WHY did the authors of your article – the researchers – decide to do this research? 

ORIGINAL ARTICLE

Social phobia, depression and eating disorders during middle adolescence: longitudinal associations and treatment seeking

Klaus Rantaa, Juha V€a€an€anenb, Sari Fr€ojdc, Rasmus Isomaad, Riittakerttu Kaltiala-Heinob,e and Mauri Marttunena,f,g

aDepartment of Adolescent Psychiatry, Helsinki University Central Hospital, Helsinki, Finland; bDepartment of Adolescent Psychiatry, Tampere University Hospital, Tampere, Finland; cSchool of Health Sciences, University of Tampere, Tampere, Finland; dCity of Jakobstad, Department of Social Services and Health Care, Jakobstad, Finland; eMedical School, University of Tampere, Tampere, Finland; fDepartment of Adolescent Psychiatry, University of Helsinki, Helsinki, Finland; gDepartment of Mental Health and Substance Use Services, National Institute for Health and Welfare, Helsinki, Finland

ABSTRACT Background: Longitudinal associations between social phobia (SP), depression and eating disorders (EDs), and the impact of antecedent SP and depression on subsequent treatment seeking for EDs have rarely been explored in prospective adolescent population studies. Aim: We aimed to examine these associations in a large-scale follow-up study among middle adolescents. Method: We surveyed 3278 Finnish adolescents with a mean age of 15 years for these disorders. Two years later, 2070 were reached and again surveyed for psychopathology and treatment seeking. Longitudinal associations between the self-reported disorders and treatment-seeking patterns for self- acknowledged ED symptoms were examined in multivariate analyses, controlling for SP/depression comorbidity and relevant socioeconomic covariates. Results: Self-reported anorexia nervosa (AN) at age 15 years predicted self-reported depression at age 17 years. Furthermore, self-reported SP at age 15 years predicted not seeking treatment for bulimia nervosa (BN) symptoms, while self-reported depression at age 15 years predicted not seeking treat- ment for AN symptoms during the follow-up period. Conclusions: Adolescents with AN should be monitored for subsequent depression. Barriers caused by SP to help seeking for BN, and by depression for AN, should be acknowledged by healthcare professio- nals who encounter socially anxious and depressive adolescents, especially when they present with eating problems.

ARTICLE HISTORY Received 13 September 2016 Revised 1 June 2017 Accepted 7 August 2017

KEYWORDS Social phobia; depression; eating disorders; adolescents; comorbidity; treatment seeking

The mean onset of eating disorders (EDs) in the population is during adolescence. The peak of incidence for both bulimia nervosa (BN) and anorexia nervosa (AN) is between 15 and 19 years [1–3]. Both population [1,2] and clinical studies [4,5] indicate that BN is associated with a range of anxiety, mood, disruptive and substance use disorders in both adolescents and adults. Of the anxiety disorders, social phobia (SP) may be especially comorbid with adolescent BN [6]. Adolescent AN, as identified in the community, seems less strongly asso- ciated with other disorders [2], but clinical studies indicate that mood disorders are frequently involved [7]. In young adult populations, AN seems broadly comorbid with anxiety, mood and alcohol use disorders [1,8].

Longitudinal studies suggest that the onset of comorbid anxiety disorders generally occurs prior to that of EDs. In clin- ical studies, this order is found in �60–90% of cases [4,6,8,9], offering support for the notion of an anxiety pathway in the genesis of EDs. The association between SP and EDs in par- ticular has been found to be largely of this type [6].

With regard to depression, there is evidence of bidirec- tional connections, depression elevating the risk for EDs [10],

as well as of a heightened risk for subsequent depression in adolescents with AN and BN [11,12]. Studies with long fol- low-up spans indicate that depression may both precede and follow AN [13]. Thus, although not entirely consistent, research does indicate somewhat different types of longitu- dinal associations between SP, depression and EDs.

The mechanisms by which SP, depression and EDs are lon- gitudinally linked are not well-known [14]. Social anxiety may predispose an adolescent to subsequent EDs by participants’ preoccupation with the fear of being negatively evaluated, high performance standards, dependence on others’ opinions or negative beliefs about the public self [15,16]. In girls, a preoccupation with appearance [15] while pubertal physical changes progress may lead to ‘social appearance anxiety’ and symptoms of EDs, such as the strict pursuit of ideal body weight [17].

Regarding associations between depression and EDs, such factors as a self-deprecating thinking style, linked with both [18], may mediate bidirectional links. Moreover, biological and psychological consequences of malnutrition (e.g. apathy and numbness) could lead to the development of depression

CONTACT Klaus Ranta klaus.ranta@hus.fi Department of Adolescent Psychiatry, Helsinki University Hospital, P.O. Box 590, FI-00029 HUS, Helsinki, Finland � 2017 The Nordic Psychiatric Association

NORDIC JOURNAL OF PSYCHIATRY, 2017 VOL. 71, NO. 8, 605–613 https://doi.org/10.1080/08039488.2017.1366548

following AN [19]. Antecedent disorders may also affect treat- ment seeking for EDs. It remains unknown whether comor- bidity in fact increases or decreases treatment seeking or engagement in treatment [20,21]. Treatment seeking may vary according to the type of comorbidity.

SP could act as a barrier to seeking treatment for EDs because only very few, �9% of adolescents with SP, seek help for their anxiety [22]; furthermore, SP is typically associated with feelings of shame that may inhibit treatment- seeking behavior [21]. Indeed, some evidence suggests that high social anxiety prevents participants from engaging in treatment for EDs [23], but no studies addressing this issue in general population samples or among adolescents have been conducted [21,24].

Associated with depression, decreased activity or apathy [25], pessimism about the possibility of being helped [21,26] or negative self-evaluations and internalized feelings of shame [21] may prevent treatment seeking. Indeed, the find- ings of one large-scale adolescent population study indicated that the higher the level of depression symptoms, the lower the possibility that the adolescent will seek help [27]. The available research on treatment seeking for EDs suggests that emotional suffering or disorders may even facilitate treatment seeking. However, all studies do not report such an effect [21]. As this research is conducted mainly among adults with EDs using retrospective analyses, they likely miss a significant part of the disorders in the population. Overall, a mixed and contradictory picture emerges from this field. Studies examining treatment seeking for EDs in adolescents, in prospective designs and large population samples, are lacking.

