Clinical Effectiveness of Intravitreal Therapy With Ranibizumab vs Aflibercept vs Bevacizumab for Macular Edema Secondary to Central Retinal Vein Occlusion

This randomized clinical trial compares the clinical effectiveness of intravitreal therapy with ranibizumab vs aflibercept vs bevacizumab for the management of macular edema secondary to central retinal vein occlusion.


Type of trial:
A multicentre Phase III, double-masked, randomised, active-controlled, clinical trial

Trial design and methods:
This is a phase III randomised controlled double-masked non-inferiority clinical trial to evaluate the relative clinical and cost-effectiveness of intravitreal bevacizumab and aflibercept compared to ranibizumab in MO due to CRVO at 100 weeks.
One eye only of 459 adult participants (18 years and above) with MO due to CRVO of ≤ 12 months duration will be randomized to bevacizumab [1.25mg in 50ul] vs aflibercept [2.0mg/50ul] vs ranibizumab [0.5mg/50ul] (1:1:1). After mandated administration of treatment in all arms at baseline, 4, 8, and 12 weeks, further intervention will be based on pre-defined MO retreatment criteria.
The primary outcome will be the change in BCVA ETDRS letter score from baseline to 100 weeks . Secondary outcomes will include additional BCVA outcomes, OCT central macular thickness measurements, change from baseline in visual function questionnaire, VFQ-25, use of resources and adverse events. The primary outcome of cost effectiveness analysis will be mean incremental cost per QALY.

Trial duration per participant: 100 weeks
Estimated total trial duration: 178 weeks (recruitment period + follow up period)

Planned trial sites:
Multi-centre study of approximately 40 sites. 1. Subjects of either sex aged ≥ 18 years.
6. SD-OCT central subfield thickness (CST) > 320μm (Spectralis) predominantly due to MO secondary to CRVO in the study eye. See appendix 1 for equivalent CST value for alternative SD-OCT machines.

7.
Media clarity, pupillary dilatation and subject cooperation sufficient for adequate fundus imaging of the study eye.

8.
In cases of bilateral CRVO, if both eyes are potentially eligible, unless the patient prefers otherwise the worst seeing eye will be recruited.

Exclusion Criteria:
The following apply to the study eye only and to the non-study eye only where specifically stated: 1. Macular oedema considered to be due to a cause other than CRVO (e.g. diabetic macular oedema, Irvine-Gass syndrome).
2. An ocular condition is present that, in the opinion of the investigator, might affect macular oedema or alter visual acuity during the course of the study (e.g. vitreomacular traction) 3. Any diabetic retinopathy or diabetic macular oedema at baseline clinical examination of the study eye. 4. Moderate or severe non proliferative diabetic retinopathy (NPDR) or quiescent, treated or active proliferative diabetic retinopathy (PDR) or macular oedema in the nonstudy eye. Note: Mild NPDR only is permissible in the non-study eye.
5. History of treatment for MO due to CRVO in the past 90 days with intravitreal or peribulbar corticosteroids or in the last 60 days with anti-VEGF drugs or >6 prior anti-VEGF treatments in the previous 12 months. 6. Active iris or angle neovascularisation, neovascular glaucoma, untreated NVD, NVE and vitreous haemorrhage or treatment for these conditions in the last 1 month. . Systemic exclusion criteria are: 9. Uncontrolled blood pressure defined as a systolic value > 170mmHg and diastolic value > 110mmHg 10. Myocardial infarction, stroke, transient ischaemic attack, acute congestive cardiac failure or any acute coronary event < 3 months before randomisation. 11. Women of child bearing potential unless using effective methods of contraception throughout the study and for 6 months after their last injection for the trial.
12. Pregnant or lactating women. 13. Males who do not agree to an effective form of contraception for the duration of the study and for 6 months after their last injection for the trial (see section 6.2 for effective methods of contraception).
14. Hypersensitivity to the active ingredients aflibercept, bevacizumab or ranibizumab or any of the excipients of these drugs. 15. Hypersensitivity to Chinese Hamster Ovary (CHO) cell products or other recombinant human or humanised antibodies 16. A condition that, in the opinion of the investigator, would preclude participation in the study. 17. Participation in an investigational trial involving an investigational medicinal product within 90 days of randomisation.

Statistical methodology and analysis:
A detailed statistical analysis plan is in place for the trial. This has been approved by the DMEC and TSC. Further details of statistical considerations can be found in with visual acuity ≥ 6/12 typically retain good vision whilst 80% of those who present with visual acuity ≤ 6/60 96 do not improve to better than 6/60. CRVO is a predominantly unilateral disease but presents bilaterally in 5% 97 of cases. The risk of developing RVO in the contralateral eye is about 5% in 12 months.

