10

Metabolic and Cardiovascular Disorders

Chapter Overview

Based on new evidence and a review of prior studies, the current committee found that new scientific evidence combined with previously reviewed studies has led to changes in the categorizations of association for two metabolic and cardiovascular outcomes outcomes: type 2 diabetes and hypertension.

• After extensive deliberation concerning the new evidence and the results of studies reviewed in previous updates, the committee was unable to reach consensus as to whether the evidence of an association between exposure to the chemicals of interest (COIs) and diabetes met the criteria for being considered sufficient or whether concerns about chance, bias, and confounding lead to the conclusion that the evidence regarding diabetes is limited or suggestive.
• There is sufficient evidence of an association of hypertension with the COIs.

The committee found that the current evidence supports the conclusions reached by committees responsible for earlier updates concerning the other metabolic and cardiovascular outcomes reviewed in the Veterans and Agent Orange (VAO) series of reports:

• There is limited or suggestive evidence of an association between the COIs and ischemic heart disease and stroke.
• There is inadequate or insufficient evidence to determine whether there is an association between the COIs and any other adverse metabolic or cardiovascular outcome examined.

This chapter summarizes and presents conclusions about the strength of the evidence from epidemiologic studies regarding an association between exposure to the COIs—2,4-dichlorophenoxyacetic acid (2,4-D), 2,4,5-trichlorophenoxy-acetic acid (2,4,5-T) and its contaminant 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), picloram, and cacodylic acid—and metabolic and cardiovascular disorders. The committee also considers studies of exposure to polychlorinated biphenyls (PCBs) and other dioxin-like chemicals to be informative if their results were reported in terms of TCDD toxic equivalents (TEQs) or concentrations of specific congeners. Studies that report TEQs based only on mono-ortho PCBs (which are PCBs 105, 114, 118, 123, 156, 157, 167, and 189) are considered even though their TEQs are several orders of magnitude lower than those of the non-ortho PCBs (77, 81, 126, and 169), based on the revised World Health Organization (WHO) toxicity equivalency factor (TEF) scheme of 2005 (La Rocca et al., 2008; van den Berg et al., 2006). The lower TEQs of the mono-ortho PCBs, however, may be counterbalanced by their abundance, which is generally many orders of magnitude higher than that of the non-ortho PCBs (H. Y. Park et al., 2010).

The first part of the chapter covers the metabolic disorder type 2 diabetes and the second part reviews a variety of cardiovascular and blood disorders. Cardiovascular conditions explored in this chapter are based on the categories in the 9th and 10th revisions of the International Classification of Diseases (ICD; ICD-9 390–459 and ICD-10 I00–I99, respectively) and include hypertension (ICD-9 401–404; ICD-10 I10–I13), ischemic heart disease (ICD-9 410–414; ICD-10 I20–I25), heart failure (ICD-9 428; ICD-10 I50), and cerebrovascular disease (ICD-9 430–438; ICD-10 I60–I69).

TYPE 2 DIABETES

Diabetes mellitus is a group of heterogeneous metabolic disorders characterized by hyperglycemia and a quantitative or qualitative deficiency in insulin action (Orchard et al., 1992) and classified as E08–E13 by ICD-10. Although all forms of diabetes share hyperglycemia, the pathogenic processes involved in the development of the various types of diabetes differ. Most cases of diabetes mellitus are classified as being in one of two categories: type 1 diabetes or type 2 diabetes. The long-term complications of both types can include cardiovascular disease (CVD), nephropathy, retinopathy, neuropathy, and increased vulnerability to infections.

When referring to diabetes, most research will not distinguish between type 1 and type 2 because type 2 diabetes accounts for 90% to 95% of all cases. Type 1 diabetes (ICD-10 E10) is characterized by a lack of insulin caused by the immunologically mediated destruction of insulin-producing cells in the pancreas (β cells), which often occurs during childhood but can occur at any age. As with many autoimmune diseases, genetic and environmental factors both influence its pathogenesis. Some viral infections are believed to be important environmental

triggers for the autoimmunity associated with type 1 diabetes. Type 1 diabetes most often occurs in children, which is only of relevance in offspring of adults exposed to the COIs.

Type 2 diabetes exhibits both resistance to the actions of insulin and inadequate secretion of insulin (called relative insulin deficiency). The prevalence of diabetes in the United States has been increasing among nearly all races and ethnic groups. In 2015 there were approximately 30.3 million people ages 18 years and older who had diabetes in the United States, or 9.4% of the U.S. population. Of these, 23.1 million had officially been diagnosed, whereas 7.2 million (23.8%) were considered to be prediabetic (CDC, 2017c). Although the modern ICD-10 classification system recognizes that type 2 diabetes can occur in children and can require insulin treatment, most cases of type 2 diabetes occur in adults.

Onset can occur before 30 years of age, and the incidence increases with age. The main risk factors for type 2 diabetes are age, obesity, abdominal fat deposition, a history of gestational diabetes (in women), physical inactivity, ethnicity, and family history, but the relative contributions of each of those risks to the overall prevalence are not known. Type 2 diabetes affects women and men nearly equally (11.7 million versus 11.3 million, respectively); however, the Centers for Disease Control and Prevention (CDC) estimates that there are more undiagnosed diabetes cases among men than among women (4.0 million versus 3.1 million, respectively). Table 10-1 shows the estimated prevalence of diagnosed and undiagnosed diabetes among adults over 18 years by age and sex. The prevalence of type 2 diabetes among U.S. adults aged 18 years and older is greatest in American Indians/Alaska Natives (15.1%), followed by non-Hispanic blacks (12.7%) and Hispanics (12.1%). The lowest prevalence rates are for non-Hispanic whites (7.4%) and Asians (8.0%). As of 2015, an estimated 1.5 million new cases of diabetes are diagnosed annually, for an incidence rate of approximately 6.7 per 1,000 persons. In non-Hispanic blacks the incidence was 9.0 per 1,000 persons and 8.4 per 1,000 persons among people of Hispanic origin (CDC, 2017c).

The etiology of type 2 diabetes is unknown, but three major components have been identified: peripheral insulin resistance in target tissues (muscle, adipose tissue, and liver), a defect in β-cell secretion of insulin, and the overproduction of glucose by the liver. When a person enters a state of insulin resistance, insulin secretion is initially higher for each concentration of glucose than in people who are not insulin resistant. This hyperinsulinemic state compensates for peripheral resistance and in many cases keeps glucose concentrations normal for years. Eventually, however, β-cell compensation becomes inadequate, and there is a progression to overt diabetes with concomitant hyperglycemia. Why the β cells cease to produce sufficient insulin is not known.

Pathogenic diversity and diagnostic uncertainty are some important problems in conducting epidemiologic studies of diabetes mellitus. There are multiple pathogenic mechanisms that are likely to play a role in the development of diabetes mellitus, including various genetic susceptibilities (as varied as autoimmunity and obesity) and a variety of potential environmental and behavioral factors (such as viruses, nutrition, and physical activity). Because in some populations up to half of all cases of diabetes remain undiagnosed for an extended period of time, the potential for ascertainment bias in population-based surveys is high, and groups that are more intensively followed for signs and symptoms of diabetes or those with more frequent health care contact, in general, are more likely to be diagnosed in time to receive appropriate treatment; this points to the need for formal standardized testing to detect cases in epidemiologic studies.

Metabolic syndrome is a cluster of at least three of the following—increased blood pressure or hypertension, insulin resistance, excess body fat around the waist, and abnormal serum cholesterol or triglyceride levels—that occur together, increasing the risk of heart disease, stroke, and diabetes. Although it is not a disease entity itself, metabolic syndrome is associated with a five-fold increased risk of developing type 2 diabetes and a doubling of the risk of developing CVD (Alberti et al., 2009). Swaminathan et al. (2015) found that pesticide exposure was a contributing factor to hyperglycemia and type 2 diabetes even after adjusting for body mass index (BMI). There is a growing literature on the association between the COIs and metabolic syndrome and its components. Given its strong linkage with type 2 diabetes, studies on metabolic syndrome will be discussed in this section.

Conclusions from VAO and Previous Updates

The committee responsible for the first report in the VAO series (IOM, 1994) concluded that there was inadequate or insufficient information to determine whether there is an association between exposure to the COIs and diabetes mellitus. Additional information available to the committees responsible for Update 1996 (IOM, 1996) and Update 1998 (IOM, 1999) did not change that conclusion.

In 1999, in response to a request from the Department of Veterans Affairs, the Institute of Medicine convened a committee to conduct an interim review of the scientific evidence regarding exposure to the COIs and type 2 diabetes. That review focused on information published after the deliberations of the Update 1998 committee and resulted in the report Veterans and Agent Orange: Herbicide/Dioxin Exposure and Type 2 Diabetes (IOM, 2000b). The committee responsible for that report determined that there was limited or suggestive evidence of an association between exposure to at least one COI and type 2 diabetes, based on positive associations reported in several mortality studies (Pesatori et al., 1998; Steenland et al., 1992, 1999; Vena et al., 1998), which may underestimate the incidence of diabetes, and positive associations reported in most of the morbidity studies (AFHS, 2000; Calvert et al., 1999; CDVA, 1998a,b; Cranmer et al., 2000; Longnecker and Michalek, 2000; Michalek and Tripathi, 1999) reviewed by that committee. However, that committee also noted that the studies indicate that the increased risk, if any, from herbicide or dioxin exposure appeared to be small and that the known predictors of diabetes risk—family history, physical inactivity, and obesity—continued to greatly outweigh any suggested increased risk from wartime exposure to herbicides. The committees responsible for updates since that report have upheld that finding. Reviews of the considered studies are found in the earlier reports.

Several epidemiologic studies have been published that examined type 2 diabetes in different populations of Vietnam veterans: Air Force Health Study (AFHS), female U.S. Vietnam-era veterans, and Korean Vietnam veterans. Publications from the AFHS have found a statistically significant increased risk of diabetes among the Ranch Hands (Henriksen et al., 1997; Kern et al., 2004; Longnecker and Michalek, 2000; Michalek and Pavuk, 2008). Most analyses of the Army Chemical Corps (ACC) veterans did not find increased risk of diabetes. Using mortality data on U.S. women veterans, Kang et al. (2014a) found that the adjusted relative risk of diabetes mortality was non-significantly lower for the female veterans deployed to Vietnam than for their non-deployed counterparts, and when the analysis was restricted to nurses only, the adjusted relative risk of diabetes mortality was effectively the same for deployed and non-deployed nurses.

No statistically significantly increased risk of diabetes was found in several studies of Australian Vietnam veterans. In an exceptionally large epidemiologic study of Korean veterans who served in the Vietnam War, using data from the Korea National Health Insurance Service, the risk of type 2 diabetes mellitus was found to be nominally higher for those with a high potential for herbicide exposure than for those with low–exposure opportunity index scores, and there was a small, but statistically significant association for veterans with non-insulin-dependent diabetes mellitus (Yi et al., 2014a). In the mortality study of 180,639 Korean Vietnam veterans, Yi et al. (2014b) found that after adjustment, high exposure (compared with low exposure) was not associated with mortality from all forms of diabetes (ICD-10 E10–E14).

Occupational studies of diabetes have found slightly elevated but generally not statistically significant associations between exposure to the phenoxy herbicides or TCDD and a risk of diabetes. Results from the Agricultural Health Study (AHS) found an increased risk of developing diabetes in the wives of pesticide applicators who themseleves had reported personally mixing or applying pesticides or the ever use of 2,4,5-T or 2(2,4,5-trichlorophenoxy) propionic acid (2,4,5-TP). Ever use of 2,4-D was much more prevelant, but it was not associated with a risk of diabetes (Starling et al., 2014). No difference or decreased risk of diabetes was found among the AHS pesticide applicators in both morbidity and mortality studies.

Several studies of environmental exposures to the COIs and diabetes have been reviewed in the VAO series. The findings are mixed, and the studies have used different designs and methods for adjusting effect estimates for confounding.

Update of the Epidemiologic Literature

Two new studies of Vietnam veterans and diabetes have been identified since Update 2014. Several occupational, environmental, and case-control studies of diabetes and exposure to the COIs were also identified for this update. Table 41, which can be found at www.nap.edu/catalog/25137, summarizes the results of studies related to diabetes and related health outcomes.

Vietnam-Veteran Studies

Using data from the AFHS, Mazur et al. (2014) studied 991 Ranch Hands and comparison veterans who completed all six medical examinations over the 20 years of data collection (1982, 1985, 1987, 1992, 1997, 2002) to see if low testosterone is a risk factor for high fasting glucose and for a type 2 diabetes diagnosis. Subjects ranged from 32 to 68 years of age at the onset of the study. Fasting glucose was tested in the morning for all men at each exam. Hemoglobin A1c (HbA1c), which is indicative of blood glucose levels over the preceding 120 days, was measured for all men in cycles 4 and 5. Among the men, gradual increases in obesity, fasting glucose, and type 2 diabetes and a decline in mean testosterone levels occurred over the 20 years of study. The authors noted that men who would not be diagnosed with type 2 diabetes until cycles 5 or 6 already showed by cycle 1 that their mean fasting glucose was significantly higher than normal. A statistically significant (p < 0.001) correlation was found between fasting glucose in all six cycles and both HbA1c in the fourth and fifth cycles and a diagnosis of type 2 diabetes in any of the six cycles. Men who had high testosterone levels and low BMI had a mean fasting glucose level of 100 mg/dl, whereas men with a high BMI and low testosterone levels had a fasting glucose level of 116 mg/dl. The multivariate analyses showed that testosterone was inversely related to glucose, independently of BMI and age, but its effect on glucose was small. The

study showed that levels of low testosterone were related to high fasting glucose, independent of the participant’s age or obesity. Results also indicated that low testosterone is a relatively poor predictor of a type 2 diabetes diagnosis. However, this study is of limited value to this update because the results were not stratified by exposure status (Ranch Hand versus comparison) and because serum concentrations of dioxin were not included in the analysis.

Cox et al. (2015) used hospital discharge records from 1988 to 2009 to identify prevalent health conditions in 2,783 male New Zealand veterans who served in Vietnam. Age-specific hospitalization rates were calculated using the total number of annual hospitalizations published by the Ministry of Health and the average annual resident population. Standardized hospitalization rates and 99% confidence intervals (CIs) were calculated for the veteran cohort and the general population and reported for diabetes and other outcomes. Although the risk estimate was slightly elevated for diabetes (standardized hospitalization ratio [SHR] = 1.14, 99% CI 0.90–1.38), it was not statistically significant compared with the general population. However, diabetes is not usually a condition that requires hospitalization, thus likely only those with the most severe disease were identified, and the number of cases available for study would be expected to be low, limiting the statistical power for this type of analysis considerably. Exposure to the COIs was not validated through serum measurements, and the study did not control for smoking or ethnicity or other potentially important risk factors.

