Abstract
Exposure to aberrant parenting appears to be a risk factor not only to childhood psychiatric disorders but also to anxiety and depressive disorders in adulthood [1–5]. Nevertheless, uncertainty persists as to whether deviant parenting constitutes a generic risk factor to a wide range of anxiety and depressive disorders in the offspring, or whether exposure to specific parental styles, such as lack of care or overprotectiveness, can be linked to particular psychiatric outcomes in the subsequent generation [6].
Bowlby [7, [8]] suggested that maternal overprotectiveness may be a key factor in the genesis of separation anxiety (SA) in affected children. He further argued that SA is the underlying anxiety in cases of school phobia in early life and agoraphobia in adulthood. Earlier studies based on clinical observations [9–11] tended to support Bowlby's assertion that patients with agoraphobia came from closely knit family backgrounds characterised by parental over-protectiveness. However, the majority of more recent systematic studies [12–14] have tended to show that patients with agoraphobia, like those with other anxiety and depressive disorders [4, [15]], characterise their parents as having been uncaring or negligent, with overprotectiveness being a more variable finding [6, [16]].
Several reasons may account for the discrepancy between Bowlby's initial observations and the findings of recent research. First, parental overprotectiveness was assessed by clinical impressions in earlier studies [9–11], whereas in recent studies, indices of care and overprotectiveness have been derived from psychometrically tested measures (the Parental Bonding Instrument [17]; the Egna Minnen Betraffinde Uppfostran [13]). The benefit of using such instruments is that the results are quantifiable and reproducible, but it is also possible that some of the nuances of clinical observations may be lost. Nevertheless, studies have demonstrated sound levels of reliability and validity for the Parental Bonding Instrument (PBI) and the measure appears to provide as accurate a measure of the dimensions of care and overprotection as more comprehensive interview approaches [18]. A second factor that may account for discrepancies in studies over time is the historical shift in the diagnostic criteria and hence the population being defined by the category of agoraphobia. Increasingly, the diagnosis of agoraphobia has been restricted to a specific constellation of avoidance behaviour that occurs secondary to the development of panic attacks [19–21]. It seems likely, therefore, that the way Bowlby [7, [8]] used the term ‘agoraphobia’ was sufficiently broad to include cases of adult SA, whereas the overlap may not be as exhaustive when contemporary DSM diagnostic criteria for agoraphobia are used.
Such a shift in the boundaries of agoraphobia thus may account in part for the persisting uncertainties about the relationship of early SA to panic disorder and agoraphobia (PD-Ag). Klein [22] proposed that early SAwas a specific precursor to the development of PD-Ag in adulthood, but several studies have shown that early SA is evident in only 40–50% of PD-Ag patients [6, [3], [24]]. Also, some studies [25–28] have suggested that early SA is not exclusively linked to PD-Ag, but that it may be a generic risk factor to severe anxiety in adulthood. It has been suggested [29] that one possible reason for the ongoing doubt about the outcome of early SA is that researchers have failed to include a category of separation anxiety disorder in the possible range of anxiety subtypes in adulthood.
Separation anxiety in adulthood
Two recent reports [30, [31]] have suggested that a constellation of SAsymptoms can persist from childhood into the adult years. Core SAsymptoms identified in adulthood included extreme anxiety when separated from major attachment figures and fears that harm might befall those close family members [30, [31]]. In a recent study [31], comorbid disorders such as major depressive disorder and PD-Ag were common but, in most instances, SAdisorder predated the additional diagnoses. If adult separation anxiety disorder (AS AD) is an identifiable category, then it is possible that it may overlap with or be misdiagnosed as agoraphobia.
The present study aimed to examine whether persons with a putative diagnosis of ASAD reported a distinctive pattern of heightened SA in early life and exposure to parental overprotection, thus confirming a trend that was found in previous studies [Silove D: unpublished data]. Anticipating substantial overlap between patients with ASAD and PD-Ag, we selected two samples, each giving differential priority to one of the target disorders. The first sample consisted of volunteers who responded to a media advertisement aimed at recruiting adults with high levels of SA, while the second sample consisted of patients referred to an anxiety clinic for the treatment of PD-Ag.
