Objectives: To examine the relationship between occupational exposures and spontaneous abortion in female veterinarians.
Methods: The Health Risks of Australian Veterinarians project (HRAV) was a questionnaire-based survey of all graduates from Australian veterinary schools from 1960 to 2000. Of 5748 eligible veterinarians sent the questionnaires, 2800 replied including 1197 females (42.8%). The response rate was 59% of women veterinarians eligible to participate. The pregnancy of women was defined as the unit of analysis. We restricted analyses to pregnancies of those women who reported being employed when the pregnancy began and were working only in clinical practice. Of 1355 pregnancies, 940 were eligible for the final analysis. Self-reported occupational exposures to anaesthetic gases, x rays, pesticides and long working hours in relation to spontaneous abortion were examined.
Results: In a multiple logistic regression controlling for 12 potential confounders, there was a more than twofold significant increase (OR 2.49, 95% CI 1.02 to 6.04) in the risk of spontaneous abortion in women exposed to unscavenged anaesthetic gases for ⩾1 h per week. Veterinarians who reported performing more than five radiographic examinations per week had a statistically significant elevated risk of spontaneous abortion compared to those who performed five or less (OR 1.82, 95% CI 1.17 to 2.82). There was also approximately a twofold significant increased risk of spontaneous abortion in women who used pesticides at work (OR 1.88, 95% CI 1.18 to 3.00).
Conclusion: Female veterinarians, particularly those of childbearing age, should be fully informed of the possible reproductive effects of unscavenged anaesthetic gases, ionising radiation and pesticide exposure and reduce their exposure by using protective devices when they are planning to become pregnant and during pregnancy.
Statistics from Altmetric.com
Spontaneous abortion has been linked to occupational exposure to anaesthetic gases,1–4 radiation4 5 and pesticides6 7 during pregnancy. Veterinarians are exposed to these hazards and the increasing percentage of women employed in veterinary practice during their reproductive years has raised concerns about the possible adverse effects of occupational exposures for the embryo and fetus.
In a recent cross-sectional survey of the same cohort population as this survey (HRAV project), we addressed the prevalence of potentially harmful occupational exposures including radiation, anaesthetic gas and pesticides and showed that there is considerable variability of exposure within the profession.8 Anaesthetic gas exposure was more prevalent in small (92%) and mixed (94%) animal practices compared with large animal practices (42%), but 22% of women in small animal practices and 34% of women in mixed animal practices did not have waste anaesthetic gas scavengers. Exposure to pesticides was more common in mixed animal practices (54%) and small animal practices (47%) compared with large animal practices (17%). x Ray exposure was more prevalent in small and mixed animal practices (90%) compared with large animal practices (37%). While taking x rays, 56% of women reported physically restraining animals, and only one in five used film holders and lead screens. The results of exposure data from this survey were similar to the results of American surveys.9–12
Data on the reproductive effects of exposure to anaesthetic gases and radiation among veterinary personnel were published in the USA in 1987–1996.4 5 10 No association was found between exposure to anaesthetic gases and spontaneous abortion in female veterinarians in those studies,5 10 but an association between the use of diagnostic x rays in veterinary practice and an increased risk of spontaneous abortion was reported.4 5 The authors state that the lack of an association between exposure to anaesthetic gases and spontaneous abortion should not be interpreted as evidence that it does not exist but rather that it is likely to have been a reflection of study limitations.10 They used the pregnancies of both unemployed women and employed non-exposed women to form the reference group due to the small size of the employed non-exposed group. They recommended further study of spontaneous abortions among female veterinarians, with intensive efforts directed at more refined ascertainment of exposures and stricter criteria for the reference group.10
No recently published study has evaluated the association of occupational exposures with spontaneous abortion in veterinarians. The aim of the present study was to examine the relationship between occupational exposures and spontaneous abortion in female veterinarians. The present study has addressed some of the limitations of previous surveys.
