Interventional procedure overview of alcohol-mediated perivascular renal sympathetic denervation for resistant hypertension
Closed for comments This consultation ended on at Request commenting lead permission
Appendix C Meta-analysis
The 5 studies in the key evidence (tables 2 and 3) were included in the meta-analysis.
Data analysis
The effects of RDN on 24-hour systolic ABPM and office systolic BP were examined using weighted mean differences, with 95% CI. The random-effects generic inverse variance model was used. The treatment effects consisted of BP reductions between RDN and sham controls (using data at a similar follow-up timepoint), and after RDN from baseline within the RDN groups (using data at the longest follow-up timepoints due to data availability and to reduce potential heterogeneity). Sensitivity analyses were conducted to explore the influence of different doses used (0.3 ml or 0.6 ml alcohol infused per treated renal artery) and hypertension stages (indicated by the numbers of antihypertensive medications taken) on the effect sizes and to investigate heterogeneity where appropriate.
Heterogeneity was assessed using Chi2 and I2 statistics. Forest plots were used to display the meta-analysis results. All analyses were performed using Reference Manager V5.
Results
24-hour systolic ABPM
Of the 5 studies, there were only 2 RCTs (Kandzari 2024; Pathak 2023) that compared RDN with sham controls and reported 24-hour systolic ABPM at 2 to 3 months postprocedure. The pooled mean difference from the 2 RCTs (n=407) showed a statistically significant difference in 24-hour systolic ABPM reduction, favouring RDN (-2.37 mmHg; 95% CI, -4.61 to -0.13; p=0.04), with no detectable heterogeneity (I2=0%; figure 1). However, the effect of RDN on BP lowering decreased at 6 months (-0.55 mmHg, p=0.6964) and was potentially inferior to that of the sham procedure at 12 months (5.3 mmHg, p=0.0775) in Pathak (2023) whereas Kandzari (2024) is still ongoing.
In terms of BP reduction after RDN from baseline within the RDN groups, 4 studies assessed and reported 24-hour systolic ABPM at the follow-up durations of 3 to 24 months. The pooled mean difference after RDN from baseline was -7.92 mmHg (95% CI, -12.25 to -3.59; n=253; figure 2a). The effect on BP reduction was statistically significant (p=0.0003), and reached the MCID of -5 mmHg. But, there was significant heterogeneity (I2=84%).
The amount of alcohol infused per treated renal artery differed across studies (0.3 versus 0.6 ml per artery). Only Janas (2020) used 0.3 ml per artery whereas other studies applied 0.6 ml per artery. A sensitivity analysis of different doses was carried out. The results showed a statistically significant reduction in 24-hour systolic ABPM in the high dose subgroup (-10.18 mmHg; 95% CI, -11.94 to -8.42; I2=0%; p<0.00001) but not in the low dose subgroup (-1.00 mmHg; 95% CI, -4.72 to 2.72; p=0.60). There was a statistically significant difference between subgroups (I2=95%, p<0.0001) as illustrated in figure 2b.
For hypertension stages, 2 studies focused on resistant hypertension, 1 study included both uncontrolled and resistant hypertension, and 1 study emphasised uncontrolled hypertension. A sensitivity analysis demonstrated a statistically significant reduction in 24-hour systolic ABPM across 3 subgroups (high medication load: -5.29 mmHg [95% CI, -14.10 to 3.52]; medium medication load: -10.00 mmHg [95% CI, -12.29 to -7.71]; low medication load: -10.60 mmHg [95% CI, -13.79 to -7.41]). This effect was also clinically relevant. There was no detectable subgroup difference (I2=0%, p=0.54) as shown in figure 2c.
Although 24-hour systolic ABPM reductions after RDN from baseline within the RDN groups were statistically significant and reached the MCID, the effect of RDN compared with sham controls was uncertain and its duration of effect was unclear. In addition, different factors might influence the effect sizes, such as the amount of alcohol infused per renal artery and hypertension stage. Furthermore, follow-up durations were relatively short. So, careful interpretation of the outcomes is needed.
