Course of social functioning after remission from panic disorder

COMPREHENSIVE PSYCHIATRY (Official Journal of the American Psychopathological Association)

VOL. 42, NO. 6

NOVEMBER/DECEMBER 2001

Course of Social Functioning After Remission From Panic Disorder Robert L. Stout, Regina Dolan, Ingrid Dyck, Jane Eisen, and Martin B. Keller While symptom status and social functioning have been observed to be correlated in many cross-sectional studies, little is known about the time course of change in functioning after a change in diagnostic status. Using data from a large longitudinal study of anxiety disorders, we present analyses of the time course of seven domains of social functioning up to 18 months before and after remission from panic disorder with or without agoraphobia. The effect of remission from panic disorder varies by domain. Four domains show a change in outcome, usually positive, after remission. The presence of major depressive dis-

order (MDD) affects the course of functioning for two domains. Generalized anxiety disorder (GAD) was observed to have effects on five of seven domains. For some domains there is improvement at approximately the same time as change in diagnostic status, although progressive change was seen in others. The amount of improvement was modest on average, indicating that other factors beyond panic symptoms may limit post-remission improvement in social functioning. Copyright © 2001 by W.B. Saunders Company

N

years, 60% had been out of episode (of anxiety disorders) for at least 6 months. However, 19% of the “recovered” group continued to have moderate impairment at work, only 3% reported improved relationships with friends, and 30% of these subjects received global social adjustment ratings from their therapist of moderate to poor.7 On the other hand, at least for some disorders and for some psychosocial functioning variables, improvement in symptomatic status is observed to correlate with improvement in functioning.7-10 These varied results have been obtained with differing populations, disorders, measures of functioning, treatment status, and time frames, making generalization difficult. While changes in symptom status and functioning are at least sometimes correlated over time,

EARLY 30 YEARS of research has demonstrated cross-sectional relationships between psychiatric symptom status and psychosocial functioning (i.e., social and role functioning).1-5 Indeed, in clinical and research settings it is now taken for granted that individuals with psychiatric diagnoses (or, in some cases, symptoms without diagnoses) have worse functioning than do individuals free from such distress. However, we know relatively little about how these variables relate to one another over time. Furthermore, cross-sectional studies cannot determine the extent to which poor functioning is due to symptom status per se, versus an exogenous process that affects both symptoms and functional status. The existing literature on symptom status and functioning over time is inconsistent. Some research suggests that even after periods of remission, functioning can remain impaired for some patients.6,7 Of 240 subjects with unipolar depression who were followed for 5 years, 46% had no major depression for the last 2 years of follow-up; however, their impairment in terms of occupation, sexual functioning, recreation, and overall satisfaction continued to be similar to that of the total group of depressed subjects.6 In a study of 53 subjects with anxiety disorders followed for 2

From the Department of Psychiatry and Human Behavior, Brown University, Providence, RI. Supported in part by the Upjohn Co, and by NIMH Grant No. MH51415. Address reprint requests to Robert L. Stout, Ph.D., Decision Sciences Institute, 120 Wayland Ave, Suite 7, Providence, RI 02906-4318. Copyright © 2001 by W.B. Saunders Company 0010-440X/01/4206-0003$35.00/0 doi:10.1053/comp.2001.27894

