Original Article

12-month longitudinal cognitive functioning inpatients recently diagnosed with bipolardisorder

Torres IJ, Kozicky J, Popuri S, Bond DJ, Honer WG, Lam RW, YathamLN. 12-month longitudinal cognitive functioning in patients recentlydiagnosed with bipolar disorder.Bipolar Disord 2014: 16: 159–171. © 2013 John Wiley & Sons A/S.Published by John Wiley & Sons Ltd.

Objectives: Although cognitive deficits are observed in the early stages ofbipolar disorder, the longitudinal course of neuropsychologicalfunctioning during this period is unknown. Such knowledge couldprovide etiologic clues into the cognitive deficits associated with theillness, and could inform early treatment interventions. The purpose ofthe present study was to evaluate cognitive change in bipolar disorder inthe first year after the initial manic episode.

Methods: From an initial pool of 65 newly diagnosed patients withbipolar disorder (within three months of the end of the first manic ormixed episode) and 36 demographically similar healthy participants, 42patients [mean age 22.9 years, standard deviation (SD) = 4.0] and 23healthy participants [mean age 22.9 years (SD = 4.9)] completedbaseline, six-month, and one-year neuropsychological assessments ofmultiple domains including processing speed, attention, verbal andnonverbal memory, working memory, and executive function. Patientsalso received clinical assessments, including mood ratings.

Results: Although patients showed consistently poorer cognitiveperformance than healthy individuals in most cognitive domains,patients showed a linear improvement over time in processing speed(p = 0.008) and executive function (p = 0.004) relative to the comparisongroup. Among patients, those without a history of alcohol/substanceabuse or who were taken off an antipsychotic treatment during the studyshowed better improvement.

Conclusions: The early course of cognitive functioning in bipolardisorder is likely influenced by multiple factors. Nevertheless, patientswith bipolar disorder showed select cognitive improvements in the firstyear after resolution of their initial manic episode. Several clinicalvariables were associated with better recovery, including absence ofsubstance abuse and discontinuation of antipsychotic treatment duringthe study. These and other factors require further investigation to betterunderstand their contributions to longitudinal cognitive functioning inearly bipolar disorder.

Ivan J Torresa,b, Jan Kozickya,Swetha Popuria, David J Bonda,William G Honera,b, Raymond WLama and Lakshmi N Yathama

aDepartment of Psychiatry, University of British

Columbia, bBritish Columbia Mental Health and

Addictions Services, Vancouver, BC, Canada

doi: 10.1111/bdi.12154

Key words: antipsychotic – cognitive –

executive function – longitudinal – mania –

memory – neurocognition – neuropsychology –

substance abuse

Received 4 May 2012, revised and accepted for

publication 29 June 2013

Corresponding author:

Ivan J. Torres, Ph.D.

Department of Psychiatry

University of British Columbia

2255 Wesbrook Mall, Room 2C7

Vancouver

BC V6T 2A1

Canada

Fax: 604-822-7922

E-mail: ivan.torres@ubc.ca

Significant cognitive deficits are present during allstages of bipolar disorder, including shortly afterthe onset of the first manic episode (1), during themiddle course (2), and in elderly patients (3). How-ever, the longitudinal course of cognitive function-

ing throughout the illness is poorly understood.The only longitudinal studies that have evaluatedcognitive change in bipolar disorder have beenrestricted to middle-aged or elderly patients. Withlittle exception (3), these studies reveal that across

159

Bipolar Disorders 2014: 16: 159–171 © 2013 John Wiley & Sons A/SPublished by John Wiley & Sons Ltd.

BIPOLAR DISORDERS

periods ranging from one to four years, there is nosignificant difference in longitudinal cognitivechange in patients relative to healthy controls (4–9). By contrast, the extent and nature of longitudi-nal cognitive change during the critical early stagesof the illness are unknown as there are no longitu-dinal studies that have investigated cognitive func-tioning during this period of the illness.

Although recently diagnosed clinically stableadults with bipolar disorder show significant cogni-tive deficits relative to healthy controls (1, 10, 11),the cause(s) or contributing factors underlyingthese early cognitive deficits are unclear, and couldreflect genetic or developmental abnormalities, dis-ease processes associated with the onset of illness,residual subsyndromal mood symptoms, oradverse treatment effects. The temporal stability ofthese early cognitive deficits (i.e., whether they per-sist, improve, or decline across time) is alsounknown. Increasing the understanding of the tra-jectory of cognitive functioning early in the courseof illness is critical as this could provide furtherclues into the etiology of these early cognitive defi-cits. For example, the presence of a decline in cog-nition after the first manic episode could point toseveral potential factors, including the adverse cog-nitive side effects of treatment or illness progres-sion early in the illness. Indeed, contemporaryviews conceptualize bipolar disorder as a potentialneuroprogressive illness (12). Although the timeframe and specific nature of any underlying neuraldegeneration remains unclear, the most prominentmechanisms that have been proposed includeinflammation, oxidative stress, and neurotrophicfactors (13–15).

Conversely, the observation of cognitiveimprovement across time could suggest that suchgains result from resolution of subsyndromalsymptoms, beneficial early treatment effects, orother factors. Relative stability in cognitive func-tioning could suggest that cognitive deficits areassociated with causes that result in more fixed orstable cognitive impairments, such as genetic vul-nerabilities impacting cognition or other stable dis-ease-related factors. Most importantly, however,increased understanding of whether early cognitivedeficits are fixed, transient, or even potentiallymodifiable may help to direct treatment or rehabil-itative efforts toward improving these early cogni-tive impairments. The clinical significance oftargeting cognitive symptoms early in the disorderis underscored by the finding that cognitive func-tioning predicts functional outcome even at thisstage of illness (16).

The purpose of the present study was to evaluatelongitudinal cognitive functioning in a sample of

clinically stable patients with recently diagnosedbipolar disorder in the first year after resolution oftheir first manic episode, and to contrast their cog-nitive performance to demographically similarhealthy individuals. A second goal was to deter-mine whether any observed cognitive gains orlosses in patients can be attributed to clinical vari-ables such as underlying mood symptoms, medica-tion effects, or other variables.

Materials and methods

Participants

Sixty-five patients meeting DSM-IV-TR criteriafor bipolar I disorder were initially recruited fromthe Systematic Treatment Optimization Programfor Early Mania (STOP-EM) at the University ofBritish Columbia (UBC) and affiliated sites (17).The STOP-EM program treats individuals referredby physicians and psychiatrists in the communityand local hospitals. Diagnosis was based on a com-prehensive clinical interview by a board-certifiedpsychiatrist and a Mini International Neuropsychi-atric Interview (MINI) (18). For enrollment intothe study, patients were required to be within threemonths of resolution of their first documented life-time manic (or mixed) episode. Inclusion criteriawere broad in order to capture a representativesample of patients presenting for treatment. Thus,patients with either a history of alcohol or sub-stance abuse, depressive or hypomanic episodepreceding the first manic episode, or psychosisassociated with the first manic episode wereincluded. Patients presenting with a history ofmajor medical or neurological illness underlyingtheir manic symptoms were excluded. Participantsin the current study were recruited as part of alarger ongoing longitudinal study investigating abroader range of biological, clinical, and psychoso-cial outcomes in first episode mania (17). Based onrecruitment patterns for this larger study we wereable to estimate that approximately 55% ofscreened first-episode patients were excluded forthe following reasons: 36% were determined tohave had previous manic episodes, 45% weredetermined not to have bipolar I disorder (hadbipolar II disorder or another Axis I diagnosis),and 18% for medical reasons. Of an initial pool of78 eligible patients who met study criteria andsigned informed consent, a further 13 decided towithdraw their participation, yielding a finalsample of 65.

