FDA NeuroPace Panel Presentation (2-21) FINAL · Deaths 35 Study Days Implanted Days of Stimulation...

1

FDA Introduction

Kristen A. Bowsher, Ph.D.FDA/CDRH/ODE/DNPMD

NeuroPace RNS® System for Epilepsy (P100026)Neurological Devices Panel

February 22, 2013

2

P100026 FDA Review Team• Kristen Bowsher, Ph.D. - Engineering• Ann Costello, Ph.D., D.M.D. – Clinical• Molly Ghosh, Ph.D. - Toxicology• Joseph Jorgens, D.Sc. – Software• Wolfgang Kainz, Ph.D. - MRI• Scott Miller, Ph.D. – Statistical• Lawrence Rodichok, M.D. - Clinical • Myra Smith, MS – Team Leader/Microbiology• Federico Soldani, M.D., S.M., Ph.D. - Postmarket• Donald Witters, Ph.D. & Joshua Guag – EMC/EMI

3

• Clinical Trial for RNS® System for Epilepsy (IDE G030126)

• NeuroPace, Inc. premarket approval application (PMA) (P100026)

» Safety and Effectiveness Data

• Panel Input and Recommendations

Purpose of Meeting

4

• Introduction• Device Description

» Kristen Bowsher, Ph.D.• Clinical Background, Study Overview, and Safety Data

» Lawrence Rodichok, M.D.• Effectiveness Data

» Ann Costello, Ph.D., D.M.D• Statistical Summary

» Scott Miller, Ph.D.• Post-Approval Study Considerations

» Federico Soldani, M.D., S.M., Ph.D.• Summation

» Lawrence Rodichok, M.D.

FDA Presentation Outline

5

NeuroPace RNS® System

Implanted Components

6

• Primary external components include a clinician programmer and a patient remote monitor

• Detection Algorithm» 3 Detection Tools (area, line-length, & bandpass)

• Stimulation Output Settings

• Adequate data not provided for MRI compatibility

NeuroPace RNS® System

7

FDA Overview of Epilepsy

Lawrence Rodichok, M.D.CDRH/ODE/DNPMD


February 22, 2013

7

8

Clinical Features of Partial Epilepsy

• Mesial temporal origin» Typical clinical features» 5-15 seizures per month» Other diagnostic studies consistent with

mesial temporal focus• Extra-temporal origin

» Clinical features differ» Many seizures per day» Localization can be difficult

8

9

Treatment Options

•Pharmacologic:» 60-70% with sustained seizure freedom for one year

or more» 12-25% difference between active and placebo

Median reduction in seizure frequency, and50% responder rate

» 1.5-3.6% likelihood of seizure freedom with new drug•VNS

» 13 to 17% difference in mean reduction with active compared to “low” stimulation

9

10

Consequences of Poorly Controlled Epilepsy

• Overall increased mortality• Increased SUDEP rates• Injuries

» Head injury in 2.7% of seizures» Subdural or epidural hemorrhage in 1 per

15,000 seizures

Russell-Jones and Shorvon, 1989

10

11

SUDEP• Generally accepted criteria

» Deaths reviewed by qualified committee in NeuroPace trials

• SUDEP rates per 1000 patient-years » 6 – 9: surgical candidates» 1 – 3: AED trials

• Risk reduced by effective treatment

11

12

FDA Overview of Clinical Studies



February 22, 2013

12

13

Overview of NeuroPace Studies

• Feasibility study• Pivotal study

» Blinded phase» Open label phase

• Long term treatment (LTT)

13

14

Feasibility Study

• Inclusion/exclusion criteria comparable to pivotal• First 4 subjects per site open label

» Then randomize 1:1» 70 enrolled» 65 subjects implanted» 42 open label; 23 randomized

• Stimulation start within 28 days• Effectiveness period: last 84 days• Responder rate >13% adequate to justify pivotal

trial14

15

Feasibility Study Results

• Safety» SAE to 1 month: 6.2% (upper CB 14.8%)» SAE to month 3: 9.2% (upper CB 18.7%)

• Responder rate» Open (N=42): 27%» Open+Active (N=51): 24%» Sham (N=14): 36%

• All groups met pre-specified futility assessment to proceed to pivotal trial

15

16

Pivotal Study

• Double blind, randomized, multicenter trial• Inactive sham• Changes from feasibility

» 3 seizures per month; sensory excluded» Post-implantation stabilization – 28 days» Stimulation optimization period – 28 days» Blinded evaluation period – 84 days

16

17

Sample Size

• Sample size based on 50% responder rate: responder rate in the treated group of 40%, a rate of 20% in the sham group for a difference of 20%.

