There has been suggestive evidence for many years that elevations in levels of some persistent organic pollutants (POPs) are associated with elevated levels of serum lipids. Since elevated serum lipids are a major risk factor for cardiovascular disease [1], this association, if causal, may have significant effects on human health. In 1978 two laboratories reported that feeding rats PCBs and/or DDT resulted in elevation of serum cholesterol [2,3]. Oda et al. [4] reported that feeding rats PCBs resulted in hypercholesterolemia, and suggested that PCB exposure resulted in a stimulation of de novo synthesis of liver lipids. Azaïs-Braesco et al. [5] studied the effects of two different PCB congeners and found that the changes in lipid profiles varied with congener structure. Monkeys fed Aroclor 1254 developed elevated triglycerides, but decreased total cholesterol, HDL and LDL cholesterol [6]. Sanyal et al. [7] reported elevations in lipid synthesis by the liver in DDT-treated monkeys.

In human populations exposed to PCBs, several authors have reported elevations in triglycerides [8,9] and total cholesterol [10-12]. Other studies of occupationally exposed persons noted elevations in rates of cardiovascular deaths [13]. In a Native American population higher PCB levels were associated with elevations of both triglycerides and total cholesterol. The elevated PCB levels were also associated with an elevation in self-reported cardiovascular disease which appeared to be secondary to the elevation in serum lipids [14], although that study was limited by the population size and the self-reported nature of the diagnosis. However, General Electric scientists, while reporting an association between levels of serum PCBs and serum lipids, have argued that the PCB elevation is a consequence, not a cause, of the hyperlipidemia [15,16].

Anniston, Alabama, is the home of one of two facilities in the US operated by Monsanto Chemical Company to produce PCBs; PCBs were made in Anniston from 1929 until 1971. In 2003 ATSDR funded a study of the health of Anniston residents through a consortium of universities. The serum PCB levels in 758 Anniston residents, ages 19–93 years, ranged from 0.1 to 170.4 ng/g (ppb), with a median of 3.2 ng/g [17]. These concentrations can be compared to the statement by ATSDR [18] that the average PCB concentration in persons in the United States that are not unusually exposed is between 0.9 and 1.5 ng/g.

The objective of the current study was to assess the relationship between the serum levels of total PCBs, different groups of congeners and single congeners as well as total chlorinated pesticides and pesticide groups with total serum lipids, serum triglycerides, total cholesterol, LDL and HDL cholesterol among residents of Anniston who were not on lipid-lowering medication.

Information about the study population, sampling methods, data collection techniques and laboratory analyses have been described in detail elsewhere [17,19], and therefore are presented only briefly here.

The study was reviewed and approved by the Institutional Review Board (IRB) of the University of Alabama at Birmingham (protocol number X050610004) with concurrence from the University at Albany IRB. Written informed consent was obtained from all participants for use of all questionnaire and analytical data for analysis and publication.

Table 1 shows demographic characteristics of the study population. All essential data were available for 575 study participants, none of whom were on any lipid-lowering medication that might obscure the results, although smoking and physical activity data were missing for a few subjects. The mean age was 52.2 and the median was 52. More than seventy percent were within the age range of 37–74 years. Those under 36 and older than 75 were 18.43% and 9.04% respectively, and graphically variable age was considered as normally distributed. Seventy percent of subjects were female. There were approximately equal numbers of Caucasians and African Americans. Nearly 70% of subjects did not drink alcohol (defined as at least one drink in the past 30 days), while nearly half smoked (defined as having smoked more than 100 cigarettes over the lifetime) and had regular exercise (defined as having daily moderate physical activity of at least 10 min duration).

Table 2 shows information on BMI and analytical values for lipids and contaminants for the study population. The mean BMI of 30.88 kg/m², with standard deviation of 8.03, indicates a high rate of obesity. Mean serum concentrations of total cholesterol, LDL cholesterol, HDL cholesterol and triglycerides were within “normal” ranges (triglycerides <200 mg/dL; total cholesterol, 120–200 mg/dL; HDL cholesterol, 35–86 mg/dL; LDL cholesterol, 80–130 mg/dL) [31], although values both lower and higher were found in some individuals. The range of exposures was wide, with serum PCB concentrations ranging from 0.1 to 170.4 ppb. Total pesticide concentrations ranged from 0.15 to 39.07 ppb wet weight. When PCB congeners were grouped by numbers of ortho chlorines or total chlorines, the highest concentration was observed for di-ortho PCBs (mean concentration, 3.95 ppb) and for hexachloro PCBs (mean concentration, 2.59 ppb). Regarding organochlorine pesticides, DDT and its derivatives composed almost 85% of mean total concentrations of pesticides. The other pesticides were in much lower concentrations.

