Neurocognitive impairment and psychosocial functioning in bipolar II disorder
Eduard Vieta and Anabel Martínez-Arán, Bipolar Disorders Program, Clinical Institute of Neuroscience, University Clinic Hospital of Barcelona, CIBERSAM, Villarroel 170, 08036-Barcelona, Spain.
E-mail: email@example.com; firstname.lastname@example.org
Solé B, Bonnin CM, Torrent C, Balanzá-Martínez V, Tabarés-Seisdedos R, Popovic D, Martínez-Arán A, Vieta E. Neurocognitive impairment and psychosocial functioning in bipolar II disorder.
Objective: There is a growing body of evidence on neurocognitive impairment in euthymic bipolar patients, but this issue has been studied mostly in bipolar I disorder, data on bipolar II (BD-II) are scant and discrepant. The two aims of this study were to ascertain whether strictly defined euthymic BD-II patients would present neurocognitive disturbances and to evaluate their impact on functional outcome.
Method: Forty-three BD-II patients and 42 demographically and educationally matched healthy subjects were assessed with a comprehensive neuropsychological test battery and with the Social and Occupational Functioning Assessment Scale (SOFAS). The euthymia criteria were reduced (Hamilton Rating Scale for Depression score ≤6 and a Young Mania Rating Scale score ≤6) to minimize the influence of subdepressive symptomatology on cognition and functioning.
Results: BD-II patients showed a significantly lower performance on several measures of attention, learning and verbal memory, and executive function compared with healthy controls. The presence of subthreshold depressive symptomatology and one measure related to executive function (Trail Making Test, part B) was the variables that best predicted psychosocial functioning measured with the SOFAS.
Conclusion: This report provides further evidence that euthymic BD-II patients present cognitive impairment which may impact psychosocial functioning.
- • Patients with bipolar II disorder in remission show cognitive dysfunction in attention, verbal memory, and executive functions.
- • The presence of subdepressive symptomatology and executive dysfunction may influence these patients’ psychosocial functioning.
- • Neurocognitive rehabilitation should be considered in each subtype of bipolar disorder to improve the functional outcome.
- • The cross-sectional design did not allow to draw conclusions regarding the stability of cognitive deficits in euthymic bipolar II patients.
- • The effect of medication and the impact of psychosis on neurocognition were not controlled for.
- • The directionality of the relationship between subclinical depressive symptoms and psychosocial dysfunction remain unclear.
It is well-documented that bipolar II disorder (BD-II) is underdiagnosed in clinical practice (1). The difficulty in diagnosing may be due to some limitations of current diagnostic criteria for this subtype of bipolar disorder (2). Hence, it is not surprising that many aspects of BD-II are still understudied such as pharmacological treatment, clinical course, and neurocognitive impairment. Regarding the latter issue, few studies have specifically investigated neuropsychological profile in BD-II patients. A recent systematic review reported working memory problems and deficits in inhibitory control in BD-II subjects and subtle neurocognitive differences with bipolar I (BD-I) patients (3). Similarly, a recent published meta-analysis showed that, with the exception of memory and semantic verbal fluency (which are more impaired in BD-I), cognitive impairment in BD-II is as severe as in BD-I patients (4). Hsiao et al. (5) also found different profiles of neurocognitive impairment between BD-I and BD-II patients, in the sense that BD-II patients did not show impairments in verbal memory compared with BD-I patients. However, psychomotor speed and working memory deficits were present in both groups. Similarly, another study, conducted by our group some years ago (6), found that patients with BD-II showed an intermediate level of performance in verbal and executive functions when compared with the BD-I group. In a subsequent report, neurocognitive dysfunctions were found in euthymic BD-I and BD-II patients; however, non-significant neuropsychological profiles were identified between both subtypes (7). Other studies have been conducted comparing BD-I and BD-II patients, using, also, other comparative groups; however, non-rigorous criteria for euthymia were established and patients with mild-to-moderate symptomatology were included (8–14). In conclusion, the results of these studies are inconsistent, and further research is needed to elucidate the nature of neurocognitive impairment in BD-II.
