Multiple pregnancies among women engaged in agriculture in Norway , 1967-91

Background. The worldwide occurrence of dizygotic (DZ) multiple pregnancies has increased since 1980 as a consequence of assisted fertilization techniques. An opposite influence by environmental factors yet unidentified has been a suspected explanation of decreased multiple pregnancies in several countries during 1960-1980. The aim of our study was to describe multiple, in particular DZ, pregnancy patterns in the Norwegian agricultural population between 1967 and 1991, and to investigate the hypothesis that grain production under climatic conditions favoring fungal growth causes a decline in DZ pregnancies. Methods. By record linkage of national registers in Norway, we identified 246,043 farm holders and spouses born between 1925 and 1971 in agricultural censuses and the population register. In this population, 190,258 pregnancies were recorded in the Medical Birth Registry 1967-91. The prevalences of multiple pregnancy, subdivided in DZ and monozygotic (MZ) pregnancies, were examined in strata of several determinants. Exposure, defined as the combination of grain farming and categories of seasonal fungal warnings, was based on data on farm activity and on local fungal warnings in the growth seasons. Adjusted prevalence ratios (PR) served as estimates of association. Results. We identified 2,131 (11.2 per 1,000) multiple pregnancies, 1,322 (6.9) were classified as DZ. The prevalence of DZ pregnancies was declining until the late 1970s, and increased thereafter. The agricultural population of Norway is heterogeneous along two dimensions, with a difference between farmers and nonfarmers (mainly engaged in forestry), and a difference according to degree of maternal work input on the farm holding. Farmers, in particular active farmers, had more DZ pregnancies than non-farmers early in the study period; this ratio was reversed late in the study period. Non-farmers had a much higher temporal increase in the proportion of first-time mothers who were ≥30 years, and were more likely to undergo in vitro fertilization (IVF). Combined grain farming and seaso nal late blight warnings did not influence the prevalence of DZ pregnancies. Unexpectedly, the same climate variable, in combination with horticulture, was negatively associated with DZ pregnancies (PR 0.3; 95% confidence interval 0.1–0.7). Conclusion. Active Norwegian farmers seem to have a somewhat higher natural occurrence of DZ pregnancies than mothers in non-farming sectors of the agricultural population. After the late 1970s this situation was reversed, mainly because the latter group consisted of more older first-time mothers, and probably with a higher need of assisted fertility treatment. Interpretations are difficult concerning the inverse relation between DZ pregnancy and climatic conditions in horticulture.


INTRODUCTION
Throughout the world, prevalences of multiple pregnancies have varied considerably during the last decades. 13][4][5][6][7] In the decades preceding the use of these fertility treatment techniques, Western European and several other countries experienced a decline in dizygotic (DZ) twin rates.8, 9 This was a matter of concern, because the decline was assumed to be a sign of declining fecundity.The causes of the observed DZ decline remained obscure; one hypothesis being that environmental factors disturb the hormonal regulation of ovulation.9, 10 In Norway, the prevalence of multiple pregnancies was stable (about 10 per 1,000) between 1967 and the mid-1980s, but increased by 30% in the years to follow.
In a study of perinatal outcomes in births to Norwegian farmers, delivery during gestational weeks 16-27 was associated with climatic conditions in grain farming favoring growth of field fungi and formation of mycotoxins. 12Several mycotoxins prevail in grain farming in Norway, [13][14][15] of which some have immune and endocrine-modulating properties 16,17 that might influence DZ pregnancy tendency.
The purpose of the present study was twofold: to describe secular patterns of multiple, in particular DZ, pregnancies in the total Norwegian agricultural population and subgroups of this population, secondly, to pursue the hypothesis that conditions favoring fungal growth and mycotoxin formation in grain farming are inversely related with DZ pregnancies.

MATERIAL AND METHODS
Statistics Norway conducted three agricultural and two horticultural censuses between 1969 and 1989.In total, 149,254 farm holders born after 1924 were identified in the censuses, and 104,370 spouses were identified after linkage with the Central Population Register.We linked the file of farmers and spouses to the Medical Birth Registry of Norway to identify their births.The establishment of the cohort and linking procedures are described in more detail elsewhere.