In sum, gaps exist in knowledge on independent pro- spective associations between specific EDs (AN/BN) with SP and depression. Moreover, little is known about the possible impact of antecedent SP and depression on treatment seek- ing for EDs in adolescence. There is a need for longitudinal, controlled studies.

Aims

To address the knowledge gaps in prior research, we aimed to answer two research questions. First: Are there independ- ent, longitudinal associations between self-reported SP and depression at age 15 years and BN/AN at age 17 years, or conversely, do self-reported BN and AN symptoms at age 15 years predict self-reported SP and depression at age 17 years? Second: Are self-reported SP or depression at age 15 years associated with a reduced rate of seeking help for self-acknowledged symptoms of BN/AN at a two-year follow-up?

Three hypotheses based on the literature search were tested. 1. SP at age 15 years will increase the risk of BN and AN at age 17 years. 2. Depression at age 15 years will increase the risk of BN and AN at age 17 years, but BN and AN at age 15 years will also increase the risk for depression at age 17 years. 3. SP at age 15 years will be associated with a heightened risk of not seeking help for BN and AN during the follow-up period. We made no specific hypothesis

concerning the role of depression as affecting help-seeking behavior because the research is inconclusive and contradictory.

Material and methods

Design, procedure and sampling

The data came from the Finnish Adolescent Mental Health Cohort Study [28]. At baseline (T1), ninth-grade students aged 15–16 years from all secondary schools in the Finnish cities of Tampere and Vantaa completed a person-identifiable questionnaire containing measures on mental health (i.e. depression, anxiety, conduct disorder, EDs and substance use), health behavior and associated psychosocial variables in their schools. The survey assessment tool took one lesson (i.e. 30–45min) to complete.

The T1 sample consisted of 3278 adolescents (1609 girls, 1669 boys) with a mean age of 15.5 (SD ¼0.39) years. The response rate for the baseline survey was 94.4%. Participants were contacted after a 2-year follow-up period (T2) by mul- tiple methods: surveys in high/vocational schools, surface mail and finally by email. In all, 2082 responses were received. Ten responses were excluded due to obvious facetiousness and due to double responding. The study was approved by the Ethics Committee of Tampere University Hospital (Study Approval No. 9924).

The 2070 adolescents who responded both times (repre- senting 63.1% of those who responded at T1) comprise the participants of the present study. At T2, their mean age was 17.6 (SD¼ 0.41) years, and 56.4% (1167) were girls. Of the participants studying at T2, 62% were in high school and 38% in vocational school. The sample represents the Finnish general adolescent population, as in Finland more than 90% of children and adolescents attend Grades 1–9 in public schools.

Instruments

Self-reported SP: SP was assessed with a 17-item self-report instrument, the Social Phobia Inventory (SPIN) [29], which covers all DSM-IV symptom clusters for SP. The reliability and validity of this instrument to detect symptoms of SP in the adolescent age group have been demonstrated in several studies across many countries, such as USA, Canada, Germany, Finland, Spain and Brazil [30]. Using a cutoff score of 24 points, a sensitivity of 81% and a specificity of 85% relative to a diagnosis of SP has been demonstrated for the SPIN in a large Finnish adolescent population sample com- parable to the present sample [31]. Scores of 0–23 represent no self-reported SP, and scores �24 represent self-reported SP in the present study.

Self-reported depression: The 13-item Beck Depression Inventory (short version) (BDI) [32] was used as the measure of depression. The BDI has been examined in several adoles- cent samples [33]. A recent meta-analysis found 17 psycho- metric studies in this age range. A pooled estimate of internal consistency was 0.86, and the pooled estimates for both sensitivity and specificity for detecting depression

606 K. RANTA ET AL.

against a clinical interview were 0.81 [34]. Bennett et al. [35] have also reported acceptable sensitivity and specificity indi- ces for the 13-item version in adolescents. The Finnish ver- sion [36] has been found to possess good internal consistency in this age group [37]. A cutoff score of �8 points suggested by Beck et al. [38] was used to identify ado- lescents with self-reported depression, and scores of 0–7 rep- resented no self-reported depression. This cutoff has resulted in a sensitivity of 0.93 and specificity of 0.88 in detecting depression, as established through a structured psychiatric interview [39].

Self-reported EDs: We assessed self-reported BN and AN with questionnaires formulated according to the DSM-IV-TR [40] criteria, which have been extensively used in prior Finnish ED research. BN was recorded if the respondent: (i) reported having had frequent binge-eating episodes; (ii) while binge-eating had a feeling that his/her eating was out of control; (iii) reported binge-eating episodes at least twice per week over a 3-month period; (iv) reported compensatory behaviors to prevent weight gain (strict diet or fasting, heavy exercise, vomiting, misuse of laxatives or diuretics); and (v) as a consequence of binge-eating, felt bad or unsuccessful as a person unless one was thin or fit.

AN was recorded if the respondent reported: (i) an inten- tional dieting episode to lose weight; (ii) during that episode a significant fear of weight gain; (iii) had felt fat despite the obviously low weight noted and commented on by family and peers; and (iv) during the episode reported amenorrhea. Furthermore, participants’ height and weight at T1 were available, from which the body mass index (BMI) at T1 was calculated.

To exclude participants who were at T1 normal weight or recovered from an earlier AN episode, we used the standard definition for normal weight (BMI of �20) proposed by British Royal College of Psychiatrists [41]. This conservative approach for setting the cutoff is consistent with findings on the weight at the return of menstruation in adolescent girls who have recovered from AN [42]. Thus, T1 AN was recorded when the participant fulfilled all DSM-IV-TR criteria 1–4 and had a BMI of <20. The DSM-IV-derived AN items have been found to possess acceptable properties to detect a lifetime episode of AN in the Finnish adolescent population [43]. At T2, no records of height and thus BMI were available; thus, AN was recorded for those fulfilling criteria 1–4.

