A Markov model has been selected for this economic evaluation and developed to project quality-adjusted life years (QALYs) and costs associated with Impella and VA-ECMO for managing patients with CS (as defined by INTERMACS Class 1-2-3 or SCAI Class C-D-E [14]). The analysis has been reported according to the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) (see Supplementary Material) [15]. Ethics approval and informed consent were not needed for this study because the data used for the analyses were derived exclusively from studies published in the literature.

The health states considered in the implemented model were: (i) “CS”, (ii) “Adverse events”, and (iii) “Death for disease” (Fig. 1). Adverse events that include stroke, bleeding, limb ischemia, and renal failure, all directly linked with healthcare resource consumption (and consequently costs), have been considered as part of the current economic evaluation. To ensure outcome comparability, only access-site bleeding and major bleeding (or equivalent) were considered. A major bleeding refers to a significant loss of blood, generally requiring transfusion, that can result in serious complications, including shock or death, if not treated promptly.

Representation of the model (CS cardiogenic shock)

The model considers a hypothetical cohort of patients with CS with a mean age of 59 years, as reported in the literature [16] (see next paragraph “Comparative clinical data”). They start the Markov process in the “CS” state. Patients may stay in this state or, in case of adverse events, may move to the “Adverse events” state. It has been assumed that adverse events happen in the first 30 days of treatment with Impella or VA-ECMO, considering that most of the studies used in a meta-analysis by Ardito et al. [16] referred to the hospitalization period. An exponential function has been applied for overall survival data emerging from the same meta-analysis [16]. The model also takes into account background mortality according to Italian life tables [17]. The model calculations ensure that the overall mortality is always higher than, or equal to, the background mortality for the Italian population. A cycle length of 30 days and a lifetime horizon have been chosen for the baseline analysis. A discount rate of 3% has been applied to costs and health outcomes [18]. The model has been implemented using Microsoft Excel.

This study is a literature-based CEA. A recently published systematic literature review with meta-analyses has been used as the source of data for the comparison of Impella versus VA-ECMO used as a standalone device (not in combination) in the treatment of CS [16]. This study considered randomized controlled trials (RCTs), prospective clinical trials and observational studies. The study selection criteria for the meta-analyses were designed to maximize the homogeneity of device usage characteristics, specifically including studies where a single device was used for managing adult patients with cardiogenic shock between 2017 and 2022. This approach aimed to minimize biases related to potential technological advancements over time. The population considered in the systematic literature review had an average age of 59 years and was mainly composed of male patients (70.8%). Mortality at 1 year was reported as 48% and 52% for Impella and VA-ECMO, respectively. Major and access site bleedings during the hospitalization totaled 19% for Impella and 23% for VA-ECMO, while limb ischemia was present in 6% of patients treated with Impella and 10% of patients treated with VA-ECMO. Finally, ischemic stroke during hospitalization was more likely to occur in patients treated with VA-ECMO (6%) than in those treated with Impella (2%). Likewise, renal failure events were slightly more frequent in the populations treated with VA-ECMO (38%) compared to Impella (34%).

Regarding the possibility of device malfunction, the DanGer Shock trial [10], which compared Impella to medical treatment in patients with cardiogenic shock CS, reported a device malfunction rate of 1.2%. This rate was incorporated into the model, including the associated re-implantation cost. To maintain a conservative approach, failure rates for VA-ECMO were not included. Table 1 summarizes the model inputs.

Direct healthcare resource consumption (direct costs) was accounted for in the model (Euros, 2024). To reflect the Italian NHS perspective, cost data were derived from national diagnosis-related group (DRG) reimbursement rates for the utilization of the devices (Table 2). Since the frequencies of complications considered in the model pertain to the hospitalization period for managing CS, no additional DRG reimbursements needed to be applied, since the cost for their management is included in either Impella or VA-ECMO DRGs.

