|Home | About | Journals | Submit | Contact Us | Français|
It has recently been suggested that cognitive impairment should be included in the diagnostic criteria of schizophrenia. One of the main arguments in support of this suggestion has been the hope that cognitive impairment can help distinguish schizophrenia from bipolar disorder (BD). However, recent evidence shows that cognitive deficits occur in BD and persist beyond euthymia. Further, mood disorders with psychotic features might be expected to manifest greater cognitive impairment, which further complicates the potential to differentiate these disorders. The goal of the current meta-analysis was to examine the magnitude and characteristics of cognitive impairments in affective psychoses (AP). A systematic search of the existing literature sourced 27 studies that met the inclusion criteria. These studies compared cognitive performances of 763 patients with AP (550 BD and 213 major depressive disorder) and 1823 healthy controls. Meta-regression and subgroup analyses were used to examine the effects of moderator variables. Meta-analyses of these studies showed that patients with AP were impaired in all 15 cognitive tasks with large effect sizes for most measures. There were no significant differences between the magnitude of impairments between the BD and major depressive disorder groups. The largest effect size was found for symbol coding, stroop task, verbal learning, and category fluency, reflecting impairments in elementary and complex aspects of attentional processing, as well as learning and memory. In general, the pattern of cognitive impairments in AP was similar to reported findings in euthymic patients with BD, but relatively more pronounced.
Cognitive impairment is one of the cardinal features of schizophrenia with affected patients typically performing 0.8–1.5 SDs below the level of healthy controls across many cognitive domains.1–4 According to the results of several meta-analyses, the most pronounced cognitive deficits in schizophrenia are in digit symbol coding, verbal learning, and general IQ (each with large to very large effect sizes).1–4 Cognitive deficits are prevalent in schizophrenia with up to 80% of patients being cognitively impaired.5 More importantly, cognitive deficits are clinically relevant because they are considered to contribute to social impairments and treatment outcome in schizophrenia.6,7 Recently, severe cognitive impairment was proposed to be part of the diagnostic criteria for schizophrenia in the Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) (DSM-V) and International Classification of Diseases, Eleventh Revision (ICD-11).5
One of the arguments raised in favor of including cognitive impairment within the diagnostic criteria for schizophrenia is the potential that it can help to discriminate schizophrenia from affective psychoses (AP).5 However, there is also evidence for persistent cognitive deficits in bipolar disorder (BD), which complicates the potential utility of cognitive impairment to differentiate between these disorders.8,9 For example, several meta-analyses have recently examined cognitive deficits in euthymic BD and found medium to large effect size impairments for several domains that have also been found to be impaired in schizophrenia including response inhibition, verbal memory, category fluency, set-shifting ability, and sustained attention.10–12 Still, some authors suggest that the severity and frequency of cognitive impairment in schizophrenia can differentiate between these 2 disorders. For example, it has been suggested that there are differing patterns of cognitive deficits in the major psychoses, with deficits being less persistent and variable in BD compared with schizophrenia.5,13
Psychotic symptoms are one of the factors that are likely to contribute to the cognitive deficits observed in mood disorders. A single meta-analysis confirmed that patients with psychotic depression are cognitively more impaired compared with their nonpsychotic counterparts.14 Studies that investigated the effect of psychotic features on cognitive deficits in BD found less consistent findings: while some studies reported more severe cognitive deficits in working memory, verbal memory, and executive functions, others did not find a between-group difference.15–18 There is emerging evidence indicating that cognitive deficits are quite common in AP. For example, a recent study reported that 84% of the schizophrenia patients, 58.3% of psychotic major depression (MDP) patients, and 57.7% of psychotic bipolar patients were cognitively impaired (1 SD below healthy controls in at least 2 domains).19 Moreover, a meta-analytic study comparing the cognitive performances of schizophrenia and BD found that schizophrenia patients scored approximately 0.5 SD lower than BD patients.20 However, BD samples in these studies included a mixture of psychotic and nonpsychotic patients. A recent meta-analysis concluded that differences between schizophrenia and AP (psychotic depression and psychotic bipolar disorder [BDP]) might be even smaller.21 Furthermore, results of this meta-analysis suggested that cognitive differences between schizophrenia, schizoaffective disorders, and AP are driven by a subgroup of schizophrenia patients who are predominantly male and have severe negative symptoms.21
To our knowledge, meta-analytic methods have not been used to estimate the magnitude and pattern of cognitive impairments in AP compared with healthy controls. This may be relevant for the discussions related to inclusion of cognitive impairment criteria for schizophrenia and other psychiatric disorders. The goal of the current meta-analysis was to examine the magnitude and characteristics of cognitive impairments in AP (MDP and BDP). Additional aims were to examine the effect of potential moderator variables on cognitive performance of patients with AP, including clinical variables (duration of illness, age of onset, and symptomatic status of the sample) and key demographic variables (gender, age, and education).
