Cost-Effectiveness of Alirocumab

Abstract

Alirocumab is a fully human monoclonal antibody that inhibits proprotein convertase subtilisin/kexin 9 (PCSK9). It was approved in 2016 by the U.S. Food and Drug Administration for patients with heterozygous familial hypercholesterolemia or preexisting atherosclerotic cardiovascular disease who require additional lipid-lowering despite maximally tolerated doses of statin therapy. This approval was based on trials showing a 50% to 60% reduction in low-density lipoprotein cholesterol (LDL-C) levels (1). Since its approval, alirocumab has been available at a wholesale acquisition cost of more than $14000, a price that is not cost-effective on the basis of the cardiovascular benefit that may be expected from the amount of LDL-C lowering (2). In March 2018, the results of the ODYSSEY Outcomes (Evaluation of Cardiovascular Outcomes After an Acute Coronary Syndrome During Treatment With Alirocumab) trial were announced at the American College of Cardiology Annual Scientific Sessions (3). This was the first trial powered to evaluate the effect of alirocumab on cardiovascular events. The study participants had a history of acute coronary syndrome in the previous 4 to 52 weeks and an LDL-C level of 1.81 mmol/L (70 mg/dL) or greater despite statin therapy. The results established that when alirocumab was added to high-intensity statin therapy, at a median follow-up of 2.8 years participants had a 15% reduction in a composite outcome of nonfatal myocardial infarction (MI), ischemic stroke, hospitalization for unstable angina, and coronary heart disease death, as well as a 15% reduction in all-cause mortality (3, 4). Here, we report how we used these results to examine the cost-effectiveness of alirocumab in the population of U.S. patients with a recent history of acute coronary syndrome by projecting the lifetime incremental health gains, costs, and cost-effectiveness of adding alirocumab to high-intensity statin therapy. We performed this analysis initially to coincide with the public presentation of the trial results in March 2018, and we now confirm the analysis on the basis of the peer-reviewed article published by the ODYSSEY trialists in November 2018 (3, 4). Methods Model Overview The Cardiovascular Disease Policy Model is an established state-and-transition computer simulation program that projects the incidence, prevalence, and costs associated with coronary heart disease and stroke among U.S. adults aged 35 to 94 years (Supplement Figure 1) (2, 57). The model is programmed in Lahey Fortran 95; Monte Carlo simulations are programmed in Python (Python Software Foundation). Supplement. Supplementary Material For persons who have incident angina, MI, stroke, or cardiac arrest, the model projects clinical outcomes that include hospitalizations, revascularization procedures, and death; quality-adjusted survival; and direct medical costs associated with the event. In the population that survives the initial cardiovascular event, the model predicts recurrent cardiovascular events, quality-adjusted survival, and direct medical costs associated with inpatient and outpatient care up to age 95 years or death from any cause, whichever occurs first. The model adopts a health system perspective, capturing all direct health care costs and health benefits, regardless of who accrues them, and a lifetime analytic horizon, following all persons until they die or reach the age of 95 years, whichever is sooner. Costs and utilities are assigned to each clinical event and health state in annual cycles, and future costs and outcomes are discounted at 3% per year (Supplement Figure 1 and Supplement Tables 1 to 4) (8). We derived model inputs from epidemiologic studies, claims data, randomized trials, vital statistics, and the U.S. Census (Table 1) (3, 927), and calibrated the model to reproduce U.S. national data on MI, stroke, and death from cardiovascular causes or any cause in 2010 as observed in the National Hospital Discharge Survey and national vital statistics (Supplement Table 4) (22, 23). The institutional review board at the University of California, San Francisco, approved research with the model. This analysis was performed independent of the ODYSSEY Outcomes sponsor and academic steering committee. Table 1. Selected Input Parameters for the Cardiovascular Disease Policy Model Target Population For this analysis, we adapted the model to approximate the inclusion criteria for the ODYSSEY Outcomes trial. We modeled a cohort of U.S. adults aged 40 years and older who had an MI 1 to 12 months before enrollment and had an LDL-C level of 1.81 mmol/L (70 mg/dL) or greater despite statin therapy. We based the initial characteristics of this cohort on the 2005 to 2012 NHANES (National Health and Nutrition Examination Survey) (20). With regard to participants who met other inclusion criteria but were not receiving statins, we first modeled them as initiating