MEMORANDUM
`
`
`
`April 6, 2005
`
`John K. Jenkins, M.D.
`Director, Office of New Drugs (OND)
`
`and
`
`Paul J. Seligman, M.D., M.P.H
`Director, Office of Pharmacoepidemiology and Statistical Science
`(OPaSS)
`
`Steven Galson, M.D., M.P.H.
`Acting Director, Center for Drug Evaluation and Research
`
`NDA files 20-998, 21-156, 21-341, 21-042
`
`
`
`
`
`
`
`
`
`
`DATE:
`
`FROM:
`
`
`
`
`
`
`
`
`
`
`
`
`THROUGH:
`
`
`
`
`TO:
`
`SUBJECT:
`
`
`
`
`
`Analysis and recommendations for Agency action regarding non-
`steroidal anti-inflammatory drugs and cardiovascular risk
`
`
`Executive Summary
`
`Following a thorough review of the available data we have reached the following
`conclusions regarding currently approved COX-2 selective and non-selective non-steroidal
`anti-inflammatory drugs (NSAIDs)1 and the risk of adverse cardiovascular (CV) events:2
`
`
`• The three approved COX-2 selective NSAIDs (i.e., celecoxib, rofecoxib, and
`valdecoxib) are associated with an increased risk of serious adverse CV events
`compared to placebo. The available data do not permit a rank ordering of these
`drugs with regard to CV risk.
`• Data from large long-term controlled clinical trials that have included a comparison
`of COX-2 selective and non-selective NSAIDs do not clearly demonstrate that the
`COX-2 selective agents confer a greater risk of serious adverse CV events than non-
`selective NSAIDs.
`
`
`1 A list of the non-selective NSAIDs is available on http://www.fda.gov/cder/drug/infopage/cox2/default.htm.
`2 The degree of COX-2 selectivity for any given drug has not been definitively established, and there is
`considerable overlap in in-vitro COX-2 selectivity between agents that have been generally considered to be
`COX-2 selective (e.g., celecoxib, rofecoxib, valdecoxib, parecoxib, lumiracoxib, etoricoxib) and older NSAIDs
`that have been considered to be non-selective (e.g., diclofenac, ibuprofen, naproxen). For purposes of
`simplicity of discussion and comparisons, this document maintains the traditional separation between COX-2
`selective and non-selective agents, but our use of this nomenclature should not be considered as FDA
`endorsement of such designations.
`
`
`
`1
`
`LUPIN EX. 1024
`Lupin v. iCeutica
`US Patent No. 8,999,387
`
`

`
`
`
`• Long-term placebo-controlled clinical trial data are not available to adequately assess
`the potential for the non-selective NSAIDs to increase the risk of serious adverse CV
`events.
`• Pending the availability of additional long-term controlled clinical trial data, the
`available data are best interpreted as being consistent with a class effect of an
`increased risk of serious adverse CV events for COX-2 selective and non-selective
`NSAIDs.
`• Short-term use of NSAIDs to relieve acute pain, particularly at low doses, does not
`appear to confer an increased risk of serious adverse CV events (with the exception
`of valdecoxib in hospitalized patients immediately post-operative from coronary
`artery bypass (CABG) surgery).
`• Controlled clinical trial data are not available to rigorously evaluate whether certain
`patients derive greater relief of pain and inflammation from specific NSAIDs
`compared to others or after failing to respond to other NSAIDs.
`• The three approved COX-2 selective drugs reduce the incidence of GI ulcers
`visualized at endoscopy compared to certain non-selective NSAIDs. Only rofecoxib
`has been shown to reduce the risk of serious GI bleeding compared to a non-selective
`NSAID (naproxen) following chronic use. The overall benefit of COX-2 selective
`drugs in reducing the risk of serious GI bleeding remains uncertain, as does the
`comparative effectiveness of COX-2 selective NSAIDs and other strategies for
`reducing the risk of GI bleeding following chronic NSAID use (e.g., concomitant use
`of a non-selective NSAID and a proton pump inhibitor).
`• Valdecoxib is associated with an increased rate of serious and potentially life-
`threatening skin reactions (e.g., toxic epidermal necrolysis, Stevens-Johnson
`syndrome, erythema multiforme) compared to other COX-2 selective agents and is
`the only NSAID with a boxed warning for this adverse event in its approved package
`insert. In the absence of any demonstrated advantage over other NSAIDs, the overall
`benefit versus risk profile for valdecoxib is unfavorable for marketing.
