<PAGE> 1
Exhibit 99.1
Gliatech Inc.
Responses to F.D.A. Form 483 dated August 23, 2000 9/14/00
OVERVIEW
On August 23, 2000, Gliatech received inspectional observations (Form
FDA 483) relating to the conduct of a U.S. ADCON(R)-L Adhesion Control in a
Barrier Gel (ADCON-L) Clinical Study. We have addressed both the specific
observations and described the additional measures to improve ongoing operations
and systems of the company.
We believe the animal and human experience with ADCON-L to date has
demonstrated the effectiveness of the product in inhibiting peridural scar.
Nevertheless, we take the observations made during the FDA inspection seriously,
and we intend to withdraw PMA Supplement 7 of PMA P960057, submitted 3/31/99,
with the MRI data in question, so that we can resubmit a PMA Supplement with the
final report of the U.S. Clinical Study and draft labeling incorporating the
revised results. Moreover, to remove any questions regarding the MRI scar
analysis, we have attached a proposed MRI analysis protocol in which all the
MRIs from this study would undergo a separate review by a blinded, trained
evaluator under controlled conditions. This review will be conducted by an
independent Contract Research Organization (CRO). We intend to obtain
concurrence from the Agency on all aspects of the study protocol prior to its
initiation. This approach is intended to respond to the concerns expressed in
observations 1-4.
Regarding observations 5, 6 and 7, the entire patient population will
be reevaluated systematically using the inclusion/exclusion criteria to
establish the final evaluable MRI patient population. A new PMA Supplement with
the final study report will be submitted as mentioned above.
In addition to responding to each item in the Form FDA 483, Gliatech
has addressed our internal systems (SOPs and resources) in order to enhance Good
Clinical Practice (GCP) compliance. In general, over the last 2 years but after
the completion of the study in question, we increased our resources in the
clinical management area. We hired a new Director of Clinical Research (November
1999) and an additional Clinical Project Manager (March 2000) responsible for
clinical studies. Our new Director of Clinical Research has been reviewing our
standard operating procedures (SOPs) and improving them since his arrival in
1999. Additionally, resources from Gliatech's Quality Assurance department were
assigned (June 2000) to assist in the on-going monitoring of clinical study
data.
Over the last several months, additional reviews of our procedures were
performed, and new SOPs were added that address the concerns raised in the Form
FDA 483. These new SOPs are also described within the text of the specific
responses, and copies are attached for review. We are currently in the process
of retaining consultants to perform independent assessments of ongoing US
clinical studies. These consultants will also perform a comprehensive assessment
of the systems and resources for managing Gliatech clinical studies, as well as
the re-training of all involved employees in Good Clinical Practices (GCP). We
believe that these actions will enhance our internal procedures and further
improve the manner in which we conduct clinical trials.
<PAGE> 2
1. THE SCAR SCORE DATA IN THE FINAL REPORT FOR THE U.S. ADCON-L STUDY ARE
NOT THE ORIGINAL SCORES FOR 115 OF 324 SUBJECTS. INSTEAD, SCAR SCORES
FROM AN INTRAOBSERVER RELIABILITY STUDY WERE SUBSTITUTED FOR THE
ORIGINAL DATA.
As a condition for an approval recommendation by the members of the
Orthopaedics and Rehabilitation Devices Advisory Panel to the FDA at the
December 1997 meeting, an intraobserver reliability study was to be submitted.
The company, in an attempt to rapidly satisfy the panel recommendations,
initiated and completed an intraobserver reliability study in January, 1998. The
purpose of the study was to assess the ability of the blinded neuroradiologist
to reproducibly score the quantity of scar seen on MRI. 117 patient MRIs that
were available since the submission of the interim report to the FDA were
re-evaluated for scar under blinded conditions, of which 114 patient MRIs were
used for the intraobserver reliability study (the differences in the patient MRI
numbers given above are explained in Attachment A).
