Neuropsychological Change in Early Phase Schizophrenia During 12 Months of Treatment With Olanzapine, Risperidone, or Haloperidol

Background  The purpose of this investigation was to test the efficacy of novel antipsychotic medications in the treatment of cognitive impairment in early phase schizophrenia.

Methods  Sixty-five patients in this multicenter double-blind study were randomly assigned to olanzapine (5-20 mg), risperidone (4-10 mg), or haloperidol (5-20 mg). Standard measures of clinical and motor syndromes were administered, as well as a comprehensive battery of tests to assess (1) motor skills, (2) attention span, (3) verbal fluency and reasoning, (4) nonverbal fluency and construction, (5) executive skills, and (6) immediate recall at baseline and after 6, 30, and 54 weeks of treatment.

Results  The general cognitive index derived from the 6 domain scores revealed a significantly greater benefit from treatment with olanzapine relative to haloperidol and olanzapine relative to risperidone, but no significant difference was shown between risperidone and haloperidol. The improvement related to olanzapine was apparent after 6 weeks and enhanced after 30 and 54 weeks of treatment. Exploratory within-group analyses of the 6 cognitive domains after a conservative Bonferroni adjustment revealed a significant improvement with olanzapine only on the immediate recall domain, and similar analyses of the 17 individual tests revealed a significant improvement with olanzapine only on the Hooper Visual Organization Test.

Conclusions  These data suggest that olanzapine has some superior cognitive benefits relative to haloperidol and risperidone. A larger sample replication study is necessary to confirm and generalize the observations of this study and begin evaluation of the implications of this change to cerebral function and quality of life for people with schizophrenia.

COGNITIVE IMPAIRMENT has been intrinsically linked to schizophrenia since Kraepelin's first descriptions of dementia praecox, and it has been documented in many cognitive domains, including attention, executive functioning, verbal recall, visuospatial abilities, and fine motor skills.^1-3 The deficits may represent a core pathophysiological feature because they are apparent early in the course of the illness and cannot be entirely attributed to chronicity, medications, or florid symptoms.^4-9 Several studies have also demonstrated an association between these deficits and poor social and occupational outcomes.^10-12 Treatment with the potential to diminish cognitive impairment may thus have implications for a core pathophysiological feature of the disorder with relevance to long-term outcome.

Traditional antipsychotic medications have no efficacy in the treatment of cognitive impairment beyond occasional improvement in attention.^13-15 A presumption that the lack of efficacy was secondary to cerebral damage was initially supported by an association between deficient mental status and atrophy on findings from computed tomography.^16,17 A challenge to this view has resulted from more recent preliminary demonstrations of circumscribed benefits to verbal production, visuomotor tracking, and immediate verbal recall with clozapine, and to attention and verbal working memory with risperidone.^18-24 Olanzapine is a new atypical antipsychotic agent with a potential for cognitive improvement suggested by converging lines of preclinical evidence showing dose-dependent increases in prefrontal cortex dopamine and norepinephrine, and increased early gene (ie, c-fos) expression in the prefrontal cortex, outcomes that are similar to those found in treatment with clozapine but distinct from treatment with haloperidol.^25,26 Olanzapine also antagonizes phencyclidine disruption of prepulse inhibition, which is presumed sensitive to prefrontal cortical function.^27,28 Cognitive improvement with clozapine may be related to alternations of prefrontal metabolism or neurotransmission, and thus similar alternations may suggest cognitive benefits from olanzapine as well.²⁹

The purpose of our study was to assess the comparative efficacy of olanzapine, risperidone, and haloperidol on cognitive impairment early in the course of schizophrenia. The primary hypothesis was that the 3 treatments would result in differential improvement on a general cognitive index derived from 6 domains of cognitive function: (1) motor skills, (2) attention span, (3) verbal fluency and reasoning, (4) nonverbal fluency and construction, (5) executive skills, and (6) immediate recall. Exploratory analyses were undertaken to assess the time course, domain specificity, and test specificity of any observed changes. Subsidiary analyses assessed the relation between a change in cognitive skill and changes in clinical or motor syndromes, as well as an association with anticholinergic medication.

