Social Anxiety Disorder and the Risk of Depression:  A Prospective Community Study of Adolescents and Young Adults FREE

SOCIAL ANXIETY disorder (SAD) (also known as "social phobia") is a prevalent disorder with its onset almost universally in childhood or adolescence.^1- 3 Recent estimates indicate that between 4% and 8% of adults in the general population suffer from SAD in a given year, with even higher rates when lifetime prevalence is considered.^4- 5 In a community study of adolescents and young adults aged 14 to 24 years, from which the current report is derived, we found a lifetime prevalence of DSM-IV SAD of 9.5% in females and 4.9% in males.^4,6- 7

Another characteristic feature of the longitudinal course of SAD, in addition to its early onset, is its frequent co-occurrence with depressive illness.^6- 12 Social anxiety disorder is reported to be the most commonly occurring comorbid anxiety disorder among patients with depressive disorders.¹³ Furthermore, when comorbidity does occur, SAD almost always starts first, often many years prior to the onset of depression.^14- 15 This consistent finding has spurred interest in the study of SAD as a possible risk factor for major depression.

A possible link between social anxiety and earlier onset of major depression has been reported in several studies.^16- 19 In a longitudinal study, anxiety disorders in early adolescence predicted clinically significant depressive and anxiety disorders (especially SAD) in early adulthood.²⁰ Furthermore, the association between early-onset (eg, prepubertal) anxiety and depression in young adulthood is evident when looking at family patterns of transmission in depressive high-risk families.^21- 22 These observations have sparked interest in the possibility that early identification of and intervention with socially anxious children or adolescents might reduce their risk for depressive disorders in later life.^6,19

In this report, we examined data from the Early Developmental Stages of Psychopathology (EDSP) Study,²³ focusing on the longitudinal relationship between SAD and depressive disorders. We hypothesized that SAD at baseline would predict onset and severity of subsequent depressive disorders.

Data were collected as part of the EDSP study. The EDSP is a prospective longitudinal study designed to collect data on the prevalence, risk factors, comorbidity, and course of mental and substance use disorders in a representative sample, which consisted of 3021 subjects aged 14 to 24 years at baseline. The study consists of a baseline (T0) survey, 2 follow-up surveys (T1 and T2), and a family history component.

The baseline sample was drawn in 1994 from government registries in metropolitan Munich, Germany, of registrants expected to be aged 14 to 24 years at the time of the baseline interview in 1995. Because the study was designed as a longitudinal panel with special emphasis on early developmental stages of psychopathology, 14- to 15-year-old individuals were sampled at twice the probability of people aged 16 to 21 years, and 22- to 24-year-old people were sampled at half the probability of the 16- to 21-year-old people.

Individuals were contacted first by letter, and then by telephone to arrange a meeting. Most interviews took place in the respondents' homes or, in some instances, at another location preferred by the respondent. Approximately one third of the sample received a financial incentive (US $10-$20) to participate. Participants provided informed consent; parental consent was provided for respondents aged 18 years and younger.

The demographic distribution of the sampled population and the respondents has been reported elsewhere.^23- 24 Briefly, among the sampled subjects, a total of 3021 interviews were completed, resulting in a response rate of 70.8%. At baseline, refusal to participate in the survey (18.2%) was by far the most frequent reason for nonresponse, followed by a reported lack of time (3.3%), failure to contact anyone in the identified household (3.1%), and failure to contact the sampled individual in the household (3.0%).

The first follow-up survey was conducted only for subjects aged 14 to 17 years at baseline, whereas the second follow-up survey was conducted for all subjects. At the first follow-up survey 14 to 25 months after baseline (mean interval, 20 months; SD, 3 months), a total of 1228 interviews were completed, resulting in a response rate of 88%. From the 3021 subjects of the baseline survey, a total of 2548 interviews were completed at the second follow-up survey 34 to 50 months after baseline (mean duration, 42 months; SD, 2 months), resulting in a response rate of 84%. A more detailed description of the study is presented elsewhere.²⁴

For those probands aged 14 to 17 years at baseline, the follow-up status is assessed from the aggregation of information obtained from the first and second follow-up interviews. For the probands older than 17 years at baseline, the follow-up status is assessed from the the second set of follow-up questions, which refer to the time between baseline and the second follow-up.

