Article Text

Download PDFPDF

Original article
Psychosocial job strain and risk of adverse birth outcomes: a study within the Danish national birth cohort
  1. Ann D Larsen1,2,
  2. Harald Hannerz1,
  3. Mette Juhl3,
  4. Carsten Obel4,
  5. Ane Marie Thulstrup2,
  6. Jens Peter Bonde5,
  7. Karin S Hougaard1
  1. 1National Research Centre for the Working Environment, Copenhagen, Denmark
  2. 2Department of Occupational Medicine, Aarhus University Hospital, Aarhus, Denmark
  3. 3The Women and Children Research Unit, Juliane Marie Centret, Rigshospitalet, Copenhagen, Denmark
  4. 4Department of Public Health, Faculty of Health Sciences, Aarhus University, Aarhus, Denmark
  5. 5Department of Occupational and Environmental Medicine, Copenhagen University Hospital Bispebjerg, Copenhagen, Denmark
  1. Correspondence to Ann Dyreborg Larsen, The National Research Centre for the Working Environment, Lersoe Parkalle 105, Copenhagen DK-2100, Denmark; adl{at}nrcwe.dk

Abstract

Objective A number of studies examined the effects of prenatal stress on birth outcomes with diverging and inconclusive results. We aimed to examine if working with high job strain during pregnancy measured in week 16 was associated with risk of giving birth to a child born preterm or small/large for gestational age (SGA/LGA), and second, if social support affected any associations.

Design Study population was 48 890 pregnancies from the Danish National Birth Cohort. Multinomial logistic regression estimated ORs. Covariates included: maternal age, BMI, parity, exercise, smoking, alcohol and coffee consumption, manual work, serious maternal disease, parental height and gestational age at interview. In accordance with Good Epidemiological Practice, a protocol outlined the study design before analyses were initiated.

Results High job strain was associated with significantly lower odds of being born LGA (OR=0.81, CI 0.70 to 0.92) when compared to low-strain jobs, but no associations between high strain and preterm birth or SGA were found. Stratification by social support showed a non-statistically significant tendency of higher odds of preterm birth when exposed to high strain and low social support.

Conclusions In spite of the high statistical precision, we did not find any clear association between job strain and preterm birth nor between job strain and SGA.

Statistics from Altmetric.com

What this paper adds

  • Literature suggests that stress exposure during pregnancy affects the unborn child by preterm delivery and lowered birth weight.

  • However, most studies are small, retrospective and do not include job strain.

  • Based on a large and prospective Danish cohort, results show no clear association between job strain and preterm birth, nor between job strain and small for gestational age.

Background

Stress has for several years been studied in relation to adverse health, for example, cardiovascular disease.1 For women of childbearing age, effects may extend beyond the exposed individual and affect pregnancy, birth and health of the child. Studies show that pregnant women, who experienced distress (eg, depression, anxiety, adverse life events, daily hassles or pregnancy-specific stress) are more likely to deliver preterm2–7 and give birth to children with lower birth weight.5 ,7–9 Importantly, pregnancy outcomes such as low birth weight and preterm birth are considered to be risk factors of later morbidity and mortality.10 Life course epidemiologists suggest that particularly birth weight may be associated with longer-term risk of disease, for example, coronary heart diseases.11

Some studies on adverse psychosocial work environment and pregnancy outcomes corroborate the findings from studies of generalised stress, but the studies are very diverging in conclusions. A Dutch study with data from the Amsterdam Born Children and Their Development (ABCD) study found high job strain in early pregnancy was associated with low birth weight, particularly if the mother worked more than 32 h per week,12 and a Korean study with data from the Mothers and Children's Environmental Health (MOCEH) cohort also found a relationship between job strain and birth weight, but only on decision latitude (control dimension).13 A Canadian study with representation from urban, semiurban and rural populations reported association between high demands and low control (high strain) and preterm birth, but only when combined with low or moderate social support.14 Two American studies on register data observed only a very modest increase in risks for preterm birth and low birth weight,15 ,16 and another study from the USA found an association on work-related stress based on job titles among a nationally representative sample and risk of preterm birth and low birth weight, but only when the women did not want to remain in the work force.17 Three studies—a Danish with birth cohort data, an Irish with use of the Lifeways cohort and a Spanish among hairdressers—found no association on either preterm birth or low birth weight when being in the high-strain group,18 or by self-reported stress.19 ,20 Overall, findings are not conclusive as to the adverse influence of work strain on pregnancy outcomes. This was also the message from a recent critical review. Here the authors called for future studies of prospective design, large study samples and with work-related stress exposure data.21

We therefore aimed to study the association between psychosocial job strain (high demands, low control) and the risk of preterm birth and birth weight related to gestational week, in a large, prospective population-based Danish pregnancy cohort with linkage to national registers.