• Minimum 180 subjects to complete Blinded Evaluation Period (BEP)

17

18

Key Inclusion Criteria

• Diagnostic testing localized no more than two epileptogenic regions» No specifed types or concordance of studies

18

19

Key Exclusion Criteria

• Non-epileptic seizure like events• Status epilepticus (SE) in the preceding year• Primary generalized epilepsy• Active psychosis, major depression or

suicidal ideation in the last year• Active Vagus Nerve Stimulation (VNS)• Need for MRI scan

19

Study Periods

Baseline Period

Post-operative

Stabilization Period

Stimulation Optimization

Period

Blinded Evaluation

Period (BEP)

Open Label Period

84 days 28 days 28 days 84 days 84 weeks

Implantation Randomization

20

21

Implantation• Up to 4 leads; no more than 2 depth leads• No specified criteria for placement• No specified criteria for modification of

detection parameters• Sensitivity and specificity of detection

algorithm during the trial not reported• No specific directions for modification of

initial programming parametersPanel Questions 4b & 4c

21

22

Disposition of Subjects(June 4, 2010)

• 240 patients enrolled• 191 subjects implanted and randomized

» One death (sham)» One withdrawal (treatment)

• 187 of 189 completed end of Blinded Evaluation Phase (BEP)

22

23

Baseline Clinical Characteristics

Balanced for» Age (years): 34.9 ± 11.6 (18 - 66)» Gender: 48% female» Epilepsy duration (years): 20.5 ± 11.6 (2 - 57) » Number of AEDs: 2.8 ± 1.2 (0 - 8)

23

Baseline Seizure Frequencyper Month

Treatment Sham

N Mean Median(Min - Max) N Mean Median

(Min - Max)

BaselinePeriod 97 33.5 8.7

(3.0 - 294.7) 94 34.9 11.6(3.0 - 338.0)

2424

25

Subsets of Interest

1 p-value per chi-square2 Characteristics used as strata in adaptive randomization algorithm

CharacteristicsAll Implanted

(N = 191)

By Randomization Group

Treatment(N = 97)

Sham(N = 94) P-value1

Mesial Temporal Lobe Only (v. other)2 50% 49% 50% 0.943

Bifocal (v. unifocal)2 55% 49% 62% 0.089Prior therapeutic surgery for epilepsy2 32% 35% 30% 0.437

Prior EEG monitoring with intracranial electrodes 59% 65% 53% 0.098

Prior VNS 34% 31% 36% 0.443Anatomical brain abnormality 67% 68% 66% 0.759

Benzodiazepine use (acute) 36% 31% 41% 0.129

25

26

Baseline Seizure Frequency by Subgroup

% ofSubjects

Pre-Implant Seizure Frequency

(Mean ± SD, seizures/month)Seizure Onset ZoneMesial Temporal Lobe 50% 15.8 ± 26.8Other 50% 52.4 ± 79.3

Number of Seizure FociUnifocal onset 45% 53.8 ± 84.7Bifocal onset 55% 18.5 ± 25.3

Prior Therapeutic Surgery for EpilepsyPrior Surgery 32% 56.4 ± 85.3No Prior Surgery 68% 23.5 ± 43.2

26

27

Leads ImplantedAll Implanted

(N = 191)

By Randomization GroupTreatment

(N = 97)Sham

(N = 94)Number of Leads: % of Subjects

1 0% 0% 0%2 58% 57% 59%3 14% 14% 13%4 29% 29% 29%

Types of Leads: % of SubjectsCortical Strip Leads Only 31% 31% 31%Depth Leads Only 39% 37% 40%Cortical Strip and Depth Leads 30% 32% 29%

27

FDA Overview of Safety Results



February 22, 2013

2828

Primary Safety Endpoint

Proportion of subjects with a Serious Adverse Event (SAE) during:

» acute post-implantation – 28 daysUpper Confidence Bound (CB) not to exceed 20%*

» short term chronic post-implantation – 12 weeks

Upper CB not to exceed 42%*

29

* see references in executive summary

29

Primary Safety Endpoint

Pivotal %[upper CB]

Pooled %[upper CB]

Comparator %[upper CB]

Acute(Surgery-Week 4)

12.0% [16.5%]

10.5%[14.9%]

15%[20%]

Short-term Chronic(Surgery-Week 12)

18.3%[23.4%]

16.0%[21%]

36%[42%]

3030

Adverse EventsImplantation to 20 Weeks

% of subjects reporting ≥ 1 AE, Implant through BEP2 (20 Weeks)

