Improving the monitoring of chronic heart failure in Argentina: is the implantable pulmonary artery pressure with CardioMEMS Heart Failure System cost-effective?

Background The CardioMEMS® sensor is a wireless pulmonary artery pressure device used for monitoring symptomatic heart failure (HF). The use of CardioMEMS was associated with a reduction of hospitalizations of HF patients, but the acquisition cost could be high in low-and-middle income countries. Evidence of cost-effectiveness is needed to help decision-makers to allocate resources according to “value for money”. This study is aimed at estimating the cost-effectiveness of CardioMEMS used in HF patients from the third-party payer perspective -Social Security (SS) and Private Sector (PS)- in Argentina. Methods A Markov model was developed to estimate the cost-effectiveness of CardioMEMS versus usual medical care over a lifetime horizon. The model was applied to a hypothetical population of patients with HF functional class III with at least one hospitalization in the previous 12 months. The main outcome was the incremental cost-effectiveness ratio (ICER). To populate the model we retrieved clinical, epidemiological and utility parameters from the literature, whilst direct medical costs were estimated through a micro-costing approach (exchange rate USD 1 = ARS 76.95). Uncertainties in all parameters were assessed by deterministic, probabilistic and scenario sensitivity analysis. Results Compared with the usual medical care, CardioMEMS increased quality-adjusted life years (QALY) by 0.37 and increased costs per patient by ARS 1,081,703 for SS and ARS 919,051 for PS. The resultant ICER was ARS 2,937,756 per QALY and ARS 2,496,015 per QALY for SS and PS, respectively. ICER was most sensitive to the hazard ratio of HF hospital admission and the acquisition price of CardioMEMS. The probability that CardioMEMS is cost-effective at one (ARS 700,473), three (ARS 2,101,419,) and five (ARS 3,502,363) Gross Domestic Product per capita is 0.6, 17.9 and 64.1% for SS and 5.4, 33.3 and 73.2% for PS. Conclusions CardioMEMS was more effective and more costly than usual care in class III HF patients. Since in Argentina there is no current explicit threshold, the final decision to determine its cost-effectiveness will depend on the willingness-to-pay for QALYs in each health subsector.

The Markov model was applied to a hypothetical population of 1000 patients similar to those that met the inclusion criteria of the CHAMPION trial [21]. Inclusion criteria were patients with diagnosis of HF class functional (CF) III (according to the classi cation of the New York Heart Association (NYHA) [34]) regardless of the systolic ejection fraction, with at least one hospitalization for decompensation of HF during the previous 12 months.
The outcome of our model is the incremental cost-effectiveness ratio (ICER), which compares the difference in costs divided by the differences in Quality Adjusted Life Years (QALY) of CardioMEMS versus usual medical care. We considered cost-effectiveness thresholds from one (ARS 700,473) to ve (ARS 3,502,363) GDPs per capita per QALY gained, due to the potential heterogeneity in willingness to pay among the healthcare sub-sectors of Argentina, and no currently recommended explicit threshold. We applied an annual discount rate of 5% for costs and QALY, as recommended by economic evaluation guidelines for countries member of MERCOSUR [35], and we followed the CHEERS guidelines to report our ndings [36].

Model structure
To de ne the model structure, we reviewed specialized literature and economic models published previously [27][28][29][30]37]. Subsequently, we iterated with cardiology experts in order to validate the structure of the model to the local context. Our Markov model considered four states: "HF outpatient stable", "HF hospital admission", "Post HF hospital admission" and "Death" (Figure 1). There are two cohorts in the model, those who receive the CardioMEMS device (treatment cohort) and those who receive the usual medical care (control cohort). For both cohorts, HF patients transitioned through states in monthly cycles, accruing costs and QALYs. Before entering the model, the individuals in the treatment cohort accrue an initial incremental cost regarding the implantation of the device and its complications (including acute mortality). We included the eight CardioMEMS complications reported in the Manufacturer and User Facility Device Experience (MAUDE) database, which collects mandatory and voluntary reports of device-related malfunctions, injuries or deaths received by the FDA [38]. In each cohort, patients enter to model and may remain outpatient/stable or they may require a HF hospital admission or die. After a cycle, patients who entered the HF hospital admission state and are still alive enter the post HF hospital admission cycle and then return to an outpatient/stable cycle or require another hospital admission. Patients hospitalized and in the post HF hospital admission cycle have higher odds to die and accrue a different rate of QALY in comparison to patients who remain outpatient/stable.
Other assumptions were that the model did not consider non-HF hospital admission, because we assume they were equal in both cohorts. In addition, no adverse events associated with the pharmacological treatment received were considered, since the medication scheme was assumed the same in both cohorts.