98
There are two main subtypes of CRVO-ischaemic and non-ischaemic. Ischaemic CRVO is diagnosed if the 99 total area of angiographic non-perfusion is at least 10 disc areas and it has a poorer prognosis than the non-

112
In a minority of cases, spontaneous resolution of MO occurs without treatment but it does not typically show a 113 corresponding improvement in visual acuity. Therefore, prompt treatment is advocated. Ranibizumab is a 114 humanized, affinity-matured VEGF antibody fragment that binds to and neutralizes all isoforms of VEGF-A 115 and their biologically active degradation products. Ranibizumab was the first anti-VEGF therapy to 116 demonstrate improved visual outcomes in patients with neovascular age related macular degeneration 117 (nvAMD) and is now approved by the FDA and EMA for MO due to CRVO. This is based on the CRUISE 118 study data (3) that showed monthly intraocular ranibizumab therapy improved mean BCVA by +15 ETDRS 119 letters at 6 months and PRN regimen with monthly monitoring improved mean BCVA by +14 letters at 12

124
Aflibercept is a fusion protein of the key domains of VEGF receptors 1 and 2 and human IgG Fc that blocks all 125 VEGF-A isoforms and placental growth factor. It is FDA approved for CRVO based on the GALILEO (6) and 126 COPERNICUS (7) studies that showed a mean gain of +16.2 letters BVCA at 12 months with 60% gaining ≥ 127 15 letters at 12 months. Cataract occurred in < 2% and glaucoma in < 0.58% of patients at 12 months.

156
To date, bevacizumab has been found to be non-inferior to ranibizumab for all visual acuity primary and iii.
There is no comprehensive data regarding the effect of bevacizumab in pregnancy and it is 253 therefore contraindicated. Females of child bearing age will require a negative urine pregnancy test before 254 enrolment in the study and will be advised to use an effective form of contraception throughout the trial 255 and for 6 months after their last trial injection. Participants will also be reminded to notify their local study 256 team if they fall pregnant during this time. The drug will be stopped immediately if a subject does become 257 pregnant. The pregnancy will be reported using a pregnancy form and followed up until outcome. The

346
After participant study eligibility has been confirmed, the date of the milestone visits at weeks 0, 12, 24, 52, 76 347 and 100 weeks will be calculated and agreed. Visits at weeks 4 and 8 will also be fixed to ensure that the 348 patient is able to attend these treatment visits. Following confirmation that the participant is able to attend 349 these visits, randomisation can occur.

350
From this point forward, all intervening follow up visits after week 12 will be flexible and designed to fit around 351 the milestone visits. Since intravitreal injection is not recommended less than 4 weeks after the previous 352 injection, it is likely that there will be 'slippage' of the study visit schedule. This is acceptable and it may be

361
In the context of a non-inferiority study, the protocol is designed to be as flexible as possible to accommodate where visit 'slippage' has occurred, is permissible and not considered a protocol deviation.

366
After mandated administration in all arms at baseline, 4, 8, and 12 weeks, further PRN intervention if 367 retreatment criteria (Section 8.14.2) are met will be administered at weeks 16 and 20.

368
From week 24 to week 96, intervals will initially be 4 weekly (with a -14 to + 14 day visit window) with the 369 potential to increase to 8 weekly (with a -14 to + 14 day visit window) if criteria for 'stability' are achieved.

370
'Stability' is defined as three successive visits from week 16 onwards at which retreatment criteria are not met 371 (Section 8.14.3) and so the first time at which treatment could be deferred for 8 weeks is week 24.

372
Similarly 'Success' is defined as an ETDRS letter score > 83 letters and if present at any retreatment visit then 373 treatment should not be given at that point and the patient reviewed at either 4 or 8 weeks depending on their 374 pre-existing visit schedule. The > 83 letter ETDRS letter score criteria for further study participation is 375 illustrated in Section 8.14.4.

376
At each visit between weeks 24 and 96 inclusively, temporary discontinuation criteria maybe met (See Section 377 8.14.5). If so, the PI or his designee at their discretion can withhold treatment to prevent therapy in a 378 participant who has not responded to at least their last three injections.

379
If retreatment criteria are met at an 8 weekly or unscheduled visit, then 4 weekly visits will be resumed until 380 retreatment criteria are not met again on three occasions and an 8 weekly visit is re-established. If a patient 381 achieves the criteria for success or temporary discontinuation then treatment will be discontinued until 382 retreatment criteria are met again.

402
The following apply to the study eye only and to the non-study eye only where specifically stated: 403 1. Macular oedema considered to be due to a cause other than CRVO (e.g. diabetic macular oedema,
An ocular condition is present that, in the opinion of the investigator, might affect macular oedema or 406 alter visual acuity during the course of the study (e.g. vitreomacular traction)              Individuals that do not meet other modifiable inclusion criteria, e.g. blood pressure, may be re-450 screened a minimum of 2 weeks after the last screening visit.

452
All assessments performed at the initial screening visit should be repeated during the rescreening visit their initial entry on the eCRF system should be updated rather than creating a 'new' patient on the system.

456
This will avoid 'double counting the patients in the CONSORT diagram.' 457 7

Recruitment 458
The study will recruit from approximately 40 centres over an 18 month recruitment period. Recruitment will be 459 competitive; however each site will be allocated a minimum target number of patients to recruit. Sites will be 460 set up strategically to ensure the recruitment period is fully utilised. Patient Identification Centres may be set 461 up to maximise and boost recruitment. Eligible patients will be invited to participate via their local clinics, or 462 via an invitation letter. Eligible participants can either respond directly to the recruiting centre (with notification 463 to their doctor) or the local PIC site via phone or response slip

465
Within each site patients will be identified from subspecialty retina clinics, general clinics, and eye casualty 466 clinics and at which clinical examination and discussion of a study will be undertaken and the PIS provided.

