- Top of page
Kleinhaus K, Harlap S, Perrin M, Manor O, Margalit-Calderon R, Opler M, Friedlander Y, Malaspina D. Prenatal stress and affective disorders in a population birth cohort. Bipolar Disord 2012: 00: 000–000. © 2012 John Wiley & Sons A/S.Published by Blackwell Publishing Ltd.
Objectives: Pregnant women exposed to an acute traumatic event are thought to produce offspring with an increased incidence of affective disorders. It is not known whether there are specific times in pregnancy which confer increased vulnerability, or if psychosocial stress alone can increase the incidence of affective disorders in offspring. We examined the relationship of the timing of an acute psychosocial threat during pregnancy to the incidence of affective disorders in offspring using data from a large birth cohort.
Methods: Using data on 90079 offspring born in Jerusalem in 1964–1976 and linked to Israel’s psychiatric registry, we constructed proportional hazards models to evaluate the link between gestational age during the Arab–Israeli war of June 1967 and incidence of mood disorders.
Results: Those in their first trimester of fetal development during the war were more likely to be admitted to hospitals for any mood disorders [relative risk (RR) = 3.01, 95% confidence interval (CI): 1.68–5.39, p = 0.0002]; for bipolar disorder the risk was doubled (RR = 2.44, 95% CI: 0.996–5.99, p = 0.054) and for all ‘other’ mood disorders the risk was tripled (RR = 3.61, 95% CI: 1.68–7.80, p = 0.001). Mood disorders were also increased in offspring whose mothers had been in the third month of pregnancy in June of 1967 (RR = 5.54, 95% CI: 2.73–11.24, p < 0.0001).
Conclusions: A time-limited exposure to a severe threat during early gestation may be associated with an increased incidence of affective disorders in offspring. The third month of fetal development was a moment of special vulnerability.
- Top of page
Of the original cohort of 92408 there were 949 (1.0%) stillbirths and 1380 (1.5%) whose ID numbers could not be traced in the Population Registry, leaving 90079 offspring available for this study. Table 1 summarizes the demographic characteristics of the updated cohort as a whole, as well as of those offspring who were in utero during the first, second, or third trimester of gestation during the Arab–Israeli war of June 1967. Table 2 shows the number and percent of offspring admitted to hospital for any mood disorder, bipolar disorder, other mood disorders, or schizophrenia in the Jerusalem Cohort as a whole and by trimester of exposure to the war. In the cohort there were also 89 offspring admitted for organic mental disorders (F00–F09), 100 for mental disorders due to drug use (F10–F19), 195 for mental disorders due to stress (F40–F48), 44 for behavioral symptoms associated with physiologic disturbances (F50–F59), 353 offspring admitted for personality disorders (ICD-10 code F60–F69), 80 for mental retardation (F70–F79), 37 for behavioral disorders childhood/adolescent onset (F90–F98), 25 for disorders pertaining to psychological development (F80–F89), and 10 for unspecified mental disorder (F99).
Table 2. Number of offspring with psychiatric diagnoses by trimester of gestation in June 1967
| || ||Trimester|
| Any mood disorder||231||12||1||3|
| Bipolar disorder||120||5||1||1|
| Other mood disorders||111||7||0||2|
| Other diagnoses||822||17||13||22|
|Any psychiatric admission||1621||43||23||35|
|Total no. offspring||90079||1423||1409||1466|
Table 3 shows the number of offspring admitted with any mood disorders, bipolar disorders, or other mood disorders by month of exposure to the war. Evident in Table 4, only those offspring whose mothers were in the first trimester of pregnancy during the war had a significantly increased RR for mood disorders as compared to other trimesters. For that first trimester, our Cox model showed a significantly increased incidence of admission to hospital for any mood disorders (RR = 3.01, 95% CI: 1.68–5.39, p = 0.0002), bipolar disorder (RR = 2.44, 95% CI: 0.996–5.99, p = 0.051), and other mood disorders (RR = 3.61, 95% CI: 1.68–7.80, p = 0.001) compared to the rest of the cohort. The RRs for our subcategories did not differ significantly from each other. Stress in either the second or third trimester was linked to a decreased incidence of all three diagnoses as seen in Table 4, although none of these decreases were statistically significant. Offspring in utero during month 3 of pregnancy at the time of the war subsequently experienced a very high incidence of admission for all mood disorders (RR = 5.54, 95% CI: 2.7–11.24, p < 0.0001), bipolar disorder (RR = 5.45, 95% CI: 2.01–14.79, p = 0.0009), and other mood disorders (RR = 5.64, 95% CI: 2.07–15.34, p = 0.0007) in comparison to other months of pregnancy; effects for these groups did not differ significantly from each other. Although the findings for the third month are based on small numbers, we report them because they are highly significant.
