aided by the development and implementation of validated clinical questionnaires to estimate contact with diabetogenic chemicals, informed clinicians may be able to identify glucose-disrupting exposures and offer patients targeted interventions to improve diabetes outcomes.

Reducing Exposures to Nonpersistent Pollutants

For patients exposed to nonpersistent diabetogenic pollutants, practices that increase exposure to diabetogenic toxicants offer insights into potential interventions. For example, fast food intake increases phthalate exposures (80), whereas consuming water from polycarbonate bottles (81) or soup from cans (82) increases urinary BPA levels. Built upon this knowledge, clinical trials have attempted to lower BPA and phthalate levels (Table 2). In an intervention focused on personal care products, attention to product contents reduced levels of various chemicals, including MEP (83). A trial focused on eating food with limited packaging reduced levels of DEHP and BPA (84), whereas hand washing and reduction of the use of plastic cups lowered phthalate levels in children (85). Although these studies are encouraging, challenges in advising patients remain. For example, avoiding packaged and fast foods may be impractical in individuals with low food security. Moreover, even careful efforts can be confounded by unexpected exposures as illustrated by a failed intervention during which DEHP levels rose because of unexpectedly high phthalate levels in milk and ground coriander (86). Collectively, these data suggest that limiting contact with plastics and packaging, encouraging hand washing, and increasing awareness of diabetogenic toxicants can reduce exposures; however, these efforts must be supported by regulatory action to ensure adequate labeling of consumer products. Finally, evidence that insulin sensitivity rapidly shifts with changes in air quality (40,41) suggests that advising patients to avoid exercise near busy streets or during peak traffic hours to limit contact with air pollutants may afford metabolic benefits, whereas community interventions to improve air quality (e.g., access to public transit, reduced wood and leaf burning, expanded use of clean energy sources, tree planting) may reduce diabetes risk.

Table 2.

Interventional studies that lowered levels of nonpersistent and persistent diabetogenic EDCs in humans

Reference	Population	Intervention and assessment	Results
Nonpersistent pollutants
Harley et al., 2016 (83)	100 Latina adolescents from the Health and Environmental Research on Makeup of Salinas Adolescents study	Mean percent change (95% CI) in urinary concentrations after 3-day intervention with personal care products devoid of chemicals under study	MEP: −27.4% (−39.3 to −13.2)*; Methylparaben: −43.9% (−61.3 to −18.8)*; Propylparaben: −45.4% (−63.7 to −17.9)*; Triclosan: −35.7% (−53.3 to −11.6)*; Benzophenone-3: −36.0% (−51.0 to −16.4)*
Rudel et al., 2011 (84)	10 children and 10 adults from the San Francisco Bay Area, California	Mean urinary concentrations of BPA and phthalates before and during 3-day dietary intervention with fresh and organic foods that were not canned or packaged in plastic	BPA: 3.7 vs. 1.2 ng/mL; −66%*; MEHP: 7.1 vs. 3.4 ng/mL; −53%*; MEOHP: 27 vs. 12 ng/mL; −55%*; MEHHP: 57 vs. 25 ng/mL; −56%*
Chen et al., 2015 (85)	30 Taiwanese girls with previously recorded high urinary phthalate metabolite concentrations	Mean urinary concentrations (μg/g) of creatinine (95% CI) of eight phthalate metabolites before and after 1 week of seven different interventions: hand washing, not using plastic containers, not eating food wrapped in plastic, not microwaving food, not taking nutritional supplements, reducing the use of cosmetics, and reducing the use of personal care products (results are for those who were compliant with the intervention)	MMP: 10.4 (3.49–29.7) vs. 4.54 (2.97–17.3)*; MEP: 58.6 (9.08–650) vs. 16.4 (4.66–200)*; MBP: 123 (57.9–482) vs. 84.7 (36.3–236)*; MBzP: 8.52 (1.87–58.2) vs. 4.67 (1.25–17.8)*; MEHP: 14.4 (4.34–38.3) vs. 6.95 (3.42–24.7)*; MEOHP: 55.2 (21.2–207) vs. 26.9 (15.5–85.6)*; MEHHP: 115 (40.3–398) vs. 61.2 (29.1–202)*; MECPP: 124 (34.7–320) vs. 52.9 (30.1–161)*; ΣDEHP: 0.98 (0.36–3.21) vs. 0.49 (0.27–1.57)*
Sathyanarayana et al., 2013 (86)	21 individuals from Seattle, Washington, with high potential for BPA and phthalate exposures	Geometric mean urinary DEHP concentrations (nmol/g creatinine) (95% CI) before and at completion of 5-day intervention with complete dietary replacement with fresh and organic foods prepared without plastics	DEHP: 283.7 (154.6–520.8) vs. 7,027.5 (4,428.1–11,152.68)*
POPs
Geusau et al., 1999 (87)	2 female patients with chloracne	Fecal excretion of 2,3,7,8-tetrachlorodibenzo-p-dioxin before and after a 38-day intervention of dietary supplementation with olestra chips by using five different dosing regimens (15–66 g olestra daily)	Patient 1: 134 vs. 1,350 ng/day; Patient 2: 29 vs. 240 ng/day
Jandacek et al., 2014 (88)	23 participants from Anniston, Alabama, with PCB levels above the national 50th percentile	Elimination rate (ng/g lipid/year; mean ± SEM) of 37 serum PCBs before and after a 1-year double-blind placebo-controlled trial of 15 g/day dietary olestra vs. placebo (vegetable oil)	Olestra: −0.00864 ± 0.0116 vs. −0.0829 ± 0.0357/year*; Placebo: −0.0283 ± 0.0096 vs. −0.0413 ± 0.0408/year
Redgrave et al., 2005 (89)	1 obese male patient with diabetes	Adipose levels of the PCB mixture aroclor 1254 before and after 2 years of dietary supplementation with olestra (16 g/day)	Aroclor 1254: 3,200 vs. 56 mg/kg; Body weight: 101 vs. 83 kg; BMI: 33.0 vs. 27.1 kg/m2; Cholesterol: 8.6 vs. 3.7 mmol/L; Triglycerides : 11.8 vs. 1.4 mmol/L; Blood glucose: 17 vs. 5.3 mmol/L
Arguin et al., 2010 (90)	37 obese men undergoing weight loss trial	Plasma concentrations (μg/L) of the OC pesticide β-HCH (mean ± SD) before and after a 3-month weight loss intervention; subjects randomized to standard treatment (n = 13), fat-reduced diet (n = 14), and olestra-substituted diet (33% of dietary fat) (n = 10)	Standard treatment: 0.009 ± 0.019; Fat-reduced group: 0.015 ± 0.035; Olestra group: −0.009 ± 0.034*
Guo et al., 2016 (92)	15 healthy women from the San Francisco Bay Area, California	Blood levels of five PCBs and two OC pesticides (ng/g lipid) (mean ± SEM) before and after 2 months of supplementation with 1,000 mg/day ascorbic acid (vitamin C)	PCB 74: 4.04 ± 0.57 vs. 4.00 ± 0.62*; PCB 118: 6.87 ± 0.97 vs. 6.79 ± 1.01*; PCB 138: 10.85 ± 1.66 vs. 10.52 ± 1.65*; PCB 153: 21.16 ± 4.16 vs. 20.68 ± 4.01*; PCB 180: 20.47 ± 4.99 vs. 20.07 ± 4.85*; 4,4′-DDT: 8.31 ± 0.96 vs. 8.18 ± 1.11*; 4,4′-DDE: 344.06 ± 58.40 vs. 338.77 ± 57.65*

