Ranibizumab versus Bevacizumab to Treat
`Neovascular Age-related Macular
`Degeneration
`One-Year Findings from the IVAN Randomized Trial
`
`The IVAN Study Investigators*
`Writing Committee: Usha Chakravarthy, PhD, FRCS,1 Simon P. Harding, MD, FRCS,2 Chris A. Rogers, PhD,3
`Susan M. Downes, MD, FRCS,4 Andrew J. Lotery, MD, FRCS,5 Sarah Wordsworth, PhD,6
`Barnaby C. Reeves, DPhil3
`
`Purpose: To compare the efficacy and safety of ranibizumab and bevacizumab intravitreal injections to treat
`neovascular age-related macular degeneration (nAMD).
`Design: Multicenter, noninferiority factorial trial with equal allocation to groups. The noninferiority limit was
`3.5 letters. This trial is registered (ISRCTN92166560).
`Participants: People ⬎50 years of age with untreated nAMD in the study eye who read ⱖ25 letters on the
`Early Treatment Diabetic Retinopathy Study chart.
`Methods: We randomized participants to 4 groups: ranibizumab or bevacizumab, given either every month
`(continuous) or as needed (discontinuous), with monthly review.
`Main Outcome Measures: The primary outcome is at 2 years; this paper reports a prespecified interim analysis
`at 1 year. The primary efficacy and safety outcome measures are distance visual acuity and arteriothrombotic events
`or heart failure. Other outcome measures are health-related quality of life, contrast sensitivity, near visual acuity,
`reading index, lesion morphology, serum vascular endothelial growth factor (VEGF) levels, and costs.
`Results: Between March 27, 2008 and October 15, 2010, we randomized and treated 610 participants. One
`year after randomization, the comparison between bevacizumab and ranibizumab was inconclusive (bevaci-
`zumab minus ranibizumab ⫺1.99 letters, 95% confidence interval [CI], ⫺4.04 to 0.06). Discontinuous treatment
`was equivalent to continuous treatment (discontinuous minus continuous ⫺0.35 letters; 95% CI, ⫺2.40 to 1.70).
`Foveal total thickness did not differ by drug, but was 9% less with continuous treatment (geometric mean ratio
`[GMR], 0.91; 95% CI, 0.86 to 0.97; P ⫽ 0.005). Fewer participants receiving bevacizumab had an arteriothrom-
`botic event or heart failure (odds ratio [OR], 0.23; 95% CI, 0.05 to 1.07; P ⫽ 0.03). There was no difference
`between drugs in the proportion experiencing a serious systemic adverse event (OR, 1.35; 95% CI, 0.80 to 2.27;
`P ⫽ 0.25). Serum VEGF was lower with bevacizumab (GMR, 0.47; 95% CI, 0.41 to 0.54; P⬍0.0001) and higher
`with discontinuous treatment (GMR, 1.23; 95% CI, 1.07 to 1.42; P ⫽ 0.004). Continuous and discontinuous
`treatment costs were £9656 and £6398 per patient per year for ranibizumab and £1654 and £1509 for
`bevacizumab; bevacizumab was less costly for both treatment regimens (P⬍0.0001).
`Conclusions: The comparison of visual acuity at 1 year between bevacizumab and ranibizumab was
`inconclusive. Visual acuities with continuous and discontinuous treatment were equivalent. Other outcomes are
`consistent with the drugs and treatment regimens having similar efficacy and safety.
`Financial Disclosure(s): Proprietary or commercial disclosures may be found after
`Ophthalmology 2012;119:1399 –1411 © 2012 by the American Academy of Ophthalmology.
`
`the references.
`
`*Group members listed online in Appendix 1 (available at http://aaojournal.org).
`
`Neovascular age-related macular degeneration (nAMD)
`is a common bilateral condition that affects older adults
`and causes severe impairment of central vision. It
`is
`currently treated by intravitreal injection of ranibizumab
`or bevacizumab, an antibody fragment and antibody re-
`spectively to vascular endothelial growth factor (VEGF).
`These treatments maintain vision in ⬎90% of patients,
`but do not cure nAMD. They are expensive because
`patients need monthly review and frequent retreatment
`for ⱖ2 years.
`
`Ranibizumab has been evaluated in multiple trials,1,2
`whereas bevacizumab, originally developed to treat cancer and
`available earlier, has gained widespread acceptance for treating
`nAMD, but without marketing authorization.3– 6 The Compar-
`ison of AMD Treatment Trials (CATT)7 studied monthly or
`as-needed ranibizumab or bevacizumab (4 groups). The CATT
`reported that distance visual acuity after 1 year was equivalent
`for the 2 drugs within each treatment regimen. Ranibizumab as
`needed and monthly were equivalent; the comparison between
`monthly and as-needed bevacizumab was inconclusive. The
`
`© 2012 by the American Academy of Ophthalmology
`Published by Elsevier Inc.
`
`ISSN 0161-6420/12/$–see front matter
`http://dx.doi.org/10.1016/j.ophtha.2012.04.015
`
`1399
`
`Celltrion Exhibit 1051
`Page 1
`
`

