A systematic review of the current application status of decision-analytical models in the pharmacoeconomic evaluation of targeted therapies for pulmonary arterial hypertension

Abstract

Background

The implementation of targeted drug therapy results in a significant improvement in both survival rates and quality of life among patients diagnosed with pulmonary arterial hypertension (PAH), concurrently imposing a greater financial burden on them. The use of pharmacoeconomic evaluation based on decision-analytical models is extensively employed in the rational allocation of healthcare resources.

Objectives

The present study conducted a systematic review of the literature on the pharmacoeconomic evaluation of drugs for treating PAH, with a focus on summarizing the composition and sources of parameters in decision-analytical models. This study aims to provide methodological guidance for future economic research.

Methods

The review was conducted across six databases (PubMed, Embase, the Cochrane Library, CNKI, VIP, WanFang Data) and two health technology assessment agency websites (NHS EED, INAHTA). The characteristics of each study and the compositional details of the decision-analytical models are extracted.

Results

In total, 13 published studies were included. The pharmacoeconomic evaluation methods employed in the studies included cost-effectiveness analysis (CEA) and cost-utility analysis (CUA). The decision analysis models employed in all 13 studies were Markov models. The models were all constructed on the basis of the World Health Organization (WHO) functional class, with variations in parameter settings and sources.

Conclusions

All 13 Markov models provided useful insight into PAH modeling. Future research in this field can employ these research methods according to diverse research objectives. The utility values were derived from a single source; therefore, future studies should evaluate the quality of life in patients with PAH across varying disease severities.

Introduction

Pulmonary arterial hypertension (PAH) is a syndrome characterized by progressive elevation of pulmonary vascular resistance resulting from remodeling of the pulmonary vasculature, potentially leading to right ventricular failure or even mortality [1]. In the era of conventional treatment, long-term oxygen therapy, diuretics, digoxin, anticoagulant therapy, and calcium channel blockers (CCBs) have been employed to ameliorate symptoms; however, patients exhibit a bleak prognosis and a low survival rate [2]. With advancements in the understanding of PAH pathogenesis, drugs that target the endothelin receptor, NO, and prostacyclin signaling pathways have been approved [3]. In the era of targeted drug therapy, there has been a significant increase in both the survival rate and quality of life for patients with PAH.

The incidence and prevalence of PAH are relatively low, rendering the development of therapeutic drugs for PAH a formidable challenge in terms of both complexity and cost. Consequently, this factor also contributes to the elevated prices of targeted therapy medications. Patients with PAH not only face the challenge of impaired work capacity due to physical limitations but also bear the substantial financial burden associated with drug treatment. Concerning medication expenses for PAH patients, the median total per patient per day and 3-year total expenditures were $56 and $50,599, respectively [1]. The chronic and incurable nature of PAH necessitates lifelong administration of targeted drugs to effectively manage disease progression, thereby imposing a substantial economic burden on both patients and society [4,5,6].

Pharmacoeconomics is an interdisciplinary field that aims to optimize the allocation of limited drug resources to maximize health outcomes. The construction of decision analytical models is the predominant approach in pharmacoeconomic evaluation for comparing the cost-effectiveness of different intervention strategies. These models include the decision tree model, partition survival model, Markov model, and discrete-event simulation model [7]. The Markov model categorizes a disease into distinct health states on the basis of its natural progression and simulates the long-term advancement of the disease by incorporating transition probabilities between these states, along with associated costs. This facilitates the computation and comparison of cumulative costs and health outcomes for different intervention strategies. The Markov model, extensively employed in economic research on chronic disease treatment drugs, has the ability to simulate long-term disease progression by utilizing short-term transition probability data [8].

The structure and parameters of the Markov model have crucial influences on the outcomes of pharmacoeconomic evaluation. Currently, pharmacoeconomic evaluations of various targeted drugs for PAH exist; however, discrepancies can be observed in terms of model construction methodologies, model assumptions, variable settings, and parameter sources. Therefore, the aim of this study was to conduct a systematic review of the literature concerning the pharmacoeconomic evaluation of drugs for treating PAH, with a focus on model construction and parameter sources. The goal is to provide methodological guidance for future economic research and enhance the quality of economic evaluations.

Methods

This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Supplementary Table 1) [9].

Search strategy

The PubMed, Embase, Cochrane Library, NHS EED, INAHTA, CNKI, VIP and WanFang databases were systematically searched from the inception of the database until January 25, 2023. Furthermore, we conducted a supplementary search in April 2024. The search terms included economic evaluation, economic study, pharmacoeconomic evaluation, pharmacoeconomic study, pharmacoeconomic analysis, cost-minimization analysis, cost‒benefit analysis, cost-utility analysis, cost-effectiveness analysis, pulmonary hypertension and pulmonary arterial hypertension (Supplementary Table 2).

