Cumulative probability of remission from generalized anxiety disorder in patients with and without avoidant personality disorder.
Cumulative probability of remission from generalized anxiety disorder in patients with and without dependent personality disorder.
Cumulative probability of remission from social phobia in patients with and without avoidant personality disorder.
Massion AO, Dyck IR, Shea MT, Phillips KA, Warshaw MG, Keller MB. Personality Disorders and Time to Remission in Generalized Anxiety Disorder, Social Phobia, and Panic Disorder. Arch Gen Psychiatry. 2002;59(5):434-440. doi:10.1001/archpsyc.59.5.434
Copyright 2002 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2002
This investigation assessed the effect of personality disorders (PersDs) on time to remission in patients with generalized anxiety disorder, social phobia, or panic disorder.
Selected Axis I and II predictors of time to remission during 5 years of follow-up were assessed in 514 patients with 1 or more of these anxiety disorders who participated in the Harvard/Brown Anxiety Research Program, a multisite, prospective, longitudinal, naturalistic study.
The presence of a PersD predicted a 30% lower likelihood of generalized anxiety disorder remission, a 39% lower likelihood of social phobia remission, and no difference in likelihood of panic disorder remission. More specifically, a lower likelihood of remission from generalized anxiety disorder was predicted by the presence of avoidant PersD (34% lower) and dependent PersD (14% lower). The presence of avoidant PersD predicted a 41% lower likelihood of social phobia remission. The presence of major depressive disorder did not account for these findings.
Our findings provide new data on the pernicious effect of PersDs on the course of generalized anxiety disorder and social phobia but not panic disorder, suggesting that PersDs have a differential effect on the outcome of anxiety disorders.
STUDIES EXAMINING the effect of comorbid personality disorders (PersDs) on the outcome of Axis I disorders generally have found that the presence of a PersD is associated with worse outcome of Axis I disorders,1,2 particularly for depression.3 However, to our knowledge, few prospective, naturalistic, longitudinal studies have assessed the effect of Axis II disorders on the course of panic disorder, only 1 study4 has assessed the effect on generalized anxiety disorder(GAD), and none have assessed the effect on social phobia (SP). The need for such studies has been emphasized in the literature.5,6
Three prospective, longitudinal (1-3 years), naturalistic studies of panic disorder7- 10 found that the presence of a PersD was associated with worse outcome of panic disorder symptoms and functioning; 1 study found an association with a lower likelihood of remission.9,10 These studies are limited by use of self-report PersD instruments,7,9,10 only 1 follow-up assessment,8 follow-up of only1 year,7,9,10 and retrospective recall.8 Only 1 study9,10 used DSM-III-R criteria for the PersDs, although with a self-report instrument.
Regarding GAD, Yonkers et al4 reported that the presence of cluster B or C PersDs predicted a lower likelihood of remission. The only other longer-term, larger-sample studies on PersD and GAD or SP were treatment studies,11- 13 both of which showed that PersDs had a negative effect on outcome; specifically, avoidant PersD predicted a worse outcome for SP.12,13
The present study used a prospective, naturalistic design to assess the effect of PersDs on the outcome of anxiety disorders, which has not been addressed in our previous studies.4,14- 18 Unlike other studies, this study has the advantages of longer follow-up (5 years), a larger sample (n = 514), assessment of several anxiety disorders, and use of multiple assessments over time with standard structured or semistructured instruments. Also, we accounted for the effect of major depression and analyzed time to remission separately for GAD, SP, and panic disorder, enabling us to determine differential course outcomes. Because panic disorder has a more episodic course than does GAD or SP, we hypothesized that the effect of a comorbid PersD would be less.
