SUMMARY MINUTES
MEDICAL DEVICES DISPUTE RESOLUTION PANEL
December 15, 2006
Hilton
MEDICAL DEVICES DISPUTE RESOLUTION PANEL MEETING
December 15, 2006
Attendees:
Chairman:
Scott D. Ramsey, M.D., Ph.D.
University of
Industry Representative:
Melissa Walker, M.D., RAC
Stereotaxis, Inc.
Consumer Representative:
Allen A. Hughes, Ph.D.
Voting Member:
Jonathan D. Sackner-Bernstein,
M.D., F.A.C.C.
Clinilabs, Inc.
Consultants:
Michael D. Crittenden, M.D.
David DeMets, Ph.D.
Warren K. Laskey, M.D.
Executive Secretary:
Nancy Collazo-Braier, Ph.D.
Ombudsman:
Les Weinstein, Esq.
Chairman Ramsey called the hearing to order at 8:03 a.m. and had the
participants around the table introduce themselves. The meeting was being held at the request of
Acorn Cardiovascular, Inc. to resolve a scientific dispute between the Sponsor
and the Office of Device Evaluation in PMA P040049 for the CorCap Cardiac
Support Device.
He
read the Ombudsman’s summary of the scientific issues in dispute, which
describes the CorCap Cardiac Support Device as a non-resorbable polyester mesh
implanted around the heart to provide ventricular support and reduce
ventricular wall stress. It is indicated
for patients with dilated cardiomyopathy who are worsening despite optimal
medical management. On August 5, 2005,
the FDA issued a not-approvable letter, citing lack of reasonable assurance of
safety and effectiveness. Acorn
amended the PMA with a post-hoc analysis to identify a sub-population in which
CorCap CSD is safe and effective. FDA
issued a non-approval determination to the amended PMA on February 2,
2006. The letter asked for clinical
validation from a prospective study on the focused cohort. The Dispute Resolution Panel was charged with
reviewing and making a recommendation to the CDRH Center Director as to the approvability
of the PMA as amended.
Executive
Secretary Collazo-Braier read the letters
deputizing Drs. Ramsey, Sackner-Bernstein, DeMets, Laskey, and Crittenden as
voting members for the duration of the meeting.
She then read the conflict of interest statement. No waivers were issued.
Mercedes K.C. Dullum, M.D., an investigator on the Acorn trial at
Robert D. Dowling, M.D., an investigator on the Acorn trial at the
Karen Becker, Ph.D., of Becker & Associates Consulting described the device and said
that there was considerable science underlying the design, including lab studies,
proof of concept studies, and animal models.
She gave a history of the device from the beginning of the trial in 1999
through the granting of the appeal. She
said that more information was available than was available to the Circulatory
System Devices Panel (CSD), since the amended PMA provides data on potential
investigator bias, supplemental information on safety at reoperation, extended
follow-up on mortality, a postmarket study design for safety and long-term
outcomes, post-hoc analysis of patients likely to benefit, and a revised
indication for use based on ventricular size.
There are also more data on the imputation methodology used, a blinded
validation of that methodology, and testimony from experts. She said that the data provides reasonable
evidence of safety and effectiveness for the device and that the statistical
methods applied to the trial are sound.
The primary endpoint result was clinically relevant and beneficial. The secondary endpoint results support the
hypothesis. The safety profile is
reasonable in relation to the benefit.
Since the original patient population adequately addressed safety and
effectiveness questions, the focused cohort analysis was post-hoc and not
necessary for approval.
Mariell Jessup, M.D., of the
Regarding safety outcomes, she said that in 2004, 81.1
percent of treatment patients and 77.6 percent of control patients had serious
adverse events. The numbers and types of
events were not significantly different between the two arms. There were no device-related events. In 2005, 83.1 percent of treatment and 78.9
percent of control patients had had serious adverse events. There was a significant difference in
mortality. She said that the device’s
reshaping effect allows the heart to function better. She said the secondary endpoints (left
ventricular end diastolic volume, left ventricular end systolic volume,
sphericity, Minnesota Living With Heart Failure Questionnaire and the SF-36)
supported the primary endpoint, that the safety profile was acceptable, and
that the conclusions provided a reasonable assurance of safety and
efficacy.
