1. Clinical Trial Pathways for Renal Denervation in Moderate to Resistant Hypertension
Laura Mauri, M.D., M.Sc.
Brigham and Women’s Hospital
Harvard Medical School
Boston, MA

2. Disclosures - Laura Mauri, M.D., M.Sc.
Research grants from Abbott, Boston Scientific, Cordis, Medtronic, Eli Lilly, Daiichi Sankyo, Bristol-Myers Squibb, Biotronik, and Sanofi-Aventis
Consulting Medtronic, Eli Lilly, Biotronik, ReCor, St. Jude

3. Study design in medical device evaluation
Each study with a specific objective Designed to address key areas of uncertainty Tested in a population with unmet need Comparison of efficacy and safety to current standards of care Feasible to conduct Reliable to make inference from to clinical practice
What is the key objective at each phase? Feasibility – signal of efficacy, Pivotal, reliable assessment of efficacy in the intended population.

4. Early efficacy and safety Procedural guidance Pivotal
Feasibility
Proof of concept
Early efficacy and safety
Procedural guidance
Pivotal
Efficacy in intended population
Post-Market
Effectiveness and safety in practice
Spectrum of potential studies: focussed to real world:
Single arm (out), randomized non blinded, randomized blinded sham control, pragmatic trial
Take home: 1. still need randomized blinded, sham control until efficacy is clear; safety question depends on similarity to tested therapy (e.g. HTN 3 results good, but may not be generalizable). Once efficacy is clear, broader trials : e.g. I and E; endpoints, more practical (lack of blinding) are reasonable.
objectives: first show efficacy, second show safety, third what is the role in patient care? This is the biggest question that defines the study design, and the most challenging.
-is RDN better than medications in patients who remain hypertensive on a maximally dosed multidrug regimen (HTN 3) – critique is that it applies to a narrowly defined population and the control arm is not being treated as they would be under optimal non clinical trial care.
-is RDN as good as medications, but easier for patients to comply with (potential of preventing clinical endpoints, cost savings for treatment of HTN and complications) this is a non inferiority design
Question of population follows from there – it is as clinically relevant in low and high risk patients: and in both need to see early determination of efficacy before embarking on larger trials.
Prior feasibility studies provided insufficient evidence
Next set of feasibility studies must meet higher bar for efficacy
Line between objectives of feasiblity and pivotal blurred for RDN

5. Clinical Trials for Renal Denervation Device Evaluation: Objectives
Resolve uncertainty Address unmet need Understand technology Feasible conduct Reliable result
Does it work?
In whom does it work?
How does it work?
Study Execution
Avoiding Bias
Target population for a pivotal trial?

6. Clinical Trials for Renal Denervation Device Evaluation: Lessons Learned
Resistant HTN is an extreme form of HTN Improved BP control occurs in the control arm Medication changes are common Clarity on physiologic mechanism was limited Sham control is necessary
Target population for a pivotal trial?

7. Bhatt et al. NEJM 2014. Bhatt et al. NEJM 2014.
63% screen failure rate
Goal of 2 week screening was to ensure stable medications on maximal tol doses – this is important to allow stable medications during the analysis period. Was this sufficient?
Bhatt et al. NEJM 2014.
Bhatt et al. NEJM 2014.

8. Changes in Antihypertensive Medication
Denervation group
(n=364)
Sham group
(n=171)
Baseline number of medications
5.1 ± 1.4
5.2 ± 1.4
6 month number of medications
5.0 ± 1.4
5.2 ± 1.6
Medication change SV1 to SV2
18 (4.9%)
13 (7.6%)
Any medication change between baseline and 6 months
139 (38.2%)
72(40.3%)
>1 change between baseline and 6 months
119 (32.7%)
60 (35.1%)
Decreased number of medications or doses
52(14.3%)
23 (12.8%)
Increased number of medications or doses
31 (8.5%)
17 (9.9%)
Undetermined
56 (15.3%)
32 (18.7%)
Medication change related to an adverse event or symptom change
98 (26.9%)
53 (31.0%)
Medication change related to SBP <115 mm Hg
13 (3.6%)
2(1.2%)
Medication change related to SBP increase >15 mm Hg
14 (3.8%)
7 (4.1%)
Other reasons (physician discretion)
72 (19.8%)
41 (24.0%)
The screening period resulted in 5 meds on average, but changed in many, and frequently over 6m.
Kandzari D, Bhatt DL, Brar S, et al. Eur Heart J 2014 Nov 16
Data are mean (SD) or n (%); SV = screening visit

9. Medication Changes During Trial
~40% (n = 211) of trial subjects required medication changes between baseline and primary efficacy endpoint assessment:
69% of first medication changes were medically necessary
121 patients had a med change due to an adverse event
80 patients had a med change due to a drug side-effect
~69% were changes in drugs at maximally-tolerated dose
Patients With Medication Changes (%)
But medications were not stable during the analysis period of the study
Update references
N = 139
N = 72
Kandzari D, Bhatt DL, Brar S, et al. Eur Heart J 2014 Nov 16

