Plasma levels of polychlorinated biphenyls and risk of cutaneous malignant melanoma: a preliminary study
A number of epidemiologic studies have suggested that exposure to polychlorinated biphenyls (PCB) and other organochlorine compounds (OCC) increase risk of cutaneous malignant melanoma (CMM). However, these studies have generally had no biologic measure of OCC exposure, and have been unable to control for sun exposure, the major known environmental risk factor for this disease. This preliminary study examined the relationship between OCC residues in plasma and risk of CMM adjusting for sun sensitivity and sun exposure. A case–control study of 80 CMM patients and 310 control subjects was conducted. Lifetime sun exposure information, along with data on pigmentation variables and sun sensitivity data was collected, along with a blood sample. Cases and controls were assayed for plasma levels of 14 PCB congeners and 11 organochlorine pesticide residues using gas chromatography. Strong associations were seen between risk of CMM and plasma levels of non-dioxin-like PCBs (Adjusted OR = 7.02; 95% CI: 2.30–21.43 for highest quartile) and several PCB congeners, organochlorine pesticides or metabolites. These associations persisted after control for sun sensitivity and sun exposure. Results from this investigation require independent confirmation in larger studies. However, they suggest that environmental factors other than UV radiation may play a role in genesis of CMM, and indicate that it may be productive to search for further agents which might increase risk.
The principal known environmental risk factor for cutaneous malignant melanoma (CMM) is ultraviolet (UV) radiation1 from sunlight2 or artificial sources.3 It has been estimated that 65% of CMM worldwide is attributable to this one environmental agent.4 Few studies have been conducted to look for other environmental and occupational risk factors for CMM, but among these, several5–13 have shown an elevated risk in occupations suspected to be exposed to organochlorine compounds (OCC) such as polychlorinated biphenyls (PCBs) and chlorine-based pesticides. The investigations showing a positive relation between CMM risk and OCC exposure were mostly conducted to study risk of more common cancers within occupational cohorts, and the findings for CMM were largely incidental, and based on small numbers of events. As well, most studies had no measure of OCC exposure among the study subjects; exposure was imputed through job title. Finally, little or no information was available to enable control of potential confounding by sunlight exposure, and sun sensitivity; the major risk factors for melanoma.
To investigate the relationship between plasma levels OCCs and risk of CMM, controlling for sun exposure and sun sensitivity, we conducted a small case–control study in British Columbia, Canada.
Material and Methods
The protocol for this investigation was approved by the Research Ethics Board of the University of British Columbia and the BC Cancer Agency.
This study used plasma specimens and sun exposure data from melanoma cases originally recruited to evaluate the effect of UV exposure and gene variants on risk of melanoma (GEM study)14, 15 and controls from a previous case–control study designed to assess the effect of solar UV exposure and plasma OCCs on risk of non-Hodgkin lymphoma.16 Both studies recruited participants, using the population-based BC Cancer Registry, and the same computer-assisted telephone interview (CATI) protocol to collect information on sun exposure, sun sensitivity and phenotype factors. A total of 153 patients with histologically verified CMM contributed data and buccal swab DNA for the GEM study. Eighty-six patients were later re-contacted and contributed the blood specimens used in the current study. Blood specimens for the CMM cases were collected during the same time period (2000–2004) as those for controls for the NHL study. Blood sample collection and processing used the same methods for both melanoma cases and NHL controls as previously described.16 All organochlorine assays were performed at the Centre de Toxicologie in Quebec, Canada; assays for the controls between 2002 and 2005, and for CMM cases in 2008.
During the initial investigation of OCCs and risk of NHL,16 plasma samples from 460 controls were assayed for 14 PCB congeners and 11 chlorine-based pesticides or their metabolites. After matching within 5-year age group to CMM cases; and limiting to Caucasian participants (as all the CMM cases were Caucasian), and restricting to residence in the same geographic area as the CMM cases; data and assay values for 309 controls were available for the current analysis. A total of 83 CMM plasma specimens were assayed for the same PCB congeners, and pesticides or their metabolites as controls. Three (3) melanoma cases were excluded due to inability to match to controls within 5-year age group, which resulted in 80 CMM cases being available for the current analysis.
