Cognitive Enhancement Therapy for Schizophrenia: Effects of a 2-Year Randomized Trial on Cognition and Behavior

Background  Deficits in social cognition and neurocognition are believed to underlie schizophrenia disability. Attempts at rehabilitation have had circumscribed effects on cognition, without concurrent improvement in broad aspects of behavior and adjustment.

Objective  To determine the differential effects of cognitive enhancement therapy (a recovery-phase intervention) on cognition and behavior compared with state-of-the-art enriched supportive therapy.

Design  A 2-year, randomized controlled trial with neuropsychological and behavioral assessments completed at baseline and at 12 and 24 months.

Setting  An outpatient research clinic housed in a medical center's comprehensive care service for patients with severe mental illness.

Patients  A total of 121 symptomatically stable, non–substance-abusing but cognitively disabled and chronically ill patients with schizophrenia or schizoaffective disorder.

Interventions  Cognitive enhancement therapy is a multidimensional, developmental approach that integrates computer-assisted training in neurocognition with social cognitive group exercises. Enriched supportive therapy fosters illness management through applied coping strategies and education.

Main Outcome Measures  Six highly reliable summary measures—Processing Speed, Neurocognition, Cognitive Style, Social Cognition, Social Adjustment and Symptoms—were tested using analysis of covariance and linear trend analysis.

Results  At 12 months, robust cognitive enhancement therapy effects were observed on the Neurocognition and Processing Speed composites (P<.003), with marginal effects observed on the behavioral composites. By 24 months, differential cognitive enhancement therapy effects were again observed for the 2 neuropsychological composites and for Cognitive Style (P=.001), Social Cognition (P=.001), and Social Adjustment (P=.01). As expected, no differences were observed on the residual Symptoms composite. Effects were unrelated to the type of antipsychotic medication received. Enriched supportive therapy also demonstrated statistically significant within-group effect sizes, suggesting that supportive psychotherapy can also have positive, although more modest, effects on cognitive deficits.

Conclusion  Many cognitive deficits and related behaviors of patients with stable schizophrenia are improved when sufficient exposure to relevant rehabilitation is provided.

Schizophrenia is a leading cause of disability throughout the world.¹ Numerous, but often cross-sectional, studies havenow shown correlations between neurocognitive deficits and behavior.² Concurrent impairment in social cognition has alsobeen proposed as an important rate-limiting factor to recovery.^3-6 Theseobservations have stimulated various neurocognitive rehabilitation approaches.^7-18 Samplesizes have been small (30-90 individuals), and many have included severelyimpaired, hospitalized patients. Treatment exposure has been limited (median,approximately 4 months), and no trial has exceeded 9 months. Successful attemptsto improve neuropsychological (NP) test performance have been reported,^19-21 as has some evidenceof generalization to untrained NP tests.^12,16,22,23 However,concurrent effects on broad aspects of functioning and behavior have beenviewed as small and lacking generalization.^15,18

Cognitive enhancement therapy (CET) was developed and piloted in theearly to middle 1990s as an integrated approach to the enhancement of neurocognitiveand social cognitive abilities. It attempts to capitalize on a presumed neuroplasticityreserve believed to respond to enriched cognitive experiences.²⁴ Cognitiveenhancement therapy is a recovery-phase intervention for symptomatically stableschizophrenic outpatients with reduced relapse risk (who nevertheless remainsocially and cognitively disabled), an increasing population in the modernera of atypical antipsychotic medications.²⁵ A2-year, randomized study was undertaken between January 1995 and February2002, together with a 1-year follow-up. This article describes the resultsof the 2-year controlled trial.

