Lessons from Vioxx

1
Lessons from Vioxx

• DAVID EGILMAN MD, MPH
• - Clinical Associate Professor
• - Brown University
• - Consultant to Plaintiffs in Vioxx Litigation
• - degilman_at_egilman.com
• A copy of the presentation with hyperlinks to
  most of the documents cited can be found at
• http//www.vioxxdocuments.com/browse.php?displayl
  istdiregilman_arcoxia/

2
Lessons from Vioxx

• COX-2s increase Mortality
• COX-2s may cause Alzheimer's
• Safety Data Unreliable
• Anti-Platelet Trial Collaborative (APTC) Strategy
• Standard Operating Procedure (SOP) was not
  standard
• Adjudications often ad hoc
• Delayed Key Data Submission

3
The Big Picture(Things the Public Thinks the FDA
Considers but it Doesnt)

• What is the best way to treat RA OA?
• RA There is no evidence that NSAIDs do anything
  but treat pain. NSAIDs do not treat the
  underlying disease process.
• OA There is evidence NSAIDs makes OA worse (not
  including Cox-2 effects on fracture). 1
• Worried about GI bleeds and CV/T ?
• (Think whole patients here.)
• NSAIDs combination with PPI or H2 blocker is
  safer and cheaper than Cox-2.

1. Huskisson EC, Berry H, Gishen P, Jubb RW,
Whitehead J. Effects of anti-inflammatory drugs
on the progression of osteoarthritis of the knee.
J Rheumatol 1995221941-6.
4
Arcoxia Less Safe than Comparators
From Mercks Arcoxia Briefing Document

• More Hypertension Pg 155-63
• More Renal complications Pg 162
• More CHF Pg 142
• More Strokes Pg 163
• More MIs Pg 164
• More Arrhythmias Pg 93
• (Despite the fact that Arrhythmias were
  retroactively deleted from adjudication event
  list.)

FDA Arthritis Drug Advisory Committee Meeting,
ARCOXIA (Etoricoxib 30 and 60 mg For Symptomatic
Treatment of Osteoarthritis.Briefing Document
(Background Package) April 12,2007.
5
Arcoxia Less Safe than Comparators
Arcoxia Raises Blood Pressure
Etoricoxib FDA ACM Background Package, 155.
6
Arcoxia Less Safe than Comparators
Arcoxia Raises Blood Pressure
Etoricoxib FDA ACM Background Package, 155.
7
Arcoxia Less Safe than Comparators
Arcoxia Raises Blood Pressure
Etoricoxib FDA ACM Background Package, 158.
8
Arcoxia Less Safe than Comparators
Arcoxia Causes Congestive Heart Failure
Etoricoxib FDA ACM Background Package, 142.
9
Arcoxia Less Safe than Comparators
Arcoxia Causes Atrial Fibrillation
Etoricoxib FDA ACM Background Package, 93.
10
Arcoxia Less Safe than Comparators
Arcoxia Causes Renal Impairment
Among OA patients on etoricoxib 90 mg in the
MEDAL Study had a Statistically significant
higher incidence of renal abnormalities versus
diclofenac 150 m.
Etoricoxib FDA ACM Background Package, 161.
11
Arcoxia Less Safe than Comparators
Arcoxia Causes Renal Impairment
A statistically significant excess of renal
problems was noted for etoricoxib 90 mg compared
to diclofenac among RA patients in the EDGE II
Study for one or more values with absolute
increase 0.5 and gtULN
Etoricoxib FDA ACM Background Package, 161.
12
Based on observational data, we know Arcoxia can
cause stroke
Arcoxia Less Safe than Comparators
Current use of etoricoxib was reported to be
associated with an OR of 2.38 (95 CI 1.10 to
5.13). Other ORs (95 CI) for current use of
NSAIDs or COX-2 selective inhibitors were
rofecoxib 1.71 (1.33 to 2.18), celecoxib 1.07
(0.79 to 1.44), diclofenac 1.32 (1.101.57),
ibuprofen 1.12 (0.911.37), and naproxen 1.16
(0.801.70). In general, the ORs for the study
drugs were higher with higher doses and longer
duration of use.
Etoricoxib FDA ACM Background Package, 163.
13
Some NSAIDS take away the pain but make the
disease worse Arcoxia never evaluated for this.
Arcoxia Less Safe than Comparators
No statistically significant difference (p
0.308) was found between tiaprofenic acid and
placebo at the 7th interim analysis, the
conclusion of the study. CONCLUSION Indomethacin
increased the rate of radiological deterioration
of joint space in patients with OA of the knee
tiaprofenic acid did not.
Effects of anti inflammatory drugs on the
progression of osteoarthritis of the knee. LINK
Study Group. Longitudinal Investigation of
Nonsteroidal Antiinflammatory Drugs in Knee
Osteoarthritis. Huskisson EC, Berry H, Gishen P,
Jubb RW, Whitehead J.
14
Arcoxia studies did not evaluate efficacy
relative to the disease process (joint space
deterioration)
Arcoxia Less Safe than Comparators
Four adequate and well-controlled studies
demonstrated the efficacy of 30 mg of etoricoxib
in treating osteoarthritis for 12 weeks. However,
these studies only evaluated pain, physical
function, and stiffness
Etoricoxib FDA ACM Background Package, 148.
15
Arcoxia studies did not evaluate efficacy
relative to the disease process (joint space
deterioration)
Arcoxia Less Safe than Comparators
Indocin also works on the joint space
deterioration, but after a year Indocin treated
patients had more joint space deterioration than
placebo patients. Huskisson EC, Berry H,
Gishen P, Jubb RW, Whitehead J. Effects of
anti-inflammatory drugs on the progression of
osteoarthritis of the knee. J Rheumatol
1995221941-6
16
Arcoxia Less Safe than Comparators

• Arcoxia in no better for patients than Tylenol,
  naproxen or Celebrex.
• The burden is on Merck the FDA to assure that
  it is as safe prior to approval.