8
Study procedures and schedule of assessments 468

8.1
Informed consent procedure

469
The Principal Investigator or designated sub-investigator will be responsible for ensuring that a patient is fully 470 consented following adequate explanation of the aims, methods, anticipated benefits and potential hazards of 471 the study. Patients will be advised that any data collected will be held and used in accordance with the Data

472
Protection Act 1998. Patients will be given at least 24 hours after receiving the patient information sheet (PIS)

473
to consider taking part. The PI or designee will record in the medical notes date the patient information sheet 474 was given to the patient. The PI or designee will explain that patients are under no obligation to enter the trial 475 and that they can withdraw at any time, without giving a reason. No clinical trial procedures will be conducted 476 prior to taking consent from the participant and consent will not denote enrolment into the trial. A copy of the 477 signed informed consent form will be given to the patient. The original signed form will be retained at the 478 study site and a copy placed in the medical notes. If new safety information results in significant changes in 479 the risk/benefit assessment, the patient information sheet will be reviewed and updated if necessary and 480 subjects will be re-consented as appropriate.

482
A patient identification number (PIN) will be generated by registering the patient on the MACRO eCRF system 483 (InferMed Macro), after consent has been signed. This unique PIN will be recorded on all source data 484 worksheets and used to identify the patient throughout the study. Randomisation will be via a bespoke web 485 based randomisation system hosted at the KCTU. Authorised site staff will be allocated a username and 486 password for the randomisation system by the Trial Manager. An authorised staff member who will typically be 487 the PI or designee will log into the randomisation system (www.ctu.co.uk and click 'randomisation -advanced' Once a patient is randomised, the system will automatically generate emails to key staff within the study.

491
Unmasked e-mails sent to site pharmacies will alert them to a patient's treatment arm: ranibizumab, 492 aflibercept or bevacizumab. The pharmacy department will use the email to cross check the trial prescription 493 to ensure that the correct medication is being dispensed for the correct patient. Additional masked emails will 494 be generated from the randomisation system to key trial site staff, and unmasked e-mails to the emergency 495 unmasking service (eSMS Global) and unmasked trial management staff.

501
The study drug the patient will receive will be transferred in a masking bag to the dedicated injection room.

502
Prior to leaving the Pharmacy a unique seal will be attached to the bag. The non-transparent masking bag, 503 designed to securely and safely transport medication, will have a safe zipped compartment containing a pre-

504
printed form detailing the participants unique PIN, date of birth, date drug dispensed and injection batch 505 number. Prior to the participant entering the injection room, the unmasked experienced injector will break the 506 seal, take the drug out of the masking bag. In the case of bevacizumab, this will be in a prefilled syringe but 507 ranibizumab and aflibercept are currently provided in a vial and will be drawn into a syringe, by the unmasked 508 injector. The syringe will then be placed on the injection trolley, out of view of the patient, who will then be 509 invited into the room, to lie on the injection bed and the injection administered to the patient. Ranibizumab and

510
Aflibercept may be provided in a unique prefilled syringe by the manufacturer during the course of the trial and 511 vials cease to be available. In this situation, the unmasked injector will take care not to allow the subject sight 512 of the syringe either before or after the injection has been given. This will be done by performing the injection 513 with the patient lying down and the injection given via the pars plana in any quadrant of the eye with the 514 syringe being brought to and taken away from the injection site from the patients inferotemporal field of vision 515 so that it is not passed across their line of sight. The unmasked injector will sign the source notes to the effect 516 that the treatment in the masked bag has been administered to the patient, without specifying the treatment, 517 and will also sign the pre-printed form within the masking bag to the effect that the drug detailed on the form 518 has been given to specified participant. The empty drug syringe with needle and vial will be disposed of in the 519 injection room. The masking bag and completed pre-printed will be returned to pharmacy, for drug 520 accountability purposes (See Section 9.8). The drug outer packaging will be disposed of in the injection room.

522
The clinical assessment team including the site PI, optometrist i.e. assessor of the primary outcome, site trial 523 co-ordinator, the clinical investigator, clinical assessment study nurse and ophthalmic technician will therefore 524 remain masked throughout the study as there will be no record of the subjects' treatment arm in the source to ensure no repetition occurs and undertake not to convey this information either to the participant or others 530 involved in the project. Certain secondary outcomes e.g. interpretation of fluorescein angiography will occur at 531 the remote NetwORC UK Reading Centre where the assessors will be masked as to the treatment allocation.

532
These masking procedures will avoid both performance and detection bias. We will describe the 533 completeness of outcome data for each outcome, including any unmasking in error, reasons for attrition and 534 exclusions from the analysis.

544
All participants will be provided with an emergency code break card providing details of the 24 hour 545 emergency code break service undertaken by Guy's Medical Toxicology Unit (eSMS Global). If a request for 546 code break is received from a physician (e.g. the patient's general practitioner) outside the research team, 547 eSMS will attempt to contact the research team to verify the request before the code is broken.

549
If the code is broken, because it is deemed necessary for the immediate management of the participant, 550 details including patient study number, the date code break was performed, the person who broke the code,

551
and reason for code break shall be recorded by the emergency code break service and retained. The Trial

552
Manager will be informed of the unmasking event. If clinically indicated, the participant will be withdrawn from 553 study medication.

555
Accidental unmasking will be dealt with on a case by case basis, if and when they arise. The patient's data 556 should continue to be collected according to the study assessment schedule, even in the event of unmasking 557 or withdrawal from study medication, unless the patient refuses.

