12/18/04 Interim Review
VIOXX
(rofecoxib)
Review of NDA 21042/s030 (Update of Cardiovascular thrombotic events
in Alzheimer’s studies 078 and
091)
Reviewer: Lourdes Villalba ,
MD.
Clinical Team Leader: James Witter, MD, PhD.
Disclaimer: This is a preliminary review of data submitted
1) Background
NDA 21-042 was approved in May 1999 for the treatment of the
signs and symptoms of OA, acute pain and dysmenorrhea. In year 2000, a large post-marketing outcome
study (VIGOR) showed that the risk of serious GI complications was halved but
the risk of cardiovascular thrombotic (CV/T) events was double for Vioxx 50 mg
as compared to naproxen. Because of the study design, generalization of the
study findings to other populations (such as those patients taking low dose aspirin),
lower doses of Vioxx, and other non-selective NSAIDs comparators was not
possible. At the risk of oversimplifying the picture, potential explanations to
these findings were a plausible pro-thrombotic effect of Vioxx (unopposed
inhibition of prostacyclins), sustained anti-platelet effects of naproxen, or a
combination of both. Additional studies
consistently showed a trend for higher CV/T events against Vioxx 25 mg as
compared to naproxen. However, available
data from placebo-controlled studies for the prevention of Alzheimer’s disease
did not suggest an excess of CV/T events for Vioxx. Labeling changes to include data from VIGOR
and preliminary data from the Alzheimer’s studies were implemented in April,
2002. The current submission of
Summary of
Alzheimer’s studies
Protocol
091 was a placebo-controlled,
parallel-group, multicenter, 15-month double-blind study to evaluate the
efficacy and safety of rofecoxib 25 mg to slow the progression of symptoms of
Alzheimer’s Disease. Patients of either gender who were ≥50 years of age
with established, possible or probable Alzheimer’s
Disease were eligible to participate. Patients using NSAIDs for ≥7
days/month for the 2 months immediately prior to entry were not eligible.
Patients were excluded if they were living in a nursing home or skilled nursing
facility. Eligible patients were
randomized to rofecoxib 25 mg or placebo for 12 months. This was followed by an
additional 3-month treatment phase in which 90% of the patients initially
assigned to rofecoxib were treated with placebo while the other patients
remained on their initial treatment. Safety and tolerability were assessed at
each visit (screening, Months 1, 3, 6, 9, 12, 13.5, and 15). At the time of the Safety Update Report
(SUR)(
Protocol 078 was a placebo-controlled, parallel-group,
double-blind, multi-center study to evaluate the effects of rofecoxib 25 mg on
the prevention of Alzheimer’s Disease and cognitive decline in patients
≥65 years of age with mild
cognitive impairment. Eligible patients were randomized to receive
rofecoxib 25 mg or placebo for 2 years or until 220 cases of clinically
diagnosed probable or possible Alzheimer’s Disease were observed, whichever
came later. Safety and tolerability were to be assessed at all visits. At the
time of the
Protocol
126 was similar in size and
design to study 091. Since study 091 failed to show efficacy, study
126 was considered futile and terminated early, with an approximate median exposure of five months. At the time of the
All the Alzheimer’s studies initially excluded
patients at high cardiovascular risk taking low dose aspirin. The Sponsor later
amended the protocols to allow low dose aspirin for cardiovascular prophylaxis
for those patients who need it. The following are some of the exclusion
criteria used in the Alzheimer’s studies (protocol 078):
1
Patient with
a history (within 2 years) or current evidence of major stroke,
multiple
lacunar infarcts or transient ischemic events.
2
Patient with
a history of angina or congestive heart failure with symptoms at rest.
3
Patients
with a history of myocardial infarction or coronary artery bypass grafting,
angioplasty, or stent placement within 1 year prior to study start.
4
Patients
taking the following medications:
-
Warfarin,
heparin, ticlopidine.
-
NSAIDs
(including salicylates or other aspirin-containing compounds) on a
chronic basis (defined as =7 total days out of the last 30 days for 2
consecutive months prior to potential study entry).
-
Estrogen
replacement therapies (excluding topical cream preparations)
2.