The Harvard/Brown Anxiety Research Program is a prospective, naturalistic, longitudinal study of patients with DSM-III-R–defined anxiety disorders.14 Participants had at least1 past or current episode of 1 or more of the following disorders: (1) panic disorder with agoraphobia (PDA) or without agoraphobia (PD), (2) agoraphobia without a history of panic disorder, (3) GAD, or (4) SP. Individuals with agoraphobia without a history of panic disorder were omitted owing to the small number of patients (n = 8). No attempt was made to affect the treatment received. Participants were recruited from and interviewed at 1 of 10 Boston area, central Massachusetts, or Rhode Island hospitals or clinics from the public and private sector (a complete list of these sites appears at the end of this article). Exclusion criteria were age younger than 18 years, organic mental disorder, schizophrenia, or psychosis within 6 months before study enrollment.
From January 1, 1989, to November 1, 1991, 711 individuals were enrolled, all of whom gave written informed consent. This article reports on the 514 individuals who had (1) a diagnosis of GAD, SP, PD, and/or PDA at study enrollment and (2) a PersD assessment and course data available for at least 6 months to 5 years (Table 1). Of the 514 participants, 232 had follow-up data for all 5 years and 282 had mean (SD) follow-up of 83 (66) weeks (range, 1-251 weeks). These participants have less than 5 years of follow-up data because of attrition or the fact that assessments were not yet due at the time of data analysis. There was no significant difference in PersD rates among participants with less than 5 years vs 5 years of follow-up data (χ21 = 1.21; P = .27).
Analyses examining the predictive effect of PersDs on course were done for the following current-at-enrollment diagnostic groups: 155 patients with GAD, 141 with SP (70 with specific and 71 with generalized SP), 73 with PD, and 313 with PDA. Except for PD and PDA, participants may have had 1, 2, or all 3 of these anxiety disorders. Of participants with GAD, 51 (33%) had comorbid SP, 16 (10%) had PD, and 65 (42%) had PDA. Of participants with SP, 50 (36%) had comorbid GAD, 14 (10%) had PD, and 50 (36%) had PDA. Of participants with PD, 16 (22%) had comorbid GAD and 12 (16%) had SP. Of participants with PDA, 65 (21%) had comorbid GAD and 52 (17%) had SP.
Axis I disorders were diagnosed at study enrollment using the SCALUP,19 which consists of items from the Structured Clinical Interview for the DSM-III-R, Patient Version (Nonaffective Disorders Section and Psychosis Screen)20 and the Research Diagnostic Criteria Schedule for Affective Disorders–Lifetime.21,22 The Global Assessment Scale also was completed at enrollment.23 Follow-up interviews were conducted at 6-month intervals for 3 years and then annually using the Longitudinal Interval Follow-up Evaluation,24,25 which assesses information on course of illness, psychopharmacological treatment received, and psychosocial functioning. The Longitudinal Interval Follow-up Evaluation rates severity of psychopathologic symptoms over time using 6-point Psychiatric Status Rating (PSR) scales for each disorder based on DSM-III-R criteria; these are scored on a week-by-week basis during the interview period (6 or 12 months). A score of 5 or 6 indicates that full criteria for the disorder are currently met (a score of 6 indicates more severe symptoms than a score of 5); a score of 3 or 4 denotes subthreshold symptoms; and a score of 1 or 2 denotes remission.
Remission was defined as occasional mild symptoms or no symptoms (PSR of 2 or 1, respectively) for 8 consecutive weeks after the episode of illness at enrollment, which is the definition used in the National Institute of Mental Health Collaborative Depression Study.26 For participants with PDA, remission was defined as a PSR of 2 or less for either panic disorder or agoraphobia for at least 8 consecutive weeks. Probability of first remission was analyzed for participants who met criteria for a current episode of the index disorder at enrollment.
Receipt of psychosocial treatments was assessed every 6 months for 2 years and annually thereafter using the Psychosocial Treatments Interview,27 a reliable and valid instrument.
