1. Introduction
The classical, kraepelinian classification of mental disorders makes an explicit, fundamental, distinction between dementia praecox (schizophrenia) and manic–depressive psychosis (bipolar disorder). However, in clinical practice, some patients present a mixture of schizophrenic and affective signs and symptoms. This has led to the introduction in more recent nosologic systems, of a new diagnostic category – schizoaffective disorder. Theoretically, schizoaffective disorder could be considered as a form of schizophrenia, a form of bipolar disorder, an independent disorder or a disorder intermediate between bipolar disorder and schizophrenia, within a psychotic continuum in which differences are more quantitative than qualitative in nature [Reference Gooding and Tallent14].
Given the obvious importance of this concept for modern nosology, several studies have explored these possibilities, with the aim of validating the diagnostic categories. These studies have mostly been of four types: (a) clinical studies comparing the symptoms and/or outcomes (including medication response) of patients diagnosed with schizophrenia, schizoaffective or bipolar disorder; (b) epidemiological and genetic studies; (c) brain imaging (morphological and functional) studies; and (d) studies of cognitive performance in subjects with these diagnoses. Two recent reviews [20,22] analysed data from the first three types of studies but reached different conclusions. Lake and Hurwitz [Reference Lake and Hurwitz22] considered that data are in favour of the hypothesis of a single disease with a broad spectrum of severity. Ketter and co-workers [Reference Ketter, Wang, Becker, Nowakowska and Yang20] considered that a more complex, mixed, dimensional/categorical model could better account for the available data.
Surprisingly, in the two cited reviews data from the cognitive studies were not analysed. However, because cognitive dysfunctions in schizophrenia are considered core deficits of this disorder [Reference Seidman, Cassens, Kremen, Pepple and White37] and because they are the most useful characteristics in differentiating schizophrenic and normal subjects [Reference Heinrichs16], numerous studies explored if cognitive performances were also useful in differentiating subjects with different psychotic or affective disorders.
Several of these studies compared the cognitive performances of SZ and SZAff patients [3,5,13,14,27,40]. The results of these studies suggest that the cognitive deficits of SZAff patients do not differ significantly from those of SZ patients. However, in the absence of comparisons with subjects with bipolar disorder, no conclusions can be drawn from these results as to whether schizoaffective disorder is a form of schizophrenia or an intermediate disorder between bipolar disorder and schizophrenia.
Evans et al. [Reference Evans, Heaton, Paulsen, McAdams, Heaton and Jeste9] compared cognitive performances in three groups of subjects: SZ patients, SZAff patients and patients with affective disorders without psychotic symptoms. They observed similar deficits in SZ and SZAff patients, and these deficits were more severe than those in affective patients. Based on the results of one of their previous studies, showing similar deficits in SZ patients and in subjects with unipolar depression with psychotic symptoms [Reference Jeste, Heaton, Paulsen, Ercoli, Harris and Heaton18], they suggested that there may be a cognitive deficit common to the “psychotic spectrum”, regardless of the specific diagnosis.
Glahn and colleagues [Reference Glahn, Bearden, Cakir, Barrett, Najt and Serap Monkul11] recently compared working memory performances in five groups of subjects: SZ, SZAff, BDP, BD and NC. In this study, no significant differences in performance were observed between SZAff and BDP patients or between SZ and SZAff patients. The lack of significant differences between SZ, SZAff and BDP may be due to the existence of a “psychotic spectrum”, as suggested by Evans et al. [Reference Evans, Heaton, Paulsen, McAdams, Heaton and Jeste9], but may also reflect a lack of statistical power due to small sample sizes (15 SZ, 15 SZAff and 11 BDP).
Based on these data, schizophrenic and schizoaffective subjects were pooled together in some studies of cognitive function [8,21]. In other studies [Reference Gilvarry, Russell, Jones, Sham, Hemsley and Murray10], patients with psychotic disorders (according to the DSM classification) and patients with affective disorders presenting psychotic symptoms were pooled together. However, the hypotheses underlying this pooling have not yet been formally tested.
