Dimensional diagnosis of depression: Adding the dimension of course to severity, and comparison to the DSM

Dimensional Diagnosis of Depression: Adding the Dimension of Course to Severity, and Comparison to the DSM Stewart A. Shankman and Daniel N. Klein It has long been debated whether depression is best classified with a categorical or dimensional diagnostic system. There has been surprisingly little discussion, however, of what the contents of a dimensional classification should include, with most studies employing a single dimension based on symptom severity. The present study explored whether a dimension based on prior course of depression increases the validity of a dimensional model based on symptom severity alone and whether the two dimensions combined improve upon the present categorical system (DSM). The sample consisted of 133 patients with a broad spectrum of depressive diagnoses. External

validators included family history of mood disorder, assessed using the family history method, and course over a 6-month prospective follow-up period. Prior course contributed significant incremental validity over and above symptom severity in predicting subsequent course and family history of mood disorder. However, the linear combination of symptom severity and prior course provided only a minimal increase in predictive power over and above the DSM diagnoses of major depressive disorder (MDD) and dysthymia. Copyright 2002, Elsevier Science (USA). All rights reserved.

O

a century, with much of the controversy focusing on whether the endogenous and neurotic subtypes were qualitatively distinct or represented different ends of a severity continuum.6,7 More recently, the categorical versus dimensional debate has focused on whether subclinical forms of depression lie on a continuum with clinical cases8,9 or if subthreshold and clinical cases represent qualitatively distinct latent classes.10 Much of the literature on this topic, however, has been complicated by methodological issues, such as sampling (clinical v college students) and instruments (diagnostic interviews v self-report inventories).11-14 Nonetheless, a considerable body of data has emerged indicating significant links between subclinical and full syndromal forms of depression.15-18 Despite the arguments for the validity of dimensional models of depression, there has been surprisingly little consideration of what the contents of such a dimensional classification system of depression should include.19 Most of the literature in this area has focused on a single dimension of current symptom severity, typically operationalized by summing the number of symptoms rated in an interview or endorsed on an inventory. Only a few investigators have proposed multidimensional models or considered the value of incorporating nonsymptom features in a dimensional classification system. One of the more sophisticated dimensional classification systems for depression is Clark and Watson’s20 tripartite model, which was developed to account for the high comorbidity between depression and anxiety. The tripartite model includes three dimensions: positive affect, negative affect,

NE OF THE GOALS of any classification system is to describe a phenomenon as it naturally occurs in the world, that is, to borrow the Platonic phrase, to “carve nature at its joints.”1 Debates about whether mental disorders are discrete or continuous, and whether they should be classified using a categorical or a dimensional system have been longstanding, and often heated.2 The current classification system for psychopathology, the DSM-IV, is categorical, although it does include some dimensional features (e.g., severity descriptors) and acknowledges that psychiatric disorders may not be discrete. However, critics have argued that dimensional systems may better reflect the true nature of many forms of psychopathology, as well as convey more information and have greater reliability.3 The argument for dimensional models of classification is probably most developed in the area of personality disorders, where a number of specific dimensional models have been proposed, and empirical data on their validity is beginning to accumulate.4,5 In depression, the debate over categorical versus dimensional models can be traced back for almost From the Department of Psychology, and the Department of Psychiatry and Behavioral Science, State University of New York at Stony Brook, Stony Brook, NY. Supported by National Institute of Health Research Grants No. RO3 MH39782 and RO1 MH45757 (to D.N.K.). Address reprint requests to Stewart Shankman, M.A., Department of Psychology, SUNY-Stony Brook, Stony Brook, NY 11794-2500. Copyright 2002, Elsevier Science (USA). All rights reserved. 0010-440X/02/4306-0001$35.00/0 doi:10.1053/comp.2002.35902 420

