Abstract
Despite the frequency with which depressive symptoms occur in schizophrenia, we continue to lack a coherent and integrated understanding of where they fit within the clinical conceptualization of the disorder. They have been recognized since the characterization of the disorder at the turn of the last century and described at all phases of the illness process, for example, during initial onset (for example [1]), following the resolution of acute psychotic symptoms (for example [2]) and during illness relapse. Depression in schizophrenia has been conceptualized as an integral part of the disorder (this notion is supported by the finding of depressive symptoms in unmedicated first-onset patients [3, 4]), as a psychological reaction to psychosis [5] or relating to medication treatment. Medication treatment has been proposed to produce depression directly (‘pharmacogenic depression’ [6]) or through the development of an akinetic parkinsonian state mimicking retarded depression [7]. Importantly, depressive symptoms have been associated with poorer prognosis in a number of domains including relapse risk and suicide (for example [8–10]). The frequency of depressive symptoms in schizophrenia has been the focus of many studies with a high variation in rates with the consistent recognition of depressive symptoms in at least 25% of subjects [11, 12].
The relationship between depressive symptoms and the positive and negative symptoms of schizophrenia has also been the focus of considerable study but remains unclear. Although some studies report the lack of a relationship between total depression ratings and total negative symptom ratings [13–17] many others have found the opposite result (for example [18],19). One study described a relationship between depression and negative symptoms after, but not before, antipsychotic therapy [20]. Several studies have found relationships between specific depressive and negative symptom items, especially between vegetative depressive symptoms and negative symptoms [13, 15, 21]. Negative symptoms that appear to have the closest relationship with depressive symptoms include avolition-apathy, anhedonia-asociality [19, 22] and stereotyped thinking [17]. Fewer studies appear to report a relationship between depressive and positive symptoms although this literature also reveals conflicting results (for example [14, 17, 22] report an association where as [16, 18, 23] do not). At least one study showed a differential association between depressive and specific psychotic symptoms such as auditory hallucinations [24].
It is possible that the overlap between depressive, negative and positive symptoms may be explained as the result of the properties of the rating scales used rather than the characteristics of the illness itself. For example the overlap between depressive and negative symptoms has been reported with the three most commonly applied rating scales used in schizophrenia studies (the Positive and Negative Syndrome Scale [PANSS], Scale for Assessment of Negative Symptoms (SANS) [22] and Brief Psychotic Rating Scale (BPRS) [21]). The Positive and Negative Syndrome Scale [25] is one of the most commonly applied rating scales in schizophrenia research and is widely used in treatment studies. In the original form, the PANSS scoring was based upon the use of three factors: positive, negative and general psychopathology. It has subsequently been subjected to at least 20 factor analytic studies with considerably divergent results generating from four to seven factors. Recently, the scoring guidelines for the PANSS have been reviewed based upon an analysis of all published factor structures and a new exploratory and confirmatory factor analytic study [26]. This revision suggests the use of five factors described as the ‘pentagonal model’ [27]. This model includes a dysphoric mood factor as well as positive symptom, negative symptom, activation and autistic preoccupation factors. The inclusion of the mood item may allow the more clear delineation of depressive from positive and negative symptoms although this has not yet been confirmed.
As is apparent from this description, although there is widespread recognition of the importance of depression in schizophrenia, our capacity to independently measure these constructs must remain in doubt. The primary aim of this study was to investigate whether the use of the pentagonal PANSS model would better discriminate depressive, positive and negative symptoms of schizophrenia in a mixed hospitalized and community-based sample of patients. We hypothesized that there would be a relationship between depressive and negative symptoms using the three-factor, but not the pentagonal model. We also hypothesized that there would be less of a relationship between extrapyramidal side-effects and negative symptoms when the pentagonal model was utilized to interpret the PANSS data. We are unaware of any other studies that have looked at these relationships using the newer PANSS scoring model.
Method
Patients for this study were recruited as part of the Schizophrenia Care and Assessment Program (SCAP), a prospective observational study conducted at the Dandenong Psychiatry Research Centre in Victoria, Australia. This study involves the longitudinal study of patients with schizophrenia from the catchment area of the Dandenong Area Mental Health Service (DAMHS). This mental health service provides psychiatric care for patients in an urban and semirural region with a population of approximately 360 000. This study involves the cross-sectional analysis of data from the initial assessments of the first 346 patients recruited into the SCAP.