In the present study, our aim was to investigate whether the results of the cognitive assessment supported the current diagnostic classifications and especially the existence of schizoaffective disorder as an entity separate from schizophrenia and also from bipolar disorder. To answer this question, we compared the performances of subjects with schizophrenia, schizoaffective disorder, bipolar disorder with psychotic features and bipolar disorder without psychotic symptoms in two tests of executive functions: the Wisconsin Card Sorting Test (WCST) and the Trail-making Test (TMT). We chose to explore executive functions because several studies and meta-analyses suggested that this is one of the most impaired cognitive domains both in schizophrenic and euthymic bipolar subjects [17,34]. We decided to use data from these two tests because they have been the most frequently used tests to explore executive functions in SZ [Reference Heinrichs and Zakzanis17] and among the most used in bipolar patients [Reference Robinson, Thompson, Gallagher, Goswami, Young and Ferrier34]. For this reason, the manner in which different diagnostic groups could be pooled together is of a particular importance. Furthermore, a previous study [Reference Yurgelun-Todd and Kinney42] suggested that, although schizophrenic subjects have impaired performances in both these tests, this impairment is probably the consequence of independent etiological factors.
2. Methods
2.1. Subjects
Consecutive patients meeting DSM-IV [1] criteria for bipolar disorder, schizoaffective disorder or schizophrenia were recruited at two university-affiliated hospitals (the Pitié–Salpétrière and Albert Chenevier hospitals, Paris). They were included in the study just before discharge. To confirm the diagnosis of bipolar disorder, schizoaffective disorder or schizophrenia subjects were interviewed by an experienced psychiatrist with the French version of the Diagnostic Interview for Genetic Studies (DIGS) [29,32]. Based on the information provided by the DIGS concerning the life-time occurrence of psychotic symptoms (delusions and hallucinations), bipolar subjects were classified as bipolar with psychotic signs and symptoms or bipolar without psychotic signs and symptoms (life-time). For all patients, additional information, if required, was obtained from the patient's medical file.
The control subjects were blood donors at Pitié–Salpétrière Hospital. These subjects were included after being interviewed with the DIGS and the Family Interview for Genetic Studies (FIGS) [Reference Maxwell25] to confirm the absence of a personal and family history of DSM-IV axis I or II disorders.
For inclusion in this study, all subjects had to be normothymic, as evaluated by the Montgomery and Asberg Depression Rating Scale (MADRS) [Reference Montgomery and Asberg28] and the Bech and Rafaelsen Mania Rating Scale (MAS) [Reference Bech, Bolwig, Kramp and Rafaelsen4]. All patients with scores above the thresholds for clinically significant depression (MADRS = 15) or mania (MAS = 6) were excluded from this study. Patients were also required to be in a stable state, with no change in medication or symptoms for at least two weeks before cognitive evaluation. They were included only if they were aged between 18 and 60 years, and had no history of neurological disease or current substance abuse.
The research ethics board of Salpêtrière Hospital reviewed and approved the study. The study was described in detail to the subjects, who then gave written informed consent for participation.
2.2. Cognitive assessment
We used the classic form of the WCST [Reference Heaton15], with four stimulus cards differing in three characteristics: colour (yellow, green, red, blue), shape (triangle, star, cross, circle) and number (1–4), and two identical sets of 64 response cards. The test was discontinued after the completion of six categories or when no more response cards remained.
The two measures most often used to assess WCST performance are the number of categories completed and the number of perseverative errors. However, as the number of categories shows an important ceiling effect, we analysed only the number of perseverative errors. This score is considered a measure of cognitive flexibility.
