Subjects

Subjects were 520 consecutive patients attending an outpatient anxiety clinic in Sydney, Australia, between 1999 and 2004. The clinic is the only public service of its kind in the catchment area, providing cost-free outpatient cognitive behavioural treatments for the full range of adult anxiety disorders. The diagnostic profile of patients attending the clinic is similar to that of comparable services in other English-speaking countries [11]. Patients in the study were mainly referred by primary care providers with non-specific diagnoses of "anxiety". Eligibility for intake is not influenced by either the duration of symptoms or history of prior treatment. At the initial intake assessment, psychologists at the clinic administered the anxiety and mood disorder modules (A and F) of the Structured Clinical Interview [SCID-I/P, [12]] to assign relevant DSM-IV-TR diagnoses. The depression module was included because of the known pattern of comorbidity within the affective disorders. Psychologists recorded all DSM-IV-TR anxiety and depressive diagnoses. If more than one disorder was identified, they used their clinical judgement to decide which disorder represented the primary problem, based on symptom severity, patient-perceived salience of the problem and associated disability. If a depressive disorder was judged to be the dominant problem, patients were referred to other relevant services. In addition, a comprehensive clinical interview was undertaken to detect other disorders such as psychosis (rarely presenting to the clinic), and if detected, these patients were referred to other services. All psychologists had received extensive training in the application of the SCID-I/P and they were required to achieve 100% inter-rater reliability with the senior clinical psychologist (at the time of the study, RW, who had over 20 years of clinical experience) prior to undertaking assessments at the clinic.

Initial examination of the data indicated relatively low numbers with a primary diagnosis of obsessive compulsive disorder (OCD, n = 23) and post-traumatic stress disorder (PTSD, n = 18). The low referral pattern for these disorders was most likely due to the availability of specialist clinics for these two conditions in Sydney. Hence, those referred to our clinic would not be typical of a help-seeking population with the relevant diagnoses, and the small cell sizes would not allow these categories to be validly included in the statistical analyses we intended to undertake. For these reasons, the categories of OCD and PTSD were excluded from further consideration in the present study. Hence, the primary DSM-IV anxiety diagnoses included in the present study were: panic disorder (PD), panic-agoraphobia (PD-AG), generalised anxiety disorder (GAD) and social phobia (SP). Comorbid mood disorders included major depressive disorder, major depressive episode and dysthymia. Because of the limited numbers assigned to each of these depressive categories, they were collapsed into a composite grouping, "current depression". Following the clinical interview undertaken at the first intake session, patients were familiarised with, and where there was a need, guided through the completion of a number of self-report questionnaires (see hereunder).

All patients signed consent forms in accordance with the ethics requirements of the Sydney South West Area Health Service.

Measures

Modules A and F of the Structured Clinical Interview for DSM-IV-TR - SCID-I/P [12] were used. The SCID-I/P is a clinician-administered semi-structured interview for diagnosing Axis 1 disorders. Reliability coefficients from other studies have yielded kappa coefficients ranging from 0.77 to 0.95 for the relevant anxiety disorders [13].

The Adult Separation Anxiety Symptom Questionnaire - ASA-27 [14] is a 27-item self-report measure with items rated on a scale from 0 (this never happens) to 3 (this happens all the time). The psychometric characteristics of the measure have been described previously [14]. The measure has been compared with a semi-structured clinical interview (the Adult Separation Anxiety Semi-structured Interview), modelled on the SCID. A high area under the curve (AUC) value of 0.9 [14] indicated an excellent level of concordance between the two instruments.

ASAD diagnoses were based on an algorithm derived from DSM-IV-TR symptom criteria for separation anxiety disorder [15], excluding the provision that symptoms had to commence in childhood. Additional file 1 shows the items in the measure that correspond to the relevant DSM-IV-TR criteria. As an example, question 2 in the ASA-27 inquires about anxieties about leaving home, reflecting the DSM-IV-TR criterion of recurrent excessive distress when separation from home or major attachment figures occurs or is anticipated. We then applied the DSM-IV-TR threshold of three or more symptoms (derived from the childhood-onset category) to assign a diagnosis of ASAD.