Epidemiological parameters
All the epidemiological parameters were retrieved by performing a literature search of the evidence regarding the patients and the intervention of study. We select the evidence that best ts and represents the local context. Epidemiological parameters are reported in Table 1.
To estimate the global mortality rate of the patients included into the model we used both life tables by age of the Argentine population and mortality reported in CF III HF patients with CardioMEMS. The mortality at 12 month of follow-up considered was 16% (95% CI: 14% to 18%), based on the largest study published with the device: a multi-center, prospective, open-label, observational, single-arm trial, that assessed the e cacy and safety of CardioMEMS with 1200 patients [39]. Therefore, the model was calibrated through an intermediate correction factor that represents the excess of risk associated with HF in Argentina. The resulting correction factor, equal to 7.7, means that patients with HF CF III have an increased relative risk to die of 7.7 in comparison to those patients without HF. The baseline mean age of patients entering the model was based on the weighted age of patients with chronic HF reported in the Heart Failure Registries in Argentina [40]. We performed a meta-analysis to obtain the weighted mean baseline risk of HF hospital admission, including the CHAMPION trial [21] and two multi-center, prospective, open-label, observational single-arm trials that evaluated the e cacy and safety of CardioMEMS in patients with HF class III, in the USA and Europe [39,41].

Treatment effect of CardioMEMS
Due to the uncertainty of CardioMEMS regarding the direct effect on mortality, and since mortality in patients with HF increases during hospitalization and the subsequent month [42,43], we indirectly modelled the effect on mortality through the decrease in the probability of HF hospital admission. We incorporated an increased risk of mortality [44] for HF hospital admission state and post HF hospital admission state (Table 1), a similar approach adopted by others [37].
CardioMEMS effect on HF hospital admission was modeled using the hazard ratio obtained in our meta-analysis, that included 3 studies [21,39,41]. For the base case analysis, these bene ts were assumed to last ve years, and disappeared after this period. This conservative decision was made due to the uncertainty regarding the long-term effects of the device, similarly to a previous economic evaluation [30].
Health-related quality of life Data regarding baseline health-related quality of life (HRQoL) at 6 and 12 months were based on the EuroQol Quality of Life Five Dimensions instrument (EQ-5D) [45] and were collected in the CHAMPION trial [21]. Based on this data, the monthly change of utility for month 1 to 6 and for month 7 to 18 were estimated for both cohorts. In the model, we assumed that the effect of CardioMEMS on HRQoL at 12 month is carried forward to the 18 month, equal to the follow-up time of the CHAMPION trial. After 18 month, the monthly change of utility of the 7 to 18 month was carried forward to the 60-month model in both cohorts. After the 60-month model, we assumed that the differences in utilities disappeared between both cohorts. The impact of HF hospital admission in utilities (or disutilities) was based on Schmier et al [28]. We compared the Minnesota Quality of Life Questionnaire scores reported by CHAMPION trial with these reported in a local study [46], and the values were similar, suggesting EQ-5D scores could be applicable to Argentina.

Direct Medical Costs
We applied a micro-costing approach to estimate the direct medical costs for a third-party payer perspective (SS and PS). Identi cation and quanti cation of healthcare resources were made by a local literature review, validated by local experts' consultation, whilst unit cost estimations were made using the Healthcare Cost Database of the Institute for Clinical Effectiveness and Health Policy [47]. This database contains unit cost information based on the tariffs of medical resources and practices for SS and PS subsectors. Direct medical costs were updated to September 2020 and were expressed in Argentinian pesos (ARS) (exchange rate USD 1 = ARS 76.95).
The CardioMEMS device acquisition price was provided by the manufacturer and was ARS 1,398,182. The cost of CardioMEMS implantation was assumed to be equivalent to the cost of the right heart catheterization and angiography, a similar assumption made in another economic evaluation [27].

Sensitivity analysis
Uncertainty was assessed with one-way deterministic sensitivity analysis (Table 1). Due to the heterogeneity in the HF population, we assessed the baseline mortality risk, the baseline risk of hospital admission, the average age of HF population, and we also included the acquisition cost of the device. A probabilistic sensitivity analysis was performed using 1000 Monte Carlo simulations, in which, with each simulation we sampled from the distributions of each input parameter. The range of values incorporated in these analyses is shown in Table 1.

Scenario analysis
We assessed the uncertainty of CardioMEMS effect on patients through two analyses of alternative scenarios. In the rst scenario, the treatment effect of CardioMEMS was extended to a lifetime horizon. In the second scenario, the effect of CardioMEMS until 18 months is the same as in the base-case, and from this point, the effect progressively declines until the fth year, when it nally disappears.

Base case results
The parameters used in the model are reported in Table 1. In the base case analysis (lifetime horizon) with a discount rate of 5%, CardioMEMS HF system increased QALYs in the treatment cohort in comparison to the control cohort by 0.37. CardioMEMS also increased costs compared with usual medical care by ARS 1,081,703 for SS and ARS 919,051 for PS (  Figure 2 shows the one-way sensitivity analyses performed for both the SS and PS perspective. The ICER was most sensitive to the HR of HF hospital admission obtained in our meta-analysis, the acquisition cost of CardioMEMS and the mortality due to HF hospital admission, among other parameters.

