ATTACHMENT
C
TABLES
PRESENTED BY LIFECORE AT THE
JANUARY,
2000
GENERAL
AND PLASTIC SURGERY
ADVISORY
PANEL MEETING
I have attached the
sponsor’s study protocol (Attachment A, dated 1/26/95) for your review. Basically, it states the following:
1) the pilot study (see Attachment A, page 34 of 38, under
section H. Power) found a difference of adhesion score of 4.0 between treated and control groups (mean
adhesion scores of 1.7 and 5.7, respectively, with corresponding standard
deviations of 1.4 and 2.7, respectively);
2) the sponsor agreed to do an intent-to-treat analysis with
the lost to follow-up being treated as failures and getting the worst score (a
16 according to their scoring system);
3) the sponsor expected an unequal lost to follow-up for the 2
groups: 20% lost to follow-up for the treated group and 10% for the control;
4) they used an intent-to-treat (ITT) analysis extrapolated
from the pilot study with the above lost to follow-up rates (20% for Intergel
and 10% for control treated patients) which yielded a mean adhesion score of
4.6 (st. dev. of 5.9) for the Intergel patients and 6.7 (standard deviation of
4.1) for the control group;
5) which results in a total of 180 patients necessary to
detect, with 80% power at the a=0.05 significance level, a difference in adhesion score of
2.0 given an expected standard deviation of 5.0 for both groups; and
6) due to the skewness of the data (patients scores can range
from 0 to 16 with an average around observed average adhesion score of 2 to 3 and with an observed standard
deviations of 3.5 to 4.5), the sponsor said they would perform a nonparametric
analysis.
The sponsor requested
they be able to perform the study in both the US and Europe. They would perform an interim analysis to
determine if 120 US and 80 European subjects were combinable to obtain a total
of 200 subjects. If they were not
combinable by the procedures laid out in the attached protocol (page 31 to 33
of 38, Attachment A), they would continue the US study and use a total of 200
US subjects (100 per arm) for their analysis and not use the European data.
We looked at the
interim data from the US and European studies and determined that the subjects
were not combinable. The baseline
adhesion scores and incidence of adhesion scores were very different (see
Tables 1 and 2 below). The difference in change from baseline for the 2 groups
was also large. The US patients had a
doubling to tripling of their baseline score at 2nd look while the
European patients score changed much less from baseline to 2nd
look. Also the type of patients
differed greatly. The U.S. had mostly
non-adhesiolysis patients while the Europeans were mostly adhesiolysis
patients. The sponsor disagreed with
our conclusion that the two groups were different but continued to enroll
patients until they had enrolled 200 US patients. They had also completed their complement of 81 European
patients. Though the sponsor sized the
trial for 180 to 200 patients, they based their analyses on all 281 U.S. and
European patients.
The data in Table 1
below presents the intent-to-treat (ITT) summary statistics for the mAFS score
and number of adhesions at baseline and 2nd look adjusted (for
baseline adhesions not lysed) for both the patients treated with Intergel (IG)
and the control, Lactated Ringer’ Solution. These summary statistics were provided by the sponsor and
extracted from the back of their December 1999 Panel Pack (Panel Pack II, Data
1, page 1) for the General and Plastic Surgery Advisory Panel.
|
|
United States |
Europe |
||||
|
|
Intergel N=102 |
Control N=98 |
Difference (IG-C) |
Intergel N=41 |
Control N=40 |
Difference (IG-C) |
MAFS
score
|
|
|
|
|
|
|
|
Baseline |
0.78 |
0.68 |
+0.10 |
1.57 |
1.95 |
-0.38 |
|
2nd Look
Adj |
2.63 |
2.76 |
-0.13 |
2.01 |
2.12 |
-0.11 |
|
Number of Adhesions |
|
|
|
|
|
|
|
Baseline |
2.49 |
2.27 |
+0.22 |
6.00 |
6.40 |
-0.40 |
|
2nd Look
Adj |
7.92 |
7.73 |
+0.19 |
6.63 |
7.33 |
-0.70 |
Note that there are
no statistically significant results between the 2 groups for both mAFS and
number of adhesions for the US patients or the European patients. In fact we see no apparent difference
between the treatment and control groups.
