Environment International 113 (2018) 20–25

Contents lists available at ScienceDirect

Environment International
journal homepage: www.elsevier.com/locate/envint

Plasma levels of polychlorinated biphenyls and risk of cutaneous malignant T

melanoma: A hospital-based case-control study☆
⁎
Michele Magonia, , Pietro Apostolib, Francesco Donatoc, Ausilia Manganonid, Pietro Combae,
Lucia Fazzoe, Fabrizio Speziania, Lucia Leonardia, Grazia Orizioa, Carmelo Scarcellaa,
Piergiacomo Calzavara Pintond, and Brescia Melanoma-PCB Working Group1
a
ATS Brescia (Brescia Health Protection Agency), Brescia, Italy
b
Department of Medical and Surgical Specialties Radiological Sciences and Public Health, Institute of Occupational Health and Industrial Hygiene, University of Brescia,
Italy
c
Department of Medical and Surgical Specialties Radiological Sciences and Public Health, Unit of Hygiene, Epidemiology, and Public Health, University of Brescia, Italy
d
Section of Dermatology, Department of Clinical and Experimental Sciences, University of Brescia, Italy
e
Department of Environment and Health, Istituto Superiore di Sanità, Rome, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: Polychlorinated biphenyls (PCB) have been classified by the International Agency for Research on Cancer (IARC)
Melanoma in Group 1 as carcinogenic to human, based on sufficient evidence in humans of an increased risk of cutaneous
Polychlorinated biphenyls malignant melanoma (CMM), however few studies have been done in the general population. This study ex-
Environmental risk factors amined the relationship between PCB plasma levels and risk of CMM adjusting for sun sensitivity and sun
exposure in a province of Northern Italy (Brescia), where a chemical factory produced PCBs from 1938 to 1984
causing human contamination. A case–control study of 205 CMM patients and 205 control subjects was con-
ducted. Cases and controls were assayed for plasma levels of 33 PCB congeners. No associations was found
between risk of CMM and plasma levels of total PCB (OR = 0.81; 95% CI: 0.34–1.96 for highest vs lowest
quartile) or specific congeners. The study confirmed the association with light skin colour (OR = 3.00; 95% CI:
1.91–4.73), cumulative lifetime UV exposure (OR = 2.56; 95% CI: 1.35–4.85) and high level of education
(OR = 1.45; 95% CI: 1.03–2.05). This case-control study does not support the hypothesis of an association
between current plasma levels of PCBs and CMM development in the general population.

1. Introduction
The global incidence rate of cutaneous malignant melanoma (CMM)
in 2015 was estimated to be about five cases per 100,000 persons, with
the greatest burden of incidence and mortality in the caucasic popu-
lation of Australasia, North America and Europe (Karimkhani et al.,
2017); in Italy is the third most frequent cancer in both sexes in the
under 50 years old population (AIOM and AIRTUM, 2016).
The most important risk factors for CMM include exposure to en-
vironmental or artificial ultraviolet radiation (UVR) (International
Agency for Research on Cancer (IARC), 1992), genetic predisposition,
phenotypic features including fair phototype, number of acquired
naevi, and aging (Fava et al., 2015; Caini et al., 2009).
However, the increasing incidence of melanoma in Italy and other
European countries in the last decades cannot be fully explained by the
hypothesis of an increase in cumulative lifetime UVR exposure and
number of episodes of sunburns (AIOM and AIRTUM, 2016; Holterhues
et al., 2013).
Most of current knowledge of melanoma aetiology derives from
study of patients in populations of European descent, for whom the use
of sun protection tools and screening procedures have appreciably de-
creased mortality. However, some melanoma subtypes that most com-
monly develop in other populations are not associated with exposure to
ultraviolet (UV) light, suggesting a different disease aetiology. Further

☆
The authors declare no potential conflicts of interest.
⁎
Corresponding author.
E-mail address: michele.magoni@ats-brescia.it (M. Magoni).
1
Members of the Brescia Melanoma-PCB Working Group: Carmen Terraroli (Chief for coordination and clinical research, ASST Spedali Civili), Arianna Coniglio (Surgical Clinic,
Department of Experimental and Clinical Sciences, University of Brescia - Division of General Surgery, Brescia Hospital), Guido Zarattini (Orthopaedic Clinic University of Brescia, II
Division of Orthopaedic and Traumatology Spedali Civili of Brescia, Italy), Giorgio Manca (Department of Plastic Surgery, ASST Spedali Civili, Brescia, Italy), Jacopo Fostinelli
(Department of Medical and Surgical Specialties Radiological Sciences and Public Health, Institute of Occupational Health and Industrial Hygiene, University of Brescia), Roberto Festa,
Maria Conti, Stefania Micheletti (Brescia Health Protection Agency).

