Differences in tumor-associated protein levels among middle-age Flemish women in association with area of residence and exposure to pollutants.

Table 1 shows internal exposure data (concentrations in blood, serum, and urine). We found that residents from Hoboken had significantly higher blood lead, serum zinc, and serum PCB levels than residents from other municipalities. Also, compared with residents from the other municipalities, residents from Peer had significantly higher serum cadmium levels, higher dioxin-like activity in blood fat (significantly so compared with Wilrijk), and higher 1-hydroxypyrene levels (significantly so compared with Hoboken). Crude data are summarized in Table 2 . Depending on the type of TAP, between 91.4% and 100% of participants showed clinically normal values, except with respect to BTA, with only 80.7% normal values. ANCOVA revealed that place of residence was significantly associated with differences in c-erbB-2, BTA, and PDGF-BB levels and, after correction for multiple comparisons, with differences in BTA and PDGF-BB levels. Using a Fisher PLSD test, we found significantly higher TAP levels for both industrial suburbs but not for the rural municipality of Peer. Women residing in Hoboken had the highest levels of c-erbB-2, TPS, and hPLAP. For c-erbB-2, the difference was significant for both Peer ( p = 0.0099) and Wilrijk ( p = 0.026); for TPS ( p = 0.049) and hPLAP ( p = 0.015), the difference was significant only for Wilrijk. Although residents from Hoboken had lower PDGF-BB levels than those from Wilrijk, their levels were significantly higher than those of residents from Peer ( p = 0.0021). Residents of Wilrijk showed significantly higher levels of BTA than those of Hoboken ( p = 0.0006) and those of Peer ( p < 0.0001) and significantly higher PDGF-BB levels than those of Hoboken ( p = 0.023) and those of Peer ( p < 0.0001). Adjustment for potentially confounding personal and lifestyle parameters (age; level of education; smoking status; total number of pregnancies; total number of months of breast-feeding; BMI; serum levels of selenium, vitamin A, vitamin E, total fat, total cholesterol, and triglycerides; and all available parameters related to food intake except those related to consumption of locally grown food, a total of 21 parameters) had no substantial effect on the association of TAP levels with area of residence ( Table 3 ). Results of ANCOVA were similar, except that area of residence was now significantly associated with differences in hPLAP levels but no longer with differences in c-erbB-2 levels. Using the Fisher PLSD test, the observed differences between residents of the different area’s remained significant. The I tap , both without correction (data not shown) and after correction for place of residence, all available personal and lifestyle parameters, and all available food-intake parameters, was higher in residents of Hoboken (adjusted value = 0.074) than in residents of Wilrijk (−0.020) or Peer (−0.040), and the difference with Peer was close to statistical significance ( p = 0.058 after correction for confounding). We used the levels of TAPs as dependent variables and all 29 available personal, lifestyle, and internal exposure parameters as independent variables in multiple regression analysis ( Table 4 ). We found a positive correlation of blood lead with serum levels of anti-p53, CEA, and TPS and with I tap ; a positive correlation of serum zinc with c-erbB-2 and with TPS; a positive correlation of dioxin-like activity in serum with CA 125; and a negative correlation of serum copper with IGF-I and c-erbB-2. The I ex showed a positive association with serum levels of CEA and of anti-p53. If the total study population was divided into two classes according to the I ex [women with an I ex higher than the median versus the others, both without correction (data not shown)] and after correction for place of residence, all available personal and lifestyle parameters, and all available food-intake parameters, the class with higher I ex levels had significantly higher values for I tap (adjusted mean, +0.075 vs. −0.078; p = 0.010). FLEHS entailed measurements of HPRT mf in peripheral lymphocytes for a subset of the women who participated in the study ( van Larebeke et al. 2004 ). For 11 of the 12 TAPs and for I tap , we found a nonsignificant positive association with HPRT m f , whereas for PDGF-BB we found a non-significant negative association (chi-square, p = 0.004). In multiple regression with all available personal, lifestyle parameters, and internal exposure parameters, CEA showed a significant positive correlation with HPRT mf (squared partial correlation = 0.108, p = 0.018).