Thirty-six healthy individuals, who were compa-rable to the patient sample with regard to age, gen-der, ethnicity, and premorbid IQ, were recruited

160

Torres et al.

from the community through advertisementsposted at UBC, local hospitals, and other commu-nity locations. Healthy subjects were also assessedusing the MINI, and exclusion criteria included apersonal or family history of major psychiatric dis-order in first- or second-degree relatives, or majormedical or neurological illness affecting cognition.Out of the healthy individuals screened, approxi-mately 50% were excluded for the following rea-sons: medical history (30%), psychiatric history(30%), both medical and psychiatric history(20%), or other reason (20%). Ethics approval wasreceived from the UBC Clinical Research EthicsBoard, and written informed consent was obtainedfrom all subjects prior to participation.

Neuropsychological assessment

The cognitive measures employed in the presentstudy were selected based on demonstration of theirrelevance to bipolar disorder (2, 19). The batteryconsisted of standardized clinical neuropsychologicalmeasures, including select subtests from the Cam-bridge Neuropsychological Test Automated Battery(CANTAB) (20). Although the individual cognitivetests were the same as those used in our previousstudies (1, 16), the grouping of the tasks into cogni-tive domains was slightly modified to better reflectcurrent understanding of cognitive functioning inbipolar disorder. Our selection of cognitive domains(processing speed, attention, verbal memory, non-verbal memory, working memory, executivefunction) and inclusive tests was thus based heavilyon the approach taken with the Measurement andTreatment Research to Improve Cognition inSchizophrenia Consensus Cognitive Battery(MCCB) (21). Although this battery was developedfor use in schizophrenia, its utility for use in bipolardisorder is based on both the similarity in cognitiveprofile between schizophrenia and bipolar disorder(22), and also its empirical validation in bipolar dis-order (23). The individual measures included withineach domain and the rationale for their inclusioninto each domain are presented below:

Processing speed. The Trailmaking Test time tocomplete Part A (24) was included in this domainas in the MCCB battery. Letter fluency numbercorrect (25) was included in this domain owing toits observed psychometric similarity with categoryfluency (included in the MCCB). This similaritywas documented during the development of theMCCB (21). The Stroop test (26) word- and color-naming trials (number correct) were also includedin this domain as these measures often emergeunder ‘processing speed’ factors in factor analytic

neuropsychological studies in schizophrenia (27)and bipolar disorder (28).

Attention. The measures included in the attentiondomain were the CANTAB Rapid Visual Informa-tion Processing (RVIP) discriminability score andlatency score. Although these measures are notderived from exactly the same Continuous Perfor-mance Test (CPT) task employed in the MCCB, theRVIP is also a test of sustained attention thatrequires the individual to respond to specified visualtarget stimuli that are sequentially presented over aspan of several minutes. A review of prior factoranalytic studies in schizophrenia that was con-ducted as a precursor to the development of theMCCB demonstrated that different variants of con-tinuous performance tasks frequently load highlyon attention/vigilance neuropsychological factors(27). Moreover, prior research in bipolar disordersupports the sensitivity of the included RVIP mea-sures in bipolar disorder (1, 29).

Verbal memory. For verbal memory, the Califor-nia Verbal Learning Test, second edition (CVLT-II) (30) recall trials 1–5 and CVLT-II delayed freerecall trial were included. The CVLT-II wasincluded as a candidate measure of verbal memoryin the development of the MCCB (21), and theincluded CVLT-II measures show high sensitivityand relevance to bipolar disorder (2, 19, 22).

Nonverbal memory. All the measures included inthis domain require nonverbal learning and long-term retention. The specific measures were CAN-TAB Spatial Recognition Memory percent correct,CANTAB Pattern Recognition Memory percentcorrect, and CANTAB Paired Associate Learningtotal errors adjusted score. Factor analytic data inhealthy individuals reveal that these tests loadmost strongly on a visual memory and learningfactor (31).

Working memory. Consistent with the MCCB, theWechsler Memory Scale, third edition (32) Letter/Number Sequencing score was included in theworking memory domain. The CANTAB SpatialWorking Memory (SWM) between errors scorewas also included in this domain as this task wasmodeled after the self-ordered pointing test, whichis a classic test of nonverbal working memory (33).The SWM between errors measure has also beenfound to correlate with spatial span tasks and toload on a working memory factor (34).

Executive function. The measures included in thisdomain tap into several aspects of the broad

161

Longitudinal cognitive functioning

construct of executive function, and these func-tions are largely supported by the integrity of dor-sal prefrontal brain regions (35). The CANTABmeasures included in this domain included theCANTAB Intra Extra Dimensional set shiftingtask number of extra-dimensional shifting errors,and the Stockings of Cambridge problems solvedin the minimum number of moves. These measuresassess mental flexibility/set shifting and planningability, respectively, and are analogs of the Wis-consin Card Sorting Test and Tower of LondonTest, which were among the candidate executive/problem-solving tests considered for inclusion inthe MCCB (21). Two further measures that tapinto the executive components of attentional setshifting and response inhibition included the Trail-making Test B time and Stroop C/W trial numbercorrect, respectively. Inclusion of these measuresinto the executive domain are supported by theadoption of similar versions of these tasks intoexecutive function batteries (36), and by the obser-vation that such measures have been found to loadhighly on executive functioning factors identifiedthrough factor analysis (37).

Procedures

Patients were evaluated at baseline (within threemonths of the end of their first manic episode), andat six months and one year post-baseline. At studyentry, patients received a comprehensive clinicalevaluation including the MINI and a wide range ofpsychiatric symptom and mood rating scales(Table 1). Patients received open-label mainte-nance treatment based on established best practiceguidelines (38). Modifications in treatmentsthroughout the study were made as clinically indi-cated on an individual basis, based on currenttreatment standards. Although patients were seenat six-month intervals, some patients requiredunscheduled visits in between sessions if clinicallyindicated.

Neuropsychological testing was conducted bygraduate student research assistants trained by theclinical neuropsychologist on the team (IJT). Test-ing was conducted when patients were clinicallystable, and mean mania and depression ratings atall three time points confirm that patients showeda low level of mood symptoms (Table 2). Theduration of the cognitive battery was approxi-mately 2.5 hours, and subjects were provided peri-odic breaks during testing. The same cognitivebattery was used at all time points, although alter-nate forms were used during the six-month assess-ment period for the CANTAB memory tasks,CVLT-II, and verbal fluency to minimize practice

effects. The intellectual [Kaufmann Brief Intelli-gence Test] (39) and premorbid intellectual [NorthAmerican Adult Reading Test] (40) measures wereonly administered at baseline.