Potential articles were identified through a literature search in PubMed, Scopus, Psychinfo, and EMBASE in the period between 1987 and May 2009. For the literature search, combinations of the following keywords were used: bipolar disorder, manic-depress*, mania, depress*, psychosis, affective psychosis, cognit*, neuropsycholog*. The last 2 terms were also replaced with the keywords describing cognitive domains (attention, executive, or memory). The reference lists of the published articles were also cross-referenced and reviewed.
The following criteria were used to select studies for review: (1) assessed cognitive abilities using reliable neuropsychological testing methods and were published in English peer-reviewed journals (2) reported test scores (means and SDs) of both groups or other data to calculate effect sizes for group comparisons (3) compared adult patients with AP (depression or BD with current psychotic symptoms or history of psychotic symptoms during mood episode) (ICD-9, ICD-10, DSM-III, DSM-III-R, or DSM-IV) with healthy controls. If there were several relevant articles based on a single sample, studies with a larger sample were selected. Studies that examined cognitive deficits of late-onset depression were excluded because the pathophysiology of late-onset depression is probably different from young-onset cases.
In addition to cognitive data, we recorded information concerning the sample characteristics and potential moderator variables. The clinical variables included diagnosis, age of onset, and duration of illness. We also coded whether samples included first-episode (FE) or multi-episode patients. Strictly euthymic patients were coded as well. Demographic variables included means and SDs of age and years of education for both groups. Gender was coded as percentage of males. We also coded whether the studies included a schizophrenia control sample in addition to healthy subjects because one could hypothesize that sites conducting research only in affective patients could misdiagnose some schizophrenia patients as psychotic mood disorder patients. Further, we coded the studies that included an affective control group without a history of psychosis. We made an attempt to code variables related to medication and positive and negative symptoms; however, because very few studies reported comparable data, we did not use these data for further analyses.
More than 1000 articles were identified through the database searches. In the first instance, abstracts and titles of articles were reviewed to identify relevant studies and eliminate the studies that obviously failed to meet inclusion criteria. This resulted in 101 articles that were examined further. Of these, 43 studies investigated the cognitive profile of AP. Thirteen further studies were excluded because they did not provide sufficient data to calculate effect sizes, did not include a healthy control group, or investigated cognitive deficits of late-life depression. The remaining 30 studies were checked for sample overlap. Three studies were excluded because they were based on the same sample used in other studies. This resulted in 27 studies included in the analysis, which are summarized in table 1.15–18,22–44 In these studies, patients were diagnosed with DSM-III, DSM-III-R, DSM-IV, or ICD-10. In most of the studies, a structured clinical interview was used (table 1). Fifteen of these studies included BDP patients, 7 of these studies included MDP patients, and 5 studies included both BDP and MDP patients. Two of these 5 studies with mixed samples reported separate scores for BDP and MDP groups, so it was possible to obtain 17 BDP-control and 9 MDP-control comparisons. Seven of the studies included FE samples with psychotic mood disorders based on structural clinical assessment. Diagnoses of the FE patients were BDP31,32,38,42,43 and/or MDP.27,32,33,38,42 Psychosis was defined as psychotic symptoms during current/last mood episode in 14 studies and as history of past psychotic mood episodes in another 13 studies (table 1). Regarding current psychotic and mood symptoms, 11 AP studies included currently psychotic patients in acute mood episodes and 5 AP studies included currently nonpsychotic patients who were euthymic. In another 9 studies, the AP group included a mixture of symptomatic (depressive or manic) and euthymic patients and all these studies included some patients with current psychosis (table 1). There was no information about current psychotic or mood symptoms in 2 studies. Fifteen of these 27 studies also included BD or MD groups without a history of psychosis15–18,23,27,28,33,34,36,37,39–41,44; however, we only included the psychotic samples (current or history) in our meta-analysis.