Initially, 132 outpatients who satisfied DSM-III-R or DSM-IV criteria for schizophrenia or schizoaffectivedisorder and who currently satisfied Research DiagnosticCriteria for a Selected Group of Functional Disorders²⁶ wereenrolled. (Another 12 referred patients were not enrolled because of mentalinsufficiency or organicity.) Referrals came from the University of PittsburghMedical Center's Comprehensive Care (outpatient) Service, a local mental healthcenter, and a University of Pittsburgh Medical Center satellite clinic followinga preliminary population screen of potentially eligible patients. Most patients(76%) were in full or partial remission of positive psychotic symptoms atbaseline. For individuals with persistent positive symptoms, stability criteriaindicated no serious effect on daily activities. By design, most patients(88%) were more than 1 year past their last hospitalization (median, 46 months)and thus were at reduced risk of relapse.²⁷ Eightof the 132 patients (4 CET recipients and 4 enriched supportive therapy [EST]recipients) withdrew consent before treatment exposure. After assessment,3 additional patients were judged to be ineligible by reason of mental insufficiency(IQ <80) or organic brain disorder. The final study sample contained 121patients who met the criteria for cognitive disability, which consisted ofthe impairments, functional disabilities, and social handicaps associatedwith 1 of 3 dysfunctional cognitive styles,²⁸ andthe criteria for social cognitive disability (Table 1). The styles are continuous rather than categorical: 91%of patients met the criteria for a single style (33% were impoverished, 41%were disorganized, and 17% were rigid) and 9% for multiple styles. These eligibilitycriteria were assessed during a videotaped, semistructured interview. Patientswere fluent in English, aged 18 to 60 years, treated with a Food and DrugAdministration–approved antipsychotic medication, and free of seriousalcohol or drug abuse in the preceding 6 months, with an IQ of 80 or greater.Final eligibility was determined by team consensus after a review of all diagnostic,historical, and interview materials. Table2 describes the sample characteristics. Before intake, patientsprovided a signed informed consent form, and the study was reviewed annuallyby the University of Pittsburgh institutional review board.

Patients were randomized by the project statistician (R.U.) to receiveCET or EST and were treated under these conditions for 2 years. To more rapidlyform the 11 CET social cognitive groups, 67 patients were randomized to theCET condition and 54 to the EST condition. Ten patients (6 CET recipientsand 4 EST recipients) who experienced an interim relapse (n = 5) or medicalillness (n = 5) were given a brief "timeout," were stabilized, and were readmittedto their assigned treatment group. Fourteen patients (12%) terminated treatmentearly: 8 patients between 4 and 6 months (4 EST recipients and 4 CET recipients)and 6 patients at 12 months (4 EST recipients and 2 CET recipients). Fourof these patients became medication refractory (2 EST recipients and 2 CETrecipients) and 7 withdrew consent (3 EST recipients and 4 CET recipients);1 EST patient became alcohol dependent and 2 EST patients died (cancer andheart disease). All 121 patients met the criteria for minimum treatment exposureand were included in the intent-to-treat analyses at 12 months. Before analyzingdata, it was deemed inappropriate to carry forward to 24 months the ratingsof the 4 CET and 4 EST patients who terminated treatment between 4 and 6 months.

Cognitive enhancement therapy theory and practice are described in detailelsewhere.^5,28 This therapy wasinfluenced by the holistic program for traumatic brain-injured patients developedby Ben-Yishay and colleagues,²⁹ the integratedcognitive approach of Brenner et al,⁸ and acontemporary theory of human cognitive development.³⁰ Cognitiveenhancement therapy attempts to facilitate the attainment of adult socialcognitive milestones, such as perspective taking and social context appraisal,^31,32 by shifting an alleged early developmentalreliance on effortful, serial, and verbatim (concrete) cognitive processing³³ to a more "gistful," spontaneous abstraction of socialthemes through structured but unrehearsed in vivo social interactions. Cognitiveenhancement therapy is a small-group approach that combines approximately75 hours of progressive software training exercises in attention, memory,and problem solving with 1.5 hours per week of social cognitive group exercises(approximately 56 sessions). A group (6 patients) began to meet 4 to 6 monthsafter the initiation of attention training. Software exercises required thatpatients work in pairs, offer mutual support and encouragement, respond toonline Socratic coaching, and use the cueing and fading of prompts until theprinciples underlying test performance were mastered. Three attention trainingcomponents of Ben-Yishay's Orientation Remediation Module³⁴ wereused (the Attention Reaction Conditioner, Zero Accuracy Conditioner and TimeEstimates) that are graduated in difficulty and designed to enhance vigilance,selective attention, the ability to shift between auditory and visual modalities,and rapid decision making. Attention training was followed by 7 memory routinesand ultimately 7 problem-solving exercises from the Bracy PSSCogReHab program³⁵ that represented the Trail Trace, Spatial Memory(sequenced and objects and location), Recall Recognition, Visual/Spatial Memory,Paired Associates and Verbal Memory (categories) exercises. The goal was todevelop memory skills through the enhancement of a categorizing capacity,an abstracting attitude, cognitive flexibility, and decision making. Problem-solvingexercises included Designer Patterns, Numbers Manipulation I and II, Reversals,Logicmaster, Deduction, and Checker Exchange. These engaging exercises targetedanalytic logic, effortful decision making, strategic and foresightful planning,and the intuitive thinking that supports social cognition.