Treatment seeking: We assessed treatment seeking for self- acknowledged BN/AN during the follow-up period at T2 with two separate questions: ‘Have you suffered from bulimia nerv- osa (AN) during the past 2 years without seeking help for the symptoms?’. The participants were asked to respond with either ‘yes’ or ‘no’ to each question. All participants who answered the BN/AN questions were included in the respective analyses, regardless of whether they met criteria for AN or BN at T1 or T2. See the ‘Statistical analysis’ section below as well.

Covariate measures: The residential stability of the adoles- cents was assessed with an item from the Life Events Checklist (LEC) [44]. At T2, adolescents reported whether they had moved during the past year with response alternatives: no (¼0), yes (¼1). Parental unemployment was assessed at T2 with an LEC item eliciting whether their parents had been

unemployed during the past year, with response alternatives: neither (¼0), or one/both (¼1) had been unemployed. These two covariates were included because of their disruptive effect on the socioeconomic conditions of adolescents and their associations with internalizing psychopathology, EDs and treatment seeking in adolescence [26,45,46].

Statistical analysis

Associations between baseline and drop-out key variables were studied using Chi-square statistics or Fisher’s exact test when appropriate. The percentages and numbers of cases for self-reported SP, depression, BN, and AN at T1 and T2 were calculated, and for those who reported not having sought treatment for BN and AN during the past 2 years at T2. The incidence (the proportion of those with a given disorder at T2 who did not have the same disorder at T1 out of all those without the condition at T1) of SP, depression, BN and AN was calculated as the percentages and the number of cases. The change in frequency for SP, depression, BN and AN in repeated measures between T1 and T2 was tested using McNemar’s test.

Longitudinal associations between the self-reported disor- ders: These were examined in logistic regression analyses (LRAs). The simple cross-tabulated frequencies, as well as unadjusted odds ratios (ORs) and adjusted odds ratios (AORs) with 95% confidence intervals (CIs) are reported for bivariate and controlled, respectively. First, the unadjusted bivariate associations between T1 SP/depression and T2 BN/AN were calculated, then associations between T1 BN/AN and T2 SP/ depression. The multivariate LRAs were constructed to con- trol for (i) the SP/depression comorbidity (at either T1 or T2,

Table 1. Cross-tabulated frequencies and bivariate associations between self- reported SP, depression and EDs at ages 15 and 17 years.

% (n/N) OR (95% CI) p

Predicting BN at age 17 years Age 15 SP -> age 17 BN

Yes 5.3 (10/188) 3.3 (1.6–6.9) .001 No 1.7 (31/1877)

Age 15 depression -> age 17 BN Yes 6.9 (13/188) 4.9 (2.5–9.6) <.001 No 1.7 (31/1877)

Predicting AN at age 17 years Age 15 SP -> age 17 AN

Yes 1.1 (2/185) 1.2 (0.3–5.1) .827 No 0.9 (17/1850)

Age 15 depression -> age 17 AN Yes 2.7 (5/187) 3.6 (1.3–10.1) .015 No 0.8 (14/1841)

Predicting SP at age 17 years Age 15 BN -> age 17 SP

Yes 20.0 (5/25) 2.0 (0.8–5.5) .161 No 11.0 (223/2032)

Age 15 AN -> age 17 SP Yes 15.8 (3/19) 1.5 (0.4–5.2) .517 No 11.0 (223/2032)

Predicting depression at age 17 years Age 15 BN -> age 17 depression

Yes 32.0 (8/25) 5.2 (2.2–12.2) <.001 No 8.3 (169/2035)

Age 15 AN -> age 17 depression Yes 31.6 (6/19) 5.1 (1.9–13.6) .001 No 8.3 (161/1936)

Note: OR (95% CI) unadjusted OR with 95% CI. Group with no disorder (No) is treated as the reference group in each LR analysis.

NORDIC JOURNAL OF PSYCHIATRY 607

depending on the respective analysis), (ii) the effect of T1 EDs in the two LRAs where the outcome variables were T2 EDs, and (iii) for the effects of confounding socioeconomic covariates. In these LRAs, T2 BN, T2 AN, T2 SP and T2 depres- sion were each treated as the dependent outcome variables. See the footnotes of Tables 2 and 3 for the description of how each variable was entered in each sequential step of the LRAs (Model 1, Model 2 and Model 3).

Treatment seeking for self-reported BN and AN: Next, we examined predictors of not seeking treatment for self- acknowledged symptoms of BN/AN during the follow-up period. These were calculated as follows. (i) Bivariate LRAs were performed with T1 SP and T1 depression each in turn as predictors. (ii) Multivariate LRAs were performed by entering T1 SP and T1 depression in the first step (Model 1), T1 BN/AN in the second step (Model 2) and T2 family relocation and par- ental unemployment in the third step (Model 3). All analyses were performed using SPSS 19.0. (SPSS Inc., Chicago, IL).

Attrition

Of the participants, 37% who had responded to the baseline survey dropped out by follow-up. Attrition was influenced by participants’ gender (girls 28% versus boys 46%, p< .001), the family structure (intact 27% versus divorced parents 33%, p< .001) and parents’ educational level (low 23% versus high 29%, p< .001). Of the main target variables, T1 SP was not more frequent among the dropouts than among those who completed the study (8 versus 9%, p¼ .51), neither was T1 AN (1% in both groups, non-significant) nor T1 BN (1.1 versus 1.6%, non-significant). However, T1 depression was more fre- quent among dropouts (12 versus 9%, p¼ .020).

Results

Frequency and incidence of self-reported SP, depression and EDs

Of the participants, 9.1% (187) at T1 and 11.1% (221) at T2 had SP (change in frequency, p¼ .014), the incidence of SP

being 7.9% (147/1865). In total, 9.2% (188) had depression at T1, while 8.6% (176) had depression at T2 (change in fre- quency, non-significant), and the incidence of depression was 5.3% (99/1861). With regard to BN, 1.2% (25) adolescents at T1 and 2.0% (41) adolescents at T2 reported BN (change in frequency, p¼ .033); the incidence for BN was 1.6% (33/ 2045). Finally, 1.0% (19) adolescents at T1 and 0.9% (19) ado- lescents at T2 reported AN (change in frequency, non-signifi- cant), the incidence for AN being 0.7% (14/1919).