Table 1. Baseline demographic and clinical variables for participantscompleting all time pointsa

Patients(n = 42)

Healthycontrols(n = 23)

Age, years, mean (SD) 22.9 (4.0) 22.9 (4.9)Education, years, mean (SD) 13.5 (2.0) 14.5 (2.3)North American Adult ReadingTest, mean (SD)

106.6 (7.7) 107.2 (7.4)

Gender, male, n (%) 19 (45) 9 (39)Ethnicity, n (%)

Caucasian 32 (78) 15 (68)Asian 5 (12) 6 (27)Other 4 (10) 1 (5)

English as first language, n (%) 36 (88) 18 (82)Premorbid socioeconomic status, n (%)

Student 23 (55) 15 (65)Part-time work 2 (5) 0 (0)Full-time work 13 (31) 7 (30)Self-employed 1 (2) 0 (0)Unemployed 3 (7) 1 (4)

Age at illness onset, years,mean (SD)b

19.8 (5.0)

Age at depression onset,years, mean (SD)

18.2 (5.6)

No. of previous depressiveepisodes, mean (SD)

1.1 (1.6)

No. of previous hypomanicepisodes, mean (SD)

0.6 (1.8)

Psychiatric rating scales, mean (SD)PANSS–Positive score 7.8 (1.5)Young Mania Rating Scale 1.3 (3.2)Hamilton Depression RatingScale–29 item

7.3 (7.5)

Brief Psychiatric Rating Scale 23.1 (5.9)Global Assessment ofFunctioning Scale

65.0 (12.9)

History of depression, n (%) 21 (50)History of psychosis, n (%) 33 (79)History of alcohol orsubstance abuse, n (%)

22 (52)

Medications, n (%)Mood stabilizers 38 (91)Lithium 18 (43)Divalproex 22 (52)Lamotrigine 1 (2)

Atypical antipsychotic agents 34 (81)Olanzapine 7 (17)Quetiapine 11 (26)Risperidone 19 (45)

PANSS = Positive and Negative Syndrome Scale; SD = stan-dard deviation.aNo significant patient–control group differences were noted onany variable based on two-tailed independent samples t-tests orv2 analyses.bAge at first mood episode of any type.

162

Torres et al.

Statistical analysis

For each primary cognitive measure, raw scoreswere converted into z-scores ranging from �4 to 4based on demographics-adjusted normative dataderived from the test manuals. Demographic-corrected normative values were used to facilitatecalculation of cognitive domain scores and becausenormative databases are based on much largersamples than our healthy control sample. The sixcognitive domain scores were derived for each sub-ject by calculating the mean z-score of all primarymeasures within each cognitive domain. Patientversus control demographics were compared usingeither t-tests or v2-statistics as appropriate.Changes in patient clinical variables across thethree time periods were assessed using repeatedmeasure analysis of variance (ANOVA) for nor-mally distributed variables, Friedman’s v2-test fornon-normally distributed ordinal variables, andCochran’s Q-test for dichotomous variables(frequencies).

The major analytic strategies consisted of (i) adoubly multivariate repeated measure ANOVA ononly the subjects who participated in all three timepoints, and (ii) mixed linear models including allavailable participants at all time points. The dou-bly multivariate ANOVA was conducted on the sixcognitive domain scores using group (patient ver-sus control) as a between-subject factor and time

(baseline, six-month, one-year) as a within-subjectrepeated measure. Because the major hypothesisinvolved assessment of patient–control differencesin cognitive change across time, we focused onevaluation of the group 9 time interactions. Poly-nomial trend analysis (linear and quadratic) wasalso conducted with a focus on the group 9 timeinteractions. To assess whether specific clinicalvariables (change in depressive symptoms, onset ofnew mood episodes, medication, substance abusehistory) could predict change in cognitive function-ing in patients, we conducted trend analysis evalu-ating the interaction between time and each ofthese clinical predictors for relevant cognitivedomains in just the patients.

Mixed linear models were also used to assesscognitive change over time as these methods canhandle using data from all participants (not justthose without missing data), and are more appro-priate when assumptions of sphericity are violated,as often occurs in the case of repeated measure-ments (41). The first model (model 1), tested on allpatients and controls, consisted of an uncondi-tional growth model aimed at determining whethera linear or quadratic trend fitted the data bestacross time for each cognitive domain score. Foreach cognitive domain, the growth curve modelwas estimated using time (coded 0, 1, 2 for base-line, six-month, and one-year assessments, respec-tively) and the quadratic term (coded 0, 1, 4) aslevel 1 predictors, treating the intercept and slopeparameters as random factors. If the quadraticterm was not significant it was dropped frommodel 2. In model 2, group (coded patient = 0,control = 1) was added as a level 2 predictor. Theprimary fixed effect of interest was the interactionbetween time and group (or the interactionbetween the quadratic term and group if it wasincluded in the model). For the models describedabove, multiple repeated (level 1) covariance struc-tures were tested (identity, diagonal, unstructured,autoregressive), and the random effects (level 2)covariance structure was unstructured. All analy-ses, including mixed linear modeling (41), were per-formed using IBM SPSS (v19.0) (SPSS, Chicago,IL, USA).

Results

A total of 65 patients and 36 controls participatedin the study, with the following breakdown of par-ticipants at each time point: baseline: 64 patients,36 controls; six months: 49 patients, 25 controls;one year: 47 patients, 26 controls. Overall, 42patients and 23 controls participated in all threetime periods. For participants who did not have

Table 2. Clinical and medication variables for patients who completedall time points (n = 42)

Baseline Six months One year

Psychiatric rating scales, mean (SD)PANSS–Positive score 7.8 (1.5) 7.2 (1.0) 7.3 (1.9)Brief PsychiatricRating Scale

23.1 (5.9) 20.1 (3.1) 19.7 (3.8)

Young Mania RatingScalea

1.3 (3.2) 1.0 (3.2) 1.5 (4.4)

Hamilton DepressionRating Scale–29 itema

7.3 (7.5) 4.9 (5.7) 3.0 (5.2)

Medications, n (%)Mood stabilizers 38 (91) 37 (90) 34 (83)

Lithium 18 (43) 17 (41) 15 (37)Divalproex 22 (52) 20 (49) 20 (49)Lamotrigine 1 (2) 1 (2) 1 (2)

Atypical antipsychoticagents

34 (81) 27 (66) 22 (54)

Olanzapine 7 (17) 6 (15) 7 (17)Quetiapine 11 (26) 12 (29) 9 (22)Risperidone 19 (45) 9 (22) 6 (15)

PANSS = Positive and Negative Syndrome Scale; SD = stan-dard deviation.aFor 40% of the sessions, mood ratings were obtained on theday of cognitive testing; for 55% of sessions, within three days oftesting; for 77%, within two weeks of testing; and for 91%, withinone month of testing.

163

Longitudinal cognitive functioning

cognitive data at all three time points, the reasonswere as follows: (i) 11 patients and three controlsmissed or declined a scheduled appointment; (ii)eight patients and three controls dropped out ofthe study; (iii) two patients and seven controls hadnot yet reached a scheduled time point, and (iv)two patients were too clinically unstable to providevalid test data during at least one time point.Patients who completed all three time points(n = 42) were comparable to patient noncomplet-ers (n = 23) with regard to age [completers: 22.9,standard deviation (SD) = 4.0 years; noncomplet-ers: 23.3 (SD = 5.3) years (t63 = 0.35, p = 0.73)],education [completers: 13.5 (SD = 2.0) years; non-completers: 13.7 (SD = 2.5) years (t63 = 0.27, p =0.79)], gender [completers: 45% male; noncomplet-ers: 52% male (v2 = 0.29, p = 0.59)], premorbidIQ [completers: 106.6 (SD = 7.7); noncompleters:107.5 (SD = 6.6) (t63 = 0.49, p = 0.63)], age of ill-ness onset [completers: 19.8 (SD = 5.0) years; non-completers: 19.9 (SD = 5.6) years (t61 = 0.05,p = 0.96)], and history of substance abuse [compl-eters: 52%; noncompleters: 45% (v2 = 0.30,p = 0.59)].

