PREDICTION OF RESPONSE TO ANTIDEPRESSANT MEDICATION BY A SIGN-BASED INDEX OF MELANCHOLIA Gordon Parker, Dusan Hadzi-Pavlovic

We examine the capacity of a new melancholia index (focussing on CORE signs indicative of psychomotorchange) to predict responseto antidepressant medication. In a naturalistic study, depressed patients were assessed and rated by their treating psychiatrist, and we focus on those who received antidepressants for three weeks or more. Those subjects who returned high CORE scores (putative melancholia) at baseline improved more over the initial six weeks, but analyses suggested that the differential improvement outcome was independent of medication. Thus, CORE scores predicted outcome for those receiving antidepressant medication, but possible mechanisms remain unresolved. Australian and New Zealand Journal of Psychiatry 1993; 2756-61 We have previously suggested that melancholia (or “endogenous depression”) may be differentiated from residual depressive disorders more by certain signs than by symptoms [l]. The signs respect the longrecognised suggestion, summarised in two reviews, [2,3] that psychomotor disturbance is a key feature of and is perhaps specific to melancholia, and our earlier research will be briefly reviewed. On the basis of two multivariate (principal components and latent class) analyses of data collected from two independent groups of clinically depressed patients, fifteen CORE signs were constructed [I], based on clinicians’ ratings of psychomotor disturbance derivedfrom a structured mental status schedule (assessing presence vs absence, and, if present, severity of each feature). Clinically diagnosed “psychotic” and “endogenous” dcprcssives received

School of Psychiatry, University of New South Wales, Psychiatry Unit, Prince of Wales Hospital, Randwick, New South Wales Gordon Parker MD, PhD, FRANZCP, Professor of Psychiatry Dusan Hadzi-Pavlovic BSc. MPsychol, Lecturer in Psychiatry

much higher CORE scores (summed for each of the 15 items) than diagnosed “neurotic” and “reactive” depressives, and the CORE scores were more successful at differentiating clinical diagnostic groups than summed “endogeneity” symptom scores. DSM-111 melancholic, RDC and “Newcastle” endogenously depressed subjects also returned much higher CORE scores than their non-melancholic or non-endogenous comparison groups, offering concurrent support for the view that the CORE system could well be regarded as a melancholia index. Other validating strategies were also supportive of the CORE index. We reported [ 11, for instance, that high CORE subjects were less likely to report predisposing and precipitating life event stressors, and had significantly higher dexamethasone suppression test (DST) non-suppression rates, consistent with the view that biological factors have greater relevance (and psychosocial factors less relevance) to melancholia than to nonmelancholic depression. Sub-typing of depression is of most relevance to the clinician if there are treatment implications. While it has long been held that melancholia is selectively

GORDON PARKER, DUSAN AHDZI-PAVOLVIC

Table I . Mean Zung and iisual analogue sample and sample sub-grouped as “high” or “low”CORE, over the first sis w>eeX-s

Table 2. Clinician-judged improvement or not against receipt of antidepressant medication or not,for separate high and low CORE subjects

scores for

Scale

Group

Base

Six Change weeks %

Whole (n=33) High CORE (n=18) Low CORE (n=l5)

54.1 54.6 53.4

42.3 38.9 46.3

20.7 27.3 12.8

analogue Whole (n=41) High CORE (n=21) Low CORE (n=20)

57.2 57.9 56.5

37.6 31.0 44.5

32.2 46.0 17.8

Zung

57

CORE group

Received Clinician-judged outcome antiImproved Not improved depressants

High High

Visual

Low Low

I

responsive to “biological” treatments [4], DSM-111-R is, however, the only classificatory system including positive response to biological treatment as a criterion. To test “treatment validation” of the CORE system, we designed two naturalistic studies examining its capacity to predict response to ECT and to antidepressant medication. A high CORE score was a significant predictor of ECT response in patients with “retarded” but not “agitated” melancholic disorders [S], and appeared to subsume two previously suggested strong predictors of ECT outcome - severity and psychotic features. The second naturalistic study is reported here.

Method The sample came from one of the two samples used to develop the CORE system. In essence, 300 depressed inpatients or outpatients were assessed by one of some 30 volunteering Sydney psychiatrists who undertook all baseline and follow-up assessments. The psychiatrists were aware that the study design prioritised the collection of baseline data (in seeking to derive a new typology of depression based on clinical features), while no objectives for the longitudinal data were stated. At baseline, the depressed patient completed two self-report measures of depression severity, the Zung [6] and a visual analogue (VA) measure (assessing nine differing depressive cognitions), while the psychiatrist generated DSM-111 and clinical diagnoses (i.e. with the latter having options of “psychotic”,

No

“endogenous”, “neurotic” or “reactive” depression), collected symptom data and rated the subject on the mental state signs. The summed CORE score can be used dimensionally (i.e. scores ranging from 15 to 45) or categorically in analyses, with the subsequently established cut-off [ I ] of 25 or more assigning a subject to the “high CORE’ sub-group. Subsequently, if the patient was treated by that psychiatrist, we requested progress to be monitored weekly, with assessments of severity (using the same VA scale) and treatment modalities to be recorded. Four, six and ten weeks after baseline assessment, the patients were requested to complete Zung scales again, and the clinicians to judge the level of improvement since initial assessment (i.e. “recovered”, “considerably improved”, “slight improvement”, or “no improvement”). As this was a naturalistic study, data could only be recorded for those weeks when the patient actually consulted the clinician.

Results Partial or complete longitudinal data were available for 38% (114/300) of the patients, now termed the “respondents”, the response rate for the longitudinal component being low due to failure to collect data at given time points, brief treatment courses and/or infrequent consultations. The mean age of respondents was 47.2 (SD=16) years, 58% were females, the mean age at first episode was 32.6 (SD=18) years, while the mean duration of episode was 35 and median 10 weeks. Respondents did not differ from non-respondents in terms of age (t = 1.34), percentage of females (x2=0.8I),age when first depressed(t= I.lS),length

PREDICTION OF RESPONSE TO ANTIDEPRESSANT MEDICATION

58

Table 3. Regression analyses assessing three predictors of percentage reduction in Zutig scores

Diagnostic Variables Multiple Equation Beta R weight independent entered predictor F ratio p

CORE score 0.31 0.33 0.35

3.40 0.07 +0.26 1.86 0.17 +0.18 1 3 7 0.27 -0.14

Clinical diagnosis 0.32 Antidepressant 0.32 Psychotherapy 0.34

3.45 0.07 +0.26 1.73 0.19 +0.11 1.25 0.31 -0.13

DSM-Ill diagnosis 0.04 Antidepressant 0.16 Psychotherapy 0.27

0.06 0.81 -0.07 0.38 0.68 +0.26 0.74 0.54 -0.25

CORE score Antidepressant Psychotherapy CIinical diagnosis

DSM-Ill diagnosis *t

* Here, those diagnosed as having psychotic or

endogenous depressions are contrasted with residual diagnoses ** Here, major depression with psychosis or melancholia are contrasted with residual DSM-Ill diagnoses

of present episode (t = 1.18), or initial Zung (t = 0.66) scores, but did have slightly higher CORE scores (29.1 vs 26.5, t = 2.53, p