Comprehensive Psychiatry, Vol. 42, No. 6 (November/December), 2001: pp 441-447

441

442

STOUT ET AL

there has been little or no research on the specific relationship over time between these two aspects of mental health. A number of basic questions have never been specifically addressed, including: Does functioning change before symptom status changes, afterwards, or roughly in synchrony? Does functioning recover gradually or quickly? How long does it take before social functioning recovers to approximately normal levels? The purpose of this report is to begin to answer some of these questions with regard to the time course of social functioning after remission from panic disorder. Using a database from an ongoing study in which most of the participants have been followed for 5 years or more, we will characterize the time course of changes in several areas of functioning relative to changes in diagnostic status. METHODS The data for this study were derived from the Harvard/Brown Anxiety Disorders Research Project (HARP). This is a prospective, naturalistic, longitudinal study of adults with present or past anxiety disorders. The sample consists of 711 subjects from 11 sites. Inclusion criteria are that patients have at least one of the following DSM-III-R current or past diagnoses (designated target disorders): panic disorder (with or without agoraphobia), agoraphobia without a history of panic disorder, generalized anxiety disorder (GAD), or social phobia (SP); however, in the analyses reported below we include only those subjects exhibiting panic disorder (with or without agoraphobia). Subjects had to have been at least 18 years of age, willing to participate voluntarily in the study, and willing to sign a written consent form. Exclusion criteria were the presence of an organic brain syndrome, a history of schizophrenia, or psychosis at the time of intake. The initial comprehensive evaluation assesses lifetime history using selected items from the Personal History of Depressive Disorders11; the Structured Clinical Interview for the DSMIII-R Non-Affective Disorders-Patient Version (SCID-P)12; and the Research Diagnostic Criteria (RDC) Schedule for Affective Disorders-Lifetime (SADS-L).13 Items of the SCID-P and SADS-L were combined to create the SCALUP, a structured interview used to assess diagnoses at intake.14 Other instruments, not relevant to the current report, were also given. Follow-up evaluations were initially conducted at 6-month intervals over the course of 2 years but, as the overall follow-up period lengthened to 3 years, evaluations subsequent to 12 months were done at annual intervals. Diagnostic follow-up data are gathered using the Longitudinal Interval Follow-Up Evaluation-Upjohn (LIFE-UP)14 to collect detailed information on the course of illness and psychosocial functioning. Other follow-up instruments are also assessed, but are not relevant to the current analyses. The follow-up data used here were for each participant’s first 5 years. A number of studies have been done to assess inter-rater reliability, subject recall, and validity of the LIFE-UP Psychiatric Status Ratings (PSRs), used to assess the course of all

disorders.15 The results of these studies have indicated good to excellent inter-rater reliability, acceptable to excellent 1-year retest reliability for all index disorders and for major depressive disorder (MDD), as well as good concurrent and discriminant validity. All intake interviews were conducted in person, while most follow-up interviews are conducted via telephone, by experienced clinical interviewers whose training and certification ranges from B.A. to M.D. Longitudinal reliability is assured by a rigorous program of training and testing with tapes of early interviews. Raters must sustain a median intraclass correlation coefficient of 0.80 for the most common disorders and for the Global Assessment Scale (GAS) scores or else must retrain. The HARP study protocol has been described in more detail by Massion et al.,16 and results from the HARP study on the course of panic disorder have been reported previously.17-19

Statistical Methods The symptomatic status of HARP subjects is tracked by PSRs.14 PSRs are recorded on a weekly basis for each disorder. PSRs are a rating on a scale from 1 (no symptoms) to 6 (one panic attack or more per day). A participant being followed for panic disorder was judged to be “remitted” for the purposes of this study if he/she experienced 8 successive weeks at PSR ratings no higher than 2—a rating of 2 indicates that no panic attacks have occurred, although the subject may have felt on the verge of such an attack. This definition of “remission” has been widely used in studies of affective and other disorders.20-22 Because there is some controversy about the length of time subjects would need to have low scores to qualify for remission, we also repeated our analyses using a criterion of 6 months of PSR ratings no higher than 2. Our results from these analyses were not materially different from the results we obtained using the 8-week criterion; hence, for brevity we will not reproduce the results from the 6-month criterion runs here. Subjects who did not remit, or whose remissions occurred at times when we had less than 3 months of either pre- or postremission functioning data (see below), were not used in our analyses. Monthly functioning ratings were rated using the LIFE-UP. We examined seven of these ratings: (1) GAS,23 (2) Rater Assessment of Impairment, (3) Patient Satisfaction, (4) Work Impairment, (5) Household Impairment, (6) Marital Impairment, and (7) Relationships With Friends. The GAS and Rater Assessment of Impairment both are general measures; however, the Rater Assessment is designed specifically to be independent of symptom status, whereas in making GAS ratings, symptom status is given consideration. For GAS ratings, a higher score is associated with better functioning. For all other measures, a higher score is associated with higher impairment.