https://doi.org/10.1016/j.envint.2018.01.018
Received 14 November 2017; Received in revised form 15 January 2018; Accepted 20 January 2018
0160-4120/ © 2018 Elsevier Ltd. All rights reserved.
M. Magoni et al.
study of these subtypes is necessary to understand their risk factors and
genomic architecture, and to tailor therapies and public health cam-
paigns to benefit patients of all ethnic groups (Ossio et al., 2017).
At the same time, other risk factors for melanoma have been in-
vestigated, including non-UV related environmental and occupational
carcinogens. In 2016, the International Agency for Research on Cancer
(IARC) upgraded the classification of Polychlorinated Biphenyls (PCBs)
to Group 1 “Carcinogenic to humans”, based on sufficient evidence in
humans of an increased risk of CMM. The decision was also supported
by experimental data on animals and the evidence of aryl hydrocarbon
receptor (AhR)-mediated mechanisms of carcinogenicity for dioxin-like
PCB congeners (IARC, 2016).
Two more recent meta-analyses, however, found no, or limited,
evidence of association between PCB exposure and risk of CMM and
concluded that more epidemiological, clinical and laboratory studies
are needed (Boffetta et al., 2016; Zani et al., 2017).
A chemical factory located in Brescia, an industrialized town in
Northern Italy, produced PCBs and other organochlorines from
the1930s to the 1980s, causing heavy PCB pollution of soil, water and
local food (Donato et al., 2006; Turrio-Baldassarri et al., 2007; Turrio-
Baldassarri et al., 2009). Some studies, performed in 2001–03 docu-
mented high levels of PCBs in the population living in the most polluted
area of Brescia and in people who had eaten local products (Donato
et al., 2006; Turrio-Baldassarri et al., 2007; Turrio-Baldassarri et al.,
2009). Subsequently, various public health interventions have been
established for reducing the environmental impact of PCB pollution. In
fact, PCB serum levels decreased by more than half, on the average, in
people living in the most polluted areas (approximately 25,000 re-
sidents), from 2003 to 2013 (Magoni et al., 2016)
The present study aimed to evaluate the association between cur-
rent serum levels of total PCBs and specific PCB congeners and CMM in
the population living in Brescia, taking into account the main risk
factors for the disease. Secondary objectives were evaluation of the
possible risk of CMM for PCB exposure according to CMM histology,
gender, age, and the presence or absence of the main risk factors for
CMM.
2. Materials and methods
The study protocol was approved by the local Ethics Committee
(Protocol Number 1695, 12/04/2014).
The present hospital-based case-control study took place between
July 2014 and November 2016 at the Spedali Civili Hospital of Brescia.
CMM cases were recruited among patients attending a tertiary referral
centre for CMM. We enrolled consecutively adult, Italian, Caucasian
CMM patients who had first diagnosis of the disease based on histolo-
gical lab results (incident cases); 19 of them reported history of a pre-
vious different melanoma that had been cured successfully, and there-
fore the present was considered as a new, incident, case. The CCM cases
included also melanoma in situ.
Age and sex matched controls were recruited in the Surgical and
Orthopaedic Departments of the same hospital among patients without
history of cancer, hepatic, endocrine or autoimmune diseases. Enrolled
people signed an informed consent, were face-to-face interviewed by a
trained nurse using a structured questionnaire and gave a blood sample
for laboratory analysis.
The interview lasted 30–45 min and collected information on the
subject's demographic variables, residential and occupational history,
smoking habit, and exposure to the main risk factors for melanoma.
According to study design, the enrolment of 200 cases and 200
controls would have provided 90% power of showing a statistically
significant odds ratio of 2 or more for the highest compared to the
lowest quartile of PCB distribution using a two-sided test with
alpha = 0.05.
21
