AND 00:1–8 (2014) Research Article DISORDERS AND THE PERSISTENCE OF ANXIETY DISORDERS IN A NATIONALLY REPRESENTATIVE SAMPLE ∗ Andrew E. Skodol, M.D.,1,2,3 Timothy Geier, M.S.,2,4 Bridget F. Grant, Ph.D.,5 and Deborah S. Hasin, Ph.D.1,2,6

Background: Among individuals with anxiety disorders, comorbid personality disorders (PDs) increase cross-sectional symptom severity and decrease function- ing. Little is known, however, about how PDs influence the course of anxiety disorders over time. The purpose of this study was to examine the effect of PDs on the persistence of four anxiety disorders in a nationally representative sample in the United States. Methods: Two waves of data were collected on 34,653 participants, 3 years apart. At both waves, participants were evaluated for gen- eralized (GAD), social and specific , and . Predictors of persistence included all DSM-IV PDs. Control variables included demographics, comorbid PDs, age at onset of the anxiety disorder, number of prior episodes, duration of the current episode, treatment history, and cardinal symptoms of exclusionary diagnoses for each anxiety disorder. Results: Any PD, two or more PDs, borderline PD, schizotypal PD, mean number of PD criteria met, and mean number of PDs diagnosed predicted the persistence of all four anx- iety disorders. Narcissistic PD predicted persistence of GAD and panic disorder. Schizoid and avoidant PDs also predicted persistence of GAD. Finally, avoidant PD predicted persistence of social . Particular patterns of cross-cluster PD were strong predictors of the persistence of individual anxiety dis- orders as well. Conclusions: In this national sample, a variety of PDs robustly predicted the persistence of anxiety disorders over 3 years, consistent with the results of recent prospective clinical studies. Personality should be assessed and addressed in treatment for all patients with anxiety disorders. Depression and Anxiety 00:1–8, 2014. C 2014 Wiley Periodicals, Inc.

Key words: personality disorder; anxiety disorder; comorbidity; ; clinical course

INTRODUCTION able psychosocial impairment,[4, 5] and have surprisingly low rates of remission,[2,6,7] suggesting a chronic bur- Anxiety disorders are prevalent in the clinic and [1–3] den. Among individuals with anxiety disorders, comor- in the community, are associated with consider- bid personality disorders (PDs) increase cross-sectional

Contract grant sponsor: NIAAA; Contract grant numbers: 1Department of , College of Physicians and Sur- U01AA01811 and K05 AA014223; Institutional support: New York geons, Columbia University, New York, New York State Psychiatric Institute. 2New York State Psychiatric Institute, New York, New York 3Department of Psychiatry, University of Arizona College of ∗ Correspondence to: Andrew E. Skodol, 2626 E. Arizona Biltmore Medicine, Tucson, Arizona Circle, #29, Phoenix, AZ 85016. E-mail: [email protected] 4Department of Psychology, University of Wisconsin, Milwau- kee, Wisconsin Received for publication 28 August 2013; Revised 20 May 2014; 5Laboratory of Epidemiology and Biometry, National Institute Accepted 27 May 2014 on Alcohol and Alcoholism, Bethesda, Maryland DOI 10.1002/da.22287 6 Department of Epidemiology, Mailman School of Public Published online in Wiley Online Library Health, Columbia University, New York, New York (wileyonlinelibrary.com).

C 2014 Wiley Periodicals, Inc. 2 Skodol et al.