Method
Subjects
The first sample (the ‘media group’) comprised 34 subjects who responded to an advertisement aimed at recruiting adults whose major anxieties focused on separation from key attachment figures. The characteristics of this sample have been described in detail elsewhere [31]. In brief, advertisements were placed in newspapers and these in turn led to a series of radio interviews. Respondents were screened briefly for SAover the telephone prior to being interviewed. The second group (the ‘clinic group’) comprised 37 patients referred to an anxiety clinic with a diagnosis of panic disorder with or without agoraphobia (PD-Ag; n=37). Both groups were blind to the hypotheses of the study.
As a comparison group for parental antecedents, a sample of 71 subjects (not screened for psychiatric symptoms) matched for sex and age (within 2 years), were drawn from a larger pool of consecutive atten-ders at general practice clinics [17]. Another group of consecutive general practice attenders (n=79) described in a previous study [32] was used as a normative group for comparing early SA scores. Sixty-six percent (n=54) of the second control group were females with a mean age for the whole sample being 35 (SD=16) years.
All subjects were aged over 18 years and had given their written consent on a form approved by the University of New South Wales Ethics Committee.
Measures
The Parental Bonding Instrument (PBI) [17, [33]] is an extensively used retrospective measure that allows adults to rate their early parenting on dimensions of ‘care’ and ‘overprotection’. Test-retest and split-half studies have demonstrated the measure's high level of reliability with scores remaining consistent in spite of changes in respondents' mood states [17, [33], [34]]. Validity studies using corroborative witnesses (e.g. mothers, siblings) [34] have indicated that perceived parenting as recorded on the PBI does not differ substantially from observer reports.
All subjects completed the Separation Anxiety Symptom Inventory (SASI) [32], a 15-item self-report inventory developed to record adults' memories of SAexperiences over the first 18 years of their lives. Individual items are rated on a four-point scale ranging from ‘0’(I never had this feeling) to ‘3’(this feeling occurred very often). A summary score on the measure is derived by adding item scores and applying a square root transformation. The measure has a coherent factor structure, high internal consistency (Cronbach's α ranging from 0.84 to 0.88) and sound test-retest reliability (intraclass correlation coefficients ranging from 0.86 to 0.98). Separation Anxiety Symptom Inventory scores were found to remain stable over repeated tests in patients whose state anxiety and depression scores changed over a period of 18 months [32]. Furthermore, close agreement was achieved between SASI scores and lifetime diagnoses of juvenile separation anxiety disorder made by independent clinical interviewers [32]. In a further test of the SASI's validity, each member of twin pairs was asked to complete a checklist of descriptors depicting early insecure behaviours observed in the other member of the pair. Separation Anxiety Symptom Inventory scores corresponded with observer twins' ratings on the proband, thus providing a crude index of the SASI's concurrent validity [32].
The SCID-P [35] was used to assess lifetime PD-Ag according to DSM-III-R. The diagnostic interview was conducted by clinical psychologists and a psychiatrist, all of whom had previously achieved a high level of interrater reliability on the SCID with the first investigator (VM).
For the media sample, assignment to the provisional category of ASAD was based on an interview conducted by the first author (VM) using the Adult Separation Anxiety Semi-Structured Interview (ASA-SI) [31]. The ASA-SI consists of 27 questions designed to assess the extent of subjects' anxieties about their attachments to persons identified as close to them. The interview enables a global assignment (present or absent) to be made according to the provisional category of ASAD. Inter-rater reliability for such assignments was high (100%), and close concordance was demonstrated between global assignments made by interview and DSM-IV criteria for juvenile separation anxiety disorder modified for adulthood [31].
For the anxiety clinic patients, assignment to the ASAD category was made using a shortened version of the Adult Separation Anxiety Checklist (ASA-CL) [31]. Items correspond to relevant questions on the ASA-SI but are self-rated on a four-point scale ranging from ‘0’ (‘this has never happened’) to ‘3’ (‘this happens very often’). A study of adult anxiety clinic patients found that a predetermined cut-off score of 17 or greater on the 16-item version of the ASA-CL corresponded closely to ASA-SI assignments (specificity and sensitivity exceeding 80%) [36].