The MRIs were rescored by the neuroradiologist in the intraobserver
reliability study during a two-day controlled reading session. The
neuroradiologist was blinded to treatment during the initial read for the
clinical study, remained blinded to treatment during the intraobserver
reliability study, and remains blinded to treatment as of today. The identical
scoring methodology was utilized for the intraobserver reliability study that
was used for the original readings. The conditions under which scar assessments
were made did vary from the original reads, in that the neuroradiologist was
asked to score 117 MRIs under controlled conditions and over a defined period of
time (2-day sessions). The MRI readings had originally been read by the
neuroradiologist over a lengthy period of time (approximately 2 years)
individually or in small batches. For the intraobserver reliability study,
Gliatech Medical Department personnel coordinated the scoring session and
provided clerical service in order to complete the MRI readings in two sessions.
The neuroradiologist read the MRIs and verbally assigned scar scores, which were
transcribed by Gliatech's former Vice President, Medical Affairs onto the scar
scoring sheets (form FA-6, Attachment B). The neuroradiologist signed all
individual score sheets at the conclusion of the intraobserver reliability
study.
The results of this intraobserver reliability study were presented to
the FDA during a 1/26/98 meeting at the Agency and later submitted as a separate
report in a PMA amendment dated 2/12/98 (Attachment C). The study results
indicated a 0.94 weighted kappa for all five MRI slices, a 0.93 weighted kappa
score for the middle three MRI slices, and weighted kappas ranging from between
0.9 and 0.99 for individual quadrants of each MRI slice.
The 114 reread MRI scores from the intraobserver reliability study,
included in the final data analysis for the U.S. clinical study, represented the
most recent scoring and were performed under controlled conditions. To assure
that all the data are collected under the same conditions we propose to have all
the MRIs reread. The protocol for this reread is being submitted to FDA for
review (Attachment D).
In addition, to ensure the inclusion of all support studies in the
final report such as the Intraobserver Reliability Study, an SOP regarding
clinical report preparation has been created. (Attachment E).
<PAGE> 3
a. THE WORST SCAR SCORE FOR 32 OF THESE 115 SUBJECTS CHANGED FROM
THE ORIGINAL SCORES. EIGHTEEN OF THE TWENTY WHOSE NEW SCORE
RESULTED IN A LESSER SCORE WERE ADCON-L PATIENTS. ELEVEN OF
THE TWELVE SUBJECTS WHOSE NEW SCORE RESULTED IN A MORE SEVERE
SCORE WERE CONTROL PATIENTS.
RESPONSE TO 1a.
This observation should be considered in the context of the overall
study. The intraobserver reliability study consisted of 114 patients and 20
separate reads for each of them (5 contiguous MRI slices x 4 quadrants), for a
total of 2280 scar quadrants that were assigned scar scores. As the observation
notes, changes were seen in 32 patients (19 ADCON-L and 13 Control) with regards
to maximum scar score (Attachment F). Of the 2280 scar scores, a total of 208
scar scores were different from the neuroradiologist's reread than they had been
scored on the original read. This is less than 10% (208/2280 x 100) of all the
scores read.
In the final analysis, 27 of the 32 patients were used out of a total
study population of 279. The other 5 patients were not included in the final
evaluable analysis, because they did not meet the selection criteria in the
clinical protocol. All of the 5 excluded patients had maximum scar changes that,
if included, would have favored treatment differences. Of the 27 patients
included in the final analysis (16 ADCON-L and 11 Control), fifteen of sixteen
ADCON-L patients had maximum scar scores showing scar improvement, and nine of
eleven control patients had maximum scar scores showing scar worsening.
Of the 32 subjects whose scar score changed, the magnitude of the
change was 1 unit in 26 cases (greater than 80%). In only 4 cases was the
magnitude of the change 2 units, and in only 2 cases was the magnitude of
change 3 units. In 1 of the 2 cases in which the magnitude of change was 3
units, the patient [*] was considered non-evaluable for other reasons, and not
included in the final analysis. In summary, the individual changes in max scar
score appeared small in magnitude in the majority of cases and would not be
unexpected for a subjective measurement.
