Spontaneous abortion among women using video display terminals

NIELSEN CV, BRANDT LPA. Spontaneous abortion among women using video displayterminals. Scand J Work Environ Health 1990;16:323-8. This case-base study with a source population of 214108 com mercial and clerical workers investigated the risk of adverse pregnancy outcome among women working with a video display terminal (VDT) in Denmark. A subpopulation was obtained by register linkage be tween a trade union membership file, the Medical Birth Register, and the National Register of In-Patients. In two years 24 352pregnancy outcomes were registered, 2248 spontaneous abortions were recorded, and a base sample of 2252 pregnancies was randomly selected. Data on VDT use, job stress, ergonomic fac tors, and life-stylefactors were collectedwith questionnaires sent to 6212women and 426employers. There was no increased risk of medically verified spontaneous abortion among women with VDT use. The rela tive risk for women exposed to any degree of use was 0.94 (95 % confidence interval 0.77-1.14). Ergo nomic work load and job stress were not confounders.

Clusters of adverse pregnancy outcomes have been reported among women using a video display terminal (VDT), and, although such clusters could occur by chance (1), there has been increased concern about the potential reproductive hazards of VDT use. The main concern has, thus far, been the exposure to extremely low-frequency electromagnetic fields. Studies of animals and human cells provide, however, only limited support for the presumption that exposure to lowfrequency electromagnetic fields results in reproductive failure (1).
Four large-scale epidemiologic studies investigating the risk of spontaneous abortion among VDT users have been published (2)(3)(4)(5). Ericson & Kallen (2, 3) found a very weak association between increased VDT use and abortion. Furthermore, they found a weaker overall association between VDT use at any time and spontaneous abortion. None of their findings were statistically significant at the 5 070 level. McDonald et al (4) found a statistically significant association between VDT use for 1-29 h/week and the occurrence of spontaneous abortion. Goldhaber et al (5) found a statistically significant association between spontaneous abortion and VDT exposure for more than 20h/week.
The present study of medically verified spontaneous abortions was part of an investigation initiated by the Union of Commercial and Clerical Employees in Denmark. Its purpose was to estimate the risk of ad-verse pregnancy outcome for VDT-using members of the Union. The results with respect to the association between congenital malformations and VDT use have been presented elsewhere (6).

Selection of subjects and information
The investigation was designed as a case-basestudy (7). The source population was defined as all women in Denmark between the ages of 15and 44 years who had been a member of the Union of Commercial and Clerical Employees for at least one month from 1982 to 1985. A total of 214108 women fulfilled these criteria. Approximately 80 0J0 of all Danish women employed in commercial or clericaljobs (ie, various job categories in retail, wholesale, production, and service) belong to this trade union. In fact, the source population in the present study comprises about one-fifth of all gainfully employed Danish women between the ages of 15 and 44 years.
By means of the women's personal identification numbers, register linkage was achieved between the trade union file and two nationwide health registers, namely, the Medical Birth Register and the National Register of In-Patients. The Medical Birth Register provides information on pregnancy outcome and all recorded diagnoses for more than 95 0J0 of all births in Denmark, and it is based on both the midwives' and physicians' reports. The National Register of In-Patients provides information on all admissions to any Danish hospital, and it includes all diagnoses made by physicians in hospitals.
The population experience was defined as all pregnancy outcomes in the source population, with the ex-ception of twins, tr iplets , a nd induced abortions. The study was restricted to a two-year period from I Jul y 1983 to 30 June 1985 for spontaneous a bo rtions and from 1 January 1984 to 31 December 1985 for birth s.
Eight case groups were defined, covering various adverse pregnancy outcomes, that is, spontaneous abortions, light for dates, preterm deliveri es, low birthweights , congenital malformations, stillbirths, perinatal dea th s, and infant death s. The base sample (reference gro up) was a 9.2 % random sample (N = 2252) of the source population. The base sample included almost the same number of cases as the largest case group (ie, spontaneous abo rtions, N = 2248). Tab le 1 Table 1. Distribution of pregnancy outcomes in the source popula tion of 214108 women and a base samp le of the source population observed over a per iod of two years. The data on the source population are given for the purpose of comparison.

Source
Base a The base sampl e is a 9.2 % random sample of th e source populat ion. A pregnancy outco me in t he base sample can also appear in one or more of the case grou ps.
b These groups were the case groups in the investi gation. c Birthweight in the lowes t fifth percent il e for gestational age of each se x. d <2500 g.
• Befor e week 37 of gestati on.
shows the number and distribution of pre gnan cies in the sour ce population, in the case groups, and in the base sample. It can be seen that some cases were included in the base sample. Table 2 sho ws the distribution of one dia gnosi s or more of the tot al number of pregnancies in the base sample and the case groups.
Th e emplo yers of women using a VDT at any time dur ing their pregn ancy were identified by register linkage of the case and base sampl e groups and the Labou r Market Supplementary Pen sion Fund.

