Gammon, MD, MS Wolff, AI Neugut, SM Eng, SL Teitelbaum, JA Britton, MB Terry, B Levin, SD Stellman, GC Kabat, M Hatch, R Senie, G Berkowitz, HL Bradlow, G Garbowski, C Maffeo, P Montalvan, M Kemeny, M Citron, F Schnabel, A Schuss, S Hajdu, V Vinceguerra, N Niguidula, K Ireland and RM Santella. 2002. Environmental Toxins and Breast Cancer on Long Island. II. Organochlorine Compound Levels in Blood. Cancer Epidemiology Biomarkers & Prevention 11: 686-697.

A 2001 review of breast cancer epidemiology
Press coverage of this report

This new study led by Marilie Gammon finds few associations between the risk of breast cancer and serum levels environmental toxins measured around the time of diagnosis. While that might seem to be good news, the study has severe limitations which undermine the policy implications of its conclusions.

Launched in 1993 as a result of a law mandating the National Cancer Institute to study links between environmental contaminants and breast cancer, the study examined risk of breast cancer as a function of current body burden levels of a series of compounds of traditional, historic concern (DDT, DDE, PCBs, chlordane and dieldrin).

The authors conclude that "it seems unlikely that breast cancer risk is associated with organochlorines when measured close to the time of breast cancer diagnosis." They imbed this conclusion in a series of cautionary statements, the most important of which is:

This possibility... that breast cancer risk may be elevated by exposure to organochlorine exposures several decades earlier... is the heart of the endocrine disruption hypothesis: exposure during crucial times in development causes developmental errors which (in the case of breast cancer) increase the likelihood of cancer later in life. Hence the results presented by Gammon et al., while potentially reassuring (but see below) about the impact of current exposures, are very limited what they tell us about endocrine disruption's contribution to breast cancer risk, even with respect to the chemicals they studied.

Recent work by Warner et al. showing signficant links between dioxin exposure and breast cancer in Seveso, Italy, residents reinforces this point. Breast cancer cases decades later are more likely in women with increased dioxin, measured in serum samples shortly after exposure. More...

Gammon et al. face another methodological limitation, however, which is at least as important. Their study focuses principally on a small number of traditional organochlorines (listed above). Women on Long Island are exposed to a significantly larger list of OCs, as well as non-organochlorine compounds that are active endocrinologically, including as estrogens... compounds like bisphenol A and nonylphenol. There are many more.

Gammon et al.'s work tells us nothing about the contribution to breast cancer risk that these compounds make. That's no surprise. But the presence of these compounds introduces another, less obvious flaw into this study which directly undermines their central conclusion, that there is no link between current OC exposure and breast cancer risk.

This is because of the ubiquity of mixtures. Everyone is exposed to multiple estrogen-like contaminants but the precise composition of the mixtures vary from person to person depending upon their individual history of exposure.

If the link between compounds and breast cancer risk is that they share the ability to bind with the estrogen receptor (for example), then what matters is likely to be the estrogenicity of the mixtures, not the individual compounds that are present. If there are many estrogen-like compounds in the mixture, and there is significant variability in the precise composition of the mixtures from woman to woman, then epidemiology will have a very difficult time discovering a link between one compound and breast cancer risk if measurements are only made of a small number of the estrogen-like contaminants.

Consider the following hypothetical but very plausible situation: In one woman, the mix of xenoestrogens that has led to breast cancer might be composed of one suite of compounds; in another, the causal mix could be different. If the epidemiological analysis focuses on individual components of mixtures, it easily could miss finding a statistically signficant link between any one compound and breast cancer, even though their presence in some women was the causal agent while the presence of other xenoestrogens in other women caused their cancers.

Unfortunately, this reality of mixtures complicates epidemiological research to the point that real impacts can be literally undetectable, at least by epidemiology as it is currently practiced. More...

This flaw is not Gammon et al's fault. It is a reality that we face because of mixtures are the rule, not the exception, and because these mixtures are made up of tens, if not hundreds, of compounds. And recent research has demonstrated unequivocally that the combined impact of even a small number of estrogen-like compounds in a mixture can be very significant. To complicate matters still further, contaminants sometimes interact with viruses in ways that can dramatically increase risk. No published studies of breast cancer epidemiology have tested explicitly for this possibility.

What this means in the real world is that epidemiological studies like this are virtually useless to disprove links between hormonally-active compounds and disease. The converse, however, is not true. A positive result of a carefully-conducted case-control study with large sample sizes should be taken very seriously. If a signal emerges out of the statistical noise that characterize these systems, despite all the limitations of epidemiology, then the links are likely to be very strong.

What did they do? Gammon et al. conducted a case-control study involving 1508 cases and 1556 controls from Nassau and Suffolk Counties on Long Island, New York. Most participants (93%) were white. Cases were recruited shortly after diagnosis of breast cancer. Controls were chosen using a randomizing procedure designed to match the age structure of the cases.

Blood samples for chemical analysis were selected from blood samples gathered on the basis of several criteria: (1) random selection of 415 cases and 406 controls; (2) all African American samples; (3) an additional 184 cases diagnosed with in situ disease that made blood sample donations.

The chemical analysis determined serum concentrations of p,p'-DDE, p,p'-DDT, oxychlordane, trans-nonachlor, dieldrin, and 24 PCB congeners.

In the data analysis, factors known to affect breast cancer risk on the basis of previous studies (e.g., history of breast feeding, alcohol and tobacco use, family history of breast cancer, etc.) were controlled statistically.

What did they find? Gammon et al. analyzed their data in several different ways. Simply comparing the concentrations of the measured organochlorines in cases vs. controls, they found no differences. Then using a multivariate model to control for the multiple variables known to influence breast cancer rate (such as those listed above), they found a suggested elevation of breast cancer risk for women in the top quintile of exposure of DDE, DDT and dieldrin, but not for PCBs or chlordane. There was not indication of a dose response curve for any of the OCs.

There were suggestions of higher risk for women that had not given birth and had higher levels of chlordane (a greater than 2 fold elevation in risk) but the sample size was small.

The authors conclude: "In this population-based case-control study conducted among women on Long Island, there was little evidence of an increased risk of breast cancer in relation to DDE, DDT, PCBs, chlordane, or dieldrin."

What does it mean? As noted above, little can be concluded with certainty from this study. Perhaps its greatest value will be to push future research in two directions: first, to test explicitly for links between breast cancer risk and exposure during crucial windows of development, instead of contamination levels at the time of diagnosis; and second, toward tackling the problem of mixtures. Our understanding of the contribution of contaminants to breast cancer risk await progress on these issues.