Sensitivity analysis
Reducing the HR of HF hospital admission to 0.38 (the base HR reported in Angermann et al [41]) resulted in an ICER of ARS $2,225,077 and ARS 1,728,223 per QALY for SS and PS respectively. On the other hand, increasing the HR to 0.67 (the HR reported in the CHAMPION trial [21]) increases the ICER to ARS 4,685,811 and ARS 4,382,352 per QALY for SS and PS respectively. The ICER was also sensitive to the acquisition cost of the device. When we reduced the acquisition cost of CardioMEMS by 25%, ICER resulted in ARS 1,924,020 per QALY for SS and ARS 1,457,693 for PS. A discount of 10% in the acquisition cost of CardioMEMS yields an ICER below 3 GDP per capita.
Another parameter that in uenced the ICER result was HR of risk of mortality due to HF hospital admission (varied between 1 and 5). The resulting ICER varied from ARS 2,683,044 to ARS 4,235,255 per QALY for SS and varied from ARS2,379,068 to ARS 3,193,044 per QALY for PS. Lastly, when we varied the utility of CardioMEMS by its standard deviation, the ICER varied from ARS 2,503,120 to ARS 3,555,040 per QALY for SS and varied from ARS 2,126,734 to ARS We also assessed the uncertainty of CardioMEMS effect on patients by performing two scenario analysis. Under the rst scenario, in which the CardioMEMS effect is lifelong lasting, the resulting ICER was ARS 1,804,312 per QALY for SS and was ARS 1,407,515 per QALY for PS. Under the second scenario in which the effect of CardioMEMS lasts 18 months and henceforth the effect progressively declines until year ve when it disappears, the resulting ICER was ARS 3,966,795 per QALY for SS and was ARS3,484,989 per QALY for PS. Figure 3 shows the results of the probabilistic sensitivity analysis. For SS, all the points in the scatter plot fall in the right upper quadrant, con rming the high certainty that CardioMEMS is both more effective and more expensive than the usual medical care. These ndings are quite similar for the PS sector. For both sectors, at one GDP per capita as a cost-effectiveness threshold almost the totality of the points are above the threshold. On the other hand, at ve GDP per capita as a cost-effectiveness threshold, the majority of the points in both sectors fall below the curve.

Probabilistic analysis
Due to the potential heterogeneity in willingness-to-pay among different third-party payers from the same sectors in Argentina, we assessed several thresholds of willingness to pay. For the SS sector, the probability that CardioMEMS be cost-effective at one (ARS 700,473), three (ARS 2,101,419,) and ve (ARS 3,502,363) GDP per capita is 0.6%, 17.9% and 64.1%. For the PS sector, the probabilities are 5.4%, 33.3% and 73.2% respectively (Figure 4).