10. Non-African American and African American Subgroups
It wasn’t just that it didn’t work in AA, it was that the sham group improved so dramatically. Why? Suggests larger hawthorne effect in AA. Not necessarily driven by race, but by medications that were difficult to comply with more frequently prescribed to AA (vasodilators). Suggests that methods to use uniform medication regimens will be important in next trials
P=0.012
P=0.641
Baseline SBP, mm Hg
179.5
178.6
180.6
183.9
Kandzari D, Bhatt DL, Brar S, et al. Eur Heart J 2014 Nov 16

11. Time since screening (days)
Trial Recruitment Improves Glycemic Control: Type II DM 3 US Studies, 611 Subjects
0.25% reduction at >28d
Effect inversely proportional to baseline HbA1C
Change in HbA1C (%)
Observed in medical trials of HTN. Not unique to this study, or HTN. Authors conclude that because effects were present in both Type I European trials and Type II (Type I with selection for good control) – not explained by regression to the mean. But altered behavior from education (not specified in protocols) or increased interest and motivation.
Time since screening (days)
Gale, E. A. Diabetes Care Dec;30(12):

12. Renal Denervation: Randomization and Blinding
Many device trials are single arm, especially feasibility
Why have a blinded sham?
Reliable determination of efficacy is necessary for RDN
Unblinded studies are appropriate when outcome is predictable in the study population, and there is minimal potential effect of placebo
For renal denervation, impact of placebo effect or changes in medication compliance could both confound assessment of efficacy if blinding absent

13. Not every medical device trial needs randomization and blinding
Not every medical device trial needs randomization and blinding. Single arm trials for coronary stents, for heart valves, why not RDN? Slide showing positive unblinded trial positive results RDN, others. HTN 2, HTN 3.

14. Key Study Design Choices
Inclusion Criteria
Choose for proof of concept but also to anticipate intended population
Screening
Ensure homogeneity of baseline BP on stable medical regimen
Randomization and Blinding/Sham
Control regimen and follow up
Protocol specified, fixed or recommended medications, with clear directive for changes
Reduce drop outs but preserve safety
Assessment of futility
Not planning to show specific designs, because there is enough uncertainty in target population
But these choices may apply to both feasilbity and pivotal trials

15. Controversial Design Choices
Target population(s)
Optimal follow-up duration
Assess maximal effect
Assess durability of effect, or late effects
Cross over
Incentive for enrollment limits ability to detect late effects
Assessing medication compliance
Assessing mechanism of action of RDN

16. Study population Proof of Concept or Intended population Denovo HTN
Medication replacement
Moderate HTN
Resistant HTN
False dichotomy of proof of concept vs intended population
Results in one population may not predict results in another, e.g. vast difference in de novo and severe resistant.

17. Clinical Trial Design for Renal Denervation Conclusions (1)
Randomization and blinding remain important
Target population and role of therapy are the most challenging questions
RDN to improve control over max tolerated medication regimen in severe resistant hypetension may be unrealistic even for an effective therapy

18. Clinical Trial Design for Renal Denervation Conclusions (2)
Next feasibility data are needed to consider effect in varied populations – potential benefit in early and late hypertension
Once efficacy is established, effectiveness can be tested more broadly (e.g. non-inferiority of replacement of medications, but easier for patients to comply with than medication, and therefore, potentially superior)
First, it may be reasonable to reconsider the target population for RDN. While patients with resistant hypertension represent an important unmet clinical need, the reduction in BP associated with trial participation, even in the sham-procedure arm, suggested that some of these patients were not truly treatment-resistant. In addition, whether RDN is best suited to the physiology of individuals on heterogeneous pharmacologic regiments of 4 drugs or more, versus those with more moderate hypertension, remains unknown, but could be of interest if procedural risks are determined to be low. Study participants with higher starting BPs have had a greater likelihood for improvement mathematically in single arm studies; however, in a controlled clinical trial, it is not known whether more moderate levels of pressure, with fewer antihypertensive agents, might be more responsive to RDN.
Finally here, BP has been an acceptable surrogate endpoint to determine the effectiveness of new pharmacologic therapies for over 4 decades. In therapies also documented to be safe, it is reasonable to expect that the same mechanism of benefit would apply to RDN. While the SYMPLICITY HTN-3 study required a superiority margin (not just a statistically significant improvement) to conclude superiority, if the risk of a candidate RDN therapy is documented to be minimal such a margin perhaps should not be a requirement. In prior studies, crossing over of study participants initially randomized to the sham procedure has been included to encourage participation, but this occurrence also prevents detection of later treatment effects.
The one determinant of trial design that must be emphasized here is the value of a sham control if BP is the primary endpoint, lest the value of RDN be overestimated without a proper control. Past device research has clearly demonstrated the importance of a control group in a well-designed cardiovascular technology trial of angina pectoris [33]. However, it is likely that the sham effect that was measured in SYMPLICITY HTN-3 is not just a placebo effect, but also the effect of differing medication adherence in a trial setting. Therefore, in considering the potential value of RDN, it may be important in future studies to consider that better “compliance” with therapy, e.g. the potentially more durable effect of RDN, may be an important value to RDN versus medical therapy in a real world setting. Finally, if RDN technologies mature in the future, noninferiority designs comparing RDN with medical therapy might be reasonable to consider as patients and physicians may place value on replacement therapy. A major challenge to this approach would be demonstrating a financial value to this strategy given the number of low cost medications that currently exist to treat hypertension.