Comprehensive Psychiatry, Vol. 43, No. 6 (November/December), 2002: pp 420-426

DIMENSIONAL DIAGNOSIS OF DEPRESSION

and physiological arousal. In this model, depression is characterized by low positive affect, anxiety is characterized by high physiological arousal, and both depression and anxiety share a high level of negative affect. A number of recent studies have reported empirical support for this model,21,22 although there have also been some mixed findings.23 A dimensional classification system based exclusively on symptomatology is unlikely to suffice, however, because levels of depression severity typically wax and wane over time,8,24 individual symptoms are not stable across episodes,25 and there is poor concordance between patients’s symptom reports and behavioral ratings from trained observers.26,27 In addition, a dimensional system based exclusively on cross-sectional symptomatology ignores the marked heterogeneity evident in the longitudinal course of depression.28 Heterogeneity in course is also associated with differences on clinically and etiologically relevant variables that cannot be accounted for by differences in symptom severity.29 For example, patients with episodic major depressive disorder (MDD) and MDD superimposed on dysthymic disorder (“double depression”)30 are similar with respect to number and types of symptoms.31 However, the two groups differ not only on prior course (i.e., dysthymia or not), but also on rates of comorbidity, family history of depression, history of childhood adversity, subsequent course, and response to treatment.32,33 Conversely, while “double depression” is associated with a higher level of symptomatology than “pure” dysthymic disorder, the two conditions are similar with respect to most of the above variables.32,33 Recently, Angst et al.29 argued for a dimensional classification system of depression that included measures of longitudinal course as well as symptomatology. They proposed that recurrence and duration be considered along with symptom severity, and argued that a number of forms of depression (MDD, dysthymic disorder, recurrent brief depression, and minor depressive disorder) could be described as variations along these dimensions. The present report addresses two aspects of the debate on categorical versus dimensional models of depression. First, we consider the contents of a dimensional system. Following Angst et al.,29 we explore whether a dimension based on longitudinal

421

course will increase the validity of a classification system based on symptom severity alone. In addition to number of episodes and duration, we included a third course variable: age of onset. Each of these variables has at least several consistent correlates that are clinically and/or etiologically significant. A history of recurrent MDD is associated with a higher risk of relapse/recurrence34,35 and an increased familial risk for depression.36-38 Duration of depression (i.e., chronicity) is associated with lower recovery rates,33,39 greater comorbidity,40,41 increased familial risk for depression,37 and poorer response to treatment.42 Finally, age of onset has been found to be associated with greater comorbidity40,43 and an increased familial risk for depression.36,38 Second, if a dimensional model is to be a viable alternative to the present categorical system, it must improve upon the existing system. Hence, we examine whether a two-dimensional model outperforms the two central unipolar depression categories in the DSM (MDD and dysthymic disorder) in predicting several clinically and etiologically relevant variables: family history of mood disorders and subsequent course. METHOD

Subjects Subjects were drawn from a larger study of the assessment, classification, and course of chronic affective disorders.44 As part of the larger project, a consecutive series of 550 adult outpatients at a community mental health center and a university-based training clinic completed the General Behavior Inventory (GBI),45 a screening inventory for chronic and recurrent unipolar and bipolar affective conditions. Using a stratified random sampling method that disproportionately sampled from the upper range of the distribution of GBI scores, 177 patients were selected and administered a structured diagnostic interview and a battery of inventories. The patients were highly representative of the larger population from which they were drawn with respect to age, sex, race, marital status, occupation, and education. Compared with blind diagnoses based on structured interviews, the GBI exhibits good to excellent negative predictive power, and acceptable positive predictive power.45,46 Patients were included in the present analyses if they met current or probable Research Diagnostic Criteria (RDC)47 for MDD, intermittent depressive disorder, or minor depressive disorder. The RDC were used for inclusion criteria because they contain a broader spectrum of depressive diagnoses than the DSM-III, the official classification system in use at the time of data collection. Patients were excluded from all analyses if they met current or lifetime DSM-III criteria for bipolar disorder, cyclothymia, schizophrenia, or schizoaffective disorder. Of the 177 patients, 133 met these criteria and were used in all sub-