Our primary goal was to analyze individual measures rather than broader cognitive constructs like executive functioning, because we aimed to compare cognitive impairment in AP with the findings in remitted BD and schizophrenia. Fifteen individual neuropsychological variables were included in the study. Similar to other meta-analyses in BD and schizophrenia, we combined data across very similar tests in some instances. For example, we combined different tasks of immediate verbal memory such as the Rey Auditory Verbal Learning Test and the California Verbal Learning Test. The same approach was used for delayed verbal memory tests. Different IQ measures were combined as well, and the digit symbol coding test and its variants were combined under the symbol coding measure. Table 2 lists the 15 cognitive measures we used in this meta-analysis and their references.45–54 We also summarized the findings for these tasks from other meta-analyses conducted in schizophrenia and remitted BD.1–4,10–12
However, we still combined individual tasks under 6 cognitive domains of the MATRICS (Measurement and Treatment Research to Improve Cognition in Schizophrenia) because there were not a sufficient number of studies to examine individual tests in BDP and MDP separately.55 We also calculated a summary measure, “global cognition,” by averaging the effect sizes from each cognitive domain.
For each cognitive test, an effect size and standard error was calculated. Effect sizes for each cognitive variable from each study were calculated as the mean difference between AP and healthy control performance divided by the pooled SD. Effect sizes were weighted using the inverse variance method. We used a random-effects model. When multiple tests were reported for one cognitive measure, one combined effect size was calculated for these tests. Whenever patients with AP performed more poorly than healthy controls, we reported between-group differences as positive effect sizes. Therefore, the effect sizes for the relevant variables were multiplied by −1. Homogeneity of the resulting mean weighted effect sizes was tested using the Q test. Publication bias was assessed by graphically examining funnel plots and using Egger test. Meta-analyses were performed using MIX software.56
The effects of moderator variables on observed between-group differences were analyzed by subgroup analyses and meta-regressions. We used subgroup analyses for examining the effect of diagnosis (BD or depression), absence/presence of a schizophrenia control group, FE/multi-episode samples, definition of psychosis, and current psychosis on the magnitude of patient-control differences. We also calculated homogeneity statistics Qw and Qbet. Qbet was used to test the significance of differences in effect size magnitude between moderator subgroups, analogously to the F statistic. Meta-regression analyses were used to estimate the impact of between-group differences of age (26 studies), gender (27 studies), and education (21 studies) on cognitive impairment in AP. The effects of age of onset (14 studies) and duration of illness (14 studies) were also examined using meta-regression. We were not able to analyze the effects of medication and symptoms on cognitive deficits because most of the studies did not report relevant data. Meta-regression analyses were conducted in SPSS 11.0. These weighted generalized least squares regressions were performed by using the macros written by David B. Wilson (http://mason.gmu.edu/~dwilsonb/ma.html). Meta-regression analyses performed with random-effects modelling were performed using the restricted-information maximum likelihood method with a significance level set at P < .05.
The 27 studies included in the meta-analysis compared the cognitive performances of 763 patients with AP (550 BDP and 213 MDP) and 1823 healthy controls. The patient and control groups were matched for age (d = 0.14, −0.05 to 0.33, z = 1.47, P = .14) and gender (48.4% of controls and 45.1% of the patients were males; RR = 1.08, CI = 0.99–1.19, z = 1.73, P = .08). There was a small but significant between-group difference for education. Specifically, patients were less well educated (d = 0.21, CI = 0.04–0.39, z = 2.35, P = .02) than controls.
There was no significant effect of diagnosis on any of the 6 cognitive domains and global cognition (table 3). Patient-control differences were not significantly different for MDP and BDP. Studies that had a schizophrenic control group were not more impaired than the studies that did not. The definition of psychosis did not significantly influence the magnitude of the cognitive deficit in AP (table 3). While FE samples were also impaired compared with controls, the magnitude of this deficit was significantly smaller than multi-episode patient-control differences. We excluded FE patients from subsequent individual task analyses because there were no sufficient data available.
In all the 15 analyses conducted for individual cognitive measures, AP patients performed more poorly than healthy controls (table 4). Effect sizes for 11 of these tasks were large (d > 0.8). A medium degree of impairment was found for 2 measures (visual memory and letter fluency), while the degree of impairment for premorbid IQ and current IQ was small. The largest effect size was found for symbol coding (d = 1.02).
The Q tests revealed significant heterogeneity for 7 of the tasks (stroop, symbol coding, Wisconsin card sorting test [WCST] perseveration/category, category fluency, verbal memory delayed/immediate). Several studies significantly contributed to heterogeneity. The study of Selva et al15 significantly contributed to heterogeneity for symbol coding, stroop, category fluency, and the WCST perseveration/category and after removal of this study, Q test results became insignificant for all these measures (d values after removal were 0.91, 0.91, 0.94, 0.80, and 0.74, respectively). The study of Basso et al. significantly contributed to heterogeneity of delayed and immediate verbal memory, and Q tests were nonsignificant after removal of this study (d values were 0.80 and 0.83). There was no evidence for a significant publication bias for any of the tasks.