17
CHF Adjudicated Post Hoc for EDGE after
Freezing (Un-blinding)

• FDA says Development of Congestive Heart
  Failure (CHF) CHF was handled similarly to the
  CV and GI events with an adjudication committee.
• NOT TRUE
• Merck deleted CHF (atrial Fib) from list of
  events to be adjudicated Dec 1999 (post dated Oct
  3).
• Merck added CHF to ADJ event list around 12/2006
• This data is unreliable since the CHF data was
  not collected systematically throughout the
  trials.

18
Placebo/Tylenol/methotrexate Long-term Trials
are Ethical

• Bombardier - 1 year placebo trial in 1999
• On the other hand, VIGOR WAS unethical
• 50 of the patients in VIGOR were on steroids.
  Scolnick wrote, this is like testing Mevacor
  for liver safety in patients with hepatitis.
• Long term placebo trials are necessary to
  establish safety there is no excuse for not
  doing them

Arthritis Rheum. 1999 Sep42(9)1870-8. Comment
in Arthritis Rheum. 2000 Nov43(11)2615-6.
Function and health-related quality of life
results from a randomized controlled trial of
leflunomide versus methotrexate or placebo in
patients with active rheumatoid arthritis.
Leflunomide Rheumatoid Arthritis Investigators
Group. Strand V, Tugwell P, Bombardier C, Maetzel
A, Crawford B, Dorrier C, Thompson A, Wells G.
MRK-NJ0130089
19
The Big Picture Arcoxia MIs occur early

• Little understood A caused case is a case that
  occurs before it would normally occur.
• For example, if you expect 3 cancer cases in
  patients who are over 80 and the exposed group
  has just 3 cases, but they occur in patients who
  are under 30, these are all caused cases.

Greenland S. Relation of probability of causation
to relative risk and doubling dose a
methodologic error that has become a social
problem. Am J Public Health. 1999
Aug89(8)1166-9. Egilman, D, Howe S., Corporate
Obstruction of Public Health via Manipulation of
Epidemiology Int J Occup Environ
Health200713118124.
20
The Big Picture Arcoxia MIs occur early

• Reviewers comments
• Of note, there are 11 cardiac events in the
  etoricoxib group vs. 2 in the non-naproxen
  combined group which is made up of diclofenac
  events only (there were no events in the
  ibuprofen treatment arm). Three of the 4 events
  in the non-naproxen group occurred at greater
  than 3 years of exposure (as noted in a review of
  the KM plots). All but one of the events in
  etoricoxib occurred within the first 2 years.
  There are also 5 MIs in the etoricoxib group vs.
  1 in the diclofenac group. Although there is no
  difference in the absolute numbers of CNS events
  the rates are higher in the diclofenac group
  because of the fewer patient-years of exposure.
  Numbers are small and firm conclusions are
  difficult.

Schiffenbauer ACM Briefing February 2005
21
The Big Picture

• Evaluation MUST consider
• Number Needed to Treat to Harm/Help

22
More Deaths on COX-2 Treatment
23
COX-2 Increase Mortality Effects are complex

• Merck claims AD deaths should be ignored because
  theres no pattern.

• Pattern excess deaths were a result of selective
  Cox-2 blocking and accidents due to AD
  conversion.
• Pneumonia Cox-2 is part of response to
  infection.
• GI Bleed deaths Cox-2 needed for ulcer healing.
  
• AD patients have increased rates of accidental
  deaths. Cox-2 treatment increases AD conversion

24
COX-2 Increase Mortality Deaths in Arcoxia
Trials Too

• For CV related deaths there appears to be an
  excess of cases due to etoricoxib compared to
  placebo although the exposure to placebo is
  limited. However, the data related to naproxen
  clearly shows an excess of CV mortality related
  to etoricoxib (and this is consistent with
  comparisons of naproxen to rofecoxib in other
  studies). The explanation for these results is
  not clear and may or may not be related to a
  protective effect of naproxen (not
  demonstrated).
• - Schiffenbauer - February 2005 ACM

25
COX-2s Increase Mortality

• Death Warning Needed on Label for Arcoxia

26
COX-2s may cause Alzheimers Disease
27
COX-2s may cause Alzheimers (AD)

• Relative Risk of AD Conversion Vioxx/Placebo
• Trial would have been stopped by a DSMB, but
  Merck eliminated the DSMB from the protocol.