The Health Risks of Australian Veterinarians study (HRAV) was approved by the Human Research Ethics Committee of The University of Western Australia. This study was the first complete national survey of the health aspects of veterinary practice in Australia.8 13–15 Of the potential respondents (n = 7928), an estimated 27% were ineligible to participate because their addresses were wrong, they were deceased or they were living abroad. Of 5748 eligible veterinarians who were sent the questionnaires (73% of the entire cohort), 2800 replied (49%). Just under half the respondents (42.8%) were females (n = 1197), representing 59% of women veterinarians eligible to participate.
The mailed self-administered questionnaire contained items about demographic details, a section on all veterinary jobs held for more than 6 months since graduation, smoking and reproductive history for women. Information obtained for each job included start date, end date, job type, practice type and work hours, number of x ray films taken per week, restraint of animals during x rays, number of hours doing surgery per week, use of scavenger equipment for anaesthetic gases (yes, no, don’t know, not applicable), driving (number of hours per week) and use of pesticides at work (daily, weekly, rarely, never).
Reproductive history was assessed by asking (for up to six pregnancies) the number of months trying to become pregnant, the year the pregnancy ended, the number of weeks the woman was pregnant, whether it was a single, twin or multiple birth, and the sex of each child. Reproductive outcomes were assessed by using information about live births, stillbirths, terminations (induced abortions), miscarriages (spontaneous abortions) and birth defects. The outcome of spontaneous abortion was defined as spontaneous fetal death prior to 20 weeks of gestation.
We reorganised the structure of the data file such that pregnancy became the unit of observation. Work history during pregnancy was assessed by cross-referencing the date of the end of the woman’s pregnancy to work history data for the year prior to the end pregnancy date. We therefore evaluated exposure for each pregnancy from a few months before conception to the end of the pregnancy.
We excluded the following pregnancies: those current at the time of the survey (n = 23), those that commenced before graduation from veterinary college (n = 48), twin pregnancies (n = 14), those where the mother had not worked up to the time of conception (n = 163) and those where the mother was not employed exclusively as a clinical practice practitioner at the time of conception (n = 167). Therefore, 940 pregnancies (of 1355 pregnancies in the file) in 442 women were eligible for analysis.
Crude risks of spontaneous abortion overall, crude relative risks and adjusted odds ratios of spontaneous abortion by clinical practice type and occupational exposures were estimated. To estimate the crude risk of spontaneous abortion, the number of spontaneous abortions referred to the total number of pregnancies (ie, births+spontaneous abortions+induced abortions).16 Crude relative risks and confidence intervals were calculated.17
An unconditional multiple logistic regression procedure in the Stata v 9 software package was adopted to calculate the adjusted odds ratio of spontaneous abortion. All pregnancies were considered for the analysis and we used the Huber-White sandwich estimator of variance to take into account the potential autocorrelation created by multiple pregnancies in a single woman.18 19
The following occupational variables were examined as exposures: working hours (number of hours per week), x ray exposure (⩽5 films/week, >5 films/week), pesticides (never or rarely, at least once a week), animal restraint (never or rarely, at least once a week), surgery (none, surgery in the presence of a gas scavenger, surgery in the absence of a gas scavenger) and driving (yes, no). The following variables were treated as potential confounders: practice type (small, mixed, large), years in the job (<2 years, ⩾2 years), maternal age (⩽35 years, >35 years), graduation year, university of graduation, smoking (have you ever smoked?), age at the time of the survey and birth order. Pregnancy history was not included in the model.
The associations of potential confounders with exposure variables were assessed by cross-tabulations, χ2 tests and the Mantel-Haenzel adjusted odds ratio. If the difference between the crude odds ratio and the adjusted Mantel-Haenzel odds ratios was less than 10%, the covariate was not considered a confounder. After confirming the confounders, an unconditional multivariate logistic model was implemented, adding each of the variables previously selected, including those that changed the odds ratio by 10% or more. The associations of spontaneous abortions with occupational exposures (anaesthetic gases, long working hours, x rays and pesticides) were assessed adjusting for practice type, years in the job, maternal age, graduation year, smoking, birth order, restraint of animals and driving.
The analysis was repeated for the subgroup of pregnancies reported by veterinarians employed exclusively in small animal practices, for first pregnancies and for those who graduated between 1980 and 2000. Interaction effects of factors related to spontaneous abortions were tested in Stata, but none contributed significantly to the model.