Occupational Studies

Collins et al. (2016) added follow-up time for a retrospective cohort of 2,192 workers exposed to dioxins during trichlorophenol (TCP) and pentachlorophenol (PCP) production at a chemical manufacturing plant in Midland, Michigan. Workers were compared to the U.S. population in order to calculate standardized mortality ratios (SMRs), and work history records provided information about the length of exposure. Serum samples used to measure levels of six types of dioxins were collected from 431 TCP and PCP workers. Concentrations for each dioxin congener were calculated based on the median concentration in the serum samples and the known half-lives associated with each congener. Complete vital status follow-up was achieved for the cohort, and there were 1,198 deaths during the entire study period (1979–2011). Compared with the standardized U.S. population, no statistically significant difference in mortality for diabetes was found for the TCP workers (n = 19; SMR = 0.92, 95% CI 0.55–1.43) or the PCP workers (n = 9; SMR = 0.97, 95% CI 0.44–1.84). Among the combined 1,198 TCP and PCP workers, there were 28 deaths from diabetes (SMR = 0.97, 95% CI 0.64–1.42). Additionally, there are other concerns with the diabetes outcome in this analysis. First, mortality data for diabetes likely under-reports disease prevalence because diabetes is often a contributing factor to but not the actual cause of a death. Second, using the general U.S. population as the comparison

is problematic because of the “healthy worker effect,” that is, some of the comparison subjects are not in the workforce and will typically have, on average, poorer general health than individuals in the workforce. In this case, it would imply a control group with rates of diabetes greater than those in the workforce. In contrast, an internal comparison of workers with high- versus low-dioxin concentrations, which was not presented by the authors, would have avoided this type of selection bias.

't Mannetje et al. (2018) conducted a morbidity survey among a subset of workers who were employed at the New Plymouth, New Zealand, phenoxy herbicide production plant for at least 1 month between 1969 and 1984. The plant produced 2,4,5-T, and workers were potentially exposed to 2,4,5-T, intermediates of TCP and other chlorophenols, and TCDD. Workers had previously been recruited and examined as part of the international cohort of producers of phenoxy herbicides led by the International Agency for Research on Cancer (IARC) (Kogevinas et al., 1997); see Chapter 5 for more details on the IARC cohort and the New Zealand phenoxy producers. This study extended the follow-up period of these workers to approximately 30 years from the last 2,4,5-T production exposure. From the original cohort of 1,025 workers, 631 were living, had a current address in New Zealand, and were below 80 years of age on January 1, 2006. For the current follow-up, 430 of the 631 workers were randomly selected and invited to participate in the morbidity survey, of which 245 (57%) participated. The survey was administered in 2007–2008 by face-to-face interview, and it collected information on demographic factors and health information, including doctor-diagnosed conditions and the year of diagnosis. A blood sample was also collected at that time and analyzed for TCDD, lipids, thyroid hormones, and other parameters. For 111 participants, a neurological examination was conducted. Associations between exposure and health outcomes were assessed using logistic regression models that controlled for age, gender, smoking, BMI, and ethnicity using two methods: working in a TCDD-exposed job (based on occupational records) and serum TCDD concentration ≥ 10 pg/g lipid (18%). Mean TCDD concentrations were 19 pg/g lipid in the 60 men directly involved in phenoxy/TCP production and 6 pg/g lipid in the 141 men and 43 women who worked in other parts of the plant. Compared with the people in the non-highly exposed jobs, the people who had ever worked in a highly exposed job at the plant had an elevated risk of doctor-diagnosed diabetes (n = 13; odds ratio [OR] = 3.98, 95% CI 1.03–15.4), although the estimate was imprecise. When compared by serum TCDD concentration, no difference in risk of diabetes was found for workers in the high- versus low-exposure groups (n = 7; OR = 3.05, 95% CI 0.87–10.67), although the estimate was again imprecise.

Cappelletti et al. (2016) performed a retrospective study of 331 male electric arc foundry workers at a single plant in Trentino, Italy, to determine if they experienced excess mortality from all causes or were at increased risk for several other diseases due to occupational exposures to foundry dust. An analysis of the

dust emissions found that it contained metals (including iron, aluminum, zinc, manganese, lead, chromium, nickel, cadmium, mercury, and arsenic), polycyclic aromatic hydrocarbons (PAHs), PCBs, and polychlorinated dibenzo-p-dioxin/dibenzofurans (PCDD/Fs) (reported as TEQs). Therefore, the authors could not determine which of the agents were associated with a specific outcome or to what extent. The men had worked at the factory for at least 1 year and, for the diabetes analysis, were compared with 32 presumed non-exposed workers (clerks, managers, and watchmen) or the general population of Region Trentino-Alto Adige (where the factory was located) because there were few non-exposed foundry workers and high attrition rates. Company and medical records were used to determine vital status; the cause of death was determined from death certificates or other registries. Requests for exemption health care fees were used as a surrogate measure to identify the most prevalent morbid conditions in the general population, which were then applied to the cohort to compute relative risks for each of the conditions, which included diabetes. The workers were followed from March 19, 1979 (or their first day of employment), through December 31, 2009, or date of death. The analysis for diabetes was limited to the 235 living workers, and effect estimates (prevalence ratios) were calculated using Mantel-Haenszel estimator adjusted for age group (20–64, 65–74, ≥ 75 years). Compared with the age-adjusted provincial population, a statistically significant increased risk of diabetes was found among electric arc furnace workers (n = 25; relative risk [RR] = 2.39, 95% CI 1.67–3.41). This study is most limited by the fact that foundry dust is a complex mixture, resulting in an inability to discern the impact of the specific contaminants of the foundry dust on the health outcomes of those exposed workers. Estimates were adjusted only for age group and not for other risk factors, such as tobacco use, BMI, or other jobs or activities that could result in similar exposures. Possible exposure to foundry dust by the general population that was used for comparison is not discussed, although the foundry appears to be in the local vicinity, and emissions from it were reported to be present within a 2-kilometer radius of it.

Yamamoto et al. (2015) performed a cross-sectional study to investigate the health outcomes that 698 male workers sustained while employed at 36 municipal and private waste incineration plants in Japan; serum dioxin measurements were obtained for 678 of the workers. These workers were employed from 2000 to 2007. First, a questionnaire was distributed to the participants to determine their medical history. Blood samples were taken from each subject to evaluate serum dioxin and PCB levels. Both clinical and physiological examinations were also performed. Participants were then categorized into four groups: workers whose jobs did not involve working directly in an incineration facility, workers whose jobs did involve work inside the incineration facility (but only handling solidified fly ash and slag or residues that were nonflammable), workers whose jobs involved helping with incineration-related work inside an incineration facility, and workers whose jobs mainly involved operation and maintenance of an

incinerator including a furnace, electric dust collector, and wet scrubber inside an incineration facility. Subjects were tested for diabetes, hypertension, hyperlipidemia, and liver dysfunction. The total serum dioxin concentration level for incinerator workers was 13.7 (median) or 17.2 (arithmetic mean) pg TEQ/g lipid, while the (arithmetic mean) average total dioxin concentration was 19.4 pg TEQ/g lipid for the general population in Japan, indicating no real difference between the two groups. However, in the exposed workers, the duration of employment was positively associated with TEQ levels (9.8 and 21.7 pg TEQ/g lipid for those employed 9 years and 15.5 years, respectively). Coplanar PCBs were significantly correlated (p < 0.05) with 14 parameters of laboratory and physiological tests. Diabetes was statistically significantly associated with PCDDs, PCDFs, coplanar PCBs, and total dioxins based on adjusted odds ratios between the first and fourth quartile, and tests for trend were also significant for each of these COIs. Positive associations were found between serum levels of total dioxins and the prevalence of diabetes in the incinerator workers even though, as the healthy worker effect would predict, the overall diabetes prevalence in the workers was not as high as in the general Japanese population (6.6% versus 9.0%, p = 0.07).

Environmental Studies

Everett and Thompson (2016) measured three chlorinated dibenzo-p-dioxins, one chlorinated dibenzofuran, and four dioxin-like PCBs in human blood and assessed their cross-sectional associations with nephropathy, specifically diabetes-related nephropathy using data collected between 1999 and 2004 from the National Health and Nutrition Examination Survey (NHANES). Two groups of people were included: the teen and young adult sample (12–30 years old) and the all-adult sample (≥ 20 years old). Nephropathy was defined as a urinary albumin-to-creatinine ratio > 30 mg/g, representing both microalbuminuria and macroalbuminuria. Diagnoses of diabetes were determined by individual, self-reported answers to the NHANES question: “Other than during pregnancy, have you ever been told by a doctor or health professional that you have diabetes or sugar diabetes?” Those who answered “borderline” were considered to not have diabetes. Having undiagnosed diabetes was defined as participants who had A1c levels ≥ 6.5% but had not been diagnosed as having diabetes. Total diabetes was defined as either diagnosed or undiagnosed diabetes. Several differences were noted between the teen/young adult and the all-adult samples. One or more of the eight dioxin-like chemicals were elevated among 9.9% of the teen/young adult sample (at-risk population) and among 53.5% of the all-adult sample. The proportions for nephropathy and pre-diabetes were, respectively, 7.8% and 1.8% in the teen/young adult sample versus 5.5% and 9.7% in the all-adult sample. The unweighted number of people with nephropathy in the teen/young adult sample was 87 females and 35 males. Only three people had pre-diabetes with nephropathy in the teen/young adult sample, which was too few for analyses.

None of the eight dioxin-like chemicals were associated with pre-diabetes, but in most instances there were no cases of pre-diabetes in the elevated concentration range. In the highest TEQ8 category, which included 2.6% of the sample, there was a strong but very imprecisely estimated association with pre-diabetes among females (OR = 11.9, 95% CI 1.6–87.2), but no associations with pre-diabetes at any level of elevated TEQ8 in the entire sample.

Two reports from a study conducted by Aminov et al. (2016a,b) examined diabetes among 601 people of the Mohawk Nation between the ages of 18 and 84 years living at Akwesasne, which spans New York, Ontario, and Québec. The most detailed reporting of this data is found in Aminov et al. (2016a). Serum samples were taken from the subjects to test for PCB and pesticide concentrations as well as to determine each individual’s fasting glucose levels. A subject in this study was considered to have diabetes if he or she had a fasting glucose concentration of > 125 mg/dL or if it was reported that a physician had made a diabetes diagnosis. Several models were used to estimate risks. Model 1 adjusted for sex, age, BMI, and serum lipid concentrations. In model 2, the total concentrations of PCBs were adjusted for the variables in model 1 and also for total pesticides (the sum of the concentrations of 101 PCB congeners, DDE, hexachlorobenzene [HCB], and mirex). Multiple analyses were undertaken, and in the minimally adjusted model 1, both total PCBs (highest quartile OR = 3.45, 95% CI 1.4–8.5) and total pesticides (highest quartile OR = 3.12, 95% CI 1.12–8.65) showed statistically significant associations with diabetes. These associations were attenuated in model 2, which adjusted for total pesticides, with total dioxin-like PCBs estimates having an odds ratio that was elevated but did not reach statistical significance (highest quartile OR = 1.82, 95% CI 0.61–5.40), while the total pesticides estimate (highest quartile OR = 5.01, 95% CI 1.76–14.24) was statistically significantly elevated but quite imprecise. In model 3, PCB concentrations were divided into three groups by numbers of chlorines (3–4, 5–6, and ≥ 7) and by numbers of ortho-substitute chlorines and each pesticide was considered alone. In addition to adjustment for sex, age, BMI, and total lipids, each result was also adjusted for the concentrations of all other groups of contaminants. Results using model 3 suggested a low-dose effect. In general, there was an association between PCB exposure and diabetes, but this was reported to be mainly due to low-chlorinated, non-dioxin-like congeners.

Grice et al. (2017) assessed the association of type 2 diabetes with persistent organic pollutants (POPs), including dioxin-like PCBs (PCB 105, 118, 156, 157, and 167) and HCB, using a nested case-control design from a relatively small prospective study conducted in American Indians in Arizona. A sample of 100 men and 200 women who were at high risk of diabetes were enrolled between 1965 and 1974. Participants underwent a baseline examination that included a blood draw and took an oral glucose tolerance test. The oral glucose test was repeated at each follow-up. Levels of PCBs and pesticides in the serum samples were assessed by CDC. Over 8 years of follow-up, 149 people developed diabetes

(cases), and 151 remained non-diabetic (controls). Cases were followed from diabetes onset to end stage renal failure, death, or 2013. When models were adjusted for age, sex, BMI, 2-hour postload plasma glucose concentration, sample water loss, sample storage time, cholesterol, and triglycerides, the odds of diabetes were statistically significantly decreased for HCB exposure (OR = 0.64, 95% CI 0.41–0.99). In fully adjusted models, most dioxin-like PCBs were positively associated with incident diabetes, but none was statistically significant: PCB 105 (OR = 1.14, 95% CI 0.81–1.60), PCB 118 (OR = 1.16, 95% CI 0.82–1.65), PCB 156 (OR = 1.06, 95% CI 0.73–1.53), PCB 157 (OR = 1.06, 95% CI 0.75–1.49), and PCB 167 (OR = 1.11, 95% CI 0.80–1.55). This study is limited by the small sample size and the fact that serum concentrations of COIs were measured only once, yet the participants were followed for diabetes incidence for 8 years.

During 2007–2008, Singh and Chan (2017) conducted a cross-sectional survey study of 33 Canadian Inuit coastal communities and three inland communities. The survey included questions about health status, chronic diseases, and such behaviors as alcohol consumption, smoking, and exercise habits. Clinical tests such as fasting glucose and blood levels of PCBs and organochlorine pesticides supplemented the survey. In total, 2,172 Inuit people aged 18 years or older participated and provided blood samples. Of the PCB congeners measured in the blood, the dioxin-like mono-ortho PCBs 105, 118, and 156 are of interest for the committee’s charge. Associations of diabetes with individual PCB congeners, the sum of dioxin-like PCBs, the sum of non-dioxin-like PCBs, and total PCBs were determined. Of the 2,172 respondents, 147 (5.7%) self-reported diabetes, with nearly the same proportions among males and females (5.5% and 5.8%, respectively). As expected, the percentage of people with diabetes increased as age increased (2.9% in the 31- to 50-year-old category and 22.1% in the 71- to 90-year-old category), and diabetes was clustered within families (either parent having diabetes: 19.7% in diabetics versus 10.4% in non-diabetics; or having a sibling with diabetes: 20.4% in diabetics versus 4.5% in non-diabetics). All contaminant concentrations were higher in respondents with diabetes. The concentrations of congeners were divided into quartiles, and odds ratios—adjusted for age, sex, BMI, high density lipoprotein-C, omega-3/omega-6 ratio, and education—were calculated for quartile 4 versus quartile 1 of self-reported diabetes. Statistically significant (p < 0.05) positive ORs were found for PCB 105, PCB 118, and PCB 156 as well as for some of the non-dioxin-like PCBs. With lipid adjustment, statistically significant positive ORs were found for dioxin-like PCB 118 and PCB 156 (information taken from graph; exact numbers not reported). The study also reported fasting glucose β coefficients for PCB congeners that were statistically significant based on lipid standardization or lipid adjustment; the congeners included PCB 105 and PCB 118. These models were adjusted for age, sex, BMI, high density lipoprotein-C, triglycerides, alcohol intake, smoking, omega-3/omega-6 ratio, selenium, and education. The β coefficients for wet-weight analyses were positive and were statistically significantly associated with

an increase in fasting glucose for PCB 105 (6%, 95% CI 3%–9%) and PCB 118 (7%, 95% CI 4%–10%). The primary limitation of this study is its cross-sectional design, but the half-lives of these two congeners are estimated at 4.5–5.5 years (Grandjean et al., 2008), and the results of this study agree with other similar studies that found positive associations between PCBs and diabetes in different exposed populations.

