Schizophrenia Research 107 (2009) 165–172

Contents lists available at ScienceDirect

Schizophrenia Research

j ourna l homepage: www.e lsev ie r.com/ locate /schres

Long-term patterns of subjective wellbeing in schizophrenia: Cluster,predictors of cluster affiliation, and their relation to recovery criteria in 2842patients followed over 3 years

Martin Lambert a,⁎,1, Benno G. Schimmelmann b,1, Alexander Schacht c, Anne Karowa,Thomas Wagner c, Peter M. Wehmeier c, Christian G. Huber a, Hans-Peter Hundemer c,Ralf W. Dittmann c,d, Dieter Naber a

a Psychosis Centre, Centre for Psychosocial Medicine, Department of Psychiatry and Psychotherapy, University Medical Center Hamburg-Eppendorf, Germanyb Department of Child and Adolescent Psychiatry, University Duisburg-Essen, Germanyc Medical Neuroscience Department, Lilly Deutschland GmbH, Bad Homburg, Germanyd Department of Child and Adolescent Psychosomatics, Centre for Psychosocial Medicine, University Medical Center Hamburg-Eppendorf, Germany

a r t i c l e i n f o

⁎ Corresponding author. Psychosis Centre, Centre fcine, Department for Psychiatry and PsychotherapyCenter Hamburg-Eppendorf, Martinistr. 52, 20246 Ha+49 40 42803 7670; fax: +49 40 42803 5455.

E-mail address: lambert@uke.uni-hamburg.de (M.1 Denotes equal contribution to the study.

0920-9964/$ – see front matter © 2008 Elsevier B.V.doi:10.1016/j.schres.2008.08.035

a b s t r a c t

Article history:Received 17 March 2008Revised 12 August 2008Accepted 25 August 2008Available online 8 October 2008

Objective: To study the longitudinal patterns of subjective wellbeing in schizophrenia usingcluster analysis and their relation to recovery criteria, further to examine predictors for clusteraffiliation, and to evaluate the sensitivity and specificity of baseline subjective wellbeing cut-offs for cluster affiliation.Methods: Data was collected in an observational 36-month follow-up study of 2842 patientswith schizophrenia in Germany. Subjective wellbeing was assessed using the SWN-K scale.Cluster analyses were applied based on Ward's procedure. Predictors were analyzed usinglogistic regression models. Optimal SWN-K total score cut-off points for cluster affiliation wereanalyzed using Cohen's kappa.Results: 4 distinct clusters were identified: a stable low (33%), a stable moderate (31%), a stablehigh (16%), and a cluster with distinct initial improvement and then stable high subjectivewellbeing (20%). Highly concordant patterns were also observed for symptoms, socialfunctioning, and quality of life. Sensitivity and specificity of SWN-K total score cut-offs atbaseline were 82.8% and 63.8% for ≤60 points for the stable low cluster and 84.7% and 95.4% for≥80 points for the stable high cluster. Affiliation to the stable low cluster was related to a 0.6%chance of being in recovery at 3-year endpoint.Conclusions: Long-term patterns of subjective wellbeing are stable and highly concordant withcourse of symptoms, functioning level, and quality of life. Baseline subjective wellbeing cut-offpoints were found to be sufficient predictors of outcome, which, particularly in case of impairedsubjective wellbeing and low baseline functioning level, make early treatment adaptationsmandatory.

© 2008 Elsevier B.V. All rights reserved.

Keywords:SchizophreniaSubjective wellbeingQuality of lifeCluster analysisRecovery

or Psychosocial Medi-, University Medicalmburg, Germany. Tel.:

Lambert).

All rights reserved.

1. Introduction

During past decades, researchers focused on how recoveryin schizophrenia should be defined and summarized that abroader definition than just reduction in psychotic symptomsis warranted (Andreasen et al., 2005; Bellack, 2006; Kaneet al., 2003; Liberman and Kopelowicz, 2005; Liberman et al.,2002). Despite the lack of a consensually validated definition

166 M. Lambert et al. / Schizophrenia Research 107 (2009) 165–172

of recovery, a state of symptomatic remission and adequatefunctioning sustained for a minimum duration of 2 years arecommonly viewed as dimensions of recovery (Bellack, 2006;Drake et al., 2006; Harrison et al., 2001; Liberman andKopelowicz, 2005; Liberman et al., 2002; Robinson et al., 2004).Problematically, quality of life or subjective wellbeing receivedmuch less attention despite being an established distinctoutcome measure in schizophrenia (Karow et al., 2005; Lambertand Naber, 2004; Lambert et al., 2003, 2006).

Notably, quality of life has a central position in the recoverydefinition from the patients' perspective (Bellack, 2006; Resnicket al., 2005), which has been promoted in the US recently by thePresident's New Freedom Commission on Mental Health (Pre-sident's New Freedom Commission on Mental Health, 2003).Both quality of life and subjective wellbeing may be regarded asan overall subjective measure of illness experience, (antipsycho-tic) treatment response, and life satisfaction (Lambert and Naber,2004; Lambert et al., 2007, 2003, 2006), as well as a valuablepredictor of medication adherence and outcome (Karow et al.,2007; Lambert et al., 2007). Therefore, long-term data onsubjective wellbeing or quality of life in schizophrenia arerelevant sources in the process of extending the objective andsubjective definitions of recovery and of developing appropriateinterventions.