`
`
`Based on these conclusions, we recommend the following regulatory actions to further
`improve the safe and effective use of these drugs by prescribers, patients, and consumers:
`
`
`• The agency should ask Pfizer to voluntarily withdraw Bextra (valdecoxib) from the
`U.S. market. In the event Pfizer does not agree to a voluntary withdrawal, the
`agency should initiate the formal withdrawal procedures; i.e., issuance of a Notice of
`Opportunity for Hearing (NOOH).
`• The professional labeling for all prescription NSAIDs should be revised to include a
`boxed warning highlighting the potential increased risk of serious adverse CV events.
`The boxed warning should also include the well described NSAID class risk of
`serious, and often life-threatening, GI bleeding, which is currently contained in a
`bolded warning.
`• Pending the availability of additional data, the labeling for all prescription NSAIDs
`should include a contraindication for use in patients immediately post-operative from
`CABG surgery.
`
`
`
`2
`
`

`
`
`
`• A class NSAID Medication Guide should be developed to inform patients of the
`potential increased risk of serious adverse CV events and the risk of serious GI
`bleeding.
`• The labeling for non-prescription NSAIDs should be revised to include more specific
`information about potential CV and GI risks and information to assist consumers in
`the safe use of these drugs.
`• The boxed warning for Celebrex (celecoxib) should specifically reference the
`available data that demonstrate an increased risk of serious adverse CV events and
`other sections of the labeling should be revised to clearly reflect these data.
`• The agency should carefully review any proposal from Merck for resumption of
`marketing of Vioxx (rofecoxib). We recommend that such a proposal be reviewed
`by the FDA Drug Safety Oversight Board and an advisory committee before a final
`decision is reached.
`• The agency should request that all sponsors of non-selective NSAIDs conduct and
`submit for FDA review a comprehensive review and analysis of available controlled
`clinical trial databases to further evaluate the potential for increased CV risk.
`• The agency should work closely with sponsors and other interested stakeholders (e.g.,
`NIH) to encourage additional long-term controlled clinical trials of non-selective
`NSAIDs to further evaluate the potential for increased CV risk.
`
`
`Background
`
`Vioxx (rofecoxib) was voluntarily withdrawn from the market by Merck in September 2004
`following the observation of an increased risk of serious adverse CV events compared to
`placebo in a long-term controlled clinical trial. Subsequent to that action, reports of
`additional data from controlled clinical trials became available for other COX-2 selective
`NSAIDs that also demonstrated an increased risk of serious adverse CV events compared to
`placebo. These new data prompted the agency to conduct a comprehensive review of the
`available data and to present the issue for review at a joint meeting of FDA’s Arthritis and
`Drug Safety and Risk Management Advisory Committees on February 16-18, 2005.
`
`Following the joint meeting, CDER conducted a thorough internal review of the available
`data regarding cardiovascular (CV) safety issues for COX-2 selective and non-selective non-
`steroidal anti-inflammatory drugs (NSAIDs). This memorandum summarizes the major
`issues considered in that review, our conclusions regarding the interpretation of the available
`data, and our recommendations for regulatory actions necessary to further improve the safe
`and effective use of these drugs by prescribers, patients, and consumers.
`
`Participants in the CDER review included staff from the Division of Anti-Inflammatory,
`Analgesic, and Ophthalmologic Drug Products, the Division of Over-the-Counter Drug
`Products, the Offices of Drug Evaluation II and V, the Office of New Drugs, the Office of
`Drug Safety, the Office of Biostatistics, the Office of Pharmacoepidemiology and Statistical
`Science, the Office of Medical Policy, the Office of Regulatory Policy, and the Office of the
`Center Director. Materials reviewed included the regulatory histories and the NDA and
`postmarketing databases of the various NSAIDs, FDA and sponsor background documents
`prepared for the Advisory Committee meeting, all materials and data submitted by other
`
`
`
`3
`
`

`
`
`
`stakeholders to the Advisory Committee meeting, presentations made at the Advisory
`Committee meeting, the discussions held by the Committee members during the meeting,
`and the specific votes and recommendations made by the joint Committee.
