Cost-effectiveness of 3-month paliperidone therapy for chronic schizophrenia in the Netherlands

Abstract

Background: A new depot formulation of paliperidone has been developed that provides effective treatment for schizophrenia for 3 months (PP3M). It has been tested in phase-3 trials, but no data on its cost-effectiveness have been published.

Purpose: To determine the cost-effectiveness of PP3M compared with once-monthly paliperidone (PP1M), haloperidol long-acting therapy (HAL-LAT), risperidone microspheres (RIS-LAT), and oral olanzapine (oral-OLZ) for treating chronic schizophrenia in The Netherlands.

Methods: A previous 1-year decision tree was adapted, based on local inputs supplemented with data from published literature. The primary analysis used DRG costs in 2016 euros from the insurer perspective, as derived from official lists. A micro-costing analysis was also conducted. For the costing scenario, official list prices were used. Clinical outcomes included relapses (treated as outpatients, requiring hospitalization, total), and quality-adjusted life-years (QALYs). Rates and utility scores were derived from the literature. Economic outcomes were the incremental cost/QALY-gained or relapse-avoided. Model robustness was examined in scenario, 1-way, and probability sensitivity analyses.

Results: The expected cost was lowest with PP3M (8,781€), followed by PP1M (10,325€), HAL-LAT (11,278€), RIS-LAT (11,307€), and oral-OLZ (13,556€). PP3M had the fewest total relapses/patient (0.36, 0.94, 1.39, 1.21, and 1.70, respectively), hospitalizations (0.11, 0.46, 0.40, 0.56, and 0.57, respectively), emergency room visits (0.25, 0.48. 0.99, 0.65, and 1.14, respectively) and the most QALYs (0.847, 0.735, 0.709, 0.719, and 0.656, respectively). In both cost-effectiveness and cost-utility analyses, PP3M dominated all other drugs. Sensitivity analyses confirmed base case findings. In the costing analysis, total costs were, on average, 31.9% higher than DRGs.

Conclusions: PP3M dominated all commonly used drugs. It is cost-effective for treating chronic schizophrenia in the Netherlands. Results were robust over a wide range of sensitivity analyses. For patients requiring a depot medication, such as those with adherence problems, PP3M appears to be a good alternative anti-psychotic treatment.

Keywords:

• Paliperidone 3-month
• Schizophrenia
• Cost-effectiveness
• Netherlands
• Long acting injection

Introduction

Schizophrenia is a debilitating chronic mental health problem afflicting individuals worldwide, with no prospects of a cure. The 2010 Global Burden of Disease Study¹ estimated that 21.5 million people are affected worldwide (∼3.1 per 1,000 inhabitants). That number (projected using population estimates)² would have become 23.1 million in 2016. However, rates vary across regions and differ between the sexes and ethnic groups^3,4. This disease affects men more than women, with a relative risk of 1.42 (95% CI = 1.30–1.56)^3,5. In a national survey from the Netherlands, the estimated lifetime prevalence of schizophrenia was 4/1,000 in males and 3/1,000 in females⁶. Estimates of the point prevalence in that country vary widely, ranging from 0.6–6.7 per 1,000^7–9. However, there is much variation, even among neighborhoods¹⁰.

This disease impacts society strongly and negatively¹. People with schizophrenia have a substantially decreased level of functioning¹¹. Their quality-of-life is decreased when compared to individuals not afflicted with the disease^12,13. Schizophrenia is also associated with an increased risk of premature mortality. Tenback et al.¹⁴ reported a 2.6-fold increase in risk of death in persons with schizophrenia as compared with those not having the disease. Notably, there was no association between mortality and age at onset of the disease or current prescription of anti-psychotics. In a meta-analysis of RCTs that included 17,416 patients, Kishi et al.¹⁵ concluded that mortality was associated with the disease and not the treatment drugs.

Schizophrenia further impacts society by the substantial amount of resources consumed by the healthcare system^1,16, and by those incurred by caregivers^17,18. In the Netherlands, van der Lee et al.¹⁹ estimated an average overall cost for psychiatric care of 31,071€per patient over 3 years. Inpatient care comprised 60% of the total, as compared with 31% for outpatient care and 9% for drugs. Clearly, avoiding hospitalizations is of major benefit in managing these patients.