`

`Ophthalmology Volume 119, Number 7, July 2012
`
`Considered for inclusion in trial (n=693)*
`
`
`
`PaƟents excluded (n=65)
`Ineligible (n=28)$:
`Best corrected visual acuity <25 leƩers (5)
`No neovascular lesion involving the foveal center (8)
`Fibrosis >50% of the total lesion(6)
`A greatest linear diameter >6000 μm (2)
`8 or more diopters of myopia (1)
`Other acƟve ocular disease causing concurrent vision loss (6)
`Not new referral (5)
`Unknown (1)
`Other (n=37)
`No reason given/ no data in study database (37)
`
`
`
`Randomized (n=628)
`
`Allocated to ranibizumab (n=323)
`
`Allocated to bevacizumab (n=305)
`
`Withdrawals# (n=9):
`PaƟent ineligible (0)
`PaƟent withdrew consent (0)
`On clinical advice (1)
`Randomized in error (1)
`Other, reason unknown (7)
`
`Withdrawals# (n=9):
`PaƟent ineligible (0)
`PaƟent withdrew consent (0)
`On clinical advice (0)
`Randomized in error (4)
`Other, reason unknown (5)
`
`Treatment received (n=314)
`
`Treatment received (n=296)
`
`Withdrawals before
`compleƟon of first 3
`injecƟons (n=2):
`PaƟent withdrew consent (1)
`On clinical advice (0)
`Reason unknown (1)
`
`Withdrawals before
`compleƟon of first 3
`injecƟons (n=2):
`PaƟent withdrew consent (0)
`On clinical advice (1)
`Reason unknown (1)
`
`Completed first 3 injecƟons (n=312)
`
`Completed first 3 injecƟons (n=294)
`
`Allocated to
`conƟnuous treatment
`(n=157):
`Followed up to 12 months
`(n=141)
`Exited trial before visit 12
`but aŌer visit 2 (n=16)
`
`Allocated to
`disconƟnuous
`treatment (n=155):
`Followed up to 12 months
`(n=146)
`Exited trial before visit 12
`but aŌer visit 2 (n=9)
`
`Allocated to
`conƟnuous treatment
`(n=149):
`Followed up to 12 months
`(n=136)
`Exited trial before visit
`12 but aŌer visit 2 (n=13)
`
`Allocated to
`disconƟnuous
`treatment (n=145):
`Followed up to 12 months
`(n=138)
`Exited trial before visit 12
`but aŌer visit 2 (n=7)
`
`
`
`Notes: The exclusions section is incomplete as not all sites have entered full screening data.
`
`* Patients had to consent before they could be considered for the trial; data characterizing patients who withheld consent
`
`could not be collected.
`
`$. Some patients may be ineligible for more than one reason.
`
`# Of the patients who did not drop out, not all of them completed all 3 treatments
`
`Figure 1. Participant flow through the trial.
`
`1400
`
`Celltrion Exhibit 1051
`Page 2
`
`