(Remember in the pilot study we saw a difference of 4.0 and, after
accounting for lost to follow-up as specified in the protocol, the study was
designed to detect a difference of 2.1)
Thus as designed, the trial does not show the device to be
effective. In fact, all the differences
we see in this table are very small, except for the difference in the baseline
scores between the European and US patients and the difference in change from
baseline to 2nd look adjusted between the US and European
patients. European baseline scores are
2 to 3 times larger than the corresponding US scores. The US patients also have a 3- to 4-fold increase over baseline
in both adhesion score and number of adhesions while the European patients have
a very small increase (of only 10% to 30%) over baseline. Thus the patients from the 2 continents are
very different at baseline and appear to respond very differently to treatment
with respect to adhesion and mAFS score indicating the 2 groups are not
combinable. Therefore, it was
determined that the 200 US patients comprised the appropriate patient group to
analyze for device. effectiveness.
The sponsor presented
the summary results based on evaluable patients as opposed to the
intent-to-treat patients. Their
analysis violated many of the premises upon which the study was designed (as
stated above in the protocol section):
1) The sponsor used only evaluable patients (which excludes
those patients lost to follow-up) instead of including lost to follow-up
patients as described in the study protocol.
2) The sponsor used all 265 evaluable patients from the US and
Europe (which is greater than the 200 US patients allowed or the 180 patients
for which the study was sized).
3) Since the patients lost to follow-up were removed from the
analysis, the observed standard deviations for the mAFS scores (of 1.5 and 2.6
for the Intergel and control groups, respectively) are much smaller than the
expected intent-to-treat standard deviation (of 5.0) for which the study was
designed.
All these conditions lead
to a vastly overpowered study that could result in finding statistically
significant differences between groups that may not be clinically
meaningful. So, if it were appropriate
to combine patients across continents and use evaluable patients only, the test
would provide 80% power to detect a difference of only 0.75 difference in the
mAFS. This is much smaller than the
agreed upon 2.1 clinical difference in mAFS score that the ITT study was
designed to detect.
In summary, using the
ITT design and analysis presented by the sponsor in their study protocol, there
is no statistical difference between the Lactated Ringer’s and control groups
with respect to mAFS score or adhesion score at second look.
Next, consider the
new post-hoc analysis that the sponsor has submitted in their post-panel
meeting PMA amendment (P990015/A10). This report presents shift tables for the
American Fertility Scoring System (AFS) which scores adnexal adhesions only and
that these scores were obtained retrospectively by a method which approximates
the AFS scoring system. Also, note that
this data was presented by the sponsor to the experts at the January 2000
Advisory Panel meeting (Attachment C) at which the Panel concluded that the
product did not provide a clinically meaningful benefit. In this Table (in Attachment C) and in their
analysis the sponsor presents data from all evaluable patients in both continents
but ignore those patients lost to follow-up.
Again, it should be emphasized that the study was not designed to
analyze AFS data and that the sponsor is performing post-hoc analyses on data
that was already presented to the Panel.
Table 3 (below) is
the ITT presentation of the shift table for sponsor’s (retrospectively
calculated) AFS data for US patients.
(I had to combine the first two categories none and minimal and mild
since intent-to-treat shift tables, stratified by continent, were not provided
by the sponsor.) The denominator is the
number of subjects in each subgroup (none/minimal/mild and moderate/severe)
having the baseline adhesion status specified by that subgroup. The numerator is the number of subjects
whose adhesion status at 2nd look is moderate/severe. Exploratory
analysis of the retrospective data in Table 3 found no statistical differences
between the treatment groups whether analyzed by subgroup or as a whole.
TABLE 3.
Number of Patients with Moderate or Severe AFS Scores (>10)
at Secondlook: Intent-to-Treat Patients
Adhesion
Status at Baseline Intergel Control p-value*
None, Minimal or Mild 12/97 11/91 0.99
Moderate or Severe 0/5 3/7 0.23
* p-value based on Fisher’s Exact test for comparison
of 2 proportions
In the sponsor’s
analysis in Amendment 11, they present an imputation scheme to account for the
patients lost to follow-up instead of their original intent-to-treat analyses
for which the study was designed. Their
post-hoc method relies on deleting data from patients who did not have 2nd
looks and did not have any complaints.