Longitudinal associations between self-reported EDs, depression and SP

Predictors of BN at age 17 years: In the bivariate analyses, SP at age 15 years alone and depression at age 15 years alone each predicted BN at age 17 years (Table 1). In the final multivariate LRA model, no independent predictive role for either SP or depression at age 15 years for BN at age 17 years was found; however, BN at age 15 years was associated with an elevated risk for this outcome (p< .001), as was recent relocation (p< .05) (Table 2).

Predictors of AN at age 17 years: In bivariate analyses, depression at age 15 years predicted AN at age 17 years (Table 1). However, in the multivariate LRA, only AN at age 15 years was associated with AN at age 17 years in the final model (p< .001).

Predictors of SP at age 17 years: In the bivariate analyses, neither BN at age 15 years nor AN at age 15 years was associated alone with SP at age 17 years (Table 1). Predicting SP at age 17 years by BN at age 15 years in the multivariate LRA showed a non-significant result. Only SP at age 15 years and depression at age 17 years were associ- ated with SP at age 17 years (p< .001). The multivariate LRA regarding AN at age 15 years as a predictor of SP at age 17 years resulted similarly in non-significant findings: the final model only showed persistent associations of SP at age 15 years and depression at age 17 years (p< .001) with SP at age 17 years (Table 3).

Predictors of depression at age 17 years. In the bivariate LRAs, both BN at age 15 years and AN at age 15 years

Table 2. Risk (AOR with 95% CIs) for age 17 years self-reported BN and AN by age 15 years self-reported SP and depression in multivariate LRAs.

Model 1, AOR (95% CI) Model 2, AOR (95% CI) Model 3, AOR (95% CI)

Outcome variable: age 17 years BN

Predictor variables T1 SP 1.8 (0.8–8.3) 1.5 (0.6–3.9) 1.5 (0.6–4.0) T1 Depression 3.8 (1.8–8.3)�� 2.4 (1.0–5.8) 1.9 (0.7–4.9) T1 BN # 15.4 (5.5–43.3)��� 16.5 (5.8–46.8)��� T2 family moved within last 12 months # # 2.6 (1.0–6.3)� T2 parental unemployment during last 12 months # # 1.0 (0.3–3.6)

Outcome variable: age 17 years AN

T1 SP 0.6 (0.1–2.9) 0.6 (0.1–3.4) 0.5 (0.1–3.1) T1 depression 4.2 (1.4–12.9)� 2.7 (0.8–9.3) 3.1 (0.9–11.0) T1 AN # 38.3 (11.5–127.9)��� 41.6 (12.2–141.2)��� T2 family moved within last 12 months # # 0.0 (0.0–)a

T2 parental unemployment during last 12 months # # 0.9 (0.1–7.5)

In the first models, SP and depression at age 15 years were entered. In the second models, baseline homotypic ED (i.e. BN/AN, depending on the analysis) was controlled for. In the third models, family relocation and parental unemployment during last year were additionally controlled for. T1: first assessment point at age 15 years. T2: second assessment point at age 17 years. �p< .05, ��p< .01, ���p< .001. #Not entered in the model. aUpper limit of CI not estimable. None of those whose family had moved had AN at age 17. The stability of Model 3 was additionally tested by dropping T2 family relocation from it. This modification did not change the significance of any one association found in Model 3.

608 K. RANTA ET AL.

predicted depression at age 17 years (Table 1). Predicting depression at age 17 years by BN at age 15 years in the multivariate LRA showed that depression at age 15 years, SP at age 17 years (p< .001) and parental unemployment during the last year (p< .01) were each associated with depression at age 17 years. Finally, predicting depression at age 17 years by AN at age 15 years in the multivariate LRA resulted in AN at age 15 years retaining an independent association with depression at age 17 years [AOR 3.6 (95% CI 1.1–11.6); p< .05], along with depression at age 15 years, SP at age 17 years (p< .001) and parental unemployment during the past year (p< .05).

Treatment seeking for self-reported symptoms of BN and an

Predictors for not seeking treatment for BN during the follow- up. In the bivariate LRAs, both SP and depression at age 15 years predicted a participant having not sought help for BN during the past 2 years (Table 4). In the multivariate LRA, the final model showed that SP at age 15 years independ- ently predicted a subject not having sought treatment for BN during the follow-up period [AOR 2.4 (95% CI 1.0–5.6), p< .05], as did relocation during the previous year (p< .01) (Table 5).

Predictors for not seeking treatment for AN during the fol- low-up. In bivariate LRAs, SP at age 15 years did not, but depression at age 15 years did predict a subject not having sought help for AN during the follow-up period (Table 4).

In the multivariate LRA, the final model revealed that depression at age 15 years [AOR 4.8 (95% CI 1.9–11.9), p< .001] and AN at age 15 years (p< .001) predicted a sub- ject not having sought help for AN during the follow-up period (Table 5).

Discussion

The first aim of this study was to examine longitudinal asso- ciations between SP, depression and EDs in adolescents fol- lowed-up from 15 to 17 years of age. The second aim was to examine whether SP and depression at age 15 years would be associated with a heightened risk of not seeking help for self-reported BN and AN symptoms during the follow-up period. We found an independent predictive association only for AN at age 15 years as a predictor of depression at age 17 years. Second, SP and depression at age 15 years had dif- ferent effects on adolescents’ treatment seeking for ED symp- toms during the follow-up period. SP was associated with not seeking help for symptoms of BN, and depression was associated with not seeking help for symptoms of AN. Our findings extend the existing knowledge of the longitudinal associations between emotional disorders and EDs as well as patterns of treatment seeking for ED symptoms by introduc- ing results from a large study in a general adolescent popula- tion sample.

Temporal order of SP, depression and EDs. Results from the uncontrolled bivariate analyses hinted at bidirectional longi- tudinal associations between depression and both AN and

Table 3. Risk (AOR with 95% CIs) for age 17 years self-reported SP and depression by age 15 years self-reported BN and AN in multivariate LRAs.

