Introduction
The significance of early negative life events and especially the loss of a significant figure as possible etiological factors in the development of subsequent adult psychopathology has been subject of much speculation and theory for years (Bowlby, 1963). In many studies, an association was reported between EPL due to death (Perris et al., 1986), or other reasons (Dennehy, 1966, Faravelli et al., 1986 and Tennant et al., 1982) and the development of affective disorders. Early work suggested a higher incidence of parental death in the history of depressed individuals while later work on this issue was less conclusive (Lloyd, 1980) yet suggested that the loss of the mother either but death or separation may be more significant then the loss of the father (Brown et al., 1997, Kendler et al., 1992 and Roy, 1985). These studies are also inconclusive about the significance of the age of loss. Degree of coping with EPL is an important moderator of future psychopathology as demonstrated by Breier et al. (1988) who reported that psychological suffering and the success or lack of coping with the loss of a parent was predictive of future psychopathology.
A large case control study found the incidence of EPL before the age of 17 due to death or separation to be almost 4-fold greater in hospitalized patients suffering from major depression compared to a control group. Moreover, this difference increased to 11-fold when EPL occurred at an age of under 9, and divorce as a cause for separation was found to be more significant than separation due to death of a parent (Agid et al., 1999). The number of studies looking at divorce as the cause for child–parent separation is small. In cases of divorce, compounded onto the need to cope with a “loss” of a parent, research has shown that divorce and life in a single parent family make the child more prone to exposure to stressful life events (Amato and Keith, 1991 and Hetherington et al., 1985) and that inadequate parenthood may increase the incidence of depressive illness in adulthood (Parker, 1983). A study in Israeli adolescents whose parents have divorced demonstrated higher levels of psychopathology, and a correlation was found between the attachment pattern with their parents and psychiatric symptoms (Canetti et al., 2000). In fact, there is a growing body of literature showing that there may be a considerable difference between long-term effects of bereavement and divorce, and that the impact of divorce is potentially more complicated to adequately assess due to a multitude of potential environmental moderating factors involved in the process (Luecken and Appelhans, 2005 and Mack, 2001).
Abnormal function of the Hypothalamic–Pituitary–Adrenal axis (HPA axis) has been well documented in a number of major psychiatric syndromes such as major depression and posttraumatic stress disorder. Hypercortisolemia, a condition that is associated with stress (Kalin and Takahashi, 1988) has been reported to induce glucocorticoid receptor changes and a reduction in hippocampus volume in animal models (Herman et al., 1995, Lopez et al., 1998 and Sapolsky, 1996) and in humans (Heim and Nemeroff, 1999 and Sheline et al., 1996). It has been suggested that early negative life events may cause hypercortisolemia, which in turn may cause damage to the developing brain during critical periods of development (Gunnar, 1998). In babies, there is a period of decreased reactivity of the HPA axis (Gunnar et al., 1996 and Suchecki et al., 1993), which has been found to be dependent on the quality of the baby-parent relationship (Nachmias et al., 1996). Furthermore, in a number of animal models, states of maternal deprivation caused long-term changes of the HPA axis demonstrating a pattern of hyperstimulation (Ladd et al., 1996 and Plotsky and Meaney, 1993). Findings in non-human primates exposed to deprivation have not been consistent, with some showing increased (Coplan et al., 1996 and Fahlke et al., 2000) and others showing decreased (Clarke, 1993) basal and stress-induced cortisol levels, or decreased morning cortisol levels in monkeys exposed to maternal separation (Dettling et al., 2002). In humans, while a number of studies have demonstrated long-term effects of childhood maltreatment and sexual abuse on the HPA axis, mainly showing increased stress reactivity (De Bellis et al., 1994, Heim et al., 2000 and Heim et al., 2001), only a few studies have focused on the effects of parental loss or separation on the HPA axis. In children, studies have been contradictory, some reporting increased morning cortisol in adopted children (Gunnar et al., 2001) and elevated diurnal cortisol secretion in institutionalized children (Kaufman, 1991), and others showing lower morning cortisol in socially deprived children (Carlson and Earls, 1997). In adults with EPL due to parental death, Luecken reported that increased cortisol responses during a psychological stressor were moderated by the quality of the parent-child relationship (Luecken, 2000 and Luecken and Appelhans, 2006). Furthermore, whereas, Breier et al. (1988) have reported elevated afternoon cortisol plasma levels only in adults with EPL who had a history of psychopathology, recently, elevated salivary cortisol levels throughout the day but especially in the morning, were reported in adult men who experienced parental death, regardless of psychopathology (Nicolson, 2004). However, a recent study in healthy young adults with EPL (death or separation), found decreased morning salivary cortisol levels (Meinlschmidt and Heim, 2005), while another study did not find a differential cortisol response in adults exposed to various childhood traumatic events (Otte et al., 2005). In the case of divorce as an early life adverse event, there is no literature regarding the long-term effects of this stressor on the HPA axis, though it has been proposed that HPA axis dysregulations may play an important role in the development of mood and behavioral maladjustment to the consequences of divorce (Troxel and Matthews, 2004).
