This cohort of RNs who worked in BC between 1974 and 2000 showed a slightly increased but non-significant risk of all cancers if they had potential exposure to antineoplastic drugs. The risk of breast cancer was significantly elevated among those who had ever worked in oncology nursing units or for a cancer center, which showed a possible trend of increasing incidence with increasing years of work. However, this increase was elevated but not statistically significant in RNs classified as potentially exposed to antineoplastic drugs, based on a survey-based exposure assessment protocol. An increased risk of cancer of the rectum was observed in RNs determined to have had a probability of exposure to antineoplastic drugs. No increased risk of any other cancer type, including leukemia, was observed using either exposure assessment method.
It has been well established that nurses in oncology who handle antineoplastic drugs have increased frequencies of biomarkers of exposure, including chromosomal aberrations and sister chromatid exchanges [19–22]. However, to date, very few relevant epidemiological studies have estimated RNs' cancer risk related to exposure to antineoplastic drugs. A linkage study of Danish nurses who worked in oncology departments, preparing and administering antineoplastic drugs, reported a statistically significantly increased relative risk (RR = 10.65) for one site, leukemia, based on 2 cases . A nested case-control study of 59 breast cancer cases from an Icelandic cohort study of female nurses found that the risk estimates were highest (although not statistically significant and based on 7 cases) among those who had ever handled cytotoxic drugs (OR = 1.65, 95% CI = 0.53 - 5.17), after adjustment for year of birth, breast cancer in a first-degree relative, marital status, and nulliparity . Our finding of a significantly increased risk of breast cancer among RNs who had ever worked in oncology or a cancer center provides some additional evidence of this association and supports the relevance of further investigating occupational risk factors for nurses. In future research about nursing and cancer incidence, consideration should also be given to the possible influence of mammography screening rates. It is possible that oncology nurses, who directly care for women with breast cancer, are predisposed to mammography screening participation and thus are subject to an over-detection bias. It is recognized that some breast cancers, if undetected and thus not treated, would not progress and that some women are "over-diagnosed" [23, 24].
The mutagenic effects of antineoplastic drugs may increase the risk of congenital anomalies and stillbirth among exposed nurses. Of three case-control studies, one showed an odds ratio of 4.7 (95% CI = 1.2 - 18.1) for congenital anomalies among those exposed to antineoplastic drugs at least once a week , one showed a relative risk of 1.7 (95% CI = 1.0 - 2.8) for spontaneous abortions among occupationally exposed nurses , and a third showed an odds ratio of 2.3 (95% CI = 1.2 - 4.4) for fetal loss among nurses exposed during their first trimester of pregnancy . A cross-sectional study observed 8 congenital anomalies among offspring of 152 physicians and nurses who administered antineoplastic drugs during pregnancy (4.05 expected; p = 0.05) . More recent studies have shown few reproductive risks from antineoplastic drug exposure. A questionnaire-based study found log-linear odds ratios of 1.20 (95% CI = 0.98 - 1.47) for stillbirth and 0.97 (95% CI = 0.86 - 1.09) for congenital anomalies among the offspring of nurses exposed to antineoplastic drugs . Two case-control studies found odds ratios of 1.02 (95% CI = 0.47 - 2.06) and 2.2 (95% CI = 0.7 - 7.2) for congenital anomalies among the offspring of nurses who handled antineoplastic drugs, compared with nurses without exposure [12, 28]. A recent meta-analysis of four studies of nurses' exposure to antineoplastic drugs estimated an odds ratio for congenital anomaly incidence of 1.64 (95% CI = 0.91 - 2.94) .
In our study, the risk of congenital anomalies of the eye was significantly increased in the offspring of RNs who had ever worked in oncology nursing units or for cancer centers during pregnancy (method 1), although this estimate was based on only 3 cases. The risks for other congenital anomalies among this employment group were also notably elevated, but not statistically significantly and were limited by the low frequency of cases. The comparatively higher risk of all congenital anomalies in the employment group may suggest an oncology/cancer agency-specific effect that is perhaps diluted in the broader department-based exposure assessments. The risks of stillbirth, low birth weight, and prematurity were not significantly increased among RNs potentially exposed to antineoplastic drugs, assessed through either their employment in oncology or estimated weighted duration of exposure.