566
The baseline visit and patient randomisation must be performed no later than 10 days after the screening visit.

583
A within window flexibility to complete the assessments and treatment is permitted.

605
The central sub-field thickness and total macular volume in both eyes will be recorded from the SD-OCT      treatment follow up to be accommodated. All data from the study milestone visits will be entered into the 662 eCRF. For regular treatment visits only BCVA, OCT CST, whether an injection was given, and if no injection 663 was given the reason it was not given will be entered into the eCRF. The latter will be entered onto a single

719
If Retreatment Criteria are met at an 8 weekly or unscheduled visit, then 4 weekly visits will be resumed.

775
The criteria for restarting therapy after 'Non-responder Treatment Suspension'' are NOT incorporated 776 into the week 24 to 96 Retreatment Algorithm,

809
Participants have the right to withdraw from the study at any time and for any reason, without providing a 810 reason. The investigator also has the right to withdraw participants from the study in the event of inter-current 811 illness, AEs, SAEs, SUSARs, protocol violations or other reasons. Should a participant decide to withdraw 812 from the study, they will be asked to volunteer a reason for withdrawal but are at liberty not to do so.

813
Should a participant withdraw from study drug only, efforts will be made to continue to obtain follow-up data,

814
with the permission of the participant.

815
Subjects who withdraw from treatment early will be encouraged to return to the study site for an early 816 termination assessment. All patients who withdraw from treatment will be encouraged to attend the milestone

882
A study medication dispensing and return log will be maintained by the centre pharmacies. Administration 883 records from the trial centres will be retained by the pharmacy department and monitored by the Trial

884
Manager, to ensure that accurate CRF data are recorded. The randomization system will be linked to the IMP 885 supply. The Hospital Site Pharmacy will also be responsible for appropriate storage, dispensing, disposal and  Study medication will be prescribed by an authorised study physician according to the protocol, using a trial 892 specific prescription. Medication will be dispensed according to local pharmacy practice. Documentation of 893 prescribing, dispensing and return of the pre-printed form signed by the unmasked investigator will be kept in 894 the pharmacy file and reconciled with the investigator site file at end of study. A study specific prescription 895 must be submitted to pharmacy as early as possible after randomisation. The pharmacy will have received an 896 email from the randomisation service at the time of randomisation, which must be printed and filed with the 897 dispensing records and which will be referred to by the dispensing pharmacist to confirm whether the 898 participant is correctly randomised to receive the drug.

899
Masking of treatment allocation: see Section 8.3. preprinted form which will detail the participant's unique pin number, date of birth, date drug dispensed and 904 injection batch number. After performing the intravitreal injection, the unmasked injector will sign this form to 905 confirm the drug has been given to the allocated patient and return it in the masking bag to the pharmacy. All 906 used drug vials and syringes will be disposed of in the injection room and not returned to pharmacy.

907
Pharmacy departments in each centre will maintain a study medication dispensing log, including date 908 dispensed, batch number, expiry date and return log. The latter will be compiled from the form signed by the 909 unmasked injector. In addition, the study specific prescriptions will be maintained in the pharmacy file for audit 910 purposes. Any administration errors will be reported to the CI and trial statistician. In the event that an 911 injection is not given as scheduled, reasons must be documented in the patients' notes and CRF. The study 912 monitor will check the pharmacy records against the eCRF. All records will be reconciled at the end of the three drugs is not recommended to be less than 4 weeks and this will be adhered to throughout this study.

961
Subjects exiting the trial who continue to require therapy for the condition will be followed up within their local 962 NHS Trust Hospital clinical service and receive standard care for the condition which is likely to be either 963 Ranibizumab (0.5mg/50ul) or Aflibercept (2.0mg/50ul).

967
NHS practice and will be recorded as a concomitant medication for the study. Dose, duration and frequency will be in accordance with local practice. Normal NHS prescribing practice in both primary and secondary 969 care will apply with no special arrangements. then be seen at two weekly intervals until sufficient PRP is applied. The participant will also continue to attend 995 all study visits until the end of study. PRP to the fellow eye will be recorded as a concomitant procedure.

996
Other planned procedures may be required in the study and non-study eye e.g. surgical repair of ptosis. The 997 ptosis will be reported as an AE.

998
Please refer to section 10.2.1 Planned hospitalisation, non-emergency procedures and SAE reporting

1000
If macular oedema due to any retinal disease is present in the non-study eye, it is advocated that macular 1001 laser therapy be given as the first line therapy if appropriate. However, the participant can be treated with 1002 intravitreal anti-VEGF therapy or steroid therapy as per discretion of the treating physician. New onset 1003 macular oedema will be recorded as an adverse event. Laser therapy or intravitreal injection to the non-study 1004 eye will be recorded as a concomitant procedure.

1011
Other injection related procedures e.g. retinal tear, retinal detachment and lens damage will be recorded as a

1012
Serious Adverse Event and will be reported as detailed in section 10.2.

1014
Ischaemic CRVO, NVA, NVI, NVG, NVE and NVD in the study eye will be recorded as adverse events.

1015
Diagnosis and management of these complications of CRVO in the study is based on investigator discretion 1016 and local practice. Laser therapy will form the mainstay of therapy and will be recorded as a concomitant 1017 procedure. Anti-VEGF agents in the study eye for NVG should be avoided.  Adverse Reaction (AR) Any untoward and unintended response in a subject to a study intervention which is related to any dose administered to that subject.

1047
All Principal Investigators will be informed of all SAEs assessed as fulfilling criteria as a SUSAR (ie, possibly,

1048
probably or definitely related to either study intervention and unexpected as per the SPC or the protocol.

