Abstract
There is now substantial evidence that clozapine is an effective treatment for patients with refractory psychosis, at least in the short term [1,2]. Some studies of patients with a high level of hospital use have provided evidence of medium- and long-term clinical improvement (e.g. [3–10], and/or cost-savings [3,9–11] following clozapine treatment. The findings of such studies do not necessarily reflect the clinical effectiveness, or cost-effectiveness, of clozapine in the majority of treatment-refractory patients, who typically reside in the community [12].
Indeed, Rosenheck et al. [13] have suggested that cost savings reported for high hospital-use patients may not generalise to low hospital-use patients and have proposed that further research employing more ‘representative samples’ should be undertaken.
No long-term clozapine studies focusing specifically on patients living in the community have been reported and only two studies have investigated the effect of clozapine on community patients in the medium term [14,15]. Breier et al. [14] reported improvements in clinical symptoms and reduced relapses and hospital bed use among a highly selected sample of 30 outpatients at 12-month follow-up, and we [15] have also reported reduced hospital bed use after 12 months among a cohort of 22 patients, the majority of whom were normally resident in the community prior to commencing clozapine. However, in both studies outcomes were confined to those patients who remained on clozapine and neither group of investigators examined the impact of clozapine on health costs.
Compared with typical antipsychotics, clozapine tablets are expensive and clozapine involves the additional expense of mandatory blood monitoring. Moreover, clozapine is complicated to prescribe due to the requirements of the Clozaril Patient Monitorin SYSTEM (CPMS). These factors combined with a paucity of evidence concerning the long-term costs and effectiveness of clozapine for lower hospital-use patients may contribute to a reluctance to use clozapine in community practice.
This study is an investigation of the long-term (3 years) clinical and financial outcomes of clozapine use among an entire cohort of patients encountered in community practice within the Australian Capital Territory (ACT; regional population about 400 000). Clozapine has been used in the ACT since early in 1993 where its use has been facilitated by a Commonwealth government-funded Clozapine Coordinator. All patients who are prescribed clozapine in the ACT are registered by the Clozapine Coordinator who maintains contact with them and ensures compliance with the requirements of the CPMS. Moreover, information on these patients is readily accessible by virtue of the small number of inpatient facilities and the close working relationships in the mental health system within the relatively closed community of the ACT and the surrounding region.
Few patients with refractory psychosis who are treated in the ACT have ever been long-stay inpatients, the majority spending most of their time living in the community. It was, therefore, considered that the present study would provide evidence as to whether or not the cost and effort involved in prescribing clozapine are justified in community practice.
Method
Subjects
Participants were all those patients who, not having had clozapine elsewhere, commenced clozapine in the ACT before 1 July 1994, who had a recorded and appropriately documented diagnosis of schizophrenia or schizoaffective disorder, and whose psychiatric history was known, particularly for the 2 years before commencing clozapine.
Design
This was an open-label retrospective study of community-based clinical practice. The primary outcome measures were changes in hospital admissions, occupied bed-days and costs of treatment in the 3 years postclozapine (years 1, 2 and 3 postclozapine) compared to the 2 years preclozapine (years 1 and 2 preclozapine); perceived changes in clinical state after 3 years on clozapine as assessed by the treating psychiatrist; and change in living circumstances and employment status after 3 years on clozapine compared to the 6-month period preclozapine. Baseline (preclozapine) and progress Brief Psychiatric Rating Scale (BPRS) scores (1–7 rating) [16] were accessed for those patients to whom this scale had been administered. Preclozapine demographic and clinical characteristics and history, and data on dosage and side effects were also collected. There were no assessments of quality of life.
The Cohort (‘intent to treat’) was divided into two groups (continuers and discontinuers) on the a priori basis of whether or not they remained on clozapine until the end of the study period (study's end) without having a total period off clozapine in excess of 3 months in any 1 year.
Data were obtained from records maintained by the ACT hospital and community health system and health services in nearby areas of New South Wales, from other hospital sources when appropriate, and from health workers and the CPMS. Every possible effort was made to examine all relevant data sources and to document progress for all subjects, including those who discontinued clozapine during the study period.
However, it should be noted that data were not available to enable any estimation of the amount (or costs) of community services utilised by subjects, either before or after the use of clozapine.
Outcome measures
Clinical status
Change in clinical status was rated retrospectively for each patient by their treating psychiatrist. Using a seven-point scale (‘marked’, ‘moderate’, and ‘slight’ deterioration, ‘no change’, and ‘slight’, ‘moderate’ and ‘marked’ improvement), the psychiatrist was asked to provide his/her impression of the change in clinical status at the study's end (3 years after commencing clozapine) compared with the period immediately prior to commencing clozapine.
