CENTER FOR DRUG EVALUATION AND
`RESEARCH
`
`
`APPLICATION NUMBER:
`
`202155Orig1s000
`
`SUMMARY REVIEW
`
`
`
`
`

`

`Deputy Office Director Decisional Memo
`
`
`From
`Robert Temple. MD
`
`Subject
`Deputy Office Director Decisional Memo
`NDA/BLA #
`202155
`Su lement #
`A licant Name
`
`Bristol-M ers Suibb
`
`Date of Submission
`PDUFA Goal Date
`
`S tember 28, 2011; Resubmission Set 17, 2012
`June 8. 2012 (initial). March 17. 2013
`
`Eliquis (apixaban) tablets
`Proprietary Name /
`
`Established (USAN) Name
`Dosage Forms / Strength
`Proposed Indication(s)
`
`Tablet 5 mg and 2.5 mg
`Reduction in the risk of stroke and systemic embolism in
`atients with non-valvular atrial fibrillation.
`
`Material Reviewed/Consulted
`0ND Action Packa - e, includin- :
`
`Medical Officer Review
`
`Medical Team Leader Review
`
`Names of discipline reviewers
`
`Nhi Beasley. PharmD
`Martin Rose. M.D.. JD
`Thomas Marciniak. MD
`
`
`
`_—
`
`Zama Patel — Patient Labelin
`
`_—
`
`De Dir for Safe Review
`
`OND=Office of New Drugs
`OPDP=Office of Prescription Drug Products
`DSI=Division of Scientific Investigations
`CDTL=Cross—Discipline Team Leader
`OSE= Office of Surveillance and Epidemiology
`DEPi= Division of Epidemiology
`DMEPA=Division of Medication Error Prevention and Analysis
`DRISK=Division of Risk Management
`
`Reference ID: 3237200
`
`

`

`I. Introduction
`
`NDA 202155 was submitted on 9/28/2011 and was given a CR on 6/22/12, primarily because of concern
`with the finding that a substantial fraction of patients might have been given the wrong treatment (active
`drug instead of placebo or vice versa). The questions we posed in the 6/22/12 CR letter are repeated in the
`Clinical Review of 12/10/12 (Rose and Beasley) and the applicant’s responses described in detail. An
`addendum to the clinical review dated 12/17/12 addresses a number of additional issues, notably concerns
`raised by Dr. Marciniak about
` (in addition to the drug’s effect on stroke and systemic
`embolism, the primary study endpoint) and recommends approval. Dr. Stockbridge’s Divisional memo of
`12/26/12 summarizes results of two large studies intended to support approval: ARISTOTLE, a non-
`inferiority study comparing apixaban and warfarin, titrated to INR of 2-3, and AVERROES, a superiority
`study (on stroke and systemic embolism) comparing apixaban to aspirin in patients with a perceived need
`to avoid warfarin. Dr. Stockbridge also recommends approval of apixaban.
`
`Whether AVERROES might alone have supported approval, in the absence of a comparison with
`warfarin (which was known to be superior to aspirin in AF) was discussed in the review of the original
`submission, but did not need to be resolved as results of ARISTOTLE became available and were
`submitted in the 9/28/11 NDA.
`
` A
`
` late issue has been whether an effect of apixaban on overall survival has been shown with sufficient
`strength to support inclusion in labeling. Dr. Stockbridge believes it has been credibly shown, but that this
`conclusion refers most clearly to its advantage over placebo/no treatment, not to a clear advantage over
`warfarin. He notes similar findings with dabigatran. It is of interest that, once again, as with dabigatran
`and rivaroxaban, the advantage of apixaban over warfarin on stroke is primarily on hemorrhagic stroke
`with no substantial advantage of apixaban on ischemic stroke. A mortality benefit thus might arise from
`an effect of all of the anticoagulants on ischemic stroke (not clearly greater with apixaban than warfarin)
`and from a lower rate of hemorrhagic strokes than warfarin. Apixaban also showed a clear advantage over
`warfarin on major bleeding.
`
`II. Effectiveness Results
`
`A. Dispensing Errors
`
`
`As noted, the principal reason for our CR response was an apparent high rate of dispensing errors, in
`as much as 7.3% of apixaban patients and 1.2% of warfarin patients. As nicely summarized in Dr.
`Grant’s 6/22/12 review (p 9) there were many opportunities for actual dispensing errors or apparent
`(recording) errors, magnified by the fact that all patients received two bottles (one apixaban or
`apixaban-placebo, one warfarin or warfarin-placebo). A principal source of errors was what was
`written into the electronic CRF (eCRF) as the bottle serial number, possibly reflecting not very clear
`and readable tear-off labels or perhaps just errors in data entry. Subsequent examination of the actual
`tear-off labels in two large samples of patients totalling about 35.5% of all bottles dispensed (possible
`because in the first half of the study the tear-off labels were placed into a paper CRF, and in the
`second half of the study were retained at the site, where they could subsequently be collected). In the
`resubmission the applicant included a 12% random sample collected in response to an EMA request
`and a further 20% random sample in response to the CR letter, ultimately yielding the 35.5% total
`random sample. As detailed in the Rose/Beasley Dec 10 review, about 99.3% of labels at the random
`sites were found and 99.9% of those were visually or barcode legible. Using a variety of analyses,
`including worst case analyses (p 13-22) the reviewers concluded that the findings for the primary
`endpoint (superiority) and bleeding rates (lower with aspirin) are robust.
`
`
`
`
`Reference ID: 3237200
`
`2
`
`(b) (4)
`
`