Model 1, AOR (95% CI) Model 2, AOR (95% CI) Model 3, AOR (95% CI)

Outcome variable: age 17 years SP

Predictor variables/by BN T1 BN 2.0 (0.8–5.5) 0.4 (0.1–1.4) 0.4 (0.1–1.5) T1 SP # 7.8 (5.3–11.4)

��� 7.4 (5.0–11.0)���

T2 depression # 14.5 (10.0–21.1)��� 14.7 (10.1–21.4)��� T2 family moved within last 12 months # # 1.2 (0.7–2.2) T2 parental unemployment during last 12 months # # 0.8 (0.4–1.5)

Outcome variable: age 17 years SP

Predictor variables/by AN T1 AN 1.5 (0.4-5.2) 0.4 (0.1–2.2) 0.5 (0.1–2.2) T1 SP # 8.2 (5.5–12.1)��� 7.9 (5.3–11.7)��� T2 depression # 14.7 (10.0–21.5)��� 14.8 (10.0–21.8)��� T2 family moved within last 12 months # # 1.2 (0.6–2.2) T2 parental unemployment during last 12 months # # 0.8 (0.4–1.6)

Outcome variable: age 17 years depression

Predictor variables/by BN T1 BN 5.2 (2.2–12.2)��� 1.7 (0.6–5.2) 1.7 (0.6–5.1) T1 depression # 7.7 (5.1–11.8)��� 7.6 (5.0–11.6)��� T2 SP # 12.0 (8.3–17.5)��� 12.4 (8.5–18.1)��� T2 family moved within last 12 months # # 1.6 (0.9–3.0) T2 parental unemployment during last 12 months # # 2.3 (1.2–4.1)��

Outcome variable: age 17 years depression

Predictor variables/by AN T1 AN 5.1 (1.9–13.6)�� 3.6 (1.1–11.5)� 3.6 (1.1–11.6)� T1 Depression # 8.0 (5.3–12.2)��� 7.8 (5.1–12.0)��� T2 SP # 11.8 (8.0–17.4)��� 12.2 (8.2–17.9)��� T2 family moved within last 12 months # # 1.7 (0.9–3.1) T2 parental unemployment during last 12 months # # 2.1 (1.1–3.9)�

In the first models, BN/AN at age 15 years was entered. In the second models, T2 SP/depression comorbidity was controlled for. In the third models, family relocation and parental unemployment during last year were additionally controlled for. T1: first assessment point at age 15 years. T2: second assessment point at age 17 years. �p< .05, ��p< .01, ���p< .001. #Not entered in the model.

NORDIC JOURNAL OF PSYCHIATRY 609

BN. The more robust, controlled multivariate analyses yielded a different picture. In these analyses, only the predictive asso- ciation of AN at age 15 years with depression at age 17 years persisted. This was contrary to our study hypothesis, but con- sistent with some earlier findings in adolescents [11] and the malnutrition account for depression following AN [19].

In their critical review, Mattar et al. [19] concluded that the malnutrition—depression link has primarily been sup- ported by evidence from few, relatively small clinical studies. These studies mainly show that depressive symptoms tend to diminish among individuals with AN when their weight is restored. Our finding of an independent pathway from ante- cedent AN to subsequent depression across middle to late adolescence replicate the findings from the study by Ivarsson et al. [11]; however, they examined a clinical sample with matched community controls, whereas we performed a large population survey study.

Future studies among both community and clinical sam- ples using rigorous methodology are clearly needed to con- firm the observed, longitudinal AN-to-depression association. With regard to identifying possible causal mechanisms explaining the emergence of depressive symptomatology fol- lowing AN, our study unfortunately does not provide answers. Possible explanatory factors that have been pre- sented to account for this sequence include, among others, malnutrition-induced dysfunction of the serotoninergic sys- tem [47], or pronounced social and emotional withdrawal in individuals with AN [48], which both may act as depresso- genic correlates of AN.

Thus, despite their rarely acknowledged need for help [24], adolescents with AN nevertheless go on to suffer from emerging depressive symptoms. This finding extends results from clinical studies [11,19] and may even point to AN’s causal role in the development of subsequent depression. Whether these symptoms disappear with the recovery from AN [13] remains unresolved.

The predictive association between SP at T1 and BN at T2 found in the bivariate analysis did not hold after controlling for depression in the LRA. This finding was contrary to our hypothesis. However, it suggests that the co-presence of SP and depressive symptoms may play a role in the develop- ment of BN. Indeed, some research on the core beliefs of individuals with BN and social anxiety indicate they are both depressive and anxious in content [49]. Our results call for replication with rigorous, interview-based methods in popula- tion samples. The potential role of social anxiety/SP as a risk or even a causal factor in the development of BN suggested by several earlier studies [6,14,15], and, on the other hand, the corresponding role of depression in increasing the risk for ED development—also observed in prior research [10,18]—may indeed hold only in the context of more gen- eral emotional distress.

The data were gathered within a period during which, on average, most cases of SP have had their onset (i.e. 75% of cases have had their onset by age 15 years), while the risk for incident depression and for both AN and BN remains quite high [3,50]. Thus, SP’s predictive role may be more detectable in the present prospective study than in studies

Table 4. Frequency and risk of not having sought treatment for self-reported BN or AN during the 2- year follow-up period by age 15 years SP and depression.

Not sought treatment for BN during the follow-up

Not sought treatment for AN during the follow-up

% (n/N) OR (95% CI) % (n/N) OR (95% CI)

Age 15 years SP Yes 5.3 (10/188) 3.5 (1.7–7.2)�� 2.1 (4/188) 1.4 (0.5–4.1) No 1.6 (30/1877) 1.5 (28/1877)

Age 15 years depression Yes 5.3 (10/188) 3.4 (1.7–7.2)�� 4.8 (9/188) 4.0 (1.8–8.9)��� No 1.6 (30/1871) 1.2 (23/1871)

OR (95% CI): unadjusted OR with 95% CI. T1: first assessment point at age 15 years. T2: second assessment point at age 17 years. ��p< .01, ���p< .001. Group with no disorder (no) is treated as the reference group in each LR analysis.