`

`Chakravarthy et al
`
`䡠 One-Year Outcomes of the IVAN Randomized Trial
`
`Demographics
`
`Age, yrs
`Male gender (n, %)
`Blood pressure, mmHg
`Systolic
`Diastolic
`Nonocular past history (n, %)
`Angina
`Dyspnea*
`Myocardial infarction
`Transient ischemic attack†
`Stroke‡
`DVT/PE§
`Current or past smoker#
`Ocular details
`Best-corrected visual acuity, letters㛳
`Near visual acuity, logMAR**
`Reading index (median, IQR)††
`Contrast sensitivity, letters‡‡
`Total thickness at the fovea, ␮m§§
`Foveal retinal plus subfoveal fluid, ␮m§§
`Foveal center involvement (n, %)
`Choroidal neovascularization##
`Fluid㛳㛳
`Hemorrhage㛳㛳
`Other㛳㛳
`No choroidal neovascularization or
`unable to grade***
`Area of lesion (median, IQR), optic
`disc area㛳㛳
`Serum VEGF (median, IQR), pg/mL†††
`Below lower limit of detection (n, %)
`Quality of life
`EQ-5D state score (median, IQR)‡‡‡
`
`Table 1. Patient Demographics and Past History
`
`Randomized to
`Ranibizumab
`(n ⴝ 314)
`
`Randomized to
`Bevacizumab
`(n ⴝ 296)
`
`Randomized to
`Continuous
`(n ⴝ 308)
`
`Randomized to
`Discontinuous
`(n ⴝ 302)
`
`Overall
`(n ⴝ 610)
`
`77.8
`129
`
`141.9
`76.4
`
`35
`56
`24
`20
`7
`16
`200
`
`7.6
`41%
`
`19.5
`10.2
`
`11%
`18%
`8%
`7%
`2%
`5%
`65%
`
`77.7
`115
`
`143.0
`77.1
`
`51
`60
`22
`9
`7
`18
`185
`
`7.2
`39%
`
`19.5
`9.9
`
`17%
`20%
`7%
`3%
`2%
`6%
`63%
`
`77.8
`126
`
`143.2
`77.4
`
`45
`56
`26
`15
`4
`16
`194
`
`8.0
`41%
`
`19.8
`10.1
`
`15%
`18%
`8%
`5%
`1%
`5%
`64%
`
`77.6
`118
`
`141.7
`76.2
`
`41
`60
`20
`14
`10
`18
`191
`
`6.8
`39%
`
`19.1
`10.0
`
`14%
`20%
`7%
`5%
`3%
`6%
`64%
`
`77.7
`244
`
`142.5
`76.8
`
`86
`116
`46
`29
`14
`34
`385
`
`7.4
`40%
`
`19.5
`10.1
`
`14%
`19%
`8%
`5%
`2%
`6%
`64%
`
`15.0
`61.8
`0.34
`0.66
`47.3 (18.6, 85.7)
`26.2
`6.2
`468
`187
`271
`129
`
`15.6
`61.1
`0.33
`0.67
`43.8 (17.5, 90.9)
`26.3
`5.8
`465
`184
`264
`131
`
`15.5
`60.0
`0.34
`0.70
`41.7 (17.0, 87.0)
`26.1
`6.0
`474
`188
`263
`127
`
`15.0
`62.9
`0.32
`0.63
`51.8 (20.4, 88.9)
`26.4
`5.9
`459
`182
`272
`134
`
`15.3
`61.4
`0.33
`0.66
`46.2 (18.2, 88.2)
`26.2
`6.0
`466
`185
`268
`130
`
`148
`154
`52
`45
`7
`
`56%
`53%
`18%
`16%
`2%
`
`153
`154
`38
`30
`8
`
`59%
`56%
`14%
`11%
`3%
`
`161
`149
`45
`39
`4
`
`61%
`51%
`16%
`13%
`1%
`
`140
`159
`45
`36
`11
`
`54%
`57%
`16%
`13%
`4%
`
`301
`308
`90
`75
`15
`
`58%
`54%
`16%
`13%
`3%
`
`3.30 (1.16, 7.86)
`
`3.97 (1.48, 8.38)
`
`3.64 (1.28, 7.81)
`
`3.86 (1.39, 8.66)
`
`3.71 (1.37, 8.10)
`
`173 (102, 289)
`22
`7%
`
`203 (111, 319)
`22
`7%
`
`193 (100, 308)
`23
`7%
`
`178 (118, 298)
`21
`7%
`
`183 (106, 304)
`44
`7%
`
`0.81 (0.73, 1.00)
`
`0.85 (0.73, 1.00)
`
`0.85 (0.73, 1.00)
`
`0.85 (0.73, 1.00)
`
`0.85 (0.73, 1.00)
`
`Data are presented as mean values and standard deviation, unless otherwise stated.
`DVT ⫽ deep venous thrombosis; IQR ⫽ interquartile range; logMAR ⫽ log(minimum angle of resolution); PE ⫽ pulmonary embolism; VEGF ⫽ vascular
`endothelial growth factor inhibitor.
`Missing data (numbers for ranibizumab continuous, bevacizumab continuous, ranibizumab discontinuous, bevacizumab discontinuous groups,
`respectively): *2 patients with missing values (1, 1, 0, 0); †34 patients with missing data (9, 8, 11, 6); ‡1 patient with missing data (0, 0, 0, 1); §2 patients
`with missing data (0, 0, 1, 1); #6 patients with missing data (3, 0, 2, 1); 㛳1 patient with missing data (0, 1, 0, 0); **7 patients with missing data (3, 2, 0,
`2); ††14 patients with missing data (5, 4, 4, 1); ‡‡4 patients with missing data (3, 1, 0, 0); §§57 patients with missing data (12, 17, 15, 13); ##87 patients
`with missing data (24, 20, 25, 18); 㛳㛳43 patients with missing data (8, 10, 16, 9); ***29 patients with missing data (7, 8, 10, 4); †††54 patients with missing
`data (13, 16, 12, 13); ‡‡‡7 patients with missing data (3, 0, 3, 1).
`
`CATT found no evidence of differences by drug in the fre-
`quency of serious adverse events previously associated with
`anti-VEGF drugs. There were slightly more serious systemic
`adverse events in the bevacizumab groups.
`We have reported herein the 1-year findings of the “alter-
`native treatments to Inhibit VEGF in Age-related choroidal
`Neovascularization” (IVAN) randomized trial, which also
`compares monthly or as-needed ranibizumab or bevacizumab.
`Although the IVAN trial was conceived and designed at the
`same time as the CATT, there are important differences be-
`tween the 2 trials. The IVAN trial has a factorial design, an
`alternative as-needed regimen requiring 3 treatments if active
`disease was detected, measured near visual acuity, reading
`speed, health-related quality of life, and collected serum sam-
`ples at specified times for analysis of VEGF concentrations.
`The IVAN also obtained information on resource use and cost
`
`for a detailed economic evaluation. Moreover, we report a
`meta-analysis of key outcomes from available trials.
`
`Methods
`
`Study Design, Participants, and Setting
`The IVAN is a multicenter, factorial, noninferiority, randomized
`trial with equal allocation to each of 4 groups formed by all
`permutations of 2 drugs and 2 treatment regimens. Allocation to
`drug was masked. Allocation to treatment regimen was not
`masked. Further details are described in the protocol (Appendix 2,
`available at http://aaojournal.org).
`Adults ⱖ50 years old with previously untreated nAMD in the
`study eye and best corrected visual acuity ⱖ25 letters on the Early
`Treatment Diabetic Retinopathy Study chart were eligible.8,9 Di-
`agnosis was confirmed by fluorescein angiography. Participants
`
`1401
`
`Celltrion Exhibit 1051
`Page 3
`
`