`

`B. Study Results
`
`
`1. ARISTOTLE
`a. Primary Endpoint – stroke & systemic embolism.
`
`The ARISTOTLE study is fully described in the Rose/Beasley review dated
`5/22/12. Apixaban inhibits Factor X (FXa), which cleaves prothrombin to generate
`thrombin, which converts fibrinogen to fibrin, the fibrous protein that polymerizes
`to form a clot, together with platelets. Apixaban has an apparent half-life of about
`12 hours after oral administration (lengthened by prolonged gut absorption) and
`was given twice daily in ARISTOTLE. There is no available drug to reverse its
`anti-Xa activity.
`
`ARISTOTLE was a randomized, parallel group, double-blind, double-dummy
`comparison with warfarin titrated to a target INR of 2-3, designed to demonstrate
`non-inferiority on a composite endpoint of stroke and systemic embolism in
`subjects with non-vascular AF. The trial included 18,201 patients and was carried
`out worldwide, about 25% in North America (20% US), 19% in Latin America,
`40% in Europe (10% Russia, about 20% Western Europe), and 16% Asia. The trial
`used a target of 448 adjudicated primary endpoint events. Patients had documented
`AF or AFI at enrollment or at least twice, 2 weeks apart, in the year preceding
`enrollment, and at least one risk factor for stroke (age > 75, prior stroke or TIA,
`CHF or LV dysfunction, diabetes, treated hypertension), which would give them a
`CHADS2 score of ≥ 1. There were numerous exclusion criteria (see 5/22/12
`Rose/Beasley review, p 72-73), most related to recent events, bleeding risk, or
`other risks. Randomization was stratified by site and by whether patients were
`already receiving warfarin (naïve or experienced); if they were receiving warfarin,
`it was stopped till INR fell below 2. The apixaban dose was 5 mg bid in most
`patients, but 2.5 mg bid in patients with 2 of the following risk factors for bleeding
`(because of higher blood levels of apixaban): age ≤ 80, weight ≤60 kg, or serum
`creatinine ≥ 1.5 mg/dl.
`
`Events were very thoroughly assessed and classified (see 5/22/12 Rose/Beasley
`review). Of note, strokes were classified (CT scan or MRI strongly urged) as
`ischemic, ischemic with hemorrhagic transformation, hemorrhagic, or uncertain.
`Major bleeding, another specified study endpoint, was defined as an acute bleed
`with decrease in Hb of ≥ 2 g/dL, transfusion of ≥ 2 units of packed red calls,
`bleeding that was intracranial, intraspinal, intraocular, pericardial, intra-articular,
`intramuscular with compartment syndrome, retroperitoneal, or fatal. Clinically
`relevant non-major bleeding was bleeding not meeting the above criteria for major
`bleeding, but that led to hospital admission, need for medical or surgical treatment,
`or need for a change in anti-thrombotic treatment.
`
`The specified NI margin was an increased HR of 1.38 (the effect of warfarin is
`quite large, allowing this large margin, representing ruling out a 50% loss of
`warfarin effect) to be ruled out with 95% CI. As will be seen, superiority was
`shown, rendering the planned NI margin unimportant. The planned analysis was of
`time to first event, although the components, as well as many other endpoints, were
`examined (kind of stroke, AMI, mortality, cause-specific mortality, various kinds
`of bleeds). Ordered endpoints were:
`
`
`Reference ID: 3237200
`
`3
`
`