Sixty-five clinically stable outpatients were recruited for this study and gave signed informed consent to participate at 1 of 19 medical centers across Canada that had received approval from a local institutional or ethics review board listed in the acknowledgments. The diagnosis of schizophrenia as defined by DSM-IV was confirmed on clinical interview by the principal investigator at each site. The sample included men and women aged 18 to 65 years who were within 5 years of their first exposure to neuroleptic treatment and had symptom severity at least in the mild range. Medical, physical, and psychiatric examinations were completed prior to random assignment to a treatment group, and participants were excluded from the study if they were pregnant or lactating, had prior medical histories of central nervous system disease or severe head injury, or if they had active serious illness or substance abuse disorders in the previous 30 days.

The study consisted of a 1-month prestudy stabilization period, a 2- to 9-day medication washout period, and a 54-week double-blind randomized treatment period. During the prestudy period, antipsychotic medications were slowly titrated down to the minimum effective dose prior to cessation of treatment. At the end of the neuroleptic-free washout period, the participants underwent a complete baseline examination of clinical symptoms, motor signs, and neuropsychological status. Assessments of clinical symptoms and motor signs were repeated weekly for the first 6 weeks and monthly thereafter. The neuropsychological examination was repeated 6, 30, and 54 weeks after random assignment to treatment. Early discontinuation patients were evaluated on their last study visit. A computer-generated random number table was used to assign each medication to a subject number assigned in sequential order of recruitment. During the first week of double-blind treatment, the olanzapine and haloperidol groups each received 10 mg per day of study medication, and the risperidone group was titrated up from 2 mg to 6 mg per day in daily 2-mg intervals. Throughout the subsequent 53 weeks, the maintenance dose was adjusted at the discretion of the study clinicians within a daily range of 5 mg to 20 mg of olanzapine or haloperidol, and 4 mg to 10 mg of risperidone. No additional antipsychotic medications were allowed, but patients could receive other adjunctive treatments as required.

Psychopathologic symptoms were rated with the Positive and Negative Syndrome Scale³⁰ and motor signs were rated with the Extrapyramidal Symptom Rating Scale.³¹ An estimate of premorbid intellect was provided by the Peabody Picture Vocabulary Test–Revised.³² The neuropsychological battery consisted of a series of standardized tests with published normal control data that have shown a wide range of test scores in subjects with schizophrenia.^1-3 The 17 individual test scores were reduced to 6 cognitive domains based on clinical judgment as discussed in detail in a previous report.³³ (1) Motor skills were measured with the Grooved Pegboard Test and the Finger Tapping Test, and standardized scores were adjusted for handedness, age, sex, and education.^34-39 (2) Attention span was measured with the Verbal and Nonverbal Span tests from the Attention Index of the Wechsler Memory Scale–Revised⁴⁰ adjusted for age. (3) Verbal fluency and reasoning were measured with the Controlled Oral Word Association Test⁴¹ adjusted for age and sex and the Similarities Subtest of the Wechsler Adult Intelligence Scale–Revised⁴² adjusted for age. (4) Nonverbal fluency and construction were measured with the Design Fluency Test,^43,44 the Hooper Visual Organization Test,⁴⁵ and the age-adjusted Rey-Taylor Complex Figure Copy Test.^46-48 (5) Executive function, defined here as the ability to initiate, sustain, and shift among numbers and symbols, numbers and letters, and categorical sets, was assessed with the Digit Symbol Subtest of the Wechsler Adult Intelligence Scale–Revised⁴² adjusted for age, Trailmaking B^34,35 adjusted for age, sex, and education, and the Wisconsin Card Sorting Test Perseverative Errors⁴⁹ adjusted for age. (6) Immediate recall was measured with the serial list learning procedures of the Rey/Crawford Auditory Learning Test⁵⁰ adjusted for age and sex, the Rey Design Learning Test^51,52 adjusted for age, the stories from the Auditory Comprehension Test,^41,53 the figures from the Visual Reproduction Tests of the Wechsler Memory Scale^54,55 adjusted for age, and the Rey-Taylor Complex Figure Immediate Recall Test^46-48 adjusted for age. Paired alternate forms counterbalanced for order of presentation were used for all tests of immediate recall, and a unique form of the Auditory Comprehension Test was used on each testing occasion, also in paired counterbalance. The tests were administered by a psychology assistant (trained by S.E.P.) at each site and monitored by direct on-site supervision from the first testing session until remote monitoring was sufficient to ensure reliability.