Non-adherence and partial adherence have long been identified as major contributors to hospitalization^20,21. De Hert et al.²² conducted a thorough meta-analysis of 36 placebo-controlled studies that examined 4,657 patients. Those receiving placebo had a risk of relapse that was 5-times higher than those treated with anti-psychotics. Furthermore, they reported an odds ratio of 3.36 for relapse in patients who had been stabilized, but either discontinued their drugs or took them intermittently.

Unfortunately, non-adherence and partial adherence to medications are not uncommon. One study reported a prevalence of 42.3% “non-compliant behaviors”²³. High rates of problems with adherence were also reported by Emsley et al.²⁴ in a survey of 4,120 nurses, mostly (>70%) from Europe. They estimated that 54% of all patients with schizophrenia were non-adherent or partially non-adherent. In a similar research, Kulkarni and Reeve-Parker²⁵ reported that Australian psychiatrists estimated that 51% of all patients were either non-adherent or partially adherent. Thus, the problem is widespread and problematic.

Noordraven et al.²⁶ identified two factors strongly associated with reduced adherence which were motivation to continue medications and patients’ insight. To overcome these problems, depot formulations, or long-acting therapy (LAT) were developed. The first atypical anti-psychotic to appear on the market was risperidone (Risperdal Consta; RIS-LAT), which requires biweekly administration²⁷. The next LAT was paliperidone palmitate (Xeplion; Invega Sustenna) in 2011²⁸, which may be injected monthly. These products have generally proven themselves to be effective, but some challenges and opportunities remain. For example, there are still problems with adherence and drug continuation, as well as polypharmacy^29,30.

A novel advance to the treatment algorithm has been the production of a 3-month formulation (PP3M; Trevicta)³¹. This drug has been tested in clinical trials, which have recently been published. Berwaerts et al.³² compared PP3M (n = 160) to placebo (n = 145) over a period of more than 1 year. PP3M significantly delayed time to relapse in persons with schizophrenia. In a large (n = 1,016) 48-week trial, Savitz et al.³³ found that PP3M was non-inferior to PP1M, both in efficacy and tolerability.

PP3M and PP1M are both long-acting formulations of paliperidone, the only difference being their durations of action. On the other hand, the cost-effectiveness of this new 3-month formulation is presently unknown. In addition to that, there is no model published that analyzes and compares the cost-effectiveness of anti-psychotic drugs in The Netherlands. Therefore, the aim of this project was to determine the relative cost-effectiveness of PP3M vs four widely used anti-psychotic drugs in patients with chronic schizophrenia in The Netherlands.

Methods

This research was performed generally according to the recommendations for the economic evaluation of health technologies in The Netherlands³⁴. The exception was that the analyses took the insurer perspective, including all direct medical care costs. Indirect costs were not included because they are not relevant to the analytic viewpoint. Also, the great majority of these patients are not fully employed, so relatively little cost is involved^35,36. As validation for not using the societal approach, we point to the work by Evers and Ament⁷, who estimated that such indirect costs comprised less than 8% of the total cost of care. Although they did not include caregiver costs, that burden would be expected to be similar across drugs. Also, we did not consider the costs of treating adverse events due to the findings of two recent pharmacoeconomic analyses^37,38. De Moor et al.³⁷ concluded that adverse events “contributed only marginally” to the overall cost, while Arteaga Duarte et al.³⁸ reported that the overall difference in treatments costs for adverse events amounted to a mere 7€. Along with other anti-psychotics, two previous studies have examined the three other drugs (RIS-LAT, HAL-LAT, and oral-OLZ)^39,40. Edwards et al.³⁹ reported that the cost of treating adverse events was very low, comprising 0.8% of the total cost for RIS-LAT, 0.5% for HAL-LAT, and 1.3% for oral-OLZ. Obradovic et al.⁴⁰ produced quite similar results: 0.3% for RIS-LAT, 1.0% for HAL-LAT, and 0.7% for oral-OLZ. Therefore, it was felt that none of these costs was worth considering.

We adapted a previous decision tree model using an Excel spreadsheet, by adding a branch for PP3M⁴¹. Figure 1 depicts the model structure. Local experts provided input into the standard practices involved in the management of the care of these patients. PP3M was the primary drug of interest; comparison drugs included PP1M, RIS-LAT, HAL-LAT, and oral olanzapine (oral-OLZ), which are the drugs that are commonly used in The Netherlands for patients with schizophrenia. We excluded two other LATs, olanzapine pamoate and aripiprazole, because neither drug is used extensively in The Netherlands, and real-life data are not available for them⁴².

