Cost Effectiveness of Inclisiran in Atherosclerotic Cardiovascular Patients with Elevated Low-Density Lipoprotein Cholesterol Despite Statin Use: A Threshold Analysis

The base-case modeled the cost effectiveness of inclisiran in US patients with ASCVD and LDL-C ≥ 70 mg/dL, despite maximum tolerated doses of statins (with or without ezetimibe) (ORION-10 population) [18]. Patient characteristics were based on ORION-10 (Table 1).

Model results are based on a weighted average population, meaning the ASCVD population was stratified into subpopulations representing events that constitute ASCVD, and further stratified by time from previous event. Each subpopulation passed through the model separately, with results weighted by the real-world distribution of these subpopulations to derive population estimates of cost effectiveness.

A lifetime Markov model was developed, including 15 mutually exclusive states (Fig. 1). Patients entered the model in any of three initial states, based on time from previous ASCVD event (0–1; 1–2; 2+ years) [19]. Three non-fatal CV (NF-CV) event states were included, i.e. UA, MI, and stroke, stratified based on time post-event (0–1; 1–2; 2+ years). Other states included revascularization and death (CV- or non-CV-related). Cycle length was 1 year, with half-cycle correction implemented based on the life-table method [20].

The model allowed annual transitions based on risks of CV events (fatal and NF) and non-CV death. Following movement to a post NF-CV event state, patients remained at risk of subsequent events (fatal or NF). Patients only moved health states when a ‘worse’ event occurred, to avoid illogical outcomes leading to quality-of-life improvement (events in order of severity are stroke, MI, UA, revascularization). Less severe events were captured as acute events, with one-off quality-adjusted life-year (QALY) loss and costs applied.

Costs and QALYs were discounted at 3% annually [21]. The model assumed the payer perspective, considering all direct health care costs, and the cost of inclisiran without any potential rebates/discounts (see electronic supplementary material [ESM] e1 for more model structure details).

US real-world data from Optum Clinformatics^® Data Mart (a retrospective cohort study of patients with ASCVD from 2015 to 2020) informed baseline NF-CV event probabilities (see ESM e2a) [22]. The distribution across the various subpopulations from the Optum data (Table 1) was used to calculate weighted average model results. Each subpopulation entered the model in the relevant health state, e.g. 100% of the stroke 0–1 population would start in the initial state (year 0–1).

Fatal event rates were taken from published literature, as mortality was not recorded in the Optum database (Table 1 and ESM e2b).

Baseline risk of cardiovascular events required adjustment because they represented a population whose age and LDL-C level (eTable 1 in the ESM) did not match the modeled population (see ESM e2c/e2d).

Non-CV mortality was estimated from the Centers for Disease Control and Prevention [23]. CV mortality was identified using the CVD International Classification for Diseases, Tenth Revision (ICD-10) code as per the American Heart Association definition [24]. The proportion of non-CV-related deaths was applied to the latest US life-table data (2017). This was calculated separately for men and women, and weighted by the modeled gender split.

The comparator was SoC, represented by the ORION-10 placebo arm. The distribution of SoC treatments were from ORION-10, split into eight mutually exclusive categories; those taking or not taking ezetimibe were split into those not taking statins or those taking low/moderate/high-intensity statins. Baseline LDL-C associated with each stratification was from ORION-10, allowing calculation of an overall SoC weighted average LDL-C, to enable LDL-C to be recalculated each cycle as people discontinue treatments.

Inclisiran was modeled adjunctive to SoC. LDL-C reduction with inclisiran was from ORION-10, based on the mean difference in LDL-C reduction between arms at day 510 (52.3%) [18]. This LDL-C reduction was applied to the baseline (SoC) LDL-C to determine the LDL-C level on inclisiran.

A link has been demonstrated between exposure time and treatment effect, with observed rate ratios per mmol/L reduction in LDL-C being smaller in the first year of treatment [25]. An adjustment was applied to reflect this (see ESM e3).

Patients who discontinued statins reverted to the LDL-C of patients receiving no statins at baseline in ORION-10, likewise for ezetimibe. Statin and ezetimibe discontinuation rates were assumed to be 23% per year [26]. A rate ratio of 0.5 versus statin discontinuation was applied to determine inclisiran discontinuation (see ESM e4) [27]. A discontinuation cap was included to prevent patients remaining on treatment falling indefinitely to zero; this was 20% for statins and inclisiran and 5% for ezetimibe (Campbell C, unpublished data, January 2022).