The design of the current study was based on the hypothesis that if EPL indeed causes permanent alterations in the HPA axis, these alterations will precede and may mediate subsequent psychopathology. If this hypothesis is correct, such alterations in the HPA axis are expected to be present even in the absence of psychiatric symptoms (in contrast to the previously described Breier study). To test this hypothesis we studied the long-term effects of childhood parental separation due to divorce on the HPA axis, using the Corticotropin Releasing Hormone (CRH) test, in a young adult population without DSM-IV axis I psychopathology, whose parents divorced (with one of them leaving home) before the subject turned 10 years of age. These criteria were selected to separate the effect of early stress due to divorce from that of current mental condition on the HPA axis. Our main hypotheses were: (a) divorce in early childhood would have long-term effects on the HPA axis in healthy young adults in the form of lower or normal basal cortisol levels, but an increased HPA axis response to CRH stimulation, (b) the effect of divorce on HPA axis response to CRH stimulation will be moderated by level of stress at home prior to the divorce and by the quality of bonding between the child and each one of the parents during childhood.
Results
First, we examined the degree of matching between the two groups at baseline. Table 2 presents means and standard deviations of symptom levels at baseline. As can be seen, baseline symptom levels, as assessed with the BSI, were rather low in both groups. This is not surprising considering that both groups were screened for Axis I diagnoses. As could be expected, subscale scores were slightly higher, though not significantly so, in the study (divorce) group across all the clinical measures. The only exception to this general rule was “hostility” which was significantly higher in the divorce group; however, this difference did not remain significant after adjusting the p-value for multiple comparisons using the Bonferroni correction.
Table 2.
BSI symptom scores according to subscales
Symptom subscale Separation group mean (SD) Control group mean (SD) Wilcoxon
Somatization 0.20 (0.36) 0.07 (0.13) 0.92
Obsessive–compulsive 0.52 (0.55) 0.37 (0.44) 0.84
Interpersonal sensitivity 0.66 (0.84) 0.38 (0.39) 0.70
Depression 0.58 (0.68) 0.39 (0.49) 0.48
Anxiety 1.36 (0.79) 1.09 (0.29) 1.27
Hostility 0.51 (0.56) 0.26 (0.35) 1.69
Phobia 0.42 (0.68) 0.19 (0.34) 1.10
Paranoia 0.53 (0.85) 0.33 (0.36) 0.01
Psychoticism 0.36 (0.56) 0.27 (0.43) 0.47
Total score 0.46 (0.56) 0.27 (0.27) 0.82
* = p < 0.05.
Table options
Next, we tested the first hypothesis regarding an increased response of the HPA axis to CRH stimulation in the divorce group. Fig. 1 presents the change over time in ACTH and cortisol levels in the 2 groups. A set of two-way ANOVAs revealed significant time effects for mean plasma ACTH response to an iv bolus of hCRH (F5,37 = 19.5, p < 0.01) and for cortisol (F5,39 = 16.2, p < 0.01).
Plasma ACTH and cortisol levels after stimulation by CRH (mean+SD). Inset—total ...
Fig. 1.
Plasma ACTH and cortisol levels after stimulation by CRH (mean + SD). Inset—total (left bars) and net (right bars) area under the curve (striped—subjects, white—controls). *Subjects vs. Controls 5 min after CRH injection; p = 0.025.
Figure options
While the group by time interaction effect was non-significant for ACTH (F5,39 = 0.5, p = 0.68), it approached significance for cortisol (F5,39 = 1.94, p = 0.06). Similarly, while the effect size (using Cohen's conventions, Cohen, 1988) for ACTH was small (ES = 0.20) the effect size for cortisol was medium (ES = − 0.41). Group means showed significantly lower cortisol plasma levels at 5 min after CRH injection for the separation group compared to the control group (F1,43 = 5.3, p = 0.025) ( Fig. 1). This overall pattern of results remains unchanged after repeating these analyses with ACTH levels at baseline as a covariate. Similarly, the results remained unchanged when the data wee reanalyzed by gender. In summary, while the lower cortisol level found in the separation group was as hypothesized, activation of the HPA axis subsequent to CRH challenge was negative for ACTH and only at a strong trend level for cortisol. Consequently, contrary to our hypothesis, the difference between HPA axis activation in the two groups was not significant.
