3.4. Discussion
This further analysis of the PEI Breast Cancer Study was one of only a few case-control epidemiological investigations of fungicide exposure and breast cancer risk and the first to examine gene-environment interactions for genetic polymorphisms in p450 enzymes in relation to fungicide exposure. Though the present investigation did not identify a statistically significant association between fungicide exposure and breast cancer risk, toxicological evidence exists to support the role of fungicides as carcinogens and endocrine disruptors [
7,
8]. Chlorothalonil, one of the primary fungicides used on PEI, has been classified as a probable carcinogen by the U.S. EPA [
7]. The primary metabolite of chlorothalonil (4-hydroxy-2,5,6-trichloroisophthalonitrile), known as DS-3701, has greater toxicity than its parent compound, as observed from wildlife studies [
31]. The two other predominantly used fungicides on PEI, metiram and mancozeb, are both ethylene bisdithiocarbamates fungicides (EBDC) and have been classified as endocrine disruptors. Toxicological studies have observed that exposure to mancozeb induces apoptosis, thus providing a mechanistic link between mancozeb and cancer [
32,
33]. Moreover, the metabolite of EBDC fungicides, ethylene thiourea, has been classified as a carcinogen [
34]. A cohort study that examined the association between maneb or chlorothalonil exposure and breast cancer risk did not identify an association between ever
vs. never use of either chlorothalonil and breast cancer risk though this study was also limited by small samples sizes of exposed cases [
9]. Metiram and mancozeb, the two other primary fungicides used in PEI were not assessed in the Iowa/North Carolina cohort study.
The present investigation’s observed that fungicide exposure was not statistically significantly associated with breast cancer risk. The trend in the models with binary and categorical exposure variables was towards a reduced risk of breast cancer risk among women with elevated levels of exposure. A biologically plausible explanation for this finding could be that fungicides inhibit the
CYP p450 system and thus lessen the concentration of estrogen metabolites. Toxicological literature has reported that maneb, an EBDC fungicide with a chemical structure similar to mancozeb, inhibits
CYP p450 enzymes, yet other supportive literature is lacking [
35]. On the contrary, it is more likely that the potential carcinogenic effects of fungicides were not observed due to the limitations in sample size, bias, and misclassification. Further toxicological and epidemiological evidence is needed to clarify the relationship between fungicide exposure and adverse health outcomes.
The analysis of the interaction between fungicide exposure and CYP1A1*2A did not identify a significant product term for any of the exposure variables. As a main effect, presence of the heterozygous and homozygous variant CYP1A1*2A alleles suggested an inverse association with breast cancer risk, though was not statistically significantly. Biological interpretation of these observed findings is again limited due to the low sample size and potential influences of bias.
While no previous research reported in the literature has examined the interaction between fungicide exposure and the
CYP1A1*2A allele in relation to breast cancer risk, studies of PCB exposure demonstrated that neither
CYP1A1*2C status nor PCB exposure were independently associated with increased breast cancer risk [
10,
11]. However, women with the variant
CYP1A1*2C allele who had the highest level of PCB exposure experienced an increased risk of breast cancer. Similarly, Li
et al. [
36] reported a trend among women with elevated levels of pesticide exposure towards increased breast cancer risk among subgroups of women with the
CYP1A1*2C and
CYP1A1*3 alleles.
In seeking an explanation for our findings, it is necessary to consider the dual role of
CYP1A1 activity. On the one hand, as
CYP1A1 is induced by many environmental contaminants, induction may increase the rate of estrogen metabolism toward the 2-hydroxyestrogens and away from the more estrogenic metabolites [
12]. In this scenario, the interaction between the environmental exposure and the variant
CYP1A1*2A allele could play a protective role because of an even greater activity of
CYP1A1. On the other hand,
CYP1A1 has the potential to activate environmental toxins leading to an increased exposure to potentially toxic metabolites [
37]. This latter role of
CYP1A1 is particularly relevant to the present investigation as the metabolites of both EBDC fungicides (ethylene thiourea) and chlorothalonil (DS-3701) are more toxic than the parent compounds. In this case, women with the variant allele may be at an increased risk of breast cancer due to an elevated concentration of these toxic metabolites. The possible relationships or interactions are therefore complex, and it is not possible to speculate from the current data on biological mechanisms or the significance of the putative association.