Treatment(N=97)

Sham(N=94) P-Value1

Serious AEs 16.5% 23.4% 0.377

Non-Serious ("Mild") AEs 92.8% 89.4% 0.917

Total AEs 92.8% 93.6%1 Comparison of percentage of subjects with AEs, Treatment vs. Sham, per Fisher's exact test2 Blinded Evaluation Period

3131

SAEs by Study Period

Study PhaseStudy Time

Period(weeks)

Treatment Group% (subjects)

[#events]

Sham Group% (subjects)

[#events]Post-operative Stabilization 0-4 12% (23/191)

[27]Stimulus Optimization 4-8 6.2% (6/97)

[10]6.4% (6/94)

[7]Blinded Evaluation 8-20 4.2% (4/96)

[4]5.4% (5/93)

[7]

Total 0-20 16.5% (16/97) 23.4% (22/94)

3232

SAEs During Post-operative Stabilization

• Implant site – 7» 5 infections» 1 sub-galeal effusion» 1 discharge

• Intracranial hemorrhages – 4» Extradural: 2» Subdural: 1» Intraparenchymal: 1

Panel Question 1a3333

SAEs by study period

34

Study PhaseStudy Time

Period(weeks)

Treatment Group

% subjects[#events]

Sham Group% subjects[#events]

Post-operative stabilization 0-4 12% (23/191)

[27]

Stimulus optimization 4-8 6.2% (6/97)

[10]6.4% (6/94)

[7]Blinded Evaluation 8-20 4.2% (4/96)

[4]5.4% (5/93)

[7]

Total 0-20 16.5% (16/97) 23.4% (22/94)

34

Deaths

35

StudyDays

ImplantedDays of

StimulationSUDEP / Cause of

Death as AdjudicatedStimulation Status at

Time of Death

Feasibility 473 361 Definite SUDEP Enabled

Pivotal 594 450 Possible SUDEP Enabled

Pivotal 36 0 Possible SUDEP Disabled

Pivotal 351 323 Definite SUDEP Enabled

Pivotal 324 295 Definite SUDEP Enabled

Pivotal 230 25 Suicide Enabled

Pivotal 371 344 Lymphoma Enabled

LTT 856 459 Definite SUDEP Disabled

LTT 1139 533 Suicide Disabled

LTT 1295 1147 Status epilepticus Enabled

LTT 1276 1129 Probable SUDEP Enabled

Panel Question 1b 35

SUDEP

* (Leetsma et al., 1997); ** (Ryvlin et al., 2011); *** (Dasheiff, 1991)

36

RNS® System Trials Population SUDEP Rate(per 1,000 implant years)

As of June 4, 2010 8.5 [95% CI: 3.8 – 18.9]

As of October 24, 2012 6.6 [95% CI: 3.0 – 14.8]

Medically Refractory Populations SUDEP Rate(per 1,000 patient years)

Candidates for AED trials * 3.5

Meta-analysis of AED trials ** 6.9 [95% CI: 3.8 – 11.6]

Candidates for epilepsy surgery *** 9.3

Panel Question 1b36

Pooled Safety Population

• All implanted who are or were participating in RNS studies as of submission cut-off date (June 4, 2010)» 59 feasibility subjects completed» 98 pivotal subjects had completed

• 256 subjects over 708.1 implant years and 632.3 patient stimulation years

3737

Safety at One YearPooled Safety Population (N=256)

SAEs in 34.0%» 161 events in 87 subjects» Most common: % of population [# of events]

Complex partial seizures increased: 4.7% [13]Implant site infection 2.3% [7]EEG monitoring: 2.3% [6]Tonic clonic seizures increased: 2.0% [6]Tonic clonic seizures exacerbated: 2.0% [5]Device lead damage: 2.0% [5]

38Panel Question 1b

38

Intracranial HemorrhagesPooled Safety Population (N=256)

39

Adverse Event

All Adverse Events(Serious and Non-

Serious)

SeriousAdverse Events

Preferred Term

% Subjects with events(# subjects)

# Events% Subjects with SAEs

(# subjects)# SAEs

Summary of all intracranial hemorrhage and hematoma events 5.1% (13) 13 4.3% (11) 11

Cerebral haemorrhage 2.0% (5) 5 1.6% (4) 4

Extradural haematoma 1.2% (3) 3 0.8% (2) 2

Subdural haematoma (dts*) 1.2% (3) 3 1.2% (3) 3

Subdural haematoma 0.4% (1) 1 0.4% (1) 1

Traumatic intracranial haemorrhage (dts*) 0.4% (1) 1 0.4% (1) 1

* dts = due to seizuresData as of June 4, 2010. Panel Question 1b 39

InfectionsPooled Safety Population (N=256)