The aims of this post-hoc analysis of the European Schizo-phrenia Outpatient Health Outcomes study (SOHO) (Haro andSalvador-Carulla, 2006; Karow et al., 2007; Lambert et al., 2006)were to study the existence and types of longitudinal patterns ofsubjective wellbeing in schizophrenia using cluster analyses andits relation to recovery criteria in a large cohort of 2842 patientswith schizophrenia followed over 3 years. Further aims were toassess predictors for cluster affiliation, and to evaluate thesensitivity and specificity of baseline SWN-K total score cut-offpoints for cluster affiliation over the 3-year follow-up period.

2. Methods

2.1. Context and sample

Data was collected in a prospective, non-randomized,observational study of 10,972 patients with schizophrenia,recruited in 10 European countries between January andDecember 2001 by 1096 participating physicians. Patients hadto meet the following inclusion criteria: DSM-IV or ICD-10diagnosis of schizophrenia (APA, 1994), at least 18 years of age,IQN70, and initiationofnewor switchof antipsychotic treatment.

Analyses presented here were restricted to patients recruitedfrom 350 outpatient centers in Germany with sufficient dataregarding the primary endpoint (n=2842 of originally n=2960)because (i) subjectivewellbeingwas onlyassessed in theGermanstudypopulation, because (ii) themental health system includingstandards of care is homogeneous for all participating patients,and (iii) because the functional variables such as employment arestrongly influenced by the socio-economic status of a givencountry (Marwaha and Johnson, 2004). The local ethics researchcommittees approved the study.

2.2. Assessments and measures

Assessmentswere carriedout at baseline andat 3, 6,12,18, 24,30, and 36-month follow-ups. Subjectivewellbeingwas assessed

with the SubjectiveWellbeingunderNeuroleptic Treatment scale(SWN-K) (Naber et al., 2001). The SWN-K, a self-rating Likert-scale with six response categories (absent to very much), covers20 statements (10 positive and 10 negative) on emotionalregulation, self-control, mental functioning, social integration,and physical functioningwith aminimum total score of 20 and amaximum total score of 120 points; higher scores indicate betterwellbeing. The SWN-K scale is associated with various Quality oflife scales (e.g., EQ-VAS from the EuroQoL-5D (Prieto et al., 2003),Short-form36 [SF-36] (Karowet al., 2005), orHeinrichsQualityofLife scale [QLS; r=0.71] (Wehmeier et al., 2007a,b)). The EQ-VASfrom the EuroQoL-5D (Prieto et al., 2003)was used asmeasure ofquality of life.

The ‘predictor for cluster affiliation’ variables assessedwere (i) predictors at baseline, (ii) predictors in the early courseof treatment defined as fulfilled cross-sectional recoverycriteria met at 3-month, and (iii) predictors during the courseof treatment defined as variables assessed continuouslythroughout the 36 months study period.

The predictors at baseline were (i) age, gender, andduration of illness; (ii) symptomatic status assessed withthe expanded version of the Clinical Global Impressions-Severity of Illness scale (CGI-Schizophrenia; CGI-SCH (Guy,1976; Haro et al., 2003)) including an overall severity scoreand 4 sub-scores for the severity of positive, negative,cognitive, and depressive symptoms. The CGI-SCH subscaleswere reported to correlate satisfyingly with their respectivePANSS-subscales (Pearson-correlations: PANSS-posi-tive=0.86; PANSS-negative=0.80; PANSS-cognitive/disorga-nized=0.78; PANSS-depressive=0.61; Haro et al., 2006); (iii)functional variables including the “occupational/vocational”status, “independent living”, and “social relationships”assessed bymeans of 1-item questions with yes/no categories(definitions see below); (iv) variables related to antipsychotictreatment (all yes/no categories) were no prior antipsychotictreatment for schizophrenia at baseline, presence of extra-pyramidal motor symptoms (EPMS) and/or tardive dyskinesia(TD) and type of initial antipsychotic medication for schizo-phrenia separated into second- generation (atypical) anti-psychotics ( i.e. olanzapine, risperidone, amisulpride,quetiapine, or clozapine) and conventional antipsychotics(i.e. oral or depot antipsychotics).

The predictors in the early course of treatment were cross-sectional symptomatic and functional recovery criteria at 3-month without the 24 months duration requirement (seebelow).

The predictors during the course of treatment were co-morbid substance use disorder (SUD) assessed according toDSM-IV criteria (APA, 1994) and categorized into (i) no SUD(no SUD at baseline and throughout follow-ups), (ii) remittedSUD (SUD at baseline and remission at all follow-up visits),(iii) persistent SUD (SUD reported at any post baseline visit).In accordance to the criterion proposed by Kane et al. (2003)non-adherence with antipsychotic medication was defined asmissing ≥50% of medication over at least 4 weeks.