Figure 1. Decision tree used to model the economic analyses. Abbreviations. HAL, Haloperidol long acting injectable; RIS, risperidone long acting injectable; OLZ, olanzapine oral.

We assumed that all patients who discontinued due to adverse events or intolerance to a drug were immediately switched to second line treatment. In the case of lack of efficacy, patients could either have their doses of drug maximized or switched to alternate (second line) therapy. For PP3M, PP1M, RIS-LAT, and oral-OLZ, second line treatment was HAL-LAT; for HAL-LAT it was oral-OLZ. The rationale for using HAL-LAT as second line for PP3M, PP1M, and RIS-LAT is that these three drugs are all related (the first two consist of paliperidone, which is a metabolite of RIS)²⁸. Therefore, a different molecule is required, and an LAT is preferred in these patients. In the event of intolerance to or lack of efficacy from the second line drug, patients would be switched to clozapine, as per NICE recommendations⁴³.

The clinical inputs for this analysis were derived from the published literature. To the extent possible, we used data from The Netherlands for the analysis. Rates (calculated at 1 year) were largely based on real world experience, with data observed in actual practice in The Netherlands⁴². In that study, 14.5 million prescriptions from The Netherlands were analyzed, representing 75% of the total market. Drugs assessed included PP1M, RIS-LAT, and HAL-LAT. To determine rates for PP3M, we applied results from the head-to-head trial between PP3M and PP1M³³, using the ratio between PP3M:PP1M and applying it to the rates calculated for PP1M. Finally, we estimated rates for oral-OLZ using the 1-year ratio between RIS-LAT and oral-OLZ from the study by Olivares et al.⁴⁴ and multiplied that ratio by the calculated rates for RIS-LAT. All of these rates are presented in Table 1.

Table 1. Clinical inputs and their sources.