symptom severity and decrease functioning.[8–10] Identi- representative sample in the United States: GAD, social fying factors that contribute to a chronic course in anx- phobia, specific phobia, and panic disorder. The 3-year iety disorders is important for prognosis and treatment follow-up provides a unique opportunity to determine planning. Little is known about how PDs influence the rates of persistence of the four anxiety disorders and the course of anxiety disorders over time.[1] specific effects of all DSM-IV PDs, and representations Three prospective studies examined associations be- of PD severity, while controlling for other potentially tween PDs and the course of anxiety disorders in pa- prognostic factors. These data allow examination of the tients. In a 6-year follow-up of Norwegian outpatients, hypothesis generated in clinical populations that PDs ex- borderline PD predicted the presence of any anxiety ert strong, independent, adverse effects on the course of disorder, obsessive–compulsive PD predicted panic dis- common anxiety disorders. order, and avoidant PD predicted social phobia.[11] In the Harvard/Brown Anxiety Research Program (HARP) 5-year study, DSM-III-R dependent and avoidant PDs MATERIALS AND METHODS decreased the likelihood of remission from generalized PARTICIPANTS anxiety disorder (GAD) and avoidant PD decreased the Participants were respondents in Waves 1 and 2 of the Na- [12] likelihood of remission from social phobia. No PD tional Epidemiologic Survey on Alcohol and Related Conditions affected the course of panic disorder, while borderline, (NESARC).[19,20] The target population was the civilian noninsti- schizotypal, and obsessive–compulsive PDs were too in- tutionalized population 18 years and older in households and group frequent to analyze. The Collaborative Longitudinal quarters (e.g., college quarters, group homes, boarding houses) in the Study of Personality Disorders Study (CLPS) prospec- United States. Blacks, Hispanics, and adults ages 18–24 were over- tively examined the effects of DSM-IV PDs on the sampled, with data adjusted via sample weights for oversampling and course of several anxiety disorders over 7 years,[13] find- household- and person-level nonresponse, in order to make standard ing that a number of PDs adversely impacted the clini- errors of estimates accurate, and to make the findings representative of the general population.[20] Of the 43,093 respondents interviewed at cal course of anxiety disorders. Specifically, schizotypal Wave 1, census-defined eligible respondents for Wave 2 re-interviews PD influenced the course of social phobia, PTSD, and included those not deceased (N = 1,403); deported, mentally or phys- GAD; avoidant PD influenced social phobia and OCD; ically impaired (i.e., too impaired to participate in an interview) (N = obsessive–compulsive PD influenced GAD, OCD, and 781); or on active military duty (N = 950). In Wave 2, 34,653 of 39,959 ; and borderline PD impacted the course of eligible respondents were re-interviewed, for a response rate of 86.7%. OCD, GAD, and panic disorder with agoraphobia. Sample weights further adjusted for Wave 2 nonresponse.[20] Overall, In addition, interest is growing in PD severity as a crit- 4,755 had GAD, social phobia, specific phobia, and/or panic disorder ical variable in PD psychopathology, stimulating efforts at Wave 1. Of these, 4,010 participated in Wave 2 and constitute the to revise the PD sections of both DSM-5[14] and ICD- present sample: 746 respondents meeting criteria for Wave 1 current 11.[15] Reviews of levels of personality functioning[16] and GAD, 989 for social phobia, 2,579 for specific phobia, and 775 for panic [17] disorder, with or without agoraphobia. Demographic characteristics of of other representations of PD severity suggest that the entire Wave 2 sample and of those with each anxiety disorder are severity might be the key PD factor in predicting con- shown in Table 1. In general, most respondents with anxiety disorders current and prospective functioning. Thus, associations were female, White, over age 40, married or cohabiting, and college of disorder chronicity with various representations of educated. PD severity, including PD cluster, number of PD cri- [17] teria, number of PDs, and “complex PD” may be PROCEDURES significant. In-person interviews were conducted at both waves by experienced Prospective studies of patient samples provide im- lay interviewers with extensive training and supervision.[19,20] All pro- portant information, but may be biased by numerous cedures, including informed consent, received full ethical review and confounds and selection factors differentiating those re- approval from the U.S. Census Bureau and U.S. Office of ceiving or not receiving treatment. Thus, generalization and Budget. from clinical samples may be difficult and prognostic fac- tors obscured or distorted. One prospective community ASSESSMENT AND VARIABLES [18] study showed that having any PD by early adulthood Interviewers administered the NIAAA Alcohol Use Disorder elevated the risk of any anxiety disorder by middle adult- and Associated Disabilities Interview Schedule-DSM-IV Version hood. However, the sample size precluded examination (AUDADIS-IV),[21] a structured diagnostic interview developed to as- of specific PDs and specific anxiety disorders. Larger sess substance use and other mental disorders in large-scale surveys. prospective studies of general population samples are Computer algorithms produced diagnoses of DSM-IV Axis I disorders needed to better understand the relationship of PDs to and all DSM-IV PDs. persistent anxiety disorders. Anxiety Disorders. GAD, social phobia, specific phobia, and panic disorder (with or without agoraphobia) were defined accord- ing to DSM-IV inclusion criteria, including all symptom, duration, AIMS OF THE STUDY and clinical significance (i.e., distress or impairment) criteria. Diag- noses additionally required that the disorders be “primary,” that is, The present study prospectively examines the impact not substance-induced or due to a general medical conditions.[3,22–24] of specific PDs on the persistence of four anxiety disor- Other exclusionary criteria were controlled for in subsequent analyses ders assessed at two points in time in a large nationally (see below).

Depression and Anxiety Research Article: Personality Disorders and Anxiety Disorders 3

TABLE 1. Demographic characteristics of the full sample and by anxiety disorder at Wave 1

Full sample Social phobia Specific phobia Panic disorder N = 34,653 GAD N = 746 N = 989 N = 2,579 N = 775 Demographic characteristic N %(SE) N %(SE) N %(SE) N %(SE) N %(SE)

Sex Male 14,564 47.92 (0.3) 194 29.65 (2.2) 340 37.28 (1.8) 710 31.47 (1.1) 206 28.75 (2.0) Female 20,089 52.08 (0.3) 552 70.35 (2.2) 649 62.72 (1.8) 1,869 68.53 (1.1) 569 71.25 (2.0) Race/ethnicity White 20,174 70.93 (1.5) 486 76.49 (2.0) 663 77.75 (1.9) 1,586 75.06 (1.4) 504 77.21 (2.1) Non-White 14,479 29.07 (1.5) 260 23.51 (2.0) 326 22.25 (1.9) 993 24.94 (1.4) 271 22.79 (2.1) Age 18–29 6,719 21.80 (0.4) 150 22.99 (1.9) 227 25.20 (1.7) 577 25.12 (1.2) 162 23.42 (2.0) 30–39 7,299 20.10 (0.3) 185 24.18 (1.8) 216 21.61 (1.7) 570 20.97 (1.1) 206 25.81 (1.8) 40–49 7,146 20.77 (0.3) 190 25.99 (1.9) 254 24.90 (1.7) 563 22.95 (1.0) 211 27.13 (1.9) 50+ 13,489 37.33 (0.5) 221 26.84 (1.8) 292 28.28 (1.7) 869 30.96 (1.1) 196 23.63 (1.9) Education Less than college 15,699 43.68 (0.6) 355 47.80 (2.3) 495 48.79 (2.2) 1,161 43.53 (1.3) 356 44.39 (2.2) College or higher 18,954 56.32 (0.6) 391 52.20 (2.3) 494 51.21 (2.2) 1,418 56.47 (1.3) 419 55.61 (2.2) Marital status As if/married 18,413 63.06 (0.5) 308 54.20 (2.1) 468 58.35 (1.8) 1,341 62.36 (1.1) 363 57.94 (2.0) Not married 16,240 36.94 (0.5) 438 45.80 (2.1) 521 41.65 (1.8) 1,238 37.64 (1.1) 412 42.06 (2.0)

