Serum concentrations of Persistent Organic Pollutants and the Metabolic Syndrome in Akwesasne Mohawks, a Native American Community

Abstract

The metabolic syndrome (MetS) is a group of diseases that tend to occur together, including diabetes, hypertension, central obesity, cardiovascular disease and hyperlipidemia. Exposure to persistent organic pollutants (POPs) such as polychlorinated biphenyls (PCBs) and organochlorine pesticides (OCPs) has been associated with increased risk of development of several of the components of the MetS. The goal of this study is to determine whether the associations with POPs are identical for each of the components and for the MetS. The subject population was 601 Native Americans (Akwesasne Mohawks) ages 18 to 84 who answered a questionnaire, were measured for height and weight and provided blood samples for clinical chemistries (serum lipids and fasting glucose) and analysis of 101 PCB congeners and three OCPs [dichlorodiphenyldichloroethylene (DDE), hexachlorobenzene (HCB) and mirex] Associations between concentrations of total PCBs and pesticides, as well as various PCB congener groups with each of the different components of the MetS were determine so as to ask whether there were similar risk factors for all components of the MetS. After adjustment for other contaminants, diabetes and hypertension were strongly associated with lower chlorinated and mono-ortho PCBs, but not other PCB groups or pesticides. Obesity was most closely associated with highly chlorinated PCBs and was negatively associated with mirex. High serum lipids were most strongly associated with higher chlorinated PCBs and PCBs with multiple ortho-substituted chlorines, as well as total pesticides, DDE and HCB. Cardiovascular disease was not closely associated with levels of any of the measured POPs. While exposure to POPs is associated with increased risk of most of the various diseases comprising the MetS, the specific contaminants associated with risk of the component diseases are not the same.

Introduction

The prevalence of diseases based on disrupted metabolism has increased dramatically and continues to grow among adults in both developed and developing countries. Central obesity (CO), high level of serum LDL-cholesterol and triglycerides, diabetes with insulin resistance (IR), high blood pressure (HBP), and cardiovascular diseases (CVD) are metabolic disorders of great concern to the public health system because of their significant contribution to morbidity and mortality and the consequent increased cost of health care.

The metabolic syndrome (MetS) has been defined as a set of metabolic disorders. The National Cholesterol Education Program’s Adult Treatment Panel III recommends diagnosing patients with the MetS if at least three of above-mentioned disorders exist in any one individual (Expert Panel, 2001). The MetS is a multi-causal condition. The most common risk factors are low physical activity, high calorie diet, alcohol abuse and genetic predisposition. Recent epidemiological and experimental studies suggest exposure to certain persistent organic pollutants (POPs) such polychlorinated biphenyls (PCBs) and organochlorine pesticides (OCPs) are also associated with an increased risk of development of the MetS. Exposure to highly chlorinated PCBs, especially multiple ortho chlorine substituted congeners, and pesticides is associated with elevated serum lipid concentrations (Aminov et al., 2013; Goncharov et al., 2008). The presence of POPs in serum has also been shown to be positively correlated with IR in animal studies (Ruzzin et al., 2010, Mazzetti et al., 2004) and with diagnosed type II diabetes (T2D) in many epidemiologic studies (Aminov et al, 2016; Lee et al., 2006, 2007a, 2007b, 2011; Rignell-Hydbom et al., 2009; Michalek et al., 1999). Individuals living near to hazardous waste sites containing POPs are hospitalized with a diagnosis of hypertension more often than those not living near to such sites (Huang et al., 2006). Goncharov et al. (2010; 2011) reported that high serum concentrations of PCBs are associated with elevated systolic and diastolic blood pressure in a population highly exposed to PCBs. Only a few reports have been published on the association between exposure to POPs and CVD. Ha et al. (2007) reported higher risk of CVD among National Health and Nutrition Examination Survey (NHANES) 1999–2000 survey participant with higher serum concentrations of PCBs and dioxins. Sergeev and Carpenter (2005) reported elevated risks of coronary heart disease and myocardial infarction among individuals living near to POPs-contaminated waste sites, and later (Sergeev and Carpenter, 2011) reported elevated hospitalization with diagnosis of the MetS in individuals living near waste sites containing POPs.

In spite of the many studies conducted to study associations between exposure to POPs and metabolic disorders, there has been little attention on the questions of whether each of the components of the MetS shows a similar response to specific chemicals. One might expect that all of the components would show a similar sensitivity to xenobiotics. However to determine the relative contribution of individual POPs to development of the MetS is complicated and difficult to do because of the fact that most are lipophilic and therefore migrate together (Aminov et al., 2013).

Akwesasne is a territory of approximately 12,000 Native American Mohawks located on both sides of the St. Lawrence River. The territory has been polluted with PCBs, which were used as hydraulic fluids in three local aluminum foundries operated by Reynolds Metals, ALCOA, and General Motors. PCBs, like other POPs, are highly persistent, bioaccumulate and biomagnify through the food-chain, especially in fish. The Mohawks are traditionally a fish eating community and have become highly exposed to PCBs. Previous studies showed that Mohawks have a higher prevalence of diabetes (Aminov et al, 2016; Codru et al., 2007), higher risk of expressing elevated serum lipids and self-reported risk of CVD (Goncharov et al., 2008). They also have lower serum testosterone (Goncharov et al., 2009), and lower thyroid hormones (Schell et al., 2008, 2009) in relation to elevated serum PCBs and chlorinated pesticides. While the primary route of exposure until recently has been thought to be consumption of PCB-contaminated fish, the results of Aminov et al. (2016) suggest that inhalation of vapor-phase PCBs is an important route of exposure as well.