3.5. Strengths and Limitations
The primary strength of this investigation is the analysis of an understudied class of pesticides in a relatively high-exposure intensity geographical region. Moreover, the availability of detailed lifestyle and covariate data from the PEI case-control study allowed this investigation to evaluate potential confounding due to established breast cancer risk factors and the presence of genetic polymorphism data facilitated the analysis of gene-environment interactions.
In addition, PEI provides a valuable study setting due to the presence of a national healthcare system and relatively homogenous population. All residents of the island are provided with universal access to the national healthcare system, thus eliminating any financial or insurance barriers to screening or treatment. At the time of the study, the Queen Elizabeth Hospital was the primary oncology clinic and all breast cancer diagnoses on the island were recorded in the PEI cancer registry. During the study time period, 345 breast cancer diagnoses were recorded in the PEI cancer registry. The case-control study recruited 207 (60%) of these breast cancer cases, the geographic distribution of residential addresses of the case population was examined and shown to be representative of the overall pattern of disease on the island. In contrast to communities lacking a national healthcare system, women who reside in PEI may be more likely to seek routine mammography screening. Rates of mammography screening have been reported at 56% on the island [
38]. The QEH mammography clinic is one of only two screening centers on the island. The QEH mammography clinic is located in the capital city, Charlottetown, and has two diagnostic imaging machines, whereas the screening clinic in Summerside is located in a community with lower population density and only one machine. It is estimated that two-thirds of PEI women undergo mammography screening at the QEH clinic. Moreover, more than 50% of the consolidated census subdivisions in PEI were represented by the study population (as indicated by a case or control residing in that subdivision). Thus, cases and controls provide a reasonable representation of the same underlying source population on the island.
The present investigation is, however, challenged by some exposure misclassification, and small sample size. While the use of the Agricultural Census data was a feasible means of assessing regional levels of fungicide exposure retrospectively, these data did not permit analysis of individual level exposure to specific fungicides. Moreover, reliance on postal code rather than individual residential address may have inappropriately assigned certain individuals to a CCS if they lived at the border of the CCS or collected their mail in a different location than where they resided, leading to errors in the percentage of hectares receiving fungicides assigned to some participants. These limitations highlight the challenges of utilizing a geographic based approach to exposure assessment. The inclusion of time-activity personal exposure histories and measurements of specific fungicides in environmental media would enhance precision of such analyses in future studies, but was not possible in this retrospective analysis of secondary data.
There are other considerations that must be taken into account in interpreting these results. PEI is relatively small, and given the extent to which fungicides are sprayed across the Island, the entire population may have in fact been exposed to elevated fungicide concentrations, as suggested by fungicide ambient air quality studies conducted by Environment Canada [
39]. The inclusion of an external, off-Island control group would have strengthened the interpretation of study findings, but an external control group was not part of the original study. Moreover, while recruitment of the control population from the mammography clinic did minimize the potential for outcome misclassification among controls, this strategy does present potential selection bias as it may inadvertently exclude women who do not seek healthcare from the study population.
Lastly, the study was underpowered to detect associations between the estimated fungicide exposure and breast cancer risk, although the sample size was limited by the sample size of the original study. Utilizing the percent of exposure among cases and controls from the binary exposure variable, the study had 25% power to detect the observed associations, thus the likelihood of a Type 2 error is notably high. Moreover, the analysis of the interaction between CY1A1*2A polymorphisms and fungicide exposure also lacked sufficient power to make conclusive observations about the findings.