Persistence was defined identically for all four anxiety disorders. At TABLE 2. Frequencies of individual personality Wave 1, criteria for each anxiety disorder were assessed in two time disorders and other personality disorder severity frames: (1) current, that is, during the last 12 months; and (2) prior to variables by anxiety disorder at Wave 1 the last 12 months. At Wave 2, 3 years later, these criteria were again assessed in two time frames covering the time period between Waves Generalized 1 and 2: (1) current, last 12 months; and (2) prior to last 12 months, anxiety Social Specific Panic but since Wave 1. From these data, the outcome variable was created. Personality disorders disorder phobia phobia disorder = = = = Persistent anxiety disorder was defined as meeting full criteria for current and severity variables N 746 N 989 N 2,579 N 775 disorder at Wave 1, and full criteria for the same disorder during each NNNN time period of the 3-year follow-up. The AUDADIS has fair to good Antisocial 16 15 32 18 test–retest reliability for these anxiety disorders (K = 0.40–0.52),[25] Avoidant 155 285 243 138 similar to or better than other instruments used in epidemiological Borderline 216 223 333 206 studies.[26,27] Dependent 41 49 48 41 Predictors of Outcomes. The main predictors of anxiety disor- Histrionic 79 92 175 71 der persistence were all 10 individual DSM-IV PDs and the PD Clus- Narcissistic 115 125 275 119 ters A, B, and C. In addition, various representations of PD severity OCPD 241 318 524 214 were tested, including having more than one PD, cross-cluster comor- Paranoid 241 299 427 191 bidity or “complex PD,”[15,17,28] mean number of PD criteria,[17,29] Schizoid 150 218 262 133 and mean number of PDs.[17] Frequencies of the individual PDs and Schizotypal 137 179 229 139 other PD variables by each anxiety disorder are shown in Table 2. Con- Any PD 474 639 1,068 449 trol variables included (1) demographic characteristics, (2) comorbid 2+ PDs 351 443 638 313 PD, (3) clinical characteristics of anxiety disorder course, (4) treatment Any Cluster A only 48 76 131 51 history, and (5) cardinal symptoms of exclusionary diagnoses for each Any Cluster B only 48 34 163 61 anxiety disorder (see below). Any Cluster C only 59 122 204 55 Personality Disorders. PDs, except for antisocial, were assessed Any Cluster A + B only 66 57 123 59 with an introduction and repeated reminders asking respondents to Any Cluster B + Conly 24 42 64 29 answer about how they felt or acted “most of the time, throughout Any Cluster A + C only 85 125 161 71 your life, regardless of the situation or whom you were with.” Respon- Any Cluster A + B + C 144 183 222 123 dents were instructed not to include symptoms occurring only when depressed, manic, anxious, drinking heavily, using drugs, recovering from the effects of alcohol or drugs, or physically ill. PD criteria items were adapted from items in the DSM-IV versions of semistructured with adult criteria reassessed at Wave 2. Antisocial PD was considered diagnostic interviews (e.g., the Structured Clinical Interview for DSM- present if respondents met criteria for lifetime disorder at Wave 1 and IV Personality Disorders and Personality Disorder Examination). For at least three adult criteria persisted at Wave 2. NESARC test–retest all symptoms coded positive, respondents were asked about distress and studies indicate reliability from fair (paranoid, histrionic, avoidant, social/occupational dysfunction: “Did this ever trouble you or cause K = 0.40–0.45) to very good (schizotypal, antisocial, narcissistic, bor- problems at work or school, or with your family or other people?” derline, K = 0.67–0.71),[25,35] generally comparable to the range re- Scoring algorithms for diagnoses required associated distress or so- ported for patient studies.[36] Good convergent validity for these di- cial/occupational dysfunction, in addition to the specified number of agnoses is indicated by significant associations with separate measures criteria.[30–34] of impairment.[30–34] Avoidant, dependent, histrionic, obsessive–compulsive, paranoid, Because not all PDs were measured at Wave 1, we investigated and schizoid PDs were assessed at Wave 1; borderline, narcissistic, their validity at both waves in the present sample, using two methods. and schizotypal were assessed at Wave 2. Lifetime antisocial PD (in- The first used weighted linear regressions to compare respondents cluding both childhood and adult criteria) was assessed at Wave 1, with each PD to those with no PD on impairment at Waves 1 and 2,

Depression and Anxiety 4 Skodol et al.

