A cohort of rubber manufacture employees has been established, using personnel records from rubber plants, in all 12 production facilities all over Sweden. In all of the facilities, there was production of general rubber goods. One of the facilities also produced tyres. The cohort includes all employees first employed 1965 or later, employed for at least 3 months, in total 12,014 men and 6,504 women. Information on periods of blue-collar employment was available for all subjects. Information on job tasks varied in complexity and completeness between plants, and was not considered to have enough accuracy for use in this study. Statistics Sweden was able to identify all but 1% of the women, and 1% of the men.
In the year 2001, the Food Worker’s Union provided a list of all female members, 35,757 women from all over the country. Of these, Statistics Sweden was able to identify all but 8 women. All women were blue-collar workers. Information on duration of employment and specific exposures was not available.
Linkage to the Swedish Population Registry to establish cohorts of mothers, fathers and children, and to registers of reproductive outcome
The rubber workers cohort and the female members of the Food Workers Union were linked to the Swedish Population Register by Statistics Sweden. Also, cross-checking with the registries of deaths and births was performed. Thus, the identities of all children born to these women and men between 1973 and 2001 were obtained. Altogether, 17,918 children to rubber workers and 33,487 children to female food industry workers were identified.
In a next step, these children were identified in the Medical Birth Register, which includes almost every infant born in Sweden since 1973. The registry is based on copies of records and forms from the maternity health care, the delivery and the pediatric examination of the newborn. In 1982, several new variables were introduced into the register, for example, information on maternal smoking in early pregnancy. Also, all children were matched to the Register of Congenital malformations, which includes serious congenital malformations reported within 6 months after birth.
In the present study, we restricted the cohort of rubber workers children. The restriction of employment period that was considered for exposure of the child was based on the assumption that there are no accumulated effects of exposure in the rubber industry that affect reproductive outcome. For female workers, only continuous employment as a blue-collar worker during 9 months before the birth of a child was consider as an exposed pregnancy. For male rubber workers, we similarly considered the entire period between 12 and 9 months before the birth of a child as an exposed sperm production period, assuming 3 months for maturation of spermatozoa, and a full term pregnancy. The various combinations of mother’s and father’s rubber work and the number of children in each study group are shown in Table 1. There were altogether 2,828 live-born children with maternal and/or paternal employment during the entire 3 or 9-month period. Children with no parental employment in the rubber industry during these periods constituted the internal reference cohort (n = 12,882). Children with partial parental employment (n = 2,208) during these periods were not included in the present study.
In a second step, we restricted the study to first-child only. In a third step, a restriction within the rubber worker cohort was made including only siblings with contrasting exposure, thus enabling an exposure crossover design. There were 222 children with maternal rubber work during the pregnancy (with or without paternal rubber work), having altogether 255 siblings with neither maternal nor paternal rubber work during the pregnancy and sperm maturation period.
Among food industry workers, 231 children with a father or mother who had ever been a rubber cohort member were excluded. Thus, 33,256 children remained in the study group.
The reproductive outcomes studied were offspring sex ratio, birth weight, preterm birth (gestational length ≤ 37 weeks), small for gestational age (SGA) (Källén 1995), large for gestational age (LGA) (Källén 1995), length at birth, head circumference at birth, multiple births, all malformations and stillbirths (week 28 and later). Also, involuntary childlessness for 1 year or more, ever, reported at the pregnancy under study was investigated.
Characteristics of the cohorts
Descriptive maternal data are given in Table 1. The annual number of children with both parents employed in the rubber industry was highest during the 1970s. In contrast, in the food industry workers as well as in the other rubber workers study groups, the annual number of births peaked during the late 1980s and 1990s.
The maternal age distribution was median 26–27 years in the rubber workers study groups, and slightly lower among food industry workers, median 25 years. Accordingly, the proportion of young mothers <20 years was slightly higher among food industry workers. Around 40% of all children were first-born. The maternal height and weight during early pregnancy did not differ between study groups. Information on the maternal native country was available only for births after 1978. Among children with both parents as active rubber workers, a slightly higher proportion of the mothers were immigrants from Europe. In contrast, more food industry workers were non-European. The rubber plants were situated in different parts of Sweden, all of them in provincial towns. The majority in the reference cohort, 79%, resided outside the urban areas of Stockholm, Gothenburg and Malmö.
Information on maternal smoking during early pregnancy was available from 1983. The proportion of non-smokers among females employed in the rubber industry during the actual pregnancy was 64%, compared to 62% among food industry workers.
The effect of cohort membership, with the food industry workers’ cohort as a reference category, was investigated using linear regression analysis for continuous outcomes (i.e. birth-weight) and logistic regression analysis for binary outcomes (i.e. multiple births, gender of child, involuntarily childlessness). Mother was incorporated as a random effect in these regression models in order to account for the dependence in outcome within a set of siblings. Only live births were included. As potential confounders, we considered child’s sex, smoking status (non-smoker, smoker), maternal age (−24, 25–29, 30+) and parity (1, 2, 3+) kept together, calendar year of birth (5 year intervals) and maternal ethnicity (Sweden, other Scandinavia, other Europe, non-European). We used the method suggested by Greenland (1989) for deciding which of the potential confounders that should be included in the final multivariate model. Potential covariates were entered into bivariate and multivariate models if they changed the effect estimate by >10%. All regression analyses were conducted using PROC MIXED and PROC NLMIXED in SAS version 8.2. For analyses of first-child only, SPSS was used for the linear and logistic analyses.
In addition, an exposure–crossover design was explored, comparing siblings in rubber worker families with and without maternal exposure during the pregnancy. Linear and logistic analyses were performed without mother as random effect, adjusting only for sex, using SPSS.