Bed use
Hospital and hostel bed-days directly or indirectly attributable to a psychiatric illness or its treatment were recorded for each of the 2 years prior to commencement of clozapine and for each of the 3 years following commencement of clozapine. Admissions indirectly due to a psychiatric illness included those due to side effects of treatment or to self-injury. Prior to 1 July 1994 in Australia, all patients were required to be under 24 h supervision (in a hospital or a hostel) when commencing clozapine. An admission to hospital specifically to commence clozapine was attributed to the first year on clozapine as was the time after commencing clozapine if this occurred during an admission for some other reason. Hostel bed-days included all days spent in a fully staffed (24-h coverage) hostel in the ACT or elsewhere.
Costs
Bed costs were the sum of the costs of bed occupancy in hospitals and hostels. The sum of bed and clozapine use costs was estimated for each patient for each year of the study by adding bed costs to the costs of clozapine tablets and blood monitoring and the costs of employing the Clozapine Coordinator. The costs for the Clozapine Coordinator were estimated to be 0.75, 0.50 and 0.33, in years 1, 2 and 3 postclozapine, these figures representing the proportion of time given to the present cohort of participants. Within the cohort the number of pathology tests was used as the basis for attributing individual costs of the Clozapine Coordinator. No attempt was made to cost either other medications or other community services used.
All costings were made at 1996-1997 values (Australian dollars) using figures supplied by the ACT Health Department. The hospital bed-day figure was based on costs for a bed-day in a psychiatry ward and included direct costs (nursing, pathology, imaging, medical, allied health, pharmacy, supplies, oncosts and other costs) and indirect costs (overhead). The hostel bed-day figure was the ratio of the direct total expenses for the ACT psychiatric hostels and the total number of actual patient days.
Employment and accommodation
The living circumstances and employment status in the 6 months prior to commencing clozapine and during the 6 months prior to the study's end were documented from recorded information and/or from information supplied by relevant mental health workers. From this information, each patient was categorised as either ‘employed/studying’ (employed, studying or both) or ‘unemployed/not studying’ (neither employed nor studying). Similarly, each patient was categorised as either ‘institutionalised’ (hospitalised or in a hostel) or ‘not institutionalised’.
Side effects
Side effects after 3 years on clozapine were assessed retrospectively for each patient by the treating psychiatrist using a side-effect checklist developed by the authors. Each psychiatrist was asked to complete the checklist indicating if the side effect was absent, mild, moderate, severe or if they did not know. Side effects on the checklist included: drowsiness/somnolence (daytime); increased sleep (night time); incontinence of urine (night time); dizziness, increased sweating; increased salivation (daytime); increased salivation (night time); constipation; extrapyramidal signs; weight gain; nausea; blurred vision; tachycardia; hypotension; and other (specify). The psychiatrist was also asked to indicate if the patient had experienced epileptic seizures in the periods before or after the commencement of clozapine. Further information concerning epileptic seizures was extracted from hospital records.
Analyses
Statistical comparisons of demographic and clinical characteristics between continuers and discontinuers were based on Fishers exact probability tests and Mann-Whitney and t-tests. A separate variance t-test was used when Levene's test revealed a significant difference in variance for the two groups. Otherwise a pooled variance t-test was used. The effect of clozapine on hospital admissions was analysed using a Cochran Q-test followed by binomial tests. The effect of clozapine on time spent in hospitals and hostels, and treatment costs, was examined using Friedman analyses followed by Wilcoxon tests where the former were statistically significant. Where data for years 1 and 2 preclozapine did not differ significantly from each other, these were averaged (the preclozapine period) before further analyses were undertaken (comparing preclozapine with postclozapine). The possible effect of clozapine on living circumstances, employment and clinical status was analysed using binomial tests. The analyses involving repeated measures were conducted separately for the cohort, continuers and, where appropriate, the discontinuers. Subsidiary comparisons between continuers and discontinuers on pre- and postclozapine outcome measures were conducted using Mann-Whitney, Chi-squared and Fishers exact probability tests.
Results
Characteristics of the sample
The effective sample size, on an intent to treat basis, was 37. Although 42 patients met the requirements of entry to the study, five were excluded from consideration; one (female aged 64) had marked improvement in her mental state but died from lymphoma while still on clozapine, one (who ceased clozapine after only a few weeks) was admitted to a nursing home during the study period because of a pre-existing senile dementia, and three were lost to follow-up after moving interstate (two of whom were still on clozapine at the time of transfer) with their families.