`

`1. N1 for time to stroke/embolism.
`
`2. Superiority for time to stroke/embolism.
`3. Superiority for time to major bleeding event.
`4. Superiority for time to all-cause mortality.
`
`The primary endpoint was an I'IT analysis following all patients during the intended
`treatment period, but data were clearly not fiilly available for patients lost to follow-up,
`making it more like an on—treatment analysis of the primary endpoint. This is apparent
`fiom the Rose/Beasley Addendum of 12/1 7/ 12 (p 3), which shows the HT and on-
`treatrnent analyses. 1T1" has far more (almost double) fatal events than the on-treatment
`analyses (about twice as many). reflecting the ability to assess vital status in patients off-
`therapy, but there are many fewer additional strokes (about 20% more in the ITT) —
`which analysis is most appropriate is always a matter ofjudgment. HT is often preferred
`in difference-showing trials because it protects against informative censoring, but given
`that an effecting agent is no longer given in the post-treatment period, the I'I'l~ analysis is
`conservative (reducing the apparent effect of an effective treatment). This is a serious
`problem in NI or safety trials, where the ITT analysis. including periods ofl‘-treatment,
`could lead to a finding of no-difference between treatments when there was in fact
`inferiority.
`
`In any event, the ITT results for the primary endpoint (first event) were (Rose/Beasley,
`May 22, p 133).
`
`Table l
`
`Apixaban
`(N=9120)
`
`Warfarin
`(N=9081)
`
`p-value
`
`-——I__—
`
`—:s_————
`_———_—
`
`A p-value of 0.0114 is reasonably low, plain evidence of an effect in a N1 trial and fairly
`strong evidence of superiority. It is notable that most of the advantage of apixaban is on
`hemorrhagic stroke, 38 of the overall advantage of 53, and the percent reduction in
`hemorrhagic stroke is about 50%, vs about 8% for ischemic stroke.
`
`This is even more striking when the events are broken down further (Rose/Beasley, p
`133); note that these are events at any time and total 214 (apixaban) and 267 (warfarin),
`i.e. 2 additional events for each drug.
`
`Reference ID: 3237200
`
`

`

`Table 2
`
`
`
`
`Apixaban
`
`Warfarin
`
`Any stroke
` Ischemic stroke
` Ischemic stroke with hemorrhagic conversion
` Hemorrhagic stroke
` Stroke of uncertain type
`Systemic embolism
`
`199
`140
`12
`40
`14
`15
`
`250
`136
`20
`78
`21
`17
`
`
`In this analysis, essentially all of the advantage of apixaban is on hemorrhagic stroke or
`ischemic stroke with hemorrhagic conversion, 46 of a total difference of 53.
`
`This is of interest because both rivaroxaban and dabigatran also had their largest effects
`on hemorrhagic stroke (no data on ischemic stroke with hemorrhagic conversion). There
`thus may be a tendency for warfarin to induce hemorrhagic strokes that is not fully shared
`by the newer agents.
`
`Analyses were also conducted of shorter follow-up periods than ITT, notably last dose
`plus 2, 7, and 30 days. Not surprisingly, given relatively less time on a therapy that had
`an advantage with the longer follow-up, results are somewhat stronger (Rose/Beasley,
`5/22/12, p 134) for analyses of periods (last dose plus 2 or 7 days) with more time on-
`therapy treatment. As noted, most of the primary endpoint events in the ITT analysis
`occurred on treatment.)
`
`
`Table 3
`
`
`
`
`Primary Endpoint
`
`Last dose + 2 days
`Last dose + 7 days
`Last dose + 30 days
`
`b. Bleeding
`
`Apixaban
`(N=9088)
`
`Warfarin
`(N=9052)
`
`HR
`
`
`p-value
`
`
`176
`184
`218
`
`225
`236
`255
`
`0.77 (0.63, 0.93)
`0.76 (0.63, 0.93)
`0.84 (0.70, 1.00)
`
`0.008
`0.006
`0.05
`
`Major bleeding, as defined above, was less frequent on apixaban, largely because of
`large advantage on intracranial bleeding.
`
`
`
`
`
`
`
`
`
`
`
`
`Reference ID: 3237200
`
`5
`
`