Table 3. Number of offspring with all mood disorders, other mood disorders, and bipolar disorders in the Jerusalem Cohort and by month of gestation during the Arab–Israeli war of June 1967
| ||Gestational month|
|All mood disorders||2||2||8||0||0||1||0||1||2|
|Other mood disorders||1||0||4||0||0||0||0||1||1|
|Total no. offspring||419||498||506||452||480||477||492||486||488|
Table 4. Relative risk (RR), 95% confidence interval (CI) and p-values for hospital admission for mood disorders according to trimester of gestation during the Arab–Israeli War of June 1967
|Trimester during war||First||Second||Third|
|RR||95% CI||p-value||RR||95% CI||p-value||RR||95% CI||p-value|
|Any mood disorders||3.01||1.68–5.39||0.0002||0.23||0.33–1.67||0.14||0.69||0.22–2.15||0.52|
|Other mood disorders||3.61||1.68–7.80||0.001||–||–||–a||0.94||0.23–3.79||0.94|
The RR for hospital admission for male offspring did not differ substantially from that for female offspring for all mood disorders (RR = 1.18, 95% CI: 0.91–1.53, p = 0.21), or for bipolar disorder (RR = 0.92, 95% CI: 0.64–1.32, p = 0.66). Male offspring had only 65% of the incidence in female offspring of other mood disorders (RR = 0.65, 95% CI: 0.44–0.95, p = 0.02). However, the RRs of male versus female offspring for bipolar disorder and other mood disorders did not differ significantly from each other (RR = 1.04, 95% CI: 0.61–1.75). There were too few cases when analyzing risk by month for us to adjust our Cox models for sex, but there were sufficient numbers for us to analyze incidence by trimester. Sex of offspring did not influence the relationship of trimester of stress and mood disorders (data not shown), so the final proportional hazards model for trimesters was not adjusted for sex. No other covariates that we evaluated met criteria (described above) for inclusion in the final model, including parental age, country of birth and history of hospitalization for psychiatric illness, socioeconomic status, and secular trend.
- Top of page
Women in our cohort who were in their first trimester of pregnancy during the Six Day War of June 1967 delivered children with an increased incidence of hospital admission for mood disorders. This effect was particularly noticeable in the offspring of those women who had been in their third month of pregnancy. The short and discrete duration of the war made residual stress in our population unlikely because relief due to Israel’s military victory immediately followed the war. This enabled us to pinpoint a narrow window of vulnerability. In other work the stressors studied were likely to have a chronic component so that effects of acute stress on the incidence of mood disorders would be hard to disentangle from those of ongoing stress. Our study is also distinctive because the war of 1967 was chiefly a psychosocial stressor unaccompanied by environmental disruption or famine. In many other studies additional types of significant stress accompanied mental stress.
