- Pieter van 't Veer, assistant professor of cancer epidemiologya,
- Irene E Lobbezoo, research scientista,
- José M Martín-Moreno, directorb,
- Eliseo Guallar, head of department of epidemiology and biostatisticsb,
- Jorge Gómez-Aracena, associate professorc,
- Frans J Kok, head of departmenta,
- Alwine F M Kardinaal, product manager, analytical epidemiologyd,
- Lenore Kohlmeier, professore,
- Blaise C Martin, head of division of epidemiology of non-communicable diseasesf,
- John J Strain, professor of human nutritiong,
- Michael Thamm, research scientisth,
- Piet van Zoonen, deputy head of laboratoryi,
- Bert A Baumann, head of pesticide departmenti,
- Jussi K Huttunen, director generalj
- a Department of Human Nutrition and Epidemiology, PO Box 8129, 6700 EV Wageningen, Netherlands
- b National School of Public Health, E-28029 Madrid, Spain
- c Department of Preventive Medicine, University of Malaga, Teatinos s/n, E-29071 Malaga, Spain
- d Department of Epidemiology, TNO Nutrition and Food Research Institute, PO Box 360, 3700 AJ Zeist, Netherlands
- e Department of Epidemiology and Nutrition, University of North Carolina, Chapel Hill, NC 27599-7400, USA
- f Institute of Social and Preventive Medicine, University of Zurich, 8006 Zurich, Switzerland
- g Human Nutrition Research Group, Department of Biological and Biomedical Sciences, University of Ulster, Coleraine, Londonderry BT52 ISA
- h Department for Health Risks and Prevention, Robert Koch Institute, D-12101 Berlin, Germany
- i Laboratory of Organic and Analytical Chemistry, National Institute of Public Health and Environmental Protection, PO Box 1, 3720 BA Bilthoven, Netherlands
- j National Public Health Institute, FIN-00300 Helsinki, Finland
- Correspondence and reprint requests to: Dr van 't Veer
- Accepted 25 April 1997
Objective: To examine any possible links between exposure to DDE (1,1-dichloro-2,2-bis (p -chlorophenyl)ethylene), the persistent metabolite of the pesticide dicophane (DDT), and breast cancer.
Design: Multicentre study of exposure to DDE by measurement of adipose tissue aspirated from the buttocks. Laboratory measurements were conducted in a single laboratory. Additional data on risk factors for breast cancer were obtained by standard questionnaires.
Setting: Centres in Germany, the Netherlands, Northern Ireland, Switzerland, and Spain.
Subjects: 265 postmenopausal women with breast cancer and 341 controls matched for age and centre.
Main outcome measure: Adipose DDE concentrations.
Results: Women with breast cancer had adipose DDE concentrations 9.2% lower than control women. No increased risk of breast cancer was found at higher concentrations. The odds ratio of breast cancer, adjusted for age and centre, for the highest versus the lowest fourth of DDE distribution was 0.73 (95% confidence interval 0.44 to 1.21) and decreased to 0.48 (0.25 to 0.95; P for trend=0.02) after adjustment for body mass index, age at first birth, and current alcohol drinking. Adjustment for other risk factors did not materially affect these estimates.
Conclusions: The lower DDE concentrations observed among the women with breast cancer may be secondary to disease inception. This study does not support the hypothesis that DDE increases risk of breast cancer in postmenopausal women in Europe.
Organochlorines such as polychlorinated biphenyls and DDT may increase the risk of breast cancer in women
DDE concentrations among the women with cancer were lower than among the controls, and there was an inverse risk gradient with higher DDE concentrations which remained significant after adjustment for risk factors for breast cancer
These results are clearly incompatible with an increased risk of breast cancer at increased concentrations of DDE, although associations with other organochlorines cannot be excluded
Environmental oestrogenic compounds, such as the organochlorines DDT (dicophane (2,2-bis (p -chlorophenyl)-1,1,1-trichloroethane)), polychlorinated biphenyls, and dioxins, have been linked to altered sexual development in various species, to a decrease in semen quality, and to an increased risk of breast cancer in women.1 2 3 Their weak oestrogenic effects may result from altered metabolism and competition for binding to cytosolic and nuclear receptors of steroid hormones.4 5
Because of their lipophilicity and long half lives, organochlorines accumulate in the food chain.6 Typically, consumption of fish, meat, and milk is held responsible for the age related increase in concentrations of DDT and its even more persistent metabolite DDE (1,1-dichloro-2,2-bis (p -chlorophenyl)ethylene) in adipose tissue.7 Since the ban on DDT in Western countries 20-25 years ago, DDE concentrations in adipose tissue and mothers' milk dropped in successive cohorts but less so in the generations exposed before the ban.6 8 9
Apart from some early reports,10 11 12 one nested case-control study in the United States showed a fourfold increased risk of breast cancer at high plasma DDE concentrations,13 which was not confirmed by a larger prospective study, also from the United States.14 We determined the association between DDE and breast cancer among 606 women from five European countries on the basis of DDE measured in adipose tissue aspirated from the buttocks.