1072
Whilst the Principal Investigator is responsible for resolving any queries that arise during the completion of the 1073 AE log and eCRF, queries can also be directed to the Chief Investigator and Trial Manager.

1074
The following categories will be used to define the causality of the adverse event: 1075 Not related There is no evidence of any causal relationship.
Not Assessable Unable to assess on information available.

1076
Category Definition

Expected
An adverse event that is classed in nature as serious and which is consistent with the information about the study intervention listed in the SPC clearly defined in this protocol (aflibercept, ranibizumab and bevacizumab).

Unexpected
An adverse event that is classed in nature as serious and which is not consistent with the information about either study intervention in the SPC or clearly defined in this protocol (aflibercept, ranibizumab and bevacizumab).
The reference document to be used to assess expectedness against the study interventions and comparator 1078 is the SPC and the protocol. The protocol will be used as the reference document to assess disease related 1079 and/or procedural expected events and will take preference over the SPC.

1080
Expected adverse events may be classified into ocular (study eye and non-study eye will be reported 1081 separately) and non-ocular. Ocular adverse events may be due to disease progression, injection procedure 1082 related, study intervention related or any other related event that the investigator deems clinically significant.

1111
The Chief Investigator/delegate will prepare and submit an annual progress report (APR) to the REC within 30 1112 days of the anniversary date on which the favourable opinion was given, and annually until end of trial has 1113 been declared.

1115
In the event a female participant becomes pregnant, this should be reported to KCTU via fax or email (Fax: 1116 020 7848 5229, email: ctu@kcl.ac.uk) using a pregnancy form as soon as the Investigator becomes aware of 1117 it. The pregnancy will be monitored to determine outcome. Any information related to the pregnancy 1118 following the initial report should be reported on a follow up pregnancy form.

1119
Further treatment with any anti-VEGF therapy should be stopped on becoming aware of pregnancy but 1120 collection of outcome data should continue to the end of the study provided the participant is willing to do so.

1121
Participants who wish to withdraw should be withdrawn as described in section 8.15 Withdrawal of Subjects

1122
This process should also be followed if a female participant becomes pregnant within 6 months after their last 1123 trial injection regardless of whether they are still in follow up or not.

1125
Any SAEs experienced during the pregnancy must be reported on an SAE form as described in 10.2

1126
Procedures for recording and reporting Serious Adverse Events.

1128
In the event that a higher does is given to a participant, the site should notify the Chief Investigator/delegate.

1129
Follow up action will be decided on a case by case basis. Participants do not need to be withdrawn from the 1130 study and should remain on treatment and in follow up. Sites will be instructed to complete the adverse event 1131 form if such an event occurs.

1133
Any urgent safety measures taken should be immediately reported to the Chief Investigator or her assignee.

1134
Any queries that arise should be promptly resolved by the site to ensure reporting timelines are adhered to.

1135
The Chief Investigator /Sponsor shall immediately and in any event no later than 3 days from the date the  Any ongoing AEs during the patient's participation in the study will be followed up until resolution. AEs, ARs,

1140
SAEs, SARs and SUSARs will be reportable for up to 30 days after the last intervention session.

1142
A "serious breach" is a breach which is likely to effect to a significant degree -

1163
Written informed consent will be obtained prior to screening and any other study specific procedures are 1164 performed.

1166
SAE data will be collected on paper SAE report forms and emailed or faxed to the KCTU. Summary details 1167 of SAEs will be transcribed to the adverse event section of the eCRF. For all other data collected, source 1168 data worksheets will be used for each patient and data will be entered onto the eCRF database. Source

1175
It will be the responsibility of the Principal Investigator and his team to ensure the accuracy of all data entered 1176 in the worksheets and the eCRF are in accordance with Good Clinical Practice. The delegation log will identify all those personnel with responsibilities for data collection and handling, including those who have 1178 access to the trial database. The Principal Investigator will be responsible for ensuring that source data 1179 worksheets are filed in a suitably secure location to ensure source data verification can be undertaken 1180 throughout the study.

1182
All study data and site files will be kept at site in a secure location with restricted access.

1183
The study will employ an eCRF created using the InferMed MACRO database system. Data will be 1184 managed via this system.

1186
Database Website Address: 1187 Go to www.ctu.co.uk and click the link to MACRO EDC V4 towards the top of the screen.

1189
The eCRF will be created in collaboration with the trial statistician and the CI and maintained by the KCTU.

1196
The study incorporates a range of data management quality assurance functions. The eCRF system will 1197 contain a range of validations defined by the trial team that will alert sites to inconsistencies in the 1198 data being entered which will be monitored by the Trial Manager. The Trial Manager w i l l provide study 1199 training, ongoing study support and w i l l conduct regular monitoring visits at each centre, checking 1200 source data for transcription errors. Any necessary alterations to entered data will be date and time 1201 stamped within the eCRF A detailed monitoring plan and data management plan will be developed and 1202 updated as the trial progresses, detailing the quality control and quality assurance checks to be undertaken.

1205
Prior to database lock, the Trial Manager will review any outstanding warnings on the eCRF and resolve or 1206 close these as appropriate before database lock. Local study personnel should resolve any queries that 1207 arise promptly. Once all queries have been resolved no further changes will be made to the database 1208 unless specifically requested by the Study Office in response to the statistician's data checks. The study PI 1209 will review all the data for each participant and provide electronic sign-off to verify that all the data are 1210 complete and correct. At this point, all data will be formally locked for analysis. At the end of the trial, each 1211 centre will be supplied on a CD-ROM containing the eCRF data for their centre. This will be filed locally for any future regulatory inspection or internal audit.

