OUP user menu

★ Editor's choice ★

Potential lifetime cost-effectiveness of catheter-based renal sympathetic denervation in patients with resistant hypertension

Marc Dorenkamp, Klaus Bonaventura, Alexander W. Leber, Julia Boldt, Christian Sohns, Leif-Hendrik Boldt, Wilhelm Haverkamp, Ulrich Frei, Mattias Roser
DOI: http://dx.doi.org/10.1093/eurheartj/ehs355 451-461 First published online: 22 October 2012


Aims Recent studies have demonstrated the safety and efficacy of catheter-based renal sympathetic denervation (RDN) for the treatment of resistant hypertension. We aimed to determine the cost-effectiveness of this approach separately for men and women of different ages.

Methods and results A Markov state-transition model accounting for costs, life-years, quality-adjusted life-years (QALYs), and incremental cost-effectiveness was developed to compare RDN with best medical therapy (BMT) in patients with resistant hypertension. The model ran from age 30 to 100 years or death, with a cycle length of 1 year. The efficacy of RDN was modelled as a reduction in the risk of hypertension-related disease events and death. Analyses were conducted from a payer's perspective. Costs and QALYs were discounted at 3% annually. Both deterministic and probabilistic sensitivity analyses were performed. When compared with BMT, RDN gained 0.98 QALYs in men and 0.88 QALYs in women 60 years of age at an additional cost of €2589 and €2044, respectively. As the incremental cost-effectiveness ratios increased with patient age, RDN consistently yielded more QALYs at lower costs in lower age groups. Considering a willingness-to-pay threshold of €35 000/QALY, there was a 95% probability that RDN would remain cost-effective up to an age of 78 and 76 years in men and women, respectively. Cost-effectiveness was influenced mostly by the magnitude of effect of RDN on systolic blood pressure, the rate of RDN non-responders, and the procedure costs of RDN.

Conclusion Renal sympathetic denervation is a cost-effective intervention for patients with resistant hypertension. Earlier treatment produces better cost-effectiveness ratios.

  • Cost-effectiveness
  • Resistant hypertension
  • Renal sympathetic denervation
  • Prevention
  • Cardiovascular disease


Systemic arterial hypertension continues to be a major public health challenge, affecting approximately 25% of the adult population in developed countries.1 It is a strong independent risk factor for cardiovascular (CVD) and cerebrovascular disease morbidity and mortality.2,3 Given its prevalence and severe consequences, the economic burden of hypertension is enormous.4,5

In most individuals, hypertension can be effectively treated by a combination of lifestyle changes and medication.6 There is a significant portion of patients, however, who are unable to achieve adequate blood pressure control despite optimized therapy and a number of them are considered to have resistant hypertension. Commonly, resistant hypertension is defined as blood pressure that remains above goal while adhering to full doses of an appropriate regimen of three antihypertensive drugs from different classes, one of which is a diuretic.7 The true prevalence of resistant hypertension is unknown, but current estimates suggest that between 10 and 30% of hypertensive patients are resistant to drug therapy.8 Among such patients, treatment options are limited and involve interventional or device-based therapies.8 One recent approach relies on a percutaneous, catheter-based procedure that selectively targets and disrupts renal sympathetic nerves using radiofrequency energy.9 Clinical safety and efficacy studies of renal sympathetic denervation (RDN) have shown substantial reductions of blood pressure without serious procedure-related complications, even though this intervention is not part of current guidelines.6,1012

As the costs of RDN differ significantly from continual costs of antihypertensive medications, the economic effect of the procedure must be considered. Our current analysis therefore aimed to determine the benefits, costs, and cost-effectiveness of catheter-based RDN for treatment of resistant hypertension.


Study design

A Markov model was used to estimate costs and survival among patients who underwent either RDN or who received failed best medical therapy (BMT). The model was designed to be applied to a population of patients with resistant hypertension. Entry systolic blood pressure (SBP) was 160 mmHg or more despite compliance with at least three antihypertensive drugs (including one diuretic). Secondary causes of hypertension were excluded. Cost-effectiveness was evaluated from the payers' perspective and included costs related to treatment of hypertension, adverse CVD events, and end-stage renal disease (ESRD). Given the absence of trials comparing hard clinical endpoints in patients treated with RDN vs. BMT, our model was based upon the assumption that the reduction in SBP associated with RDN would translate into reductions in event rates derived completely from other studies. Moreover, the model assumed that the reduction in SBP derived from RDN was sustained over the lifetime of the patient.