The 14 PCB congeners were classified according to their chlorination content and included IUPAC nos. 28, 52, 99, 101, 105, 118, 128, 138, 153, 156, 170, 180, 183 and 187. The 11 pesticides examined were aldrin, p,p′-DDT, hexachlorobenzene (HCB), β-hexachlorocyclohexane (β-HCCH), mirex and four chlordane compounds (α-chlordane, γ-chlordane, cis-nonachlor and trans-nonachlor). Two other analytes, p,p′-DDE, a metabolite of DDT, and oxychlordane, a metabolite of chlordane, were also examined.
Free cholesterol (FC), total cholesterol (TC), triglycerides (TG) and phospholipids (PL) were measured in each plasma sample using enzymatic methods. Total lipid concentration was determined using the Akins summation formula,17 and lipid-adjusted organochlorine concentrations were calculated for each sample by dividing the whole-weight measurements by the total lipid concentration. Although analyses were conducted on whole-weight and lipid-adjusted plasma concentrations, only lipid-adjusted values are reported.
Relationships between the different lipid-adjusted PCB congeners and pesticides or pesticide metabolites were examined using Spearman rank correlations for organochlorines with more than 80% of samples above the detection limit. Samples with organochlorine concentrations below the limit of detection were assigned a value of the detection limit divided by √2.18
Aldrin, α-chlordane, γ-chlordane and PCB congeners nos. 28, 52, 101 and 128 had fewer than 5% of samples with detectable levels and were excluded from further analysis. Lipid-adjusted plasma levels were categorized before data analysis based on values among the controls. OCC analytes with fewer than 25% of samples below the detection limit were categorized into quartiles according to their distribution in the control samples. For organochlorine analytes with more than 25% of the samples below the detection limit (cis-nonachlor, p,p′-DDT, mirex and PCB congeners nos. 105 and 183), concentrations were categorized as either detectable or nondetectable. The lowest quartile or category was used as the reference category and values below the detection limit were always included in the lowest quartile.
Because of the time period between assessment of OCC levels in the cases and controls, consistency of the OCC assay process was examined by repeat measuring of OCCs in 2009 in 44 controls from the NHL study with two blood samples; and comparing the levels with those obtained earlier. Measured levels were comparable for all OCCs except PCBs 99 and 105, which showed significantly lower levels in the 2009 measurement (data not shown). These congeners were excluded from further analysis.
The total sum of all PCB congeners, dioxin-like PCB congeners (PCB nos. 118 and 156) and non-dioxin-like PCB congeners19 were computed by summing the individual serum values of each PCB congener. After lipid adjustment, the summary PCB congener values were categorized into quartiles according to the distribution in controls.
Odds ratios (OR) and 95 percent confidence intervals (95% CI) for risk were estimated using unconditional logistic regression relationships between sun sensitivity, recreational sun exposure, pigmentation variables and risk of CMM were assessed in cases and controls adjusting for age, gender and education level. Tests for trend across quartiles were conducted by creating a continuous variable with assigned values equal to the median level among controls within each category. To determine the median levels, values below the detection limit for individual analytes were assigned a value of the detection limit divided by √2, as noted. The change in estimate criterion was used to select confounders, with more than a 5% change in the estimate considered important. Possible confounders included age, gender, education, eye color, hair color, skin color, degree of freckling, number of sunburns, sun sensitivity and sun exposure. Age was examined in nine categories (<40, 40–75 in 5-year age groups and 75+), in five categories (<40, 40–49, 50–59, 60–69 and 70+) and in four categories (20–49, 50–59, 60–69 and 70+). Age in four categories along with, gender, education, sun sensitivity (skin reaction to repeated sun exposure) and recreational sun exposure (total lifetime beach, pool, boating and sunbathing recreational activity hours) were included as covariates in the final-adjusted models to estimate the association for each individual OCC and the total summed PCB groups. The addition of other covariates, or finer age categories did not materially alter the ORs.
Interactions between covariates and each of the PCB congeners and pesticide analytes were examined by entering the interaction terms in the logistic regression model. Significance was based on the likelihood ratio p value for the interaction terms.