2.3. Definition of recovery criteria

Recovery as defined in this observational study requiredsimultaneous fulfillment of the following criteria over aperiod of at least 24 months.

167M. Lambert et al. / Schizophrenia Research 107 (2009) 165–172

➢ Symptomatic recovery was defined as achieving a CGI-SCHscore of no worse than “mild” (score ≤3) in the assess-ments of overall severity, positive, negative, and cognitivesub-scores and no hospitalization in the respective timeperiod.

➢ Functional recovery was defined as fulfillment of thefollowing three criteria:Occupational/vocational status, i.e. paid or unpaid full- orpart-time employment, being an active student inuniversity or head of household with employed partner,Independent living, i.e. living alone, with partner, or withpeers, andSocial relationships, i.e. having ≥2 social contacts duringthe last 4 weeks or having a spouse.

Table 1Baseline variables, early course of treatment, course of treatment andoutcome variables (N=2842)

Variable Value for totalsample a

Pre-morbid and baseline variablesAge, mean (SD), years 42.1 (13.8)b30 years 537 (19.4)30–b40 years 845 (30.5)40–b50 years 683 (24.6)N=50 years 709 (25.6)

Sex, N (%), male 1376 (49.5)Duration of illness, median (quartiles), years 7.7 (1.7–16.8)CGI—Schizophrenia severity score, mean (SD)Overall score 4.4 (1.0)Positive subscore 3.7 (1.4)Negative subscore 4.0 (1.3)Cognitive subscore 4.0 (1.2)Depressive subscore 3.7 (1.3)

Functioning N (%)Employed 1215 (43.0)Independent living 1836 (64.7)Social relationships 1566 (56.4)

SWN-K total score, mean (SD) 61.9 (17.3)Extrapyramidal motor symptoms and/or tardivedyskinesia, N (%)

1039 (37.5)

No prior antipsychotic treatment for schizophrenia, N (%) 381 (13.4)First-line antipsychotic, N (%)Atypical antipsychotic 2400 (86.5)Conventional antipsychotic 374 (13.5)

Early course of treatment variables (at month 3)Cross-sectional symptomatic recovery, N (%) 851 (30.8)Cross-sectional functional recovery, N (%) 1753 (61.7)

Course of treatment variablesSubstance Use Disorder (SUD), N (%)No SUD 2133 (80.5)Remitted SUD 302 (11.4)Persistent SUD 216 (8.2)

Adherence, N (%) 1771 (62.3)

Fulfilled recovery criteria at 3-year endpointSymptomatic recovery N (%) 1002 (36.0)Functional recovery N (%) 591 (20.8)Adequate subjective wellbeing N (%) 973 (34.2)Combined recovery N (%) 226 (8.1)

Abbreviations: CGI-SCH = Clinical Global Impressions—Severity of Illnessscale, expanded version; SWN-K = Subjective Wellbeing under NeurolepticTreatment scale, short-version.

a Percentage were based on non-missing observations.

➢ The criterion for adequate subjective wellbeing was metif an SWN-K total score of ≥80 points was achieved(Lambert et al., 2007, 2006). This score corresponds to anaverage rating of “marked” positive subjective wellbeingin the positive SWN-K items and ratings of only “mild”impairment of subjective wellbeing in the negative SWN-K items.

2.4. Data analysis

All analyses were carried out using SAS following theintention to treat (ITT) principle. All 2842 SOHO patients fromGermany with a valid SWN-K total score at baseline wereincluded in the analyses and last observation carried forward(LOCF) was used to impute missing values in visits followingdrop out or due to missing observations.

A cluster analysis was applied based onWard's procedure.With this method, the within-cluster sum of squares is mi-nimized over all partitions obtainable by merging twoclusters from the previous generation. Sensitivity analyseswith other methods of cluster analyses and using differentnumbers of clusters were performed for the 3-year data andconfirmed the clusters. Further sensitivity analyses wereperformed using only observed values instead of LOCF valuesand with only 1-year, 2-year and 3-year follow-up data.

Patient variables were described using means (M) andstandard deviations (SD), or by absolute and relativefrequencies based on non-missing observations, which maylead to varying denominators for computing proportions.Overall univariate comparisons between the resulting clustersregarding baseline variables were performed using ANOVA orχ2-tests.

Baseline characteristics for clusters were also comparedusing the logistic regression technique. Full models includingall covariates as well as forward and backward selectionmodels, the latter of which is reported, using p=0.1 as athreshold for single variables, were performed. Odds ratioswith the respective 95%-confidence intervals are given and p-values were based on Wald's test.

To find optimal cut-off points regarding the SWN-K totalscore at baseline between clusters, Cohen's kappas with 95%confidence intervals were calculated for varying cut-offs ofthe mean SWN-K total score values. The maximum kappa andrespective conditional probabilities are presented. The con-ditional probabilities are calculated in the same way as thetest diagnostic measures: sensitivity and specificity as well aspositive and negative predictive values.