Patients were recruited from two sources. Firstly, the files of all consecutive admissions to the two acute inpatient units of DAMHS were screened and patients meeting criteria for the study were approached. Patients regarded as being too unwell for inclusion into the study were approached later during their admission or following discharge. Secondly, outpatients and patients receiving case management from the adult community psychiatry services of DAMHS were approached following random selection from case lists.
Inclusion criteria for the SCAP study included: a diagnosis of schizophrenia, schizoaffective disorder or schizophreniform disorder as defined by DSM IV, age greater than 18 years, sufficient capacity to communicate with the investigators in simple spoken and written English and an ability to understand the nature of the study and its implications. Exclusion criteria included an involvement in clinical drug trials in the 30 days prior to the study and a lack of availability for follow up. Only patients with a diagnosis of schizophrenia or schizophreniform psychosis have been included in the study presented in this paper.
Clinical study measures undertaken during the SCAP that are relevant to this study included the following: the PANSS [25], the Montgomery-Asberg Depression Rating Scale (MADRS) [28] and the Extrapyramidal Side-effects Rating Scale (ESRS) [29]. Three trained research assistants made ratings. Interrater reliability was established prior to the conduct of the study and was reviewed on a 6-monthly basis for the PANSS, MADRS and EPRS (ratings with a coefficient of variance of < 0.9).
Statistical analysis was conducted using the Statistical Package for Social Sciences package (SPSS 9.0.1, 1998). T-tests were used to assess differences in depressive symptoms in demographic groups. Pearson correlations were used to examine the relationship between symptom clusters and individual scale items. A comparison between the magnitude of the correlations seen between the two models of positive and negative symptoms and depressive symptoms was conducted using the formula described by Cohen and Cohen [30]. Partial correlations were used to study the relationship between parkinsonian symptoms and negative symptoms controlling for the effect of total MADRS score. The group data were examined for equal variance prior to the application of parametric analysis. Where comparisons were made on individual symptom items a p-value of less than 0.01 was considered significant to adjust for multiple comparisons.
The study was described in writing and verbally to all subjects and written informed consent was obtained. The Southern Health Care Network Human Research and Ethics Committee approved the study.
Results
Five hundred and twenty-three patients met inclusion criteria and 309 consented to participate. Those who consented to participate did not differ in age or gender from those who refused consent (age: p = 0.59, gender: p = 0.64).
Of the 309 patients there were 198 males (mean age 31.2 ± 9.8) and 111 females (36.9 ± 11.4). The females were significantly older (t = − 4.5, df = 307, p = 0.000). The mean age of illness onset reported by the participants was 24.42 ± 8.68 years and 47.1% of the group had been hospitalized in the previous year.
The group mean score for the MADRS was 14.6 (SD 9.07); 46% of the patients had moderate or severe depressive symptoms as defined by a MADRS score of greater than 15 [31].
Inpatients had significantly higher MADRS scores than outpatients (16.6 vs 12.6, p = 0.000). Inpatients also had higher PANSS positive scores (20.2 vs 14.6, p = 0.000). A marginal difference was seen between the groups in the PANSS negative symptoms (20.5 vs 18.6, p = 0.06).
The relationship between depressive, positive and negative symptoms
Correlations between MADRS total scores and PANSS dimensions are presented in Table 1. Significant correlations were seen between MADRS scores and all PANSS factors, with both the original 3-factor PANSS scoring and with the pentagonal model. The presence of positive correlations between both positive and negative symptom scores and total MADRS scores was consistent for outpatients only (positive – depressive symptoms: outpatients r = 0.41, p = 0.000, inpatients r = 0.04, negative – depressive symptoms: outpatients r = 0.50, p = 0.000, inpatients r = 0.26, p = 0.01). The correlation between depressive and negative symptoms was of a similar magnitude for both PANSS models (0.38 and 0.41). The correlation between depressive and positive symptoms was also of a similar magnitude for both PANSS models (0.33 and 0.32). The differences between both sets of correlations were not significant (t = 0.907, df 308, p < 0.05 for negative symptoms; t = 1.278, df 308, p < 0.05 for positive symptoms).