Methods and design

Study cohort

This study was conducted as described in the study protocol ‘Testing the association between psychosocial job strain and adverse birth outcomesdesign and methods’.22 The protocol provides a detailed description of methods, design, study population, exposure, outcomes, covariates and planned statistics prior to conduction of the analyses.

The study was based on the Danish National Birth Cohort (DNBC). This nationwide longitudinal survey was established in 1996 and included prospective interview data of more than 100 000 pregnancies in Denmark. All participating women should be able to read and understand Danish well enough to complete the interviews. The original cohort protocol of the DNBC included four interviews: two in weeks 12–14 and 30–32 of pregnancy and two interviews postnatal. Further information on the structure and aim of DNBC may be obtained from Olsen et al.23

Totally, 100 418 pregnancies were included in the DNBC. First inclusion criteria for this study were confirmation that the woman was still pregnant and working in the baseline interview (63 739pregnancies). Including only singletons, the first sibling in the cohort, and complete data on psychosocial job strain, birth outcomes and included covariates, reduced the number of eligible pregnancies to 48 890. This can be seen in the flowchart,22 although small discrepancies appear due to removal of outliers in birth weight. Among the included pregnancies (n=48 890), the prevalence of preterm birth, defined as delivery before 37 completed gestational weeks was 4.9% (n=2375). The prevalence of small for gestational age (SGA) and large for gestational age (LGA) equalled, by definition, approximately 10% for each outcome.

Exposure

From the first interview (range: gestational week 6–21, median: 15.6 weeks), the psychosocial exposure-variable was based on two questions on the psychosocial work environment:

  • Do you have too many tasks at your work? (Demand dimension)

  • Do you have the opportunity to influence your tasks and working conditions? (Control dimension).

(Both questions had the response categories: often, sometimes and seldom.)

Combination of the answers to these questions allowed for evaluation of psychosocial load in the working environment by use of the Demand Control Model/Job Strain Model.24 In agreement with the model, the women were divided into four categories based on their answers to the above questions: low strain (low demands, high control), active (high demands, high control), passive (low demands, low control) and high strain (high demands, low control).

In order to maximise exposure contrast, high strain was defined by the answer categories ‘often’ when enquiring about demands, and ‘seldom’ when enquiring about control. This combination was of our primary interest, as high demands and low control at work were expected to predict the highest degree of mental strain.24 The reference group was the low-strain group, defined by the answer categories ‘sometimes’ or ‘seldom’ to the demand question and ‘often’ or ‘sometimes’ to the control question as illustrated in figure 1. This model has been applied previously in DNBC research, in assessment of the effects of work-related stress on pelvic pain and late fetal loss.25 ,26

Figure 1

The demand-control model divided into the four categories: passive, active, low and high strain based on the two questions from the Danish National Birth Cohort.

For explorative purposes, the question ‘Do you get any help from colleagues when you have troubles in the work?’ with the response categories: often, sometimes and seldom was included, to study social support in relation to the Job Strain Model. Low social support has been hypothesised to alleviate the effect of job strain (iso-strain-model) as introduced by Johnson and Hall.27

Outcomes

Birth and pregnancy data were extracted from the Danish Medical Birth Register by use of the personal identification number given to all Danish citizens at birth, allowing for combination of data across all registers in Denmark.

Gestational age was calculated as the number of days from the first day of the last normal menstrual period (LMP) to the day of birth. Mean gestational age when baseline interviews were conducted was 15.6 weeks.

SGA was defined as the 10% lightest babies at each gestational week for each gender within the present study population. For gestational weeks with less than 10 births in each group (week 22–24), SGA equals the lowest birth weight in the group. LGA was defined by the 90th percentile or above.

In the published protocol, we failed to exclude outliers in birth weight when calculating SGA and LGA. In the actual analysis, we excluded 227 outliers (<125 g or >6000 g) in birth weight according to established growth charts.28

Preterm birth was defined as a delivery after 22 weeks and up to 36 completed weeks of gestation. A delivery before 22 completed weeks was defined as miscarriage and was therefore excluded from analysis. Full term includes gestation week 37 to 44.

