Pharmacoeconomic analysis of paliperidone palmitate for treating schizophrenia in Greece

Unlike a clinical trial, patients are not recruited into this research. Rather, a decision model was used to represent the average patient being treated using standard approaches.

Therefore, it is necessary to define the population to whom results would apply. The population of interest consisted of patients having chronic schizophrenia with multiple relapses, frequent hospitalizations, and problems with adherence to prescribed medications. At initiation of the analysis, all patients were stable and treated as outpatients with maintenance doses of their LAIs. For the purposes of this analysis, comorbidities were not considered even though they are common in this population [21].

The primary drug of interest was PP-LAI, which was compared against RIS-LAI. Long term use of PP-LAI has been investigated in a number of randomized controlled trials [13‐17]; one of these trials involved a comparison with RIS-LAI [15]. As previously mentioned, the European Medicines Agency has approved PP-LAI for monthly dosing [12], while RIS-LAI is administered every two weeks [10]. It should be noted that at the time of this analysis PP-LAI was not marketed in Greece and there was no local clinical experience.

Data were modelled for one year using a previously validated decision tree [22], which appears in Figure 1, adapted for use in Greece. An expert panel was recruited and interviewed to provide clinical input describing patterns of patient management in this country. To enter the model, an average patient with chronic relapsing schizophrenia must be an outpatient with stable disease treated with either PP-LAI or RIS-LAI. The patient can be either adherent or non-adherent, according to published rates and expert opinion. Patients can remain stable or can relapse, with treatment either in the emergency room or in hospital for severe cases. Those who cannot tolerate the primary drug or refuse to take it are switched to olanzapine oral tablets. In the event of a subsequent failure on that drug, clozapine oral tablets are prescribed [23, 24].

Given the challenge of collecting valid local data for populating the model, which has also been identified in the relevant literature [25], data on resource utilisation, frequency and duration of relapses were mainly extracted from an expert panel consisting of hospital psychiatrists. Other clinical rates and associated data inputs were determined from the literature (Table 1) [10, 12‐16, 23, 24, 26‐40]. The doses of drugs actually administered in long term trials of these drugs were used, rather than the Daily Defined Doses (DDDs) as published by the World Health Organization. DDDs represent the average dose for the drug when used in its most common indication, which does not represent our target population [41]. In the present analysis, the research hypothesis is focused on frequently relapsing patients, necessitating hospitalization and intensive intervention. Therefore, DDDs may underestimate the doses used in the real world when treating these patients, whereas doses from the clinical trials could be considered as better proxy for real world dosing.

Costs were considered from the perspective of the National Health Service of Greece. We included only direct costs of care while indirect costs, such as time lost from work, were excluded (Table 2) [20, 42, 43]. We did not apply discounting because the analysis had a time horizon of one year. Prices were taken from official bulletins or from the literature, then inflated to 2011 Euros using the Consumer Price Index for Greece [44].

For each drug, we calculated the average cost per patient treated. Patient outcomes analyzed included average days with stable disease, numbers of hospitalizations, emergency room visits, and quality-adjusted life-years (QALYs). To derive utilities (i.e., the quality weights) for this quality adjustment, preference based estimates were obtained from the literature [45‐49]. Each of the three primary health states (i.e., stable disease, exacerbation requiring emergency room treatment, and hospitalization) were weighted using the average of the reported utility scores. QALYs were then estimated for each drug by multiplying the amount of time in each health state by the quality value assigned to that health state. Therefore, a cost-utility analysis was conducted, which involves calculating the incremental treatment cost per QALY as the pharmacoeconomic outcome. In addition, a cost-effectiveness analysis was performed in order to assess other important clinical outcomes, such as the number of stable and relapse days as well as rates of hospitalisation and emergency room visits.

Several sensitivity analyses were performed to determine if alterations in clinical or cost inputs would influence outputs. One-way sensitivity analyses were done to identify break-even points, that is, what the values would have to be in order for PP-LAI to cost more than RIS-LAI. We conducted one-way sensitivity analyses on important parameters such as rates of adherence, hospitalization, and emergency room visits as well as drug acquisition costs. Finally, all variables were varied over plausible ranges in a probabilistic sensitivity analysis (also called a Monte Carlo simulation) with 10,000 iterations. That analysis reproduces results for a large group of patients and gives a projection of what average costs and outcomes would be.