The objective of current study was to assess the relationship between exposure to different POPs and metabolic disorders and to determine whether all components of the MetS show similar associations with the various individual POPs. Since POPs are very different in their structure and properties, the next objective is to determine which group(s) of POPs is (are) responsible for the relationships. The last objective is to determine and discuss possible mechanisms which could explain the relationships.

Materials and Methods

Complete data on serum biochemical characteristics, exposure status, and medical history was available for 601 Mohawks with mean age of 43.9 (range 18 to 84). Details of how study participants were sampled and recruited, interviews conducted, blood samples were obtained, stored and analyzed for serum biochemical parameters and exposure status have been described in previous studies (Aminov et al., 2016; Santiago-Rivera et al., 2007; Goncharov et al., 2008; DeCaprio et al., 2000; 2005). The study was approved by the Institutional Review Board of the University at Albany and the Akwesasne Task Force on the Environment. The original data collection was supported by the National Institute of Environmental Health Sciences (ES04913) with subsequent support from the Institute for Health and the Environment.

Serum concentrations of 101 PCB congeners and three pesticides [hexachlorbenzene (HCB), dichloro-diphenyl-dichloro-ethylene (DDE), and mirex] were measured to determine exposure status (DeCaprio et al., 2000). Values below the method detection limit (MDL) for individual PCB congeners and OCPs were substituted with MDL/√2.

In order to conduct more comprehensive assessment of the relationships we were interested in, we have used two ways of grouping PCB congeners, by total number of chlorine atoms and by number of ortho-substituted chlorine atoms in the molecule. The individual congeners in each group are listed in Table 1. Since the group of mono-/dichloro biphenyl congeners were at very low concentrations, with a high prevalence of samples below the detection limits values, they are not presented in any of the tables. To test the importance of dioxin-like activity of PCBs, the group of non-/mono-ortho PCBs were further separated into two groups: congeners identified as having dioxin-like toxic equivalent factors (TEFs) [PCB77 (0.0001), PCB105 (0.00003), PCB114 (0.00003), PCB118 (0.00003), (PCB123+PCB149) (0.00003), PCB156 (0.00003)] (Van den Berg et al., 2006); and non-dioxin-like non-/mono-ortho PCBs (the remaining PCB congeners of the group).

Table 1:

PCB congener groups based on numbers of chlorines and numbers of ortho- chlorines. Individual congeners are identified by number given them by the International Union of Pure and Applied Chemistry (IUPAC):

Mono-/dichloro PCBs:	1, 2+4, 3, 6, 7, 8, 9, 13, 15.
Tri-tetrachloro PCBs:	19, 18, 17, 24+27, 32+16; 29, 26, 25, 31, 28, 33, 53, 51, 22, 45, 46, 52, 49, 47+59, 44, 42, 71, 64, 40, 67, 63, 74, 70, 66, 56, 77.
Penta-/hexachloro PCBs:	95, 91, 92, 84, 90+101, 99, 83, 97; 87, 136, 110, 151, 144, 109+147, 123+149, 118, 134, 114, 146, 153, 132, 105; 141, 137, 130, 164+163+138, 158, 129, 128, 156.
Hepta-/octa-/nona-/decachloro PCBs:	179, 176, 187, 183, 174, 177, 171, 201, 172, 180, 200, 170, 190, 199, 203, 196, 195, 194, 206.
Non-/mono-ortho PCBs;	3, 13, 15, 77, 1, 7, 9, 6, 8, 29, 26, 25, 31, 28, 33, 22, 67, 63, 74, 70, 66, 56, 118, 114, 105, 156.
Di-ortho PCBs:	4+2, 18, 17, 24+27, 32+16, 52, 49, 47+59, 44, 42, 71, 64, 40, 92, 90+101, 99, 83, 97, 87, 110, 146, 153, 141, 137, 130, 164+163+138, 158, 129, 128, 172, 180, 170, 190, 194.
Tri-/tetra-ortho PCBs:	19, 53, 51, 45, 46, 95, 91, 84, 151, 144, 147+109, 123+149, 134, 132, 187, 183, 185, 174, 177, 171, 199, 203, 196, 195, 206, 136, 179, 176, 201, 200.

Since certain MetS-related parameters have not been measured during the study period, some of them have been estimated. Only serum concentrations of cholesterol and triglycerides were measured; therefore serum total lipids concentrations were estimated using the following formula: Total lipids (mg/dL) = 2.27 Total cholesterol + Triglycerides + 0.623 (Phillips et al., 1989). The participants are considered to have high serum lipids if the concentrations of serum total lipids exceeded 500 mg/dL. Waist circumference was not measured, and therefore we used body mass index (BMI) to differentiate participants with central obesity (Nyamdorj et al., 2008). BMI was calculated using the formula weight (kg) over squared height (in meters). BMI was considered in six categories. Underweight was a BMI <18.5, normal weight was a BMI between 18.5 and 24.9, overweight was BMI 25.0 to 29.9, obesity I was BMI 30.0 to 34.9, obesity II was BMI 35.0 to 39.9 and obesity III was BMI 40.0 and greater. For most analyses the three obesity groups were combined.

Diabetes was defined by self-report of having been diagnosed with diabetes by a physician or by measured serum fasting glucose level of>125 mg/dL. HBP and CVD also have been defined subjectively from self-report by study participants.

Mean serum biochemical characteristics, concentrations of POPs, and body parameters were tested for differences in groups with and without metabolic disorders using the Student’s t-test with adjustment for age, since all those variables are highly correlated with age.

Multiple log-binomial and logistic regression were used to assess the relationships between exposure variables and each metabolic disorder. Since prevalence of CVD was 11.0% in our study population we used multiple logistic regression, while for HBP, diabetes, obesity, and elevated serum lipids we used multiple log-binomial regression, since they were more common (28.6%, 18.5%, 46.1%, and 71.4% respectively). When defined as having at least three metabolic disorders (Ervin, 2009), 24.8% of the study population were found to have the MetS; therefore its relationship with serum concentrations of POPs was assessed using multiple log-binomial regression.