Steven Piantadosi, M.D., Ph.D., of
Donald B. Rubin, Ph.D. of
Randall C. Starling, M.D., MPH of Cleveland Clinic, an investigator and member of
the steering committee, gave a clinical comment on the primary endpoint, which
he said was statistically significant and clinically meaningful. Patients with the CorCap were more likely to
feel better, evidenced by higher NYHA scores; had fewer transplants and
ventricular assist devices; and showed no increase in mortality. Whether or not to perform major cardiac
procedures (MCP) is decided by committee, so it is unlikely that bias would
creep in. If it did and patients were
not getting needed surgeries, heart failure and mortality rates in the
treatment arm would have increased. The
study showed a clinically meaningful endpoint with no demonstrable evidence of
bias.
Dr. Piantadosi
continued the discussion of primary endpoints, explaining the stratification
used in the trial. Patients were
stratified by MVR versus no-MVR in order to reduce the variance, control
heterogeneity in the samples, and to balance the condition of the patients in
the two arms of the trial. This was done
to increase the precision of the results.
Stratification does not require or dictate stratum treatment effects be
part of the trial, since the goal of the study is to estimate the average
relative treatment effects across the strata.
Within both strata, however, CorCap is favored over control, and the
treatment effect is clinically significant.
Dr. Piantadosi
moved on to discuss secondary endpoints, starting with a statistical
comment. The secondary endpoints were
chosen for their clinical utility. They
were explicitly named as secondary outcomes, not primary. The protocol specified how they were to be
analyzed, the timepoints, and the methodology, using longitudinal regression
models. Proper interpretation of the
secondary endpoints depends upon direction and magnitude of differences, not
just p-values. Multiplicity adjustments
are not necessary or required, since the purpose is to give supportive evidence
to the primary endpoint. Secondary endpoints
are not expected to independently establish safety or effectiveness. Variations in statistical significance are to
be expected. The major secondary
endpoints are supportive of the primary endpoint. A Hochberg adjustment was done in response to
the FDA request to pre-specify a type I error rate. With the adjustment, LVEDV meets the
collective success criteria. The
treatment differences still favor CorCap.
Douglas L. Mann, M.D., of the Baylor College of Medicine spoke on the
clinical relevance of the secondary endpoints.
Secondary endpoints do three things.
First, they reflect the mechanism of the device. The device decreases transmural stress,
leading to reverse remodeling. The end
diastolic and end systolic volumes show that the device is working as
anticipated. Second, secondary endpoints
should show fidelity with other devices doing a similar thing. The end-systolic and end-diastolic volumes
reflect that. The third thing a
secondary endpoint should do is show benefit to the patient. This is reflected in the questionnaire
scores.
Steven F. Bolling, M.D. of the University of Michigan, a site PI during the
study, and Michael A Acker, M.D. of the University of Pennsylvania, a
member of the steering committee, spoke on safety, both perioperative
mortality, and risk of re-operation. Dr.
Bolling said that operative mortality in MVR CorCap patients was extremely
low, less than 2 percent in already-ill patients. There were five deaths in the no-MVR stratum,
one prior to the operation. The four
perioperative deaths were hemodynamically unstable at the time of the
surgery. These results were early in the
trial. Dr. Acker explained that
in very sick patients with low ejection fraction, the enlarged heart can become
unstable during the operation. The
deaths were reviewed in 2002, and it was decided to use cardiopulmonary bypass
and intra-aortic balloon in patients with extremely enlarged hearts. After the meeting, the mortality rate dropped
from16.7 percent to 3 percent. Adhesions
were another safety issue to consider.
While CorCap causes adhesions, adhesions are common to any surgery, and
these adhesions do not affect safety.
There were seven transplants in the CorCap group and no deaths in those
patients. Of the 16 control patients with
transplants, 2 died. The adverse event
rates are similar to control. Since
heart surgeons are used to adhesions, there is no evidence that adhesions have
significant impact on patient outcomes.