28
COX-2s may cause AD AD Conversion and Evidence
of Dose Response

• Patients who were more compliant on Vioxx had a
  greater risk of developing AD than those who were
  less compliant.
• ITT (includes all patients regardless of
  compliance)
• AD conversion RR 1.46 p0.01
• For those 80 compliant RR 1.72 ss
• This evidence for dose response was removed from
  published paper.

MRK-I2690008681
29
COX-2s may cause Alzheimers Disease

• AD Warning Needed on Label for Arcoxia Class
  Effect
• Cox-2 drugs should have a warning that states
  that studies have found a relationship between
  use and conversion to AD.

30
Unreliable Safety Data Anti-Platelet Trial
Collaborative (APTC) Strategy
31
Unreliable Safety Data APTC STRATEGY

• In October 2000, cardiology consultants review
  VIGOR meta-analysis.
• Dr. Gertz Notes

Alise - reduce the emphasis on MI- sic
subdivision and subdivision
MRK-NJ01528953 Exh 184 Hermans
32
Unreliable Safety DataAPTC STRATEGY

• The Original Endpoints in SOP
• Primary Event Type
• MI (fatal and non-fatal)
• Unstable angina pectoris
• Ischemic stroke (fatal and non-fatal)
• Secondary Event Type
• Arterial thrombo-embolism (fatal and non-fatal)
• Sudden cardiac death
• Resuscitated cardiac arrest
• Transient Ischemic attack

THE SOP, MRK-AHR0073092
33
Unreliable Safety DataAPTC STRATEGY

• Rationale for Changing Endpoints from Dr.
  Reicins
• Power point

What should be the primary endpoint for the
analysis? Option l- Endpoint as defined in
current CV adjudication plan (confirmed arterial
and venous thromboembolic CV events) PROS
Endpoint as stated in the CV plan All events
adjudicated as planned Most inclusive/greatest
number and therefore the most power Option2--Use
the Anti-platelet trialists definition of events
(MIs CVAs ischemic, thrombotic and hemorrhagic,
vascular death includes sudden death, death due
to unknown cause and death due to GI
bleed) CONS Not predefined in the current
SOP Not all endpoints are being adjudicated
(i.e. unknown cause of death) Includes
"bleeding" endpoints which mixes risk/benefit
(could be seen as a pro since this could work in
our favor) Smaller number of events would
reduce the power of the analysesEmphasis
added.
MRK-NJ0363443
34
Unreliable Safety DataAPTC STRATEGY

• Actually there is no APTC endpoint

MRK-AFL0028614
35
Unreliable Safety DataAPTC STRATEGY

• APTC mixes strokes, MIs bleeding. So when Merck
  changed the endpoint from MI related events (in
  the SOP) to APTC, they did this to hide the MIs. 

36
Unreliable Safety Data Merck knew MI was the
only meaningful endpoint

• One Merck physician stated
• I would have thought we would see some
  difference in the totality of non-MI events, even
  with these small numbers. Apart from the
  ever-present possibility of small-number
  artifact, another possibility is that the
  ''thromboembolic'' terms other than MI were so
  non-specifically used that they represent
  "noise", making "MI" the only category defined
  specifically enough to be meaningful.

MRK-ACF0005856.
37
Unreliable Safety DataDeaths adjudicated but if
CV/T not analyzed
Why is this necessary when all Deaths are
included in APTC Endpoint?
From Merck via Schiffenbauer
38
Unreliable Safety DataAPTC STRATEGY

• Causes of death were subdivided into
  "non-vascular" (that is, definitely non-vascular)
  and "vascular" (that is, definitely or possibly
  vascular, which includes all deaths attributed to
  cardiac, cerebral, hemorrhagic, embolic, other
  vascular, or unknown causes).

APTC Reported Results for ALL Deaths.
BMJ. 1994 Jan 8308(6921)81-106. Collaborative
overview of randomised trials of antiplatelet
therapy--I Prevention of death, myocardial
infarction, and stroke by prolonged antiplatelet
therapy in various categories of patients.
antiplatelet Trialists' Collaboration.
39
Unreliable Safety DataAPTC altered to hide GI
bleeds
Fatal GI bleeds moved to other category
MRK-AJA0004277
40
Unreliable Safety Data APTC Were never going
to be consistent
Sometimes we include deaths sometimes we dont.
MRK-AJA0171877
41
Unreliable Safety Data APTC STRATEGY
42
Unreliable Safety DataAPTC STRATEGY

• Where is the MI/SD only analysis?

43
Unreliable Safety Data Manipulating the
Standard Operating Procedure (SOP)
44
Unreliable Safety Data Manipulating the
Standard Operating Procedure (SOP)
Lots of SOP changes most are post hoc follow
changes in practice.
45
Unreliable Safety Data MANIPULATING THE SOP

• CV SOP has No Cut-off Rule for CV Events

46
Unreliable Safety Data MANIPULATING THE RESULTS

• Study was un-blinded and results were distributed
  3-9-2000. Bad CV data.

