Introduction

Bipolar disorder (BD) is a recurrent and highly disabling affective disorder characterised by mood swings as well as cognitive disturbances. Impairments in psychomotor speed, attention, working memory, long-term memory, and executive functions have been consistently reported both during acute episodes of depression and mania [1] as well as during euthymia, albeit in a milder form [2]. In contrast to extensive investigations of neurocognitive functioning in euthymic bipolar patients (BP), relatively few studies have characterised changes in cognition across different phases of the illness. A meta-analysis of cognitive functioning in euthymic, manic/mixed, and depressed BP revealed severe impairments across neuropsychological domains in all illness phases, with a moderate worsening of a subset of deficits in acute states [1]. Cross-sectional studies demonstrated that manic, mixed, and depressed BP perform worse (especially in verbal memory and executive functioning) compared to healthy controls (HC), whereas no or only marginal differences have been reported between depressed and (hypo-)manic BP [3–5]. This is surprising regarding the clinical presentation of cognitive patterns in the two phases: during mania, hyperactivity, thought racing, flight of ideas, disinhibited behaviour and pressure of speech predominate, whereas depression is typically accompanied by psychomotor slowing, fatigue, egocentric restriction of thinking, and ruminations, all of which impair cognitive speed. Several neuropsychological studies aimed to objectify cognitive dysfunctioning during acute episodes of BD and found similar deficits in attention, memory, and executive functions in bipolar and unipolar depression [6, 7], while cognitive deficits in acute mania seem to overlap with impairments found in psychosis [8, 9].

In addition, an emerging body of research demonstrates that cognitive disturbances in BD persist during remission [10], at least in a subgroup of BP [11–13]. However, up to now studies in euthymic BP are inconsistent and have failed to report cognitive deficits specific to euthymic BD. The identification of trait-related variables, however, would be helpful in studies on the pathophysiology of BD. Furthermore, stable trait markers could help to identify persons with high risk of the disease and thus could facilitate early diagnosis, prevention, and treatment options. Furthermore, cognitive disturbances that persist in the absence of clinical symptomatology may point to a genetically driven neurocognitive endophenotype [14]. Within-subject longitudinal studies across different mood states are the gold standard to determine whether cognitive deficits in BD are trait- or state-dependent. However, so far there is little systematic research on the longitudinal course of cognitive functioning in different clinical states in BD patients. Chaves et al. [15] investigated clinically stable but symptomatic BD outpatients in a pre- and post-testing over a period of 3 months. The authors assessed changes in mood symptoms in relation to cognitive test performance at baseline and follow-up. They found that BP showed persistent impairments in psychomotor speed and attention over time. Neuropsychological measures were not influenced by variations in affective symptoms with the exception of verbal fluency and verbal long-term memory, both of which were influenced by depressive symptoms. In contrast, Malhi et al. [16] investigated mood state-related cognitive deficits in BD over a period of 30 months across all three phases of the illness. Due to the small sample size, only four patients could be followed up in all three phases (depression, mania, and euthymia), and eight patients in two phases of the disorder. The authors found modest impairments in attention, memory, and executive tasks in depressed and manic patients, while cognitive functions were largely normal during euthymia, showing only subtle impairments in attention and memory. Similarly, a recent longitudinal study by Xu et al. [17] compared 223 bipolar patients and 293 unipolar depressed patients at baseline and after 6 weeks of treatment. In comparison with healthy controls (HC), bipolar and unipolar depressed patients were characterised by similar patterns of cognitive impairment (reduced speed, memory, verbal fluency, and executive functioning). During clinical remission, test performance in both patient groups improved, with the exception of deficits in speed and visual memory, which points to trait-like cognitive impairments in those domains. However, it should be noted that patients in this study were in a relatively unstable state as follow-up took place immediately after remission.