Environment International 113 (2018) 20–25
2.1. Laboratory analyses on serum samples
The PCB analyses were performed at the Institute of Occupational
Health and Industrial Hygiene of the University of Brescia (Italy) using
the same methodology described in previous studies carried out in the
same population (Magoni et al., 2016; Raffetti et al., 2017).
A 20 ml blood sample was collected for each subject under fasting
conditions for determination of PCB congeners and some biochemistry
parameters after the diagnosis confirmation and before starting che-
motherapy. The following 33 PCB congeners were determined: 28, 31,
52, 74, 77, 81, 99, 101, 105, 114, 118, 123, 126, 128, 138, 146, 153,
156, 157, 167, 169, 170, 172, 177, 180, 183, 187, 189, 194, 196, 201,
203, 206 and 209. We adopted, with minor adjustments, a previously
published PCB analysis method (Turci et al., 2002), using an Agilent
Technologies 6890N gas chromatograph coupled with an Agilent
Technologies MSD 5973 (electron impact ionization, mass filter:
quadrupole). A PONA column (Agilent Technologies; 50 m × 0.20 mm
ID) was used for chromatographic separation with carrier gas Helium. A
2 ml injection at 250 °C was performed by a 7683 Series Injector (Agi-
lent Technologies) in splitless mode with a salinized injection liner
(Agilent Technologies; 4 mm, 78.5 × 6.5 OD).
The limit of quantification (10 times the signal-to-noise ratio peaks)
varied among PCBs but was generally < 0.1 ng/ml for each congener.
PCB serum levels are reported as volumetric values (ng/ml). The
total PCB value was computed by summing the serum values of each
PCB congener. Lipid-adjusted PCB concentrations are also reported and
expressed as ng/g lipid. Total serum lipid concentration was computed
by serum cholesterol and triglyceride levels, using the formula pro-
posed by Phillips et al. (1989): total serum lipid (g/l) = 2.27 ∗ serum
cholesterol (g/l) + triglycerides (g/l) + 0.623. A twin version of Tables
4, 5,and 6 with PCB values expressed in ng/g lipid is provided as
Supplementary tables.
2.2. Statistical analyses
Due to asymmetric, non-normal distribution of PCB values, the
medians, range and 75th and 95th centile of the distribution are re-
ported together with arithmetic means and standard deviations (SDs).
Non-parametric statistical methods for distribution comparison and
Spearman's correlation coefficient were used for continuous variables
and chi square for linear trend tests were used for categorical variables.
The odds ratios (ORs) and their 95% confidence intervals (95% CIs)
were estimated for assessing the associations between CMM and pig-
mentation variables, family and personal history of skin cancer, sub-
ject's sun sensitivity, sun exposure, education level and working history
in agriculture and chemical factory, adjusted for gender and age by
logistic regression analysis. A multivariable logistic regression model
was fitted to assess the risk of CMM for PCBs serum levels, including
age, gender and the investigated risk factors retained in the final model
as possible confounders. The final model was chosen by a stepwise
backward procedure with a p value of 0.1 for removing variables.
PCB serum concentration was considered both as a continuous, log
transformed, variable and an ordinal variable using quartiles of PCB
distribution in all subjects. These analyses were performed for total PCB
and single PCB congeners detectable in at least 50% of the subjects. The
lowest quartile was used as the reference category and subjects with
PCB values below the detection limit were included in the lowest
quartile.
A sub-group analysis was performed in people younger and older
than the median age of the sample (56.67 years).
All the statistical tests were two sided and the corresponding p va-
lues are reported; 95% confidence intervals (CIs) of the ORs were
computed using the commonly suggested methods. Statistical analyses
were performed using the STATA software (Stata Statistical Software
release 12.1, 2013; Stata Corporation, College Station, Texas).
M. Magoni et al. Environment International 113 (2018) 20–25