Markov model structure

We adopted a recently proposed Markov state-transition model that has been developed to assess the cost-effectiveness of CVD interventions using TreeAge Pro 2012 (TreeAge Software, Inc., Williamstown, MA, USA) (Figure 1).13,14 The present model had two arms: one that followed patients who were subjected to RDN, and the other which followed patients who were treated with BMT. For modelling purposes, it was assumed that patients in both arms received an antihypertensive regimen consisting of metoprolol, ramipril, and torasemid at maximum doses. This three-drug therapy was continued after RDN and simultaneously represented BMT in the comparator arm. The model could run from age 30 to 100 years or death, with a Markov cycle length of 1 year with half cycle correction. It was designed to simulate cohorts of men or women 30–99 years of age free from prior CVD or renal disease. Patients could remain in this disease-free state or could have a primary CVD event [myocardial infarction (MI), angina, stroke, or heart failure (HF)] or could transition to ESRD. During the first year, patients might experience more than one event (e.g. the occurrence of HF after MI). After an event, patients transitioned into one of the following chronic health states: post-MI (asymptomatic), post-angina (asymptomatic), post-stroke (no sequelae), long-term post-stroke sequelae (moderate or severe), chronic HF, or ESRD. While being in these states, patients might suffer from further CVD events. In addition to these primary events, secondary events also included worsening of HF and reinfarction. Subsequent to a secondary event, patients might move to other chronic health states. Transitions from severe to comparably less severe health states (e.g. from permanent stroke sequelae to an asymptomatic post-infarction state) were not modelled as they would involve unrealistic cost savings (see below). In each cycle, patients faced a risk of CVD or non-CVD death.

Figure 1

Markov model tree structure diagram. Shown is the Markov decision model pertaining to the treatment of resistant hypertension (HTN). The square on the left is a decision node with two branches representing the two treatment options: renal sympathetic denervation (RDN) and best medical therapy (BMT). Chance nodes are represented by circles; an encircled ‘M’ represents a Markov node with a 1-year cycle length. Triangles on the right-hand side are terminal nodes. For space reasons, identically structured subtrees are collapsed and not visible (indicated by a plus sign). CV, cardiovascular; CVD, cardiovascular disease; ESRD, end-stage renal disease; HF, heart failure; MI, myocardial infarction.

Transition probabilities

Our Markov model was calibrated to predict German CVD mortality. Thus, incidences of primary CVD events were based on German or, in the case of their unavailability, on North European registry data.13,1517 Tables of age- and gender-specific CVD incidences are provided in the Supplementary material online, Table S1. The incidence of ESRD (213 patients per million population) was derived from the German QuaSi-Niere renal registry.18 The probability of dying from non-CVD causes was obtained by subtracting the German CVD mortality (International Classification of Diseases, ICD-10 I00-I99) from all-cause mortality (Supplementary material online, Table S2).19,20