The Depression Anxiety Stress Scale - DASS-21 [16] is a 21-item self-report measure that provides continuous scores on three subscales of depression, anxiety and stress, recorded for the past week. Items are scored from 0 (did not apply to me at all) to 3 (applied to me very much, or most of the time). High levels of severity on this measure are indicated by scores of 20, 14 and 26 or greater for depression, anxiety and stress, respectively. In the development of the measure, individual scales yielded Cronbach's alphas of 0.94 (depression), 0.87 (anxiety) and 0.91 (stress) [17].

The Work and Social Adjustment Scale - WSAS [18] is a self-report measure comprising subscales assessing functional impairment in the areas of work, home management, social leisure activities, private leisure activities (eg reading, gardening, etc) and close relationships. Items are rated on a Likert scale from 0 (affected "not at all") to 8 (affected "very severely, I never do these activities"). The measure has sound test-retest reliability and convergent validity [18]. A total score above 20 indicates high levels of functional impairment associated with a severe disorder; scores of 10 - 20 indicate significant impairment associated with mild to moderate level disorders; and scores below 10 are typical of a non-clinical population.

The Revised NEO Personality Inventory - NEO PI-R [19] is a self-completed scale measuring five personality traits: neuroticism, extraversion, openness, agreeableness and conscientiousness. Responses are coded on a five point scale ranging from "strongly disagree" to "strongly agree". Psychometric testing has supported the internal reliability of the scales. Normative data have been provided elsewhere [19]. In the present study, in order to facilitate statistical analysis, the personality dimensions were analysed as continuous indices.

The Separation Anxiety Symptom Inventory - SASI [20] is a 15-item self-report measure assessing separation anxiety symptoms retrospectively, based on experiences prior to 18 years of age. Items are scored from 0 to 3 on a frequency scale. The SASI has been shown to have sound internal (Cronbach's alpha = 0.88) and test-retest reliability over 24 months (intraclass correlation coefficient = 0.89). In the development of the measure, distributions were found to be skewed, a pattern adjusted for by applying a square root transformation. Hence, a raw score of 16 generates a transformed score of 4, whereas a score of 9 transforms into a score of 3. In past studies, mean transformed SASI scores of 4 or more have been associated with reports of past childhood separation anxiety disorder and/or school refusal, offering some evidence of the concurrent validity of the measure [21].

We also applied the DSM-IV-TR criteria for childhood separation anxiety disorder as reported retrospectively, in order to assess its occurrence prior to the age of 18 years.

Statistical analyses

Three sets of analyses were undertaken for the whole sample and then by gender. Model 1 compared the conventional SCID-derived adult anxiety subcategories (ie PD, PD-AG, GAD and SP). In Model 2, those meeting criteria were assigned to the ASAD category, with all residual patients being grouped into a single category for comparison (ie ASADs and non-ASADs). Model 3 compared ASADs with all residual patients remaining in their initial diagnostic groups (ie PD, PD-AG, GAD, SP and ASAD).

Initial analyses indicated some variation in the number of comorbid anxiety and/or depressive disorders across primary anxiety categories (mean number of comorbid disorders associated with ASAD = 1.3, compared to 0.9 for PD, 1.0 for PD-AG, 0.9 for GAD and 0.9 for social phobia; p < .01 for all comparisons against ASAD). Since comorbidity generally is associated with severity of disorder [22], that factor could confound any comparisons we made, for example in contrasting ASADs with other anxiety categories in relation to indices of symptom severity and functional impairment. To address that issue, we entered the number of disorders (anxiety or depressive) per patient as a covariate in analyses involving continuous measures of the SASI, DASS, WSAS and NEO PI-R.

SPSS version 15 was used for all analyses [23]. Univariate analysis of variance was applied for continuous data with post hoc contrast testing. Categorical data were analysed using chi square tests. Significance levels were set at p < .01.

Model 1

The primary anxiety subcategories identified by the SCID were: PD (n = 121, 23% of total sample), PD-AG (n = 162, 31%), GAD (n = 135, 26%) and SP (n = 102, 20%). The mean age across all groups was 36 (SD = 12) years, with the SP group being younger (p < .01) compared to all other groups (see Additional file 2).

With the exception of those with SP, females predominated in all groups. Just under half the sample were married or in a cohabiting relationship (n = 244, 47%) except for the SP group, where only 22% (n = 22) were married, differing significantly from all other groups. Just over half the sample (n = 265, 51%) were employed, with PD-AGs being over-represented in the unemployed group (n = 72, 44%, p < .01) compared to those with PD and SP. Most were born in Australia and spoke English at home (all tests NS across groups for both indices). Additional file 3 shows that anxiety subcategories were similar in their reports of both indices of early separation anxiety.