Although possible, it is unlikely that a selection bias occurred regarding participation in Wilrijk, Hoboken, and Peer. The quite low response rate may be because only one letter was sent and participants were required to give blood and urine; only 12 of 130 respondents from Wilrijk or Hoboken and only 14 of 125 from Peer actually refused to participate. Concerning a series of items on which respondents provided information (described in “Materials and Methods”), we found no significant differences between participants and persons who refused or could not participate. The study participants generally considered themselves to be in good health (data not shown). We considered it preferable not to exclude persons presenting with complaints or with a disease, in order to avoid introducing a bias by eliminating persons whose illness or complaints might be influenced by environmental factors. Because only a limited percentage of cancer cases can be ascribed to hereditary factors ( Czene et al. 2002 ; Verkasalo et al. 1999 ), because the populations in the areas under study are genetically probably quite similar (distances are small, the landscape permits easy traveling, no historical divisions, all participants spoke the same language), and because selection bias is unlikely, we consider it improbable that the observed differences in TAP levels as a function of area of residence or of internal exposure are due to hereditary factors. Residents from Hoboken had significantly higher blood lead and serum zinc levels than residents from other municipalities, which was to be expected in view of the presence of a large nonferrous smelter; they also showed significantly higher serum PCB levels, which was less expected. Surprisingly, however, compared with residents from the other municipalities, residents of the rural municipality of Peer had higher serum cadmium levels, higher dioxin-like activity in blood fat (significantly so compared with Wilrijk), and higher 1-hydroxypyrene levels (significantly so compared with Hoboken). We are not aware of a reliable explanation of these high internal exposure levels in residents of Peer; some of them may have originated from previous exposures (cadmium, dioxin-like activity). Also it remains possible that both intensive agriculture (pesticides, fertilizers) and some industrial activities at a distance of about 15 km were involved. In 2000–2001, episodes of air pollution with polycyclic aromatic hydrocarbons and high mutagenic activity were observed in Peer ( Du Four et al. 2004 ). These data and also the data from an ongoing Flemish biomonitoring program involving other rural areas ( Milieu en Gezondheid 2006 ) indicate that pollution is widespread in Flanders and suggest that exposure to pollutants may be quite homogeneous in some developed Western nations. Whether differences in levels of TAPs, within clinically normal ranges, are of importance in terms of health risks is not known with certainty. There are, however, a series of observations suggesting that exposure to carcinogenic or tumor-promoting agents leads to a slight increase in these levels, including in those of TAPs examined in this study ( Koshida et al. 1990 ; Krajewska et al. 1998 ; Luo et al. 1999 ; Lutz et al. 1997 ; Wong et al. 2002 ). Also, in a number of studies, higher levels of TAPs, still in the clinically normal range, were associated with a higher risk of cancer. Such an association was clearly demonstrated for prostate-specific antigen and risk of prostate cancer ( Ito et al. 2003 , 2005 ) and has also been observed for the TAPs used in this study ( Bohlke et al. 1998 ; Brandt-Rauf and Pincus 1998 ; Giovannuci 1999 ; Hankinson et al. 1998 ; Husgafvel-Pursiainen et al. 1997 ; Jacobs et al. 1996 ; Jeyarajah et al. 1999 ; Kobayashi and Kawakubo 1994 ; Oka et al. 1994 ; Wolk et al. 1998 ). The trend toward a positive association between levels of TAPs and HPRT mf is consistent with the hypothesis that these parameters are biomarkers of exposure to carcinogenic agents or of cancer risk. For the whole study population, certain parameters of internal exposure known or suspected to be associated with an increased risk of cancer showed a positive association with the serum or urinary levels of some TAPs or with the I tap . There is ample experimental evidence that inorganic lead is carcinogenic ( Silbergeld et al. 2000 ), and it is classified by the International Agency for Research on Cancer (IARC) as group 2B, possibly carcinogenic to humans ( IARC 1987 ). In the general U.S. population, a dose–response relationship between quartile of blood lead and all cancer mortality