Renal Denervation vs Pharmacotherapy for Resistant Hypertension: A Meta‐Analysis

Abstract

The effect of renal denervation (RD) for resistant hypertension remains controversial because of the conflicting results of finished and ongoing studies. The authors performed a meta‐analysis of case‐control studies to identify whether renal sympathetic denervation or pharmacotherapy (PHAR) was more effective for resistant hypertension. A systematic Internet database search of relevant papers written in English was performed. A total of nine studies met the inclusion criteria, with a total of 1096 patients. When comparing the RD group with the PHAR group, there was a significant decrease in systolic blood pressure (SBP) (weighted mean difference, −12.81 mm Hg; 95% confidence interval [CI], −22.77 mm Hg to −2.85 mm Hg; P=.01) and diastolic blood pressure (DBP) (weighted mean difference, −5.56; 95% CI, −8.15 mm Hg to −2.97 mm Hg; P<.0001). This pooled analysis shows that for patients with resistant hypertension, RD is more effective in reducing SBP and DBP than PHAR. RD may be more effective in special subgroups of patients, which needs to be identified in future investigations.

Hypertension is prevalent all over the world, with 29.2% of adults predicted to have hypertension by 2025.1 About 5% to 10% of patients with high blood pressure (BP) have resistant hypertension, defined as BP above the target level despite the use of three or more antihypertensive drugs, with at least one being a diuretic.2 The sympathetic nervous system plays an important role in the occurrence and development of resistant hypertension.3 Renal denervation (RD) delivers radiofrequency energy or other forms of energy to break down afferent and efferent renal sympathetic nerve signalling and reduces total sympathetic nerve activity.4 Many observational5, 6 and case‐control studies have shown that RD achieves a significant reduction on BP. In recent years, RD treatment for resistant hypertension has been used clinically in more than 80 countries, including parts of Europe, South America, Australia, and Canada.7, 8, 9 A consensus of experts from the European Society of Cardiology proposed that RD is an optimal treatment for resistant hypertension and provided guidance on the efficacy, safety, limitations, and potential benefits of RD and on who is most suitable for the procedure.10 However, there are a few studies11, 12, 13, 14 that do not support the viewpoint that RD is superior to pharmacotherapy (PHAR) in treating resistant hypertension, particularly the Symplicity HTN‐3 study, which is a prospective, single‐blinded, randomized, sham‐controlled trial. In a meta‐analysis of five studies (three randomized controlled trials and two nonrandomized controlled trials, n=800),15 results showed that RD was superior to maximal medical therapy in lowering BP. Because of the small number of studies performed, however, the author did not provide further analysis. Here, our aim was to provide an update and subgroup analysis of RD's effect on systolic BP (SBP) and diastolic BP (DBP) at 6‐month follow‐up in patients with resistant hypertension.

Methods

Design

We searched PubMed, Embase, Web of Science, and the Cochrane Library to evaluate the effect of RD on resistant hypertension. References were searched for relevant articles. We used the following keywords and their combinations: “RD,” “resistant” or “refractory,” and “hypertension” or “blood pressure”. Inclusion criteria were: (1) controlled or randomized controlled trials; (2) patients with uncontrolled BP despite treatment with three maximally dosed antihypertensive medications with least one being a diuretic16; (3) one group receiving RD and the other PHAR; (4) reports on BP change at 6 months of follow‐up; and (5) complete data. Exclusion criteria were: (1) observational studies, uncontrolled trials, and case reports; (2) incomplete data and irretrievable data; and (3) unpublished literature or in other languages. Clinical data were extracted by two investigators and checked by other authors. The concordance rate of the two investigators was 92%. Divergences were resolved by discussion with the other authors. The following data were abstracted: study design, first author, patient number, sex, age, body mass index, method of BP measurement, and BP at baseline and 6 months after procedure. We contacted authors when the data were incomplete. We used the Newcastle‐Ottawa Scale including selection, comparability, and exposure to assess the quality of the included studies.17

Outcome

The primary end point was the magnitude of BP drop at 6‐month follow‐up.