All assessments were conducted by experienced interviewers with 3 to6 months of training in the study instruments. Most had a bachelor of science or bachelor of arts degree, several had a master of arts degree, and 2 had a doctor of medicine degree. To monitor interrater reliability, raters independently rated 1 videotaped interview per month. Three substudies assessed interrater reliability, participant recall, and validity of the Longitudinal Interval Follow-up Evaluation PSR score.25 The PSR scores had very good reliability,25 with test-retest intraclass correlations ranging from 0.62 to 0.88 for PD, 0.49 to 0.98 for agoraphobia, 0.78 to 1.00 for GAD, and 0.75 to 0.89 for SP.25
Personality disorders were assessed at the 1-year follow-up interview using an earlier version of the International Personality Disorder Examination, a reliable semistructured clinical interview based on DSM-III-R criteria,28,29 which differs from the later expanded version only in its lack of International Classification of Diseases, 10th Revision, diagnoses (referred to herein as PDE). This instrument assesses characteristics typical throughout the participant's life and uses scoring algorithms to determine the presence or absence of specific PersDs.
Rates of paranoid, sadistic, and antisocial PersDs were expected to be low and therefore were excluded from the PDE to reduce participant fatigue. Participants with a probable PersD diagnosis were combined with those with a definite diagnosis because rates of PersDs were low.
Training for use of the PDE was conducted by its developer, Armand W. Loranger, PhD, and included studying case examples, watching and rating tapes, and conducting mock interviews. Raters were supervised by a trained rater during the first 3 PDE assessments with study participants.
We examined the following as predictors of first remission for GAD, SP, and PDA: any PersD and avoidant, passive-aggressive, obsessive-compulsive, histrionic, dependent, narcissistic, and borderline PersDs. We do not report on PersD clusters because cluster A and B disorders were excluded because rates of schizoid and schizotypal PersDs (cluster A) were low. We tabulated but did not analyze (as course predictors) schizoid, schizotypal, narcissistic, borderline, and self-defeating PersDs because rates were low. Similarly, because of the low rates, we did not analyze passive-aggressive or histrionic PersDs for participants with SP, PDA, or both. To investigate the overlap between generalized SP and avoidant PersD, we (1) separately analyzed generalized and specific SP for predictors of remission and (2) further analyzed predictors of remission for participants with generalized SP with and without avoidant PersD. Also, we examined the effect of comorbid major depression at enrollment on the relationship between PersDs and course of the anxiety disorders because previous data30 suggested that the presence of major depression is associated with a more chronic course of anxiety disorders. Specifically, we examined the effect of major depression at enrollment on overall remission and when occurring in the week before remission from the anxiety disorder. Last, we separately analyzed the effect of past or current substance abuse or dependence on remission.
Statistical analyses were conducted using a software program (SAS Version6.07; SAS Institute Inc, Cary, NC) and PROC FREQ, PROC PHREG, PROC TTEST, and PROC LIFETEST. We used an α cutoff level of .01, or 95% confidence intervals (CIs), or both. To examine whether PersD variables or major depressive disorder predicted the course of GAD, SP, or PD/PDA, we analyzed for individual predictors of time to remission (which includes probability of remission) over time using Kaplan-Meier survival curves. Kaplan-Meier life tables were constructed for PersDs and substance abuse disorders to determine whether the probability of remission over time was significantly different for presence vs absence of the disorder. For comparison of demographics and comorbidity between groups, χ2 tests were used to examine categorical data, and the Fisher exact test was used when expected cell counts were small. Noncomorbid anxiety disorders were not compared because of the relatively large numbers of participants who had comorbid disorders.
To determine whether participants with and without a PersD received different treatment, we assessed the percentage of participants receiving1 or more of the following: any psychotropic medication, selective serotonin reuptake inhibitors, monoamine oxidase inhibitors, tricyclic antidepressants, benzodiazepines, atypical antidepressants, any psychosocial treatment, relaxation training, 1 or more cognitive techniques, 1 or more behavioral techniques, or psychodynamic therapy.
Table 1 gives the frequency of comorbid PersDs.31 There were no across-the-board differences in the rates of treatment for participants with and without PersDs(Table 2, including baseline Global Assessment Scale scores).
The probability of remission/number of first remissions from the anxiety disorders at 5 years was 0.38/53 for GAD, 0.29/38 for SP, 0.35/102 for PDA, and 0.65/45 for PD. Regarding remission from GAD, participants with a PersD had a lower probability of remission at 5 years (probability, 0.18; 95% CI, 0.07-0.29) than those with no PersD (probability, 0.50; 95% CI, 0.39-0.61; log-rank test, 12.52; P<.001). The presence of2 specific PersDs, avoidant or dependent, primarily explained this finding(Figure 1 and Figure 2). No other PersD had a statistically significant effect on remission of GAD.
