Risk of spontaneous abortion among nurses handling antineoplastic drugs.

Risk of spontaneous abortion among nurses handling antineoplastic drugs. 1990;16:102-7. The purpose of the study was to evaluate the frequency of spontaneous abortion in two groups of women. One group consisted of women regularly involved in the preparation of cancer chemotherapy perfusions and therefore considered to be exposed to cytostatic agents; the other consisted of women not occupa tionally exposed to such agents. The study was carried out in four French hospitals. Four hundred and sixty-six women were involved; 534 pregnancies were described in which 139 were exposed and 357 were unexposed. The results showed that the frequency of spontaneous abortion was 26 % for the exposed pregnancies and 15 070 in the unexposed ones (odds ratio 2.0). These results do not seem to be due either to the classic risk factors of spontaneous abortion (age, cigarette consumption during pregnancy, preg nancy order) (adjusted odds ratio I. 7) or to possible errors concerning the retrospective evaluation of prior gynecologic and obstetric history. significant increase in urine mutagenicity in exposed subjects as compared with unexposed subjects Other studies ha ve that work preparing cytostatic perfusions can lead to an increase in the average number of sis ter chromatid exchanges and to an increase in the fre qu ency of chromosome ano malies in lymphocytes 14). our knowledge four studies provide some epidemiologic information on the relationship between cytostatic exposure and reproductive hazards

Cytostatics are chemical agents which, because of their antimitotic action, are often used in cancer chemotherapy. Their mutagenicity, ca rcinogenicity, and teratogeni cit y have been amply documented by in vitro and in vivo experiments (1,2). Some particular agents, esp eciall y me t ho trexate, have been shown to be teratogenic in pr eg nant women being treated by cancer chemotherapy (3,4). Moreover the re lat ionship between the mec ha nisms of mutagenesis and teratogenesis have been emphasized by several authors (5,6) .
Several stud ies concerned with occupational exposure to these agents have shown a significant increase in urine m utagenicity in exposed subjects as compared with unexposed subjects (7-1 1). Other st ud ies ha ve demon strated that work preparing cyto static perfusion s can lead to an increase in the average number of sister chromatid excha nges and to an increa se in the frequ ency of chro mosome a no m a lies in lymphocytes (12)(13)(14). To our knowledge fo ur st udies provide some epidemiologic information on the relationship between cyto static exposure and reproductive hazards (15)(16)(17)(18).
McDonald et a l, in the frame of a ca se-referent survey on spo n ta neo us abortions in a la rge ge neral populat ion , did not find a significant relationship between cytostatic exposure and spontaneous abortion (15). T aski nen et a l observed an increased risk among potentially exp osed pregna ncies, which was however no t sta tistically significa nt, in a ca se-referent study in the p harmaceutical industry (16). T he ca se-referent study o f Hemminki et a l (17) was performed wit h in a cohort of hospital nurses and did not show any increased risk o f spontaneous a bortion among nurses exposed to cyto st atics. H o wever , re stri cting their population to specia lized departments where cy tos ta tic exposure can be thought to be higher, Hemminki et al (17) sho wed a sign ificantly higher risk of co ngenita l malformations in relation to this expo sure. Finally, Selevan et al (18), who perform ed a ca se-referent study wit hin a cohort o f n urses wo rking in oncolog ic depar tm ents, o bserved a two fol d inc rease in the risk of spontaneous a bo rtio n among exposed nurses.
The present stu dy was aimed at evalua tin g th e po ssible mo d ifi cations of parameters characteristic o f the reproductive funct ion in relat ion to the ha ndl ing of cytos tatic drugs. The presented result s concern the frequency of spo nta neo us abor tion in two gro u ps of nurses, one in whi ch th e nu rses prepared ca ncer chemotherapy perfusion s a nd th e other one in which they did not.

Subjects and methods
The frequency of spo nta neo us abortion was co m pa red between th e pr evious pregnancies of a gro up of nurses handling perfusion s in cancer treatment unit s and those o f a gro up o f nur ses who were no t so expo sed . The dat a were co llected by inter view from May 1985 to Ma y 1986 in three Fren ch hospitals and in a large center for ca ncer treatment.

Choice of medical departments 'and eligibility of the subjects
T he ca ncer units selected were those in which the number of cyto sta tic ph ials distrib uted a nnually by the hospital pharm ac y co rr espo nded to a minim um of ten cytostat ic perfu sions prep a red per week a nd per nurse. Cons ulti ng medi cal , ca rdiolog ic, e ndocri no log ic, a nd genera l med ica l units were chos en as the reference units. All the female daytime nursing sta ff (hea d nurses , registered nurses , practical nur ses) aged 45 years or less an d wor king in one of the selected un its were ask ed to pa rticipate .