measured with the Mental Component Summary of the SF-12v2.[37] RESULTS Some between-wave change in scores occurred, consistent with prior research, but the participants with any of the PDs, with the excep- GENERALIZED ANXIETY DISORDER tion of antisocial PD, consistently had significantly greater impairment Of the respondents with GAD in Wave 1, 119 (15.1% < (P .01) in both waves, regardless of the wave in which the disorder [SE = 1.6]) had persistent GAD in Wave 2, meeting was assessed. The second method used logistic regression to compare criteria for the disorder during each time period of the respondents with each PD to those with no PD on Wave 1 and Wave 2 life events suggesting impaired functioning, such as breaking up of re- 3-year follow-up. Of the demographic and clinical vari- ables, only age (40–49) (OR = 2.45, 95% CI = 1.14– lationships; problems with friends, employer, or finances; being fired = = or laid off; and being unemployed and looking for work. In general, re- 5.27) and treatment (OR 1.86, 95% CI 1.09–3.16) spondents with PDs were more likely to experience the events at both predicted persistence in univariate analyses. In the final Waves 1 and 2, regardless of when their disorder was diagnosed. The multiple regression models, any PD and two or more consistency of association of both impairment indicators with PDs at PDs were significantly related to GAD persistence (see Waves 1 and 2, regardless of when the disorder was assessed, supported Table 3). Cluster B PD only predicted persistence (OR = the validity of the PD diagnoses. [Details available on request.] 3.62, 95% CI = 1.02–12.88). Of the specific PDs, nar- Demographic Characteristics. These included gender, race/ cissistic (OR = 3.28, 95% CI = 1.58–6.77), schizotypal ethnicity, age, education, and marital status. (OR = 2.88, 95% CI = 1.50–5.51), schizoid (OR = 2.59, Clinical Characteristics of Anxiety Disorder Course. Since 95% CI = 1.45–4.64), borderline (OR = 2.54, 95% CI = early onset, recurrence, and previous chronicity predicted persistent = = anxiety disorder in other studies,[38,39] analyses included age at first on- 1.36–4.77), and avoidant (OR 2.26, 95% CI 1.15– set, number of prior episodes, and duration (years) of current episode. 4.44) PDs all predicted persistence of GAD controlling Treatment. Current treatment for internalizing (mood or anx- for demographics, any other PD, age at onset, number iety) disorder was examined, including outpatient services (counselor, of episodes, duration of current episode, treatment, and therapist, physician, or other professional), inpatient services (hospi- final controls for cardinal symptoms of disorders that ac- talized overnight or longer), and prescribed medication.[40] cording to DSM-IV would exclude GAD (see Table 3). Of the PD severity measures, comorbidity across Clus- STATISTICAL ANALYSES ters A and B only (OR = 4.27, 95% CI = 1.48–12.28); A and C only (OR = 3.43, 95% CI = 1.41–8.38); and A, B, Weighted means, frequencies, and univariate associations were = = computed for GAD, social phobia, specific phobia, and panic disor- and C (OR 5.14, 95% CI 2.76–9.55); mean number der. Relationships between each predictor (PD and control) and the of PD criteria (OR = 1.08, 95% CI = 1.05–1.11); and binary outcome variables representing the persistence of each anxiety mean number of PDs (OR = 1.58, 95% CI = 1.35–1.85) disorder were tested with separate multiple logistic regression models, predicted persistence of GAD. producing adjusted odds ratios (ORs) and 95% confidence intervals (CIs). Standard errors and 95% CIs for the predictors were estimated SOCIAL PHOBIA with SUDAAN, using Taylor series linearization to adjust for design effects of complex sample surveys. Of the respondents with social phobia in Wave 1, To test the effect of PDs on the persistence of each anxiety disorder, 174 (17.6% [SE = 1.5]) had persistent social phobia in we first tested individual models for each PD, PD cluster, two or more Wave 2. No demographic variable was associated with PDs, and cross-cluster , adjusting for demographic fac- persistence. Of other clinical variables, only receiving tors. Next, we tested a model incorporating demographic factors and treatment was associated with persistence (OR = 2.91, a variable (yes/no) representing PD comorbidity beyond the PD vari- = [41] CI 1.64–5.19). Any PD and two or more PDs were able being tested. Given the comorbidity of PDs with each other, significantly related to social phobia persistence (see this model was needed to test the effect of each PD (PD cluster, and Table 3). Any Cluster C PD only predicted persistence pattern of PD comorbidity) adjusting for any other to detect if any = = particular PD (or combination) had unique effects. For the PD clus- (OR 2.26, 95% CI 1.03–4.99). Specific PDs asso- ters, control was for any PD outside of that cluster; for patterns of PD ciated with persistent social phobia included schizotypal comorbidity, control was for a PD outside of that pattern. No comor- (OR = 2.77, 95% CI = 1.64–4.66), borderline (OR = bidity controls were possible for continuous PD variables (i.e., number 2.59, 95% CI = 1.49–4.40), and avoidant PD (OR = of PDs or PD criteria counts). To then establish the robustness of dis- 1.96, 95% CI = 1.20–3.20). Clusters A and B comor- tinct PD effects, we added age at first onset, number of prior episodes, bidity (OR = 10.31, 95% CI = 4.29–24.78), Clusters B duration of current episode, and Wave 1 treatment utilization for each and C comorbidity (OR = 3.66, 95% CI = 1.30–10.28), individual anxiety disorder. Finally, to determine whether the complex mean number of PD criteria (OR = 1.04, 95% CI = DSM-IV diagnostic hierarchies for the anxiety disorders affected the 1.02–1.06), and mean number of PDs (OR = 1.30, 95% findings, we tested a final model with cardinal symptoms of exclusion- = ary disorders or other conditions added to the above covariates. CI 1.16–1.45) predicted persistence of social phobia. The cardinal symptoms variable was established via DSM-IV hi- erarchical rules. For GAD, this variable was positive if symptoms of OCD, separation anxiety, weight gain, physical complaints, serious ill- Of respondents with specific phobia at Wave 1, 571 ness, major depression, , or psychotic disorder occurred during = all episodes. For , this variable was positive if symptoms (21.9% [SE 1.0]) persisted over 3 years. Having re- ceived treatment was associated with persistence (OR = of separation anxiety, body dysmorphic, or eating disorders; PTSD; = OCD; or physical illness occurred during all episodes. For specific 1.97, 95% CI 1.13–3.44). Any PD and two or more phobia and panic disorder, this variable was positive if symptoms of PDs were significantly related to specific phobia per- OCD, PTSD, or separation anxiety occurred during all episodes. sistence (see Table 3). Cluster B PD only predicted

Depression and Anxiety Research Article: Personality Disorders and Anxiety Disorders 5

TABLE 3. Personality disorders (PDs) as predictors of anxiety disorder persistence among those who had anxiety disorders at Wave 1: odds ratios (OR) and 95% confidence intervals (CI)

Generalized anxiety disordera Social phobiaa Specific phobiaa Panic disordera N = 119 N = 174 N = 571 N = 115 Personality disorder ORb 95% CI ORb 95% CI ORb 95% CI ORb 95% CI