Participants with a PersD also were significantly less likely to remit from SP than were those with no PersD (probability, 0.39; CI, 0.15-0.91; P = .03). In particular, participants with avoidant PersD were significantly less likely to remit from SP (Figure 3). Furthermore, the presence of avoidant PersD predicted remission of generalized (log-rank χ21, 4.05; P = .04) but not specific (log-rank χ21, 1.88; P = .17) SP. None of the other PersDs had a significant effect.
The presence of a PersD did not have a differential effect on probability of remission from PD or PDA.0
The presence of major depression at enrollment was not a significant predictor of partial (PSR of 3 or 4) or full (PSR of 1 or 2) remission of PD, SP, or GAD. However, the presence of major depression during the week before remission resulted in a lower probability of remission from PDA (risk ratio, 0.88; 95% CI, 0.79-0.99; χ2 = 4.42; P = .04) but not from SP (risk ratio, 1.11; 95% CI, 0.93-1.33; χ2 = 1.42; P = .23). There was only a trend for lower probability of GAD remission (risk ratio, 0.80; 95% CI, 0.64-1.01; χ2 = 3.49; P = .06). Panic disorder without agoraphobia was not analyzed because there were too few patients in that group.
The presence of past or current alcohol abuse or dependence was associated with a lower probability of SP remission (present: probability, 0.84; 95% CI, 0.74-0.94, 8/54 remitted vs absent: probability, 0.6; 95% CI, 0.53-0.74,30/87 remitted; log-rank χ2, 6.09; P= .01) but not GAD or PD/PDA remission.
This is the first study, to our knowledge, of a large sample of patients with multiple assessments and prospective longitudinal data for up to 5 years of follow-up that examines the effect of PersDs on probability of remission from 3 anxiety disorders. The presence of 1 or more PersDs was a significant predictor of lower probability of remission of GAD and SP but not PD/PDA, taking into account the effect of major depression. Regarding specific PersDs, avoidant PersD predicted a lower likelihood of remitting from GAD and SP but not from PD/PDA, and dependent PersD predicted a lower likelihood of remitting from GAD.
The finding that avoidant PersD predicts a lower likelihood of remitting from SP is consistent with the overlap in the diagnostic criteria for these disorders and with previous reports that individuals with comorbid SP and avoidant PersD tend to have more symptoms and avoidance than those with SP without avoidant PersD.32 In addition, because avoidant PersD and generalized SP tend to be chronic, the 2 would not be easily distinguished by their course.33 The finding that avoidant PersD also predicted a lower probability of GAD remission was unexpected, as this has not been previously investigated. Because 12 of the31 individuals with GAD and avoidant PersD did not have comorbid SP, and only1 of these had a later onset of SP (between the enrollment and PDE assessments), this finding cannot be completely explained by the presence of comorbid SP. This supports a distinction between SP and avoidant PersD, supported by the DSM-IV criteria that most people with generalized SP also will have avoidant PersD, but not necessarily the reverse.
Our findings are inconsistent with those of previously cited naturalistic studies7- 10 of PD/PDA, which showed that the presence of a PersD predicted a worse outcome for PD/PDA. A possible explanation may be that 2 of the studies7,9,10 used self-report PersD instruments, which overdiagnose PersDs compared with semistructured interview instruments.34 The other study8 had only a single follow-up interview after 3 years, which involved retrospective recall. Also, because these patients originally had participated in a treatment study, they might not be broadly representative of patients with PD/PDA.
Because the rate of lifetime alcohol abuse or dependence was low for participants with SP (12 of 141 individuals), we could not assess whether alcohol abuse or dependence accounted for the predictive effect of avoidant PersD on remission of SP. The rates of all substance abuse or dependence disorders in this sample were low, thus limiting our ability to assess the predictive effect.