Reliability of domain scores

Test–retest reliabilities between baseline and sixmonths were computed for each domain score inthe healthy sample, and these were as follows: pro-cessing speed r = 0.93; attention r = 0.71; verbalmemory r = 0.68; nonverbal memory r = 0.73;working memory r = 0.71; executive functionr = 0.77.

Data analysis on participants completing all three

time points

Patients were comparable to controls on age, pre-morbid IQ, and gender (Table 1). At baseline,91% and 81% of patients were treated with moodstabilizers and atypical antipsychotic agents,

respectively. Patients also showed a low level ofmood symptoms based on baseline depression andmania ratings. Overall, 52% of the sample metDSM-IV criteria for history of alcohol or sub-stance abuse or dependence, with 43% reportingcurrent use (marijuana 33%, alcohol 10%, other10%). Nearly all patients presented with a first epi-sode of pure mania (three with mixed episode), andpsychosis was present in 79%. Mean duration ofillness (since mood episode of any type) was2.9 years (SD = 4.4). One-half of the patients hada history of at least one depressive episode prior tothe first manic episode. Patient clinical characteris-tics are detailed further in Table 1.

Patient–control differences in cognitive changeacross time. Mean cognitive domain scores forpatients and controls across all time points arepresented in Table 3.

The doubly multivariate ANOVA revealed a sig-nificant multivariate group by time interaction oncognitive domain scores (Wilk’s lambda = 0.61,F12,52 = 2.8, p = 0.006). Univariate tests revealed asignificant group by time interaction for processingspeed (F2,126 = 5.9, p = 0.003) and executivefunction (F2,126 = 4.0, p = 0.02), but not for othercognitive domains (all interactions p > 0.20).Group main effects indicating poorer patient per-formance relative to controls were present forverbal memory (F1,63 = 10.4, p = 0.002), nonver-bal memory (F1,63 = 4.9, p = 0.03), and workingmemory (F1,63 = 5.5, p = 0.02). Trend analysisrevealed a significant linear group by time interac-tion for processing speed (F1,63 = 9.0, p = 0.004)and executive function (F1,63 = 7.5, p = 0.008), butnot other cognitive domains. There were no inter-actions between group and the quadratic term forany of the cognitive measures (all p > 0.10). Fig-ure 1 illustrates that for both processing speed andexecutive function, patients showed a significantlysteeper linear increase in performance across timethan healthy controls. In addition, post-hoc

Table 3. Cognitive performance across time periods in patients and controls

Baseline Six months One year

Patientsmean (SD)

Controlsmean (SD)

Effectsize

Patientsmean (SD)

Controlsmean (SD)

Effectsize

Patientsmean (SD)

Controlsmean (SD)

Effectsize

Processing speed �0.51 (0.73) �0.08 (0.87) 0.54 �0.19 (0.74) �0.01 (0.90) 0.22 �0.03 (0.73) 0.03 (0.92) 0.07Attention �0.15 (0.88) 0.09 (0.63) 0.30 0.23 (0.82) 0.61 (0.61) 0.50 0.36 (1.09) 0.86 (0.65) 0.51Verbal memory �0.07 (1.04) 0.54 (0.94) 0.60 �0.35 (1.03) 0.62 (0.92) 0.97 0.36 (1.12) 1.03 (1.00) 0.61Nonverbal memory 0.02 (0.78) 0.42 (0.56) 0.56 0.24 (0.73) 0.54 (0.61) 0.43 0.15 (1.01) 0.58 (0.66) 0.48Working memory �0.29 (0.93) 0.31 (0.96) 0.64 0.00 (0.82) 0.38 (0.74) 0.48 0.06 (0.80) 0.44 (0.75) 0.48Executive function �0.11 (0.77) 0.45 (0.66) 0.75 0.11 (0.70) 0.46 (0.67) 0.50 0.32 (0.72) 0.47 (0.81) 0.20

SD = standard deviation.

164

Torres et al.

pairwise comparisons showed significantly poorerprocessing speed in patients relative to controls atbaseline (p = 0.04) but not at other time points.For executive function, patients showed poorerperformance than controls at baseline (p = 0.004)and a trend toward poorer performance at sixmonths (p = 0.06), but no differences at one year.

Change in patient clinical variables across time.There was no significant difference in the YoungMania Rating Scales scores across the three timeperiods (Friedman’s v2 = 0.95, p = 0.62) (Table 2).By contrast, patients showed a significant improve-ment in Hamilton Depression Rating Scale–29item (HAM-D) ratings across time (Friedman’sv2 = 10.4, p = 0.006) that was mirrored byimprovement in Brief Psychiatric Rating Scalescores (Friedman’s v2 = 13.3, p = 0.001). Psy-chotic symptoms remained low at all time points asonly six patients at baseline [mild (n = 4), moder-ate (n = 1), moderate/severe (n = 1)], one at sixmonths (mild), and none at one year endorsed ascore of 3 (mild) or more on any Positive and Neg-ative Syndrome Scale item. Overall, 55% ofpatients experienced at least one new mood episodein the year after baseline. Treatment with moodstabilizers was highly consistent across the threetime periods (Table 2) as there was no significantdifference in the proportion of patients treatedwith mood stabilizers at each time point (Coch-ran’s Q = 2.0, p = 0.37) During the study, threepatients were started on a new mood stabilizer [val-proate (n = 1), lithium (n = 1), lamotrigine(n = 1)], seven patients were taken off an existingmood stabilizer [valproate (n = 3), lithium (n = 3),lamotrigine (n = 1)], and two patients wereswitched to a new mood stabilizer.

In contrast to mood stabilizer treatment, atypi-cal antipsychotic treatment was more variable and

decreased across time (Cochran’s Q = 11.6,p = 0.003), with 81%, 66%, and 54% of patientsreceiving treatment during the baseline, six-month,and one-year periods, respectively. Among the spe-cific antipsychotic agents, there was a particularreduction in the proportion of patients treated withrisperidone (45%, 22%, and 15% across the threetime periods, respectively). At the individual level,one patient was started on a new antipsychoticagent (quetiapine), 11 patients were taken off of anexisting antipsychotic agent [risperidone (n = 6),risperidone and olanzapine (n = 1), quetiapine(n = 2), quetiapine and loxapine (n = 1), quetia-pine and olanzapine (n = 1)], and 12 were switchedfrom one or more antipsychotic agents to at leastone other (six different combinations) at somepoint during the study.

Given the naturalistic treatment approach andthe multiple combinations of medication changesthat occurred throughout the study (especiallyantipsychotic agents), it was not possible to com-prehensively evaluate the influence of a change intreatment in all individual agents on longitudinalcognitive functioning. Nevertheless, we were ableto divide patients into those who were taken offantipsychotic agents during the course of the studyversus other patients, in order to compare theirlongitudinal cognitive performance (see below).

Clinical predictors of change on longitudinal cogni-tive functioning in patients. To investigate whetherthe steeper increase in processing speed and execu-tive function in patients might be explained by lon-gitudinal change in various clinical variables, weconducted trend analysis on the processing speedand executive function measures across the threetime points (time). Specifically, we evaluatedwhether either linear or quadratic components ofthe time effect interacted with the following clinical

Fig. 1. Patient and control cognitive functioning across three time points for (A) processing speed and (B) executive function. Valuesrepresent mean scores and error bars are standard errors. *Significant group difference at p < 0.05.