Determining Time Points For those participants who had a remission from a given anxiety disorder during our period of observation, we designated the month the remission occurred for that individual as month 0. Functioning ratings were then assembled relative to that time point, up to 18 months prior to the remission and 18 months afterward. We did not have data for all subjects for as much as 18 months in either direction because of the beginning of follow-up, end of follow-up, relapse to the anxiety disorder being examined, or missing data. Because the focus of our

COURSE OF FUNCTIONING AFTER PANIC REMISSION

analysis was on change in functioning before and after remission, we also required that subjects have at least 3 months of nonmissing functioning data immediately before, and immediately after, month 0. We used latent growth analysis (random effects regression) to determine the impact of remission on outcome. In latent growth analysis, the time course of recovery for individual cases is the focus of the analysis. For a general introduction to latent growth analysis, see Bryk and Raudenbush.24 In these analyses, modeling is done at two levels. In the first level, latent growth curves are fit to the available data on outcome over time for each subject. Different subjects may have data for different sets of time points; however, the more nonmissing time points are available for a subject, the more weight that subject receives in the analysis. At the second level of the model, the parameters of the latent growth curves are related to predictor variables such as diagnostic status. We modeled the functioning over time for each subject as an interrupted time series,25 with the interruption occurring between months 0 and 1. The models employed four random effects: (1) intercept (the subject’s estimated level at time 0), (2) secular linear trend (linear change in functioning level over time regardless of whether before or after remission), (3) remission level change (a shift in the mean level starting in month 1), and (4) remission slope change (a change in the slope of the functioning curve over time starting in month 1). The last two random effects were used to test the hypothesis that there were functioning changes postremission. These postremission changes could take the form of either a change in level, or a progressive change over time (slope), or both. The secular linear trend term is necessary to rule out the possibility that apparent postremission change might have begun well before remission. There were no constraints on the covariation across time points. To take comorbidity and other factors into account, we covaried gender and the presence of major depression (whether at baseline or during follow-up), GAD, and SP. Our subjects were a mixture of cases with simple panic and those with agoraphobia. In preliminary analyses, however, we were unable to detect any effect of the presence or absence of agoraphobia on our results; therefore, this variable was not included as a covariate in the analyses below. Because major depression is known to have a strong effect on functioning,5 we also tested for whether major depression affected not only level of functioning but also the secular linear trend, the remission level change, and the remission slope change. We also considered taking treatment

443

effects into account in our analyses. In naturalistic studies it is difficult to ascertain treatment effects unambiguously; indeed, because of well-known biases for the worst-prognosis patients to receive the most treatment,26 treatment effects often appear to be negative. We therefore elected not to include treatment terms in our analyses.

RESULTS

Of the 711 HARP study subjects, 467 had at least one episode of panic with or without agoraphobia during the first 5 years of follow-up, of whom 153 had a remission from panic during that period. Of the 153 subjects who exhibited a remission, 98 had nonmissing monthly functioning data for a minimum of 3 months before and 3 months after the month of remission. Not all of these cases had nonmissing functioning data for all roles, however. The sample size for each role analyzed was: GAS, 98; Rater Assessment of Functioning, 76; Patient Satisfaction, 76; Employment Functioning, 57; Household Functioning, 75; Marital Functioning, 47; and Relationships with Friends, 76. Of the 98 subjects who were in the functioning analyses, 65 (66%) were female, 24 (24%) met criteria for GAD at intake (24%), 20 (20%) met criteria for SP, and 30 (31%) met criteria for MDD. Latent Growth Analysis Results Table 1 contains the results of the random effects analyses of postremission change in functioning. Of the seven functioning measures studied, three showed statistically significant changes in mean level postremission: GAS, Rater Assessment of Impairment, and Employment Impairment. All of these measures show an improvement. For three of the measures—Rater Assessment, Employment, and Marital Impairment—there was also a statistically significant change in linear trend postremis-

Table 1. Random Effects by Dependent Measure Effect Linear Time Trend, Overall

Mean Change Postremission

Change in Linear Trend Postremission

Variable

Coefficient

P

Coefficient

P

Coefficient

P

GAS Rater Assessment Patient Satisfaction Employment Household Marital Relations Friend Relations