Using data collected as part of cross-sectional health study of Taiwan residents living near a closed PCP-producing factory, C. Y. Huang et al. (2015) conducted an analysis of exposure to PCDD/Fs in order to evaluate the association between such exposures and type 2 diabetes. More details on the factory, exposure, recruitment, and collection of health information are provided in Chapter 5, but, briefly, the environmental contamination took place 25–40 years prior to the study. Of the 2,898 participants included in this analysis, 1,143 had a serum dioxin level between 20 and 63 pg WHO98-TEQ_DF/g lipid, which was considered a “high dioxin level”; and 284 subjects had a dioxin level of 64 pg WHO98-TEQ_DF/g lipid or higher. Of the 2,898 participants included in this analysis, 425 had diabetes, defined as having a fasting plasma glucose level above 126 mg/dL or an existing diagnosis. Subjects who had diabetes were compared with those who did not have diabetes; these two groups were comparable with regard to sex and reported family history of diabetes, but differed in age and BMI (high BMI was defined as ≥ 24 kg/m²), factors that were controlled in analyses. Serum dioxin levels of 20–63 pg WHO98-TEQ_DF/g were associated with statistically significantly increased odds of diabetes (OR = 4.4, 95% CI 3.4–5.7), and the association was even stronger for serum dioxin levels above 64 pg WHO98-TEQ_DF/g lipid (OR = 7.8, 95% CI 5.6–10.9). After adjusting for age and BMI in multiple logistic regression models, high dioxin levels of 20–63 pg WHO98-TEQ_DF/g lipid (OR = 2.1, 95% CI 1.5–2.8) and ≥ 64 pg WHO98TEQ_DF/g lipid (OR = 2.7, 95% CI 1.9–4.0) were found to increase the risk for diabetes. When the dioxin level was included as quartiles in the model, adjusted ORs remained elevated and statistically significant, with a statistically significant and steep upward trend (p < 0.001). Additional analyses replaced age with the duration of residency in the endemic area in the model, and after adjustment for the duration of residency and BMI, a high dioxin level of 20–63 pg WHO98TEQ_DF/g lipid (OR = 2.2, 95% CI 1.6–3.0) and ≥ 64 pg WHO98-TEQ_DF/g lipid (OR = 3.6, 95% CI 2.4–5.4) remained strongly associated with risk of diabetes. The duration of residency in the endemic area was also an independent risk factor for diabetes, with a positive trend (p = 0.01). The strengths of this study include its large population (including a substantial proportion with diabetes, n = 425, the largest number of cases in studies newly reviewed in this volume); serum measurements of dioxin concentration; adjustments for age, BMI and residency; and a clear definition of diabetes and control groups. General limitations for this study and all studies that used the cross-sectional data collected from this population include an unknown age at first exposure to PCDDs and PCDFs, unknown

cumulative exposure dose, the cross-sectional design, and the lack of additional data on other potentional confounders, such as waist circumference, dietary intake, and socioeconomic status.

Van Larebeke et al. (2015) describes how dioxin-like activity and serum HCB affected the risk of diabetes in both men and women (n = 1,583; 775 men and 808 women) enrolled by biomonitoring programs organized by the Flemish Centre for Environment and Health. Associations were based on self-reported health outcomes obtained in questionnaires administered in 2011. Blood and urine samples provided measures of internal exposure to organochlorine pollutants, PCB 118, and cadmium; samples had been collected from these subjects between September 2004 and June 2005. The researchers also assessed dioxin-like activity in pg TEQ/g fat. Among the participants, 30 men reported having diabetes (6.4%), and 62 men (13.2%) reported having either diabetes or a related condition; while 29 (5.8%) women reported having diabetes, and 55 (10.9%) women reported having diabetes or a related condition. A significant association (p ≤ 0.05) was found between internal exposure to the above mentioned pollutants and diabetes or a related condition when both men and women were combined; however, when the estimates were adjusted for additional covariates and stratified by sex, serum dioxin-like activity showed a significantly positive association with diabetes and with diabetes or a related condition for men and women separately. Results from this study suggest that exposure to dioxin-like chemicals increases the risk of diabetes or a related condition.

Other Identified Studies

Two other studies of diabetes were identified but either lacked exposure specificity (Swaminathan and Thangavel, 2015) or examined the association of diet in diabetics on serum levels of persistent organic pollutants (Kahleova et al., 2016), which is not considered relevant to the committee’s charge.

Biologic Plausibility

Data from cell culture and animal models support the potential diabetogenic effects of TCDD in humans. TCDD is known to modify the expression of genes related to insulin transport and signaling and to inflammation (Ambolet-Camoit et al., 2015; M. J. Kim et al., 2012), which play a pivortal role in diabetes progression. In previous VAO updates, several studies have been reviewed that support the postulate that TCDD is mechanistically implicated in an increased risk of insulin resistance and in the development of diabetes. C. Wang et al. (2011) found that mice that lacked the aryl hydrocarbon receptor (Ahr) had enhanced insulin sensitivity and enhanced glucose tolerance, suggesting that Ahr has a physiologic function in glucose metabolism. This is supported by studies that found that sustained activation of Ahr by dioxin-like chemicals could contribute to diabetes,

including a study by Kurita et al. (2009), who found that exposing mice to dioxin significantly reduced insulin secretion after a glucose challenge. Recent data also demonstrate the importance of Ahr in glucose and fat metabolism, showing that the chemical inhibition of Ahr leads to decreased obesity and fatty livers in both male and female mice (Moyer et al., 2017). In an in vitro study of differentiated adipocytes, TCDD was found to significantly reduce insulin-stimulated glucose uptake (Hsu et al., 2010). Furthermore, P. Lu et al. (2015) found that activation of the AHR leads to dysregulation of the link between fatty liver and insulin resistance, suggesting a mechanism involving fibroblast growth factor 21. A recent study used a transcriptomic approach to identify potential molecular mechanisms of TCDD-induced changes in the insulin secretion of pancreatic islets and β cells (Lai et al., 2017). Taken together, these data provide potential mechanims for the metabolic-modulating effects of AHR and dioxins that are critical for the progreassion of metabolic diseases, including diabetes.

Among the studies included in the current literature review, K. S. Kim et al. (2014) explored the relationships of diabetes and insulin resistance with 14 organochlorine insecticides and 22 PCBs in visceral adipose tissue and subcutaneous adipose tissue in 50 patients with or without type 2 diabetes who underwent surgery for either cancer or benign liver or gallbladder lesions. The researchers reported that persistent organic pollutants in visceral or subcutaneous fat were significantly associated with both diabetes and insulin resistance. These findings are consistent with experimental animal studies that have reported that exposure to persistent organic pollutan mixtures through contaminated fish oil induces a severe impairment of whole-body insulin action (e.g., Ibrahim et al., 2011). Thus, on balance, there is biological plausibility for the COIs being causally implicated in the development of insulin resistance and diabetes. No new content has been identified since Update 2014.

Synthesis

The considerable amount of new evidence regarding type 2 diabetes reviewed and considered by the committee in forming its judgment included studies on male Vietnam veterans from the United States and New Zealand and studies of occupational cohorts and residential population-based studies of exposure. The analysis of AFHS data on testosterone levels and incident diabetes was considered to be of limited value to the committee’s charge because the results were not stratified by exposure status and because serum concentrations of dioxin were not included in the analysis (Mazur et al., 2014). Cox et al. (2015) calculated annual rates of hospital discharge records for New Zealand Vietnam veterans (presumed to be exposed to herbicides based on deployment to Vietnam) compared with the general residential population of New Zealand. Although slightly elevated among the veterans, the standardized hospitalization ratio for diabetes was not statistically significant. However, because diabetes is not usually a condition that

requires hospitalization, it is likely that only those with the most severe form of the disease were identified, and the prevalence of diabetes is expected to be higher. Moreover, this analysis is limited because exposure was not confirmed through serum measurements, and hospitalization estimates were adjusted only for age and not for other important risk factors, such as BMI or ethnicity.

The four occupational studies that were examined presented mixed results concerning an association between exposure to the COIs and diabetes mortality or prevalence. Among U.S. workers who were exposed to dioxins through the manufacturing of TCP or PCP, no difference in the rate of death from diabetes was found between either group of workers and the general U.S. population (Collins et al., 2016). Although serum dioxin measurements were collected, no results based on those measurements were presented for diabetes; instead the authors used employment records to categorize exposure, which may have introduced exposure misclassification, and use of the general U.S. population as a comparison likely introduced selection bias. Moreover, mortality is a poor measure of diabetes prevalence since while diabetes is often a contributing factor, it is not the actual cause of death and so may not be listed on the death certificate. Three international cohorts of workers examined the prevalence of diabetes. New Zealand workers who were employed in a phenoxy herbicide production plant that produced 2,4,5-T were also potentially exposed to the intermediates of TCP and other chlorophenols, and TCDD. 't Mannetje et al. (2018) extended the follow-up period of the 245 men and women in this cohort to approximately 30 years from their last 2,4,5-T production exposure. Comparisons of workers by job title (high- versus low-exposure jobs) found that workers with highly exposed jobs had an elevated risk of doctor-diagnosed diabetes after controlling for age, gender, smoking, BMI, and ethnicity, but when the comparison of diabetes prevalence was based on serum TCDD concentration, no difference in the risk of diabetes was found for workers in the high- versus low-exposure groups, although both of the estimates were imprecise. A cross-sectional study of Japanese incinerator plant workers that included serum dioxin measurements found that the duration of employment was positively associated with TEQ levels (Yamamoto et al., 2015). Diabetes was statistically significantly associated with PCDDs, PCDFs, coplanar PCBs, and total dioxins based on adjusted odds ratios between the first and fourth quartile, and tests for trend were also significant for each of these COIs. Finally, among 235 Italian electric arc furnace workers exposed to multiple agents including metals and PCDD/Fs, PAHs, and PCBs, as sampled in the foundry dust, the prevalence of diabetes among the workers was statistically significantly increased compared with the general population; however, this study is quite limited because foundry dust is a complex mixture, and the authors were unable to discern the impact of the specific contaminants of the foundry dust on the health outcomes of those exposed workers. Estimates were adjusted only for age group and were not adjusted for other risk factors or activities that could affect the association (similar to Cox et al., 2015). It is likely

that many of the occupationally exposed study populations received co-exposures to metals and chemicals other than those that the committee was charged with specifically reviewing. Co-exposure to metals is a possible confounder that may affect the estimates and associations reported in those studies but none of them attempted to adjust for this factor.

Several studies that examined environmental exposures to the COIs or related chemicals were also reviewed. Four of the studies used U.S. populations. An analysis of NHANES data that examined serum concentrations of three chlorinated dibenzo-p-dioxins, one chlorinated dibenzofuran, and four dioxin-like PCBs and diabetes-related nephropathy found that none of the eight dioxin-like chemicals were associated with pre-diabetes (Everett and Thompson, 2016). Two other studies, both of U.S. American Indian populations, were examined for diabetes prevalence (Aminov et al., 2016a,b) and incidence (Grice et al. 2017) with exposure to the COIs. Aminov et al. (2016a,b) found an association between pesticide (including HCB, a dioxin-like chemical) and PCB exposures and diabetes among a sample of people from the Mohowk Nation; however, analyses of individual dioxin-like chemcials or congeners were not presented. In a study of American Indian men and women in Arizona, Grice et al. (2017) examined incident type 2 diabetes with a one-time baseline measurement of serum concentrations of POPs, including dioxin-like PCBs and HCB, and found that the adjusted odds of diabetes were statistically significantly decreased for HCB exposure and that none of the measured dioxin-like PCBs were statistically significantly associated with incident diabetes.

Three international studies of environmental exposures to the COIs and diabetes were reviewed for the current volume. The largest study of diabetes reviewed in this volume was a population-based study of diabetes using residents living in close proximity to a closed PCP-producing factory in Taiwan that had contributed to a high dioxin (PCDD/Fs) contamination of the environment between 1965 and 1979 (C. Y. Huang et al., 2015). Blood samples and health information were collected on 2,898 residents in 2005–2007; 1,143 participants had dioxin levels between 20 and 63 pg WHO98-TEQ_DF/g, and 284 subjects had a dioxin level of 64 pg WHO98-TEQ_DF/g lipid or higher (considered to be the high-exposure groups). Serum dioxin levels of 20–63pg WHO98-TEQ_DF/g were associated with statistically significantly increased odds of diabetes in both crude and adjusted models (adjusted for age, BMI, and years of residence in the contaminated area), and the association was even stronger for serum dioxin levels above 64 pg WHO98-TEQ_DF/g lipid. When the dioxin level was analyzed in terms of quartiles in the model, the adjusted estimates remained elevated and statistically significant, with a statistically significant and steep upward trend (p < 0.001). In a cross-sectional survey of health outcomes that included a blood draw to measure the serum concentrations of PCBs (including dioxin-like mono-ortho congeners 105, 118, and 156) and organochlorine pesticides in 2,172 Canadian Inuits aged 18 years and older, 147 respondents self-reported diabetes

(Singh et al., 2017). All contaminant concentrations were higher in respondents with diabetes. The concentrations of congeners were divided into quartiles, and the adjusted odds ratios for diabetes were found to be statistically significant for quartile 4 versus quartile 1 for PCB 105, PCB 118, and PCB 156. In the third international environmental study, Van Larebeke et al. (2015) conducted a general population study in Beligium and described how dioxin-like activity and serum HCB affected the risk of diabetes in both men and women. They assessed dioxin-like activity in pg TEQ/g fat and found a statistically significant association for both sexes between internal exposure to the COIs as well as dioxin-like activity in pg TEQ/g fat and diabetes or a related condition.

Several lines of toxicologic evidence support mechanisms by which dioxins and dioxin-like chemicals could increase risk for diabetes. First, these chemicals modify the expression of genes related to insulin transport and signaling and to inflammation (Ambolet-Camoit et al., 2015; Kim et al., 2012). Studies in Ahr knockout mice demonstrate increased insulin resistance and glucose tolerance (J. Wang et al., 2011), findings that are consistent with an experiment that found dioxin-exposed mice to have significantly reduced insulin secretion after a glucose challenge (Kurita et al., 2009). More recently, Lai et al. (2017) used a transcriptomic approach to identify the potential molecular mechanisms of TCDD-induced changes in the insulin secretion of pancreatic islets and β cells.

In the aggregate, the epidemiologic studies provide new evidence supporting the previously observed associations between dioxins or dioxin-like PCBs and increased risk for type 2 diabetes. Although some studies had substantial limitations or weaknesses, the Taiwanese study was large and demonstrated that even after adjustment for the factors associated with diabetes in their study, the odds ratios for dioxin-like chemicals equivalents had a strong monotonic trend for higher risk. The committee was divided on whether the newly reviewed studies were adequate to move the level of evidence of association of the COIs and diabetes to sufficient versus leaving the conclusion as limited or suggestive.

Conclusion

Following extensive deliberations regarding the strengths and weaknesses of the new evidence and evidence from studies reviewed in previous VAO reports, the current committee could not reach a consensus on whether the body of literature continues to constitute limited or suggestive evidence of an association between exposure to the COIs and diabetes or whether it now meets the criteria, shown below, for sufficient evidence of an association:

Studies quite consistently show a relationship between exposure to dioxin and dioxin-like chemicals, characterized via serum levels, occupation, or subject self-report, and measures of diabetes health outcomes. These studies include multiple, independent studies of Vietnam veteran populations as well as studies of diverse cohorts of men and women with occupational or environmental exposure to the COIs. Much is known about the risk factors for diabetes, such as age, obesity, and family history, and these have by and large been controlled for in the analyses of most studies reviewed. The studies of Ranch Hand veterans reviewed by previous committees and the Taiwan study examined by this committee demonstrate strong dose–response trends and excellent control for confounding. The disease is, unfortunately, common enough that it has been feasible for a number of investigators to conduct epidemiological investigations in worker or general populations with sufficient statistical power to allow for conclusions to be drawn from the results. Importantly, there is a separate scientific literature that has identified candidate biologic mechanisms that would account for the observed health outcomes in humans.

