Maternal occupational exposure and oral clefts in offspring

Study design and population

To examine the possible association between maternal occupational exposure and oral clefts in the offspring a case-malformed control study was performed. Cases and malformed controls were selected from the European Registration of Congenital Anomalies and Twins database of the Northern Netherlands (Eurocat NNL). This population-based registry has been monitoring congenital anomalies in about 18,000 births annually in the provinces of Groningen, Friesland and Drenthe since 1981. In addition to live births (up to 10 years of age at notification), stillbirths, miscarriages and terminated pregnancies because of a congenital anomaly, are registered in the database. Children and foetuses are only registered in Eurocat NNL after parents give informed consent. In general, the informed consent rate is around 80% for all types of congenital anomalies.

Coding and classification of congenital anomalies are performed according to Eurocat guidelines [22]. In this study, Eurocat NNL data of children and foetuses born from 1997 until 2013 was used.

Data collection

Since 1997, parents have been asked to complete a written questionnaire to supply information about the pregnancy. The questionnaire includes a question about maternal occupation and the workplace (e.g. the company where the mother worked) at the beginning of the pregnancy. In addition, information is gathered concerning medical history, demographic characteristics and maternal pre-pregnancy weight and height. For smoking habits, alcohol consumption, and the use of medication, information is gathered from three months before pregnancy until the end of pregnancy. After parental consent, data on prescribed medication is retrieved from the pharmacy. Ambiguities in the questionnaire, actual use of medication and for which period it was used, were verified in a telephone interview with the mother.

Definition of cases and controls

Cases were defined as non-syndromic clefts, either occurring isolated or together with other major congenital anomalies. Children with a Pierre Robin sequence were included in the case group. International Classification of Diseases 9th revision (ICD-9, 749) was used for births up until 2001 and the ICD-10 classification (Q35-Q37) was used for births since 2002. A total of 679 cases with an oral cleft were selected for this study. Cases with a cleft that were also labelled as having a chromosomal or monogenic disorder were excluded (n = 89), because these clefts may be part of that specific syndrome. Additionally, cases with anencephaly, arhinencephaly and holoprosencephaly were excluded (n = 9) because these anomalies are often associated with oral clefts. In total, 95 cases (14%) were excluded because mothers’ occupation was unknown (e.g. the questionnaire was not returned). In this study only mothers with a paid job were included, which led to an exclusion of 99 cases (e.g. housewives).

Non-malformed children are not registered in the Eurocat database. Infants and foetuses born with chromosomal/monogenic disorders, not accompanied by oral clefts, were used as controls, because the etiology of these malformations is known. In total, 1764 chromosomal controls were selected for this study. We excluded 357 controls (20%) because mothers’ occupation was unknown and another 272 controls were excluded because their mothers had no paid job. Hereafter we refer to this group as chromosomal controls.

Analyses were performed with a second control group, because chromosomal controls can introduce bias through higher maternal age. This second control group is defined as all other babies/foetuses registered in Eurocat with non-chromosomal/non-monogenic disorders, and no malformation accompanied by an oral cleft. A total of 6847 babies/foetuses were selected for the non-chromosomal malformed control group. Because mothers’ occupation was unknown, 1626 controls (24%) were excluded. Furthermore, 869 controls were excluded because mother had no paid job. Hereafter we refer to this group as non-chromosomal controls.

This resulted in a total of 387 cases, 1135 chromosomal controls and 4352 non-chromosomal controls. Cases were further subdivided in a group of CP (n = 124) and a group of CL(P) (n = 263).

Exposure assessment

A community-based JEM (ALOHA+ JEM) is applied to translate self-reported information about mothers’ occupation during the periconceptional period (three months before conception through the first trimester) into occupational exposures to solvents, pesticides, metals and more generic categories like mineral and organic dust, and gases and fumes. The ALOHA+ JEM is built specifically for use in community-based studies [23]. Given that specific occupational exposures are relatively rare in the general population, specificity in exposure assignment was preferred over sensitivity when elaborating the ALOHA+ JEM [24].

Jobs were coded by two of the authors (NS and HK) into the International Standard Classification of Occupations 1988 (ISCO88) without knowledge of case/control status [25]. The ALOHA+ JEM assigned occupational exposure to solvents (aromatic, chlorinated and other [e.g. alkanes, alcohols, and esters]), pesticides (fungicides, herbicides and insecticides), metals, dust (organic and mineral), and gases and fumes. Based on the mothers’ occupation, the JEM assigned no (0), low (1) or high (2) exposure to solvents, pesticides, metals, dust, and gases and fumes. For mothers who had two or more jobs with different exposures, the highest exposure category was selected.