Moreover, it is known that cognitive impairment may have negative implications in the functional outcome of bipolar patients (15–19). In BD-II, this relationship has been studied to a lesser extent (6, 20). However, at least two reports have shown that BD-II patients are as functionally disabled as BD-I patients (21, 22). Therefore, it may be useful to identify specific neurocognitive predictors of functional outcome in BD-II, because it is not known to which extent both subtypes share the same pattern in neurocognitive deficits and functional outcome.
To the best of our knowledge, only one study has focused on the research of neurocognitive dysfunction in two groups, a BD-II group vs. a healthy control group, without including other comparative groups (23). However, the sample comprised not only euthymic patients but also hypomanic and depressed participants.
Aims of the study
Our primary aim was to study a large sample of strictly defined BD-II euthymic patients to compare it with a demographically and educationally matched sample of healthy controls to elucidate the neuropsychological profile of this bipolar subtype. The second objective was to study the relationship between neurocognitive performance and functional outcome. Based on previous studies, we hypothesized that BD-II patients would show deficits in several cognitive domains, especially with respect to attention, verbal memory, and aspects of executive functions.
Material and methods
A total of 85 participants, 43 of whom were BD-II patients and 42 healthy controls, from the Barcelona Bipolar Disorder Program at Hospital Clinic were screened for inclusion in the study. Inclusion criteria were represented by the following: i) BD type II diagnosis, according to DSM-IV as assessed by the Structured Clinical Interview for DSM-IV (SCID) (APA, 1994); ii) strict euthymia, defined by a Hamilton Rating Scale for Depression (HDRS) score ≤6 (24, 25) and a Young Mania Rating Scale (YMRS) score ≤6 (26, 27) during at least 6 months prior to neuropsychological evaluation; and iii) age between 18–65 years. Exclusion criteria were as follows: i) history of mental retardation or any clinical condition that could affect cognitive performance (significant physical or neurological illness, neurodegenerative disorder or endocrine disorders, except for corrected hypothyroidism); ii) axis I comorbidity; iii) electroconvulsive therapy during the year prior to neuropsychological assessment. Of 70 potential BD-II patients, 27 patients were excluded because they did not meet the strictly defined euthymia criteria (n = 21) or were older than 65 years (n = 6). Twenty-one patients of the 106 subjects (19%) assessed in the Torrent et al. 2006 (6) report participated in this study as well. All participants provided written informed consent, and the study was conducted in accordance with the guidelines laid down in the Declaration of Helsinki.
We decided to reduce the HDRS criteria for euthymia to minimize the influence of subdepressive symptomatology on cognitive functioning, as it has been shown that it has a negative impact on cognition and psychosocial outcome.
A group of healthy controls without psychiatric or neurological history were recruited from a pool of volunteers. It included a demographically and educationally matched sample of students, workers, and housewives. The matching was performed on group level. It was ensured that none in the control group had first-degree relatives with an affective or a psychotic disorder. Controls were screened for Axis I psychiatric disorders using the SCID (28).
Patients and healthy volunteers were tested during euthymia with a comprehensive neuropsychological battery. It consisted of: i) Premorbid intelligence quotient (IQ), estimated with the WAIS Vocabulary subtest (29); ii) verbal learning and memory: California Verbal Learning Test (CVLT) (30); iii) Executive functions: Wisconsin Card Sorting Test (WCST) (31); Stroop Colour–Word Interference Test (SCWT) (32); FAS and Animal Naming (Controlled Oral Word Association Test) (33); Digits Backwards from the WAIS (29); and Trail Making Test – part B (34); iv) Attention: Digits Forward; (29) Trail Making Test – part A (34). This battery takes approximately 60 min to administrate.