The group curriculum²⁸ contained activitiessuch as categorization exercises; formation of gistful, condensed messages;solving of real-life social dilemmas; abstraction of themes from the editorialpages of USA Today; appraisal of affect and socialcontexts; initiating and maintaining conversations; play writing; and thecenter stage exercises Introduce Yourself/Friend adapted from the curriculumof Ben-Yishay.³⁶ Sessions typically containeda homework review, a psychoeducation topic, an exercise by a patient or pair,feedback from other patients and coaches, and a new homework assignment basedon the education topic. Although a group contained patients with differentcognitive deficits, relative emphasis was placed on the prominent cognitivestyle deficit of a patient. To provide the best test of CET and EST, mostCET patients (74%) were treated by 3 psychologists (S.F., D.G., and A.G.)who, by inclination and training, were disposed toward cognitive interventions.Two master's-level psychiatric nurses (S.C. and A.L.D.), each with 9 years'experience providing personal therapy (PT),^37,38 onwhich EST was based, treated most EST patients.

A review of patient performance on each core exercise indicated thatall CET patients improved over baseline before proceeding to a new exercise.All patients except 1 with early treatment termination completed the 3 attentionexercises, and more than 90% of all memory exercises and problem-solving routineswere completed. Excluding 6 early terminators (who accounted for most noncomplianceand missing data), the 61 CET recipients who completed 2 years of treatmentkept a mean (SD) of 91.3% (4.4%) of their scheduled or rescheduled appointmentsper therapist. Missed group sessions were reviewed on videotape, and missedcomputer sessions were rescheduled. Cognitive enhancement therapy was an open,structured, manual-guided intervention that was supervised by 2 CET designers(G.E.H. and S.F.). Computer performance data and adherence to a written groupagenda served to ensure fidelity.

Enriched supportive therapy included most practice principles of thebasic and intermediate phases of the demonstrably effective PT approach.^37,38 The intent was to provide a stringenttest of CET. Enriched supportive therapy encouraged illness self-managementthrough the control of subjective cues of distress that might lead to destabilizationor social dysfunctioning. Phase 1 provided psychological and material support,psychoeducation regarding the nature and treatment of schizophrenia, resumptionof instrumental tasks, role restructuring, and basic skills training in stressavoidance. Phase 2 included a personalized education concerning vulnerabilityto stress, adjustment to disability, identification of early signs of decompensation,and stress management strategies. (These EST components were also made availableto CET patients through the group curriculum.) Strategies related to socialcognition, such as perspective taking, criticism management, and most PT Advanced-Phasetechniques, were strictly avoided. Enriched supportive therapy was intendedto be applied weekly in phase 1 and biweekly in phase 2. No attempt was madeto match EST to CET hours. (Artificially increasing the hours of a controlintervention has been shown to have an adverse effect,¹⁷ andhours of psychosocial treatment have most often not been associated with greatereffects in large, well-controlled, multisite studies.^39-41)A mean (SD) of approximately 90.8% (6.1%) of scheduled or rescheduled sessionsper therapist were kept by the 46 EST patients who completed 2 years of treatment.Enriched supportive therapy was a manual-directed, office-based interventionsupervised by a codesigner of the PT approach. Random audiotaped sessionswere scale rated for fidelity during the initial years of study. Cost considerationsand the absence of therapist drift led to discontinuation of monitoring inthe final year of grant support.

All patients were administered a Food and Drug Administration–approvedantipsychotic medication. Medication was not controlled, and compliance washigh among all patients, as intended. At baseline, 33.5% of patients receivedclozapine, 28.9% received an atypical antipsychotic medication (mostly risperidoneor olanzapine), and 35.5% received a conventional neuroleptic, typically atthe minimum effective dose. By 2 years, these percentages had changed to 41.6%,33.6%, and 24.8%, respectively. Analyses of medications at baseline and at1 and 2 years revealed no treatment group differences in either the type ordose of medication or in compliance. Thus, the differential psychosocial treatmenteffects reported herein cannot be attributed to medication differences. Medicationwas prescribed by an experienced research psychiatrist (H.P.).