Table 1 Table 2
Characteristics of cutaneous malignant melanoma cases and controls. Phenotypic and constitutional characteristics, sun sensitivity, UV exposure, level of
education and occupational history in melanoma cases and controls and odds ratio (OR)
Cases (n = 205) Controls (n = 205) estimates with their 95% confidence intervals (95% CIs) adjusted for age and gender by
logistic regression analysis.
N % N %
Characteristics Cases Controls OR 95% CI p for trend
Age at diagnosis (years)
Mean (SD) 55.1 (14.2) 54.9 (14.0) N N
Age (years)
20–49 77 37.6 74 36.1 Skin colour
50–59 50 24.4 55 26.8 Dark 59 121
60–69 48 23.4 48 23.4 Fair 146 84 3.56 2.36–5.38 < 0.0001
70+ 30 14.6 28 13.7 Hair colour
Gender Dark 67 95
Male 119 58.1 117 57.1 Light brown 96 89 1.53 1.00–2.34
Female 86 41.9 88 42.9 Blond/red 42 29 2.98 1.60–5.52 < 0.0001
Residence Melanoma family historya
Brescia town 53 25.9 58 28.3 No 175 189
Brescia Province excluding town 136 66.3 136 66.3 Yes 30 15 2.17 1.13–4.17 0.02
Outside province 11 5.4 16 7.8 History of melanoma
No 186 204
Yes 19 1 21.3 2.8–161.3 0.003
3. Results History of non-melanoma skin
cancer
No 165 198
A total of 205 cases and 205 controls were recruited; participation Yes 40 5 9.59 3.70–24.9 < 0.0001
was 75% among CMM patients and 69% among controls. Age and Cumulative lifetime sun
gender distribution in cases and controls were almost identical exposures
Low 165 184
(Table 1) as expected by study design. The residence at recruitment was
High 40 21 2.13 1.20–3.77 0.009
also similar in cases and controls: the percentage of those living in town Sun burns history
was 25.9% among cases and 28.3% among controls. No 87 125
At histological examination, 153 (74.6%) patients had a superficial Yes 118 79 2.28 1.51–3.43 < 0.0001
spreading melanoma, 27 (13.2%) in situ melanoma, 21 (10.2%) lentigo Sun burns in childhood
No 164 187
maligna melanoma and 4 (2.0%) nodular melanoma (data not shown in
Yes 41 18 2.61 1.44–4.73 0.002
Tables). Sunlamps use
The associations between CMM and phenotypic and constitutional Never 140 156
characteristics, UV exposure, level of education and professional his- Occasionally 55 45 1.47 0.90–2.41
Often 10 3 3.93 1.05–14.7 0.027
tory, adjusted for age and gender, are shown in Table 2: all the con-
Holidays in tropics
stitutional factors (skin and hair pigmentation, family and subject's No 126 149
history of skin cancer) as well as UV exposure factors (cumulative life Yes 77 54 1.70 1.11–2.59 0.014
sun exposure, sun burns history, sunlamps use and holidays in tropics) Regular use of sun protection
showed a strong positive association with CMM with odds ratio 95% methods
No 170 174
confidence interval clearly above one; regular use of sun protection
Yes 35 29 1.30 0.74–2.31 0.4
methods did not show any association. CMM risk was also associated Education
with high level of education. Previous occupations in the agricultural < High school 81 105
sector (OR = 1.69) and in any chemical factory (OR = 2.23) showed High school 95 81 1.60 1.04–2.47
University 29 19 2.05 1.07–3.95 0.010
associations with CMM although with wide confidence intervals. The
Work in agriculture
analyses for residential history were poorly informative, because of the No 195 199
limited number of subjects who had resided in the most polluted area or Yes 10 6 1.69 0.60–4.77 0.32
in those at intermediate level of PCB pollution. Work in a chemical factory
The serum levels of total PCBs and single PCB congeners in cases No 188 197
Yes 17 8 2.23 0.94–5.28 0.069
and controls are reported in Tables 3 and 4. A wide range of values was
History of residence in the
found in both cases and controls, from 0.1 to 16 and −22 ng/ml, re- most polluted area
spectively. The distribution of lipid-adjusted PCB serum levels is also No 203 204
shown (Table 3 and Supplementary Table 4b): a high correlation was Yes 2 1 1.89 0.10–26.4 0.6
History of residence in the
found between volumetric and lipid-adjusted measures (Spearman's
most and intermediate
rho = 0.98). The distribution of total PCBs serum values was highly polluted areas
skewed and did not approximate the normal distribution (Skewness/ No 184 188
Kurtosis test: p < 0.0001; Shapiro-Wilk test: p < 0.0001), with most Yes 21 17 1.16 0.60–2.40 0.6
subjects having low values but a few ones extremely high values. Sev-
a
1st and 2nd degree familiarity.
eral PCB congeners were undetectable in all or almost all subjects
whereas congeners 138, 153 and 180 were found in > 90% of both
cases and controls. No differences in serum values of both total PCB and transformed total PCB serum levels whereas significant associations of
single PCB congeners were observed between cases and controls. The CMM with well-known risk factors (skin and hair colour, history of
PCB values were strongly associated with age (Spearman's rho = 0.74 melanoma and other skin cancers, cumulative lifetime sun exposures
in cases and rho = 0.73 in controls: p < 0.0001 for both), with people and sun burns history, and education) were observed (Table 5).
over 70 years having 10–15 times higher median values than people Table 6 shows results for quartiles of total PCBs and of the PCB
under 40 years. No association was found with gender. congeners detectable in at least 50% of the study population: no asso-
No significant association was found between CMM and log- ciation with CMM was found, apart from an inverse association for
congener 118, with the lowest risk of CMM in subjects with the highest