When the between-group comparisons were restricted to BDP, effect sizes were very similar to the ones obtained for the combined AP group (table 5). To examine the effect of current mood and psychotic symptoms on cognitive impairment in AP, we compared the mean effect sizes of studies that included psychotic patients with acute mood episodes (d = 0.75, CI = 0.56–0.95, z = 7.6, P < .001) and nonpsychotic euthymic patients (d = 0.69, CI = 0.50–0.88, z = 7.1, P < .001). Current symptoms did not significantly influence the severity of the cognitive dysfunction (Qbet = 027, P > .05). We repeated this analysis by excluding FE patients because only the currently psychotic group included FE patients. This analysis confirmed that currently psychotic patients (d = 0.85, CI = 0.60–1.09, z = 6.8, P < .001) were not significantly more impaired than euthymic patients (Qbet = 1.41, P> .05).
Meta-regression analyses did not reveal any significant relationships between cognitive deficits and demographic confounders (age, gender, and education) except for a relationship between duration of education and 3 cognitive measures. The patients who had a shorter duration of education than controls had a more pronounced deficit in letter fluency (z = 2.38, P = .02), WCST categories (z = 2.31, P = .02), and the stroop (z = 2.23, P= .03). Differences in age of illness onset and duration of illness did not significantly contribute to the results.
To our knowledge, this is the first meta-analysis to examine the magnitude of cognitive deficits in AP. The current meta-analysis combined results from 27 studies consisting of 763 patients with AP and 1183 healthy controls and found large effect size impairments for most of the measures. Patients performed at least 0.8 SD lower than controls on 11 of 15 cognitive measures with the largest effect sizes being observed for symbol coding, stroop interference, and verbal learning and category fluency tasks.
In general, the pattern of cognitive impairment in AP was similar to findings reported in euthymic patients with BD, despite being minimally more pronounced.10–12 For some measures (eg, verbal and visual memory, stroop, WCST per, premorbid IQ), the observed between-group differences were about 0.2 SD higher than the results reported for BD-control comparisons. Digit symbol coding was relatively more differentially affected in AP compared with remitted patients with BD (1.02 vs 0.75). One reason for the minimal increased severity of cognitive impairment in AP might be explained by the influence of psychosis. Another reason might be the effect of symptoms on cognitive impairment in AP because many patients in these samples were not assessed in a remitted phase of the illness. This is an important issue because some authors believe that cognitive impairments in AP are nonpersistent symptoms of the illness.5 In this study, a preliminary analysis suggested that the magnitude of cognitive impairment was very similar in euthymic and currently psychotic AP. These findings are in line with previous meta-analyses that found substantial impairment in remitted BD patients.10–12 However, these findings do no rule out a significant role for symptoms in the cognitive impairment of psychotic depression because there are no controlled studies in patients with remitted major depression who also have a history of psychosis. Other factors like severity of illness and number of mood episodes could also contribute to cognitive impairment, but we were not able to examine the effect of these variables due to lack of reported data.57,58 Long-term effects of antipsychotics could also play a role because psychotic patients are likely to use these medications more frequently.
According to the results of our meta-analysis, cognitive deficits for many of the measures in AP are comparable with the impairments reported for schizophrenia.1–4 The severity of impairments in executive functions, sustained attention, and working memory are all similar to those observed in schizophrenia. This result could have implications for the inclusion of cognitive impairment in the diagnostic criteria for psychoses in DSM-V. However, there are other factors that should be considered in interpreting these findings. Most importantly, cognitive deficits could worsen in AP but not in schizophrenia during acute episodes and psychotic episodes. If schizophrenia patients differ from AP when they are nonpsychotic, this would suggest that these 2 diagnostic entities are somewhat discrepant with regard to cognitive impairment. Our previous meta-analysis, which examined studies comparing cognitive functioning in schizophrenia and AP, provide relevant data regarding this issue.21 That meta-analysis suggested that cognitive differences between schizophrenia and AP were very minimal in both inpatient and outpatient samples. However, further research is clearly needed to investigate the effect of symptomatic status on the observed differences between schizophrenia and AP.