Subjects and methods
This investigation is an extension of the European study on antioxidants, myocardial infarction, and cancer of the breast (EURAMIC). This multicentre case-control study included 347 women with breast cancer from Germany, the Netherlands, Northern Ireland, Switzerland, and Spain as well as 374 population and hospital controls.15 16 In short, apparently healthy postmenopausal women, aged 50 to 74 years, with a stable dietary pattern and no history of breast cancer were eligible. To achieve efficiency in study conduct and data analysis we aimed at a similar number of cases and controls by study centre and age using group matching. Incident cases with histologically confirmed ductal breast cancer, primary tumour <5 cm, axillary lymph nodes stage <N3, and no clinical indication of distant metastases at discharge were included. Response rates among women with cancer were 75% (Germany), 76% (Northern Ireland), 92% (Switzerland), and 97% (Netherlands, Spain). Controls were obtained from the hospital catchment area by using registries of local municipalities (Switzerland and Germany with response rates of 22% and 45%) and general practitioners as the sampling frame (Northern Ireland, Netherlands, and Spain with response rates of 46%, 50%, and 91%). Information on factors relevant to breast cancer was obtained by similarly formatted centre specific questionnaires, with a priori specified variables for central data analysis. Procedures for sampling and data collection were approved by local ethics committees in each participant's country. Informed consent was obtained from eligible subjects.
Needle aspirates were taken from the subcutaneous fat of the buttocks in 317 women with cancer and 367 controls.15 16 17 For the cancer patients the fat aspirates were taken within seven days after hospital admission. After the exclusion of five subjects for whom relevant background data were missing, fat aspirates of 265 cases and 341 controls were available for laboratory analysis—that is, DDE was analysed in 84% and 93% of the aspirates obtained. This percentage is lower in the cases because for one of the centres the aspirates had to be collected in several hospitals by different staff, less skilled in fat aspiration. To facilitate the standardising and instructing of the skills for fat aspiration, a videotape was prepared and distributed among all centres before the start of the project. In the four other centres this apparently led to good results as sufficient amounts of fat could be aspirated from 91% of the cases and 93% of the controls in these centres. After local storage at -70°C and shipment on dry ice, samples were analysed centrally for vitamins and fatty acids.15 16
Laboratory analyses of DDE were conducted at the National Institute of Public Health and Environmental Protection (Bilthoven, Netherlands), with sample vials randomised before laboratory analyses and staff blinded to disease status. The initial iso-octane extracts contained the methylated fatty acids and heptadecanoic acid (C17:0) as an internal standard. This solution was subjected to hyphenated chromatography, in which liquid chromatography was coupled on line to capillary chromatography. This analysis was carried out on a commercially available liquid chromatography-gas chromatography system (Dualchrom 2000; Carlo Erba Instrument, Milan, Italy) equipped with a Ni63 electron capture detector. The liquid chromatography injection volume was 50 μl. The on column interface was used to transfer a volume of 200 μl into the column. The chromatographic columns were a liquid chromatography column of 50 mm × 1 mm internal diameter packed with 3 μm hypersil silica and a DB5MS gas chromatography column (30 m × 0.32 mm internal diameter, film thickness 0.5 μm). The C17:0 content was determined by gas chromatographic analysis. When the concentration of DDE was below the detection limit (n=19) half the lowest concentration in the remaining samples was entered. The coefficient of variation within each batch varied from 1% to 5% for C17:0 and from 1% to 8% for DDE and between batches from 7% to 8% for C17:0 and from 8% to 11% for DDE. Concentrations of DDE are expressed in micrograms of DDE per gram of fatty acids in the aspirate.