Selection of studies

The study inclusion criteria are outlined in detail in Table 1. According to the predefined inclusion criteria, two researchers (WXD and ZZ) independently conducted literature screening and data extraction, ensuring that the results were cross-validated for accuracy. In cases of disagreement, a third researcher (CP) was consulted to facilitate resolution.

Table 1 Inclusion criteria for review

Data extraction and quality appraisal

The extracted data were categorized into two sections: study characteristics and model characteristics. Study characteristics included the first author’s name, publication date, country, target population, intervention and control measures, as well as the evaluation type employed. Model characteristics include the perspective, time horizon, cycle length, state partitioning of the model structure, cost, transition probabilities, discount rate, source of utility, and uncertainty analysis.

The reporting quality of economic evaluations was systematically assessed using the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) checklist [10]. This 28-item instrument required reviewers to classify each criterion as: “Yes” (fully addressed), “Partially” (partially addressed), “No” (not addressed), or “NA” (not applicable). Two independent researchers (WXD and ZZ) performed the evaluations.

Results

Characteristics of the included studies

A total of 2443 articles were identified through a comprehensive search of databases and health technology assessment websites. After 225 duplicate articles were excluded, 2218 articles remained for further evaluation. Following screening conducted by two researchers, a final selection of 13 literature sources was made [12,13,14,15,16,17,18,19,20,21,22,23,24]. The detailed process of literature screening is visually presented in Fig. 1.

Fig. 1

Preferred reporting items for systematic reviews and meta-analyses (PRISMA) diagram for the inclusion of relevant studies. CNKI China National Knowledge Infrastructure, VIP China Science and Technology Journal Database, NHS EED The National Health Service Economic Evaluation Database, INAHTA International Network of Agencies for Health Technology Assessment

The reporting quality of the 13 included studies was assessed using the CHEERS checklist. Most studies adequately addressed key items such as objectives, methodology, and outcomes, with all studies clearly reporting titles, abstracts, and model structures. However, partial reporting was observed in domains like study perspective [13] and funding sources [14, 15, 17, 19]. Notably, all studies were theoretical model analyses, with parameters derived from published literature and clinical trial data, and model structures designed based on the natural history of the disease. None of the studies directly involved patients or stakeholders; thus, CHEERS checklist item 21 (“Approach to engagement with patients and others affected by the study”) was not applicable to any of the included studies.

The initial publication of the first economic evaluation of drugs for PAH treatment via a Markov model can be traced back to 2009, as mentioned in Table 1 [12]. The latest study regarding this subject matter was published in 2023 [24]. Among the 13 studies included, six were carried out in developed nations across Europe and the United States [12,13,14,15,16, 18], whereas the other seven originated from developing Asian countries [17, 19,20,21,22,23,24]. Notably, Canada has established itself as a significant global frontrunner in regard to conducting pharmacoeconomic research [15, 19]. Among the 13 included studies, the evaluation encompassed all five classes of targeted drugs utilized in PAH treatment, with a primary focus on comparing the cost-effectiveness between targeted drugs and standard therapy.

The pharmacoeconomic evaluation methods employed in the studies included cost-effectiveness analysis (CEA) and cost-utility analysis (CUA). Upon examination of the complete texts, it was observed that all studies employed utility values for the computation of quality-adjusted years (QALYs) as the health outcome assessment metric. Therefore, all studies should be referred to as cost‒utility analyses. The decision analysis models employed in all 13 studies were Markov models (See Tables 2 and 3).

Table 2 Summary of identified studies

Table 3 Modeling approach and the sources of parameter values in the included studies

Modeling approach and the sources of parameter values

Study perspective

The pharmacoeconomic evaluation guidelines of numerous countries worldwide advocate prioritizing the social perspective and the healthcare system perspective when conducting pharmacoeconomic evaluations [25,26,27,28]. Studies based on the social perspective should encompass direct medical costs, direct nonmedical costs, and indirect costs, whereas healthcare system-based studies require only the inclusion of direct medical costs [27]. Among the 13 economic evaluation articles included in this study, eight articles adopted the healthcare system perspective [12, 14,15,16, 21,22,23,24], whereas two articles took the social perspective [18, 20], potentially attributed to challenges associated with capturing direct nonmedical costs and indirect expenses. Furthermore, one article adopted the payer perspective [19], whereas two articles did not explicitly specify the study perspective [13, 17].