However, the human evidence base also has its weaknesses. Although positive associations have been observed, some of the relative risks reported are low. A number of studies examined cohorts exposed to mixtures of both dioxin and dioxin-like chemicals and, importantly, a number of other chemicals that could plausibly influence diabetes outcomes. This lack of exposure specificity complicates any attribution of the outcome to the COIs. While most studies adjusted for the primary risk factors for diabetes, several investigations relied on self-reported information that might affect the development of the disease, rendering any adjustment for confounders possibly less effective. The studies of diabetes mortality are of limited utility because death from diabetes, either as a primary or a contributing cause, is underdiagnosed, which could introduce bias. Finally, some committee members felt that it is not yet possible to dismiss the notion that an as yet unidentified systematic bias, including confounding, may be influencing the observed results.

Given these observations, it was not clear to all committee members that a category change was appropriate. Thus, the committee was unable to reach consensus as to whether the evidence regarding exposure to the COIs and diabetes should be classified as “limited or suggestive” or “sufficient.”

CIRCULATORY DISORDERS

Circulatory diseases are a group of diverse conditions, of which hypertension, coronary heart disease, and stroke are the most prevalent. CDC reports that

the number one cause of death for people 65 years of age and older is diseases of the heart. In the United States, 28.4 million adults (11.7% of the population) have a physician-diagnosed heart disease. American Indian and Alaska Natives have the highest incidence of heart disease, estimated at 13.7%, followed by whites with 11.3% and blacks with 9.5% (CDC, n.d.).

In 2016 more than 630,000 people died as a result of diseases of the heart, and more than 140,000 people died as a result of cerebrovascular disease and stroke (CDC, 2016). In addition to family history, the major risk factors for circulatory diseases include age, male gender, smoking, hyperlipidemia, diabetes, and hypertension (World Heart Federation, 2018). Ideally, epidemiologic investigations of circulatory diseases would consider the conditions in this category separately rather than as a group because they all have different patterns of occurrence, and many have different etiologies. However, many mortality studies follow the ICD-9 rubric and report deaths from circulatory diseases together as a single category. Deaths from coronary or ischemic heart disease (IHD), heart failure, and, to a lesser extent, stroke predominate. Many of the reports also break out subcategories such as cerebrovascular disease and hypertension. The American Heart Association reports mortality related to coronary heart disease, not to its symptoms, which include angina and myocardial infarction. The relative importance of heart failure is determined by the age of the cohort. In younger age groups, most of the deaths in this category are expected to be from IHD. In most cases, cerebrovascular deaths are deaths from strokes, which can be classified as either ischemic or hemorrhagic. In the U.S. population, the great majority of strokes are of the ischemic type.

The methods used in morbidity studies of circulatory diseases may involve a direct assessment of the circulatory system, including an analysis of symptoms or history, a physical examination of the heart and peripheral arteries, ultrasound measurements of the heart and arteries, electrocardiography (ECG), chest radiography, cardiac computed tomography (CT), and, more recently, cardiac magnetic resonance imaging (MRI). Carotid ultrasound, echocardiogram, CT, MRI, and nuclear medicine studies can be used to visualize the heart and related vasculature directly as well as to assess function. ECG can be used to detect heart conditions and abnormalities, such as arrhythmias (abnormal heart rhythms), heart enlargement, and ischemia (acute or past myocardial infarctions). Chest radiography can be used to assess the consequences of IHD and hypertension, such as the enlargement of the heart often seen in heart failure. It is sometimes difficult to determine the time of onset of clinical findings, making the temporal relationship between exposure and disease occurrence uncertain. Although the ideal way to study cardiovascular outcomes would be to prospectively following a well-characterized cohort and monitor for the incidence of discrete clinical events, researchers often must rely instead on medical records (which may or may not have all CVD-related information and outcomes) or death certificates (which may list nonspecific causes of death [e.g., respiratory arrest] and may not

contain all comorbid or contributing factors of death). New-onset angina or the performance of a revascularization procedure in a person who has no history of disease is also used as evidence of incident disease.

The committee responsible for Update 2006 began evaluating hypertension separately from other circulatory diseases, and the committee responsible for Update 2008 began the practice of evaluating IHD separately from other cardiovascular outcomes. Beginning in Update 2012, stroke and cerebrovascular disease were also considered separately from discussions of “other circulatory diseases.” The current committee has continued the practice of evaluating the literature related to these outcomes as independent subsections.

A number of studies of different populations that received potentially relevant exposures were identified in the literature search, but the studies did not characterize exposure with sufficient specificity for their results to meet the committee’s criteria for inclusion in the evidentiary database (see Chapter 3), and they are only briefly mentioned under the heading of “Other Identified Studies.” For instance, this rubric would apply to the occupational study conducted by Ruder et al. (2014) in which 24,865 eligible workers from capacitor manufacturing, repair, and maintenance sites in the United States were exposed to arochlor 1254, 1242, and 1016, among others (mixed PCBs), and the authors sought to examine the relationship between PCB exposure and different causes of mortality. However, the study did not name specific dioxin-like PCBs, and no TEQs or other quantification of relevant exposures was presented. Likewise, Ljunggren et al. (2014) assessed serum concentrations of dioxin-like chemicals in patients diagnosed with CVD but did not define how this determination was made, and therefore the work is of little utility to the commmittee.

Table 42, which can be found at www.nap.edu/catalog/25137, summarizes the results of studies related to circulatory disorders.

Biologic Plausibility

Studies have demonstrated that both the vasculature and the adipose tissue are targets of TCDD toxicity, and these studies have provided a mechanistic understanding of how TCDD exposure increases the risk of circulatory diseases, such as hypertension, IHD, and stroke. Exposing cultured endothelial cells or cultured adipocytes to TCDD induces major changes in gene expression and leads to substantial increases in oxidative stress and inflammatory markers (Andersson et al., 2011; S. G. Han et al., 2012; Ishimura et al., 2009; Kerley-Hamilton et al., 2012a; M. J. Kim et al., 2012; Kopf and Walker, 2010; Majkova et al., 2009; Puga et al., 2004; Qin et al., 2015; Zhou et al., 2017). Studies also indicate that the exposure of cultured endothelial cells to TCDD results in the down-regulation of genes involved in blood pressure regulation and the up-regulation of genes involved in the myocardial infarction pathways (Qin et al., 2015) and can promote endothelial cell apoptosis (Q. Liu et al., 2017). In animal models, developmental

exposure to TCDD has been shown to cause structural, molecular and functional cardiac abnormalities, and altered heart physiology in mouse embryos and to predispose adults to cardiac disease (Carreira et al., 2015). Furthermore, the loss of AHR, as happens in Ahr knockout mice, is associated with decreases in blood pressure (modeling hypotension), while sustained activation of AHR resulting from dioxin exposure leads to increases in blood pressure (Agbor et al., 2011). N. Zhang et al. (2010) showed that the genetic loss of AHR from all tissues or solely from endothelial cells results in hypotension. In contrast, Kopf et al. (2010) demonstrated that the chronic exposure of mice to TCDD induces hypertension that is associated with significant increases in vascular oxidative stress and decreases in vascular relaxation. Those changes in vascular function and blood pressure could be mediated in part by increases in the metabolism of arachidonic acid to vasoconstrictive and inflammatory eicosanoids (Bui et al., 2012; Diani-Moore et al., 2014). Studies have also demonstrated that exposure to AHR agonists, including TCDD and benzo[a]pyrene, increases the incidence, severity, and progression of atherosclerosis, a primary cause of IHD and stroke (Dalton et al., 2001; Kerley-Hamilton et al., 2012a; D. Wu et al., 2011). Furthermore, D. Wu et al. (2011) demonstrated that TCDD mediates those effects in part by increasing vascular inflammation, a finding that is suppored by a recent study in chick embryos showing that TCDD-cardiotoxicity is eliminated with exposure to a COX-2 inhibitor (Fujisawa et al., 2014). Another study in human coronary artery smooth muscle cells indicates that the AHR pathway can cooperate with the Tcf21 pathway to initiate the expression of pro-inflammatory genes (J. B. Kim et al., 2017). In addition to the vasculature, studies also suggest that the heart is a target of TCDD. TCDD exposure increases the hypertrophy of rat cardiac cells in culture (Zordoky and El-Kadi, 2010) and impairs the differentiation of mouse embryonic stem cells into cardiomyocytes (Neri et al., 2011).

In addition to the direct effects of TCDD on the vasculature and heart, there is evidence that TCDD influences other CVD risk factors, for example, by promoting obesity (Brulport et al., 2017; Kerley-Hamilton et al., 2012b), accumulating macrophage lipid, inducing lipid mobilization, and altering lipid metabolism. Thus, on the basis of animal models, there appear to be several overlapping and potentially contributing pathways that link TCDD exposure and increased CVD risk.

Arsenic exposure has also been linked to heart effects in a study examining the relationship between the expression of AHR and CYP1A1 in humans exposed to arsenic in drinking water. Data from this study indicate that arsenic exposure was associated with an increased expression of AHR and CYP1A1 in the blood and that increased CYP1A1 was associated with a prolonged corrected QT interval following long-term exposure to arsenic (Cui et al., 2016).

Long-term exposure to oxidative stress is suspected to be etiologic to many chronic diseases, including CVDs. A variety of data, some discussed above, demonstrates a link between TCDD exposure and oxidative stress (reviewed in

Mohsenzadeh et al., 2018). A recent small study by Lerro et al. (2017) indicates a link between 2,4-D and biologic markers of oxidative stress in 30 farmers who participated in the AHS and who applied pesticides occupationally and in 10 non-farming controls who did not apply pesticicides; all were non-smoking men ages 40–60 years. Multivariate linear mixed-effect models for repeated measures were used for each pesticide–oxidative stress marker combination, and all estimates were adjusted for age, farmer or control subject, study time point, and creatinine to account for pesticide metabolites and oxidative stress markers in urine. Other confounders included were BMI, smoking history and duration, alcohol use, regular physical activity, multivitamin or vitamin C supplementation, infection or symptoms at sample collection, allergy symptoms at sample collection, and history of cancer. Farmers had significantly higher urinary 2,4-D levels compared with controls, and this was associated with elevated levels of 8-OHdG, a marker of oxidative stress (β = 0.066, 95% CI 0.008–0.124), and 8-isoPGF, a product of lipoprotein peroxidation (β = 0.088, 95% CI 0.004–0.172). Thus, this study provides one route to plausibility for the 2,4-D association with cardiovascular outcomes, although studies will need to be specifically conducted in individuals with known clinical endpoints (e.g., hypertension) in order to implicate this mechanism directly.

Hypertension

Hypertension, typically defined as blood pressure above 140/90 mmHg, affects more than 70 million adult Americans and is a major risk factor for coronary heart disease, myocardial infarction, stroke, and heart and renal failure. The major quantifiable risk factors for hypertension are well established and include family history, age, sex, race, obesity, reduced nephron number, high dietary salt intake, tobacco use, excessive alcohol intake, and physical inactivity (CDC, 2014b). The strongest conclusions regarding a potential increase in the incidence of hypertension come from studies that have controlled for these risk factors. CDC estimates that in the United States, 64% of men and 69% of women ages 65–74 years have hypertension. When stratified by race/ethnicity and sex, the prevalence of diagnosed hypertension is highest among African American men and women (43.0% and 45.7%, respectively), followed by white men and women (33.9% and 31.3%, respectively). For the groups reported, hypertension is lowest among Mexican Americans (27.8% for men and 28.9% for women) (CDC, 2018).

Conclusions from VAO and Previous Updates

The committee responsible for Update 2006 first concluded that the available evidence placed hypertension in the limited or suggestive category, based primarily on consistent evidence from several studies of Vietnam veterans that used measured serum concentration of dioxin. Additional evidence reviewed

in Update 2008, Update 2010, Update 2012, and Update 2014 reaffirmed this conclusion.

Several studies of hypertension among Vietnam veterans have been reviewed in the VAO series. These have included well-designed studies of incidence, prevalence, or mortality in the U.S. ACC (Cypel and Kang, 2010; Kang et al., 2006) and the AFHS (AFHS, 1995, 2000, 2005) cohorts that have consistently reported increased hypertension with increasing levels of serum dioxin. Other studies of U.S. Vietnam veterans that did not use serum dioxin concentrations as markers of exposure also reported an increased prevalence of hypertension associated with presumed exposure to herbicides.

Among international cohorts of Vietnam veterans, the prevalence of and mortality due to hypertension have been assessed among Australians and South Koreans. A statistically significant increased prevalence was found among the Australian veterans compared with standardized population controls. Two prevalence studies of hypertension among the Korean Vietnam veteran cohort did not find an increased prevalence of hypertension (Yi, 2013; Yi et al., 2014a), but the studies were limited because they do not include veterans who died or relocated between their Vietnam service and the start of the investigation. Thus, the validity of the calculated exposure–outcome relationship is based on the strong assumption that the observed relationships in those included are similar to those who were not included, which is doubtful. In some cases, 40% of the relevant data for the population are missing. In addition, the determination of hypertension was either by self-report or through health insurance claims. It cannot be certain that all participants with hypertension were detected because no standardized blood pressure assessment was done.

Mortality studies that report hypertension are rarely informative because hypertension is so prevalent in the adult population and many more people die with hypertension than from hypertension. For those with hypertension listed as the cause or a contributing cause of death, it is uncertain how representative those who died from hypertension are of all people who may have developed it. Several mortality studies that included hypertension have been reviewed by VAO committees, but they have shown inconsistent findings, likely because they suffer from the limitations listed above as well as from a lack of adjustment for many important confounding factors. A decreased, but not statistically significant, risk of mortality from hypertension was found in the study of U.S. women veterans as well as when the population was limited to nurses only (Kang et al., 2014a), whereas a slight, but again, not statistically significant, increase was reported among Korean veterans (Yi et al., 2014b).

The studies of occupational cohorts reviewed by previous VAO update committees rarely reported hypertension as a discrete outcome; among those that did, the results were mixed, with reports of both increased and decreased risk, but none of the risk estimates were statistically significant. Several studies did not define hypertension, therefore making it difficult to draw conclusions on its

association with the COIs. Similar mixed and not statistically significant findings were reported for the environmental studies that have been reviewed.

Update of the Epidemiologic Literature

Six new studies of exposure to the COIs and hypertension have been published since Update 2014.

