Psychosocial Stress Related to the Loss of a Close Relative the Year Before or During Pregnancy and Risk of PreeclampsiaNovelty and Significance
The role of stress in the pathogenesis of preeclampsia has only been investigated in a few studies, and the findings are not conclusive. We analyzed whether maternal bereavement shortly before or during pregnancy is associated with an increased risk of preeclampsia. We conducted a cohort study of singleton births in Denmark during 1978–2008 and in Sweden during 1973–2006 (n=4 122 490) by linking national population-based registers. Mothers were considered exposed to bereavement if they lost a parent, a sibling, a partner, or a child the year before or during pregnancy (n=124 553). The risk of preeclampsia was slightly increased for women who lost a close relative during the 6 months before conception (odds ratio [OR], 1.14; 95% confidence interval [CI], 1.06–1.23) or during the first trimester of pregnancy (OR, 1.15; 95% CI, 1.03–1.29). Exposure during these periods tended to be more closely related to early preeclampsia (delivery before 34 weeks of gestation; OR, 1.37; 95% CI, 1.12–1.67) than to late preeclampsia (OR, 1.13; 95% CI, 1.06–1.20). The strongest association was observed between loss of a child and early preeclampsia when the exposure window was from 6 months before pregnancy until start of second trimester (OR, 4.03; 95% CI, 2.46–6.61). Our results related to timing of exposure suggest that severe stress may influence early placentation. However, the public health implications of our findings are limited in populations with a low prevalence of severe stress exposures.
Preeclampsia, a common syndrome characterized by hypertension and proteinuria presenting after the 20th week of gestation, is an important risk factor for maternal and infant morbidity and mortality.1 The first stage in preeclampsia consists of poor placentation, and the second stage is characterized by endothelial dysfunction, systemic inflammation, and the clinical features of the disease.1,2 However, the syndrome may also present in women with proper placentation, who have endothelial dysfunction or are prone to inflammation.1,2 Despite its relatively well-described pathophysiology, decades of intensive research, and its important consequences, the actual causes and the means to prevent the disease are still quite unknown.1,2
Preeclampsia has been characterized as “a state of sympathetic overactivity,”3,4 which could be present already in the preclinical phase of the disease.5 Psychosocial stress induces sympathetic arousal, activates the hypothalamic–pituitary–adrenal axis, and may thus promote endothelial dysfunction,6 proinflammatory activity,7,8 and blood pressure elevations.9 Only a few studies have investigated the association between antenatal stress and preeclampsia and their findings are not conclusive, perhaps because of low statistical power,10–13 exposure misclassification,10,12,14–16 induced in part by coping, and the risk of reverse causation.16
Death of a close family member is considered one of the most severe sources of stress exposure,17 especially, if the loss is unexpected or if it involves children or the spouse17; it is likely to lead to a hormonal response irrespective of coping capacity.18–20,23
In a population-based cohort study of >4 million singleton pregnancies, we investigated whether death of a close relative the year before or during pregnancy was associated with an increased risk of preeclampsia, and whether this risk was higher after unexpected deaths or losses of children or spouses. Because it has been suggested that the associations between stressful life events and risks of other placental dysfunction disorders, such as preterm birth18 and fetal growth restriction among preterm births,19 may be stronger for exposure shortly before placentation than later in pregnancy, we expected to observe a similar exposure time window of importance also for preeclampsia.
Study Population and Design
We studied live births during 1978−2008 recorded in the Danish Medical Birth Register and all births during 1973−2006 in the Swedish Medical Birth Register (SMBR); both registries cover ≥98% of the births in their countries and contain prospectively collected pregnancy- and delivery-related information.21,22 To retrieve additional data on pregnant women and their families, the cohort was linked to other nationwide registers (Table S1 in the online-only Data Supplement) through the unique personal number issued in both countries to all residents.23 A total of 5 499 333 births were recorded during the study period in the 2 birth registries. After excluding infants who could not be linked to their father (n=127 305) or to their mother’s parents (n=1 085 917), who had an unrealistically short, long (see the online-only Data Supplement), or missing gestational age data (n=231 151), or whose mother had a previous preeclamptic pregnancy (n=82 985) or missing identification number (n=24 061), a multiple index pregnancy or missing information on the number of fetuses (n=300 282), 4 122 490 births remained. Compared with women in the studied group, excluded women were more likely to be from Denmark, to be ≥35 years old, multiparous, to have lower education, and to originate from another country than their country of residence. The higher rate of preeclampsia in the excluded than in the included group (3.1% versus 2.1%) was primarily because of twin pregnancies and a preeclampsia history.