Of the cohort of 37 patients, 25 (68%) were continuers and 12 (32%) were discontinuers. The clinical and demographic characteristics of each of these groups is summarised in Table 1. The only significant difference between the preclozapine demographic and clinical characteristics of the continuers and discontinuers was that the latter had a history of fewer hospitalisations.
All of the continuers remained on clozapine for the entire 3-year follow-up period except in several cases for brief periods of interruption (less than 1 month over 3 years due to non-compliance and, in one case, a pregnancy subsequently terminated).
Three of the 12 discontinuers were taking clozapine at the study's end, of whom two had spent a substantial continuous period off clozapine during the study period (33 months and 21 months) and the third, who was officially off clozapine continuously for 4 months, claimed to have stopped taking the medication 4 months prior to de-registration and presented an ongoing problem of non-compliance. Eight of the other nine discontinuers permanently ceased clozapine use within 1 year of commencing and the remaining patient was off clozapine for a total period of 1 year over the 3-year period of the study. Reasons for temporarily or permanently discontinuing clozapine were lack of therapeutic response and/or noncompliance (n = 6) or intolerable side effects (n = 6, 16.2% of the total sample). Major side effects (sometimes multiple) causing discontinuance were agranulocytosis (n = 1), pancreatitis which reappeared on rechallenge (n = 1), sedation (n = 3), vomiting (n = 2) and dizziness (n = 2).
Mean (SD) of the clinical/demographic characteristics of the sample
Among the continuers, the mean maximum dose of clozapine was 410 mg/day (range = 150–600) and the mean dose at the study's end was 359 mg/day (range = 25–575).
Outcomes: cohort and continuers
Clinical status
At 3 years postclozapine, there was a significant improvement in clinical status (as rated by the treating psychiatrist) for both the cohort (p < 0.0001) and the continuers (p < 0.0001); 86.1% of the cohort (sample n = 36; rating unavailable for one continuer) showed moderate or marked improvement and all of the continuers (n = 24) showed improvement (half moderate and half marked).
Brief Psychiatric Rating Scale
There were too few progress (postclozapine) BPRS scores to permit an analysis of the effect of clozapine on BPRS.
Hospital and hostel use
The percentage of subjects admitted to hospital for at least 1 day in each year of the study (see Table 2) changed significantly over time (p < 0.0001) for both the cohort and the continuers, but there was no significant difference between years 1 and 2 preclozapine for either group. Compared with years 1 and 2 preclozapine, there was a significant reduction in the percentage of subjects admitted to hospital in year 2 and year 3 postclozapine for the continuers (p < 0.01 for all comparisons). For the cohort, compared with years 1 and 2 preclozapine, there was a significant reduction in the percentage of subjects admitted to hospital in year 3 postclozapine (p < 0.05).
There was also a significant change over time in days spent in hospital (see Table 2) for the cohort (p = 0.004) and for the continuers (p = 0.0001). There were no significant differences between years 1 and 2 preclozapine or between the preclozapine period and year 1 postclozapine. There was a reduction in hospital bed-days by the cohort, which was significant in year 2 postclozapine (p = 0.023) and approached significance in year 3 (p = 0.051), compared with the preclozapine period. For the continuers, there was a reduction in hospital bed-days in years 2 and 3 postclozapine compared with the preclozapine period (p = 0.0006 for year 2 and p = 0.002 for year 3). One discontinuer was almost constantly in hospital during the three years which followed his temporary use of clozapine. If data from this extreme outlier are excluded, the mean hospital bed-days for the cohort for year 2 and 1 preclozapine and years 1, 2 and 3 postclozapine become 21.1, 26.8, 18.5, 7.1 and 11.3, respectively.
Bed use for psychiatric reasons by the cohort, continuers and discontinuers: (a) mean (SD) days occupied in each year of the study for hospitals and hostels; and (b) percentage of subjects admitted to hospital in each year
Estimated cost (mean [SD] values; A$1000s) of psychiatric treatment for the cohort, continuers and discontinuers for each year of the study
There were no significant differences in hostel bed use over time for either the cohort or the continuers.
Treatment costs
Estimates of bed costs and the sum of bed and clozapine use costs are presented in Table 3.
Bed costs (combined hospital and hostel) changed significantly over time both for the cohort (p = 0.001) and for the continuers (p = 0.0002), but there was no significant difference in bed costs between years 2 and 1 preclozapine for either group. Compared with the preclozapine period there were significant reductions in bed costs in year 2 postclozapine for the cohort (p = 0.021) and in years 2 and 3 postclozapine for the continuers (p = 0.006 and p = 0.002, respectively). For the continuers, this represented an average saving of $12 632 (or 36%) per patient per year.