β-HCH, β-hexachlorocyclohexane; MBP, monobutyl phthalate; MBzP, monobenzyl phthalate; MECPP, mono-2-ethyl-5-carboxypentyl phthalate; MEHHP, mono-2-ethyl-5-hydoxyhexyl phthalate; MEHP, mono-2-ethylhexyl phthalate; MEOHP, mono-2-ethyl-5-oxohexyl phthalate; MMP, monomethyl phthalate.

*P < 0.05.

Clinical Strategies To Reduce the Body Burden of Persistent Pollutants

For people exposed to persistent organic pollutants (POPs), evidence suggests that interventions can reduce diabetogenic EDC levels (Table 2). In several small studies, the nonabsorbable fat olestra facilitated elimination of lipophilic toxicants, including the dioxin 2,3,7,8-tetrachlorodibenzo-p-dioxin, in two patients with chloracne (87). Moreover, olestra was shown to accelerate the elimination of 37 noncoplanar PCBs in 11 individuals from Anniston, Alabama (88). With regard to the metabolic impact of these changes, one case study of OC toxicity showed that 2 years of olestra resulted in weight loss and improvements in glycemic control (89). Whether these metabolic improvements resulted from the elimination of POPs or were simply a consequence of weight loss requires further study. Olestra may not universally lower POP levels, however. In subjects who underwent a weight loss intervention, olestra decreased levels of β-hexachlorocyclohexane but did not attenuate the expected weight loss–induced rise in other OCs (90). Collectively, these findings raise the possibility that other agents that interrupt the enterohepatic circulation of lipophilic toxicants similarly lower the body burden of POPs and mitigate their diabetogenic effects. The glycemic benefits of the bile acid sequestrant colesevelam could partially reflect clearance of metabolism-disrupting chemicals, but this hypothesis requires formal testing. Other approaches to reduce the body burden of diabetogenic EDCs also have been tried. Supported by cross-sectional data suggesting that fruit and vegetable consumption attenuates the PCB-diabetes association (91), a study of 15 healthy women showed that 1,000 mg/day of ascorbic acid for 2 months reduced levels of six PCBs and two OC pesticides (92).

