Does Participation in Psychosocial Treatment Augment the Benefit of Clozapine? FREE

THIS STUDY examines the relationship of concomitant participation in psychosocial treatment to the outcome of pharmacotherapy with clozapine. During the last decade, the efficacy of clozapine has been well established in refractory patients.^1- 7 A recent review⁸ concluded that clozapine reduces positive symptoms (eg, hallucinations, delusions), while its efficacy for negative symptoms (eg, blunted affect, motor retardation) is subject to debate.^9- 12 In a recent 15-site, 12-month randomized trial⁷ of clozapine and haloperidol, we found significant improvement in quality of life (QOL) among patients who continued treatment for a full year. A major advantage of clozapine is that patients experience a far lower incidence of extrapyramidal symptoms (EPS) than with other medications,^7,13 potentially improving treatment compliance.

The ultimate goal of treatment for a severely disabling illness such as schizophrenia is the improvement of social functioning and overall QOL. Findings suggesting a positive effect of clozapine in these areas are thus of particular importance. There are 4 nonmutually exclusive mechanisms through which clozapine may affect QOL and functional outcomes: (1) by reducing psychotic symptoms; (2) by a direct effect on motivation and functioning, independent of its effect on symptoms; (3) by its association with lower EPS, thus improving relative motor functioning; and (4) by facilitating participation in psychosocial rehabilitation activities that in turn may enhance functional capacity. The fourth mechanism is of particular practical and policy importance. There is growing concern that, as health care budgets tighten during the coming years, resources available for psychosocial rehabilitation programs will decrease.¹⁴ If the benefit of novel pharmacotherapies such as clozapine is enhanced by participation in concomitant psychosocial treatment, it will be important that such services be made available.

Several experimental studies have previously found that combined pharmacological and psychosocial treatment in schizophrenia is more effective than pharmacotherapy alone. In a series of related studies,^15- 17 Hogarty and associates reported that (1) rehabilitation therapy is associated with reduced relapse rates among patients treated with chlorpromazine; (2) social therapy enhances the effectiveness of fluphenazine in preventing relapse; and (3) the combination of family treatment, social skills training, and drug therapy is more effective than drug treatment alone. A variety of family treatments has also been found more effective than drug treatment alone.^18- 19 In each of these studies, participation in psychosocial treatment was carefully structured through a standard experimental protocol. No study, however, has examined whether pharmacotherapy is associated with enhanced participation in psychosocial treatment, or whether pharmacological facilitation of psychosocial treatment contributes to long-term pharmacotherapeutic effectiveness, whether with clozapine or other antipsychotic medications.

This study uses data from a 15-site, 12-month randomized clinical trial⁷ of clozapine and haloperidol to examine intermediate factors, especially participation in psychosocial treatment, that may contribute to clozapine's long-term effect on symptoms and QOL. While the original study was a randomized trial of clozapine and haloperidol, participation in psychosocial treatment was based on clinical decision making and patient choice, not on random assignment, and our examination of intermediate factors is thus based on nonexperimental secondary analyses of study data.

Patients with refractory schizophrenia at 15 Veterans Affairs (VA) medical centers were randomly assigned, during a 2-year recruitment period, to clozapine or haloperidol and treated for 12 months. Patients were currently hospitalized with a history of high inpatient service use, defined as at least 30 days' hospitalization for schizophrenia and no more than 364 days' hospitalization during the previous year. Eligibility criteria, described in detail elsewhere,⁷ included (1) a DSM-III-R diagnosis of schizophrenia using the Semi-structured Clinical Interview for Diagnosis²⁰;(2) refractoriness defined as persisting psychotic symptoms despite 2 adequate treatment trials; (3) severe symptoms indicated by specified scores on the Brief Psychiatric Rating Scale²¹ and the Clinical Global Impressions Scale²²; and (4) serious social dysfunction for the previous 2 years. Patients were excluded if they were unable to give informed consent, had a previous trial of clozapine, had a current myeloproliferative disorder, or were pregnant.

Random assignment was made to double-blind treatment with clozapine (100-900 mg/d) or haloperidol (5-30 mg/d). Dose adjustments were made as clinically indicated, using 12 fixed dosage levels. Haloperidol-treated patients also received benztropine mesylate (2-10 mg/d) for EPS and participated in weekly red blood cell counts as required for clozapine treatment. Clozapine-treated patients received placebo benztropine capsules.