The TMT [Reference Reitan and Wolfson33] is a pencil and paper test assessing psychomotor speed, attention and set alternation. Part A requires the subjects to connect 25 consecutively numbered circles as quickly as possible. In Part B, the subjects have to connect 25 consecutively numbered and lettered circles by alternating between the two sets. The time taken to complete Parts A and B of the TMT is recorded in seconds. As set alternation is required only in Part B, the time taken to complete Part A is subtracted from the time taken to complete Part B to eliminate performance variation due to psychomotor speed [Reference Lezak24] and the resulting value is used as a measure of executive function.
2.3. Statistical methods
We compared the demographic characteristics of the five groups of subjects by analysis of variance (ANOVA) for age, and by the Chi-square test for education level and sex.
The influence of diagnosis on executive performance was assessed by multiple regression, after adjustment for all the demographic variables for which significant differences between groups had been identified. Bilateral comparisons between groups were carried out using contrasts. In order to reduce type I errors due to multiple tests we used for contrasts the Bonferroni correction for 10 comparisons, resulting in the threshold for significance being set at p = 0.005. All analyses were carried out with the SAS® V9 package.
3. Results
We included 214 subjects, of whom 48 were SZ, 26 SZAff, 40 BD, 52 BDP and 48 were normal controls (NC). The demographic characteristics of these subjects are summarized in Table 1. The highest completed school grade (education level) was recorded, according to usual conventions [Reference Pichot, Lebeaux, Penhouët and Simon31] as a trichotomous variable (1 = elementary school; 3 = at least high school completed; and 2 = intermediate between 1 and 3). However, we finally decided to pool the first two levels together because level 1 was infrequent in all groups.
Demographic characteristics of the five groups of subjects
SZ = subjects diagnosed with schizophrenia, SZAff = subjects diagnosed with schizoaffective disorder, BDP = subjects diagnosed with bipolar disorder with psychotic symptoms, BP = subjects diagnosed with bipolar disorder without psychotic symptoms and NC = normal controls.
Significant differences were found between the groups for all demographic characteristics. We therefore adjusted for all these variables when comparing cognitive performances between the diagnostic groups.
The results of cognitive assessment, after adjustment for differences in demographic characteristics, are summarized in Table 2 and Figs. 1 and 2.
Cognitive performances in the five groups of subjects before and after adjustment for differences in age, sex and education
SZ = subjects diagnosed with schizophrenia, SZAff = subjects diagnosed with schizoaffective disorder, BDP = subjects diagnosed with bipolar disorder with psychotic symptoms, BD = subjects diagnosed with bipolar disorder without psychotic symptoms and NC = control subjects.
a Least square means (adjusted for age, sex and education level) ± standard error.
Perseverative errors in the WCST in the five groups (SZ, SZAff, BDP, BD, NC).
Difference between parts B and A of the TMT in the five groups (SZ, SZAff, BDP, BD, NC).
Significant differences were found between the groups for the two executive function measures, following adjustment for age, sex and education level. For the TMT, the performances of SZ and SZAff patients were very similar, as were those of BD and BDP patients (Table 2 and Figs. 1 and 2). A different pattern was observed for the WCST, with the best performances obtained from BD patients, followed by BDP patients, SZAff patients and finally SZ patients.
Diagnostic and education levels significantly (p < 0.05) influenced both cognitive variables. Age significantly influenced the TMT (p = 0.014) but not the WCST PE (p = 0.16), whereas sex had no significant effect on either score (p > 0.7 for both).
Contrast analysis showed that both SZ and SZAff patients had significant deficits, for both tests (significant differences between these groups and normal controls after Bonferroni correction for 10 comparisons). The only significant differences between the four pathologic groups were for perseverative error scores (in the WCST) between SZ and BD patients and between SZ and BDP patients. Differences between BP or BDP and controls did not reach the significance threshold after Bonferroni correction, for either of the tests.
The patterns of results suggest that for one measure (TMT scores) SZAff subjects resemble SZ subjects and are different from bipolar subjects. However, for the other measure there is a continuum of severity with deficits gradually increasing from BD to BPD, to SZAff and finally to SZ.