The study was approved in Denmark by the Data Protection Agency and the Scientific Ethics Committee of Central Region Jylland and in Sweden by the Research Ethics Committee at Karolinska Institutet, Stockholm. The boards did not request informed consent for participation in the study.
We considered women to be exposed to bereavement during pregnancy if they lost a child, a partner, a sibling, or a parent the year before or during pregnancy.23 Family members were identified in the Danish Civil Registration System and in the Swedish Multi-generation Register, as described previously.23 In Sweden, mothers’ register linkage to their partner was possible only in case of live births.23 Date and cause of death for the relatives were obtained from the Danish Civil Registration System and from the Swedish Cause of Death Register.
Health and Pregnancy-Related Characteristics
Medical conditions and causes of death have been coded in both countries, according to the International Classification of Diseases (ICD). We retrieved data on maternal preeclampsia, chronic hypertension, pregestational diabetes, renal disease (in Sweden only), and placental abruption from the Danish National Hospital Register and from the SMBR; information on maternal psychiatric diagnoses was obtained from the Danish Central Psychiatric Register and from the Swedish Patient Register. Data on diagnoses of cardiovascular diseases (only in Denmark) and diabetes mellitus in mothers’ parents and siblings were provided by the Danish National Hospital Register and the Swedish Patient Register. The diagnostic codes used to identify these medical conditions are presented in Table S2. A validation study of the preeclampsia diagnoses in the SMBR reported a positive predictive value of 50% in case of ICD-8 and 93% in case of ICD-9 codes24; a similar study regarding ICD-10 preeclampsia diagnoses in the Danish National Hospital Register found a positive predictive value of 74%.25
Information on maternal age at delivery, parity, smoking in early pregnancy (since 1991 in Denmark and since 1982 in Sweden), date of delivery, length of gestation, birth weight, previous stillbirths, the number of fetuses in the pregnancy, prepregnancy obesity (in Sweden only, since 1992), and the infant’s sex was obtained from the Danish Medical Birth Register and the SMBR.
The Danish Integrated Database for Longitudinal Labor Market Research provided data on mothers’ country of origin and educational attainment. In Sweden, information on these variables was obtained from the SMBR and the Education Register, respectively.
The association between bereavement during pregnancy and preeclampsia was analyzed using logistic regression. Our primary multivariate models included the following confounders: country, year of birth, maternal education, country of origin, age at delivery, parity, chronic hypertension, psychiatric disorder, and family history of diabetes mellitus before the exposure period. In additional subanalyses, we also adjusted for pregestational diabetes, a measure that in Sweden was identifiable from 1987 onwards; maternal renal disease, available only for the Swedish cohort; early pregnancy body-mass index, available in Sweden only from 1992 onwards; and history of cardiovascular disease in parents and siblings of the mother before the exposure period, available to us in Denmark only. Criteria for including these confounders in our multivariate models were a known or an a priori considered plausible association with the death of a relative and with preeclampsia, and not being on the causal pathway between exposure and the outcome.
To investigate whether the association between bereavement and preeclampsia varied according to the type of loss, we categorized the exposed group according to the mother’s relationship to the deceased. Furthermore, we analyzed the association between bereavement and preeclampsia after natural and unexpected deaths; the definition of this latter category is presented in Table S2. In case of several losses during the exposure period, the first death was considered in the analysis.
To study whether stress may influence placentation, we first divided our exposure window into 5 periods: (1) 7 to 12 months before pregnancy, (2) 0 to 6 months before pregnancy, and (3) first, (4) second, and (5) third trimester.23 Furthermore, we categorized the outcome as early versus late preeclampsia, defined as preeclampsia with birth before 34 weeks versus at 34 weeks or later.
