Repository logo

Menstrual cycle characteristics in women exposed to atrazine in drinking water


Introduction. Atrazine is the most commonly used herbicide in the United States and a wide-spread groundwater contaminant. Concern regarding potential health effects of human exposure to atrazine is based on its well recognized designation as an endocrine disruptor. Studies have shown that menstrual cycle characteristics are markers for reproductive conditions. The specific hypothesis tested in this research was: Exposure to atrazine in municipal drinking water is associated with menstrual cycle abnormalities which, in turn, are modulated through a diminution of the pre-ovulatory luteinizing hormone surge. In addition, the following secondary hypothesis was tested: There is agreement between retrospective menstrual cycle questionnaire data and data obtained prospectively from menstrual cycle diaries. This study was the first to examine the effects of drinking water exposure to atrazine on menstrual function in humans and the first to examine the underlying mechanism of this association.
Methods. The state of Illinois was selected as the exposed study site location because of its intensive atrazine use. In 2005, 87% of Illinois corn was treated with 0.38 pounds of active ingredient applied per acre. The state of Vermont was selected as the comparison site because of its low atrazine use. In 2005, 0.009 pounds of active ingredient per acre were applied. The study population was comprised of women 18 to 40 years old residing in Illinois and Vermont. Women participated by either answering a retrospective questionnaire describing their menstrual cycle characteristics, maintaining a prospective menstrual cycle diary and/or collecting daily urine samples through two menstrual periods. Participants provided two first morning urine voids (one on each day) and a total of four home tap water samples (two on each day) were collected for analyses of atrazine and atrazine degradation by-products. Participants also collected urine voids daily through two or more menstrual bleeding periods for determination of reproductive hormone levels and phase length. Results of municipal plant analyses (Syngenta Crop Protection, Inc. 2005) were obtained from the Illinois Environmental Protection Agency (EPA).
Crude and multivariable unconditional logistic regressions were used to assess the relationship between atrazine exposure and the menstrual cycle characteristics. Differences in means and both crude and multivariable linear regression were used to evaluate the relationship between drinking water exposure to atrazine (or markers of atrazine exposure) and menstrual cycle length as reported by the prospective menstrual cycle diary. Differences in means and crude and multivariable linear regression were also used to evaluate the potential relationship between drinking water exposure to atrazine (or markers of atrazine exposure) and the urinary concentrations of reproductive hormones including luteinizing hormone, pregnanediol 3-glucuronide, and estrone 3-glucuronide levels. Percent agreement, Cohen's kappa and the prevalence index were used to compare retrospectively reported menstrual cycle characteristics with prospectively maintained menstrual cycle diary data.
Results. One hundred and two women participated in the study by answering a retrospective questionnaire (53 Illinois women and 49 Vermont women). Sixty seven of these women (65.7%) also maintained menstrual cycle diaries (30 Illinois women and 37 Vermont women). Thirty nine of these 102 women (38.2%) also provided daily urine samples for hormone analyses (18 Illinois women and 21 Vermont women).
Overall, levels of atrazine and atrazine metabolites were low in 2005, the year of data collection for this study, relative to previous and subsequent years. Atrazine levels in tap water were higher among Illinois women compared to Vermont women (p-value (p) = < 0.001). According to municipal plant monitoring (Syngenta Crop Protection, Inc.), atrazine averaged 0.21 parts per billion (ppb) in Mount Olive, Illinois and 0.29 ppb in Gillespie, Illinois in 2005.
Menstrual cycle length irregularity was associated with atrazine exposure, as estimated in several ways. Using state of residence as an exposure marker, women living in Illinois were more likely to report irregular menstrual cycle lengths (OR = 4.69; 95% CI: 1.58 - 13.95). In addition, a significant association, although imprecise, was observed between cycle length irregularity and residing more than four years in current home in Illinois (OR = 6.88; 95% CI: 2.08 - 22.78). As the number of years in the current home increased among Illinois women, increasing odds ratios (ORs) were observed in a dose response manner. Although uncertainty exists because of the wide confidence intervals, a possible dose response association was also observed between the amount of unfiltered water consumed and menstrual length irregularity among Illinois women (< 2 cups OR = 4.10, 95% CI: 1.24 - 13.51; > 2 cups OR = 5.73, 95% CI: 1.58 - 20.77 with Vermont women the comparison group).
Going more than six weeks without a menstrual period was significantly associated with residence in Illinois (OR = 6.16; 95% CI: 1.29 - 29.38). Elevated odds ratios, although not in a dose response manner, were also observed for years in current home and going more than six weeks without a menstrual period (< 4 years OR = 9.68, 95% CI: 1.83 - 51.22; > 4 years OR = 3.76, 95% CI: 0.64 - 21.97 with VT women living in Vermont the comparison group).