Table 3 presents results of multiple linear regression models of association between five outcome variables and serum total concentrations of both PCBs and pesticides. Separate regression analyses were fitted for each outcome variable. Results of analyses for total PCBs and total pesticides were considered significant at p < 0.05. Each outcome variable was regressed on PCBs and pesticides with additional adjustment for the confounders (age, gender, race, BMI, alcohol consumption, smoking and exercise) (Model 1). Race and age were significant covariates in both Models 1 and 2, but gender, alcohol consumption, smoking and exercise were not. Both total PCBs and total pesticides were significantly and positively associated with elevations in total lipids, total cholesterol and triglycerides, but not with HDL or LDL cholesterol. In order to evaluate the unique effect of PCBs we therefore adjusted for total pesticides as well as the other confounders, and adjusted results for total pesticides for total PCBs as well as other confounders (Model 2). Total PCBs were significantly and positively associated with total lipids, while total pesticides were significantly and positively associated with levels of total lipids, total cholesterol and triglycerides. There was no significant association between HDL or LDL cholesterol with either total PCBs or total pesticides. The relationship between serum total POPs (the sum of total PCBs plus total pesticides) with each of five outcome variables is presented in Additional file 2. When total PCB concentration was adjusted for total pesticides (Model 2), the significant relationship with total cholesterol and triglycerides disappeared. In contrast, when total pesticides was adjusted for total PCBs, the significant associations with total lipids, total cholesterol and triglycerides, but not LDL or HDL cholesterol, did not change markedly from results of Model 1.

Table 4 shows results of the associations between serum lipid concentrations with concentrations of PCB congeners classified by number of ortho chlorines. Each outcome variable was regressed with 27 PCB congeners categorized into 3 groups based on number of ortho chlorines. Results were considered significant at p < 0.0019 (0.05/27) [30]. Separate results are presented with adjustment only for confounders other than the contaminants (Model 1) and with additional adjustment for all other groups of POPs (total pesticides and both of the other PCB groups) (Model 2). With Model 1 there were significant positive associations between concentrations of mono-ortho congeners and total lipids, di-ortho congeners with total lipids and triglycerides, and tri-plus tetra-ortho congeners with total lipid, triglycerides and total cholesterol. No association was observed with HDL or LDL cholesterol. When adjusted for other POPs (Model 2), only tri- plus tetra-ortho PCBs showed statistically significant positive correlations with total lipids and triglycerides.

Among the results shown in Table 4, there are 10 separate instances in which a reversal of signs of the regression coefficients from positive to a negative occurs between the analyses of Model 1 and Model 2. We call the reader’s attention to the fact that these results should be treated most cautiously, as they are the result of suppression effects due to severe multicollinearity among the predictors. We give further details of these statistical artifacts in the Additional file 1.

The relationships between serum lipids with PCB congeners grouped by numbers of total chlorines are shown in Table 5. Each outcome variable was regressed with 27 PCB congeners categorized into 5 groups based on number of chlorines. Results were considered significant at p < 0.0019 (0.05/27) [30]. Prior to adjustment for other POPs (Model 1), all groups showed significant associations with total lipids and triglycerides. Associations with total cholesterol were observed for groups with three or four, six, seven, and eight or more chlorines. The only significant association with LDL cholesterol was with eight or more chlorines, and there were no associations with HDL cholesterol. These associations disappeared in Model 2 except for the significant positive association between total lipids and PCBs with three or four chlorines.

Table 6 provides the associations obtained with single PCB congeners without adjustment for other POPs (Model 1). Each outcome variable was regressed with 27 separate PCB congeners. Results were considered significant at p < 0.0019 (0.05/27) [30]. The mono-ortho substituted congeners 74 and 156 were significantly associated with total lipids, total cholesterol and triglycerides, as were the higher chlorinated di-ortho congeners 172 and 194, as well as the tri-and tetra-ortho congeners 199, 196–203, 206 and 209. Associations with total lipids and triglycerides only were observed for mono-ortho congeners 28 and 189, di-ortho congeners 170 and 180, and tri-ortho congeners 178, 187, and 195. Mono-ortho congener 157, di-ortho congeners 146 and 153, and tri-ortho congeners 177 and 183 were only associated with total lipids. Mono-ortho congeners 66, 118, 105, and 167 and di-ortho congeners 87, 99, and 138–158 were not associated with any lipids. Only PCBs 194 and 209 had positive associations with LDL but not HDL cholesterol at alpha <0.0019. Application of Model 2 to single congeners was not feasible due to extreme multicollinearity.