Antisocial 3.82 0.58–25.19 1.97 0.41–9.58 2.12 0.57–7.94 1.63 0.19–13.72 Avoidant 2.26 1.15–4.44 1.96 1.20–3.20 1.21 0.77–1.90 1.18 0.52–2.69 Borderline 2.54 1.36–4.75 2.59 1.49–4.50 1.68 1.15–2.46 4.54 2.36–8.71 Dependent 2.40 0.84–6.83 0.98 0.31–3.10 2.06 0.91–4.67 2.13 0.59–7.71 Histrionic 1.04 0.39–2.78 1.40 0.70–2.79 1.21 0.71–2.04 1.12 0.42–2.96 Narcissistic 3.28 1.58–6.77 1.71 0.96–3.03 1.13 0.73–1.75 2.22 1.11–4.44 OCPDc 1.84 0.97–3.52 0.75 0.44–1.28 0.94 0.68–1.29 1.55 0.82–2.95 Paranoid 1.94 1.00–3.74 1.18 0.71–1.96 1.31 0.86–2.01 0.69 0.33–1.46 Schizoid 2.59 1.45–4.64 1.12 0.66–1.91 1.05 0.70–1.57 1.45 0.65–3.32 Schizotypal 2.88 1.50–5.51 2.77 1.64–4.66 2.53 1.67–3.85 2.26 1.03–4.96 Any PD 3.82 1.96–7.45 3.07 1.72–5.50 1.59 1.22–2.07 2.14 1.08–4.24 2+ PDs 4.00 2.17–7.39 3.01 1.77–5.11 1.90 1.41–2.55 3.06 1.66–5.63 Any Cluster Ad — — 1.47 0.55–3.95 1.53 0.89–2.63 0.47 0.14–1.63 Any Cluster B 3.62 1.02–12.88 1.21 0.34–4.32 1.90 1.18–3.05 1.83 0.65–5.19 Any Cluster C 2.10 0.69–6.40 2.26 1.03–4.99 0.98 0.61–1.58 1.21 0.34–4.29 Any Cluster A + B 4.27 1.48–12.28 10.31 4.29–24.78 2.30 1.32–4.02 5.70 2.30–14.12 Any Cluster B + C 0.27 0.04–1.65 3.66 1.30–10.28 1.30 0.64–2.65 7.26 2.27–23.26 Any Cluster A + C 3.43 1.41–8.36 1.84 0.86–3.96 1.81 1.12–2.94 0.71 0.12–4.07 Any Cluster A + B + C 5.14 2.76–9.55 1.53 0.92–2.53 1.37 0.90–2.09 2.11 0.90–4.92 PD symptoms (0–79)e,f 1.08 1.05–1.11 1.04 1.02–1.06 1.03 1.02–1.04 1.05 1.03–1.08 Total number of PDs (0–10)e,g 1.58 1.35–1.85 1.30 1.16–1.45 1.18 1.09–1.29 1.30 1.11–1.52 aSubstance and illness etiologies excluded. bControlled for age, sex, education, race/ethnicity, comorbid personality disorder, age at first onset, current episode duration, number of prior episodes, treatment for internalizing (mood or anxiety) disorder in past 12 months at Wave 1, and cardinal symptoms. ORs in bold indicate significant associations. cOCPD, obsessive–compulsive personality disorder. dOnly one person had a Cluster A PD only and persistent generalized anxiety disorder. eORs represent change in odds of persistence given one unit change. fWeighted mean PD criteria count by Wave 1 anxiety disorder: among those with W1 GAD, mean = 22.91 (SE = 0.70); among those with W1 social phobia, mean = 22.98 (SE = 0.52); among those with W1 specific phobia, mean = 16.99 (SE = 0.31); among those with W1 panic disorder, mean = 21.23 (SE = 0.52). gWeighted mean number of PDs by Wave 1 anxiety disorder: among those with W1 GAD, mean = 1.86 (SE = 0.11); among those with W1 social phobia, mean = 1.83(SE = 0.08); among those with W1 specific phobia, mean = 0.99 (SE = 0.04); among those with W1 panic disorder, mean = 1.54 (SE = 0.08). persistence (OR = 1.90, 95% CI = 1.18–3.05). Among panic disorder persistence (see Table 3). No PD cluster the specific PDs, schizotypal (OR = 2.53, 95% CI = predicted persistence. The specific PDs associated with 1.67–3.85) and borderline (OR = 1.68, 95% CI = 1.15– panic disorder persistence were borderline (OR = 4.54, 2.46) were associated with persistence of a specific pho- 95% CI = 2.36–8.71), schizotypal (OR = 2.26, 95% CI = bia. Clusters A and B comorbidity (OR = 2.30, 95% CI = 1.03–4.96), and narcissistic (OR = 2.22, 95% CI = 1.11– 1.32–4.02), Clusters A and C comorbidity (OR = 1.81, 4.44). Clusters A and B comorbidity (OR = 5.70, 95% 95% CI = 1.12–2.94), number of PD criteria (OR = CI = 2.30–14.12), Clusters B and C comorbidity (OR = 1.03, 95% CI = 1.02–1.04), and mean number of PDs 7.26, 95% CI = 2.27–23.26), mean number of PD crite- (OR = 1.18, 95% CI = 1.09–1.29) predicted persistence ria (OR = 1.05, 95% CI = 1.03–1.08), and mean number of a specific phobia. of PDs (OR = 1.30, 95% CI = 1.11–1.52) predicted per- sistence of panic disorder.