Clinical and sociodemographic assessment
We gathered all the relevant clinical and sociodemographic data through a clinical interview that belongs to our Program’s Protocol (35) based on the SCID for DSM-IV (36). The self-reported patient information was checked with the clinical history. Hence, age, gender, educational level, occupation, diagnosis, the number and type of episodes, chronicity, age at first hospitalization, number of hospitalizations, number of suicide attempts, history of psychosis [defined by delusions and/or hallucinations during mood episodes (37)], and history of rapid cycling and symptomatology at the time of assessment (depressive and manic) were also collected.
Functional outcome was assessed using the Social and Occupational Functioning Assessment Scale (SOFAS-DSM-IV). The SOFAS was used to measure psychosocial functioning 1 month prior to neuropsychological assessment. This scale differs from the Global Assessment of Functioning scale as it focuses exclusively on the individual’s level of social and occupational functioning, whereas clinical symptoms are not taken into account. Functioning was only measured in patients to find out clinical and cognitive predictors of poor functioning.
All data were analyzed with the Statistic Package for Social Sciences [spss v.16 for Windows (spss Inc., Chicago, IL, USA)]. First, we performed a one-way analysis of variance (anova) for continuous variables or chi-square for the categoric ones, to find differences in clinical, sociodemographic, and neuropsychological variables between both groups (BD-II patients vs. healthy control group). Statistical significance was set at P < 0.05 level.
As neuropsychological tests are naturally correlated, this procedure was considered better than Bonferroni inequality correction, which would have increased type II error. The effect sizes (Cohen’s d value) were calculated to estimate the magnitude of the differences between both groups (0.2 = small, 0.5 = medium, 0.8 or greater = large). If any clinical or sociodemographic variables were significantly different between both groups, these were included as covariates in an analysis of covariance (ancova) to investigate if group differences in neurocognitive performance remained significant after controlling for.
The next step was to run Pearson’s correlations to identify the clinical and neurocognitive variables linked to functional outcome. As a final step, the clinical and neuropsychological variables that were significantly correlated with the SOFAS scores were introduced in a linear regression model using a stepwise method to find the variables that could best predict the patients’ functional outcome.
BD-II patients and healthy controls did not differ with respect to age, gender, educational level, and estimated premorbid IQ, as shown in Table 1. Regarding the mean SOFAS scores, our sample was characterized by a reasonably good psychosocial functioning with a mean score of 71.70 [14.15 standard deviation (SD)], as a value equal or higher to 70 stands for scarse difficulties in social, occupational, or academic activities. However, it is noteworthy that nearly 35% of patients was unemployed at the time of assessment.
Table 1. Clinical and sociodemographic data
|Educational level (years)||13.36||4.24||13.15||3.19||0.06||0.804|
|Duration of illness (years)||15.49||10.19|| || || || |
|Age at first hospitalization||44.25||12.16|| || || || |
|Number of hospitalizations||0.37||0.97|| || || || |
|Number of depressive episodes||6.85||6.01|| || || || |
|Number of hypomanic episodes||6.51||5.93|| || || || |
|Total episodes||13.49||11.52|| || || || |
|Number of suicide attempts||0.41||1.09|| || || || |
|SOFAS scores||71.70||14.15|| || || || |
| ||n||%|| || || || |
| Women||23||53.5||27||64.3|| ||0.380|
| Not working||15||34.9|| || || || |
|Family history of affective disorders|
| Yes/positive||25||58.1|| || || || |
|History of rapid cycling|
| Yes||6||14.0|| || || || |
|Prior psychotic symptoms|
| Yes||5||11.6|| || || || |
When anova was performed, 13 of 15 neurocognitive comparisons between-group differences were statistically significant. Only non-significant differences between groups were found regarding FAS and SCWT interference score. We found that subsyndromal symptomatology (YMRS scores) was statistically different between groups (P = 0.030). As this variable may influence neuropsychological performance, we decided to apply an analysis of covariance (ancova) controlling for the confounding effect of subsyndromal hypomanic symptoms but overall the results did not significantly vary. ancova yielded Pillai’s F = 1.814, df = 15, 68 (P = 0.050) for the main effect, indicating that there were overall differences in neuropsychological performance between groups. For 12 of 15 comparisons, the differences were statistically significant (P < 0.05) (see Table 2). In general, the patients performed poorer on most neuropsychological measures than the healthy control group, particularly on those measures related to attention (Digits Forward and TMT-A), executive function (WCST perseverative errors) and verbal learning and memory (CVLT delayed recall) with a greater statistical significance. Other measures in which BD-II patients significantly differed from controls are WCST categories, Digits Backwards subtest, TMT-B, semantic verbal fluency (Animal Naming), and CVLT learning task, short cued recall and recognition. Neuropsychological data and effect sizes are provided in Table 2. It is noteworthy that effect sizes were, in general, moderate, except for the WCST categories with a small effect.