It was assumed that average incidences of CVD events implied average levels of CVD risk factors. As resistant hypertension is associated with an increased risk of CVD morbidity and mortality, the CVD risk was adjusted upwards by using the Systemic Coronary Risk Evaluation (SCORE) risk estimation system.21 In general, the SCORE equations can be used to estimate the increase in CVD events associated with an increase in SBP. For example, based on the SCORE risk equations, a SBP elevation of 20 mmHg is associated with a relative risk of 1.43 for death from coronary heart disease and of 1.55 for death from non-coronary CVD (stroke) (Table 1).13,14,21 The relative effect on morbidity was assumed to be the same as on mortality.13,14 As SCORE does not give confidence intervals, it was assumed that the logarithm of the standard deviation was one-quarter of the logarithm of the relative risk for mortality and, because of greater uncertainty, one-third of the logarithm of the relative risk for morbidity (Table 1).13,14 The relative risk of developing ESRD with various SBP levels is given in Supplementary material online, Table S3.22 Based on average blood pressure levels in Germany, which are listed in Supplementary material online, Table S4, the SBP level was increased by 30 mmHg or at least to 160 mmHg in order to allow a simulation of baseline clinical characteristics described in the most relevant RDN studies for resistant hypertension.1012,23 For instance, and according to this adjustment, 60-year-old men entered the model with an average SBP of 180 mmHg and 60-year-old women with an SBP of 183 mmHg. In the base-case scenario, RDN was assumed to result in a sustained SBP reduction of 20 mmHg.11,12 In contrast, the elevated SBP level in the BMT arm remained unchanged. In elderly patients, age becomes the dominant risk factor and the risk fraction attributable to hypertension declines accordingly.3 The present model accounted for this decline by reducing the relative risk of CVD events and mortality (Table 1) by 2.2% per year of age.3,13,14

View this table:
Table 1

Estimated relative risk associated with systolic blood pressure elevation of 20 mmHg, based on SCORE equations

CVD eventEstimated relative riskRangeReferences
CHD mortality1.431.20–1.7013,14,21
CHD incidence1.431.13–1.8113,14
Stroke mortality1.551.25–1.9213,14,21
Stroke incidence1.551.16–2.0613,14
  • Relative risks refer to patients aged 60 years and a systolic blood pressure (SBP) elevation of 20 mmHg (for risk adjustment to other age groups and SBP elevations, see text). By definition, coronary heart disease (CHD) comprised myocardial infarction, angina, and heart failure. CVD, cardiovascular disease; SCORE, systemic coronary risk evaluation.21

Each of the above-described primary events might be followed by secondary events or chronic conditions. The transition probabilities for subsequent events within the first year after a primary event were drawn from large registries or randomized controlled trials (Table 2).13,14,16,2234 If necessary, 1-year probabilities were derived from in-hospital and/or 6-month probabilities by using previously described formulas.13,14 Briefly, if only in-hospital event probabilities were given, they were assumed to be half of 1-year probabilities, and if in-hospital and 6-month probabilities were available, 1-year probabilities were calculated by doubling event probabilities during the first 6 months subtracted by in-hospital probabilities. With respect to MI, the ratio of the incidence of ST-elevation and non-ST-elevation MI was assumed to be unity. The risk of MI-related mortality was modelled as age-dependent, with half of these deaths occurring in the pre-hospital phase.29 Reinfarction was defined as having a new MI within the first year after an MI. Later infarctions were regarded as new and thus as secondary events. Some probabilities were not modelled because of lack of data (e.g. MI or angina during the first year after stroke; Table 2). It was assumed that HF secondary to MI might be reversible within the first year in some patients, while the rest transitioned to chronic HF states.

View this table:
Table 2

Transition probabilities during the first year after a primary event

Myocardial infarction (MI)
 Angina pectoris0.10200.0871–0.116925,26
 Heart failurea0.26700.2593–0.274728
 Mortality after MIcvaries by age (0.06–0.4)29
 Mortality after reinfarctiona0.4840.270–0.73830
Angina pectoris
 MI (men)0.01530.0096–0.021031
 MI (women)0.01730.0101–0.024531
 Stroke (men)0.01190.0068–0.017031
 Stroke (women)0.01100.0053–0.016731
 Heart failure (men)0.01530.0096–0.021031
 Heart failure (women)0.01810.0108–0.025431
 Mortality (men)0.01080.0060–0.015631
 Mortality (women)0.01340.0071–0.019731
 Moderate stroke sequelae0.0720.060–0.08413,14
 Severe stroke sequelae0.1690.158–0.18013,14
Heart failure (HF)
 Continued HF (after MI)0.5000.333–0.75013,14
 Mortality secondary HF0.2900.240–0.34013,14
 Mortality primary HF (men)0.1580.119–0.19733
 Mortality primary HF (women)0.1500.106–0.19533
End-stage renal disease
 MortalitydVaries by age (0.041–0.410)34
  • aIn-hospital event probabilities were assumed to be half of 1-year probabilites.13,14

  • bOne-year probabilities were calculated by doubling event probabilities during the first 6 months subtracted by in-hospital probabilities.13,14

  • cMortality during the first year after myocardial infarction (MI): age 30–49 years, 0.06; age 50–79 years, 0.281; age >80 years, 0.40.