Two types of models, unadjusted and adjusted, were constructed. Unadjusted models contained one of the exposure variables and covariates (age, gender, BMI, and serum total lipids). The models were further adjusted for serum total concentrations of PCBs (for pesticides) or pesticides (for PCBs) (adjusted models). The regression model for obesity did not include BMI as a covariate, and the regression model for the high serum lipids did not include serum concentrations of total lipids as a covariate. Exposure variables (serum concentrations of POPs) were categorized using tertiles and the age variable was categorized using quartiles to keep the variability within the categories.

Results

Table 2 shows the demographics of the study population and the serum concentrations of glucose, lipids and POPs. Age of the study population ranged from 18 to 94. Participants with metabolic disorders were older than those who did not have them. The least difference in age observed was 8.4 years (SE=1.2) for participants who had elevated serum lipids compared to those who didn’t have elevated serum lipids. The study participants who had any CVD were 15.3 years older (SE=1.8) than those who did not. There was no age difference by obesity status. The mean value for BMI for the total study population was 30.3 kg/m², which can be classified as the first level obese. Age adjusted BMI was not significantly different when groups with CVD and elevated serum lipids were compared, but study participants with HBP, diabetes, and those with at least three metabolic disorders had significantly older age-adjusted BMI than those who did not. Age-adjusted serum concentrations of glucose were higher in participants with any metabolic disorder in comparison to participants without any disorder other than elevated serum total lipids.

Table 2.

Demographic and serum biochemical characteristics of the study population.

	Mean for study population N=601	High blood pressure N=172		Diabetes N=111		Obesity (BMI≥30 kg/m2) N=277
	(min-max)	β±SE	PR>|t|	β±SE	PR>|t|	β±SE	PR>|t|
Age (years)	43.9 (18.0–84.0)	10.1±1.2	<0.0001	9.5±1.5	<0.0001	0.0±1.17	0.9672
BMI (kg/m2)	30.3 (14.6–59.8)	2.9±0.6	<0.0001	3.8±0.7	<0.0001	9.5±0.4	<0.0001
Serum glucose (mg/dL)	107.2 (54.0- 480.0)	25.5±4.2	<0.0001	84.6±3.5	<0.0001	19.6±3.6	<0.0001
Serum cholesterol (mg/dL)	194.7 (101.0- 410.0)	4.6±3.5	0.1914	−3.8±4.0	0.3498	1.7±3.0	0.5682
Serum triglycerides (mg/dL)	152.1 (30.0- 746.0)	24.7±8.7	0.0045	43.9±9.8	<0.0001	23.2±7.4	0.0019
Serum total lipids (mg/dL)	594.7 (270.0- 1378.3)	35.2±13.9	0.0115	35.3±15.9	0.0264	27.1±11.9	0.0235
Total PCBs (ppb wet weight)	5.26 (1.56- 49.09)	0.54±0.30	0.0761	1.14±0.35	0.0010	0.41±0.26	0.1156
Mono-/dichloro	0.27 (0.24–1.36)	0.00±0.00	0.6854	0.01±0.01	0.5062	−0.01±0.01	0.1964
Tri-/tetrachloro	0.99 (0.50–6.58)	0.12±0.05	0.0104	0.20±0.05	0.0001	0.11±0.04	0.0066
Penta-/hexachloro	2.62 (0.54- 26.62)	0.28±0.17	0.0907	0.62±0.19	0.0011	0.31±0.14	0.0308
Hepta-/octa- /nona-/decachloro	1.49 (0.22- 15.63)	0.13±0.11	0.2409	0.31±0.13	0.0175	0.00±0.10	0.9971
Non-/monoortho	1.34 (0.48- 14.71)	0.18±0.09	0.0440	0.37±0.10	0.0003	0.22±0.08	0.0049
Dioxin-like	0.58 (0.10–8.59)	0.09±0.05	0.0709	0.19±0.06	0.0012	0.14±0.04	0.0021
Non-dioxin-like non-/monoortho	0.80 (0.40–6.34)	0.09±0.04	0.0293	0.19±0.05	0.0001	0.09±0.04	0.0156
Diortho	2.92 (0.58- 25.26)	0.26±0.17	0.1306	0.59±0.19	0.0026	0.16±0.15	0.2849
Tri-/tetraortho	1.06 (0.40–9.07)	0.10±0.06	0.1141	0.18±0.07	0.0128	0.04±0.05	0.4920
Total pesticides (wet weight)	3.37 (0.20- 22.79)	0.57±0.25	0.0245	1.03±0.29	0.0004	0.28±0.22	0.2017
HCB	0.08 (0.01–0.33)	0.13±0.00	0.0001	0.02±0.00	<0.0001	0.01±0.00	0.0001
DDE	3.21 (0.04- 22.51)	0.57±0.25	0.0232	1.00±0.28	0.0005	0.27±0.21	0.2045
Mirex	0.12 (0.01–1.67)	−0.01±0.01	0.4373	0.01±0.01	0.3528	−0.01±0.01	0.6237
Gender (women/men)	377/224	96/76		70/41		167/110

Serum concentrations of total cholesterol, triglycerides, and total lipids did not differ in those who had or did not have CVD after adjustment for age. Mean serum concentration of total cholesterol was 10.1 mg/dL higher only in the group with at least three metabolic disorders (SE=3.9). Higher age-adjusted serum concentrations of triglycerides and total lipids were significantly associated with the presence of all remaining metabolic disorders.

There were no significant differences in serum concentrations of total PCBs or OCPs between obese and non-obese study participants. Participants with high blood pressure had non-significantly higher levels of total PCBs (p=0.0761), but significantly higher concentrations of pesticides (p=0.0245). The group of study participants with CVD, diabetes, elevated serum lipids, and with at least three metabolic disorders had significantly higher concentrations of both total PCBs and pesticides after adjustment for age.

