|Modelling the impact of compliance on the costs and effects of long-acting risperidone in Canada
|Chue P S, Heeg B M, Buskens E, van Hout B A
This is a critical abstract of an economic evaluation that meets the criteria for inclusion on NHS EED. Each abstract contains a brief summary of the methods, the results and conclusions followed by a detailed critical assessment on the reliability of the study and the conclusions drawn.
The study compared three treatment options for schizophrenia. The three treatment options differed in the starting treatment:
haloperidol depot, a long-acting conventional antipsychotic agent, at a daily dose of 4.76 mg;
long-acting risperidone, a long-acting atypical antipsychotic agent, at a daily dose of 2.3 mg; or
oral risperidone, an oral atypical antipsychotic agent, at a daily dose of 4 mg.
In all three scenarios, the second and third stage of treatment consisted of olanzapine (15 mg/day) and clozapine (384 mg/day), respectively.
Economic study type
The hypothetical cohort comprised 3,000 patients with schizophrenia at high risk of non-compliance. Only patients who had suffered relapses at home, had fully recovered from multiple episodes (i.e. two or more relapses), and had had no or only minor impairment between the episodes, were included in the study. Patients who had partly recovered and experienced (increasing) impairment with each of several episodes, but did not return to normal between multiple episodes, were also included in the study. The male-to-female ratio was 1, and the average age was 34.4 year
As this was a modelling study, the setting was not explicitly reported at the outset. The economic study was carried out in Canada.
Dates to which data relate
The specific studies that provided effectiveness evidence were not referenced, making it impossible to know the exact date range of these studies. In addition, the date of the report composed by a panel of local experts was also not reported. The cost data were derived from official sources published between 2003 and 2004. All costs were reported for the price year 2003.
Source of effectiveness data
The effectiveness data were derived from a synthesis of studies, augmented by estimates based on experts' opinion where data were not available.
A discrete event model was constructed using Microsoft Excel (Microsoft Corp., Redmond) and @-Risk 4.5 (Palisade Corp. New York) to represent the progression of the disease and to estimate the costs and effects of the three treatment options. The relationship between the individual patient's history and the expected health states and costs of treatment were also modelled. The time horizon of the model was 5 years. Further details of the model were given in another study (Heeg et al. 2005, see 'Other Publications of Related Interest' below for bibliographic details). The model was based on the assumptions that patients were either compliant or not compliant, and that there was no differentiation between changed different degrees or modules of non-compliance.
Outcomes assessed in the review
The authors did not specify whether a systematic or an ad hoc review of the literature was undertaken to identify the model parameters. The model took the following fixed patient characteristics from specified distributions:
age of the patients (when entering the model and death; a continuous variable);
gender, patient profile (partial or complete recovery between episodes) and potential risk (if the patient is a risk to himself or to society), (all binomial variables);
severity of the disease (non, medium or very severe patient profile; a categorical variable); and
side effects (i.e. whether the patient will experience side effects from a specific antipsychotic agent; transition probabilities).
The following time-dependent variables based on patients' history were included in the model:
duration of relapse;
time between relapses;
Positive and Negative Syndrome Scale (PANSS) score;
appointments with a psychiatrist;
Disorganization Index (DI; index of a patient's ability to take care of him or herself);
actual risk of self-harm or harming others; and
The primary health outcome measures were the average cumulative number of relapses, the average PANSS score during and between relapses, and the cumulative time spent in psychosis. The authors reported that a discount rate of 5% was used for outcomes.
Study designs and other criteria for inclusion in the review
The authors only reviewed randomised controlled trials. No further inclusion criteria were reported.
Sources searched to identify primary studies
Criteria used to ensure the validity of primary studies
Methods used to judge relevance and validity, and for extracting data
Number of primary studies included
The actual number of primary studies that provided effectiveness evidence was not explicit.
Methods of combining primary studies
Investigation of differences between primary studies
The authors do not seem to have investigated differences between the primary studies.
Results of the review
The results of all input parameters and time-dependent variables were not reported in the current study.
The mean cumulative number of relapses was 4.54 with the long-acting risperidone option, 5.08 with the oral risperidone option, and 4.82 with haloperidol depot option.
Time in psychosis was 3.00 years with the long-acting risperidone option, 3.36 years with the oral risperidone option, and 3.18 years with the haloperidol depot option.