Dr. Mann
addressed the risk of pericardial constriction.
There was no evidence of it in the animal models, no consistent
echocardiograph evidence of it in patients, and no clinical evidence in the
follow-up with the trial patients.
Monitoring of the patients will continue out to five years.
Dr. Starling
gave a clinical risk/benefit analysis.
He said the device provides benefit, since the primary endpoint was
met. It is safe, since the mortality
rates are neutral and the secondary endpoints show a benefit. Functional class, quality of life, need for
major cardiac procedures (MCP), and heart size are all significantly
improved. He concluded the presentation
by pointing out that patients eligible for the device have few treatment
options and that the device represented a therapy to address an unmet need.
Aron Yustein, M.D., introduced the presentation. Bram
D. Zuckerman, M.D., gave an introduction to the FDA review. He described the device and explained that
the new indications--for use in adult patients who have been diagnosed with
dilated cardiomyopathy, are symptomatic despite treatment with optimal heart
failure medical management, have a dilated heart, and have a LVEF less than or
equal to 35 percent or 45 percent if mitral valve repair or replacement is
planned—were based on the post hoc focused cohort analysis. There are key disagreements between the FDA
and the Sponsor on the history of the process.
The primary endpoint for the Acorn trial was a composite of mortality,
additional major cardiac procedures, and change in NYHA class, but 58 percent
of patients did not have a blinded NYHA assessment at baseline. This is due to Acorn enrolling patients
before the Sponsor and FDA had reached an agreement on the primary
endpoint. Acorn changed its assessment
method for blinded NYHA during the trial due to procedural and validity
concerns. While FDA was in discussions
with Acorn about these concerns, Acorn continued to enroll patients, leading to
172 patients being enrolled prior to the implementation of the new, blinded
assessment.
FDA did not pre-specify imputation. Because the site and core lab NYHA
assessments had poor agreement, FDA proposed imputation for the missing
data. The Sponsor rejected the suggestion
as clinically invalid. The Sponsor
changed its position two months after being unblinded to the results.
Following the Advisory Panel vote of not approvable,
Acorn proposed a reanalysis of the data to find a patient profile in which the
risk benefit ratio was maximized. The
FDA said the post-hoc analysis would be useful in generating a hypothesis for
future studies and that post-hoc analysis would have to be prospective. When Acorn submitted the focused cohort
analysis, it contained no new data, so a second non-approval letter was sent.
FDA’s key concerns about the focused cohort submission
are that the primary endpoint is not interpretable, the secondary endpoint
results are problematic and not supportive of trial success, the safety
concerns raised by the FDA and Advisory committee have not been adequately
addressed, the risk-benefit profile is unacceptable in both strata, and a
focused cohort analysis must be prospectively validated with additional
clinical data.
William H. Maisel, M.D., MPH, Chair of the Circulatory System Medical Device
Advisory Panel, summarized the June 22, 2005 Panel meeting. The Panel consisted of 15 participants of
diverse relevant expertise. This Panel
voted 9-4 not approvable. The Panel had
a number of concerns: device effectiveness, both with primary and secondary
endpoint issues; safety concerns; and concerns about missing data and
subgroups. No one issue sank the
application. The combined issues led to
the determination of absence of reasonable assurance of safety and
effectiveness. In the primary endpoint,
NYHA assessment was missing over half of the baseline data, MCP was possibly
biased, and mortality was not statistically different from control. Dr. Maisel disagreed with the Sponsor’s
suggestion that the Panel did not understand imputation, noting that the
Sponsor’s statistician at the 2005 meeting had commented on the difficulty of
making assessments with so much missing data.
The bias in MCP is that the decision to operate or not is subjective. The reduction in mortality over the course of
the study was encouraging but implied that there is a learning curve to the
procedure. In the secondary endpoints,
there were significant missing data, and the patients were not blinded. He noted that these were not end-stage,
hopeless patients. One third of the
control patients got better, so there is a danger of harming patients who may
get better. The study was underpowered
to detect important differences sufficiently to identify a subgroup that might
benefit.