To assess the potential effectiveness of clozapine in the context of enriched clinical practice, a predefined set of psychotherapeutic and rehabilitative treatments was offered through a "broker-advocate" case management system implemented uniformly at each site. Services were offered on the basis of clinical need and patient interest, and both treatment recommendations and patient responses were systematically recorded.

As contrasted with intensive models in which case managers provide the majority of the services directly,²³ in the broker-advocate model, case managers establish links between the patient and other service providers and encourage and facilitate their use. The broker-advocate model was used because it reflects the type of service commonly available to schizophrenic patients.

The basic study treatment was a 15-minute weekly meeting with the patient, the case manager, and physician that focused on medication management. Patients were allowed to continue previous psychosocial treatments, but only 20% of the clozapine group and 17% of the haloperidol group (a nonsignificant difference) were involved in such treatments at study entry.

To standardize access to additional services, a comprehensive menu was developed of services to which patients could be readily referred. Every month after entry into the treatment protocol, the case manager met with each study subject for a 3-phase, structured treatment review and planning meeting. First, all health care services received during the previous month were reviewed, along with current housing, financial, and employment arrangements. Second, the case manager and the patient together reviewed the services available on their local resource menu. Third, the case manager recommended participation in appropriate services and offered to facilitate enrollment of the patient into the desired programs. The patient's highest current level of participation and his or her response to the recommendations were documented.

Before the study was implemented, services available on each of the local menus were classified into 1 of 5 categories, ranked in order of increasingly demanding social behavior. The 5 categories were (1) office-based psychological support or psychotherapeutic intervention (eg, individual therapy, group therapy, family therapy); (2) low-expectation, high-frequency supportive services at least 2 times per week (eg, mental health day program, prevocational activities, volunteer work, participation in a psychosocial club or community-based case management); (3) high-expectation supported activity at least 10 hours per week (eg, supported employment, social or vocational rehabilitation programs); (4) mainstream community training program or supervised employment (eg, school, college or high school equivalency program, community vocational training program not specifically designed for the handicapped); and (5) competitive employment (at least 10 hours per week). Each site reviewed its categorizations of each specific service with the investigators (R.R. and J.C.) to assure consistent classification of programs. Interrater reliability between clinical staff and the central investigators on these ratings was moderately high (intraclass correlation=0.73).

Symptom outcomes were assessed with the structured clinical interview for Positive and Negative Syndrome Scale²⁴ for schizophrenia. Quality of life was evaluated with the Quality of Life Scale (QOLS),²⁵ a clinician-rated scale of social functioning, interpersonal relationships, and intrapsychic well-being. Medication adverse effects were assessed with the Barnes scale²⁶ for akathisia, the Abnormal Involuntary Movement Scale²⁷ for tardive dyskinesia, and the Simpson-Angus scale²⁸ for EPS. Because clozapine had its strongest effect on the Simpson-Angus scale,⁷ we selected that measure as representative of EPS effects for the analyses presented here. Interviews were conducted at 6 weeks and 3, 6, 9, and 12 months after random assignment.

During the course of the 12 months of follow-up, some patients stopped taking study medication because of lack of efficacy or adverse effects and switched to other treatments. Altogether, 83 (40%) of 205 patients assigned to clozapine discontinued use by the 48th week of the trial and crossed over to a standard antipsychotic medication. Although 157 (72%) of the 218 patients receiving haloperidol also discontinued blinded treatment, only 49 (22%) received clozapine treatment for 4 or more weeks during the trial. The rest continued to receive conventional antipsychotic medications (including haloperidol).

Since the inclusion of crossover patients in our analyses dilutes our ability to detect the effect of clozapine, especially at the 12-month mark, all the analyses presented here were conducted with crossover cases excluded. Thus, all patients in the clozapine group in this study were treated with clozapine for at least 48 weeks, while patients in the control group received conventional antipsychotic medications for the entire period.

Examination of baseline sociodemographic and clinical status measures showed that crossover patients had significantly lower baseline QOLS scores (P=.01), but did not differ on any other measures. Among patients who did not cross over, there were no significant differences between treatment groups at baseline (Table 1). Thus, although the original study was based on random assignment and no significant differences were noted between treatment groups, the analyses presented here must be regarded as coming from a nonequivalent control group design.

First, analysis of covariance was used to identify differences between groups in symptoms (using the Positive and Negative Syndrome Scale) and QOL at each assessment point.

