Fecundity and the use of video display terminals

and the Scand association between prolongedwaitingtime to pregnancy and the use of a videodisplayterminal(VDT)wasinvestigated amongcommercial and clericalemployees in Denmark. Information on 24352pregnancies wasobtained through registerlinkagewith two national registers. A random sampleof 2252 pregnancies wasdrawn, and the womenwereasked for information on waiting time to pregnancy,occupationalexposures, and life-style factors. The overallexposureto a VDTindi cated a slightly increased associationwith prolonged waitingtime to pregnancy. The relativerisks for waitingmorethan oneyearwere1.23,0.77,and 1.61 for the womenwhoworkedwitha VDTfor 1-10, 11-20, and 2:21 h perweek, respectively. Onlythelastrelative riskwasstatistically significant. Theresults couldbeexplainedbymethodological shortcomings suchas recallbiasor confoundingbyalternativefac tors that might affect time to pregnancy, or simplyby the lack of statistical power.

The possible effects of the use of a VIdeo display terminal (VDT) on reproduction have been investigated during the last decade with the main focus on spontaneous abortion and congenital malformations. No scientific evidence has been found to support the early case reports on clusters of spontaneous abortions and infants with congenital malformations among women working with a VDT (1-10). Two of the studies (4,6) reported, however, an increased risk of spontaneous abortion among women working more than 20 h per week with a VDT, but the authors explained their findings as a result of selection bias (6) or job stress (4).
Other aspects of reproduction, apart from spontaneous abortion and congenital malformations, may however be at risk. For instance, no research has been done on a possible association between the use of a VDT and reduced fecundity. Reduced fecundity indicates lack of conception or implantation or the early loss of pregnancy at any stage from gametogenesis to early unrecognized spontaneous abortion. The number of menstrual cycles (in the absence of any form of contraception) before a couple conceives (ie, time to pregnancy) can be used as an estimate of fecundity (the probability of becoming pregnant within a cycle), given regular unprotected sexual intercourse. Baird et al (II) have suggested the use of time to pregnancy as a screening tool for identifying factors that might interfere with the reproductive process. This paper presents results on fecundity, measured as time to pregnancy, and the use of video display terminals among commercial and clerical female employees in Denmark.

Subjects and methods
The reproductive histories of 214 108 female members of the Union of Commercial and Clerical Employees in Denmark were obtained for 1983-1985 by register linkage with the Medical Birth Register and the National Register of In-patients. The women's personal identification number was used as the identification key. All of the women were between the ages of 15 and 44 years. A total of 24 352 pregnancies was registered, from which a random sample of 2252 pregnancies was drawn (9.2070). More-detailed information on data collection has been given previously (1, 2).
A questionnaire was sent in September 1987 requesting information on time to pregnancy, occupational status, job title, job stress, ergonomic factors, exposure to organic solvents, VDT use, life-style factors, and health status. The response rate was 75.5% (N = 1700). A random sample of nonrespondents was drawn (N = 196). These women were contacted by telephone and interviewed in accordance with the questionnaire.
The question relating to time to pregnancy in this study was "How long did it take you to become pregnant, that is, what was the interval between stopping contraceptive use and becoming pregnant?" The alternative responses were 0-6 months, 7-12 months, 13-24 months, 2 years or more, do not remember." Only gainfully employed women were included in the analysis (N = 1365). The statistical analysis was based on Mantel Haenszel's stratified analysis, obtaining adjusted relative risks (12). Table 1 shows the a recent Danish study among pharmacy assistants (0-6 months 78%, 7-11 months 9%,~12months 13%) (14).
Information on exposure was not collected prior to the occurrence of the outcome (waiting time to pregnancy). The exposure information concerned early pregnancy, and therefore represents a proxy variable for the proper exposure time. The validity of the proxy variable depends on the stability of the women's jobs over a period of 6-12 months. It seems that exposure was rather stable over time in that 83% (N = 58) of the women with two pregnancies in the study period had the same exposure during both pregnancies. Still, bias due to a misclassification of exposure could exist if the women who had an increased waiting time to pregnancy tended to stop work when they became pregnant. However, no such tendency was found among the 17% of the women who had had an altered exposure from the first to the second pregnancy.
Lack of response to the questionnaire might have resulted in nonresponse bias. We studied the poten-potential confounders that were considered in the analysis. Information from the random sample of the nonrespondents was used to estimate the direction and magnitude of nonresponse bias from calculations of the selection odds ratios (13).
To study the potential for recall bias, another measure of work with a VDT was used, that is, occupational titles were used as a job-exposure matrix. The amount of VDT work within each job reported by women with a waiting time to pregnancy of less than seven months was used to divide the occupational titles into two groups, in which less than 5Ofo or more than 60% of the women worked with a VDT. The relative risks for increased waiting time to pregnancy were then calculated by comparison among these two groups.