The primary aim of our study was to test the hypothesis that the 3 treatments would result in differential improvement on cognitive function as measured by a general cognitive index. The general cognitive index was computed for each subject from the mean of 6 domains of cognitive function corresponding to (1) motor skills, (2) attention span, (3) verbal fluency and reasoning, (4) nonverbal fluency and construction, (5) executive skills, and (6) immediate recall. The domain scores consisted of the average test scores within each domain after z score standardization using published normal^34-55 control data (available on request from S.E.P.). The average of the standardized domain scores provided a general cognitive index interpreted as the number of SD units of disparity of each participant from the expected score of a healthy control sample. Differences in the general cognitive index between baseline and end point with the last observation carried forward were compared using analysis of variance and between-treatment contrasts. The model contained a fixed term for pooled investigative site and used each patient's baseline score as a covariate. This was followed by a series of t test comparisons of baseline score with the cumulative score at each of 3 follow-up assessments. To ensure that any changes observed in this analysis could not be attributed to differential rates of attrition, pairwise comparisons using only subjects who completed all 3 follow-up assessments were made, and a repeated-measures analysis of covariance was computed using all subjects in a general linear mixed model with an unstructured covariance matrix, including fixed-effect terms for time, treatment, pooled investigator × treatment interaction, and a random-effect term for subject within treatment by investigator. Subsidiary correlational analyses examined the relative contributions of clinical syndromes and motor syndromes to changes in the general cognitive index. Specific domain and test scores were examined with exploratory within- and between-group comparisons of the change from baseline. The critical α for the exploratory comparisons was derived from a conservative Bonferroni adjustment resulting in a critical α of P<.0012 for evaluation of the 42 domain score comparisons (Table 1) and P<.0004 for the 126 test score comparisons (Table 2).

Potential differences in demographic characteristics were tested with the Pearson χ² statistic for categorical scores and analysis of variance for continuous scores. Analysis of variance was also used to compare clinical and motor syndromes at baseline prior to a comparison of change from baseline to end point, with the last observation carried forward using analysis of covariance and terms entered for investigative site and baseline syndrome score. The clinical and motor syndrome data were then analyzed with pairwise and between-group t test comparisons. Preliminary within-group comparisons were used to assess contributions from adjunctive use of anticholinergic medication. Two-tailed tests of significance were used for all statistical tests described in the "Results" section.

Sixty-five patients completed the baseline assessment and were randomized to 1 of 3 treatment groups (olanzapine, n = 21; haloperidol, n = 23; risperidone, n = 21) (Table 3). Subjects in the baseline sample were early in the course of the illness and consisted of relatively young, predominantly male, high school–educated patients with average premorbid intellect and mild impairment on the general cognitive index. There was no significant difference in baseline general cognitive index scores of patients tested in eastern, middle, or western Canada (F_2,60 = 0.49, P = .62), and there was no difference between patients tested at each site by the study monitor or patients tested by the site examiners (t₆₁ = 1.35, P = .18). There were no significant differences at baseline between the olanzapine, haloperidol, or risperidone treatment groups with respect to age, sex, education, premorbid intellect, age of onset, or duration of illness (Table 3).