showed relative risks of 1.24, 1.33, and 1.5 for the second, third, and fourth quartiles, respectively, compared with the first quartile, but this relationship was not significant ( p -trend = 0.16) ( Jemal et al. 2002 ). Workers exposed to lead have been reported to have an increased incidence of lung cancer ( Anttila et al. 1995 ). In our study on women without occupational exposure and with blood lead levels within “normal” limits, blood lead showed a significant positive correlation with serum levels of anti-p53, CEA, and TPS and with I tap , this in multiple regression analyses with all available personal and lifestyle parameters and with all available parameters of internal exposure. In the same cohort, blood lead also showed a positive correlation with HPRT mf ( van Larebeke et al. 2004 ). In the vicinity of a nonferrous smelter in Sweden, an increased incidence of lung cancer in men has been found ( Besso et al. 2003 ). The findings presented here and in our previous report ( van Larebeke et al. 2004 ), taken together, suggest that lead could be an important carcinogen for at least part of the population included in our study. Dioxins are human carcinogens associated with an increase in total cancer incidence in humans ( IARC 1997 ) and with endometriosis in apes ( Rier et al. 1993 , 1995 ) and in women ( Koninckx et al. 1994 ; Mayani et al. 1997 ); we found a positive association between dioxin-like activity in serum and serum CA 125. Serum CA 125 concentrations are often higher in women with endometriosis ( Somigliana et al. 2004 ). Zinc deficiency may enhance carcinogenesis, and zinc is under consideration as a chemopreventive agent ( Riboli et al. 1996 ). However, zinc also has a genotoxic potential; it has induced cell transformation in vitro , and an oversupply of zinc may increase the risk of prostate cancer ( European Commission 2003 ). In the present study, serum zinc level showed a positive correlation with c-erbB-2 and TPS. The relationship between copper and the risk of cancer is complex ( Theophanides et al. 2002 ). In our study, serum copper showed a weak but significant positive correlation with CA 125, but a quite strong negative correlation with c-erbB-2 and IGF-I levels. The I ex , integrating parameters related to lead, cadmium, PCBs, dioxin-like activity, and 1-hydroxypyrene, was calculated for each participant as described in “Materials and Methods.” Although such an index does not take into account possible interactions, it might give a somewhat more integrated view on the internal exposure of an individual. The I ex showed a positive association with I tap and with serum levels of anti-p53 and CEA. For 5 of the 12 TAPs assessed in this study (i.e., BTA, c-erbB-2, PDGF-BB, TPS, and hPLAP), we observed significant differences in serum levels between residents of the three municipalities. After correction for all 21 available personal or lifestyle parameters, the differences remained significant. For each of these five TAPs, we observed the highest levels in one of the industrial suburbs, although, compared with residents from the other municipalities, residents of the rural municipality of Peer had higher serum cadmium levels, higher dioxin-like activity in blood fat, and higher urinary 1-hydroxypyrene levels. Even when correction through ANCOVA analysis included all available internal exposure parameters in addition to all 21 available personal or lifestyle parameters, these differences remained significant (data not shown). This suggests that the higher levels of some TAPs in the industrial suburbs might, at least in part, be caused by exposures other than those measured in this study, or by other unknown factors somehow associated with residence in these suburbs.

Although we found levels of internal exposure to pollutants to be quite homogeneous in Flanders, and although we observed the highest levels of some pollutants in the rural municipality of Peer, we found significantly higher levels of TAPs only in the industrial suburbs. This suggests that, in biomonitoring, effect biomarkers might indeed be important in addition to biomarkers of internal exposure. Also, this study provides some evidence indicating that levels of internal exposure such as those present in Flanders, lead in particular, are indeed associated with biologic effects and that even relatively small differences in these levels are associated with observable differences in such effects. Our observations suggest that more research into the use of TAPs to assess, in an integrated manner, the biologic effects of exposure to carcinogenic or tumor-promoting agents is of interest.