In summary, longitudinal studies on the course of cognitive impairments in BD are rare, inconsistent, depend on small samples and interpretation is limited due to methodological issues. The aim of the present study was therefore to detect changes in cognitive performance from acute episodes to euthymia in BD. We investigated BP suffering from acute depression or (hypo-)mania longitudinally (follow-up after at least 3 months of remission). Based on previous findings, we expected cognitive impairments in acutely ill BP compared to a sample of HC [1]. Specifically, we hypothesised that depressed BP would show psychomotor slowing and attentional deficits [7], while (hypo-)manic BP would be impaired in frontal-executive functions like working memory, cognitive flexibility, and response inhibition [8]. Regarding cognitive functioning over time, we assumed BP to improve during remission in all cognitive domains, but to still show mild cognitive disturbances compared to HC. Neuropsychological performance in HC should be stable over time. Furthermore, based on previous findings, we hypothesised that only a subgroup of remitted BP show cognitive deficits [11]. Finally, we were interested in the relationship between cognitive performance and demographic, clinical, and treatment variables of the bipolar sample.

Materials and methods

This study was performed at the Bipolar Disorder Program of the Department of Psychiatry, Psychosomatics and Psychotherapy, University Hospital Wuerzburg. All procedures followed the Declaration of Helsinki in its latest version and were approved by the ethical committee of the medical faculty at the University of Wuerzburg. Written informed consent was obtained from all participants.

Participants

A total of 55 inpatients with BD (diagnosed according to DSM-IV criteria by an experienced psychiatrist) and 55 HC participated in the study. Groups were matched for sex, age, and years of education. Participants were 18–55 years old and native German speakers. Intellectual abilities were estimated via the German multiple-choice word test (MWT-B) [18]. BP were initially assessed (baseline) during an acute episode of depression (N = 35), mania (N = 15), or hypomania (N = 5), all leading to inpatient treatment. Within the acutely depressed sample, 20 patients were diagnosed with BD type I and 15 with BD type II. Of the (hypo-)manic patients, 15 were diagnosed with BD type I and 5 with BD type II. Follow-up assessment took place the first time 3 months after discharge. Patients had to be euthymic for at least 3 months before the follow-up testing was conducted. If patients were not fully remitted at this 3-month mark, they were frequently contacted until they reported having been euthymic for at least 12 weeks. On average, follow-up took place after 21.1 ± 5.3 weeks. At both, baseline and follow-up, affective symptoms of BP were measured with clinical interviews and questionnaires. We applied the observer-rating scales Montgomery–Asberg Depression Ratings Scale (MADRS) [19], the Young Mania Rating Scale (YMRS) [20], and the self-rating questionnaire Beck Depression Inventory, Second Edition (BDI-II) [21]. Criteria for euthymia were rating scores of MADRS < 12, BDI-II < 15 and YMRS < 5 points. Furthermore, patients had to be on a stable medication for at least 4 weeks at follow-up to qualify for follow-up assessment. Exclusion criteria for BP were previous head trauma, neurological illnesses, schizoaffective disorder, or present substance abuse. Patients who had received electroconvulsive therapy (ECT) in the preceding 6 months were excluded as well. HC (N = 55) were recruited via Web-based announcements. They were screened for history of DSM-IV Axis I disorder by the Mini International Neuropsychiatric Interview, German Version 5 [22]. Furthermore, persons with a history of substance abuse/taking of psychotropic drugs and neurological diseases were excluded. Similar to BP, HC were tested in an initial and a follow-up assessment of approximately 3-month interval (13.3 ± 1.3 weeks).

At the follow-up testing, 26 of 55 BP could not be retested in a euthymic mood state for the following reasons: 7 BP (27 %) did not reach full remission during the full observation period, 9 BP (35 %) had relapsed, 6 BP (23 %) were lost during follow-up, and 4 BP (15 %) did not wish to participate at follow-up. Comparison of completers and non-completers revealed that patients who did not take part again were characterised by a higher number of episodes in their life (t = −1.79, p = .079), especially of depressive phases (t = −2.34, p = .024). Furthermore, they displayed higher MADRS scores at the initial assessment (t = 1.90, p = .063), indicating a higher load with depression. Groups did not differ in other clinical variables.