Methods
Demographic characteristics
Parental Bonding Instrument comparisons
The combined anxiety group (n=71) differed from primary care clinic controls (n=71) on all PBI sub-scale scores with means, standard deviations (SD) and paired two-tailed t-tests as follows (anxiety group reported first in all contrasts): maternal care 21.4 (SD=9.7) versus 28.8 (SD=6.5) (t=5.1, df=70, p< 0.001); maternal overprotection 17.3 (SD=8.2) versus 11.9 (SD=7.1) (t=4.3, df=70, p<0.001); paternal care 18.7 (SD=9.7) versus 23.9 (SD=7.4) (t=3.5, df=70, p<0.001); and paternal overprotection 16.1 (SD=8.9) versus 12.3 (SD=7.3) (t=2.8, df=70, p<0.01). Significant differences were also found in comparisons between SASI scores for the combined anxiety group (n=71) and the second primary care comparison group (n=79; 3.6 vs 3.1, respectively; t=2.2, df= 138, p<0.05).
More distinctive PBI and SASI patterns emerged, however, once comparisons were made within the combined anxiety group when it was dichotomised according to ASAD and PD-Ag status, respectively. A power analysis indicated that the anxiety subgroups were sufficiently large to detect a significant difference (p<0.05) for a moderate effect size (0.3). Anxious subjects were first assigned to the ASAD category and then compared with the residual anxiety group on SASI and PBI scores. Subjects assigned to the ASAD category returned statistically higher SASI scores than those not so assigned (3.8 vs 2.4; t=3.2, df=69, p<0.01). No differences emerged between the two groups on PBI dimensions of paternal or maternal care. In contrast, subjects assigned to the ASAD category reported statistically higher levels of maternal overprotection (18.1 vs 12.8; t=2.0, df=69, p<0.05) than the residual anxious group. It is noteworthy that the residual group returned maternal overprotection scores that were no different from those of the general practice controls (12.8 vs 11.9). A subgroup of subjects with the sole diagnosis of ASAD and no comorbid PD-Ag (n=29) returned similarly higher maternal overprotection scores (19.4 vs 12.8; t=2.4, df=38, p<0.05) compared to residual anxious subjects, suggesting that the effect was not due to comorbidity.
Using assignment to the ASAD category as the dependent variable, the four PBI subscale scores were entered into a logistic regression analysis. The overall Chi-squared for the regression equation was statistically significant (Chi-squared=3.9, p<0.05) with maternal overprotection being the only significant predictor of ASAD status (B=0.08, p<0.05).
Parental Bonding Instrument (PBI) subscale scores for subjects assigned to categories of adult separation anxiety and panic disorder-agoraphobia
Discussion
The most striking finding from the present study was that subjects assigned to the ASAD category reported high levels of maternal overprotection. Past research has found such elevated levels of parental overprotection in conjunction with normal levels of care in only a few psychiatrically affected groups such as those characterised by high levels of ‘dependency’ [34] and hypochondriasis [34, [37]]. In contrast, subjects in the present study who were assigned to the ASAD category reported no deviations in parental care even though deficits in that domain constitute the key bonding problem reported in a wide range of other anxiety and depressive disorders [4, [6]]. Furthermore, parental overprotection reported by subjects assigned to the ASAD category only involved mothers and not fathers.
In comparison, PD-Ag subjects showed fewer distinctive parental bonding characteristics. No statistically significant differences were evident on the overprotection subscales and maternal overprotection did not differ from levels reported by controls. Although somewhat lower levels of paternal over-protection and higher levels of paternal neglect were reported, these differences disappeared in the multivariate analysis.
In addition, differences emerged in relation to early SA. Subjects assigned to the ASAD category reported substantially increased levels of early SA (SASI score=3.8) than those not so assigned (SASI score=2.4) (p<0.01). It should be noted that SASI scores are subjected to a square root transformation so that the absolute difference in raw scores was substantially greater (14.4 vs 5.8). Adult separation anxiety disorder subjects' scores were among the highest recorded in patient samples [24], whereas those not so assigned scored in the range of non-psychiatric normative groups [32]. In contrast, PD-Ag patients reported a mean SASI score (3.6) that was identical to that for non-PD-Ag anxiety patients (3.6).