Although the kappa scores indicate a high degree of reliability in
reading the MRI, these observed changes are not surprising. One probable reason
for differences in the assessment of scar in the intraobserver reliability study
was that the multiple MRI reads (~60 MRIs) were done during each of the two
defined sessions. This would lend itself to a greater consistency in scoring
when compared to reading over a longer period of time (approximately 2 yrs) and
with limited samples at each read. The conditions of the reading session would
be expected to influence the variability of the neuroradiologist. An improvement
in reader consistency would tend to favor a more accurate reading and would
establish treatment differences that may be missed under less stringent
conditions. The improved consistency would also be particularly important in
measurements in which ranges of scar are being determined (i.e., 0-25%, 26-50%,
51-75%, 76-100% scar within the quadrant of each MRI slice).
-------------
* indicates portions that have been intentionally omitted
<PAGE> 4
Nevertheless, to remove any questions regarding the MRI scar analysis,
we have attached a proposed MRI analysis protocol (Attachment D) in which all
the MRIs from this study would undergo a separate review by a blinded, trained
evaluator under controlled conditions. The review will be conducted by an
independent CRO.
b. ONLY SUBJECTS WHOSE SCAR SCORES HAD NOT PREVIOUSLY BEEN
SUBMITTED TO FDA IN AN INTERIM REPORT FOR THE U.S. ADCON-L
STUDY WERE CHOSEN FOR THIS RELIABILITY STUDY. THE MRI'S FOR
THE OTHER STUDY SUBJECTS WERE NOT RE-READ.
RESPONSE TO 1b.
The patients that were included in the interim report were in a closed
and locked databases, and all analyses were complete. Rather than reopening a
closed, locked data set, which generally is not practiced, Gliatech used the
available patients not included in the interim report. Gliatech believed, and
this belief was borne out, that the available data set would be sufficient for
purposes of conducting the intraobserver reliability study.
c. THE SUBSTITUTION OF THE RE-READ SCORES FOR THE ORIGINAL SCORES
WAS NOT DISCUSSED IN THE FINAL U.S. ADCON-L STUDY REPORT
SUBMITTED TO FDA.
RESPONSE TO 1c.
The failure to note that the reread MRIs were used in the final
clinical study report was an oversight, which was not corrected by Gliatech or
the CRO during the process of preparing the report. Gliatech has developed SOPs
that require GCP audits of regulated clinical studies (Attachment G), and
require reference in final clinical reports of support studies (Attachment E).
2. THE INTRAOBSERVER RELIABILITY STUDY WAS PERFORMED TO ASSESS THE
RELIABILITY OF THE SCAR SCORES GIVEN BY [*]. HE WAS TO RE-READ
SOME OF THE MRI'S THAT HE HAD ORIGINALLY READ AND SCORED DURING THE
U.S. ADCON-L STUDY. HIS RE-READ SCORES WERE TO BE COMPARED TO HIS
ORIGINAL SCORES.
a. THERE WAS NO PROTOCOL DETAILING SUCH THINGS AS HOW THE
SUBJECTS WOULD BE SELECTED OR WHAT RESULTS WOULD BE
ACCEPTABLE.
RESPONSE TO 2a.
In order to address the December 1997 panel's requests, the
intraobserver reliability study started almost immediately after the panel
meeting. In addition, the study methodology was thought of as being an extension
of the clinical study already conducted prior to the interim report. The timing
of this intraobserver reliability study immediately after the FDA panel meeting,
and the use of a previously established standard method for MRI assessment,
-------------
* indicates portions that have been intentionally omitted
<PAGE> 5
contributed to the absence of a protocol. A separate intraobserver reliability
report was generated with a methods section describing the study and use of the
clinical protocol's method for the MRI scoring. This report was presented to FDA
at a meeting between FDA and Gliatech on 1/26/98 and was submitted to the FDA on
2/12/98.