Data Synthesis and Statistical Analysis

Database management and analysis were conducted with Review Manager 5.3 (RevMan, The Nordic Cochrane Centre and The Cochrane Collaboration, Copenhagen, Denmark). The weighted mean differences (WMDs) in SBP and DBP change at 6‐month follow‐up in the RD group were compared with those in the PHAR group, and then all included studies were pooled if there were sufficient data. The results were reported as means±standard deviations. Data were combined with a random effects model of DerSimonian and Laird with inverse variance.18 Estimates were recorded as WMDs with 95% confidence intervals (CIs). The significant difference between two groups was identified with a two‐side P value <.05. The heterogeneity was tested with Cochran's Q statistic (chi‐square, P<.10 for significance) and I ² statistic, because I ² represented the percentage of variability in point estimates from heterogeneity instead of sampling error.18 I ² values <25% were deemed to have low heterogeneity, I ² values 25% to 50% were considered to have moderate heterogeneity, and values >50% were considered to have high heterogeneity.19 If the heterogeneity was high, we would conduct subgroup analysis to investigate the potential sources. The potential publication bias was assessed with a funnel plot and Egger's linear regression test. A P value <.05 indicated significant publication bias.20 Sensitivity analysis was conducted by this means that each study was removed by turns and the others were analyzed to assess whether the result was affected by the omitted one.

Results

Study Selection

With the searching method previously described, we retrieved a total of 1961 articles (Figure 1). Two investigators viewed titles and abstracts independently and finally identified 18 articles concerned with our aim. We then reviewed all full texts. Three studies were excluded because they showed data for long‐term follow‐up and the data at 6‐month follow‐up were recorded in other articles, which were included in our analysis. Five trials had a follow‐up of 3 months and one trial did not provide enough data and received no reply after trying to contact the authors. Thus, nine trials11, 12, 13, 14, 21, 22, 23, 24, 25 were included in the analysis. We assessed risk of bias through Review Manager 5.3 and the result is shown in Figure 2. After confirming with the researchers of each trial or physicians, a total of 1096 patients were diagnosed with resistant hypertension. A total of 719 patients received treatment with RD and drug therapy and 377 patients received only PHAR. The characteristics of the studies and participants are shown in Tables 1 and 2, respectively.

Figure 1.

Flow diagram of included trials.

Figure 2.

Risk of bias of nine included trials.

Table 1.

Characteristics of Included Studies

First Author	Year	Region	Design	Blinding	Definition of RH	BP Measurement	Catheter Type	RD, No.	PHAR, No.	Nonresponders, %
Bhatt	2014	US	RCT	Single	SBP ≥160 mm Hg	O and A	Simplicity catheter	364	171	NR
Fadl	2014	Norway	RCT	OL	SBP ≥140 mm Hg	O and A	Symplicity Catheter System (Ardian, Mountain View, CA)	9	10	78
Mahfoud	2012	Germany	RCT	OL	SBP ≥160 mm Hg	O	Symplicity and Flex	88	12	8
Mahfoud	2014	Germany, Australia	CT	DB	SBP ≥140 mm Hg	O	Simplicity Flex system	55	17	NR
Rosa	2015	Czech Republic	RCT	OL	24‐h ABPM SBP ≥130 mm Hg	O and A	Symplicity System	52	54	NR
Esler	2010	Australia,Europe,New Zealand	RCT	OL	SBP ≥160 mm Hg	O and A	Simplicity catheter	49	51	16
Pokushalov	2012	Russia	RCT	DB	SBP ≥161 mm Hg	O	Navistar ThermoCool		14	0
Ewen	2014	Germany	CT	OL	SBP ≥140 mm Hg	O	Simplicity catheter	50	10	14
Zhang	2014	China	CT	OL	SBP ≥160 mm Hg	O	5F standard radiofrequency catheter	39	38	21

Abbreviations: ABPM and A, ambulatory blood pressure monitor; CT, controlled trial; DB, double‐blind; NR, not reported; O, office blood pressure; OL, open‐label; PHAR, pharmacotherapy; RCT, randomized controlled trial; RD, renal denervation; RH, resistant hypertension; SBP, systolic blood pressure; US, United States.