We also used a Cox regression including both parametric and semiparametric models and performed a test of the constant proportional hazards assumption based on Schoenfeld and scaled Schoenfeld residuals to evaluate the possibility of using a Cox model for this study and to compare the results with logistic models.
Respondents were born between 1938 and 1976 with a median age of 39.8 years in 2002 and 95% were >30 years old. Nearly 86% of respondents were born in Australia and 6% were born in the UK. Approximately 72% of respondents had graduated after 1980 with a median graduation year of 1984 and a median graduation age of 23 years.
Of the 940 pregnancies, 764 (81.3%) resulted in a live birth, six (0.6%) were stillbirths, 146 (15.5%) ended in spontaneous abortion and 24 (2.6%) were terminations of pregnancy. All of the live births in the study cohort were singletons and the percentages of male and female infants were approximately the same.
The overall crude risk of spontaneous abortion in women in veterinary clinical practice was 16% (table 1). The risk of spontaneous abortion was highest in women who graduated in Queensland, had a maternal age of >35 years, had one or two children and had more than two pregnancies (table 1).
The crude relative risk, compared to those unexposed to that specific factor, was non-significantly higher in veterinarians who worked in small animal practices compared with those in large animal practices (table 2). There was no excess risk of spontaneous abortions in women working for 2 years or more in clinical practice compared with those who had worked for less than 2 years. Long working hours were associated with only a slight excess risk of spontaneous abortion. Risks were highest in veterinarians who undertook more surgery, especially in the absence of a gas scavenger, and in veterinarians who used pesticides at work. There were no excess risks of spontaneous abortions in women who were exposed to radiation, restrained animals or drove more.
A multiple logistic regression analysis (table 3) showed that after adjusting for other variables, there was a more than twofold significant increase in the risk of spontaneous abortion in women exposed to unscavenged gases for ⩾1 h per week compared with women who did no surgery. When we examined the variable surgery/gas scavenger in two categories, the risk of spontaneous abortion for veterinarians who carried out surgery in the absence of gas scavengers increased (OR 2.82, 95% CI 1.82 to 4.37) compared with the reference group (veterinarians who had never done surgery and those who had done surgery in the presence of a gas scavenger) (data not shown in table 3). There was also an approximately twofold significant increased risk of spontaneous abortion in women who reported performing more than five radiographic examinations per week compared with those who took five or fewer x rays films per week (OR 1.82, 95% CI 1.17 to 2.82). There was also a slight significant risk of spontaneous abortion when we examined exposure to x rays as a continuous variable (OR 1.03, 95% CI 1.00 to 1.05) (data not shown in table 3). We also examined the effect of long working hours on spontaneous abortion, but there was no significant change in results. There was a slight non-significant increased risk of spontaneous abortion for women who worked more than 45 h per week (OR 1.09, 95% CI 0.69 to 1.71) compared to those who worked for 45 h/week or less (data not shown in table 3).
In a separate multivariate analysis restricted to veterinarians employed exclusively in small animal practices, the adjusted odds ratio increased significantly for occupational exposures to unscavenged anaesthetic gases, radiation and pesticides. We also restricted the analysis to those who graduated between 1980 and 2000 and for the first pregnancy only and found similar results to those for the whole cohort (table 3).
There were also associations between spontaneous and non-occupational factors such as maternal age >35 years (OR 2.66, 95% CI 1.66 to 4.27, p = 0.001) and graduation year (OR 1.04, 95% CI 1.00 to 2.26).
Using a Cox regression including both parametric and semiparametric models, we found results very similar to the logistic models. For example, the key results from the semiparametric Cox model for key variables such as “surgery in the absence of scavenger gases” (HR 2.49, CI 1.70 to 3.65), “taking more than five x rays” (HR 1.64, 95% CI 1.14 to 2.36) and “pesticides use” (HR 1.73, 95% CI 1.13 to 2.66) adjusted for other variables were very similar. However, after performing a test of the constant proportional hazards assumption based on Schoenfeld and scaled Schoenfeld residuals and another post-estimation command for the Cox model using Stata plots, we found evidence that the model violates the proportional hazards assumption. The results from plots indicated that the lines for the first two variables were not parallel. We also examined various solutions for the breach of proportional hazards (eg, to include the time-dependent variable for the non-proportional predictors or to stratify on the non-proportional predictors). For example, when we used stratification on the non-proportional predictors, the parameter estimates were not the same for each level of the variables, which further indicated that the variables really were non-proportional. Therefore, the results of logistic models only are presented in this paper.
