The toxicity of pesticides on human reproduction is largely unknown--particularly how mixtures of pesticide products might affect fetal toxicity. The Ontario Farm Family Health Study collected data by questionnaire on the identity and timing of pesticide use on the farm, lifestyle factors, and a complete reproductive history from the farm operator and eligible couples living on the farm. A total of 2,110 women provided information on 3,936 pregnancies, including 395 spontaneous abortions. To explore critical windows of exposure and target sites for toxicity, we examined exposures separately for preconception (3 months before and up to month of conception) and postconception (first trimester) windows and for early (< 12 weeks) and late (12-19 weeks) spontaneous abortions. We observed moderate increases in risk of early abortions for preconception exposures to phenoxy acetic acid herbicides [odds ratio (OR) = 1.5; 95% confidence interval (CI), 1.1-2.1], triazines (OR = 1.4; 95% CI, 1.0-2.0), and any herbicide (OR = 1.4; 95% CI, 1.1-1.9). For late abortions, preconception exposure to glyphosate (OR = 1.7; 95% CI, 1.0-2.9), thiocarbamates (OR = 1.8; 95% CI, 1.1-3.0), and the miscellaneous class of pesticides (OR = 1.5; 95% CI, 1.0-2.4) was associated with elevated risks. Postconception exposures were generally associated with late spontaneous abortions. Older maternal age (> 34 years of age) was the strongest risk factor for spontaneous abortions, and we observed several interactions between pesticides in the older age group using Classification and Regression Tree analysis. This study shows that timing of exposure and restricting analyses to more homogeneous endpoints are important in characterizing the reproductive toxicity of pesticides. Key words: atrazine, carbaryl, developmental toxicity, epidemiologic methods, glyphosate, herbicides, pesticides, phenoxy acetic acid herbicides, spontaneous abortion, thiocarbamates, triazine, windows of vulnerability.
**********
Farm residents may be exposed to several types of pesticides from various chemical families (e.g., phenoxy acetic acids, triazines, carbamates, and organophosphates) during the course of a growing season. Several studies have reported positive associations between occupational pesticide exposure and fetal death (spontaneous abortion or still-birth) (1-3). However, little is known about the human reproductive toxicity of specific pesticide active ingredients and even less about mixtures of pesticides and how they may interact with other risk factors.
In addition to the nature of the chemical and its target, the consequences of exposure to chemical agents depend on the timing of exposure relative to critical windows in development of the fetus or reproductive system (4,5). In a recent article (6), we noted that the risk of spontaneous abortion in farm families varied depending on when exposure to phenoxy herbicides occurred and on whether the abortion occurred earlier (< 12 weeks) or later (12-19 weeks) in the pregnancy. Previous analyses had also discussed the role of male pesticide exposure on pregnancy outcomes (7) and time to pregnancy (8). In this analysis we used the data from our study of farm families to explore further the critical windows of exposure, the target sites and interaction among the pesticides, and other risk factors for spontaneous abortion.
Subjects and Methods
The Ontario Farm Family Health Study collected data retrospectively by questionnaire from farm operators and eligible couples living on the selected farms, as described in detail elsewhere (6,9). To be eligible, the couple had to be living year round on the study farm and the wife had to be 44 years of age or younger (to reduce the length of recall of reproductive events). At least one member of the couple had to be working on the farm. Three questionnaires were designed to collect relevant information from the farm operator, husband, and wife on demographic and lifestyle information; pesticides currently and historically used on the farm and around the home; medical history; and a complete reproductive history.
The women in the study were asked to recall all their pregnancies, starting with their first. For spontaneous abortions, the woman was asked how many weeks pregnant she was (based on the last menstrual period) at the time of the abortion. We calculated the estimated calendar month of conception by subtracting the gestational age at abortion or delivery from the delivery date. The outcome of interest in this analysis was self-reported spontaneous abortion of less than 20 weeks' gestation. We examined subgroups of spontaneous abortions of less than 12 weeks' and 12-19 weeks' gestation to provide an indirect estimate of risk by likely frequency of chromosomal anomaly, a more common cause in early abortions (10). Pregnancies occurring when the woman was not living on the study farm and thus had unknown exposure status were excluded, as were pregnancies for which the study husband was not likely the father.
We pooled pesticide exposure information from the farm operator (the person responsible for the day-to-day operations of the farm, if different from the husband or wife), husband, and wife to construct a history of monthly agricultural and residential pesticide use. For each pesticide reported, we identified the active ingredients and uses using a database of registered pesticide products in Canada. Where possible, we categorized the active ingredients into chemical families. We divided all pesticides reported into four major classes of use: herbicides, insecticides, fungicides, and miscellaneous others (including those that could not be classified). We identified the active ingredients and chemical families that were most frequently used on the farms in the study, as well as those most likely to have adverse reproductive effects according to the literature. This categorization produced 17 pesticide unit variables that we examined in this study (Table 1).
Because only couples living on the farm were eligible for the study, the exposure assessment in this analysis was intended to capture potential occupational (direct) and residential (indirect) exposures. Because indirect exposures were possible, we could not completely separate the exposure statuses of the men and women. Most pesticide applications were done by the husband, with only 20% of the wives reporting handling of farm pesticides. No other information was available to validate the exposure assessments; however, we used both open-ended and checklist questions to obtain as complete a recall as possible.
We merged reproductive and pesticide exposure history data to create pesticide unit variables for months preceding and during each pregnancy. Exposure to pesticides was analyzed for two windows: preconception, the 4-month period from 3 months before conception to the calendar month of conception (consistent with potential sperm-mediated effects); and postconception, the 3-month period from the first calendar month after conception to the end of the first trimester (consistent with a fetotoxic effect). Exposures that occurred after a pregnancy loss but within the period of interest (i.e., first trimester) were not considered in assessing exposure status. We also created pregnancy-specific variables for all other time-related factors (parental age, smoking, farm activities, and alcohol and caffeine intake).
Statistical Analysis
We calculated crude odds ratios (ORs) using logistic regression for each combination of pesticide unit, exposure window, and gestational age at abortion category. Because no strong confounders were evident in previous analyses of these data (6) and our sample size was limited, we did not estimate adjusted risks. Nonexposed pregnancies were those not exposed during the time window to the pesticide unit of interest.
To assess the importance of the timing of exposure to the risk of spontaneous abortion, we compared preconception exposures to postconception in a combined model where preconception exposures were coded 1 and postconception exposures were the referent. Pregnancies exposed to a pesticide unit in both windows were excluded from this analysis. Similarly, we used an indicator to distinguish early (< 12 weeks' gestation) and late (12-19 weeks' gestation) fetal age at abortion to identify the major target site for pesticide toxicity (embryo or fetus). In this latter model, which analyzed only spontaneous abortions, we used the 12-19 weeks' gestational age abortions as the referent group.