The mean PANSS score during relapses was 137 with the long-acting risperidone option, 149 with the oral risperidone option and 147 with the haloperidol depot option.
The mean PANSS score between relapses was 82 with the long-acting risperidone option and 91 with the oral risperidone and haloperidol depot options.
Methods used to derive estimates of effectiveness
Some of the effectiveness estimates were based on expert opinion. A panel of local experts on psychiatric diseases was formed in order to comment on a model that had been developed for the UK. The input parameters of the model were adjusted according to the panel's comments.
Estimates of effectiveness and key assumptions
The duration of a relapse was assumed to be independent of treatment.
Measure of benefits used in the economic analysis
The authors did not derive a summary measure of health benefit. The analysis was therefore categorised as a cost-consequences study.
Adopting the health system perspective, the direct costs included in the analysis were for the depot injection, treatment in different locations (e.g. home, sheltered living, day care, hospital, long-term hospitalisation) and medications. The costs and the quantities were not analysed separately. The costs were derived from official published sources, while the quantities of resources used were derived from the model. All costs were reported for the year 2003 and, as the time horizon of the model was 5 years, discounting was appropriately conducted.
Statistical analysis of costs
The costs were treated deterministically.
The indirect costs were not included in the analysis.
Canadian dollars (CAD). The conversion rate to US dollars ($) was CAD 1 = $0.71.
A one-way sensitivity analysis was performed to estimate variability in the data. The input parameters tested were:
the level of hospital care (six different levels of care according to increasing nursing attendance and daily costs varying from CAD 192 at level 1 to CAD 1,248 at level 6);
the treatment regimen;
the choice of conventional depot;
the patient compliance rates;
the effect of atypical agents on the PANSS score;
the definition of DI;
the effect of sex on risk potential;
the PANSS threshold to present a risk to society;
the calculation of the PANSS score;
the probability of side effects of each treatment;
the effect of non-compliance on time between relapses
the difference between atypical and conventional treatment in terms of the duration of relapse; and
the discount rate.
The method used to select the ranges was not explicitly reported. A sub-group analysis was also performed to test the robustness of the results. Patients at high risk of relapse were either assumed to fully recover ("total recovery patients") or get worse ("partial recovery patients"). The results were tested when the analysis was limited to either sub-group.
Estimated benefits used in the economic analysis
See the 'Outcomes Assessed in the Review' section.
The total costs were reported per patient. At year 5, the long-acting risperidone treatment resulted in a total cost per patient of CAD 171,181, the oral risperidone in CAD 187,226 and haloperidol depot in CAD 179,839.
It was reported that, over the 5-year period, starting with long-acting risperidone resulted in cost-savings of CAD 8,657 per patient (CAD 6,908 discounted) in comparison with haloperidol depot and CAD 16,044 per patient (CAD 13,130 discounted) in comparison with oral risperidone.
It should be noted that the cost of side effects were not considered in the economic analysis.
Synthesis of costs and benefits
The costs and benefits were not combined.
When the sub-group analysis was limited to total recovery patients, the long-acting risperidone option resulted in additional costs in comparison with oral risperidone (-CAD 7,104) or haloperidol depot (-CAD 10,184). On the other hand, when the analysis was limited to patients with partial recovery, the long-acting risperidone option resulted in additional savings in comparison with oral risperidone and haloperidone depot (savings of CAD 21,185 and CAD 15,353, respectively).
The one-way sensitivity analysis demonstrated that when the additional efficacy of atypical agents for reducing the PANSS score (0.83) compared with conventional formulations (0.89) was decreased from 6 to 0%, long-acting risperidone remained more effective but did not result in cost-savings. For symptoms and relapses, the most influential parameter was the compliance rate of the patients. For example, when the base-case difference in compliance between oral atypical agents and long-acting risperidone was decreased by 50%, long-acting risperidone did not result in cost-savings.
"Long-acting risperidone was the dominant strategy, being more effective and less costly than oral risperidone or haloperidol depot."
CRD COMMENTARY - Selection of comparators
The choice of the comparators was explicitly justified. Starting with an oral atypical antipsychotic agent (e.g. oral risperidone) as first-line treatment would seem to represent current practice in the authors' setting. You should decide if these represent a commonly used health technology in your own setting.