14 Day CV cut off rule follows unblinding.
MRK-NJ0121088
47
Unreliable Safety Data MANIPULATING THE SOP
14 Day Rule I knew you would say this was wrong
MRK-NJ0121088
48
Unreliable Safety Data MANIPULATING THE SOP
AD study 078 True ITT but we only use 14 day
data.
MRK-ACF0004015
49
Unreliable Safety Data 14 Day Rule
Post hoc post un-blinding justification
MRK-NJ0121090
50
Unreliable Safety DataRETROACTIVE CHANGES to
events to be adjudicated after unblinding
(unblinded Oct 3 changed Dec 29)
CHF and PUMONARY EDEMA GONE - 21 DELETIONS
12/1999
51
Unreliable Safety Data
CHF removed from adjudication events list too
many events low not no yield.
MRK-AJA0137782
52
Unreliable Safety Data MANIPULATING THE SOP

• SOP re-written to post hoc to match procedures

2003 Chen memo regarding serious AEs eligible
for adjudication, I used the following wording in
my DAP Serious adverse experiences eligible for
adjudication include All deaths, including
all-cause mortality events and deaths not
eligible for inclusion in the all-cause mortality
analyses Serious adverse experiences that the
clinical monitors feel may potentially be
thromboembolic, even though the
investigator-reported terms would not normally be
eligible for adjudication (e.g., a case with a
term of "neurological disorder"). Could you
confirm that bullet 3 was actually done for the
AD studies? If not, Ill simply delete it.
Emphasis added
Chen, J. Email communication, Examples of
spreadsheets for internally adjudicated confirmed
safety endpoints for Vioxx ad. MRK-AFV0210573.
May 28, 2003.
53
Unreliable Safety Data Lots of Post Hoc SOP
Changes

• Standard Operating Procedure for the
  Surveillance, Monitoring, and Adjudication of
  Acute Thrombotic and Embolic Vascular Events and
  Deaths in Clinical Trials of COX-2 Specific
  Inhibitors
• Revised 16-February-1999
• Revised 30-August-1999
• Revised 17-September-2003
• Revised 22-August-2005
• Revised 22-February-2006

54
Unreliable Safety Data CHF under reported
CHF is serious by definition if it was a
diagnosis that was Eligible for Adjudication.
Catch 22 Merck deleted CHF post hoc from list of
eligible events.
See Arcoxia transcript page 161. MRK-AQZ0039486
55
Unreliable Safety DataCHEATING ON
ADJUDICATIONS and MANIPULATION OF INDIVIDUAL
CASES
56
Unreliable Safety DataCHEATING ON ADJUDICATIONS
Not all adjudicated cases made the analysis.
MRK-AJJ0062955
57
Unreliable Safety DataCHEATING ON ADJUDICATIONS

• AN 0158 Alzheimer's Trial

Three out of four adjudicators call 0158 sudden/
unknown death
Barr
Reported to FDA as insufficient data to
adjudicate on VIOXX
MRK-afh0017170
58
Unreliable Safety DataCHEATING ON ADJUDICATIONS

• Merck adjudication loses more Vioxx cases than
  Placebo cases it happens study after study.

59
Unreliable Safety Data Biased adjudication

• Vioxx cases confirmed less often than placebo
  cases. SS 1999 2000
• Vioxx label only included data from through March
  2001.
• Study 078 was not supposed to be adjudicated.

60
Unreliable Safety DataWhy the different rate of
Adj success?

• What Happened 1999-2000?
• 078 began in April 1998 the CV SOP" only
  applied to studies that started after 6/98.
• The SOP was not written until 7/99 it was
  constantly revised after that.
• Although the SOP called for adjudication of all
  deaths, this was not done until sometime in 2001
  under the outside adjudication process. Deaths
  were "internally adjudicated" before after this
  date. 
• Merck did not include adjudicated death (other
  case) diagnosis in analyses if they did not have
  a term that indicated that they should be sent
  for adjudication in the first place (initially
  CRISP, later ad hoc changed to MedDRA with
  numerous ad hoc deletions). This in not in DAP or
  SOP but was used to justify exclusion of case
  5005 others, including all deaths without
  proper terms even if they were adjudicated to MI
  or other endpoint.

61
Unreliable Safety Data Rush to Adjudicate AD
cases for Label
MRK-NJ0124950
62
Unreliable Safety Data Arcoxia Medal Trial

• Reviewers comments
• Of note, there are 9 events in the etoricoxib
  group that had insufficient information vs.
  only 1 in the naproxen group.
• - Schiffenbauer - February 2005 ACM

63
Unreliable Safety DataArcoxiaCT/T Adjudicated
cases missing from analysis
Despite the fact that the CV SOP called for the
adjudication of ALL Deaths Merck excluded CV/T
cases that were reported with ineligible terms
from the analysis of adjudicated cases.
Between 1999 and 2006 Merck deleted over 100
terms from the list of eligible terms.
All deaths in this category APTC were
reported with terms ineligible for adjudication,
but were adjudicated to determine if they
qualified as APTC endpoints. The remaining 16
deaths in the etoricoxib program had terms
eligible for adjudication and thus are included
under Investigator-Reported Events
Schiffenbauer - February 2005 ACM
64
Is the safety data reliable?