The results of the log-binomial regression model analyses of the associations between the serum concentrations of POPs and risk of metabolic diseases are presented in tables 3 and 4. Both contain the results obtained from assessment of relationships between exposure variables and each metabolic disorder separately. Table 5 contains the results of the logistic regression model analyses of the associations between the serum concentrations of the POPs and risk of the MetS and CVD.

Table 3.

Results of log-binomial regression analyses of the relationships between serum concentrations of different groups of PCBs and metabolic disorders. Statistically significant values are shown in bold. Adjusted results indicates adjustment for total pesticides.

		High blood pressure		Diabetes		Obesity		High serum lipids
		II tertile β (p-value)	III tertile β (p-value)	II tertile β (p-value)	III tertile β (p-value)	II tertile β (p-value)	III tertile β (p-value)	II tertile β (p-value)	III tertile β (p-value)
Total PCBs	Unadjusted	0.34 (0.1623)	0.54 (0.0316)	0.15 (0.6676)	0.74 (0.0322)	0.05 (0.6751)	0.03 (0.8528)	0.20 (0.0103)	0.34 (<0.0001)
	Adjusted	0.27 (0.2882)	0.39 (0.1480)	−0.17 (0.6239)	0.29 (0.4217)	0.03 (0.8046)	−0.08 (0.6447)	0.11 (0.2070)	0.19 (0.0447)
Tri- and tetrachloro PCBs	Unadjusted	0.58 (0.0185)	0.79 (0.0013)	1.45 (0.0442)	1.77 (0.0136)	−0.01 (0.9658)	0.13 (0.3039)	0.13 (0.0833)	0.16 (0.0340)
	Adjusted	0.54 (0.0304)	0.72 (0.0043)	1.39 (0.0613)	1.63 (0.0282)	−0.02 (0.8829)	0.09 (0.4987)	0.07 (0.3623)	0.06 (0.4424)
Penta- and hexachloro PCBs	Unadjusted	0.18 (0.4337)	0.44 (0.0686)	0.09 (0.7981)	0.78 (0.0258)	−0.04 (0.7481)	0.01 (0.9210)	0.28 (0.0005)	0.39 (<0.0001)
	Adjusted	0.07 (0.7609)	0.23 (0.3803)	−0.46 (0.3526)	0.42 (0.3938)	−0.07 (0.6215)	−0.08 (0.6398)	0.20 (0.0201)	0.26 (0.0098)
Hepta-, octa-, nona-, and decachloro PCBs	Unadjusted	0.41 (0.1047)	0.60 (0.0287)	0.0.42 (0.1960)	0.84 (0.0133)	−0.09 (0.4386)	−0.36 (0.0286)	0.31 (<0.0001)	0.42 (<0.0001)
	Adjusted	0.33 (0.2123)	0.46 (0.1122)	0.12 (0.7111)	0.47 (0.1898)	−0.16 (0.2428)	0.51 (0.0044)	0.24 (0.0056)	0.30 (0.0021)
Non- and monoortho PCBs	Unadjusted	0.28 (0.2533)	0.67 (0.0073)	0.65 (0.1658)	1.25 (0.0065)	0.09 (0.4629)	0.29 (0.0362)	0.22 (0.0034)	0.29 (0.0005)
	Adjusted	0.26 (0.3063)	0.62 (0.0229)	0.58 (0.2328)	1.06 (0.0328)	0.09 (0.4828)	0.26 (0.0847)	0.14 (0.0741)	0.16 (0.0724)
Dioxin-like PCBs	Unadjusted	0.28 (0.2617)	0.54 (0.0358)	0.15 (0.6668)	0.74 (0.0305)	0.12 (0.3421)	0.24 (0.0960)	0.16 (0.0359)	0.32 (0.0001)
	Adjusted	0.23 (0.3639)	0.42 (0.1338)	−0.20 (0.5614)	0.23 (0.4877)	0.13 (0.3193)	0.21 (0.1790)	0.08 (0.3245)	0.20 (0.0317)
Non-dioxin-like non- and monoortho PCBs	Unadjusted	0.42 (0.0743)	0.70 (0.0033)	1.24 (0.0595)	1.77 (0.0062)	0.15 (0.2100)	0.18 (0.1818)	0.16 (0.0272)	0.19 (0.0177)
	Adjusted	0.36 (0.1338)	0.60 (0.0171)	1.10 (0.0820)	1.57 (0.0131)	0.14 (0.2567)	0.14 (0.3175)	0.09 (0.2452)	0.07 (0.4257)
Di-ortho PCBs	Unadjusted	0.26 (0.2642)	0.42 (0.0898)	0.10 (0.7501)	0.60 (0.0653)	0.03 (0.7805)	−0.06 (0.6874)	0.30 (0.0002)	0.42 (<0.0001)
	Adjusted	0.16 (0.5219)	0.23 (0.3946)	−0.27 (0.4262)	0.10 (0.7674)	0.00 (0.9876)	−0.18 (0.2973)	0.23 (0.0075)	0.30 (0.0016)
Tri- and tetraortho PCBs	Unadjusted	0.69 (0.132)	0.77 (0.0087)	0.34 (0.3250)	0.81 (0.0181)	0.01 (0.9348)	−0.15 (0.3423)	0.13 (0.0979)	0.33 (<0.0001)
	Adjusted	0.63 (0.0285)	0.64 (0.0366)	0.09 (0.7926)	0.46 (0.1908)	−0.03 (0.8387)	−0.25 (0.1356)	0.06 (0.4491)	0.22 (0.0149)

The models for the high blood pressure and diabetes the models were adjusted for age, gender, BMI, and serum total lipids.