An SOP describing the requirements and approval process for Gliatech
study protocols was revised in June 2000 to include clinical activities
(Attachment H). As previously mentioned, to remove any questions regarding the
MRI scar analysis, we have attached a proposed MRI analysis protocol in which
all the MRIs from this study would undergo a separate review by a blinded,
trained evaluator under controlled conditions. The review will be conducted by
an independent CRO.
b. THE SCAR SCORE RESULTS OBTAINED DURING THE RE-READING OF THE
MRI'S WERE RECORDED IN PENCIL AND HAD ERASURES AND
WRITE-OVERS. THREE OF THE ERASURES CAUSED ADCON-L PATIENTS TO
HAVE THE WORST SCAR SCORE IMPROVE FROM a 4 TO a 3. THREE
WRITE-OVERS (FROM a 1 TO a 4) CAUSED THE WORST SCAR SCORE FOR
CONTROL PATIENTS TO WORSEN TO a 4.
RESPONSE TO 2b.
The intraobserver reliability study was conducted in a manner in which
the neuroradiologist read the MRIs and called out the scores. Gliatech's former
VP, Medical Affairs captured the neuroradiologist's response on the scar score
sheets in pencil. A likely explanation is that the neuroradiologist reviewed the
MRIs and occasionally corrected his response, necessitating Gliatech's former
VP, Medical Affairs to correct his written response by erasure and/or writeover.
Other changes appear to be simply corrections of transcription error. An SOP was
in place at the time that described the method in which data was to be recorded,
i.e. a preferred use of ink. The SOP regarding data capture has been modified to
require ink (preferably black) (Attachment I). An SOP has also been established
to require an ongoing GCP audit program regarding procedural aspects of clinical
studies (Attachment G). Further, an SOP has been established to ensure
appropriate training in GCPs (Attachment J).
The new scar scores were transferred from the original pencil-written
scar sheets to a Microsoft(TM) Excel spread sheet form by Gliatech personnel
within 2 days of the rereads. The Excel spread sheet was emailed to [*]
(statistical CRO for the intraobserver reliability study, [*]) and utilized for
statistical comparison to the original scores. Approximately 8 months following
completion of the intraobserver reliability study, the MRI scores were submitted
to [*] (CRO for the U.S. ADCON-L clinical study) for incorporation into the
final database and clinical report. The data from the intraobserver reliability
study was not modified between the time the data was sent to [*] and [*]
(Attachment F).
The nature of the erasures appears to be corrections or transcriptional
errors made during the course of the re-read sessions. For example, columns of
numbers were entered into the wrong column, erased and rewritten in the correct
column. Eighty-eight erasures were identified from the scar assessment forms, of
which a total of 6 resulted in a change in
-------------
* indicates portions that have been intentionally omitted
<PAGE> 6
maximum scar score (3 FDA observations plus 3 more from Gliatech's internal
review = 6 total). The three patients identified by the FDA were patient #s
[*] and [*] and were ADCON-L patients. These erasures resulted in the
ADCON-L patients having their maximum scar score lowered (improvement) by 1 unit
from 4s to 3s. Additional patients not mentioned in Form FDA 483 were patient
#[*]. These 3 patients were Control patients, of which 2 patients had lowered
(improved) scar scores. All 6 patients' maximum scar scores changed only 1 unit
in magnitude. Thus, the majority of erasures (82/88:93%) had no bearing on the
primary outcome measure. Of the 6 maximum scar scores that were changed due to
erasures, 4 were in favor of treatment differences and 2 were against treatment
differences.
Three apparent overwrites were cited in the Form FDA 483. We are unable
to determine whether these are authentic overwrites. Gliatech has not been able
to establish from its former employee the authenticity of these apparent
overwrites.