Clinical assessment

The following clinical characteristics (lifetime presence) of BP were recorded via structured interviews with the patients, their relatives, and clinical records: number of episodes (depression, mania/hypomania, mixed episodes), bipolar subtype, age at illness onset, number of hospitalisation, polarity of the first episode, attempted suicides, substance abuse, psychotic symptoms, medication, ECT, family history of mental disorders, and comorbid somatic (e.g. thyroid diseases) or mental illnesses, especially anxiety disorders and attention deficit disorder (ADHD). Patients were asked whether they had persistent sleep disturbances (initial or middle insomnia) in the week preceding the interview. Additionally, MADRS item 4 (reduced sleep) was analysed separately in order to assess sleep disorder in our sample. Furthermore, current mood of all participants was recorded using the self-report Positive and Negative Affect Scale (PANAS) [23] prior to neuropsychological testing. In order to assess psychosocial functioning, all patients were evaluated by the clinician with the Global Assessment of Functioning Scale (GAF) [24].

Neurocognitive assessment

We conducted a 60-min neuropsychological testing (including a 5-min break). Five standardised tests were administered in a fixed order by the same trained, experienced clinical psychologist (JV). Except the California Verbal Learning Test, all tests were computerised and parallel forms were used for the follow-up testing. Tests were selected based on evidence of sensitivity to cognitive deficits in bipolar disorder (see Table 1).

Table 1 Neuropsychological tests
Full size table

Statistical analysis

All statistical tests were performed using the Statistical Package for Social Sciences (SPSS), version 22.0 (IBM Corporation, Armonk NY, USA). For all tests, the significance level was set at p < .05 and Bonferroni–Holm correction for multiple testing was applied. Chi-square tests and analyses of variance (one-way ANOVAs) were used to compare demographic variables and current mood by group. In case of a significant F test, post hoc t tests were calculated. Neurocognitive and clinical variables were checked for outliers and normal distribution (Kolmogorov–Smirnov test). Given that all test measures (except compatible trials of Stroop Test) were not normally distributed, we conducted nonparametric tests. We used the Kruskal–Wallis one-way analysis of variance to test whether the three groups [depressed BP, (hypo-)manic BP, and HC] differed from each other on the baseline assessment. In case of significant group differences, we conducted post hoc Mann–Whitney U tests.

In a subsequent step, Wilcoxon signed-rank tests were used to check for differences in the repeated measurements of each group. In order to check the test–retest reliability of the neuropsychological tests, we examined performance of HC over time and calculated Pearson correlation coefficients for each test between the initial and follow-up assessments. Test–retest reliability was considered good if baseline and follow-up were significantly correlated (p < 0.01).

Furthermore, we compared the clinical characteristics of completers and non-completers (dropout analyses) using Student t tests. Finally, exploratory correlational analyses (Pearson coefficients) were conducted to assess associations between test performance and demographic/clinical and treatment variables of BP. In order to control the influence of age on cognitive measures, we calculated partial correlations with age as the covariate. We did not apply a correction of p values in this exploratory analysis, because our sample of euthymic BP was very small, and the correlational analysis was not the main hypotheses of the paper.

Results

Sample characteristics

Demographic characteristics and current mood at baseline and follow-up of depressed/(hypo-)manic BP and HC are presented in Table 2. There were no group differences regarding age, gender, verbal IQ, and years of education between groups. Mood ratings (measured by PANAS score) at the initial assessment showed a significantly reduced positive and increased negative affect in depressed patients compared to (hypo-)manic patients and HC. At follow-up, BP showed reduced positive and increased negative affect compared to HC, despite being euthymic.

Table 2 Demographic characteristics of depressed and (hypo-)manic bipolar patients (BP) and healthy controls (HC)
Full size table

Depressed and (hypo-)manic BP did not differ from each other in their drug regimen with the exception of antidepressant add-on therapy, which was prescribed significantly more often in acutely depressed patients (see Table 3). Overall, 11 (37.9 %) of patients had an altered drug regimen at follow-up as compared to the initial measurement. HC were completely medication free. At follow-up, patients reported being in a euthymic state over a period of 15.4 (SD 2.8) weeks. Mood ratings in the acute and remitted sample are demonstrated in Table 3. GAF scores indicated a marginal reduced psychosocial functioning in euthymic BP (M = 76.5; SD 9.9).

