Introduction

Hypertension is one of the most important cardiovascular risk factors and a major health problem worldwide [1, 2]. The age-adjusted prevalence of hypertension among United States (US) adults ≥ 20 years of age has been estimated as 46.7% [3]. Lifestyle modification and pharmacotherapy represent the two major strategies for hypertension [4, 5], and the importance of specifically managing early morning and nocturnal hypertension has also been reported [6, 7]. Despite the establishment of these treatment options, many hypertensive patients continue to have poorly controlled hypertension due to factors such as suboptimal adherence, economics, medication intolerance, and clinical inertia [5]. Therefore, renal denervation (RDN), the denervation of renal efferent and afferent sympathetic nerves, is being investigated as a non-pharmacologic approach, either alone or as an adjuvant therapy. Although early proof-of-concept studies reported significant reductions in blood pressure (BP) after RDN [8,9,10], the first randomized controlled trial (RCT) with a sham group (Symplicity HTN-3) failed to confirm these early findings [11]. Recent RCTs have shown significant BP reduction after renal denervation in patients with hypertension [12,13,14], but the efficacy has not been shown to be consistent [15]. Meanwhile, the ESH2023 guidelines recommend the use of RDNs for uncontrolled Class II hypertension [16], and the Recor’s Paradise™ RDN and Medtronic’s Symplicity Spyral™ RDN systems were recently approved by the US Food and Drug Administration (FDA) for use as an adjunctive treatment in hypertension patients in whom lifestyle modifications and antihypertensive medications do not adequately control BP. This means that the day is not far off when RDN will be introduced into clinical practice.

Our previous meta-analysis showed statistically significant and clinically meaningful reductions in all ambulatory blood pressure monitoring (ABPM) and office BP parameters for patients with hypertension [17]. However, the previous meta-analysis had several major limitations. First, the number of patients included in the subgroup analysis was relatively small. Second, the analysis was performed over a short follow-up period (2–6 months), making it impossible to determine long-term effects. Finally, several important RCTs on the effect of RDNs in hypertension [18, 19], as well as the introduction of a novel RDN system using alcohol [20], were published after completion of the previous meta-analysis. Therefore, the purpose of this study is to perform a systematic review and an updated meta-analysis to examine the antihypertensive efficacy and safety of RDN in patients with hypertension.

Methods

Data sources and search strategy

This meta-analysis adhered to the PRISMA statement [21] and was registered prospectively in PROSPERO (ID CRD42022368297). We conducted searches in MEDLINE and the Cochrane Central Register of Controlled Trials from their inception until November 15, 2023, without language limitations (Supplementary Table 1). Our inclusion criteria comprised randomized sham-controlled trials that focused on RDN in hypertension patients and included 24 h ABPM changes from baseline. Trials without random allocation or a sham control group, and those exploring the effects of RDN without transcatheter devices were excluded.

Screening and data extraction

Two independent reviewers performed literature screening for each candidate study and then extracted the data from eligible studies. Any discrepancies between them were resolved by a third reviewer. Titles and abstracts were reviewed for eligibility in the first screening, and full-text screening was conducted in the second screening. If there were multiple publications on a single study, only the publication with a complete dataset was included. The following data were extracted from all eligible publications: study characteristics (study design, patient selection, inclusion and exclusion criteria, follow-up duration, number of patients, and RDN catheters and endpoints); baseline characteristics (age, sex, country of residence, body mass index (BMI), presence of resistant hypertension, BP-lowering medication, and comorbidities); and outcomes mentioned in the following paragraph. The risk of bias and study quality were determined using version 2 of the Cochrane risk of bias tool for randomized trials [22].

Outcome measures

The primary outcome was the mean change in 24 h ambulatory systolic blood pressure (ASBP) from baseline in the RDN group, compared with the mean change in the sham-control group. Secondary outcomes included mean changes in 24 h ambulatory diastolic blood pressure (ADBP), daytime systolic and diastolic BP (DSBP and DDBP), nighttime systolic and diastolic BP (NSBP and NDBP), office systolic and diastolic blood pressure (OSBP and ODBP), and home systolic and diastolic blood pressure (HSBP and HDBP) from baseline in the RDN group, as compared with the mean change in the control group. Safety outcomes included complications of the RDN procedure and serious adverse events (SAEs). If both results of intention-to-treat and per-protocol analysis were available, those of intention-to-treat analysis were extracted for each endpoint.