Although further work is needed, these small intervention trials provide clinicians and patients with intriguing evidence that therapeutic approaches may be devised to mitigate exposures to diabetogenic toxicants and potentially reverse their adverse effects. On the basis of these data and knowledge of common exposure sources, physicians can aid patients wishing to take a precautionary approach by providing guidance on exposure-reduction strategies (Table 3 and Healthcare Provider Guide in Supplementary Data).

Table 3.

Sources of diabetes-promoting EDCs and exposure reduction strategies

Chemical	Source	Exposure reduction strategy
PCBs	Contaminated fish, meat, and dairy products, including bottom-feeding freshwater fish that consume PCB-laden sediment	Consult local guidelines regarding which sport fish are safe to consume; Trim fat from meat and skin from fish and cook on a rack that allows fat to drain away
	Dusts contaminated with low levels of PCBs can coat the surfaces of fruits and vegetables	Wash fruits and vegetables before consumption
	Contaminated drinking water arising from PCB leaching from toxic waste sites or old submersible pumps containing PCBs (development of an oily film or fuel odor in water wells)	Check submersible pumps for failure and, if so, replace pumps and clean the well
	Older fluorescent lights with transformers or ballasts containing PCBs	Replace old PCB-containing fluorescent bulbs
	Deterioration of old building materials, including some paints and caulking	Remove deteriorating building materials; Repair damaged areas with new, safer alternatives
OC pesticides	Some high-fat meats and dairy products as well as some fatty fish	Trim fat from meat and skin from fish and cook on a rack that allows fat to drain away
	Dust and soil contaminated from historical use	Regularly clean floors and remove dust with a damp cloth; Wash hands often, especially before eating or preparing food; Wash fruits and vegetables before consumption
Air pollutant	Burning of fossil fuels, including power plants, motorized vehicles, lawn care equipment, chemical plants, factories, refineries, and gas stations	Check local air pollution forecasts and avoid outdoor exercise when pollution levels are high; Avoid exercise near high-traffic areas; Use hand-powered or electric lawn care equipment; Encourage local schools and municipalities to reduce bus emissions by eliminating idling; Plant trees
	Gas appliances, paints, solvents, tobacco smoke, and household chemicals, including cleaning supplies	Choose electrical appliances and paints low in volatile organic compounds; Limit use of household chemicals; Avoid places that permit smoking
	Combustion of organic materials, including fireplaces, wood stoves, charcoal grills, and leaf burning	Do not burn wood, leaves, or trash
BPA	Polycarbonate plastics, including some water and baby bottles, compact discs, impact-resistant safety equipment, and medical devices	Avoid plastic containers designated #7 on the bottom; Do not microwave polycarbonate plastic food containers; Opt for infant formula bottles and toys that are labeled BPA- free; Opt for glass, porcelain, or stainless steel containers when possible, especially for hot foods and drinks
	Epoxy resins coating metal products, such as food cans, bottle tops, and water supply pipes	Eat fresh and frozen foods while reducing use of canned foods; Prepare more meals at home and emphasize fresh ingredients
	Thermal paper, including sales receipts	Minimize handling of receipts and thermal paper
	Some dental sealants and composites	Consult dentist about alternative options
Phthalates	Plastic food and beverage containers	Opt for glass, porcelain, or stainless steel containers when possible, especially for hot food and drinks
	Personal care products, such as perfumes, hair sprays, deodorants, nail polishes, insect repellants, and most consumer products containing fragrances, including shampoos, air fresheners, and laundry detergents	Read labels and avoid products containing phthalates; Choose products labeled phthalate-free; Avoid fragrances and opt for cosmetics labeled no synthetic fragrance, scented only with essential oils, or phthalate-free
	Contaminated food and water	Purchase, if possible, organic produce, meat, and dairy products; Avoid food known to be especially high in contaminants; Consider using a water filter
	Plastic toys; plastic coatings on wires, cables, and other equipment; plastic shower curtains; PVC-containing products; carpeting and vinyl flooring; and medical devices, including intravenous bags, tubing, and some extended-release medications	Choose nonplastic alternatives whenever possible, especially avoid plastics labeled #3 and #7; Avoid hand-me-down plastic toys

Conclusions

African Americans, Latinos, and the socioeconomically disadvantaged have long been recognized to bear a higher burden of diabetes, but the reasons for these disparities are not completely understood. We provide evidence that higher exposure to diabetogenic pollutants is an important contributor. Although further work is required to validate the EDC-diabetes link and better quantify exposure disparities, current evidence suggests that improvements in environmental health could reduce diabetes risk and disparities. As additional data accumulate and the field matures, the practicing diabetologist and endocrinologist will be uniquely positioned to address exposure to diabetogenic environmental toxicants as part of individualized diabetes care plans to reduce disease risk and to improve diabetes outcomes across the population.