Table 5. Risk (AOR with 95% CIs) for not having sought treatment for self-reported BN and AN during the 2-year follow-up period in multivariate analyses.

Model 1, AOR (95% CI) Model 2, AOR (95% CI) Model 3, AOR (95% CI)

Outcome variable: not sought treatment for BN symptoms during the follow-up

Predictor variables T1 SP 2.4 (1.0–5.5)� 2.3 (1.0–5.4) 2.4 (1.0–5.6)� T1 depression 2.4 (1.0–5.5)� 2.0 (0.8–4.9) 1.8 (0.7–4.4) T1 BN # 3.6 (0.9–14.0) 3.5 (0.9–13.7) T2 family moved within last 12 months # # 3.6 (1.7–8.0)�� T2 parental unemployment during last 12 months # # 1.7 (0.6–4.5)

Outcome variable: not sought treatment for AN symptoms during the follow-up

T1 SP 0.7 (0.2–2.2) 0.5 (0.1–1.9) 0.4 (0.1–1.7) T1 depression 4.5 (1.9–10.8)�� 4.5 (1.8–11.2)�� 4.8 (1.9–11.9)�� T1 AN # 13.4 (3.9–46.1)��� 13.9 (4.0–48.5)��� T2 family moved within last 12 months # # 0.3 (0.0–2.6) T2 parental unemployment during last 12 months # # 1.8 (0.5–6.4)

In the first models, SP and depression at age 15 years were entered. In the second models, the baseline homotypic ED (i.e. BN/AN, depending on the analysis) was controlled for. In the third models, family relocation and parental unemployment during last year were additionally controlled for. T1: first assessment point at age 15 years. T2: second assessment point at age 17 years. �p< .05, ��p< .01, ���p< .001. #Not entered in the model.

610 K. RANTA ET AL.

among samples with higher mean ages, in which a recall bias concerning the timing of the onset of symptoms may be found due their retrospective design. The community con- text of the study may be beneficial in revealing the associa- tions between SP, depression and EDs, avoiding the bias in clinical studies, which may involve only the most severe and comorbid cases and altogether miss individuals who do not seek treatment.

Treatment seeking for self-reported EDs. While preliminary in nature, the results of this follow-up study suggest there may be specificity in the role that antecedent emotional dis- orders have in subsequent treatment seeking for the two main types of EDs: AN and BN. The results were only partially consistent with our hypotheses on treatment seeking. Bivariate analyses suggested that T1 depression may hinder treatment seeking for both later BN and AN symptoms and T1 SP for later BN symptoms. However, the multivariate LRAs showed that the associations were specific to the type of ED.

The role of SP in blocking the help-seeking pathway for BN is consistent with previous findings on the role of SP as preventing engagement in treatment for EDs [23]. In the pre- sent study, we observed that SP prevented treatment seeking specifically for BN symptoms over a two-year follow-up period. Goodwin and Fitzgibbon [23] observed in their small pilot study that symptoms of social anxiety predicted lower engagement in treatment among individuals who had already registered at an ED outpatient clinic due to their AN or BN symptoms. Our results extend these findings and sug- gest that social anxiety may altogether prevent adolescents from seeking help in the first place, especially for their BN symptoms. The fears of negative evaluation, social avoidance and possibly even of authority figures, all typical for adoles- cents with SP [15,16,51], may contribute to this finding. It is noteworthy that feelings of shame, pronounced in both SP and BN, have been reported as common reasons for not seeking treatment among individuals with social anxiety and both main types of EDs [52,53].

Our finding that antecedent depression prevented treat- ment seeking for incident AN symptoms may be related to prior findings that among participants with EDs, beliefs that others cannot help are frequent reasons for not seeking help [21,26]. Such beliefs likely represent a depressive thinking style. Furthermore, depression is typically characterized by participants’ negative, global self-attributions, such as worth- lessness and inferiority [54], further exacerbating hopeless- ness, whereas SP more typically involves beliefs of vulnerability and outsiders’ criticality [49]. Among depressive adolescents, the behavioral symptoms of AN (e.g. restrictive dieting and the strict control of one’s body shape) may rep- resent a solution to the need for controlling negative feelings associated with such self-attributions [54]: once controlled, the need to seek help may also disappear. Our results extend the scarce knowledge base on the correlates of treatment seeking in adolescents with EDs [21,24].

Role of covariates. In the LRAs, family relocation predicted subsequent BN, and not seeking help for BN symptoms, while parental unemployment predicted subsequent depres- sion. These findings support the notion that family economic

hardships and social transitions both contribute to emotional adjustment and treatment seeking in adolescents [26,45,46].

Methodological considerations. This study has a number of limitations. Self-report instruments assessing psychopath- ology, treatment seeking and BMI were used. In psychopath- ology assessments, the self-report method carries the risk of including false positives in the analyses [33]. Identifying ‘disorders’ from participants’ self-reported symptoms is obvi- ously not as reliable a method as performing a clinical inter- view. Thus, caution is needed when comparing our results with the results of studies using structured clinical assess- ments. Furthermore, using survey questionnaires consisting of multiple self-report instruments is an inherent weakness of broad-range survey studies.

To identify a present episode of AN and to exclude recov- ered participants, we combined the DSM-IV criteria for an AN episode with a conservative normal weight BMI cutoff at T1. While aiming to be comprehensive, this method may lead to including some adolescents who are already in remission from AN and to an inflation of the prevalence estimate. Due to the lack of weight and height reports at T2, there was a discrepancy between the AN definition at T1 and T2, which should be considered a methodological weakness. With regard to treatment seeking for EDs, we used two questions that relied on the self-identification of an ED. The reliability and validity of these questions is unknown. Thus, the results regarding treatment seeking should be interpreted with caution.

Despite repeated attempts to acquire responses from the non-responders, a number of participants were lost at the follow-up. We reached 63% of the participants at T2. It is possible that a higher rate of psychopathology would have been detected among the dropouts than the completers at that time point. The attrition analysis cannot fully determine whether this is the case. However, we did observe a differ- ence in the rate of depression at T1, but not in other disor- ders at T1 between the dropouts and completers. We recommend caution in interpreting the results, as depressive disorders may be underrepresented among the completers. It is quite difficult to estimate how this may have impacted the results. It can be assumed that this at least weakened our ability to detect all possible longitudinal associations between the disorders.