Statistical analysis

We employed random-effects models to summarize absolute differences in BP changes between groups across trials. Heterogeneity was assessed using I2 statistics, with potential publication bias evaluated through funnel plots. Heterogeneity in treatment effects between subgroups was tested using a meta-regression analysis. Subgroups included: use of antihypertensive medications [on-med group] or no use [off-med group]; outcome BPs measured at trough or non-trough blood concentration of drugs; radiofrequency, ultrasound, or alcohol-mediated RDN; and use of first- or second-generation RDN. First-generation RDN was defined as that using a monopolar Symplicity FLEX and main renal artery ablation without branch ablation, while second-generation RDN was defined as that using advanced radiofrequency ablation techniques (with multipolar catheter and ablation extension beyond the main renal artery into renal-artery branch vessels), endovascular ultrasound, and modified medical treatment regimens. The REINFORCE trial [23] was not included in the subgroup analysis by generation of RDN because it employed a balloon-based bipolar radiofrequency-based Vessix system without branch ablation. Office BP measured at the trough blood concentration of drug was defined as that measured before ingestion of a participant’s regular antihypertensive drugs or measured after healthcare providers witnessed participants taking their drugs at the office in the morning.

The outcomes of the two long-term (36 months) follow-up studies, the SPYRAL HTN-ON MED trial (2022) [24] and SYMPLICITY HTN-3 (2022) trial [25], were analyzed separately. In the SYMPLICITY HTN-3 trial [25], the control group was divided into crossover and non-crossover groups, with only the crossover group undergoing RDN at six-month follow-up. Stata SE version 16 (StataCorp (2019) Stata Statistical Software, Release 16; StataCorp, College Station, TX) was used for statistical analyses. A two-tailed p-value < 0.05 was considered statistically significant.

Results

Description of included trials

A total of 707 publications were identified in the initial search, of which 580 citations were screened (the first screening) after excluding 127 duplicates (Supplementary Fig. 1). The majority of the selected studies were excluded during the first screening. The remaining 278 studies were selected for full-text screening. After excluding 264 studies during the second screening, 14 trials with 2222 participants (RDN, 1295; sham-control, 927) were included in the meta-analysis [11,12,13, 15, 18,19,20, 23,24,25,26,27,28,29], with 2 of them constituting extensions of studies included in our previous meta-analysis. These trials used radiofrequency (9 trials), ultrasound (4 trials), or an alcohol-mediated device (1 trial). The most common sham procedure was renal angiography.

At baseline, the mean ASBP and OSBP ranged between 130 and 160 mmHg, and 146 to 180 mmHg, respectively (Table 1). Among the included studies, except for the long-term follow-ups, the mean follow-up period ranged from 3 months to 3 years. The mean age of participants fell between 53 and 64 years, with an average BMI of 29–34 kg/m2. Male participation varied from 40% to 80%, while diabetes prevalence ranged from 5% to 54%. The mean number of antihypertensive drugs varied from 0 to 5.1. The risk of bias across the trials was generally low (Supplementary Table 2).