Table 2. Neuropsychological performance across groups
| Perseverative errors||18.84||17.68||8.98||6.30||11.61||0.001||8.20||0.005||0.73|
|Subtest digits WAIS|
| Digits backwards||4.19||1.16||4.79||1.15||5.68||0.019||6.75||0.011||0.51|
| Trail B||105.16||57.22||78.76||36.89||6.35||0.014||5.25||0.024||0.54|
|Subtest digits WAIS|
| Digits forward||5.37||1.32||6.31||1.29||10.83||0.001||13.96||<0.001||0.72|
| Trail A||43.05||19.12||31.48||11.26||11.47||0.001||9.55||0.003||0.73|
|Verbal learning and memory|
| List A (total)||48.16||11.13||54.29||9.42||7.47||0.008||5.53||0.021||0.59|
| Short free recall||10.23||2.91||11.64||3.15||4.58||0.035||3.25||0.075||0.46|
| Short cued recall||11.56||2.34||13.07||2.36||8.79||0.004||5.93||0.017||0.64|
| Delayed free recall||10.58||2.98||12.76||2.90||11.64||0.001||9.58||0.003||0.74|
| Delayed cued recall||11.35||2.61||13.33||2.55||12.51||0.001||9.18||0.003||0.76|
Pearson’s correlations showed that patients with higher YMRS scores had a poorer performance on CVLT short cued recall (r = −0.30, P = 0.044); patients with a higher number of hypomanic episodes had a higher premorbid IQ (r = 0.35, P = 0.027); longer illness duration was positively correlated with TMT-A (r = 0.36, P = 0.018), and negatively correlated with semantic fluency (r = −0.38, P = 0.013) and phonemic fluency (r = −0.36, P = 0.017).
Functional outcome, assessed by means of the SOFAS, was found to correlate with HDRS (r = −0.57, P = 0.001) as well as with performances on the TMT-A (r = −0.36, P = 0.040) and TMT-B (r = −0.47, P = 0.006). No further correlations were found between SOFAS scores and other clinical, sociodemographic, or neurocognitive variables.
The subsequent stepwise multiple linear regression analysis, using SOFAS scores as dependent variable, showed that the HDRS scores (t = −3.94, β = −0.523, df = 31, P < 0.001) together with TMT-B (t = −3.13, β = −0.416, df = 31, P = 0.004) were the variables that could best predict functional outcome. This model reached significance (F = 14.34, df = 2, 31, P < 0.001) with an adjusted r2 = 0.46. The HDRS scores (mild depressive symptoms) explained 30.4% of the variance of psychosocial functioning and the TMT-B (an executive function measure) explained 15.9% of the variance of psychosocial functioning as measured through the SOFAS.