Information on the number of samples measured above the detection limit for each analyte is shown in Table 1. Plasma levels of all PCB congeners and pesticide analytes were significantly correlated at p < 0.001, with correlations ranging between 0.19 and 0.98 (data not shown). OC levels for all analytes were higher in older individuals; the median concentrations in subjects over 60 were almost double the median value in subjects younger than 60 for PCB congeners and pesticides (data not shown).
Table 1. Number of samples measured above limit of detection (unadjusted serum levels for PCB congeners and pesticide analytes)
Characteristics of eligible CMM cases and controls are shown in Table 2. Cases were roughly similar to controls with respect to age and sex, but cases had a lower level of education than controls as indicated by the smaller proportion of individuals with a postsecondary degree.
Table 2. Characteristics of melanoma cases and NHL controls
Associations between cases and controls by pigmentary characteristics, sun sensitivity and recreational sun exposure are shown in Table 3. A higher proportion of controls had missing data on phenotype factors (hair/skin/eye color, freckling) than cases. Compared to controls, cases were more likely to tan only mildly or not at all with repeated sun exposure. A significant positive association was observed with total lifetime recreational sun exposure hours (calculated as the sum total of beach, pool, boating and sunbathing activities) for the highest versus the lowest quartile. CMM cases showed a slightly higher prevalence estimate for lifetime and childhood sunburns than controls, although the differences were not statistically significant.
Table 3. Comparison of phenotypic characteristics, sun sensitivity and sun exposure for melanoma cases and NHL controls
Overall, there was a significant association between total summed PCB level and CMM (Table 4), with a sixfold increased risk for those in the highest quartile compared to those in the lowest quartile (OR = 6.02, 95% CI: 2.00–18.17). A strong association was observed for the summed non-dioxin-like PCB level, with those in the highest quartile having an OR of 7.02 (95% CI: 2.30–21.43) compared to those in the lowest quartile. The association between melanoma and summed total dioxin-like PCB levels was weaker, but still increased for the highest versus the lowest quartile (OR = 2.84, 95% CI: 1.01–7.97).
Table 4. Association between melanoma and summary and individual PCB congeners
Table 4 also presents results for individual PCB congeners. Increased ORs were observed for individual dioxin-like PCB congeners (PCB congener nos. 118 and 156) and all six of the non-dioxin-like PCB congeners (PCB congener nos. 138, 153, 170, 180, 183 and 187). The strongest association was observed for PCB congener no. 187 with an OR for the highest versus the lowest quartile of 11.47 (95% CI: 3.32–39.68).
Table 5 shows results for CMM and chlorine-based pesticides and their metabolites. Overall, five of the eight pesticide analytes exhibited a statistically significant positive dose-response trend. The strongest relationship was observed for trans-nonachlor (OR = 4.26, 95% CI: 1.37–13.26). There was only a weak nonstatistically significant association observed with p,p′-DDE and no association with p,p′-DDT.
Table 5. Association between melanoma, and pesticides and pesticide metabolites
No significant interactions were observed between any of the PCB congeners or pesticide analytes, including total summed PCB levels, and measures of sun sensitivity or recreational sun exposure, although the power to detect such interactions was limited due to the relatively small number of melanoma cases.
The International Agency for Research on Cancer's evaluation of PCBs is that they are probable carcinogens (group 2A) based on sufficient evidence in animals, but only limited evidence in humans.20 The role PCBs might play in carcinogenesis in melanoma is suspected to be that of a promoter rather than an initiator. The results seen in the current analysis suggest that PCB exposure may increase risk of melanoma.