40

Adverse Event All Adverse Events(Serious and Mild)

SeriousAdverse Events

Preferred Term

% Subjects with events

(# subjects)# Events

% Subjects with SAEs (#

subjects)# SAEs

Summary of all implant or incision site infections 8.6% (22) 26 5.9% (15) 17

Implant site infection 5.5% (14) 16 4.7% (12) 14

Incision site infection 2.7% (7) 8 0.4% (1) 1

Implant site infection (dts*) 0.8% (2) 2 0.8% (2) 2

* dts = due to seizuresData as of June 4, 2010. Panel Question 1b 40

Other SAEs of Special InterestPooled Safety Population (N=256)

41

SAE Group

Serious Adverse Events% Subjects with

SAEs(# subjects)

# SAEs

Changes in seizures 14.8% (38) 64Seizure-related injury 8.2% (21) 28 Psychiatric 7.4% (19) 30Status epilepticus 3.5% (9) 17

Panel Question 1b41

Changes in SeizuresPooled Safety Population (N=256)

42

New Increased Exacerbated% subjects[# events]

% subjects[# events]

% subjects[# events]

Simple partial motor - 1.6% [9] 0.4% [1]

Complex partial 0.4% [1] 6.3% [18] 1.2% [3]

Generalized tonic clonic 0.4% [1] 5.9% [19] 3.5% [9]

Panel Question 1b42

Seizure Related InjuriesPooled Safety Population (N=256)

• Non-serious:46.1%» Serious 8.2%

• Contusion: 16%» 0.4% serious

• Skin laceration: 16.0%» 2.0% serious

• Head injury: 9.0%» 0.8% serious

43Panel Question 1b 43

Psychiatric Adverse EventsPooled Safety Population (N=256)

• 15 SAEs in 10 subjects related to suicidality• 2 completed suicides• 4 events of an acute psychosis in 2 subjects

44Panel Question 1b

44

Status EpilepticusPooled Safety Population (N=256)

• 18 events; 3.9% of subjects» 17 in 9 subjects serious

• One subject with 9 episodes

4545

Safety Conclusions• Peri-procedural complication rate

» Met pre-specified upper bound» No worse than the risks of other implantations

in brain• No apparent risk of neurostimulation• Long term safety

» Compatible with poorly controlled epilepsy» Concern for intracranial hemorrhage due to

seizure/injury and seizure related injuries

Panel Questions 1a-c, 44646

FDA Effectiveness Presentation

Ann Costello Ph.D., D.M.D.FDA/CDRH/DNPMD

47


February 22, 2013

Prespecified Primary Effectiveness Endpoint

• Demonstrate superiority of the Treatment group over the Sham group in reducing the frequency of total disabling seizures during the BEP.

• Modeled seizure count data using generalized estimating equations (GEE) model

48

Adaptive Randomization• Used to balance variables that might influence the

clinical response

• Variables (listed in order of priority) » Investigational site» Seizure onset zone location » Number of seizure foci» Previous resection

49

Mean, Median and Range of Seizures during Baseline and Blinded Evaluation Period (BEP)

Treatment Sham

N MeanMedian

(Min - Max) N MeanMedian

(Min - Max)

BaselinePeriod 97 33.5 8.7

(3.0 – 294.7) 94 34.9 11.6(3.0 – 338.0)

BEP 96 22.4 5.8(0 – 247.0) 93 29.9 7.6

(0 – 799.0)

Difference 11.5 2.7 5.0 4.6

50

Range of Seizures per Month

51

Range of Seizures per Month

52

Observed Mean Seizure Counts

Panel Question 2

Months0-1 1-2 2-3 0-1 1-2 2-3 3-4 4-5

53

Baseline Post-op Blinded Evaluation Period


Panel Question 2

Months0-1 1-2 2-3 0-1 1-2 2-3 3-4 4-5

54



Panel Question 2

Months0-1 1-2 2-3 0-1 1-2 2-3 3-4 4-5

55


StimulationON


Panel Question 2

Months0-1 1-2 2-3 0-1 1-2 2-3 3-4 4-5

56


StimulationON

Observed Mean & Median Seizure Counts

Months0-1 1-2 2-3 0-1 1-2 2-3 3-4 4-5

Mean

Median

57



Months0-1 1-2 2-3 0-1 1-2 2-3 3-4 4-5

Mean

Median

58


SurgicalEffect


Months0-1 1-2 2-3 0-1 1-2 2-3 3-4 4-5

Mean

Median

59


SurgicalEffect

Primary Effectiveness AnalysisPrespecified Generalized Estimating Equations

(GEE) Model• Daily seizure counts• Poisson distribution• Compare 84 days of the baseline period to 84 days

of the BEP • Two standard error estimation methods: empirical

and model-based• Model-based p <0.0001• Empirical p = 0.15

Panel Question 260

Discussion of Effectiveness EndpointNeuroPace and FDA

• Discussion followed unblinding of the data and analysis of the prespecified primary effectiveness endpoint