`
`Summary of available data
`
`The most persuasive evidence in support of an increased risk of serious adverse CV effects
`of the COX-2 selective NSAIDs is derived from a small number of long-term placebo- and
`active-controlled clinical trials in patients with arthritis or in the disease prevention setting.
`We will briefly summarize the available data from the long-term controlled clinical trials for
`the three approved and two investigational COX-2 selective agents. We will also briefly
`summarize the available data from long-term controlled clinical trials to assess the potential
`for increased CV risk for the non-selective NSAIDs. Finally, we will briefly summarize the
`available data from observational studies that have sought to assess the potential for
`increased CV risk for NSAIDs. We will focus our discussion on the combined endpoint of
`death from CV causes, myocardial infarction (MI), and stroke, as that is a widely accepted
`endpoint in assessing the benefits and risks of a drug for CV outcomes. It should be noted
`that the exact definitions and adjudication procedures for this combined endpoint vary to
`some degree across the trials discussed below.
`
`Celecoxib
`
`The strongest data in support of an increased risk of serious adverse CV events for celecoxib
`comes from the National Cancer Institute’s Adenoma Prevention with Celecoxib (APC) trial
`in patients at risk for recurrent colon polyps. In the APC trial a 2-3 fold increased risk of
`adverse CV events was seen for celecoxib compared to placebo after a mean duration of
`treatment of 33 months. There was evidence of a dose response relationship, with a hazard
`ratio3 of 2.5 for celecoxib 200 mg twice daily and 3.4 for celecoxib 400 mg twice daily
`compared to placebo for the composite endpoint of death from CV causes, myocardial
`infarction (MI), or stroke.
`
`The results from the APC trial were not replicated, however, in the nearly identical
`Prevention of Spontaneous Adenomatous Polyps (PreSAP) trial. Based on preliminary,
`unpublished data presented by the PreSAP investigators at the AC meeting, the hazard ratio
`was 1.1 for celecoxib 400 mg once daily compared to placebo for the composite endpoint of
`death from CV causes, MI, or stroke. It is worth noting that the dosing interval differed
`between the APC trial (twice daily) and the PreSAP trial (once daily), although both trials
`included a total daily dose of celecoxib of 400 mg. It remains unclear what, if any, role this
`difference in dosing interval may have played in the disparate findings between the two
`trials.
`
`Another long-term controlled clinical trial of celecoxib versus placebo, the National Institute
`of Aging’s Alzheimer’s Disease Anti-Inflammatory Prevention Trial (ADAPT) in patients at
`
`3 The hazard rate is a measure of risk per unit of time in an exposed cohort (e.g., the event rate per month).
`The hazard ratio is the ratio of the hazard rates from the treatment group relative to the control group, and is
`often used to represent the relative risk when the relative risk is constant over time.
`
`
`
`4
`
`

`
`
`
`risk for Alzheimer’s disease, also does not appear to have shown an increased risk for
`celecoxib 200 mg twice daily compared to placebo for the composite endpoint of death, MI,
`or stroke. Preliminary, unpublished data shared with FDA by the ADAPT investigators
`showed no increased relative risk for celecoxib compared to placebo.4 Finally, there was a
`small one-year trial comparing celecoxib 200 mg twice daily to placebo in patients with
`Alzheimer’s disease that did not demonstrate a significantly increased risk of serious
`adverse CV events, but did show a trend toward more CV events in the celecoxib treatment
`arm.
`
`The only available data from a long-term comparison of celecoxib to non-selective NSAIDs
`come from the Celebrex Long-Term Arthritis Safety Study (CLASS) in which celecoxib 400
`mg twice daily was compared to diclofenac and ibuprofen in approximately 8000 patients
`with osteoarthritis or rheumatoid arthritis. No differences were observed for serious adverse
`CV events between celecoxib and the two non-selective NSAID comparators in this trial.
`
`The ADAPT trial also included naproxen as an active control and will provide an additional
`comparison of celecoxib to a non-selective NSAID when the final study results become
`available. Preliminary, unpublished data shared with FDA by the ADAPT investigators
`showed that celecoxib was intermediate between placebo (lowest incidence) and naproxen
`(highest incidence) for the composite endpoint of death, MI, or stroke.