Note that this method is not appropriate, nor can it be statistically
justified because there is no way of knowing how the patients without
complaints or who did not return really fared. Furthermore, their method discards data from 8 of 12 of those
Intergel patients lost to follow-up group, while only deleting 1 of the 4
control patients who were lost to follow-up; this approach biases the results
in favor of the sponsor.
In their PMA
Amendment 11, the second part of their proposed Indication for Use states “INTERGEL Solution was also shown to
reduce adhesion reformation to sites in addition to adnexa; and adhesion
formation at surgical sites, including the anterior abdominal incision.” In the discussion and analysis of adhesion
reformation and surgical site adhesions presented on page 46-7 of Section III
of Amendment 11, they only present evaluable data for all US/European patients
combined (ignoring the appropriate intent-to-treat analyses stratified by
continent). Furthermore, after having failed to show effectiveness of the
primary study endpoint, the sponsor has chosen two of several secondary
endpoints (surgical site and reformed adhesions) for which they claim Intergel
is superior. Therefore, Amendment 11
proposes the situation where, not only is the appropriate intent-to-treat
analysis stratified by continent discarded, but a couple of several secondary
endpoints defined in the original study are evaluated without any adjustment of
the significance level of the statistical tests. A proper multiplicity adjustment is required to lower the
significance level of the tests to adjust for the multiple subgroups. In addition, clinically meaningful
differences for these endpoints were not defined a priori, and thus, the study was neither designed nor powered to
assess them in a statistically valid fashion.
The intent-to-treat results for these endpoints (reformed and surgical
site adhesions), as well as denovo adhesions, are presented in Table 4
below. These summary statistics were
provided by the sponsor and extracted from the back of their December 1999
Panel Pack (Panel Pack II, Data 1, page 1) for the General and Plastic Surgery
Advisory Panel. No statistical
differences were found for any of these endpoints, even without a multiplicity
adjustment.
|
|
United States |
||
|
Mean Incidence |
Intergel N=102 |
Control N=98 |
Difference (IG-C) |
# adhesions lysed*
|
2.09 |
1.89 |
+0.20 |
|
Reformed Adhesions |
3.18 |
3.51 |
-0.38 |
|
DeNovo Adhesions |
6.71 |
6.34 |
+0.37 |
|
Surgical Site
Adhesions |
2.29 |
2.62 |
-0.33 |
*Average incidence at baseline minus
average incidence after 1st surgery
Using the statistical
analysis plan from the study protocol (intent-to-treat analysis on the US
patients), the sponsor was unable to demonstrate that patients treated with
Intergel had statistically lower mAFS score or statistically fewer adhesions
than the control, Lactated Ringer’s solution.
In fact, in the United States, both products showed approximately
equivalent increases from baseline in both adhesion incidence and mAFS score
and these increases were substantial.
Furthermore, the sponsor was unable to demonstrate that it is valid to
combine the data across continents as both the baseline and change from
baseline for both mAFS score and incidence of adhesions were very different for
the Intergel and the control. The January
2000 General and Plastic Surgery Advisory Panel determined that there was not
reasonable assurance that the product was safe and effective. In their subsequent PMA Amendment, the
sponsor did not present any new data, but only selectively re-analyzed (in an
unplanned, post-hoc fashion) data already presented at the earlier panel
meeting.
ATTACHMENT A
SPONSOR’S STUDY PROTOCOL
Lifecore Biomedical, Inc.
Investigational New Device Clinical
Protocol No. PTL-OO13, Rev.1 Page 28
of 38
X.
STATISTICAL
METHODS
XI.
A. PATIENT POPULATIONS
1. The
intent-to-treat efficacy and safety populations will consist of all patients
who receive LUBRICOAT Gel or Lactated Ringer’s Solution.
2. A subset of
the intent-to-treat efficacy population will exclude patients who refuse the
second-look laparoscopy for reasons unrelated to the device.
3. The evaluable
efficacy population will consist of all patients who receive a second-look
laparoscopic evaluation.
Patients who are randomized but
do not receive treatment will be described but will not be otherwise analyzed.