The survey staff throughout the entire study period (including the family history component of the study) consisted of 57 clinical interviewers, most of whom were clinical psychologists with extensive experience in diagnostic interviewing, including the Munich-Composite International Diagnostic Interview (M-CIDI).^25- 26 At baseline, 25 professional health research interviewers recruited from a survey company were also involved. Formal training with the M-CIDI took place for 2 weeks, followed by at least 10 closely monitored practice interviews and additional 1-day booster sessions throughout the study.

The M-CIDI allows for the assessment of symptoms, syndromes, and diagnoses of 48 mental disorders, along with information about onset, duration, severity, and psychosocial impairment. Diagnostic findings reported in this article were obtained by using the M-CIDI/DSM-IV diagnostic algorithms. Test-retest reliability and validity for the full M-CIDI have been reported elsewhere,^26- 28 along with descriptions of the M-CIDI format and coding conventions.

Social anxiety disorder is defined here as one meeting DSM-IV criteria per the M-CIDI diagnostic algorithm.⁶Depressive disorder is defined as one meeting DSM-IV criteria for one or more episodes of major depression or dysthymia. One-week test-retest reliability of these diagnostic categories was acceptable (all κ values, > 0. 64),²⁶ as was their validity (κ values range, 0.54 for dysthymia to 0.96 for single depressive episodes). Descriptors of the course of depression (eg, number of depressive episodes), which was also derived from the M-CIDI, refer to the time between baseline and follow-up. Severity descriptors (eg, number of depressive symptoms) refer to the self-identified worst episode of depression during this interval. The variable "total duration of depression" was estimated in weeks by multiplying the number of depressive episodes by the duration of the longest depressive episode. The variable "suicidal ideas" refers to the number of endorsed items from a total of 4 possibilities: (1) frequent thoughts of death, (2) desire for death, (3) suicidal thoughts, and (4) concrete suicidal plans or attempts.

Data were weighted to consider different sampling probabilities as well as systematic nonresponse at baseline. The Stata Software²⁹ package was used to compute robust confidence intervals (eg, by applying the Huber-White sandwich matrix in the case of regression models)³⁰ required when basing analyses on weighted sample sizes. Logistic regressions with odds ratios (OR) were used to describe associations with onset and stability of depressive disorders, recognizing confounding variables such as subjects' age, sex, or substance abuse or dependence.³¹ We also conducted most analyses omitting the subjects (n = 600) who suffered from alcohol abuse or illicit drug abuse or dependence at any time point. The results of these analyses did not differ meaningfully from those in which the full sample was included; we have therefore elected to report only results for the full sample. The quantitative outcomes of severity of depressive disorder considered here (eg, number of depressive episodes) constitute count variables with a strongly positively skewed distribution. For this, negative binomial regressions were used with multiplicative effects described by so-called incidence rate ratios (IRR) (ie, the factor by which the mean differs from the one in the comparison group). Negative binomial regressions allow for extra-Poisson variation or overdispersion that is likely to be owing to unobserved heterogeneity in the outcome between subjects as well as correlated events that are counted (eg, symptoms),³² and 95% confidence intervals (CI) are used throughout this article.

Sociodemographic characteristics of the sample at baseline are summarized in Table 1. A total of 3021 cases were available at baseline, with data from the second follow-up available for 2548 cases.

Baseline prevalence rates (lifetime and past 12 months) of SAD and depressive disorder are presented in Table 2; 27.3% of cases with lifetime social phobia were of the generalized subtype. These disorders are categorized into 3 mutually exclusive combinations (ie, SAD without depressive disorder, depressive disorder without SAD, depressive disorder with SAD) to permit comparison of their longitudinal outcomes at follow-up. Depressive disorders (OR = 1.9; 95% CI, 1.5-2.5) and SAD (OR = 2.0; 95% CI, 1.4-2.9) were more common in women than men; associations were therefore adjusted for sex.

Rates of depression during the period between baseline and the second follow-up are presented in Table 3. Persons with SAD but no depression (current or previous) at baseline were significantly more likely (OR = 3.5; 95% CI, 2.0-6.0) than persons with no mental disorder to have experienced a depressive disorder during the follow-up period. This effect, however, could be detected only for the older (ie, aged 18-24 years at baseline) (OR = 5.3; 95% CI, 1.6-3.8), but not the younger (ie, aged 14-17 years at baseline) subsample (OR = 0.6; 95% CI, 0.1-2.7).