• Considerations
• Same players
• Dr. Reicin
• Dr. Laine
• History of Bad Conduct
• VIGOR ASA/Non-ASA lie
• Approve 18 month error
• Hiding the AD ITT analysis

65
Is the safety data reliable?
ACM Relies on Merck Data. If the Merck safety
data is not reliable, then any evaluation of
Arcoxia is not reliable.
66

• Giving the MI Data the Best Face Possible

67
Giving the MI Data the Best Face Possible
I am trying to give this the data the best
face possible that is compatible with the truth
of the data. One can emphasize what one wants
Again- why are we not trying to at least make hay
while the sun is shining. - Gertz of Merck
MRK-AAC0128698.
68
Giving the MI Data the Best Face Possible

• Additionally, Merck used the number of patients
  as the denominator for calculating the MI rates
  but used patient-years for calculating the all
  other rates.
• Mercks own statistician, Joshua Chen, criticized
  this methodology.

69
Giving the MI Data the Best Face Possible

• To make the MI rate look lower than it was Merck
  reported the rate using persons NOT person-years
  in the denominator.
• Merck did not disclose this in the Vigor
  publication.
• The paper did not provide the number of MIs
• All other rates were in person-years actual
  numbers were provided
• MI rate in person years was .63
• (.74 with the
  missing MIs)
• MI rate in persons was .4 (In paper)

70
Giving the MI Data the Best Face Possible
To support the Naproxen hypothesis in Vigor
(Naproxen prevents MIs, rather than Vioxx causing
excess MIs), Merck claimed that the risk
difference in MI rates was not statistically
significant in low risk (non-ASA- indicated
patients). Only true if the three MIs were
deleted.
71
Giving the MI Data the Best Face Possible
Since Merck used Total-Patients (rather than
Patient-Years) to calculate the rate difference
for MI, they had to use Total-Patients to make
the Aspirin Sub-Group Calculations. Excluding
the 3 MIs from the Non-ASA indicated eliminated
the statistical significance of the MI increase
in the non-ASA indicated patients.
Comparison of Non-ASA patients (without the 3
extra MIs)
Comparison of Non-ASA patients (with the 3 extra
MIs)
Totals
No MI
MI
3877
3865
12
Vioxx
3838
3834
4
Naproxen
7715
7699
16
Totals
Chi-square 3.93 p lt 0.05
Chi-square 1.876 0.05 lt p lt 0.20.
72

• Data Hidden from the FDA

73
Hiding Data from the FDA

• Goal Keeping the AD death data off the label.

74
Hiding Data from the FDA AD Deaths CV/T
In the midst of label negotiations, FDA requests
the updated AD death CV/T data (12-5-2001).
75
Hiding Data from the FDA AD deaths CV/T

• Dec. 7, 2001 Merck email and spreadsheet of
  additional Alzheimers data collected since March
  15 shows 39 additional CV/T including 13
  additional deaths.
• At least 5 of the 13 deaths were CV/T.
• At least 6 of the 13 deaths were adjudicated.

MRK-afv0070119
76
Other Lessons from Vioxx

• Advantage Trial
• Secret Internal Adjudications process
• Missing Data Safety Monitoring Board (DSMB)
• Hiding the ITT Data
• VACT Studies
• Misrepresentations in Published Literature, Thal
  et al.
• How Merck really feels about the FDA.
• Other Hidden Analyses

77
ADVANTAGE TRIAL
78
Advantage A Seeding Trial

• Key issues
• Study did not support Naproxen defense theory
• Study did not support high dose defense
• ASA allowed (25 mg) short term (3 month).
• Reported results
• 7 MI/SD on Vioxx to 1 on Naproxen
• Reported results as not statistically
  significant.

79
INTERNAL ADJUDICATIONS
80
Unreliable Safety DataCHEATING ON ADJUDICATIONS
Case 5005 was UNBLINDED on 8/3/2000. FRANTIC
EFFORT TO RE-ADJUDICATE CASE 5005 TO AVOID
STASTICALLY SIGNIFICNT MI/SD
MRK-NJ0124427
81
Case 5005 is unblinded on 11/3/2000, making the
MI/SD statistically significant. Dr. Reicin asks
Barr to change the diagnosis on Nov 8, 2000.
INTERNAL ADJUDICATIONS

• Since the APTC analysis does not change whether
  we call it case 5005 unknown cause of death or
  sudden death I would prefer unknown cause of
  death so that we dont raise concerns. I still
  strongly feel that this should not be an MI - how
  do you know it wasnt a massive stroke, primary
  arrhythmia? Pulmonary embolus etc.
• I am not sure the adjudication process can be
  changed going forward but it is worth discussing
  with Dougbut that does not help us for ADVANTAGE
  OR VIGOR. Emphasis added

MRK-NJ0124427
82
INTERNAL ADJUDICATIONS

• Case 5005 is classified as Unknown Cause of
  Death in published literature.