3. Results

3.1. Subject characteristics

The demographic and clinical characteristics of the 2842subjects with schizophrenia are outlined in Table 1. The meanagewas 42.1 years, 49.5%weremale, and themedian durationof illness was 7.7 years. The mean SWN-K total score atbaseline was 61.9 points. 13.4% of subjects had received noprior antipsychotic treatment for schizophrenia, 86.5% wereinitially treated with atypical antipsychotics, all others withoral or depot conventional antipsychotics (for further treat-ment details see Lambert et al., 2006).

168 M. Lambert et al. / Schizophrenia Research 107 (2009) 165–172

3.2. Cluster analysis

The cluster analyses based on 1-year LOCF data revealed fourclusters of patients with different subjective wellbeing patterns:(i) patients with stable low (n=932, 32.8%), patients with stablemoderate (n=890, 31.3%), patients with stable high (n=456,16.1%), and patients with early improvement followed by stablehigh subjective wellbeing (henceforth early improvement,n=564, 19.9%). Fig. 1 displays the SWN-K LOCF data over3 years for these four clusters. The clusters remained stablewhen only observed cases (OC), or 1-year follow-up data (basedonn=2378), 2-year- (based onn=2067), or 3-year follow-updata(based on n=1775) were used instead of LOCF data, indicatingthat the observed follow-up stability of subjective wellbeing wasnot due to bias introduced by LOCF. Because values and cluster-

Fig. 1. Scatter plots of subjectivewellbeing (SWN-K) total scores at consecutive visits.the vertical axis; the clusters aremarked in colors: stable lowSWN= red, stablemoderSubjective Wellbeing under Neuroleptic Treatment scale, short-version. N=2842 (LO

membership were consistent across imputation methods andnumber of visits used, the initially chosen clusters based on 1-year LOCF data were used in all subsequent analyses.

3.3. Subjective wellbeing cluster and 3-year patterns ofsymptoms, quality of life, and social functioning

Fig. 2 shows the conformity of the four subjective well-being clusters with regard to course of subjective wellbeing,symptoms, quality of life, and functioning level. Obviously, thepatterns of symptoms, quality of life, and subjectivewellbeingover time are remarkably similar. However, the course offunctioning level appears to deviate, in that patients in theearly improvement cluster with low initial subjective well-being already exhibit a very good functioning level at baseline.

Footnotes: The earlier visits correspond to the horizontal and the later visits toate SWN=black, stable high SWN= green, and improved SWN=blue. SWN-K =CF).

Fig. 2. Course of subjectivewellbeing, severity of illness, combined functioning, and Quality of life in the 4 SWN cluster over the 36months study period. Footnotes:Abbreviations: CGI-SCH = Clinical Global Impressions—Severity of Illness scale, expanded version (for schizophrenia); SWN-K = Subjective Wellbeing underNeuroleptic Treatment scale, short-version. EQ5D-VAS = 1 item from the EuroQoL. aCombined functioning = % of patients with cross-sectional functional recoveryat the respective time point.

169M. Lambert et al. / Schizophrenia Research 107 (2009) 165–172

3.4. Affiliation to the high subjective wellbeing cluster andsubsequent recovery rates at 3-year follow-up

Based on the course of SWN-K total score in the fourclusters, displayed in Fig. 2, it was hypothesized that thestable high subjective wellbeing cluster can be satisfactorilydistinguished from the other three clusters by means of theSWN-K total score at baseline only. To find an optimal cut-offpoint between this and all other clusters, Cohen's kappaswere calculated for all baseline SWN-K total scores between60 and 90. Cohen's kappa reached a maximum at a meanbaseline SWN-K total score of ≥80 points (kappa=0.7730; 95%CI 0.7413 to 0.8047). With this cut-off point, 386 (77.8%) pa-tients of the 496 patients with SWN-K total score ≥80 pointswere correctly predicted to be in the stable high subjectivewellbeing cluster (positive predictive value) and 2276 (97.0%)patients of the 2346 patients with SWN-K total score b80points were correctly predicted to be not in this cluster(negative predictive value). The sensitivity of this cut-offwas 84.7% and the specificity 95.4%. Accordingly, the stablehigh cluster affiliation was well distinguished from all otherclusters by the SWN-K total score at baseline. Thus, in thesearch for predictors of cluster affiliation, the stable high

cluster was not considered in the subsequent predictionanalyses.

The rate of symptomatic recovery at endpoint of the 3-yearobservationperiodwas 50.5% in the stable high cluster and 33.2%in the other clusters. The rate of functional recoverywas 25.1% vs.21.0%, the rate of adequate subjective wellbeing was 79.2% vs.25.7%, and the rate of combined recovery was 14.7% vs. 6.9%.