Partially correlated behavioral assessments of social cognition, cognitivestyle, symptoms, and adjustment were selected (or developed) before the studythat reflected the observations of clinicians, patients, and family members.Neuropsychological tests of attention, memory, and problem solving were chosenthat had historically indicated schizophrenia cognitive impairment.^42,43 Neuropsychological testing was administeredunder highly standardized conditions by a psychometrician (R.Z.), who wasuninvolved with treatment, and served as a more objective, longitudinal outcomemeasure on which patients had not been trained.

Before analysis, highly reliable, multivariately derived indices ofpsychosocial behavior and cognitive functioning were constructed. These "composites"provided protection against the potential error contained in extensive univariatetesting that could not be realistically accommodated with Bonferroni corrections.Within each domain, we first selected measures that at face value representedhypothesized outcomes. Following suggestions for constructing summary scores⁴⁴ and approaches that would increase power by reducingwithin-group variance,⁴⁵ stringent reliabilitycriteria were applied. Candidate variables for a composite had to satisfytest-retest reliability criteria of 0.40 or greater, using the time seriespanel method across baseline and 1 and 2 years⁴⁶;mean interitem correlations within a composite of 0.30 or greater; mean item-totalscore correlations of 0.50 or greater (with no item <0.30), and an α≥.80for the composite itself. Table 3 describesthe composites and their reliabilities. The (Bracy) version of the ContinuousPerformance Test⁵⁹ used by us and the measureof Gist Extraction Deficits failed retest reliability. Also, the Brief PsychiatricRating Scale Thought Disorder factor, the Major Role Adjustment InventoryRelationships in the Home and Sexual Relationships measures also failed retestreliability because of low variance. Although they satisfied retest reliabilitycriteria, all family measures and most patient measures showed inadequateinternal consistency within the assigned composite (or any other composite).Similarly, the Benton Verbal Fluency Test⁶⁰ wasinsufficiently correlated with other neurocognitive measures and was alsoexcluded.

During pilot studies (36 patients and 6 raters), ongoing revisions ofnew instruments were made until interrater reliability coefficients were 0.77or greater, with insignificant rater F ratios. The Cognitive Styles EligibilityCriteria (Table 1) were derivedfrom studies^61-63 ofacute schizophrenia symptoms reflecting impoverishment (negative symptoms),disorganization, and reality distortion. (During remission, the latter constructseemed to represent a type of rigid, inflexible cognition.²⁸)These styles reflected the cognitive processes with which patients approachedinstrumental and relationship tasks. Selected NP and neurobiological datasupport the validity of the constructs.^61,62,64-66 Thestyles serve to reduce the heterogeneity in schizophrenia diagnosis⁶² and, we believed, the diverse adaptations to cognitiveimpairments.²⁸ Clinician retest reliabilitycoefficients were 0.78, 0.64, and 0.69 for the 3 styles, with high independenceof styles (intercorrelations ≤0.10). The associated Cognitive Styles Inventorywas a 46-item micromeasure of the 3 styles, which were also well discriminated.

Social cognition was a more difficult concept to measure, with morethan 100 definitions available.⁶⁷ We abandonedselected "proxy" measures, previously developed on low-functioning, hospitalizedpatients (eg, the Social Cue Recognition Test⁶⁸),that showed a ceiling effect among our stable outpatients. We attempted todistinguish social cognition (awareness of relationship aspects) from socialadjustment (instrumental role performance or capacity). We relied on the SocialCognition Profile, a 50-item checklist of relevant behaviors gleaned fromthe literature, for example, Selman and Schultz.⁶⁹ Aprincipal components analysis of the item correlations across rating periods(using >1500 ratings from patients, families, and clinicians) yielded 4 factorswith loadings greater than 0.40 and few split-loaded items. Although the factorstructure was nearly identical across different observers (specific variance),when scored, the patient and family factors were highly redundant (commonvariance). However, the clinician factors were well discriminated. The SocialCognitive Deficit Eligibility Criteria (Table 1) represented our judgment of functionally incapacitatingdeficits. Because few patients worked full time, we scaled the 4 criterionareas (50 items) used by the Social Security Administration⁵⁸ fordetermining employment disability (see Social Adjustment in Table 5). Component areas met high interrater (>0.77), retest (>0.62),and internal consistency (>0.87) criteria. Information on these scales isavailable from one of us (G.E.H.). The results provided herein represent initialvalidity data.