`
`Rofecoxib
`
`The strongest data from a long-term placebo-controlled trial for an increased risk of serious
`adverse CV events with rofecoxib come from the Adenomatous Polyp Prevention on Vioxx
`(APPROVe) trial in which rofecoxib 25 mg once daily was compared to placebo for up to
`three years. A relative risk of approximately two was seen for rofecoxib compared to
`placebo for serious adverse CV events. It is noteworthy that the rofecoxib and placebo CV
`event curves in a Kaplan-Meier plot did not appear to begin to separate until after
`approximately 18 months of treatment. In contrast to the results seen in APPROVe, two
`long-term placebo-controlled trials in patients with early Alzheimer’s disease, including up
`to four years of treatment in a small number of patients, did not show a significant difference
`in CV events between rofecoxib 25 mg once daily and placebo.
`
`The only long-term controlled clinical trial comparison of rofecoxib to a non-selective
`NSAID comes from the Vioxx GI Outcomes Research (VIGOR) trial in which rofecoxib 50
`mg once daily was compared to naproxen for up to 12 months. In VIGOR, rofecoxib was
`associated with a hazard ratio of approximately two compared to naproxen based on the
`composite endpoint of death, MI, or stroke. In contrast to the findings in APPROVe, in
`VIGOR the Kaplan-Meier CV event curves for rofecoxib and naproxen began to separate
`after approximately two months of treatment.
`
`Valdecoxib
`
`4 Relative risk is defined as the cumulative risk in the treatment group (e.g., number of events per the number
`of individuals in this group) divided by the cumulative risk in the control group. The term relative risk is often
`used interchangeably with the hazard ratio.
`
`
`
`5
`
`

`
`
`
`
`No long-term controlled clinical trials have been conducted comparing valdecoxib to either
`placebo or non-selective NSAIDs. Data are available from two short-term placebo-
`controlled trials of early dosing with intravenous parecoxib (a pro-drug for valdecoxib)
`followed by oral valdecoxib in patients immediately post-operative from coronary artery
`bypass graft (CABG) surgery. In both studies, valdecoxib was associated with an
`approximately two-fold increased risk of serious adverse CV events compared to placebo.
`In contrast, a short-term placebo-controlled trial of intravenous parecoxib followed by oral
`valdecoxib in patients undergoing various types of non-vascular general surgical procedures
`showed no differences for serious adverse CV events.
`
`Investigational COX-2 Selective Agents
`
`Data from long-term controlled clinical trials are also available for two investigational
`COX-2 selective agents (lumiracoxib and etoricoxib), and were presented at the AC meeting.
`These data are summarized here as they provide further insights regarding the issue of CV
`risk for COX-2 selective agents and the comparison of CV risks between COX-2 selective
`drugs and non-selective NSAIDs.
`
`The Therapeutic COX-189 Arthritis Research and Gastrointestinal Event Trial (TARGET)
`compared lumiracoxib 400 mg once daily to naproxen and ibuprofen for one year in
`approximately 18,000 patients with osteoarthritis. TARGET was designed as two sub-
`studies and the planned primary analysis was to be the combined lumiracoxib groups
`compared to the combined naproxen and ibuprofen groups. The study design, however, did
`not clearly reflect this intent since randomization occurred at the sub-study level rather than
`across the entire study. For reasons that are not entirely clear, but possibly related in part to
`the randomization schema, the event rates for serious adverse CV events in the lumiracoxib
`groups in the two sub-studies were very different, i.e., 1.1 events per 100 patient years in the
`naproxen sub-study versus 0.58 events per 100 patient years in the ibuprofen sub-study. The
`event rates for serious adverse CV events for naproxen and ibuprofen were very similar in
`the two sub-studies; i.e., 0.76 events per 100 patient years for naproxen and 0.74 events per
`100 patient years for ibuprofen.
`
`The pre-specified primary analysis of TARGET found no difference in serious adverse CV
`events between the combined lumiracoxib groups and the combined naproxen and ibuprofen
`groups. The validity of combining the two lumiracoxib groups for purposes of the primary
`analysis is debatable, however, given the study design and the very different lumiracoxib
`event rates in the two sub-studies. It is unfortunate that the study design did not call for
`randomization of treatment assignment across the entire study, which would have allowed
`for a much more powerful comparison of lumiracoxib to the two non-selective NSAIDs.