The literature on the effect of depression on anxiety disorder outcome is mixed,8,9,35 with1 review36 reporting that 3 of 5 studies showed that comorbid major depression did not affect panic disorder outcome. Our findings on the effect of comorbid major depression are similar to those of2 previously cited studies9,35 (ie, that comorbid depression was not associated with outcome of panic disorder at 1-year follow-up). However, we did find that the presence of major depression in the week before remission predicted a lower probability of remission of PDA but not GAD or SP. Possibly we were not able to detect an effect for GAD or SP because there was such a low rate of remissions for these two; after10 years of follow-up and more remissions, we may be able to determine the presence or absence of such an effect.
We found that other comorbid individual PersDs, such as borderline PersD, did not predict a worse outcome for any anxiety disorder. This finding, which previously has not been investigated, is contrary to what clinical wisdom might suggest. However, because few participants had certain individual PersDs, this finding may reflect type II error.
The strengths of this study include a large sample size; a prospective, naturalistic, longitudinal design with multiple assessments over time; the use of structured clinical interviews to diagnose both DSM-III-R Axis I and II disorders; detailed course information over5 years of follow-up; and assessment of the predictive effect of individual PersDs on the course of anxiety disorders. Because this was not a treatment study that excluded comorbid conditions, our sample may be more representative of patients treated in clinical settings. Therefore, we cannot draw conclusions about the course of these disorders in a nonclinical setting. Another limitation is that our participants had a high rate of comorbid anxiety disorders, making statistical comparisons of remission rates for noncomorbid anxiety disorders unfeasible. Last, the PDE was given 1 year after enrollment, and a state effect(eg, remission at 1 year from an Axis I disorder) could have affected the results.
Despite these limitations, the strength of the association between the presence of a PersD and longer time to remission from GAD and SP but not PD/PDA, and the findings on specific PersDs, suggest that there is a differential effect of PersDs on the outcomes of anxiety disorders. Future studies are needed to confirm these findings and to address the limitations of this study. Because anxiety disorders and PersDs are relatively common conditions that are associated with considerable morbidity, this clinically important issue deserves further investigation.
Submitted for publication March 3, 1999; final revision received August8, 2001; accepted September 4, 2001.
Dr Shea has received honorarium from Organon, West Orange, NJ. Dr Phillips has a significant interest that derives from her service to the following entities: Eli Lilly and Co, Indianapolis, Ind; Solvay Pharmaceuticals, Marietta, Ga; Forest Laboratories, Inc, St Louis, Mo; Gate Pharmaceuticals, North Wales, Pa; Wyeth-Ayerst Pharmaceuticals, Philadelphia, Pa; Bristol-Myers Squibb, Co, Wallingford, Conn; Glaxo-Wellcome Inc, Research Triangle Park, NC; and SmithKline Beecham Pharmaceuticals, Philadelphia. Dr Keller has a significant interest that derives from his service to the following entities: Pfizer, Inc, New York, NY; Bristol-Myers Squibb, Co; Forest Laboratories, Inc; Wyeth-Ayerst Pharmaceuticals; Merck & Co, Inc, Whitehouse Station, NJ; Janssen Pharmaceutica, Titusville, NJ; Eli Lilly and Co; Organon; Pharmacia/Upjohn Inc, Kalamazoo, Mich; SmithKline Beecham Pharmaceuticals; Zeneca Pharmaceuticals, Wilmington, Del; Mitsubishi Pharmaceuticals, Tokyo, Japan; Scirex, Philadelphia; Vela Pharmaceuticals Inc, Princeton, NJ; Sepracor Pharmaceuticals, Marlborough, Mass; Somerset Pharmaceuticals, Inc, Tampa, Fla; and Sanofi-Synthelabo, New York.
This study was supported in part by Wyeth-Ayerst Pharmaceuticals through its Global Research Program on Anxiety Disorders; grant MH 51415 from the National Institute of Mental Health, Bethesda, Md; and the Department of Psychiatry and Human Behavior, Brown University.