Vietnam-Veteran Studies Cypel et al. (2016) analyzed the results of a 2013 survey of 3,086 (80.3% response rate) ACC veterans that compared outcomes in both Vietnam-deployed (n = 1,477) and non-Vietnam-deployed (n = 1,609) herbicide sprayers and non-sprayers. Eligibility for the ACC cohort was restricted to men who had a minimum of 18 months of active U.S. Army service from July 4, 1965, to March 28, 1973, and who were alive in October 2011 and whose health allowed them to participate. ACC veterans were specifically involved in chemical operations in Vietnam and were directly exposed to herbicides and other chemicals, including tear gas and napalm (see description of the ACC veterans in Chapter 5). Participants self-reported physician-diagnosed hypertension, but the diagnosis was evaluated and confirmed by blood pressure measurements taken by trained medical technicians and by medical record reviews for a subset of 468 individuals. Overall agreement between the medical records review and self-reported hypertension was 89%. Additionally, individual exposure to herbicides was collected by self-report to questions of mixing, handling, and spraying herbicides while in the military and was verified by using measured serum TCDD levels for a subset of 636 individuals who had participated in an earlier health survey of the cohort (1999–2000) and who had had blood drawn and serum TCDD (ppt or pg/g lipid) measurements made at that time. A history of herbicide spraying was much higher among the Vietnam-deployed (62.0%) than non-Vietnam-deployed veterans (28.0%) (p < 0.0001). A greater percentage of Vietnam-deployed veterans were current or former smokers (72.3%) than non-Vietnam-deployed veterans (64.0%) (p < 0.0001). Hypertension was highest in Vietnam-deployed sprayers (81.6%), followed by non-Vietnam-deployed sprayers (77.4%), Vietnam-deployed non-sprayers (72.2%), and non-Vietnam-deployed non-sprayers (64.6%). Overall, self-reported hypertension was found in 78.0% of all participants who served in Vietnam and in 68.2% of all participants who did not deploy to Vietnam, indicating a possible effect of Vietnam tour of duty over and above exposure to herbicides. To demonstrate a dose–response relationship, the authors showed that the mean serum TCDD levels were significantly higher among Vietnam herbicide sprayers (mean = 3.5 ppt, range: 0.5–30.6) than among Vietnam non-sprayers (mean = 2.5 ppt, range: 0.7–17.7) (p < 0.0001), but no differences in mean TCDD level were observed between non-Vietnam-deployed sprayers (mean = 2.4 ppt, range: 0.7–9.6) and non-sprayers (mean = 2.2 ppt,

range: 0.4–12.5) (p = 0.69). Using logistic regression and adjusting for Vietnam service status, rank, age at the time of the survey, tobacco use, alcohol use, race, and BMI, having sprayed herbicides (OR = 1.74, 95% CI 1.44–2.11) and having deployed to Vietnam (OR = 1.26, 95% CI 1.05–1.53) were both strongly associated with self-reported hypertension. The association was strongest when comparing Vietnam-deployed sprayers to non-Vietnam-deployed non-sprayers (OR = 2.21, 95% CI 1.76–2.77). Finally, among Vietnam-deployed veterans, a significantly elevated association between the odds of hypertension for sprayers and non-sprayers remained after an adjustment was made for potential confounders (OR = 1.77, 95% CI 1.35–2.30). Similarly, for those veterans who did not deploy to Vietnam, self-reported hypertension was significantly elevated when sprayers were compared with non-sprayers (OR = 1.72, 95% CI 1.31–2.26).

This was a well-designed study with a large sample size and conducted among the most relevant population (Vietnam veterans with known herbicide exposure) which included several levels of exposure (herbicide sprayers and non-sprayers and Vietnam-deployed and non-Vietnam-deployed) and an attempt to quantify it in the participants. Although serum TCDD concentrations were not available for all participants and were collected at least 25 years after Vietnam-era service, for those with serum TCDD levels available, self-reported herbicide spray status had high agreement with the measured levels. The highest mean serum TCDD level was observed among sprayers deployed to Vietnam, and the lowest mean TCDD level was found for non-Vietnam non-sprayers, as would be expected, with a significant dose–response association. Likewise, there was high agreement (89%) between self-reported hypertension and in-person blood pressure measurements and medical records review for a subsample of study participants. The analyses controlled for important risk factors for hypertension, including age, race (white versus others), BMI, tobacco smoking status, rank, Vietnam service status, and alcohol intake, but did not collect information on (and therefore did not control for) other risk factors such as diabetes, a family history of hypertension, and dietary intake of sodium and fat. A major strength of this analysis was using the non-Vietnam-deployed ACC veterans as a comparison group because they were similar to members of the study group with respect to branch, length and time period of service, military occupation, and duties except for deployment in Vietnam, which has the effect of minimizing unmeasured exposures and confounders of concern and bias. Additionally, because all of the men who served in ACC units were stationed at Fort McClellan for at least some time, and Fort McClellan is in close proximity to Anniston, Alabama, where Monsanto operated a plant that produced PCBs, all ACC veterans were likely exposed to at least low levels of these and other chemicals. Therefore, comparisons using deployed and non-deployed ACC men are likely to be biased toward the null due to this baseline of increased exposure, but despite that, the adjusted effect estimate when Vietnam-deployed sprayers were compared with non-Vietnam-deployed non-sprayers was still more than twice as high, precise, and statistically

significant. Although the exact types and quantities of the various chemicals these ACC veterans were possibly exposed to during the Vietnam War are unknown and may include chemicals other than the herbicides (such as insecticides, diesel and jet fuels, cleaning solvents, tear gas, napalm, and antimalarial medications), there is statistically significant support for an association between herbicide exposure and self-reported, physician-diagnosed hypertension.

Occupational Studies Cappelletti et al. (2016) performed a retrospective study of 331 male electric arc foundry workers at a single plant in Trentino, Italy, to determine if they experienced excess mortality from all causes or were at an increased risk for several other disases including complicated (with end organ damage) and uncomplicated (no end organ damage) hypertension, diabetes, and CVD due to their occupational exposure to foundry dust. An analysis of the dust emissions found that the dust contained metals (including iron, aluminum, zinc, manganese, lead, chromium, nickel, cadmium, mercury, and arsenic), PAHs, PCBs, and PCDD/Fs (reported as TEQs). Because foundry dust is a mixture, it is not known which of the agents were associated with a specific outcome or to what extent. The men included in the study had worked at the factory for at least 1 year. For the hypertension analysis, 235 living workers, were compared with the standardized general population of Region Trentino-Alto Adige (where the factory was located) because there were few non-exposed foundry workers and high attrition rates. Company and medical records were used to determine vital status. Requests for exemption health care fees were used as a surrogate measure to identify the most prevalent morbid conditions in the general population, which were then applied to the cohort to compute relative risks for each of the conditions. The workers were followed from March 19, 1979 (or their first day of employment), through December 31, 2009 or date of death. Effect estimates (prevalence ratios) were calculated using Mantel-Haenszel estimator adjusted for age group (20–64, 65–74, ≥ 75 years). Compared with the age-adjusted provincial population, statistically significantly elevated risks of both non-complicated hypertension (n = 30; RR = 2.44, 95% CI 1.75–3.40) and complicated hypertension (n = 14; RR = 2.22, 95% CI 1.35–3.65) were found among the electric arc furnace workers. The strengths of the study include its long-term follow-up of the workers. This study is most limited by the fact that foundry dust is a complex mixture, which made it impossible to discern the impacts of the specific contaminants of the foundry dust on the health outcomes of the exposed workers. Estimates were adjusted only for age group and were not adjusted for other risk factors such as tobacco use, BMI, or other jobs or activities that could result in similar exposures. The possible exposure to foundry dust by the general population that was used for comparison is not discussed, although the foundry appears to be in the local vicinity and emissions from it were reported to be present within a 2-kilometer radius of it.

Yamamoto et al. (2015) performed a cross-sectional study to investigate the health outcomes that 698 male workers sustained while employed at 36 municipal

and private waste incineration plants in Japan; serum dioxin measurements were obtained for 678 of the workers. These workers were employed from 2000 to 2007. First, a questionnaire was completed by participants on lifestyle and medical history. Blood samples were taken from each subject to evaluate serum dioxin, PCDD, PCDF, and coplanar PCB levels (77, 81, 126, 169, 105, 114, 118, 123, 156, 157, 167, and 189). Serum concentrations of dioxin isomers were expressed as pg/g lipid, converted to TEQ, and expressed as TEQ/g lipid. Both clinical and physiological examinations were also performed. Participants were then categorized into four groups: workers whose jobs did not involve working directly in an incineration facility, workers whose jobs did involve work inside the incineration facility (but only handling solidified fly ash and slag or residues that were nonflammable), workers whose jobs involved helping with incineration-related work inside an incineration facility, and workers whose jobs mainly involved the operation and maintenance of an incinerator including a furnace, electric dust collector, and wet scrubber inside an incineration facility. Subjects were tested for diabetes, hypertension, hyperlipidemia, and liver dysfunction. The total serum dioxin concentration level for incinerator workers was 13.7 (median) or 17.2 (arithmetic mean) pg TEQ/g lipid, while the (arithmetic mean) average total dioxin concentration was 19.4 pg TEQ/g lipid for the general population in Japan, indicating no real difference between the two groups. However, in the exposed workers, the duration of employment was positively associated with TEQ levels (9.8 and 21.7 pg TEQ/g lipid for those employed 9 years and 15.5 years, respectively). Coplanar PCBs were significantly correlated (p < 0.05) with 14 parameters of laboratory and physiological tests. Serum concentrations of total dioxins were higher in workers whose jobs involved operation, maintenance, and other incinerator work inside the facilities regardless of the duration of their employment as compared with workers without these job duties. Serum concentrations of dioxin congeners (PCDDs, PCDFs, coplanar PCBs, and total dioxins) were used to divide the participants into quartiles, and odds ratios for hypertension were computed for each category of dioxin congener by comparing each quartile to the lowest quartile (referent) and adjusting for age, survey year, BMI, smoking, and alcohol consumption. Although the adjusted associations between quartiles 2 and 3 and the referent were not consistently significantly elevated, the comparisons of quartile 4 with the referent group for PCDDs (OR = 1.66, 95% CI 1.00–2.78, p < 0.05), PCDFs (OR = 1.90, 95% CI 1.12–3.25, p < 0.05), coplanar PCBs (OR = 2.31, 95% CI 1.33–4.02, p < 0.01), and total dioxins (OR = 1.92, 95% CI 1.12–3.28, p < 0.05) showed elevated, statistically significant associations between the serum levels of PCDD/Fs and total dioxins with the prevalence of hypertension. The test for trend was also significant (p < 0.05) for each of these COIs. Overall, there was no difference in the prevalence of hypertension among the workers of all ages and the Japanese population (44.8% versus 46.2%, respectively). However, when stratified by age groups and compared with the general Japanese population, the prevalence of hypertension was

statistically significantly increased in in workers between the ages of 29 to 49 years. Nonetheless, there were no statistically significant differences in any of the age groups in the total dioxin concentrations between the incinerator workers and the general population, suggesting that something other than dioxin may be contributing to the increased risk for hypertension among this younger population. The strengths of this study include a large sample size, the homogeneity of study subjects with respect to ethnicity and workplace, the measurement of exposure for individuals, significance for each congener, and the adjustments of multiple confounders in the analysis.

Environmental Studies Shiue et al. (2014) used the U.S. population-based NHANES study data from 2011–2012 to examine whether different sets of environmental chemical concentrations were risk factors of high blood pressure. The study used previously collected demographics and blood pressure readings along with the concentrations of various environmental chemicals (14 heavy metals and 20 industrial chemicals, including arsenic compounds and 2,4-D) measured in urine. The urine samples were available for only a subsample of the study population (20–30%). High blood pressure was defined as a systolic blood pressure ≥ 140 mmHg or a diastolic blood pressure ≥ 90 mmHg. The analysis used a total sample size of 9,756 participants who were 20 years of age or older, of whom 3,035 (31.1%) had high blood pressure. This sample is quite young, with 63% of participants being under age 39 years, a group in whom hypertension would be uncommon. The associations of the environmental contaminents with high blood pressure were adjusted for urinary creatinine, age, sex, ethnicity, and BMI and were weighted for the survey design using logistic regression. After full adjustment, even though a number of heavy metals and chemicals were associated with high blood pressure, none of them were the COIs. Of note, however, although total arsenic was marginally associated with high blood pressure in the weighted model (OR = 1.13, 95% CI 0.99–1.29, p = 0.066), trimethylarsine oxide (OR = 2.47, 95% CI 1.27–4.81, p = 0.011) and dimethylarsonic acid concentrations (OR = 1.42, 95% CI 1.12–1.79, p = 0.006) were associated with high blood pressure. Thus, in the general U.S. population, arsenic and its related compounds were found to elevate the risk for hypertension, even in this low at-risk NHANES sample, and the correlation was likely detected because of the large sample size and accurate determinations of exposure concentrations.

Van Larebeke et al. (2015) prospectively studied the associations of exposure to organochlorine pollutants, HCB, dioxin-like (PCB 118) and non-dioxin PCBs, and cadmium with a variety of self-reported health conditions, including diabetes and hypertension, as part of a Flemmish biomonitoring program. In 2004–2005, height, weight, urine, and serum were collected from the participants. Dioxin-like activity in pg TEQ/g fat was also assessed. Subjects filled out a survey collecting demographic data and information on education and tobacco and alcohol use. In 2011, participants in the program were mailed a second survey to collect updated

information on general health, smoking and alcohol consumption behaviors, medical conditions, and current medications. There were 973 respondents, 504 women and 469 men (response rate 65.6%), whose information was used for the current analysis. Serum was assayed for HCB, dioxin-like PCB 118, and other chemicals. Medical diagnoses were all self-reported and not confirmed. For the diagnosis of hypertension, subjects were asked, “Do you suffer or have you suffered from arterial hypertension in the past 10 years?” Although 162 (34.5%) men and 172 (34.1%) women reported hypertension, it is likely underreported in this population of adults aged 50–65 years when recruited in 2002–2006. Overall, there were no differences in the levels of dioxin-like activity (pg/TEQ/g fat) or PCB 118 (ng/g fat) between participants and non-participants of the follow-up survey, emphasizing the representativeness of the sample. However there was a difference for HCB (p = 0.051). In the combined sample, after adjusting for BMI, exercise in minutes per week, level of education, and glasses of alcoholic beverages per week, dioxin-like activity was positively associated with a risk of hypertension (OR = 1.61, p = 0.014), as were HCBs (OR = 1.99, p = 0.0005); PCB 118 was not reported. PCBs with dioxin-like activity approached statistical significance in men only (p = 0.066), while HCBs were significant for both men and women in at least one model. In a crude model, the serum concentration of PCB 118 was not associatied with hypertension. In a model accounting for confounders and with additional adjustments made for correlated exposures (HCB, p,p′-DDE, cadmium, and non-dioxin PCBs), there was a significant association between serum PCB 118 levels and hypertension in men, but not women. The odds ratio for men for the 90th percentile vs. 10th percentile was 3.27 (p = 0.0011). In a model with additional adjustments for covariates and correlated exposures (HCB, p,p′-DDE, cadmium, and non-dioxin PCBs), the association with hypertension was marginally significant for men: the OR for 90th percentile versus 10th percentile was 1.95 (p = 0.09). The strengths of the study included its prospective design, its large sample size, the representativeness of the older adult general population, and the use of objective measures of exposure. However, the findings are limited by the absence of validated medical diagnoses. Although HCB was associated with hypertension in several models, mono-ortho dioxin-like PCB 118 was not associated with hypertension.

P. M. Lind et al. (2014) examined hypertension in participants of the PIVUS (Prospective Investigation of the Vasculature in Uppsala Seniors) cohort, who were men and women aged 70 years or older and residents of Uppsala, Sweden. Of 2,025 eligible subjects, only half (n = 1,016) agreed to participate. Participants completed a questionnaire to assess their medical history, smoking history, and medication use. A clinical exam including blood pressure measurement was performed, and fasting blood work was obtained for lipid and glucose analysis. Hypertension was defined as having a systolic blood pressure > 140 mmHg, a diastolic blood pressure > 90 mmHg, or using antihypertensive medication. General exposure to 23 persistent organic pollutants (POPs) was quantified from

stored serum samples collected at the time of entry into the PIVUS study using high-resolution gas chromatography/mass spectrometry. Normalized POPs were estimated using individualized lipid levels and TEQs measured for dioxin-like PCBs and OCDD combined and separately for the dioxin-like coplanar non-ortho PCBs (PCB 126 and PCB 169) and the dioxin-like mono-ortho PCBs (105, 118, 156, 157, and 189). As the participation rate in this cohort was only 50%, an evaluation of cardiovascular disorders and medications in 100 consecutive non-PIVUS participants was also performed to serve as a control. Although both PIVUS participants and non-participants were broadly similar, the prevalence of diabetes, congestive heart failure, and stroke was higher among non-participants. Overall, 72% of the participants in this study had hypertension. When exposures were adjusted for gender, BMI, smoking status, exercise habits, and education levels, none of the dioxin-like PCBs were associated with prevalent hypertension: PCB 105 (OR = 1.23, 95% CI 0.96–1.60; p = 0.11), PCB 118 (OR = 1.26, 95% CI 0.95–1.67; p = 0.11), PCB 156 (OR = 0.9, 95% CI 0.63–1.3; p = 0.58), PCB 157 (OR = 0.9, 95% CI 0.65–1.26; p = 0.55), PCB 189 (OR = 0.87, 95% CI 0.69–1.09; p = 0.22), PCB 126 (OR = 1.1, 95% CI 0.94–1.3; p = 0.24), and PCB 169 (OR = 0.86, 95% CI 0.61–1.2; p = 0.36). When exposure was classified by individual POPs, PCB 105 (OR = 1.5, 95% CI 1.19–1.89) and PCB 118 (OR = 1.56, 95% CI 1.2–2.01) were significant when adjusted for gender only. This was a large and well-designed and analyzed study of elderly adults that when adjusted for gender, BMI, smoking staus, exercise habits, and education showed no associations between hypertension and dioxin-like PCBs.