Prior to offering an interpretation of the results, the limitations of the study need to be acknowledged. As expected, there was overlap between subjects assigned to the diagnoses of ASAD and PD-Ag, making statistical comparisons more complex. The relatively small sample size meant that the three diagnostic subcategories (those with PD-Ag alone, those with both diagnoses, and those with ASAD alone) could not be entered into a multivariate analysis. Such an approach undertaken on a larger sample might allow more definitive conclusions to be drawn. Also, the control general practice groups were not screened for anxiety. Nevertheless, the inclusion of anxiety patients in the general practice sample and the overlap of diagnostic groups would both favour the null hypothesis by reducing the likelihood of finding between-group differences on the core developmental variables measured.
The design of the study makes it difficult to offer an aetiological explanation for the development of ASAD. It is possible that an underlying factor, such as a particular personality style, may result in the development of ASAD and also contribute to a bias towards reporting parental overprotection. Although research has been carried out to test the validity of both the PBI and the SASI, retrospective bias remains a potential confounder that cannot be entirely dismissed. Nevertheless, global biases in reporting would not account for the differential patterns obtained, especially the selective reporting of high maternal overprotection but not parental neglect by those assigned to the ASAD category.
As indicated earlier, the diagnosis of ASAD can only be regarded as provisional, even though it has received some support from recent research [30, [31]]. That research has suggested that, in the majority of cases, ASAD is assigned as the primary diagnosis, with PD-Ag, where present, being a secondary complication. Nevertheless, substantial further work will be necessary to confirm the independent nosological status of ASAD as a subtype of adult anxiety disorder. It should also be noted that different methods were used to assign subjects to the ASAD category in the two samples. However, a previous study [36] has found close agreement between the two methods for assignment to the ASAD category (specificity and sensitivity exceeding 80%). Samples used in the present study were not randomly selected. The media sample comprised a self-selected group who acknowledged high levels of SA. The second sample was obtained from a specialist clinic for anxiety disorders. Although an attempt was made to balance the two samples in a manner which gave similar priority to each diagnosis, it cannot be assumed that the composite sample was representative of cases in the general population. Thus, as yet, the results yielded cannot be generalised to other populations.
In spite of these limitations, the findings of the study provide some tentative evidence that may lead to a reconsideration of developmental theories of SA. As indicated, Bowlby [8] suggested that maternal overprotectiveness may lead to SA in the affected child, who was then at risk of developing school phobia in the early years and agoraphobia in adulthood. Later studies, using a more restricted definition of agoraphobia have indicated that subjects with PD-Ag more commonly report exposure to parental neglect rather than parental overprotectiveness [6]. Also, although there is a tendency for PD-Ag patients to report higher levels of SA in childhood [24], only approximately 40% of PD-Ag sufferers report such early symptoms. The present data may assist in clarifying these previous observations by suggesting that early SA and parental overprotection may not be directly associated with risk to later PD-Ag but rather with the persistence of SAsymptoms into adulthood.
Under conditions of interpersonal stress, persons with ASAD may be at risk of developing secondary symptoms of panic, which then may be complicated by symptoms of agoraphobia [30].
An important corollary is that a substantial number of patients with PD-Ag do not report heightened SA either in childhood or adulthood and that group does not characterise their mothers as having been over-protective. Thus, taken together, the findings suggest that there may be heterogeneous developmental pathways leading to PD-Ag in adulthood. It is possible that further research may assist in identifying more clearly an ‘at risk’ SA group who could be targeted for early family intervention programs.
Conclusions
The present study reported higher levels of maternal overprotection in subjects with a putative diagnosis of ASAD. Those subjects also reported higher levels of early SA. In contrast, the diagnosis of PD-Ag in itself was not associated with reports of heightened early SA or parental overprotectiveness, although there was a trend for such subjects to report a degree of paternal neglect. Although limited by the provisional nature of the ASAD diagnosis, overlapping diagnostic groups, and the retrospective nature of some of the data, the study does support the possibility that there may be differential developmental pathways leading to adult SAsymptoms and PD-Ag.