Data analyses were performed with sas statistical software.18 Although skewed distributions of DDE required log transformations in data analysis, results presented are geometric means. Crude means as well as means adjusted for age and centre were calculated for DDE and major risk factors for breast cancer. To identify potential confounders, mean levels of risk factors for breast cancer were compared among fourths of DDE concentration in controls. Odds ratios (and 95% confidence intervals) for breast cancer were obtained from multivariate logistic regression analysis by modelling DDE, firstly in fourths and then as a continuous variable, and contrasting the risk for subjects on the 75th and 25th centiles. In tests for trend, the median DDE concentration in each fourth was used. Interaction terms between DDE and body mass index, waist : hip ratio, fat mass, parity, hormone replacement therapy, oestrogen receptor status, time since menopause, previous benign breast disease, family history of breast cancer, and current alcohol use were tested, but none was significant.
Table 1 shows risk factors among cases and controls. Age was similar because of the matched design. Women with breast cancer had significantly higher body mass indices, waist : hip ratio, age at first birth, and family history of breast cancer. No other significant differences were found.
Among controls, body mass index (kg/m2) was positively associated with DDE concentration and increased from 24.9 (SD 4.0) in the lowest fourth to 28.2 (4.7) in the highest fourth (P<0.01). Waist : hip ratio increased from 0.83 (0.08) to 0.86 (0.06) (P=0.03). Reported history of benign breast disease differed between fourths of DDE (P=0.05), ranging from 13.1% (10.2%) in the second to 3.6% (10.8%) in the fourth. Current alcohol use was more prevalent in the lowest than in the highest fourth (57.4 (7.9%) v 40.2 (8.5%); P<0.01).
Mean DDE concentration was 1.35 μg/g and 1.51 μg/g among cases and controls, respectively, a difference of -10.5%, which changed little after adjustment for age and centre (-9.2%; P=0.36; table 2). Concentrations were lower among cases than among controls in all but one centre (the Netherlands). DDE concentrations ranged from 2.5-fold to 3-fold between countries, being lowest in Northern Ireland and highest in Spain. Within controls, the median exposure in the lowest (0.40 μg/g) versus the highest fourth (5.07 μg) differed 12.5-fold (table 3). The odds ratio for breast cancer, adjusted for age and centre, for the highest versus the lowest fourth of DDE was 0.73, but the test for trend was not significant (P=0.16). Adjustment for body mass index, age at first birth, and current alcohol consumption strengthened the inverse association (P for trend=0.02), with the odds ratio in the highest fourth showing a significant 52% risk reduction. The largest change in the odds ratio occurred when body mass index was entered into the model, which made further adjustment for waist : hip ratio superfluous. When DDE was modelled as a continuous variable, the odds ratio for a subject at the 75th versus the 25th centile (3.46 v 0.86 μg/g) was 0.84 (95% confidence interval 0.73 to 0.97). Further adjustment for waist : hip ratio, history of previous benign breast disease, family history of breast cancer, and years since menopause did not materially affect the risk estimates.
Our study did not confirm the hypothesis that DDE concentrations in adipose tissue are higher in women with breast cancer. In contrast, we observed an overall inverse association between DDE and breast cancer, which was consistent in most of the study centres.
Interpretation of results
In the EURAMIC study, we assessed DDE in fat aspirates rather than plasma, and subjects who had lost over 5 kg of body weight in the past year were not eligible. These characteristics are important for diminishing random misclassification and for preventing information bias in the case-control design. Response rates of women with cancer ranged from 75% to 97% compared with 22% to 45% among population controls (Switzerland, Germany) and from 46% to 91% when control recruitment was mediated by general practitioners (see methods).
Case-control differences in the DDE concentrations were largest in the centre that contributed the smaller number of subjects to the dataset (table 2). Although part of this difference may have resulted from a less successful aspiration technique in this centre, it might be due to chance as well, and because of the small number of subjects it will not have affected the main findings and conclusions. Moreover, the case-control differences in DDE concentrations specific for centre were not obviously related to response rates or to control selection procedures.
The overall ratio of the DDE concentration in cases and controls was 0.89 (1.35:1.51; see table 2), and it was slightly closer to 1.0 in the three centres that had higher response rates because controls were obtained by general practitioners (Spain 0.82, Northern Ireland 0.96, Netherlands 1.05) than in the centres that obtained their controls through population registers (ratio in Switzerland 0.89, in Germany 0.68). Although this might suggest some bias towards an inverse association between DDE and breast cancer, multivariate adjustment for age at first birth, alcohol, and body mass index will have attenuated biases resulting from subject recruitment and risk factors for breast cancer. Finally, the observed associations between breast cancer and body mass index, reproductive factors, and familial risk are consistent with those reported in the literature, adding credibility to the internal validity of our results.