Co ffection of inf ormation
In each center , interviews were ca rried out by a n investigating ph ysician . The qu estio n naire consisted of the fo llowing three pa rt s: (i) general char act eristi cs o f the subject; (ii) des cription of pregnancies a nd classica l risk fac to rs invo lved in spo nta neo us a bo rtio ns (pregnan cy order, age of the mot her at pregnan cy, a nd cigar ette co nsum pt io n); a nd (iii) job held by th e subjec t at th e time of the study a nd previou s jo bs , includi ng those o uts ide th e medi ca l field . T o be able to eva lua te the relia bility of th e spo nta neous a bo rtions reported, the subjects were asked to define th e method used to diagnose the pregna ncy (lab oratory exam ination , pharmaceut ical test , physician examinat io n o r late menstrua l peri od), the num be r o f wee ks of gesta tio n a t abortio n, a nd the oc curren ce o f hosp italizati on o r curettage. A spo nta neo us a bo rt io n was defin ed as a pregnan cy which ended before 28 weeks o f ges ta tio n . In or der to avo id th e underreporting o f ind uced a bo rtion s, we preferred to exclude them fro m the interview. Th erefo re , the freq uency of sponta neous a bort ion was defined as the rate of the num ber o f sponta neo us abortio ns over the number of pr egn a ncies, excluding indu ced a bo rtions.
T he q uestionnaire enabled us to date pr ecisely occupat ion a l exposur e to cytostat ics dr ugs, a nes thetic agen ts, a nd formaldehyde. An y pre gn a ncy occ urr ing a fter o r during a period of exposure to cytos ta tics was co nside red to be " exposed. " Sim ilarl y, pregnancies occurri ng after expos ure to fo rmaldeh yde or a nes thetic age nts were co nsidered to be exposed to th e respective substa nce. Exposure to cytos ta tics was fur ther describ ed ac co rding to th e fr equ en cy (regular, occasio na l, rar e) and total durat ion o f expos ure . At th e time of the study, no special protective eq uipme nt, such as a fume hood , was in op erati o n . Furth ermor e, the use o f personal pro tective equipmen t (surg ical gloves a nd masks) was not un ifo rm a mong the nurses, co ncern ed o nly a few drugs (methotr exate), a nd was a recen t co nce rn fo r the nurses.

Stat istical analysis
In all the sta tistical analyses the different pregn an cies o f the sa me wo ma n were co nsidered to be sta tistica lly ind ependent , even though th is is no t strict ly true. Statisti cal analysis was th us also carried out so lely for first pre gnancies to take into accou nt the va ria bilit y in the risk of spo nt a neo us abo rtio n between women .