statin therapy so that their mean LDL-C level equaled that of patients receiving statin therapy; then, we included persons in the simulation if their LDL-C level remained at 1.81 mmol/L (70 mg/dL) or higher despite statin therapy. Treatment Strategies We conducted our baseline analysis with patients who were receiving only a statin, as identified in the 2005 to 2012 NHANES (20). We modeled 2 additional treatment strategies: the addition of ezetimibe to statin therapy (ezetimibe is the recommended second-line agent for lipid lowering) and the addition of alirocumab to statin therapy. These approaches enabled us to compare a statin alone with either a statin plus ezetimibe or a statin plus alirocumab and to compare a statin plus alirocumab with a statin plus ezetimibe. Table 1 describes how we modeled the effect of adding ezetimibe or alirocumab to statin therapy. We assumed that any reduction in all-cause mortality was mediated through a reduction in the risk for death related to coronary heart disease or stroke. We also estimated that 3.8% of patients receiving alirocumab would have injection site reactions, which would produce a small quality-of-life penalty without an increase in costs or treatment discontinuation (3, 4). Cost and Quality-of-Life Estimates We estimated direct health care costs from U.S. national data (Table 1) and adjusted them to 2018 U.S. dollars by using the medical component of the Consumer Price Index (912, 28). In the base case, we assumed that annual drug costs were equal to the U.S. prices in 2018 for brand-name ezetimibe and for alirocumab net of rebates and discounts before the change in May 2018, but we varied these prices in sensitivity analyses. We used health-related quality-of-life weights for atherosclerotic cardiovascular disease states based on the Global Burden of Disease 2010 Study (1719). Outcomes Primary outcomes were the projected number of events averted in the population and the incremental cost-effectiveness ratio (ICER) measured in 2018 U.S. dollars per quality-adjusted life-year (QALY) gained over the lifetime analytic horizon. Secondary outcomes were the number of patients who would need to receive treatment for 5 years to avert 1 major adverse cardiovascular event (MACE, defined in this study as a composite of cardiovascular death, nonfatal MI, or nonfatal stroke) and the price of alirocumab at which it would become cost-effective (relative to a statin alone or to a statin plus ezetimibe) at a willingness-to-pay threshold of $100000 per QALY. We also estimated the change in total health care expenditures if all patients meeting the inclusion criteria of the ODYSSEY Outcomes trial received alirocumab. For this analysis, we assumed the perspective of a health plan whose beneficiaries are similar to the general U.S. population, and we estimated incremental health care spending over 5 years per 100000 beneficiaries aged 40 to 94 years. Sensitivity Analysis We performed deterministic and probabilistic sensitivity analyses (across the ranges shown in Table 1) to examine the effect of uncertainty in input parameters on model results. In 1-way sensitivity analyses, we varied 1 input parameter at a time, holding all other parameters at their base values. We substantially varied the drug prices, including lowering the cost of ezetimibe to the median U.S. price for available generic formulations. We also varied the discount rate for future costs and benefits (8). In addition, we identified the price at which alirocumab would be cost-effective at the conventional willingness-to-pay threshold of $100000 per QALY. We repeated this analysis for the subgroup of patients with a baseline LDL-C level above 2.59 mmol/L (100 mg/dL), assuming a higher baseline risk of events but an identical relative reduction in the risk for MACE (4). In probabilistic sensitivity analyses, we varied several input parameters across prespecified statistical distributions 1000 times to derive 95% uncertainty intervals (UIs). Statistical Analysis Outcomes were analyzed by using Python, QB64 (Microsoft), and Excel 2011 (Microsoft); statistical analyses were performed by using SAS, version 9.4 (SAS Institute), and R, version 3.4 (The R Foundation). Role of the Funding Source The funders had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript. Results Base-Case Analysis We modeled a population of 215000 U.S. adults who had an incident MI in the 1 to 12 months before enrollment and had an LDL-C level of 1.81 mmol/L (70 mg/dL) or greater despite statin therapy (Table 2). This population was 39.5% female and had a mean age of 67.3 years and a mean LDL-C level of 2.67 mmol/L (103.2 mg/dL). Approximately 30.9% had diabetes mellitus, and 74.7% had hypertension. The patients in this population had more comorbid conditions than those enrolled in the ODYSSEY Outcomes trial (3, 4). The rate of the composite end point was 6.2 per 100 patients per year in the model population