However, if PCBs and chlorine based pesticides are involved in the genesis of CMM, an explanation must be found as to why incidence rates are an order of magnitude higher in Caucasians than in more heavily pigmented ethnic groups21; who are unlikely to have substantially lower plasma levels of OCCs. Results from analytic studies indicate that differences in melanoma risk are due not only to sun sensitivity and sun exposure2, 22 but also to differences in nevus density.23 Studies of the evolution of nevi indicate that nevus density increases with age and exposure to solar UV throughout childhood among white children.24, 25 Although nevus density also increases slightly in non-white children, ultimately, density in non-white adolescents (age 16–18) is substantially lower than among Caucasian adolescents.26 It is also known that a significant proportion of melanomas may arise from existing nevi.27 If nevus density in childhood and adolescence is an indicator of the number of melanocytic clones which have undergone the initial step in carcinogenesis, and PCB exposure acts as a promoter of these transformed clones of cells, then perhaps the ultimate probability of developing melanoma among people from lightly and heavily pigmented groups might be determined primarily by the density of cells transformed by early solar UV exposure, which would be amenable to the promotion activity of PCBs. Such a model would assume no significant interaction between UV and PCBs, as seen in our study.
One of the questions that arise from the results presented here is the origin or source of high plasma levels of PCBs and other OCCs. As manufacture of PCBs and use in industrial applications ceased some years ago, it is likely that the main source is diet.28, 29 Measurable PCB levels are found in cream and cheese, butter, meat and especially, fish.30, 31 Although levels of these compounds in food are declining over time, living animals including humans bioaccumulate and store the compounds in fatty tissue.29, 32, 33 Studies of diet and melanoma have—for the most part—not found consistent differences between melanoma cases and controls for dietary intake of foods known to carry high levels of PCBs.34–45 However, this may not be unexpected as bioaccumulation of PCBs and other OCCs through very small difference in daily dietary intake; undetectable by today's dietary intake instruments; could result in major differences in plasma levels over time. In any case, our study is unable to evaluate this, as no dietary history was taken for cases or controls.
As noted in the introduction to this article, the original suggestion for an association between PCBs and CMM arose from occupational studies, suggesting that specific occupations might be the source of exposure.8–13 However, occupations with high PCB exposures are relatively uncommon, and it seems unlikely that occupation could account for the increased PCB levels seen in a study such as ours, which recruited both male and female cases and controls from the general population. No data were collected on occupational background in our study.
Strengths of the current study include population-based ascertainment of cases and controls, a well-recognized biomarker of PCB and pesticide burden, and good control for sun exposure, sun sensitivity and phenotypic characteristics known to affect risk of melanoma. The fact that known relationships between CMM and pigmentary variables, sun exposure and sun sensitivity were observed in comparisons between the cases and controls gives confidence that the study participants are similar to those in previous population-based etiologic studies of melanoma.
However, the study has several weaknesses. The number of CMM cases available for analysis was relatively small, leading to fairly wide confidence intervals around risk estimates, and, in addition, the study had limited power to detect interactions. Since the cases and controls came from different studies there could be concerns about comparability. However, controls were slightly older than cases, and with levels of PCBs and other OCCs known to rise with age, study results should, if anything, be conservative. Although adjusted ORs for PCBs and other OCCs were presented, the possibility of residual confounding remains. The cases and controls were interviewed during the same time period concerning their sun exposure and sun sensitivity using the same questions, and their blood specimens were collected during the same time period using the same methods and were stored under identical conditions (−80°C). However, OCC assays for the controls were conducted 3 years before those for the CMM cases, and slight changes in assessment methods over this period could potentially have affected comparability of results. However, as noted earlier, plasma from 44 controls assayed in 2005 and again in 2008 showed substantially the same values for the PCBs and OCCs included in this analysis.
Finally, since blood specimens for CMM cases were collected after disease development, there is a possibility that the disease process or its treatment may have affected plasma levels of OCCs. It is known that weight loss due to chemotherapy may lead to increases in plasma levels as the bioaccumulated OCCs stored in fat are mobilized.46 However, none of the melanoma cases had undergone chemotherapy. The primary mode of treatment for melanoma is localized surgery, which is unlikely to result in significant weight loss. In addition, blood samples were collected from CMM patients well after any minimal weight loss from treatment might have occurred.
In conclusion, results of this preliminary study suggest that environmental factors apart from UV radiation may play a part in CMM. The results need confirmation in larger investigations, perhaps carried out internationally to take into account ethnic differences in melanoma risk. The results also indicate that it may be productive to revisit the results from past occupational and environmental studies47 to generate hypotheses concerning factors other than UV radiation, which might affect melanoma risk.