`

`Ophthalmology Volume 119, Number 7, July 2012
`
`Table 2. Outcomes at 1 Year*
`
`Best corrected visual acuity, letters†
`Number of treatments (median, IQR)¶
`Near visual acuity, logMAR§,¶
`Reading index (median, IQR)#
`Contrast sensitivity, letters㛳
`Total thickness at fovea, ␮mⴱⴱ,‡
`Retinal thickness plus subfoveal fluid, ␮m**
`Fluid on OCT (n, %)
`Present
`Absent
`Missing data
`Dye leakage on angiogram (n, %)
`Present
`Absent
`Missing data
`Area of lesion (median, IQR), optical disc
`area††
`Serum VEGF (median, IQR), pg/mL‡‡
`Below lower limit of detection (n, %)
`Blood pressure, mmHg§§
`Systolic
`Diastolic
`EQ-5D state score (median, IQR)***
`
`Randomized to
`Ranibizumab
`(n ⴝ 287)
`
`16.0
`69.0
`10 (6, 12)
`0.57
`0.38
`73.8 (27.7, 122.0)
`28.3
`5.19
`322
`139
`172
`78
`
`Randomized to
`Bevacizumab
`(n ⴝ 274)
`
`17.4
`66.1
`11 (7, 12)
`0.62
`0.41
`67.5 (13.7, 120.0)
`28.6
`5.42
`325
`134
`180
`92
`
`Randomized to
`Continuous
`(n ⴝ 277)
`
`17.4
`66.8
`12 (11, 12)
`0.60
`0.39
`73.8 (15.8, 117.9)
`28.6
`5.46
`311
`126
`173
`82
`
`Randomized to
`Discontinuous
`(n ⴝ 284)
`
`16.1
`68.4
`7 (6, 9)
`0.41
`0.58
`70.9 (25.5, 126.5)
`28.4
`5.14
`335
`145
`178
`88
`
`Overall##
`(n ⴝ 561)
`
`16.7
`67.6
`10 (7, 12)
`0.59
`0.40
`71.8 (19.6, 121.6)
`28.5
`5.30
`323
`136
`176
`85
`
`126
`119
`42
`
`44%
`41%
`15%
`
`131
`93
`50
`
`48%
`34%
`18%
`
`109
`123
`45
`
`39%
`44%
`16%
`
`148
`89
`47
`
`52%
`31%
`17%
`
`257
`212
`92
`
`46%
`38%
`16%
`
`29%
`82
`45%
`129
`26%
`76
`0.39 (0.00, 2.44)
`
`31%
`86
`41%
`113
`27%
`75
`0.51 (0.00, 3.06)
`
`24%
`67
`49%
`135
`27%
`75
`0.30 (0.00, 2.17)
`
`36%
`101
`38%
`107
`27%
`76
`0.88 (0.00, 3.41)
`
`30%
`168
`43%
`242
`27%
`151
`0.46 (0.00, 2.94)
`
`151 (100, 277)
`29
`10%
`
`83 (59.5, 157)
`79
`29%
`
`114 (71.0, 196)
`60
`22%
`
`131 (76.9, 263)
`48
`17%
`
`125 (73.8, 215)
`108
`19%
`
`17.3
`138.1
`9.7
`74.5
`0.85 (0.73, 1.00)
`
`18.0
`138.8
`9.6
`75.0
`0.85 (0.73, 1.00)
`
`18.2
`138.4
`9.2
`74.9
`0.85 (0.73, 1.00)
`
`17.1
`138.5
`10.0
`74.5
`0.85 (0.73, 1.00)
`
`17.6
`138.4
`9.6
`74.7
`0.85 (0.73, 1.00)
`
`IQR ⫽ interquartile range; OCT ⫽ optical coherence tomography; logMAR ⫽ log(minimum angle of resolution); VEGF ⫽ vascular endothelial growth
`factor inhibitor.
`*Data are presented as mean values and standard deviation, unless otherwise stated.
`‡The total thickness at the fovea includes the retina, subretinal fluid, choroidal neovascularization, and retinal pigment epithelial elevation.
`Missing data (numbers for ranibizumab continuous, bevacizumab continuous, ranibizumab discontinuous, bevacizumab discontinuous groups, respectively):
`†36 patients with missing data (5, 11, 9, 11); §55 patients with missing data (10, 16, 12, 17); #67 patients with missing data (12, 18, 16, 21); 㛳50 patients
`with missing data (7, 16, 13, 14); **82 patients with missing data (16, 20, 23, 23); ††148 patients with missing data (37, 35, 37, 39); ‡‡21 patients with
`missing data (6, 6, 5, 4); §§38 patients with missing data (5, 12, 10, 11); ***63 patients with missing data (12, 17, 17, 17). ##49 patients withdrew or died
`before 1 year.
`¶Includes all 610 patients.
`
`without a subfoveal (within 200 ␮m) neovascular component were
`eligible if subretinal fluid or serous pigment epithelial detachment
`was subfoveal. To avoid including inactive or advanced disease,
`lesions comprising ⬎50% fibrosis or blood were excluded. Only 1
`eye from each participant was studied.
`We recruited participants from 23 teaching and general hospi-
`tals in the United Kingdom (UK) (Appendix 1, available at http://
`aaojournal.org). A UK National Health Service (NHS) Research
`Ethics Committee gave approval (reference 07/NIR03/37). This
`trial is registered (ISRCTN92166560).
`
`Interventions
`After informed written consent, participants were allocated to 1 of 4
`combinations of the 2 treatment factors: intravitreal injections with ranibi-
`zumab or bevacizumab and continuous or discontinuous regimens.
`Drug doses were ranibizumab 0.5 mg,1,2 bevacizumab 1.25
`mg.7,10,11 Ranibizumab and bevacizumab were procured commer-
`cially. Bevacizumab was repackaged in prefilled syringes in an
`aseptic manufacturing facility.
`The protocol required all participants to attend monthly (win-
`dow, 28 –35 days) for clinical examination, optical coherence
`tomography (OCT), and fundus photography. All participants were
`treated at visits 0, 1, and 2. Participants randomized to the contin-
`uous regimen were treated monthly thereafter. Participants ran-
`domized to the discontinuous regimen were not retreated after visit
`
`2 unless prespecified clinical and OCT criteria for active disease
`were met. If retreatment was needed, a further cycle of 3 doses
`delivered monthly was required.
`Retreatment criteria were any subretinal fluid, increasing in-
`traretinal fluid, or fresh blood. If there was uncertainty about these
`criteria and visual acuity had dropped by ⱖ10 letters, retreatment
`could be initiated. In the absence of fluid on OCT or visual acuity
`deterioration, fluorescein leakage ⬎25% of the lesion circumfer-
`ence or expansion of choroidal neovascularization was required to
`initiate retreatment.
`Decisions about eligibility and retreatment were made on the
`basis of ophthalmologists’ interpretation of OCTs, fluorescein
`angiograms, and fundus photography.
`
`Outcome Measures
`The primary endpoint is at 2 years (follow-up is ongoing), but the
`protocol specified an interim analysis at 1 year. The primary
`outcome measure is best-corrected distance visual acuity measured
`as Early Treatment Diabetic Retinopathy Study letters. Secondary
`outcome measures include (1) adverse effects; (2) EQ-5D (generic
`health-related quality of life assessment);12 (3) cumulative re-
`source use and costs; (4) contrast sensitivity,13 near visual acu-
`ity,14 and reading index;15 (5) lesion morphology and metrics from
`angiograms and OCTs; and (6) serum VEGF levels (sandwich
`enzyme-linked immunosorbent assay, R & Dsystems, Abingdon,
`
`1402
`
`Celltrion Exhibit 1051
`Page 4
`
`