Data sources
The Medical Birth Registry of Norway, comprising all deliveries from 16 completed weeks of gestation since 1967, provided data on multiple pregnancies.We identified 190,258 pregnancies (192,417 births) 1967-91, to 90,578 mothers (median year of birth, 1948) who were farm holders or spouses.Data on sex of the newborn, year of birth, maternal age and birth order were derived from the Registry.Year of birth was categorized into six periods, maternal age into four categories and birth order into four categories (Table 1).Pregnancies after IVF treatment are registered since the late 1980s by the Medical Birth Registry in cooperation with the IVF laboratories. 7In the study population, 39 pregnancies (2.0 per 10,000) were to mothers who had been subject to IVF treatment.
Data on farm activity from the agricultural and horticultural censuses were linked to all pregnancy records.Dichotomous categories of types of farming were assigned to each pregnancy, dependent on information in the census closest in time to the conception: farming activities defined as animal farming (cattle, sheep/goats, pigs, poultry, horses), and cultivation of potato, grain, horticultural products (orchards, berries, greenhouses) or field vegetables (Table 1).Pregnancies of mothers on a farm with none of the above mentioned activities came almost exclusively from forestry holdings, and were classified as «no farming activity» pregnancies.Data on farm location (six geographical regions) and acreage (three categories), and maternal work input on the farm were derived from the censuses.Maternal work input was dichotomized into <500 and ≥500 annual hours, the occupational criterion for a farmer of Statistics Norway.18   The Norwegian Meteorological Institute has since 1958 provided daily fungal warnings in the growing season (mid June to mid September) as a tool against late blight growth in potato. 19Warnings have been based on daily climate data at local weather stations.Criteria for a late blight warning applied for two consecutive days and remained unchanged until the 1990s: minimum temperature ≥10 0 C, maximum temperature 17-24 0 C, relative humidity ≥75%, and precipitation.Warning data during 1973-90 were computerized, and data from 46 stations were allocated to all study farms based on their location.The local warning count in a specific season was assigned to pregnancies conceived between June in the same season until May next year.Thus, only pregnancies conceived between June 1973 and May 1991 were classified on seasonal warnings.

Analysis
Our observation unit was the pregnancy, conditional on birth after 16 weeks of gestation.The analyses were performed in the Epicure software package. 20We computed multiple pregnancy prevalences per 1,000, that were divided into DZ and MZ pregnancies on the probabilistic assumption that twice the number of opposite-sexed twin pairs and 8/6 the number of unequal-sexed triplets are DZ (Weinberg's differential). 21DZ pregnancy was the main study outcome.Assignments of DZ and MZ pregnancies were made on the group level, after stratification for all determinants in the covariate matrix, allowing full adjustment in analysis.We used prevalence ratio estimates (PR) as measure of association, computed in contingency tables that were stratified by the determinants listed in Table 1.We fitted Poisson regression models, controlling for these variables, and computed approximate 95% confidence intervals (CI).
The analysis of the effects of the main environmental exposure indicator, combined grain farming and seasonal late blight warnings, was restricted to the pregnancies with season-specific warning data and known gestational age.Using DZ pregnancies as main outcome in this analysis created a conceptual problem since it can be influenced by opposing forces.2][3][4][5][6] We tried to avoid combining natural and iatrogenic DZ pregnancies in one outcome category in the investigation of environmental influence, but, apart from the few IVF pregnancies, we had no data on assisted fertilization treatment.Therefore, we used two measures to minimize the influence of iatrogenic DZ pregnancies: restrict analysis to 1967-1982, when those fertilization techniques scarcely were used, and, exclude first-time mothers ≥30 years, supposedly the main high consumer group of fertility treatment.