• Agreements» Difference in p-values indicated poor fit of the model to the

data» Alternative analysis model needed » Both the pre-specified and alternative analysis models

could be provided in PMA61

NeuroPace’s Post-hoc Effectiveness Analysis

• Daily → Monthly seizure counts• Poisson → Negative binomial distribution • Add the clinical covariates

» Seizure onset» Number of seizure foci» Prior resective surgery

62

NeuroPace’s Post-hoc GEE Model

Percent Change from Pre-Implant Period [95% confidence interval]

Treatment -37.9%[-46.7%, -27.7%]

Sham -17.3%[-29.9%, -2.3%]

Difference -20.6%

63

Model-based p-value = 0.0056Empirical p-value = 0.012

Uncertainties with Post-hoc Primary Effectiveness Analysis

• Prespecified effectiveness endpoint: » Model-based: p<0.0001 » Empirical: p=0.15

• Post-hoc effectiveness endpoint:» Model-based: p=0.056» Empirical: p=0.012

• Some alternative post-hoc GEE models do not achieve statistical significance.

• None of the prespecified secondary endpoints achieve statistical significance

• No observed data analyses achieve statistical significance64Panel Question 2

Prespecified Secondary Effectiveness Endpoints

• Responder analysis (50% reduction from baseline)

• Change in mean seizure frequency• Proportion of seizure-free days• Self-reported seizure severity according to the

Liverpool Seizure Severity Scale inventory

65

Secondary Effectiveness Endpoint Results

The study was powered on the expectation of a 20 % difference inresponder rates between the treatment and sham groups.

No interim analysis was performed.

Effectiveness Endpoint Treatment Sham P-value

50% Responder Rate 29% 27% 0.727

% Change in Mean Seizure Frequency -24% -17% 0.238

% Change in Days with Seizures -18.9% -18.3%Not

Reported

Liverpool Seizure Severity Score -4.7 -5.8 0.574

Panel Questions 2 & 3a(i)66

Post-hoc AnalysisMonthly Responder Rates

BEPTreatment

% Responders(n/N*)

Sham% Responders

(n/N*)Difference P-value

Expected 40% 20% 20%

Entire BEP 29% (28/96) 27% (25/93) 2% 0.727

Month 2-3 34% (33/96) 39% (36/93) - 5% 0.536

Month 3-4 39% (37/95) 31% (28/90) 8% 0.264

Month 4-5 34% (32/95) 30% (27/91) 4% 0.557* N represents the number of subjects for which any seizure data are

available for that time period.

67

Observed Median Percent Changein Seizures

Effectiveness Endpoint Treatment Sham Difference

% Median Change in Seizures - 28% - 19% - 9%

Panel Question 3a(i)68

Percent of Subjects with Increase or Decrease in Seizure Counts during BEP

Seizures

TreatmentN=96

% Subjects(#subjects)

ShamN=93

% Subjects(#subjects)

Difference

No change or increase 24% (23) 30% (28) - 6%

> 0 to < 50% reduction 47% (45) 43% (40) 4%

≥ 50% reduction 29% (28) 27% (25) 2%

> 90% reduction 5% (5) 4% (4) 1%

69

Prespecifed Additional Study Measures

• Subset analyses of clinical covariates » Seizure onset zone » Number of seizure foci» Prior resective surgery

• Quality of Life in Epilepsy (QOLIE)-89

70

Seizure Onset ZoneResponder Rates

Treatment Sham Difference(Treatment – Sham)

Mesial Temporal Onset (N=95)

Responders 33% (16/48) 26% (12/47) 7%

Other Onset (N=96)

Responders 24% (12/49) 28% (13/47) - 4%

Panel Question 3a(ii)71

Seizure FociResponder Rates


One Seizure Focus (N=85)Responders 31 % (15/49) 28% (10/36) 3%

Two Foci (N=106)

Responders 27% (13/48) 26% (15/58) 1%

72Panel Question 3a(ii)