As expected, selected behavioral composites were moderately correlatedacross periods (Cognitive Style and Symptoms, r =0.25; Social Adjustment and Social Cognition, r =0.50; and Cognitive Style and Social Cognition, r =0.44), indicating some redundancy but nevertheless unique variance. The Neurocognitionand Processing Speed composites were only modestly correlated at baseline(r = 0.30). Each composite correlated higher withitself across periods than with any other composite.

Composite scores were standardized according to a baseline mean (SD)of 50 (10). The regressed composite change scores between baseline and 1 and2 years were tested using 2-tailed (P<.05) analysisof covariance (ANCOVA) and confirmed by a linear trend analysis of treatmentdifferences across 2 years. In addition to treatment and sex, length of illness(≤15 vs >15 years), baseline positive symptoms (less than mild vs greaterthan mild), and IQ (≤98 vs >98) dichotomized at the mean were entered as(moderator) factors; the baseline mean served as the covariate. Only the simpleinteractions of a moderator and treatment condition were examined. Univariate(ANCOVA) analyses were undertaken to identify the relative item contributionswithin the significant composites at 2 years.

At baseline, the CET and EST groups were not different on any of thedemographic or historical variables examined or on the ANCOVA composites orcomposite items except for the following 4 variables: EST assignees were ratedon 3 variables as being less withdrawn, being more socially engaged, and havinga higher Global Assessment Scale score by clinicians. Recipients of CET hada somewhat higher estimated average IQ than EST patients (99 vs 95) on thecombined Wechsler Adult Intelligence Scale (WAIS) subtests.

Table 4 describes the treatmenteffects on the 6 composites at 1 and 2 years. At 12 months, a robust effectof CET is observed on Processing Speed and Neurocognition, with marginal CETeffects observed on the behavioral domains of Cognitive Style, Social Cognition,and Social Adjustment. No effect was seen on the Symptoms composite, as expected.By 2 years, all composites reflected significant differential improvementfor CET recipients except for Symptoms. (Linear trend analysis of treatmentdifferences across 2 years yielded the same results.) Effects of CET are broadlyapplicable to patients because there were no significant (moderator) ANCOVAor linear trend interactive effects with treatment.

Effects of CET were unrelated to the receipt of an atypical antipsychoticmedication. A 6-factor ANCOVA that entered baseline drug type as an additionalmoderator (clozapine, atypical antipsychotic, or conventional antipsychotic)revealed no treatment interactions. Drug type had a significant main effecton Cognitive Style and Social Adjustment at 2 years, but not in the expecteddirection.⁷⁰ These effects indicated greaterimprovement overall among patients receiving the minimum effective dose ofa conventional antipsychotic agent.

Effect sizes can provide a better indication of clinical significancethan P values.⁴⁴Figure 1 illustrates treatment effect sizes(Cohen d) that are appropriately calculated on the basis of variance associatedwith the actual regressed change scores tested by ANCOVA⁷¹ ratherthan baseline variance. (Within and between groups, an effect size >0.40 issignificant at the P = .05 level, with n = 50 pergroup.)

By 2 years, effect sizes for CET exceeded 1 SD for all composites exceptSymptoms. Large effect size differences (>0.50) between treatments favoringCET occur on the Processing Speed (0.92) and Neurocognition (0.63) compositesat 1 year and again on all composites at 2 years except Symptoms. Recipientsof EST also show clinically meaningful change by 2 years on many composites,particularly Neurocognition (0.94), an unexpected effect for a psychotherapythat was devoid of targeted cognitive strategies. However, Processing Speedwas unaffected by EST. As suggested elsewhere,⁷² apsychotherapeutic approach could lower the stress (arousal) that might exacerbateNP deficits.

The contribution of the 45 component items to significant compositeeffects at 2 years are presented in Table5. Various measures of reaction time, except for the embedded stimulustest, show uniform contributions to the Processing Speed effect. Verbal memoryimprovement accounts for most change in Neurocognition, with some evidenceof improved cognitive flexibility and problem solving. The WAIS Digit Symbol(an indicator of psychomotor speed, cognitive flexibility, and vigilance)also contributes. Regarding Cognitive Style, impoverished cognition improvesmost and rigid thinking improves least, with disorganization showing intermediateimprovement. Differential improvement in Social Cognition is uniform acrossthe composite measures except for the Support factor. Few measures of SocialAdjustment achieve statistically significant levels of differential improvement,with measures of improved work capacity indicating relatively greater improvementthan actual role performance.