No significant inter-group differences were found for either the calculated total and net AUC for ACTH and cortisol between the study and control group. A significant main effect of sex (female > male) on total AUC was found for both ACTH (F1,43 = 14.4, p < 0.01) and cortisol (F1,43 = 12.1, p < 0.01) but no significant total or net AUC by sex by group interaction was detected for either hormone.
Next, we tested the second hypothesis regarding the moderating role of level of stress at home on the effect of divorce on the HPA axis. To test this hypothesis, we reanalyzed our data with parental conflict (assessed with the DAS) as an interacting variable. First, we looked at the relationship between DAS and ACTH AUC and cortisol AUC. The Pearson correlation between these two variables was significant for ACTH AUC only (r = − 0.32. p < 0.05). A particularly strong correlation was found between ACTH levels and the measure of dyadic agreement. Next, we divided the group into high-conflict versus low-conflict using the median DAS score as a cut-off point. As would be expected, a chi-square analysis revealed a strong association between divorce status and level of parental conflict (χ2 = 23.3; p < 0.01). Finally, we conducted a hierarchical regression analysis with three steps. On the first, group (i.e., divorce status) only was entered. The obtained R-squares were 0.040 for ACTH and 0.037 for cortisol. On the second step, the effect of stress at home (DAS) was added to that of group. The resulting R-squares were 0.063 for ACTH and 0.038 for cortisol. Lastly, on the third step, the group by stress interaction term was added to the equation. The obtained R-squares were 0.065 for ACTH and 0.038 for cortisol. These results do not provide support for a moderating role of stress at home on the effect of divorce on HPA axis. However, at least for ACTH, they do suggest that stress at home has an important additive effect over and above that of divorce alone. In fact, for ACTH, stress at home was the best single predictor (R-square = 0.062).
Finally, we examined the potential moderating role of parental care and controlling patterns as measured by the PBI on the effect of divorce on HPA axis. First, we looked at the relationship between the PBI and ACTH AUC and cortisol AUC. No association was found between the Parental Care and Control subscales and total or net AUC for neither ACTH (r range = 0.02 to 0.30; p range = 0.92 to 0.06) nor cortisol (r range = 0.02 to − .19; p range = 0.90 to 0.23).
Second, we examined the effect of parental caring and controlling patterns on clinical symptoms in adulthood. Table 3 presents means and standard deviations on the PBI in the two groups. As can be seen, while a caring bond with the mother (care mother) is described in equal terms in both groups, in the early separation group there was a trend for the mother to be perceived as more controlling. The relationship with the father in the separation group was perceived as less caring and more controlling. For the whole group, a significant negative correlation was found between the care-father subscale of the PBI and total BSI score (r = − 0.38, p < 0.01) and a trend for significance between the control-mother subscale and total BSI scores (r = 0.28, p < 0.06).
Table 3.
Bonding pattern with parents; the Parental Bonding Instrument (PBI mother/father)
Parameters Separation group mean (SD) Control group mean (SD) Wilcoxon
Care-mother 25.64 (4.85) 26.82 (4.99) 0.50
Control-mother 11.95 (7.85) 7.72 (6.21) 3.63ϕ
Care-father 14.55 (7.97) 23.09 (6.65) 10.13**
Control-father 13.52 (6.96) 6.91 (8.06) 7.36**
ϕ = p < 0.06.
** = p < 0.01.
Table options
Lastly, we conducted a hierarchical regression analysis with three steps. On the first, group (i.e., divorce status) only was entered. The obtained R-squares were 0.047 for ACTH and 0.030 for cortisol. On the second step, the effect of mother attachment and father attachment was added to that of group. The resulting R-squares were 0.24 for ACTH and 0.07 for cortisol. On the third step, the group by mother and father interaction terms were added to the equation. The obtained R-squares were 0.26 for ACTH and 0.10 for cortisol. These results do not provide support for a moderating role of parental attachment on the effect of divorce on HPA axis. However, at least for ACTH, they do suggest that parental attachment, specifically maternal one, has an important additive effect over and above that of divorce alone. In fact, in this model for ACTH, maternal bonding was the best single predictor (R-square = 0.17).