If any patients are incorrectly randomized, alternative analyses will be
performed with those patients analyzed in the treatment group or the assigned
group.
B. EFFICACY VARIABL ES
The primary efficacy variable
will be a total adhesion score using the Adhesion Scoring Method of the
American Fertility Society (AFS) applied to 24 anatomical sites. Adhesions
occurring at each of the 24 potential adhesion sites will be scored as mild (a
filmy avascular adhesion) or severe (a dense organized cohesive vascular
adhesion). The extent of adhesions will be graded as Localized (<1/3 of the
site covered), Moderate (1/3-2/3 of the site
covered) or Extensive (>2/3 of the site covered). The extent of adhesions
will not be scored for the small bowel, omentum and left and right large bowel
since their size precludes adequate visualization. These sites will be assigned
a classification of Moderate in order to determine the total adhesion score.
Lifecore Biomedical, Inc.
Investigational New
Device Clinical Protocol No. PTL-OO13, Rev.1
Page 29 of 38
For each adhesion site, the
adhesion score will be derived from severity and extent scores as follows:
No Adhesion
Severity: Mild Extent: Localized 1
Severity:
Mild Extent: Moderate 2
Severity:
Mild Extent: Extensive 4
Severity:
Severe Extent: Localized 4
Severity:
Severe Extent: Moderate 8
Severity:
Severe Extent: Extensive 16
Scores from all potential
adhesion sites will be averaged to yield a total adhesion score. Adhesions will
be characterized as de nova if the
site had no pre-existing adhesions and as reformed if the
site had adhesions that were lysed during the original surgery. Sites with de novo adhesions will also be characterized
as surgical versus non-surgical.
These analyses will be conducted
for all sites as well as for pelvic and abdominal site groupings. Pelvic sites
include the caudal anterior peritoneum, anterior and posterior uterus,
cul-de-sac, right and left pelvic sidewall and all tube, ampulla and ovarian
sites. Abdominal sites include the right and left cephalad anterior peritoneum,
smaII bowel, omentum, right and left large bowel, rectosigmoid and the anterior
peritoneum incision.
The proportion of sites with adhesions
will be analyzed as a secondary efficacy variable. This will be a mean
proportion based on the number of sites with adhesions divided by the number of
possible adhesion sites. As above, adhesions will be characterized as de nova versus reformed, surgical versus
nonsurgical, and pelvic versus abdominal.
In addition, adhesion sites will
be categorized by the presence or absence of endometriosis, use of sutures and
the method of adhesiolysis (sharp dissection, blunt dissection, cautery,
laser). Each anatomical site will also be analyzed.
Additional secondary variables
will include the extent and severity of all categories 9f adhesions. Severity
will be scored on a three-point scale where 0 =
None, 1 = Mild and 3 = Severe. Extent
will be scored on a four-point scale where 0 =
None, 1 = Localized, 2 = Moderate and 3
= Extensive.
Lifecore Biomedical, Inc.
Investigational New
Device Clinical Protocol No. PTL-001 3, Rev.1 Page
30 of 38
C. SAFETY VARIARBLES
Safety variables will include the
proportions of patients reporting adverse events categorized using COSTART
terms. Laboratory values will be presented as mean change from baseline and as
transition tables showing the proportions of patients above, below and within
the normal range before and after treatment.
D. DEMOGRAPHIC, PRETREATMENT AND
SURGICAL VARIABLES
Age, race, height, weight, blood
pressure, previous and concomitant medications (categorized by AHFS codes),
presence of endometriosis, surgical procedures (categorized by CPT codes), estimated
blood loss, operative time, baseline adhesion scores and length of hospital
stay will be analyzed. Use of these variables to determine combinability with a
European study (Protocol PTL-0022) is described in Section G.
E. STATISTICAL ANALYSIS
Second-look adhesion scores will
be analyzed using factorial analysis of covariance where one factor is
treatment group (LUBRICOAT Gel versus Lactated Ringer’s Solution), the other
factor is center and baseline adhesion score is a covariate. This will allow analyses
of the effect of treatment, the effect of center and the interaction of
treatment with center. Homogeneity of slopes will be tested by examination of
interactions between baseline adhesion score and treatment group.