Adverse events were excluded, as the ORION program showed inclisiran had a similar adverse event profile to placebo [18]. The exception was injection-site reactions, which clinical experts suggested would not be associated with significant cost or health-related quality of life (HRQoL) impact, as these were found to be mild and did not persist in the ORION trials [18]. This is consistent with other CV risk reduction models [19]. However, a disutility associated with having injections themselves, has been included, as literature showed that treatment modality can have an effect on health state utility [28].

For SoC, the average cost of 10 mg and 20 mg simvastatin represented both low- and medium-intensity statins, while 40 mg atorvastatin represented high-intensity statins. A weighted average cost of SoC per year was calculated, given the distribution of ORION-10 patients using each SoC component (Table 1).

All patients with ASCVD, regardless of therapy type, should receive, at minimum, one outpatient appointment per year (not included in the analysis [11]). Therefore, with two doses per year, one outpatient appointment and two injection procedures were costed annually. An additional appointment and injection procedure cost were included in year 1 to account for the additional dose in the first year. Zero administration costs for SoC were assumed.

The costs for UA, MI, stroke, revascularization events [29], and CV death [30] were from published literature and are inflated to 2021 US dollars using the medical consumer price index (Table 1) [31]. Acute event costs were applied as one-off costs to acute events less severe than the patient’s current health state.

Population age-adjusted utilities for patients with no history of CVD were based on the regression from Bhatt et al., [30] which analyzed patient-level data from the alirocumab ODYSSEY study and identified baseline age-adjusted utilities (controlling for age, sex, baseline EQ-5D-3L utility, CV history, and diabetes at baseline). Utility multipliers to represent baseline subpopulations were obtained from previous cost-effectiveness models in this disease area [32‐34], and applied to population age-adjusted utilities.

Disutilities from CV events (Table 1) were applied for all health states in all years, except for revascularization, where disutility was assumed non-permanent and only applied to the cycle where the event occurred [30]. For subsequent CV events, for the same or worse event, the patient transitioned to the more severe state and the disutility from the second event was applied as the patient proceeded through the model; for a less severe event, a one-off short-term disutility was applied [30].

An annual disutility associated with having an injection has been included, i.e. − 0.00150 [35], based on a study investigating utilities associated with attributes of migraine prevention treatments, including route of administration (see ESM e5).

The primary outcome was the price at which inclisiran is cost effective over the lifetime analytic horizon, at WTP thresholds of $50,000, $100,000, and $150,000 per QALY [36].

An incremental cost-effectiveness ratio (ICER) is based on the difference in costs between treatments divided by the difference in QALYs (to give the cost per additional QALY) [Eq. 1].

A threshold analysis rearranges this equation and finds the total incremental cost (by varying the cost of inclisiran) that would lead to an ICER at a particular threshold (Eq. 2).

One-way sensitivity analysis was undertaken, varying parameters within their 95% CI, or ± 15%, where this was not available.

Several prespecified scenario analyses were also performed (Table 2), including a scenario that uses the public list price of inclisiran [37].

Joint parameter uncertainty was explored through probabilistic sensitivity analysis (PSA), where all parameters were assigned distributions and varied jointly.

Table 1 presents the range and type of distribution utilized for each input parameter. A total of 1000 Monte Carlo simulations were recorded as results had stabilized by this point, with variation of < 0.01% between simulations. PSA was undertaken at various annual prices of inclisiran. Cost-effectiveness acceptability curves (CEAC) were generated, with the proportions of simulations where inclisiran is optimal (at each WTP threshold) plotted as a function of the price of inclisiran.

The model was constructed in Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA). Statistical analyses of event rate data were undertaken using SAS (see the ESM for model validation details [Sect. e6]).

Of the ASCVD patients analyzed, around 90% were taking statins and 10% were taking ezetimibe. Average LDL-C was 104.97 mg/dL and average age was 66 years, with 45% having diabetes, and 31% were female. Independently of the inclisiran price modeled, inclisiran as an adjunct to SoC produces an additional 0.43 QALYs and incremental costs (excluding the cost of inclisiran) of $104,471. The annual cost of inclisiran at each WTP threshold of $50,000, $100,000, and $150,000 per QALY was found to be $6383, $9973, and $13,563, respectively (price per dose of $3191, $4987, and $6782, respectively), leading to incremental costs of $21,498, $42,990, and $64,482, respectively. Table 3 presents the base-case analysis results in the weighted average ASCVD population.

The parameters with the most impact on the ICER at each threshold price were the rate ratios used for the relationship between LDL-C reduction and reduction in CV events, and the efficacy of inclisiran. The tornado diagrams at each threshold are shown in eFig. 4, eFig. 5, and eFig. 6 in the ESM.