There were 9 significant interactions with treatment (P≤.05). Among patients who were ill for less than 15 years, CETrecipients improved more than EST recipients on measures of social functioning,the WAIS Digit Symbol Test, and impoverished cognition. Among patients whowere ill for more than 15 years, CET recipients improved more than EST patientson simple reaction time. Patients in the CET group with fewer positive symptomsat baseline improved more than less symptomatic EST patients on the GlobalAssessment Scale, Wisconsin Card Sorting Test % Conceptual Level Response,and Impoverished cognition. Recipients of CET with a higher IQ improved moreon Trails B than EST patients with a higher IQ, and CET patients with a lowerIQ increased their WAIS Vocabulary score more than EST patients with a lowerIQ. There were no significant interactions with sex.

After 2 years of treatment, CET demonstrated significant differentialeffects on all domains of behavior and cognition tested except residual symptoms,the latter likely speaking to the clinical stability of patients at baselineand the efficacy of PT in symptom management.³⁸ ModestCET effects on the behavioral composites at 1 year possibly reflect minimalexposure to the CET social group curricula. At the moment, it is unclear howthe cognitive changes are causally "linked" to changes in behavior.⁷³ Mediator analyses⁷⁴ thatshould help clarify causal links are planned.

Processing speed is a primary mediator candidate because improvementwas also observed immediately after training, was sustained over time, andwas reserved exclusively for CET recipients. This result supports the observationsof other researchers.^10,15,75 Verbalmemory is another mediator candidate because it improved most in the firstyear among the neurocognitive measures and is thought to best predict functioning.⁷⁶ Computerized cognitive training may have been necessaryfor the differentially superior CET effects; however, substantive gains werealso made by EST patients on neurocognition and behavior.

Intelligence is important for understanding treatment effects, particularlythose involving cognitive deficits, because higher-IQ patients improved morethan lower-IQ patients in both treatments on 12 measures. Although many patientswith schizophrenia have a "below average" IQ, and some are alleged to experiencea decline in IQ after schizophrenia onset, most are thought to remain withinthe range of "normal" intelligence and often recover intellectual functioningafter successful treatment.⁷⁷ Much of the therapeuticpessimism concerning functional recovery in schizophrenia could be attributedto the minority of patients who have concurrent mental insufficiency (IQ <80),organic brain disorder, or persistent positive symptoms. In our experience,most patients with schizophrenia learn from developmentally appropriate experiences,including their "mistakes." Recommendation for a specific psychosocial interventionin schizophrenia can likely be best made in terms of patient characteristics:intelligence, length of illness, and phase of illness.³⁸

There are potential limitations to our results. Most behavioral findingsreflect the assessments of "unblinded" clinicians, although equally robusteffects occurred on the blinded NP tests. The relative absence of strong effectsin blinded psychosocial treatment trials could speak either to the controlof a possible clinician bias or to the issue of questionable validity whenassessments are decontextualized.⁷⁸ Studiesare needed that evaluate the success of blinding psychosocial treatments andthat establish the validity of blinded and clinician assessments against independentcriteria (eg, NP tests and neuroimaging changes). Replication studies of CETmight profitably blind ratings by using structured interview assessments,such as the Social Adjustment Scale II.⁷⁹ Otherwise,relevant and valid measures of social cognition among stable outpatients werenot available at the start of the study and had to be constructed. Furtherdevelopment of these and other proxy measures is an important task.