  • dEnd-stage renal disease mortality during the first year of dialysis: age 30–44 years, 0.041; age 45–64 years, 0.09; age 65–74 years, 0.18; age 75–84 years, 0.312; age ≥85 years, 0.41.

More than 1 year after a primary event, patients were in symptomatic health states (stroke sequelae, HF, or ESRD) or in health states which were more or less asymptomatic (post-MI, post-angina, or post-stroke without specific long-term consequences). In any of these health states, patients were at a risk of new CVD events. All probabilities were expressed as relative risks. These were calculated by comparing the probabilities from registries and a range of randomized clinical trials with the incidences in the underlying population (Supplementary material online, Tables S1 and S2). The relative risk estimates are given in Table 3.13,14,25,31,33,3549 For the sake of conservative estimates, the probabilities of new events in any of the post-CVD states were not adjusted according to the different SBP levels. In addition, again due to the lack of data, some probabilities were assumed to be zero (e.g. risk of HF more than 1 year after MI or angina; Table 3).

View this table:
Table 3

Relative risks more than 1 year after a primary event

Health state and eventEstimated relative riskRangeReferences
Post myocardial infarction (MI)
 Angina pectoris21.715.8–27.613,14,25
 Mortality (all-cause)2.091.85–2.3437
Post angina pectoris
 Stroke (men)6.814.23–10.3931
 Stroke (women)5.633.17–9.3031
 Mortality (all-cause)1.210.97–1.4839
Post stroke (no sequelae)
 Mortality (all-cause)5.154.77–5.5540
Moderate stroke sequelae
 Heart failure21–413,14
Severe stroke sequelae
Heart failure (HF)
 MI (men)1.50.6–3.844
 MI (women)4.11.8–9.344
 HF worsening9.458.49–10.4947
 Mortality second year after HF (men)5.443.65–7.7933
 Mortality second year after HF (women)8.065.41–11.5733
 Mortality third year after HF (men)4.913.25–7.1333
 Mortality third year after HF (women)5.793.89–8.3133
 Mortality later year after HF (men)2.381.28–4.0733
 Mortality later year after HF (women)2.491.21–4.5733
End-stage renal disease
 Mortality (all-cause)5.95.4–6.548
  • Unless stated otherwise, mortality refers to cardiovascular mortality.

Quality of life

The Markov model incorporated adjustments for the quality of life of each health state. Utility values were derived from published studies (Table 4).5057 By convention, utilities are rated on a scale of 0 to 1, with 0 representing death and 1 a state of ideal health. The procedural disutility of RDN was not available. Hence, a previously reported utility of 0.94 for percutaneous coronary interventions was assigned to the 1-year period following the RDN procedure (i.e. approximately 3 weeks deducted from overall survival).57 Quality-adjusted life-years (QALYs) were calculated by multiplying the length of time spent in a certain health state by the utility associated with that particular health state.

View this table:
Table 4

Health-state utility estimates assigned to the different disease states in the Markov model

Myocardial infarction and angina
 First year0.880.80–0.9651,52
 Subsequent year0.900.80–0.9551,52
 Stroke without sequelae0.880.84–0.9253
 Moderate stroke sequelae0.710.68–0.7453
 Severe stroke sequelae0.310.29–0.3453
Heart failure0.690.30–0.9054
End-stage renal disease0.700.50–0.9055,56
Disutility of RDN procedurea0.060–0.1057
  • aUtility decrement over 1 cycle. RDN, renal sympathetic denervation.