`

`Chakravarthy et al
`
`䡠 One-Year Outcomes of the IVAN Randomized Trial
`
`UK) with detection limits of 2000 to 32 pg/mL. All outcomes
`except EQ-5D and serum VEGF were measured at baseline and
`visits 3, 6, and 12. The EQ-5D was measured at baseline, visits 3
`and 12 and serum VEGF at baseline, visits 1, 11, and 12 (Appendix
`3, available at http://aaojournal.org).
`Adverse events were recorded at each visit. The primary safety
`outcome measure was the occurrence of an arteriothrombotic event
`or heart failure. Events were reviewed and classified using the
`Medical Dictionary for Regulatory Activities [MedDRA] version
`14.1. All serious adverse events were reviewed by senior clinicians
`(U.C., S.P.H., S.M.D., A.J.L.) masked to treatment allocation.
`
`Randomization and Masking
`Randomized allocations were computer generated by a third party
`in blocks and stratified by center. Research teams at sites recruited
`participants, and accessed a password-protected website to ran-
`domize participants. Allocations were concealed until participants’
`eligibility and identities were confirmed.
`We intended that drug allocation should be concealed by having
`separate masked assessment and unmasked treating teams. This sys-
`tem was achieved by 14 sites. At the other 9 sites, staffing levels could
`not support this system and an unmasked staff member prepared
`ranibizumab in a syringe identical to those containing bevacizumab
`and did not perform assessments. To assess the adequacy of masking,
`ophthalmologists and participants stated at visits 3 and 12 (and at exit
`visits if participants withdrew early), whether they knew the allocated
`drug (don’t know/Lucentis/Avastin).
`Lesion morphology was assessed by independent graders,
`masked to drug and treatment regimen, in the UK Network of
`Ophthalmic Reading Centers. Serum VEGF analyses were also
`masked to drug and treatment regimen. Because independent as-
`sessment of lesions could not be done immediately, some random-
`ized participants were subsequently found to be ineligible.
`
`Statistical Analysis
`We specified a noninferiority limit of 3.5 letters, assuming there
`would be no interaction between drug and treatment regimen,
`visual acuity would be analyzed by a mixed model and at least 2
`postrandomization visual acuity measures would be analyzed. We
`set a target sample size of 600, giving 90% power to detect
`noninferiority (significance 2.5%, 1 sided).
`Intention-to-treat analyses were performed. Drugs and dosing reg-
`imens were compared using logistic regression (binary variables) and
`linear mixed model regression (continuous variables), except where
`otherwise noted. Centers were classified into 7 strata with respect to
`the numbers of participants recruited. Analyses adjusted for these
`strata, combining adjacent strata if necessary to allow models to be
`fitted. For continuous variables measured at baseline, values were
`modeled jointly to avoid having to exclude or impute cases with
`missing baseline measures. Interactions with follow-up time were
`fitted and differences between groups are described at 1 year. Model
`validity was checked using standard methods.16 If a model fitted
`poorly, transformations were explored. Outcomes analyzed on a log-
`arithmic scale were transformed back to the original scale after anal-
`ysis and results presented as geometric mean ratios (GMR). For
`Euroqol EQ-5D and lesion area at 1 year, no suitable transformation
`could be found; data were dichotomized, (EQ-5D score, 1 vs ⬍1;
`lesion present vs absent) and analyses adjusted for the baseline value.
`For serum VEGF concentrations below the detection limit for the
`assay (32 pg/mL), values in the range of 16 to 32 pg/mL were
`imputed. Numbers of serious adverse events were compared by drug
`and treatment regimen when ⬎10 participants experienced the event
`(Appendix 3, available at http://aaojournal.org). Likelihood ratio tests
`were used to determine statistical significance.
`
`1403
`
`Figure 2. Best-corrected visual acuity. A, Mean and standard deviation
`of the visual acuity at each visit during the first year of follow-up (by
