Analysis of the development of psychopathology in a large longitudinal sample of offspring of parents with and without panic disorder and major depression revealed that separation anxiety disorder significantly increased the risk for the subsequent development of agoraphobia, generalized anxiety disorder, panic disorder, and major depression, while agoraphobia significantly increased the risk for subsequent generalized anxiety disorder. These findings indicate that separation anxiety disorder may help identify a group of children at very high risk for a wide range of adverse psychopathological outcomes within a population already at risk by virtue of parental psychopathology.
Even after controlling for the presence of other anxiety disorders, separation anxiety disorder was the best predictor of panic disorder. These findings are consistent with an early literature by Gittelman and Klein (Gittelman and Klein, 1984
, Klein, 1995
) and by Weissman and colleagues (Weissman, et al., 1984
), that suggested a link between childhood separation anxiety disorder and risk for panic disorder in adulthood, and with more recent findings in referred and non-referred adults showing that separation anxiety disorder was the best predictor of early onset panic disorder (Biederman, et al., 2005
). This latter result confirms previously reported findings in the literature (Goodwin, et al., 2001
) and indicates that the association between separation anxiety disorder and early onset panic disorder is not due to referral bias, since this was an non-referred sample.
Our finding that separation anxiety disorder conferred risk for disorders other than panic as well is consistent with other studies (Aronson and Logue, 1987
, Moreau and Follett, 1993
, Otto, et al., 1994
, Pollack, et al., 1996
). For example, in a 7-year prospective follow-up study of referred children and adolescents diagnosed with separation anxiety disorder, generalized anxiety disorder, or social phobia (Aschenbrand, et al., 2003
), children with separation anxiety disorder did not show elevated rates of panic disorder compared to children with generalized anxiety disorder or social phobia, but did show higher rates of specific phobia, obsessive compulsive disorder, and posttraumatic stress disorder. Our results are also consistent with the literature from retrospective studies of adults. For example, Lipsitz et al (Lipsitz, et al., 1994
) found a significantly higher prevalence of childhood separation anxiety disorder in adults with two or more anxiety disorders suggesting that separation anxiety disorder may be a risk factor for multiple anxiety syndromes in adulthood. In addition, Biederman et al. (Biederman, et al., 2005
) found in samples of referred and non-referred adults that both separation anxiety disorder and overanxious disorder were independent predictors of subsequent panic disorder. This discrepancy with the current findings with regard to GAD may reflect a difference between pediatric and adult onset panic disorder. In fact, Biederman et al. (Biederman, et al., 2005
) showed that separation anxiety disorder was the best predictor of panic disorder in the adults who had an onset of panic disorder before age 18.
Our findings support the idea that childhood separation anxiety disorder may reflect an underlying anxiety diathesis that increases susceptibility to a range of subsequent adult anxiety disorders. Alternatively, separation anxiety in childhood may influence learned responses to anxiety and tendencies toward avoidance (Otto, et al., 2001
On the other hand, longitudinal findings by Pine et al. (Pine, et al., 1998
) did not find adolescent separation anxiety disorder (average age at baseline=13.7 years) to be a significant predictor of anxiety disorders or major depression at follow-up. Our younger sample (average age at baseline=6.4 years) enabled us to assess the longitudinal effects of childhood separation anxiety disorder.
The finding that pediatric agoraphobia was a predictor of subsequent generalized anxiety disorder is novel. Since this was not a hypothesized outcome, our results could be attributed to chance. As such, this finding should be considered preliminary until replicated and confirmed in future research.
The finding that social phobia did not predict subsequent panic disorder is discrepant with retrospective data reported by Otto et al (Otto, et al., 2001
, Otto, et al., 1994
). These investigators reported that social phobia was the most prevalent reported childhood anxiety disorder in a sample of 100 adult patients with panic disorder. Although the reasons for these discrepant finding are unknown, they could reflect differences between referred and non-referred individuals or the fact that panic disorder emerged in our sample in childhood whereas in the Otto et al. study (Otto, et al., 2001
, Otto, et al., 1994
) it did so in adulthood suggesting that different risk factors may be responsible for early versus late onset panic disorder. Other prospective studies have found that social phobia is a significant risk factor for subsequent depression (Bittner, et al., 2004
, Stein, et al., 2001
). Although our results did not support these previous findings, the young age of our sample meant that we could not fully assess the children's chance of developing these disorders. Following these offspring into later adolescence and adulthood may reveal the link between social phobia and major depression or between social phobia and adult-onset panic disorder.
