There are several reasons, beginning with the fact that few epidemiological studies even try to examine the effects of mixtures, and instead consider contaminants one at a time. Yet mixtures are the rule. No one encounters one contaminant at a time.

This is especially problematic for epidemiology as it is currently practiced. Mixtures can make it difficult to detect impacts of contaminants even when the individual components of the mixtures cause strong effects individually.

If a suite of contaminants share the ability to cause the same health effect, for example breast cancer, then the fact that people have mixtures that vary from person to person depending upon their personal histories will impair epidemiology's ability to demonstrate cause and effect. For epidemiological studies that focus on individual components within mixtures, the fewer the number of similarly-acting contaminants in mixtures and the stronger their individual impacts, the more likely it is that a cause-effect relationship can be established.

Consider the following exaggerated analogy. Police investigate what appears to be a serial murderer who has a signature way of killing and has dispatched 50 people, all the same way. They identify a series of 50 suspects by establishing that each of them were the last person seen with one of the victims. But then each of them has a strong alibi for the 49 other murders. Hence the police eliminate each of them as a potential suspect. Yet it turns out they were working together, each one committing only one of the murders.

In the case of breast cancer, life-time exposure to estrogen (natural estrogens exist in the body in several forms) has been established as a risk factor. Xenoestrogens (contaminants that behave as estrogens) become a logical suspect, and considerable attention has been paid to testing the impact of individual xenoestrogens, one at a time. The reality is, however, that there are many xenoestrogens, they always occur in mixtures, but the precise composition of the mixtures vary from person to person. And, recent research has established that mixtures of xenoestrogens, each at levels at which they cause no effect, can dramatically increase the effect of natural estrogen.

This becomes even more challenging because of the likelihood of long delays between when breast cancer is initiated and when it is detected. This delay may involve decades; indeed some research is even exploring the possibility that events in the womb set breast cancer in motion. This is clearly the case for testicular cancer.

Most studies of breast cancer that have attempted to look at links betwen contaminants and risk measure contamination levels at the time of diagnosis (or later). These all assume that this measurement of contamination is an indicator of contamination levels at the time the disease was initiated. For the more persistent compounds like DDE and PCBs note below, this is at least plausible, although to the degree that it is wrong it will weaken the strength of the association.

But many xenoestrogens are not persistent, like bisphenol A. Their concentrations will rise and fall over the short term in ways that cannot be indicated by samples taken long after-the-fact. If the key attribute in determining whether a contaminant is a risk for breast cancer is its estrogenicity, and if xenoestrogens act in mixtures the way that current research indicates, this will render epidemiological studies of individual contaminants virtually useless for detecting risks.

Note: DDE and several PCB congeners have become the much-studied standard suspects in epidemiological studies purporting to test whether xenoestrogens are breast-cancer risks. The problem is that DDE and the commonly-studied most persistent PCBs act as an anti-androgen and anti-estrogens, respectively, not estrogens. Findings that indicate these contaminants are not associated with breast cancer risk are completely irrelevant to the hypothesis that xenoestrogens may induce breast cancer.

In the forms in which they are manufactured, DDT and PCB mixtures show weak estrogenic activity. Over time, however, as the compounds are metabolized and less persistent forms disappear, this weak estrogenicity also disappears.

The most active component of commercial DDT at the time of sale is p,p'-DDT. This chemical does have estrogenic activity. Both it and the metabolite to which it is converted in the human body, however, disappear years after exposure. What doesn't disappear is p-p'-DDE. This is the most persistent metabolite of commercial DDT. It is the form that lingers the longest. And rather than being a weak estrogen, p-p'-DDE is an anti-androgen.

Analogous changes occur with PCB mixtures. It turns out that many of the estrogenic PCBs (and their more active metabolites) are not persistent. The most persistent of the PCB metabolites that are found in human tissue actually are anti-estrogens.

What this means is that the studies measuring DDE and PCBs at the time of diagnosis (or afterward) are actually studying anti-androgens, when the original estrogenic forms of the compounds from which they were converted metabolically have long-since disappeared. To conclude that negative findings using this approach establishes that xenoestrogens are not risk factors for breast cancer defies scientific logic.