Record keeping and archiving 1214
The Chief Investigator will be custodian for the data generated from the study. The Chief Investigator will be 1215 responsible for archiving the original data. All data will be archived for at least 5 years from the end of the trial 1216 and will be archived in accordance with Sponsor and regulatory requirements. Principal Investigators will be 1217 responsible for securely archiving local data generated, essential documents and source data in accordance 1218 with local requirements, but for at least 5 years from the end of the study. Investigators should provide 1219 archiving details to the Chief Investigator/delegate and will be instructed that authorisation from the Chief 1220 Investigator should be obtained before study data or study documentation is destroyed. Essential documents 1221 held by the KCTU will be returned to the Chief Investigator for archiving by the Sponsor organisation. eCRF 1222 data will also be exported and provided to the Chief Investigator for archiving.

Statistical Considerations 1224
The trial statisticians will be responsible for all statistical aspects of the trial from design through to analysis 1225 and dissemination.
To determine differences between arms in mean change in best corrected visual acuity at 100 weeks

1280
The primary method of analysis will include all available refracted data of the primary outcome up to and 1281 including 100 weeks, including data from the 15% of patients we anticipate could be missing the 100 weeks 1282 primary outcome endpoint, thereby giving flexibility to provide increased power or a higher dropout allowance 1283 for the stated power without having to amend the sample size in this event.

1285
Approximately 40 sites will be opened and recruit into this study. It is anticipated that 459 participants will be 1286 recruited over an 18 months period. The DMEC will receive recruitment updates and based on committee 1287 recommendations new sites will be added as needed.    Secondary outcome analyses will be on an ITT basis only, and assessed with tests at the two-sided 5% level 1317 of significance. Continuous outcomes will be compared between arms using a linear mixed effects model, as 1318 specified for the primary outcome ITT analysis, incorporating prior measurements of the outcome over time.

1319
Binary outcomes will be compared between arms using logistic regression. Continuous and binary outcomes 1320 will be reported as adjusted differences in means or odds ratios respectively. All tests will be two-sided at the 1321 5% significance level and interpreted cautiously with a focus on interpreting effect sizes with 95% confidence 1322 intervals. Safety outcomes will be reported as unadjusted patient proportions and rates within and between 1323 arms with 95% confidence intervals using exact methods where appropriate.

1325
Sensitivity to the missing at random assumption made in the primary outcome analysis will be undertaken to 1326 assess sensitivity to the handling of missing 100-week data, and to the use of concomitant treatments, and 1327 will be detailed in the statistical analysis plan.

1328
If non-inferiority is concluded for either of the investigational treatments, then superiority will be assessed. If 1329 non-inferiority is concluded for both the investigational treatments then there will be a formal test of superiority 1330 to compare these two investigational treatments.

1332
Only one eye can be randomised into the trial. In 95% of cases, one eye will be affected by CRVO and will be 1333 the 'worst seeing eye' and will therefore be randomised. On rare occasions, some patients may have bilateral

1334
CRVO that meet the eligibility criteria. In these cases the worst-seeing eye will be randomised unless the 1335 patient opts for the 'better seeing eye' to be randomised.

1336
Randomisation will be via a bespoke web based randomisation system hosted at the KCTU. 459 adult 1337 patients with MO due to CRVO will be randomised 1

1351
The TSC is the Committee, responsible for monitoring the overall integrity, conduct and safety of the trial. It  Ethics Committee concerned. All data reviewed by the DMEC will determine safety issues. All serious adverse 1370 reactions will be reported to the KCTU within 24 hours of learning of their occurrence.

Trial Management Group (TMG)
1372 The TMG will be responsible for monitoring the delivery of the trial on a day to day basis and will be supported

1382
Site monitoring will include:


Reviewing all consent forms within the site file and medical notes.

1384
 Source data verifying serious adverse events against medical records and a proportion of the 1385 primary outcome measure.

1386
 Checking essential documents in the investigator site file and study files.

1388
Central reviews will include:

1394
The investigator(s)/ institution(s) will also permit trial-related monitoring, audits, REC review, and regulatory 1395 inspection(s), providing direct access to source data/documents. Trial participants are informed of this during 1396 the informed consent discussion. Participants will consent to provide access to their medical notes.

1402
The precise risks and benefits of participating in the clinical study will be outlined in patient information sheets, 1403 formulated with service user involvement.
Both ranibizumab and aflibercept are licensed for use in this indication. Bevacizumab is not licensed for 1405 intravitreal use but it is the most widely used anti-VEGF agent worldwide and has been shown to be non-1406 inferior to ranibizumab in nvAMD.

1407
Any breach of confidentiality will be minimised by adherence to the UK Data Protection Act 1998 and the 1408 approved protocol

1421
Within 90 days after the end of the trial, the Chief Investigator/Sponsor will ensure that the main REC and the 1422 MHRA are notified that the trial has finished. If the trial is terminated prematurely, those reports will be made 1423 within 15 days after the end of the trial.

1424
The Chief Investigator will supply the Sponsor with a summary report of the clinical trial, which will then be 1425 submitted to the MHRA and main REC within 1 year after the end of the trial.

Finance 1430
The study is funded through the NIHR HTA CET Programme.