Second, χ² tests were used to compare (1) the percentage of patients participating in psychosocial rehabilitation treatments each month (level 2 or higher), (2) the percentage of patients receiving new psychosocial treatment recommendations, (3) the patients' recommended behavioral levels, and (4) the percentage of patients verbally accepting any recommendations.

Third, we used structural equation modeling to examine sequential relationships, at each of 3 points, between sets of variables, ie, (1) baseline clinical status (symptoms, QOL, and EPS), participation in psychosocial treatment, and assigned treatment condition (clozapine vs haloperidol); (2) intermediate outcomes of these variables (eg, at 3 or 6 months); and (3) the final outcomes (eg, at 6 or 12 months).

Structural equation modeling is an extension of multiple regression analysis²⁹ in which each variable is analyzed as having both direct effects on the final outcome and indirect effects through its influence on temporally intermediate variables. Two separate sets of causal models were tested, one addressing factors mediating early outcomes (6 months) and the other addressing final outcomes (12 months). These analyses were conducted using the CALIS procedure in SAS software.³⁰ An α of P=.05 was used as the criterion for statistical significance.

Comparison of sociodemographic and clinical characteristics of patients assigned to clozapine (n=122) and haloperidol (n=169) showed no significant differences (Table 1).

Controlling for baseline scores, patients assigned to clozapine had significantly lower symptom scores than control patients by the 6-week assessment (mean Positive and Negative Syndrome Scale score, 79.6 for clozapine group vs 85.8 for haloperidol group; F₁₂₇₀=11.8; P=.007) and at all assessments thereafter. Differences in QOLS scores were not statistically significant at 6 weeks or 3 months, but were significant at 6 months (mean QOLS score, 46.6 for clozapine group vs 42.7 for haloperidol group; F₁₂₇₀=6.25; P=.01) and all assessments thereafter.

Figure 1 shows that by the third month of treatment, patients receiving clozapine were significantly more likely than those assigned to haloperidol to participate in psychosocial rehabilitation treatments involving at least low-expectation supportive services (level 2 in the classification described above), and these differences remained significant at each level for the rest of the trial. At the sixth month, for example, clozapine-treated patients were more likely to participate in low-expectation, high-frequency supportive services (33% vs 24%), high-expectation supportive services (14% vs 7%), and mainstream community training programs or competitive employment (3% vs 0%).

There were no significant differences at any point in either the proportion of patients who received a clinical recommendation from their case manager or the treatment level at which recommendations were made. Thus, although patients receiving clozapine participated at higher levels of treatment, there were no differences in their case managers' treatment recommendations, and it does not appear their case managers had higher expectations of them. Verbal acceptance of referrals during treatment planning sessions was generally low and decreased from 21% during the first month of the trial to less than 10% by the 10th month. During months 6 through 8, however, acceptances increased in both groups, but patients receiving clozapine were significantly more likely than controls to consider and accept treatment recommendations (47%-48% accepted recommendations in the clozapine group vs 26%-28% among controls) (6-month assessment: χ²₂=9.3, P=.009; 7-month assessment: χ²₂=6.8, P=.03; 9-month assessment: χ²₂=6.3, P=.04).

Four structural equation models of the effects of clozapine are presented in Figure 2, Figure 3, Figure 4, and Figure 5. Goodness-of-fit indexes for these models were all greater than 0.90, with parsimony indexes ranging from 0.36 to 0.39.

Figure 2 shows that after 3 months of treatment, the clozapine group had lower symptom levels and fewer adverse effects and showed greater participation in psychosocial treatment. Both 3-month symptom level and 3-month QOLS scores were statistically associated with symptom scores at 6 months (symptoms positively and QOL negatively). There was no direct correlation from clozapine treatment or from participation in psychosocial treatment at 3 months to symptom scores at 6 months. Thus, it appears that the association of clozapine treatment with reduced symptom scores at 6 months is mediated through its effect on symptom scores at 3 months.

The next model (Figure 3) shows significant statistical associations between 3-month QOLS and symptom scores and 6-month QOLS scores and shows also a nonsignificant trend linking psychosocial treatment at 3 months with improved QOLS scores (P<.10).