On one hand, in accordance with previous studies (5–7), the present report provides further evidence that euthymic BD-II patients differ from healthy control subjects with respect to neuropsychological performance. On the other hand, these findings suggest that these patients’ functional outcome can be partially explained by two main factors, subthreshold depressive symptoms and executive functioning.
With regard to attention, in line with our findings, euthymic BD-II patients were found to have a poorer performance when compared with healthy controls in previous reports (6, 7). Moreover, in a meta-analytic study, attention/processing speed appears as a trait-related neuropsychological deficit in euthymic bipolar patients (38). However, in our study, attention domain was assessed only with two measures and probably the use of a widespread measure of sustained attention could be helpful to confirm this disturbance, like the Continuous Performance Test (39). As mentioned earlier, the TMT-A, a measure related to attention, was correlated with psychosocial functioning, although it was not found to be a predictor of functioning.
We were able to replicate the results of the study by Torrent et al. with regard to deficits in verbal learning and memory tasks. However, other studies failed to detect these deficits in euthymic BD-II subjects (5–7). We hypothesize that these discrepancies between studies in verbal memory impairment might be due to different memory tasks used. The CVLT test might have a frontal executive component, because semantic strategies are needed to encode information (40). Recently, a committee of experts has proposed a preliminary neuropsychological battery of tests for research in the field of cognition in BD (41), and the CVLT was considered an appropriate tool to detect verbal learning/memory difficulties in patients with BD. Nevertheless, other factors could also explain these discrepant findings: some studies emphasize the role of psychotic symptoms in bipolar disorder, as history of psychosis could explain the neurocognitive impairment better than the diagnostic subtype (37, 42). Simonsen et al. (42) found that diagnostic subtype only had significant main effects on two verbal recall measures, while history of psychosis had significant main effects on most neurocognitive measures. Similarly to the previously mentioned study, Martinez-Aran et al. (37) suggested that psychosis may partly account for the cognitive dysfunction seen in euthymic patients, especially with regard to persistent verbal memory dysfunction as well as in some executive dysfunctions. Unfortunately, the small number of patients with history of psychosis in our sample did not allow us to investigate the impact of this variable on neurocognition. Moreover, the impact of psychosis in BD-II patients is difficult to study as they can only be present in depressive episodes (43). The extent to which BD-I and BD-II patients differ in the pattern of impairment in verbal learning and memory could not be assessed in this report. Nevertheless, a meta-analytic study suggests that BD-I patients may be more impaired owing to the manic episodes, which are known to be neurotoxic, or to secondary features such as psychosis, age of illness onset and treatment (4).
Regarding executive functions, BD-II participants achieved a lower number of categories and committed a significant higher number of perseverative errors in the WCST compared with healthy controls. Our results confirm the trend toward a higher number of perseverative errors previously observed in Torrent and colleagues’ report (6).
The time spent in the TMT-B was higher in patients compared with healthy controls. This latter finding is in line with the study conducted by Dittmann et al. (7), and a trend was also found in the study by Torrent et al. (6).
Other measures like the semantic verbal fluency and working memory were also impaired in our BD-II sample. The former finding is not surprising because other reports and some meta-analyses have also emphasized this deficit (6, 13, 44, 45), and concerning working memory, our finding is also in line with the literature (5–7, 23), although we assessed working memory only by using a simple index: digits backward (WAIS); hence, this result should be taken as preliminary.
We did not find a significant difference between groups in the Stroop interference task. This was an unexpected result because most of the studies assessing this measure in BD-II subjects reported impaired inhibitory control (6, 13, 23). However, in support to our findings, Andersson et al. (23) discuss the significance of the interference deficit because when controlling for psychomotor speed in the Stroop condition, the differences between groups disappeared. Moreover, the authors suggested that the deficits in executive function may be more related to psychomotor than to executive functioning. Phonemic fluency was also preserved as confirmed by other studies (6, 12, 13, 23). Only one report found impaired phonemic fluency, but their finding may not generalize beyond the subtype because the sample was formed by depressed suicide attempters, which represents more severely ill individuals (10). In this regard, impaired phonemic verbal fluency has been reported as a more state-dependent neuropsychological measure in other studies (46–48).