Drug	Category	Item	Rate	Source
PP3M	Clinical rate	Hospitalized	0.141	PP1M rate pro-rated via Savitz et al.^33
		Non-hospitalized relapse	0.152	PP1M rate pro-rated via Savitz et al.^33
		Switch due to adverse events	0.063	Pro-rated via Savitz et al.^33
		Switch due to lack of efficacy	0.087	Pro-rated via Savitz et al.^33
		Dropouts (non-AE, non-LoE)	0.328	PP1M rate pro-rated via Savitz et al.^33
	Drug dosing	Dose (mg)	332	PP1M dose ×3.5, as per EMA Trevicta Summary^67
		Dosing interval	Every 3 months	EMA Trevicta Summary^67
		Restarting regimen after missing doses	If <4 months, give usual dose; if 4–9 months, stabilize with equivalent dose of PP1M immediately and 1 week later, then PP3M every 3 months	EMA Trevicta Summary^67
PP1M	Clinical rate	Hospitalized	0.194	Pro-rated from RIS-LAT using the ratio of hospitalization rates from 6 trials (n = 1883) (Baser et al.^68; Taylor and Olofinjana^69; Fu^70; Hough et al.^64; Savitz et al.^33; Zhang et al.^71)
		Non-hospitalized relapse	0.132	Decuypere et al.^42
		Switch due to adverse events	0.089	7 trials (n = 1976) (Attard et al.^70; Zhang^72; Alphs et al.^73; Fu et al.^74; Hough et al.^64; McEvoy et al.^75; Savitz et al.^33)
		Switch due to lack of efficacy	0.123	6 trials (n = 1988) (Attard et al.^70; Zhang et al.^71; Alphs et al.^73; Fu et al.^74; McEvoy et al.^75; Savitz et al.^33)
		Dropouts	0.463	Decuypere et al.^42
	Drug dosing	Dose (mg)	95 mg	Decuypere et al.^42
		Dosing interval	monthly	EMA Summary Xeplion^28
		Restarting regimen after missing doses	Same dose immediately and 1 week later, then monthly	EMA Summary Xeplion^28
RIS-LAT	Clinical rate	Hospitalized	0.267	20 observational studies (n = 7,033) (Bitter et al.^76; Chan et al.^77; Chang et al.^78; Chue et al.^79; Deslandes et al.^80; Huang et al.^81; Koczerginski and Arshoff^82; Lammers et al.^83; Leal et al.^84; Llorca et al.^85; Lloyd et al.^86; Mihaljevic-Peres et al.^87; Niaz and Haddad^47; Olivares et al.^88; Ren et al.^89; Rossi et al.^90; Taylor and Olofinjana^69; Taylor et al.^91; Williams et al.^92; Yu et al.^93)
		Non-hospitalized relapse	0.145	5 observational studies (n = 848) (Chan et al.^77; Chang et al.^78; Koczerginski and Arshoff^82; Leal et al.^84; Yu et al.^93)
		Switch due to adverse events	0.147	Decuypere et al.^42
		Switch due to lack of efficacy	0.203	Decuypere et al.^42
		Dropouts	0.444	Decuypere et al.^42
	Drug dosing	Dose (mg)	37.7 mg	Decuypere et al.^42
		Dosing interval	Every 2 weeks	Risperdal Consta EMA^94
		Restarting regimen after missing doses	Resume prior regimen	Risperdal Consta EMA^94
HAL-LAT	Clinical rate	Hospitalized	0.214	3 observational studies (n = 744) (Conley et al.^95; Huang et al.^81; Yu et al.^93)
		Non-hospitalized relapse	0.370	Yu et al.^93
		Switch due to adverse events	0.113	Decuypere et al.^42
		Switch due to lack of efficacy	0.156	Decuypere et al.^42
		Dropouts	0.480	Decuypere et al.^42
	Drug dosing	Dose (mg)	100 mg	Decuypere et al.^42
		Dosing interval	55% every 4 weeks, 11% every 3 weeks, 25% every 2 weeks, 9% weekly	Shi et al.^96
		Restarting regimen after missing doses	Same
oral-OLZ	Clinical rate	Hospitalized	0.314	10 studies (n = 12,913) (Ascher-Svanum et al.^97; Baser et al.^68; Bitter et al.^76; Conley et al.^95; Patel et al.^98; Vanasse et al.^99; Williams et al.^100; Yu et al.^101; Lieberman et al.^102; Meltzer et al.^103)
		Non-hospitalized relapse	0.470	Yu et al.^93
		Switch due to adverse events	0.228	RIS-LAT rate adjusted via Olivares et al.^44
		Switch due to lack of efficacy	0.165	RIS-LAT rate adjusted via Olivares et al.^44
		Dropouts	0.498	RIS-LAT rate adjusted via Olivares et al.^44
	Drug dosing	Dose (mg)	12.8	18 observational studies in 17 papers (n = 14,152) (Chan et al.^77; Dossenbach et al.^104; Garcia-Cabeza et al.^105; Gaviria et al.^29; Gibson et al.^106; Gomez et al.^107; Kasper et al.^108; Kilian et al.^109; Kraemer et al.^110; Kumar et al.^111; Mladsi et al.^112; Rascati et al.^113; Sacristan et al.^114; Snaterse and Welch^115; Takahashi et al.^116; Tiihonen et al.^117; Williams et al.^100)
		Dosing interval	Daily
		Restarting regimen after missing doses	restart using same regimen
clozapine	Clinical rate	Success	0.402	4 observational studies (n = 433) (Buchanan et al.^118; Essock et al.^119; Kane et al.^120; Rosenheck et al.^121)
	Drug dosing	Dose (mg)	418	21 trials in 20 papers (n = 798) (Bitter et al.^76; Bondolfi et al.^122; Breier and Hamilton^123; Buchanan et al.^118; Conley et al.^95; Henderson et al.^124; Hong et al.^125; Josaissen et al.^126; Kane et al.^120^,^127; Klieser et al.^128; Krakowski et al.^129; Kumra et al.^130; McGurk et al.^131; Moresco et al.^132; Naber et al.^133; Sacchetti et al.^134; Tollefson et al.^135; Volavka et al.^136; Wahlbeck et al.^137)
		Dosing interval	Daily	clozapine opinion^138
placebo (oral)	Clinical rate	Relapses	0.683	9 trials in 7 papers (n = 564) (Chen et al.^139; Crow et al.^140; Hogarty and Goldberg^141; Hogarty and Ulrich^142; Kane et al.^143; McCreadie et al.^144; McGlashan et al.^145)
placebo (1-monthly)	Clinical rate	Relapses	0.415	4 trials (n = 511) (Eklund and Forsman^146; Fu et al.^74; Hough et al.^64; Kane et al.^27)
placebo (3-monthly)	Clinical rate	Relapses	0.290	Berwaerts et al.^32

Abbreviations. HAL-LAT, haloperidol decanoate therapy; OLZ, olanzapine; PP1M, paliperidone monthly injection; PP3M, paliperidone 3-monthly injection; RIS-LAT, risperidone long-acting therapy.