Upon further review, we have found five potential write-overs using the
similar subjective criteria as the FDA investigators (Attachment K). The five
patients were patients #[*]. In one case the apparent overwrite had no bearing
on the maximum scar score. In the 4 other cases, the maximum scar score
increased in both ADCON and control patients from 3 to 4 (in 3 patients) and 2
to 4 (in one ADCON patient) on the score sheet (Form FA-6) due to the apparent
overwrite. These scores were maintained in the final report in Appendix E.14
(3/19/99). In all cases, the scar scores indicated increased quantity of scar.
3. THE "ORIGINAL" SCAR SCORE SHEET FOR PATIENT 39 AT CENTER 6 WAS A
PHOTOCOPY. QUADRANTS 3b, 3d, 4b, AND 4d SHOWED HEAVY SHADING BUT WERE
SCORED 1, 0, 1, AND 1, RESPECTIVELY. THIS SHEET, LIKE ALL THE OTHER
SCORE SHEETS FOR THE ORIGINAL READS, WAS NOT SIGNED BY [*]. WHEN THE
MRI WAS RE-SCORED ON 8/4/00 THE SHADINGS WERE VERY SIMILAR TO THE FIRST
SHEET BUT THE SCORES WERE 3, 3, 4, AND 3, RESPECTIVELY. THIS WAS AN
ADCON-L PATIENT.
RESPONSE TO 3.
A search to locate the missing form did not turn up the missing
original. Regarding the form FA-6, (scar score sheet) page 2 of the scar score
sheets contains a diagram, and shadings may or may not have been included by the
neuroradiologist indicating the areas involved with scar. The diagram and the
shadings were meant as a visual aid for the neuroradiologist to establish in
which MRI quadrants scar was found. The degree of shading was not intended to be
indicative of severity or extent of scar. The neuroradiologist signed on page 1
of Form FA-6 as required by the protocol.
The neuroradiologist under blinded conditions was provided patient 39's
MRI and new scar score sheets on 8/4/00 to reread. The neuroradiologist
confirmed the quadrants 3b, 3d, 4b and 4d for the presence of scar and indicated
the severity was 3, 3, 4, and 3.
-------------
* indicates portions that have been intentionally omitted
<PAGE> 7
Since the original MRI was rescored, this result should be used in
place of the photocopy. Despite the differences in severity, the presence of
scar was found in the identical quadrants as the photocopy.
4. MANY SCAR SCORE SHEETS (ORIGINALS AND RE-READS) HAD SCORES IN QUADRANTS
THAT WERE NOT SHADED AND HAD NO SCORES IN QUADRANTS THAT WERE SHADED.
RESPONSE TO 4.
The scar score sheets from only the original MRI reading and not the
MRI rereading contained shadings in the graphic representation of forms FA-6
(scar score sheets). In other words, the neuroradiologist did not utilize the
graphic representation during the intraobserver reliability study. A complete
review of the original Forms FA-6 in the neuroradiologist's binder found 10
forms in which discrepancies were observed (Attachment L) between the scores and
shaded diagram. In 5 patients, the score was placed in the wrong quadrant column
(a,b,c,d). In 1 patient, scores were placed in the wrong level row (1,2,3,4,5).
In 1 patient, a quadrant was scored, and no shading was seen in the graphic
representation. In 3 patients, shading in the graphic representation was seen
and no score was seen in the corresponding quadrant. The shaded diagram was only
a visual tool for the neuroradiologist's own use. The shaded diagram was not
intended to be indicative of the presence or extent of scar. The
neuroradiologist was not required to shade the diagram, and, while he routinely
did, the absence of shading is not conclusive nor significant.
5. SIX OF THE 280 SUBJECTS DETERMINED TO BE EVALUABLE FOR SCAR SCORING DID
NOT MEET ALL THE CRITERIA TO BE INCLUDED IN THE EVALUABLE GROUP. THE
MRI'S FOR 30 SUBJECTS SHOWED HERNIATIONS AT THE SURGERY SITE (15 MARKED
AS RECURRENT, 3 AS RESIDUAL, AND 12 NOT SPECIFIED) BUT THESE SUBJECTS
WERE NOT EXCLUDED FROM THE EVALUABLE GROUP.