⫹.021 ⫺.030 ⫺.021 ⫺.014 ⫺.008 ⫺.025 ⫹.012

.8791 .0003 .0376 .1705 .4112 .0379 .2468

⫹4.47 ⫺.256 ⫺.004 ⫺.407 ⫺.158 ⫺.147 ⫺.018

.0034 .0014 .9673 .0003 .0801 .1333 .8260

⫹.098 ⫹.029 ⫹.011 ⫹.035 ⫹.013 ⫹.045 ⫺.009

.6067 .0074 .4033 .0113 .3379 .0046 .5063

444

STOUT ET AL

Table 2. Effects of Major Depression by Dependent Measure Overall Effects Intercept

Post-Panic-Remission Effects

Linear Trend

Change in Mean

Change in Linear Trend

Variable

Coefficient

P

Coefficient

P

Coefficient

P

Coefficient

P

GAS Rater Assessment Patient Satisfaction Employment Household Marital Relations Friend Relations

⫺6.41 ⫹.567 ⫹.439 ⫹.803 ⫹.290 ⫺.078 ⫹.508

.0309 .0012 .0495 .0024 .1449 .8118 .0231

⫺.000 ⫺.046 ⫺.039 ⫺.027 ⫺.032 ⫺.052 ⫺.000

.9888 .0080 .0518 .1742 .1152 .0285 .9961

⫺.313 ⫺.201 ⫹.199 ⫺.408 ⫺.098 ⫺.009 ⫹.126

.9163 .2001 .2785 .0620 .5848 .9617 .4492

⫹.113 ⫹.057 ⫹.038 ⫹.032 ⫹.036 ⫹.070 ⫺.012

.7655 .0087 .1619 .2452 .1865 .0235 .6653

sion suggestive of a longer-term change process. Patient Satisfaction exhibits an overall linear trend toward improvement. For two measures, Household Impairment and Relationships with Friends, there were no statistically significant relationships between the onset of remission and functioning status. Table 2 contains the results of tests for the impact of MDD on change over time relative to remission from panic. The presence of MDD as expected has a generally negative effect on functioning for five of the seven measures. MDD did not have statistically significant effects on the change in functioning occurring post-remission. For two measures, however, Rater Assessment and Marital Impairment, MDD was found to affect the linear trend in the time course of functioning both before and after panic remission. Of the other covariates tested, gender and SP were not found to have significant effects on any of the functioning measures. GAD, however, was found to have negative effects on the level of six of the seven measures, Marital Impairment being the only exception. Figures 1 to 5 show the time course for each functioning rating relative to month of remission from panic disorder. Because MDD was found to have an effect on the time course for Rater Assessment and Marital Impairment, in these figures separate curves are shown for cases with versus without comorbid MDD. The data in these figures are group average data, and do not necessarily represent the course of recovery for individual cases. Nonetheless, we believe they accurately show the general trends detected in the data by the latent growth analyses. For the two general measures of functioning, GAS and Rater Assessment (Figs 1 and 2), there is

a statistically significant pre-/postremission improvement in functioning. For the Rater Assessment measure, improvement seems to begin prior to remission, as evidenced by the statistically significant secular linear trend. On the GAS scale, improvement may start a few months prior to remission, although the data are noisy. For those subjects who do not have MDD, Rater Assessment scores seem to continue to improve after remission; however, those subjects who have MDD do not seem to exhibit this gradual improvement trend. For Patient Dissatisfaction (Fig 3), our data seem to indicate that there is a long-term gradual improvement (secular linear trend), but no particular change at or around the time of diagnostic remission. Employment data (Fig 4) are somewhat noisier because of the smaller number of subjects; however, overall there is statistically significant pre-post improvement in functioning. There is, however, some trend for Employment Impairment

Fig 1. sion.