To investigate potential confounding attributable to recurrence of placental dysfunction disorders across pregnancies, we reran our analysis regarding the relation between loss of a child and preeclampsia after excluding infants whose mother had a record of a placental dysfunction disorder in a previous pregnancy, that is, placental abruption, stillbirth, or a small-for-gestational age infant (<2 SD below the mean, according to the sex-specific Swedish reference curve for normal fetal growth).26 We also examined potential mediation by smoking by repeating our primary model after excluding women who reported smoking in early pregnancy. To study interaction between bereavement attributable to death of any relative and other exposures, we performed tests of multiplicative interaction.
The data were analyzed using SPSS for Windows 19 and SAS 9.2.
Mothers in the bereaved group (n=124 553) tended to be older, multiparous, smokers, to have shorter education, a family history of diabetes mellitus and of cardiovascular disease, a record of a placental dysfunction disorder in a previous pregnancy, and to deliver their index child earlier in the study period than their unexposed counterparts (n=3 997 937). The 2 groups did not substantially differ with respect to other characteristics (Table S3).
There were 85 889 pregnancies (2.1%) with preeclampsia in the study group. Compared with being unexposed, any loss of a first degree relative the year before or during pregnancy was associated with an 8% increased risk of preeclampsia in the adjusted analysis (Table 1). Further adjustment for maternal pregestational diabetes, renal disease, obesity, or cardiovascular diseases in parents and siblings of the mother before the exposure window did not substantially alter the point estimate for the association between death of any relative and preeclampsia (data not shown). The increased preeclampsia risk was confined to deaths occurring in the 6 months before or in the first trimester of pregnancy. Only bereavement after natural deaths was associated with preeclampsia (Table 1).
Women who lost a parent the year before or during pregnancy had an increased risk of preeclampsia; the corresponding associations for the loss of a sibling or a child were of borderline statistical significance (Table 2). The odds ratios were generally slightly higher for exposure before pregnancy or in the first trimester than in later exposure periods. Death of partner was not associated with preeclampsia.
Because only deaths within 6 months before pregnancy and in the first trimester were associated with preeclampsia (Table 1), we focused on this period in our analyses of early and late preeclampsia. Loss of any close relative within 6 months before pregnancy or in the first trimester was somewhat more closely related to early preeclampsia than to late preeclampsia. Death of a child during this time period was associated with a 4-fold increased risk of early but not with late preeclampsia (Table 3). Losses in other periods were not related to early or to late preeclampsia (data not shown).
Exclusion of infants whose mother had a record of other placental dysfunction disorders (placental abruption, stillbirth, or a small-for-gestational age infant) in a previous pregnancy somewhat attenuated the association between the death of a child and early preeclampsia; the odds ratio for loss of a child the year before or during pregnancy decreased to 2.43 (95% confidence interval, 1.34 to 4.40), whereas that corresponding to the 6 months before pregnancy or the first trimester decreased to 3.40 (95% confidence interval, 1.76–6.56). The relation between bereavement of any relative and preeclampsia did not differ according to country, infant’s sex, and year of birth, ICD revision used at the time of delivery, maternal age, education, country of origin, chronic hypertension, and psychiatric history (P>0.05). Exclusion of women who smoked in early pregnancy did not change the association between bereavement of any relative and preeclampsia (data not shown).
Death of a close relative the 6 months before or in the first trimester of pregnancy was associated with a slightly increased risk of preeclampsia, especially of preeclampsia with an early onset. This association was strongest for the loss of a child and risk of early preeclampsia.
Comparison With Previous Studies
Our results are in line with earlier investigations, suggesting that the association between antenatal stress and preeclampsia may exist but may be modest. Because of their smaller sample sizes and lower exposure contrasts, all earlier follow-up11,13,27 and several retrospective studies10,12 had limited statistical power to detect a modest association of the magnitude we reported. In 3 of the 5 retrospective studies in this field, stress exposure during pregnancy was reported after delivery;12,15,16 thus the observed associations may have been overestimated due to differential recall12,15,16 or may be—at least partly—attributable to reverse causation.16 In the other 2 of the 5 investigations using retrospective data,10,14 women were assigned job stress scores based on their occupational titles, which may have led to a nondifferential misclassification of exposure and thus a likely underestimation of the association.