Prior SurgeryResponder Rates


No Prior Resection (N= 129)Responders 32% (20/63) 26% (17/66) 6%

Prior Resection (N= 62)Responders 24% (8/34) 29% (8/28) -5%


QOLIE-89Proportion of subjects with > 5 point improvement

Treatment Sham P-value

QOLIE-89 Overall Score 36.6% (34/93) 39.1% (34/87) 0.760

1 Year 2 Years

QOLIE-89 Overall Score 38% (63/166) 44% (68/154)

Panel Question 3a(i)74

Seizure SubtypesResponder Rates

Seizure TypeTreatment

(N=96)Sham(N=93)

Difference(Treatment – Sham)

SimplePartial- motor 55% (11/20) 29% (6/21) 26%

Complex Partial 31% (26/83) 38% (31/81) - 7%

Generalized Tonic-clonic 46% (19/41) 33% (14/42) 13%

Total Disabling 29% (28/96) 27% (25/93) 2%


Sham Group Observed Mean Seizure Counts (Baseline through month 9)

Baseline 0-1 4-53-42-31-2 8-97-86-75-6

76

Months Post-opPost-op Blinded Evaluation

PeriodOpen Label Months 5 - 9

Panel Question 3b


Baseline 0-1 4-53-42-31-2 8-97-86-75-6

77



Panel Question 3b


Baseline 0-1 4-53-42-31-2 8-97-86-75-6

78



Panel Question 3b

Baseline 0-1 4-53-42-31-2 8-97-86-75-6

79

StimulationON



Panel Question 3b


Observed Mean Seizure Counts in Sham Comparison of Months 6 to 9 to Baseline and BEP

Time Period

Mean Seizure Frequency

Months 6-9 to Baseline(N = 91)

Month 6-9 to BEP(N = 91)

Change - 7.8 - 2.5

p-value 0.04 0.39

80

Sham subjects receive stimulation for the first time at month 5.

Open Label DataObserved Mean Seizure Frequency Over Time with 95% Confidence Intervals

81

Baseline BEP 5-8 8-11 11-14

20-2317-2014-17 23-26

Open Label Panel Question 3b


82


20-2317-2014-17 23-26

Stimulation ON



83


20-2317-2014-17 23-26

Stimulation ON

Stimulation ON



84


20-2317-2014-17 23-26

Stimulation ON

Stimulation ON


Concerns with Open Label Data

• Subjects know they are receiving stimulation which may cause them to overestimate the benefit

• Regression to the mean• Changes to AEDs• Missing data/drop outs

85Panel Question 3b

AED Changes in Open Label Period (N=183)

AED Change % Subjects (N)Decrease 7.7% (14)Increase 21.9% (40)Increase and Decrease 16.4% (30)No change 54.1% (99)

Panel Question 3b86

All subjects continue to take AEDs.

Subject Discontinuations (43/256)• 8 Infection or Hemorrhage• 9 Deaths • 3 Lost to Follow-up• 23 Elective

» 14 Pursue other treatments» 4 Reduction in seizures not sufficient » 3 Not wish to undergo IPG replacement » 1 withdrawn because of psychiatric issues » 1 reason for withdrawal not specified.

87

FDA Statistical Presentation

Scott W. Miller, Ph.D.FDA/CDRH/OSB/DBS

88


February 22, 2013

Outline• Overview of trial• GEE model

» What is it measuring?» Improvement metrics

• 3 sources of uncertainty:1. Alternative GEE models2. Differential response by baseline seizure

count3. Influential Sham subjects

• Statistical conclusions89

Overview of Clinical Trial• 2 arm, randomized (1:1) concurrent sham-control

double-blind clinical trial• Sample size based on secondary outcome

» 50% responder rate in seizure count from baseline• Primary effectiveness endpoint

» Seizure counts over 3 months of baseline and 3 months of blinded evaluation

• Primary effectiveness analysis» Generalized Estimating Equations (GEE) for

longitudinal count dataAnalyzes mean seizure counts over time on the same subjects

90

Overall Mean Number of Seizures

Phase MonthTreatment

(N=97)Sham(N=94)

Seizures per Month

Baseline (BL)

-3 34.5 28.7-2 34.3 34.5-1 31.7 41.5

Blinded Evaluation

Period (BEP)

2-3 22.9 27.13-4 22.8 28.9

4-5 21.4 35.4

91

92

GEE Model – What is it measuring?

92

9393

GEE Model – What is it measuring?