Review of current submission of
At the
time of the April 2002 labeling negotiations, only Vioxx studies 078 and 091,
were included in the label. These studies together had a median exposure of at
least 14 months. The thought was that adding a few CV/T events from a study
with shorter exposure such as study 126, could dilute the results. The current
submission includes updated results of studies 078 and 091 only. This reviewer
has requested additional information from studies 078, 091 and 126 individually
and combined. This information is pending at the time of this review.
a.
Demographics
Demographic
characteristics, co-morbid conditions and concomitant medications were similar
in general in both treatment groups. The studies included an elderly population (mean age 75 years, range 49
to 95 years); 60% were male and 95% were Caucasian. Approximately 6% of
patients in each treatment arm used low dose aspirin.
There was a mild imbalance in the number of
patients with a prior history of MI (3.1 % vs. 4.5% in the Vioxx 25 and placebo
groups, respectively). Also, there was a mild imbalance in the number of
patients with hyperlipidemia: 4.6% and 5.8% in the Vioxx and placebo groups, respectively.
However, there were more patients with greater than one risk factor for
coronary artery disease in the Vioxx group (57.5%) as compared to placebo
(54.8%). It is unclear how these factors may have influenced the results
related to cardiovascular outcomes.
Table 1
shows baseline cardiovascular risks factors and history of a CV/T event
Table
1. Patients with CV risk factors and history of CV/T (ITT)
Source:
Table 2 of s030. Although
a patient may have had 2 or more secondary diagnoses, the patient is counted
only once within a category. The same patient may appear in different
categories.
† Patients with a past history of myocardial infarction, angina pectoris,
angina unstable, coronary artery
disease, coronary artery
occlusion, coronary artery stenosis, ischemic heart disease, ventricular
tachycardia, ST-segment
depression, carotid artery occlusion, carotid artery stenosis, cerebral
infarction, cerebrovascular
accident, cerebrovascular disorder, lacunar infarction, transient ischemic
attack, vertebrobasilar
insufficiency, pulmonary edema, angioplasty, aortic aneurysm repair, cardiac
operation, coronary bypass,
endarterectomy, thromboendarterectomy, or vascular bypass graft.
b) Exposure data
Based on the first dose of study drug, the first patient
entered the trial on
The
cut-off date for the
Table 2. Alzheimer’s studies 078 and 091. Exposure up to
|
|
Rofecoxib 25 mg |
Placebo |
||||
|
091(completed) 078 (ongoing) |
N |
Pt/years at risk |
Median duration (days) |
N |
Pt/years at risk |
Median duration (days) |
|
346 721 |
301 996 |
366 520 |
346 729 |
366 1098 |
448 577 |
|
|
|
1067 |
|
|
1075 |
|
|
Source: Sponsor’s tables. SUR submitted
It is unclear why the number of patients
randomized to each treatment group in the July 12, 2001 submission (1067 and
1075 for Vioxx and placebo, respectively) differs from the numbers in the current
submission (1069 and 1074, for Vioxx and placebo, respectively) and the number
of patients in Table 2 of the ISS for s030 (1071 and 1067, respectively). A request for clarification has been sent to
the Sponsor and is pending at the time of this review.
The mean duration of exposure to study drug as per
the
c) Safety results
1. Deaths
1.1 Total cause mortality
There were 41 and 23 deaths for all causes in the Vioxx 25 and placebo groups respectively in studies 078 and 091 combined, including on-drug deaths (Table 3.a) and deaths that occurred within 14 days off-drug (Table 3.b). Review of the causes of death did not suggest a particular pattern, with the exception of cardiovascular deaths, as noted below. A detailed review of case reports forms has not been conducted by this reviewer yet.
Table 3. Deaths in Alzheimer’s studies 078 and 091 (as
per
Table 3.a Patients with Death in studies 078 and 091, as reported by the investigator. On-Drug population.
Source. Table 11, ISS, s030.
Although a patient may have had 2 or more
clinical adverse experiences, the patient is counted only once within a
category. The
same
patient may appear in different categories.
† This count does not include 5 patients (ANs
0290, 0536, and 1113 from Protocol 078 ; ANs 0376 and
0964 from Protocol 091) with fatal off-drug adverse experiences
that may have been related to previous non-fatal, on-drug adverse experiences. ‡ This count does not include 1 patient (AN 0832) from Protocol 091
with a fatal off-drug adverse experience that may have been related to a
previous non-fatal, on-drug adverse experience.