To estimate the time to relapse in those who discontinued from each drug, we used results from patients who discontinued active treatment and were switched to placebo in the Kim et al.⁴⁵ study. That study presents the median number of days to relapse (50th percentile), as well as days for 25% of the patients to relapse (25th percentile). These numbers can then be used to determine the time for relapse in those who discontinue each drug. In the rare cases where data were not available (i.e. four cases), observed rates from Decuypere et al.⁴² were pro-rated using ratios between drugs as reported in RCTs, using the approach described by Bucher et al.⁴⁶. That method adjusts rates from different studies by calculating a ratio of each drug against a common comparison drug. The ratio for each drug is then multiplied by a known standard rate. For PP3M, we projected rates from PP1M using the ratio of rates from the Savitz et al.³³ trial; for PP1M and oral-OLZ, RIS-LAT was used as the reference drug as there were many RCTs comparing these drugs. Clinical inputs and their sources are presented in Table 1.

To determine the length of stay of hospitalizations, we used data from Niaz and Haddad⁴⁷ and Spill et al.⁴⁸, who directly compared RIS-LAT with (mostly) oral anti-psychotics in mirror image studies. Patients receiving oral therapy stayed 61.7 days in hospital, while those on RIS-LAT stayed 31.6 days. We assumed that the other three long-acting treatments would have stays equal to that of RIS-LAT.

Two different approaches were taken for the costing of treatments. The primary analysis was based on the Dutch system of diagnosis related groups (DRGs). Under that system, clinics are reimbursed according to the number of minutes per year that healthcare providers spend with a patient. There are nine categories of reimbursement, the lowest of which (Category 230) pays €1,229.01 for 250–799 min, and the highest (Category 186) pays €67,531.381 for patients requiring >30,000 min per year⁴⁹. Tariffs are presented in Table 2. In the second approach, we used standard microcosting methods for services rendered that would be paid by insurers. Prices of these resources were obtained from official lists issued by various governmental agencies in The Netherlands. These prices appear in Table 3, which are all expressed in 2016 euros. Estimates of resources costed in previous years were adjusted to 2016 values via the Consumer Price Index for The Netherlands⁵⁰.

Table 2. List of DRG codes in The Netherlands used to reimburse treatments for schizophrenia and other major psychoses, time allotments and reimbursement amounts for each DRG category.

DRG code	Minutes/patient/year	Cost/patient/year
230	250–799	$1,232.38
184	800–1,799	$2,370.33
066	1,800–2,999	$4,422.96
067	3,000–5,999	$7,656.82
068	6,000–11,999	$14,200.24
143	12,000–17,999	$24,171.91
144	18,000–23,999	$34,256.88
185	24,000–23,999	$41,468.24
186	30,000–29,999	$67,716.51

Abbreviations. DRG, diagnosis-related group.

Data source: DBC tarivien 2016⁴⁹.

Table 3. Costs of resources used in the analysis^a.

Cost centre	Resource	Cost^b	Source
Drugs	Paliperidone 3-month injection	3.5577 €/mg	Staatscourant^147
	Paliperidone 1-month injection	4.1974 €/mg	Staatscourant^147
	Risperidone long-acting injection	3.7702 €/mg	Staatscourant^147
	Haloperidol decanoate injection	0.5002 €/mg	Staatscourant^147
	olanzapine tablets	0.0212 €/mg	Staatscourant^147
	Clozapine tablets	0.0020 €/mg	Staatscourant^147
Medical care	Psychiatrist	113.31 €/visit	ZIN^148
	Injection fee	65.00€	ZIN^148
	Primary care physician	66.77€	ZIN^148
	Social worker	65.76€	ZIN^148
Institutional care	Hospital acute psychiatric unit	2,038.60€	ZIN^148
	Acute care (DRG)	374€	NZA^149
	Day hospital	164.91€	ZIN^148
Tests	Clozapine level	18.38€	SHL^150
	Blood test	4.12€	ZIN^148

^a All costs are expressed in 2016 euros.

^b Costs for drugs are the average price across all available strengths.

Utilities were derived from those used in previous research⁴¹, but adjusted using new information. Using a time-trade-off technique, Osborne et al.⁵¹ determined patient preferences and associated utilities for receiving injectable drugs having different administration intervals. For stable disease, we subtracted scores between adjacent pairs of administration intervals, using those differences as disutilities. We assumed that PP3M would not incur a disutility. Disutility scores were 0.05 for monthly administration, a further 0.045 for biweekly, and another 0.005 for greater than biweekly. Utility for hospitalization was fixed at 0.490, and no disutilities were applied. To determine the utility for non-hospitalized relapses, we used the mid-point between stable disease and relapse, under the assumption that utilities representing quality-of-life are linear, as has been done by others⁵². Utilities are presented in Table 4.