PANIC DISORDER Of respondents with panic disorder at Wave 1, 115 DISCUSSION (13.3% [SE = 1.5]) persisted. Of the demographic or This study provides a rigorous test of the impact other clinical variables only older age (OR = 3.67 for of PDs on the persistence of four anxiety disorders 30+ vs. 18–29, 95% CI = 1.69–7.97) was associated. Any in a large, nationally representative sample assessed PD and two or more PDs were significantly related to with a well-established instrument. Participants were

Depression and Anxiety 6 Skodol et al.

ascertained independently of treatment status and but owing to the broader range of types of anxiety dis- reevaluated 3 years later with excellent retention to orders studied in the CLPS (e.g., OCD, PTSD), other determine the rates of persistence of GAD, social and associations with PDs were found in that study. specific phobias, and panic disorder. This study tested Although all 10 PDs in Section II of DSM-5 will be di- the prognostic significance of individual PDs, PD agnosed using the same criteria as in DSM-IV, the results clusters, patterns of cross-cluster comorbidity, number of the current study lend some support to the retention of PD criteria met, and number of PDs diagnosed as of select PDs in the alternative model for PDs in DSM-5 present, while controlling for demographic factors, other Section III, “Emerging Measures and Models,”[43] based PD comorbidity, other clinical factors that could have on their prognostic significance. In addition to border- impacted the course of the anxiety disorders, treatment line PD, schizotypal, narcissistic, and avoidant PDs all history, and symptoms of exclusionary disorders and had significant effects on the persistence of multiple anx- conditions. The large sample allowed for multivariate iety disorders. Of the four PDs recommended for dele- tests of predictors in a logical progression that enabled tion as specific PDs, schizoid was the only one to show the untangling of the effects of these multiple factors in an (isolated) effect in this study; paranoid, histrionic, and a manner not possible in previous smaller studies. dependent PDs showed no effects. Although prognostic Rates of persistence of anxiety disorders varied from significance is not the only consideration in determin- 13.3% (panic disorder) to 21.9% (specific phobia) in ing which PDs should be retained as specific types in this study. These figures correspond to rates of per- the DSM, it has often been mentioned as an important sistence of the same anxiety disorders in the CLPS[13] element of clinical utility for psychiatric diagnoses.[44] over 7 years from 16.5% (GAD) to 22.1% (panic disor- In addition, this study provides some support for dif- der with or without agoraphobia), but are considerably ferent representations of PD severity, as opposed to PD lower than rates of persistence of these anxiety disorders type, as predictors of anxiety disorder persistence. For in the HARP study over 12 years.[2] Since the method- example, a cross-cluster comorbidity severity construct ologies of these two clinical studies are quite similar, the that has been under consideration for ICD-11[15, 17] is differences in rates of persistence presumably are due to supported by the strong predictive effects of Clusters A the nature of anxiety disorders in patients recruited for and B, B and C, and A and C comorbidity for several PDs (CLPS) versus those recruited for primary anxiety anxiety disorders. Two or more PDs, the total number disorders (HARP). Our finding that treatment was asso- of PD criteria met, and the number of PDs present were ciated with indicators of a more chronic course of each of also significant predictors for all four anxiety disorders. the anxiety disorders other than panic disorder is consis- So-called “complex PDs” and severity measures based on tent with the common observation in naturalistic studies criteria counts have been associated with poor outcomes that more treatment is received by patients who do not in other epidemiologic[45, 46] and clinical[28, 29] studies. recover. The relationship of avoidant PD and social phobia In contrast to previous NESARC findings that border- over time has been demonstrated,[47] and suggests patho- line PD was the single most robust predictor of persis- logical processes common to both disorders. Other PD tence of major depressive disorder,[42] effects for a num- and anxiety disorder relationships found here are, per- ber of PDs on the persistence of anxiety disorders were haps, less expected. The “affective instability” criterion found in this study. Borderline and schizotypal PDs pre- for borderline PD includes short-term anxiety and the dicted the persistence of all four anxiety disorders. Nar- criteria for schizotypal PD include one for “excessive so- cissistic PD predicted persistence of GAD and panic dis- cial anxiety.” The findings of the current study suggest order. In addition, schizoid and avoidant PDs predicted that anxiety proneness is more widespread in these two persistence of GAD. Finally, avoidant PD predicted the PDs than their current criteria imply. The criteria for persistence of social phobia. Interestingly, controls for narcissistic PD do not suggest that anxiety is a major is- symptoms that might have excluded an anxiety disorder sue. However, the concept of “vulnerable ,” diagnosis according to DSM-IV did not significantly af- which often underlies overt , reflects many fect the association of personality with the traits in the domain of , and anxious- persistence of the anxiety disorders. ness in particular.[48] These findings are consistent with those from the Strengths and limitations of the study are noted. CLPS regarding the effects of PDs on the course of Strengths include a large nationally representative anxiety disorders in clinical settings,[13] where a num- sample of adults with anxiety disorders ascertained in- ber of PDs impacted the course of different anxiety dis- dependently of treatment seeking; use of reliable and orders. For example, both studies found that avoidant standardized measures; high retention rates over a 3- PD increased persistence of social phobia. In addition, year follow-up; and multivariate analyses controlling for the previously unrecognized importance of schizotypal a comprehensive set of potential predictors. Potential PD in influencing the course of several anxiety disorders limitations include data obtained with structured inter- was highlighted in both studies. There were also differ- views administered by trained lay interviewers rather ences, however, in that more associations were found in than clinicians; only one follow-up time point 3 years the present study than in the CLPS with the persistence later; a single outcome for each anxiety disorder, rather of the four anxiety disorders common to both studies, than multiple possible indicators of outcome, which