If the two groups differ on any important
demographic or surgical variables or if these pre-treatment variables appear to
strongly predict second-look adhesion scores (as determined using multiple
linear regression with treatment group forced into the model as a dummy
variable), these variables may be added to the model as covariates. Homogeneity
of slopes will be tested by examination of interactions between covariates and
treatment group. Pretreatment variables may be transformed in order to yield
homogeneous slopes.
The mean proportion of sites with
adhesions at second look will be analyzed in the same fashion as the mean
second-look adhesion scores.
Lifecore Biomedical, Inc.
Investigational New
Device Clinical Protocol No. PTL-OO13, Rev.1 Page
31 of 38
Other continuous variables will
be analyzed using factorial analysis of variance where one factor is treatment
group and the other factor is center. Analyses to determine combinability will
also use continent (US versus Europe) as a factor (see Section G).
Categorical variables will be
analyzed using the Cochran-MantelHaenszel test with individual sites as
strata. Determination of combinability of the US and European data will use
categorical models as described in section G. Proportions with small expected
event rates (e.g. adverse events) will be analyzed using Fisher’s exact test.
Laboratory value transition tables will be compared using 2x9 Fisher’s exact
tests.
Two-sided p
values will be reported and p values less than 0.05 will be considered to be
statistically significant.
F. INTENT-TO-TREAT ANALYSIS
As requested
by FDA, an intent-to-treat analysis will be performed in which patients treated
with LUBRICOAT Gel or Lactated Ringer’s Solution who do not have a second-look
laparoscopy will be considered to be treatment failures. This will be
accomplished by assigning them second-look adhesion scores of 16 (the worst
possible score). Because this will produce a highly skewed distribution, scores
will be transformed to ranks prior to statistical analysis.
G. EVALUATION OF COMBINABILITY
After at least 120 patients have
completed the study, the possibility of combining these patients with a
concurrent European study (Protocol PTL-0022) will be considered. The European
study is expected to have enrolled approximately 80 patients by this time.
Combinability will be based on three factors.
1. There should be no significant
interaction between location (US versus Europe) and treatment efficacy.
Lifecore Biomedical, Inc.
Investigational New
Device Clinical Protocol No. PTL-001 3, Rev.1 Page
32 of 38
2. The US and
European population should be similar on demographic and pre-treatment
variables and the level of medical care. Variables examined will include:
· Age
· Race
· Body weight
· Baseline adhesion score
· Previous and concomitant medications
(AHFS classification2)
· Presence of endometriosis
· Surgical procedures performed (CPT
classification)
· Estimated blood loss
· Operative time
· Baseline
clinical laboratory values
· Length of hospital stay (expected to be
longer in Europe)
· Time to second-look laparoscopy
· Number of patients lost to follow-up
(by reason for discontinuation)
Continuous
variables will be analyzed using factorial analysis of variance where one
factor is treatment group (LUBRICOAT Gel versus Lactated Ringer’s Solution) and
the other factor is location (US versus Europe). All statistically significant
effects involving location will be considered as possible sources of
non-homogeneity that might preclude combination of the US and European data.
Categorical variables will be analyzed using categorical models equivalent to
analysis of variance with factors for treatment group, location and the
interaction between treatment group and location.
3. The US and
European control groups should be similar on second-look adhesion scores. This
variable can serve as a proxy for subtle differences in medical treatment. The
95% confidence intervals of the difference between the US and European control
groups will be presented.
For each of these factors, data
will also be analyzed and presented by individual center within the US and
Europe.
2 McEvoy, G. K., Ed. American
Hospital Formulary Service Drug Information.
American Society of
Health-System Pharmacists, Inc., Bethesda, MD, 1995.
Lifecore Biomedical, Inc.
Investigational New
Device Clinical Protocol No. PTL-OO13, Rev.1 Page
33 of 38
These data will be presented to
FDA and the US and European data will not be combined unless Lifecore, Inc. and
FDA agree that there are no clinically significant differences that preclude
that combination.
If the US and European centers
are combinable, then the study will terminate as soon as that decision is made.
All patients currently enroIled in the study will be followed to second-look
laparoscopy and added to the database for the final statistical analysis.