The models for the central obesity were adjusted for age, gender, and serum total lipids.

The models for the high serum lipids were adjusted for age, gender, and BMI.

Table 4.

Results of multiple regression analyses of the relationships between serum concentrations of OCPs and metabolic disorders. Statistically significant values are shown in bold. “Adjusted” indicates adjustment for total PCBs.

		High blood pressure		Diabetes		Obesity		High serum lipids
		II tertile β (p-value)	III tertile β (p-value)	II tertile β (p-value)	III tertile β (p-value)	II tertile β (p-value)	III tertile β (p-value)	II tertile β (p-value)	III tertile β (p-value)
Total pesticides	Unadjusted	0.43 (0.1435)	0.65 (0.0290)	0.71 (0.1343)	1.21 (0.0110)	0.01 (0.9356)	0.18 (0.2554)	0.30 (0.0006)	0.44 (<0.0001)
	Adjusted	0.36 (0.2167)	0.51 (0.0980)	0.57 (0.2105)	0.95 (0.0528)	0.01 (0.9681)	0.22 (0.2165)	0.24 (0.0123)	0.33 (0.0040)
HCB	Unadjusted	0.21 (0.2619)	0.32 (0.1180)	0.06 (0.8572)	0.81 (0.0033)	0.12 (0.3362)	0.27 (0.0532)	0.39 (<0.0001)	0.47 (<0.0001)
	Adjusted	0.12 (0.5260)	0.15 (0.5032)	−0.10 (0.7599)	0.56 (0.0629)	0.13 (0.3016)	0.37 (0.0227)	0.37 (<0.0001)	0.42 (<0.0001)
DDE	Unadjusted	0.44 (0.1285)	0.65 (0.0264)	0.63 (0.1844)	1.26 (0.0069)	0.02 (0.8876)	0.22 (0.1583)	0.29 (0.0011)	0.41 (<0.0001)
	Adjusted	0.38 (0.1878)	0.52 (0.0829)	0.54 (0.2433)	1.08 (0.0277)	0.01 (0.9208)	0.26 (0.1387)	0.23 (0.0174)	0.30 (0.0066)
Mirex	Unadjusted	0.30 (0.0496)	0.19 (0.2370)	0.45 (0.1807)	0.30 (0.1032)	−0.20 (0.0636)	−0.22 (0.0755)	0.12 (0.564)	0.10 (0.1348)
	Adjusted	0.19 (0.2142)	−0.00 (0.9882)	0.20 (0.2962)	−0.12 (0.5559)	−0.25 (0.0317)	−0.29 (0.0403)	0.04 (0.5810)	−0.07 (0.4110)

The models for the high blood pressure and diabetes the models were adjusted for age, gender, BMI, and serum total lipids.

The models for the central obesity were adjusted for age, gender, and serum total lipids.

The models for the high serum lipids were adjusted for age, gender, and BMI.

Table 5.

Results of logistic regression analyses of the relationships between serum concentrations of persistent organic pollutants and MetS (presence of at least three metabolic disorders). Statistically significant values are shown in bold. Adjusted values indicate adjustment for total pesticides for PCB values and adjustment for total PCBs for pesticide values.

		Metabolic syndrome		Cardio-vascular diseases
		II tertile β (p-value)	III tertile β (p-value)	II tertile β (p-value)	III tertile β (p-value)
TOTAL PCBs	Unadjusted	0.79 (0.0148)	1.14 (0.0008)	0.38 (0.5360)	0.65 (0.3067)
	Adjusted	0.52 (0.1440)	0.66 (0.0822)	0.20 (0.7520)	0.46 (0.5029)
Tri- and tetra-chloro PCBs	Unadjusted	1.01 (0.0018)	1.38 (<0.0001)	−0.03 (0.9467)	0.13 (0.7845)
	Adjusted	0.85 (0.0100)	1.12 (0.0008)	−0.14 (0.7727)	−0.01 (0.9835)
Penta- and hexa-chloro PCBs	Unadjusted	0.57 (0.00805)	1.05 (0.0015)	0.77 (0.2621)	1.15 (0.1012)
	Adjusted	0.25 (0.4768)	0.53 (0.1527)	0.64 (0.3697)	1.03 (0.1739)
Hepta-, octa-, nona-, and deca-chloro PCBs	Unadjusted	0.67 (0.0358)	0.73 (0.0329)	0.05 (0.9339)	1.00 (0.1442)
	Adjusted	0.33 (0.3393)	0.19 (0.6142)	−0.14 (0.8371)	0.84 (0.2480)
Non- and mono-ortho PCBs	Unadjusted	0.85 (0.0101)	1.44 (<0.0001)	−0.20 (0.6944)	0.16 (0.7577)
	Adjusted	0.67 (0.0510)	1.12 (0.0018)	−0.37 (0.4782)	−0.02 (0.9662)
Dioxin-like PCB TEFs	Unadjusted	0.55 (0.0755)	1.08 (0.0007)	−0.29 (0.5728)	0.16 (0.7545)
	Adjusted	0.31 (0.3393)	0.65 (0.0561)	−0.47 (0.3781)	−0.03 (0.9618)
Non-dioxin-like non-and mono-ortho PCBs	Unadjusted	0.99 (0.0026)	1.49 (<0.0001)	−0.39 (0.4081)	−0.02 (0.9601)
	Adjusted	0.83 (0.0155)	1.20 (0.0006)	−0.56 (0.2534)	−0.23 (0.6522)
Di-ortho PCBs	Unadjusted	0.73 (0.0229)	0.98 (0.0031)	0.79 (0.2461)	1.04 (0.1410)
	Adjusted	0.43 (0.2195)	0.47 (0.1999)	0.64 (0.3631)	0.89 (0.2377)
Tri- and tetra-ortho PCBs	Unadjusted	1.03 (0.0022)	1.23 (0.0005)	0.70 (0.3044)	1.24 (0.0797)
	Adjusted	0.81 (0.0238)	0.82 (0.0309)	0.58 (0.4059)	1.14 (0.1231)
TOTAL PESTICIDES	Unadjusted	0.87 (0.0163)	1.47 (0.0001)	0.76 (0.2737)	0.83 (0.2466)
	Adjusted	0.63 (0.1031)	1.14 (0.0063)	0.66 (0.3574)	0.63 (0.4145)
HCB	Unadjusted	0.32 (0.2268)	0.85 (0.0010)	0.16 (0.7330)	0.54 (0.2647)
	Adjusted	0.14 (0.5962)	0.59 (0.0411)	0.08 (0.8729)	0.38 (0.4804)
DDE	Unadjusted	0.84 (0.0210)	1.50 (<0.0001)	0.37 (0.5567)	0.53 (0.4062)
	Adjusted	0.60 (0.1177)	1.19 (0.0039)	0.25 (0.7034)	0.32 (0.6388)
Mirex	Unadjusted	0.29 (0.1729)	0.20(0.3623)	0.81 (0.0611)	0.74 (0.0923)
	Adjusted	−0.01 (0.9640)	−0.26 (0.2868)	0.76 (0.1011)	0.63 (0.226)