`ranibizumab and bevacizumab at the top and by continuous and dis-
`continuous treatment regimen below). The circles and squares indicate
`the mean and the bars 1 standard deviation either side of the mean.
`The numbers in parentheses are the number of observations. B, Dif-
`ferences between ranibizumab and bevacizumab (top) and between
`continuous and discontinuous treatment regimen (bottom) in mean
`visual acuity at 1 year (estimated using data from visits 0, 3, 6, and 12,
`adjusted for center size). The circles indicate the mean difference and
`the bars 95% confidence intervals. Negative values reflect a greater
`mean visual acuity at 1 year in the ranibizumab or continuous groups.
`Confidence intervals within ⫺3.5 and ⫹3.5 letters (dashed vertical
`lines) indicate that the 2 groups are equivalent (continuous vs discon-
`tinuous treatment regimen). Confidence intervals extending beyond
`the noninferiority limit of ⫺3.5 letters indicate that the comparison of
`the 2 groups is inconclusive (ranibizumab vs bevacizumab). MD ⫽
`mean difference; SD ⫽ standard deviation; VA ⫽ visual acuity. The
`numbers in brackets give the 95% confidence interval.
`
`Celltrion Exhibit 1051
`Page 5
`
`

`

`Ophthalmology Volume 119, Number 7, July 2012
`
`1
`
`0
`
`0
`
`0
`
`1
`
`1
`
`1
`
`1
`
`1
`
`1
`
`GMR 0.92 (0.84, 1.00), P=0.058
`
`MD 0.20 (-0.47, 0.87), P=0.56
`
`MD -5.53 (-14.59, 3.54), P=0.23
`
`MD 0.50 (-2.06, 3.06), P=0.70
`
`MD 0.10 (-1.37, 1.57), P=0.89
`
`OR 0.82 (0.54, 1.25), P=0.36
`
`OR 0.83 (0.55, 1.24), P=0.36
`
`OR 0.76 (0.52, 1.11), P=0.16
`
`OR 0.82 (0.54, 1.26), P=0.36
`
`GMR 0.97 (0.89, 1.05), P=0.40
`
`GMR 0.97 (0.91, 1.04), P=0.42
`
`GMR 0.47 (0.41, 0.54), P<0.001
`
`1.2
`
`2
`
`20
`
`4
`
`2
`
`2
`
`2
`
`2
`
`1.2
`
`2
`
`Near
`visual
`acuity
`
`0.8
`
`Contrast
`sensitivity
`
`Reading
`index
`
`Systolic
`BP
`Diastolic
`BP
`
`EQ-5D
`score=1
`
`Lesion
`present
`
`Fluid on
`OCT
`
`Dye
`leakage on
`angiogram
`
`Retinal
`thickness
`at fovea
`Total
`thickness
`at fovea
`
`VEGF
`
`-2
`
`20
`
`-4
`
`0
`
`0
`
`0
`
`0
`
`0.8
`
`0
`
`Favors ranibizumab
`
`Favors bevacizumab
`
`Figure 3. Secondary outcomes. Differences between ranibizumab and bevacizumab (above) and between continuous and discontinuous treatment regimen
`(next page) in the secondary functional outcomes at 1 year. The circles indicate the mean difference (MD), geometric mean ratio (GMR), or odds ratio
`(OR) and the bars 95% confidence intervals. Negative values for the MD or ratios less than 1 reflect better functional outcomes at 1 year in the
`ranibizumab or continuous groups. The numbers in brackets give the 95% confidence interval. BP ⫽ blood pressure; OCT ⫽ ocular coherence tomography;
`VEGF ⫽ vascular endothelial growth factor.
`
`Results are reported as effect estimates with 95% confidence
`intervals (CI). Comparisons between drugs are only reported sep-
`arately for the continuous and discontinuous regimens if the inter-
`action of drug and dosing regimen reached a prespecified level of
`statistical significance (5% for foveal thickness and presence of
`fluid on OCT, for which the CATT suggested a possible interac-
`tion,7 or 1% otherwise).
`
`Cost Analysis
`The differences in total 1-year costs between continuous and
`discontinuous dosing regimens were estimated for each drug
`
`from a UK NHS perspective, using established guidelines.17
`Quantities of concomitant medications, hospitalizations, and
`ambulatory consultations attributable to expected adverse
`events were measured using questionnaires completed by par-
`ticipants. Detailed costing of consultations to administer VEGF
`and/or monitor outcomes were carried out using data from 14 of
`the 19 IVAN centers that recruited more than 10 patients.
`Resources were valued in 2010 and 2011 pounds Sterling using
`national sources.18 –21 Ranibizumab was costed at UK list price
`(£742.17/dose18) and bevacizumab at the price charged by the
`trial manufacturing facility (£49/dose). Data were analyzed
`using bootstrapping, allowing for withdrawals and deaths using
`
`1404
`
`Celltrion Exhibit 1051
`Page 6
`
`