422

SHANKMAN AND KLEIN

Table 1. Characteristics of Sample at Baseline (N ⴝ 133) Mean age, yr (SD) Percent male Mean years completed in school (SD) Percent caucasian Percent married Mean age of onset of depressive syndrome, yr (SD) Mean chronicity rating (SD) Mean no. of previous MDEs (SD) Mean depression severity (SD)

28.8 (8.1)* 24.1% 14.1 (2.2) 94.0% 30.1% 18.8 (8.8)† 4.00 (1.6) 1.68 (1.5) 37.4 (6.0)

NOTE. Mean chronicity rating ⫽ 6-point scale of chronicity of depression (see text); Mean depression severity ⫽ sum of ratings of all DSM-III MDD and melancholia symptoms during the worst period in the index episode. *Range, 17 to 58. †Range, 5 to 55.

sequent analyses. Written informed consent was obtained for all participants. Descriptive characteristics of the sample are presented in Table 1. One hundred patients (75.2%) met criteria for definite or probable RDC MDD. Seventy-six patients (57.1%) met criteria for RDC intermittent or chronic minor depression. Eight patients (6.0%) met criteria for RDC acute minor depression. Of the 133 patients in the sample, 79 (59.4%) met criteria for DSM-III MDD and 57 (42.9%) met criteria for DSM-III dysthymia. Thirty-one (23.3%) of the patients met criteria for both dysthymic disorder and MDD (i.e., double depression).

Diagnostic Assessment Diagnostic interviews were based on a modified version of the Schedule for Affective Disorders and Schizophrenia (SADS).48 Data on psychopathology in all first-degree relatives over the age of 17 were systematically collected using the Family History Research Diagnostic Criteria (FH-RDC) interview guide.49 Interviews were administered within the first few treatment sessions. The interviews typically required between 2 and 3 hours to complete and were tape-recorded. A clinical psychology faculty member and several advanced graduate students conducted the interviews. Diagnoses were derived by the second author. Family members were diagnosed according to the FH-RDC. In approximately half of the cases where a student conducted the interview, the second author reviewed the audiotape of the interview before assigning a diagnosis. In the remaining instances, diagnoses were based on the interviewer’s presentation of the case and a review of the interview protocol. When necessary, patients were recontacted for further information. To assess interrater reliability, the audiotapes of 20 consecutive interviews conducted by one of the authors (D.N.K.) were independently rated by the primary graduate student interviewer. Kappas were .87 for both MDD and dysthymia. For family history of depressive disorder, kappa was .64.

episode on a 4-point scale. The severity of depression dimension was constructed by summing these ratings. Interrater reliability for depression severity was high (r ⫽ .91). The course variables included age of onset of the depressive syndrome, number of previous major depressive episodes (MDEs), and a variable representing duration (chronicity). Due to the positive skewness and kurtosis of number of previous MDEs, we truncated this variable at four or more previous episodes. Chronicity was defined according to the following scale: 1 ⫽ longest period of depression in the last year was less than 2 weeks; 2 ⫽ longest period of depression in the last year was 2 to 4 weeks; 3 ⫽ longest period of depression in the last year was 4 weeks to 1 year; 4 ⫽ patient was depressed for the past 1 to 2 years; 5 ⫽ patient was depressed for the past 2 to 10 years; and 6 ⫽ patient was depressed for the more than 10 years. As would be expected, the three course variables were significantly intercorrelated: r (age of onset, chronicity) ⫽ -.47; r (age of onset, previous number of MDEs) ⫽ -.49; and r (chronicity, previous number of MDEs) ⫽ .29. As it is impractical to have a classification system with three different dimensions for course alone, we conducted a principle components analysis to determine whether we could reduce the course variables into a smaller number of dimensions. The principal components analysis indicated that the three course variables loaded on a single factor that accounted for 61.3% of the total variance (factor loadings were all ⬎ .74). Therefore, we converted the variables to Z scores and summed them to create a composite score for prior course.