Ileana L. Pina, M.D., FACC,
FAHA, FACP,
gave the FDA clinical review of P040049.
The primary endpoint was a composite of mortality, major cardiac
procedure for worsening heart failure, and change in NYHA class. She elaborated on FDA’s concerns with the
PMA. For the interpretability of the
primary endpoint, she questioned the validity of the NYHA measure and noted the
35 percent of patients missing outcome data.
There was no observed benefit in mortality or NYHA. The MCP component in the no-MVR group is the
only sign of potential clinical benefit, and it could be the result of bias and
placebo effect. On the issue of
secondary endpoints, she said that there is a lack of clinical or statistical
significance in the majority of key structural and functional measures; there
is a large amount of missing data for several secondary endpoints; the
available objective data for functional endpoints actually favor the control
group; and the correlation between changes in structure and function are
low. The unaddressed safety concerns
were the high (7.8 percent) perioperative mortality compared to control, dense
adhesions and the related difficulty in reoperation, CorCap patient
ineligibility for CABG, and possible future pericardial constriction. The risk/benefit profile was acceptable in
neither strata. The focused cohort
analysis may have identified a patient population with an better risk/benefit
profile; however, this post hoc analysis would have to be prospectively tested.
Laura Thompson, Ph.D., gave a statistical review, focusing on the FDA’s
concerns with primary endpoints, secondary endpoints, and the focused cohort
analysis. Because the trial was begun
before the primary endpoint was agreed upon and the level of agreement between
site-assessed NYHA and core lab NYHA was so low, there was considerable missing
baseline information. FDA suggested
multiple imputation for the missing data, to be accompanied by the primary
endpoint analysis using complete data.
The Sponsor chose to use imputation after the data was unblinded. It is important to decide on the analyses
before unblinding to avoid bias. The
device patients did have fewer patients with additional MCP, which is where the
difference in primary endpoint occurs for the device. NYHA and mortality rates are not responsible
for statistically significant results.
Secondary endpoints only matter if the primary
endpoint was met, and the unblinding of the data prior to choosing a method of
analysis calls that primary endpoint into question. Of the secondary endpoints, only the p-value
for LVEDV is statistically significant, according to the Hochberg
procedure.
In the focused cohort analysis, the Sponsor identified a subgroup that may benefit from CorCap. The subgroup excludes high-risk patients. The same data is used to choose the cohort as to statistically test the cohort. This analysis cannot be the basis of an approval, since the perceived benefit can be chance or a Type I error. She showed how similar results can be found using data that shows no treatment difference between experiment and control.
Clyde W. Yancy, M.D., FACC, FAHA, FACP, a primary reviewer from the June 2005 meeting and a
consultant to ODE, gave the perspective from that meeting. The fundamental question was whether
surgically-induced remodeling achieved with the CorCap resulted in similar
safety and efficacy as medically-mediated reverse remodeling. The NYHA and mortality portions of the
composite primary endpoint were not helpful, and the trial was unblinded. Referral bias in the third portion of the
endpoint concerned the first Panel. The
original rationale for the device was that Class III and IV patients had few
treatment options. Today, however, there
are now several treatment options, many of which provide reverse
remodeling. The premise was that
surgical constraint of the left ventricle results in a decrement of
Many issues
concerned the Advisory Panel in 2005.
The Composite endpoint was achieved, but there was insufficient evidence
of reverse remodeling, such as changed
Since other treatment options exist, the hypothesis
was not proven, the lack of convincing data, more data was needed before moving
forward. That was the opinion of the
Panel at the meeting.
Dr. Yustein
returned to reiterate the FDA’s concerns and close the presentation.
Dr. Becker
gave the Sponsor’s rebuttal. She
reiterated that the original cohort provides valid scientific evidence to
support approval of the product. The
focused cohort study was not intended to stand alone. However, the PMA was modified at the FDA’s
suggestion, and this Panel had data the previous Panel did not have. She said the Sponsor was flexible about the
labeling.