CHANGED AS REQUESTED
Braunstein, N, Polis, Ann Intern Med. 2005 Jul
19143(2)158-9.
83
Mercks excuse for not counting Case 5005 as
MI/SD the term hypertensive heart disease
did not trigger adjudication in the existing
standard operating procedure. Counting Case
5005 would have made Vioxx/placebo difference
statistically significant.
INTERNAL ADJUDICATIONS
84
SOP called for external adjudication of all
unknown deaths. Internal adjudication not
mentioned.
INTERNAL ADJUDICATIONS
MRK-AHR0073092
85
Missing Data Safety Monitoring Board (DSMB)
86
Missing Alzheimers Data Safety Monitoring Board
(DSMB)
December 18 2001 Letter to FDA from Merck
MRK-01420167265
87
FDA missed the DSMB removal no surprise given
Mercks disclosure
Missing Alzheimers DSMB
MRK-AFO0021822
88
FDA missed the DSMB removal no surprise given
Mercks disclosure
Missing Alzheimers DSMB
MRK-AAF000764
89
Merck knew they needed a DSMB
Missing Alzheimers DSMB
MRK-ARP0047036
90
Missing Alzheimers DSMB

• Nov. 6, 2001 Merck amends 10/15/01 FDA- LABEL
  proposed label, adding favorable language that
  CV/T events occurred in 25 Vioxx patients and 39
  placebo patients.
• Merck proposal omits CV deaths.

Martin Report MRK-NJ0065201
91
Missing Alzheimers DSMB

• December 18, 2001 Silverman responds to FDA.
• Data not given. FDA is told that there is no DSMB
  despite the fact that the protocol called for
  one.
• FDA does NOT respond. Merck Response appears to
  be lost!

MRK-01420167265
92
Hiding the ITT Data
93
Alzheimers Data Analysis Plan (DAP) called for
ITT Analysis
Hiding the ITT Data
MRK-AAD0245328
94
Alzheimers ITT was Pre-Specified Analysis
Hiding the ITT Data
MRK-NJ0186461
95
Merck had it but you never saw it
Hiding the ITT Data
Chen analysis, April 2001.
MRK-AAX0000710
96
Hiding the ITT Data

• Published study A total of 39 deaths occurred
  in patients who were taking study treatment or
  from fatal adverse events that started within 14
  days of the last dose (24 or 3.3 for rofecoxib
  and 15 or 2.1 for placebo).
• ITT mortality analysis
• All cause 41 vs 20
• CVT 17 vs 4

97
CV/T deaths placed on 4/2002 label used
3/16/2001 cut off 8 events on Vioxx/asa vs.
3 on Placebo/asa
Hiding the ITT Data
Alzheimers ITT analysis was not given to FDA and
not in the label.
98
Vioxx vs. Comparator Drugs
99
Merck argues that Arcoxia should be approved
because it MAY work better for patients who fail
on other drugs.
Vioxx vs. Comparator Drugs

• Although Vioxx has shown efficacy in some
  patients, population, data does not show
  superiority of Vioxx over other NSAIDS.

There is no data on the superiority of Arcoxia
compared to existing drugs.
100
Protocol 906 evaluated Vioxx in Celebrex failures
Merck made the same argument for Vioxx and tested
this hypothesis
Protocol 906 found NO Differences in Efficacy but
906 Powerpoint. MRK-AFK0253305
101
June 22, 2001 FDA requests All studies (105,
150, 112, 116, 905, and 906), including complete
analyses of CV events.
FDA asked for the Results
MRK-AAF0004045
102
Protocol 906 Keep Alarming Results tight
Merck never reported the Alarming results to
the FDA
this is a very serious result and you will
hardly be surprised by the idea of keeping this
VERY TIGHT for the moment. -Andreas Moan
Email Communication July 23, 2001. MRK-NJ0199449
103
Vioxx vs. Comparator Drugs
Protocol 906 Alarming Results
I would like to inform you at this moment that
results are not as expected- key efficacy and
safety results are summarized in the attached
document. Please note that the information
included in this document is confidential apart
from both of you and William, other individuals
have not been made aware of the study results.
Email Communication, July 23, 2001.
MRK-NJ0199449.
104
Merck did not Give the FDA Key Adverse Event
Data
Vioxx vs. Comparator Drugs
906 Powerpoint, MRK-AFK0253304.
105
Lessons Learned from Vioxx
Vioxx vs. Comparator Drugs

• Are there similar crossover studies on Arcoxia?
• Is there evidence that Arcoxia works when others
  fail?
• Approval should be not be based on
  unsubstantiated anecdotal theory.

106
Unpublished Placebo Data
107
CV Events and Outcomes Rofecoxib v. Placebo
Merck had 5-2002 Given to FDA
3-2004
Unpublished Placebo Data for the diseases Vioxx
was approved to treat less than 3 month data
3.57 (0.43, 164.23)
2.84 (0.97, 8.38)
108
Mercks pooled, placebo-controlled data for
approved uses showed a 3.29 RR (1.27,8.47) for
TCVSAEs on Vioxx
Unpublished Placebo Data

• Analysis was cut from final Merck slides for 2005
  Advisory Comm. Meeting

109
Hidden evidence that Cox-2s cause AD
110
Misrepresentations in Published Literature, Thal
et al.
Merck writers along with Mercks favorite To be
determined lead author
MRK-AFV0431065
111
Misrepresentations in Published Literature, Thal
et al.
112
Misrepresentations in Published Literature, Thal
et al.

• In the draft paper Merck makes the lack of dose
  response a central argument against the
  proposition that that Cox-2s causes Alzheimers.