Table 1 Baseline characteristics of the included studies
Full size table

Ambulatory BP

Across 12 studies (after excluding the 2 previously included studies), RDN significantly reduced 24 h ASBP and ADBP compared to sham procedures (ASBP: − 2.81 mmHg, 95% confidence interval [CI] − 4.09 to − 1.53, p < 0.001; ADBP: − 1.47 mmHg, 95% CI −2.39 to −0.56, p = 0.002) (Fig. 1a, b). However, some degree of heterogeneity (I2 = 31.4%, p = 0.140 for ASBP and 47.8%, p = 0.033 for ADBP) was noted, and asymmetry in funnel plots were noted, potentially indicating publication bias (Supplementary Fig. 2). The long-term follow-up studies also showed that RDN significantly reduced 24 h ASBP and ADBP (ASBP: − 13.82 mmHg, 95% CI − 20.09 to − 7.54, p = 0.001; ADBP: − 8.83 mmHg, 95% CI − 13.99 to − 3.66, p = 0.001), compared with sham procedure (Supplementary Fig. 4). There was a significant heterogeneity for 24 h ADBP change (I2 = 62.9%, p = 0.101 for ASBP; 78.5%, p = 0.031 for ADBP). Publication bias could not be assessed because only 2 studies were included in the meta-analysis (Supplementary Fig. 3).

Fig. 1
figure 1

Effects of renal denervation on 24 h ambulatory blood pressure and office blood pressure. a 24 h ambulatory systolic blood pressure; b 24 h ambulatory diastolic blood pressure. c Office systolic blood pressure. d Office diastolic blood pressure. Horizontal lines in the graphs display the 95% CIs with the point estimate positioned at the center of the respective box. Each box within every subplot is proportionate to the sample size of the corresponding study. Diamonds are utilized to depict summary data centered around the combined estimates, and their width encompasses the corresponding 95% CIs. CI confidence interval; RDN renal denervation

Full size image

Office BP

Ten studies demonstrated significant reductions in OSBP and ODBP with RDN, compared to sham procedures (OSBP: − 4.95 mmHg, 95% CI − 6.37 to − 3.54, p < 0.001; ODBP: −  2.79 mmHg, 95% CI − 3.67 to − 1.90, p < 0.001) (Fig. 1c, d). There was no significant heterogeneity (I2 = 0.0% for both). The OSBP and ODBP data appeared to exhibit asymmetry in the funnel plots (Supplementary Fig. 2). As for long-term follow-up studies, RDN significantly reduced 24 h ASBP by − 15.66 mmHg (95% CI, − 29.24 to − 2.07; p = 0.024) and ADBP by − 8.39 mmHg (95% CI, − 16.28 to − 0.50; p = 0.037), compared with sham procedure (Supplementary Fig. 4). There was significant heterogeneity for both 24-h ASBP and ADBP changes (I2 = 85.7%, p = 0.008 for ASBP; 83.6%, p = 0.013 for ADBP).

Daytime and nighttime BP

The definitions of daytime and nighttime during the 24 h ABPM in each trial are shown in Supplementary Table 3. RDN exhibited significant reductions in DSBP, DDBP, NSBP, and NDBP, compared to sham procedures across studies (DSBP: −3.17 mmHg, 95% CI − 4.75 to − 1.58, p < 0.001; DDBP: − 1.88 mmHg, 95% CI − 3.08 to − 0.68, p = 0.002; NSBP: − 3.41 mmHg, 95% CI − 4.69 to − 2.13, p < 0.001; NDBP, − 1.61 mmHg, 95% CI − 3.05 to − 0.17, p = 0.028) (Fig. 2a–d). There was some degree of heterogeneity for DDBP (I2 = 41.2% for DSBP, p = 0.074; 51.2% for DDBP, p = 0.037; 0.0% for NSBP, p = 0.595; 48.0% for NDBP, p = 0.052). Publication bias was visually suggested for the NSBP data (Supplementary Fig. 2). In the long-term follow-up studies, RDN significantly reduced DSBP by − 13.75 mmHg (95% CI, − 21.88 to −  5.61; p = 0.001), DDBP by − 8.36 mmHg (95% CI, − 15.90 to − 0.83; p = 0.030), NSBP by − 14.76 mmHg (95% CI, − 18.53 to − 10.98; p < 0.001), and NDBP by − 10.06 mmHg (95% CI, − 13.41 to − 6.72; p < 0.001), compared with sham procedure (Supplementary Fig. 5). There was a significant heterogeneity for DDBP (I2 = 72.2% for DSBP, p = 0.058; 88.5% for DDBP, p = 0.003; 0.0% for NSBP, p = 0.340; 31.7% for NDBP, p = 0.226).