The discrepant findings from case-control studies may be attributable to subject recruitment and choice of biomarker. In early studies, DDE concentrations in plasma, mammary fat, or tumours from patients with breast cancer were compared with those in healthy subjects and patients with other diseases.10 11 12 19 20 Although some of these studies reported higher concentrations of DDE among cases, given the study design and small number of subjects (not exceeding 44 cases and 33 controls) it is difficult to regard the conclusions as definitive.
Comparison with other research
Table 4 summarises the main results as well as population and design characteristics of the three largest studies on DDE and breast cancer, including the EURAMIC study. The negative findings from Krieger's prospective study arise from data which, having been collected before the ban on DDT and many years before diagnosis of breast cancer, showed high concentrations of DDE with a wide range.14 The results from the two case-control studies relate to exposure levels about 10-fold lower and a range in exposure 5-fold to 6-fold narrower, clearly relating to women who were exposed for a shorter period of their lives.13 16
The small nested case-control study by Wolff et al (table 4) was a preliminary report which used prevalent, subclinical cases, in which blood samples had been taken from the patients six months or less before diagnosis. We used adipose DDE concentrations to obtain a long term preclinical reference period within one week after diagnosis and excluded subjects with clear weight loss, thereby providing a more robust assessment of exposure. Although serum and adipose DDE concentrations are highly correlated in healthy women (r =0.94),22 this may not extend to women with disease. Metabolic changes secondary to disease inception but preceding clinical diagnosis may mobilise fat stores, leading to increased plasma concentrations and even to some decrease in DDE concentrations in adipose tissue.
To explain the discrepancy between the Wolff and Krieger papers, it has been postulated that previously higher concentrations of the more potent antioestrogen dioxin have masked the risks of the weakly oestrogenic DDE in the Krieger study,23 but our results do not support this explanation. Similarly, one could argue that the weakly oestrogenic effects of DDE might be measurable only at low background concentrations of endogenous oestrogens—that is, after menopause. No evidence of weaker associations after menopause was reported by Wolff or Krieger, and the inverse association we observed in our study among postmenopausal women does not support this idea either.
As lactation is a major route of excretion of DDE in women, parity and duration of lactation have both been associated with lower concentrations of DDE in women.24 25 26 27 With the possible exception of prolonged lactation, breast feeding is not considered to provide substantial protection against breast cancer within the populations represented in this study and had not been included in our questionnaires. Therefore, like Krieger et al,14 we adjusted for parity and age at first birth as proxy variables, but the odds ratio was not materially affected. In the case-control study by Wolff et al, which did provide data on lactation, the positive association between DDE and breast cancer increased by almost 40% after adjustment.13 In our data, this would increase the odds ratio of 0.48 (for highest versus lowest fourth of DDE) to about 0.7, consistent with our conclusion of lack of association.
When we consider the characteristics of the epidemiological studies on DDE and breast cancer, the apparently conflicting results may be due to a combination of chance and mobilisation of energy from fat stores in the cases. Although these results do not support complex biological interactions between DDE and other environmental or endogenous (anti)oestrogens, the recent observation of a 1000-fold potentiation by combinations of two environmental oestrogens28 (but not DDE) suggests that these substances may be related to a wide range of health effects. Whatever the reality, the results of this large case-control study are clearly incompatible with a substantially increased risk of breast cancer among European women with high DDE concentrations.
Funding: The study on DDE and breast cancer was supported by EC DG-V “Europe Against Cancer” grant No 2216. The larger framework of the case-control study (EURAMIC) was supported by the commission of the European Union, as a Concerted Action within DG-XII, with additional support from DG-V “Europe Against Cancer.” The national studies were financed by the Dutch Ministry of Health, Ulster Cancer Foundation, Milk Intervention Board (corresponsibility Levy Disbursement Reg EEC 110/90 contract 77.2), Swiss Cancer Research, Swiss National Research Foundation (grant 32-9257-87), Spanish Health Research Fund (FIS), Ministry of Science and Education, Swiss NRF, and German Federal Health Office.
Conflict of interest: None.