Recording of spontaneous abortions
Most women who have experienced a recognized spontan eou s ab ortion are referred to a hospital for diagnostic confirmation and treatment. The dia gnosi s is made by the physician in charge and is recorded in the National Register of In-Patients as prescribed in the eight h revision of th e Internation al Classification of Diseases (ICD-8). In this study the ICD -8 cod es 642, 634.61, and 645.1 were used for the identification and selection of spontaneous abortion s.
Since data on the week of gestation in which the spontaneous abort ion occurred were not available , the time of conception was estimate d to be thr ee mon ths prior to the spon ta neous abortion on the average. All cases in the source population were included; 2248 spontaneous abortions were identi fied.

Assessment of exposure
The pre valence of VDT use vari es over time. In order to make a more preci se comparison of th e pregnancies ending in spontaneous abortion with successful pregnan cies, a special mat chin g technique was used. Th is techniqu e takes the time of estimated conception, rath er than the termination of pregnan cy, as the starting point when measuring th e prevalence of VDT use. Table 2. Distribution of one d iagnos is or mor e of the pregnan cy outcomes among 214108 Danish women observed over a period of two years in def ined case groups (total N =4660) and a 9.2 % random samp le (base sample) (N =2252) of the sour ce population .  Table 3. The variables (potential confounders) used in the analysis .

Results
The model used for the statistical analysis was the Mantel-Haen szel stratified analysis modified to casebase studies by Greenland (8).
In the base sample, 80.6 % were employed. Of the employed, 33.5 % worked with a VOT. The proportion of VDT users in the subgroup of clerical and administrative employees was 55.2 %. The subgroup of clerical and administrative female employees formed the basis for the rest of the analysis made in this study.
The proportion of spontaneous abortions for women gainfully employed during pregnancy can be calculated for the base sample as the total number of spontaneous abortions divided by the sum of the total number of spontaneous abortions plus the total number of deliveries [119/(119+ 1224)], the resulting percentage being 8.9 % . The corresponding proportion of spontaneous abort ions for the unemployed women in the base sample [28/(28 + 302)] equaled 8.5 %. Table 4 shows that VOT use during pregnancy did not increase the frequen cy of spontaneous abortions. The relative rate ratio was 0.92 for VOT use at any Heavy lifting (yes/no); silting more than 1 h at a tim e during a workday (yes/no) ; standing more than 1 h at a time during a workday (yes/no); variat ion in work position during a workday (yes/no); standing, sitting, moving (0-2 hId, 3-4 hId. 5-6 hId and 2:7 hId); treatment of cervicobracial disorders during pregnancy (yes/no) Data on the week of gestation at the time of the spontaneous abortion were not available, and thus the time of conception was estimated to be three months prior to the spontaneous abortion. An estimate of the degree of exposure to a VOT was made for the followup period of two years.
In September 1987 a questionnaire with 72 questions was mailed to all women in the base sample and the case groups. The women were asked to give information on occupational status , job title, job stress, ergonomic work load, exposure to organic solvents, VOT use, and life-style factors (eg, alcohol consumption, smoking, and general state of health) for every pregnancy from 1983 to 1985. One question concerned the number of hours of VOT work during pregnanc y: "How many hour s (on the average per week) did you work with a VDT during pregnancy?" The response rate was 75.40/0 (1699 pregnancies) for the base sample and 61.0 % (1371 pregnancies) for the case group of spontaneous abortions. A random sample of 215 was taken of the nonresponders of the base sample (N = 945) and of the spontaneous abortion case group (N = 645). In order to explore the possibility of selection bias, interviewers who did not know the outcome of the pregnancy interviewed the nonresponders who had a telephone (84.8 %). Among these nonresponders 27.5 % in the group of spontaneous abortions and 19.3 070 in the base sample still refused to answer the questionnaire.
Selectiverecall is a known potential problem in casereferent and case-base studies. To control for this problem, we questioned the employers of the pregnant women about the extent of VOT use. All the women who had used a VDT during their pregnancies were asked permission to contact their employer. A total of 185 women in the case group of spontaneous abortions and 284 women in the base sample gave their consent. By register linkage of the women's personal identification number and the Labour Market Supplementary Fund, the employers were identified . A questionnaire with six questions concerning VOT use was sent to a random sample of 63.9 % of the employers in March 1988. A question concerning hours of VOT work corresponded to the question given to the women . The respon se rate was 93.6 % and 88.5 %, respectively, for the group of spontaneous abortions and the base sample group.