113
From Draft
Misrepresentations in Published Literature, Thal
et al.

• Since the above analyses all included patients
  whether or not they were taking study medication,
  we also performed a post hoc analysis excluding
  subjects who were less than 80 compliant with
  taking treatment, as well as those who were
  protocol violators, in order to determine whether
  the effect of rofecoxib was increased in this
  group with greater time-adjusted exposure to
  rofecoxib (as would be expected if the primary
  findings represented a true effect of rofecoxib).
  There was no evidence for an increased effect of
  rofecoxib in this analysis (hazard ratio 1.45,
  P 0.052)."

The original calculated hazard ratio for ITT was
1.46 and that for gt 80 compliant was 1.45, so
Merck argued there was no evidence of dose
response but.
MRK-AFV0431065
114
Misrepresentations in Published Literature, Thal
et al.

• Merck performed the analysis wrong

The real results support causation and show
dose-response 1.46 in the ITT vs. 1.72 in the
gt 80 compliant
MRK-AFS0017022
115
Misrepresentations in Published Literature, Thal
et al.

• Merck deletes this analysis from published paper
  never tells the FDA.
• They substitute an ITT analysis that helps them

The finding that the treatment difference was
not further increased in the analysis restricted
to the on-drug population was also not supportive
of a true treatment effect.
Merck only publishes most favorable data.
116
Aisen et al. also publish that COX-2 makes AD
Worse
Effects of rofecoxib or naproxen vs. placebo on
Alzheimer disease progression a randomized
controlled trial. Aisen PS, Schafer KA, Grundman
M, Pfeiffer E, Sano M, Davis KL, Farlow MR, Jin
S, Thomas RG, Thal LJ Alzheimer's Disease
Cooperative Study. JAMA. 2003 Jun
4289(21)2819-26
117
How Merck Really Feels about the FDA.
118
How Merck Really Feels about the FDA.
April 22, 1999
I had her very close to buying the Merck
position when Ed S. came up and nearly strangled
her and her supervisorThat hurt badly. They
were less willing to talk afterward.
119
How Merck Really Feels about the FDA.
April 6, 2001
I have already told them I think their review
system is an anachronism because they cannot
possibly keep up with science given their hiring
constraints.
I have never seen being nice to the FDA- except
on rare occasions- pay off. Ed Scolnick
MRK-ACR0014514
120
How Merck Really Feels about the FDA.
May 14, 1999
THIS GROUP IS DYSFUNCTIONAL AND WILL NOT RESPOND
TO YOU IN PERSON TELECONS.
No one will remember any of those if Vioxx label
comes out as it is now or is late sic because
we cannot agree with them by may 22nd. And Merck
WILL become a completely different company.
MRK-ABH0015578
121
How Merck Really Feels about the FDA.
October 16, 2001
it is ugly cubed. They sic are bastards
Be assured we will not accept this label.
MRK-ABW0004799
122
How Merck Really Feels about the FDA.
January 21, 2001
we should put all of the data in an
easy-to-digest table make it easy for the
reader, who is overworked, underpaid, and short
on time.
MRK-NJ0443267
123
How Merck Really Feels about the FDA.
April 11, 2001
All in all, the agency seemed more reasonable
than usual (except for Vallalba, who was ignored
by the rest of the group).
MRK-ACR0009153
124
How Merck Really Feels about the FDA.
February 13, 2001
You were FANTASTIC. You made them look like
grade d high school students and you won big huge
and completely. You should be proud happy and
exhausted. Enjoy having done an impossible job
superbly. -Ed Scolnick
MRK-ACT0018064
125
How Merck Really Feels about the FDA.
November 8, 2001
twice in my life I have had to say to the FDA
That label is unacceptable, we will not under
any circumstances accept it. You WILL have to
do that on the cardiac warning for Vioxx. I
assure you.
And I assure you I will NOT sign off on any
lable sic that had a cardiac warning. -Ed
Scolnick
MRK-ACR0009287
126
How Merck Really Feels about the FDA.
April 9, 2001
I think giving them Advantage was not wise. The
alzheimers data vs placebo is helpful. Advantage
is not. numbers are too small. they will data
dredge as they did on original submission and we
will end up with bad labeling. -Ed Scolnick
MRK-ACR0009151
127
Other Hidden Analyses
128
Hiding the MI-only Meta-Analysis

• In October 2000, Merck performed an MI-only
  meta-analysis. Never given to FDA or MCA, who
  asked for it specifically.
• Vioxx vs. all comparators 1.66 (1.05, 2.63)

MRK-ACF0000845
129
Hiding the MI-only Meta-Analysis

• Dr. Scolnick says it should have been given to
  the FDA.
• Merck claims it was withheld because it failed
  test for Homogeneity but Merck gave FDA other
  data from the same set of meta-analyses that also
  failed the test for homogeneity.
• Even if it failed Merck could have used random
  effects model meta-analysis.