A similar analysis for the groups of pesticides is shown in Table 7. Results were considered significant at p < 0.0019. All of the pesticide groups showed a positive association with total lipids in Model 1. In addition, DDT was positively associated with triglycerides, while total chlordane, HCB, and HCCH were positively associated with total cholesterol and triglycerides, with HCB also associated with LDL cholesterol (at alpha < 0.0019). When adjusted for other POPs (total PCBs and all other pesticides) (Model 2), most associations of lipid levels with total DDT and other pesticides disappeared. However, total chlordane remained significantly associated with triglycerides, and all associations from Model 1 for HCB remained unchanged, showing strong positive associations with all lipids except for a non-significant negative association with HDL-cholesterol.

Additional file 3 presents Pearson correlation coefficients for all of the groups of PCBs and pesticides, and demonstrates the high degree of multicollinearity among some of the groups. Additional file 4 presents Pearson correlation coefficients among the single PCB congeners. Note that the correlation coefficients are greater for higher than lower chlorinated congeners.

We observed that elevated serum concentrations of lipids were significantly associated with serum concentrations of ΣPCBs and summed pesticides in analyses adjusted for age, race, gender, BMI, alcohol consumption, smoking and exercising status. The strongest associations were seen for PCB congeners with three or four ortho and/or at least eight substituted chlorines. Serum concentrations of HCB and chlordane also had strong associations with lipid components. The results reported here are consistent with conclusions made in study of a Native American population where PCB and pesticide levels were found to be positively correlated with serum total cholesterol and triglyceride concentrations (14).

Elevated concentrations of serum cholesterol and triglycerides and low concentrations of HDL cholesterol are major risk factors for cardiovascular disease [1,32]. There is supportive evidence from both human [25,28,33] and animal [3,34,35] studies that exposure to POPs increases risk of cardiovascular disease.

While numerous studies examined association of PCBs with serum lipids as part of other analyses [9,15,16,25,26], the only previous study that has systematically investigated individual concentrations of POPs in relation to levels of serum lipids is that of Lee et al. (27). They studied a population of 90 individuals over 20 years of age, enriched with overweight and obese persons, and reported associations between concentrations of eight chlorinated pesticides and 22 PCB congeners with serum triglyceride and HDL cholesterol levels. They found heterogeneous relationships between serum concentrations of toxicants and lipids. As in our study they report significant association between concentrations of HCB and serum triglyceride levels. Some of the associations for other POPs they report were non-linear when the POPs were presented in quartiles. There are, however, significant differences in some of their results and ours. This is particularly so in relation to the effects of DDE, which they report to be associated with an elevation in triglycerides and suppression of HDL. We did not detect any effect of total DDTs, including DDE, after adjustment for other POPs and we found no significant effect on HDL cholesterol by DDE. The reason for this difference is uncertain, but may reflect the differences in the populations being studied and/or differences in adjustment for other POPs.

Obesity and change in weight also have profound effects on concentrations of lipids and levels of organochlorines. Lind et al. [36] studied life-time weight change in relation of concentration of POPs in the PIVUS study. The average estimated weight change over 50 years was 14.4 kg. Both the sum of OC pesticide concentrations (4.3 kg more weight gain in quintile 5 vs. quintile 1, p < 0.0001) and the sum of the less-chlorinated PCBs were positively related to the estimated weight change (3.7 kg more weight gain in quintile 2 vs. quintile 1, non-linear relationship p = 0.0015). In contrast, the sum of concentrations of highly-chlorinated PCBs were inversely related to estimated weight change (8.4 kg less weight gain in quintile 5 vs. quintile 1, p < 0.0001). Differences in mode of action, toxicokinetics, non-linear relationships and reverse causation might explain these discrepancies [36]. In the analysis of abdominal obesity of this cohort [37], twenty-one plasma POPs (16 polychlorinated biphenyl (PCB) congeners, 3 organochlorine (OC) pesticides, 1 brominated diphenyl ether (BDE), and 1 dioxin) were measured at baseline in 970 participants aged 70 years with prospective analyses in 511 participants re-examined after 5 years. Abdominal obesity was defined by an increased waist circumference. In the cross-sectional analyses, concentrations of the less chlorinated PCBs, OC pesticides such as p,p’-DDE and dioxin had adjusted odds ratios of 2 to 3 for abdominal obesity. Many relations had inverted U-shapes rather than being linear, particularly in women. In contrast, concentrations of highly chlorinated PCBs were strongly inversely associated with abdominal obesity. In a single model including summary measures of the less chlorinated PCBs, highly chlorinated PCBs, and OC pesticides, both the positive associations and inverse associations strengthened. Similar but somewhat weaker associations were seen between POPs and risk of development of abdominal obesity in the prospective analyses.