Table 2.

Baseline Characteristics of Study Patients

First Author	Year	Treatment Group	Age, y	Men	BMI, kg/m2	DM, %	CAD, %
Bhatt	2014	RD	57.9±10.4	215 (59.1)	34.2±6.5	47	28
		PHAR	56.2±11.2	110 (64.3)	33.9±6.4	41	25
Fadl	2014	RD	57±10.9	7 (78)	29±5.3	22	11
		PHAR	62.7±5.1	10 (100)	30±5.3	30	60
Mahfoud	2012	RD	61.6±10.3	54 (61)	29.9±7.5	17	NR
		PHAR	61.9±12.5	7 (58)	28.1±6.6	33	NR
Mahfoud	2014	RD	65±10	39 (71)	29.2±4.3	47	NR
		PHAR	70±9	10 (59)	28.6±5.3	41	NR
Rosa	2015	RD	56±12	40 (77)	31.2±4.3	22	6
		PHAR	59±9	34 (63)	33.4±4.7	17	7
Esler	2010	RD	58±12	34 (65)	31±5	40	19
		PHAR	58±12	27 (50)	31±5	28	7
Pokushalov	2012	RD	57±8	11 (85)	28±6	8	15
		PHAR/PVI	56±9	10 (71)	28±5	14	14
Ewen	2014	RD	64.7±7.1	39 (78)	30.7±4.2	50	24
		PHAR	68.4±3.8	8 (80)	28.6±3.2	30	40
Zhang	2014	RD	58.6±14.1	24 (62)	29.9±1.8	18	NR
		PHAR	62.9±12.6	20 (53)	30.1±0.9	16	NR

Abbreviations: BMI, body mass index; CAD, coronary artery disease; DM, diabetes mellitus; NR, not reported; PHAR, pharmacotherapy; RD, renal denervation group. Values are expressed as mean±standard deviation or absolute number (percentage).

Meta‐Analysis of Studies

Results from pooling all nine studies in the analysis showed that the RD group achieved a significant reduction in office SBP (WMD, −12.81 mm Hg; 95% CI, −22.77 mm Hg to −2.85 mm Hg; P=.01, I ²=92%) compared with the PHAR group. Eight pooled studies were included in the analysis, which showed that there was a significant decrease in DBP (WMD, −5.56; 95% CI, −8.15 mm Hg to −2.97 mm Hg; P<.0001; I ²=63%) in the RD group compared with the control group (Figure 3). We conducted subgroup analysis. As shown in Figure 4, six studies were designed as randomized and pooled analysis and showed no significant decrease in office SBP (WMD, −9.72 mm Hg; 95% CI, −22.72 mm Hg to 3.2 mm Hg; P=.14; I ²=94%) in the RD group compared with the PHAR group. Data pooled from three nonrandomized studies showed a significant reduction in office SBP (WMD, −21.30 mm Hg; 95% CI, −27.48 mm Hg to −15.12 mm Hg; P<.00001; I ²=11%) in the RD group compared with the PHAR group.

Figure 3.

Forest plot of meta‐analysis of office systolic blood pressure (SBP; A) and diastolic blood pressure (DBP; B). RD indicates renal denervation; PHAR, pharmacotherapy; IV, weighted mean difference; SD, standard deviation; CI, confidence interval.

Figure 4.

Forest plot of subgroup meta‐analysis. A forest plot of the randomized subgroup (A), a forest plot of the nonrandomized subgroup (B), and a forest plot of the randomized subgroup after sensitivity analysis (C). RD indicates renal denervation; PHAR, pharmacotherapy; IV, weighted mean difference; SD, standard deviation; CI, confidence interval.