There were no statistically significant associations between atrazine exposure and menstrual cycle length as measured with menstrual diary data. Since the latency period from atrazine exposure to menstrual cycle disruption is unknown, it is uncertain whether data collected from prospective diaries would reflect the present exposure or exposure months earlier.
A statistically significant association was present between municipal plant chlorotriazine monitoring data and follicular phase length in the adjusted linear model (P = -0.019; 95% CI: -0.04 - 0.00). When municipal plant data were used to calculate estimated 'dose' (atrazine concentration x volume of unfiltered water ingested per day), a statistically significant increase in follicular phase length was observed with atrazine estimated 'dose' (p = -0.021; 95% CI: -0.04 - 0.00) and with chlorotriazine estimated 'dose' (P = -0.023; 95% CI: -0.04 - 0.00). These associations remained significant in the adjusted model.
Increased estimated 'dose' of atrazine and chlorotriazine were associated with decreased mid-luteal phase estrone 3-glucuronide levels for several markers. When municipal plant data were used to calculate estimated 'dose', mean mid-luteal phase estrone 3- glucuronide levels decreased but not significantly (atrazine: 29.45 ng/mg Cr compared to 20.62 ng/mg Cr, p = 0.36; chlorotriazine: 32.99 ng/mg Cr compared to 20.94 ng/mg Cr, p = 0.09). Using CDC (Centers for Disease Control and Prevention) analyzed samples to estimate 'dose', mid-luteal phase estrone 3-glucuronide levels were significantly decreased (both atrazine and chlorotriazine: 35.67 ng/mg Cr compared to 24.39 ng/mg Cr,/? = 0.01). For the linear regression analysis, state of residence was imprecisely associated with mid-luteal phase estrone 3-glucuronide (β = -0.32; 95%CI: -0.68 - 0.04). Estrone 3-glucuronide was also imprecisely associated with atrazine exposure as determined by concentrations in residential tap water (β = -0.32; 95%CI: -0.68 - 0.04). When the amount of tap water consumed was taken into consideration to estimate 'dose', this association became stronger and statistically significant (β = -0.46; 95%CI: -0.82 - 0.10).
Exposure to atrazine through drinking water also appeared to have an effect on the concentration of progesterone during the luteal phase. Mean mid-luteal phase exposure to atrazine through drinking water also appeared to have an effect on the concentration of progesterone during the luteal phase. Mean mid-luteal phase pregnanediol 3-glucuronide levels decreased with increasing atrazine (7.92 jug/mg Cr compared to 12.44 μg/mg Cr, p = 0.02). Results of the linear regression analysis were in agreement as mid-luteal phase pregnanediol 3-glucuronide was statistically significantly associated with atrazine estimated 'dose' when using municipal plant data (β = -0.57; 95% CI: -1.06 - -0.09).
Although not statistically significant, atrazine exposure appeared to be associated with small but consistent reductions in preovulatory luteinizing hormone concentrations across the various atrazine exposure variables.
For the retrospective versus prospective menstrual cycle data analysis, a regular menstrual cycle was defined two different ways (definition 1 = 25-30 days; definition 2 = 25-35 days). A high overall agreement was observed between retrospective questionnaires and prospective diaries (69% and 75% for definitions 1 and 2, respectively) but unadjusted Cohen's kappas were low (0.31 and 0.43 for definitions 1 and 2, respectively). The prevalence indices were -0.33 for definition 1 and -0.66 for definition 2.
Conclusions. Although the majority of atrazine concentrations in municipal drinking water measured in this study were below the EPA standard for drinking water (3.0 ppb), exposure to atrazine was associated with altered menstrual cycles. Menstrual cycle length irregularity, increased follicular phase length and increased cycle length were significantly associated with atrazine exposure. Moreover, the reproductive hormone results provided further support of the menstrual cycle findings and offer the possibility of reduced fecundability in women exposed to atrazine. Given the dependence of reproductive hormones on one another, any hormone modifications could lead to a menstrual cycle alteration.
In addition, there was agreement between data for menstrual cycle characteristics reported retrospectively from questionnaires and data obtained prospectively from menstrual cycle diaries. Therefore, although Cohen's kappa was low, it was shown to have been kept deceptively low by the high likelihood of chance agreement resulting from the high prevalence effect. The demonstration of the effect of prevalence on the kappa statistic was a major finding of this study. The major strengths of the study were availability of tap water samples, urinary analyses of atrazine, urinary concentrations of reproductive hormones and both retrospective as well as prospective measurement of menstrual cycle activity. The major limitations were the relatively low levels of atrazine measured in Illinois drinking water during 2005 and the small number of subjects. Further studies, on larger populations, are needed to confirm the findings of this study.


Rights Access


endocrine disruptors
menstrual cycle
reproductive hormones
water contamination


Associated Publications