In the 12-month models (Figure 4 and Figure 5), we were able to evaluate the association of clinical changes at 3 months with participation in psychosocial treatment at 6 months and the association of psychosocial treatment at 6 months with outcomes at 12 months. Figure 4 shows that assignment to the clozapine group, QOLS scores at 3 months, and reduced EPS at 3 months are all associated with greater participation in psychosocial treatment at 6 months. Symptom scores at 1 year are associated with symptoms at 3 months (a positive effect), QOLS scores at 3 months (a negative effect), and participation in psychosocial treatment at 6 months (a negative effect).

The final model (Figure 5) shows that QOLS scores at 1 year are associated with lower symptom scores at 3 months, higher QOLS scores at 3 months, and, once again, with participation in psychosocial treatment at 6 months. The model thus suggests that the association of clozapine treatment with improved QOLS scores at 1 year, like its association with symptom score, is mediated through 3 pathways: reduced symptoms, participation in psychosocial treatment, and reduced EPS, which is also associated with enhanced participation in psychosocial treatment. In each of our 12-month models, the benefits of clozapine are at least partly mediated by its association with participation in psychosocial treatment.

This study illustrates a statistical approach to exploring the processes through which antipsychotic medications affect symptoms and QOL in patients with schizophrenia. More specifically, it identifies the relationship of enhanced participation in psychosocial treatment to the effectiveness of clozapine. The analyses presented here confirm well-established findings that clozapine has a beneficial effect on symptoms of schizophrenia and on EPS. The analyses also show that these effects emerge rapidly—within the first 6 weeks of treatment. In addition, these data indicate that, when participation in psychosocial treatment is allowed to emerge naturally from clinician-patient interaction (rather than being standardized through the research protocol), clozapine treatment is robustly associated with such participation by the third month of treatment. Participation in psychosocial treatment also appears to independently augment the pharmacotherapeutic benefit of clozapine at 12 months.

It is notable that although enhanced participation in psychosocial treatment emerges relatively quickly (by the third month of treatment), 6-month symptom improvements attributable to clozapine do not appear to be mediated by enhanced participation in such treatment. In contrast to the rapid symptom response, improvement in QOL relative to standard medication does not emerge until the sixth month of treatment, more than 4 months after the emergence of differential symptom improvement and 3 months after the appearance of significant differences in participation in psychosocial treatment. The association of clozapine and improved QOL at 6 months is mediated primarily through symptom reduction, although an effect attributable to psychosocial intervention begins to appear at 6 months as a marginally significant trend.

Most likely as a result of more extensive exposure to psychosocial treatment, both symptom and QOL gains associated with clozapine at 12 months are associated with participation in psychosocial treatment. A third pathway also emerges in the 12-month models, indicating that clozapine-related reductions in EPS are associated with participation in psychosocial treatment (presumably because of reduced motor restriction), further contributing to both symptom and QOL gains at 12 months.

These findings suggest that psychosocial treatment should be made available to maximize the long-term benefits of clozapine. The exact nature and optimal intensity of such psychosocial treatment is not clear from our data, but available services should probably include both day treatment and opportunities for vocational or prevocational rehabilitation.

Four potential methodological limitations of this study must be considered. First, it might be argued that the greater participation in psychosocial treatment associated with clozapine treatment might have resulted from prior improvements in social functioning and QOL, rather than vice versa, and thus are secondary to earlier clozapine effects. This seems unlikely since clozapine's association with greater participation in psychosocial treatment emerged 3 months before evidence of its association with improved QOL.

Second, although the study from which the data were obtained was based on random assignment, the analyses presented included clozapine-treated patients only if they continued their assigned treatment for a full year, but allowed haloperidol-treated patients to be included if they did not take clozapine. The clozapine group might thus have been more motivated for change than controls, and differences in participation in psychosocial treatment could reflect selection bias rather than pharmacological effect. The positive relationship between clozapine and participation in psychosocial treatment, however, was also statistically significant in the original, unbiased intention-to-treat sample (with no crossovers excluded). We conducted our analyses on the present sample because the 1-year outcome effects for clozapine were considerably attenuated by crossovers and our objective was to consider whether participation in psychosocial treatment plays a mediating role in the relationship between clozapine and outcomes. The lack of any significant baseline differences between the groups in the analyzed sample also suggests limited risk of selection bias.

Third, our measure of participation in psychosocial treatment has not been validated by on-site observation of the available services, and we lack evidence that either the classification of services or the actual provision of services was comparable across sites. One would expect, however, that imprecision in measurement would add variability and uncertainty to the analysis, thus reducing the chance of finding significant relationships between participation in psychosocial treatment and outcomes. The statistical significance of our findings, in fact, lends credibility to our measure.