Dr. Piantadosi
reiterated the potentially misleading pitfalls he’d described earlier:
overly-strict reliance on p-values where relative odds ratios can show
treatment effect, taking subgroup findings as primary and definitive,
deconstruction of the primary outcome, misunderstanding the role of
stratification, and misinterpretation of the potential effects of missing
data. He elaborated that missingness of
data can be informative, depending on how it is used.
Dr. Mann
addressed the question of CorCap’s role in reverse remodeling. Reverse remodeling is a biologic surrogate
for myocardial recovery. He said that
the device does lead to reverse remodeling, but the trial was too small to show
the mortality benefit. The beta blocker
trial showed mortality benefit, but with 3000 patients. The device procedure should not be compared
to SAVER or DOR procedures, since remodeling occurs at the molecular
level. The pressure volume data is not
consistent with pericardial constriction.
Dr. Becker
gave examples of missing data in other heart failure device trials to
demonstrate that the amount missing in this trial was actually less than in
most studies.
The Panel asked why the Sponsor had changed its mind
on using imputation. Steven Anderson,
Acorn Regulatory Officer, explained that the initial refusal to use imputation
was done hurriedly, based on input from statisticians who were not imputation
experts. Input was limited because the
regulatory affairs director and staff were still blinded. More information on imputation was obtained
in July. The director was still blinded
when the decision was made, in August.
Dr. Matt Hillebrenner said that the FDA had sufficient statistical
expertise and that the actions of the Sponsor or the FDA are not in
dispute. He said the issue comes down to
two questions: Do the overall treatment
results justify placement of the CorCap concomitant with mitral valve surgery?
and Do the overall treatment results justify performing a median sternotomy
just to place the device? The FDA
answers both questions in the negative.
The trial has not shown safety and effectiveness.
Dr. Sackner-Bernstein asked about the FDA’s concern about pericardial
constriction. Dr. Zuckerman said
that it was a possible safety concern that may arise later.
Dr. Robert Brewer, M.D., of
Eugene Grossi, M.D. of
The Panel was charged to answer the following
question: “Does the PMA as amended provide valid scientific evidence that
demonstrates a reasonable assurance of the safety and effectiveness of the
CorCap CSD for its intended use in the original patient population and/or the
focused cohort?” In considering that
recommendation, the Panel was asked to determine the following: “1) Whether the overall trial results for the
primary effectiveness endpoint are interpretable and clinically meaningful, 2)
Whether the secondary endpoint results are supportive of the safety and
effectiveness of the device, 3) Whether FDA’s safety concerns have been
addressed by the data provided, 4) Whether the data submitted by Acorn
adequately address FDA’s safety and effectiveness concerns for the original
patient population and/or focused cohort.”
Dr. Sackner-Bernstein asked if the use of the missing at random assumption
in the NYHA imputation was valid. Dr.
DeMets said that the assumption was strong because the missing data was
systemic. There are times in trials that
patient characteristics change. When
that happens, the assumption is bad. Dr. Laskey asked about the statistical
robustness of the categories improved, same, and worsened. Dr. DeMets said that such endpoints
are difficult to interpret.
Dr. Ramsey
asked Acorn about the low degree of clinical benefit in the endpoints. Dr. Mann said that the primary
endpoint shows good benefits if the primary endpoint is analyzed as a
whole. Dr. Jessup added that
simply not getting worse is a benefit to these patients, who are already on
maximal therapy. Dr. Starling
added that end diastolic volume and remodeling improvements improve over time,
so long-term data will show more benefit.
Dr. Ramsey further asked if the main driver of statistical
significance in the primary endpoint was the change in major cardiac
procedures. Dr. Piantadosi said
that was true but that the primary endpoint was not designed to be
deconstructed. Dr. Ramsey asked
the FDA to respond. Dr. Zuckerman
said that the Sponsor was referring to data outside of the scope of the
meeting. Dr. Hillebrenner said
that the FDA was not comfortable with the primary endpoint if the components
could not be analyzed individually.
Dr. Sackner-Bernstein asked about the mortality confidence intervals. Dr. Piantadosi said that the risk
ratios for control and device were equal, about 1.0. Dr. Zuckerman said the sample size was
chosen to protect the primary endpoint but that the confidence interval was
wide. Dr. Thompson added that it
was not a time to event analysis.