Statistical analysis

A multinomial logistic regression analysis as a function of job strain (high strain, active and passive vs low strain) estimated the ORs with 95% CIs for being preterm, full term but SGA and full term but LGA. Full term and normal weight for gestational age served as reference. The advantage of this outcome categorisation was a clean reference outcome (full term and normal weight for gestational age) with which the three types of cases can be contrasted. Similar analysis was done when stratifying on social support.

The analysis was conducted with the procedure ‘proc logistic’ in the computer package SAS V.9.2. A likelihood ratio was used to test the overall null hypothesis, which stated that the outcome vector (being either preterm, SGA or LGA) was independent of job strain. The null hypothesis would be rejected if p≤0.05. An estimated OR would be accepted as statistically significant if (1) the overall null hypothesis was rejected and (2) the 95% CI of the OR did not include one.

Covariates

The analysis was controlled for several covariates (see protocol22 for detailed description): maternal age (<25, 25–29, 30–34, >35 years) due to elevated risk of low birth weight when mother is above 35 years.29 BMI (15–18.4, 18.5–24.9, 25–29.9 and 30.0–49.9) as maternal overweight and obesity are known risk factors, for example, elective preterm delivery.30 Parity (0, 1, ≥2 births), as women generally deliver lighter babies in their first compared with the following pregnancies.31 Gestational age at interview (<16, 16–21), to crudely control for potential differences due to the healthy worker effect. Exercise (none, less than 3½ h and 3½ h or more of exercise per week) due to its beneficial effects on adverse health outcomes such as gestational diabetes.32 Smoking habits (never, sometimes, daily), as several studies show cigarette smoking is causatively related to low birth weight and SGA.33 Alcohol consumption measured in units where one unit is equal to 12 grams of alcohol which corresponds to one normal beer (33 centilitres), one glass of wine (12.5 centilitres) or 4 centilitres spirits (none, less than 1 unit of alcohol per week, 1–2 units of alcohol per week, and more than 2 units of alcohol per week), as it is documented that alcohol increases the risk of adverse birth outcomes.34 Coffee consumption (yes/no) as a high caffeine intake (>300 mg per day) has been associated with reduced birth weight.35 Type of work (manual vs non-manual), as women with manual work might have a different risk of adverse pregnancy outcomes than those performing non-manual work. Additionally, this factor is associated with a person's opportunity to influence work tasks and working conditions.36 Serious maternal disease, diabetes and epilepsy (either, neither) as especially diabetes has been associated with increased risk of congenital malformations, obstetric complications and neonatal morbidity.37 Parental height (combined into a continuous variable) as we assumed that the probability of a child being SGA or LGA at birth depended on the length of child, which again depended on how tall the parents were. Table 1 present the descriptive data.

Table 1

Descriptive table for the included covariates

Results

The crude prevalence of preterm birth was 4.85% in the high-strain group, 4.65% in the passive group, 5.04% in the active group and 4.82% in the low-strain group. The corresponding percentages for SGA and LGA, respectively, were 9.80, 9.35, 8.44, 9.11 and 8.03, 9.40, 9.80, 9.59. The cut-points for SGA and LGA that were used in the present study were compared with a table based on an intrauterine fetal weight standard.28 There were no significant differences.

Analysis was initiated by an overall likelihood ratio test, where we tested if the outcomes were independent of job strain. The null hypothesis was rejected with a statistically significant p-value at 0.02, and this permitted further analyses of the model. The use of a multinomial logistic regression model allowed for testing of the hypothesis: no relationship between job strain and each outcome; preterm (p=0.68), SGA (p=0.09) and LGA (p=0.01)), indicating that only LGA was statistically significant related to job strain, and is therefore, the only outcome interpretable.

Table 2 presents the crude results of the multinomial logistic regression. After adjustment for covariates (table 3), an effect of high strain on LGA was observed (OR=0.81, CI 0.70 to 0.92) when compared to low-strain jobs. The table also includes the number of children for each outcome and strain group.

Table 2

Crude data; multinomial logistic regression analysis on the four job strain categories and the odds of having a child born preterm, small for gestational age, or large for gestational age

Table 3

Results from multinomial logistic regression analysis on the four job strain categories and the odds of having a child born preterm, small for gestational age or large for gestational age*

Comment: In tables 2 and 3, the relationship between active style and reduced risk of SGA appeared to be statistically significant according to the 95% CI. A similar apparent significance was found in the ‘high support’ stratified group. These results should, however, be regarded as non-significant, according to the overall p value of the relationship between SGA and job strain (p=0.09).