Follow-up Assessment Follow-up assessments were conducted 6 months after entry into the study. In almost all cases, the follow-up and intake interviews were conducted by different interviewers. Follow-up data were obtained for 71.4% (N ⫽ 95) of the original sample. The follow-up was naturalistic in that treatment was not controlled. However, all treatment received during the follow-up period was recorded. The follow-up assessment included a semistructured interview based on the Longitudinal Interval Follow-up Schedule (LIFE).50 Two measures of course and outcome were derived from the interview; mean level of depression across the 6 months, and whether or not the patient recovered from all mood disorders during the follow-up period. According to the LIFE conventions, recovery was defined as a minimum of 8 consecutive weeks with no more than one or two mild depressive symptoms. Interrater reliabilities were kappa ⫽ .86 for recovery and r ⫽ .85 for level of depression at follow-up. Patients who were and were not available for follow-up were compared on the following baseline variables: age, sex, race, marital status, education, socioeconomic status, score on the Global Assessment Scale, age of onset of depression, percent with dysthymia, percent with MDD, and the severity and previous course dimensions. Patients who were followed up were significantly older (mean ⫾ SD age, 29.7 ⫾ .6 years) than those who were not followed up (age, 26.7 ⫾ 6.2 years; t(131) ⫽ 1.98, P ⬍ .05). None of the other comparisons were significant.

Data Analysis Dimensions Interviewers rated the severity of all DSM-III MDD and melancholia symptoms during the worst period in the index

The Weinberg abridged method51 was used to calculate the age-corrected prevalence rates (or morbid risk) of mood disorders in relatives. The risk period used was 18 to 59 years. Often,

DIMENSIONAL DIAGNOSIS OF DEPRESSION

423

Table 2. Zero-Order Correlations Between Diagnostic Dimensions, DSM-III Diagnoses, and Validators

Predictors

Dimensions Depression severity Previous course Diagnoses MDD Dysthymic disorder

Family History of Depression

Family History of Severe Depression

Recovery

Mean Depression During Follow-up

⫺.24* ⫺.39†

.25* .30†

.13 .18*

.23† .27†

⫺.02 ⫺.41†

.06 .24*

⫺.12 .21*

.10 .26†

NOTE. Depression severity ⫽ sum of interviewer ratings of the severity of depressive symptoms; previous course ⫽ the sum of the Z scores of age of onset, no. of episodes and duration/chronicity; MDD ⫽ whether or not the patient had a current DSM-III diagnosis of major depression; Mean depression during Follow-up ⫽ average level of depression across the 6-month follow-up, as determined by the Longitudinal Interval Follow-up Schedule; N ⫽ 133. *P ⬍ .05. †P ⬍ .01.

differences in morbid risk are calculated by pooling all relatives from all families and computing a modified chi-square statistic.52 However, because this violates the assumption of independence of observations, age-corrected prevalence rates were calculated separately within the family of each proband. Thus, the family, rather than the relative, was used as the unit of analysis. In addition to examining mean within-family risk for mood disorder, we also created a variable for family history of severe mood disorder, which was operationalized as whether or not any of the patient’s relatives had ever been hospitalized for a mood disorder. The relations between dimensions, diagnoses and validators were examined using bivariate correlations. Logistic regression analysis was used for categorical validators (recovery from all mood disorders and family history of severe mood disorder) and multiple linear regression analysis was used for continuous validators (familial risk for mood disorder and mean level of depression across the follow-up period).

RESULTS

Are the Severity and Prior Course Dimensions Associated With Family History and Prospective Course? First, we determined if the severity and prior course dimensions were associated with the family history and prospective course variables. To provide a comparison, we also examined the correlations between DSM-III diagnoses of MDD and dysthymia and the four external validators. These correlations are presented in Table 2. Severity significantly predicted three of the four validators (recovery, mean level depression over the follow-up, and family history of severe mood disorder). Previous course was significantly associated with all four validators. While a diagnosis of dysthymia was also significantly related to all four validators, an MDD diagnosis was not significantly

