A case study of ex ante, value-based price and reimbursement decision-making: TLV and rimonabant in Sweden

Abstract

Value-based pricing (VBP) is a method of setting prices for products based on perceived benefits to the consumer. When information is symmetric and freely available and agency is perfect, VBP is efficient and desirable. Because of substantial information asymmetries, medical insurance distortions, and the prescribing monopoly of physicians, VBP is rare for prescription drugs, though a number of countries have recently moved in this direction. Because the potential benefits can be sizable, it is high time for a review of actual VBP-based decision-making in practice. Sweden, with its pharmaceutical benefits board (TLV), was an early adopter of VBP decision-making. We illustrate actual decision-making, thus, using the case of Acomplia^® for the treatment of obesity in Sweden, with and without the presence of co-morbid conditions. This example has a number of features that will be useful in illustrating the strengths and weaknesses of VBP in actual practice, including multiple indications, a need for not just one but two economic simulation models, considerable sub-group analysis, and requirements for additional evidence development. TLV concluded, in 2006, that Acomplia was cost-effective for patients with a body mass index (BMI) exceeding 35 kg/m² and patients with a BMI exceeding 28 kg/m² and either dyslipidemia or type 2 diabetes. Because of uncertainty in some of the underlying assumptions, reimbursement was granted only until 31 December 2008, at which time the manufacturer would be required to submit additional documentation of the long-term effects and cost-effectiveness in order to obtain continued reimbursement. Deciding on reimbursement coverage for pharmaceutical products is difficult. Ex ante VBP assessment is a form of risk sharing, which has been used by TLV to speed up reimbursement and dispersion of effective new drugs despite uncertainty in their true cost-effectiveness. Manufacturers are often asked in return to generate additional health economic evidence that will establish cost-effectiveness as part of ex post review. The alternative is to delay the reimbursement approval until satisfactory evidence is available.

Appendix

The economic models used in the submission for reimbursement for Acomplia

The models

Both economic models are deterministic, based on Markov health states, and run using Microsoft Excel^® (Seattle, WA) and Visual Basic^®. The key features of obesity that were captured in the model included poor cardiovascular risk factor values (blood pressure, cholesterol values, and glucose tolerance and diabetes), elevated risks of cardiovascular disease, and the risk of developing T2DM for non-diabetics. T2DM further puts individuals at even higher risk for macrovascular events as well as for the development of microvascular complications (including retinopathy, nephropathy, and neuropathy). These outcomes are serious and a major cause of premature mortality [21].

The IHE T2DM model

The economic simulation model of T2DM follows closely the structure of the seminal National Institute of Health (NIH) model [22]. Patients in two treatment arms transit separately between health states reflecting retinopathy, nephropathy, neuropathy, and cardiovascular disease (CVD) on an annual basis for a user-defined time horizon up to 50 years. The microvascular complication health states, the associated transition probabilities, and the relationship with HbA1c were based on the NIH model [22]. Risk equations for CVD (separately by myocardial infarction, ischemic heart disease, congestive heart failure, and stroke) and mortality were taken from the UKPDS [23]. Each health state is accompanied by a QoL utility weight. The default values were taken from CODE-2 [9], a large-scale study of 4,641 diabetic patients in five different European countries, including Sweden. Moreover, time-trade-off methodology was used, and BMI was included explicitly as a continuous variable. State (follow-up) and event (acute) costs are also attached to each of the health states. The default values are presented in Appendix Table 1. The costs of the key adverse events (anxiety, nausea, dizziness, insomnia, diarrhoea, mood alterations with depressive symptoms, vomiting, depressive disorders, contusion, and tendonitis), including medicine and physician consultation, were calculated as SEK 1,066 for Acomplia and SEK 573 for the comparator [24, 25].

Table 1 Diabetic complication treatment costs per event and annual follow-up costs, SEK in 2006 prices

Patient cohorts are defined at baseline according to age, gender, smoking status, BMI, systolic blood pressure (SBP), total cholesterol, high-density lipids (HDL), and HbA1C. Progression through the health states is linked to treatment via changes in a number of these patient characteristics. Specifically, the microvascular complication transition probabilities change with changes in the HbA1c level [22]. The risks of the four types of CVD depend additionally upon BMI, HbA1c, systolic blood pressure, and the ratio of total to HDL cholesterol [UKPDS 68 paper]. Treatment effects for Acomplia versus placebo were taken after 1 year of treatment from RIO Diabetes (see Appendix Table 2) [15]. Duration of treatment and duration of fade-out effect following discontinuation are user-defined.

Table 2 Change in CVD risk factors in RIO-Diabetes clinical trial for T2DM and RIO-Lipid clinical for dyslipidemia

The model outcomes include health benefits (life-years and QALYs) and costs (aggregated as well as disaggregated by treatment and complication type) over the user-specified time horizon for each treatment arm. The model then generates the incremental costs per LY gained and incremental cost per QALY gained. Both health benefits and costs are discounted at a user-defined rate.

The RAINBOW model

The economic simulation model used for overweight or obese patients initially free of T2DM was designed and constructed by Health Economics and Disease Management (HEDM) in Belgium. Patients in two treatment arms transit between three formal health states: (1) overweight or obese without having experienced any CVD event, (2) overweight or obese and having experienced one or more CVD event, and (3) death. Patients can become diabetic (or experience HbA1c normalisation), experience coronary or cerebrovascular events (fatal or non-fatal), and stop smoking during each 6-month cycle. The time horizon is user-defined up to 20 years. Macrovascular risks and microvascular risks, for those who became diabetic, were derived from a number of sources [23, 26–30].

Each patient is given a QoL utility during each cycle based on gender, BMI, transient ischemic attack, stroke, coronary heart disease, and microvascular disease. The default values used in the submission to LFN were taken from HODAR [10]. State and event costs were the same as described above.

Patient cohorts are defined at baseline according to age, gender, smoking status, BMI and/or waist circumference, SBP, percent diabetic, total and HDL cholesterol, triglycerides, and blood glucose (mg/dl). Progression through the model is linked to treatment via changes in BMI or waist circumference, SBP, and cholesterol. Treatment effects for Acomplia versus placebo were taken after 40 years of treatment from RIO-Lipids clinical trial, and they are presented in Appendix Table 2. Duration of treatment and duration of treatment fade-out effect following discontinuation are user-defined.

The model calculates incremental costs plus three measures of health benefits (life-years gained, QALYs gained, and CVD events avoided) for rimonabant versus the comparator. The model then generates ICERs for each outcome.