Results
There were statistically significant associations between prolonged time to pregnancy (~7 months) and smoking > 10 cigarettes daily [relative risk (RR) 1.63], age 25-34 years (RR 1.61), age~35 years (RR 3.93), and one or more previous pregnancies (RR 0.55). The corresponding figures using prolonged time to pregnancyas 13 months or more were 1.75, 1.99,7.35, and 0.39, all again statistically significant. Table 2 shows that there was no difference in the distribution of time to pregnancy within the groups of working and nonworking women.
There was a statistically significant increased relative risk for prolonged waiting time to pregnancy ( 4). When using the job-exposure matrix, we found no associations between VDT work and increased time to pregnancy (table 5).

Discussion
The distribution of time to pregnancy among the study population (

299
Scand J Work Environ Health 1992, vol 18, no 5 Table 3. Work with a video display terminal (VOn and tim e t o pregnancy -adjusted relat ive risk and 95 % confidence intervals (95% CI) for prolonged waiting time to pregnancy (;0,:7 months or ;0,: 13 mon th s compared wi th < 7 mo nth s). Time to pregnancy Table 5. Work with a video display term inal (VOn , as measured by a [c o-exposure matri x, and ti m e to pregnancy -adjus ted relative risk and 95 % confidence int ervals (95% CI) for prolonged wai ti ng time to pregnanc y (;0,: 7 months or ;0,: 13 mon ths compared with < 7 mon ths).
tial risk of selection bias by applying the distribution of VDT use in the random sample of nonresponders to th e whole group of nonresponders. The selection prob abilities for the exposed and unexposed women with a time to pregnan cy of less than seven months and seven months or more were estimated , and the selection odds ratio was calculated to be 1.01, which mean s no selection bias due to nonrespon se.
;,,7 months Using time to pregnan cy as a measure of subfecundity means that women who did not become pregnan t were not included, and the association between exposure and sterility was not investigated .
Information was not sought on the use of contraception nor on the frequency and timing of int ercourse. The presence of organic disorders causing a prolonged waiting time for pregnancy, such as salpingitis and anovulation, was not known. Confounding might have been caused by such disord ers.
Another source of confounding might have been an "infertile worker effect" (ie, if women with reduced fertility tend to stay in jobs and women who have proved their fertility by giving birth to children do not). However, the distribut ion of waiting time to pregnancy was the same among the working and non working women (table 2). But a selection effect could have still been operating within the group of workin g women if nulliparous women tended to work longer hou rs per week than women who alread y had children. The for mer group would have had a higher probability of high exposure , which would lead to con founding due to an association between a nulliparous state and VDT exposure . Ho wever the results for the nulliparas and ;,,13 months Occupational groups with less than 5% VDr work Occupational groups with more than 60% VDr work the women working :2:35 h per week (table 4) showed almost the same relative risks as all the working women together.
Smoking, number of previous pregnancies, and age have been found to be associated with prolonged time to pregnancy in other studies (11, [15][16][17] . The present study found the same associations, a result suggesting a high reliability of the measure' 'time to pregnancy." The present study only showed a statistically significant association between prolonged waiting time to pregnancy (:2: 13 months) and self-reported high exposure to VDT use (:2: 21 h per week), but this finding was based on a very small number of persons . When the job-exposure matrix was used, no increased risk for prolonged time to pregnancy was found. Use of the job-exposure matrix may have introduced a nondifferential misclassification of exposure, which would have led to a bias towards the null (18,19). However, the exposure matrix was based on information from noncases and reflects true exposure. Therefore, nondifferential misclassification cannot explain the finding of no effects at all (18,19). Recall bias thus remains a possible explanation. In particular, recall bias might be expected among those waiting :2: 13 months to conceive. Therefore, bearing the shortcomings of the study in mind, one should interpret the results with caution.