Other Identified Studies One other occupational study was identified that reported deaths from hypertension with underlying heart disease, but it was limited by a lack of exposure specificity (Ruder et al., 2014). An environmental study was also identified (Akahane et al., 2017) that examined the prevalence of self-reported long-term health outcomes (including high blood pressure) in people exposed to PCBs, dioxins (e.g., PCDD/Fs), and dioxin-like chemicals through the ingestion of contaminated rice bran oil (Yusho accident) compared with an age-, sex- and residential-area-matched group. Because no TEQs or other quantification of relevant exposures was presented, the study was not considered further.

Biologic Plausibility

The biological mechanism for dioxin’s impact on hypertension is being investigated in animal models and human cell cultures, and it has shown clear effects on gene expression, vascular function, and lipid glucose metabolism. The exposure of human endothelial cells to TCDD causes the down-regulation of genes involved in blood pressure regulation (Qin et al., 2015), suggesting a potential molecular mechanism for TCDD’s impact on hypertension. Recent data also indicate that the increase in hypertension following AHR activation is linked

to the inactivation of endothelial nitric oxide synthases, which is expressed in endothelial cells and functions to modulate vascular blood pressure (C. C. Chang et al., 2017). Data also demonstrate a link to the Ahr pathway using mouse models, demonstrating that sustained Ahr activation by dioxins results in increased blood pressure, which is associated with significant increases in vascular oxidative stress and decreases in vascular relaxation (Kopf et al., 2010). Conversely, hypotension is associated with Ahr loss in knockout models, either knocked out in the whole animal or specifically in endothelial cells, (Agbor et al., 2011; N. Zhang et al. 2010). Those changes in vascular function and blood pressure could be mediated in part by increases in the metabolism of arachidonic acid to vasoconstrictive and inflammatory eicosanoids (Bui et al., 2012; Diani-Moore et al., 2014).

A recent study on genetic polymorphisms of the AHR signaling pathway genes in a human cohort recruited from central Russian (Kursk) cardiology and neurology clinics also suggests a plausible link between exposure to TCDD and dioxin-like chemicals and essential hypertension (Polonikov et al., 2017). Seven common polymorphisms in the AHR pathway genes AHR, ARNT, AHRR, CYP1A1, CYP1A2, CYP1B1, and NQO1 were genotyped from venous blood samples of 1,341 cases and 819 controls. A polymorphism in ARNT was shown to be associated with increased essential hypertension, whereas a single nucleotide polymorphism in CYP1A2 showed a decreased risk of essential hypertension in a recessive genetic model. Additional log-likelihood ratio tests showed epistatic interactions on essential hypertension susceptibility for all single nucleotide polymorphisms. These results are generally consistent with the suggestion that the AHR pathway is involved in hypertension and that TCDD activation of this pathway can influence hypertension in the human population.

Synthesis

Hypertension, defined as a systoloic/diastolic blood pressure exceeding 140/90 mmHg, affects approximately 75 million Americans, or one in every three adults. This trait remains one of the main contributing risk factors to cardiovascular, peripheral vascular, and cerebrovascular disease. Risk factors include family history, age, sex, race, obesity, reduced nephron number, high dietary salt intake, tobacco use, excessive alcohol intake, and physical inactivity. Owing to its frequency, assessing whether there is increased risk with exposure to the COIs has been challenging. However, the committee for the current update believes that there are enough new data to move the category of association to sufficient evidence.

Cypel et al. (2016) offers the most compelling evidence for the change. The well-designed study was conducted in the population of interest, U.S. Vietnam veterans, had a large sample size, appropriate controls (non-Vietnam-deployed sprayers, Vietnam non-sprayers, and non-Vietnam-deployed non-sprayers), and validated endpoints (self-reported physician-diagnosed hypertension that was

confirmed with measurements and medical record reviews). The study also quantified exposures using serum TCDD measurements that were validated in a subset of participants. Finally, the statistical analyses conducted are robust, used state-of-the art methods, and adjusted for appropriate confounders. This study clearly demonstrated that self-reported physician-diagnosed hypertension rates were the highest among Vietnam-deployed sprayers (81.6%) compared with non-Vietnam-deployed sprayers (77.4%), Vietnam-deployed non-sprayers (72.2%), and non-Vietnam-deployed non-sprayers (64.6%), representing a significant association with exposure.

The five additional studies reviewed were occupational (Cappeletti et al., 2016; Yamamoto et al., 2015) and environmental (Lind et al., 2014; Shiue et al., 2014; Van Larebeke et al., 2015) exposure investigations. Each of these has one or more significant study design deficiencies as compared to Cypel et al. (2016) and would not be considered adequate to change the level of association individually. However, at least a portion of the effect model results corroborate the positive, elevated risk between exposure to the COIs and hypertension using a variety of study designs, populations, and measurements of exposure.

Briefly, among the occupational cohorts, Yamamoto et al. (2015) found hypertension to be statistically significantly associated with serum concentrations of PCDDs, PCDFs, dioxin-like PCBs, and total dioxins based on adjusted odds ratios between the first and fourth quartile of each of those COIs among incinerator workers, and tests for trend were also significant for each of these COIs. Among 235 Italian electric arc furnace workers exposed to multiple agents, including metals and PCDD/Fs, PAHs, and PCBs, as sampled in the foundry dust, the prevalence of both complicated and non-complicated hypertension among the workers compared with the general regional population showed statistically significant increased risks; however, this study is quite limited because foundry dust is a complex mixture, and the authors were unable to discern the impact of the specific contaminants of the foundry dust on the health outcomes of those exposed workers (Cappelletti et al., 2016). Estimates were only adjusted for age group, and were not adjusted for other risk factors or activities that could affect the association. It is likely that workers of both of these occupationally exposed study populations received co-exposures to metals and chemicals other than those that the committee was charged with specifically reviewing that may be possible confounders that may affect the true estimate of association.

The results of the three environmental exposure studies had mixed findings. In their analysis of 2011–2012 NHANES data, Shiue and colleagues (2014) found an increased risk of hypertension with urinary levels of a number of heavy metals and other chemicals, including trimethylarsine oxide and dimethylarsonic acid, after adjusting for several risk factors of hypertension. Among the participants of a Flemish biomonitoring program, Van Larebeke et al. (2015) found an elevated risk of self-reported (and unconfirmed) hypertension for dioxin-like activity and HCB after adjusting for BMI, exercise in minutes per week, level of education,

and glasses of alcoholic beverages per week that was significant for men and women combined. However, when stratified by sex, the risk of hypertension was only marginally significant for men exposed to PCBs with dioxin-like activity, while HCBs were significant for both men and women in at least one model. In the third environmental exposure study of elderly residents of Upsala, Sweden, significant positive associations between hypertension and dioxin-like PCBs were found only for PCB 105 and PCB 118 when analyses were adjusted for gender only. When models were fully adjusted for the risk factors of hypertension, no association between dioxin-like PCBs and hypertension was found (Lind et al., 2014).

In addition to the new studies, the committee re-examined the studies reviewed from the previous VAO reports, specifically those among Vietnam veterans (AFHS, 1995, 2000, 2005; CDC, 1988a; Cypel and Kang, 2010; Kang et al., 2006, 2014; J. S. Kim et al., 2003; O’Toole et al., 2009; Yi et al., 2014a,b) when making its decision of the strength of the evidence of exposure to the COIs and hypertension. Furthermore, data from animal and human cell-culture models support the hypothesis that TCDD activation of AHR increases the development of hypertension, and suggest plausible molecular mechanisms for hypertension.

Conclusion

After an examination of the literature that had been previously reviewed with the additional new evidence, the committee determined that there is sufficient evidence of an association between at least one of the COIs and hypertension.

Ischemic Heart Disease

IHD refers to a loss of blood flow and lack of oxygen to the heart muscle. It is also referred to as cornonary heart disease or myocardial ischemia and includes the conditions of stable angina, unstable angina, myocardial infarction, and sudden cardiac death. It is often the result of an atherosclerotic narrowing of the blood vessels that supply the heart muscle. Risk factors include smoking, hypertension, hyperlipidemia, obesity, family history, age, and male sex. It is one of the leading causes of death in the United States, at a rate of 97.2 per 100,000 deaths (Healthypeople.gov, 2018a). IHD was first addressed as a separate oucome in Update 2008.

Conclusions from VAO and Previous Updates

The committee responsible for Update 2008 revisited the entire body of evidence on TCDD exposure and heart disease and concluded that the evidence supported moving IHD to the limited or suggestive category. That conclusion was based on evidence of a dose–response relationship in the occupational cohorts,

evidence of an increased risk of myocardial infarction in Vietnam veterans, supporting cross-sectional survey data, and a strong biologic rationale. Evidence reviewed for Update 2010 and Update 2012 continued to support that classification. A number of studies of potential relevance were reviewed for Update 2014, including several studies of Vietnam veterans. The studies of New Zealand, Korean, and female U.S. veterans did not find an increase in IHD mortality (Kang et al. 2014a; McBride et al. 2013; Yi et al. 2013a, 2014a,b). However, because IHD is not a uniformly fatal disease, mortality rates may not be the most accurate marker of prevalence. Information on IHD gleaned from death certificates represents not only disease occurrence but also disease severity—the validity of which depends on those dying of the disease being a fair representation of all persons developing IHD. For example, in the Korean Veterans Health Study of veterans who served in Vietnam, the mortality analysis included 843 deaths from IHD, but analyses of disease prevalence in the same population identified more than 20,000 persons with this condition.

Studies comparing mortality among veteran populations to that among the general population may also be biased by the so called “healthy warrior effect,” in which veterans have a health advantage over the general population across a range of health outcomes. Furthermore, only the study of Korean veterans quantified possible herbicide exposure, whereas the New Zealand and U.S. veteran studies assumed deployment to be synonymous with herbicide exposure. Each of the Vietnam veteran cohort studies was limited by not adjusting the estimates for various relevant confounding variables. Additional studies of CVD, which includes IHD, in well-conducted and nationally representative samples, such as NHANES (Lin et al., 2012), found an increase in the risk of CVD death in those with higher levels of dioxin TEQs, but other studies of CVD did not demonstrate such relationships; thus IHD has remained in the limited or suggestive category.

Update of the Epidemiologic Literature

Results of new studies of IHD and exposures to the COIs are summarized below. No case-control studies of exposure to the COIs and IHD have been published since Update 2014.

Vietnam-Veteran Studies Cox et al. (2015) used hospital discharge records from 1988 to 2009 to identify prevalent health conditions in 2,783 male New Zealand veterans who served in Vietnam. Age-specific hospitalization rates were calculated using the total number of annual hospitalizations published by the Ministry of Health and the average annual resident population. Standardized hospitalization rates and 99% CIs were calculated for the veteran cohort and the general population and reported for CVD overall as well as for acute myocardial infarction, coronary atherosclerosis, chest pain, cardiac arrest, dysrhythmia, and congestive heart failure. CVD was broken down into

acute myocardial infarction (SHR = 1.16, 99% CI 1.00–1.33), coronary atherosclerosis (SHR = 1.27, 99% CI 1.14–1.39), chest pain (SHR = 1.35, 99% CI 1.16–1.53), cardiac arrest (SHR = 1.54, 99% CI 0.22–2.86), dysrhythmia (SHR = 1.27, 99% CI 1.06–1.48), and congestive heart failure (SHR = 1.02, 99% CI 0.78–1.26). Acute myocardial infarction, coronary atherosclerosis, and chest pain were further examined by 4-year periods of time. In each case, the time period 2006–2009 had the highest SHR, which was also statistically significant. The authors concluded that there was a small but significant increase in the number of hospitalizations for New Zealand Vietnam veterans, with modest increases in hospitalization for common conditions such as CVD. Exposure to the COIs was not validated through serum measurements, and the study did not control for smoking or ethnicity or other potentially important risk factors.

Occupational Studies Collins et al. (2016) offers additional follow-up time to a retrospective analysis of a cohort of 2,192 workers exposed to dioxins during TCP and PCP production at a chemical manufacturing plant in Michigan. The U.S. population was used as the comparator for standardized mortality ratios. Work history records were used to determine the length of exposure. Serum samples to measure levels of six types of dioxins, which were collected from 431 TCP and PCP workers. The historical concentrations for each dioxin congener were calculated from the median concentrations from the serum samples and the known half-lives associated with each congener. Complete vital status follow-up was achieved for the cohort, and there were 1,198 decedents through the entire study period (1979–2011). For IHD, there were 371 deaths among all workers combined (SMR = 1.10, 95% CI 0.99–1.22). No difference in mortality from IHD was found among the TCP workers (n = 256; SMR = 1.07, 95% CI 0.95–1.21), but there was a slight increase of death from IHD among the PCP workers (n = 150; SMR = 1.20, 95% CI 1.01–1.41). When the SMRs for IHD were evaluated by exposure levels of each constituent congener, only the middle estimates for HpCDD and OCDD were statistically significantly elevated, but the trend was not.

't Mannetje et al. (2018) conducted a morbidity survey among a subset of workers who were employed at the New Plymouth, New Zealand, phenoxy herbicide production plant for at least 1 month between 1969 and 1984. The plant produced 2,4,5-T, and workers were potentially exposed to 2,4,5-T, intermediates of TCP and other chlorophenols, and TCDD. Workers had previously been recruited and examined as part of the international cohort of producers of phenoxy herbicides led by IARC (Kogevinas et al., 1997); see Chapter 5 for more details on the IARC cohort and the New Zealand phenoxy producers. This study extended the follow-up period of these workers to approximately 30 years from the final 2,4,5-T production exposure. From the original cohort of 1,025 workers, 631 were living, had a current address in New Zealand, and were below 80 years of age on January 1, 2006. For the current 't Mannetje et al. follow-up, 430 or the

631 workers were randomly selected and invited to participate in a morbidity survey, of whom 245 (57%) participated. The survey was administered in 2007–2008 by face-to-face interviews and collected information on demographic factors and health information, including doctor-diagnosed conditions and the year of diagnosis. A blood sample was also collected at that time and analyzed for TCDD, lipids, thyroid hormones, and other parameters. A neurological examination was conducted for 111 of the participants. Associations between exposure and health outcomes were assessed using logistic regression models that controlled for age, gender, smoking, BMI, and ethnicity using two methods: working in a TCDD-exposed job (based on occupational records) and serum TCDD concentration ≥10 pg/g lipid (18%). Mean TCDD concentrations were 19 pg/g lipid in the 60 men directly involved in phenoxy/TCP production and 6 pg/g lipid in the 141 men and 43 women who worked in other parts of the plant. Compared with the people in the not highly exposed jobs, the people who had ever worked in a highly exposed job at the plant did not have any difference in risk of doctor-diagnosed heart disease (n = 10; OR = 2.71; 95% CI 0.65–11.4), although the estimate was imprecise. When compared by serum TCDD concentration, again there was no difference in the risk of heart disease for workers in the high- versus low-exposure groups (n = 5; OR = 1.64; 95% CI 0.41–6.57), although the estimate was again imprecise.