Results
Of the 535 wome n eligible fo r th e study, 39 co uld not be co ntac ted du e to temporar y lea ve of ab sence (sick lea ve, materni ty leave, hol ida y, pro fession al train ing), and 30 re fused to participate. The study therefore invo lved 466 women (rate of par ticipatio n 87 % ), 271 belon gin g to units in which cyt os tatics were handled and 195 to control units. The gene ra l charac teri stics o f th e women who p a rticipated are presented in ta ble I, and those of th e 534 reported pregnancies are sho wn in tabl e 2.
Fo r 38 pregn a ncies there was no availabl e information to enab le the determinati o n of cytostatic expo sure . Among the 496 remaining pregn a ncies, fo r which exposure info rma tio n was a vaila ble, t here wa s no ex- Becau se of the sma ll numbers, it was impossible to analyze the effects of exposu re before and during pregnancy separately. We therefore combined a ll the exposed pregnancies (N = 139).
The freque ncy of spontaneous abortion was significantly higher in the exposed group (25.9 %) than in the unexposed one (15.1 0/0) (P<O.OI). This result was also observed when the analysis was limited to the first pregnancy of each woma n (table 3).
A comparison of the reported gestational ages at the time of the abortion did no t revea l any significant difference between the exposed and unexposed groups .
The relationships between th e occurrence of spontaneous abortion an d well-known risk factors are described in table 4. All of them were found to be statistically significant in the present sample. The occurrence of some of these factors was significantly different between the exposed and unexposed groups (table 5). Adjustment for smoking during pregnancy on one hand, and for age at pregnancy on the other, did not explain the observed relationship between exposure and the frequency of spontaneous abortion. Furthermore this relationship was still stati stically significant when restricted to the first pregnancy of each woman . Adjustment for all these confounders (age at pregnancy, pregnancy order, smoking during pregnancy) simultaneously in a logistic regression decreased the odd s ratio of the association from 2.0 [95 % confidence interval (95 % CI) 1.2-3.1] to 1.7 (95 % CI 1.0-2.8) without modifying the statistical significance .
The pregnancies could have been exposed to formaldehyde or to anesthetic agents. A pregnancy was considered exposed to formaldehyde if it occurred during or after expos ure , defined as daily or freq uent, to formo l in connection with th e disinfection of rooms and /or equipment. P revious exposure was th us fo un d to be negative for 383 pregnancies and positive for 50, the forma ldehyde exposure of the remaining 63 being unknown . When the 113 pregnancies for which formaldehyde exposure was positive or unknown were excluded , the results concerning cytostatics were not modified (table 3). Similarly, excluding the 24 pregnancies with positive (N = 10) or unkn own (N = 14) previous exposure to anesthetic agents from the analy sis did not modify the results either (table 3).
Fina lly, in order to imp rove the social homogeneity of the sample, the analysis was restricted to pregnancies of head nurses and registered nurses only (practica l nurses being excluded). The results again demonstrated a statistically significant excess of spontaneo us abortions among the exposed pregnancies in comparison with the rate among the unexposed ones .
In order to reduce the likelihood of bias due to an underreporting of spontaneous abortion, we excluded the ten spontaneous abortions corresponding to pregnancies diagnosed solely by the women themselves on the basis of late menstrual periods and spontaneous abortions which occurred after less than eight weeks without a menstrual period from the analysis. In both cases the frequency of spontaneous abortion was twice as high among the exposed pregnancies as among the unexposed ones (P<O.OI). The same observation was made after the exclusion of extrauterine pregnancies. Finally, all these results were substantiated when only first pregnancies were assessed alone.
The frequency of spontaneous abortions did not displaya statistically significant association with duration of exposure to cytostatic drugs prior to pregnancy (28, 24, and 24 010 of the spontaneous abortions for 1-24, 25-60, and> 60 months of exposure prior to pregnancy, respectively). Excluding the 24 pregnancies exposed only prior to pregnancy did not modify the results (rate ratio 1.7, 95 010 CI 1.1-2.5). Such was also the case for the relationship between spontaneous abortions and the frequency of exposure to cytostatic drugs during pregnancy.

Discussion
The purpose of this study was to compare the frequency of spontaneous abortion in relation to the handling of cytostatics. A pregnancy was defined as being exposed if the woman had, prior to or during pregnancy, given perfusions in departments where the handling of cytostatics was frequent at the time of the study (18 perfusions per week per nurse prepared on the average).
Spontaneous abortions were approximately twice as frequent among the exposed pregnancies than among the unexposed ones. McDonald's population-based case-referent study did not show any significant relationship between cytostatic exposure and spontaneous abortion (15). It may be thought however that McDonald's definition of cytostatic exposure could only be much broader than the one adopted in the present study, which concerned oncologic departments only. Taskinen et aI's study (16) of women working in the pharmaceutical industry showed an odds ratio of 2.8 in relation with cytostatic exposure. The number of exposed subjects in their study was small however, and this association was not statistically significant. Hemminki et aI's case-referent study (17) on spontaneous abortion was performed in a population of hospital nurses, and no significant association with cytostatic exposure was observed. However, restricting their population to nurses working in departments with frequent exposure to cytostatics, this group of researchers found a significant excess of malformed children among the exposed nurses. Finally, in a casereferent study performed within a population of nurses working in oncology departments, Selevan et al (18) observed an increase in the frequency of handling cytostatics in the group of subjects who had experienced spontaneous abortions in comparison with a group of subjects who had given birth. The results of their study, which is the most closely comparable to the present one as far as the definition of exposure is concerned, showed a risk of spontaneous abortion that was approximately twice as high for the exposed women as in our study.

Selection bias
In this study it was not technically possible to include subjects having previously worked in the departments in question and having since left the hospital. Axelsson (19) observed that women leave work more frequently after giving birth than after a spontaneous abortion (19). The rate of spontaneous abortion could therefore have been overestimated among the women remaining in the departments. It is however very unlikely in the present study that the influence of the outcome of pregnancy on subsequent employment differed for the exposed and unexposed women. It can therefore be thought that, if the rate of spontaneous abortion is overestimated without differential biases, the true relative risk will be in fact underestimated.
The study population was socially homogeneous, as only nursing staff were included (head nurses, registered nurses, practical nurses). Moreover restricting the analysis to the head and registered nurses did not modify the findings.