To study the effect of social support on our model, we stratified by the three levels of social support (often, sometimes and seldom) in the multinomial logistic regression analysis. An overall test, which assessed the independency of social support on job strain, indicated that only ‘often’ was statistically significantly related to job strain, as the results were ‘often (received help from colleagues)’ (p=0.00), ‘sometimes’ (p=0.63) and ‘seldom’ (p=0.90). Table 4 presents the results from the stratified analysis, showing practically no changes in ORs for preterm birth, SGA or LGA when replying ‘often’ to social support compared to the non-stratified analysis. Results from the other social support categories were not statistically significant, but did show a tendency of higher odds of preterm birth when exposed to high strain and low social support. The results should be interpreted with care as some outcome categories included very few births.

Table 4

Multinomial logistic regression analysis on the four job strain categories and the odds of having a child born preterm, small for gestational age or large for gestational age, stratified on the three levels of social support, controlled for covariates*

Discussion

This study examined the relationship between maternal psychosocial job strain measured early in pregnancy and the risk of preterm birth or having a child with low birth weight in relation to gestational age.

Analyses did not confirm previous findings of an association between work-related stress and preterm birth.14–17 When stratifying by social support analysis indicated higher odds (OR: 1.39, CI 0.86 to 2.23) of having a child born preterm when being in the high-strain group with low social support compared to the low-strain group, however, the results are not statistically significant.

In our study, we used the birth weight relative to gestational week. Furthermore, we distinguished between being born SGA and LGA leaving us with a ‘normal group’ to which we could compare the other groups. We did not find association between high strain and having a child born SGA, as other studies have,12 ,13 ,16 ,17 but did see that high maternal strain lowered the odds of having a child born LGA, both with and without adjustment of covariates.

Regarding the use of SGA as a way of determining child growth: ideally, a longitudinal monitoring of weight of the child during the pregnancy, and calculations of growth rate in a given time interval, would give the best estimates to distinguish between a genetically small child and a pathologically small child. Several issues need to be addressed in this matter. First, we do not have ultrasound measurements from several times in pregnancy, also there is an uncertainty related to ultrasound estimated weights used in diagnostic context. Second, the relative increase in weight is small and decreases with rising gestational age as indicated at the ‘lulugram’ by Lulu Lubcheno et al.38 Third, we do not have enough information on the individual variation of the child growth. The Danish Society of Obstetrics and Gynaecology therefore recommends use of SGA as a way to measure growth retardation.39 The decision of a 10% percentile makes this study comparable with similar studies, for example, the study by Henriksen et al.22

The present study has several strengths. The number of participants is large (48 890 pregnancies). The prospective study design eliminated recall bias. Hindsight bias was eliminated since the statistical model, and hypotheses were defined and peer reviewed before analyses of the relationship between the exposure and outcome variables in our data material.22 Mass significance was avoided since all outcomes were incorporated into one single model and tested with the overall likelihood ratio test.

Other studies in the field of psychosocial workplace factors and adverse pregnancy outcomes are relatively small (248–1033 pregnant women),13 ,15 ,19 ,20 use case-control/retrospective designs,14 ,15 ,20 or registers/job titles in assessment of maternal exposure to psychosocial job strain.16 ,17 Two studies are both large and have a prospective design and are therefore comparable. One is the prospective study by Henriksen et al including 8000–9000 pregnancies where they found no associations between high strain and preterm birth and birth weight.18 Another is the study from Vrijkotte et al,12 where the researchers find that the combination of job strain and long workweek reduced the birth weight by 150 g and heightened the risk of SGA. Our findings are in line with the findings from Henriksen et al who also studied births among Danish women from 1989–1991, and concluded ‘the influence of work-related psychosocial strain on the risk of SGA and preterm delivery seems to be small in countries with highly developed social support systems and few other work-related hazards’.18

A weakness in this study is the use of only two questions on the exposure in assessment of job strain, and not the original questionnaire (Job Content Questionnaire; JCQ40) developed for Karasek's Job-Demand-Control-Model. Also the Job-Demand-Control-Model was developed in the late seventies, primarily as an attempt to assess job strain among industrial workers.24 It may therefore be questioned if the model is applicable in assessment of job strain in pregnant women of which many work outside the industrial setting, 25–30 years later. We acknowledge that this may preclude detection of the full spectrum of job strain, but these were the data available in the DNBC dataset. Another line of questioning might have provided a more liberal criterion with regard to whom to include in the high-strain group. More subjects in the high-strain group would have increased the statistical power. It would, however, also maybe decrease the exposure contrast.