Time horizon and cycle length

Owing to the incurable and progressive nature of PAH, all the included studies were designed with a prolonged time frame to simulate the cumulative costs and health outcomes over an extended period of time. The time horizon for the six studies encompassed the entire lifetime, resulting in 99% of patients transitioning into a death state [17, 18, 20,21,22,23]. The time horizon was set to 30 years in five articles [12, 15, 16, 19, 24], while one article specified a limit of three years [14], and another article limited the time horizon to one year [13]. The cycle length set in each study remained consistent, with a 3-month cycle observed in 12 of the studies [12,13,14,15,16, 18,19,20,21,22,23,24]. The rationale behind this lies in the fact that the majority of clinical trials evaluating the efficacy of drugs for PAH typically have a duration of 12 weeks, posing challenges in obtaining therapeutic effect data over an extended period. Therefore, adopting a 3-month cycle length more closely aligns with the practicality observed in clinical settings.

State partitioning of the model structure

The literature included in the study was classified into different Markov states on the basis of functional class by the World Health Organization (WHO), with death considered the absorbing state. The initial state distributions incorporated into the model are derived primarily from hypotheses [12,13,14,15,16, 18,19,20,21], with only a limited number of studies employing initial state distributions on the basis of real-world data [17, 22,23,24]. The age and sex distributions of the initial cohort were obtained from the PAH registry [24, 29, 30].

Cost composition and source of values

The cost categories incorporated in the model should be aligned with the study perspective. All the literature included in this study covered direct medical costs, such as expenses related to medication, charges for outpatient services, expenditures during hospitalization, fees for follow-up examinations, and costs associated with adverse reaction treatment. Studies adopting the social perspective also consider direct nonmedical costs and indirect costs. Different studies feature varying cost categories on the basis of data availability.

The sources of drug cost parameters include hospital data, data provided by pharmaceutical manufacturers, drug prescription records, and medical insurance data. The sources of outpatient cost parameters include hospital data, medical insurance data, published literature, and other relevant resources. Notably, the utilization of hospital data in the literature accounts for approximately 50% of hospital data [13, 17,18,19,20, 22]. The sources of hospitalization cost parameters encompass a variety of data, including hospital records, medical insurance information, and published literature. Notably, the majority of literature relies primarily on hospital data. The sources of follow-up examination cost parameters include hospital data, medical insurance reimbursement data, and published literature. Significantly, the majority of studies rely heavily on medical insurance reimbursement data. The sources of parameters for ADR treatment costs include medical service price lists, data on medical insurance reimbursements, published literature, and expert consultations. Owing to the unavailability of primary data, both direct nonmedical costs and indirect costs were collected via questionnaires administered to patients diagnosed with PAH in the included studies [18, 20].

Transition probability and parameter source

On the basis of the classification of model states, the transition probability within each state is divided into two components: the probability of transitioning between each FC state and the probability of mortality for each FC state. The transition probability parameters of FC states are derived primarily from RCTs and meta-analyses, with a limited number of studies utilizing data sourced from published literature. The mortality probability is determined by combining the mortality risk ratio of PAH patients with varying disease severities, as reported in the published literature, with the life table for the general population [31]. Notably, the calculation of the transition probability in Fan et al.‘s study differs from that in other studies. The transition probability between FC states is derived from data obtained through the Bosentan Charity Project, whereas the death probability is calculated on the basis of data collected from PAH registries across various countries [17].

Discount rate

The costs and health outcomes of literature with a study duration exceeding one year were discounted via discount rates recommended by local health technology assessment agencies or pharmacoeconomic evaluation guidelines. Fan et al. did not discount health outcomes for unspecified reasons [17].

Utility

The utility value most commonly utilized in the literature included was the SF-36 scale-based utility value converted by Keogh et al. [32] A questionnaire survey was used to examine the impact of bosentan on the quality of life of patients with PAH. The utility value of each FC status measured with bosentan, as assumed in all the literature, is presumed to be applicable to other interventions.

Evaluation methods

Antonio Roman et al. [14] clearly distinguished between the ICER and the ICUR in their study. They conducted a CEA by utilizing cumulative costs and life years to calculate the ICER while simultaneously performing a CUA using cumulative costs and QALYs to determine the ICUR. The remaining 12 articles exclusively consisted of CUA, which evaluated the cumulative costs and QALYs, followed by computation of the ICUR. The ICER mentioned in the original literature is derived from the calculation of costs and QALYs. Therefore, within these 12 studies, while ICER and ICUR differ only in their expression, they share the same meaning, as both represent ICUR.

Uncertainty analysis

The literature reviewed in this study included uncertainty analyses of the research findings, with a primary focus on deterministic sensitivity analyses and probability sensitivity analysis as the key evaluation criteria. The scenario analysis was conducted in three articles [12, 19, 24]. Several studies have performed deterministic sensitivity analysis to validate the stability of parameters and models by substituting specific parameters [15, 17].