The economic analysis was conducted from the perspective of a health care payer (German statutory health and nursing care insurance system). Data on costs were obtained from multiple sources, including the 2012 version of the German Diagnosis Related Groups (G-DRG) system, German pharmaceutical price lists, and German fee schedules for doctors and outpatient visits.5860 The values used in the model are presented in Table 5 (further details are provided in Supplementary material online, Table S5). The cost of RDN was based on G-DRG F54Z (€4424) and included further costs in terms of periprocedural bleeding complications (e.g. groin hematoma) at a rate consistent with other percutaneous interventional procedures.61 Continuous therapy with three antihypertensive drugs was assumed in both treatment arms (RDN and BMT). Drug costs reflected the fixed maximum amounts covered by German health insurance increased with the prescription costs on the assumption that four prescriptions were issued each year.59 Cost of further medication was added as required (e.g. clopidogrel after drug-eluting stent implantation). End-stage renal disease requiring long-term dialysis was considered the most expensive health condition (Table 5). All costs are given in Euro (€) for the year 2012 and were rounded to the nearest whole amount. Future costs and outcomes were discounted at an annual rate of 3%. Incremental cost-effectiveness ratios (ICERs) were defined as the difference in costs divided by the difference in life-years or in QALYs.

View this table:
Table 5

Health care costs

ParameterValue (€)Range (€)
Therapy of hypertension
 Renal sympathetic denervation44463335–5558
 Antihypertensive treatment (per year)610229–991
Myocardial infarction (MI)
 Treatment of acute MI10 7168037–13 395
 Fatal pre-hospital MI672504–840
 Treatment post-MI (per year)343257–429
Angina pectoris
 Treatment of one attack of angina49203690–6150
 Treatment post-angina (per year)343257–429
 Treatment of stroke (first year)20 50415 378–25 630
 Post-stroke (no sequelae) (per year)347260–434
 Moderate stroke sequelae (per year)60674550–7584
 Severe stroke sequelae (per year)24 07918 059–30 099
Heart failure (HF)
 Developing HF (initial diagnosis)37072780–4634
 Worsening of HF37072780–4634
 Long-term treatment of HF (per year)21391604–2674
End-stage renal disease (ESRD)
 Long-term treatment of ESRD (per year)53 96140 471–67 451

Sensitivity analyses

One-way deterministic sensitivity analyses were performed to evaluate whether the results were affected by changes in the model assumptions. Input parameters were varied over the ranges given in Tables 25. The relatively wide range of annual costs of antihypertensive treatment shown in Table 5 (€229–€991) allowed the model to simulate other medication regimens that included more or other drug classes than those used in the base-case scenario (two alternative treatment regimens are exemplified in Supplementary material online, Table S5). Additionally, alternative scenarios were examined and the model was re-run after: (i) decreasing and increasing the SBP lowering effect of RDN (10 and 30 mmHg, respectively); (ii) assuming a non-responder rate to RDN of 30% (whereas non-response was defined as failure to demonstrate any SBP reduction at all); and (iii) discounting costs and outcomes at 0 and 5% per annum, respectively.

Probabilistic sensitivity analysis (PSA) was undertaken to assess the overall uncertainty in the values used in the model. In the PSA, appropriate probability distributions are placed over parameter values. All transition probabilities during the first year as well as health utilities were characterized as beta distributions. Relative risks were incorporated as log-normal and costs as normal distributions. Baseline estimates from Supplementary material online, Tables S1 and S2 were not varied in the PSA because they were assumed to be fixed. The PSA was carried out using a Monte Carlo approach with 10 000 iterations. The willingness-to-pay (WTP) threshold (€ per QALY) was set at an internationally accepted level of €25 000 to €35 000 (about £20 000 to £30 000), indicated, for example, by the National Institute for Health and Clinical Excellence (NICE).62


Model validation

Our Markov model was validated by comparing model-projected life expectancies with actual German life tables. Specifically, the remaining lifetime (in undiscounted life-years) was estimated for patients with average CVD risk levels at ages 30, 40, 50, 60, 70, 80, and 90 years of age, respectively. For both men and women, a very close concordance was found between model predictions and national life tables (Supplementary material online, Table S6).