Two other studies in the DNBC have used the exposure questions the same way as in this study,25 ,26 but as far as we know, no studies have looked at the validity of using global questions in regards to JCQ. Although not using the same aspect of job strain as us, a study of Littman et al, compares two single-item measures on stress with three more fully validated multi-item instruments on perceived stress, daily hassles and life events, which assessed the same underlying constructs as the single-item measures. The two constructed questions (ability to handle stress, and rate the amount of stress) were considered reliable at measuring stress with validity similar to longer questionnaires.41

Another limitation is collection of exposure data by self-report in the telephone interview. This precluded inclusion of objective measures, such as collecting blood or urine samples for cortisol42 relating to the psychosocial work environment, as called for in the review of Mutambudzi et al.21 Our exposure data were, however, collected prospectively, early in pregnancy, before the women knew the pregnancy outcome, and some researchers find that subjective measurement of stress is the best predictor of effect on birth weight.43

The present study does not allow for clarification as to whether the lack of association between job strain and preterm birth truly reflects absence of effect, or if it arises as a result of preventive measures directed toward the pregnant woman at work (eg, sick leave), as we cannot withdraw the help and actions from the healthcare system, labour market and so on from the equation. Moreover, only one measurement of job strain was available (around pregnancy week 16). Hence, we had no information about whether the strain was constant during the pregnancy, and it is reasonable to assume that pregnant women may have reduced tasks towards the end of the pregnancy although some women might also experience increased stress towards the end of the pregnancy in order to prepare task, for example, a substitute or finish ongoing tasks. The analyses were, however, made from a public health perspective. We wanted to know if the risks for the studied pregnancy outcomes were higher among women in high-strain jobs compared with low-strain jobs. Our primary interest was not if job strain has an effect per se but if an effect of job strain resides after preventive measures have been taken, that is, does the preventive system work, and are the measures taken sufficient to protect the woman.22 It is possible that a protective approach to pregnant women in the work force may have reduced associations between job strain and negative birth outcomes to an undetectable level, on the other hand we know from Danish legislation that, for example, the recommendations regarding heavy lifting and time standing has been eased. In the recommendations from 2002, lifting a maximum 1000 kg per day and no lifting of more than 10–12 kg at one time were included. In the recommendations from 2009, there are no upper limits of lifting, just information on the risk of preterm birth and fetal growth retardation when physically loaded.44

Also the lack of associations might be due to selection bias among workers in high-strain jobs also called the healthy worker effect. The women that are most affected by job strain might be less likely to work and to participate in the quite time-consuming study interviews. Thus, women with low socioeconomic resources in terms of education, occupation, income and civil status are under-represented in the DNBC compared with the background population.45 As decision authority, which is one of the dimensions in the Job Strain Model, can be expected to be related to education, occupation and income, those with the highest job strain may not be fully represented in the study.

Conclusion

In spite of the high statistical precision, we did not find any clear association between job strain and preterm birth, nor between job strain and SGA.

Acknowledgments

The MINERVA working group is thanked for fruitful discussions, for further information see http://www.minervanet.dk.

References

Footnotes

  • Contributors ADL, KSH and HHA designed and wrote the first draft. This was modified and adapted by MJ, AMT, CO and JPB.

  • Funding The Danish National Research Foundation has established the Danish Epidemiology Science Centre that initiated and created the Danish National Birth Cohort. The cohort is furthermore a result of a major grant from this Foundation. Additional support for the Danish National Birth Cohort is obtained from the Pharmacy Foundation, the Egmont Foundation, the March of Dimes Birth Defects Foundation, the Augustinus Foundation, and the Health Foundation. The MINERVA project is supported by the Danish Work Environment Research Foundation (Grant no: 20080016458/4). This specific study is supported by grants from the MINERVA project, Aarhus University, and the Danish Graduate School in Public Health Science (GRASPH).

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement This study is based on data from the Danish National Birth Cohort. Researchers can apply for use of these data at http://www.dnbc.dk.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.