Issues Concerning Assessing Risks Posed by EDCs to Human Health

Chemo Secrets From a Breast Cancer Survivor

Breast Cancer Survivors

Get Instant Access

All of these findings suggest that it is likely that EDCs pose a significant threat to human health that classical epidemiological methods may not have the sensitivity to detect. Human studies on EDCs, which fall under the broader heading of environmental epidemiology, share many features with studies of environmental exposures, such as to radon or total suspended particulates, which are not endocrine disruptors. However, they differ from non-EDC studies in several important ways (study hypothesis, exposure(s), effect(s), model selection, analysis, and interpretation) that make detection of effects more difficult. We will consider these points and then examine a recent study that circumvents at least some of these problems.

What triggers an investigation between an EDC and a human health effect? Traditional ('classical') epidemio-logical studies were often designed to investigate unusual patterns of human health outcomes. Perhaps the most dramatic of these was the investigation of diethylstilbestrol (DES) in response to a cluster of seven cases of a rare vaginal cancer (clear cell adeno-carcinoma) in young women. Similarly, an awareness of increasing rates of lung cancer triggered the first studies of smoking and lung cancer. Some epidemiological studies of EDCs have similar origins. Indeed, DES itself is a quintessential EDC, and current research into possible EDC involvement in breast cancer causation and fertility impairment have been provoked by observations of human trends.

Many epidemiological questions raised by EDCs have their origins, however, in observations of impacts on laboratory animals and wildlife. These include the possible role of EDCs in increases in hypospadias, the effects of phthalates on male fertility, and the impact of polybrominated diphenyl ethers (PBDEs) on neurocog-nitive development. In each of these cases, and many more, pronounced laboratory and field effects provoke questions about human impacts based on animal observations.

All else being equal, the ability of an epidemiological study to identify the cause of an adverse outcome decreases as the prevalence of the outcome and the number of causal factors increase. For example, the identification of DES as the cause of clear cell vaginal adenocarcinoma in young women was relatively easy because vanishingly few cases of this rare cancer had ever been documented in this age group, and no other cause had ever (before or since) been identified. Conversely, causes of breast cancer are notoriously hard to find, not only because it is a complex, multifactorial disease but because of its extremely high lifetime incidence (one in eight women). The metaphor of signal detection may be helpful in clarifying this point; high background levels of a disease contribute to background 'noise', as do alternative causes, errors in exposure identification and diagnoses) and make detection of the 'signal' (the association under investigation) difficult to identify. Epidemiology handles diseases of low incidence, and strong associations well, but multifactorial diseases of high incidence only poorly.

Consider the following thought experiment (Figure 1). Imagine a population of 5000 women with a significant but not unusual spontaneous miscarriage rate of 10%, normally distributed. In that hypothetical population one would expect 500 miscarriages, ±42. In this experiment, expose 1% of the women to a

Elevation in risk (-fold)

Figure 1 The expected number of miscarriages in a population of 5000 women as a function of risk elevated by exposure to a hypothetical contaminant. See text for parameters.

Elevation in risk (-fold)

Figure 1 The expected number of miscarriages in a population of 5000 women as a function of risk elevated by exposure to a hypothetical contaminant. See text for parameters.

contaminant that increases the risk of that abortion, on average by X-fold, with X increasing from 1 (no effect on risk) to a tenfold increase. Elevation in risk would have to be more than ninefold before the signal of exposure-induced miscarriage rose above background noise.

A crucial feature of EDCs is their 'stealth' nature. Several recent studies have demonstrated that the general population has been exposed to, and currently carries measurable levels of, tens to hundreds of EDCs. The subject has no knowledge of these exposures; so the classical tools of epidemiologists (questionnaires, vital records, occupational histories, etc.) provide no information. These presuppose the subject's (or physician's or employer's) knowledge of exposure. Instead, it is necessary to obtain biological measures of exposure (biomarkers).

Biomarker studies require that subjects agree to provide a biological sample (e.g., blood, urine, or saliva) and give permission for its use in such a study. Obtaining subjects willing to do this, Institutional Review Boards willing to approve these protocols, and funding for such studies is becoming increasingly challenging. An increasing number of studies are taking this approach; we describe a recent example below. But for many EDCs, the analytical chemistry that would permit body burden measurements has not yet been developed, and for many for which it has, the chemical analyses are very costly, limiting sample size and thus statistical power. Moreover, the rapid metabolic degradation of some compounds means that single exposure measurements, for example from cord blood at birth, may completely miss critical exposures during pregnancy. Finally, there is now clear evidence that for some environmental chemicals, the detection limits for chemical analysis methods are far above the concentrations that are able to cause biological effects.

Whether distinguishing between exposure and non-exposure in cases and controls or estimating changes in risk as a function of increasing exposure, epidemiological studies traditionally assume monotonic dose response curves. Higher exposure levels are assumed to produce larger effects. Laboratory work with EDCs clearly shows, however, that nonmonotonic curves are commonly found. The result of use of inappropriate models, such as those that assume monotonicity of dose response, and absence of low dose effects, will result in 'false negatives'.