22
M. Magoni et al. Environment International 113 (2018) 20–25

Table 3 Table 5
Distribution of total PCB serum levels in melanoma cases and controls expressed in ng/ml Association of melanoma with PCB serum levels (log-transformed), phenotypic and
and in ng/g lipid. constitutional characteristics, history of melanoma and other skin cancers, sun exposure
and sensitivity, education, age and gender, odds ratio (OR) estimates by multiple logistic
Congeners Cases Controls regression analyses.

Total PCBs (ng/ml) Independent variable OR 95% CI p for trend

Mean (SD) 3.42 (3.18) 3.53 (3.45)
25th centile 1.14 0.96 Total PCB (log transformed) 0.94 0.70 1.26 0.671
Median 2.46 2.68 Skin colour 3.00 1.91 4.73 < 0.0001
75th centile 4.55 4.96 Hair coloura 1.48 1.07 2.06 0.018
95th centile 10.12 9.97 History of melanoma 11.36 1.36 94.67 0.025
Range Not detectable-15.86 Not detectable-22.19 History of non-melanoma skin cancer 7.20 2.63 19.68 < 0.0001
Total PCBs (ng/g lipid) Cumulative lifetime sun exposure 2.56 1.35 4.85 0.004
Mean (SD) 521 (481) 520 (501) Sun burns history 1.82 1.15 2.90 0.011
25th centile 169 160 Educationb 1.45 1.03 2.05 0.032
Median 368 371 Age (continuous) 1.01 0.98 1.03 0.528
75th centile 670 728 gender (F vs M) 0.94 0.59 1.52 0.805
95th centile 1542 1449
Range Not detectable −2380 Not detectable −3090 a
Three levels with dark hair as reference.
b
Three levels with “ < High school” as reference.
Mann-Whitney test: p = 0.9.
Since many subjects resided outside the town, we also restricted the
Table 4 analysis to residents in Brescia town and in those residents in Brescia
PCB congeners distribution in melanoma cases and controls: number and percentage of
province: the results showed no substantial differences compared to the
subjects with detectable serum levels and mean values (ng/ml) for each congener.
main analysis. The median age of the study population was 55.67 years
Congeners Cases Controls and two separate analyses were conducted in younger (103 cases and
102 controls) and older (102 cases and 103 controls) subjects. The re-
With detectable Mean With detectable Mean sults are reported in Supplementary Table 7: no association with CMM
levels value levels value
was found for total and any PCB congeners in both groups.
N° % (ng/ml) N° % (ng/ml) Secondary analyses were conducted according to gender, presence
or absence of at least one of the UV-related risk factors for CMM, and
28 0 0.0% 0.000 2 1.0% 0.001 histology (excluding lentigo maligna and nodular melanomas): all these
31 0 0.0% 0.000 0 0.0% 0.000
subgroup analyses showed results similar to those obtained for all
52 0 0.0% 0.000 0 0.0% 0.000
74 53 25.9% 0.023 54 26.3% 0.026 subjects (data not reported).
77 0 0.0% 0.000 0 0.0% 0.000
81 0 0.0% 0.000 0 0.0% 0.000
99 53 25.9% 0.026 63 30.7% 0.031 4. Discussion
101 0 0.0% 0.000 0 0.0% 0.000
105 1 0.5% 0.001 2 1.0% 0.002 The present study aimed to assess the association of current PCB
114 0 0.0% 0.000 0 0.0% 0.000
118 107 52.2% 0.068 120 58.5% 0.091
serum levels with CMM using a case-control study design. The main
123 0 0.0% 0.000 0 0.0% 0.000 statistical analyses so as the subgroup analyses performed according to
126 0 0.0% 0.000 0 0.0% 0.000 gender, age, CMM histology and presence or absence of UV-related risk
128 0 0.0% 0.000 1 0.5% 0.000 factors showed no association of current serum levels of total PCBs and
138 191 93.2% 0.452 186 90.7% 0.468
single PCB congeners with CMM.
146 89 43.4% 0.043 78 38.0% 0.035
153 200 97.6% 0.819 199 97.1% 0.846 This finding is in general agreement with those of two recent meta-
156 141 68.8% 0.083 139 67.8% 0.090 analyses that have not found definite evidence of association between
157 6 2.9% 0.002 10 4.9% 0.003 PCB exposure and the risk of CMM (Boffetta et al., 2016; Zani et al.,
167 22 10.7% 0.009 25 12.2% 0.011 2017), but does not support the conclusions of the International Agency
169 0 0.0% 0.000 0 0.0% 0.000
170 165 80.5% 0.268 154 75.1% 0.274
for Research on Cancer Working Group (IARC, 2016) that has classified
172 39 19.0% 0.015 42 20.5% 0.017 PCBs as carcinogenic to humans based on sufficient evidence of an in-
177 3 1.5% 0.002 3 1.5% 0.002 creased risk of cutaneous malignant melanoma. However, the conclu-
180 198 96.6% 0.984 195 95.1% 0.998 sions of the Working Group of the evaluation of epidemiologic studies
183 77 37.6% 0.041 85 41.5% 0.041
in humans were based on occupational cohort studies without in-
187 155 75.6% 0.135 149 72.7% 0.129
189 2 1.0% 0.001 5 2.4% 0.001 dividual measures of PCB exposure, all except one showing increased
194 125 61.0% 0.205 126 61.5% 0.215 risks of melanoma for PCB exposure, and two case-controls studies, one
201 125 61.0% 0.099 124 60.5% 0.093 of which was based on biologic measures of PCB exposure (Gallagher
206 2 1.0% 0.001 5 2.4% 0.003 et al., 2011; Behrens et al., 2010).
209 48 23.4% 0.038 43 21.0% 0.047
196 + 203 101 49.3% 0.106 98 47.8% 0.106
The study by Gallagher et al. (2011) was considered preliminary by
the authors, because they enrolled 80 cases and 310 controls from two
separate case-control studies, with some concerns regarding compar-
levels of the congener (OR = 0.43; CI95% = 0.21–0.84 and p for ability. Furthermore, the study by Gallagher et al. (2011) found that
trend = 0.02). This paradoxical finding was also observed for total PCBs were strongly associated with pesticides, some of which were also
PCBs and congeners 153, 156, 187 and 201, although all the confidence associated with CMM, making difficult to understand if the increased
intervals were wide. risk of CMM was due to PCBs, pesticides or a combination of both.
The computation of the odds ratio for melanoma according to Since PCBs are a mixture of several congeners, which have different
quartiles was also performed using lipid-adjusted PCB serum levels: no activities and possible effects to living organisms, the type and sources
differences were found compared to the analysis by volumetric mea- of PCB exposure may also play a relevant role in their carcinogenic
sures (Supplementary Table 6.b). capacity. We searched for several PCB congeners but some of them were
not present in any subject. However, the PCB congeners found in most