Time trend for GAS relative to month of remis-

COURSE OF FUNCTIONING AFTER PANIC REMISSION

Fig 2. Time trend for Rater Assessment of Impairment relative to month of remission, by major depression status.

to rise after remission. Marital Role Impairment (Fig 5) seems to begin to improve in the months prior to remission (hence the statistically significant secular linear trend); unfortunately, this improvement seems to cease shortly after remission, being replaced by a slight trend toward deterioration. DISCUSSION

The picture of social functioning relative to symptom status that emerges from these analyses is complex. In some cases, recovery of social functioning seems to precede diagnostic recovery, but in other instances little if any recovery is apparent. Even though our data show that in some do-

Fig 3. Time trend for Patient Dissatisfaction relative to month of remission.

445

Fig 4. Time trend for Employment Impairment relative to month of remission.

mains there is improvement after remission, even in these cases significant impairment remains for at least 18 months after diagnostic recovery. These findings are consonant with earlier studies indicating that symptoms alone do not appear to account for all, or even most, of the variation in functioning.1,3,8,27 There are a number of other possible influences on functional improvement that may need to be taken into account before we can expect to see substantial increments in functioning. Comorbidity is an important area to examine. Other studies have indicated that subjects with greater diagnostic complexity (i.e., multiple axis I disorders or the addition of a personality disorder) have worse social adjustment and role functioning than patients with less diagnostic complexity.28 In pop-

Fig 5. Time trend for Marital Impairment relative to month of remission, by major depression status.

446

ulations such as HARP’s, comorbidity levels tend to be high, and it is plausible that recovery from all current disorders, rather than just one, may be necessary to achieve substantial improvement in functioning. Other variables potentially related to functioning include duration of illness prior to remission, axis II comorbidity, personality variables (e.g., temperament), and medical comorbidity. Our results on the effects of major depression on functioning indicate that it is an important factor that deserves further study. Some of the variations in the effect of panic remission on functioning across domains have relatively direct potential explanations. Household Functioning is probably less affected than other domains by agoraphobia symptoms, in particular since it infrequently requires leaving the house. Relationships with Friends can also be maintained in many ways in the face of limited mobility. GAS scores should be strongly affected if only because they depend in part on symptom status; Rater Assessment, which is nominally independent of symptom status, also shows a relatively large effect, however. The fact that Patient Dissatisfaction shows little if any change around the time of remission is somewhat of a surprise, but this measure may reflect patients’ perspectives that other factors may be at least as important as symptoms. This study has some limitations that need to be discussed to put the results into context. One limitation is that because functioning data were recorded monthly, changes occurring over a shorter time scale would be missed. Since the same interviewer tracked both diagnostic status and functional impairment, we cannot definitively rule out the possibility that the functioning ratings may have been swayed by the diagnostic ratings, or vice versa. There are no comparable studies of which we are aware in which diagnostic and functioning ratings are done independently. We do know, however, that inter-rater reliability for both types of ratings are high.15 That there are many instances in which diagnostic improvement occurs without a substantial change in functioning argues against the hypothesis that bias in one rating was a major influence on the other. Despite the limitations of the data set, we are not aware of the existence of any data better suited to address questions such as those we have posed.