Variable definition

Potential confounders applied in our analyses were child sex (boy or girl), number of babies/foetuses delivered (1 or ≥2), previous births (0, 1 or ≥2 births), maternal age at delivery (15–19, 20–24, 25–29, 30–39, ≥40 years old), maternal BMI (underweight [<18.5 kg/m²], normal [18.5–25 kg/m²], overweight [25–30 kg/m²], obese [>30 kg/m²]), maternal education level (low [primary school, lower vocational education, pre-vocational education], middle [secondary vocational education, general secondary education or pre-university education] or high [higher professional education or academic education]), maternal smoking (no, yes/some period during pregnancy), maternal alcohol use during pregnancy (no, yes/some period during pregnancy), folic acid use (no/wrong period, yes/periconceptional period [400 μg folic acid/day from 4 weeks before until 8 weeks after conception [26]), fertility problems (no, yes [self-reported fertility problems or fertility treatment]) and positive family history (yes/no). A positive family history means a first-degree family member with the same condition as the baby/foetus under study, e.g. if a child has an oral cleft, the family history is positive when a first degree family member has an oral cleft as well.

Statistical analyses

The associations between specific maternal occupational exposures and oral clefts were assessed using univariate and multivariate logistic regression models to estimate crude odds ratios (OR) and adjusted ORs. We adjusted multivariate models for potential confounders, based on significance using Chi Square tests. Confounders for the analyses with chromosomal controls were child sex, maternal age at delivery, pre-pregnancy BMI, education level, smoking and alcohol use during pregnancy, and family history. Analyses with non-chromosomal controls were corrected for child sex and previous births as confounders. Separate subgroup analyses were conducted for CP and CL(P) alone compared with both control groups.

From literature is known that the prevalence of CL(P) is higher among male infants. Therefore, an additional analysis was performed stratified by child’s gender. Due to the small number of mothers with high exposure, low and high exposure were merged into one ‘any exposure’ group for all types of occupational exposures. Additionally, for specific exposure categories with a high prevalence of exposed cases, it was possible to evaluate no, low, and high exposure categories separately. P-values of <0.05 were considered statistically significant. Statistical Package for the Social Sciences version 22 (SPSS V22) was used to perform all analyses.

Strengths and limitations

A major strength of this study is the use of data from the population-based Eurocat registry. Ascertainment of oral cleft cases by Eurocat NNL was virtually complete for birth years 1997–2009, with a consent rate for registration of over 90% [8].Data in the Eurocat NNL database are of high quality and congenital anomalies are classified according to high standards and ICD codes. This made it possible to accurately distinguish between isolated clefts, clefts occurring together with other major congenital anomalies and syndromic clefts. Moreover, because both cases and both control groups had anomalies, recall bias is not expected to play a role in our study design.

Another strength is the use of the ALOHA+ JEM. The benefit of using a JEM is that it avoids recall bias since the mother is not directly asked about her occupational exposure during pregnancy. Besides, results in occupational exposure estimates are that are less prone to differential misclassification of exposure compared to self-reported exposures [24, 28].

The Eurocat NNL questionnaire includes questions about job title and workplace during pregnancy, but did not include questions about the actual job tasks that were performed. It is therefore possible that women avoided certain activities during the periconceptional period in order to decrease exposure to potential teratogenic agents. Their actual exposure could therefore have been lower or absent from what was assigned by the JEM based on their job. Another limitation of using a JEM, compared to expert assessment, is that JEMs have often low sensitivity. Partly, this low sensitivity is due to the variability in exposure across time which is not taken into account by the JEM [29].

In our study a relatively low numbers of cases are exposed to pesticides. This has resulted in a lower power. Besides, our study could not address exposure intensity for all subcategories of exposure as assigned by the JEM (low or high exposure) separately in our analyses, due to the low numbers of highly exposed women. This precluded an exposure-response evaluation.

Finally, we used malformed controls and could therefore not compare with healthy children. It is known that occupational exposure to pesticides is possibly associated with chromosomal aberrations [30]. Furthermore, residential exposure to solvents or metals has been suggested to be associated with an increased risk of chromosomal anomalies in the offspring of older women [31]. Given our design, if these associations between occupational exposure and chromosomal anomalies would have been present, this would have resulted in attenuated risk estimates of maternal occupational exposures for the risk of oral clefts in the offspring.