Publication bias was evaluated with funnel plots. The funnel plots showed symmetric distribution, indicating no significant publication bias (Figure 5). We performed sensitivity analysis as previously mentioned. No single study significantly influenced the analysis of office SBP and DBP. In the randomized subgroup, when we removed the study by Fadl,12 the reduction of office SBP became significant when comparing RD with PHAR (Figure 4).

Figure 5.

Funnel plot of meta‐analysis of office systolic blood pressure (SBP) and diastolic blood pressure (DBP). A funnel plot of the office SBP group (A), a funnel plot of the office DBP group (B), a funnel plot of randomized subgroup (C), and a funnel plot of the nonrandomized subgroup (D).

Discussion

In all nine trials, five supported the findings that RD was better than PHAR in reducing office SBP,21, 22, 23, 24, 25 while four showed opposite results.11, 12, 13, 14 This analysis pooled a total of nine trials (six randomized controlled trials and three controlled trials) to investigate the difference between RD and PHAR in treating resistant hypertension. It showed significant decreases in both office SBP and DBP compared with PHAR, demonstrating that the RD procedure was more effective than PHAR in reducing BP. We performed sensitivity analysis and found that no single study significantly influenced the pooled analysis of office SBP and DBP, indicating that the result was steady and robust. We came to the conclusion that RD was efficient in controlling drug‐resistant hypertension.

The results of the Symplicity HTN‐3 trial,11 which used a prospective, single‐blind, randomized, sham‐controlled design, did not support the findings that RD was beneficial in treating resistant hypertension. In this trial, approximately one fourth of patients (26.2%) were African American, who were not included in other studies. When patients were analyzed separately, non–African Americans with RD achieved a significant drop in office SBP that was greater than that seen in the sham‐control group; however, changes were not seen in African American patients.26, 27, 28 The stable use of antihypertensive agents was required to be enrolled into this trial; however, two of five patients asked for medicine changes during the study, which challenged the qualification of “resistant hypertension.”28 In addition, the use of spironolactone was found in 28.7% of patients in the sham procedure and in 22.5% of patients in the RD group, while in the Symplicity HTN‐2 trial, the proportion was 17% in both groups. Spironolactone's mechanism was similar to RD,29 which may reflect the slight difference in office SBP between the groups. The denervation procedure fared poorly in Symplicity HTN‐3,30 with 60% of patients who had just one or no “ablation notches.” This may be partly explained by the deficiency of the first‐generation Medtronic device (Minneapolisc, MN). Presented at EuroPCR 2014, many of the US operators conducted RD for the first time in the Symplicity HTN‐3 study. Because of the multiple participating sites (88 sites in the United States) and the lack of operators' experience and methods of testing effective denervation, we could not confirm whether each procedure was up to standard. Moreover, Messerli and Bangalore31 pointed out that the BP change in the RD group had wide variance (−14.13±23.93 mm Hg in the RD group vs −11.74±25.94 mm Hg in the control group), indicating that the procedure may be effective in specific subgroups and variable in the time to respond to RD. Identifying these specific subgroups in future trials may be meaningful in the application of RD in clinical practice.

When restricting the analysis to six randomized trials,11, 12, 14, 21, 23, 24 there was no significant reduction in office SBP in the RD group compared with the PHAR group. We then performed sensitivity analysis and found that when the study by Fadl12 was removed, the result became statistically significant. In the study by Fadl,12 much attention was paid to the PHAR group, which may have increased the adherence of treatment as well as produced the Hawthorne effect. In addition, the number of participants was small (n=19), which may have led to selection bias and lack of statistical power to identify the effect of RD (n=9) and PHAR (n=10). Furthermore, the definition of resistant hypertension in this trial was set at an office SBP ≥140 mm Hg, while in the others in this subgroup (except Rosa and colleagues14) it was set at an office SBP ≥160 mm Hg. Thus, we omitted this article in the analysis of the randomization group and the result was still favorable for office SBP reduction comparing RD with PHAR. After pooling three nonrandomized trials,13, 22, 25 the analysis showed significant differences in office SBP comparing RD with PHAR. Heterogeneity (I ² values were 92% and 63% in the office SBP group and office DBP group respectively; 94% and 11% for randomized and nonrandomized groups respectively) was high for this analysis, therefore we performed a random effects model to evaluate whether RD or PHAR was more effective in resistant hypertension. We considered the high heterogeneity to be attributable to patient nationality, the types of catheters used, the implementer of the procedure, and the definition of resistant hypertension. Some trials were open‐labelled12, 14, 21, 22, 24, 25 or nonrandomized, which may produce selection bias.