According to the BC Cancer Agency Benefit Drug List, there was an increase in the number of different antineoplastic drugs used, from 66 in 1986 to 88 in 2000. Standardized safety practices for handling antineoplastic drugs and other hazardous materials were established in 1985. However, a recent review of methods used to monitor occupational exposure to antineoplastic drugs concluded that, despite the introduction of safety guidelines and protective measures, health-care workers can still be exposed . Nonetheless, the adoption of these practices may partly account for the predominantly null or non-significant risks observed in this and other recent studies [15, 12, 28].
Whereas many previous studies were questionnaire-based and were, therefore, susceptible to exposure recall and selection biases, our record linkage used comprehensive recruitment of an employment group based on registry-derived data and allowed for complete ascertainment of outcomes according to standardized medical reporting systems. Furthermore, the historical prospective design of the study meant that exposure variables were assessed independently from health outcomes.
Despite our large sample size, limitations of our study arise from limited statistical power associated with the small numbers of cases and the low prevalence of exposure, possibly leading to missed associations, as well as the testing of multiple hypotheses, such that a statistically significant association may have occurred by chance alone.
As is common in historical cohort studies, we had no information about potential confounding factors related to lifestyle; however, all comparisons were within the nursing profession, a narrow, well educated, socioeconomic stratum. Nurses have been documented to have healthier habits than the population as a whole, but there have been temporal patterns, for example strong declines in smoking rates among US nurses in the last 30 years (33% smokers in 1976 vs. 8.4% in 2002/2003 ). There also were temporal patterns in antineoplastic drug use in the British Columbia healthcare agencies during the study period (e.g., hand mixing of the drugs was completed in 45% of the facilities in the 1970 s and 1980 s, but only 8% in the 1990 s and later). If smoking and antineoplastic drug exposures were correlated and both related to the outcomes of interest, there could be uncontrolled confounding in the study results. In a recent study of 1,147 live births among nurses, controlling for parity, smoking, alcohol, coffee, multivitamin, and folic acid intake did not materially change the effect estimates for congenital anomalies .
We were not able to adjust for other chemical exposures in the RNs' workplaces or other environments. Unfortunately, there is little if any data available about the presence and usage of such chemicals. We are left to assume that the RNs exposed to antineoplastic drugs were similar to the unexposed RNs with respect to their opportunity to having been exposed to other toxins.
The exposure assessment methods used in the study were crude and may have resulted in misclassification of exposure. For example, in method 1, most nurses working in oncology departments were outside of the cancer centers and were not identifiable in the registry, prior to 1996. This likely led to underestimation of the number of exposed RNs but would have been highly specific. The survey (method 2) identified relevant departments in general hospitals and treatment centers that administered antineoplastic drugs and ascribed individual nurses therein with a probability of exposure. Within these departments, some of the nurses ascribed a probability of exposure may not have been exposed and, within other departments, some of the nurses may have been missed because of poor information recall. This likely resulted in misclassification, which would have biased the effect estimate toward the null. Nonetheless, the method was novel in its attempt to more precisely characterize exposure estimates of a cohort of nurses through the use of relevant questions about employment history that were simple enough to assume sufficient recall for an approximation of exposure; that is, the classification was broadly based.
The premise underlying method 1 was that the oncology departments, as cancer treatment facilities by definition, would have consistently used antineoplastic drugs, whereas other departments (in general hospitals and other agencies) may not have used them, or when they did, perhaps they did so with less regularity. This method more definitively assigned exposure through the use of reliable employment records, limited the exposure group, and introduced some uncertainty about the unexposed group (it may have contained nurses from nursing departments with possible exposure).
Consequently, these limitations of the assessments of exposure may have affected their accuracy and could possibly have resulted in an underestimation of the risks associated with exposure to antineoplastic drugs or have distorted the shape of the dose-response relationship. For instance, the relative risk found for breast cancer using method 2, as opposed to method 1, was elevated, but not statistically significant.