RESULTS
We registered 2,131 multiple pregnancies (2,103 twin, 28 triplet; 11.2 per 1,000); 1,322 (6.9) were classified as DZ and 809 (4.2) as MZ pregnancies.Figure 1 shows the crude temporal trend of multiple pregnancies.The prevalence was stable around 11.0 before 1980, but increased thereafter until a maximum of 14.2 during 1987-91.This trend was mainly influenced by DZ pregnancies: the prevalence declined from 6.5 (1967-70) to 5.6 (1979-82), followed by a maximum of 10.6 in 1987-91.By contrast, the MZ prevalence was rather stable during 1967-91.
Table 2 illustrates that the DZ prevalence increased considerably with year of birth, maternal age and birth order.Twelve out of 39 IVF pregnancies (307.7 per 1,000) were multiple (10 twin, 2 triplet).The DZ prevalence changed only moderately by geographical region, acreage and maternal work input on the farm.The adjusted model showed clearly that maternal age was a strong determinant whereas birth order scarcely had any independent effect.In the adjusted model, PRs changed little over time but were moderately increased for births during 1987-91.Type of farming was not associated with DZ pregnancies, with the exception of horticulture (PR 0.7; 95% CI 0.6-0.9).The overall DZ prevalence in the «no farming activity» category did not deviate from the total prevalence.However, stratification by year of birth showed different trends in pregnancies according to farming activity (    The maternal age and birth order distribution in the population obviously had an influence on DZ prevalence.Therefore, we made a closer examination of the proportions of total pregnancies which were to first-time mothers ≥30 years (Table 5).This proportion increased almost three-fold during the study period, was highest in the eastern region, among residents on small farms and mothers in the no farming activity category.Low proportions were found in animal farming and potato cultivation, and particularly in pregnancies of mothers with ≥500 annual work-hours on the farm (2.7%).The increasing trend of first births to women ≥30 years was more moderate in the high work input category, from 2.2% to 4.1% between 1967-70 and 1987-91, compared with the low work input category (from 2.9% to 8.0% between 1967-70 and 1987-91).Table 6 data show the association between combined grain and seasonal climate data, and DZ pregnancy.Analysis was restricted to pregnancies conceived between June 1973 and May 1991, firsttime mothers ≥30 years of age were excluded.DZ prevalences declined slightly in pregnancies conceived in a season with >1 late blight warning, but this was not restricted to grain farms.Accordingly, was no apparent association between combined grain farming and warnings, and DZ pregnancies.Due to the negative association with horticulture, the effect of horticulture was investigated in combination with late blight warnings, revealing that DZ pregnancy was negatively associated with combined horticulture and fungal warnings.On farms with horticultural cultivation and any seasonal warnings, 6 DZ pregnancies were observed (2.1 per 1,000; PR 0.3; 95% CI 0.1-0.7).Results of analyses in all pregnancies con-