From the hospital perspective, costs were based on the cost function linked to the different production factors. Since the IRCCS (Istituto di Ricovero e Cura a Carattere Scientifico—Italian acronym for research hospitals) San Raffaele Scientific Institute in Milan is a tertiary care center for CS in Italy, the hospital costs of this clinical center have been considered a proxy for hospital costs in the Italian setting. In particular, for the management of the different adverse events, a list of activities and healthcare resources based on the clinical practice were identified by the clinicians of the center (MP, MS) (Table 3), based on clinical record charts review. Therefore, costs derived from the administrative office of San Raffaele Scientific Institute were applied.

To consider long-term costs for the management of patients with stroke or renal failure in Italy, we referred to the published literature. In particular, costs of 1420€, 358€ and 185€ (uplifted to year 2024) were retrieved for the monthly post-discharge follow-up of patients with stroke in the periods 1–3 months, 4–6 months and 7–12 months, respectively [22]. Regarding renal failure, from the Italian study by Ingrasciotta and colleagues [23], the monthly management cost of 4610€ (uplifted to 2024) has been retrieved. These follow-up costs have been considered for both NHS and hospital perspectives due to the lack of more specific data. For the NHS perspective, only direct costs borne by the NHS were included (and not societal or informal costs).

Regarding the devices, the model accommodated the use of either Impella CP® heart pump or Impella 5.5® heart pump, with associated list prices for the Italian market provided by the manufacturer (Impella CP: €17,500 + 4% VAT; Impella 5.5: €37,000 + 4% VAT). The usage percentage of the Impella CP (90%) was based on 5-year extrapolations from data reported in the study by Chieffo and colleagues [24], which reported statistics on the multicenter observational Italian registry IMP-IT for the management of patients with CS; considering that versions Impella 2.5® heart pump and Impella 5.0® heart pump are currently out of market, the remaining market share was assigned to the Impella 5.5 (10%). Since the implantation of the Impella 5.5 is performed in the operating room, an additional cost of 1431.76€ was considered for this type of device.

The hospital cost for VA-ECMO has been estimated as the mean cost per patient reported in a few public tenders in Italian Regions. This cost is composed by a component for consumables (9463€ per patient) and a fixed amount for the device (2466€ per patient), for a total cost per patient of 11,930€. These costs are summarized in Table 4.

Studies reporting data on patients’ quality of life have been searched in the literature as well, and three relevant sources were retrieved. The study by Roos et al. [19] reported a utility coefficient of 0.67 for the CS state and utilities of 0.31 for stroke and 0.30 for major bleeding. Another study reported a disutility for limb ischemia of 0.059 [21]. For renal failure a utility value of 0.66 was applied [20] to the CS state, leading to a value of 0.44. Disutilities were applied for the duration of events, estimated in 30 days according to the duration of hospitalization. All the model inputs are summarized in Table 1.

Costs and QALYs for the strategies considered were estimated to calculate the incremental cost-effectiveness ratio (ICER) and incremental cost-utility ratio (ICUR) of Impella versus VA-ECMO. In Italy, willingness-to-pay (WTP) thresholds vary between 25,000€ and 60,000€ [25, 26], therefore in the context of the present analysis an intermediate WTP of 50,000€/QALY has been applied.

Model parameters were incorporated along with specific probability distributions: a beta distribution for utilities and the proportions of patients experiencing adverse events, and a gamma distribution for costs. Variations in model parameters were based on 95% confidence intervals, standard deviations, or ranges reported in meta-analyses [16] and other reference studies. For parameters without reported variation information, a ± 50% variation from the baseline value was applied.

To assess the robustness of the model results, both deterministic sensitivity analysis (DSA) and probabilistic sensitivity analysis (PSA) were conducted. In PSA, second-order Monte Carlo simulations (10,000 iterations) were performed to account for parameter variability. Additionally, one-way sensitivity analyses were carried out using the same parameter variations as in PSA, except for the discount rate, which was varied from 0 to 10%.