The strengths of the study include our use of well-studied psychopathology measures that have been validated for use both in Finland and in several other countries. In addition, we gathered a large population-based sample covering all the secondary schools in two major urban areas in Finland.

Clinical implications. The clinical implications of the pre- sent study are that healthcare professionals, for example in school health care, should monitor adolescents with AN for the possible emergence of depressive symptoms. In addition, clinicians should enquire about possible eating problems when encountering adolescents with SP and depression. If such problems are noted, the adolescents should be informed of accessible treatment options.

In conclusion, our study shows that AN in middle adoles- cence increases the risk for the development of depression 2 years later in late adolescence. Furthermore, SP and

NORDIC JOURNAL OF PSYCHIATRY 611

depression, both independently and in specific ways, increase the risk that an adolescent will not seek treatment for emerg- ing ED symptoms. Specifically, SP may delay treatment seek- ing for BN, and depression may delay treatment seeking for AN.

Disclosure statement

No potential conflict of interest was reported by the authors.

References

[1] Hudson JI, Hiripi E, Pope HG, et al. The prevalence and correlates of eating disorders in the National Comorbidity Survey Replication. Biol Psychiatry. 2007;61:348–358.

[2] Swanson SA, Crow SJ, Le Grange D, et al. Prevalence and corre- lates of eating disorders in adolescents: results from the National Comorbidity Survey Replication Adolescent Supplement. Arch Gen Psychiatry. 2011;68:714–723.

[3] Hoek HW, van Hoeken D. Review of the prevalence and incidence of eating disorders. Int J Eat Disord. 2003;34:383–396.

[4] Fischer S, Le Grange D. Comorbidity and high-risk behaviors in treatment-seeking adolescents with bulimia nervosa. Int J Eat Disord. 2007;40:751–753.

[5] Kaye WH, Bulik CM, Thornton L, et al. Comorbidity of anxiety dis- orders with anorexia and bulimia nervosa. Am J Psychiatry. 2004;161:2215–2221.

[6] Godart NT, Flament MF, Curt F, et al. Anxiety disorders in subjects seeking treatment for eating disorders: a DSM-IV controlled study. Psychiatry Res. 2003;117:245–258.

[7] Salbach-Andrae H, Lenz K, Simmendinger N, et al. Psychiatric comorbidities among female adolescents with anorexia nervosa. Child Psychiatry Hum Dev. 2008;39:261–272.

[8] Godart NT, Flaming MF, Lecrubier Y, et al. Anxiety disorders in anorexia nervosa and bulimia nervosa: co-morbidity and chron- ology of appearance. Eur Psychiatry. 2000;15:38–45.

[9] Bulik CM, Sullivan PF, Fear JL, et al. Eating disorders and ante- cedent anxiety disorders: a controlled study. Acta Psychiatr Scand. 1997;96:101–107.

[10] Sihvola E, Keski-Rahkonen A, Dick DM, et al. Prospective associa- tions of early-onset Axis I disorders with developing eating disor- ders. Compr Psychiatry. 2009;50:20–25.

[11] Ivarsson T, Råstam M, Wentz E, et al.. Depressive disorders in teenage-onset anorexia nervosa: a controlled longitudinal, partly community-based study. Compr Psychiatry. 2000;41:398–403.

[12] Zaider TI, Johnson JG, Cockell SJ. Psychiatric disorders associated with the onset and persistence of bulimia nervosa and binge eat- ing disorder during adolescence. J Youth Adolesc. 2002;31: 319–329.

[13] Godart NT, Perdereau F, Rein Z, et al. Comorbidity studies of eat- ing disorders and mood disorders. Critical review of the literature. J Affect Disord. 2007;97:37–49.

[14] Pallister E, Waller G. Anxiety in the eating disorders: understand- ing the overlap. Clin Psychol Rev. 2000;28:366–386.

[15] Bulik CM. Anxiety disorders and eating disorders: a review of their relationship. NZJ Psychol. 1995;24:51–62.

[16] Clark DM. A cognitive perspective on social phobia. In: Crozier WL, Alden LE, editors. International handbook of social anxiety: concepts, research and interventions relating to the self and shy- ness. New York: John Wiley & Sons; 2001. p. 406–430.

[17] Levinson CA, Rodebaugh TL. Social anxiety and eating disorder comorbidity: the role of negative social evaluation fears. Eat Behav. 2012;13:27–35.

[18] Schlesier-Carter B, Hamilton S, O’Neil PM, et al. Depression and bulimia: the link between depression and bulimic cognitions. J Abnorm Psychol. 1989;98:322–325.

[19] Mattar L, Huas C, Duclos J, et al. Relationship between mal- nutrition and depression or anxiety in anorexia nervosa: a

critical review of the literature. J Affect Disord. 2010;132: 311–318.

[20] Hughes EK. Comorbid depression and anxiety in childhood and adolescent anorexia nervosa: prevalence and implications for out- come. Clin Psychologist. 2012;16:15–24.

[21] Ali K, Farrer L, Fassnacht DB, et al. Perceived barriers and facilita- tors towards help-seeking for eating disorders: a systematic review. Int J Eat Disord. 2017;50:9–21.

[22] Burstein M, He J-P, Kattan G, et al. Social phobia in the National Comorbidity Survey – Adolescent Supplement: prevalence, corre- lates, and comorbidity. J Am Acad Child Adolesc Psychiatry. 2001;50:870–880.

[23] Goodwin RD, Fitzgibbon ML. Social anxiety as a barrier to treat- ment for eating disorders. Int J Eat Disord. 2002;32:103–106.

[24] Hart LM, Granillo MT, Jorm AF, et al. Unmet need for treatment in the eating disorders: a systematic review of eating disorder spe- cific treatment seeking among community cases. Clin Psychol Rev. 2011;31:727–735.

[25] Galbaud du Fort G, Newman SC, Boothroyd LJ, et al. Treatment seeking for depression: role of depressive symptoms and comor- bid psychiatric diagnoses. J Affect Disord. 1999;52:31–40.