RD is a safe and effective method for treating resistant hypertension.10, 32 Side effects of RD and adverse events during 6 months of follow‐up (Table S1) were reported in all eight studies except the study by Ewen and colleagues.25 The only reported side effect was post‐punctual pseudoaneurysm, which is common for other interventional therapies. No significant serious adverse renal artery or kidney function changes have been noted at periods ranging from 6 months to 3 years after the procedure.33 RD has also shown benefits on glucose metabolism and insulin resistance.34 Exercise heart rate is an independent predictor of cardiovascular mortality and morbidity,35 especially in patients with hypertension.36 In the study by Ewen and colleagues,25 researchers found that RD could reduce heart rate, improve mean workload, and increase exercise time. Left ventricular hypertrophy, which may increase cardiovascular disease morbidity and mortality, occurs in patients as a result of hypertension and sympathetic stimulation.13, 37, 38 In a 2014 study,13 researchers found that left ventricular end‐systolic volume decreased significantly, ejection fraction remained unchanged in a control group but increased significantly after RD, and left ventricular mass index to height^1.7 significantly decreased by 7.1%. This suggests potential benefits for left ventricular hypertrophy and cardiac function. There is evidence that RD decreases arrhythmia burden and reduces arterial stiffness.39 Sympathetic overactivity is a cardinal feature of heart failure, which may be reduced by RD. Davies and colleagues40 point out that in the first study of RD for chronic systolic heart failure, all seven patients described symptomatic improvement and increased 6‐minute walk distance at 6‐month follow‐up. Few studies provided enough data to discuss the cardiovascular outcomes in more detail.

The previous meta‐analysis15 evaluated the effect of RD on decreasing BP, but the number of trials included was small (a maximum of five). We included four trials and performed subgroup analyses using PRISMA guidance to improve reporting quality (Table S2).

Study Limitations

This study has several limitations. First, because the original data were not available, this analysis was based on data from articles that may not evaluate the confounding factors appropriately. Second, the number of patients in several of the trials was small. The fewer participants may be underpowered to assess differences between RD and PHAR. Third, the definition of resistant hypertension was different in the nine trials. Five trials defined resistant hypertension as an office SBP ≥160 mm Hg despite three antihypertensive medicines, including a diuretic, and the other four trials defined resistant hypertension as an SBP ≥140 mm Hg. This difference may result in inconformity of participants and high heterogeneity. Fourth, these trials reported only 6‐month follow‐up data, therefore long‐term analysis was not possible. Finally, only four trials provided ambulatory BP data, which is more powerful in assessing BP. Because of lack of data, analysis of ambulatory BP could not be performed. Therefore, more trials using ambulatory BP monitoring are needed to better assess the effect of RD.

Conclusions

This meta‐analysis demonstrated that RD was more effective than PHAR in decreasing office SBP and DBP in patients with resistant hypertension. This procedure may be more effective in special groups of patients, which should be identified in future studies.

Disclosures

The authors report no specific funding in relation to this research and no conflicts of interest to disclose.

Supporting information

Table S1. Side Effects of Renal Denervation

Table S2. PRISMA 2009 Checklist

J Clin Hypertens (Greenwich). 2016;18:733–740. DOI: 10.1111/jch.12742. © 2015 Wiley Periodicals, Inc.