Dr. Ramsey
said that the primary endpoint is a composite of safety and efficacy data. He asked for more data on the potential bias
in the category of MCP. Dr. Mann
said that if patients had not received the procedures they required, there
would have been more Class IV heart failures in the treatment arm. Dr. Bolling added that major heart
surgery decisions are made by committee, so an individual’s bias would be
mitigated. Dr. Thompson showed
data on the inconsistent NYHA measures. Dr.
Kubo, speaking for Acorn, said that although NYHA is common, it is not
standardized, so different clinicians score differently. In the core lab, the questions were
standardized and made consistent; that is why the lab and site results were not
comparable. The patient cannot be
blinded to the questionnaire, but the assessor is blinded, since he or she does
not see the patient.
Dr. Crittenden
asked why ejection fraction did not improve.
Dr. Bolling said that EF actually improved in CorCap patients and
worsened in the control group, but the difference was not statistically
significant. MVR patients also skew the
results. Dr. Crittenden asked
about the rings used in MVR surgery in relation to reverse remodeling. Dr. Bolling said the trial had been held
before the rings were available, so any remodeling is due to the device or the
surgery. Dr. Laskey asked if the
association between treatment and outcome could be confounded by the site. Dr. Acker said there was no site
difference.
Dr. Crittenden
asked if the FDA had a response to Dr. Kubo’s statement. Dr. Pina said that the FDA had
struggled with trying to figure out what the core lab NYHA was measuring and
the difference between the core lab and site assessments. Dr. Mann added that the assessor at
the site was blinded and unable to see if there had been a surgery or not. Dr. Piantadosi said that any problem
with the assessment would be equally present in the control group. Dr. Zuckerman disagreed.
Dr. Hughes
asked if there was any difference between mechanical or biological valves in
MVR patients. Dr. Acker said
there was no difference between replacement and repair.
Dr. Sackner-Bernstein asked for the change in NYHA class over the trial in
relation to the other components of the primary endpoint. Dr. Zuckerman said the trial tried to
recruit Class III patients, but the core lab assessed 50 percent of the
patients as Class IV. Dr. Thompson
said he did not believe the relative difference between CorCap and control were
the same in site NYHA compared to core lab NYHA. Additionally, patients who had a MCP did not
have a final NYHA assessment. Dr.
Rubin said Dr. Thompson’s assessment was invalid, since it stratifies
by outcome variables. Dr. Thompson
said it did not affect the numerical result.
Dr. Sackner-Bernstein asked about the different imputation modes, wondering
how using four different models affected the results. Dr. Rubin said that it can lead to increased
variability, and sampling has to be done at the end of the calculations. Dr. Pina reiterated the differences
between the site and core lab assessed patients.
Dr. Laskey
noted that the majority of heart failure is related to coronary heart disease,
and the trial had no coronary heart disease patients. He also noted that MVR complicated the
question of cause and effect in the study.
Dr. Ramsey agreed that he could not tell if the MVR procedure and
the CorCap device were additive, synergistic, or unrelated. Dr. Sackner-Bernstein commented that
the procedure’s risk for mortality is high.
Dr. Ramsey
said that the amount of missing data that had to be imputed was
extraordinary. Dr. Hughes said
that after such data mining and complex calculations it is important to
prospectively test the findings. Dr.
DeMets said he was not troubled by the imputation so much as by the
endpoints showing potential for bias and little difference between the device
and control. Dr. Sackner-Bernstein
found the missing at random assumption troubling.
Ms. Walker
asked the Panel members if they could think of a patient subset that would
potentially benefit from the therapy. Dr.
Ramsey said the focused cohort was promising. Dr. Sackner-Bernstein said that the
next step should be to plan a validation strategy for what the focused cohort
analysis has shown. Dr. Laskey
said that until the issue of the integrity of the parent trial was resolved it
is difficult to make conclusions on subsets of the trial. Dr. Ramsey said that, despite the
missing data, the secondary endpoints are the least subject to bias. Dr. Laskey noted the closeness of the
control and treatment outcomes. Dr.