`

`0.72 (0.19. 1.79)
`
`Major bleeding
`
`GI
`Intracranial
`Intra-ocular
`Fatal
`
`0.69 (0.60-0.80)
`
`0.89 (0.70. 1.14)
`0.41 (6.30. 0.57)
`1.42 (0.83. 2.45)
`
`c. Mortality
`
`There has been considerable discussion of the mortality findings in ARISTOTLE. As
`noted, all cause mortality was a specified secondary endpoint and, as all prior endpoints
`were successful. it could be considered.
`
`The mortality results for the ITT analysis are shown in the following table
`(Rose/Beasley. 5/22/12. p 137).
`
`Table 5
`
`Endpomt
`
`Apixaban
`(N=9l20)
`
`E
`
`HR
`
`(N=9081)
`
`(95% CI)
`
`603
`
`308
`W 00
`
`NW
`NW
`
`r—r—-N
`
`All-cause death
`
`CV death
`Caused b w
`
`S stemic embolism
`
`Sudden death
`Heart failure
`Other CV cause
`Unobserved death
`Non-CV death
`Caused b 9
`Bleedin-
`Mali m anc
`Infection
`Trauma
`failure
`Res 1 irato
`Other non-CV cause
`
`Reference ID: 3237200
`
`Unknown cause of death \l0‘
`
`0.89
`
`(0.80. 1.00)
`0.89
`0.76. 1.04
`
`0.93
`0.77. 1.13
`
`0.84
`
`(0.64. 1.09)
`
`ONkl!
`
`0N
`NN
`
`\l
`
`)—r—A LN
`
`LIIN
`
`NWLIILII
`
`M
`
`O\\l
`
`N W
`
`

`

`The deaths have some notable features:
`
`1. Most, but by no means all, of the advantage (difference of 66) is for CV deaths (diff = 36)
`and almost all of that difference (27/36 or 75%) is the difference in fatal stroke, a very
`plausible advantage for apixaban.
`2. The nominal p—value is driven across p=0.05 by the addition of the non-CV deaths. a
`difference of 12, and the unknown cause deaths, a difference of 18. The non-CV deaths
`
`are not plausibly affected by apixaban, but it is likely that the unknowns include some
`undetected CV deaths, (CV plus unknown would have a difference of 54, and an HR of
`about 0.88, probably nominally significant (although I did not calculate it).
`3. A nominally significant effect on CV deaths would plainly have been a more persuasive
`finding (although that was not the identified secondary endpoint).
`4. Many of the deaths occurred long off therapy, as an analysis by Rose/Beasley (addendum
`12/17/12) shows. This analysis examines overall mortality results for populations on
`therapy, or 7 or 30 days off therapy, shown in table below. Primary endpoint events are
`included also, for comparison purposes.
`
`Table 6
`
`-—m Maw-m
`
`.00
`
`.0 “m
`
`200.0
`
`255/9052
`
`Deathm
`
`Death Tx
`
`Death TxLD+7
`
`Death TxLD+
`Stroke SE HT
`550905222
`
`smsmrm
`
`603/9120
`
`265 / 9088
`
`330 I 9088
`
`429 / 9088
`212 / 9120
`270/9020
`
`184/9088
`
`669/9081
`
`295 / 9052
`
`372 / 9052
`
`471 / 9052
`265/9081 __m
`225/9052
`
`220/9052
`
`50020505 2221200
`
`210/9025
`
`As noted above, the table shows that the on-therapy results (or on-therapy plus 7 days) look
`stronger for the primary endpoint (not surprising, as talking the drug provides the benefit) and
`most of the events occur on therapy (perhaps because off-therapy strokes, for patients no longer
`participating actively, are simply not reported). In contrast, a very large number of deaths occur
`off drug. not on face plausibly related to treatment.
`
`Dr. Rose explains why the on-therapy death effect seems weaker, however, in his (12/ 17/ 12, p 5)
`addendum. He shows that in 102 patients with a fatal stroke (on apixaban or warfarin), the stroke
`occurred at the time of discontinuation or within one day in 71%. Their deaths, however, occurred
`afler the stroke in about 70% of patients, although 55% occurred within 7 days. This could
`explain why treatment plus 30 days adds relatively few events to the primary endpoint (about
`18%) but much more (about 60%) to mortality; i.e. people who stop treatment with a stroke are
`relatively likely to die. As Dr. Rose notes. a similar difl'erence between post—treatment primary
`endpoint events and post—treatment deaths was seen with rivaroxaban (ROCKET) and dabigatran
`(RE-LY).
`
`Reference ID: 3237200
`
`