3.5. Prediction of affiliation to the stable low, stable moderate,and early improvement clusters

The three models assessing predictors of cluster affiliation,separated into stable low, stable moderate, and earlyimprovement subjective wellbeing clusters, are presented inTable 2. Stable low cluster affiliation (vs. the two otherclusters) was mainly predicted by a low SWN-K total score atbaseline, initial treatment with conventional antipsychotics,and further bymedication non-adherence as well as not beingin cross-sectional early symptomatic and functional recoveryat 3 months. Stable moderate cluster affiliation was predic-ted by being employed, a higher SWN-K total score at base-line, and further by non-adherence with medication. Earlyimprovement cluster affiliationwasmainly predicted by a good

Table 2Predictors of subjective wellbeing cluster affiliation for stable low, stable moderate, and improved cluster vs. the respective other two clusters— logistic regressionwith backward selection (N=2842)

Predictora Low subjectivewellbeing cluster

Moderate subjectivewellbeing cluster

Improved subjectivewellbeing cluster

OR 95% CI p-value OR 95% CI p-value OR 95% CI p-value

Baseline predictorsAge (years)

b30 yearsb

30–b4040–b50N=50

Sex (male vs. female)Duration of illness (years) 1.02 1.01 to 1.03 0.001 0.99 0.98 to 1.00 0.055Increase of 1 in CGI—Schizophrenia severity

Overall score 0.88 0.77 to 1.02 0.091Positive subscore 0.90 0.82 to 0.99 0.028 1.15 1.05 to 1.26 0.003Negative subscoreCognitive subscore 0.89 0.79 to 1.01 0.069 1.18 1.05 to 1.33 0.006Depressive subscore

FunctioningIndependent living 1.61 1.25 to 2.07 b0.001Employed 1.30 1.04 to 1.63 0.021Social relationships 1.43 1.13 to 1.81 0.003

SWN-K total score (in 20 points)c 0.09 0.07 to 0.12 b0.001 5.66 4.54 to 7.07 b0.001 1.31 1.07 to 1.60 0.009Extrapyramidal motor symptoms and/or tardive dyskinesiaNo prior antipsychotic treatment for schizophrenia 1.45 1.04 to 2.02 0.031First-line AP (atypical vs conventional) 0.43 0.30 to 0.61 b0.001 2.50 1.66 to 3.77 b0.001

Early course of treatment predictorsCross-sectional symptomatic recovery 0.15 0.11 to 0.21 b0.001 0.08 0.62 to 1.02 0.066 5.03 3.92 to 6.45 b0.001Cross-sectional functional recovery 0.50 0.39 to 0.64 b0.001

Course of treatment predictorsSubstance use disorder (SUD)

No SUDb … … … … … … … … …

Remitted SUDPersistent SUD

Adherence 0.71 0.55 to 0.91 0.007 1.29 1.04 to 1.61 0.024

Abbreviations: CGI-SCH = Clinical Global Impressions—Severity of Illness scale, expanded version; SWN-K = Subjective Wellbeing under Neuroleptic Treatmentscale, short-version, AP = antipsychotics.aSignificant predictors are marked bold.bReference category.cSWN-K total score was rescaled in units of 20 points.

170 M. Lambert et al. / Schizophrenia Research 107 (2009) 165–172

functional level at baseline (meeting criteria for independentliving and social relationships), initial treatment with atypicalantipsychotics and further by cross-sectional symptomaticrecovery at three months.

3.6. Affiliation to the stable low subjective wellbeing cluster andsubsequent recovery rates at 3-year follow-up

As reported above, the SWN-K total score at baselinewas oneof the main predictors for cluster affiliation in the logisticregression analyses. However, the mean SWN-K total scores ofthe stable low, stable moderate, and early improvement clusterswere all in the range of 50 to 65 points at baseline. Therefore, theoptimal mean SWN-K total score cut-off point at baselinebetween the stable low cluster and the other two clusters wasassessed. Cohen's kappawas calculated for the SWN-K total scorevalues between50 and65. Cohen's kappa reached amaximumatamean SWN total score of ≤60 points at baseline (kappa=0.5016;95% CI 0.4704 to 0.5327). With this cut-off point, the positivepredictive value was 59.4% and the negative predictive value85.2%. The sensitivity was 82.8% and the specificity 63.8%.

The rate of symptomatic recovery at endpoint of the 3-year observation period was 15.2% in the stable low clusterand 44.7% in the other two clusters (stablemoderate and earlyimprovement). The rate of functional recovery was 13.5% vs.24.2%, the rate of adequate subjective wellbeing was 2.6% vs.40.4%, and the rate of combined recovery was 0.6% vs. 10.9%.

4. Discussion

To the authors' knowledge, the present study is the first tostudy longitudinal patterns of subjective wellbeing in schizo-phrenia and its relation to recovery in a 3-year follow-upstudy. It further assessed predictors of the three clinicallymost important clusters and the sensitivity and specificity aswell as the positive and negative predictive values of baselineSWN-K total score cut-off points for cluster affiliation.