23
M. Magoni et al. Environment International 113 (2018) 20–25

Table 6 occupation in agriculture and in a chemical plant were observed, but

Odds ratios (ORs) for melanoma according to quartile distribution of serum levels of total the 95% confidence intervals were wide, probably because of the small
PCB and PCB congeners by logistic regression analysis.
number of subjects with occupational exposure. Similarly, the limited
Quartiles (ng/ml) Cases Controls ORa 95% CI p for trend number of subjects who had resided in the highest and intermediate
polluted areas included in the present investigation did not allow us to
N N assess if having resided in the most polluted areas of the town was
associated with risk of CMM.
Total PCB summed
≤1.045 47 55 1.00 0.44 We have no explanation for the apparent paradoxical finding of
1.045–2.544 58 44 1.42 0.71 2.88 lower ORs for the highest quartile of PCB congener 118 and others
2.545–4.789 54 51 1.18 0.54 2.60 congeners. Our hypothesis is that this finding is due to chance, when
> 4.789 46 55 0.81 0.34 1.96 considering the relatively high number of statistical analyses per-
PCB 118
formed. However, it is noteworthy that also a recent case-control study
=0.000 98 85 1.00 0.02
> 0.000–0.059 10 12 0.54 0.18 1.58 on PCB plasma levels and non-Hodgkin lymphoma showed no asso-
0.060–0.109 53 50 0.70 0.38 1.30 ciation of total PCB levels with non-Hodgkin lymphoma but significant
> 0.109 44 58 0.43 0.21 0.84 inverse trends for some PCB congeners (Kelly et al., 2017). In the
PCB 138
present study, all the most commonly found PCB congeners were more
< 0.188 50 52 1.00 0.11
0.188–0.340 60 43 1.21 0.62 2.38 chlorinated than the congener 138. This finding, similarly to what ob-
0.341–0.614 46 57 0.56 0.26 1.19 served in a recent general population survey in Brescia province
> 0.614 49 53 0.64 0.28 1.45 (Magoni et al., 2016), is not surprising, because most human intake of
PCB 153 these chemicals occurred in the past: more chlorinated congeners are
< 0.332 50 52 1.00 0.16
more stable, have longer half-life and are therefore more persistent in
0.332–0.618 56 47 1.13 0.57 2.26
0.619–1.132 52 51 0.76 0.35 1.67 human body than less chlorinated congeners.
> 1.132 47 55 0.63 0.27 1.48 The present investigation confirms the role of exposure to ultra-
PCB 156 violet radiation as well as constitutional risk factors in the pathogenesis
< 0.000 64 66 1.00 0.47
of CMM. People with fair skin had about 3 times higher risk than people
0.000–0.076 43 32 1.26 0.63 2.53
0.077–0.129 53 50 1.04 0.52 2.11
with dark skin. Concerning hair colour, the blond/red phenotype had
≥0.129 45 57 0.77 0.35 1.70 the highest risk and light brown an intermediate risk. In addition, the
PCB 170 study confirms the role of sun exposure and sunburn, so as frequent use
< 0.111 47 55 1.00 0.74 of sunbed and history of melanoma and non-melanoma skin cancer, as
0.111–0.214 56 47 1.52 0.74 3.14
reported by others (International Agency for Research on Cancer
0.215–0.394 54 49 1.12 0.51 2.45
> 0.394 48 54 1.06 0.44 2.58 (IARC), 1992; Chen et al., 2008; Vogel et al., 2014; Koh et al., 1996).
PCB 180 Artificial UV rays are another risk factor for CMM. Furthermore, our
< 0.349 46 56 1.00 0.84 study showed that frequent use of sunbed was associated with CMM in
0.349–0.718 55 48 1.51 0.73 3.11
agreement with Vogel et al. who provided evidence that indoor tanning
0.719–1.385 53 50 1.34 0.61 2.96
> 1.385 51 51 1.26 0.51 3.08
is a risk factor for melanoma even without burns (Vogel et al., 2014).
PCB 187 Also the association of CMM with high education is coherent with
< 0.000 50 56 1.00 0.47 the association with high socioeconomic status already documented
0.000–0.102 54 45 1.29 0.64 2.59 and could be related to more intense recreational sun exposures and to
0.103–0.175 54 49 1.27 0.58 2.79
a better healthcare leading to more frequent dermatologic examinations
> 0.175 47 55 0.79 0.33 1.87
PCB 194 and hence diagnosis of initial CMMs (Idorn and Wulf, 2014; Jiang et al.,
< 0.000 80 79 1.00 0.89 2015).
0.000–0.152 23 23 1.43 0.64 3.20 This study has some strengths and limitations. First, the study was
0.153–0.318 50 53 0.96 0.48 1.90
implemented in one of the most PCB polluted areas in the world, where
> 0.318 52 50 1.05 0.49 2.27
PCB 201
a part of the population had particularly high serum levels of total PCBs
< 0.000 80 81 1.00 0.50 (up to 271 ng/ml in subjects examined in 2001) and where the range of
0.000–0.071 25 19 2.12 0.94 4.78 PCB levels among individuals (0.1–22.16 ng/ml in present study) is
0.073–0.146 50 53 0.93 0.47 1.85 greater than that found in other populations living in non-polluted and
> 0.146 50 52 0.90 0.41 1.96
also polluted areas (Donato et al., 2006; Turrio-Baldassarri et al., 2007;
a
Adjusted for age and sex, skin colour, hair colour, history of melanoma and non-
Turrio-Baldassarri et al., 2009; Magoni et al., 2016). Such wide varia-
melanoma skin cancers, cumulative lifetime sun exposures, sun burns history and edu- tion of PCB exposure would have allowed us a better capacity to assess
cation level. the CMM risk for relatively high, compared to low, serum levels of
PCBs, than studies performed in unexposed populations in North
melanoma cases and healthy controls in our study (PCB 138, 153, 156, America or Europe.
170, 180, 187) are those with the largest diffusion and persistence in Second, both cases and controls were recruited prospectively in the
the environment, that have been detected in most case-control and same hospital, most of which coming from the same population basis, as
cohort studies finding positive associations between PCB exposure and shown by the similar proportions of cases and controls living in the
cancer so far (IARC, 2016). Non-occupational exposure to PCBs mainly town and province of Brescia. The blood samples of both cases and
occurs through the food chain (Malisch and Kotz, 2014), while the main controls were taken in parallel and stored for a short time before un-
routes of occupational exposure are dermal and respiratory. In fact one dergoing laboratory analyses, performed in blind. Finally, the relatively
of the cited meta-analyses (Boffetta et al., 2016) showed different re- high number of cases enrolled (n = 205) provided the study a sufficient
sults according to type of PCB exposure: the summary risk estimates for power to detect statistically significant ORs of 2 or more between PCBs
melanoma were above one for occupational cohorts and below one for exposure and CMM, which corresponds to the lowest 95% confidence
community-based studies. For a skin cancer like melanoma is plausible limit estimated by Gallagher et al. for the highest quartile of total PCBs
that dermal exposition is more dangerous than other ways of exposure. (OR = 6.02; 95% CI: 2.00; 18.17), based on 29 cases and 75 controls
In the present investigation, moderate associations of CMM with exposed (Gallagher et al., 2011). It is noteworthy that the Canadian
population included in the study by Gallagher et al. had lower PCB