165

Longitudinal cognitive functioning

variables: taken off an antipsychotic agent at somepoint after baseline (yes = 11, no = 29), experi-enced a new mood episode after baseline (yes = 22,no = 18), history of alcohol or substance abuse(yes = 22, no = 20), history of depression prior tomanic episode (yes = 21, no = 21), and change inHAM-D rating score between baseline and one-year time points (continuous variable). Forprocessing speed, none of the interactions reachedsignificance (all p > 0.10). For executive function,there was a significant linear interaction betweentime and both history of substance abuse(F1,40 = 11.8, p = 0.001) and whether patients weretaken off an antipsychotic agent during the study

(F1,38 = 8.4, p = 0.006). Specifically, patients witha history of alcohol or substance abuse showedpoorer cognitive recovery than nonabusers, andpatients taken off an antipsychotic agent showedbetter recovery than the rest of the patient sample(Fig. 2).

Data analysis on all participants using mixed

effect models

The model 1 analyses (Table 4) revealed a signif-icant linear increase in performance across timefor all cognitive domains except executivefunction.

Fig. 2. Longitudinal cognitive functioning among different patient groups. Comparison between (A) patients who discontinued anti-psychotic treatment during the study versus those who did not, and (B) patients with substance abuse history versus no such history.Values represent mean scores and error bars are standard errors.

Table 4. Summary of mixed linear model analyses

Modela Fixed effectc

Parameter estimates (standard error)

Processing speed Attention Verbal memoryNonverbalmemory

Workingmemory

Executivefunction

1 Intercept �0.36 (0.07)f �0.13 (0.08) 0.06 (0.11) 0.10 (0.07) �0.10 (0.09) 0.03 (0.07)Time 0.28 (0.08)f 0.59 (0.11)f �0.53 (0.20)e 0.31 (0.12)e 0.30 (0.13)d 0.21 (0.11)Quadratic �0.06 (0.04) �0.14 (0.05)e 0.37 (0.10)f �0.12 (0.06)d �0.08 (0.07) �0.02 (0.06)

2b Intercept �0.46 (0.09)f �0.27 (0.10)e �0.23 (0.12) �0.06 (0.09) �0.30 (0.11)e �0.16 (0.09)Time 0.21 (0.03)f 0.58 (0.14)f �0.67 (0.23)e 0.40 (0.14)e 0.18 (0.05)e 0.23 (0.04)f

Group 0.33 (0.15)d 0.41 (0.16)d 0.82 (0.20)f 0.44 (0.14)e 0.63 (0.18)e 0.54 (0.14)f

Time 9 group �0.13 (0.06)d 0.04 (0.23) 0.42 (0.39) �0.22 (0.25) �0.08 (0.09) �0.18 (0.07)e

Quadratic – �0.15 (0.06)d 0.45 (0.11)f �0.17 (0.07)d – –Quadratic 9 group – 0.01 (0.11) �0.23 (0.19) 0.13 (0.12) – –

aReported results are for models employing identity repeated (level 1) covariance matrix and unstructured random (level 2) covariancematrix. The same pattern of results was obtained for fixed effects when diagonal, unstructured, and autoregressive repeated covariancematrices were used.bQuadratic terms were dropped from model 2 if the quadratic term was not significant in model 1.cGroup coded as follows: 0 = patients; 1 = controls.dp < 0.05 indicates significant fixed effect.ep < 0.01 indicates significant fixed effect.fp < 0.001 indicates significant fixed effect.

166

Torres et al.

However, when the nonsignificant quadraticterm was removed from the model for executivefunction, the linear effect of time was also signifi-cant (F1,81.0 = 24.7, estimate = 0.16, standarderror = 0.03, p < 0.001). In addition, for attention,verbal memory, and nonverbal memory, the signif-icant linear effect was accompanied by a significantquadratic effect (Table 4); therefore, the quadraticterm was retained in the model 2 analyses for thesecognitive domains.

In the model 2 analyses, group was added as apredictor, and the interactions between group andeither the linear or quadratic effect of time wereevaluated. Table 4 shows that the only significantinteractions observed were the group by linear timeinteraction for processing speed (F1,78.8 = 5.1,p < 0.05) and executive function (F1,77.8 = 7.3,p < 0.01). Thus, consistent with the repeated mea-sure ANOVA analysis, the results revealed differ-ential between-group linear trajectories forprocessing speed and executive function.

Discussion

The primary finding of the present study is that,relative to demographically comparable healthyindividuals, patients with bipolar disordershowed evidence of selective improvement incognitive functioning in the first year after reso-lution of their first lifetime manic episode. Spe-cifically, patients demonstrated improvement inprocessing speed and executive function. Forboth of these domains, patients showed signifi-cantly poorer performance than healthy controlsat baseline, but, due to improvements acrosstime, there were no significant group differencesin these domains by the one-year point. Consis-tent with prior studies in early bipolar disorder(1, 10, 11), patients also showed poorer perfor-mance than healthy controls in other domains,including working memory, nonverbal memory,and verbal memory. However, unlike processingspeed and executive function, these other difficul-ties remained consistent across time.

The pattern of findings suggests that the process-ing speed and executive deficits noted shortly afterthe first manic episode are somewhat transient andvariable across time, at least early in the course ofillness. These changes are thus more likely to reflectunstable, functional, or neuroplastic mechanisms.Discussion of the potential factors that may be orare unlikely to be driving these cognitive changes ispresented later. By contrast, verbal learning/mem-ory deficits and, to some extent, difficulties in non-verbal memory and working memory continue topersist throughout the first year after diagnosis. In

particular, verbal learning/memory difficulty mayrepresent a particularly stable and fixed cognitivedeficit characterizing bipolar disorder, and thus acandidate endophenotype of the illness. This isconsistent with reports that verbal learning/mem-ory deficits are observed in children (42), adults (2,43), and elderly individuals (44) diagnosed with theillness. Moreover, the finding that unaffected first-degree relatives of patients can also exhibit dimin-ished verbal learning (45, 46) suggests that thiscognitive vulnerability may have at least a partialdevelopmental or genetic basis (47, 48).

Given the documented association between neu-ropsychological functioning and functional out-come both in first manic episode (16) and moreestablished bipolar illness (49, 50), the finding ofpotential cognitive improvement shortly after diag-nosis is encouraging. In our previous study thatinvestigated the cognitive predictors of six-monthlongitudinal functional outcomes after the firstmanic episode, verbal learning/memory wasobserved to be most robustly associated with psy-chosocial functioning six months after cognitiveassessment (16). From this perspective, it is unfor-tunate that we did not observe patients improve inthis domain over follow-up as this could beexpected to have the most beneficial impact ondaily functional outcome. Nevertheless, given thefrequently reported association between executivefunctioning or processing speed and functionaloutcome (50), it can be surmised that the cognitivegains that patients exhibited in the present studyhave the potential to result in meaningful improve-ment in daily patient functioning.

To follow-up on the finding that patients andhealthy individuals showed differential longitudi-nal patterns of change in processing speed andexecutive function, a secondary goal of the presentstudy was to evaluate the degree to which certainvariables might predict differential cognitive trajec-tories among the patients. In other words, werethere any variables that predicted better or worsecognitive recovery in processing speed or executivefunction in patients? There was no evidence thatmood symptom variables predicted improvementin these cognitive functions. Specifically, there wasno association between change in depression rat-ings between baseline and one year, and improve-ment in executive functioning or processing speed,nor was there a difference in cognitive trajectoriesbetween those who experienced a new mood epi-sode during the study and those who did not. Inone regard, these findings are not surprising aspatients were clinically stable at all three timepoints, as indicated by mood ratings taken at eachinterval. Based on these findings, cognitive

167

Longitudinal cognitive functioning

improvement in the patient sample was not likelydue to resolution of subsyndromal symptoms.