4.1. Key findings

Four different clusters of subjective wellbeing wereidentified in the 3-year follow-up study: a stable low, a stable

171M. Lambert et al. / Schizophrenia Research 107 (2009) 165–172

moderate, a stable high, and, interestingly, a cluster withinitial low subjective wellbeing, early improvement in thefirst 3 to 6 months and subsequent long-term stable highsubjective wellbeing (henceforth early improvement cluster).Importantly, these clusters were consistent across differentmethodological approaches (e.g., LOCF, OC, or varying num-bers of visits included) and highly concordant with 3-yearpatterns of symptoms, quality of life, and, in part, psychosocialfunctioning. The concordance between low subjective well-being, high severity of symptoms, low functional level, andlow quality of life are known from various cross-sectionalstudies (Karow et al., 2005; Lambert et al., 2003). However, todate there is only one study, which addressed this issuelongitudinally (Lambert et al., 2007). In the latter study, lack ofsubjective wellbeing improvement in the first 4 weeks of(antipsychotic) treatment, defined as more than 20% increasein the SWN-K total score from baseline, classified a subgroupof approximately 30% of patients who were found to have noor only minimal response in positive and negative symptoms,functional level, and overall severity of illness throughout a3 months study period. Interestingly, in the present study,again 30% of patients of the stable low SWN cluster achievedno or only minimal improvement in symptoms, functioning,and quality of life, yet in a much longer observational period.Across multiple long-term outcome studies, there is a sub-group of approximately 30% to 40% of patients who do not oronly minimally respond to treatment with continuouspsychotic symptoms as well as poor functional level (Flycktet al., 2006; Jobe and Harrow, 2005).

We further identified predictors for cluster affiliation. Ofclinical importance are the predictors for the stable lowclusterand the early improvement cluster, particularly since baselinesubjective wellbeing is severely impaired in both groups.Stable low cluster affiliation was mainly predicted by a lowSWN-K total score at baseline (OR 0.091) and further by a longduration of illness and not meeting criteria for early cross-sectional symptomatic and functional recovery at 3 months.The early improvement SWN cluster was mainly predictedby early symptomatic recovery at 3 months (OR 5.028) andfurther by a good functioning level at baseline. Duration ofillness, level of functioning, and early response are well-known prognostic variables for overall outcome in schizo-phrenia (Flyckt et al., 2006; Jobe and Harrow, 2005; Lambertet al., 2006; Leucht et al., 2007; Siegel et al., 2006). However,the finding that the level of functioning was highly predictivefor early improvement cluster affiliation indicates that a goodfunctional level at baseline might separate patients withand without a good chance of recovery, even in case of highbaseline illness severity, level of symptoms, and impairedsubjective wellbeing. The result that initial treatment withconventional vs. atypical antipsychotics was predictive forcluster affiliation and thereby for outcome should be inter-preted with caution, as antipsychotics were not randomlyassigned and only the initial medication choice was asses-sed. It cannot be excluded that the choice of conventionalvs. atypical antipsychotic was influenced by factors, whichwere not assessed such as previous treatment response,thereby favoring the group receiving second-generationantipsychotics.

The study further explored the sensitivity and specificityas well as the positive and negative predictive values of

subjective wellbeing levels at baseline for cluster affiliation.Nearly 85% of patients with a SWN-K total score of ≥80 pointsat baseline belong to the stable high subjective wellbeingcluster, which is related to a good chance of symptomatic(50.5%), functional (25.1%), subjective wellbeing (79.2%), andcomplete recovery (14.7%). With the previously reportedSWN-K total cut-off score of ≤60 points (Lambert et al., 2007),corresponding to a severely impaired subjective wellbeing atbaseline, nearly 83% of the patients were correctly categorizedinto the stable low cluster. Contrary to the stable high cluster,the affiliation to the stable low cluster was associated with avery low chance of symptomatic (15.2%), functional (13.5%),subjective wellbeing (2.6%), and complete recovery (0.6%).The good predictive values of these SWN-K cut-off pointsindicate their usefulness for early outcome prediction andthereby early treatment planning and adaptation.

4.2. Limitations and strengths

As this was a large observational study in daily clinicalpractice, several methodological limitations such as thelack of assessment of inter-rater reliability or the availabilityof more differentiated measures, such as PANSS or neuro-psychological assessments for the detailed assessment ofcognitive deficits, were unavoidable. Furthermore, general-izability of the results to epidemiological or first-episodesamples is limited by the selection of an outpatient popula-tion with initiation of new or switch of antipsychotictreatment, resulting in a sample with a long mean durationand only moderate severity of illness, a comparatively higherprevalence of males and patients with employment atbaseline. Strengths of the present study are the large numberof patients, the naturalistic design without selection biasrelated to randomized controlled designs, the long follow-upperiod, and the relatively low dropout rate compared to manyprevious trials, especially long-term studies (Lambert et al.,2006).