The death of a close relative is a well-defined source of stress that will be very stressful, regardless of coping skills or social support.18–20,23 Deaths of children and suicidal deaths of spouses are rated among the most severe sources of stress, as catastrophic life events, by the axis IV of the Diagnostic and Statistical Manual of Mental Disorders, third edition,17 one of the most widely used classification systems of stressful life events. Death of a child was in our study more closely related to early preeclampsia than death of other relatives, and this is in line with this classification system. Furthermore, our findings corroborate those of 3 earlier investigations showing that prenatal exposure to severe negative life events in children (severe illness or death) was associated with a higher risk of giving birth to a child that was preterm,18 extremely preterm, and growth restricted,19 or with a congenital malformation20 than exposure to similar life events in other relatives. The lack of association between the death of a partner and preeclampsia is likely to be attributable to the fact that these losses occurred after placentation, a likely crucial period for the development of preeclampsia.
Potential Explanations for the Association Between the Death of a Relative and Preeclampsia
There may be several explanations for the association between the death of a close family member the year before or during pregnancy and the risk of preeclampsia. It could reflect confounding because of familial and maternal health-related28 and sociodemographic characteristics. Nevertheless, adjustment for sociodemographic factors and for familial and/or maternal cardiovascular, metabolic, or renal diseases did not eliminate the relation between bereavement of any relative and preeclampsia. Similarly, a link between the loss of a child and early preeclampsia was observed after excluding women with a record of a placental dysfunction disorder in a previous pregnancy.
Our finding that the increased preeclampsia risk was confined to exposure in the 6 months before or the first trimester of pregnancy indicates that the timing of the loss is important and may suggest that psychosocial stress after a relative’s death, particularly death of a child, could affect placentation and early placental development. This is supported by the result that losses during this period were more closely associated with early preeclampsia than with late preeclampsia. Poor placentation is considered to be important primarily in preeclampsia with early onset, whereas late-onset preeclampsia may be more closely related to maternal metabolic and cardiovascular risk factors.1,29 Our findings are thus consistent with 2 previous investigations, suggesting that exposure to severe life events shortly before pregnancy is more strongly associated with 2 other disorders related to poor placentation, preterm birth18 and extremely preterm birth combined with fetal growth restriction,19 than exposure in other periods. However, other studies have not found evidence for a difference in the risk of fetal growth restriction30 or of stillbirth,23 according to the time of exposure. The explanation for our finding that only natural deaths were associated with preeclampsia might also be related to the timing of stress.20,23 Severe diseases will usually precede natural deaths; thus stress is likely to have affected pregnancies of women exposed to such losses for a longer time period, possibly during the sensitive period of placental development.20,23
The pathways through which stress could affect placentation and increase preeclampsia risk are not well understood. Plausible mechanisms involve the chronic activation of the hypothalamus–pituitary–adrenal system, subsequent sleep disturbances, anxiety, and depressive symptoms; these may induce vasoconstriction, adverse immune7,8 and coagulatory31 changes, and altered utero-placental blood flow.23 Reduced or intermittent blood flow to the placenta29 may cause oxidative stress and an excessive release of placental factors in the maternal circulation2,29; these promote endothelial dysfunction, proinflammatory activity, and blood pressure surges.2,29 The systemic inflammation and the blood pressure elevation induced by placental dysfunction may be further enhanced by the direct adverse effect of stress,7,8 disturbed sleep,32 and mood disorders33 on immune (increase in proinflammatory activity and vulnerability to infections)34 and metabolic activities.35
Strengths and Limitations
Compared with previous investigations, the present study has several strengths. First, given the longitudinal design and that exposure data were registered independently and before the outcome, recall bias or reverse causation are not plausible explanations. Second, because information on deaths in the Nordic registers is of high quality and bereavement of a family member is an objective and well-defined source of stress, exposure misclassification in the studied group is very small, if it exists at all. Furthermore, the large sample size offered an opportunity to adjust for a large number of potential confounders, to reveal even modest differences in preeclampsia risk between the exposure groups, to investigate the importance of the timing of exposure, and to study both early- and late-onset preeclampsia.
