Sub-typing depression, II. Clinical distinction of ...

2 downloads 0 Views 719KB Size Report
psychomotor disturbance ('depressive stupor') over time and across diagnostic systems. The status of psychotic depression has also varied over time. An early ...
Psychological Medicine, 1995, 25, 825-832. Copyright © 1995 Cambridge University Press

Sub-typing depression, II. Clinical distinction of psychotic depression and non-psychotic melancholia G. PARKER,1 D. HADZI-PAVLOVIC, H. BRODATY, M.-P. AUSTIN, P. MITCHELL, K.W1LHELM AND I.HICKIE From the Mood Disorders Unit, Prince Henry Hospital and the School of Psychiatry, University of New South Wales, NSW, Australia

We have attempted to clarify clinical differentiating features of psychotic depression. Forty-six depressed subjects meeting DSM-III-R criteria for major depression with psychotic features were compared with (i) DSM-defined melancholic, (it) Newcastle-defined endogenous, and (Hi) a residual DSM-defined major depressive episode group. Additionally, a 'bottom up' latent class analysis (LCA) suggested a larger sample of 82 'psychotic depressive' subjects, and multivariate analyses contrasted these subjects with both LCA-identified melancholic and all residual depressed subjects. Analyses suggested that, in addition to two features with absolute specificity (delusions and hallucinations), both the DSM-defined and LCA-defined 'psychotic depressive' subjects were significantly more likely to demonstrate marked psychomotor disturbance, to report two morbid cognitions (feeling sinful and guilty; feeling deserving of punishment), as well as be more likely to report constipation, terminal insomnia, appetite/weight loss and (variable across the defined 'psychotic depressive' groups) loss of interest and pleasure. The study identifies a wider set of potentially discriminating clinical variables than previous studies, as well as both indicating the existence and assisting identification of'true' psychotic depression in the absence of formal psychotic features being acknowledged or elicited. SYNOPSIS

INTRODUCTION

The classification and status of psychotic or delusional depression remain unclear with official classificatory systems generally providing few clinical descriptors. The more recent DSM systems have generally superimposed features on a base diagnosis of major depression. Thus, DSM-III (APA, 1980) added delusions or hallucinations, but also allowed 'depressive stupor' (when the individual was mute and unresponsive), and further sub-typed on the basis of the psychotic features being mood congruent or incongruent. However, DSMIII-R (APA, 1987) deleted the capacity to reach diagnosis on the basis of 'depressive stupor', a feature also missing from the third edition of the RDC criteria (Spitzer, 1978) and from the DSM-IV criteria (APA, 1994). ICD-10 (WHO, 1 Address for Correspondence: Professor Gordon Parker, Psychiatry Unit, Prince of Wales Hospital, Randwick, NSW 2031, Australia.

1992) requires the depressive episode to be 'severe', and for either delusions/hallucinations or depressive stupor to achieve the diagnosis. These somewhat varying definitions highlight varying emphases on psychotic features and on psychomotor disturbance ('depressive stupor') over time and across diagnostic systems. The status of psychotic depression has also varied over time. An early separatist (in viewing the conditions as distinct) was Maudsley (1879), who distinguished between 'melancholia' and 'melancholia with delusions', while others (e.g. Kendell, 1976), have regarded 'psychotic depression' and 'endogenous depression' (or melancholia) as essentially synonymous. Prior to DSM-III, the view prevailing in the United States, according to Minter & Mandel (1979), was that 'psychotic depression' meant a 'depression of severe proportions, usually with endogenous symptomatology, not necessarily with symptoms of psychosis'. Such representative views (psychotic depression is, respectively, a separate type of melancholia, synony-

825

826

G. Parker and others

mous with melancholia, and overlapping with melancholia) indicate that any clarification of its status might best involve a comparison group of melancholic depressed patients. Two recent reviews (Parker et al. 1991a; Schatzberg & Rothschild, 1992) indicate, however, that most published studies seeking to clarify either the status or clinical features of psychotic depression have used 'major depression' rather than non-psychotic melancholia as the diagnostic comparison group, presumably respecting the DSM-III and DSM-III-R structure. Those reviews suggest only a few features as either being specific (i.e. delusions and hallucinations) or being over-represented (i.e. psychomotor disturbance, especially agitation; and a severer mood disturbance, evidenced in particular by features such as self-reproach and guilt), while our earlier study (Parker et al. 1991a, b) also suggested a greater chance of reporting a sustained mood disturbance (or absence of a diurnal mood variation) and a much higher rate of constipation. We seek now to clarify the relevance of several suggested latent constructs to psychotic depression assuming, on the basis of previous studies, that dimensions or domains of psychoticism, endogeneity and psychomotor disturbance underpin the condition. As in a previous study (Parker et al. 19916) we examine clinical features distinguishing those with formally diagnosed psychotic depression from non-psychotic melancholic/endogenous depression, but also those distinguishing psychotic depression from a broader' major depression' group. Additionally, as analyses in our previous paper indicated that 'psychotic' features (i.e. delusions, hallucinations) may not always be able to be elicited in those with 'psychotic depression', we reassess this possibility with analyses enabling any latent psychotic depressive class to be identified.