STOUT ET AL

Our current data set allows us to compute statistics for periods more than 18 months prior to and after remission, but the sample sizes become so small that the generalizability of the results becomes questionable. Since some of our results suggest that functioning change in some domains may begin well before remission, and in some domains may have a complex time course after remission, further study should focus on the long-term course of functioning and its determinants. We have demonstrated that carefully collected longitudinal data and modern statistical techniques can reveal important information about the time course of recovery of functioning relative to remission from panic and other disorders. Our findings raise a number of interesting questions about the causal mechanisms responsible for functional improvement. Disparate processes with distinct time courses seem to affect different domains of functioning. Further research on the longitudinal determinants of functioning should help to reveal the causal patterns underlying the limited recovery we have seen in our data on panic disorders. Knowledge of these causal patterns should help in the creation of treatment protocols to maximize functional recovery in psychiatric patients. APPENDIX The Harvard/Brown Anxiety Disorder Research Program is conducted with the participation of the following investigators: M.B. Keller, M.D. (Chairperson); J. Eisen, M.D., K. Phillips, M.D., R. Stout, Ph.D. (Decision Sciences Institute and Brown University School of Medicine); M.T. Shea, Ph.D. (Veterans Administration Hospital, Providence-Brown University School of Medicine); M.G. Warshaw, M.S.S, M.A. (Brown University School of Medicine); R.M. Goisman, M.D. (Massachusetts Mental Health Center-Harvard University School of Medicine); G. Mallya, M.D. (McLean Hospital-Harvard University School of Medicine); F. Rodriguez-Villa, M.D. (The Cambridge Hospital-Harvard University School of Medicine); M.P. Rogers, M.D. (Brigham and Women’s Hospital-Harvard University School of Medicine); R. Vasile, M.D. (New England Deaconess Hospital-Harvard University School of Medicine); A. Massion, M.D. (University of Massachusetts Medical Center); G. Steketee, Ph.D. (Boston University School of Social Work); K. Yonkers, M.D. (University of Texas, Dallas); I. Goldenberg, Psy.D.; and E. Fierman, M.D. Additional contributions from: P. Alexander, M.D.; J. Curran, M.D.; J. Cole, M.D. (McLean Hospital-Harvard University School of Medicine); J. Ellison, M.D., M.P.H. (Harvard Pilgrim Health Care-Harvard University School of Medicine); A. Gordon, M.D., S. Rasmussen, M.D. (Butler Hospital-Brown University School of Medicine); R. Hirschfeld Ph.D. (University of

COURSE OF FUNCTIONING AFTER PANIC REMISSION

447

Texas, Galveston); J. Hooley, D.Phil. (Harvard University); P. Lavori, Ph.D. (Stanford University); J. Perry, M.D. (Jewish General Hospital-McGill University School of Medicine, Montre´al); L. Peterson (Veterans Administration Hospital, Togus, ME); J. Reich, M.D., M.P.H.; J. Rice, Ph.D. (Renard HospitalWashington University School of Medicine); H. Samuelson, M.A. (Brigham and Women’s Hospital); D. Shera, M.S. (Harvard School of Public Health); N. Weinshenker, M.D.

(New Jersey Medical School); M. Weissman, Ph.D. (Columbia University); K. White, M.D.

ACKNOWLEDGMENT The Quintiles Corporation provided valuable consultation services to this project. The authors are especially grateful to Meredith Warshaw, M.A. for her contributions to the statistical and substantive aspects of the manuscript.

REFERENCES 1. Dohrenwend BS, Dohrenwend BP, Link B, Levav I. Social functioning of psychiatric patients in contrast with community cases in the general population. Arch Gen Psychiatry 1983; 40:1174-1182. 2. Mauri M, Sarno N, Armani A, Rossi VM, Zambotto S, Cassano G, et al. Differential social adjustment correlates of axis I and axis II psychopathology in anxiety and depressive disorders. Eur Psychiatry 1991;6:127-130. 3. Skodol AE, Link BG, Shrout PE, Horwath E. The revision of axis V in DSM-III-R: should symptoms have been included? Am J Psychiatry 1988;145:825-829. 4. Spitzer RL, Endicott J, Fleiss JL, Cohen J. The Psychiatric Status Schedule: a method for evaluating psychopathology and impairment in role functioning. Arch Gen Psychiatry 1970;23: 41-55. 5. Wells KB, Stewart A, Hays RD, Burnam A, Rogers W, Daniels M, et al. The functioning and well-being of depressed patients: results from the Medical Outcomes Study. JAMA 1989;262:914-919. 6. Coryell W, Scheftner W, Keller M, Endicott J, Maser J, Klerman GL. The enduring psychosocial consequences of mania and depression. Am J Psychiatry 1993;150:720-727. 7. Scheibe G, Albus M. Predictors and outcome in panic disorder: a 2-year prospective follow-up study. Psychopathology 1997;30:177-184. 8. Miller IW, Keitner GI, Schatzberg A, Klein DN, Thase ME, Rush JA, et al. The treatment of chronic depression, part 3: psychosocial functioning before and after treatment with sertraline or imipramine. J Clin Psychiatry 1998;59:608-619. 9. Perugi G, Maremmani I, McNair DM, Cassano GB, Akiskal HS. Differential changes in areas of social adjustment from depressive episodes through recovery. J Affect Disord 1988;15:39-43. 10. Stewart JW, Quitkin FM, McGrath PJ, Rabkin JG, Markowitz JS, Tricamo E, et al. Social functioning in chronic depression: effect of 6 weeks of antidepressant treatment. Psychiatry Res 1988;25:213-222. 11. Katz MM, Secunda SK, Hirschfeld RMA, Koslow SH. NIMH Clinical Research Branch Collaborative Program on the psychobiology of depression. Arch Gen Psychiatry 1979;36: 765-771. 12. Spitzer RL, Williams JB, Gibbon M, First MB. Structured Clinical Interview for DSM-III-R-Patient Version (SCIDP). New York, NY: Biometrics Research Department, New York State Psychiatric Institute, 1988. 13. Endicott J, Spitzer RL. A diagnostic interview: The Schedule for Affective Disorders and Schizophrenia. Arch General Psychiatry 1978;35:837-844. 14. Keller MB, Lavori PW, Friedman B, Nielsen E, Endicott