Hoboken and Wilrijk are located 11–13 km southeast of the chemical and petrochemical industry established in the seaport of Antwerp. Hoboken harbors a large nonferrous smelter (the cause of important lead pollution in the past), other metallurgic industries, a large printer, and plants producing electronic equipment. The mean lead concentration of suspended particles during 1997 in Hoboken was between 0.08 and 1.35 μg/m 3 , depending on the location. Wilrijk contains metallurgic industry, plastic industry, a printer, a crematorium, and two waste incinerators with annual turnovers of 23,000 and 110,000 tons; these incinerators were shut down in November 1997 because of dioxin emissions exceeding the acceptable level [0.1 ng 2,3,7,8-tetrachlorodibenzodioxin (TCDD) toxic equivalents (TEQ)/m 3 ] by 20-fold. Both Hoboken and Wilrijk are traversed by a highway traveled by > 80,000 vehicles a day. Peer (14,622 inhabitants), an area of intensive agriculture, is situated in a rural area 70 km east of Antwerp. Although there is a military airfield about 5 km north of the municipality and some polluting industries about 15 km away, measured environmental pollution was lower in Peer than in many other places in Flanders. Sources of and data on environmental pollution of these municipalities were described previously ( van Larebeke et al. 2004 ). We chose middle-age women because they spent most of their lives residing in the same region and working at home. We measured TAP levels in serum samples from 200 women from Hoboken ( n = 61), Wilrijk ( n = 39), and Peer ( n = 100) recruited between June and September 1999. To obtain the planned 200 women for the study, we used the following selection strategy. We contacted (only one attempt) all female residents 50–65 years of age (2,898 women) by letter containing a questionnaire on age, duration of residence, smoking, occupation and place of occupation, number of children, consumption of vegetables or fruit from the family’s garden, and vegetarianism. In Hoboken/Wilrijk and Peer, 40.1% and 30.8% responded, respectively. We used the following two inclusion criteria: minimal residence of 10 years in the study area and employment in the town of residence or at home for at least 10 years. We excluded smokers and persons who stopped smoking < 10 years before the start of the study as well as persons who were ever employed in jobs with specific risks of exposure. Having been diagnosed with cancer or another disease was not an exclusion criterion. We ranked the remaining 685 women by duration of residence and smoking history, with never-smokers given priority and ex-smokers prioritized by time elapsed since cessation of smoking. In order to reach 100 participants per region, we used telephone calls to invite 130 women from Wilrijk or Hoboken and 125 women from Peer to participate in the study. Of the 130 residents from Wilrijk or Hoboken, 12 refused to participate and 18 others did not participate for other reasons. Of the 125 residents from Peer, 14 refused to participate and 11 others did not participate for other reasons. We noted no significant differences between participants and nonparticipants in terms of their responses to items in the questionnaire. All participants gave their written informed consent. Throughout the study, a communication plan (with the participants and local authorities) was in force. Each participant was offered the possibility of having an interview with a physician-researcher. We complied with all applicable requirements. The study was approved by the institutional review board of the University of Antwerp. We measured height and body weight, calculated body mass index (BMI), and collected approximately 200 mL urine and 40 mL blood from each subject. Blood samples were collected in polyethylene tubes. Immediately after sampling, serum was separated. Split samples of serum, plasma, whole blood, and urine were stored at 4°C or immediately deep frozen. All laboratory analyses were performed blindly in specialized laboratories that met national and international quality-control standards. For each participant, we performed a series of routine hematologic tests and determined serum concentrations of total fat, triglycerides, cholesterol, selenium, vitamin A, vitamin E, zinc, and copper. We measured the following biomarkers of exposure in serum, blood, or urine as described previously ( van Larebeke et al. 2004 ): dioxin-like activity, expressed in picograms TEQ per gram blood fat; PCB-138, PCB-153, and PCB-180 (indicator PCBs), expressed in nanograms per gram blood fat [we calculated the sum of the concentrations of these PCBs (∑3PCBs) to use as a parameter in statistical analysis]; cadmium in urine and blood; lead in blood; and 1-hydroxypyrene in urine. TAP levels in urine [bladder tumor antigen (BTA)] or serum (all other parameters) were measured by Interlab (Evergem, Belgium), a laboratory recognized by the Belgian Ministry of Health as a reference laboratory for tumor markers. We measured the following TAPs: tissue polypeptide–specific antigen (TPS; TPS IRMA kit; BEKI Diagnostics, Bromma, Sweden); human placental alkaline phosphatase (hPLAP; Innotest hPLAP kit; Innogenetics, Ghent, Belgium); c-erbB-2 ectodomain (c-erbB-2; Human neu Quantitative ELISA kit; Oncogene Research Products, Cambridge, MA, USA); insulin-like growth factor-I (IGF-I; IGF-I ELISA, DSL-10-600 kit; DSL, Webster, TX, USA); platelet-derived