Our study had limitations as well. First, we could not investigate residual confounding from several lifestyle, genetic, or health-related factors, such as unmeasured chronic or inflammatory disorders. Second, although the rates of preeclampsia observed in our study are comparable with the incidence rates of 2% to 8% reported in several countries,1 the low positive predictive value of the ICD-8 preeclampsia diagnoses24 suggests that some misclassification of the outcome in earlier study periods is likely. We expect this misclassification to be nondifferential, which often results in a bias toward the null. Third, we do not know whether our findings may be generalized to other populations. Severe stress and preeclampsia are more prevalent in more disadvantaged populations36 than in those included in our study; whether the studied association would be different in less favorable social and health-care settings needs to be investigated in future studies. Furthermore, although exclusion of infants with unknown gestational age or who could not be linked to their father’s or to their mother’s parents reduces exposure misclassification, it also limits the generalizability of our findings. Women with parents not residing in Denmark or Sweden, the major reason for being excluded from our study, may have less social support than their studied counterparts. Social support may buffer the negative health consequences of stressful life events.37 Thus, the link between stress after bereavement and preeclampsia may be more modest among studied than among excluded women. Fourth, although our sample was large, the power for some of our subanalysis is limited.
Loss of a close relative during the 6 months before or in the first trimester of pregnancy was associated with a small increased risk of preeclampsia, especially of early-onset preeclampsia. These findings may suggest that severe stress could influence placental development. The public health implications of our findings will be modest in populations with a low prevalence of severe stress exposures unless less severe stress also plays a role in a susceptible population.
Sources of Funding
The study was supported by the Swedish Council for Working Life and Social Research (grant no. 2010-0092). J. Li and X.Q. Liu were supported by a grant from the European Research Council (ERC-2010-StG no. 260242).
The online-only Data Supplement is available with this article at http://hyper.ahajournals.org/lookup/suppl/doi:10.1161/HYPERTENSIONAHA.111.00550/-/DC1.
- Received October 31, 2012.
- Revision received November 21, 2012.
- Accepted April 1, 2013.
- © 2013 American Heart Association, Inc.
- Strawn WB,
- Bondjers G,
- Kaplan JR,
- Manuck SB,
- Schwenke DC,
- Hansson GK,
- Shively CA,
- Clarkson TB
- 17.↵American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders: DSM-III-R. Washington, DC: American Psychiatric Association; 1987.
- Khashan AS,
- McNamee R,
- Abel KM,
- Mortensen PB,
- Kenny LC,
- Pedersen MG,
- Webb RT,
- Baker PN
- 22.↵National Board of Health and Welfare. The Swedish Medical Birth Register. A summary of content and quality; 2003. http://www.socialstyrelsen.se/Lists/Artikelkatalog/Attachments/10655/2003-112-3_20031123.pdf. Accessed October 24, 2012.
- László KD,
- Svensson T,
- Li J,
- Obel C,
- Vestergaard M,
- Olsen J,
- Cnattingius S
- Ros HS
- Klemmensen AK,
- Olsen SF,
- Osterdal ML,
- Tabor A
- Redman CW
- Khashan AS,
- McNamee R,
- Abel KM,
- Pedersen MG,
- Webb RT,
- Kenny LC,
- Mortensen PB,
- Baker PN
- von Känel R,
- Mills PJ,
- Fainman C,
- Dimsdale JE
- Dolea C,
- Abou Zahr C
- Sarason IG,
- Sarason BR,
- Potter EH III.,
- Antoni MH
Novelty and Significance
What Is New?
In a cohort study involving >4 million singleton pregnancies, we investigated whether maternal exposure to bereavement during pregnancy, one of the most severe sources of stress, is associated with an increased risk of preeclampsia.
The large sample size allowed investigating the timing of stress exposure and its effect on both early- and late-onset preeclampsia.
What Is Relevant?
A better understanding of the pathogenesis of preeclampsia is needed to increase our chances to predict, prevent, and treat this condition.
Exposure to the death of a close relative during the 6 months before or in the first trimester of pregnancy was associated with an increased risk of preeclampsia, especially of preeclampsia with birth before 34 weeks of gestation. These findings suggest that severe stress may influence placentation.