All models were adjusted for age, gender, BMI, and serum total lipids.

Serum concentrations of total PCBs showed significant association with HBP (β=0.54, p=0.0316), diabetes (β=0.74, p=0.0322) in the third tertile and elevated serum lipids (2^nd tertile β=0.20, p=0.0103; 3^rd tertile β=0.34, p<0.0001) in unadjusted models (Table 2). When models were additionally adjusted for total pesticides, total PCBs in the third tertile was elevated in relation to serum lipids (β=0.19, p=0.0447).

The group of tri-chloro and tetra-chloro PCBs showed significant association with risk of HBP and diabetes in both adjusted and unadjusted models. There was no association with obesity and only in the unadjusted model were concentrations of tri-chloro and tetra-chloro PCBs associated with elevated serum lipids in the third tertile (β=0.16, p=0.0340). The group of penta- and hexa-chloro PCBs were significantly associated with risk of diabetes in the unadjusted model (3^rd tertile β=0.78, p=0.0258), and with elevated serum lipids in both models, but the relationship was stronger in the unadjusted model (2^nd tertile β=0.28, p=0.0005; 3^rd tertile β=0.39, p<0.0001). The group of highly chlorinated PCBs (hepta-/octa-/nona-/decachloro PCBs) showed significant relationship with risk of HBP and diabetes only in the third tertile when models were not adjusted for total pesticides (β=0.60, p=0.0287 for HBP; β=0.84, p=0.0133 for diabetes), while with high serum lipids both models showed significant association with higher estimates for the unadjusted model (2^nd tertile β=0.31, p<0.0001; 3^rd tertile β=0.42, p<0.0001). With obesity highly chlorinated PCBs showed a reverse association in the unadjusted model (3^rd tertile β=−0.36, p=0.0286), but the association was positive when the model was adjusted for total pesticides (3^rd tertile β=0.51, p=0.0044).

Analysis with groups of PCBs by numbers of ortho-substituted chlorines showed that when models were not adjusted for total pesticides the group of non-/mono-ortho PCBs had a significant relationship with risk of HBP (3^rd tertile β=0.67, p=0.0073), diabetes (3^rd tertile β=1.25, p=0.0065), obesity (3^rd tertile β=0.29, p=0.0362), and high serum lipids (2^nd tertile β=0.22, p=0.0034; β=0.29, p=0.0005). When models were additionally adjusted for total pesticides, non-/mono-ortho PCBs were only associated with elevated risk of HBP (3^rd tertile β=0.62, p=0.0229) and diabetes (3^rd tertile β=1.06, p=0.0328).

This group of PCBs contains some congeners that have assigned dioxin-like TEFs and some that do not, and these two groups differed in their associations. After adjustment for total pesticides the dioxin-like PCB TEFs showed significant association only with high serum lipids (3^rd tertile β=0.20, p=0.0317), while the non-dioxin-like non- and mono-ortho group showed significant associations with HBP and diabetes. The group of di-ortho PCBs showed significant association only with high serum lipids in both the unadjusted (2^nd tertile β=0.30, p=0.0002, 3^rd tertile β=0.42, p<0.0001) and adjusted models (2^nd tertile β=0.23, p=0.0075; 3^rd tertile β=0.30, p=0.0016).

The group of tri-/tetra-ortho PCBs showed significant relationships with elevated risk of HBP (unadjusted β=0.77, p=0.0087; adjusted β=0.64, p=0.0366), diabetes (unadjusted β=0.81, p=0.0181) and high serum lipids (unadjusted β=0.33, p<0.0001; adjusted β=0.22, p=0.0149).

Results of multiple regression analysis with serum concentrations of OCPs are shown in table 4. Serum total pesticides when models were not adjusted for total PCBs had significant association with higher risk of HBP (3^rd tertile β=0.65, p=0.0290), diabetes (3^rd tertile β=1.21, p=0.0110), and high serum lipids (2^nd tertile β=0.30, p=0.0006; 3^rd tertile β=0.44, p<0.0001). However when the models were adjusted for total PCBs, total pesticides had significant association with only high serum lipids (2^nd tertile β=0.24, p=0.0123; 3^rd tertile β=0.33, p=0.0040). Serum concentrations of HCB showed significant association with diabetes (3^rd tertile β=0.81, p=0.0033) when the model was not adjusted for total PCBs, but the relationship was weaker and non-significant after adjustment for total PCBs. The association between serum concentrations of HCB and risk of obesity was stronger when the model was adjusted for total PCBs (3^rd tertile β=0.37, p=0.0227). There was a robust and significant relationship between HCB with risk of high serum lipids with almost the same estimates and levels of significance in both unadjusted and adjusted models. Serum DDE concentration had significant association with high serum lipids in the unadjusted model and a bit weaker association after being adjusted for total PCBs (2^nd tertile β=0.23, p=0.0174; 3^rd tertile β=0.30, p=0.0066). DDE showed significant association with elevated risk of HBP and diabetes only in the highest tertile (β=0.65, p=0.0264 for HBP; β=1.26, p=0.0069 for diabetes) in the unadjusted model, and only with diabetes when adjusted for total PCBs (β=1.08, p=0.0277).