The overall proportion of spontaneous abortions (number of spontaneous abortions/number of pregnancies) was 146 in 940 pregnancies (15.5%), which is the same as for all Australian women.20 This rate was also comparable with that reported in occupational risk studies in US female veterinarians (11.1% and 15.1% in separate studies),5 10 French nurses (12.3%)3 and US female dentists (7%).1 However, it has been argued that variations in the frequency and patterns of induced abortion will influence the measures usually employed to estimate the rate of spontaneous abortion.21 Therefore, the ratio used in the present study may underestimate the true spontaneous abortion rate because if induced abortions had not been voluntarily terminated, a proportion of them would have been distributed to the numerator as spontaneous abortions as well as to the denominator as births after 20 weeks, which would yield a higher true spontaneous abortion rate. Although a method of correction was proposed, it is not applicable in cohort studies with low proportions of induced abortions as here.
The overall risk of spontaneous abortion in women employed in any type of veterinary clinical practice was 16%, which exceeded the 11% reported in a cohort study of 2997 female graduates from US veterinary colleges between 1970 and 1980.10 The risks of spontaneous abortion for veterinarians did not vary significantly by clinical practice type, although the crude risk of spontaneous abortion was higher in small animal practice, in contrast to a 1996 US study that suggested a higher risk in large animal practice.10
There was a more than twofold significant increased risk of spontaneous abortion in women veterinarians who were exposed to unscavenged anaesthetic gases. The association between unscavenged gas and spontaneous abortion is consistent with a study of 1465 dental assistants whose most recent pregnancy was conceived while they were working full time.1 In that study, an elevation in risk of spontaneous abortion was seen in women who worked with nitrous oxide for ⩾3 h per week in offices not using gas scavenging equipment (RR 2.6, 95% CI 1.3 to 5.0), adjusted for age, smoking and number of amalgams prepared per week. A meta-analysis of 19 epidemiological studies of the association between maternal occupational exposure to anaesthetic gases and spontaneous abortion reported a weak to moderate association (RR 1.3 to 2.4).22 The overall relative risk of spontaneous abortion for those who were exposed to anaesthetic gases compared with those who were not exposed was 1.48 (95% CI 1.4 to 1.58) for all studies and a relative risk of 2.45 (95% CI 1.25 to 5.02) was found for the subgroup of veterinarians and veterinary assistants, which is consistent with our survey. Previous studies in veterinarians have found increased risks which have not always been statistically significant.4 10
The National Institute for Occupational Safety and Health23 recommends that no worker be exposed to 8 h time-weighted average concentrations of nitrous oxide greater than 25 ppm during anaesthetic administration and concentrations of halothane and methoxyflurane agents greater than 2 ppm for a period not to exceed 1 h. Among our participants, we found that 92% were exposed to waste anaesthetic gases and among those who were exposed, 46% did not use a waste anaesthetic gas scavenging system. However, we did not have information about what types of anaesthetic gases were used or intensities of exposure.
Ionising radiation was found to be associated with increased risk of spontaneous abortion in our study (OR 1.82, 95% CI 1.17 to 2.82). This relationship has been supported by a US study of 537 female veterinarians that reported a marginally elevated risk of spontaneous abortion in veterinarians who performed five or more radiographic examinations per week compared with those who took less than five films per week.5 Our finding was also consistent with that of Steele et al10 who reported an association between spontaneous abortion risk and job-related exposure to ionising radiation (adjusted OR 1.3, 95% CI 0.8 to 2.0) when analyses were restricted to small animal practitioners. Johnson et al4 in a case–control study of reproductive outcome in veterinary personnel reported a significant association between the use of diagnostic x rays in veterinary practice and increased spontaneous abortion risk in veterinarians (OR 2.24, 95% CI 0.99 to 5.09) and veterinary assistants (OR 2.45, 95% CI 1.10 to 5.46).