94

GEE Model – Difference in Slope

94

Time

Group*Time

GEE Model: Standard ErrorsTwo standard error estimation methods:• Model-based

» Assumes variance of model correctly specified• Empirical / robust

» Calibrates model-based estimate with variance observed in the data

» Robust to mis-specification of variance of model» Potentially less efficient in some settings

• Two methods can give different results» Different results sometimes indicate poor model fit

95

Study Results

• The GEE model estimates the mean seizure counts in the BEP relative to baseline

• Ratio of seizure frequency: BEP to baseline (BL)*» Treatment: 62.1% of BL level (37.9% reduction from BL)» Sham: 82.7% of BL level (17.3% reduction from BL)

• Relative rate ratio (RR) of Treatment to Sham: » 62.1/82.7 = 75%

96

* Numbers from Table 41 of FDA Executive Summary

3 Sources of Uncertainty

1. Alternative GEE models

2. Differential response by baseline seizure count

3. Influential Sham subjects

97

98

1. Multiple Plausible GEE Models:

Panel Question 2Model-based estimateEmpirical estimate

2. Differential response by baseline seizure count:

• Does response to treatment vary by baseline seizure count?

• Statistical test of interaction» Inference on interaction terms known to have lower

power than main effects» Interactions typically need to be visualized to see

what the issue is• Grouped baseline into 4 categories by 3 cutoffs:

» 0-1 seizure per day at baseline, 1-2, 2-3, >3» Goal is descriptive, not formal statistical claim

9999

100


101


S: N=69T: N=71

Panel Question 2101

102


S: N=69T: N=71

S: N=12T: N=8

Panel Question 2102

103


S: N=69T: N=71

S: N=12T: N=8

S: N=6T: N=6

Panel Question 2103

104


S: N=69T: N=71

S:N=12T: N=8

S:N=6T: N=6

S: N=7

T: N=12

Panel Question 2104

105

Overall Response Varies by Baseline Seizure Count

N=19

N=12

N=20

N=140

Panel Question 2Model-based estimateEmpirical estimate

3. Influential subjects:• GEE

» Observation out of range of others» Subject with a different response pattern than

others• Response patterns in baseline seizures

subgroup graphs» Return to baseline in Sham in >84 subgroup» Possibility that these subjects may be influential

106Panel Question 2

107Panel Question 2

2 Sham Subjects Respond Differently

108

How influential are these 2 subjects?Overall

Panel Question 2

109

How influential are these 2 subjects?By Baseline Seizure Frequency

Panel Question 2

Summary Results

110

N=191

N=172

N=189

Model-based estimateEmpirical estimate

Statistical ConclusionsStatistical Uncertainty:• Pre-specified primary analysis model

» Not statistically significant by empirical estimate» Not a good fit to data

• Post hoc analysis model» Appropriate but not only possible model» Selected after being un-blinded

111Panel Question 2

Statistical Conclusions (2)Statistical Uncertainty:• Overall treatment effect

» Consistent but statistical significance dependent upon particular GEE assumptions

» Magnitude may be driven by subjects with very high baseline seizure frequency (>84)

» Magnitude sensitive to 2 influential Sham subjects

112Panel Question 2

• Secondary and additional effectiveness outcomes» Median seizure count» Responder rate» Median percent change

• Long-term data» Open-label» No concurrent control arm» Confounded with changes to AEDs» Potential for drop-out due to non-response

113

Statistical Conclusions

Panel Question 3113

Post-Approval Study (PAS) Considerations

Federico Soldani, M.D., S.M., Ph.D. Division of Epidemiology

Office of Surveillance and Biometrics114


February 22, 2013

Reminder• The discussion of a PAS prior to FDA determination of

device approvability should not be interpreted to mean FDA is suggesting that the device is safe and effective.

• The plan to conduct a PAS does not decrease the threshold of evidence required by FDA for device approval.

• The premarket data submitted to the Agency and discussed today must stand on their own in demonstrating a reasonable assurance of safety and effectiveness and an appropriate benefit/risk balance.

115

General Principles for Post-Approval Studies

• Objective is to evaluate device performance and potential device-related problems in a broader population over an extended period of time after premarket establishment of reasonable evidence of device safety and effectiveness

• Post-approval studies should not be used to evaluate unresolved issues from the premarket phase that are important to the initial establishment of device safety and effectiveness

116

Need for Post-Approval Studies • Gather postmarket information

» Long-term performance including effects of re-treatments & device changes

» Real-world device performance (patients and clinicians)

» Effectiveness of training programs» Sub-group performance» Outcomes of concern (safety and

effectiveness)• Account for panel recommendations

117

Post-Approval Study Components

• Fundamental study question or hypothesis• Safety endpoints and methods of assessment• Acute and chronic effectiveness endpoints

and methods of assessment• Duration of follow-up

118

Important Postmarket Issues• The need to collect safety and effectiveness

data on recipients of the RNS System who are being treated by physicians newly trained on implantation and management of the RNS System