Mirex showed a different pattern from the other pesticides. It was significantly associated with risk of HBP in the second tertile but not the third, and this disappeared after adjustment for PCBs. No significant association was observed with high serum lipids, but a significant and reverse association was observed with obesity in the model after adjustment for total PCBs (2^nd tertile β=−0.25, p=0.0317; 3^rd tertile β=−0.29, p=0.0403).

Table 5 presents data on the MetS as defined as having at least three metabolic disorders, regressed with all PCB groups and chlorinated pesticides. After adjustment for total pesticides, the significant associations with the MetS were limited to tri- and tetra-chloro congeners, non-dioxin-like non- and mono-ortho congeners and tri- and tetra-ortho congeners in both second and third tertile, and total non-and mono-ortho PCBs in only the third tertile. There were no significant associations with more highly chlorinated PCBs nor with dioxin-like TEFs after adjustment for total pesticides.

Serum total pesticides had a significant association with the MetS when the model was adjusted for total PCBs (3^rd tertile β=1.14, p=0.0063). Both HCB and DDE showed significant associations in the third tertile after adjustment for total PCBs, but there were no significant associations with mirex in either the unadjusted or adjusted models.

None of the PCB groups or pesticides showed significantly associated with risk of CVD whether or not models were adjusted for other POPs.

Discussion

More than 25% of US population over 20 years of age have metabolic disorders related to lipid and glucose metabolism (Ervin 2009; Ford et al., 2004) and the prevalence is increasing dramatically year by year. Therefore, many scientists are focused on study of the MetS. Our team has been exploring environmental risk factors for the MetS, in particular the chlorinated POPs. In our previous publications we have reported associations between serum concentrations of POPs and high serum lipids (Aminov et al., 2013; Aminov et al., 2014) among residents of Anniston, AL, and elevated risk of diabetes among Akwesasne Mohawks (Codru et al., 2007; Aminov et al., 2016).

In the current study, we tried to assess the risk of the various metabolic disorders that constitute the MetS in relation to exposure to different groups of POPs. We have observed significant associations between metabolic disorders and several groups of POPs, but find clearly that not all POPs have similar associations. By far the strongest associations with the MetS were found for the lower chlorinated, non-dioxin like PCBs (tri- and tetra-chloro and/or non- and mono-ortho). These two PCB group contain many of the same congeners. The observation that lower chlorinated PCBs show the strongest associations in a new observation, since most previous studies have focused on highly chlorinated, more persistent congeners and especially congeners with dioxin-like activity.

Obesity and elevated serum lipids were much more strongly associated with more highly chlorinated PCBs, HCB and DDE. However after adjustment the more highly chlorinated congeners did not contribute significantly to the MetS.

The observation that lower chlorinated congeners are important has significant implications for routes of exposure, since lower chlorinated PCBs are both more water soluble and more volatile (Carpenter, 2015). Thus exposure can occur via inhalation and by drinking water to a much greater degree than is the case for higher chlorinated PCBs, for which exposure is primarily by food. There is extensive PCB contamination of soil at Akwesasne as well as sources of volatile PCBs from landfills and contaminated rivers. Our results suggest that inhalation is an important route of exposure to the PCB congeners most associated with the MetS in this population. In this study we only monitored PCBs and three pesticides, but as for the general population we are all exposed to many different chemicals and chemical mixtures. While it is possible, even if unlikely, that some other chemical(s) are responsible for the associations with the MetS we observe, our evidence indicates that the most important route of exposure is inhalation.

Previous studies of diabetes in relation to dioxin-like activity and degree of chlorination of PCBs have provided somewhat conflicting results. Veterans of Operation Ranch Hand in Vietnam, who handled Agent Orange and were thus exposed to dioxin, were reported to have elevated rates of diabetes (Michalek et al., 1999). There are some published reports on relationships between exposure to POPs and elevated risk of HBP. Everett et al., (2008), using information from the National Health and Nutrition Examination Survey (NHANES) conducted during 1999–2002, reported that PCBs with fewer ortho-substituted chlorines and dioxin-like activity were more strongly associated with risk of HBP. However, Goncharov et al. (2011) found a stronger relationship of elevated systolic and diastolic blood pressure with multiple ortho-substituted PCBs. A study from Denmark reports no significant association between exposure to POPs and HBP (Valera et al., 2013). We did not have measurements of serum concentrations of dioxin, but serum concentration of dioxin-like PCBs have shown some relationship with HBP, diabetes, and high serum lipids in models not adjusted for other POPs. However with the exception of association with high serum lipids all other associations disappeared after adjustment of levels of pesticides.

Our results support the conclusion that it is primarily the non-dioxin like congeners that are important for the associations with the MetS. Because all of these lipophilic POPs migrate together within serum lipids, there are difficulties in separating actions of different congener groups (Aminov et al., 2013. This is why adjustment for other POPs is important, and may be the reason for the different interpretation of results in previous studies.