4.3. Conclusions

In the last decade, cross-sectional studies on subjectivewellbeing in schizophrenia have shown and replicated its cli-nical importance as a subjective measure of illness experience,(antipsychotic) treatment response, and overall life satisfaction.Our study indicates that there are stable long-term patterns ofsubjective wellbeing, which are highly concordant with thecourse of symptoms, quality of life, and, in part, of psychosocialfunctioning. Clinically, there are two important patterns: The‘improved cluster’ is of special interest because it shows thateven in a group of “mostly chronic” patients a significantproportion of patients (about 20%) can reach a distinct responsein a short time period and subsequent recovery of subjectivewellbeing. The discriminating factor for this “responder” groupis a good functioning level at baseline, despite a comparablyhigh baseline severity of illness and impaired subjectivewellbeing. Contrary, patients with a low subjective wellbeingand poor functioning level at baselinemost likely belong to thestable low subjectivewellbeing cluster, which itself is related toa very low chance of symptomatic, functional, subjectivewellbeing, and composite recovery. With 33%, the latter clusteris the largest subgroup in the present study possibly related to

172 M. Lambert et al. / Schizophrenia Research 107 (2009) 165–172

the selection of patients with long illness duration and thuschronicity. Clinically, the SWN-K scale, which takes approxi-mately 5 min to complete, appears to be a good tool for theprediction of response and subsequent recovery, especially ifpatients' pre-treatment functioning level is additionallyconsidered.

Role of funding sourceThis observational study was funded by Eli Lilly and Company. The first

authors (ML and BGS) developed the research question and data analysisstrategy, interpreted the data, and wrote the manuscript. The statistician ofLilly Deutschland GmbH (AS) had access to the raw data and supplied ML andBGS with the original data analysis outputs but not with the raw data. Noneof the academic authors (ML, BGS, AK, CGH, and DN) were paid for discussingor writing the manuscript. Authors AS, TW, PMW, HPH, and RWD areemployees of Lilly Deutschland GmbH and aremembers of the German SOHOstudy group and contributed to the final version of the manuscript.

ContributorsAuthors ML and BGS developed the research question and data analysis

strategy, interpreted the data, and wrote the complete manuscript. Author ASundertook the statistical analyses. Authors AK, CGH, DN, TW, PMW, RWD,and HPH are members of the German SOHO study group and havecontributed to and approved the final manuscript.

Conflict of interestPD Dr. Martin Lambert, Dr. Benno Schimmelmann, Prof. Dieter Naber, Dr.

Anne Karow, and Dr. Christian Huber have received educational grants andare on the speakers' board of Eli Lilly Company. Dr. Alexander Schacht, Dr.Thomas Wagner, Dr. Peter Wehmeier, Prof. Dr. Ralf Dittmann, and Dr. Hans-Peter Hundemer are employees of the Medical Neuroscience Department,Lilly Deutschland GmbH.

AcknowledgementsThe authors wish to thank all participating physicians and patients.

References

Andreasen, N.C., Carpenter Jr., W.T., Kane, J.M., Lasser, R.A., Marder, S.R.,Weinberger, D.R., 2005. Remission in schizophrenia: proposed criteriaand rationale for consensus. Am. J. Psychiatry 162, 441–449.

APA, 1994. Diagnostic Criteria from DSM-IV. American Psychiatric Publishing,Washington, DC.

Bellack, A.S., 2006. Scientific and consumer models of recovery in schizophrenia:concordance, contrasts, and implications. Schizophr. Bull. 32, 432–442.

Drake, R.E., McHugo, G.J., Xie, H., Fox, M., Packard, J., Helmstetter, B., 2006.Ten-year recovery outcomes for clients with co-occurring schizophreniaand substance use disorders. Schizophr. Bull. 32, 464–473.

Flyckt, L., Mattsson, M., Edman, G., Carlsson, R., Cullberg, J., 2006. Predicting5-year outcome in first-episode psychosis: construction of a prognosticrating scale. J. Clin. Psychiatry 67, 916–924.

Guy, W., 1976. ECDEU Assessment Manual for Psychopharmacology, revised.US Dept Health, Education, and Welfare Publication (ADM) 76-338.National Institute of Mental Health, Rockville Md, pp. 218–222.

Haro, J.M., Salvador-Carulla, L., 2006. The SOHO (Schizophrenia OutpatientHealth Outcome) study: implications for the treatment of schizophrenia.CNS Drugs 20, 293–301.

Haro, J.M., Kamath, S.A., Ochoa, S., Novick, D., Rele, K., Fargas, A., Rodriguez, M.J.,Rele, R., Orta, J., Kharbeng, A., Araya, S., Gervin, M., Alonso, J., Mavreas, V.,Lavrentzou, E., Liontos, N., Gregor, K., Jones, P.B., SOHOStudyGroup, 2003. TheClinical Global Impression-Schizophrenia scale: a simple instrument tomeasure the diversity of symptoms present in schizophrenia. Acta. Psychiatr.Scand. 107 (suppl. 416), 16–23.

Harrison, G., Hopper, K., Craig, T., Laska, E., Siegel, C.,Wanderling, J., Dube, K.C.,Ganev, K., Giel, R., an der Heiden, W., Holmberg, S.K., Janca, A., Lee, P.W.,Leon, C.A., Malhotra, S., Marsella, A.J., Nakane, Y., Sartorius, N., Shen, Y.,Skoda, C., Thara, R., Tsirkin, S.J., Varma, V.K., Walsh, D., Wiersma, D., 2001.Recovery from psychotic illness: a 15- and 25-year international follow-up study. Br. J. Psychiatry 178, 506–517.