A similar significant increase in odds (OR = 3.8; 95% CI, 2.6-5.5) was seen for persons with depression but no SAD at baseline. Persons with depression and SAD (current or previous) at baseline were also at significantly amplified odds for subsequent depression (OR = 8.7; 95% CI, 4.5-16.8) compared with persons with no mental disorder at baseline. In persons with depressive disorder at baseline, SAD (current or previous) at baseline approximately doubled the odds for subsequent (or persistent, as this might reflect a continuous episode from baseline to follow-up) depressive disorder (OR = 2.3; 95% CI, 1.2-4.6).

Among persons with SAD, age at onset of social anxiety symptoms (not disorder, which was unavailable)³³ was found to predict neither depressive disorder onset (OR = 1.1; 95% CI, 0.9-1.2) nor persistence or recurrence (OR = 1.0; 95% CI, 0.9-1.1).

Specificity of the effects attributable to SAD (vis-á-vis other anxiety disorders) in predicting onset and recurrence or persistence of depressive disorder was not high. Other anxiety disorders (specific phobias and generalized anxiety disorder in particular) at baseline were also associated with similarly increased odds (data not shown). We therefore examined the possibility that the associations we had linked to SAD might be due to comorbidity with other anxiety disorders. Analyses were repeated where SAD groups were stratified for the presence or absence of at least one other comorbid anxiety disorder at baseline. The association with the onset of depressive disorder was somewhat higher in cases with SAD plus another comorbid anxiety disorder (OR = 6.0; 95% CI, 2.6-13.3) than in cases with SAD alone (OR = 2.4; 95% CI, 1.2-4.9), though not significantly so. No appreciable difference in association was found for the persistence or recurrence of depressive disorder in cases with SAD alone (OR = 7.7; 95% CI, 2.5-23.7) compared with SAD plus another comorbid anxiety disorder (OR = 9.2; 95% CI, 4.2-19.9).

Several clinical characteristics and descriptors of the course of depressive illness during the follow-up period were compared between groups (Table 4). Compared with persons with no mental disorder at baseline, there were few significant differences in these parameters for persons with either a depressive disorder alone or SAD alone at baseline. The sole exception was a small but statistically significant increase in the severity of depression (as indicated by the number of depressive symptoms experienced during an episode) for persons with depression alone at baseline compared with persons not mentally ill at baseline (IRR, 1.2; 95% CI, 1.1-1.4).

A very different portrait of risk emerged, however, for persons who had both depressive disorder and SAD at baseline (Table 4). During the follow-up period, these persons were significantly more likely to experience (compared with persons with no mental disorder) more intense suicidal ideation (IRR, 2.2; 95% CI, 1.4-3.4), more depressive symptoms (IRR, 1.4; 95% CI, 1.2-1.7), and a longer duration of major depressive episode(s) (IRR, 3.2; 95% CI, 1.5-6.8). They also had much greater odds (OR = 7.0; 95% CI, 1.5-32.2) of having attempted suicide during the follow-up period. Similarly increased magnitudes of effects were seen for this comorbid group (ie, positive for both depressive disorder and SAD at baseline) in comparison with persons with depressive disorder alone at baseline (Table 4, bottom row).

In this prospective study, we found that the presence of SAD in adolescence or early adulthood is a strong risk factor for the subsequent occurrence of depressive illness during young adulthood. Moreover, the combination of depression and SAD in adolescence markedly augments the risk for subsequent depressive disorder, over and above the risk conferred by either disorder alone. Thus, in addition to confirming the findings from retrospective reports that preexisting SAD increases the risk for "early-onset" depression,^16- 17,19 our observations suggest that those persons with the combination of SAD and depression in adolescence or early adulthood are at the greatest risk for subsequent depression.

Individuals with this early form of comorbidity (ie, SAD plus depressive disorder) are not only at highest risk for subsequent depression, they also experience a more malignant course of depressive illness. This is manifested in more suicidal ideation and suicide attempts, and more depressive symptoms during episodes, as well as more frequent and/or more protracted depressive episodes. These findings are consistent with observations from other settings (eg, primary care) where the presence of SAD comorbidity predicts poorer depressive outcomes.³⁴ Remaining to be shown is whether or not these associations are unique to SAD or whether they are seen with other forms of anxiety disorder comorbidity. Our preliminary look at these data suggests that the association with SAD is not specific, but that it must be scrutinized more closely in future analyses. Regardless, the fact that SAD may be the most common form of anxiety disorder comorbidity seen in depressed patients¹³ makes this a particularly salient observation.