There are clear reservations regarding the use of NP tests as longitudinaloutcome measures,^80,81 a conditionfor which these tests were not designed. Schizophrenia cognitive deficitsseem to be more generalized than focal^82,83;thus, the specification and sequencing of training is uncertain. Neither isit clear whether a core cognitive deficit has been remediated (or compensated)or whether the improvement only reflects a tangential correlate of the deficit.⁸⁰ Practice effects are also possible, although it isunlikely that NP tests administered on an annual basis are greatly subjectto this bias.⁴⁴ Increased comfort and familiaritywith test procedures,⁸⁰ especially computerizedtests, could have positively biased CET neurocognitive and processing speedperformance given the extensive computer-based training provided. However,the most robust neurocognitive effects (Wechsler Memory Scale and CaliforniaVerbal Learning Test memory effects and the WAIS Digit Symbol) were not computerizedtests. By design, processing speed was a training goal. Otherwise, some researchersmight argue for the dismantling of a multidimensional psychosocial treatmentin search of the therapeutic component. However, attempts to deconstruct thesoftware exercises and group components of the similarly organized Ben-Yishayprogram demonstrated that the whole was clearly superior to the componentparts.⁸⁴

Although the goal of the study was to improve social and nonsocial cognition,some researchers might cite the relative absence of robust CET effects ontraditional "real-world" outcomes. Relapse (or days hospitalized) was nota hypothesized outcome because most patients were many years past their lastpsychotic episode. Expected^27,85 andobserved relapses (9 EST and 3 CET patients) were quite low. Patients wereolder and ill for longer periods than in our previous studies,^86-89 andmany had found their "niche," desiring an improved quality of life more thana job or return to school. Furthermore, most patients received disabilitybenefits and the associated health insurance. Coupled with a difficult locallabor market and the work disincentives of disability programs, many patientsfeared a loss of benefits if they increased part-time work or secured full-time,minimum-wage employment. (Our program did not provide job finding, training,or placement but relied on the referral of interested patients to the localOffice of Vocational Rehabilitation.) Nevertheless, CET gains in adjustmentwere observed, an encouraging outcome given the stable pretreatment statusof patients and the efficacy of EST. The CET trend in employment (Table 5) is attributed to the greater numberof CET patients who found volunteer positions (26% of CET and 8% of EST patients)rather than paid employment. The CET gain in Major Role Performance (Table 5) reflects the observation that60% of CET patients vs 42% of EST patients equaled or exceeded their bestpremorbid or previous functioning. The 3-year follow-up should provide a betterassessment of adjustment effects.

Finally, CET results cannot be generalized to all patients with schizophrenia.It was designed as a recovery-phase intervention for stable, non–substance-abusingambulatory patients with an IQ of 80 or higher. Behavioral, compensatory approaches¹⁷ might be preferable for the more severely impairedor intellectually compromised subpopulations of patients. Cognitive behaviortherapy, PT, skills training, and family psychoeducation seem to be more appropriatefor those in the subacute, stabilization phase.³⁸ Althoughwe lack data for a cost-benefit analysis, programs that rely on small-grouprehabilitation approaches should find that CET is cost-effective. Most patients(87%) highly endorsed their CET experience. The stable patient with rehabilitationpotential is more likely to functionally improve after extended cognitiverehabilitation. In the era of managed cost, even expert consensus opinion⁹⁰ tends to disenfranchise this substantial segmentof the schizophrenia population.

In conclusion, the cognitive disabilities of schizophrenia do not needto be the persistent deficits described in numerous naturalistic, longitudinalstudies.^91,92 Instead, many ofthese disabilities are capable of improvement after adequate exposure to cognitiverehabilitation.

Correspondence: Gerard E. Hogarty, MSW, University of PittsburghMedical Center, Western Psychiatric Institute and Clinic, 3811 O'Hara St,Pittsburgh, PA 15213 (hogartyje@upmc.edu).

Submitted for publication January 27, 2003; final revision receivedMarch 11, 2004; accepted March 16, 2004.

This study was supported by grant MH-30750 from the National Instituteof Mental Health, Bethesda, Md.

We thank the clinicians of the University of Pittsburgh Medical Center'sComprehensive Care Service and Diane Ludewig, PhD, and Prabir Mullick, MD,for patient referrals; David Kupfer, MD, chairperson of the University ofPittsburgh Medical Center Department of Psychiatry, and Robert Prien, PhD,formerly of the National Institute of Mental Health, Bethesda, for extendedsupport and encouragement at critical times; Sander Kornblith, PhD, for providingdirection and treatment fidelity for EST; Odie Bracy, PhD, for graciouslyproviding his memory and problem-solving exercises; Yehuda Ben-Yishay, PhD,for offering his attention training software and opening his program to multiplevisits by CET therapists before the study; Michele Bauer for preparing grantapplications and the manuscript and for overseeing administrative details;and the study patients for being dedicated to their recovery (which was aconstant source of inspiration and encouragement).