4. Discussion
To explore the validity of the current nosological boundaries, in this study we compared the performances of five groups of subjects: SZ (N = 48), SZAff (N = 26), BDP (N = 40), BD (N = 52) and NC (N = 48) in two tests of executive function (the WCST and the TMT). To our knowledge this is the largest study, to date, that analysed the cognitive performances in these five groups of subjects. The two tests gave different patterns of results supporting the assumption [Reference Yurgelun-Todd and Kinney42] that impairment in these tests is influenced by different, independent, factors but also suggesting that continuous and discrete differences between these nosological entities may coexist.
For one of the tests, the TMT, subjects with psychotic disorders (SZ and SZAff) obtained very similar results as did the two subgroups of bipolar patients. However, there was a clear difference between the results of psychotic and affective patients. This finding suggests that, for some characteristics, schizoaffective disorder could be considered as a form of schizophrenia and different from affective disorders. Grouping SZ and SZAff and also BDP and BD patients could be therefore acceptable.
However, a different pattern of results was observed for the perseverative errors on the WCST. For this test, results suggest that there may be a continuum of severity from SZ to SZAff to BDP and finally BD. The absence of significant differences between SZAff and both SZ and BDP could be due to lack of power and other studies in larger samples must be realised before a clear answer emerge. However, these results also suggest that pooling SZ and SZAff may not always be appropriate. Moreover, the significant difference in the number of perseverative errors between SZ and BDP patients clearly demonstrates that pooling together all patients with psychotic symptoms is not justified for this test.
A rigorous comparison of our results with published data is not possible because, to our knowledge, no other study has analyzed the same cognitive variables in the same diagnostic groups. We therefore assessed whether our results were in line with previous findings and could be generalized by calculating standardized differences in our samples and comparing them with those for WCST PE (for the TMT B − A difference, no data were available) reported in three published reviews. In these reviews, results were compared between SZ and NC [Reference Laws23], SZ and BP [Reference Daban, Martinez-Aran, Torrent, Tabares-Seisdedos, Balanza-Martinez and Salazar-Fraile6] and BP and NC patients [Reference Robinson, Thompson, Gallagher, Goswami, Young and Ferrier34]. For SZ vs NC and for SZ vs BP (both with and without psychotic symptoms) the standardized differences (Cohen's d) in our study (0.92 and 0.57, respectively) were well within the range of previous reports. The difference observed in our study between BP and NC (0.45) was slightly smaller than those reported in the cited review (0.54–1.48). However, when the heterogeneity of all reported BP vs NC differences (including our own) was assessed, our results were found not to be significantly different from those previously reported (Q statistic 7.29, p = 0.29). So, overall, when comparisons were possible, our results were found to be consistent with those previously reported.
Overall, our data suggest that, even for tests that are supposed to explore the same broad cognitive domain (in this case executive), comparing performances of SZAff with those of SZ and bipolar subjects may lead to different patterns of results. This, in turn, suggests, in line with the conclusions reached by Ketter et al. [Reference Ketter, Wang, Becker, Nowakowska and Yang20], that various theoretical views concerning schizoaffective disorder are not necessarily mutually exclusive and that some characteristics may be viewed as specific to psychotic disorders (i.e. SZ and SZAff) whereas others lie on a dimensional continuum between affective disorders, schizoaffective disorder and schizophrenia. This more complex model is compatible with the results of the study of Glahn and colleagues [Reference Glahn, Bearden, Cakir, Barrett, Najt and Serap Monkul11] that suggested that other cognitive deficits, not explored in our study, are common to all groups of subjects presenting psychotic symptoms (SZ, SZAff, BPD) but not to non-psychotic BP.