associated with any of the validators within this sample of depressed patients. Does the Prior Course Dimension Add to the Predictive Power of Depression Severity? We next examined whether the prior course dimension contributed significant variance over and above the depression severity dimension using hierarchical multiple linear and logistic regression analyses. In each analysis, depression severity was entered in the first block and prior course was entered in the second block. The linear combination of the two dimensions was significantly related to three of the four validators: mean level of depression over the follow-up (R2 ⫽ .12, adjusted R2 ⫽ .10, F[2, 92] ⫽ 6.24, P ⬍ .01), recovery (␹2[2] ⫽ 17.02, P ⬍ .001), and family history of severe depression (␹2 [2] ⫽ 12.01, P ⬍ .01). The two dimensions were also associated with family history of depression at a trend level (R2 ⫽ .04, adjusted R2 ⫽ .02, F[2, 128] ⫽ 2.48, P ⫽ .09). After controlling for depression severity, prior course significantly predicted recovery (␹2 change [1] ⫽ 11.30, P ⬍ .01), mean level of depression over the follow-up (R2 change ⫽ .06, F change [1, 92] ⫽ 5.74, P ⬍ .05), and family history of severe depression (␹2 change [1] ⫽ 5.04, P ⬍ .05). Does a Dimensional System Provide Additional Predictive Power Over and above DSM-III Diagnoses? Finally, we examined whether the two dimensions were significantly associated with the four external validators over and above the categorical

424

SHANKMAN AND KLEIN

DSM-III diagnoses. Four hierarchical multiple linear and logistic regressions were conducted. In the first block, we entered the presence or absence of a diagnosis of MDD and the presence or absence of a diagnosis of dysthymia. In the second block, the severity and prior course dimensions were entered. The dimensions provided a trend for significant increment in prediction over and above the two categorical diagnoses for mean level of depression across the follow-up period (R2 change ⫽ .05, F[2, 90] ⫽ 2.71, P ⫽ .07). However, the severity and prior course dimensions did not provide significant increments in prediction over and above the DSMIII diagnoses for the other three validators; recovery (␹2 change [2] ⫽ 3.78, P ⫽ .15), family history of depression (R2 change ⫽ .03, F[2, 126] ⫽ 1.85, P ⫽ .16), and family history of severe depression (␹2 change [2] ⫽ 2.64, P ⫽ .27). DISCUSSION

In the depression literature, there is a heated debate as to whether depression is better classified with a dimensional system rather than the current categorical diagnostic system.10-19 Nonetheless, there has been very little discussion regarding the nature of that dimensional system. The results of the present study indicated that the linear combination of severity and previous course of depression predicted recovery, mean level of depression, and family history of severe depression. Moreover, the course dimension contributed significantly over and above the severity dimension for these three validators. While most researchers have considered only a continuum based on symptoms,15-17,20 these findings suggest that a second dimension related to prior course of depression contributes predictive validity over and above a single symptom severity dimension. This is consistent with the model proposed by Angst et al.,29 which included measures of recurrence and duration. Despite evidence supporting the validity of a dimensional model, the categorical versus dimensional discussion has rarely involved a direct comparison of the predictive validity of the two systems. The present study provides such an evaluation by comparing the proposed multidimensional model to categorical DSM-III diagnoses in predicting several variables related to course and etiology. The proposed dimensional model contributed predictive validity at a trend level over and above

DSM-III diagnoses for mean level of depression over a 6-month prospective follow-up. The dimensional model did not, however, contribute incremental validity for the family history variables or for recovery. Thus, the dimensional model at best provided minimal predictive power over the categorical diagnoses. The minimal superiority of the dimensional model may be due to an often overlooked feature of most categorical diagnoses: they are inherently multidimensional (or at least multifactorial), as they include symptoms, duration, as well as exclusion criteria. Thus, although they are not as finely graded as a dimensional system, DSM diagnoses often incorporate a number of attributes that many dimensional systems fail to consider. Because of these multifactorial properties, categorical diagnoses may be harder to improve upon than advocates of dimensional systems assume. The diagnostic criteria for dysthymic disorder places a greater emphasis on the dimension of course than the criteria for MDD31 and thus may have somewhat better predictive validity. In this sample, a diagnosis of MDD was not predictive of any of the validators, while a diagnosis of dysthymic disorder was significantly related to all of them. As previously reported from this sample,44 patients with dysthymic disorder had a poorer course at follow-up and a greater morbid risk for affective disorders than patients with episodic MDD. It is important to note that a different sample might have yielded different results. The present study employed a clinical sample, hence subthreshold cases, single episode cases, and cases with very brief episodes were underrepresented. In addition, due to the original purpose of the study, chronic and recurrent cases were oversampled–with almost half of the sample meeting criteria for dysthymic disorder. Interestingly, most of the support for dimensional models of depression comes from community samples, which are generally nonchronic, and tend to provide finer gradations at the lower/ milder end of the spectrum. It is possible that a categorical diagnosis is better suited for clinical populations and moderate/severe depression, and a dimensional system is more helpful for nonclinical populations and milder/subthreshold forms of depression. There are several limitations to the present study. Depression in relatives was determined