Fig. 2
figure 2

Effects of renal denervation on ambulatory daytime and nighttime blood pressure. a Daytime systolic blood pressure. b Daytime diastolic blood pressure. c Nighttime systolic blood pressure. d Nighttime diastolic blood pressure. Horizontal lines in the graphs display the 95% CIs with the point estimate positioned at the center of the respective box. Each box within every subplot is proportionate to the sample size of the corresponding study. Diamonds are utilized to depict summary data centered around the combined estimates, and their width encompasses the corresponding 95% CIs. CI confidence interval, RDN renal denervation

Full size image

Home BP

RDN led to significant reductions in HSBP and HDBP across five studies (HSBP: − 4.64 mmHg, 95% CI − 7.44 to − 1.84, p = 0.001; HDBP: − 2.28 mmHg, 95% CI − 4.30 to − 0.26, p = 0.027), albeit with substantial heterogeneity for both HSBP and HDBP (I2 = 68.6% for HSBP, p = 0.013; 78.4% for HDBP, p = 0.001) (Fig. 3a, b). Funnel plots of both appeared to exhibit substantial symmetry (Supplementary Fig. 2)

Fig. 3
figure 3

Effects of renal denervation on home blood pressure. a Home systolic blood pressure. b Home diastolic blood pressure. Horizontal lines in the graphs display the 95% CIs with the point estimate positioned at the center of the respective box. Each box within every subplot is proportionate to the sample size of the corresponding study. Diamonds are utilized to depict summary data centered around the combined estimates, and their width encompasses the corresponding 95% CIs. CI, confidence interval; RDN, renal denervation

Full size image

Safety

Complications and SAEs in each study are summarized in Supplementary Table 4. The long-term follow-up studies showed higher death rates, ranging from 2.4% to 7.9%, compared to the other studies.

Subgroup analysis

There was no significant interaction between the use of antihypertensive drugs (on- off-med studies), the blood concentration of drugs (trough or non-trough), the energy source of RDN (radiofrequency, ultrasound, or alcohol-mediated), or the generation of RDN device (first or second generation) and the effects of RDN on 24 h ambulatory and office BP changes (on/off-med: p for interaction = 0.122 for ASBP, p = 0.140 for ADBP, p = 0.171 for OSBP, p = 0.264 for ODBP; trough/non-trough blood concentration of drugs: p = 0.243 for OSBP, p = 0.577 for ODBP; energy source: p = 0.578 for ASBP, p = 0.499 for ADBP, p = 0.853 for OSBP, p = 0.870 for ODBP; first/second generation: p = 0.446 for ASBP, p = 0.315 for ADBP, p = 0.094 for OSBP, p = 0.535 for ODBP) (Figs. 4a–d, 5a, b, 6a–d), Supplementary Fig. 7). Meanwhile, there was a significant interaction between use of antihypertensive drugs and the effects of RDN on DDBP change (p for interaction = 0.014), suggesting that the effect of RDN on DDBP reduction in the off-med groups was greater than that in the on-med groups (Supplementary Fig. 6).

Fig. 4
figure 4

Subgroup analysis based on concomitant use of blood pressure-lowering medications. a 24 h ambulatory systolic blood pressure. b 24 h ambulatory diastolic blood pressure. c Office systolic blood pressure. d Office diastolic blood pressure. Horizontal lines in the graphs display the 95% CIs with the point estimate positioned at the center of the respective box. Each box within every subplot is proportionate to the sample size of the corresponding study. Diamonds are utilized to depict summary data centered around the combined estimates, and their width encompasses the corresponding 95% CIs. CI confidence interval, RDN renal denervation. P values for interaction were 0.122 (ASBP), 0.140 (ADBP), 0.171 (OSBP), and 0.264 (ODBP)