Epidemiology regularly compares exposed and unex-posed populations. Yet the global distribution of EDCs means that finding unexposed populations is virtually impossible. Classical epidemiological studies were designed primarily to examine isolated exposures, ignoring concurrent exposures or considering them as confounding factors to be treated as 'nuisance variables'. This is inappropriate with EDCs, however, for two reasons. First, EDCs from similar and different chemical families can work through the same mechanism. They are thus substitutable. Unless the possibility of substitution is factored into the study by measuring multiple exposures and examining their joint risk, such mixtures will increase misclassification of exposure (an important source of conservative bias) and thus increase the likelihood of false negatives.

When mixtures of EDCs have been studied they have been seen to interact, often in unpredictable ways, with subadditive, additive and even synergistic effects. It is difficult, if not impossible, to isolate exposure to a single pesticide, phenol or phthalate. As a study by Thornton and colleagues showed, it is likely that all subjects are exposed to measurable amounts of large numbers of these chemicals, many of which act along common pathways. These factors pose a significant and currently unsolved challenge to epidemiology.

Long time lags between exposure and effect, which may span decades or even generations as in the case of DES, will further complicate detection of impacts. For nonpersistent compounds, all traces of the parent compound and its metabolites will likely have disappeared. With persistent contaminants, degradation of the parent compound into different metabolites, some toxic, some not, and some working via different mechanisms (e.g., o,p'DDT is estrogenic while the metabolite p,p'DDE is antiandrogenic) will further complicate interpretation, even in cases where the study has measured biomarkers of exposure.

Aside from ecological studies, epidemiology is conducted at the individual level. Effects of classical exposures are usually binary outcomes in individuals, which are well defined and severe (cancer case vs. non-case, birth with limb reduction or not). However, wildlife data suggest that changes from EDC exposure at the level of the individual are often subtle and difficult to classify (reduced fertility, poor semen quality, more feminine play behavior, genital dysmorphology). The effect of such changes at the population level, however, can be profound. As discussed above, trends in mean values of several outcomes have been reported but other changes at the population level, which may be even more profound, are increases in population variance and an increasingly non-normal (non-Gaussian) population distribution.

While EDCs manifestly present challenges to epidemiological studies, and are likely to have led to false negatives and underestimates of true risk, some progress is being made in developing approaches that acknowledge these pitfalls and employ methods explicitly designed to avoid them. One of us (Swan) has been involved in such a study, investigating reduced semen quality in relation to pesticide exposure. This study is somewhat unusual from an EDC perspective because it focuses on what appear to be adult-mediated impacts rather than developmental impacts. This avoids the problem of long time lags noted above.

A Study of Semen Quality in Relation to Pesticide Exposure

After finding that fertile men from the general population of an agrarian area (Columbia, MO) had decreased semen quality (e.g., only 58% of the number of moving sperm as men from Minneapolis, MN), pesticide exposure was examined as a cause of poor semen quality. The authors measured urinary metabolites of eight nonpersistent, current-use pesticides in two groups of men from mid-Missouri; men with all semen parameters (concentration, % normal morphology, and % motile) below median value (cases) and men in whom all semen parameters were within normal limits (controls). Pesticide metabolite levels were particularly elevated in cases compared to controls for the herbicides alachlor and atrazine, and for the insecticide diazinon (2-isopro-poxy-4-methyl-pyrimidinol, or IMPY) (p values for Wilcoxon rank test = 0.0007, 0.012, and 0.0004, for ala-chlor, atrazine and IMPY, respectively). Men with higher levels of alachlor or IMPY were significantly more likely to be cases than men with low levels (OR = 30.0, 16.7 for alachlor and IMPY, respectively), as were men with atrazine over the LOD (OR = 11.3). The number of pesticides found in the urine at elevated levels was significantly related to the risk of poor semen quality (being a case rather than a control). These associations were seen in the general population, who were not occupationally exposed. The three pesticides most strongly associated with semen quality are among the five that have been measured most frequently in drinking water sources in the mid-West. These are not removed by routine water treatment. Therefore, drinking water is the most plausible route of exposure. These findings suggest that adult exposure to several widely used pesticides via drinking water is a likely cause of the reduced semen quality seen in fertile men from mid-Missouri.

Subject responses to questions about home and occupational pesticide use were not related to semen quality, suggesting that the relevant pesticide exposure was unknown to the subject. Therefore, collection of urine samples and assays for pesticide metabolites in the subject's urine using highly sensitive GC-MS were required to document exposure to the low levels of pesticides that were related to semen quality. In addition, effects were seen at the level of the individual, with likely more profound effects at the population level. The average decrease in sperm concentration in fertile men living in mid-Missouri relative to men living in Minneapolis, MO is 40 million sperm ml- . While the median sperm concentration for Missouri men (54 million ml-1) was within normal limits, the sperm count for about 40% of these men fell below 40 million ml-1, the point at which fertility declines significantly.

Was this article helpful?

0 0
10 Ways To Fight Off Cancer

10 Ways To Fight Off Cancer

Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.

Get My Free Ebook

Post a comment