Sackner-Bernstein said that the core lab NYHA methodology minimizes bias
and adds consistency, but it should have been developed in one population and
validated in another. Furthermore, the
studies were cross-sectional, not longitudinal.
In a best-case scenario, it could not be more than 60 percent
accurate.
Dr. Donna-Bea Tillman said that Acorn has not provided sufficient valid
scientific evidence to support approval.
To determine that Acorn has demonstrated reasonable assurance of
effectiveness, the Panel would have to find that the device would provide a
clinically significant result in a significant portion of the target
population. In this trial, the results
of the primary endpoint analysis are not interpretable, and the secondary
endpoints do not demonstrate a clinically meaningful benefit. The perioperative mortality associated with
the device cannot be outweighed by the benefit, since none was shown. The focused cohort should be viewed as
hypothesis-gathering and would require additional prospective data to test the
hypothesis. This data-gathering must be
pre-market, and the FDA is willing to work with the company in that
process.
Dr. Piantadosi
said that the study and the primary endpoint were designed to be responsive to
clinical concerns and methodological rigor.
Although there are internal irregularities in the primary endpoint, the
endpoint itself was met and demonstrates substantial consistency in
effect. While there is a chance of bias,
the investigators did everything possible to mitigate that. Imputation is not the first choice, but the
multiple imputation models demonstrate that the outcomes are not sensitive to
method of imputation. Though the data is
imperfect, it is high-quality and reasonable evidence.
Dr. Jessup
said that the risk of surgery is outweighed by the benefit of avoiding
transplant, VADs, and deteriorating clinical status, especially if the device
and procedure is held to the same standard as other heart procedures.
Dr. Paintadosi
said that is a mistake to break the primary endpoint into its components, but
if it is broken down, one must note the 41 percent decrease in the need for
transplantation as a benefit. He
reiterated that the decision on transplantations could not have been due to
bias.
Dr. Rubin
pointed out that for the missing at random assumption, the assumption was used
in comparing the treatment and control, which had the same data missing due to
the time of the NYHA assessor change. He
added that some of the FDA’s analyses were not valid by FDA standards. There is evidence that the device is needed
and little that it is unsafe.
Dr. Ramsey asked the industry and consumer representatives if
they would like to comment. Ms.
Walker said the voting members should consider whether or not there were
other opportunities to explore patient sub-populations that might be proposed
in conditions. She also said that the
labeling could be made as narrow as is necessary to mitigate the Panel’s
concerns. Dr. Hughes pointed out
that the physician is a learned agent and that approving the device would give
the physician another tool. However,
using the device precludes certain future procedures. He said that biocompatibility and
biointegrity must be considered. He also
noted that the control group seemed to do better than treatment on some
secondary endpoints. He raised issues of
avoidable risk, avoidable consequences, and avoidance of future heart
transplants.
Dr. Sackner-Bernstein said that the surgical intervention
had a high mortality risk, though how high was unclear. He said index hospitalization should be
counted, since CorCap patients had longer stays. Perhaps the advantage of the device is not
that it prevents heart surgery but that it allows the time of the surgery to be
chosen.
Executive Secretary
Collazo-Braier read the Panel
recommendation options for pre-market approval applications, including the
thresholds and definitions of these possible recommend options. Dr. Ramsey called for a motion. Dr.
Crittenden moved to recommend approvable with conditions. Dr. Laskey seconded the motion.
The Chairman called for conditions of approval. Dr.
Crittenden moved as a condition of approval that the Sponsor continue the
proposed post hoc analysis and that the FDA work with the Sponsor to narrow the
labeling. He asked for his
fellow-panelists’ help in narrowing the labeling. Dr. Sackner-Bernstein asked that the
condition be narrowed to a single condition.
Dr. Crittenden restated his first motion, that the Sponsor
continue the post hoc analysis, though he said it may be better to narrow the
labeling first. The motion was not
seconded, and seeing no one willing to second it, the Chairman closed the
discussion and called for a new motion.