Full size image
Fig. 5
figure 5

Subgroup analysis based on whether or not blood pressures were measured at the trough blood concentration of a drug. a Office systolic blood pressure. b Office diastolic blood pressure. Horizontal lines in the graphs display the 95% CIs with the point estimate positioned at the center of the respective box. Each box within every subplot is proportionate to the sample size of the corresponding study. Diamonds are utilized to depict summary data centered around the combined estimates, and their width encompasses the corresponding 95% CIs. CI confidence interval, RDN renal denervation. P values for interaction were 0.243 (OSBP) and 0.577 (ODBP)

Full size image
Fig. 6
figure 6

Subgroup analysis according to types of renal denervation device (radiofrequency, ultrasound, or alcohol-mediated device). a 24 h ambulatory systolic blood pressure. b 24 h ambulatory diastolic blood pressure. c Office systolic blood pressure. d Office diastolic blood pressure. Horizontal lines in the graphs display the 95% CIs with the point estimate positioned at the center of the respective box. Each box within every subplot is proportionate to the sample size of the corresponding study. Diamonds are utilized to depict summary data centered around the combined estimates, and their width encompasses the corresponding 95% CIs. CI confidence interval, RDN renal denervation. P values for interaction were 0.578 (ASBP), 0.499 (ADBP), 0.853 (OSBP), and 0.870 (ODBP

Full size image

Discussion

In this meta-analysis using data from 12 randomized, sham-controlled trials (2222 patients in total), including 4 new trials, RDN was associated with statistically significant and clinically meaningfully reductions in 24 h BP, daytime BP, nighttime BP, office BP, and home BP compared to the sham group in patients with resistant, uncontrolled, and drug-naive hypertension. These reductions were significant regardless of the RDN device used and the presence or absence of antihypertensive use, and in both first- and second-generation trials. To our knowledge, this is the largest meta-analysis of randomized, sham-controlled trials in RDN ; it includes results from the most recent RCTs and provides updated insight into the effects of RDNs for the future treatment of hypertension.

Antihypertensive effect of RDN on ABPM

In this meta-analysis including sham-controlled trials, RDN was shown to significantly reduce 24 h SBP by 2.81 mmHg, daytime SBP by 3.17 mmHg, and nighttime SBP by 3.41 mmHg. Although no interventional studies have examined the effect of targeting BP values assessed by ABPM on the reduction of cardiovascular events, epidemiological and observational studies have widely reported that BP pressure values assessed by ABPM have better predictive ability for cardiovascular disease risk than office BP [30,31,32]. Yang et al. found that higher 24 h SBP and higher nocturnal SBP significantly increased the risk of death and cardiovascular events and were the optimal BP parameters for estimating cardiovascular outcomes [33]. Fengler et al. reported that patients with a decrease in 24 h ASBP of 5 mmHg or more after 3 months of RDN had an approximately 50% reduction in cardiovascular events compared to patients with no decrease [34]. In a recent Spanish ABPM observational cohort study (including 59,214 participants), ABPM (and especially nocturnal ABPM) was more beneficial than office BP in predicting all-cause and cardiovascular mortality [35]. The ESH2023 guidelines also recommend the assessment of nighttime BP and nocturnal BP variability during treatment of hypertension Class II [16]. Although previous RDN trials involving sham treatment have shown significant reductions in nocturnal SBP after RDN [36,37,38,39], the present analysis showed a significant reduction in nocturnal SBP, similar to our previous meta-analysis. Narita et al. reported that nocturnal BP control is more important than daytime BP control in patients with treatment-resistant hypertension ( ≥ 3 drugs), with a 10 mmHg increase in nocturnal blood pressure being associated with a 1.45-fold increased risk of cardiovascular events [6], and Kario et al. reported that a nocturnal BP of 120/70 mmHg or higher was associated with a significant, 2.22-fold higher risk of heart failure in patients with treatment-resistant hypertension (≥ 3 drugs) [7]. In a post-hoc analysis of the SPYRAL HTN-ON MED Pilot study, RDN not only lowered 24 h BP and office BP, but also suppressed morning BP surges [40]. Morning BP surge and increased BP variability are risks for cardiovascular events [41, 42], and RDN may contribute to a reduction of cardiovascular events, especially in patients with nocturnal hypertension and early morning hypertension.