RESPONSE TO 5.
All patients had an IE-P CRF form listing the exclusion-inclusion
criteria of the study. The CRF was completed at the site. The answers to the
questions on this CRF were used as criterion by the database logic checks to
systematically identify unevaluable patients. In each case regarding the 6
patients cited below, the patient's IE-P CRF indicated the patient met all
exclusion-inclusion criteria. Data on other CRFs were not referenced by the
computer logic check. Each patient's specific case upon and further review is
described below and detailed in Attachment M.
Patient [*] was included in the MRI evaluable group even though the
patient was a multi-level operative patient. Patient [*] was enrolled, and the
IE-P form that was filled out at the site listed the patient as a single-level
operation. Further review identified that the patient had a multi-level
operation. We conclude this patient should have been considered unevaluable.
-------------
* indicates portions that have been intentionally omitted
<PAGE> 8
Patient [*] was included in the MRI evaluable group but did not have a
preoperative visual analog pain scale (VAPS) pain score. Patient [*] was
enrolled, and the IE-P form that was filled out at the site listed the patient
as having a VAPS pain score [*]. Further review identified the patient had no
VAPS pain score. We conclude this patient should have been considered
unevaluable.
Patient [*] was enrolled, and the EI-P form that was filled out at the
site listed the patient as having a VAPS pain score [*]. Patient [*] did not
have a preoperative VAPS pain score sheet filled out but did have a preoperative
pain score. We conclude this patient should have been considered unevaluable.
Patient [*] was enrolled, and the IE-P form that was filled out at the
site listed the patient as 30 yrs of age. Patient [*] was actually 29 1/2 yrs
old. Patient was enrolled on 3/12/96, and a waiver was granted on 4/8/96. We
conclude this patient should have been considered unevaluable.
Patient [*] had a discrepancy between the neuroradiologist and the
Investigator operative report regarding the operative level. A data
clarification form (DCF) was issued from the CRO. Subsequently, the
neuroradiologist and Clinical Project Manager reviewed the case. The
neuroradiologist concurred with the operative notes, and the DCF was closed. We
conclude the patient should be evaluable.
Patient [*] was included in the MRI evaluable group. The
neuroradiologist had observed spondylolisthesis at the 6 month post-op MRI.
Preoperative spondylolisthesis was listed as an exclusion criteria on form IE-P
in the protocol. The patient did not have this condition listed preoperatively
and therefore met the selection criteria for inclusion in the study. The
neuroradiologist rated the MRI as "satisfactory" for scar evaluation. We
conclude the patient should be evaluable.
In addition, the Form FDA 483 states that the neuroradiologist, in
reviewing the MRIs, observed at the surgery site reherniations (15 marked as
recurrent, 3 as residual, and 12 not specified) but these subjects were not
excluded from the evaluable group. The intent of the protocol was to exclude
reherniation or reoperations only based on clinical symptomology. It is well
known that asymptomatic reherniations are seen by MRI (Jensen, M.C. et al., N
Engl J Med 1994; 331:69-73. [Attachment N]). This observation was confirmed by
the neuroradiologist and a consulting neurosurgeon (Attachment O). We consider
all 30 patients properly included in the evaluable population.
The entire patient population will be reevaluated systematically using
the inclusion/exclusion criteria to establish the final evaluable MRI patient
population. A final report that will be provided will contain that systematic
analysis of MRI data.
-------------
* indicates portions that have been intentionally omitted
<PAGE> 9
6. ONE CONTROL SUBJECT WAS EXCLUDED BY DR. MCKINLEY BECAUSE HERNIATION AT
L1-L2 MIGHT INTERFERE WITH THE MRI FINDINGS BUT THE SURGERY WAS AT
L5-S1 AND [*], WHO READ THE MRI'S, FELT THE MRI WAS SATISFACTORY
AND GAVE SCORES OF 1.