Base-case analysis

The model was run in men and women for all age groups between 30 and 90 years of age. For clarity, only results from selected, but representative, age groups are presented (50, 60, 70, 80, 85, and 90 years of age). In the base-case scenario, treatment of resistant hypertension with RDN was consistently more effective in terms of life-years or QALYs gained. When compared with BMT, RDN was associated with increased health care costs in each age group (Table 6). For example, RDN gained 0.98 QALYs in men and 0.88 QALYs in women 60 years of age over the remaining lifetime and led to additional costs of €2589 and €2044, respectively. The resulting ICER in men (€2642/QALY) was comparable with the ICER found in women (€2323/QALY) (Table 6). Incremental cost-effectiveness ratios increased with increasing age, whereas this increase was very substantial in patients 80 years of age or older and was at the same time more pronounced in women than in men. Hence, cost-effectiveness ratios were generally more favourable in younger patients, including patients 30–49 years of age (data not shown). However, due to the higher initial treatment costs, RDN was never a cost-saving strategy in comparison with BMT with respect to the base-case scenario.

View this table:
Table 6

Health and economic outcomes of renal sympathetic denervation when compared with best medical therapy

Gender and age (years)TherapyCost (€)LYQALYICER (€/LY)ICER (€/QALY)
Men, 50BMT30 47414.5013.28
RDN32 34915.6914.5215761512
Men, 60BMT27 14912.0010.93
RDN29 73812.9911.9126152642
Men, 70BMT22 6019.048.18
RDN25 4349.728.8441664292
Men, 80BMT14 0655.945.45
RDN17 4366.275.7410 21511 624
Men, 85BMT11 7034.464.08
RDN15 1214.694.2614 86118 989
Men, 90BMT78603.353.11
RDN11 6053.473.1731 20862 417
Women, 50BMT29 59316.8415.75
RDN31 32517.9116.8616191560
Women, 60BMT26 96114.2513.24
RDN29 00515.1214.1223492323
Women, 70BMT22 11310.9810.18
RDN24 58411.5810.7641184260
Women, 80BMT13 9087.076.61
RDN17 0767.356.8511 31413 200
Women, 85BMT10 6125.184.85
RDN14 0265.364.9818 96726 262
Women, 90BMT69303.743.53
RDN10 7293.833.5642 211126 633
  • Costs and outcomes are discounted at an annual rate of 3%. BMT, best medical therapy; ICER, incremental cost-effectiveness ratio; LY, life-years; QALY, quality-adjusted life-years; RDN, renal sympathetic denervation.

One-way sensitivity analyses

The cost-effectiveness of RDN was generally robust to changes of the input variables within the estimated ranges (Tables 25). The results of one-way deterministic sensitivity analyses showed that the relative cost-effectiveness was most sensitive to the SBP lowering effect related to RDN, the rate of RDN non-responders, and the costs associated with the RDN procedure. Figure 2 presents the results of exemplary sensitivity analyses in 60-year-old men and women as tornado diagrams. Displayed are the 10 individual parameters that influence the ICER estimates most, arranged from top to bottom in order of their importance. Variations in the values of other input parameters, specifically those relating to secondary CVD events, had only a minimal impact on results, regardless of gender and age (data not shown). Although not cost-saving, RDN remained, in principal, cost-effective in all one-way scenarios examined.

Figure 2

Tornado diagrams. Exemplary tornado diagrams showing the results of deterministic one-way sensitivity analyses in 60-year-old men (A) and women (B). Each bar represents a sensitivity variable. The vertical axis intersects the horizontal axis at the base-case incremental cost-effectiveness of renal sympathetic denervation (RDN) expressed in euro per quality-adjusted life-year (QALY) gained (€2642/QALY for men and €2323/QALY for women, respectively). The width of a horizontal bar denotes the impact of each parameter's uncertainty on the base-case value. The values on either side of the bar represent the lowest or highest value simulated for each input parameter. RR, relative risk; SBP, systolic blood pressure.

Probabilistic sensitivity analysis

All probabilistic analyses evaluated 10 000 iterations of the model. Figure 3 shows exemplary incremental cost-effectiveness scatter plots of RDN vs. BMT derived from PSA in 60-year-old patients. In comparison with BMT, RDN resulted in an increase in QALYs in 99.3% of simulations in men and in 98.9% of simulations in women. With respect to incremental costs, RDN was cost-saving in 3.3% and in 7.3% of simulations in 60-year-old men and women, respectively.