DIMENSIONAL DIAGNOSIS OF DEPRESSION

425

through the family history method rather than direct interviews of family members. Second, we used DSM-III rather than DSM-IV diagnoses. It should be noted, however, that the DSM-III and DSM-IV criteria for MDD and dysthymic disorder are quite similar. Thirdly, it is possible that the 6-month follow-up period was too brief to allow for adequate variability in course. Finally, neither our dimensions nor validators should be considered the only relevant ones in the study of dimensional models of depression. For example, future studies might include personality and cognitive measures (e.g., neuroticism and self-criticism), functional impairment, biological correlates (e.g., sleep electroencephalogram [EEG], hemispheric asymmetry in resting EEG), and treatment response. The selection of validators and dimensions, however, depends largely on the purpose of

the diagnostic system. A nosology may require different criteria depending on whether it is intended for treatment, prevention/early intervention or for epidemiological, biological, or genetic research.2 Regardless of the specific validators and dimensions employed, the present study suggests that dimensional models of depression should include additional features besides symptom severity if they are to provide a viable alternative to the current categorical system. It also indicates that dimensional models may have more difficulty improving upon the existing categorical diagnostic system than many proponents of dimensional models seem to assume. Finally, the study underscores the value of head-to-head empirical comparisons in the debate over the merits of categorical versus dimensional models of depression.

REFERENCES 1. Phaedrus. Plato’s Phaedrus. Ithaca, NY: Cornell University Press, 1998. 2. Klein DN, Riso LP. Psychiatric disorders: problems of boundaries and comorbidity. In: Costello CG (ed): Basic Issues in Psychopathology. New York, NY: Guilford Press, 1993:1966. 3. Eysenck HJ, Wakefield JA, Friedman AF. Diagnosis and clinical assessment: the DSM-III. Annu Rev Psychol 1983;34: 167-193. 4. Livesley WJ, Jang KL, Vernon PA. Phenotypic and genotypic structure of traits delineating personality disorder. Arch Gen Psychiatry 1998;55:371-410. 5. Widiger TA. The DSM-III—R categorical personality disorder diagnoses: a critique and an alternative. Psychol Inquiry 1993;4:75-90. 6. Zimmerman M, Spitzer RL. From DSM-III to DSM-III-R. Am J Psychiatry 1989;146:20-28. 7. Parker G. Classifying depression: should paradigms lost be regained? Am J Psychiatry 2000;157:1195-1203. 8. Judd LL. Pleomorphic expressions of unipolar depressive disease: toward a new diagnostic paradigm. J Affect Disord 1997;45:1-3. 9. Pincus HC, Davis WW, McQueen LE. ‘Subthreshold’ mental disorder: a review and synthesis of studies of minor depression and other ‘brand names.’ Br J Psychiatry 1999;174: 288-296. 10. Solomon A, Haaga DAF, Arnow BA. Is clinical depression distinct from subthreshold depressive symptoms? A review of the continuity issue in depression research. J Nerv Ment Dis 2001;189:498-506. 11. Coyne JC. Self-reported distress: analog or ersatz depression? Psychol Bull 1994;116:29-45. 12. Flett GL, Vrendenberg K, Krames L. The continuity of depression in clinical and nonclinical samples. Psychol Bull 1997;121:395-416. 13. Tennen H, Hall JA, Affleck G. Rigor, rigor mortis, and