RESPONSE TO 6.
The control subject in question was excluded from the final report for
the ADCON-L U.S. study based on the symptomology noted on the MRI and the
medical judgement of the former VP, Medical Affairs at Gliatech. A herniation on
the MRI was noted by the neuroradiologist at L1-L2 with inflammation potentially
interfering with MRI findings at other levels. The original surgery was at L5-S1
and the neuroradiologist read the MRI and recorded scar scores of 1. The VP,
Medical Affairs conferred with the neuroradiologist, and determined that the
inflammatory response could affect the readability of the entire MRI. The
patient was excluded from the MRI analysis. On 9/6/00, the neuroradiologist
reviewed the MRI and reconfirmed both the reherniation and inflammation. In the
neuroradiologist's opinion on 9/6/00 this patient's symptomology could impact on
the MRI scar scoring. This was a judgement call. These responses are consistent
with the earlier documentation and the ultimate decision by the former VP,
Medical Affairs to exclude the patient.
7. THE MRI'S FOR 38 OF THE 280 EVALUABLE SUBJECTS WERE TAKEN MORE THAN ONE
MONTH LATE.
RESPONSE TO 7.
The inclusion/exclusion criterion of the clinical study did not include
timing of the MRI visit. The protocol states that MRIs should be taken at 6
months post-op. Furthermore, in the Schedule of Visits table ([*])" (Attachment
P). There was also language on the CRF for the sites to have patients return for
the 6 month MRI 24-28 weeks after surgery. However, the protocol did not
establish a last date by which the MRI was to be taken. The intent of the study
was to measure scar as assessed by MRI at a time when scar was mature and beyond
a point when initial operative findings might obscure accurate readings. This is
confirmed in a letter from the neuroradiologist (Attachment Q). This was also
set forth in a response to a [*] FDA letter (Attachment R). On [*] the response
stated "Several published reports demonstrate convincingly that the clinical
outcome is stable from 3 to 6 to 12 months following lumbar discectomy; there is
also robust evidence that scar tissue is substantially stable 3 months after
surgery. There is no literature to support the idea of an evolving fibrotic and
clinical picture after 6 months." The response also quotes from published
literature to support this position.
Clinical literature and practice show that scar is mature by 6 months.
Thus, the inclusion of MRIs is consistent with clinical practice and the intent
of the protocol.
-------------
* indicates portions that have been intentionally omitted
<PAGE> 10
CONCLUSIONS:
In conclusion, Gliatech takes the observations made in the Form FDA 483
seriously. While we continue to believe that the data show ADCON-L to be safe
and effective, we acknowledge the Form FDA 483 observations. Thus, we intend to
withdraw the PMA Supplement 7, PMA P960057, submitted 3/31/99, with the final
MRI data, and to resubmit at a later date. Moreover, to remove any questions
regarding the MRI scar analysis, we have attached a proposed MRI analysis
protocol in which all the MRIs from this study would undergo a separate review
by a blinded, trained evaluator under controlled conditions. This review will be
run by an independent Contract Research Organization (CRO).
Upon agreement with the FDA, the entire patient population will be
reevaluated systematically using the inclusion/exclusion criteria to establish
the final evaluable MRI patient population. A final report that will be provided
will contain that systematic analysis of MRI data.
Gliatech has addressed the internal systems (SOPs and resources) in
order to enhance Good Clinical Practice (GCP) compliance. We have performed
additional reviews based on the current observations and added new SOPs that
address many of the specific concerns that were raised in the Form FDA 483. We
are currently in the process of comprehensively assessing the systems and
resources for managing Gliatech clinical studies, as well as the re-training of
all employees involved in clinical studies in GCPs. We believe that the above
actions will enhance the existing procedures and further improve the manner in
which we conduct clinical trials.