Validity of estimate of measure of effectiveness
A systematic review of the literature was not undertaken. Although this is common practice with models, it does not always ensure that the best data available are used in the model. The sources searched for primary studies were not reported and it is impossible to identify the primary studies or references that provided each effectiveness estimate. This may limit the validity of the conduct of the study. The authors appear to have used data from the available studies selectively and they do not seem to have considered the impact of differences between the studies identified.
In the model, some effectiveness estimates were based on expert opinion. A panel of local experts on psychiatric diseases was used, but the authors did not report any criteria according to which the members of the panel were chosen. The authors carried out several sensitivity analyses in relation to the efficacy estimates. These analyses improved both the internal validity and the generalisability of the study by demonstrating the robustness of the results to changes in the base-case estimates. However, the author did not provide details of the ranges over which the parameters were varied in the sensitivity analysis.
Validity of estimate of measure of benefit
The costs and benefits were not combined. In effect, a cost-consequences analysis was performed.
Validity of estimate of costs
The analysis of the costs was performed from the perspective of the Canadian health system. As such, it appears that all the relevant categories of costs have been included in the analysis. The costs and the quantities were not reported separately. In addition, the use of summary costs made it impossible to know which aspects of the costs were included, for example, whether overhead costs were included. Thus, it is difficult to rework the analysis for other settings. The quantities of resources used were obtained from the model, while all costs were obtained from official published sources. The robustness of the cost estimates was not assessed in a sensitivity analysis. The price year was reported and currency conversions and discounting were appropriately conducted. These will aid any future reflation exercises.
The authors compared their findings of clinical effectiveness with those from other studies and found them to generally be in agreement. The issue of generalisability was not directly addressed. The authors do not appear to have presented their results selectively, but statistical tests were not performed to test the statistical significance of the results. The study enrolled patients with schizophrenia at a high risk of non-compliance and this was reflected in the authors' conclusions.
The authors reported a number of limitations to their study. First, the model included only direct medical costs, and not the costs of treating side effects or indirect costs (e.g. job loss). It is possible that the omission of these costs might have affected the authors' conclusions. Second, following the experts' recommendations, the duration of psychosis was not included as a parameter affecting a patient's treatment. Third, quality-adjusted life-years (QALYs) were not included in the model, owing to a lack of available data that connect PANSS score with QALYs. Another limitation of the model was the assumption that patients can either be compliant or non-compliant, and that differing degrees of non-compliance (most usually found among patients) were not investigated, which might have biased the results. The authors admitted that each treatment option was limited to only three medications and changes in dosage or combination therapies, or factors such as polypharmacy, were not investigated. Finally, the authors acknowledged that their definitions for "presenting a risk to oneself or society" and "disorganization" were firm and robust.
Implications of the study
The authors suggested that, according to study findings, government formularies should timely adopt long-acting risperidone as a treatment option and promote early treatment initiation. Nevertheless, they acknowledged several limitations to their study and suggested areas for future research. In particular, research should focus on identifying the most suitable treatment options for different patient sub-groups, and on estimating patients' utility in different health states and developing a measure that associates PANSS scores with QALYs.
Source of funding
Supported by Janssen Pharmaceutica N.V., Belgium, and Janssen-Ortho, Canada.
Chue P S, Heeg B M, Buskens E, van Hout B A. Modelling the impact of compliance on the costs and effects of long-acting risperidone in Canada. PharmacoEconomics 2005; 23(Supplement 1): 62-74
Other publications of related interest
Heeg BM, Buskens E, Knapp M, et al. Modelling the treated course of schizophrenia: development of a discrete event simulation model. Pharmacoeconomics 2005;23 Suppl 1:17-33.
Subject indexing assigned by NLM
Administration, Oral; Antipsychotic Agents /administration & Canada; Cost of Illness; Cost-Benefit Analysis; Delayed-Action Preparations; Economics, Pharmaceutical; Haloperidol /administration & Hospitalization /economics; Humans; Models, Economic; Monte Carlo Method; Patient Compliance; Risperidone /administration & Schizophrenia /drug therapy /economics; dosage /economics /therapeutic use; dosage /economics /therapeutic use; dosage /economics /therapeutic use
Date bibliographic record published
Date abstract record published