24
M. Magoni et al.
serum values than the cases and controls enrolled in our study: the cut-
off for defining the highest quartile of exposure based on PCB dis-
tribution among controls was 213.44 ng/g lipid, about 3-fold lower
than the corresponding cut-off in the Brescia study (4.79 ng/ml, cor-
responding to 701 ng/g lipid). However, should the risk of CMM for
PCB exposure be lower than that hypothesised for the highest PCB ex-
posure level, a greater study would be necessary to estimate it.
The case-control study design is usually considered suitable for as-
sessing the role of environmental exposure for developing human dis-
eases. Indeed, also the cohort and case-control studies performed in the
general population living in unpolluted areas were considered in-
formative by the IARC working group for classifying PCBs as human
carcinogens (IARC, 2016).
The enrolment of hospitalized controls is a well-known matter of
concern in case-control studies. We excluded from controls patients
with cancer, hepatic diseases, auto-immune disease and endocrine
diseases, selecting only patients hospitalized for traumatic causes,
vascular surgery or other minor surgery, diseases that have never been
associated with PCB exposure, therefore the risk of selection bias seems
unlikely. Finally, we measured the serum levels of PCBs at the time of
CMM diagnosis, although past exposure was probably more relevant for
disease occurrence than present exposure. However, various studies
have demonstrated that the serum levels of PCBs and other organo-
chlorines are highly related with the concentrations of these chemicals
in subcutaneous fat, which are a valuable measure of body storage and
therefore reflect cumulative lifetime exposure (IARC, 2016; Rusiecki
et al., 2005; Covaci et al., 2008; Bergonzi et al., 2009; Zani et al., 2013).
Indeed, we found a substantial decline of PCB serum levels in this po-
pulation in the last 10 years, but a high correlation between present and
past values at the individual level (Magoni et al., 2016; Raffetti et al.,
2017).
In conclusion, this case-control study does not support the hypoth-
esis of an association between current plasma levels of PCBs and CMM
development in the general population.
Acknowledgments
The study was implemented with resources of the ATS Brescia and
with a specific fund by Lombardy Region (DG X/141 on the 17/05/
2013).
We appreciated the precious collaboration of the ASST Spedali Civili
di Brescia and we are grateful to the medical doctors, nurses and health
personnel who enrolled the patients and performed the interviews.
A final but biggest thanks to all patients who accepted to partici-
pate.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://
doi.org/10.1016/j.envint.2018.01.018.
References
AIOM & AIRTUM, 2016. I numeri del cancro in Italia. In: Il Pensiero Scientifico Editore.
Stampato in Italia da Ti Printing S.r.l. (in Italian).
Behrens, T., Kaerlev, L., Cree, I., Lutz, J.M., Afonso, N., Eriksson, M., et al., 2010.
Hormonal exposures and the risk of uveal melanoma. Cancer Causes Control 21 (10),
1625–1634.
Bergonzi, R., Specchia, C., Dinolfo, M., Tomasi, C., De Palma, G., Frusca, T., Apostoli, P.,
2009 Aug. Distribution of persistent organochlorine pollutants in maternal and foetal
tissues: data from an Italian polluted urban area. Chemosphere 76 (6), 747–754.
Boffetta, P., Catalani, S., Tomasi, C., Pira, E., Apostoli, P., 2016 Oct 5. Occupational ex-
posure to polychlorinated biphenyls and risk of melanoma: a meta-analysis. Eur. J.
Cancer Prev (Epub ahead of print).
Caini, S., Gandini, S., Sera, F., Raimondi, S., Fargnoli, M.C., Boniol, M., Armstrong, B.K.,
2009 Nov. Meta-analysis of risk factors for cutaneous melanoma according to ana-
tomical site and clinico-pathological variant. Eur. J. Cancer 45 (17), 3054–3063.
Chen, J., Ruczinski, I., Jorgensen, T.J., Yenokyan, G., Yao, Y., Alani, R., Liégeois, N.J.,
Hoffman, S.C., Hoffman-Bolton, J., Strickland, P.T., Helzlsouer, K.J., Alberg, A.J.,
25
Environment International 113 (2018) 20–25
2008 Sep 3. Nonmelanoma skin cancer and risk for subsequent malignancy. J. Natl.
Cancer Inst. 100 (17), 1215–1222.
Covaci, A., Voorspoels, S., Roosens, L., Jacobs, W., Blust, R., Neels, H., 2008.
Polybrominated diphenyl ethers (PBDEs) and polychlorinated biphenyls (PCBs) in
human liver and adipose tissue samples from Belgium. Chemosphere 73 (2),
170–175.
Donato, F., Magoni, M., Bergonzi, R., Scarcella, C., Indelicato, A., Carasi, S., Apostoli, P.,
2006. Exposure to polychlorinated biphenyls in residents near a chemical factory in
Italy: the food chain as main source of contamination. Chemosphere 64, 1562–1572.
Fava, P., Astrua, C., Chiarugi, A., Crocetti, E., Pimpinelli, N., Fargnoli, M.C., Maurichi, A.,
Rubegni, P., Manganoni, A.M., Bottoni, U., et al., 2015. Differences in clin-
icopathological features and distribution of risk factors in Italian melanoma patients.
Dermatology 230 (3), 256–262.
Gallagher, R.P., Macarthur, A.C., Lee, T.K., Weber, J.P., Leblanc, A., Mark Elwood, J.,