In contrast to symptom variables, comorbidalcohol or substance abuse history in patients wasassociated with poorer recovery in executive func-tioning. This finding is generally consistent withprevious reports that adult bipolar patients withsubstance abuse comorbidity show poorer neuro-psychological functioning than those without suchcomorbidity (51–54), although there have beenconflicting findings (48, 55, 56). Nevertheless,because almost all patients in our sample with analcohol or substance abuse history reported thatthey were still using substances concurrently (atbaseline), it is unclear whether prior history of useor current use contributes more prominently to thepoor cognitive recovery that was observed. Priorresearch suggests that previous substance abusehistory, as opposed to current abuse, may be suffi-cient to associate with diminished neuropsycholog-ical functioning (51), and if so, this may be thebasis for diminished recovery in our patients. Onthe other hand, it cannot be ruled out that patientswith an alcohol or substance abuse history mayeventually recover to a level approaching theirnonabusing peers, particularly if they cease theirsubstance misuse. Further research will be neededto address this possibility.

Due to the individualized and naturalistic natureof the medical treatments in the present study, itwas not possible to evaluate the impact of medica-tion changes on longitudinal cognitive functioningin a systematic and detailed manner. In particular,assessment of the influence of mood-stabilizerchange on cognition was not feasible owing to themarked stability of treatment with these agentsacross all time points. However, because there wasconsiderably more variability in treatment withatypical antipsychotic medications across the threetime points, we were able to provide at least a grossassessment of the effect of change in treatment withthese agents on cognition. It was observed thatpatients who were taken off an antipsychotic agentduring the study tended to show better recovery inexecutive function than the remainder of patients.This finding, however, should be interpreted cau-tiously and requires replication because (i) it waspost-hoc in nature and was not observed in the con-text of a prospective randomized clinical trial, and(ii) the group of patients taken off antipsychoticagents showed poorer baseline executive function-ing than the comparison group (Fig. 2). Thus, it ispossible that the observed improvement in the for-mer group may have been somewhat exaggeratedbecause this group had more room to show recov-ery than the comparison group (ceiling effects), or

as a result of regression effects. Nevertheless, thesedata may be consistent with other emerging reportssuggesting that treatment with some atypical anti-psychotic agents may have a negative impact onneuropsychological functioning in patients withbipolar disorder (2, 57, 58), and that stable firstpsychotic episode patients maintained on antipsy-chotic agents show poorer cognitive outcomes thanthose who discontinue such treatments (59).

Taken together, the data emerging from thepresent study suggest that the early cognitivecourse of cognitive functioning in bipolar disorderis likely complex and impacted by multiple factors,with cognitive deficits and improvements likelyarising from multiple sources. For example, earlyimprovements in cognitive function may at leastpartly result from neuroprotective or neurotrophiceffects of mood stabilizers such as lithium or val-proate (60). These medications may also exertopposing acute mild cognitive side effects (61, 62).The net effect may be that existing psychotropictreatments for bipolar disorder early in the courseof illness may have both beneficial longer-term anddetrimental acute effects on cognition. Anotherpossibility is that the cognitive improvementsobserved in the present study, although not linkedto symptomatic improvement directly, may never-theless reflect gradual improvement in residualcognitive impairment due to the initial manic epi-sode. From this point of view, cognitive deficitsmay take longer to recover than acute mood symp-toms, similar to the way that poor daily function-ing persists in bipolar disorder even aftersymptomatic recovery (63).

Despite the methodological strengths of thepresent study, including the longitudinal design,the sample size, the use of a cognitive battery tai-lored for bipolar disorder, and inclusion of ahealthy control group to mitigate practice effects,there are several methodological limitations thatshould be considered. First, as was the case withour prior studies on this cohort, our participants(patients and controls) tended to be above averagein intellectual functioning. Thus, in manyinstances, patients’ domain scores did not fall inthe impaired range relative to normative data (seeTable 3), and their poorer cognitive functioningwas only revealed in reference to performance ofthe (above average) healthy control group. The useof a high average sample could impact the general-izability of the results as it is unclear whether thepresent findings extend to lower educated orpoorer functioning patients with bipolar disorder.The future study of less educated patients andhealthy volunteers could also minimize the possi-bility that the observed effects may have been

168

Torres et al.

partly due to ceiling performance on some serialtests by the control group. For example, in healthywell-educated groups, individuals could reach anasymptotic range of ‘maximal’ performance beforepatients, thus accentuating the improvementshown in the patient sample, and minimizing thepotential improvement in controls. In a similarvein, although the use of a healthy control grouphelps to minimize practice or learning effects acrossrepeated measures, this is based on the assumptionthat such effects will be comparable betweengroups. However, if learning is not uniform acrossgroups, this could have an impact on the interpre-tation of results.

A further potential confound may be that indi-viduals in the healthy control group were includedonly if they did not have a history of mental illnessin first- or second-degree relatives. While thisapproach helps to link any observed patient–con-trol differences in cognition more directly to bipo-lar illness, it does so at the potential expense ofgeneralizability as a family history of psychiatricillness is not uncommon in the general population.Thus, the possibility that the patient–control differ-ences reported here may have been overestimatedby the use of a ‘super-normal’ control group can-not be ruled out (64). A final issue potentiallyimpacting the generalizability of findings is the factthat our sample showed a high percentage of indi-viduals with a history of psychosis. Although thismay, in part, be due to our liberal definition of psy-chosis (including mood-congruent and mood-incongruent features), it suggests that further workin nonpsychotic individuals is necessary beforegeneralizing findings to the broader range ofpatients with the illness.

Another potential problem for this (and most)longitudinal studies is the possibility of selectiveattrition, which could also compromise the inter-pretation of results. In our analysis, we were notable to identify any clear clinical or demographicdifferences between the patients who completedall three time points and those who did not.Moreover, the replication of our main findingsusing linear mixed model techniques that capital-ize on the use of all collected data points makesit less likely that findings were seriously biased.Nevertheless, it is impossible to know if non-completers may have had a different cognitivetrajectory than completers, and thus the impactof selective attrition remains a possibility.Finally, our main measures of previous and cur-rent alcohol/substance abuse history were takenat the baseline time point, and we did not havea reliable measure of current or ongoing use dur-ing the study. Thus, we were not able to assess

the potential impact of ongoing substance use onthe observed cognitive changes.

Despite these limitations, the present study pro-vides desperately needed information about theearly longitudinal course of neuropsychologicalfunctioning in bipolar disorder patients after theirfirst manic episode. Findings suggest that earlycognitive outcomes are generally favorable inrecently diagnosed patients with bipolar disorder;however, more research is needed to help tease outthe multiple factors that likely contribute to cogni-tive impairment and cognitive change during theearly course of the illness.

Acknowledgements

The data for this manuscript were generated from the System-atic Treatment Optimization Program for Early Mania, whichwas supported by unrestricted grant funding from AstraZenecato LNY.