The specific cognitive deficits involved in the performance patterns observed in our study remain, at this stage, speculative. Both of the tasks used are considered to test executive function and are influenced by numerous cognitive (executive and non-executive) processes. The two variables used in our study are thought to reflect more specific executive processes. Although both variables are considered to reflect cognitive flexibility and working memory, there are also some important differences that might account for different patterns of responses observed in our study. These differences are, at least in part, due to differences in the nature of the two tests. One of these tests (the WCST) is a performance (power) test and the other (the TMT) a timed (or speed) test, and this is reflected in the measures obtained. On one hand, a good performance in the WCST depends on the capacity to recognize and to solve a problem and to initiate change; on the other hand, for the TMT measure used, the speed of mental operations is probably the most important factor [Reference Arnett and Labovitz2].
Our results should be considered in the context of our study's limitations. In particular, our analyses did not take into account some variables that might affect cognitive performance, such as treatment and clinical symptoms.
It would have been impossible to carry out a comprehensive assessment taking into account the effects of treatment, due to the large number of substances prescribed and uncertainties about their effects on cognitive performance. Most of the available data suggest that, after the initial, acute period, classic and atypical antipsychotic agents do not impair cognitive performances in schizophrenic patients and, indeed, that they may even enhance them [19,38,41]. It seems therefore highly unlikely that the observed pattern of performances (with worst results in psychotic subjects) results from differences in antipsychotic prescriptions. Mood stabilizers, including lithium salts [6,30] and anticonvulsant drugs [Reference Meador26] in particular, may have a negative effect on cognition. This would theoretically influence differences between the affective groups and the schizophrenic group, and should be explored in further studies. This issue is further complicated by the potentially different cognitive effects of medication in different clinical groups. For example, the use of antipsychotics has been associated with poorer cognitive performance in BP subjects [Reference Donaldson, Goldstein, Landau, Raymont and Frangou7], and very mild and selective negative effects of lithium have been described in normal subjects [Reference Stip, Dufresne, Lussier and Yatham39].
The effects of symptoms were minimized by including only patients in a stable and euthymic state. However, although the cognitive performances of schizophrenic [Reference Rund35] and bipolar patients [Reference Glahn, Bearden, Niendam and Escamilla12] are considered stable after the initial period of the disorder and outside acute episodes, we could not completely exclude an effect of symptoms on cognitive measures.
The lack of longitudinal data is also a limitation of the present study. This is an important issue as the diagnosis of schizoaffective disorder has been shown to be unstable over time [Reference Schwartz, Fennig, Tanenberg-Karant, Carlson, Craig and Galambos36] and the conversion of BD to BDP over time is possible. However, in the only study to date to explore cognitive differences between SZ and SZAff patients over time [Reference Preisig, Fenton, Matthey, Berney and Ferrero32], Stip et al. [Reference Stip, Sepehry, Prouteau, Briand, Nicole and Lalonde40] found differences that remained stable over a two-year period. Furthermore, the potentially changing boundaries between the two psychotic disorders and the two subgroups of BP patients may account for the results obtained in the TMT, which were similar for SZ and SZAff and for BD and BDP patients, but not for the results obtained in the WCST or differences in the patterns of results for the two tests.
Another limitation of our study is the use of only two cognitive tests. The use of a larger battery of tests measuring more diverse cognitive abilities is needed in order to confirm the existence of these patterns and/or suggests other relationships between cognitive performances in subgroups of psychotic and affective subjects.
To address these limitations further studies with larger samples of subjects, more comprehensive batteries of cognitive tests and a better assessment of confounding factors are needed.
5. Conclusion
Overall, the results of this study suggest that more complex models, combining quantitative factors and qualitative differences, might better account for the cognitive deficits in patients with affective and psychotic disorders. This study also suggests that, in the absence of clear, test-specific, evidence to the contrary, the pooling of cognitive data from different diagnostic populations is not justified.
Acknowledgments
This research was supported by grants from INSERM and “Délégation à la Recherche Clinique de l'AP-HP” (PHRC AOM 98152). A.S. and A.M. received grants from the “Fondation pour la Recherche Médicale”. We thank M.J. Pereira-Gomes and E. Abadie for technical assistance.
Comments
No Comments have been published for this article.