However, the combined costs of bed use, clozapine tablets, pathology tests, and Clozapine Coordinator for the preclozapine period were not significantly different from these costs in any of the years postclozapine for either the cohort or the continuers. These costs in years 2 and 3 postclozapine were significantly lower than in year 1 postclozapine for both the cohort (year 2 p = 0.003; year 3 p = 0.02) and the continuers (year 2 p = 0.002; year 3 p = 0.001).
Overall, there was no substantive evidence that clozapine use was associated either with an increase or with a decrease in the treatment costs detailed above.
Living circumstances and employment status
There was no change in the number of patients who were institutionalised (in hospital or in a hostel) while on clozapine but there was a significant increase in the number of subjects who were employed/studying at study's end compared with baseline (cohort 21:12, p = 0.012; continuers 18:10, p = 0.008).
Side effects
Information on side effects (as reported by treating psychiatrists) was available for 24 of the continuers (except data on sweating and night-time salivation which was available for only 23 patients). Side effects present after 3 years on clozapine were weight gain (severe, 1; moderate, 3; mild, 7), increased sleep (moderate, 5; mild, 6), drowsiness (moderate, 2; mild, 7), increased salivation at night (moderate, 1; mild, 6) and in the daytime (mild, 5), sweating (moderate, 1; mild, 5), constipation (mild, 5), dizziness (mild, 3), night-time incontinence (mild, 1), and nausea (mild, 1).
There were no deaths from suicide in the cohort over the period of 3 years after the prescription of clozapine. Four patients had at least one epileptic seizure (two had a prior history of convulsions).
Outcomes: discontinuers
Based on ratings by psychiatrists, one discontinuer showed a moderate deterioration in clinical status 3 years postclozapine, four were unchanged, three showed moderate improvement and four (including the three who were on clozapine at the study's end) showed marked improvement.
In contrast to the findings for the cohort and continuers, there were no significant reductions in the discontinuer's bed use or bed costs in the 3 years after clozapine. In fact, the discontinuers spent significantly longer in hospital in years 1, 2 and 3 postclozapine than in year 2 preclozapine (p < 0.05) and there was no significant difference in bed-days between year 1 preclozapine and years 1, 2 and 3 postclozapine. Compared with years 1 and 2 preclozapine, there was no significant change in the percentage of discontinuers admitted to hospital in years 2 or 3 postclozapine. There was no significant change over time in the number of hostel bed-days used by discontinuers or in the proportion of discontinuers admitted to a hostel. Moreover, bed costs (hospital and hostel combined) for the discontinuing group did not change significantly over time. Finally, employment and accommodation status in the discontinuers did not change significantly over the study period.
Comparisons between discontinuers and continuers
There was no evidence that discontinuers were more impaired than the continuers preclozapine. Discontinuers were hospitalised for less time than the continuers in year 2 preclozapine. There was no significant difference between the continuers and discontinuers in time spent in hospital in year 1 preclozapine, in the percentage of patients admitted to hospital or hostels in years 1 and 2 preclozapine, in the total cost of hostel and hospital admissions, in preclozapine employment/studying status or in accommodation status. However, continuers spent significantly less time than discontinuers in hospital in years 2 and 3 postclozapine (p < 0.05), and they were less likely to have been admitted to hospital than discontinuers in year 2 (p = 0.03) and were more likely to be employed/studying 3 years postclozapine (p = 0.02).
Discussion
Our community sample was similar to other samples in studies of clozapine in terms of age, sex distribution and age of onset of illness. However, as might be expected, with the exception of the Breier et al. [14] and Drew and Hodgson [15] samples, subjects had spent far less time in hospital preclozapine, and had a shorter index period of hospitalisation and (in most cases) a lower preclozapine BPRS score. The maximum daily dose of clozapine in this study (both the mean and the upper end of the range) was relatively modest and the number of patients followed to completion was comparable with most other long-term clozapine studies.
The reported improvement in clinical state following long-term clozapine treatment broadly confirms the previous findings of Breier et al. [14] in their 12-month follow-up study of outpatients. The reduction in hospital admissions and hospital bed-days during the second and third years postclozapine for the continuers, and second year (bed-days) and third year (admissions) for the cohort, is consistent both with the perceived clinical improvements and the increase in employment/studying rate at 3 years postclozapine in the cohort and the continuers. The absence of improvements in living circumstances in patients on clozapine is not surprising given the low level of institutionalisation at baseline: no patient was a long-stay hospital patient, the longest preclozapine index hospitalisation being 17 days, and only a minority of patients were hostel residents.