Disclosures

IJT has received research funding from the Canadian Institutesof Health Research (CIHR). DJB has received research grantsfrom, has been a member of advisory boards for, and has beena speaker for CIHR, the UBC Institute of Mental Health/Coast Capital Depression Research Fund, the Canadian Net-work for Mood and Anxiety Treatments (CANMAT), theCanadian Psychiatric Association, AstraZeneca, Janssen-Ortho, Bristol-Myers Squibb, and Otsuka. WGH has receivedgrant support from Eli Lilly & Co. and AstraZeneca; hasserved on advisory boards for AstraZeneca, Janssen, Pfizer,and Wyeth/Solvay; has been a consultant to AstraZeneca andIn Silico; and has received honoraria from Pfizer, AstraZeneca,and Janssen. RWL has received research funding from, and ison ad-hoc speaker/advisory boards for AstraZeneca, Biovail,Bristol-Myers Squibb, CIHR, CANMAT, Common DrugReview, Eli Lilly & Co., Litebook Company, Ltd., Lundbeck,Lundbeck Institute, Pfizer, Servier, St. Jude’s Medical, andUBC Institute of Mental Health/Coast Capital Savings. LNYhas received research grant funding from, has been a memberof advisory boards for, and has been a speaker for AstraZene-ca, Janssen, Eli Lilly & Co., GlaxoSmithKline, Bristol-MyersSquibb, Novartis, Servier, Lundbeck, Merck, and Pfizer. JKand SP have no conflicts of interest to disclose.

References

1. Torres IJ, Defreitas VG, Defreitas CM et al. Neurocogni-tive functioning in patients with bipolar I disorder recentlyrecovered from a first manic episode. J Clin Psychiatry2010; 71: 1234–1242.

2. Bora E, Yucel M, Pantelis C. Cognitive endophenotypes ofbipolar disorder: a meta-analysis of neuropsychologicaldeficits in euthymic patients and their first-degree relatives.J Affect Disord 2009; 113: 1–20.

3. Gildengers AG, Mulsant BH, Begley A et al. The longitu-dinal course of cognition in older adults with bipolar disor-der. Bipolar Disord 2009; 11: 744–752.

4. Arts B, Jabben N, Krabbendam L, van Os J. A 2-year nat-uralistic study on cognitive functioning in bipolar disorder.Acta Psychiatr Scand 2011; 123: 190–205.

169

Longitudinal cognitive functioning

5. Depp CA, Savla GN, Moore DJ et al. Short-term courseof neuropsychological abilities in middle-aged and olderadults with bipolar disorder. Bipolar Disord 2008; 10: 684–690.

6. Moorhead TW, McKirdy J, Sussmann JE et al. Progres-sive gray matter loss in patients with bipolar disorder. BiolPsychiatry 2007; 62: 894–900.

7. Mur M, Portella MJ, Martinez-Aran A, Pifarre J, Vieta E.Neuropsychological profile in bipolar disorder: a prelimin-ary study of monotherapy lithium-treated euthymic bipo-lar patients evaluated at a 2-year interval. Acta PsychiatrScand 2008; 118: 373–381.

8. Mur M, Portella MJ, Martinez-Aran A, Pifarre J, Vieta E.Long-term stability of cognitive impairment in bipolar dis-order: a 2-year follow-up study of lithium-treated euthymicbipolar patients. J Clin Psychiatry 2008; 69: 712–719.

9. Delaloye C, Moy G, de Bilbao F et al. Longitudinal analy-sis of cognitive performances and structural brain changesin late-life bipolar disorder. Int J Geriatr Psychiatry 2011;26: 1309–1318.

10. Hill SK, Reilly JL, Harris MS et al. A comparison of neu-ropsychological dysfunction in first-episode psychosispatients with unipolar depression, bipolar disorder, andschizophrenia. Schizophr Res 2009; 113: 167–175.

11. Zanelli J, Reichenberg A, Morgan K et al. Specific andgeneralized neuropsychological deficits: a comparison ofpatients with various first-episode psychosis presentations.Am J Psychiatry 2010; 167: 78–85.

12. Vieta E, Reinares M, Rosa AR. Staging bipolar disorder.Neurotox Res 2011; 19: 279–285.

13. Berk M, Kapczinski F, Andreazza AC et al. Pathwaysunderlying neuroprogression in bipolar disorder: focus oninflammation, oxidative stress and neurotrophic factors.Neurosci Biobehav Rev 2011; 35: 804–817.

14. Kapczinski F, Vieta E, Andreazza AC et al. Allostaticload in bipolar disorder: implications for pathophysiol-ogy and treatment. Neurosci Biobehav Rev 2008; 32:675–692.

15. Post RM. Mechanisms of illness progression in the recur-rent affective disorders. Neurotox Res 2010; 18: 256–271.

16. Torres IJ, DeFreitas CM, DeFreitas VG et al. Relation-ship between cognitive functioning and 6-month clinicaland functional outcome in patients with first manic episodebipolar I disorder. Psychol Med 2011; 41: 971–982.

17. Yatham LN, Kauer-Sant’Anna M, Bond DJ, Lam RW,Torres I. Course and outcome after the first manic episodein patients with bipolar disorder: prospective 12-monthdata from the systematic treatment optimization programfor early mania project. Can J Psychiatry 2009; 54: 105–112.

18. Sheehan DV, Lecrubier Y, Sheehan KH et al. The mini-international neuropsychiatric interview (M.I.N.I.): thedevelopment and validation of a structured diagnostic psy-chiatric interview for DSM-IV and ICD-10. J Clin Psychia-try 1998; 59 (Suppl. 20): 22–33.

19. Torres IJ, Boudreau VG, Yatham LN. Neuropsychologi-cal functioning in euthymic bipolar disorder: a meta-analy-sis. Acta Psychiatr Scand 2007; 116 (Suppl. 434): 17–26.

20. Robbins TW, James M, Owen AM, Sahakian BJ, McInnesL, Rabbitt P. Cambridge neuropsychological test auto-mated battery (CANTAB): a factor analytic study of alarge sample of normal elderly volunteers. Dementia 1994;5: 266–281.

21. Nuechterlein KH, Green MF, Kern RS et al. The MAT-RICS consensus cognitive battery, part 1: test selection,reliability, and validity. Am J Psychiatry 2008; 165: 203–213.

22. Yatham LN, Torres IJ, Malhi GS et al. The InternationalSociety for Bipolar Disorders–Battery for Assessment ofNeurocognition (ISBD–BANC). Bipolar Disord 2010; 12:351–363.

23. Burdick KE, Goldberg TE, Cornblatt BA et al. The MAT-RICS consensus cognitive battery in patients with bipolar Idisorder. Neuropsychopharmacology 2011; 36: 1587–1592.

24. Reitan RM, Wolfson D. The Halstead-Reitan Neuropsy-chological Test Battery. Tucson, AZ: NeuropsychologyPress, 1985.

25. Lezak MD. Neuropsychological Assessment, 2nd edn.New York, NY: Oxford University Press, 2004.

26. Golden CJ, Freshwater SM. A Manual for the AdultStroop Color and Word Test. Chicago, IL: Stoelting, 2002.

27. Nuechterlein KH, Barch DM, Gold JM, Goldberg TE,Green MF, Heaton RK. Identification of separable cogni-tive factors in schizophrenia. Schizophr Res 2004; 72: 29–39.

28. Czobor P, Jaeger J, Berns SM, Gonzalez C, Loftus S. Neu-ropsychological symptom dimensions in bipolar disorderand schizophrenia. Bipolar Disord 2007; 9: 71–92.

29. Clark L, Iversen SD, Goodwin GM. Sustained attentiondeficit in bipolar disorder. Br J Psychiatry 2002; 180: 313–319.

30. Delis DC, Kramer JH, Kaplan E, Ober BA. CaliforniaVerbal Learning Test, Second Edition, Adult Version. SanAntonio, TX: The Psychological Corporation, 2000.