Table 4. Utilities used in the research.

State	PP3M	PP1M	RIS-LAT	HAL-LAT	oral-OLZ	clozapine
Stable disease	0.890	0.840	0.795	0.790	0.790	0.790
Relapse - ER visit (outpatient)	0.690	0.665	0.643	0.640	0.640	0.640
Relapse - hospitalized	0.490	0.490	0.490	0.490	0.490	0.490

Abbreviations. ER, emergency room; HAL-LAT, haloperidol decanoate therapy; OLZ, olanzapine; PP1M, paliperidone 1-monthly injection; PP3M, paliperidone 3-monthly injection; RIS-LAT, risperidone long-acting therapy.

The primary analysis was cost-utility, with quality-adjusted life-years (QALYs) gained as the outcome. Secondary analyses involved the determination of cost-effectiveness, with relapse as the clinical outcome. Three separate analyses were performed for different aspects of relapse, including relapses requiring hospitalization, relapses that could be managed as out-patients, and total relapses. The analytic time horizon of 1 year precluded discounting of either costs or outcomes. Economic outcomes were judged against the thresholds authorized by the Dutch health technology assessment agency Zorginstituut⁵³.

Sensitivity analyses were conducted to assess the robustness of the model and its results. In scenario analyses, we varied inputs according to alternate possibilities, including the dose of PP3M, the dose of PP1M, maximizing dropouts, eliminating dropouts, and using alternate utility scores.

We also conducted 1-way sensitivity analyses on all clinical inputs and cost centers. Finally, a probability sensitivity analysis was conducted with 10,000 iterations, varying all inputs within their plausible ranges using standard distributions.

Results

Results of the economic analysis appear in Table 5. Expected costs were lower in the PP3M treatment arm in both the DRG or costing analyses. Furthermore, PP3M generated higher QALY values and lower relapse or hospitalization rates. Therefore, PP3M was the dominant treatment option (i.e. both least costly and more effective than comparison).

Table 5. Outcomes from the pharmacoeconomic analysis.

Drug	Expected cost^a using DRGs	Expected cost^a using microcosting	Total relapses	Hospital admissions	Emergency room visits	QALYs	Economic outcome
PP3M	€8,781	€12,927	0.36	0.11	0.25	0.847	Dominant
PP1M	€10,325	€13,148	0.94	0.46	0.48	0.735	Dominated
HAL-LAT	€11,278	€13,801	1.39	0.40	0.99	0.709	Dominated
RIS-LAT	€11,307	€14,650	1.21	0.56	0.65	0.719	Dominated
oral-OLZ	€13,556	€18,335	1.70	0.57	1.14	0.656	Dominated

^a Costs are expressed in 2016 euros.

Abbreviations. HAL-LAT, haloperidol decanoate therapy; OLZ, olanzapine; PP1M, paliperidone palmitate 1-monthly injection; PP3M, paliperidone 3-monthly injection; QALY, quality adjusted life-year; RIS-LAT, risperidone microspheres therapy.

Costs were consistently higher under the costing version, with an average difference of 31.9%. The difference in costs was greatest for PP3M (43% higher) and lowest for haloperidol (22% higher).

For PP3M, the driver of the costs was the acquisition cost of the drug, comprising 48% of the total, with 39% for hospital/institutional care and 13% for medical care. For all of the other drugs assessed, hospitalization was the major driver, amounting to 63%, 69%, 85%, and 92% of the overall treatment for PP1M, HAL-LAT, RIS-LAT, and oral-OLZ, respectively. Drugs ranked second for PP1M, but cost the least for the remaining three drugs.

Scenario analyses and 1-way sensitivity analyses resulted in no changes to cost-effectiveness results. In probability sensitivity analyses of the DRG analyses, PP3M remained dominant over all comparison treatments. PP3M had a lower cost than PP1M in 99.8% of the simulations with DRGs; it was dominant in 85.0% and cost-effective in 85.8%. With microcosting, results were very similar. PP3M cost less in 99.9% of the 10,000 simulations; it was dominant in 84.6%, and cost-effective in >99.9%. Scatterplots of PP3M vs PP1M are presented in Figures 2(a) (DRG analysis) and b (Microcosting analysis). For the other drugs examines, PP3M similarly dominated in very high proportions of the simulations. Scatterplots are presented as Figures A1–A3 in the Appendix.