Although bed use decreased during the later years of clozapine treatment, there was no significant reduction in bed use in the first year of treatment, a finding which contrasts with previous reports of decreased hospital bed use 1 year after initiation of clozapine [14,15]. However, only two of the patients followed up by Breier et al. [14] were reported to have been hospitalised during the first year of clozapine. Thus, either the patients must not have been routinely hospitalised for clozapine commencement or, as was the case in our previous study [15], the initial period of hospitalisation must have been excluded from the postclozapine analysis. When admissions specifically for the purpose of commencing clozapine were excluded from our current results there was a significant decrease in hospital bed-days for the continuers in year 1 postclozapine compared with the preclozapine period.
In contrast to the continuers, the discontinuers did not show a pattern of improvement in outcomes in the 3-year follow-up period. Although the discontinuers are not a suitable control group and their data cannot be used to support any positive claim concerning the use of clozapine, their results do not provide an argument against such claims (as would be the case had they shown an improvement).
The decreased level of hospital bed use led to a significant reduction in residential bed costs of more than $12000 per patient per year (35%) for the continuers. However, these gains were offset by the special costs associated with clozapine use so that for both the cohort and the continuers the sum of bed and clozapine use costs was neither significantly increased nor significantly decreased postclozapine. The fact that hostel bed costs did not include indirect costs is not a serious issue as there was no change in hostel bed use over time.
It should be noted that the estimation of treatment costs excluded costing of most elements of community mental health services. In contrast to most other studies (in which subjects were recruited as long stay inpatients and hence used additional community services when they were discharged from hospital) our subjects were normally located in the community and were high consumers of community services before they commenced clozapine. No attempt was made to assess savings in costs of community services nor to place any value on the benefits of clozapine (such as improved quality of life and employment status) in our estimate of costs associated with the continuing use of clozapine. The absence of such estimations is a serious limitation of the study.
Except for the lack of evidence of cost savings when clozapine and bed costs were combined (but see Essock et al. [6] who also reported no total cost savings), the findings of this study of community-based patients are broadly comparable to the results reported in studies of long-term hospitalised and other high hospital use patients: that is, improvements in clinical status (e.g. [3–8]) and employment [3,9], and reduced hospital bed use and hospital costs [5,9–11], the absence of any deaths from suicide over 3 years, the high compliance rate (68% of the cohort remained on clozapine for the whole 3-year period), and the absence of severe side effects in patients who continued using clozapine for 3 years [17].
In our study, as with others, caution must be exercised in attributing improvement to the use of clozapine.
Our study, like most others, is open to general criticisms (e.g. [18,19]), particularly concerning retrospectivity, use of mirror-image comparisons, the lack of blinding procedures and the absence of a control group. Caution is appropriate concerning the clinical judgements of psychiatrists, which, in the context of an open trial may have been influenced by the views of the first author, who is known as an enthusiast for clozapine. However, the fact that two-thirds of all of the cohort were still on clozapine after 3 years (in spite of the inconvenience this caused to both patients and psychiatrists) suggests that they perceived significant ongoing clinical benefit (independently of the study).
The course of schizophrenia is variable, even in the late stages, and improvements/reduced hospital use could have occurred without any special intervention [5,20].
The use of a double-blind, controlled design in the study of long-term clozapine use is not without its limitations [18,21]. For example, Hargreaves [18] points out that compliance rates might be seriously compromised among the users of traditional antipsychotics who are required to undergo weekly blood samples.
Conclusions
In common with most other studies of the long-term effects of clozapine, there are serious limitations to this study and, hence, conclusions must be treated with caution.
Uniquely, this study provides the first long-term follow-up (albeit only retrospectively) of clozapine intervention in a cohort of community-based patients. It would seem that in this cohort of patients, who had refractory psychosis but who had not experienced extensive hospitalisation, there was a significant rate of moderate to marked clinical improvement sufficient to compensate for any inconvenience posed by the use of clozapine to either the patient or the treating doctor. Further, there was no evidence to support the contention that the use of clozapine in community practice adds significantly to the cost of treatment or to support the use of expense as a barrier against prescribing clozapine in the community.
Thus, although clozapine is an expensive drug which is both complex and time-consuming to use, in community practice the added costs of using clozapine appear to be counter-balanced by savings and it seems likely that the extra effort required to prescribe clozapine will be rewarded with better patient outcome. There is a pressing need for long-term prospective studies (which include measures of change in quality of life) of the use of clozapine, notwithstanding the considerable problems to be encountered in conducting such a study and in the long-term follow-up of patients.
Footnotes
Acknowledgements
This study was funded by the Private Practice Trust Fund, the Canberra Hospital. The authors wish to thank Helen Christensen for her helpful comments on the manuscript.