It is important to emphasize that causal inferences cannot be drawn from these observational data. We may speculate, however, about the mechanism(s) by which preexisting SAD might increase the risk of subsequent depression. Certainly, common genetic risk factors may exist.^35- 36 Investigators have theorized a role of social anxiety and avoidance in contributing to demoralization and social isolation,⁸ all of which are known depressive risk factors. Socially anxious children are more likely than their less socially anxious peers to develop problems with self-esteem and lack friendships.³⁷ It is therefore reasonable to posit that a cause-and-effect relationship between social anxiety and depression exists at least in some cases, but this remains to be empirically proven.

Limitations of our study should be considered. Although we did not find selective attrition for persons with social phobia from baseline to second follow-up investigation in our sample, it is possible that persons most impaired by SAD did not participate in our study. It is also possible that our findings from this urban German sample, consisting of persons who are well educated with relatively high economic status, may not generalize to other populations. Diagnoses at baseline are retrospective and therefore subject to possible recall problems or biases, though these should be attenuated in this relatively young sample. Some of our findings (eg, increased suicide attempts in comorbid cases) rest on relatively few cases, and though statistically significant, should be interpreted with caution. Also, as mentioned earlier, it seems that the association with subsequent depressive disorder is not specific for SAD. It is also possible that depressive disorders are not the most common (or important) kinds of mental disorders predicted by SAD. We intend to conduct additional analyses to further explore these latter issues.

Our findings are consistent with other longitudinal studies showing anxiety disorders in youth to be a predictor of more serious depression in adulthood, particularly in those at risk for depression on the basis of family history.²² Though we caution once again about imparting causality to our findings, they do support the proposal made by numerous investigators that early intervention with socially phobic youth be tested as a primary prevention of depressive illness.^6,12,15,19 Furthermore, our results suggest that the union of depression with SAD (and probably other anxiety disorders as well) in adolescence or early adulthood is a particularly sinister combination that heralds an increased risk for subsequent depressive episode(s) of increased severity, with amplified suicide risk. Given the substantial morbidity and mortality risks associated with adolescent-onset major depressive disorder,^38- 39 serious efforts should be initiated to identify and test treatments for youth who fit this clinical profile (ie, early-onset anxiety and depressive disorders).

Accepted for publication September 25, 2000.

This work is part of the Early Developmental Stages of Psychopathology (EDSP) Study and is funded by the German Ministry of Research and Technology, project 01 EB 9405/6.

The principal investigators are Hans-Ulrich Wittchen, PhD, and Roselind Lieb, PhD. Current or former staff members of the EDSP group are Kirsten vonSydow, PhD; Gabriele Lachner, PhD; Axel Perkonigg, PhD; Peter Schuster, PhD; Franz Gander, PhD; Michael Höfler, DiplStat; and Holger Sonntag, DiplPsych, as well as Esther Beloch, Mag Phil; Martina Fuetsch, MagRerNat; Elzbieta Garczynski, DiplPsych; Alexandra Holly, DiplPsych; Barbara Isensee, DiplPsych; Marianne Mastaler, DiplPsych; Nina Müller, DiplPsych; Chris Nelson, PhD; Hildegard Pfister, DiplInf; Victoria Reed, DiplPsych; Dilek Türk, DiplPsych; Antonia Vossen, DiplPsych; Ursula Wunderlich, PhD; and Petra Zimmermann, DiplPsych. Scientific advisors are Jules Angst, PhD (Zurich, Switzerland); Jürgen Margraf, PhD (Basel, Switzerland); Günther Esser, PhD (Mannheim, Germany); Kathleen Merikangas, PhD (New Haven, Conn); and Ron Kessler, PhD (Boston, Mass).

Corresponding author and reprints: Murray B. Stein, MD, Department of Psychiatry, University of California San Diego, 8950 Villa La Jolla Dr, Suite 2243, La Jolla, CA 92037 (e-mail: mstein@ucsd.edu).