DISCUSSION
The population of this study comprises most of the people who were engaged in agricultural work in Norway. 12The study provides evidence that the prevalence of multiple pregnancy declined in the Norwegian agricultural population between 1967 and the late 1970s, after which it increased until 1991.These time trends were due to changes in DZ pregnancies whereas MZ pregnancies showed a stable tendency.
This register-based study has crude exposure indicators, and it is well known that the resulting misclassification may attenuate true associations.In this study, there is another potential problem in using DZ pregnancy as outcome.DZ pregnancy might be a manifestation of both decreased and increased fecundity.
1-6,9,10 Furthermore, the DZ computation is based on assumptions that may not be true: the sex ratio of DZ births may shift over time and region, and Dutch data suggest that the use of Weinberg's differential underestimated DZ rates in 1960-90.22   The declining trend in multiple pregnancies before 1980 is in accordance with other populations in the same period, 1,8,9,22 but not with the total Norwegian po-pulation who had a stable prevalence during 1967-85. 11 ][4][5][6][7][8]23 Results from East Flanders 3 suggest that this is the only significant explanation, a statement that has been disputed.24 We had data on IVF treatment but not other methods of infertility treatment.However, the DZ prevalence increase was restricted to older mothers (Figure 2) and could not entirely be accounted for by IVF that is considered to be completely recorded in the Medical Birth Registry since 1988.7 It is therefore reasonable to assume that further iatrogenic factors, not recorded in the present study, had a major impact on the trend during the later study years.
Twinning has been reported to be more frequent in rural than in urban populations in Sweden and Finland 25 but not in Denmark. 26The crude prevalence of multiple pregnancies in Norway was about 10 per 1,000 deliveries during 1967-85 and increased to 13.0 in 1990, 7 being more than 10% lower than in the agricultural population.However, this difference could be explained by a higher maternal age distribution compared with all Norwegian births. 7Nonetheless, DZ pregnancies to mothers who took part in farming activities were more common than in other sectors of agriculture during 1967-78, even after adjustment for maternal age.
The results demonstrate other interesting differences within the study population.The high DZ prevalence during the later study years among women who were not involved in farming activities, suggest that they were more commonly treated for infertility.This was the case for IVF treatment: during 1988-91, the overall proportion of IVF treatments was 33.0 per 10,000 pregnancies in Norway, 7 the corresponding proportions in the farming and no farming activity categories were 19.4 and 47.2, respectively.However, this cannot be interpreted as a difference in fecundity, but rather be explained by different proportions of older first-time mothers in the population subsets during the later study years.During 1987-91, only 5.9% of pregnancies in the «farming activity» category were to first-time mothers ≥30 years of age; the corresponding proportion for «no farming activity» pregnancies was 9.8%.
The study hypothesis that mycotoxin exposure in grain farming reduces the prevalence of DZ pregnancies was not confirmed.Thus, very preterm delivery 12 and certain hormone-dependent cancers among the mothers (Kristensen et al, unpublished results), both being associated with fungal warnings in grain farming, were not correlated with DZ pregnancies.Unexpectedly, a decrease in DZ pregnancies was associated with climatic conditions favoring fungal growth in horticulture.This result is interesting, since pesticides was a suggested cause of the widespread decline in DZ pregnancies during 1960-80. 9Furthermore, the horticulture-climate combination is associated with solid tumors in infancy (Kristensen et al, unpublished results).However, the quality of data in this study make inferences difficult, in particular since the indicator of exposure is related to fungi and mycotoxins as well as fungicides and other pesticides.Besides, DZ pregnancy has since 1980 become an ambiguous outcome unless data on assisted fertility treatment is at hand and can be accounted for.

Figure 1 .
Figure 1.Crude prevalences (per 1,000 pregnancies) of multiple pregnancy, subdivided in DZ and MZ pregnancies, by year of birth, of women in the agricultural sector, Norway 1967-91.

Figure 2 .
Figure 2. Prevalences (per 1,000 pregnancies) of DZ pregnancy by year of birth, in categories of maternal age and birth order, of women in the agricultural sector, Norway 1967-91.

Table 1 .
Items of personal data for pregnancies of 90,578 women in the agricultural sector, Norway 1967-91.

Table 2 .
Prevalence of DZ pregnancies of women in the agricultural sector, Norway 1967-91, and associations with demographic and maternal variables, and type of farming.

Table 3 .
Prevalence of DZ pregnancies of women in the agricultural sector, Norway 1967-91, by year of birth and main type of agricultural activity.

Table 4 .
Prevalence of DZ pregnancies among 7,558 first-time mothers, aged ≥30 years, in the agricultural sector, Norway 1967-91, and associations with year of birth and IVF treatment.
* Model including year of birth, IVF treatment, geographical region, acreage and maternal work input on the farm.

Table 5 .
Proportions of pregnancies which were to firsttime mothers ≥30 years of age, in the agricultural sector, Norway 1967-91, and relations with demographic variables and type of farming.No cultivation of potato, grain, horticultural products or field vegetables, and no animal farming in the agricultural census closest in time to the conception.
* Adjusted for year of birth, geographical region, acreage and maternal work input on the farm.†

Table 6 .
Prevalence of DZ pregnancies of women in the agricultural sector, Norway 1973-91*, and associations with environmental exposure indicators.