We did not find any association between CVD and serum concentrations of any POPs. Previously multiple studies have shown association between exposure to POPs and CVD. Sergeev and Carpenter (2005) and Sergeev et al., (2010) reported that New York State zip codes with close proximity to hazardous waste sites containing POPs had higher hospitalization rates with cardiovascular diseases, especially with acute myocardial infarction. The relationship between POPs and cardiovascular diseases was studied by Goncharov et al. (2008), who concluded that any elevation in risk of CVD was indirect due to the elevated serum lipids. In addition, an experimental study on mice showed that there are inflammatory pathways with atherosclerotic consequences of exposure to POPs (Wu et al., 2011). Some other epidemiologic studies presented similar relationships (Ha et al., 2007; Lind et al., 2012). Our models were adjusted for age, gender, BMI, and serum total lipids. Since, age and serum lipids are very highly correlated with serum concentrations of POPs and risk of CVD (Aminov et al., 2013; Goncharov et al., 2008), there is a possibility that our models are over-controlled for potential covariates. Clearly the positive associations we have observed between concentrations of some POPs and serum lipids, rates of diabetes, hypertension and obesity, all known risk factors for CVD, is evidence that there is at least an indirect relationship between exposure to POPs and risk of CVD.

There are notable strengths but also significant weaknesses in our study. The population studied is racially homogeneous. This is a major strength as it controls for the confounding that results from study of genetically diverse populations, although there also could be genetic susceptibility factors in this population not expressed in non-Native persons. Our ability to measure 101 PCB congeners is more than are measured by most PCB laboratories, and allows us to evaluate the relative contribution of different congeners in relation to numbers and locations of chlorines around the biphenyl ring. However there are 209 PCB congeners, and our analytical methods do not capture some of the most potent dioxin-like congeners. While we had direct measurements of fasting glucose, serum cholesterol and triglyceride levels, and BMI, we relied only on self-report of HTN and CVD. Because all of the chemicals under investigation are lipophilic, they migrate together and this colinearity poses significant difficulties when one tries to identify which specific chemical is responsible for a specific health effects. We have addressed this concern is great detail in a previous publication (Aminov et al., 2013, additional file 1). But use of this adjustment modeling has provided important insight into which congeners are most associated with the various components of the MetS and which are the most important routes of exposure.

Casual temporality between serum concentrations of POPs and metabolic disorders has been a subject of debate, especially regarding elevated serum lipids. In our previous publications we have discussed this issue (Aminov et al., 2013; 2014; Goncharov et al., 2008) and provided evidence in support of the conclusion that exposure to POPs is casually associated with elevated serum lipids and other metabolic disorders. Experimental animal studies also show that exposure to POPs can lead to insulin resistance and impaired lipids metabolism (Ruzzin et al., 2010; Mazzetti et al., 2004). In addition, Lind et al., (2004) showed that exposure to dioxin-like PCB126 can lead to elevated serum cholesterol and blood pressure. Another study reported increased risk of cardiovascular diseases in rats chronically exposed to PCB126, because of degenerative cardiovascular lesions, cardiomyopathy and chronic active arteritis (Jokinen et al., 2003). Our findings are in general consistent with the results of these experimental studies. We did find a strong relationship between exposure to PCBs and OCPs with high serum lipids and diabetes. We did find some relationship with dioxin-like PCBs and metabolic disorders as well, but the relationships were weaker for dioxin-like PCBs than for remaining non- and mono-ortho PCBs, and tri- and tetra-chloro PCBs. Dioxin and dioxin-like compounds are considered very toxic, and the majority of the experimental studies have been focused on them. However, based on our findings, more comprehensive experimental studies focused on lower chlorinated PCBs, non-dioxin-like PCBs need to be done.

We found that DDE and HCB, but not mirex, were significantly associated with elevated risk of the MetS, although not as strongly as were the lower chlorinated, non-dioxin-like PCBs. The Mohawk population does not have unusual exposure to these pesticides, and thus these results suggest that background exposure to these pesticides, which occurs primarily from consumption of animal fats, may elevate risk of the MetS in the general population. While there is some evidence for associations between levels of OCPs and diabetes (Lee and Jacobs, 2013) and obesity (Dirinck et al., 2013) there has to date been little study of pesticides and the MetS. One particular concern is evidence that prenatal exposure to DDE may increase risk of obesity later in life (Karmaus et al., 2009; Mendez et al., 2010).

There are many lipophilic POPs in addition to PCBs and the three chlorinated pesticides we have monitored. Everyone is exposed to a chemical mixture of persistent, lipophilic chemicals. We cannot exclude the possibility that some of the associations we have observed are actually due to lipophilic chemicals we have not measured.

Conclusions:

Our observation that lower chlorinated, non-dioxin-like PCBs are more strongly associated with risk of the MetS has important implications. Lower chlorinated PCBs are more volatile, therefore people are exposed to them by inhalation (Carpenter 2015). They are more water soluble, and can be present in drinking water. They are more rapidly metabolized in the human body, and thus the level of exposure is less easy to quantitate than is the case for highly chlorinated PCBs. However if a major route of exposure is continuous inhalation of vapor phase PCBs coming from a local source, the degree of health hazard may be much greater than that reflected by serum PCB concentrations. For all of these reasons, the relationship between exposure to lower chlorinated PCBs and metabolic disorders may have been underestimated in previous studies.

While there is no question but that life style factors such as diet and lack of exercise contribute to development of the MetS, these results provide evidence that exposure to PCBs and OCPs also increase risk of development of the MetS. Further studies are needed to determine the relative roles of the various risk factors.

Highlights:

PCBs trigger the metabolic syndrome in a Native American population. However the PCB congeners most closely associated with diabetes and hypertension are those with fewer chlorines, for which the likely route of exposure is inhalation of vapor phase PCBs, whereas the PCB congeners most closely associated with hyperlipidemia are those with multiple chlorines, where the most likely route of exposure is consumption of animal fats.