Occupational exposure to pesticides was also found to be significantly associated with spontaneous abortions (OR 1.8, 95% CI 1.18 to 3.00) in our study. This is not in accordance with the report by Steele et al10 who found no relationship, although several other studies have reported positive associations between occupational pesticide exposure and spontaneous abortions.7 24–26 Garry et al reported that personal use of pesticides including mixing, loading and pesticide application by the female spouses of applicators was a significant risk factor for fetal loss (OR 1.81, 95% CI 1.04 to 3.12).25 Savitz et al reported that the use of dithiocarbamate fungicides and pesticide combinations from different pesticide classes was associated with an increased risk of miscarriage.26 Our finding is also consistent with another recent report which observed a moderate increase in the risk of early and late abortions for preconception and postconception exposures to different classes of pesticides.24
This is the first study on veterinarians to evaluate the association between working hours and spontaneous abortion and no significant association was found. However, in a US cohort study of self-reported stress and reproductive health of female lawyers, Schenker et al27 reported that weekly job hours during the first trimester of pregnancy showed a strong independent association with the risk of spontaneous abortion. It is possible that we could not find an association between working hours and spontaneous abortion because we did not have information about job hours specifically in the first trimester of pregnancy when most miscarriages occur.
In the multiple logistic regression models, there was also an association between spontaneous abortion and non-occupational factors such as maternal age >35 years, which relationship with age was also found in a US cohort study of female veterinarians.28
There are several important limitations to this study, especially a possible selection bias due to a suboptimal response rate (59%). Previous studies have also reported similar response rates of 56%,29 60%30 and 69%.31 Although our response rate was relatively low, we included a large cohort of female veterinarians of all ages and from all types of clinical practice in Australia, suggesting that our study provided a reasonable representation of female veterinarians. In addition, although we could not determine the exposure and outcome status of non-responders, we endeavoured to identify the main reasons for non-participation from 195 veterinarians who acknowledged receipt of the questionnaires but refused to participate. The reasons included working for a postgraduate qualification (eg, in a research area or in teaching), being outside the profession for a long time, temporarily working overseas and being too busy or having no time to fill in the questionnaire, which reasons suggested that those not in clinical practice or veterinary work were less likely to participate.
We reduced the risk of selection bias and response bias (differential response rate) in exposed and unexposed women by restricting the analyses to homogeneous women who work only in clinical practice. The timing of exposure to the risk of spontaneous abortion and information on dose in this study were important in characterising the reproductive toxicity of anaesthetic gases, radiation and pesticides.
Information bias (using questionnaires rather than actual measurement) is another important limitation to this study. However, the detailed occupational exposure information collected from these highly educated women might have minimised this type of information bias. In addition, previous surveys have supported the use of exposure data from retrospective self-reports.31 A review of women’s reproductive health research32 based on four validation studies on exposure assessment, found little evidence for the existence of recall bias in reporting occupational exposure and assessed that if recall bias exists, its effects tend to be small. To reduce the effect of recall bias, we evaluated the association between occupational exposures and spontaneous abortion in recent graduates as well. Non-differential misclassification caused by inaccurate reporting of exposure was reported to be a greater problem.32 The use of an expert (eg, an industrial hygienist) in exposure ratings, questionnaires that address specific jobs and biological exposure markers has been suggested to improve data quality in exposure assessment and to avoid errors in using self-reports of exposure.32
With regard to the outcome data, Joffe33 found recall regarding gestational age to be accurate for 20 years or more after the pregnancy. However, miscarriage is an end point for which self-reporting can be inaccurate because of potential under-reporting, as one third of spontaneous abortions are estimated to occur before the second missed menstrual period.34 A study compared responses concerning pregnancy outcomes with data from medical records and found some errors regarding the week of pregnancy in which the spontaneous abortion occurred.35 Registry-based studies offer the advantage of using data which cannot be biased by respondents and the ascertainment of pregnancies is high.22 However, the ascertainment of spontaneous abortions is poor in registry studies as a very large proportion of spontaneous abortions do not lead to admission to hospital.36
We were able to control for 12 important variables in the analysis including maternal age, although we did not include pregnancy history in the model. Weinberg37 has argued that one should not adjust for history of pregnancy, such as previous spontaneous abortion, as these determinants may have been caused in part by exposure to anaesthetic gases or radiation and can bias results. There is also a possibility that a woman experiencing a previous spontaneous abortion might be more careful in subsequent pregnancies and avoid exposures to toxic agents.