Discussion

Currently, there are variations in research content and model construction methods across the global literature employing Markov models to assess the economic efficiency of PAH treatment drugs, as well as discrepancies in parameter source selection within the model. After the fundamental characteristics of the included studies and the specific details of Markov model construction are extracted, the following methodological references can be provided for future economic research.

The WHO FC of disease severity is a well-established and widely accepted practice in industry, rendering it highly recommended for model status classification in accordance with academic standards.

The recommended approach for determining the time horizon is to establish a standard analysis period of 30 years as the fundamental timeframe while adjusting shorter or longer durations on the basis of specific research objectives to accomplish scenario analysis.

The determination of cycle length should be based on the clinical trials referenced by the model to closely align with clinical practice [7, 8, 25, 33].

The setting of cost types to be included should align with the study perspective, and hospital databases and medical insurance reimbursement data are the primary sources that accurately reflect cost resource consumption in real-world scenarios. The most recommended and feasible approach for collecting direct nonmedical costs and indirect costs is through a questionnaire survey [28, 34].

The acquisition of transition probabilities is recommended to be sourced from published literature, which can be obtained through clinical trials and meta-analyses. Mortality rates are suggested to be acquired in conjunction with life table conversion, which is specific to the study countries [34, 35].

Pharmacoeconomic evaluations exceeding a study period of one year necessitate the application of discounting principles for costs and health outcomes, as recommended by each national guideline for pharmacoeconomic evaluation [36, 37]. The determination of the discount rate is contingent upon the specific national conditions.

Our review highlights that utility values in PAH models rely predominantly on a single historical source [32], potentially limiting generalizability given advancements in targeted therapies. Future studies should prioritize multinational registries to capture diverse utility data, employ standardized tools, and integrate patient-centered approaches to enhance reliability and clinical relevance.

The economics of interventions for chronic diseases necessitate the computation of long-term cumulative costs and QALYs, followed by the derivation of the ICUR.

The impact of uncertainty on the results of basic analysis can be explored through deterministic sensitivity analysis and probabilistic sensitivity analysis, which evaluate parameter uncertainty. Additionally, scenario analysis can be conducted to assess methodological and model uncertainty by altering assumptions and data sources if necessary [38, 39]. The literature incorporated in this study predominantly centers around the examination of parameter uncertainty while placing relatively less emphasis on methodology and model uncertainty. It is worth noting that besides parameter uncertainty, structural uncertainty in the model also deserves attention. Although widely used, Markov models rely on several structural assumptions, such as the memoryless property of state transitions and the homogeneity of patient populations within each functional category. These assumptions may not fully capture the complexity of PAH progression and treatment effects in real-world settings. Future research could explore alternative model designs, such as partitioned survival models or discrete event simulation models, which might provide different insights into the cost-effectiveness of PAH treatments. Analyzing results using different models can also enhance the robustness and validity of pharmacoeconomic evaluations.

With the expected launch of novel targeted therapies for PAH, pharmacoeconomic assessments will play a central role in guiding drug pricing and reimbursement policies. However, existing guidelines lack disease-specific recommendations for rare disease drug assessments. Our research consolidates and evaluates all modelling methods and data sourcing strategies used in PAH pharmacoeconomic studies and proposes a standardized framework to overcome challenges such as limited real-world evidence and uncertainties in model validity. This review not only synthesizes methodological advances, but also provides actionable, multidimensional perspectives to inform future research in PAH and other rare diseases.

Our systematic review emphasizes the critical need for transparent and standardized methodological approaches in decision-analytical models, such as Markov models stratified by WHO functional class, in generating robust cost-effectiveness evidence for PAH. Well-validated models enhance the credibility in pharmacoeconomic evaluations, thereby informing clinical decision making by prioritizing cost-effective therapies. Moreover, these analytical frameworks provide policymakers with essential data required for the formulation of equitable reimbursement policies.

Conclusion

We identified 13 cost-effectiveness models for inclusion in this review, all of which provided useful insight into PAH modeling. The number of studies published by developed and developing countries is approximately equal. The majority of studies compare the cost-effectiveness of targeted drugs with that of supportive treatment. The studies were all conducted with the Markov model as the foundation. The study horizon primarily covered a span of 30 years, with the cycle length determined on the basis of referenced clinical trials. The transition probabilities and costs were predominantly sourced from published clinical trials or the literature. The utility values originated from a single source, with reference literature dating back quite some time. There is an urgent need for further measurement of utility values in patients with PAH across different disease severities.