The 3 treatment groups had a similar mild severity of positive and general syndromes at baseline (Table 3), but there was a significant difference for the negative syndromes resulting from more severe ratings of the haloperidol group relative to the risperidone group (t₄₂ = 3.00, P = .005). There was no difference between groups in the amount of improvement in clinical status throughout the course of the study (Table 4). The 3 treatment groups showed an equivalent mild severity of motor syndromes at baseline and no difference in the change in dystonia or dyskinesia at end point. Parkinsonian symptoms showed differential changes between groups, with contrast comparisons showing more improvement in the olanzapine group relative to the haloperidol group (Table 4).

Ten patients were discontinued from the study after a second evaluation of clinical and motor syndromes but prior to a second examination of neuropsychological status at 6 weeks, leaving 55 patients with prospective cognitive data. The early termination sample was not different from the prospective sample regarding sex, years of education, estimated premorbid intellect, age of onset, duration of illness, general cognitive index, or severity of positive syndromes, negative syndromes, parkinsonism, or dystonia. However, the early termination sample was more likely to have been assigned to haloperidol (n = 8) than to olanzapine (n = 1) or risperidone (n = 1) (χ²₂ = 10.29, P = .006). They were also slightly younger (mean ± SD age, 23.53 ± 5.10 years) than the remaining subjects (t₆₃ = 2.26, P = .03) and showed more dyskinesia (Extrapyramidal Symptom Rating Scale score, 3.70 ± 6.31) (t₆₃ = 2.68, P = .009). An additional 27 subjects were discontinued from the study prior to the full 54 weeks of the trial, including 7 prior to a third follow-up examination. The predominant reason for discontinuation was adverse events in the haloperidol group, patient decision in the olanzapine group, and lack of efficacy in the risperidone group (Table 5).

Within the group of 55 patients with at least 1 prospective neuropsychological evaluation, there was no significant difference between treatment groups in the mean ± SD number of days from baseline to endpoint assessment (olanzapine, 275.50 ± 148.13; haloperidol, 267.87 ± 134.48; risperidone, 256.65 ± 119.01). There were also no significant treatment group differences on baseline demographic characteristics, clinical syndromes, motor syndromes, or the general cognitive index. The mean ± SD of the daily medication dose in milligrams throughout the course of the trial was 11.00 ± 4.60 mg for olanzapine, 9.70 ± 4.20 mg for haloperidol, and 6.00 ± 1.80 mg for risperidone, showing little discrepancy from the mean of the modal daily dose (11.70 ± 5.30 mg, 10.20 ± 4.90 mg, and 6.10 ± 2.00 mg, respectively), or the median modal daily dose (10 mg, 10 mg, and 6 mg, respectively). Most of risperidone-treated patients (76%) received a modal dose of 6 mg or less per day throughout the course of the study, with 33% receiving 4 mg per day and 43% receiving 6 mg per day.

The treatment groups showed a substantial difference in the amount of change in the general cognitive index from baseline to end point (F_2,43 = 9.84, P<.001) characterized by advantages of olanzapine vs haloperidol (F_1,43 = 18.04, P<.001) and olanzapine vs risperidone (F_1,43 = 9.10, P = .004) but no difference between haloperidol and risperidone (F_1,43 = 2.17, P = .15) (Table 6). Within-group comparisons of change at each follow-up assessment showed significant improvement with olanzapine at weeks 6, 30, and 54, and a significant improvement with risperidone at week 54, but no improvement with haloperidol at any of the prospective assessments. Similar results were obtained with both the analysis of cases with complete data at 3 follow-up assessments and with the repeated-measures analysis of all cases for olanzapine and haloperidol, though the benefit from risperidone was not apparent in these alternate analyses. There were no significant correlations between change in the general cognitive index and change in any of the clinical or motor syndromes from the Positive and Negative Syndrome Scale or the Extrapyramidal Symptom Rating Scale.