PSA results as a function of the price of inclisiran are presented in Fig. 2. At each WTP threshold, inclisiran had 100% probability of being cost effective if the annual price is below $6000, $9000, or $12,000, respectively. Cost-effectiveness planes at each threshold are available in the ESM (e7b).

The scenarios (see eTable 4 in the ESM) that resulted in more favorable threshold prices for inclisiran than the base-case resulted when inclisiran efficacy was increased. The scenarios that resulted in less favorable threshold prices for inclisiran than the base-case were assuming the minimum proportion remaining on any of the treatments was 40%, having no discontinuation, and having alternative health state costs.

The scenario where the publicly available price of inclisiran was used led to an ICER just above the $50,000 per QALY threshold, of $51,686.

The strengths of this analysis include the use of real-world estimates of baseline event rates, which greatly improved generalizability to a US population. Although the analysis was considered representative of NF event probabilities in a US ASCVD population, it did not inform the probability of CV death, where additional literature was sourced.

The principal limitation of this analysis was the reliance on predicted event risk reduction based on observed reductions in LDL-C. The total number of CV events observed in the ORION program were underpowered to draw meaningful conclusions regarding the benefits of inclisiran on CV outcomes [38]. Although a recognized limitation, the relationship applied from the 2019 Cholesterol Treatment Trialists (CTT) analysis represents the largest study to investigate the relationship between LDL-C reduction and event risk reduction [39]. Previous CTT versions have been used in the economic evaluation of inclisiran in other countries [38] and other therapies in this disease area [19]. This uncertainty will ultimately be addressed with the publication of the results of ORION-4, inclisiran’s CV outcomes trial.

The model utilized the results of a meta-analysis for the percentage reduction in LDL-C from inclisiran versus placebo. Treatment effects from ORION-10 were based on intention-to-treat outcomes and therefore implicitly contain the effects of discontinuation observed in the individual clinical trials. As the model also includes discontinuation as an input, then the treatment effect for inclisiran can be considered conservative.

A cost-effectiveness analysis of inclisiran plus statins versus statins alone from an Australian healthcare perspective (Markov model, based on ORION-10) reported an ICER of AU$125,732 ($87,325.35) [40] per QALY, using an inclisiran price of AU$6334 ($4399). Differences in results may be explained by lower baseline event rates from the Australian analysis (based on the 2010 CTT publication [41], in which only 41% of patients had ASCVD) and a slightly different model structure (e.g. the exclusion of stroke).

The US Institute for Clinical and Economic Review (US-ICER) also assessed the comparative clinical effectiveness and value of inclisiran and found a cost per QALY of $157,000 [42]. Key differences in the US-ICER’s methodology include slightly different health states, including a composite acute coronary syndrome and stroke health state; different population and baseline LDL-C; no discontinuation was included; and different sources for the relationships between 1 mmol/L reduction in LDL-C and reduction in CV events (using the 2010 CTT analysis [41], although an updated 2019 version was available [39]). Specifically, the US-ICER derived baseline LDL-C by identifying population characteristics from NHANES for patients with LDL-C ≥ 70 mg/dL and taking statins. For those not taking ezetimibe (95.8%), baseline LDL-C was downwardly adjusted to reflect hypothetical ezetimibe use, and those with an LDL-C still ≥ 70 mg/dL were used to determine the average baseline LDL-C in the model (calculated as 88.8 mg/dL; by contrast, ORION-10 found a baseline LDL-C of 104.97 mg/dL). This methodology assumed inclisiran will be used by a population whose LDL-C remains ≥ 70 mg/dL after trying both statins and ezetimibe. This represents only a subpopulation of patients for whom inclisiran has been studied and does not reflect the current treatment pathway (as shown by the US-ICER NHANES analysis). It also assumes patients are fully compliant with ezetimibe. As a consequence of lower baseline LDL-C, and therefore lower baseline risks, the estimates produced by ICER are likely to be a more conservative estimate of the cost effectiveness of inclisiran (see ESM e8) and less likely to represent the cost-effective profile of inclisiran in clinical practice.

Cost-effectiveness analyses of PCSK9 inhibitors have demonstrated a variety of results, with ICERs ranging from $56,655 to $503,000 [30, 33, 34, 43, 44]. This may be attributed to factors such as a change in price of the interventions over time, different model structures, different approaches to outcomes (CV endpoints from trials or surrogate endpoints of LDL-C reduction being linked to CV outcomes), and different compositions of SoC (including inclusion or exclusion of ezetimibe).