Antihypertensive effect of RDN on office BP

The antihypertensive effect and safety of RDN have been reported to persist for up to 3 years [24, 25, 38, 43,44,45,46,47,48,49] and more recently for 10 years [50,51,52]. Unfortunately, there are no multicenter, blinded, randomized, prospective outcome trials on RDN, and the efficacy of RDN in reducing cardiovascular events has not yet been established. A meta-analysis of 123 intervention trials on antihypertensive drugs (613,815 patients included in the analysis) showed that the relative risk reduction of cardiovascular events was proportional to the magnitude of BP reduction during treatment, regardless of baseline BP levels or comorbidities [53]. For every 10 mmHg reduction in OSBP, the risk of major cardiovascular events was significantly reduced by 20%, the risk of stroke by 27%, the risk of heart failure by 27%, and the risk of all-cause mortality by 13% [53]. Furthermore, a meta-analysis of large clinical trials of antihypertensive drugs (344,716 subjects included in the analysis) reported that a 5 mmHg reduction in OSBP was associated with an approximately 10% reduction in risk of major cardiovascular events, regardless of whether the patient had a previous diagnosis of cardiovascular disease [54]. The Global Symplicity Registry estimates that RDN reduces the relative risk of major cardiovascular events by 34% in patients with treatment-resistant hypertension [55]. In this study, RDN reduced OSBP by 4.95 mmHg from baseline. This reduction is comparable to the reduction in BP that was found to be effective in reducing event risk in the aforementioned meta-analysis, suggesting that RDN provides clinically meaningful reductions in BP.

Association of RDNs with drug treatment

This sub-analysis showed that the 24 h ABP and office BP-based antihypertensive effects of RDNs were consistent regardless of the presence or absence of antihypertensive medications. In particular, significant antihypertensive effects were observed in the off-med group [12, 13, 18], except in the REINFORCE which were terminated during enrollment, and alcohol RDN trials. This is the effect of confounding factors such as medication adherence was minimal, demonstrating a clear antihypertensive effect of RDN alone.

On the other hand, there are limitations to oral antihypertensive drug therapy, with the greatest being medication adherence. It has been reported that approximately 50% of patients with treatment-resistant hypertension become non-adherent within 1 year of antihypertensive initiation [56], and poor adherence in hypertensive patients increases the risk of cardiovascular mortality 1.69 times [57]. A post-hoc analysis of REQUIRE found that approximately 45% of Japanese patients with treatment-resistant hypertension were non-adherent to their medication [58]. Another limitation is that oral antihypertensive drug therapy creates a blind spot in the pharmacological effects of antihypertensive medications. As a result of the usual once-daily (often morning) dosing schedule and pharmacokinetics of drug clearance, drugs may reach sub-therapeutic trough levels at night or early in the morning. In the present study, we performed a sub-analysis of trough OSBP (no pill intake until OSBP measurement on the same day) vs. non-trough OSBP in the on-med group and found that trough OSBP was more effective in lowering BP by − 5.27 mmHg vs. trough OSBP is the trough office BP equivalent to early morning home BP measured prior to taking antihypertensive medications, reflecting control of early morning home BP that contributes to cardiovascular events. Increasing the number of antihypertensive medications improves control of office BP, but fails to improve nocturnal and early morning hypertension, which remains poorly controlled in 45–55% of patients [59]. As a result, it is difficult to maintain good control of BP for a sustained period of 24 h. In this respect, the “always on” effect of RDN may be greater than that achieved with drug therapy in the long term. The antihypertensive effect of RDN in effectively lowering 24 h BP, including nocturnal and early morning hypertension [25, 60], independent of adherence to antihypertensive treatment, may provide long-term additive cardiovascular protection as an adjunctive treatment to drug therapy.