18
Insurance 1432 The participating NHS Trusts have liability for clinical negligence that harms individuals towards whom they 1433 have a duty of care. NHS indemnity covers NHS staff and medical academic staff with honorary contracts 1434 conducting the trial. There are no arrangements for non-negligent compensation.

19
Publication policy 1436 The data will be the property of and publication will be the responsibility of the Chief Investigator. It 1437 is planned to publish this study in peer review journals and to present data at national and international 1438 meetings. Results of the study will also be reported to the Sponsor and Funder, and will be available on their 1439 web site. All manuscripts, abstracts or other modes of presentation will be reviewed by the Trial Steering

Record
The trial statistician will write the first version of the plan. After revision by the supervisor the plan will be filed as version number 2. The plan will then be discussed with the Principal Investigator for further input and filed as version number 3. The plan will then be sent to the DMC and TSC for final approvals and saved as version 4 and 5 of the plan, respectively.

Purpose and scope of the statistical analysis plan
The purpose of this Statistical Analysis Plan is to set out the study objectives and hypotheses, and the analytical approaches and procedures necessary to address these for the main trial paper and to provide guidance for further research reported in other papers, promoting consistent approaches and methods.
As there can typically be more than one analytical approach to address a hypothesis, there is the potential for different results to be produced from using alternative approaches, alternative methods, alternative outcome definitions and the alternative data that may be involved. These differences can be influential, for example, when results are of borderline statistical significance.
Therefore, this Plan records those decisions that can be made about study hypotheses, outcome definitions and statistical procedures, along with their basis and the appropriateness of the assumptions required for their use, in advance of the main trial analysis, while any access to unmasked follow-up data and to trial arm is prevented. Changes within subsequent versions of the Plan prior to analysis will be dated, with the basis for the changes reasoned, and recorded within the plan.
Other analysis decisions may need to be made later, based on viewing the observed distribution of the data. Where possible these decisions will be made prior to access to trial arm, or where necessary from control arm data alone. The main place for this will be the "pre-analysis review" phase, blinded to treatment arm (ICH E9 (1) meeting with follow-up data, indications, such as on the need for data transformation, will be made at a point when baseline data only is available without access to arm.
Decisions will be supported by reasoning and justification, and these will be appended to the Statistical Analysis Plan, and dated, to provide a record of any post-analysis decisions and their basis.
It is not intended that the strategy set out in the plan should prohibit sensible practices. However, the principles established in the plan will be followed as closely as possible when analysing and reporting the trial.

Overview of the condition and treatment 2.1 Description of the condition and its importance/scale
Retinal Vein Occlusion (RVO) is a blockage of the small retinal veins by a blood clot. When the blood cannot drain away from the retina, there is an accumulation of pressure in the blood vessels resulting in leakage of fluid and blood causing macular oedema and ischemia. This condition can affect the central retinal vein (CRVO) (formed by the union of the four retinal veins (one retinal vein drains each quarter of the eye)) or a major branch retinal vein (BRVO), where blockage occurs somewhere along the course of one of the four retinal veins.CRVO is characterised by retinal haemorrhages, venous dilatation and tortuosity in all four quadrants of the retina and is typically more severe than BRVO (8,9).
Approximately 6,860 people develop CRVO every year in England and Wales of whom 5,150 are potentially eligible for treatment. Once established, the visual impairment due to CRVO is typically profound with no tendency to improve spontaneously. Without intervention permanently impaired visual loss is likely to occur. In this study the focus will be on CRVO.

Description of the standard treatment (or placebo or current care)
Until 2011 no treatment was available to improve vision in people with CRVO. In 2011, NICE recommended the NHS an implant called "Ozurdex", which although improving vision, could cause cataracts and glaucoma with repeated use, was difficult to administer and had only moderate uptake in the UK. In the meantime another treatment, more effective with fewer side effects, Ranibizumab (Lucentis, Novartis, & Genentech), was approved by the FDA and EMA for macular oedema due to CRVO. Ranibizumab is a humanized, affinitymatured VEGF antibody fragment that binds to and neutralizes all isoforms of VEGF-A and their biologically active degradation products and it was the first anti-VEGF therapy to demonstrate improved visual outcomes in patients with neovascular age related vascular degeneration. It is EMA licensed for use in wet age related macular degeneration, diabetic macular oedema and retinal vein occlusion and NICE recommended for all three.

Description of the investigational treatments
This study aims to determine if the two anti-VEFG agents Bevacizumab or Afilbercept are as effective as Ranibizumab in reducing visual loss from MO due to CRVO, whether they have an equivalent side effect profile and whether either could be considered as a recommended NHS treatment based on non-inferior clinical effectiveness and superior cost-effectiveness.
Aflibercept or VEGF Trap-Eye (Eylea, Bayer/Regeneron), is a fusion protein of the key domains of VEGF receptors 1 and 2 and human IgGFc that blocks all VEGF-A isoforms and placental growth factor. Like ranibizumab, it is EMA licensed for nvAMD, DMO and RVO. It is FDA approved for CRVO and NICE has recommended this drug for MO due to CRVO (TA 305).
Bevacizumab (Avastin, Genetech/Roche), is a monoclonal antibody that inhibits vascular endothelial growth factor (VEGF), a mediator in the pathogenesis of common and disabling eye disorders including neovascular age related macular degeneration (nvAMD), diabetic macular oedema (DMO) and retinal vein occlusion (RVO). Bevacizumab is EMA licensed for the treatment of cancer but not for use in the eye. There is limited evidence regarding its use in central retinal vein occlusion (CRVO) with macular oedema (MO). To date, bevacizumab has been found to be non-inferior to ranibizumab for all visual acuity primary and secondary endpoints in nvAMD in the IVAN and CATT studies (10, 11).

