Effects of Renal Denervation on Cardiac Structural and Functional Abnormalities in Patients with Resistant Hypertension or Diastolic Dysfunction

The aim of the present study is to systematically evaluate the impact of RDN on cardiac structure and function in patients with resistant hypertension (RH) or diastolic dysfunction. We retrieved Pubmed, Embase and Cocharane Library databases, from inception to April 1st, 2016. Studies reporting left ventricular mass (LVMI) or left ventricular (LV) diastolic function (reflected by the ratio of mitral inflow velocity to annular relaxation velocity [E/e’]) responses to RDN were included. Two randomized controlled trials (RCTs), 3 controlled studies and 11 uncontrolled studies were finally identified. In observational studies, there was a reduction in LVMI, E/e’ and N-terminal pro B-type natriuretic peptide (BNP) at 6 months, compared with pre-RDN values. LV ejection fraction (LVEF) elevated at 6 months following RDN. In RCTs, however, no significant change in LVMI, E/e’, BNP, left atrial volume index or LVEF was observed at 12 months, compared with pharmaceutical therapy. In summary, both LV hypertrophy and cardiac function improved at 6 months after RDN. Nonetheless, current evidence failed to show that RDN was superior to intensive (optimal) drug therapy in improving cardiac remodeling and function.

Data extraction and quality assessment. After eligible articles being identified, data were extracted by two separate researcher. The characteristics of included studies, involving study design, sample size, measurements, study population and follow-up interval were extracted. Also, baseline characteristics of included participants, including age, gender, commodity diseases, body mass index, and usage of antihypertensive drugs were collected. The outcomes of interest, i.e. change in LVMI or diastolic function (including LVMI or diastolic function before and after RDN) were extracted. Any disagreement was resolved by consensus. The primary endpoints were: 1) LVMI and E/e' change following RDN in observational studies; the difference in LVMI and E/e' change between RDN and DT in RCTs. Secondary endpoints were: 1) LV ejection fraction (LVEF) response; 2) left ventricular diastolic diameter (LVDD) response; 3) left atrial volume (LAVI) response; 4) BNP response. Data at different time points were collected separately.
For observational studies, the mythological quality was assessed by means of the Newcastle-Ottawa scale. (Supplementary Table 1). For RCTs, Cochrane Collaboration Risk of Bias Tool was applied in the quality assessment. (Supplementary Table 2).

Data Synthesis and Analysis.
For observational studies, including controlled and single-arm studies, the changes in LVMI and E/e' before versus after RDN were pooled. For RCTs, the differences in LVMI and E/e' changes with RDN and PT were pooled and analyzed. The same strategy was used to handle secondary outcome parameters. Weighted mean difference (WMD) with 95% confidence interval (CI) was taken as treatment effect measure when outcome measurements in all trials were made on the same scale. The standard mean difference (SMD) was applied when the trials all assessed the same outcome, but measured it in a variety of ways (referring to Cochrane handbook). I 2 was calculated and used to assess the between-study heterogeneity. I 2 value of 25%, 50% and 75% represents low, moderate and high heterogeneity, respectively. A Fixed-effects model was applied unless the level of heterogeneity reached high. Data were mainly presented as mean ± standard deviance (SD), thus when only 95% CI or standard error (SE) was reported, we converted 95% CI or SE to SD according to the Cochrane handbook. P < 0.05 was considered statistically significant. We performed data analyses by the mean of Review Manager (version 5.2). Meta-regression analyses were conducted in STATA software (version 11.0) to evaluate the relationship between changes in LV remodeling or function and blood pressure.