Therefore, the main executive measures affected in our sample were reasoning, problem solving, semantic verbal fluency, working memory, and speed of processing.
As mentioned before, there is a scarcity of data with regard to functional outcome in BD-II patients. We found that our sample presented an acceptable functional outcome based on the mean SOFAS score. However, over one-third of patients were unemployed, suggesting that not all the patients have achieved a complete functional recovery. This result may indicate that BD-II is not a less severe illness and that these patients seem to be as disabled as BD-I in functional outcome (22). It may also mean that the SOFAS, although helpful and easy to use, is not the ideal tool to measure functioning in bipolar disorder (20, 49), as it is just a global measure of functioning. Instruments with an objective evaluation for assessing different functional abilities or areas are needed. Moreover, functioning is a complex concept that involves several real-world domains (such as occupational capacity, autonomy, interpersonal relationships, leisure time). Therefore, it would be useful to assess functioning with instruments measuring problems more specifically associated with bipolar disorders.
Neurocognitive impairment has been demonstrated to be significantly associated with lower psychological functioning in euthymic patients (17). Cognitive function may be mediating between illness course and functional outcome (18, 47, 50, 51). In previous studies, verbal memory has been to influence on psychosocial outcome in euthymic bipolar patients (16, 52, 53). However, this finding emerged from studies with mixed samples (BD-I and BD-II) or samples composed only of BD-I patients. As far as we know, few studies have assessed psychosocial functioning in BD-II patients (6, 13) but only the study by Torrent and colleagues (6) assessed predictors of functional outcome related to cognitive variables on euthymic BD-II and found that the TMT-B was a good predictor of psychosocial outcome. In the present study, the same index of executive function (TMT-B) was found to predict psychosocial functioning in BD-II euthymic patients. Hence, it may suggest that impairment in executive functions and specifically the TMT-B may be the best predictor of functional outcome for BD-II patients. It is noteworthy that two recent reports studying BD-I patients have found other predictive variables in psychosocial functioning: such as IQ together with white matter volume (54) or changes in a composite neurocognitive score, which included eight neurocognitive domains (19). All together, the evidence so far suggests that different neurocognitive variables could be involved in the functional outcome of BD-I and BD-II. However, further research studying separately both subtypes is urgently needed to identify cognitive endophenotypes and clarify which variables could best predict their functional outcome.
We have to keep in mind that executive functions and working memory are needed to develop and maintain everyday life activities as are important for successful adaptation and performance in real-life situations. Specifically, planning and generation of strategies, problem solving and flexibility, semantic memory organization are needed to perform complex tasks. In real-world life, the main executive dysfunctions found in our sample of patients, by means of performance in standard neuropsychological tests, might affect daily activities such as starting, maintaining and completing tasks in work, housework, pharmacological treatment adherence as well as difficulties in interpersonal relationships. Nowadays, the use of traditional neuropsychological tests may be questioned as they may lack of ecologic validity; however, as mentioned elsewhere (55), traditional tests are known to have veridicality. This concept refers to the degree to which classical neuropsychological tests are empirically related to measures of every-day functioning. Thus, even though the tests used in this report are not designed with ‘ecological validity’ in mind, they may still be predictive of everyday functioning, as reported in several studies (51, 56–58).
It is well established that subthreshold depressive symptoms influence both neurocognition and functional outcome (18, 59–61). A recent report by Parker and Fletcher (62) pointed out that BD-II subjects showed high rates of psychomotor retardation that could be associated with cognitive dysfunction. Once again, our results suggest that the presence of subthreshold depressive symptoms is a good predictor of psychosocial functioning in bipolar disorders. Nevertheless, those patients with a low functioning may be more likely to show subdepressive symptoms in the long term, so that the directionality of this association is not as clear in our study as it was in the follow-up study by Bonnin et al. (18).