Table 3 Mood ratings and medication of depressed and (hypo-)manic bipolar patients at baseline (acute) and follow-up (euthymic)
Full size table

Comparison of cognitive performance between acutely ill bipolar patients and healthy controls (baseline testing)

Acutely depressed and (hypo-)manic BP showed a significantly reduced test performance compared to HC in the test measures. However, some group differences did not reach level of significance after the correction of the p value according to Bonferroni–Holm method for multiple testing. The results in Table 4 demonstrate that depressed BP showed a significantly reduced psychomotor speed and a lower response inhibition (reading colours, Stroop Test) compared to HC. (Hypo-)manic BP showed a significantly reduced performance in Working Memory. Both patient groups were characterised by more omission errors in the test assessing divided attention (trend). No significant differences in total verbal learning (CVLT) and Cognitive Flexibility were observed between depressed, (hypo-)manic BP, and HC (for more detail, see Table 4). In the acutely admitted bipolar sample, 51 BP (92.7 %) showed cognitive deficits (cut-off score: at least one test measure 1.5 SD below the average of normative data group), while only four (depressed) BP (7.3 %) showed a test performance in the normal range of the normative data group. These four patients without cognitive deficits differed not in clinical or demographic variables compared to acutely ill patients with cognitive deficits.

Table 4 Group analyses of cognitive performance (baseline) in acutely depressed, (hypo-)manic bipolar patients and healthy controls
Full size table

Comparison of cognitive performance between euthymic bipolar patients and healthy controls (follow-up testing)

At follow-up, euthymic BP differed significantly from HC in Working Memory and delayed verbal memory. Although we found marginal reduced performances in the other test measures in euthymic BP compared to HC, differences did not reach level of significance (see Table 5). In comparison with the normative data of each test, 17 (58.6 %) euthymic BP performed below the average in at least one test measure (cut-off score: 1.5 SD below the average of normative data group), while 12 BP (41.4 %) showed no cognitive deficits at all. The extent of cognitive deficits during remission was not associated with the test performance during the acute episode.

Table 5 Group analyses of cognitive performance (follow-up testing) in euthymic bipolar patients and healthy controls
Full size table

Change in cognitive performance (baseline to follow-up) in bipolar patients and healthy controls

Differences in repeated measurements of each group were analysed using the Wilcoxon signed-rank test. Depressed BP showed an improvement from baseline to follow-up in the compatible trials of the Stroop Test Reading (Z = −2.98, p = .003) and Naming (Z = −2.49, p = .013), a marginal improvement in the immediate verbal recall (Z = −1.71, p = .088) and the Stroop Test Naming (Z = −1.77, p = .077). (Hypo-)manic patients showed an improvement in the Working Memory task (Z = −2.67, p = .008) and the Stroop Test Naming (Z = −2.07, p = .038). A detailed table of the test statistics of changes in cognitive performance can be found in the supplementary material (table S1). Figure 1 illustrates z score transformation of the test scores (healthy controls as reference) of acutely depressed, (hypo-)manic, and euthymic patients in order to facilitate comparison of the cognitive profile of the different clinical states of illness.

Fig. 1
figure 1

Mean z score transformations of test scores of depressed, (hypo-)manic, and euthymic bipolar patients (z scores were calculated using mean and SD values of sample of healthy controls) (*significant differences after Bonferroni–Holm correction; for details, see Tables 4 and 5)

Full size image

Performance of HC improved in the same tests as BP, namely the compatible trials of the Stroop Test (Z = −4.82, p < .001), the immediate recall of the CVLT (Z = −3.16, p = .002), and the Working Memory task (Z = −2.09, p = .036). This indicates learning and practice effects in these subtests. Thus, we calculated test–retest reliability by Pearson correlations, which revealed that with the exception of Cognitive Flexibility, initial and follow-up performance scores were highly correlated, indicating good test–retest reliability for the compatible trials of Stroop Test Reading (r 12 = .887, p < .001) and Naming (r 12 = .815, p < .001), Divided Attention (r 12 = .490, p < .001), CVLT total verbal learning (r 12 = .577, p < .001), CVLT immediate recall (r 12 = .664, p < .001), CVLT delayed recall (r 12 = .579, p < .001), Working Memory (r 12 = .556, p < .001), Cognitive Flexibility (r 12 = .152, p = .246), and the incompatible trials of the Stroop Test Reading (r 12 = .654, p < .001) and Naming (r 12 = .531, p < .001).