We thank the Quintiles Corporation, Research Triangle Park, NC, for providing valuable consultation services to this project.
Martin B. Keller, MD (Chairperson), Providence, RI; M. Tracie Shea, PhD, Providence (Veterans Administration Hospital, Brown University School of Medicine); Jane Eisen, MD, Providence; Katharine A. Phillips, MD, Providence; Robert Stout, PhD, Providence (Butler Hospital, Brown University School of Medicine); Steven E. Bruce, PhD, Providence; Risa B. Weisberg, PhD, Providence (Brown University School of Medicine); Meredith G. Warshaw, MSS, MA, Boston, Mass (Massachusetts Veterans Epidemiology Research and Information Center, Boston VA Healthcare System, and Department of Epidemiology and Biostatistics, Boston University School of Public Health); Robert M. Goisman, MD, Boston(Massachusetts Mental Health Center, Harvard University School of Medicine); Ann O. Massion, MD, Worcester, Mass (University of Massachusetts Medical School, Worcester); Malcom P. Rogers, MD, Boston (Brigham and Women's Hospital, Harvard University School of Medicine); Carl Salzman, MD, Boston (Massachusetts Mental Health Center); Gail Steketee, PhD, Boston (Boston University School of Social Work); Kimberly Yonkers, MD, New Haven, Conn (Yale University School of Medicine); Idell Goldenberg, PsyD, Medway, Mass; Gopi Mallya, MD, Belmont, Mass (McLean Hospital, Harvard University School of Medicine); Timothy Mueller, MD, Providence(Butler Hospital, Brown University School of Medicine); Fernando Rodriguez-Villa, MD, Belmont (McLean Hospital, Harvard University School of Medicine); Russell Vasile, MD, Boston (New England Deaconess Hospital, Harvard University School of Medicine); Caron Zlotnick, PhD, Providence (Butler Hospital, Brown University School of Medicine); Eugene Fierman, MD, Brookline, Mass.
Paul Alexander, MD, Providence (Butler Hospital, Brown University School of Medicine); James Curran, MD, Providence; Jonathan Cole, MD, Belmont (McLean Hospital, Harvard University School of Medicine); James Ellison, MD, MPH, Somerville, Mass (Harvard Pilgrim Health Care, Harvard University School of Medicine); Alan Gordon, MD, Providence (Butler Hospital, Brown University School of Medicine); Steven Rasmussen, MD, Providence (Butler Hospital, Brown University School of Medicine); Robert Hirschfeld, PhD, Galveston, Tex (University of Texas, Galveston); Jill Hooley, DPhil, Cambridge, Mass (Harvard University); Philip Lavori, PhD, Stanford, Calif (Stanford University); John C. Perry, MD, Montreal, Quebec (Jewish General Hospital, McGill University School of Medicine, Montreal); Linda G. Peterson, Rockport, Me (Midcoast Medicine, Rockport); James Reich, MD, MPH, Brockton, Mass (Renard Hospital–Washington University School of Medicine); John Rice, PhD, St Louis, Mo (Renard Hospital–Washington University School of Medicine); Harriet Samuelson, MA, Boston (Brigham and Women's Hospital); David Shera, MS, Somerville (Harvard School of Public Health); Naomi Weinshenker, MD, Newark, NJ (New Jersey Medical School); Myrna Weissman, PhD, New York, NY (Columbia University); Kerrin White, MD, Seattle, Wash.
McLean Hospital, Belmont; Brown Medical School (Rhode Island and Butler Hospitals), Providence; University of Massachusetts Medical School, Worcester; Cambridge City Hospital, Cambridge; Massachusetts Mental Health Center, Boston; Brighton/Allston Mental Health Center, Brighton, Mass; Brigham and Women's Hospital, Boston; Worcester VA Medical Clinic, Worcester; Beth Israel Hospital, Boston; Deaconess Hospital, Boston.
Corresponding author and reprints: Ann O. Massion, MD, Department of Psychiatry, University of Massachusetts Medical School, 55 Lake Ave N, Worcester, MA 01655 (e-mail: email@example.com).