Cappelletti et al. (2016) performed a retrospective study of 331 male electric arc foundry workers at a single plant in Trentino, Italy, to determine if they experienced excess mortality from all causes or were at an increased risk for several other disases including CVD, due to an occupational exposure to foundry dust. An analysis of the dust emissions found that the dust contained metals (including iron, aluminum, zinc, manganese, lead, chromium, nickel, cadmium, mercury, and arsenic), PAHs, PCBs, and PCDD/Fs (reported as TEQs). Because foundry dust is a mixture, it is not known which of the agents were associated with a specific outcome or to what extent. The men included in the study had worked at the factory for at least 1 year. Company and medical records were used to determine vital status. Requests for exemption health care fees were used as a surrogate measure to identify the most prevalent morbid conditions in the general population, which were then applied to the cohort to compute relative risks for each of the conditions. The workers were followed from March 19, 1979 (or their first day of employment), through December 31, 2009, or date of death. Effect estimates (prevalence ratios) were calculated using Mantel-Haenszel estimator adjusted for age group (20–64, 65–74, ≥ 75 years). Compared with the age-adjusted provincial population, no difference in death from IHD (n = 4; SMR = 1.27, 95% CI 0.35–3.26) was found among the electric arc furnace workers. However, a statistically significantly elevated risk of CVD (n = 5; RR = 1.74, 95% CI 1.07–2.82) was found. This study is most limited by the fact that foundry dust is a complex mixture, which makes it impossible to discern the impact of the specific contaminants of the foundry dust on the health outcomes of those

exposed workers. Estimates were adjusted only for age group and were not adjusted for other risk factors such as tobacco use, BMI, or other jobs or activities that could have resulted in similar exposures. Exposure to foundry dust by the general population that was used for comparison is not discussed, although the foundry appears to be in the local vicinity and emissions from it were reported to be present within a 2-kilometer radius of it.

Environmental Studies Van Larebeke et al. (2015) carried out a prospective study of the associations of exposure to organochlorine pollutants, HCB, dioxin-like (PCB 118) and non-dioxin-like PCBs, and cadmium with a variety of self-reported health conditions, including diabetes, hypertension, and atheromata (the results for diabetes and hypertension were described earlier) as part of a Flemmish biomonitoring program. In 2004–2005, height, weight, urine, and serum were collected from the participants. Dioxin-like activity in pg TEQ/g fat was also assessed. Subjects filled out a survey that collected demographic data and information on education, tobacco use, and alcohol consumption. In 2011, participants in the program were mailed a second survey to collect updated information general health, smoking and alcohol consumption behaviors, medical conditions, and current medications. There were 973 respondents, 504 women and 469 men (response rate 65.6%), whose information was used for the current analysis. The serum was assayed for HCB, dioxin-like PCB 118, and other chemicals. Medical diagnoses were all self-reported and not confirmed. For the diagnosis of atheromata, subjects were asked, “Did you have a problem with the blood circulation in the brain (cerebrovascular incident, ischemic episode)?” “Have you ever experienced a heart infarction?” and “Do you sometimes suffer from pain in the thorax during physical effort?” A positive response to at least one of these questions was considered a diagnosis. A total of 71 men (15.1%) and 62 women (12.3%) reported atheromata, and 27 men (5.8%) and 10 women (2%) reported myocardial infarction. Overall, there were no differences in the levels of dioxin-like activity (pg/TEQ/g fat) or PCB 118 (ng/g fat) in participants versus non-participants of the follow-up survey, but there was a difference for HCB (p = 0.051). In the combined sample, after adjusting for BMI, the level of education, and cholesterol concentration in blood, the researchers found that dioxin-like activity was not associated with the risk of atheromata (OR = 1.60, p = 0.083); estimates for HCBs and PCB 118 were not provided. In a model accounting for confounders and making additional adjustments for correlated exposures (HCB, p,p’-DDE, cadmium, and non-dioxin PCBs), the odds of atheromata in the 90th versus the 10th percentiles were increased in men only (OR = 1.83, p = 0.031). No individual analyses were performed for myocardial infarction. The strengths of the study include its prospecitive design, its large sample size, the representativeness of the older adult population of Flanders, and the use of objective measures of exposure. However, the findings are limited by the study’s lack of validated medical diagnoses, and the data were collected based on self-report from a survey that was not validated.

Bergkvist et al. (2015) performed a prospective population-based cohort study of 33,446 women who participated in the Swedish Mammography Cohort and who were free of CVD, diabetes, and cancer at baseline. The women were followed for 12 years to determine the association between dietary PCB exposure and the risk of myocardial infarction. Fish fatty acid consumption was also measured to determine any protective effects of the consumption of eiconsapentaenoic acid and docosahexaenoic acid. The food frequency questionnaire-based dietary PCB estimates have been extensively validated against serum PCBs in women from this cohort. During the follow-up period, 1,386 myocardial infarctions occurred, 276 of which were fatal. Cases were ascertained through a computerized linkage to the National Hospital Discharge and Cause of Death Registers in Sweden using personal identification numbers. PCB exposure was grouped into quartiles, and women in the highest quartile of dietary PCB exposure were found to be more likely to report high cholesterol levels, to use fish oil supplements, to have a slightly higher consumption of alcohol and red and processed meat, and to have up to a six-times-higher consumption of fish than those in the lowest quartile. Multivariate-adjusted analysis was used to determine the association between risk of myocardial infarction and PCB exposure. Models were adjusted for several factors, including postsecondary education, a family history of myocardial infarction before the age of 60 years, ever use of postmenopausal hormones, the use of aspirin, the use of fish oil supplements, and a weight loss of ≥ 5 kg within 1 year, which were treated as dichotomous variables (yes/no). Additional variables were also included in Cox proportional hazards regression models: smoking status (never, past, or current); waist circumference (< 80, 80–87, ≥ 88 cm); parity (0, ≥ 1 children); total physical activity (quartiles, MET-h); alcohol consumption (0, > 0–4.9, 5.0–14.9, > 15.0 g/day); energy intake (continuous, kcal/day); consumption of fruit and vegetables (quartiles, servings/week), dairy products (quartiles, servings/day), and red and processed meat (quartiles, servings/week); dietary intake of saturated fatty acids (quartiles, g/day); and dietary methyl mercury exposure (quartiles, μg/day). In additional adjustments, the dietary sum of both eiconsapentaenoic acid and docosahexaenoic acid (quartiles, g/day) was also included. Models were not adjusted for hypertension or high cholesterol levels because the authors reasoned that these intermediate risk factors may be in the causal pathway between PCB exposure and CVD. Dietary PCB exposure was associated with total myocardial infarction in the multivariable-adjusted model (RR = 1.21, 95% CI 1.01–1.45; p for trend = 0.012), and this estimate increased after adjusting the model for the sum intake of eiconsapentaenoic acid and docosahexaenoic acid (RR = 1.58, 95% CI 1.10–2.24; p for trend = 0.007). There are several limitations to the study. Although the survey was validated with a serum PCB analysis in this population, recall bias is of concern, as is the fact that this information was collected only at baseline, yet the cohort was followed for 12 years. Using a prospective design eliminates concerns of temporal ambiguity. Methyl mercury is a potentially confounding factor. Though adjustments were made for dietary intake, it is not possible to ascertain the role of

mercury in risk of CVD. Most importantly, because exposure to PCBs is grouped, it is not possible to know if any of the increased risk is due to dioxin-like PCBs and which specifically those might be.

Other Identified Studies One other study that reported deaths from IHD was identified, but it was limited by its lack of exposure specificity (Ruder et al., 2014).

Biologic Plausibility

Experimental studies demonstrate that exposure to AHR agonists, including TCDD and benzo[a]pyrene, increase the incidence, severity, and progression of atherosclerosis, a primary cause of IHD (Dalton et al., 2001; Kerley-Hamilton et al., 2012a; D. Wu et al., 2011). There is also evidence that TCDD influences other IHD risk factors by promoting obesity (Brulport et al., 2017; Kerley-Hamilton et al., 2012b), accumulating macrophage lipid, inducing lipid mobilization, and altering lipid metabolism. Thus, on the basis of animal models there appear to be several overlapping and potentially contributing pathways that link TCDD exposure and increased IHD risk.

Synthesis

This committee considered six new studies in addition to the multiple studies reviewed by prior VAO committees when determining its conclusion of the level of association. Among New Zealand veterans who had served in Vietnam, Cox et al. (2015) calculated hospitalization ratios for CVD and found elevated rates for the outcomes of acute myocardial infarction, chest pain, and coronary atherosclerosis among the veterans. The two occupational studies of heart disease mortality—one among U.S. workers producing TCP and PCP (Collins et al., 2016) and the other among Italian electric arc foundry workers exposed to dioxins (Cappelletti et al., 2016)—both found no difference in mortality from heart disease compared with their resepective general populations; however the use of general populations as a comparator likely introduced selection bias. A third occupational study of morbidity among people who had worked in a New Zealand plant that produced phenoxy herbicides also found no increased risk of heart disease based on high versus low exposure ('t Mannetje et al., 2018). Two studies of populations with environmental exposures were also examined. Bergkvist et al. (2015) followed a cohort of women for 12 years who were free of CVD at baseline and who were exposed to PCBs via dietary exposure. Even though adjustments were made for dietary intake, PCBs were grouped together, making it difficult to determine the role of individual dioxin-like PCBs in the CVD risk. Regardless, these data are not immediately generalizable, given that very few women were exposed to the herbicides. The occupational studies are further limited because

these populations of workers likely received co-exposures to metals and chemicals other than those that the committee was charged with specifically reviewing that may be possible confounders that may affect the true estimate of association. Van Larebecke et al. (2015) studied the association of exposure to organochlorine pollutants, HCB, and dioxin-like and non-dioxin-like PCBs with a self-reported history of atheromata and myocardial infarction. While there was a significant association for men reporting atheromata, this association lost significance when the analysis corrected for other risk factors known to cause atheromata. Furthermore, based on a review of available experimental data, there appear to be several overlapping and potentially contributing pathways that link TCDD exposure and an increased risk of IHD.

The above studies were limited by relatively small numbers of cases, a lack of consistent case definitions for the various types of CVD (IHD versus angina versus chest pain versus atheromata), the use of non-validated surveys for diagnosis, and the presence of mixtures of chemicals in addition to the COIs. Most of the new studies used mortality or hospitalization data, which, as discussed above, may not be the best endpoint for determining an increased risk for a particular condition. Given these considerations, the committee reaffirmed the decision of previous VAO committees to keep IHD in the limited or suggestive category.

Conclusion

After reviewing the new evidence, the present committee concurred with those for previous updates that there is limited or suggestive evidence that IHD is associated with exposure to the COIs.

Cerebrovascular Disease and Stroke

Cerebrovascular disease refers to a disorder that affects the circulation in the brain. The most common conditons are ischemic stroke, hemorrhagic stroke, and transient ischemic attack. This disease is one of the top 10 causes of death in the United States and was responsible for 5.2% of deaths overall in 2015 (CDC, 2017b). Presented by race, in 2015 whites had the greatest number of deaths from cerebrovascular disease (~117,000) followed by blacks (~18,000), Asian and Pacific Islanders (~5,000), and American Indians/Alaska Natives (~700) (CDC, 2017b). Mortality rates from cerebrovascular disease and stroke were declining, but have remained stagnant in recent years (CDC, 2017d).

Conclusions from VAO and Previous Updates

Among U.S. Vietnam veterans, stroke and cerebrovascular disease have not been found to be elevated in the AFHS, ACC, or female veteran cohorts (Cypel and Kang, 2010; Kang et al., 2014; Ketchum and Michalek, 2005). Similar to

the analysis of deployed and non-deployed U.S. female veteans, analyses of Australian Vietnam veterans compared with the general Australian population have found a decreased risk of stroke and cerebral hemorrhage. Among Korean Vietnam-era veterans, after adjustment for multiple behavioral, demographic, and service-related factors, stroke prevelance was statistically significantly elevated among the more highly exposed cohort members (Yi et al., 2014a). The association was seen for both major stroke types: cerebral infarction (ICD-10 I63) and cerebral hemorrhage (ICD-10 I60–I62). In the mortality study in the same cohort, after adjusting for only age and rank, no differences were found in mortality from cerebrovascular disease (ICD-10 I60–I69) between the groups with high and low potential for herbicide exposure or when individual exposure potential was used as a continuous variable (Yi et al., 2014b).

Few occupational cohorts have examined stroke and cerebrovascular diseases, but those that have, have reported very few cases and very few differences in mortality rates. Such studies have included Dutch herbicide production workers (Boers et al., 2010), the U.S. National Institute for Occupational Safety and Health mortality cohort (Steenland et al., 1999), PCP production workers in Midland, Michigan (Collins et al., 2009a), and Italian licensed pesticide users (Gambini et al., 1997). One study of AHS applicators and their wives (Waggoner et al., 2011) found a statistically significantly decreased risk of cerebrovascular disease.

Only two environmental studies have reported on stroke or cerebrovascular disease. Y. S. Lin et al. (2012) reported a positive, but not significant association between dioxin-like compounds in the blood and mortality from cerebrovascular disease (defined as a stroke) in an analysis of NHANES data. In the 25-year follow-up of the Seveso cohort, Consonni et al. (2008) found no difference in cerebrovascular disease between people exposed in Zone A and nonexposed residents of 11 surrounding municipalities, but slight statistically significant increases in cerebrovascular disease were found for people residing in Zones B and R compared with the nonexposed residents of 11 surrounding municipalities.

Therefore, based on these studies the evidence to determine an association between exposure to the COIs and cerebrovascular disease and stroke remains limited or suggestive.

Update of the Epidemiologic Literature

Four new studies of exposure to at least one COI and cerebrovascular disease and stroke outcomes have been identified since Update 2014.

Vietnam-Veteran Studies Cox et al. (2015) used hospital discharge records from 1988 to 2009 to identify prevalent health conditions in 2,783 male New Zealand veterans who served in Vietnam. Age-specific hospitalization rates were calculated using the total number of annual hospitalizations published by the Ministry of Health and the average annual resident population. Standardized

hospitalization rates and 99% CIs were calculated for the veteran cohort and the general population and reported for several noncancerous conditions. A total of 170 cases of acute cerebrovascular disease were found in the veteran population, which was statistically significantly elevated compared with the standardized hospitalization rate of New Zealand (SHR = 1.31, 99% CI 1.05–1.56). In addition, a total of 45 cases of aneurysm within the veteran population were identified (SHR = 1.47, 99% CI 0.91–2.04). The study was limited by its assumption that deployment to Vietnam was synonomous with herbicide exposure, which was not validated through serum measurements, and by a lack of adjustment for smoking or ethnicity or other potentially important risk factors.