Figure 2. Scatterplot of cost-utility results between PP3M and PP1M using (a) DRGs and (b) microcosting.

Discussion

PP3M represents a major advance in the treatment of chronic schizophrenia. In its development, a different particle size was used, but applying the same Nanocrystal technology as was used for PP1M. As a result, the active drug is more slowly released, providing continuous therapeutic levels over at least 3 months, with a half-life of 84–139 days^53–55. Also, if discontinued after having achieved steady state, therapeutic levels remain in the blood for 5–7 months⁵⁴.

These properties generate many clinical advantages. Having a therapeutic blood level over a long period of time provides fewer opportunities for omitting doses, which decreases the probability of relapsing. In a head-to-head trial, Savitz et al.³³ did find a lower rate of relapse in patients treated with PP3M. We also found that PP3M was associated with substantially fewer relapses than PP1M.

When comparing the costs from DRGs to those from microcosting, we found a simple average difference of 31.9%, with DRGs being lower. This difference corresponds with the results of Plehn et al.⁵⁷, who reported a difference of 32.8% for cardiac care. This is somewhat higher than the 23.8% found by Baumgart et al.⁵⁸ for ulcerative colitis and the 22.6% by Heerey et al.⁵⁹ in myocardial infarction and HIV cases. Thus, DRGs appear to have under-estimated costs in these papers. Schreyögg et al.⁶⁰ determined that (because of how they are calculated) DRGs are highly correlated with hospital length of stay, but not resource consumption or case complexity. For this, and other reasons, DRG values vary considerably across nations, and attempts are being made to standardize them for Europe⁶¹.

Four other research projects have reported on the pharmacoeconomics of PP3M, all as posters. Benson et al.^62,63 used the RCT of Berwaerts et al.³² to model costs and outcomes between PP3M and placebo over 1 year. All clinical and resource data were gathered during the conduct of that trial. Costs in 2014 USD and outcomes were reported for each phase of the trial. In the open label phase, overall costs were comparable between PP3M and placebo. In the subsequent double-blind phase, costs decreased for PP3M, but increased for placebo. Overall, PP3M dominated placebo due to consuming fewer resources, as a results of lower rates of hospitalization and relapses.

Arteaga Duarte et al.³⁸ used a 5-year Markov model in France to compare PP3M with PP1M. Data were obtained from the RCTs by Savitz et al.³³ and Berwaerts et al.³². They found cost savings and increased QALYs with PP3M. De Moor et al.³⁷ also used a 5-year Markov model to examine PP3M and PP1M in Belgium. Data were extracted from the RCTs by Berwaerts et al.³² and Hough et al.⁶⁴. Results were comparable to those from Arteaga Duarte et al.³⁸. Garcia-Ruiz et al.⁶⁵ presented an economic analysis comparing PP3M with PP1M based on the 1 year head-to-head randomized controlled trial by Savitz et al.³³. PP3M dominated PP1M in both cost-utility and cost-effectiveness analyses. Thus, results were consistent across analyses that used different inputs and different economic models.

Differences between our results and those of Benson et al.^62,63 probably arose due to the structure of the models. Those authors modeled the trial by Berwaerts et al.³², whereas we also used data from the Savitz et al.³³ trial. In the Benson et al.^62,63 analysis, patients who relapsed were censored and resources were not costed after censoring. On the other hand, we costed resource use for all patients over a full year, including those switched to other drugs and those who dropped out. Finally, the Benson calculations did not include drug costs, while the present analysis did, which explains the very large difference in reports costs.

Model validation

We elected to use a decision tree model rather than a Markov for a number of reasons. First, results would likely be quite similar using either model⁶⁶. Second, an existing decision tree was available, which required minor modifications to incorporate PP3M into the analysis. Third, the time frame of 1 year matched the duration of the two major trials of PP3M^32,33, and was adequate for assessing outcomes of interest.

To validate the outputs of our primary cost-utility model, we compared results with other similar economic analyses with respect to costs and QALYs. Our average cost to manage the psychiatric care of patients for 1 year was 10,793€, which compares favorably with the estimate of van der Lee et al.¹⁹ of 10,357€. Table 6 lists the costs and QALYs from the present studies and validating rates from other independent studies. All rates were quite close, with the exception of oral-OLZ, which had a lower cost in the economic analysis of Druais et al.⁵².