Office systolic BP
The 2 RCTs also assessed and reported the office systolic BP between RDN and sham procedure. The pooled mean difference was -3.62 mmHg (95 CI, -6.74 to -0.51; I2=0%; p=0.02; figure 3) at 2 to 3 months, in favour of RDN. However, the effect of RDN on BP reduction was potentially inferior to that of the sham procedure at 6 months (1.8 mmHg) and 12 months (2.2 mmHg), although the differences were not statistically significant (both p>0.05; Pathak 2023).
For pre-post treatment effect, all the 5 studies measured and reported the office systolic BP at the follow-up periods ranging from 3 to 24 months. The pooled mean difference after RDN from baseline was -17.09 mmHg (95% CI, -22.14 to -12.03; n=271; figure 4a). This reduction was statistically significant (p<0.00001) and met the MCID of -10 mmHg, but with significant heterogeneity (I2=71%). It is noted that this direction of effect on office systolic BP reduction is consistent across all individual studies as illustrated in figure 4a.
A sensitivity analysis of different amounts of alcohol infused per renal artery demonstrated statistically and clinically significant reductions in office systolic BP for both subgroups (studies with higher dose: -13.85 mmHg [95% CI, -18.04 to -9.65], I2=57%, p<0.00001; studies with lower dose: -24.52 [95% CI -22.14 to -12.03], I2=0%, p<0.00001). There was a significant subgroup difference (I2=85%, p=0.01; figure 4b).
For hypertension stages, a sensitivity analysis illustrated a statistically significant reduction across 3 subgroups (high medication load: -22.19 mmHg [95% CI -27.07 to -17.30]; medium medication load: -12.70 mmHg [95% CI, -15.64 to -9.76]; low medication load: -11.00 [95% CI, -22.14 to -12.03]). The effect also reached the MCID of 10 mmHg in all subgroups. The difference between subgroups was statistically significant (I2=84%, p=0.002; figure 4c).
The effect of RDN on office systolic BP after RDN from baseline within the RDN groups reached both statistically and clinically significant endpoints. However, this effect diminished and was even potentially inferior when comparing with the sham procedure. So, the effect of RDN on BP control was uncertain and its duration of effect was unclear. Similar to 24-hour systolic ABPM, the effect of RDN on office systolic BP might be affected by different doses of alcohol used and varying hypertension stages.
Summary
When comparing RDN with the sham procedure, there were considerable BP reductions in the sham controls that mitigated the between group differences. Thus, the relative differences between treatments were of uncertain clinical significance and in particular the duration of effect was unclear. The large BP reductions in the sham controls indicated a placebo effect on BP lowering.
The placebo effect on BP lowering might also contribute to pre-post treatment effects within the RDN groups, even though BP reductions after RDN were statistically significant and reached the MCIDs. The pre-post treatment effects might also be affected by a Hawthorne effect, such as improvement in adherence to antihypertensive medications and change in lifestyles. Moreover, the values on baseline and post-RDN were not independent of each other, but the value for the correlation was not available and so not reflected in the results. Hence, the pre-post treatment effects were indicative only and should be treated carefully.
Across all the studies, most people were on 2 or more antihypertensive medications, and medication adherence was only measured in 2 studies with the rate of full adherence between 50% and 60% over time. Indeed, evidence suggests that a larger number of prescribed drugs is associated with a high rate of nonadherence (Azizi 2016). Due to the issue of highly variable adherence to medication across studies, usage of antihypertensive medication was not included as an endpoint and relevant data was not pooled. In addition, the variation in the amounts of alcohol infused per renal artery and medication loads relevant to hypertension stages might influence the effects of RDN. Therefore, the effects of RND on BP lowering were inconclusive and should be interpretated with caution.
How are you taking part in this consultation?
You will not be able to change how you comment later.
You must be signed in to answer questions