`

`Chakravarthy et al
`
`䡠 One-Year Outcomes of the IVAN Randomized Trial
`
`Near
`visual
`acuity
`
`Contrast
`sensitivity
`
`Reading
`index
`
`Systolic
`BP
`Diastolic
`BP
`
`EQ-5D
`score=1
`
`Lesion
`present
`
`Fluid on
`OCT
`
`Dye
`leakage on
`angiogram
`
`Retinal
`thickness
`at fovea
`Total
`thickness
`at fovea
`
`VEGF
`
`0.8
`
`-2
`
`-20
`
`-5
`
`0
`
`0
`
`0
`
`0
`
`0.8
`
`.4
`
`1
`
`0
`
`0
`
`0
`
`1
`
`1
`
`1
`
`1
`
`1
`
`1.2
`
`2
`
`20
`
`5
`
`2
`
`2
`
`2
`
`2
`
`1.2
`
`GMR 0.97 (0.89, 1.06), P=0.53
`
`MD -0.55 (-1.22, 0.12), P=0.11
`
`MD -1.75 (-10.8, 7.31), P=0.71
`
`MD -1.76 (-4.32, 0.80), P=0.18
`
`MD -0.46 (-1.93, 1.01), P=0.54
`
`OR 0.97 (0.64, 1.49), P=0.90
`
`OR 0.60 (0.40, 0.90), P=0.014
`
`OR 0.53 (0.36, 0.77), P<0.001
`
`OR 0.51 (0.33, 0.78), P=0.0017
`
`GMR 0.97 (0.90, 1.05), P=0.47
`
`GMR 0.91 (0.86, 0.97), P=0.0048
`
`GMR 1.23 (1.07, 1.42), P=0.0044
`
`1.6
`1
`Favors continuous Favors discontinuous
`
`Figure 3. (Continued., see preceding page for figure legend).
`
`Kaplan–Meier sample averaging. Analyses were performed
`with Stata version 12 (STATA Corp, Inc, College Station, TX)
`and SAS version 9.2 (SAS, Inc, Chicago, IL).
`
`Results
`
`Participants and Treatment
`Between March 27, 2008, and October 15, 2010, we randomized 628
`participants; 18 were withdrawn before receiving the first treatment,
`leaving 610 who were treated and included in analyses (Fig 1).
`Participants’ characteristics at baseline were similar across the groups
`(Table 1). Nine participants were ineligible. One read ⬍25 letters; 8
`
`failed the independent graded angiographic eligibility criteria, al-
`though 2 had fluid on OCT suggesting an active lesion.
`Regarding adequacy of masking, ophthalmologists reported
`not knowing which drug participants were receiving on 97.9%
`of visits 3, 98.7% of visits 12, and 100% of exit visits; the
`corresponding percentages for participants were 99.3%, 98.6%,
`and 100%.
`There were some protocol deviations (Appendix 4, Tables A1
`to A3, available at http://aaojournal.org). The wrong study drug
`was administered on 2 of 6699 follow-up visits. Adherence to
`treatment regimens was excellent (according to allocation on 6576
`of 6699 visits [98.2%]). Overall, 35% of participants missed ⱖ1
`visit, but the methods of analysis allowed most participants to be
`included.
`
`1405
`
`Celltrion Exhibit 1051
`Page 7
`
`