conspiratorial views of depression research. J Pers Soc Psychol 1995;68:895-900. 14. Vredenburg K, Flett GL, Krames L. Analogue verses clinical depression: a critical reappraisal. Psychol Bull 1993; 113:327-344. 15. Chen L-S, Eaton WW, Gallo JJ, Nestadt G. Understanding the heterogeneity of depression through the triad of symptoms, course, and risk factors: a longitudinal, population-based study. J Affect Disord 2000;59:1-11. 16. Gotlib IH, Lewinsohn PM, Seeley JR. Symptoms versus a diagnosis of depression: differences in psychosocial functioning. J Consult Clin Psychol 1995;63:90-100. 17. Kendler KS, Gardner CO. Boundaries of major depression: an evaluation of DSM-IV criteria. Am J Psychiatry 1998; 155:172-177. 18. Lewinsohn PM, Solomon A, Seeley JR, Zeiss A. Clinical implications of “subthreshold” depressive symptoms. J Abnorm Psychol 2000;109:345-351. 19. Angst J, Merikangas KR. The depressive spectrum: diagnostic classification and course. J Affect Disord 1997;45:3139. 20. Clark LA, Watson D. Tripartite model of anxiety and depression: psychometric evidence and taxonomic implications. J Abnorm Psychol 1991;100:316-336 21. Brown TA, Chorpita BF, Barlow DH. Structural relationships among dimensions of the DSM-IV anxiety and mood disorders and dimensions of negative affect, positive affect, and autonomic arousal. J Abnorm Psychol 1998;107:179-192. 22. Watson D, Weber K, Assenheimer JS, Clark LA, Strauss ME, McCormick RA. Testing a tripartite model, I. Evaluating the convergent and discriminant validity of anxiety and depression symptom scales. J Abnorm Psychol 1995;104:3-14. 23. Burns DD, Eidelson RJ: Why are depression and anxiety correlated? A test of the tripartite model. J Consult Clin Psychol 1998;66:461-473 24. Judd LL, Akiskal HS, Maser JD, Zeller PJ, Endicott J, Coryell W, et al. A prospective 12-year study of subsyndromal

426

and syndromal depressive symptoms in unipolar major depressive disorders. Arch Gen Psychiatry 1998;55:694-700. 25. Young MA, Fogg LF, Scheftner WA, Fawcett JA. Concordance of symptoms in recurrent depressive episodes. J Affect Disord 1990;20:79-85. 26. Mazure C, Nelson JC, Price LH. Reliability and validity of the symptoms of major depressive illness. Arch Gen Psychiatry 1986;43:451-456. 27. Parker G, Hadzi-Pavlovic D, Boyce P, Wilhelm K, Brodaty H, Mitchell P, et al. Classifying depression by mental state signs. Br J Psychiatry 1990;157:55-65. 28. Keller MB, Klein DN, Hirschfeld RMA, Kocsis JH, McCullough JP, Miller I, et al. Results of the DSM-IV Mood Disorders Field Trial. Am J Psychiatry 1995;152:843-849. 29. Angst J, Sellaro R, Merikangas KR. Depressive spectrum diagnoses. Compr Psychiatry 2000;41:39-47. 30. Keller MB, Shapiro RW. “Double depression”: superimposition of acute depressive episodes on chronic depressive disorders. Am J Psychiatry 1982;139:438-442. 31. Klein DN, Kocsis JH, McCullough JP, Holzer CE, Hirschfeld RM, Keller MB. Symptomatology in dysthymic and major depressive disorder. Psychiatr Clin North Am 1996;19: 41-53. 32. Klein DN, Riso LP, Donaldson SK, Schwartz JE, Anderson RL, Ouimette PC, et al. Family study of early-onset dysthymic disorder: mood and personality disorders in relatives of outpatients with dysthymic disorder and episodic major depression and normal controls. Arch Gen Psychiatry 1995;52:487496. 33. Klein DN, Schwartz JE, Rose S, Leader JB. Five-year course and outcome of early-onset dysthymic disorder: a prospective, naturalistic follow-up study. Am J Psychiatry 2000; 157:931-939. 34. Belsher G, Costello CG. Relapse after recovery from unipolar depression: a critical review. Psychol Bull 1988;104: 84-96. 35. Keller MB, Shapiro RW, Lavori PW, Wolfe N. Relapse in major depressive disorder: analysis with the life table. Arch Gen Psychiatry 1982;39:911-915. 36. Bland RC, Newman SC, Orn H. Recurrent and nonrecurrent depression: a family study. Arch Gen Psychiatry 1986; 43:1085-1089. 37. Kendler KS, Gardner CO, Prescott CA. Clinical characteristics of major depression that predict risk of depression in relatives. Arch Gen Psychiatry 1999;56:322-327. 38. Weissman MM, Merikangas KR, Wickramaratne P, Kidd KK, Prusoff BA, Leckman JF, et al. Understanding the clinical heterogeneity of major depression using family data. Arch Gen Psychiatry 1986;43:430-434.