`

`All in all. these data suggest that looking only at on-treatment deaths will miss relevant fatal
`events but will be reasonably good for examining the primary endpoint (p-value is 0.008,
`compared to 0.0114 for UT).
`
`The evidence for an overall mortality advantage for apixaban is thus statistically marginal, but
`marginal mortality findings should not be dismissed In the present case the finding is
`strengthened by the observation that virtually all of the advantage of apixaban is on fatal stroke,
`and the effect of apixaban on stroke. especially hemorrhagic stroke or hemorrhagic conversion, is
`its principal advantage over warfarin. lending the mortality finding greater credibility. Dr. Rose
`finds this consistency, as well as the significant efi'ect on overall mortality, supportive of the
`finding.
`
`Dr. Stockbridge, noting data suggesting. albeit not proving, that warfarin has a mortality effect,
`considers the mortality data supportive of a mortality effect compared to no treatment, but not
`necessarily a clear advantage over warfarin. Dr. Marciniak (review 12/1 7/ 12) has collected data
`on warfarin trials (p 7 of review) and finds that it is, on its face, not very strong. Three of 6 trials
`show RR close to or above 1, with only 1 (BAATAF) showing a significant effect and 2 showing
`a “lean.” It is also noteworthy that on pure thrombotic strokes, the advantage of apixaban is
`relatively small. adding little to whatever warfarin does. Apixaban’s mortality advantage, if real,
`most probably occurs because it does NOT cause as much intracranial hemorrhage.
`
`d. Other Issues
`
`Dr. Marciniak (review dated 12/ 12/ 12) considers loss to follow-up and the marginal mortality
`finding a reason not to include that claim explicitly in indications, a point Drs. Rose and
`Stockbridge concur in. All agree that the overall findings for stroke and systemic embolism are
`strong. That leaves open the question of what to say about the data in section 14 (Clinical
`Studies). discussed fmther below.
`
`2. AVERROES
`
`AVERROES was a randomized. double-blind, double-dummy comparison of apixaban 5 mg
`bid to aspirin 81-324 mg in 5598 patients not taking coumadin because it was shown or
`expected to be unsuitable; patients had risk factors for stroke similar to ARISTOTLE. The
`endpoint was the rate of stroke and systemic embolus, with secondary endpoints including MI
`and vascular death. Results are shown below; essentially all of the difference was on
`stroke/SE (a difference of 61 events) with MI and vascular death adding little (difference of
`65). All cause mortality favored apixaban but was not statistically significant (note, however,
`that the study was stopped early).
`
`Table 7
`
`Apixaban
`(N = 2807)
`
`Aspirin
`(N = 2791)
`
`p-value
`
`0.068
`
`Stroke/SE
`Stroke, SE. MI, vascular death
`All Cause death
`
`0.45 0.32, 0.62
`0.66 0.53, 0.83
`I_IE_ 0.79 0.62, 1.02
`
`< 0.00001
`< 0.00036
`
`Reference ID: 3237200
`
`