Additional file 4 displays results for the DASS and WSAS, while Additional file 5 shows results for the NEO PI-R. Scores on these measures were not influenced by age. The anxiety subgroups returned similar scores on the DASS depression and stress scales. The PD and PD-AG groups scored higher on DASS anxiety compared to the GAD and SP categories. WSAS disability scores were higher for SPs and PD-AGs, primarily in the domains of work and social activities.

The SP group scored higher on the NEO PI-R subscale for neuroticism (see Additional file 5). SPs and PD-AGs scored lower than other groups on the extraversion and conscientiousness subscales.

Model 2

The sample was then divided according to whether or not patients met criteria for ASAD. With the inclusion of that category, the total number of anxiety/depressive diagnoses assigned (primary and comorbid) was 921 or a mean of 1.8 per person. The numbers and percentages for each diagnosis were: ASAD = 207 (23%), PD = 108 (12%), PD-AG = 100 (11%), GAD = 195 (21%), SP = 132 (14%), and current depression = 133 (14%). Hence, the prevalence of ASAD assignments was roughly similar to that of GAD or the combined categories of PD/PD-AG. The proportion of primary diagnoses initially made on the SCID that were later assigned to the ASAD grouping, once formed, were: PD: n = 42 (35% of the initial PD group were re-assigned to the ASAD grouping); PD-AG: n = 80 (49%); GAD: n = 52 (39%); and SP: n = 33 (32%). A statistically greater number of those with an initial diagnosis of PD-AG was included in the ASAD grouping (p < .01). Females were overrepresented in the ASAD grouping.

The ASAD group had higher scores on all DASS subscales, on all disability scales of the WSAS (Additional file 4), and on NEO PI-R scores for neuroticism (Additional file 5). For depression analyses, we compared the ASAD group with a residual group where patients had 2 or more anxiety diagnoses, in order to broadly match the two groupings for levels of comorbidity in relation to other anxiety categories. No differences in rates of co-occurring current depression emerged from this comparison.

ASADs scored substantially higher on both indices of early separation anxiety (Additional file 3). We then divided the ASAD sample into those with probable childhood onset (SASI scores ≥ 4) and those with probable adult onset (<4). Three quarters of the ASAD sample (n = 151) had probable early onset and 52 were probable adult onset.

Model 3

We then compared the ASAD grouping with specific subcategories of anxiety in the residual group. The ASAD and SP groups were younger, but statistically so only in relation to the GAD group. SP remained the only group with a minority of females (n = 31, 45% female), significantly so except in comparison with PD. As in Model 1, the SP group had more single and fewer married people compared to all other groups.

ASADs had higher DASS depression scores compared to all other groups, and higher stress scores than all groups but GADs. ASADs, PDs and PD-AGs reported higher levels of anxiety on the DASS than GADs and SPs. ASADs returned higher scores on both indices of early separation anxiety compared to all other groups (see Additional file 3). ASADs were more disabled on WSAS scales in relation to all other groups except on selective indices in relation to SPs.

ASADs and SPs scored higher on the NEO PI-R neuroticism scale than the other SCID anxiety categories. As in Model 1, the SPs scored lower on extraversion, while ASADs had the second lowest scores, although were significantly different only from the PD group.

Analysis by gender

As indicated (Additional files 3, 4 and 5), the key analyses were also undertaken separately for males and females. Results largely replicated those for the total sample, although the differences between females with ASAD compared to their non-ASAD counterparts were more extensive than for the comparable analyses for males. We note however the smaller number of males (total sample: n = 359 females, n = 161 males), a factor that may have restricted statistical power for comparisons involving that gender.

One key finding that emerged from the gender-based analysis was that in both Model 1 and 3, SASI scores for the male-only social phobia group were significantly higher than the PD and PD-AG male groups (Additional file 3). In contrast, in the total sample and female-only analyses, the ASAD group alone scored significantly higher on the SASI. Minor differences also emerged on the NEO PI-R. Male ASADs returned statistically lower scores on extraversion and conscientiousness in Model 2 and were lower on agreeableness compared to the PD-AG group in Model 3. It should be noted, however, that all the relevant scores fell within the low to average range according to normative data [19].