Several researchers have reported an increase in mood disorders in offspring of mothers who experienced a stressful event in their second or third trimester. Brown et al. (16) studied the incidence of hospitalization for affective psychosis in children who had been in utero during a severe famine that occurred during the winter of 1944–1945 in Holland. Affective psychosis was defined as diagnostic code 296 in ICD-9 [severe disturbances of mood (mania and/or depression) accompanied by mood-congruent psychotic symptoms] (16). The researchers found that a severe nutritional deprivation in the second trimester was associated with an increase in the incidence of affective psychosis in male offspring. However, a re-analysis of data from the same population, but with the addition of newly defined cases, showed that the incidence increased when famine began during the second and the third trimesters, for both sexes, and for both unipolar and bipolar affective disorders (5). This work is difficult to compare with ours because of the different definitions of outcome. Brown et al. used ICD-9 codes to designate their groups for analysis; their two main groups, a priori, were affective psychosis (all 296 ICD-9 codes) and neurotic depression (ICD-9 code 300.4), and these are not directly equivalent to our outcome groups. Furthermore, our cohort did not suffer significant nutritional deprivation, which triggers different physiologic responses than psychosocial stress alone. Khashan et al. (2) found that the male offspring of women who experienced death of a close relative or diagnosis of a close relative with serious disease during the second trimester had a 55% increase in the incidence of affective disorders. In that population, in contrast to ours, psychosocial disturbance was likely to continue after the initial event. An increase in admission to hospital for affective disorders, and especially for unipolar disease, was reported by Machon et al. (4) in a population exposed to an influenza epidemic during the second trimester of pregnancy. An infection would have affected the mother and her pregnancy differently than a purely psychosocial stress and this could account for their conflicting findings. Although none of the RRs were significant, in our analyses prenatal stress in both the second and third trimesters was linked to slight decreases in incidence for all three diagnoses. It is interesting to note this consistent reversal of direction for the effect of prenatal stress for the second and third trimesters. We can speculate that perhaps fetuses in later stages of development react to physiologic signals of maternal psychosocial trauma in ways that differ from reactions in the first trimester, resulting in different effects on long-term mental health outcomes.
Affective disorders are currently divided into several distinct diagnostic subgroups in the ICD-10, as noted above, and the same is true in the DSM–IV. Some consider bipolar disorder to be distinct from the other kinds of mood disorders, however, and bipolar disorders and schizophrenia are sometimes conceptualized as components of a clinical continuum with overlapping symptoms and etiologies (17). In consideration of these views we examined the subgroups of bipolar disorders and other mood disorders separately in some of our analyses. We did not find evidence for a substantial difference in effects of stress during specific time periods on bipolar disorder as compared to other mood disorders, although the increased incidence of other mood disorders, after in utero exposure to the war in the first trimester, was 1.5 times that of the increase measured for bipolar disorders. Machon et al. (4), as noted above, reported a larger increase in unipolar depression than bipolar disorder after prenatal stress. He concluded that there was a difference because one effect was significant and one was not; this approach is not necessarily definitive and statistical tests should be used to confirm this difference (14).
There are several biological mechanisms that could theoretically contribute to our findings. One process might be the actions of elevated levels of cortisol in the developing fetal brain. Increased levels of corticosteroids are part of the stress response. When a pregnant woman experiences stress, her levels of corticosteroids rise and result in a transfer of higher than normal amounts to the fetus. Although some of the cortisol secreted by pregnant women is inactivated by 11-beta-hydroxysteroid dehydrogenase-2 (11β-HSD-2) in the placenta, some maternal cortisol crosses into the fetus (18, 19). Increased levels in pregnant women are thought to cause increases in adverse neurobehavioral outcomes in offspring (20). This is not surprising because cortisol plays an important role in maturation of the human fetal brain, and glucocorticoid receptors are present throughout the fetal brain from early in development in both humans and rodents (18).
There is plentiful 11β-HSD-2 in the human fetal brain as well as in the placenta. 11β-HSD-2 appears to be silenced during specific time periods in gestation and active in others (21), which suggests an intricate, time-specific regulation of cortisol’s influence on neuronal development. This supports the idea of the importance of timing of increased levels of cortisol during the process of fetal development. Cortisol could also affect fetal neurodevelopment via the placenta rather than through direct action on the fetus’ central nervous system. Glucocorticoids have multiple actions on the placenta and fetus, each of which varies by period of gestation (18, 22). Another possible mechanism is suggested by studies in fetal rats showing that early prenatal stress can reduce expression of glucocorticoid receptors in the hippocampus of offspring (23, 24), a brain structure involved in the pathology of bipolar disorder (25, 26).
Elevated cortisol during early pregnancy could also potentially affect the mother in some way so that she raises her child differently. This possibility was suggested in one animal study in which high levels of synthetic glucocorticoids were administered to pregnant rats during early pregnancy. It was found that the treated dams displayed different nursing behaviors as compared to controls (27). Epigenetic regulation of gene expression is yet another potential biological mechanism for our findings. DNA methylation has been shown to vary due to environmental factors including nutrition, chemical exposure or psychosocial problems (28–31). Differences in DNA methylation were also shown in adult offspring who were in utero during the Dutch famine of 1944–1945. These variations were found to vary by sex of offspring and timing of the famine during gestation (32, 33). DNA methylation in the central nervous system has been connected to neural development and neuropsychiatric disorders (34). Theoretically, changes in methylation could follow prenatal exposure to stress. All of these hypothesized mechanisms require further study.