`

`
`Interestingly, and not surprisingly, the effect of apixaban compared to aspirin in AVERROES
`was primarily on ischemic stroke. Thus, although apixaban does not seem clearly superior to
`warfarin in its effect on ischemic stroke in ARISTOTLE, both drugs are effective, as the
`comparison with aspirin shows clearly for apixaban, and as is known from previous warfarin
`studies.
`
`
`
`
`FIRST EVENT
`
` Ischemic/unspecified
` stroke
`
` Hemorrhagic stroke
`
` SE
`
`MI
`
`Vascular death
`
`Non-vascular death
`
`
`Table 8
`
`Apixaban
`N=2807
`
`
`
`38
`
`5
`
`2
`
`21
`
`84
`
`27
`
`
`
`Aspirin
`N=2791
`
`HR
`
`
`
`
`94
`
`6
`
`11
`
`23
`
`96
`
`44
`
`
`
`
`
`
`
`
`
`
`
`0.87
`(0.65, 1.17)
`
` 0.62
`(0.38, 1.00)
`
`
`
`Obviously, the striking stroke effect is not surprising, but the effect on non-vascular death
`would be hard to explain.
`
`Bleeding in AVERROES was more frequent on apixaban but fatal and intracranial bleeding
`rates were similar on the two treatments.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Table 9
`
`
`
`
`Apixaban
`
`Aspirin
`
`HR
`
`Major
` Fatal
` Intracranial
`
`45
` 5
`11
`
`29
` 5
`11
`
`1.54 (0.96, 2.45)
` 0.99
` 0.99
`
`p
`
`0.07
`
`
`
`
`
`
`
`
`
`
`Reference ID: 3237200
`
`9
`
`

`

`
`III. Conclusion
`
`Apixaban shows clear effectiveness in decreasing rates of stroke or systemic embolism compared to
`warfarin in an ITT analysis (Table 1), with effect most prominent (Table 2) on hemorrhagic stroke and on
`hemorrhagic stroke plus ischemic stroke with hemorrhagic transformation. The advantage in pure
`ischemic stroke is not clear. Effectiveness is shown both for ITT analysis and on-therapy or on-therapy
`plus 2-7 days (Table 3). Major bleeding events were significantly less frequent with apixaban (Table 4).
`The nominally significant mortality advantage of apixaban should be noted in labeling. The clear
`advantage of apixaban on hemorrhagic stroke is a plausible driver of the mortality effect. It is of interest
`that all three warfarin alternatives have had their largest advantage (rivaroxaban, apixaban) over warfarin
`on the hemorrhagic stroke endpoint with lesser, but real advantages (dabigatran), or no clear advantage on
`ischemic stroke, where warfarin is very effective, AVERROES shows clearly that apixaban has a very
`substantial effect on this endpoint compared to aspirin.
`
`Apixaban should be approved for reducing the rate of stroke on systemic embolism in patients with
`nonvalvular AF. Labeling should include a warning (Boxed Warning and Section 5 warning) of the need
`for particular care when apixaban is discontinued and of increased bleeding risk with a wide range of
`other drugs (aspirin and other anti-platelet drugs, SSRI/SNRI, NSAIDs, other anticoagulants. Like
`rivaroxaban and dabigatran there is no specific reversal agent for apixaban, although activated charcoal
`can block the prolonged absorption of apixaban and speed elimination.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Reference ID: 3237200
`
`10
`
`

`

`---------------------------------------------------------------------------------------------------------
`This is a representation of an electronic record that was signed
`electronically and this page is the manifestation of the electronic
`signature.
`---------------------------------------------------------------------------------------------------------
`/s/
`----------------------------------------------------
`
`ROBERT TEMPLE
`12/28/2012
`
`Reference ID: 3237200
`
`