Specific phobia, although common at baseline, did not significantly predict development of any anxiety disorders or major depression. This is consistent with the literature suggesting that specific phobias are common disorders of childhood that do not herald the likelihood of other, more severe forms of anxiety or mood disorders (Fyer, et al., 1995
, Pine, et al., 1998
, Shaffer, et al., 1996
Although major depression is frequently comorbid with panic disorder, in this sample panic disorder did not predict subsequent major depression, consistent with results from the National Comorbitity Survey (Kessler, et al., 1998
). However, since our sample is still within the period of risk for panic disorder, a longer follow-up is needed to confirm these findings.
Our results inform the targeting of prevention studies to individuals at risk for future anxiety disorders, namely offspring of parents with panic disorder or depression who present with separation anxiety disorder or agoraphobia. Studies of prevention protocols could focus on teaching children strategies for managing anxiety without resorting to avoidance behaviors, and on teaching parents ways to model and facilitate adaptive coping (Hirshfeld-Becker and Biederman, 2002
). Studies are needed to ascertain whether treating separation anxiety disorder or childhood agoraphobia in early childhood with cognitive-behavioral therapy could prevent the onset of subsequent disorders.
Our findings should be viewed in the context of the study's methodological limitations. Any interpretation of these findings must be tempered by the fact that these children are still early in the risk period for anxiety and depressive disorders. Future waves of assessment of this sample will deepen our understanding of trajectories of disorder, and may lead to different inferences about patterns of risk. In addition, our findings derive from a group of high-risk children, many of whom were the offspring of parents with panic disorder. It is not known whether these same trajectories would be seen in children without family history of panic disorder. The time between the onset of the predictor and outcome variables was not analyzed. However, because the analysis used baseline disorders to predict follow-up disorders, we can estimate that the average time between the onset of predictor disorders and outcome disorders would be the average time between assessments (i.e., 4.5 years). While subjects with separation anxiety disorder were at increased risk for agoraphobia, we did not examine whether these subjects continue to meet criteria for separation anxiety disorder. Considering the overlapping symptoms of these disorders, future work will benefit from further assessment of this issue. Because of the wide age range of the sample, informants of older versus younger participants may have differed in their recall of early symptoms. However, since age at follow-up was covaried, this difference would not be expected to affect our results.
Although raters were blind to the ascertainment status of the offspring, parents were not. The assessment of psychopathology in the children younger than 12 years was based on interviews with the mothers only, which may have led to underestimates of the true rates of psychopathology in offspring. Psychiatric disorders in parents may have affected the reported rates of childhood symptoms. Psychiatric patients may exaggerate symptoms in their children, or alternatively, mothers without psychopathology may under-report problem behaviors (Mick, et al., 2000
). We did not directly interview children under age twelve about their lifetime diagnoses, because younger children are limited in their expressive and receptive language abilities, lack the ability to map events in time, and have limited powers of abstraction. Several studies (Edelbrock, et al., 1985
, Schwab-Stone, et al., 1994
) have documented poor reliability of reports of psychopathology by children under twelve. In addition, Breton et al. (Breton, et al., 1995
) found that children aged 9 to 11 understood only about 40 percent of the questions from the Diagnostic Interview Schedule for Children (Version 2.5). In contrast, Faraone et al. (Faraone, et al., 1995
) and others (Fallon and Schwab-Stone, 1994
) have shown high reliability for maternal reports of psychopathology, even over a one-year period. In addition, through relying on parent-reports for diagnoses in children 11 and under and combining data from parent- and child-reports in children 12 and older, we introduced method variance in our study. However, we opted for this approach, because using only the parent reports for children over 11 might result in our missing important information about symptoms that adolescents, but not their parents, can report. Moreover, because the age of subjects (and therefore the assessment method used) was covaried in all analyses, this additional method variance cannot confound our findings. Future waves of data collection, in which all children in our sample will be old enough to be directly interviewed about their lifetime symptoms, will enable us to deepen our current findings. Although our study collected data at two time points, lifetime diagnoses were based on retrospective report and may have been subject to recall bias (Angold and Costello, 1996
). Because the proband parents were clinically referred, the generalizability of our findings is limited to offspring of referred subjects. Also, because the large majority of subjects were Caucasians from intact families and higher levels of social class, further work is needed to determine if our results generalize beyond these constraints.
Despite these considerations, this large prospective follow-up study of children at risk growing up provides compelling evidence for the critical role of separation anxiety disorder as a key antecedent risk factor for subsequent panic disorder. Results also indicate that separation anxiety disorder may help identify a group of children at very high risk for a range of adverse psychopathological outcomes. Further work is needed to determine if the results found in this unique sample reflect developmental patterns in the general population.