Figure 3

Incremental cost-effectiveness scatter plots. Incremental cost-effectiveness scatter plots of renal sympathetic denervation vs. best medical therapy derived from probabilistic sensitivity analyses in 60-year-old men (A) and women (B). Each dot represents the result of one iteration. For clarity, only 1000 out of 10 000 simulated data points are shown. The ellipse delineates the 95% confidence interval, respectively. Incremental effectiveness is expressed in quality-adjusted life-years (QALY) gained. Costs and QALYs are discounted at an annual rate of 3%.

Figure 4 presents the cost-effectiveness acceptability curves for RDN compared with BMT based on multiple simulated ICERs for men and women 50, 60, 70, 80, 85, and 90 years of age. Considering a WTP threshold of €25 000/QALY, there was a 95% probability that RDN would remain cost-effective up to an age of 76 and 75 years in men and women, respectively. If the WTP threshold was increased to €35 000 per QALY gained, this age threshold increased to 78 years for men and to 76 years for women. With higher WTP thresholds, RDN was likely to be cost-effective even at advanced ages (Figure 4). At ages 80 years and older, however, cost-effectiveness was obviously more favourable in men than in women (Figure 4 and Table 6).

Figure 4

Cost-effectiveness acceptability curves. Cost-effectiveness acceptability curves indicating the age-dependent probability that renal sympathetic denervation (RDN) is cost-effective compared with best medical therapy in men (A) and women (B) at different willingness-to-pay thresholds to gain one quality-adjusted life-year (QALY). Costs and QALYs are discounted at an annual rate of 3%. In patients aged 30–49 years, RDN was highly cost-effective at a willingness-to-pay threshold below €25 000 per QALY, independent of gender (curves not shown).


Main findings

To the best of our knowledge, this is the first study to evaluate the cost-effectiveness of RDN relative to BMT for the treatment of resistant hypertension. Calculations were based on a Markov simulation model and the cost-analysis was conducted from the health care payer perspective. Our analyses indicate that RDN is a highly cost-effective treatment for resistant hypertension. Using the WTP threshold of €35 000 per QALY, there was a 95% probability that RDN would remain cost-effective up to an age of 78 and 76 years in men and women, respectively. Results were robust to extensive sensitivity analyses. Cost-effectiveness was influenced mostly by the magnitude of the effect of RDN on SBP, the rate of RDN non-responders, and the procedure costs of RDN.

Cost and cost-effectiveness analysis

Non-pharmacological treatment strategies of resistant hypertension involve novel device-based approaches such as interventional RDN and baroreceptor activation.1012,63 Despite the apparent efficacy of both approaches, data on their cost-effectiveness are scarce, as most cost-effectiveness analyses of treating hypertension focused on comparing or adding conventional therapies.14,64,65 However, a recent study found a surgically implantable baroreceptor stimulation device (Rheos®, CVRx, Inc., Minneapolis, MN, USA) to be cost-effective at a $64 400 (about €52 000) per QALY threshold.54 The target population of that study was a simulated 50-year-old US cohort with resistant hypertension and without a history of CVD events. Therapy with the Rheos® system resulted in a 20 mmHg SBP reduction from a starting SBP of 180 mmHg compared with failed medical management.54 Evidence on the clinical effectiveness of RDN mainly stems from the Symplicity clinical trial program, which investigated the effects of RDN with a dedicated radiofrequency ablation catheter system (Symplicity, Ardian, Inc., Palo Alto, CA, USA, acquired by Medtronic, Inc., Minneapolis, MN, USA) in the treatment of resistant hypertension.1012 The primary endpoint was change in office blood pressure. Recent data showed that patients (n = 24) in the Symplicity HTN-1 trial maintained at an average SBP reduction of −33 mmHg at 36 months following RDN.66 For the sake of a conservative estimate, the base case of our model incorporated a 20 mmHg decrease. A decrease of 30 mmHg was included as alternative scenario in the sensitivity analyses and resulted in an even better cost-effectiveness of RDN (Figure 2). For simulation purposes, and in accordance with a previous study, the model assumed that the 20 mmHg reduction in SBP derived from RDN was maintained over the lifetime of the patient.54 However, we found even a 10 mmHg decrease to be cost-effective (Figure 2). Moreover, RDN remained cost-effective even if 30% of patients did not respond to therapy and blood pressure levels persisted at initial elevated systolic levels (Figure 2). The sensitivity analysis undertaken around alternative procedure costs of RDN indicated an important impact of this parameter on overall cost-effectiveness. Although RDN offers great procedural safety, the present model accounted for costs incurred from potential complications related to vascular access at a rate comparable with other percutaneous interventional procedures.61 Nonetheless, as demonstrated by our results, the cost-effectiveness of RDN also persisted even with cost assumptions biasing model results against RDN. Health care costs associated with stroke care and stroke-related disability were based on conservative estimates and did, for example, not include frequently required patients' co-payments to nursing facilities, which are frequently required. Thus, our baseline assumptions were likely to underestimate the actual costs of stroke.13,14,67 Higher stroke-related costs, however, would further decrease the ICERs of RDN since this therapeutic approach prevents more strokes than BMT (Figure 2). Furthermore, we did not consider potential cost-savings from reducing or stopping antihypertensive medications in the RDN group.1012 For the sake of a conservative estimate, beneficial effects of RDN, other than those consequent to blood pressure reduction, were not entered into the model.6871