[26] Cachelin FM, Rebeck R, Veisel C, et al. Barriers to treatment for eating disorders among ethnically diverse women. Int J Eat Disord. 2001;30:269–278.

[27] Sawyer MG, Borojevic N, Ettridge KA, et al. Do helpseeking inten- tions during early adolescence vary for adolescents experiencing different levels of depressive symptoms? J Adolesc Health. 2012;50:236–242.

[28] Fr€ojd S, Marttunen M, Pelkonen M, et al. Perceived financial diffi- culties and maladjustment outcomes in adolescence. Eur J Public Health. 2006;16:542–554.

[29] Connor KM, Davidson JRT, Churchill LE, et al. Psychometric prop- erties of the Social Phobia Inventory (SPIN). New self-rating scale. Br J Psychiatry. 2000;176:379–386.

[30] Garcia-Lopez LJ, do Ceu Salvador M, De Los Reyes A. Assessment of social anxiety in adolescents. In: Ranta K, La Greca AM, Garcia- Lopez L-J, et al., editors. Social anxiety and phobia in adolescents: development, manifestation and intervention strategies. Heidelberg: Springer; 2015. p. 121–150.

[31] Ranta K, Kaltiala-Heino R, Rantanen P, et al. Screening social pho- bia in adolescents from general population: the validity of the Social Phobia Inventory (SPIN) against a clinical interview. Eur Psychiatry. 2007;22:244–251.

[32] Beck AT, Beck RW. Screening depressed patients in family prac- tice. A rapid technic. Postgrad Med. 1971;52:81–85.

[33] Young JF, Miller MR, Khan N. Screening and managing depression in adolescents. Adolesc Health Med Ther. 2010;1:87–95.

[34] Stockings E, Degenhardt L, Lee YY, et al. Symptom screening scales for detecting major depressive disorder in children and adolescents: a systematic review and meta-analysis of reliability, validity and diagnostic utility. J Affect Disord. 2015;174:447–463.

[35] Bennett DS, Ambrosini PJ, Bianchi M, et al. Relationship of Beck Depression Inventory factors to depression among adolescents. J Affect Disord. 1997;45:127–134.

[36] Raitasalo R. Mood questionnaire. Finnish modification of the short form of the Beck Depression Inventory. Helsinki: The Social Insurance Institution; 2007. [Book in Finnish]

[37] Kaltiala-Heino R, Rimpel€a M, Rantanen P, et al. Finnish modifica- tion of the 13-item Beck Depression Inventory in screening an adolescent population for depressiveness and positive mood. Nord J Psychiatry. 1999;53:451–457.

[38] Beck AT, Rial WY, Rickels K. Short form of depression inventory: a cross-validation. Psychol Rep. 1974;34:118.

[39] Fountoulakis KN, Iacovides A, Kleanthou S, et al. The Greek trans- lation of the symptoms rating scale for depression and anxiety: preliminary results of the validation study. BMC Psychiatry. 2003;3:1–8.

[40] American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders: 4th ed., Text Revision. Washington: American Psychiatric Association; 2000.

612 K. RANTA ET AL.

[41] Royal College of Psychiatrists [Internet]. Guidelines for the nutri- tional management of anorexia nervosa. London: Royal College of Psychiatrists; July 2005. (Council Report CR130). Available from: http://rcpsych.ac.uk/files/pdfversion/cr130.pdf

[42] Swenne I. Weight requirements for return of menstruations in teenage girls with eating disorders, weight loss and secondary amenorrhoea. Acta Paediatr. 2004;93:1449–1455.

[43] Isomaa R, Isomaa A, Kaltiala-Heino R, et al. Capturing clinically sig- nificant eating pathology in adolescence. Eur J Psychiat. 2013;27:122–128.

[44] Johnson J, McCutcheon S. Assessing life stress in older children and adolescents: preliminary findings with the Life Events Checklist. In: Sarason IG, Spielberger CD, editors. Stress and anx- iety. Vol 7. Washington: Hemisphere; 1980. p. 111–125.

[45] Sund AM, Larsson B, Wichstrom L. Psychosocial correlates of depressive symptoms among 12-14-year-old Norwegian adoles- cents. J Child Psychol & Psychiat. 2003;44:588–597.

[46] Vernberg EM, Abwender DA, Ewell KE, et al. Social anxiety and peer relationships in early adolescence. A prospective analysis. J Clin Child Psychol. 1992;21:189–196.

[47] Gauthier C, Hassler C, Mattar L, et al. Symptoms of depression and anxiety in anorexia nervosa: links with plasma tryptophan and serotonin metabolism. Psychoneuroendocrinology. 2014; 39:170–178.

[48] Oldershaw A, Lavender T, Sallis H, et al. Emotion generation and regulation in anorexia nervosa: a systematic review and meta-analysis of self-report data. Clin Psychol Rev. 2015;39:83–95.

[49] Hinrichsen H, Waller G, Emanuelli F. Social anxiety and agorapho- bia in the eating disorders: associations with core beliefs. J Nerv Ment Dis. 2004;192:784–787.

[50] Kessler RC, Berglund P, Demler O, et al. Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry. 2005;62:593–602.

[51] Rao PA, Beidel DC, Turner SM, et al. Social anxiety disorder in childhood and adolescence: descriptive psychopathology. Behav Res Ther. 2007;45:1181–1191.

[52] Grabhorn R, Stenner H, Stangier U, et al. Social anxiety in ano- rexia and bulimia nervosa: the mediating role of shame. Clin Psychol Psychother. 2006;13:12–19.

[53] Becker AE, Arrindell AH, Perloe A, et al. A qualitative study of per- ceived social barriers to care for eating disorders: perspectives from ethnically diverse health care consumers. Int J Eat Disord. 2010;43:633–647.

[54] Espindola CR, Blay SL. Anorexia nervosa treatment from the patient perspective: a metasynthesis of qualitative studies. Ann Clin Psychiatry. 2009;21:38–48.

NORDIC JOURNAL OF PSYCHIATRY 613

Copyright of Nordic Journal of Psychiatry is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder’s express written permission. However, users may print, download, or email articles for individual use.