130
Hiding the MI-only Meta-Analysis

• They also performed analysis wrong. Sander
  Greenland, author of Modern Epidemiology,
  comments

• Proper pooling requires stratification on study.
  That is, any summary or meta-analytic result
  should be computed keeping the data from each
  study separated, using stratified analysis
  techniques as described for example in Ch. 15
  (Stratified Analysis) of the textbook Modern
  Epidemiology (2nd ed., Rothman and Greenland,
  1998) see also Ch 32 of that book. The main
  purpose for this requirement is to prevent
  confounding (distortion of the estimated
  rofecoxib effect) by study. This confounding
  arises if the proportion exposure to rofecoxib
  and the outcome rate vary across the studies
  being pooled, which appears to be the case based
  on the data reported in the document you sent me
  entitled Vioxx Preliminary Cardiovascular
  Report.
• (Author Deborah Shapiro dated Oct. 18,2000)

131
PLANNED ANALYSISCOMBINE 0-12 MONTH 091 AND 078
MRK-JAK0016070
132
Merck Compares the Data they had to the Data they
Gave the FDA
On study 0-15 picks up 5 placebo events
Mix on study with on drug data in FDA report
October 2001 Bain Memo, MRK-NJ0186461
133
COX-2s AD DATA HIDDEN

• CV/T Deaths in Alzheimers Studies 091 078
  (March 30, 2002)

17 deaths on Vioxx to 5 on placebo (p 0.006)
Vioxx Label Reads
8 deaths on Vioxx to 3 on placebo
134
Out-takes of Mercks Vioxx Video News Release

• Dr. Laine claims Merck broadcasted bogus
  information.
• Arcoxia does not reduce hospitalizations for GI
  bleeds and serious PUBs.

135
Out-takes of Mercks Vioxx Video News Release

• Transcript
• Interviewer You know, lets just take another
  quick crack at the hospitalizations for the VNR,
  alright?
• Dr. Laine And the reason I actually think is
  because those numbers, by the way, that people
  use are totally incorrect, and theyre based on
  just extreme, totally incorrect umm data.
• Interviewer But we keep using them.
• Dr. Laine No. Everybody uses them because they
  sound good.
• Indistinct Comment from Interviewer
• Dr. Laine No, they sound good, but I mean, well
  its the same person who keeps putting them out.
• Interviewer Oh, I see.
• Dr. Laine But I mean I have recalculated also,
  so the only way you can do it is subtract those
  who do from those who dont, and that number
  doesnt take it into account. So to say its due
  to NSAIDs is also incorrect. So theres about
  five different reasons why those numbers are
  totally bogus, but umm, I agree, its in the,
  its out there in the umm common realm and
  everybody uses those numbers. Yeah, I know,
  because its a very impressive sound-byte.

136
Out-takes of Mercks Vioxx Video News Release

• Transcript
• Interviewer Does it help that were using the
  word associated with NSAIDs? Does that sort of,
  water it down a little bit?
• Dr. Laine No, I mean because the issue is umm
  part of the issue is the umm you just dont have
  any idea. Im not saying its actually wrong. The
  death-rate is probably wrong. The
  hospitalizations prob-, may be right. Umm, just
  the death rates probably wrong, but umm anyway.
• Interviewer Alright, lets, lets
• Dr. Laine But as long as we say its estimated,
  or reported its, its not me saying it, so

137
Merck keeps repeating bogus numbers in Arcoxia
ACM document
Out-takes of Mercks Vioxx Video News Release
FDA Arthritis Drug Advisory Committee Meeting,
ARCOXIA (Etoricoxib 30 and 60 mg For Symptomatic
Treatment of Osteoarthritis.Briefing Document
(Background Package) April 12,2007.
138
Out-takes of Mercks Vioxx Video News Release

• Dr. Laine claims We were cagey in how we
  published the data.

139
Out-takes of Mercks Vioxx Video News Release

• Transcript
• Interviewer What were the renal findings in the
  study?
• Dr. Laine Well, umm, thats actually not going
  to be, I mean the only thing thats in the New
  England journal article, says theres no
  difference in renal failure, or renal
  dysfunction. So I dont think you really want to
  go there, do you? Because, there are no data on
  blood pressure or hypertexcuse me, blood
  pressure or edema in the study. And the only
  thing it says specifically, and we were cagey
  about this, was related to renal failure, renal
  dysfunction. And thats not what youre looking
  at, so, I mean I would actually take that out,
  because I think you dont, no I mean I would just
  suggest that anything you do, just as an aside,
  Im gonna talk to Alise in about an hour, but you
  dont wanna talk about that because if you start
  bringing up hypertension and edema its nowhere
  in the study, so if you bring it up its not
  whats in the article.
• Interviewer I agree, I agree.
• Dr. Laine Okay.

140
What data did Merck omit when Dr. Laine said they
were being cagey ?
Out-takes of Mercks Vioxx Video News Release

• ON VIOXX
• Statistically significant increase in
  discontinuances due to hypertension (RR 4.67, 95
  CL 1.93- 11.28)
• Large and important differences in congestive
  heart failure (RR 2.11, 95 CL 0.96-4.67)
• Edema-related discontinuances (RR 1.92, 95 CL
  0.98-3.57)
• Tripled the rate for CHF for those remaining in
  the trial (RR 3, plt0.05)
• Nine fold increase in hypertension (9 on Vioxx,
  0 on Naproxen, plt0.025)
• Statistically significant increase in edema
  events (RR 1.5, plt0.01)
• Discontinuations for edema were borderline
  significant but were significant for a trend
  (plt0.1)
• Total edema rates were elevated (5.4 vs. 3.6)
  and this rate ratio was statistically significant
  (plt.001).