In general, we found that the ICER increased with the age at which RDN was administered (Figure 4). This was due to the shorter post-intervention survival time from which to accrue cost offsets (to the initial RDN costs) and QALY gains resulting from the assumed reduction in hypertension-related disease events and death. The tested base-case scenario included patients aged 60 years, which corresponds to the mean age of the patients in the Symplicity trials.1012 With respect to probabilistic analyses, we found RDN to be generally cost-effective in this age group and cost-saving in a minority of individual patients (Figure 3). In practice, many patients may be older than 60 years. However, our sensitivity analyses as a function of age at which RDN is performed indicated that the ICER of RDN compared with BMT remained below a WTP threshold of €25 000 to €35 000 at least up to the age of 75 years. In simulated scenarios involving patients of older age (>80 years), treatment costs exceeded €35 000 per QALY, and thus RDN may only be considered cost-effective at higher WTP thresholds (Figure 4). A recent study showed that treating hypertension in very elderly patients reduced their risk of adverse CVD advents and death and there is, in principle, no upper age limit for patients to undergo RDN.72,73

To evaluate the cost-effectiveness of RDN, our study used a comprehensive model that captures more CVD events and health states than most other models.13,14 In contrast to previously published models, which mainly base the risk estimation for CVD events on Framingham risk equations, the present model incorporated observed incidence rates and adjusted these rates according to the SBP levels.13,14,54 This approach avoids potential bias introduced by using risk equations and by uncertainties due to distances in geography and time.13,14 As part of the validation process, the input to the model needed some adjustment in order to fit German mortality data. This is a limitation, because the model might be more consistent if there were fewer data sources. However, in accordance with previous studies, we considered the use of old data from the USA to generate comparatively more bias.13,14 Although the model is relatively complex, it necessarily simplifies the underlying clinical reality, as do all models. As there are no trials comparing hard clinical endpoints in patients treated with RDN, our model is based completely upon the assumption that the reduction in SBP associated with RDN translates into reductions in event rates derived from other studies. The model neither accounts for all possible combinations of health states (e.g. the simultaneous presence of chronic HF and stroke sequelae) nor for all disease events commonly associated with hypertension, such as peripheral arterial disease, cardiac arrhythmias, and retinopathy. However, where cost estimates and utility values were omitted they would favour the BMT group. Therefore, results from the current model likely provide conservative estimates for the cost-effectiveness of RDN.


Our analysis suggests that at commonly used European WTP thresholds, RDN represents a cost-effective treatment option in populations with resistant hypertension in which a significant reduction in blood pressure has been demonstrated.

Supplementary material

Supplementary material is available at European Heart Journal online.

Conflict of interest: M.D. has received travel support from Medtronic for travel to meetings for the study or other purposes. W.H. and M.R. have received lecture honoraria from Medtronic. The other authors have no disclosures.


View Abstract