The association between subthreshold symptomatology, cognitive dysfunction, and functional impairment implies the use of more stringent remission criteria and, therefore, to aim higher in symptomatic treatment goals (63).
There are some limitations in the present study. First, a larger sample would have been recommendable, providing increased statistical power and firmer conclusions. Second, we cannot rule out the effect of the medication interfering with the cognitive performance, because most patients were polimedicated (64, 65). In two studies conducted on unmedicated subjects with bipolar depression, type II patients appeared to be unaffected (9, 11). However, the former had a relatively small sample size and the latter included a brief neuropsychological battery, failing to assess some neurocognitive domains. It seems that the cognitive disturbances observed in bipolar disorder are not simply effect of medication as some of the same deficits have been observed in unaffected relatives as well (45, 66). Thus, more research is needed to clarify the influence of medications on neurocognition. A further limitation is the cross-sectional design of the present study. In this regard, it would also be interesting to conduct a longitudinal follow-up study to provide more information about the progression or stability of cognitive deficits and their influence on psychosocial functioning.
Finally, other instruments, such as the Functioning Assessment Short Test (FAST) (67), would probably provide more information about specific areas of functioning than the SOFAS. The FAST involves specific functional domains, as above mentioned, which could be impaired in bipolar patients, such as occupational functioning, leisure time, autonomy and finances.
In conclusion, we have found that BD-II patients performed worse than healthy controls in attention (Digits Forward and TMT-A), executive functions (WCST categories and perseverative errors, semantic verbal fluency, digits backward and TMT-B), and on most measures of verbal learning and memory (CVLT learning task, short cued recall, delayed free and cued recall, and recognition hits). Moreover, one measure of executive function (TMT-B) and the subthreshold depressive symptomatology was found to predict the functional outcome of these patients.
This report does not allow us to conclude that there may be differences in neuropsychological profiles between the two subtypes, although we can confirm that cognitive deficits exist in BD-II. Differences on cognitive dysfunction between the bipolar subtypes may have an effect on psychosocial functioning, treatment adherence as well as the possibility to benefit from psychoeducational programs. Neurocognitive rehabilitation should consider differences in cognitive profiles to design specific programs to treat prevailing cognitive dysfunctions for each subtype and improve the functional outcome.
This work was supported by the Instituto de Salud Carlos III, Centro de Investigación en Red de Salud Mental, CIBERSAM, and the Generalitat de Catalunya to the Bipolar Disorders Group (2009 SGR 1022), and the following grant support by the Spanish Ministry of Science and Innovation (CP07/00144, PI080180 and PI08/90094). Mar Bonnin is funded by Spanish Ministry of Education through a FPU. The work was carried out ‘in part’ at the Esther Koplowitz Center (CEK) in Barcelona.
Declaration of interest
Professor Eduard Vieta has served as consultant, advisor, or speaker for the following companies: Almirall, AstraZeneca, Bial, Bristol-Myers Squibb, Eli Lilly, Forest Research Institute, Geodon Richter, Glaxo-Smith-Kline, Janssen-Cilag, Jazz, Lundbeck, Merck-Sharp and Dohme, Novartis, Organon, Otsuka, Pierre Fabre, Pfizer Inc, Sanofi-Aventis, Servier, Shering-Plough, Takeda, UBC, and Wyeth. Professor Rafael Tabarés-Seisdedos has received grants from Lilly, Pfizer and AstraZeneca companies. Dr. Balanzá-Martínez has received research grants and has served as consultant, advisor, or speaker during the last 3 years for the following companies: Almirall, AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, Grunenthal, Janssen, and Pfizer Inc. Dr Martinez-Aran has served as speaker or advisor for the following companies: Bristol-Myers Squibb, Otsuka, Pfizer. The rest of co-authors have no conflict of interest.