Association of cognitive performance and clinical variables in BP

In order to test for associations between clinical variables and cognitive functioning in the euthymic BP, exploratory correlational analyses were conducted. We found significant influences of subthreshold depressive symptoms and sleep disturbances on cognitive deficits in BD after controlling for age (see Table 6). Psychomotor speed (compatible trials of Stroop Reading) was associated with MADRS score (p = .036) and reduced sleep (p = .015); Stroop Test Naming was correlated with reduced sleep (p = .018). Divided Attention was associated with BDI-II depression score (p = .036) and reduced sleep (p = .039). Total verbal learning (CVLT) was correlated with MADRS (p = .013), BDI-II (p = .006), and reduced sleep (p = .001), and delayed recall in the CVLT was associated with MADRS (p = .025), BDI-II (p = .021), and reduced sleep (p = .009). Interestingly, short-term recall was correlated with YMRS (p = .030). Furthermore, higher BDI-II scores were associated with more errors in the Cognitive Flexibility test (p = .019). Therefore, impairment in most test measures was associated with subclinical depression and/or sleep disturbances except for Working Memory, which was significantly associated with psychotic symptoms (lifetime) (p = .015). There was no association of cognitive performance and other clinical characteristics indicating disease severity such as number of previous episodes or age of onset. Given these results suggesting that mood symptoms have a negative effect on cognition in euthymic BP, we expected to see the same pattern in acutely ill BP. However, the performance in the test measures was not correlated with depression scores in acutely depressed BP or YMRS scores in (hypo-)manic patients, respectively.

Table 6 Association (r partial) of clinical characteristics and test performance (follow-up testing) of euthymic bipolar patients
Full size table

Psychosocial functioning was assessed using the GAF. In euthymic BP, we found negative correlations between GAF score and omission errors in the Working Memory task (r = −.454, p = .013) and reaction time in compatible trials of the Stroop Test (r = −.562, p = .002). Furthermore, reduced mnestic performance was highly associated with low psychosocial functioning, as indicated by a positive correlation between GAF score and verbal learning (CVLT) (r = .529, p = .003).

Discussion

The present within-subject longitudinal study investigated the neurocognitive performance of bipolar inpatients in an acute episode of depression or (hypo-)mania and during remission. We found broad impairments in cognitive speed, attention, working memory, verbal memory, and executive functioning in acutely admitted BP compared to HC. As expected, depressed and (hypo-)manic patients showed a different cognitive profile, in that depressed patients had slowed reaction times while (hypo-)manic patients were more impaired in executive functioning. However, the differences were marginal and only the performance in the Stroop Test (Naming of colour words with incongruent ink) differed significantly between depressed and (hypo-)manic BP. After at least 3 months of remission BP performance generally improved, albeit we found persisting impairments in working memory and delayed verbal memory. Furthermore, we confirmed previous data in that at least some cognitive measures were associated with subthreshold depressive symptoms and sleep disturbances of euthymic BP, but not with clinical variables indicating severity of illness [11]. In addition, we found strong associations between worse cognitive performance and reduced daily life functioning (GAF score). This is in line with previous investigations demonstrating that neuropsychological performance predicts functional adjustment [25].

Cognitive impairments in acutely ill bipolar patients

In detail, we specified the clinical presentation of cognitive disturbances in acutely depressed BP, in that patients suffered of reduced psychomotor speed compared to (hypo-)manic patients and HC. Furthermore, patients were impaired in divided attention and response inhibition (Stroop Test Reading). This is in line with previous studies which demonstrated a cognitive slowing and attentional deficits in acutely depressed BP [7, 26–28], and also in executive functioning [29–31]. However, attention and cognitive speed are very basic processes which underlie other cognitive functions, and deficits in attention can thus lead to greater distractibility and interference effects [32]. Therefore, deficits in executive functioning in depressed BP could be secondary to attentional deficits. Similarly to the depressed sample, (hypo-)manic patients showed considerable deficits in divided attention and working memory. Accordingly, previous investigations found impaired cognition in manic patients, especially regarding deficits in frontal-executive measures [8, 33, 34]. Therefore, we confirmed a different cognitive profile during depression and (hypo-)mania, in that the test performance of depressed BP was comparable to the cognitive deficits found in major depression [6, 7], and the cognitive performance of (hypo-)manic patients was similar to deficits found in psychosis [8, 9]. Interestingly, our results showed that the differences between depressed and (hypo-)manic BP were only marginal, and with the exception of the Stroop Test, they did not reach significance. Possibly, the neuropsychological tests are not sensitive enough to depict the marginal differences of a depressive and hypomanic symptomatology.

