a.

Study Design

This study used a Markov model to extrapolate the results of the RADIANCE-HTN TRIO trial, which included patients aged 18–75 years with an office BP ≥ 140/90 mm Hg despite three or more antihypertensive medications including a diuretic [9]. All patients in the trial were treated with a single-pill triple antihypertensive combination therapy prior to their allocation to uRDN or sham control. Key baseline characteristics of the trial population have been previously reported [9] and are shown in eTable1. This economic evaluation is reported in accordance with the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) [18] and applied the perspectives of the three health care systems of Belgium, France and the Netherlands.

b.

Model Structure

Consistent with the methodology used in our previously developed economic decision-analytic model [14], a state-transition (Markov) model was used to project the impact of treatment with uRDN plus SoC compared with SoC alone over a lifetime horizon. Reflective of current guidelines [11, 12] and the TRIO trial [9], SoC across the three countries included treatment with three classes of antihypertensive medication, including a diuretic, a renin-angiotensin-system blocker and a calcium channel blocker at maximally tolerated dose. The model had a 1-month cycle length with half-cycle correction incorporated. For the Belgian and Dutch models, costs were discounted at 1.5% annually and outcomes [QALYs and life years (LYs)] were discounted at 3.0% annually in order to take into account the impact of time on valuation of costs and outcomes; for the French model, costs, QALYs and LYs were discounted at 2.5% annually up to 10 years, and 1.5% annually thereafter [19,20,21].

The model included 11 mutually exclusive health states to represent disease progression. The health states consisted of hypertension, five types of cardiovascular events (myocardial infarction, stroke, heart failure, end-stage renal disease and chronic heart disease) and death. All patients start in the hypertension health state and move to a different health state (with different health-related quality of life and costs) when an event occurs. Death is an absorbing health state and can occur at any time. In accordance with our previous cost-effectiveness model [14], we used risk equations based on Framingham and the Prospective Cardiovascular Münster (PROCAM) study to model how patients transition through the different health states [22, 23].

All patients start in the hypertension health state and move to a different health state when an event occurs, or a specific diagnosis is established. The patient pathway through various cardiovascular events is shown in Fig. 1. In accordance with the previously developed model [14], memory for end stage renal disease (ESRD) was incorporated, and a recurrent stroke health state was included; these enable consideration of different costs and health-related quality of life for the previous ESRD status of a patient experiencing an additional complication, and for patients who remain at risk for subsequent strokes. Also, consistent with previous modelling [14], the SBP reduction associated with uRDN was translated to a reduction of clinical events on the basis of the relative risks reported by the meta-analysis of Thomopoulos et al. (55 randomised controlled trials of antihypertensive medication in 195,267 individuals) [24]. We adapted and populated the economic model with data in late 2023 to early 2024.

Fig. 1

Schematic of cost-effectiveness model structure. AP angina pectoris, CHD coronary heart disease, ESRD end-stage renal disease, MI myocardial infarction. *Death is an absorbing health state that can be entered at any given time. #A memory has been incorporated to track ESRD status throughout the model time horizon. ^A memory has been incorporated to track heart failure status in stroke patients

c.

Clinical and Health‑Related Quality‑of‑Life Inputs

The key parameters used in the model are detailed in the data supplement (eTable 2). The model used office-based SBP from the RADIANCE-HTN TRIO trial since current cardiovascular risk equations [22, 23] and published meta-analyses of the clinical effect of changes in SBP [24] are calibrated using office SBP measurements. The baseline office SBP across both arms of the trial was 155.3 mmHg, with a mean reduction of 8.5 mmHg (and standard deviation of 19.1 mmHg) in the uRDN arm at 2 months [9, 10]. No sham intervention would be performed in real-world clinical practice, and any placebo effect would be part of the overall treatment effect observed for the intervention. Therefore, the base-case analysis assumes that no SBP reduction was associated with continuation of SoC (continued medical management for rHTN). Additional model clinical parameters were derived from literature searches and previously published models relevant to hypertension, and specifically, treatment with RDN. Health utilities for the specific health states identified were drawn from a variety of sources, including previous clinical trials and economic evaluations of cardiovascular interventions and health-related quality of life studies (eTables 3–5).

d.

Cost Inputs

Hypertension-related costs and event-related costs were considered throughout the patient’s lifetime. These were broken down into the following categories: procedure costs, anti-hypertensive medication costs, monitoring costs, cardiovascular event costs, drug acquisition, drug administration, resource use, cardiovascular events and treatment-related adverse events. Details on the resources and their unit costs, for each country, are presented in eTables 6–9. The procedural cost for uRDN (including all cost components associated with the one-off provision of the uRDN system plus procedural components and hospital treatment costs) was assumed to be €8500 in all three countries. Whilst this is higher than the assumed cost in our previously published UK economic evaluation [14], we wished to take a conservative position with respect to the impact of recent cost inflation, exchange rates and inter-country procedural cost variability. When required, costs were inflated to 2023 levels using country-specific inflation indices.

e.

Data Analysis

Results were reported as ICERs. This was done by calculating the ratio of the difference in mean costs and mean change in LYs and QALYs between uRDN plus SoC and SoC alone. To provide full insight into the robustness of the results, a 95% CI around the ICER has been calculated. The box method was applied as a simplified method to calculate this interval to avoid additional complexity [25].

One-way sensitivity analysis and probabilistic sensitivity analysis (PSA) were conducted to assess the uncertainty surrounding the model inputs and sensitivity of the model results to changes in parameter inputs. One-way sensitivity analysis was performed using realistic minimum and maximum individual model inputs (one at a time); for all model parameters, the minimum and maximum plausible values for univariate analysis were defined as the lower and upper 95% CIs. For the PSA, all parameters were varied simultaneously, and results were recorded for 1000 iterations, which was sufficient to provide stable results from the model. The majority of variables were assumed to have a normal distribution, with the exception of proportions, probabilities and utility estimates, which were all varied using a beta distribution. Gamma distributions were applied for hazard ratios.

In addition, several scenario analyses were used to explore the impact of the model’s validity. For insight into the cost-effectiveness of uRDN in real-world conditions, we used the 12-month results of the ACHIEVE study, which included patients treated with the uRDN system (n = 96) [26]. The ACHIEVE observational study demonstrated a 15.0 mmHg reduction in mean office SBP at 12-month follow-up. The mean baseline office SBP in ACHIEVE was 176 mmHg versus 155.3 mmHg in the RADIANCE-HTN TRIO trial [9]. A scenario in which the sham-subtracted SBP reduction of 5 mmHg from the sham-controlled RADIANCE-HTN TRIO trial was used to further test the cost-effectiveness based on conservative assumptions.

A patient-level simulation component explored the impact of modelling a heterogeneous patient population, which can cause biased results when there is a nonlinear relationship between risk factors and cardiovascular event risks (Jensen’s inequality) [27]. The simulation model uses random sampling to create a virtual patient cohort based on defined patient characteristics and the correlation between them, as found in the RADIANCE-HTN TRIO trial (eTables 1 and 25). Each patient from the cohort is then run through the model’s existing Markov structure. The results are averaged to achieve an overall cohort result to compare with the base-case deterministic results.