Information bias
Particular attention was given to the problem of information bias. The questionnaires were completed during interviews. The study was presented to each department concerned as a study whose purpose was to assess the reproductive characteristics in hospital personnel without focusing on any particular risk factors. Indeed, departments as different as consulting medical units and care units in oncology or cardiology were included. Finally, the questionnaire separately addressed the description of previous pregnancies on one hand and the work history of the subject on the other. With these precautions, a normal value for the frequency of spontaneous abortion among unexposed pregnancies was observed (15.1010). Time since pregnancy was significantly different for the exposed and unexposed pregnancies. However it was not significantly related to the frequency of spontaneous abortion when considered simultaneously with cytostatic exposure and maternal age in a logistic regression.
In addition excluding spontaneous abortions corresponding to pregnancies diagnosed only by the women themselves following a late menstrual period from the analysis did not modify our results. Con-versely, taking the somewhat restrictive hypothesis that all women are aware of their pregnancy only after eight weeks without a menstrual period , we also restricted our analysis to spontaneous abortions occurring after eight weeks or more. Once again the results were not modified. We therefore feel that it was unlikely that the increased frequency of spontaneous abortions among the exposed pregnancies resulted from information bias .

Confounders
The frequenc y of spontaneous abortions in this study was significantly correlated with the classical risk factors of age at pregnancy (20,21), smoking during pregnancy (22), and pregnanc y order. The variations in the spontaneous abortion rate with pregnancy order has been analyzed in detail (23) and involves both direct and selection phenomena. Adjustment for these three factors either separately or simultaneously in a logistic regression did not modify the results of our study. This was also the case when we took into account occupational exposure to anesthet ic agents or formaldehyde. In this stud y the existence of previous spontaneous abortions can be considered either as an indicator for the existence of risk factors predisposing to spontaneous abortion or , alternatively, as the first manife station of sensitivity to cytostatic exposure. Of the 43 exposed pregnancie s that were second pregnancies, 10 were preceded by one pre vious spontaneous abortion, and all of the earlier pregn ancies that had resulted in a spontaneous abortion had been exposed. Consequently it was not possible to distinguish women who had a greater risk of spontaneous abortion from women for whom previous sponta neous abortion was the manifestation of cytostatic exposure. It was for this reason that we also decided to restrict the anal ysis to the first pregnancy of each woman rather than taking previous spontaneous abortions into consideration in a multiple regression.

Statistical analysis
Taking into account all the pregnancies of each woman could place too much weight on a small number of women with a large number of spontaneous abortions. For this reason the analysis was also carried out for first pregnancies only. When the basic relationship, the relationship adjusted for classic risk factor s, and the relationship studied in limited categories of spontaneou s abortion were taken into consideration, a statistically significant twofold increase in the spontaneous abortion rate was ob served for the exposed pregnancies as compared with the unexpo sed ones.

Detailed analysis of exposure
An increase in the frequenc y of spontaneous abortion with an increase in the frequency or durat ion of exposur e was not observed. It must be noted however that onl y 25 010 of the exposed pregnancies were occasionally exposed. In addition no relationship with durati on of exposure is to be expected if exposure during pregnancy onl y leads to an increase in the risk of spo nta neous abort ion.
It is well known that a large variety of different chemicals is used in cancer treatm ent. Moreover they are very often used in association with the treatm ent of a given patient. Con sequentl y reliable info rmation on the exposure of nurses to parti cular dru gs would be very difficult to measure since the nurses handl ed cytostatic drugs over a long period of time and treated different types of cancer with different chemicals. For this reason we did not attempt to obt ain information on specific exposure.
Selevan et al (IS) was able to obtain such information . They found that the four most frequently used drugs were doxorubicin, cyclopho sphamide, fluorouracil, and vincristine, and indeed they observed that the exposure of pregnancies to these fou r drugs were highly corr elated. Attempting however to evaluate the relation ship between spontaneous abortion and exposure to doxorubicin , cyclophospha mide, fluorouracil, and vincristine separately, they finally concluded that it was not possible to separate the effects of the individual dru gs.
This study , together with that of Selevan et al (I S), reveals a doubling in the frequency of spontaneous abortion related to exposure to cytostatic dru gs in oncologic departm ents. This increase in risk should be confir med in other populations.
We were only able to study general exposure to cytostatics. An interesting but difficult question would be to try to identify the most hazard ous drug s among the cytostatics and the critical exposure period s.