`
`Given the study design, the data from TARGET have also been analyzed by sub-study. In
`the naproxen sub-study, a hazard ratio of 1.44 was observed for the comparison of
`lumiracoxib and naproxen for serious adverse CV events. In the ibuprofen sub-study, a
`hazard ratio of 0.79 was observed for the comparison of lumiracoxib and ibuprofen for
`
`
`
`6
`
`

`
`
`
`serious adverse CV events. The observed differences between lumiracoxib and the NSAID
`comparators were not statistically significantly different in either sub-study.
`
`Depending on which analysis of the TARGET study one considers, the conclusions may be
`very different. The pre-specified primary analysis would suggest that lumiracoxib, a highly
`COX-2 selective agent, is indistinguishable from two non-selective agents with regard to the
`risk of serious adverse CV effects. The sub-study results, however, would suggest that
`lumiracoxib may be associated with a slightly increased CV risk compared to naproxen and
`a slightly decreased CV risk compared to ibuprofen. The cross sub-study comparison of
`naproxen and ibuprofen, however, would suggest no difference in CV risk for these non-
`selective NSAIDs. Overall, this study does not support a clear distinction between
`lumiracoxib and the non-selective NSAIDs.
`
`The Etoricoxib versus Diclofenac Sodium Gastrointestinal Tolerability and Effectiveness
`Trial (EDGE) compared etoricoxib 90 mg once daily versus diclofenac for up to 16 months
`in approximately 7100 patients with osteoarthritis. The relative risk for serious adverse CV
`events was 1.07 for the comparison of etoricoxib to diclofenac (not significantly different).
`EDGE, therefore, is another large controlled clinical trial that did not distinguish COX-2
`selective and non-selective NSAIDs with regard to CV risk.
`
`Non-selective NSAIDs
`
`Long-term placebo- and active-controlled trials are generally not available for the non-
`selective NSAIDs, with the exception of the studies noted above where certain non-selective
`NSAIDs were used as active controls in studies of COX-2 selective drugs.
`
`Observational studies
`
`Data are available from a number of published and unpublished observational studies to
`address the issue of increased risk of serious adverse CV events for COX-2 selective and
`non-selective NSAIDs. These studies have utilized a variety of designs, methods, source
`databases, and comparison groups, and each study has been characterized by strengths and
`weaknesses. In most of the observational studies, the estimated relative risks of the COX-2
`selective NSAIDs have ranged from 0.8 to 1.5, with many point estimates not achieving
`statistical significance. These data were presented and discussed in detail at the AC meeting
`and the committee members generally agreed that the observational data could not
`definitively address the question of a modestly increased CV risk for the COX-2 selective
`compared to the non-selective NSAIDs, with the possible exception of data on rofecoxib 50
`mg.
`
`Overall, the most consistent finding for increased CV risk was observed for rofecoxib 50 mg,
`where statistically significant relative risks of approximately 2 and 3 were seen in two
`studies. The signal for increased CV risk for the 25 mg rofecoxib dose, however, was
`smaller and did not consistently achieve statistical significance. The relative risks in the
`seven observational studies for celecoxib ranged from 0.4 to 1.2, with statistical significance
`observed once for a lowered risk and once for a higher relative risk. The available data for
`
`
`
`7
`
`

`
`
`
`the non-selective NSAIDs from the observational studies are limited, and no consistent
`signals were observed.
`
`Analysis and Conclusions
`
`As noted above, the most persuasive evidence in support of an increased risk of serious
`adverse CV effects of the COX-2 selective NSAIDs is derived from a small number of long-
`term placebo- and active-controlled clinical trials in patients with arthritis or in the disease
`prevention setting. The data from these trials, however, are not consistent in demonstrating
`an increased risk of serious adverse CV effects for COX-2 selective drugs. Perfect
`replication of study results cannot be expected, and is not required to reach a valid scientific
`conclusion. However, the degree of inconsistency observed in the data from long-term
`controlled clinical trials has a considerable impact on our ability to reach valid conclusions
`about the absolute magnitude of increased risk and to make risk versus benefit
`determinations for particular doses of specific drugs.