Table 3.b. Patients
with Off-Drug Death possibly related to previous On-Drug adverse experiences.
Source: Table 12. ISS. S030.
As per Tables 3a and b, total cause mortality was greater for Vioxx 25 mg as compared to placebo.
1.2 Cardiovascular deaths (confirmed by CV adjudication committee)
Of all deaths, 11 and 5 were cardiovascular thrombotic deaths confirmed by the CV adjudication committee (See Table below) in the Vioxx and placebo groups, respectively.
Table 4. Updated listing
of confirmed cardiovascular deaths in Alzheimer’s Studies.
Vioxx 25 mg
(n=11) Placebo
(n= 5)
Protocol 091
1 acute MI 1
sudden cardiac death
1 sudden cardiac death 1 hemorrhagic
stroke
1 ischemic stroke 1
ruptured aortic aneurysm
Protocol
078
7 sudden cardiac death 3
sudden cardiac death
1 acute MI 1
acute MI
In addition to these deaths, there was one hemorrhagic stroke and one ruptured aortic aneurism in the placebo group, in study 091.
Although the numbers
are small, there is greater number of cardiovascular thrombotic deaths in the
rofecoxib 25 mg daily group, as compared to placebo. Of note, there were 8 vs.
4 cases of sudden cardiac death. The finding of more cardiovascular deaths in
the Alzheimer’s studies is not consistent with other studies, such as VIGOR and
APPROVe, in which the number of cardiovascular deaths was the same in Vioxx as
compared to naproxen (VIGOR) and Vioxx as compared to placebo.
Shift analyses of ECG
changes from baseline have been requested and are pending at the time of this
review.
2. Cardiovascular thrombotic events in the
Alzheimer’s studies
2.1 Investigator reported Serious Cardiovascular
Thrombotic events
All cases of investigator reported serious
cardiovascular events (within a list of pre-specified terms used by the sponsor
in prior studies), as well as all deaths (cardiovascular and
non-cardiovascular) were referred for evaluation by a blinded, independent, CV
adjudication committee, that would confirm that the cases were indeed CV
thrombotic.
This list and the number of patients with
investigator reported serious CV/T events referred for adjudication has not
been provided in this application.
2.1 Confirmed or Adjudicated CV/T events
(including fatal and non-fatal events)
Consistent with the interim results of CV/T
events submitted in
Table 5. Confirmed
cardiovascular thrombotic events in studies 078 and 091
Source: Table 20, ISS, s030.
The finding of no difference in MI (14 in each group) but
greater number of ischemic strokes in the placebo group (n=17) as compared to VIoxx
(n= 7) is not consistent with VIGOR (in which there was an increase in the
number of MI (20 vs. 4) but no difference in ischemic strokes (8 and 9) for
Vioxx 50 and naproxen, respectively) or what observed in the APPROVe study (an
excess of MI and strokes for Vioxx as compared to placebo).
(It is unclear whether the 14 acute myocardial
infarctions include the fatal cases. If not, there would be 16 and 15 MIs, not
a large difference. A request for additional information is pending at the time
of this review)
2.3 Analyses of Confirmed
CV/T events over time
Evaluation of CV/T events incidence rates over time shows that the relative risk for these events is lower on Vioxx as compared to placebo during the first 2 years but then, the incidence rate for Vioxx crosses the placebo line (Table 6 and Figure 1).
Of note, the number of patients remaining in the study after two years is relatively small for adequate evaluation of cardiovascular events (243 and 185 for Placebo and Vioxx, respectively) and there is wide overlap of the 95% confidence intervals, which hampers the interpretation of the results. At no time the differences are statistically significant.
Table 6. CV/T events over time in studies 078 and 091
Source, Table 21. ISS. s030.