J, McDonald-Scott NC, et al. The longitudinal interval follow-up evaluation: a comprehensive method for assessing outcome in prospective longitudinal studies. Arch Gen Psychiatry 1987;44:540-548. 15. Warshaw MG, Keller MB, Stout RL. Reliability and validity of the Longitudinal Interval Follow-up Evaluation for assessing outcome of anxiety disorders. J Psychiatry Res 1994; 28:531-545. 16. Massion AO, Warshaw MG, Keller MB. Quality of life and psychiatric morbidity in panic disorder versus generalized anxiety disorder. Am J Psychiatry 1993;150:600-607. 17. Keller MB, Yonkers KA, Warshaw MG, Pratt LA, Gollan JK, Massion AO, et al. Remission and relapse in subjects with panic disorder and panic with agoraphobia. J Nerv Ment Dis 1994;182:290-296. 18. Warshaw MG, Massion AO, Keller MB, Shea MT. Predictors of remission in patients with panic with and without agoraphobia: prospective five-year follow-up data. J Nerv Ment Dis 1997;185:517-519. 19. Yonkers KA, Zlotnick C, Allsworth J, Warshaw M, Shea T, Keller MB. Is the course of panic disorder the same in women and men? Am J Psychiatry 1998;155:596-602. 20. Keller MB, Shapiro RW, Lavori PW, Wolfe N. Recovery in major depressive disorder: analysis with the life table and regression models. Arch General Psychiatry 1982;39:905-910. 21. Mueller TI, Keller MB, Leon AC, Solomon DA, Shea T, Coryell W, et al. Recovery after 5 years of unremitting major depressive disorder. Arch Gen Psychiatry 1996;53:794-799. 22. Strober M, Freeman R, Morrell W. The long-term course of severe anorexia nervosa in adolescents: survival analysis of recovery, relapse, and outcome predictors over 10-15 years in a prospective study. Int J Eating Disord 1997;22:339-360. 23. Endicott J, Spitzer RL, Fleiss JL, Cohen J. The Global Assessment Scale: a procedure for measuring overall severity of psychiatric disturbances. Arch Gen Psychiatry 1976;33:766771. 24. Bryk AS, Raudenbush SW. Hierarchical Linear Models. Newbury Park, CA: Sage, 1992. 25. McDowall D, McCleary R, Meidinger EE, Hay RA. Interrupted Time Series Analysis. Beverly Hills, CA: Sage, 1980. 26. Cochran W. Planning and Analysis of Observational Studies. New York, NY: Wiley, 1983. 27. Leon AC, Shear MK, Portera L, Klerman GL. The relationship of symptomatology to impairment in patients with panic disorder. J Psychiatry Res 1993;27:361-367. 28. Fabrega, H, Pilkonis, P, Mezzich, J, Ahn, C. W, and Shea,. S. Explaining diagnostic complexity in an intake setting. Comprehensive Psychiatry 1990;31:5-14.