Differences of antihypertensive effects among different RDN devices and generations

There was no significant difference in antihypertensive effects among the radiofrequency, ultrasound, and alcohol RDN devices in this study. In the randomized RADIOSOUND-HTN trial in patients with resistant hypertension, Fengler et al. compared three groups (a radiofrequency RDN group for the main renal artery only: RFM-RDN; a radiofrequency RDN group for the main renal artery and branch artery: RFB-RDN; and an ultrasound RDN group for the main renal artery: USM-RDN), and found that at 3 months, the USM-RDN group showed a better BP reduction effect than the RFM-RDN, and the USM-RDN and RFB-RDN groups showed comparable results [61]. However, after 6 months the results were similar, and after 12 months the BP reduction effects of radiofrequency RDN and ultrasound RDN were equivalent [62]. These results suggest that radiofrequency RDN of the main renal artery and branch arteries, and ultrasound RDN, which cauterizes only the main renal artery, are equally effective. The efficacy of the ultrasound device, which has a depth of about 6 mm, might be greater than that of the radiofrequency device [63], which has a depth of about 3 mm, if the ultrasound device [63] could be used to cauterize the branching arteries, but the branching of the renal artery is complicated by about three to four branching vessels meandering toward the renal parenchyma, and there is a risk of overcauterizing the branch by covering it with neighboring branches if the depth of cauterization is too deep. In addition, the diameter of the branch vessels is small, and several devices are required to occlude them together with the main trunk, so cost is also an issue to be considered. On the other hand, Sakakura et al. reported that ultrasound devices can change the total area and depth of ablation by changing the power and time [64], and if the power and time can be adjusted to safely ablate the branch, and if the balloon size can be adjusted by pressure, more ablation could be achieved with fewer devices. There is thus a possibility that a more hypotensive effect can be obtained with fewer devices. Regarding procedure time and contrast volume, radiofrequency RDN, which includes branch ablation, requires a significantly longer procedure time and uses more contrast media than ultrasound RDN [61], which may be a consideration when selecting a device. Alcohol RDNs were reported in one study (106 patients), was not associated with a significant antihypertensive effect on the primary endpoint. Although the primary endpoint of 24 h ASBP at 8 weeks was not significantly different between the RDN and sham groups, the RDN group used significantly less antihypertensive medication, despite having similar BP to the sham patients at 12 months. The present study was conducted in an open-label fashion, and one possible reason for the smaller antihypertensive effect than previously observed is that the patients were recruited during the COVID-19 epidemic. Although observational in design, another study reported that OSBP was increased by about 5 mmHg between before and after COVID-19 [65], and such a COVID-19-associated increase in BP may have been involved in the smaller antihypertensive effect observed in the present meta-analysis. In any case, the number of cases is small, and we currently need to wait for the results of the on-going studies to be considered.

In a sub-analysis, there was a trend for second-generation RDN trials to show a greater antihypertensive effect compared to first-generation trials. This is a result related to technological advances, additional ablation sites, and revised protocols, etc.

Home BP and RDN

This study showed that HSBP in the RDN group was 4.64 mmHg lower than that in the sham group, with the difference being statistically significant. To our knowledge, this is the first report on home BP in a meta-analysis or RCT of RDN. Martine et al. reported that home BP is an alternative to ABPM for blood pressure monitoring up to 12 months after RDN [66]. Home BP is a stronger predictor of cardiac disease than office BP and is more strongly associated with organ damage than office BP [67, 68]. Home BP during antihypertensive treatment has been reported to be more strongly associated with risk of CV events and death than office BP [69, 70]. Furthermore, Kario et al. used data from the STEP study in China and estimated that a 5 mmHg reduction in early morning home BP was associated with an 18.2% reduction in cardiovascular events. However, even though the JSH guidelines strongly recommend the use of home BP as an index for antihypertensive treatment [71], it is often not used in practice due to the individual circumstances of physicians and patients. The JSH guideline states that the diagnosis of hypertension should be based on home BP, especially when there is a divergence between office BP and home BP measurements [5]. If home BP is not measured, this may result in the possibility that white coat hypertension (hypertension in the office with normal home BP) and masked hypertension (normal office BP with hypertension in the home) are missed, and that the white coat hypertension may be a result of the presence of a masked hypertension. Unlike white coat hypertension, masked hypertension is associated with higher vascular risk. In addition, self-monitoring of home BP using an information and communication technology (ICT) system and feedback of the results can bring about significant BP reduction through improved medication adherence, self-efficacy, and lifestyle improvement [72]. Currently, in ultrasound RDNs, a registry focused on home BP has been initiated, which is expected to improve the integrity of long-term follow-up and provide unique insights into BP and cardiovascular event suppression over time [73].

