Common flaws exist in published cost-effectiveness models of pharmacologic stroke prevention in atrial fibrillation

Abstract

Objectives

Decision makers use models to assist in evaluating the cost-effectiveness of pharmacologic stroke prevention in atrial fibrillation (SPAF).

Study Design and Setting

We performed a search of databases through October 3, 2012 to identify pharmacologic SPAF cost-effectiveness models.

Results

Of 30 identified models, 28 included warfarin, but only 60% assessed the impact of warfarin control on conclusions. Aspirin, dual antiplatelet, and newer anticoagulants were included in 41%, 10%, and 63% of models, respectively. Models used similar structures but included varying health states and made varying assumptions. They rarely reported performing a literature search to identify anticoagulant-specific inputs and used similar and older sources. Sixteen models used a lone randomized trial to reflect the efficacy and safety of main comparisons. One-third of models claimed a societal perspective; however, none included indirect costs. Patients typically initiated anticoagulation in the sixth or seventh decade of life and are followed for their lifetimes. Almost 70% of incremental cost-effectiveness ratios were below reported willingness-to-pay thresholds. All used deterministic sensitivity analyses and 77% conducted Monte Carlo simulation. Less than half of the models were rated “high quality,” yet were frequently published in high-impact journals.

Conclusion

Pharmacologic SPAF cost-effectiveness models have been extensively reported, but many may have flaws giving reason for decision makers to use caution. We provide 10 recommendations to avoid common flaws in SPAF cost-effectiveness models.

Introduction

What is new?

• Cost-effectiveness analyses are increasingly being used to inform decision makers about the relative value of alternative treatment strategies for chronic disease states such as atrial fibrillation.

• Warfarin has been the anticoagulant of choice for stroke prevention in atrial fibrillation (SPAF), but there has been a revitalized interest in developing newer anticoagulants (eg, oral factor Xa inhibitors, oral direct thrombin inhibitors) with the potential of superior efficacy and/or safety.

• Numerous economic models have assessed the cost-effectiveness of pharmacologic agents for SPAF; however, the methods often vary in complexity (health states included), and assumptions made (anticoagulant persistence) often differ in noteworthy ways.

• Overall structures and inputs used in these models did not appear to vary as significantly as with other chronic health state models; however, inputs were sometimes dated and selectively chosen from the literature.

• Other potential flaws of models included lack of consideration of varying international normalized ratio control in results, use of a sole randomized trial to support comparative efficacy and safety assumptions, not including indirect costs in models conducted from the societal perspective, and failure to conduct probabilistic sensitivity analysis (Monte Carlo simulation).

• More specific guidance for pharmacologic SPAF models based on the input of stakeholders, decision makers, and clinicians is needed to help relieve concerns regarding usability and transparency currently expressed by the intended publishers and users of these models.

• In this article, we provide some initial guidance to improve the usability and transparency of SPAF economic models.

Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia, and it is estimated that one in every four people >40 years of age will develop AF during their remaining lifetime [1], [2]. Current prevalence estimates suggest about 3 million Americans and over 6 million Europeans suffer from AF, and its prevalence is expected to at least double by 2050 [1], [2].

Although AF increases patients' risk of ischemic stroke by about five-fold, proper anticoagulation can significantly reduce that risk [2], [3], [4], [5]. Warfarin has been the anticoagulant of choice for stroke prevention in atrial fibrillation (SPAF), but recently, there has been a revitalized interest in developing newer anticoagulants (eg, oral factor Xa inhibitors, oral direct thrombin inhibitors) with the potential of superior efficacy and/or safety [6], [7], [8], [9].

Cost-effectiveness analyses are increasingly being used to inform decision makers about the relative value of alternative treatment strategies for chronic disease states such as AF [10], [11]. Although cost-effectiveness analyses can be conducted alongside randomized controlled trials (RCTs), more often than not, they take the form of economic models (Markov models or discrete event simulations). These models begin with the formulation of a long-term map of patients' health that depicts their transition through associated health states (dependent on the therapy received). Data or inputs underlying the probabilities of relevant outcomes, health-related quality of life, and costs are then identified and selected from published (or sometimes unpublished) literature. Finally, the incremental costs and outcomes are calculated and interpreted in context of a benchmark value. These are often depicted as a cost or quality-adjusted life-year (QALY) or life-year (LY) gained.

Numerous economic models have assessed the cost-effectiveness of pharmacologic agents for SPAF (Appendix e-references e1-e30 at www.jclinepi.com); however, the methods often vary in complexity, and assumptions made often differ in noteworthy ways. Recognizing the importance of pharmacologic SPAF models in health care decision making, as evidenced by the attention drawn to pharmacologic SPAF economic model data in recent American Heart Association/American Stroke Association and the European Society of Cardiology guidance documents, it becomes important to understand what constitutes such a model [4], [5]. A greater understanding of the similarities, differences, and potential flaws of these models may also help future researchers build more useful models and provide decision makers with the knowledge to better interpret and use them [10]. Thus, we systematically reviewed existing cost-effectiveness models of pharmacologic SPAF with a focus on model structure and methodology, data sources, handling of uncertainty, and presentation of results.