`
`The data from controlled clinical trial comparisons of COX-2 selective and non-selective
`NSAIDs do not clearly demonstrate an increased relative risk for the COX-2 selective drugs,
`despite the substantial size of these studies. Only VIGOR clearly indicates such a difference
`with CLASS and EDGE giving no suggestion of a difference and TARGET giving analysis-
`dependent results. These findings, and the absence of any long-term placebo- or active-
`controlled clinical trials for most of the non-selective NSAIDs, make it difficult to conclude
`that the COX-2 selective drugs as a class have greater CV risks than non-selective NSAIDs.
`The data from the well-controlled observational trials also have not provided consistent
`assessments of risk when comparing COX-2 selective and non-selective NSAIDs. The point
`estimates of the relative risk comparisons from these data are mostly in a range where
`interpretation may be difficult and influenced by uncontrolled residual confounding or
`biases often inherent in the design and data limitations of these studies
`
`
`Despite the limitations of the available data, overall, there is evidence, principally from a
`small number of placebo-controlled trials, that the approved COX-2 selective NSAIDs (i.e.,
`celecoxib, rofecoxib, valdecoxib) are associated with an increased risk of serious adverse
`CV events (e.g., MI, stroke, and death). It remains unclear, however, that it is the presence
`of, or the degree of, COX-2 selectivity that accounts for these observations, as some have
`hypothesized. As noted above, in various controlled clinical trials, COX-2 selective drugs
`have been indistinguishable from non-selective NSAIDs (i.e., ibuprofen, diclofenac) in
`studies of substantial size and duration. Further, although on theoretical grounds the
`addition of low-dose aspirin (a COX-1 inhibitor) to a COX-2 selective drug should resolve
`any increased CV risk caused by COX-2 selectivity, this effect has not in fact been observed
`in several studies in which such comparisons are possible. Taken together, these
`observations raise serious questions about the so called “COX-2 hypothesis,” which
`suggests that COX-2 selectivity contributes to increased CV risk. It, therefore, remains
`unclear to what extent the COX-2 selectivity of an individual drug predicts the drug’s
`potential for an increased risk of adverse CV events compared to drugs that are less COX-2
`selective.
`
`
`
`
`8
`
`

`
`
`
`After carefully reviewing all the available data, we believe that the data are sufficient to
`support a conclusion that celecoxib, rofecoxib, and valdecoxib are associated with an
`increased risk of serious adverse CV events when compared to placebo. For celecoxib and
`rofecoxib these conclusions are primarily supported by the data from the APC and
`APPROVe trials, respectively. However, for celecoxib a nearly identical long-term placebo-
`controlled trial (the PreSAP trial) and a similarly sized placebo-controlled trial in patients at
`increased risk for Alzheimer’s disease did not replicate these findings. For rofecoxib, other
`long-term placebo-controlled trials of equal or greater duration (the Alzheimer’s treatment
`trials) did not replicate the APPROVe findings. There are no long-term placebo-controlled
`trial data for valdecoxib. It is difficult to know how to extrapolate the findings from the
`parecoxib/valdecoxib CABG trials to the chronic use situation given the significant
`physiologic and traumatic impact on the coronary vasculature during and following CABG
`surgery, and the systemic pro-inflammatory response resulting from heart-lung bypass. We
`believe, however, that it is reasonable from a public health perspective to assume that
`valdecoxib does not differ from the other COX-2 selective agents with regard to increased
`CV risk with chronic use pending the availability of data from long-term controlled clinical
`trials that would indicate otherwise.
`
`The long-term controlled clinical trial data comparing COX-2 selective agents (i.e.,
`celecoxib, rofecoxib, lumiracoxib, etoricoxib) to non-selective NSAIDs are limited in
`number, but include several trials of very substantial size. They raise significant unresolved
`questions. First, rofecoxib 50 mg clearly appears to have an increased risk of serious
`adverse CV events compared to naproxen based on the data from the VIGOR trial.5 The
`absence of a placebo arm in the VIGOR trial, however, precludes a determination of
`whether chronic use of naproxen might also confer an increased risk of serious adverse CV
`events, albeit at a lower rate than rofecoxib. The VIGOR trial also does not provide a
`comparison between lower doses of rofecoxib and naproxen. Other controlled clinical trial
`data have also suggested some increased risk of serious adverse CV events for COX-2
`selective agents versus naproxen (i.e., lumiracoxib in the naproxen sub-study in TARGET
`and etoricoxib in the NDA database); however, these studies also leave unresolved the
`question of whether naproxen is itself associated with an increased CV risk. The ADAPT
`trial is the only long-term controlled clinical trial in which a COX-2 selective agent and
`naproxen have been compared to placebo. The preliminary data from the ADAPT trial,
`however, do not appear to follow the pattern of the other COX-2 selective versus naproxen
`trials, showing a trend toward a higher event rate on naproxen compared to celecoxib and
`placebo (see above). Further, the cross sub-study comparison of naproxen and ibuprofen in
`TARGET suggests no difference in CV risk between these two non-selective NSAIDs.