Depression and Anxiety Research Article: Personality Disorders and Anxiety Disorders 7

should be examined in future studies; focus on DSM-IV 11. Alnæs R, Torgersen S. A 6-year follow-up study of anxiety disor- categorical PD diagnoses (i.e., alternative dimensional ders in psychiatric outpatients: development and continuity with models were not tested); and three PDs assessed at Wave personality disorders and personality traits as predictors. Nord J 2. Although our impairment analyses supported the va- Psychiatry 1999;53:409–416. lidity of the PD diagnoses regardless of the timing of 12. Massion AO, Dyck IR, Shea MT, et al. Personality disorders and their assessment, the design introduced the small pos- time to remission in generalized anxiety disorder, social phobia, and panic disorder. Arch Gen Psychiatry 2002;59:434–440. sibility that the timing of the assessments might have 13. Ansell EB, Pinto A, Edelen MO, et al. The association of per- affected some of the results. Finally, persistent anxiety sonality disorders with the prospective 7-year course of anxiety disorders might be a cause of personality dysfunction disorders. Psychol Med 2011;41:1019–1028. rather than an effect. 14. Skodol AE. Personality disorders in DSM-5. Annu Rev Clin Psy- chol 2012;8:317–344. 15. Tyrer P, Crawford M, Mulder R, et al. The rationale for the reclas- CONCLUSIONS sification of personality disorder in the 11th revision of the Interna- tional Classification of Diseases. Pers 2011;5:246– In summary, in this nationally representative sample of 259. adults with anxiety disorders, a variety of individual PDs, 16. Bender DS, Morey LC, Skodol AE. Toward a model for assess- particular patterns of cross-cluster PD comorbidity, and ing level of personality functioning in DSM-5, part I: a review of representations of PD severity robustly predicted the theory and methods. J Pers Assess 2011;93:332–346. persistence of individual anxiety disorders, consistent 17. Crawford MJ, Koldobsky N, Mulder R, et al. Classifying personal- with results of recent clinical studies. These findings sug- ity disorder according to severity. J Pers Disord 2011;25:321–330. gest that personality psychopathology should be assessed 18. Johnson JG, Cohen P, Kasen S, Brook JS. Personality disorders by clinicians, considered in prognosis, and addressed in evident by early adulthood and risk for anxiety disorders during the treatment of all patients with anxiety disorders. middle adulthood. J Anxiety Disord 2006;20:408–426. 19. Grant BF, Stinson FS, Dawson DA, et al. Prevalence and co- occurrence of 12-month alcohol and drug use disorders and per- Conflicts of Interest. The authors have no sonality disorders in the United States: results from the National interests to disclose. Epidemiologic Survey on Alcohol and Related Conditions. Arch Gen Psychiatry 2004;61:361–368. 20. Grant BF, Goldstein RB, Chou SP, et al. Sociodemographic and REFERENCES psychopathologic predictors of first incidence of DSM-IV sub- stance use, mood and anxiety disorders: results from the Wave 2 1. Roy-Byrne PP, Cowley DS. Course and outcome in panic disor- National Epidemiologic Survey on Alcohol and Related Condi- der: a review of recent follow-up studies. Anxiety 1994;1:151–160. tions. Mol Psychiatry 2009;14:1051–1066. 2. Bruce SE, Yonkers KA, Otto MW, et al. Influence of psychiatric 21. Grant BF, Dawson DA, Hasin, DS. The Alcohol Use Disorder comorbidity on recovery and recurrence in generalized anxiety and Associated Disabilities Interview Schedule—DSM-IV version disorder, social phobia, and panic disorder: a 12-year prospective (AUDADIS-IV), 2009. study. Am J Psychiatry 2005;162:1179–1187. 22. Grant BF, Hasin DS, Stinson FS, et al. Prevalence, correlates, 3. Grant BF, Hasin DS, Blanco C, et al. The epidemiology of social co-morbidity, and comparative disability of DSM-IV generalized anxiety disorder in the United States: results from the National anxiety disorder in the USA: results from the National Epidemi- Epidemiologic Survey on Alcohol and Related Conditions. J Clin ologic Survey on Alcohol and Related Conditions. Psychol Med Psychiatry 2005;66:1351–1361. 2005;35:1747–1759. 4. Boden JM, Fergusson DM, Horwood LJ. Anxiety disorders and 23. Grant BF, Hasin DS, Stinson FS, et al. The epidemiology of DSM- suicidal behaviours in adolescence and young adulthood: findings IV panic disorder and agoraphobia in the United States: results from a longitudinal study. Psychol Med 2007;37:431–440. from the National Epidemiologic Survey on Alcohol and Related 5. Weisberg RB. Overview of generalized anxiety disorder: epidemi- Conditions. J Clin Psychiatry 2006;67:363–374. ology, presentation, and course. J Clin Psychiatry 2009;70(Suppl 24. Stinson FS, Dawson DA, Chou SP, et al. The epidemiology of 2):4–9. DSM-IV specific phobia in the USA: results from the National 6. Bowen RC, Senthilselvan A, Barale A. Physical illness as an out- Epidemiologic Survey on Alcohol and Related Conditions. Psy- come of chronic anxiety disorders. Can J Psychiatry 2000;45:459– chol Med 2007;37:1047–1059. 464. 25. Grant BF, Dawson DA, Stinson FS, et al. The Alcohol Use 7. Yonkers KA, Bruce SE, Dyck IR, Keller MB. Chronicity, relapse, Disorder and Associated Disabilities Interview Schedule-IV and illness-course of panic disorder, social phobia, and general- (AUDADIS-IV): reliability of alcohol consumption, tobacco use, ized anxiety disorder: findings in men and women from 8 years of family history of depression and psychiatric diagnostic modules in follow-up. Depress Anxiety 2003;17:173–179. a general population sample. Drug Alcohol Depend 2003;71:7–16. 8. Skodol AE, Oldham JM, Hyler SE, et al. Patterns of anxiety and 26. Wittchen H-U, Zhao S, Abelson JM, et al. Reliability and pro- personality disorder comorbidity. J Psychiatr Res 1995;29:361– cedural validity of the UM-CIDI DSM-III-R phobic disorders. 374. Psychol Med 1996;26:1169–1177. 9. Ozkan M, Altindag A. Comorbid personality disorders in sub- 27. Ruscio AM, Brown TA, Chiu WT, et al. Social fears and social jects with panic disorder: do personality disorders increase clinical phobia in the USA: results from the national comorbidity survey severity? Compr Psychiatry 2005;46:20–26. replication. Psychol Med 2008;38:15–28. 10. Newton-Howes G, Tyrer P, Anagnostakis K, et al. The preva- 28. Tyrer P, Seivewright H, Johnson T. The Nottingham Study lence of personality disorder, its comorbidity with mental state of Neurotic Disorder: predictors of 12-year outcome of dys- disorders, and its clinical significance in community mental health thymic, panic and generalized anxiety disorder. Psychol Med teams. Soc Psychiatry Psychiatr Epidemiol 2010;45:453–460. 2004;34:1385–1394.