Cognitive deficits in remitted bipolar patients

Due to these difficulties in interpreting cognitive abilities in acutely ill BP regarding the pathophysiology of the disorder, it is important to investigate the long-term course within subjects. Unfortunately, longitudinal studies about cognitive deficits in BD are rare and our study is one of the few performing a follow-up measurement after stable remission. We found that patients’ neurocognitive performance improved after 3 months of remission. Particularly, depressed BP no longer displayed a reduced psychomotor speed compared to HC. (Hypo-)manic patients improved significantly in the Working Memory task. However, even in an euthymic mood state, our sample of BP showed significant deficits in working memory and delayed verbal memory compared to HC. This coincides with previous studies which demonstrated an impaired verbal memory [35–37] and working memory [5, 38, 39] as prominent deficits of BD. Furthermore, we found that 58.6 % of our sample showed cognitive impairments, while 41.4 % had a test performance within the average of the normative data group. Therefore, we confirmed previous study results that only a subgroup of euthymic BP seems to have cognitive deficits [11, 12].

We found that performance in working memory was extensively reduced in patients with acute (hypo-)mania and in remitted patients who reported psychotic symptoms (lifetime). Working memory deficits have been demonstrated as core impairment in psychotic disorders, and BP were suggested to have comparable deficits to schizophrenic patients, albeit less severe [40, 41]. Interestingly, Simonsen [42] showed that neurocognitive dysfunction in BD and schizophrenia was determined more by the history of psychosis than by the diagnostic category or subtype. These results indicate that deficits in working memory may be related to neuronal changes related to the psychotic spectrum. In addition, there is evidence of structural abnormalities in brain regions associated with working memory in BD [43]. Neuroimaging studies have shown altered neuronal activation within the brain network involved in working memory tasks [44]. Most imaging studies in BD reported a loss of connectivity in prefrontal and parietal networks, structures that are well known to be involved in working memory [45]. Taken together, there is evidence that working memory is a trait biomarker in BD although being aggravated by acute symptomatology in our study. Also our correlational analyses are in favour of this hypothesis: psychomotor speed, attention and verbal memory were significantly associated with residual symptoms (MADRS, BDI-II, and sleep disturbances). On the contrary, working memory was not correlated with any actual mood symptoms, which is in agreement with a recently published study about predictors of cognition in euthymic BP [11]. In line with these results, impairments in executive functions including working memory were considered as salient endophenotypic components of cognition in BD as they persist in remission, appear to be heritable, and hence co-segregate within families [46, 47]. Therefore, our study provides further evidence that working memory could serve as a potential neurocognitive endophenotype in remitted BP.

Association between cognition and clinical variables

Regarding the clinical characteristics of our bipolar sample, we found no relationship between test measures and disease characteristics indicating severity of the disorder (e.g. number of episodes, age of onset, bipolar type). However, subclinical mood symptoms and sleep disturbances were associated with cognitive speed, attention, and verbal memory. Although patients with bipolar disorder have historically been characterised as returning to full remission between affective episodes, recent studies demonstrated that some residual symptoms may persist [48, 49]. It has been previously demonstrated that subthreshold mood symptoms are associated with impairments in attention and verbal memory [11, 50–52]. In a study by Bonnin et al. [53], the degree of cognitive functioning was best predicted by subdepressive symptoms, and a meta-analysis reported that more rigorously defined euthymia goes along with smaller cognitive impairments [2]. Therefore, it seems that the neuronal circuitry activated in the modulation of emotions overlaps with brain regions involved in neurocognition. Besides subclinical depressive symptoms, 37 % of our remitted bipolar sample reported sleep disturbances like initial or middle insomnia or an abnormal increased sleeping time. Rates of sleep disturbances are usually very high in remitted BP [54, 55], and BD is highly associated with sleep and circadian rhythm abnormalities [49, 56, 57]. It is well known that sleep deprivation in healthy people causes cognitive deterioration [58, 59]. Furthermore, it has been demonstrated that sleep and circadian rhythms are involved in the cognitive deficits in BD through overlapping neurobiological systems [60]. In our bipolar sample, sleep disturbances were associated with impairments in psychomotor speed, divided attention, and verbal learning. In a recently published study, we reported accordingly that several cognitive impairments in euthymic BP were predicted by sleep disorder [11]. Taken together this underscores the need for more extensive research on the influence of sleep disturbances on cognitive deficits in BD. In conclusion, our results suggest that residual symptoms negatively affect cognitive functioning in BD. Therefore, an optimised antidepressant treatment and sleep regulation should be enforced in patients complaining about cognitive impairments. Even though our results do not disprove the idea of neurodegenerative processes in BD [1], our data underscore the importance of residual symptoms. This also has implications for future research and calls for the improvement of treatment options for cognitively impaired BP.