Borugian, M., Abanto, Z., Spinelli, J.J., 2011 Apr 15. Plasma levels of polychlorinated
biphenyls and risk of cutaneous malignant melanoma: a preliminary study. Int. J.
Cancer 128 (8), 1872–1880.
Holterhues, C., Hollestein, L.M., Nijsten, T., Koomen, E.R., Nusselder, W., de Vries, E.,
2013. Burden of disease due to cutaneous melanoma has increased in the Netherlands
since 1991. Br. J. Dermatol. 169, 389–397.
IARC, 2016. Polychlorinated biphenyls and polybrominated biphenyls. In: IARC
Monographs on the Evaluation of Carcinogenic Risks to Humans. Volume 107
International Agency for Research on Cancer, Lyon, France.
Idorn, L.W., Wulf, H.C., 2014 Apr. Socioeconomic status and cutaneous malignant mel-
anoma in Northern Europe. Br. J. Dermatol. 170 (4), 787–793.
International Agency for Research on Cancer (IARC), 1992. IARC monographs on the
evaluation of carcinogenic risks to humans. Solar and ultraviolet radiation. IARC
Monogr. Eval. Carcinog. Risks Hum. 55, 1–316.
Jiang, A.J., Rambhatla, P.V., Eide, M.J., 2015 Apr. Socioeconomic and lifestyle factors
and melanoma: a systematic review. Br. J. Dermatol. 172 (4), 885–915.
Karimkhani, C., Green, A.C., Nijsten, T., Weinstock, M.A., Dellavalle, R.P., Naghavi, M.,
Fitzmaurice, C., 2017 Mar 30. The global burden of melanoma: results from the
global burden of disease study 2015. Br. J. Dermatol. http://dx.doi.org/10.1111/bjd.
15510.
Kelly, R.S., Kiviranta, H., Bergdahl, I.A., Palli, D., Johansson, A.S., Botsivali, M., Vineis,
P., Vermeulen, R., Kyrtopoulos, S.A., Chadeau-Hyam, M., 2017 Feb 16.
EnviroGenoMarkers project consortium. Prediagnostic plasma concentrations of or-
ganochlorines and risk of B-cell non-Hodgkin lymphoma in envirogenomarkers: a
nested case-control study. Environ. Health 16 (1), 9.
Koh, H.K., Geller, A.C., Miller, D.R., Grossbart, T.A., Lew, R.A., 1996 Apr. Prevention and
early detection strategies for melanoma and skin cancer. Current status. Arch.
Dermatol. 132 (4), 436–443.
Magoni, M., Donato, F., Speziani, F., Leonardi, L., Orizio, G., Scarcella, C., Gaia, A.,
Apostoli, P., 2016 Oct. Substantial decline of polychlorinated biphenyls serum levels
10 years after public health interventions in a population living near a contaminated
site in Northern Italy. Environ. Int. 95, 69–78.
Malisch, R., Kotz, A., 2014. Dioxins and PCBs in feed and food — review from European
perspective. Sci. Total Environ. 491–492, 2–10.
Ossio, R., Roldán-Marín, R., Martínez-Said, H., Adams, D.J., Robles-Espinoza, C.D., 2017
Jul. Melanoma: a global perspective. Nat. Rev. Cancer 17 (7), 393–394.
Phillips, D.L., Pirkle, J.L., Burse, V.W., Bernert Jr., J.T., Henderson, L.O., Needham, L.L.,
1989 Jul-Aug. Chlorinated hydrocarbon levels in human serum: effects of fasting and
feeding. Arch. Environ. Contam. Toxicol. 18 (4), 495–500.
Raffetti, E., Speziani, F., Donato, F., Leonardi, L., Orizio, G., Scarcella, C., Apostoli, P.,
Magoni, M., 2017 Apr. Temporal trends of polychlorinated biphenyls serum levels in
subjects living in a highly polluted area from 2003 to 2015: a follow-up study. Int. J.
Hyg. Environ. Health 220 (2 Pt B), 461–467.
Rusiecki, J.A., Matthews, A., Sturgeon, S., Sinha, R., Pellizzari, E., Zheng, T., Baris, D.,
2005. A correlation study of organochlorine levels in serum, breast adipose tissue,
and gluteal adipose tissue among breast cancer cases in India. Cancer Epidemiol.
Biomark. Prev. 14 (5), 1113–1124.
Turci, R., Angeleri, F., Minoia, C., 2002. A rapid screening method for routine congener-
specific analysis of polychlorinated biphenyls in human serum by high-resolution gas
chromatography with mass spectrometric detection. Rapid Commun. Mass Spectrom.
16 (20), 1957–1964.
Turrio-Baldassarri, L., Abate, V., Alivernini, S., Battistelli, C.L., Carasi, S., Casella, M.,
Iacovella, N., Iamiceli, A.L., Indelicato, A., Scarcella, C., La Rocca, C., 2007. A study
on PCB, PCDD/PCDF industrial contamination in a mixed urban-agricultural area
significantly affecting the food chain and the human exposure. Part I: soil and feed.
Chemosphere 67, 1822–1830.
Turrio-Baldassarri, L., Alivernini, S., Carasi, S., Casella, M., Fuselli, S., Iacovella, N.,
Iamiceli, A.L., La Rocca, C., Scarcella, C., Battistelli, C.L., 2009. PCB, PCDD and PCDF
contamination of food of animal origin as the effect of soil pollution and the cause of
human exposure in Brescia. Chemosphere 76, 278–285.
Vogel, R.I., Ahmed, R.L., Nelson, H.H., Berwick, M., Weinstock, M.A., Lazovich, D.A.,
2014. Exposure to indoor tanning without burning and melanoma risk by sunburn
history. J. Natl. Cancer Inst. 106 (7).
Zani, C., Ceretti, E., Covolo, L., Donato, F., 2017 Sep. Do polychlorinated biphenyls cause
cancer? A systematic review and meta-analysis of epidemiological studies on risk of
cutaneous melanoma and non-Hodgkin lymphoma. Chemosphere 183, 97–106.
Zani, C., Gelatti, U., Donato, F., Capelli, M., Portolani, N., Bergonzi, R., Apostoli, P., 2013
Apr. Polychlorinated biphenyls in serum, liver and adipose tissue of subjects with
hepatocellular carcinoma living in a highly polluted area. Chemosphere 91 (2),
194–199.