Findings of the current meta-analysis also suggest that impairments in several cognitive domains in AP are not as pronounced as those in schizophrenia. For example, previous meta-analyses conducted in schizophrenia reported more substantial global intellectual impairment in schizophrenia. This result is consistent with studies that found IQ deficits to be apparent before the onset of psychosis in schizophrenia but not in mood disorders.59,60 Symbol coding tasks, a measure of information processing efficiency, were also more impaired in schizophrenia. This is interesting in light of a recent meta-analysis, which found that symbol coding impairment was the most severe cognitive deficit and that this information processing inefficiency may be the central feature of many of the higher order cognitive deficits observed in schizophrenia.3 These deficits might be related more to schizophrenia-specific brain abnormalities and, consistent with this idea, both low IQ and information processing inefficiency have been reported to be associated with gray matter abnormalities.61
One of the factors that might increase the heterogeneity of the findings in AP is the effect of diagnosis. AP groups consisted of a mixture of BDP and MDP patients. Our results indicate a similar level of cognitive impairment when results were analyzed separately in patients with BDP. Moreover, MDP-control and BDP-control differences were not significantly different. However, these results should be interpreted with caution because the number of studies that included patients with psychotic depression were relatively small. Studies that compare cognitive performance of patients with psychotic depression and BDP are important in order to show possible differences between these groups. To date, only a few studies compared the cognitive functioning of these groups, and they have not provided any evidence for between-group cognitive differences.19,62
The current study does not provide any information regarding longitudinal progression of cognitive deficits from early to chronic phases of illness. Only 6 of the studies included in the meta-analysis investigated cognitive impairment in FE patients with AP.27,31,32,38 Like the findings of Albus et al,63 these studies reported significant cognitive impairments in FE patients with AP. However, the magnitude of these deficits was significantly less than in that multi-episode patients. Very few studies have investigated cognitive deficits in individuals at high risk for psychosis, providing some evidence for cognitive deficits prior to the onset of psychosis.64,65 Previous cross-sectional and longitudinal studies in schizophrenia investigated cognitive impairments in different phases of illness from preonset to FE and chronic stages.66–68 A similar approach should be applied to cognitive research of AP and mood disorders. The outcome of this research is also relevant to examining the developmental model suggested by Murray et al because this model suggests a shared genetic predisposition (and shared neurobiological markers) between BD and schizophrenia and schizophrenia-specific additional neurodevelopmental changes.60
One of the limitations of the current meta-analysis was the lack of data about potential moderator factors in some studies. One of these factors was antipsychotic treatment, which can clearly influence the observed cognitive deficits in AP. On the one hand, antipsychotics could be expected to moderately improve cognitive deficits in AP because atypical antipsychotics have been shown to be moderately beneficial for cognitive deficits in schizophrenia.69,70 On the other hand, antipsychotics could be associated with poorer cognitive deficits functioning in AP: Patients with more severe prognosis and more severe and acute symptoms could be treated with larger doses of antipsychotics. In line with this possibility, some authors reported an association between poorer cognition and antipsychotics in BD.58,71 However, we were unable to examine the influence of this variable because very few studies reported between-studies comparable data. Also, age of onset and duration of illness are among a number of other variables that could have an impact on cognitive impairments in AP. We analyzed the effect of these variables on cognitive impairments and did not find any significant influence of these factors on cognitive deficits in AP. This result should be interpreted with caution because only some of the studies reported these variables.
The results of the current meta-analysis might have important implications regarding current discussions about the inclusion of cognitive impairment as a criterion in diagnostic criteria for schizophrenia.5 First, together with the meta-analyses in remitted BD, they show that cognitive deficits are characteristic features of nonschizophrenic psychoses. Some of these observed impairments are similar in severity compared with those observed in schizophrenia, and in some others, schizophrenia patients are moderately more impaired. These results argue against the view suggesting that inclusion of cognitive impairment criteria in the diagnosis of schizophrenia can increase our ability to differentiate between schizophrenia and BD. Our results could be interpreted as being in line with the hypothesized psychosis continuum between schizophrenia, BD, and depressive disorder.72,73 Second, results of these studies suggest that there might be deficits in some cognitive domains that can be relatively more specific to schizophrenia such as processing speed and global intelligence.
In conclusion, the findings of the current meta-analysis confirm that cognitive impairment is also a common characteristic of AP. This indicates the necessity of investigating cognitive deficits of AP in more detail as has been done in schizophrenia (ie, longitudinal and FE studies). Future research efforts regarding management of cognitive deficits should also target patients with AP because these deficits are likely to be associated with functional impairments.
National Health and Medical Research Council (NHMRC) Program Grant (ID: 566529 to C.P.). NHMRC Clinical Career Development Award (ID: 509345 to M.Y.). Melbourne Neuropsychiatry Centre to E.B.