Jobe, T.H., Harrow, M., 2005. Long-term outcome of patients with schizo-phrenia: a review. Can. J. Psychiatry 50, 892–900.

Kane, J.M., Leucht, S., Carpenter, D., Docherty, J.P., 2003. Expert consensusguideline series. Optimizing pharmacologic treatment of psychoticdisorders. Introduction: methods, commentary, and summary. J. Clin.Psychiatry 64 (Suppl 12), 5–19.

Karow, A., Czekalla, J., Dittmann, R.W., Schacht, A., Wagner, T., Lambert, M.,Schimmelmann, B.G., Naber, D., 2007. Association of subjective well-being, symptoms, and side effects with compliance after 12 months oftreatment in schizophrenia. J. Clin. Psychiatry 68, 75–80.

Karow, A., Moritz, S., Lambert, M., Schoder, S., Krausz, M., 2005. PANSS syndromesand quality of life in schizophrenia. Psychopathology 38, 320–326.

Lambert, M., Naber, D., 2004. Current issues in schizophrenia: overview ofpatient acceptability, functioning capacity and quality of life. CNS Drugs18 (Suppl 2), 5–17 discussion 41-3.

Lambert, M., Naber, D., Eich, F.X., Schacht, M., Linden, M., Schimmelmann, B.G.,2007. Remission of severely impaired subjective wellbeing in 727 patientswith schizophrenia treated with amisulpride. Acta Psychiatr. Scand. 115,106–113.

Lambert, M., Schimmelmann, B.G., Karow, A., Naber, D., 2003. Subjectivewellbeing and initial dysphoric reaction under antipsychotic drugs —

concepts, measurement and clinical relevance. Pharmacopsychiatry 36(Suppl 3), 181–190.

Lambert,M., Schimmelmann, B.G., Naber, D., Schacht, A., Karow, A.,Wagner, T.,Czekalla, J., 2006. Prediction of remission as a combination of sympto-matic and functional remission and adequate subjective well-being in2960 patients with schizophrenia. J. Clin. Psychiatry 67, 1690–1697.

Leucht, S., Busch, R., Kissling, W., Kane, J.M., 2007. Early prediction ofantipsychotic nonresponse among patients with schizophrenia. J. Clin.Psychiatry 68, 352–360.

Liberman, R.P., Kopelowicz, A., 2005. Recovery from schizophrenia: a conceptin search of research. Psychiatr. Serv. 56, 735–742.

Liberman, R.P., Kopelowicz, A., Ventura, J., Gutkind, D., 2002. Operational criteriaand factors related to recovery from schizophrenia. Int. Rev. Psychiatry 14,256–272.

Marwaha, S., Johnson, S., 2004. Schizophrenia and employment — a review.Soc. Psychiatry Psychiatr. Epidemiol. 39, 337–349.

Naber, D., Moritz, S., Lambert, M., Pajonk, F.G., Holzbach, R., Mass, R.,Andresen, B., 2001. Improvement of schizophrenic patients' subjectivewell-being under atypical antipsychotic drugs. Schizophr. Res. 50, 79–88.

President's New Freedom Commission on Mental Health. Achieving thePromise: Transforming Mental Health Care in America. 2003; http://www.mentalhealthcommission.gov/index.html.

Prieto, L., Novick, D., Sacristan, J.A., Edgell, E.T., Alonso, J., 2003. A Rasch modelanalysis to test the cross-cultural validity of the EuroQoL-5D in theSchizophrenia Outpatient Health Outcomes Study. Acta Psychiatr. Scand.Suppl. 24–29.

Resnick, S.G., Fontana, A., Lehman, A.F., Rosenheck, R.A., 2005. An empiricalconceptualization of the recovery orientation. Schizophr. Res. 75, 119–128.

Robinson, D.G.,Woerner,M.G., McMeniman, M., Mendelowitz, A., Bilder, R.M.,2004. Symptomatic and functional recovery from a first episode ofschizophrenia or schizoaffective disorder. Am. J. Psychiatry 161, 473–479.

Siegel, S.J., Irani, F., Brensinger, C.M., Kohler, C.G., Bilker, W.B., Ragland, J.D.,Kanes, S.J., Gur, R.C., Gur, R.E., 2006. Prognostic variables at intake and long-term level of function in schizophrenia. Am. J. Psychiatry 163, 433–441.

Wehmeier, P.M., Kluge, M., Schacht, A., Helsberg, K., Schreiber, W., 2007a.Correlation of physician and patient rated quality of life during antipsychotictreatment in outpatients with schizophrenia. Schizophr. Res. 91, 178–186.

Wehmeier, P.M., Kluge, M., Schneider, E., Schacht, A., Wagner, T., Schreiber,W., 2007b. Quality of life and subjective well-being during treatmentwith antipsychotics in out-patients with schizophrenia. Prog. Neuro-Psychopharmacol. Biol. Psychiatry 31, 703–712.