Comparable to the influence of subdepressive symptoms on test performance of euthymic BP, we expected a relationship between severity of depression and test performance in acutely ill BP. Very surprisingly though, we did not find associations between the severity of depression or (hypo-)mania and neurocognitive measures in acutely admitted BP. This might well be due to a ceiling effect of the effect on affective measures on neuropsychological impairment. Finally, it has to be discussed how much time patients need to fully recover after an acute episode. We defined a period of 3 months of euthymia preceding neurocognitive assessment in order to ensure a relatively stable state. Nevertheless, some of the remitted BP reported residual symptoms. Interestingly, in a recent study, Torres et al. [61] investigated first-episode patients three times within the first year following their initial manic episode. Patients’ cognitive functioning was improved 6 months after discharge and even further at follow-up after 1 year. Therefore, it seems that it takes a long time to fully cognitively recover after an acute episode, and we recommend more research on BP who have been remitted for a period of at least a year although these studies are inherently difficult to conduct.

Limitations

Among the limitations of the present study is the relatively small sample size, especially at follow-up, which is inherent to these types of studies. This nevertheless raises the possibility of selective attrition and whether non-significant findings can be related to insufficient statistical power. Indeed, we found that patients who dropped out had a higher load with depression (more number of previous depressive episodes and higher depression scores at baseline). Therefore, the study completers may have a less severe progress of disease. Due to the naturalistic nature of the study, we had only limited control of pharmacological effects on cognition. Although there were no statistical differences between depressed and (hypo-)manic groups on that regard, the effects of medications for bipolar disorder on cognition are not totally understood [62]. As some BP had changes in drugs and dosing within the interval between baseline and follow-up, interpretation of changes in cognitive performance is limited. In particular, (hypo-)manic patients received antipsychotics at baseline and drugs were tapered at follow-up. A methodological problem which has to be addressed is to determine the extent of practice effects in the pre-post-testing. We thus included a sufficient sample size of HC to calculate test–retest reliability allowing to control for practice effects.

Summary and conclusions

Although these limitations warrant attention, the results of the present study add the existing literature to cognitive functioning in BD. To our knowledge, this is the first study presenting within-subject pre- and post-testing with acutely admitted BP in depression or (hypo-)mania and during euthymia. We found that cognitive deficits in BD are exacerbated during acute episodes and partially recover during remission. Working memory and verbal memory were still impaired in euthymic BP compared to HC, indicating these two cognitive domains as trait-related. The present study demonstrated that “state” factors like subclinical symptoms and persisting sleep disturbances contribute to cognitive impairment and are potential confounders in the search of cognitive endophenotypes in BD. Furthermore, our results implicate the need for interventions that improve cognitive dysfunctions in remitted BP. In the clinical setting, an optimised antidepressant treatment and sleep regulation could improve cognitive deficits which are caused by residual symptoms. Moreover, some studies demonstrated promising results in that euthymic BP benefit from a cognitive training with regard to neurocognitive and psychosocial functioning [63–65]. However, to clarify trait versus disease process effects in cognition in BD, more studies with a prospective design, especially with high-risk groups and first-episode patients, are needed.