Han et al. (2016) conducted a small hospital-based study of Korean Vietnam veterans in Seoul who had presumed Agent Orange exposure (determined by deployment to Vietnam) to examine whether exposure resulted in differences in the vascular features of acute ischemic stroke. A total of 91 veterans with preseumed exposure were compared with 288 male patients from other area general hospitals (and presumed to be unexposed to Agent Orange). Both groups of patients were compared on lifestyle factors, clinical history, and the clinical manifestations of stroke at admission and discharge. Vascularity was examined within 7 days of acute ischemic stroke onset, and stroke was confirmed by MRI within 1 week after onset. The subtype of the stroke was assessed using the acute stroke treatment protocol and the modified Rankin Scale at discharge and at 3 months after onset. The two groups were comparable for many demographic and lifestyle factors except that veterans (exposed) had higher rates of smoking and comorbid diabetes and the unexposed males had higher hyperlipidemia and history of stroke. Stroke subtypes were statistically significantly different (p = 0.014) between the two groups, and characterized by small vessel occlusion in exposed subjects and large artery atherosclerosis in unexposed subjects. These features may have led to a somewhat better short-term prognosis of exposed than unexposed members. The exposed veterans had lower scores on the stroke scale at admission than the control group (p = 0.003), but there was no significant difference between the two groups at discharge. This study was small and exposure was assumed based on deployment to Vietnam and not otherwise objectively measured or validated. Only simple comparison testing was used, but the study does indicate that studying stroke as a single category may lose statistical power especially when attempting to study the effects of potential exposure to the COIs.

Occupational Studies Collins et al. (2016) offered additional follow-up time to a retrospective analysis of a cohort of 2,192 workers exposed to dioxins during TCP and PCP production at chemical manufacturing plant in Michigan. The U.S. population was used as the comparator for standardized mortality ratios. Work history records were used to determine the length of exposure. Serum samples were collected from 431 TCP and PCP workers to measure the levels of six types of dioxins. The historical concentrations for each dioxin congener

were calculated from the median concentrations from the serum samples and the known half-lives associated with each congener. Complete vital status follow-up was achieved for the cohort, and there were 1,198 deaths through the entire study period (1979–2011). For cerebrovascular disease, there were 68 deaths among all workers combined and no difference in the mortality rate compared with the general population (SMR = 1.03, 95% CI 0.80–1.30). No difference in mortality from cerebrovascular disease was found among the TCP workers (n = 49; SMR = 1.07, 95% CI 0.79–1.41) or the PCP workers (n = 27; SMR = 1.07, 95% CI 0.701–1.55). Although serum dioxin measurements were collected, no results based on those measurements were presented for cerebrovascular disease; instead the authors used employment records to categorize exposure, which may have introduced exposure misclassification, and use of the general U.S. population as a comparison likely introduced selection bias.

Environmental Studies Van Larebeke et al. (2015) carried out a prospective study of the associations of exposure to organochlorine pollutants, HCB, dioxin-like (PCB 118) and non-dioxin-like PCBs, and cadmium with a variety of self-reported health conditions, including atheromata, as part of a Flemmish biomonitoring program. Because a diagnosis of atheromata was assessed based on three questions, one of which was “Did you have a problem with the blood circulation in the brain (cerebrovascular incident, ischemic episode)?” the results for atheromata were described in the section on IHD.

Other Identified Studies One other study that reported deaths from cerebrovascular disease was identified, but it was limited by a lack of exposure specificity (Ruder et al., 2014).

Biologic Plausibility

Although no papers were identified that specifically addressed cerebrovascular disease and stroke following exposure to the COIs, many of the mechanisms that influence stroke and cerebrovascular disease incidence and progression have been shown to be targets of TCDD and the AHR pathway, as noted in the introduction of the biologic plausibility section under the circulatory diseases heading. These mechanisms include impacts on endothelial cell function and proliferation, inflammation, and blood vessel blockage.

Synthesis

Four new studies on cerebrovascular disease or stroke were reviewed for the current update. Two studies followed Vietnam veterans. Cox et al. (2015) used hospital discharge records from 1988 to 2009 to identify the health conditions that had affected the veterans. Among New Zealand veterans, 170 cases

of cerebrovascular disease or stroke were observed, resulting in a significant increase in the standardized hospitalization rate for acute cerebrovascular disease compared with the general population. Han et al. (2016) published a hospital-based case-control study of 91 Korean veterans presumed to have been exposed to Agent Orange during their deployment to Vietnam who presented for medical attention within 7 days of the onset of an acute ischemic stroke. The key finding was a statistically different type of stroke between the presumed exposed and unexposed groups, with small vessel occlusion more common in the exposed subjects and large artery atherosclerosis more common in the unexposed subjects, resulting in a somewhat better short-term prognosis for the exposed versus the unexposed patients. While this study is small and requires replication as well as exposure validation, it raises an important issue about including all stroke subtypes in future studies of stroke and exposure to the COIs. Among a long-term follow-up of U.S. workers exposed to dioxins during TCP and PCP production, Collins et al. (2016) found no difference in the mortality rate from cerebrovascular disease between the workers and the standardized general population, but the study is limited by likely exposure misclassification and selection bias. In a prospective study of environmental exposure to organochlorine pollutants, HCB, and dioxin-like and non-dioxin-like PCBs in Belgium, Van Larebeke et al. (2015) assessed levels of exposure with a variety of self-reported health conditions, including atheromata. A small but statistically significant increase was found between the 90th versus the 10th percentile of exposure, but this was found for men only. The strengths of the study include its prospecitive design and large sample size, the representativeness of the older adult population of Flanders, and the use of objective measures of exposure. However, the findings are limited by the study’s lack of validated medical diagnoses and by the fact that the data were collected based on self-report from a survey that was not validated. No new data specifically addressing the biological plausibity of the impact of exposure to the COIs on cerebrovascular disease and stroke were identified. Therefore, on the basis of the new studies, the committee maintained the previous conclusion of limited or suggestive evidence between exposure to the COIs and cerebrovascular disease or stroke.

Conclusion

After carefully reviewing the previous evidence as well as the new evidence related to exposure to the COIs and cerebrovascular disease and stroke, the committee maintains the prior conclusion of limited or suggestive evidence of association that cerebrovascular disease and stroke are associated with exposure to the COIs.

Other Cardiovascular Outcomes

The previous sections in this chapter cover hypertension, IHD, and cerebrovascular disease and stroke. This section covers additional cardiovascular and blood disorders that have been examined in studies that have measured exposure to the COIs.

Vietnam-Veteran Studies

Cox et al. (2015) used hospital discharge records from 1988 to 2009 to identify health conditions in 2,783 male New Zealand veterans who served in Vietnam. Age-specific hospitalization rates were calculated using the total number of annual hospitalizations published by the Ministry of Health and the average annual resident population. Standardized hospitalization rates and 99% CIs were calculated for the veteran cohort and the general population and reported for cardiovascular and other outcomes. Syncope (n = 100; SHR = 1.47, 99% CI 1.18–1.76) and phlebitis (n = 51; SHR =1.67, 99% CI 1.07–2.28) were both elevated among the veterans compared with the standardized New Zealand population. No difference in the hospitalization rate for peripheral atherosclerosis (n = 70; SHR = 1.17, 99% CI 0.81–1.53) was found. Asymptomatic peripheral arterial vascular disease is more common than symptomatic disease and does not often result in death or hospitalization. This study was limited by its assumption that deployment to Vietnam was synonomous with herbicide exposure, which was not validated through serum measurements, and by the fact that important risk factors, such as smoking and ethnicity, were not controlled for in the analysis.

Occupational Studies

Laverda et al. (2015) assessed the association between weight gain (using BMI) and pesticide use using data on 8,365 male pesticide applicators from the AHS. The goal was to assess the effect of common factors beyond high-calorie diet and physical activity on BMI; pesticides were studied because many include endocrine disruptors and thus have a biological rationale for why they might affect BMI. Details of the AHS study design, exposure measurement, and data collection are presented in Chapter 5. Multivariate linear regression was used to estimate the unit change in BMI (kg/m²/d) associated with a unit change in exposure to a pesticide class/individual pesticide. To assess validity and the strength and direction of the associations, partial Spearman correlation coefficients controlling for all covariates were calculated for BMI at follow-up in relation to cumulative exposure days to each pesticide class/individual pesticide. Ordinal logistic regression was used to evaluate the exposure–response associations between BMI and pesticide exposures. Models were adjusted for BMI at age 20 years (from enrollment data), age, smoking, energy input, and energy output (from follow-up data). Results

were also adjusted for the state of residence and used multiple test corrections. In this analysis, 1,262 (15.1%) applicators reported exposure to phenoxy herbicides (2,4,5-T and 2,4,5-TP), and 1,214 (14.9%) applicators reported exposure to 2,4-D. The adjusted parameter estimate for unit change in BMI (kg/m²/d) was statistically significant for both phenoxy herbicides and 2,4-D in the entire sample (p = 0.02 for both), but when stratified by state, the two parameter estimates remained statistically significant for Iowan applicators only.

Environmental Studies

S. A. Kim et al. (2015) used serum concentrations of persistent organic pollutants, including dioxin-like and non-dioxin-like PCBs and organochlorine pesticides collected within the 1999–2004 NHANES and adjusted for fat mass in order to make associations with overall mortality, mortality from all cancers combined, and mortality from CVDs in people 70 years and older. Among the 633 people for whom there was information on PCBs, 50 had died of cardiovascular disease during the follow-up, while 56 had died among the 675 people for whom there was information on organochlorine pesticides. Models were adjusted for age, sex, race, BMI, cigarette smoking, and physical activity. When fat mass was excluded from the analysis of PCB exposure by tertiles, the adjusted hazard ratio for death from CVD in the second tertiles compared with the lowest exposure (tertile 1) was not different (HR = 1.06, 95% CI 0.47–2.40), but it was elevated for the third versus first tertile (HR = 2.37, 95% CI 1.10–5.09), and the trend for increasing exposure was statistically significant (p = 0.02). For organochlorine pesticides there was no difference in cardiovascular deaths between the second and third tertiles and the referent (p trend = 0.17). When individuals were stratified by body fat mass (< 25%, 25%–75%, and > 75%) and estimates were adjusted for the factors listed above, there were statistically significantly more deaths from CVD with exposure to PCBs only for the comparison of the third tertile with the first in the intermediate fat mass group (25– < 75%) (HR = 3.42, 95% CI 1.02–11.4), but this estimated is imprecise. This study was limited by the relatively low numbers of deaths, especially when stratified by exposure levels. This study measured death, not the incidence of disease, and it is possible that the POPs affect development of disease differently than they affect survival once the disease has been acquired.

Penell et al. (2014) investigated whether baseline circulating levels of POPs can predict future changes in circulating lipid levels. They measured lipid levels and the levels of 23 different POPs in 598 individuals in the PIVUS study at ages 70 and 75 years. No subjects were on lipid-lowering medications. POPs included non-dioxin-like PCBs, organochlorine pesticides, one dioxin (OCDD), and one brominated flame retardant compound. There were no significant difference related to the level of dioxin. The authors propose that the increase in low-density lipoproteins could strengthen the link between POP exposure and CVD.

Other Identified Studies

Zong et al. (2015) conducted a methodological study to assess the relationship between POPs and the fat mass percentage in the trunk, leg, and whole body, so the study does not have direct relevance to a health outcome. Their rationale was that since fat biopsies are used for assessing long-term lipophilic pollutant exposure, the anatomical location of the biopsy may affect the estimated body burden of these chemicals. The analysis included 2,358 men and women 20 years of age and older who had participated in NHANES 1999–2004 and who had stored serum samples. Serum measurements of 49 to 68 POPs were measured using gas chromatography/isotope dilution mass spectrometry. X-ray bone densitometry was used for dual-energy X-ray absorptiometry scans and fat mass percentage measurement. Demographics and information on lifestyle, medical history, and the history of lactation and parity for women were collected using questionnaires. Ethnicity, educational attainment, smoking and alcohol use, and physical activity were used as covariates. β-hexachlorocyclohexane, heptachlorodibenzo-p-dioxin, octachlorodibenzo-p-dioxin, and PCB 126 showed stronger positive correlations with trunk fat mass percentage than with leg fat mass percentage, suggesting a more important or greater role of trunk fat in the action and pharmacokinetics of persistent organic pollutants as endocrine disrupters. This study highlights one explanation for why studies may provide varying results on a metabolic or cardiovascular effect of COI exposure depending on which anatomical site was used for measurement.

Biologic Plausibility

Stea et al. (2016) suggest that the inconsistent associations of chronic low-dose exposure to arsenic on CVD may be due to high inter-individual variability in susceptibility to arsenic’s toxicity, a part of which is genetic. To assess this relationship, 214 healthy Italian volunteers between the ages of 20 and 46 years were studied for arsenic and related compounds in first voided urine, intima-media thickness, and genotypes for arsenic-metabolizing genes (ASIIIMT Met287Thr, GSTT1+/−, and GSTM1+/−). The samples were tested for inorganic arsenics (including trivalent and pentavalent), for methylated metabolites (i.e., mono-methylarsinic [MMA] and dimethylarsinic acid [DMA]), and for total arsenic using high-performance liquid chromatography and mass spectrometry. Intima-media thickness was measured on the common carotid artery on both sides using carotid ultrasound by a trained cardiologist. All genotypes were obtained by PCR and RFLP analysis. The main finding was that in both men and women with chronic low dose exposure to arsenic, intima-media thickness increases significantly faster with age than in the healthy population. The findings also suggest that GSTT1 and hOGG1 gene polymorphisms might play an important role in the

individual risk of arsenic-induced carotid atherosclerosis, with the same exposure providing differential outcomes.

Synthesis

In addition to outcomes related to hypertension, IHD, and cerebrovascular disease, several studies presented other cardiovascular and circulatory outcomes. However, most of these findings were of limited use because they examined different outcomes and different measures, and some of the outcomes were more methodologic or surrogates in the pathway of disease. For example, Penell et al. (2014) examined whether POPs can predict changes in lipid levels because higher blood lipid levels are a risk factor for CVD; however, no association was found for dioxin-like chemicals. S. A. Kim et al. (2015) conducted an analysis using NHANES data to look for an association between the concentrations of POPs (both dioxin-like and non-dioxin-like) and organochlorine pesticides and mortality from CVD. While there were statistically significant trends in death from cardiovascular disease among several subgroups classified by body fat, the estimates were imprecise. Stea et al. (2016) examined how exposure to arsenic may be a possible contributing factor to the individual risk of arsenic-induced carotid atherosclerosis, but a genetic component appears to play a part. The study of hospitalization risk among New Zealand Vietnam veterans used deployment to Vietnam as an indicator of presumed exposure to herbicides, and in analyses that did not control for smoking or other important risk factors of CVDs they found elevated rates of syncope and phlebitis but not peripheral atherosclerosis. Although there is similarity in the mechanisms by which atherosclerosis develops regardless of where it occurs (e.g., the heart, central nervous system, or periphery), asymptomatic peripheral arterial vascular disease is more common than symptomatic disease. The peripheral vascular disease literature is limited by high rates of asymptomatic disease, the requirement for specific and often costly testing to document clinically silent disease, death due to myocardial infarction and stroke before diagnosis, and infrequency of death or hospitalization from peripheral vascular disease alone.

In an assessment of weight gain (using BMI) and pesticide use from the AHS, Laverda et al. (2015) found that the parameter estimate for unit change in BMI (kg/m²/d) was statistically significant for both phenoxy herbicides and 2,4-D in the entire sample. However, based on the myriad studies and outcomes examined, no conclusion can be made concerning an association of exposure to the COIs and other cardiovascular disorders.

Conclusion

After carefully reviewing the previous evidence as well as the new evidence related to exposure to the COIs, the committee maintains the prior conclution that there is inadequate or insufficient evidence that blood disorders and other cardiovascular disorders are associated with exposure to the COIs.