Discussion
In contexts where resources for healthcare are scarce, health technologies assessments and economic evaluations are useful tools to evaluate the e cacy, safety, e ciency and affordability of new healthcare technologies. From the e cacy and safety evidence of CardioMEMS HF system, this study assessed the cost-effectiveness of CardioMEMS use in the Argentinian context. The analysis showed that base case ICER was between three and ve GDP per capita per QALY gained. Although Argentina does not have an explicit cost-effectiveness threshold, this result could be above the e ciency threshold suggested by the literature [35,36]. However, considering that the bene t population are patients with advanced-stage HF and poor prognosis, the decision rule to determine cost-effectiveness of the device will depend on the willingness to pay for QALYs gained from each speci c third-party payer in Argentina.
Evidence of cost-effectiveness analysis suggests that CardioMEMS is cost-effective in the USA and UK settings, although ICER ranges widely due to different time horizons adopted, assumptions made in the effects of CardioMEMS on mortality, among others methodological aspects [27][28][29][30]37]. In our estimations, survival years, QALYs and HF hospital admissions averted are consistent with the estimations made by studies that use lifetime horizons [27,37].
In our model, among all parameters, ICER estimates were most sensitive to the HR of HF hospital admission. In the base case, to synthesize all the available evidence regarding the effectiveness of CardioMEMS on the hospital admission, we performed a meta-analysis using the pivotal trial [21] and two real-world evidence studies [39,41]. In the one-way sensitive analysis we assessed the variability of this parameter by using the HR reported in Angermann et al [41] and the HR reported in the CHAMPION trial [21], as they represent the extreme values in the available literature. In addition, we assessed all the treatment effects of CardioMEMS by performing two analyses of scenarios. In the best scenario where the treatment effect of CardioMEMS is lifelong, ICER falls below the 3 GDP per capita willingness-to-pay, but in the worst scenario where the treatment effect declines from 18 month and it equal for both cohorts in the month 60, ICER exceeds the 5 GDP per capita willingness-to-pay. We consider that our model incorporates a credible scenario for the base case, given that it is not possible to know whether the bene t in avoiding hospitalizations would continue as the patients worsen given that it is a progressive disease. [50][51][52].
The results of the CHAMPION trial regarding the treatment effect of CardioMEMS on mortality is unclear [21], and studies of non-invasive remote monitoring systems have been neutral regarding the potential reduction in mortality [53,54]. Due to this uncertainty, our approach in the modelling was to capture this treatment effect indirectly from the increased risk of mortality during and one month later the HF hospital admission. This approach is more conservative than the used in other economic evaluations [28][29][30], but is similar to the approach adopted in an economic evaluation performed in the USA [37].
Our ndings could have implications for CardioMEMS pricing policies. As the largest costs of the model are driven by the acquisition cost of CardioMEMS, we varied this parameter by +/-25%. In the lowest price, ICER improves and falls below the 3 GDP per capita willingness-to-pay threshold on social security and a discount of 10% turns the ICER on 3 GDP in the private sector. This information together with budget impact analysis, is helpful to design pricing policies in the different health sector in Argentina.
This study has some limitations to note. First, all the effectiveness and safety of CardioMEMS comes from a CHAMPION trial [21] and two real world studies [39,41] that assessed effectiveness and safety at one year post implantation of CardioMEMS, thus the long-term e cacy is still unclear. To reduce the uncertainty, we performed a meta-analysis to summarize the best evidence available at the moment. In the near future, this meta-analysis should be updated as more post-surveillance evaluations become available. Second, the third-party payer perspective of this study prevented us from evaluating all the possible bene ts and costs at a societal level. For example, reductions in the rate of HF hospital admissions could reduce indirect cost for patients (labor productivity loss costs, out-of-pocket expenditures) and favors medical attention to other HF and non-HF patients by reducing waiting time queue. Third, our model does not include some indirect saving aspects that can improve the ICER but there is lack of information. For example, the time spent monitoring CardioMEMS, although not in uencing ICER results, is an important aspect to consider. CardioMEMS HF system seems plausible to require less time to monitor the device in comparison to other less advanced technologies such as telephone calls, anamnesis on body weight, diuretic rhythm and symptoms of dyspnea and fatigue. Fourth, we used a linear trajectory of utilities or disutilities; however, the relationship between patient-reported utilities and HF hospital admission may be no linear since patients with multiple HF hospital admissions during an interval of time can report higher disutilities. Unfortunately, there is no data available that describe the utilities of HF patients regarding hospital admissions, but in our sensitivity analysis we widely assessed this parameter and did not show greater impact on ICER. Given the substantial societal and economic toll of HF, it is worthy to consider the previous aspects as soon as they become available.
Despite these limitations, this study gives a clear snapshot about the cost-effectiveness of CardioMEMS in Argentina, underpin the nding with a model adapted to the local setting and using a nationally representative costs database to perform a micro-costing approach to estimate costs of health events. Furthermore, sensibility analysis was useful to examine multiple variability of ICER, thereby ndings here presented can promote the use of cost-effectiveness evidence in HF management strategy adoption at a national level. Future studies that re ne estimates of the long-term effects of the device on mortality could reduce uncertainty and improve conclusions regarding its clinical and economic impact, contributing to informed healthcare decision-making.

Conclusion
CardioMEMS was more effective and more costly than usual care in class III HF patients in Argentina, being the ICER between three and ve GDP per capita per QALY gained. The decision rule to determine the cost-effectiveness of the device will depend on the speci c willingness to pay for QALY gained from each healthcare subsector.

Declarations
Ethics approval and consent to participate. Not applicable.

Consent for publication.
Not applicable.
Availability of data and materials.
The datasets analyzed in this study are available from the corresponding author on reasonable request.

Competing interest.
This study was made possible by the support of Abbott Argentina through a research grant. The contents are the responsibility of the authors.

Funding.
This work was supported by Abbott Argentina. The sponsor of the study had no role in the study design, data collection, data analysis, data interpretation or writing the manuscript. The corresponding author had full access to all the data. Authors´ contributions.
AA coordinated the project, advised on all aspects and co-write the nal draft. AP, FA and APR advised on methods and all statistical aspects. JMG extracted and cleaned all the clinical, effectiveness, epidemiological and utility data and advised on all clinical aspects. DP constructed and implemented the costeffectiveness model and performed time to event analysis. CRR extracted and estimated all the costs data and write the nal draft. All authors contributed to the interpretation of the results and contributed to edit the nal draft.

Strategy
Per patient cumulative costs (ARS)  Figure 1 Analytic structure of the model