The total direct cost to treat one patient over the year was calculated for each drug, as presented in Table 3. Included in those calculations were drugs, medical care (visits) and hospital care, based on the units presented in Table 2. The overall cost to treat patients with PP-LAI was lower than with RIS-LAI by €166, despite having a higher acquisition cost. In the case of PP-LAI, drugs accounted for the largest proportion of the total costs (61%), while hospitalization comprised 30% and medical care the remaining 9%. Costs for RIS-LAI had a similar pattern of distribution, with 56% due to drugs, 33% hospital care and 11% medical care.

The primary pharmacoeconomic analysis was cost-utility, which simultaneously compares costs and QALYs. Along with costs, Table 3 also lists the numbers of QALYs associated with the use of each drug. In the base case, patients treated with PP-LAI have a higher QALY score, meaning that they experience more time with a higher quality of life. Because PP-LAI has a lower cost and a greater number of QALYs, it is considered dominant over RIS-LAI. That means it is the preferred treatment and should be adopted, providing it is affordable.

Table 4 displays the other clinical outcomes of interest, which were numbers of stable and relapse days as well as rates of hospitalization and emergency room visits. Patients receiving PP-LAI experienced more days with stable disease and fewer days in relapse. Fewer of them visited the emergency room or were hospitalized. Since PP-LAI has a lower cost and all outcomes were superior, it dominated RIS-LAI in all of these scenarios. These observations confirm that PP-LAI is the preferred strategy to RIS-LAI.

To test the robustness of the model and its results, an array of different sensitivity analyses were conducted. In one-way sensitivity analyses, the model was insensitive to variations in rates of hospitalization or emergency room visits. That means results would not change and favour RIS-LAI regardless of how many patients were hospitalized or treated in the emergency room. As well, for cost equality between drugs, the adherence rate for PP-LAI would have to decrease by 28.5% (while that of RIS-LAI remained the same), which is not a reasonable scenario.

In the Monte Carlo synthesis, 74.6% of the 10,000 simulations favoured PP-LAI. Figure 2 depicts the scatterplot of costs versus QALYs. All points below the horizontal line labelled €0.0 represent lower costs for PP-LAI and points to the right of the vertical line represent greater quality of life for patients treated with PP-LAI, as opposed to RIS-LAI.

While reviewing these results it is important to keep in mind the potential limitations of this analysis. An apparent limitation was the fact that local expert panel was used for estimating specific input parameters to the model, namely, those associated with resource utilization in the Greek setting. Although this approach has been followed before in other similar studies [19, 20], it could lead to potentially biased estimates. However, in the light of the absence of real life resource utilization data the expert panel could give a picture of the actual clinical setting. Furthermore, in our model we did not include the costs for treating adverse events. One reason was that the drugs are closely related, with paliperidone being a metabolite of risperidone [12]. Therefore, one might expect the efficacy and safety profiles to be quite similar. In fact, that assumption of equal side effect rates was made in a recently published pharmacoeconomic analysis in the USA that included both of these drugs [18]. Geitona and associates [20] did include some of these events, but found that the associated cost was trivial and had no impact on the model. Similar results have been reported by Vera-Llonch and coworkers [51], who estimated the monthly cost associated with side effect management for risperidone and olanzapine. It should be noted that our model captured the discontinuation and switching rates, which are also attributed to adverse events.

Efficient resource allocation has become a priority for policy makers across Europe and in Greece in particular. That applies to healthcare as well and to the management of patients with chronic schizophrenia. Health economic studies could provide significant tools for well documented and rational decision making given the scarcity of resources and the increasing control on health care and pharmaceutical expenditure. Pharmacoeconomic analyses, like the one presented, give quantitative estimates of the costs of care and identify the preferred choices for drug treatment. In the case of PP-LAI, its adoption would actually lead to savings for the system, since the overall cost of care would decrease. Both clinicians and managers within the system need to become aware of these analyses and use them to the advantage of both themselves and patients.