Exploratory analyses assessed group differences in the change on each cognitive domain and test (Table 1 and Table 2). With Bonferroni adjustment of α, the domain scores showed a benefit of olanzapine only on immediate recall (P<.0012) and the test scores showed a benefit of olanzapine only on the Hooper Visual Organization Test⁴⁵ (P<.0004). Although not significant with Bonferroni adjustment, a differential pattern of improvement was suggested by exploratory analyses without the conservative statistical correction. Uncorrected within-group comparisons suggested improvement in 5 (83%) of 6 domains with olanzapine, with medium to large effect sizes for attention span, motor skills, nonverbal fluency and construction, executive skills, and immediate recall. In contrast, improvement was suggested in only 2 (33%) of 6 domains with risperidone, with a small to medium effect size observed for verbal fluency and reasoning and immediate recall. Improvement with haloperidol was suggested for only 1 (17%) of 6 domains in a small to medium effect for attention span. Similarly, excluding the measure of premorbid intellect (Peabody Picture Vocabulary Test–Revised³²), the exploratory within-group comparisons suggested improvement in the olanzapine group on 10 (59%) of 17 tests, with a small to medium effect size on design list learning, a medium effect size on verbal fluency, and medium to large effect sizes on Grooved Pegboard,^34-39 Visual Span,⁴⁰ Wisconsin Card Sorting Test⁴⁹ Perseverative Errors, verbal list learning,⁵⁰ nonverbal fluency,^43,44 Hooper Visual Organization Test,⁴⁵ Rey-Taylor Complex Figure Copy, and Rey-Taylor Complex Figure Immediate Recall.^46-48 Exploratory risperidone comparisons suggested a beneficial change on only 3 (18%) of 17 tests, with a small to medium effect size on the Similarities Subtest, a medium effect size on nonverbal fluency, and a medium to large effect size on the Rey-Taylor Complex Figure Immediate Recall. Exploratory haloperidol comparisons suggested beneficial change on only 2 (12%) of 17 tests, with a small to medium effect size on the Hooper Visual Organization Test⁴⁵ and a medium to large effect size on the Rey-Taylor Complex Figure Immediate Recall.^46-48

A significantly different proportion of each sample received anticholinergic medications within 48 hours preceding the final evaluation (χ²₂ = 12.12, P = .002). A greater proportion of the haloperidol group required anticholinergic treatment (11/15 [73.3%]) relative to the risperidone group (9/20 [45%]), which was relatively more frequent than in the olanzapine group (3/20 [15%]). The last observation carried forward analysis after 6, 30, and 54 weeks of treatment, respectively, showed no difference over time in the proportion of patients receiving anticholinergic treatment in the haloperidol (67%, 73%, and 73%, respectively), olanzapine (15%, 15%, and 15%, respectively), or risperidone (40%, 45%, and 45%, respectively) samples. A stratification based on anticholinergic use within each treatment group showed no significant differences between subgroups receiving or not receiving anticholinergic treatment on the general cognitive index, nor on any of the individual cognitive domains after Bonferroni adjustment of α. The only suggestion of a detrimental anticholinergic effect was apparent in the risperidone group after 6 weeks (t₁₈ = 2.33, P = .03) and after 54 weeks (t₁₈ = 2.64, P = .02) for the immediate recall domain. Thus, although anticholinergic medications did not have a robust effect on cognitive change, a contribution from this supplement cannot be ruled out entirely, particularly when used with risperidone.

Olanzapine produced a substantial gain in cognitive skills greater than that observed with risperidone or haloperidol that was apparent after 6 weeks of treatment and enhanced after 30 and 54 weeks of treatment. An attempt to qualify the olanzapine improvement in relation to domains of cognitive skill was hampered by limited statistical power after conservative Bonferroni adjustment, but a significant improvement was evident for the immediate recall domain and the Hooper Visual Organization Test.⁴⁵ The lack of change observed in subjects treated with haloperidol is consistent with prior reports.^13-15 The lack of change in subjects treated with risperidone is somewhat discrepant with prior reports, but the comprehensive neuropsychological battery used in our study is not directly comparable to the restricted range of tests used in prior research with risperidone.^22-24 Although preliminary, these results implicate the value of additional research toward a neuropsychological differentiation of antipsychotic treatments.