SHANKMAN AND KLEIN

39. Keller MB, Lavori PW, Endicott J, Coryell W, Klerman GL. “Double depression”: two-year follow-up. Am J Psychiatry 1983;140:689-694. 40. Garyfallos G, Adamopoulou A, Karastergiou A, Voikli M, Sotiropoulou A, Donias S, et al. Personality disorders in dysthymia and major depression. Acta Psychiatr Scand 1999; 99:332-340. 41. Pepper CM, Klein DN, Anderson RL, Riso LP, Ouimette PC, Lizardi H. DSM-III-R axis II comorbidity in dysthymia and major depression. Am J Psychiatry 1995;152:239-247. 42. Thase ME, Reynolds CF, Frank E, Simons AD, Garamoni GD, McGeary J, et al. Response to cognitive-behavioral therapy in chronic depression. J Psychother Pract Res 1994;3: 204-214. 43. Fava M, Alpert JE, Borus JS, Nierenberg AA, Pava JA, Rosenbaum JF. Patterns of personality disorder comorbidity in early-onset versus late-onset major depression. Am J Psychiatry 1996;10:1308-1312. 44. Klein DN, Taylor EB, Dickstein S, Harding K. Primary early-onset dysthymia: comparison with primary nonbipolar nonchronic major depression on demographic, clinical, familial, personality, and socioenvironmental characteristics and shortterm outcome. J Abnorm Psychol 1988;97:387-398. 45. Klein DN, Dickstein, S, Taylor EB, Harding KJ. Identifying chronic affective disorders in outpatients: validation of the General Behavior Inventory. J Consult Clin Psychol 1989; 57:106-111. 46. Depue RA, Krauss S, Spoont MR, Arbisi P. General behavior inventory identification of unipolar and bipolar affective conditions in a nonclinical university population. J Abnorm Psychol 1989;98:117-126. 47. Spitzer RL, Endicott J, Robins E. Research diagnostic criteria: rationale and reliability. Arch Gen Psychiatry 1978;35: 773-782. 48. Spitzer RL, Endicott J. Schedule for the Affective Disorders and Schizophrenia. Ed. 2. New York, NY: New York State Psychiatric Institute, Biometrics Research, 1978. 49. Andreasen NC, Endicott J, Spitzer RL, Winokur G. The family history method using diagnostic criteria: reliability and validity. Arch Gen Psychiatry 1977;34:1229-1235. 50. Keller MB, Lavori PW, Friedman B, Nielsen E, Endicott J, McDonald-Scott P, et al. The Longitudinal Interval Follow-up Evaluation: a comprehensive method for assessing outcome in longitudinal studies. Arch Gen Psychiatry 1987;43:540-548. 51. Slater E, Cowie V. The Genetics of Mental Disorders. London, UK: Oxford University Press, 1971. 52. Breborowicz G, Trzebiatowska-Treciak O. A method for testing differences in morbidity risk for affective psychoses. Acta Psychiatr Scand 1976;54:353-358.