If the US and European centers
are not combinable, then enrollment in the US protocol will continue until 200
evaluable patients have completed the study.
The decision
to stop or continue the study will not be affected by the p values of the difference between LUBRICOAT Gel and
Lactated Ringer’s Solution. Note that:
1. If the data
are not combinable, the U.S. study will not be stopped regardless of the
statistical significance of the difference between the treated and control
groups (either in the combined US/European study or the US study alone).
2. If the data
are combinable, and the difference between the treated and control groups in
the combined US/European study is not statistically significant, the study will
not be continued, but will be stopped and considered to have failed.
Therefore, p
values required to demonstrate statistical significance will not be adjusted.
- - CONFIDENTIAL - -
Lifecore Biomedical, Inc.
Investigational New
Device Clinical Protocol No. PTL-OO13, Rev. 1 Page
34 of 38
H. POWER
Power calculations were performed
using the method described by Lachin3 using an alpha level of 0.05
arid a beta level of 0.20 (80% power). Preliminary analysis of a Phase I study indicated
a mean adhesion score of 1.7 (Standard deviation: 1 .4) for the treated group
and 5.7 (Standard deviation: 2.7) for the control group. Assuming that 20% of
the treatment group and 10% of the control group are lost to follow-up, scoring
these patients as treatment failures would yield a mean adhesion score of 4.6
(Standard deviation: 5.9) for the treated group and 6.7 (Standard deviation:
4.1) for the control group. Assuming a standard deviation of 5.0, 180 patients
would be required. Thus the 200 evaluable patients (approximately 250 total
patients) appears to provide sufficient power to reject the null hypothesis if
the observed trends are maintained.
Xl. PATIENT DROPOUT RATIONALE
Study
enrollment has been planned to allow for a worse case 30% screen failure rate
and 20% loss to follow-up rate. This correlates with our request for 350
patients to be asked to participate in the study, with 250 expected to receive
treatment, and 200 to complete second-look laparoscopy. All patients assigned
study numbers and receiving treatment will be carefully followed and all screen
failure and loss to follow-up
patients documented. All efforts will be made to keep these to a minimum.
Any patient
who fails to return for the Day 7 -
28 laboratory determination and/or
the second-look laparoscopy will be contacted and interviewed if possible
as to her reason for not returning and her medical status ascertained relative
to the effects of the study device. All attempts to contact the patient will be
documented on case report form FINAL
STATUS.
A patient may
be discontinued from the study at any time in the event of a serious or
intolerable adverse event1 the need for an excluded medication, an
intercurrent illness, a protocol violation or at the patient’s request.
3 Lachin
JM. Introduction to sample size
determination and power analysis for clinical trials. Controlled Clinical Trials 2:93-113. 1981.
ATTACHMENT B
PILOT STUDY EFFICACY DATA
(from IDE G950025/S9, Final Clinical Study Report, Attachment
10)
NUMBER AND PROPORTIONS OF ADHESIONS
|
|
LUBRICOAT
Gel (N=11) |
CONTROL (N=9) |
P-value |
||
|
|
Mean |
SD |
Mean |
SD |
|
|
BASELINE |
|
|
|
|
|
|
Number of Sites with Adhesions |
3.55 |
4.52 |
4.33 |
3.93 |
0.687 |
|
Number
of Sites where Adhesions were Lysed |
2.82 |
4.04 |
3.78 |
3.49 |
0.582 |
|
Number of
Primary Surgical Sites |
5.36 |
3.50 |
5.56 |
2.50 |
0.892 |
|
INCIDENCE
OF ADHESIONS AT “SECOND - LOOK” |
|
|
|
|
|
|
Total
Number of Sites with Adhesions |
6.09 |
4.59 |
11.00 |
3.24 1.34 |
0.015 0.706 0.023 |
|
Total
Number of Sites Possible |
17.18 |
1.66 |
17.44 |
|
|
|
Proportion |
0.364 |
0.280 |
0.629 |
0.168 |
|
*Students t-test
ATTACHMENT
C
TABLES
PRESENTED BY LIFECORE AT THE
JANUARY,
2000
GENERAL
AND PLASTIC SURGERY
ADVISORY
PANEL MEETING