`Taken together these data provide some support for the conclusion that a difference exits in
`the risk of serious adverse CV events between COX-2 selective agents and naproxen, but
`they do not provide any assurance that naproxen itself confers no increased CV risk; i.e., we
`cannot consider naproxen to be equal to or better than placebo.
`
`
`
`
`5 Rofecoxib 50 mg is not recommended for chronic use in the approved labeling for Vioxx. The higher dose of
`rofecoxib was used in the VIGOR trial to provide a “worst case” estimate of the risk of serious GI bleeding for
`rofecoxib in comparison to naproxen.
`
`
`
`9
`
`

`
`
`
`The comparisons of COX-2 selective agents to certain other non-selective NSAIDs also
`raise interesting, and in the end unresolved, questions regarding the relative risk of COX-2
`selective drugs compared to non-selective NSAIDs, despite the very large size of some of
`the trials. Several long-term controlled clinical trial comparisons of COX-2 selective agents
`to diclofenac have failed to provide evidence that diclofenac has a lower risk of serious
`adverse CV events than COX-2 selective agents (e.g., versus celecoxib in CLASS, versus
`etoricoxib in the NDA database, versus etoricoxib in EDGE). Large, long-term controlled
`clinical trial comparisons of COX-2 selective agents to ibuprofen, an unequivocally non-
`selective agent, also have failed to suggest a clear separation with regard to the risk of
`serious adverse CV events (e.g., versus celecoxib in CLASS, versus lumiracoxib in the
`ibuprofen sub-study in TARGET). While even these large studies cannot rule out a small
`true difference in CV risk between COX-2 selective agents and diclofenac and ibuprofen,
`they show no clear trend and are best interpreted as showing that the risk of serious adverse
`CV events between COX-2 selective agents and either diclofenac and ibuprofen are in fact
`very similar. The latter interpretation, taken together with the findings of an increased risk
`of serious adverse CV events from the long-term placebo-controlled clinical trials of COX-2
`selective agents, would support a conclusion that at least some of the non-selective NSAIDs
`are also associated with an increased risk of serious adverse CV events.
`
`The inability to reliably estimate the absolute magnitude of the increased risk of serious
`adverse CV events for individual COX-2 agents, combined with the inability to reliably
`draw conclusions about the risk of COX-2 agents compared to one another or to other
`NSAIDs, highlights the conundrum the Agency faces in making decisions on appropriate
`regulatory actions. There is an urgent public health need to make appropriate regulatory
`decisions because the adverse events at issue are serious and a very large number of patients
`use selective and non-selective NSAIDs to treat chronic pain and inflammation. At the same
`time, erroneous conclusions and inappropriate actions are themselves potentially harmful to
`the public health. Although the currently available data are not definitive, the Agency
`cannot await more definitive data, which may take years to accumulate from studies that
`have not even begun, before taking action.
`
`In summary, we conclude that the three approved COX-2 selective drugs are associated with
`an increased risk of serious adverse CV events, at least at some dose, with reasonably
`prolonged use. We do not believe, however, that the currently available data allow for a
`rank ordering of the approved COX-2 selective drugs with regard to CV risk. We also
`believe that it is not possible to conclude at this point that the COX-2 selective drugs confer
`an increased risk over non-selective NSAIDs in chronic use. Naproxen may be an exception,
`but the comparative data to COX-2 selective agents are not entirely consistent, we do not
`have adequate long-term placebo-controlled data to fully assess its potential CV risks, and
`the cross sub-study comparison to ibuprofen in TARGET does not suggest a lesser C