Table 6. Validation of the cost-utility model.

Parameter	Drug	Rate in present analysis	Validation rate	Difference	% difference	Validation source
Expected cost	PP3M	€8,781	€8,783	–€2	−0.02%	De Moor et al.^37
	PP1M	€10,325	€10,176	€149	1.4%	Druais et al.^52
	HAL-LAT	€11,278	€11,025	€253	2.2%	Druais et al.^52
	RIS-LAT	€11,307	€10,990	€317	2.8%	Druais et al.^52
	oral-OLZ	€13,556	€10,076	€3,480	25.7%	Druais et al.^52
Expected QALYs	PP3M	0.847	0.850	−0.003	−0.4%	Average of PP1M rate from 10 studies + PP3M-PP1M difference from Osborne^51
	PP1M	0.735	0.753	−0.018	−2.4%	Druais et al.^52
	HAL-LAT	0.719	0.753	−0.034	−4.7%	Druais et al.^52
	RIS-LAT	0.709	0.725	−0.016	−2.2%	Druais et al.^52
	oral-OLZ	0.656	0.711	−0.055	−8.4%	Druais et al.^52

HAL-LAT, haloperidol decanoate; OLZ, olanzapine; PP1M, paliperidone palmitate 1-monthly injection; PP3M, paliperidone 3-monthly injection; QALY, quality adjusted life-year; RIS-LAT, risperidone microspheres depot.

Limitations

In our analysis, we restricted the time horizon to 1 year, in parallel with the two available clinical trials. It is recognized that this disease persists throughout the life of the patient. However, no long-term data are presently available to extend the analysis with any confidence. Results could change over time.

Another limitation is that some (but not all) of the clinical data were obtained randomized controlled trials, which may not reflect what happens in actual practice. In trials, patients are highly selected and treated according to strict protocols. In “real life”, other influences affected the patients and the care delivered. We did, however, incorporate data from observational studies to account for some of that variance.

We considered only the direct costs of care that would be paid by the Dutch insurer. We did not include indirect costs of care such as decreased employment, and time spent by family or other caregivers.

As well, the analysis was done specifically for The Netherlands, which has a unique method of payment. Therefore, results may not apply to other countries or healthcare systems.

In estimating utilities, we used values obtained from the literature, which represent preferences of the global community, and were not specific to The Netherlands. It is possible that local tariffs could differ.

In addition, as previously mentioned, we excluded costs associated with treating adverse events. A further example was published by another group that examined several long-acting anti-psychotics, including PP1M and RIS-LAT. They found that the cost of treating adverse events ranged from £153–£190, comprising 0.10–0.12% of the total cost of treatment, which ranged from £157,594–£163,359 over 10 years⁵². Considering these results and those mentioned above, it appears that omission of the cost of treating adverse events has little impact on overall results.

Conclusions

In the primary analysis using DRGs, PP3M was cost-effective, dominating all of the other drugs most commonly used for chronic schizophrenia in The Netherlands. Robustness of results was confirmed in both 1-way and probability sensitivity analyses. Findings were similar when we used microcosting.

Much of the clinical advantage for PP3M may be attributed to its very long half-life, which assures a therapeutic blood level for a long period of time. Having a 3-monthly dosing interval reduces resource utilization in an already strained healthcare system. Also, it lowers the probability of failing to adhere to prescribed regimens.

The drug was the cost driver for PP3M; for the rest, hospitalization consumed the major proportion of the costs. The increased drug cost was more than offset by the savings derived for avoiding hospitalization.

A limitation is that many of the data inputs for PP3M were derived from clinical trials, which tend to reflect ideal practice, as opposed to that in the “real world”. Although we did use such data for the other drugs, that information was not yet available for PP3M. Ideally, all results should be verified using observational studies to examine the situation under actual conditions in clinical practice.

Transparency

Declaration of funding

This research was funded by Janssen.

Declaration of financial/other relationships

TE and BB received funding for the research and manuscript. FT, TD, and KVI are employees of Janssen. Peer reviewers on this manuscript have received an honorarium from JME for their review work, but have no other relevant financial relationships to disclose.

Appendix

Scatterplots of PP3M vs RIS-LAT, HAL-LAT, and oral-OLZ

Figure A1. PP3M vs RIS-LAT.

Figure A2. PP3M vs HAL-LAT.

Figure A3. PP3M vs oral olanzapine.