In addition, experimental animal studies have shown dose-dependent increases in serum levels of total cholesterol and triglycerides in response to exposure to PCBs [3,4,6,38]. Exposure to PCBs changes the patterns of lipid metabolism and synthesis [39]. In this study, animals were fed either an olive oil or a corn-oil rich diet, and then exposed to PCB 77, a coplanar, dioxin-like congener. They found that the PCB treatment in both groups caused induction of genes involved in fatty acid degradation but that there was upregulation of different genes. These observations raise another possible complication, in that while different PCB congeners and different pesticides may induce different genes, different kinds of dietary fat may influence which genes are induced. As suggested above, PCBs can initiate inflammatory responses in vivo, and this inflammation can be either exacerbated or ameliorated by nutrition. Petriello et al. [40] also showed that diets high in certain dietary lipids such as omega-6 fatty acids can worsen PCB-induced vascular toxicity while diets enriched with bioactive food components such as polyphenols and omega-3 polyunsaturated fatty acids can improve the toxicant-induced inflammation. There is evidence that bioactive nutrients protect through multiple cell signaling pathways, but they have shown that lipid raft caveolae and the antioxidant defense controller nuclear factor (erythroid-derived 2)-like 2 (Nrf2) both play a predominant role in nutritional modulation of PCB-induced vascular toxicity. Interestingly, there appears to be an intimate cross-talk between caveolae-related proteins and cellular Nrf2, and focusing on the use of specific bioactive food components that simultaneously alter both pathways may produce a more effective and efficient cytoprotective response to toxicant exposure. The use of nutrition as a protective tool is an economically beneficial means to address the toxicity of persistent environmental toxicants and may become a sensible means to protect human populations from PCB-induced vascular inflammation and associated chronic diseases [40].

The precise mechanisms responsible for the associations between serum POPs and the various serum lipid components are uncertain. Organochlorines are difficult to metabolize, and exposure causes induction of various degradative enzymes, particularly different cytochrome P450s [41,42]. In addition, DNA methylation levels in humans have been found to be altered by several persistent organic pollutants, including PCBs and chlorinated pesticides [43], indicating that these chemicals can also induce epigenetic changes. Liver size in animals actually doubles in situations of high exposure, reflecting enzyme induction [44]. Others have documented specific alternations of enzymes involved in lipid synthesis caused by exposure to single PCB congeners [45]. A variety of genes are either up-or down-regulated by different PCB congeners [46,47], p,p’-DDE [48] and by a mixture of 28 persistent organic pollutants [49]. Thus, there are multiple possible pathways that might lead to selective alteration in rates of synthesis or metabolism of some classes of serum lipids.

Serum concentrations of total PCBs (sum of 35 congeners) and total chlorinated pesticides (sum of nine pesticides) were significantly associated with elevated serum levels of total lipids, total cholesterol and triglycerides but not levels of LDL and HDL after adjustment for age, gender, race, BMI, smoking and exercise in a study population who were not on lipid-lowering medication. After adjustment for confounders, associations between different groups of POPs and serum lipids were not homogenous. The PCBs with the strongest associations were those with three, four, or eight or more total chlorines, while, of the pesticides, the most striking associations were observed for hexachlorobenzene, even though concentrations of it were low. Chlordane also was found to be positively associated with some lipid components after adjustment for other contaminant concentrations. These results suggest that, in spite of the fact that concentrations of all of these lipophilic chlorinated compounds are highly correlated, the effects on lipid levels are different for different pesticides and PCB congeners. Our results are consistent with the hypothesis that higher levels of POPs, especially higher chlorinated PCBs and some pesticides, result in elevations of serum lipids and that risk of cardiovascular disease and obesity after exposure to PCBs and chlorinated pesticides is, at least in part, a consequence of the elevations in serum lipids that result from exposure.

ATSDR: Agency for toxic substances and disease registry; BMI: Body-mass index; DDE: Dichloro-dichlorophenyl-ethylene; DDT: Dichloro-diphenyl-trichloroethane; DNA: Deoxyribonucleic acid; HCB: Hexachlorobenzene; HCCH: Hexachlorocyclohexane; HDL: High density lipoprotein cholesterol; LDL: Low density lipoprotein cholesterol; PCBs: Polychlorinated biphenyls; POPs: Persistant organic pollutants; Ppb: Parts per billion; Ww: Wet weight.

DOC has served as an expert witness in legal actions against Monsanto relating to exposure of resident of Anniston, Alabama to PCBs, with all compensation paid to the Research Foundation at the University at Albany. The other authors declare that they have no competing interests.

ZA participated in the planning of the study, carried out most of the data analysis and this study constituted a part of his PhD thesis. RH reviewed and advised on the statistical analysis and wrote the material contained in the Additional file 1. MP reviewed and advised on all aspects of the study. DOC designed the study and supervised all aspects of data analysis and preparation of the manuscript in addition to receiving the funding for the project. All authors read and approved the final manuscript.