31. Robbins TW, James M, Owen AM et al. A study of per-formance on tests from the CANTAB battery sensitive tofrontal lobe dysfunction in a large sample of normal volun-teers: implications for theories of executive functioningand cognitive aging. Cambridge Neuropsychological TestAutomated Battery. J Int Neuropsychol Soc 1998; 4: 474–490.

32. Wechsler D. The Wechsler Memory Scale–III manual. SanAntonio, TX: The Psychological Corporation, 1997.

33. Strauss E, Sherman EMS, Spreen O. A Compendium ofNeuropsychological Tests: Administration, Norms, andCommentary, 3rd edn. New York, NY: Oxford UniversityPress, 2006.

34. Lehto JE, Juuj€arvi P, Kooistra L, Pulkkinen L. Dimen-sions of executive functioning: evidence from children. Br JDev Psychol 2003; 21: 59–80.

35. Stuss DT, Levine B. Adult clinical neuropsychology: les-sons from studies of the frontal lobes. Annu Rev Psychol2002; 53: 401–433.

36. Delis DC, Kaplan E, Kramer JH. Delis–Kaplan ExecutiveFunction System. San Antonio, TX: The PsychologicalCorporation, 2001.

37. de Frias CM, Dixon RA, Strauss E. Structure of four exec-utive functioning tests in healthy older adults. Neuropsy-chology 2006; 20: 206–214.

38. Yatham LN, Kennedy SH, Schaffer A et al. CanadianNetwork for Mood and Anxiety Treatments (CANMAT)and International Society for Bipolar disorders (ISBD) col-laborative update of CANMAT guidelines for the manage-ment of patients with bipolar disorder: update 2009.Bipolar Disord 2009; 11: 225–255.

39. Kaufman AS, Kaufman NL. Kaufman Brief IntelligenceTest Manual. Circle Pines, MN: American Guidance Ser-vice, 1990.

40. Blair JR, Spreen O. Predicting premorbid IQ: a revision ofthe national adult reading test. Clin Neuropsychol 1989; 3:129–136.

41. Heck RH, Thomas SL, Tabata LN. Multilevel and Longi-tudinal Modeling with IBM SPSS: Quantitative Methodol-ogy Series. New York, NY: Routledge, 2010.

170

Torres et al.

42. Horn K, Roessner V, Holtmann M. Neurocognitive per-formance in children and adolescents with bipolar disor-der: a review. Eur Child Adolesc Psychiatry 2011; 20: 433–450.

43. Senturk V, Goker C, Bilgic A et al. Impaired verbal mem-ory and otherwise spared cognition in remitted bipolarpatients on monotherapy with lithium or valproate. Bipo-lar Disord 2007; 9 (Suppl. 1): 136–144.

44. Delaloye C, de Bilbao F, Moy G et al. Neuroanatomicaland neuropsychological features of euthymic patients withbipolar disorder. Am J Geriatr Psychiatry 2009; 17: 1012–1021.

45. Kulkarni S, Jain S, Janardhan Reddy YC, Kumar KJ,Kandavel T. Impairment of verbal learning and memoryand executive function in unaffected siblings of probandswith bipolar disorder. Bipolar Disord 2010; 12: 647–656.

46. Quraishi S, Walshe M, McDonald C et al. Memory func-tioning in familial bipolar I disorder patients and their rel-atives. Bipolar Disord 2009; 11: 209–214.

47. Balanza-Martinez V, Rubio C, Selva-Vera G et al. Neuro-cognitive endophenotypes (endophenocognitypes) fromstudies of relatives of bipolar disorder subjects: a system-atic review. Neurosci Biobehav Rev 2008; 32: 1426–1438.

48. Savitz JB, van der Merwe L, Stein DJ, Solms M, RamesarRS. Neuropsychological task performance in bipolar spec-trum illness: genetics, alcohol abuse, medication and child-hood trauma. Bipolar Disord 2008; 10: 479–494.

49. Levy B, Manove E. Functional outcome in bipolar disor-der: the big picture. Depress Res Treat 2012; 2012: 949248.

50. Wingo AP, Harvey PD, Baldessarini RJ. Neurocognitiveimpairment in bipolar disorder patients: functional impli-cations. Bipolar Disord 2009; 11: 113–125.

51. Levy B, Monzani BA, Stephansky MR, Weiss RD. Neuro-cognitive impairment in patients with co-occurring bipolardisorder and alcohol dependence upon discharge frominpatient care. Psychiatry Res 2008; 161: 28–35.

52. Sanchez-Moreno J, Martinez-Aran A, Colom F et al. Neu-rocognitive dysfunctions in euthymic bipolar patients withand without prior history of alcohol use. J Clin Psychiatry2009; 70: 1120–1127.

53. Shan C, Lee SY, Chang YH et al. Neuropsychologicalfunctions in Han Chinese patients in Taiwan with bipolarII disorder comorbid and not comorbid with alcohol

abuse/alcohol dependence disorder. Prog Neuropsycho-pharmacol Biol Psychiatry 2011; 35: 131–136.

54. van Gorp WG, Altshuler L, Theberge DC, Wilkins J,Dixon W. Cognitive impairment in euthymic bipolarpatients with and without prior alcohol dependence. A pre-liminary study. Arch Gen Psychiatry 1998; 55: 41–46.

55. Ringen PA, Vaskinn A, Sundet K et al. Opposite relation-ships between cannabis use and neurocognitive functioningin bipolar disorder and schizophrenia. Psychol Med 2010;40: 1337–1347.

56. van der Werf-Eldering MJ, Burger H, Holthausen EA, Al-eman A, Nolen WA. Cognitive functioning in patients withbipolar disorder: association with depressive symptomsand alcohol use. PLoS ONE 2010; 5: e13032.

57. Kozicky JM, Torres IJ, Bond DJ, Lam RW, Yatham LN.Comparison of neuropsychological effects of adjunctiverisperidone or quetiapine in euthymic patients with bipolarI disorder. Int Clin Psychopharmacol 2012; 27: 91–99.

58. Torrent C, Martinez-Aran A, Daban C et al. Effects ofatypical antipsychotics on neurocognition in euthymicbipolar patients. Compr Psychiatry 2011; 52: 613–622.

59. Faber G, Smid HG, Van Gool AR, Wiersma D, Van DenBosch RJ. The effects of guided discontinuation of antipsy-chotics on neurocognition in first onset psychosis. Eur Psy-chiatry 2012; 27: 275–280.

60. Yasuda S, Liang MH, Marinova Z, Yahyavi A, ChuangDM. The mood stabilizers lithium and valproate selec-tively activate the promoter IV of brain-derived neuro-trophic factor in neurons. Mol Psychiatry 2009; 14: 51–59.

61. Gualtieri CT, Johnson LG. Comparative neurocognitiveeffects of 5 psychotropic anticonvulsants and lithium.MedGenMed 2006; 8: 46–46.

62. Wingo AP, Wingo TS, Harvey PD, Baldessarini RJ.Effects of lithium on cognitive performance: a meta-analy-sis. J Clin Psychiatry 2009; 70: 1588–1597.

63. Judd LL, Akiskal HS, Schettler PJ et al. Psychosocial dis-ability in the course of bipolar I and II disorders: a pro-spective, comparative, longitudinal study. Arch GenPsychiatry 2005; 62: 1322–1330.

64. Schwartz S, Susser E. The use of well controls: anunhealthy practice in psychiatric research. Psychol Med2011; 41: 1127–1131.

171

Longitudinal cognitive functioning