`

`Ophthalmology Volume 119, Number 7, July 2012
`
`Death from
`any cause
`
`Arteriothrombotic
`or HF
`
`Any vascular
`event
`
`Any vascular
`event or death
`
`Any systemic
`event
`
`Death from
`any cause
`
`Arteriothrombotic
`or HF
`
`Any vascular
`event
`
`Any vascular
`event or death
`
`Any systemic
`event
`
`0
`
`0
`
`1
`
`2
`
`3
`
`Favors bevacizumab
`
`Favors ranibizumab
`
`2
`
`4
`
`6
`
`OR 0.86 (0.26, 2.87), P= 0.81
`
`OR 0.23 (0.05, 1.07), P= 0.03†
`
`OR 0.56 (0.22, 1.41), P= 0.21
`
`OR 0.71 (0.32, 1.56), P= 0.39
`
`OR 1.35 (0.80, 2.27), P= 0.25
`
`OR 1.23 (0.37, 4.07), P= 0.74
`
`OR 1.80 (0.52, 6.26), P= 0.34
`
`OR 1.25 (0.51, 3.08), P= 0.62
`
`OR 1.29 (0.59, 2.80), P= 0.52
`
`OR 1.30 (0.77, 2.19), P= 0.32
`
`Favors discontinuous Favors continuous
`
`Figure 3. (Continued.) Differences between ranibizumab and bevacizumab (top) and between continuous and discontinuous treatment regimen (bottom)
`in the safety outcomes at 1 year. Circles indicate odds ratios and bars indicate 95% confidence intervals. Odds ratios ⬍1 reflect fewer serious adverse events
`during the first year in the bevacizumab or discontinuous groups. HF ⫽ heart failure; OR ⫽ odds ratio. †P value from likelihood ratio test, whereas the
`95% confidence interval was derived from Wald-based normal approximation.
`
`Visual Acuity
`All 610 participants were included in the analysis of visual acuity.
`Mean acuities at 1 year were 69.0 and 66.1 letters in the ranibi-
`zumab and bevacizumab groups, respectively, and 66.8 and 68.4
`letters in the continuous and discontinuous regimens (Table 2).
`The difference between drugs (bevacizumab minus ranibizumab)
`was ⫺1.99 letters (95% CI, ⫺4.04 to 0.06) and between treatment
`regimens (discontinuous minus continuous) was ⫺0.35 (95% CI,
`⫺2.40 to 1.70; Fig 2). The comparison by drug was inconclusive;
`bevacizumab was neither inferior nor equivalent to ranibizumab
`using the 3.5 letter limit. Discontinuous treatment was equivalent
`to continuous treatment.
`
`Secondary Outcomes
`Contrast sensitivity and reading index did not differ significantly
`between drugs or treatment regimens (Fig 3; Table 2). Near visual
`acuity was 8% worse in the bevacizumab group (GMR, 0.92; 95% CI,
`0.84 to 1.00; P ⫽ 0.058), but did not differ with treatment regimen.
`Mean foveal retinal thickness, which we defined as the sum of the
`thickness of the neurosensory retina (measured from the internal
`limiting membrane to its outer boundary with the retinal pigment
`epithelium and which on the stratus OCT output is seen as the outer
`high reflectivity band) and the height of the subretinal fluid (hypore-
`flective area between the neurosensory retina and the outer high
`reflectivity band), did not differ significantly between drugs but was,
`on average, 9% less for participants receiving continuous treatment
`(GMR, 0.91; 95% CI, 0.86 to 0.97; P ⫽ 0.005; Fig 3; Table 2).
`Percentages of participants with fluorescein leakage were 29% and
`31% in the ranibizumab and bevacizumab groups (P ⫽ 0.36) and 24%
`
`and 36% in the continuous and discontinuous groups (P ⫽ 0.002; Fig
`3). Median lesion area in the discontinuous group was larger and
`significantly more participants had evidence of dye leakage, but no
`differences by drug were found.
`The median EuroQol EQ-5D state scores were identical (0.85)
`for all 4 groups and the proportion of participants reporting a state
`score of 1.0 did not differ by drug or treatment regimen (P ⫽ 0.36
`and P ⫽ 0.90; Fig 3).
`Median serum VEGF concentrations at 1 year were lower than
`at baseline in all groups (Appendix 4, Table A7; available at
`http://aaojournal.org). Median serum VEGF concentrations at 1
`year were 151 and 83 pg/mL for ranibizumab and bevacizumab,
`and 114 and 131 pg/mL for continuous and discontinuous regi-
`mens (Table 2). The VEGF concentrations were significantly
`lower at 1 year for bevacizumab than ranibizumab and higher for
`the discontinuous than continuous regimen (GMR, 0.47 [95% CI,
`0.41 to 0.54] and GMR, 1.23 [95% CI, 1.07 to 1.42], respectively;
`P⬍0.01; Fig 3).
`
`Adverse Events
`One year after randomization, 6 participants (1.9%) in the ranibi-
`zumab group and 5 (1.7%) in the bevacizumab group (P ⫽ 0.81) had
`died; 5 (1.6%) had received continuous and 6 (2.0%) discontinuous
`treatment (P ⫽ 0.74; Table 3). Fewer participants treated with bev-
`acizumab compared with ranibizumab had an arteriothrombotic event
`or heart failure (0.7% vs. 2.9%; odds ratio, 0.23; 95% CI, 0.05 to 1.07;
`P ⫽ 0.03; Fig 3), but no difference between treatment regimens was
`found (P ⫽ 0.34). One or more serious systemic adverse events
`occurred in 30 (9.6%) in the ranibizumab group and 37 (12.5%) in the
`bevacizumab group (P ⫽ 0.25; Fig 3). Similarly, 30 (9.7%) in the
`
`1406
`
`Celltrion Exhibit 1051
`Page 8
`
`

`

`