Depression and Anxiety 8 Skodol et al.

29. Hopwood CJ, Malone JC, Ansell EB, et al. Personality assessment 38. Warshaw MG, Massion AO, Shea MT, et al. Predictors of remis- in DSM-5: empirical support for rating severity, style, and traits. sion in patients with panic disorder with and without agoraphobia: J Pers Disord 2011;25:305–320. prospective 5-year follow-up data. J Nerv Ment Dis 1997;185:517– 30. Grant BF, Chou SP, Goldstein RB, et al. Prevalence, correlates, 519. disability, and comorbidity of DSM-IV borderline personality dis- 39. Beard C, Moitra E, Weisberg RB, Keller MB. Characteristics and order: results from the Wave 2 National Epidemiologic Survey on predictors of social phobia course in a longitudinal study of pri- Alcohol and Related Conditions. J Clin Psychiatry 2008;69:533– mary care patients. Depress Anxiety 2010;27:839–845. 545. 40. Keyes K, Hatzenbuehler M, Alberti P, et al. Service utilization 31. Stinson FS, Dawson DA, Goldstein RB, et al. Prevalence, cor- for axis I psychiatric and substance disorders in white and black relates, disability, and comorbidity of DSM-IV narcissistic per- adults. Psychiatr Serv 2008;59:893–901. sonality disorder: results from the Wave 2 National Epidemio- 41. Grant BF, Stinson FS, Dawson DA, et al. Co-occurrence of DSM- logic Survey on Alcohol and Related Conditions. J Clin Psychiatry IV personality disorders in the United States: results from the Na- 2008;69:1033–1045. tional Epidemiologic Survey on Alcohol and Related Conditions. 32. Pulay AJ, Stinson FS, Dawson DA, et al. Prevalence, correlates, Compr Psychiatry 2005;46:1–5. disability, and comorbidity of DSM-IV schizotypal personality 42. Skodol AE, Grilo CM, Keyes KM, et al. Relationship of per- disorder: results from the Wave 2 National Epidemiologic Survey sonality disorders to the course of major depressive disorder in a on Alcohol and Related Conditions. Prim Care Companion J Clin nationally representative sample. Am J Psychiatry 2011;168:257– Psychiatry 2009;11:53–67. 264. 33. Grant BF, Hasin DS, Stinson FS, et al. Co-occurrence of 12- 43. Skodol AE, Bender DS, Morey LC, et al. Personality disor- month mood and anxiety disorders and personality disorders in the der types proposed for DSM-5. J Pers Disord 2011;25:136– US: results from the National Epidemiologic Survey on Alcohol 169. and Related Conditions. J Psychiatr Res 2005;39:1–9. 44. Kendell R, Jablensky A. Distinguishing between the validity and 34. Grant BF, Hasin DS, Stinson FS, et al. Prevalence, correlates, utility of psychiatric diagnoses. Am J Psychiatry 2003;160:4– and disability of personality disorders in the United States: results 12. from the National Epidemiologic Survey on Alcohol and Related 45. Shea MT, Stout RL, Yen S, et al. Associations in the course of Conditions. J Clin Psychiatry 2004;65:948–958. personality disorders and Axis I disorders over time. J Abnorm 35. Ruan WJ, Goldstein RB, Chou SP, et al. The Alcohol Use Psychol 2004;113:499–508. Disorder and Associated Disabilities Interview Schedule-IV 46. Pulay AJ, Dawson DA, Ruan WJ, et al. The relationship of im- (AUDADIS-IV): reliability of new psychiatric diagnostic mod- pairment to personality disorder severity among individuals with ules and risk factors in a general population sample. Drug Alcohol specific axis I disorders: results from the National Epidemio- Depend 2008;92:27–36. logic Survey on Alcohol and Related Conditions. J Pers Disord 36. Zimmerman M. Diagnosing personality disorders: a review of is- 2008;22:405–417. sues and research methods. Arch Gen Psychiatry 1994;51:225– 47. Yang M, Coid J, Tyrer P. Personality pathology recorded by sever- 245. ity: national survey. Br J Psychiatry 2010;197:193–199. 37. Ware JE, Kosinkski M, Turner-Bowker DM, Gandek B. How to 48. Miller JD, Gentile B, Wilson L, Campbell WK. Grandiose and Score Version 2 of the SF-12 Health Survey. Lincoln, RI: Quality vulnerable narcissism and the DSM-5 pathological personality Metric Incorporated; 2002. trait model. J Pers Assess 2013;95:284–290.

Depression and Anxiety