Caution is warranted in the generalization of these results until completion of a larger sample replication study with further delineation of a primary alternation in central nervous system physiologic characteristics from secondary contributions from extrapyramidal syndromes, adjunctive use of anticholinergic supplements, changes in clinical syndromes, or repeated exposure to the neuropsychological instruments. Although this investigation showed a benefit of olanzapine over haloperidol on parkinsonian symptoms, it did not differentiate between treatments on other motor syndromes. The more frequent use of anticholinergic medications in the risperidone and haloperidol groups sustains the possibility of a contribution from extrapyramidal syndromes or anticholinergic supplements to differences in cognitive change, particularly given prior suggestions that anticholinergic treatment may impede new learning and memory.^15,56 The risperidone group gave subtle indication of potential anticholinergic effects on immediate recall, and a larger sample may thus have detected more beneficial effects of risperidone in the absence of anticholinergic treatment. Future attributions of differential effects to anticholinergic medications will also have to address the benefits of olanzapine on immediate recall despite a potent inherent anticholinergic effect.²⁵ Although our study did not show a relationship between change in cognitive and clinical syndrome scores, subtle contributions to cognitive test scores from clinical improvement may become apparent with a larger sample size. In addition, although the literature to date has rarely demonstrated any cognitive change to treatment in subjects with schizophrenia, the positive results of our study underscore the importance of including alternate test forms to control for practice effects in further research. The design of this study addressed practice effects with the use of alternative measures of immediate recall, but it did not control for potential practice effects in other domains of cognitive skill. Although the significant domain-specific effect of olanzapine was observed on the measures of immediate recall for which alternate test forms were used, caution is warranted in the attribution of cognitive change to potential differential neurophysiological effects of treatment.³³

Our results may be specific to patients with a relatively early phase of schizophrenia who have symptom severity and cognitive impairment at least in the mild to moderate range. The negative syndrome scores at baseline are consistent with published data on a large outpatient sample, but the positive syndrome scores are somewhat lower than the published results.³⁰ Caution is therefore warranted in the generalization to patients with more florid symptoms. Also, the general cognitive index is somewhat higher (ie, mean z = −1.10) than that reported in a previous study of patients with early phase schizophrenia (ie, estimated mean z = −1.57),⁵⁷ suggesting caution in generalization to patients with limited intellectual resources and more severe cognitive impairment until larger samples of patients are examined. A study of patients with more chronic and severe disease will be required to determine similar results in a broader group of patients.

Caution in the generalization of our results is also suggested by potential patient-specific characteristics that may have influenced the attrition rates in each treatment group. For example, a sample selection bias against relatively young men with low intellectual resources and more baseline negative and extrapyramidal symptoms was suggested for the haloperidol group owing to substantial attrition of this type of patient. The attrition in the haloperidol group may suggest a subsample of patients with greater cerebral dysfunction prone to adverse reactions to haloperidol. However, this potential bias may have been expected to select for continuation patients more amenable to a therapeutic response, but this group showed the least amount of cognitive change. Moreover, it is unlikely that differential drop-out rates are responsible for the beneficial effects of olanzapine and the lack of effects for haloperidol because the same result was obtained in the intent-to-treat, complete cases, and complete sample repeated-measures analyses. The diminution of statistical power caused by attrition seems only relevant to the risperidone group, in which the intent-to-treat analysis with the last observation carried forward showed a beneficial change in the general cognitive index not apparent with the complete cases or the repeated-measures analyses. It would therefore be premature to entirely rule out potential beneficial effects from risperidone until a larger sample of patients is examined.

The present results may also be specific to patients who can be maintained with approximately 12 mg of olanzapine, 6 mg of risperidone, or 10 mg of haloperidol. A flexible-dose schedule allowed the clinician at each site to select the optimal dose to maximize clinical efficacy and minimize adverse effects. It remains to be determined if more liberal dose ranges than the allowed 5 mg to 20 mg of olanzapine or haloperidol or 4 mg to 10 mg of risperidone would produce the same results. The dose ranges and titration schedules were established from the relevant product monographs, and the recommended ranges have not changed since implementation of this study, despite increasing interest in higher doses of olanzapine and the use of slower titration and lower doses for risperidone. Thus, it is possible that risperidone might result in cognitive benefits not apparent in this study if a patient can be maintained with less than 4 mg per day. However, the only published demonstration of benefits of risperidone in a double-blind clinical trial were reported on a single measure of verbal working memory after an initial fixed dose of 6 mg per day, which is consistent with the modal dose in this study and higher than the dose received by many patients in the risperidone sample.²⁴ However, it would be prudent to delay further generalization of the present results until a broader dose range of treatments has been investigated.