growth factor BB (PDGF-BB; Quantikine Human PDGF-BB kit; R&D Systems, Minneapolis, MN, USA); carcino-embryonic antigen (CEA; IRMA-coat CEA kit; Byk-Sangtec Diagnostica, Dietzenbach, Germany); cancer antigen (CA) 15-3 (IRMA-count BR-MA kit; DPC, Los Angeles CA, USA); CA 125 (IRMA-mat CA 125 II kit; DPC); α-fetoprotein (AFP; COAT-A-COUNT AFP IRMA kit; DPC); BTA (BTA TRAK kit; Bard Diagnostic Sciences, Redmond, WA, USA); anti-p53 (Anti-p53 ELISA kit; PharmaCell, Paris, France); soluble interleukin-2 receptor (sIL-2R; Milenia IL2R kit; DPC). Except for BTA, urinary measurements were standardized to 1 mmol creatinine. Coefficients of variation of the TAPs in the clinically normal range in our laboratory (expressed as percentage) were 5.0% for anti-p53, 5.0% for c-erbB-2, 8.7% for TPS, 7.1% for AFP, 9.8% for CEA, 6.5% for CA 15-3, 11.0% for CA 125, not available for hPLAP, 4.7% for BTA, 7.1% for IGF-I, 8.5% for sIL-2R, and 7.7% for PDGF-BB. All participants filled out a self-reporting questionnaire on education, residence history, food intake, smoking, alcohol habits, and health status. We classified education by the highest education received: primary school or the first 3 years of secondary school, complete secondary school, and higher education. We considered participants ex-smokers if they had smoked at least one cigarette per day for at least 1 year; participants who smoked less than this were classified as never-smokers. We used the following food-intake parameters, calculated from the questionnaires on dietary habits over the last year: ethanol in grams per day, total daily intake of animal fat in grams, frequency of consumption of dairy products per day, total daily intake of calcium in milligrams, and number of different types of locally grown food items regularly consumed. Also, we recorded frequencies of monthly consumption of meat (smoked, salted, grilled, roasted, baked, or organ meat); baked, roasted, or grilled meat; fish; smoked or salted meat or fish; fish, fish organ meat, mussels or shrimp; and vegetables from own garden. In addition, we calculated the total number of pregnancies and the total number of weeks of breast-feeding. We made no attempt to collect information on occurrence of cancer in family members. For each TAP and for each biomarker of exposure we calculated a standard or z -score for each individual by dividing the difference between the value for that individual and the mean value for the entire subject population by the standard deviation for the entire subject population. We calculated for each individual an index of mean TAP level ( I tap ), defined as the arithmetic mean of the z -scores for each type of TAP [ I tap = ( z TPS + z hPLAP + z c-erbB-2 + z IGF-I + z PDGF-BB + z CEA + z CA 15-3 + z CA 125 + z AFP + z BTA + z anti-p53 + z sIL-2R )/12]. For each subject we calculated an index of internal exposure ( I ex ), defined as the arithmetic mean of the z -scores for blood lead concentration; ∑3PCBs; dioxin-like activity in serum measured by the chemically activated luciferase expression (CALUX) bioassay; urinary excretion of cadmium per millimole creatinine; and urinary excretion of 1-hydroxypyrene per millimole creatinine [ I ex = ( z blood lead + z ∑3PCBs + z dioxin-like activity + z urinary cadmium + z urinary 1-hydroxypyrene )/5]. We performed stepwise regression, multiple regression, analysis of covariance (ANCOVA), and accompanying Fisher’s probable least-squares difference (PLSD) post hoc tests, chi-square, non-parametric Spearman rank correlation, Mann-Whitney U , and Kruskal-Wallis tests using Statview (version 5.0.1; SAS Institute, Cary, NC, USA) or Statistica (Statsoft, Tulsa, OK, USA) programs. We transformed parameters that did not show a Gaussian distribution for use as dependent variables in an ANCOVA analysis; we used the natural logarithms of the serum concentrations of anti-p53, TPS, AFP, CA 15-3, CA 125, BTA, IGF-I, and sIL-2R and the square root of the concentrations of c-erbB-2, CEA, hPLAP, and PDGF-BB. To select parameters associated with differences in TAP levels for use in ANCOVA analysis or in multiple regression, we performed a forward stepwise regression of personal characteristics and lifestyle parameters (except those related to locally grown food) for ANCOVA analysis and personal, lifestyle, and internal exposure parameters for use in multiple regression, with each TAP as a dependent variable using F = 4 for inclusion and F = 3.9 for exclusion. Adjustment through ANCOVA analysis resulted in geometric mean values. ANCOVA analyses were limited to main effects only. In accordance with the views formulated by Rothman (1986) , we present p -values without adjustment for multiple testing. However, to address concerns regarding multiple testing, we also provide the p -values required to maintain an overall type I error bound of 0.05 after adjustment for multiple comparisons according to Holm ( Aickin and Gensler 1996 ). In relation to ANCOVA and multiple regression analyses, we calculated p -values taking into account that 12 tests—one for each TAP—were performed. Results from multiple regression include the squared semipartial correlation, which is a measure of the correlation between two variables that remains after controlling for the effects of the other predictor variables and indicates the proportion of variance accounted for by the predictor X 1 relative to the total variance of Y .