Dr. Laskey moved to recommend approvable with
conditions. Dr. Crittenden seconded the
motion. Dr. Laskey moved that the
condition be continued enrollment in a clinical trial with better safety
estimates in order to give more assurance that it is a safe procedure in the
patient population. Ms. Walker asked for clarification of
the motion. Dr. Laskey said that
he wanted to see better safety results than have been presented, a lower
perioperative 30 day rate that shows steadily decreasing mortality. Chairman Ramsey observed that if more
or better PMA data were needed to approve the application, the application
would not be approvable, so the condition may have been invalid. Dr. Crittenden asked whether the
motion called for new data or for clarification of existing data, and Dr.
Laskey said his motion called for new data.
The Chairman called for a second on the condition. There
being no second, he asked for other
conditions to the motion to approve with conditions. No conditions were named, and for the sake of
procedure, he called the vote to
recommend approvable with conditions, though no conditions had been voted
upon. The vote was 3-1 to oppose the motion, with Dr. Crittenden voting in
the affirmative. The Chairman
called for other motions.
Dr. Sackner-Bernstein moved to recommend that the
application is not approvable. Dr.
Laskey seconded the motion. Dr. Ramsey opened the floor for
discussion. Dr. Hughes said that
a vote of not approvable is harsh but that post-market surveillance cannot
substitute for a pre-market condition.
He said that a prospective study is needed, which cannot be done
post-market, due to regulations. Ms.
Walker said that any previously-unseen data should be reviewed and analyzed
before the application is declared non-approvable. Dr. Ramsey agreed that the Panel
recommended that the FDA follow up on unseen data. Dr. Sackner-Bernstein agreed that the
recommendation of non-approvable was harsh but there was too much uncertainty
about safety and benefit for the regulations to allow approval. Any therapy is decided upon after weighing
risk and benefit. In this case, they
could not be weighed, due to the uncertainty.
Dr Laskey said that although the Panel members were sympathetic
to the investigators, the safety concerns could not be overlooked. Dr. Crittenden pointed out that the
patients needing the device were at risk if no action were taken as well. He said that the device was probably easier
for surgeons to approve and that a symptomatic benefit and avoidance of secondary
procedures had been shown. The Dr.
Ramsey called for further discussion.
Hearing none, he called the
question. The motion carried 3-1 with Dr. Crittenden voting in opposition.
Dr. Ramsey asked each member to explain his vote. Dr. Laskey said that he did not see
the safety signal that would allow him to overlook the ambiguity on
efficacy. Dr. Sackner-Bernstein
said that there was not enough efficacy data to outweigh the lack of safety
data. Dr. DeMets said that he was
not convinced, due to doubts about the primary outcome and the size of the
trial. Dr. Crittenden said that
some efficacy had been shown in the reduced need for LVAD placement and cardiac
transplants. With expert surgeons, this
would be a safe procedure, especially for patients with few alternatives. Dr. Weinstein thanked the Panel
members, the Sponsor, and the FDA for their work.
Dr. Ramsey polled the members on what would make the PMA
approvable. Dr. Laskey said there
would have to be another trial, though the current data could be included
through a Bayesian design. Dr.
Sackner-Bernstein said that the studies should be designed to put the upper
confidence interval of safety in an acceptable range and give convincing
evidence of a clinically meaningful benefit.
Dr. DeMets said the existing data had been fully analyzed, that a
new trial was necessary, that the new trial should address bias issues that
arose in this trial, and that the new trial should be larger. Dr. Crittenden gave no further
comment. The Dr. Ramsey thanked the
participants and adjourned the meeting at 4:47 p.m.
I
certify that I attended this meeting of the Medical Devices Dispute Resolution
Panel on December 15, 2006 and that these minutes accurately reflect what
transpired.
______________________________
Nancy
Collazo-Braier, Ph.D.
Executive
Secretary
I approve the minutes of this
meeting
as recorded in this summary.
_______________________________
Scott D. Ramsey, M.D., Ph.D.
Chairman
Summary
Prepared by
Eric M. Hendrixson
Neal
R. Gross & Company
(202) 234-4433
Tuesday, February 13, 2007