METHOD Pursuing defining features of psychotic depression The preceding paper (I) described both the recruitment of the sample and diagnostic allocations. Univariate analyses reported in Table 1 of paper I identified two ' psychoticism' symptoms (i.e. delusions; hallucinations) which

(by definition) had absolute specificity to the DSM-defined PD class, as well as a set of symptoms which were also over-represented in the PD group but had significant prevalences in the non-PD groups. This suggests three possibilities. First, that measurement error resulted in some non-PD subjects being positively rated on PD-specific items. Secondly, a dimensional view of depression (with PD being a more severe condition than melancholia which, in turn, is more severe than non-melancholia) accounting for the decreasing prevalence estimates across those three respective classes. Thirdly, that PD is a sub-class of 'melancholia', so that PDs might have some' melancholic' features and experience such symptoms at a similar or increased likelihood. In seeking 'defining' features of psychotic depression, the three possibilities need to be conceded, and required us undertaking analyses involving a wide set of potential clinical variables. In addition to that set of psychoticism/endogeneity symptoms suggested and consolidated in paper I, we included CORE scores in the analyses to allow consideration of the psychomotor disturbance (PMD) domain. Analytical strategies

Our assumption that 'psychotic depression', ' melancholia' and' non-melancholic depression' represent separate sub-groups on the basis of differentiating clinical features first required testing. In 'bottom up' analyses we assume that no diagnostic or sub-typing system is valid, and that each observation comes from one of a number of unknown populations, so that our sample is actually a mixture of cases from those populations. Various 'clustering' methods have been devised to group a set of observations by population of origin. When our measures are continuous, and we are willing to assume the distributions of our variables, then an effective method is mixture analysis. When the measures are categorical, a variant called latent class analysis (LCA) is possible. For a specific number of populations, both procedures will estimate the parameters of the distributions (such as means and standard deviations), while an LCA will estimate the probability of each possible response for each variable for an individual from a particular class. Since tests of significance for the number of latent class populations are not very satisfactory for other than a few

827

Sub-typing depression, II

variables, a mixture analysis based on total scores of the variables in the LCA can provide some support for the number of latent classes. Secondly, we report 'top down' analyses, where we assume that the DSM allocation of 46 (14%) of the sample of 327 subjects to a 'psychotic depression' diagnosis is correct, and proceed to determine distinguishing clinical features, using two multivariate analytical strategies (i.e. discriminant function and logistic regression) which have been described in paper I. RESULTS Sample and clinical features Table 1 in paper I provides data on the DSMdefined psychotic depressives (PDs). Delusions were established in 89%, and hallucinations in 24% of the PDs, and both had absolute specificity to the PD class (by definition). The PDs returned significantly higher CORE scores (both on total and on each of the three subscales) than the DSM-allocated melancholic and Newcastle-allocated endogeneous depressives, indicating significantly greater psychomotor disturbance in the PDs than in the 'melancholic' comparison groups. 'Bottom up analyses'

In our 'bottom up' approach, wefirsttested our a priori assumption of there being three 'differing' depressive groups in our sample (i.e. we assume sub-groups of' psychotic depression', 'melancholia', and a residual 'non-melancholic depression' group). We therefore undertook a mixture analysis of our provisionally refined set of clinical features (i.e. delusions and hallucinations; CORE scale scores; nine consolidated endogeneity and/or psychoticism items), examining the total score on all these items. The reduction in the log likelihood in going from one to two, and from two to three populations was significant, suggesting that fitting three classes was tenable. Table 1 reports the three-class latent class analysis (LCA) solution, with Class A (containing 82 subjects) being labelled putative 'psychotic depression', Class B (N = 115) being putative 'melancholia', and Class C (N = 130) a putative residual 'non-melancholic' class. Perhaps the most intriguing aspect is in estab-

Table 1. Latent class analysis based on all patients (three classes fitted) Item probabilities for the three latent (:kisses

Clinical variable Appetite and/or weight loss Mood and/or energy worse AM Anticipatory and/or consummatory anhedonia Terminal insomnia Loss of interest Non-reactive mood Constipation Deserving of punishment Sinful or guilty

Class A Class B Class C Mel* Non-mel* PD* 0-94 0-49 0-66

0-71 0-51 0-57

0-54 0-37 007

0-54 0-70 0-39 0-68 0-47 0-52

0-47 0-43 0-33 0-43 0-20 0-22

013 002 001 019 019 008

Hallucinations Delusions CORE retardation scale CORE agitation scale CORE non-interactiveness scale

013

000

000

0-49 0-92 0-75 0-98

000 0-39 0-21 0-62

001 000 014 013

Prevalence

25%

35%

40%

• Putative classes: PD = Psychotic Melancholia, Non-mel = Non-melancholia.

Depression;

Mel