Results
Our primary search identified 435 records, while only 16 of these articles were finally included in our analysis ( Fig. 1). There are 14 observational studies 7,12,14-25 (including 11 uncontrolled studies and 3 controlled studies) and 2 RCTs 26,27 . The baseline characteristics of included subjects was shown in Table 1. Table 2 summarized details regarding studies with 6-months follow-up (number of medications, responses to medications, blood pressure control and inclusion criteria or the time when RDN started). Mean age varied from 53.9 to 74.6 years. Number of participants ranged from 14 to 100. Mean number of antihypertensive drugs varied from 4.3 to 6.4. Echocardiography was the most used technology (15 out of 16), six studies employed cardiac magnetic resonance (CMR), and 3 trials used both echocardiography and CMR. Most observational studies featured a follow up of 6 months, while both RCTs had a follow up of 12 months. Differing from other studies, the study by Patel et.al included HF patients with preserved EF (HFpEF) 26 . As shown in Supplementary Tables 1 and 2, most studies had a low risk of bias. Impact of RDN on cardiac structure. In observational studies, echocardiography showed that LVMI was reduced at 6 months following RDN (WMD = −13.88 g/m 2 , 95% CI = −19.94 to −7.82 g/m 2 , I 2 = 0). More pronounced reduction was observed at 12 months (WMD = 16.67 g/m 2 , 95% CI = −25.38 to −7.97 g/m 2 , I 2 = 0) (Fig. 2). CMR showed a reduction in LVMI at 6months (WMD = 5.18 g/m 2 , P = 0.05) but not 12 months (P = 0.48).
LVDD was not significantly changed after RDN in observational studies (Fig. 4). The difference in LVDD change between RDN and PT was not significant in the RCT by Patel et al. (Table 3).
Meta-regression analysis. Meta-regression analyses showed that the changes in LV remodeling or dysfunction at 6 months were not significantly associated with blood pressure lowering (Fig. 7) Discussion RDN is a new interventional approach designed to treat RH. The value of RDN is still under debate, especially after failure of RDN to lower BP over and above the sham control group in the large randomized double-blinded Symplicity HTN-3 trial 28 . The secondary effects beyond BP lowering of RDN have been noticed early. The renin-angiotensin-aldosterone system (RAAS) and SNS were over-activated in patients with RH. Previous experimental studies have demonstrated that RDN could block the over-activity of SNS and RAAS, and thus exert protective effects on cardiac fibrosis and remodeling [29][30][31] . In 2012, Brandt et al. reported that RDN improved LVH and diastolic function in patients with RH using echocardiography 7 . In 2014, Mahfoud et al. also found that RDN reduced LVMI in a trial using CMR 18 . Notably, these effects of RDN occurred, at least in part, independently of BP 7,12 . Following these pilot studies, many trials have been conducted to investigate the effect of RDN on LVH and diastolic function. This is the first meta-analysis on RDN and cardiac structure and function. Prior to this analysis, there have been many meta-analyses assessing the BP-lowering effects of RDN 32,33 , and only one involved the influence of RDN on LVH and atrial size 34 . That study, however, has included conference abstracts, of which the ultimate results are not available. Differing from that study, the present study included only full-text article and evaluated not only LVH but also LV diastolic function change following RDN. Furthermore, more important relevant trials, including two recent RCTs, were included in the current study, which may increase our understanding of RDN on cardiac remodeling and function.
The group of trials in the present analysis is rather heterogeneous in many respects: 1) Some trials are observational, some are RCTs; 2) Imaging was performed by CMR or echocardiography; 3) Follow-up dates ranged from 6 months to 12 months. Therefore, it is of no surprise that we separately pooled data according to study design, imaging technology and follow-up duration.
Different cardiac imaging may produce inconsistent results 35 . In our analysis, we found that both echocardiography and CMR revealed a reduction in LVMI at 6 months, indicating a consistency of echo and CMR in LVMI measurement at this time point. However, we also observed a discrepancies between echo and CMR at 12 months regarding LVMI. The small number of studies involved in CMR analysis at one year might explain this. Our study indicates a requirement for more studies with CMR analysis and long follow-up visits. Follow-up duration is also a pivotal factor that influence the final results. We found that LVMI change at 12 months was more remarkable than that at 6 months, suggesting that the effect of RDN on LVH may be sustained up to one year. With regards to cardiac function, both E/e' and EF were improved at 6 months, but none of them changed at 12 months. One potential explanation is that the number of trials included at 12 months was too small (only 2) 21,24 , and therefore it may be difficult to derive a conclusive analysis. More future studies should consider a longer follow up duration.
Study design is the most important factor that may impact the results. In observational studies, we found that LVH and diastolic function significantly improved, however, these results cannot be reproduced in RCTs. One reason is that observational studies have overestimated the treatment effects. Another explanation is that the two RCTs included all featured a small sample size. Large RCT remains to be needed.
There were several studies demonstrating some subgroup patients without blood pressure lowering still featured an improvement in LVMI 7,18 . Our meta-regression analysis showed that the change in LV remodeling or dysfunction were not significantly associated with blood pressure lowering (Fig. 7). These indicates that RDN might improve LV remodeling or dysfunction at least partly independent of blood pressure lowering.  Difference in change (P value) P = 0.807 P = 0.962 P = 0.559 NR NR P = 0.504 Table 3. Changes in cardiac structure and function from baseline to 12 months in randomized controlled trials. Data are presented as mean ± standard deviation or median (quartile 1, quartile 3). LVMI: left ventricular mass; E/e': ratio of mitral inflow velocity to annular relaxation velocity; LLVDD: left ventricular diameter in diastolic; BNP: B-type natriuretic peptide; EF: ejection fraction; LAVI: left atrial volume index. Other abbreviations as in Table 1.
The effects of RDN on HF have been frequently investigated in animal studies, most results are promising 31 . The REACH-Pilot study comprised of 7 patients showed a trend towards improvements in 6-min walk distance and diuretic use 36 . The study of Patel et al. included in our analysis, however, showed an unexpected negative   26 . Nonetheless, there are several concerns regarding this trial, including small sample size, no sham-control or strict evaluation of adherence to medication and application of outdated single-electrode catheter 37 . Several ongoing studies may further increase our understanding of RDN in patients with HF 38 .
The paper of Rosa et al. reported one-year outcome of the randomized, controlled PRAGUE-15 study 27 , which assessed the role of adding spironolactone and RDN in RH. The authors found that RDN was not superior to intensified pharmacological treatment in improving high BP, LVH and diastolic function. Notably, the cardiac structure and function change was not the primary endpoint in the study, and sufficient number of ablations was not achieved in all patients. Similarly to the RDT-PEF trial, this study lacked a sham RDN procedure. More RCTs focusing on cardiac structure and function are required to further clarify this issue.

Study limitations.
Several drawbacks of the present study should be noticed. First, both observational studies and RCTs were included in the analysis, especially only two RCTs were finally included. Generally, RCTs constitute a higher level of data quality. Although we pooled and analyzed data from the two types of study separately, the conclusion should be interpreted with some caution. Second, despite the evaluation of diastolic function and EF, the role of RDN in HF with reduced EF was not estimated. The main reason is because there are now too few studies involving patients with reduced EF to analysis. Last, life changes may play an important role in blood pressure management, and a more complex treatment might be more effective in treating those with RH. Although RDN represents a novel and promising interventional strategy, we expect future trials to compared life changes or a more complex treatment with RDN to provide some new insights.

Conclusions
Both LVH and cardiac function improved at 6 months following RDN. However, current evidence failed to show that RDN was superior to intensive (optimal) drug therapy in improving cardiac remodeling and function. More RCTs focusing on the impact of RDN on cardiac remodeling and function are needed. Actually, there are several ongoing trials investigating the role of RDN in cardiac remodeling and HF (NCT01870310, NCT01534299 and NCT02115230), and we expect the publication of the results of these trials to provide more information and new insights into this challenging subject.