Correlations between MADRS total score and the PANSS dimensions
The relationship between individual MADRS and PANSS negative item scores is presented in Table 2 with correlations controlling for EPRS scores. Depressive symptoms correlated more frequently with items of blunted affect, social and emotional withdrawal and passive/ apathetic social withdrawal. There was little relationship with lack of spontaneity, poor rapport and difficulty in abstract thinking. The MADRS items of reduced sleep, reduced appetite and to a lesser degree pessimistic thoughts and inner tension correlated less frequently with negative symptom items with the latter two items only correlating with PANSS ‘withdrawal’ items. There were considerable correlations with the MADRS items of apparent sadness, reported sadness, concentration difficulties and inability to feel.
Correlations among negative and depressive symptoms controlling for EPRS scores
Relationships with extrapyramidal side effects
A very weak but statistically significant correlation was seen between total MADRS scores and EPRS scores (0.16) (Table 3). EPRS scores did correlate with PANSS negative scores. The magnitude of the correlation with the 3-factor and 5-factor negative symptom factors was similar (0.25 and 0.27). These correlations remained significant after controlling for MADRS total scores in partial correlations (3-factor solution: r = 0.20, p < 0.000, 5-factor solution r = 0.22, p < 0.000). EPRS scores did not correlate with positive symptoms.
Correlations between Extrapyramidal Side Effects Rating Scale and PANSS and MADRS scores
Discussion
The results of our study confirm that depressive symptoms are found commonly in schizophrenia and that there is a significant relationship between depressive, positive and negative symptoms. Significant correlations between PANSS and MADRS scores are found more consistently in outpatients than inpatients and for negative more than positive symptoms. Importantly, the pentagonal model for PANSS scoring fails to more clearly differentiate depressive from positive and negative symptoms. There was an overlap between extrapyramidal and negative symptoms but only a weak relationship between extrapyramidal and depressive symptoms.
The results of the study need to be interpreted in the light of several issues. Firstly, a single rater for each subject made concurrent assessments of negative and depressive symptoms rather than separate assessors as suggested in the literature [32] although the raters were highly trained and regularly assessed for reliability. Secondly, we have been able to consider the effect of parkinsonism but not the direct effect of antipsychotic medication on the prevalence of depressive symptoms and the relationship of depressive to negative symptoms. The cross-sectional nature of the study poses difficulty in our capacity to consider the direction of some of the causal relationships we are considering. Finally, the choice of the depression rating scale is worthy of comment. The properties of the MADRS scale are important in this clinical group as it is widely used in research with patients with schizophrenia. Differing results may have been obtained with the Calgary Depression Scale [33] that was specifically developed to distinguish depressive symptoms in schizophrenia, although this scale has also shown an overlap with negative symptoms in some samples [34].
As previously alluded to, there is a considerable lack of clarity in the research literature as to the relationship between depressive and positive symptoms. Our finding of a positive correlation between depressive and positive symptoms is consistent with the work of several groups [14, 17, 22, 35] but in conflict with others [18, 23]. The conflicting results from these studies may be due to the site and phase of treatment varying between studies as suggested by Nayaka et al., who described a correlation in the post acute but not the acute phase of psychosis [16]. Our finding of a strong relationship for outpatients and no relationship for inpatients would appear to confirm this notion as the inpatient unit within our service is predominately occupied by patients in an acute phase of their illness and has a short average duration of stay. This is somewhat in contradiction to the study by Koreen et al. who followed a clinical group for up to 5 years after the first onset of schizophrenia and found a close relationship between depressive symptoms and the occurrence of episodes of active psychosis [35].
These findings may suggest that there are two important aspects to the relationship between depressive symptoms and the occurrence of active psychosis. Firstly, as seen by Koreen et al. [35], depressive symptoms may occur in the acute illness phase and improve with treatment of the psychosis. Secondly, the finding of a stronger relationship in the post acute phase may indicate that the persistence of some positive symptoms may prove to have an important role in the genesis of depressive symptoms. The presence and persistence of positive symptoms may also prove to be more distressing when individuals are within their own environment rather than the artificial hospital setting surrounded by other individuals with similar illness experiences. Patients who experience significant ongoing active positive symptoms when living in the community are likely to frequently face circumstances that confront their perception of the reality of their experiences. They are also vulnerable to the direct impact of stigmatized attitudes in the community and faced with the impact of their symptoms on their social and occupational functioning. These findings would suggest that antidepressant therapy would be most useful in patients in whom depressive symptoms persist beyond the acute psychotic phase and that patients who experience the persistence of positive symptoms should be carefully monitored for the development of significant mood disorder.