Figure 1. Kaplan Meier estimates (95% CI) of Time to Confirmed Thrombotic Event (on drug population)
Source: Figure 6. ISS, 030 submission
Of note, this finding of an increased risk of CV/T events
for Vioxx 25 mg after two years as compared to placebo is not consistent with
VIGOR, in which the separation between Vioxx 50 mg and naproxen started at
approximately 6 weeks but was more evident after 6 months, with an increase in
hazard ratio after 8 months. The finding
in the Alzheimer’s studies, however, is consistent with APPROVe (the colon
polyp prevention study) in regards to the inflexion point for an increased CV
risk. However, APPROVe showed no differences in the number of CV deaths between
Vioxx 25 mg and placebo.
3. Other analyses related to cardiovascular safety
3.1 Hypertension-related AEs
As seen in the table below, there was greater use of anti hypertension medication in Vioxx 25 mg as compared to placebo. This is important, however, it is known that selective and non-selective NSAIDs are associated with fluid retention and edema as well as hypertension. The lack of a comparator NSAID in this study precludes any comparative statement.
Table 7: Number of patients with Hypertension related adverse experiences (on-drug population)
Patients with multiple
events may be counted more than once in different terms, but only once in each
term. † Crude incidence (100
× n/N). ‡ Events per 100 patient-years, where patient-years at risk were
calculated based on the overall endpoint. Source: Table 16, ISS, s030
Analyses of risk of HTN related events over time indicate a constant hazard ratio (see table below)
Table 8: Hypertension – Incidence rates over time (On-Drug population)
Source: Table 17, ISS, s030.
Figure 2. Kaplan Meier estimates (95% CI) of time to hypertension (On-drug population)
Source: Figure 4, ISS, s030
Information on change from baseline on systolic and diastolic blood pressure in studies 078 and 091 has been requested and is pending at the time of this review.
3.2 New use of increased dosage of antianginal medication.
There was no difference in the use of antianginal medication between Vioxx and placebo in this study.
Table 9. New use of antianginal medication in 078 and 091
Patients with multiple
events may be counted more than once in different terms, but only once in each
term. † Crude incidence (100 × n/N). ‡ Events per 100 patient-years,
where patient-years at risk were calculated based on the overall endpoint. Source: Table 24. ISS s030.
3.3 Subgroup analyses of CV/T events in patients taking low dose ASA has been requested and is pending at the time of this review.
4. Non-cardiovascular safety
4.1 Serious adverse events
There were no differences in the total number of serious adverse events (SAE) as reported by the investigator. There were 280 (26.2%) and 305 (28.4%) cases in the Vioxx 25 and placebo groups respectively. The total number of SAE by category in general was also similar between groups. Of note, as expected, more cases of GI bleeding were observed with Vioxx as compared to placebo (see Appendix 1). Unfortunately, the study did not include a non-selective NSAID as an active comparator.
4.2 Discontinuations due to AE’s
Total number of discontinuations due to adverse events was
also similar among treatment groups. There were 205 (19%) and 189 (17.6%) such
cases. Evaluation by body system
category shows that for some categories, such as Cardiac, GI and skin systems,
there were more discontinuations from Vioxx as compared to placebo (See
Appendix 2).
Of note, in the tables included in Appendix 1 and 2, patients with multiple events may be counted more than once in different terms, although only once in each term. Additionally, these tables list investigator’s reported events in preferred terms (not confirmed by the GI or CV adjudication committees).
Again, the lack of a non-selective NSAID comparator precludes any statement as to whether Vioxx 25 mg is similar, better or worse than other selective and non-selective NSAIDs regarding GI, CV or skin events.
4.3 Other AE of interest
4.3.1 Fractures
Some theoretical concerns have been raised in the past
regarding the possibility of impaired healing
of fractures with NSAIDs and selective COX-2 inhibitors. The Protocol 078 and 091 combined database
was searched for fracture terms. For the time-to-event analysis, only the first
fracture event for each patient was considered.
There were 65 (6.1 %) patients with fractures in the
rofecoxib group and 59 (5.5 %) in the placebo group (data not shown). The
patient-year adjusted incidence rates were 4.05 and 3.12 per 100 patient-years
for the rofecoxib and placebo groups, respectively. The overall relative risk
(Vioxx over placebo) was 1.27 with 95% CI (0.89, 1.80). Of note, for the 0-6 month period, the
relative risk was 1.85 (1.02, 3.35 95% CI), against Vioxx.
There were wide and overlapping CI after 6 months.
Studies 078 an 091. Updated safety. Risk of fractures
overtime
Source: Table 29, ISS, s030.