We were also able to adjust for other occupational exposures such as ionising radiation and pesticides. Furthermore, we analysed exposure to anaesthetic gases modifying for safety practices such as the use of a gas scavenging system. However, we did not have sufficient information about radiation protection for each job in the questionnaire (job history section) to modify exposure for protection against radiation. A cross-sectional analysis from this survey for the use of radiation protection in the current job, indicated that, even though protective devices were usually available in the workplace, they were not always used.8 The majority of female veterinarians used lead aprons in their current job, but only one in five veterinarians used lead screen and film holders and one half of them did not use thyroid protectors or lead gloves.
A study by Lemasters and Pinney evaluated the effect of employment in a cohort of 1535 pregnancies and found that an employment effect can be confounded with exposure status.38 They indicated that this type of confounding can increase the apparent magnitude of exposure effect when one does not exist and suggested that the best approach for controlling this bias is to compare occupationally exposed pregnancies with employed non-exposed pregnancies. To overcome these difficulties we chose to study a homogeneous population with similar working conditions in a group of women working only in clinical practice and to compare unexposed pregnancies with exposed pregnancies. This approach has also been recommended by Steele and Wilkins in their 1996 study.10
However, we did not have information about factors related to the work environment of women exposed to anaesthetic gases such as occupational stress, standing, lifting heavy weights and other physical effort. These factors may be confounders and be associated with spontaneous abortion.36 39 40 Risk of spontaneous abortion is also reported to increase with alcohol consumption,28 41 use of coffee,41 maternal height,28 history of cancer,28 paternal age42 and paternal occupational exposures,43 but this information was not available in the survey. Another important limitation was that we did not have smoking data for each pregnancy and only had information about history of smoking for each woman.
We found a statistically significant more than twofold increased risk of spontaneous abortion in female veterinarians exposed to anaesthetic gases where these agents were delivered without a gas scavenging system compared with those who did not do surgery and were therefore not exposed. These results and their agreement with previous studies suggest that there may be a real risk in practices not using gas scavenging equipment.
A retrospective cohort study was conducted to explore the relationship between occupational exposures and spontaneous abortions in female veterinarians employed in clinical practice using self-reported data.
There was an increased risk of spontaneous abortion in female veterinarians exposed to anaesthetic gases where these agents were delivered without a gas scavenging system.
There was also an association between spontaneous abortion and exposure to radiation and pesticides.
Female veterinarians, particularly those of childbearing age, should be fully informed of the possible reproductive effects of ionising radiation, unscavenged anaesthetic gases and exposure to pesticides and should reduce their exposures by use of protective devices when they are planning to become pregnant and during pregnancy.
The results from this study will be useful for public health policy and occupational risk prevention.
We also found an association between spontaneous abortion and exposure to radiation and pesticides. The results from this survey and other surveys of veterinarians suggest that there may be a risk. Female veterinarians, particularly those of childbearing age, should be fully informed of the possible reproductive effects of ionising radiation, unscavenged anaesthetic gases and exposure to pesticides and should reduce their exposures by use of protective devices when they are planning to become pregnant and during pregnancy.
Our study indicated how addressing the limitations and weaknesses of analysis from previous work improves the ability to detect associations.
The results from this study will be useful for public health policy and occupational risk prevention.
The authors wish to thank alumni organisations, the Australian Veterinary Association and the Australian State Veterinary Registration Boards. The authors also wish to thank Professor Carol Bower, Associate Professors Andrew Vizard, Ian Robertson, David Morrison and Dr Lesley Day who assisted with the development of the survey.
Competing interests: None.
Funding: Funding was provided by the Cancer Foundation of Western Australia, University of Western Australia small research grants and the National Health and Medical Research Council of Australia.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.