There is also a considerable disagreement in the literature about the nature of the relationship between negative and depressive symptoms. A large number of studies report a positive correlation (for example [18, 19, 22, 36]) and others do not [13–17]. These studies generally are based upon the analysis of total symptom scores. This literature is only partially clarified by looking at the phase of illness, despite the finding by both Goldman et al. [20] and our study of a stronger relationship in the post acute phase, as there are studies in both inpatient and outpatient groups reporting conflicting findings.
Another contributing factor may be the failure of certain items in standard rating scales to discriminate depressive and negative symptoms while other items do clearly make this distinction. There may also be a real overlap between some depressive and negative symptoms; they may overlap regardless of the use of rating scales. In our study, the most significant relationships appeared in the areas of blunted affect, social and emotional withdrawal and passive/apathetic social withdrawal.
This is consistent with studies that have generally found relationships in the areas of social withdrawal, anhedonia and ‘slowness’ [21, 22]. The items of poor rapport, lack of spontaneity and difficulty in abstract thinking appeared to distinguish most successfully between negative and depressive symptoms. Studies using the SANS [22] indicate that affective flattening and poverty of speech appear to be most stable over time and may be the best representation with that instrument of ‘primary’ negative symptoms [37]. The MADRS items of reduced sleep, reduced appetite and possibly inner tension would appear to best discriminate depressive from negative symptoms. The items relating to sadness, suicidality and pessimism distinguish poorly, perhaps due to the distress caused by the impact of disabling negative symptoms on the lives of these patients. Thus, there are clearly areas of considerable rating overlap and areas of rating distinction detectable when one considers individual rating items rather than total scores. This indicates that there are genuine differing symptom components (negative and depressive) but these either fundamentally overlap in their phenomenology or the rating tools applied are unable to sufficiently distinguish between them in the common form in which these tools are applied.
From these results, it is apparent that the use of total negative symptom cluster scores to demonstrate therapeutic effect in clinical trials appears to be problematic with this symptom overlap now being reported with the PANSS, SANS and BPRS [21]. Unfortunately, this difficulty is not overcome by the use of the 5-factor pentagonal model for the PANSS. The negative symptom factor in this version of the PANSS includes the items of blunted affect, social and emotional withdrawal and passive/apathetic social withdrawal, all of which have significant correlations with multiple MADRS items. Two of these ‘high overlap’ items are also included in the withdrawal/retardation factor of the BPRS, probably explaining its failure to distinguish depressive and negative symptoms. The pentagonal negative symptom factor ‘excludes’ the items of difficulty with abstract thinking and stereotyped thinking which have limited correlations with depressive items. In the 5-factor model these items appear in the autistic preoccupation factor. The 5-factor pentagonal model also overlaps to a similar extent to the original 3-factor model with extrapyramidal side-effect ratings. The EPSE–negative symptom overlap is not accounted for by an effect of EPSE on depressive symptoms and appears to be an independent relationship.
Unfortunately, these difficulties would appear to indicate that the capacity of the PANSS, in either format, to discriminate between negative and depressive symptoms is limited. This has considerable clinical significance, for example in the interpretation of pharmacological trials in schizophrenia, where the PANSS is widely used. This would add weight to concerns about the small effect sizes seen with negative symptom treatment with most, if not all, therapeutic modalities. Treatment may only need to produce a moderate response in the lessening of depressive symptoms to be incorrectly seen to be having an effect on negative symptoms.
Conclusions
We found that depressed mood occurred frequently in our community and hospitalized sample of subjects with schizophrenia. A clear association was found between depressive and negative symptom scores and this was not lessened by the use of the pentagonal model as compared to the traditional 3-factor model. An association was also seen between depressive scores and positive symptoms but for outpatients only. A small association was found between parkinsonian and negative scale scores and between parkinsonian and depressive scale scores. The relationship between parkinsonian and negative scale scores was not lessened through the use of the pentagonal model. Depressive symptoms are an integral part of clinical presentation of schizophrenia. There are problems with the use of rating scales in the differentiation of depressive and negative symptoms either due to a true overlap in these symptoms or due to the nature of the scales, especially when grouping the item scores in the standard accepted formats. Depressive symptoms in schizophrenia warrant continuing systematic investigation, particularly in view of their relevance for clinical treatment and rehabilitation.
Acknowledgements
The authors wish to thank all the staff and patients of Dandenong Area Mental Health Service who participated in, or aided in the conduct of this study. Funding for the SCAP study has been provided by Eli Lilly, Australia.