Figure 4. KM
estimates (95% CI) of time to fracture. On-drug population.
Source: Figure 10. ISS. S030.
Data from a bone mineral density study conducted by Merck
(083) suggested that rofecoxib does not differ from placebo with regard to
biochemical indices of bone turnover in OA patients over 3 months, and is
comparable to ibuprofen 2400 mg daily after 12 months. This information would
not support a true increase in the risk of fractures for Vioxx as compared to
placebo, but does not answer the question whether both, ibuprofen and Vioxx
have an impact on the risk of fractures as compared to placebo.
4.2 New diagnosis of Cancer
In vitro and animal data suggest that COX-2 selective
inhibitors could be useful for the prevention of different kinds of cancer. At
the time of Vioxx withdrawal from the market, several studies for cancer
prevention were ongoing and have been stopped. Currently, several cancer
prevention studies are still ongoing with Celebrex, another COX-2 selective
NSAID. The Protocol 078 and 091 combined database was searched for cancer
terms. For the time-to-event analysis, only the first cancer event for each
patient was considered.
There were 67 (6.3%) patients with cancer in the rofecoxib
group and 93 (8.7%) in the placebo group (data not shown). The patient-year
adjusted incidence rates were 4.18 and 5.02 per 100 patient-years for the
rofecoxib and placebo groups, respectively. The relative risk (rofecoxib over
placebo) was 0.83 with 95% CI (0.61, 1.14), therefore, not statistically
different.
Of note, there was some evidence for a reduced rate of
basal cell carcinomas on Vioxx (1.31) versus placebo (2.10). The incidence of
other types of cancers was too small to allow meaningful comparisons.
4.3 Efficacy for the prevention of progression of Alzheimers’s disease (AD).
Epidemiologic data suggest that NSAIDs may have a
beneficial effect in patients with AD. The efficacy of Vioxx 25 mg as compared
to placebo in the prevention of progression of established AD (study 091) and
in the prevention of progression to established AD in patients with Mild
Cognitive Impairment (study 078) has been reviewed separately by Dr. Ranjit
Mani, in the Division of Neuropharmacologic Drug Products. Vioxx failed to show efficacy in study 091
and showed a trend towards a worsening of cognitive function as compared to
placebo in study 078. The Sponsor had
scheduled a meeting with the FDA to further discuss these findings on
5. Summary
and discussion
The Alzheimer’s studies were not specifically designed or powered to
address CV outcomes, however, they provided a substantial number of MI and
cerebrovascular events for analyses.
Of note, the cut-off date of the data submitted to the FDA in
The updated data from the Alzheimer’s studies submitted
The risk of total confirmed CV/T events over time with Vioxx appeared to
be lower than placebo during the first 2 years and appeared to increase as
compared to placebo after two years (Figure 1). Additionally, total cause
mortality (41 vs. 23) and confirmed cardiovascular thrombotic deaths (11 and 5)
were against Vioxx. This information is not entirely consistent with findings
of the VIGOR and the APPROVe studies, in which there was no difference in the
number of cardiovascular deaths.
The biologically plausible pro-thrombotic effect of COX-2 selective
agents, plus the known effects of fluid retention, edema and hypertension
associated with all NSAIDs, as well as some unknown factors may be playing a
role in the detrimental cardiovascular safety profile of Vioxx as compared to
placebo and naproxen.
Given the lack of information of long-term placebo-controlled studies
with non-selective agents such as diclofenac and ibuprofen, the question will
remain of how the cardiovascular safety profile of Vioxx would have compared to
these agents.
Two additional issues that may need to be further addressed with the
selective and non-selective NSAIDs are the potential for an increased risk of
fractures and the potential for worsening of cognitive function in patients
with mild cognitive impairment.
6. Conclusions: Based on
this interim review of available data from placebo-controlled studies in
patients with either established Alzheimer’s disease or mild cognitive
impairment as submitted by Merck on
Given the lack of information from long-term placebo-controlled studies
with non-selective agents such as diclofenac and ibuprofen, the question will
remain of how the cardiovascular safety profile of Vioxx would have compared to
these agents.
Additional information and further review of adverse events from the
Alzheimer’s studies are pending at the time of this review.