Analysis
As the base sample was a random sample of the total source population, it was possible to obtain a direct estimate of the relative risk even if the rare disease assumption was not fulfilled. Control for bias induced by potential confounders was made. The classification of the potential confounding variablesl covariates used are shown in table 3. A potential confounding variable was adjusted for if it was statistically significantly associated with spontaneous abortion at a 5 % level.
Table~. Risk ratios for medicall y verified sp ontan eous abort ions accordi ng to work with a video displ ay terminal (VOT) among the gamfull y employed women. (95 % CI =95 % confidence interval of the adj usted rate ratio ) in the case group of spontaneous aborti ons and in the base sampl e.  (7). b Detailed information on hours of VOT work/week was missing for 16 pregnancies. These pregnancies have been assigned a value of zero (none).  Table 6. Estimation of select ion btas .s Among the total number of nonrespondents (pregnancies) In the source population a random sample was drawn. A tel ephone interview was made without the interviewer knowing the outcome of the pregnancy.  326 time compared to no VDT use altogether. VDT use for 11-20 and 21-30 h/ week showed an insignificant relative risk of 1.05 and 1.12, respectively, in a comparison with women with no VDT use during pregnancy.
Results similar to those in table 4 were obtained by means of a stratified analysis of the potentially confounding variables. Maternal age was the only variable that correlated significantly with the occurrence of spontaneous abortion. Ergonomic work load and job stress did not con found the results.
If the employers' reports of VDT use are correct (which is not necessarily the case), table 5 shows that women who reported VDT work for 1-10 h/ week " underreported" their VDT use. Conversely, women who stated they worked~11 h/ week with a VDT " overreported" when their responses were compared with the informa tion given by their employers. This tendency was even clearer in the base sample. Therefore selective recall was indicated which could bias the results towards high rate ratios. A test of agreement between the differences in the statements of the women and the employers in the case group and in the base sample by the rank sum test did not show any statistically significant disagreement.
Women who used a VOT and who had experienced a spontaneous abortion might be expected to be more willing to answer the questionnaires than women who gave birth to a normal child. Thi s selection bias was estimated by the odds ratios of sensitivities in table 6 (9) (selection odds ratio = 1.30).

Discussion
The 6541 pregnancies in the study represented 6212 women -5894 women had had one pregnancy, 306 had had two, 9 had had three, and only 2 women had had four pregnancies in the follow-up period. The pregnancies were treated as independent events in the analysis although some women contributed with more than one pregnancy. The low average number of pregnancies (N = 1.05) contributed by each woman was assumed to have a minor effect on the estimation and testing (10).
Recall bias has previously been mentioned as a possible explanation of findings of a positive association between VOT exposure and spontaneous abortion (1,5). If recall bias was responsible for our negative findings, the women who experienced a spontaneous abortion must have underreported VOT exposure during pregnancy in comparison with the reporting of the women in the base sample group and/or the women in the base sample group must have overreported their VOT exposure in comparison with the women in the case group. In order to clarify this point, we compared the information given on VOT exposure by the women in the base sample, as well as those women who experienced a spontaneous abortion, with the information their employers gave on VOT use. As shown in table 5, it is unlikely that recall bias could explain the finding of no association in this study.
The selection odds ratio was 1.30. This value indicates that selection bias due to nonresponse produced a slight overestimation of the rate ratios.
The duration of pregnancy averaged nine months for those women who gave birth as compared with an average of three months for women who aborted. Thus the first group would have had a greater chan ce a priori for any VOT exposure and would, therefore, bias the comparison between spontaneous abortions and births in the direction of low risk ratios. Therefore, measurement s of VOT exposure were expres sed in term s of an average, that is, average exposure to work with a VOT during pregnancy. Such information is less dependent on the duration of pregnancy, and the study, in fact, supports this observation. When the groups of women who used a VnT to various extents during pregnancy were compared, no significant differences were found between the average VOT use when working and when not working. Detailed information on hours of VOT work at various stage s of pregnancy was unfortunately not obtainable.
A woman who had previously experienced a spontaneous abortion might avoid VOT use if she planned to get pregnant again because of fear of the potential risk of adverse pregnancy outcome when using a VOT. Such behavior modification would lead to an underestimation between VOT use and spontaneous abortion. However, this study did not support the possibility that such behavior modification took place.
For ethical reasons, women who had had induced abortions were not contacted. There is, however, no reason to suspect an unequal distribution of induced abortions among the exposed and unexposed women. It should be mentioned that the relative risk, as calculated, is an unbiased estimator of the relative frequencies of spontaneous abortion of all recognized pregnancies under the nonnull hypothesis, but it is not so in the case of the null hypothesis (11). In conclusion, this study does not indicate an increased risk of medically verified spontaneous abortion for women using a VOT.