It is becoming increasingly apparent that cognitive impairment and the possible therapeutic reduction of this impairment will have an important contribution to the emotional, interpersonal, and vocational implications of schizophrenia. For example, the effect of improved cognitive skills on insight into illness may well influence vulnerability to depression, and this vulnerability may show an interaction with the chronicity of illness. Whereas it is likely that improved insight and cognitive skill will contribute to a positive mood state among patients with more acute illness, the implications for those with chronic illness are less transparent. Also, positive associations between psychosocial outcome measures and cognitive status alone,¹¹ cognitive status in relation to relevant cerebral processes,¹⁰ and cognitive changes related to clozapine treatment^18,20 all indicate the relevance to the general prognosis of treating the cognitive impairment in patients with schizophrenia. An improved cognitive profile is likely to contribute to educational and occupational opportunities that may also require the simultaneous addition of creative rehabilitative and psychotherapeutic strategies directed toward full reintegration. Moreover, the pessimistic historical approach to the presumed static nature of the cerebral dysfunction of schizophrenia should be revisited with emphasis on the possibility of a more positive prognosis than was envisioned only a few years ago.

Accepted for publication December 1, 1999.

This work was sponsored by Eli Lilly and Company, Indianapolis, Ind.

The authors wish to thank Tracy Purdon, BA, and Gayle Purdon, BSc, for site training and cognitive data reduction; Greg Anglin, PhD, of Eli Lilly Canada Inc, Scarborough, Ontario, for statistical consultation; and the research coordinators and psychology assistants across Canada responsible for data collection.

The Canadian Collaborative Group also included F. Ali, MD; L. Cortese, MD; R. Dickson, MD; M. Filteau, MD; D. Irwin, MD; S. Johnson, MD; L. Kopala, MD; W. Lit, MD; W. McKuen, MD; F. Neuman, MD; L. Oyewumi, MD; G. Remington, MD, PhD; S. Shrikhande, MD; C. Shriqui, MD; P. Silverstone, MD; M. Teehan, MD; P. Vincent, MD; and R. Williams, MD.

Institutional review board or ethics review board approval was obtained prior to initiating recruitment from 19 medical centers across Canada including the Alberta Hospital, Edmonton; Calgary General Hospital, Calgary, Alberta; Clarke Institute of Psychiatry, Toronto, Ontario; Foothills Hospital, Calgary; Grace General Hospital, Winnipeg, Manitoba; Holmwood Health Centre, Guelph, Ontario; Hopital de L'Enfant-Jesus, Quebec City, Quebec; Lion's Gate Hospital, Vancouver, British Columbia; London Psychiatric Hospital, London, Ontario; Centre Hopital Robert-Giffard, Laval, Quebec; Hopital Louis-H. Lafontaine, Montreal, Quebec; McGill University, Montreal; McMaster Cedokee Hospital, Hamilton, Ontario; Victoria Hospital, London; Queen Street Mental Health Centre, Toronto; Queen Elizabeth II Health Sciences Centre, Halifax, Nova Scotia; Royal Ottawa Hospital, Ottawa, Ontario; UBC University Hospital, Vancouver; and the Waterford Hospital, St Johns, Newfoundland.

Reprints: Scot Purdon, PhD, Neuropsychology-9 Bldg, Room 148, Alberta Hospital Edmonton, Box 307, 17480 Fort Rd, Edmonton, Alberta, Canada T5J 2J7 (e-mail: Scot.Purdon@AMHB.AB.CA).