Stress, particularly during the third month, may be tied to an increased incidence of mood disorders in particular for two reasons. Firstly, during the third month of fetal development, neuroblasts that proliferated earlier begin to differentiate into specific neuronal cell types or into microglia (35), and neurons that are the source of all of the GABAergic neurons in the mature brain migrate to the developing cerebral cortex and thalamus (35). This critical process starts and progresses throughout the third month, although these migrations peak from gestational week 12 to week 20 (35). Since disruption in GABAergic mechanisms is associated with mood disorders (36), one could hypothesize that disturbances during the third month would disrupt the migration of neurons that are the source of GABAergic neurons in the brain, thereby influencing the risk for these disorders. Secondly, the third month of gestation is a crucial period of development of portions of the brain that are directly related to affect and affective disorders. During the third and fourth months the major nuclei complete development in the limbic system (e.g., the hippocampus and amygdala) and limbic regions of the cortex (e.g., the anterior cingulate cortex) (37).
Strengths of our study include use of data from a large population-based cohort with over three decades of follow-up, and a clear source of severe psychological stress that began and ended in a well-defined and discrete time period. In studies of longer-term stressors, such as famine within the context of a long conflict, bereavement or natural disasters, it would be difficult to disentangle the effects of the exposure to stress at a specific point in gestation from exposure to chronic stress during the rest of a pregnancy. Prenatal stress and certain outcomes may be associated due to post-natal conditions including the mental health of the mother and the environment in which the infant is raised (38). Stress during infancy has itself been shown to potentially affect risk for mood disorders later in life (39, 40). Therefore, if one studies an ongoing stress, though it began during pregnancy its effects may be difficult to differentiate from effects of an altered environment after birth.
A limitation of our study, as well as of other studies that use data from national registries for hospitalization, is that people with mood disorders do not all require hospitalization. Although Israel has a very high rate of admission to hospital for mental health conditions (41), it is likely that for some these disorders manifested prior to first hospitalization, and that others in the cohort were never admitted for their disorder; therefore, this study informs solely on the incidence of hospitalization. It seems unlikely, however, that those exposed to war prenatally would be hospitalized at different rates than others in the population. In addition, any variation of incidence over time and season was considered in the analyses. Another limitation is the small numbers of cases in the cohort. Small numbers for each diagnosis in each trimester or month make it possible that we could not detect some confounders or mediators. Nevertheless, despite smaller numbers in the month-by-month analyses in particular, our findings are highly significant. Finally, we cannot analyze data on those offspring with an older age of onset. Currently a 33–41-year follow-up of offspring has been completed so that the cohort data are truncated in the late thirties. The cohort will have to mature before this question can be addressed.
While completing this study we confirmed the results of earlier work that showed a link between prenatal exposure to the war of 1967 during early gestation and the incidence of hospitalization for schizophrenia (1). In our updated cohort, exposure to the Six Day War in the second month of gestation, as in the earlier work, was associated with a significantly increased incidence of schizophrenia in offspring. Here we used a narrower definition of schizophrenia (ICD-10 code F20) in our models, which we will be using going forward, while the earlier work defined schizophrenia as hospitalization for any schizophrenia spectrum disorder (F20–F29). A narrower definition of schizophrenia should designate a more specific group of patients, which is likely to be more useful in studying the disease’s etiology.
Major depression is a leading cause of disability worldwide (42). Few genes have yet been found to explain the large proportion of the population with major depression or bipolar disorder (43–45). It has been proposed that environmental factors might interact with genetic ones to affect the risk for these diseases (46). Our work supports the hypothesis that an acutely stressful event occurring early in a woman’s pregnancy might increase the incidence of mood disorders in her offspring. Additional research on this topic is essential because of its implications for public health. As many women continue to experience acute psychosocial traumas during pregnancy, it is necessary to understand the relationship between fetal exposure to this stress at certain points in gestation and the later risk of mood disorders as a first step in developing interventions to reduce negative effects of traumatic events on the subsequent generation.