Safety and long-term outcomes of RDN

Although the studies included in this meta-analysis had shorter follow-up periods, there were few adverse events associated with RDNs. In a meta-analysis of 50 trials (5,769 participants, Symplicity FLEX or SPYRAL devices) examining adverse events and long-term prognosis using registry data from high-frequency RDNs, renal artery adverse events including stenosis and dissection were observed in approximately 0.45% of patients per year and stenting in an extremely rare 0.2% or patients per year [74]. This stenting rate is lower than the spontaneous rate previously reported in the hypertensive population. In 2022, the SPYRAL ON-MED and Symplicity HTN-3 trials reported 3-year follow-up results, respectively, both of which showed a significant sustained antihypertensive effect compared with the sham group. The results of the SPYRAL ON-MED and Symplicity HTN-3 trials showed that the RDN procedure itself was durable and safe, with no significant differences in adverse events [24, 25].

Regarding long-term efficacy, we performed a sub-analysis of two studies of radiofrequency RDN with 3-year follow-up, and found that the RDN group had significantly lower 24 h SBP by − 13.8 mmHg, OSBP by − 15.7 mmHg, DSBP by − 13.75 mmHg, and NSBP by − 14.76 mmHg than the sham group at 3 years. However, there was a high degree of heterogeneity with respect to office BP, which should be evaluated with caution. Rader et al. reported that ultrasound RDN was safe and achieved a sustained significant reduction in OSBP based on a 3-year follow-up report of the treatment arm of the RADIANCE- HTN SOLO trial [49]. Other meta-analyses have shown no change in the estimated glomerular filtration rate after RDN [75]. Thus, RDN is considered to be well tolerated as an endovascular treatment. However, compared to radiofrequency RDN, there remains a dearth of available data from clinical trials of ultrasound RDN and especially alcohol RDN, and hence further clinical data accumulation is needed.

Study limitations

Several limitations of this meta-analysis bear mention. First, the number of patients in the subgroup analyses was relatively small, particularly in the case of alcohol RDN, for which only a single trial was included, and thus the results of the subgroup analyses should be interpreted with caution. Second, the observed effect of RDN was on BP, not clinical outcomes, and the trials had short follow-up periods of 2 to 6 months. Long-term effects were analyzed using the SPYRAL ON-MED Pilot and Symplicity HTN-3 trials, which reported 3-year follow-up results, but both trials included patients who crossed over to the sham group, so caution is needed in interpreting the results. Finally, blood concentrations of antihypertensive drugs in urine and blood were measured in only some of the trials, so the impact of adherence could not be examined in detail.

Conclusions

This meta-analysis of 12 randomized, sham-controlled trials, including the most recent study, showed that renal denervation significantly reduces, even in the short term, all 24 h, office, and home BP parameters in patients with treatment-resistant hypertension, poorly controlled hypertension, and untreated hypertension. The results showed that the RDN devices significantly reduced 24 h BP, office BP, and home BP parameters in the short term. There were no differences in antihypertensive efficacy between different devices or between the use of the devices with and without antihypertensive medications, but there was a trend toward stronger antihypertensive efficacy in the second-generation trials across generations.

The long-term safety of renal denervation was clearly demonstrated, but to demonstrate its efficacy, it is necessary, above all, to perform the procedure on patients with hypertension whose BP is elevated due to sympathetic nerve activation, i.e., to identify the responder group. In addition, if an endpoint indicator can be identified, it will be possible to determine whether to add or terminate cauterization during the procedure. In the future, clinical trials with longer-term follow-up and appropriate clinical implementation based on evidence from actual clinical practice are needed.