• The need to gather additional patient years of data to contribute to the estimate of the rate of sudden unexplained death in epilepsy (SUDEP)

119

PAS1: Long Term Treatment (LTT) Study Study Objectives

To assess the ongoing safety and continue to evaluate the long-term effectiveness of the RNS System for its proposed intended use as an adjunctive therapy in reducing the frequency of seizures in individuals 18 years of age or older with partial onset seizures from no more than two foci that are refractory to two or more antiepileptic medications

Study Design Prospective, Non-Randomized, Multi-Center, 7-year follow-up

Sample Size 230 patients who completed the Feasibility and Pivotal Trials

Primary Endpoint

Total serious adverse event rate (device related and not device related)

Secondary Endpoints

Average percentage change in mean frequency of total disabling seizures relative to the Pre-Implant Period 120

PAS2: Proposed New Enrollment Study Title New Enrollment Study with 1-year follow-up

Study Objectives

Collect one year of safety data on recipients of the RNS System who are being treated by physicians newly trained on implantation and management of the RNS System Gather additional patient years of data to contribute to the estimate of the rate of sudden unexplained death in epilepsy (SUDEP)

Study Design Prospective, Non-Randomized, Multi-Center, 1-year follow-up

Study Hypothesis

For the primary endpoint: the total serious adverse event rate (device related and not device related) at 1 year is not worse than the total SAE rate observed in the first year of the RNS System Pivotal trial, which was 39% (74/191)

121

PAS2: Proposed New Enrollment StudySample Size 200 patients will be enrolled at 20 centers

Study Population

Patients 18 to 70 years of age with:• Disabling seizures (simple partial motor, complex partial and/or generalized tonic clonic seizures)• Failed treatment with a minimum of two antiepileptic medications (used in appropriate doses) with adequate monitoring of compliance and the effects of treatment, as determined by the physician investigator• Diagnostic testing as part of standard care that has identifiedno more than two epileptogenic regions

Primary Endpoint

Total serious adverse event rate (device related and not device related) at 1 year for patients treated by physicians newly trained in implantation and use of the RNS System

Secondary Endpoints

SUDEP rate

122

Assessment of Proposed Postmarket Plan and Issues for Panel Discussion /1

FDA has concerns about the New Enrollment Study as presented by the company and is seeking the panel’s input on the following points:

1. The sponsor has not proposed a newly enrolled comparison group (e.g., best medical therapy). A comparison group may indeed be critical to evaluate safety and effectiveness. (Panel Question 5a)

2. Apart from SUDEP rate, no other specific safety endpoint is proposed. Other safety endpoints in addition to SUDEP (e.g., intracranial hemorrhage and injuries) could be worth measuring in the new enrollment PAS (Panel Question 5b)

3. 1-year follow up is proposed. A longer duration of follow up (perhaps up to 10 years) may be necessary to estimate the long term safety and effectiveness of this permanently implanted device (Panel Question 5c)

4. Effectiveness data are not planned to be collected during the new enrollment study. Given that the device is a permanent implant, it would be important to monitor effectiveness in the postmarket setting (Panel Question 5d)

123

124

FDA Summation



February 22, 2013

124

FDA Summation – Safety• Procedure associated with significant risks

» No worse than implantation of intracranial or DBS electrodes

» In excess of risks with continued medical therapy• Overall risk

» Comparable to risks with poorly controlled epilepsy» Possible increased risk of intracranial hemorrhage

and injury» Data for full population complete to one year

125125

FDA Summation – Effectiveness• Significant uncertainty for the primary

endpoint analysis• Pre-specified secondary endpoints

» Not statistically significant» Do not indicate a clinically meaningful benefit

• Largest change in seizure frequency occurred prior to stimulation

• Open label uncontrolled data for effectiveness difficult to interpret

126126

Benefit to Risk Comparison

• Risks of procedure• Risks of chronic implant in epilepsy

population• Benefit uncertain• Alternatives are available

» Pharmacologic – benefit known, lower risk» VNS – benefit known, lower risk» Mesial temporal resection – benefit > risk

127

Thank You!

128

FDA NeuroPace Panel Presentation (2-21) FINAL · Deaths 35 Study Days Implanted Days of Stimulation...

Documents

Transcript of FDA NeuroPace Panel Presentation (2-21) FINAL · Deaths 35 Study Days Implanted Days of Stimulation...