Finally, there is some risk of rater bias in the assessment of patients' participation in psychosocial treatment, since the case managers who made the participation ratings were aware of the patients' functional status and, in unblinded cases, of their medication assignment. The ratings of psychosocial participation, however, were not based on qualitative judgment of the patient's participation, but rather by the patient's enrollment and attendance at the specified programs. Furthermore, because no differences appeared between the groups in the clinicians' clinical recommendations (but, rather, only in the patients' responses), bias in the treatment planning process seems unlikely.

There has been considerable emphasis in recent years³¹ on opening the "black box" of treatment to document the implementation of complex interventions and better understand their mechanisms of action. A complex disabling disorder such as schizophrenia affects numerous domains of life, and specific interventions may be needed for each of these domains. Although the impact of pharmacological interventions is typically conceptualized in biochemical terms, the conceptual approach and empirical analyses presented here show that medications can have a complex sequential impact on a variety of behavioral processes that can, eventually, operate independently of the original pharmacological effect. We believe that these processes can be differentiated by examining their correlation over time using techniques like structural equation modeling.

The issues addressed in this study also illustrate the complementary role of efficacy and effectiveness research on medical interventions.^32- 33 While efficacy studies assess the specific impact of treatment with highly controlled conditions designed to reduce the influence of extraneous factors as much as possible, effectiveness studies assess the impact of a treatment in "real world" conditions, in which issues of compliance and choice are accepted as part of clinical reality. There has been a major emphasis in recent years³⁴ on standardizing the psychosocial treatments in clinical trials. Such standardization is essential in efficacy studies in which the central research question concerns specific pharmacological impact. However, as shown here, psychopharmacological agents may also be of help to patients by facilitating use of other treatments, adding to their net therapeutic effect. In most discussions of differences between efficacy and effectiveness trials, it is assumed that the carefully controlled conditions with which efficacy trials are conducted increase the likelihood of discovering positive effects and that effectiveness studies are needed to more realistically evaluate medications in more demanding real world conditions.³⁵ This study suggests, however, that by standardizing psychosocial interventions secondary benefits of some agents may be obscured and that effectiveness studies may be needed, in addition, to identify their full therapeutic potential.

Accepted for publication March 20, 1998.

Supported by the Department of Veterans Affairs Health Services Research and Development Service. Clozapine and matching placebo were generously provided by Sandoz, a division of Novartis Pharmaceutical Corporation.

We thank Lois Ucas, Jennifer Cahill, and Dennis Thompson of the Chairman's Office. We are indebted to the Data Monitoring Board (Alan Breier, MD; Howard Goldman, MD; James Klett, PhD; and David Pickar, MD) and the Executive Committee (Boris Astrachan, MD; John Crayton, MD; Linda Frisman, PhD; Carol Fye, RPh, MS; William Hargreaves, PhD; and William Lawson, MD) for their careful overview of the progress of the trial. We are also indebted to the Research Service staff in Veterans Affairs Headquarters, Washington, DC (John Feussner, MD; Shirley Meehan, MBA; Charles Welch, PhD; Ping Huang, PhD; and Joseph Gough, MA).

Members of the Veterans Affairs Cooperative Study Group on Clozapine in Refractory Schizophrenia: John Grabowski, MD, Detroit, Mich; Denise Evans, MD, Augusta, Ga; Lawrence Herz, MD, Bedford, Mass; George Jurjus, MD, Brecksville, Ohio; Sidney Chang, MD, Brockton, Mass; Lawrence Dunn, MD, Durham, NC; John C. Crayton, MD, Hines, Ill; William B. Lawson, MD, PhD, Little Rock, Ark; Yeon Choe, MD, Lyons, NJ; Richard Douyon, MD, Miami, Fla; Edward Allen, MD, Montrose, NY; John Lauriello, MD, Palo Alto, Calif; Michael Peszke, MD, Perry Point, Md; Jeffrey L. Peters, MD, Pittsburgh, Pa; Janet Tekell, MD, San Antonio, Tex; and Joseph Erdos, MD, PhD, West Haven, Conn.

Reprints: Robert Rosenheck, MD (182), Veterans Affairs Connecticut Health Care System, 950 Campbell Ave, West Haven, CT 06516-2770 (e-mail: Robert.Rosenheck@Yale.edu).