Description of the motivation for the study / need to investigate the new treatment
This project will compare the relative clinical and cost effectiveness of the anti-VEGF agents bevacizumab, aflibercept, and ranibizumab in MO due to CRVO over 100 weeks which is of critical importance to the NHS in the next 10 years. The NICE Final Appraisal Document for ranibizumab in RVO has recommended that further research is required comparing ranibizumab and bevacizumab (12) and there are no comparisons to date of bevacizumab and ranibizumab with aflibercept. This trial will be the first well-powered Phase III trial exploring the relative effectiveness of these drugs and aflibercept in the management of MO in CRVO. This study will inform us of the potential use of the most clinically effective and cost-effective drug for this condition in the NHS by providing a better understanding of the economic and societal impact of RVO in the long-term and help decision makers evaluate and compare these medical interventions over the duration of the natural history of the condition.

Target Population
The target population, to which inferences from the end of this trial are intended to generalise, is the population of adult patients with MO due to CRVO.

Trial Population
The trial population, from which the study sample is drawn, is further defined to be adults aged 18 year or over, of less than 12 months duration who attend the 40 ophthalmology centres in the UK with expertise in retinal disorders and a proven track record in effective research. Only one eye per patient will be included in the trial. In subjects with both eyes meeting the eligibility criteria, then the 'worst seeing eye' will be enrolled unless the patients preference is for the best seeing eye (see section 7.3).

Intention To Treat (ITT)
The achieved trial sample comprises those patients who consent to participate and are actually randomised into this trial. These patients are the study subjects.
This randomised trial sample is also the trial Intention To Treat (ITT) population. The intention-to-treat principle states that every subject will be analysed according to the treatment group to which they were randomised. In this trial, subjects' data will be analysed according to the Intention to Treat Strategy (13), under which at least one analysis is recommended to be based on the ITT population.
The trial ITT population comprises all randomised participants, regardless of eligibility (inclusion/exclusion) error, post-randomisation withdrawal, and whether the correct study treatments were received, or other interventions received.

Per Protocol (PP)
Definition A per protocol set of subjects will also be included. These will be defined as the subset of the found to be eligible at entry and who had minimal sufficient exposure to the treatment regimen, defined as 4 treatments correctly assessed and received during the first 6 visits up to week 20. For each of the first four visits, a correct treatment is defined as receiving the injection. For the 5 th and 6 th visits, a correctly assessed and received treatment is defined to be the receipt of an injection where this is indicated to be required by the retreatment criteria or the non-receipt of an injection where this is indicated by the retreatment criteria..

Rationale
The main reason for having a per protocol set comes from the fact that this is a noninferiority trial and so the use of the full analysis set is generally not conservative (ICH E9 section 5.2.3 (1)). As Lesaffre 2008 (7) states, "dropouts and a poor conduct of the study might direct the results of the two arms towards each other". Although this can be interpreted as an indication that the per protocol analysis is the conservative choice for noninferiority studies Garrett AD 2003 (6) state that "The perceived conservative nature of the PP population appears to be much more a reflection of reduced patient numbers than the presence of bias, while bias can be in either direction depending on the pattern of violations". Moreover, with two active treatments it may be more likely that any bias affecting both treatments would be reduced in comparison to a placebo-controlled trial.

Prominence
Non-inferiority will only be declared if both ITT and the PP analysis are supportive of a noninferiority conclusion. This is supported by the Committee on Proprietary Medical Products Points-to-Consider (5) and several other papers (7,14).
The requirement to declare noninferiority in both the ITT and the PP analyses promotes the adherence to treatment protocol and the minimisation of exclusions, maintaining power.  13. Males who do not agree to an effective form of contraception for the duration of the study and for 6 months after their last injection for the trial. 14. Hypersensitivity to the active ingredients aflibercept, bevacizumab or ranibizumab or any of the excipients of these drugs. 15. Hypersensitivity to Chinese Hamster Ovary (CHO) cell products or other recombinant human or humanised antibodies. 16. A condition that, in the opinion of the investigator, would preclude participation in the study. 17. Participation in an investigational trial involving an investigational medicinal product within 90 days of randomisation.

Objectives, principal research question and associated hypotheses 4.1 Principal Trial objective
The objective of the LEAVO trial is to determine if Bevacizumab or Afilbercept are as effective as Ranibizumab in reducing visual loss from MO due to CRVO, whether they have an equivalent side effect profile and whether either could be considered as a recommended NHS treatment based on non-inferior clinical effectiveness and superior cost-effectiveness.

Principal Research Question
The principle research question is as follows: Is visual acuity following Aflibercept or Bevacizumab non-inferior to Ranibizumab in eyes with MO due to CRVO at 100 weeks?

Hypotheses
The hypotheses refer to the populations of relevant patients rather than study subjects.
The Working hypothesis: The so-called "working hypothesis" is the hypothesis which motivates the trial, which the trial results may or may not support. It is that the change in best corrected visual acuity is non-inferior in patients treated with either Aflibercept or Bevacizumab compared to patients treated with Ranibizumab.

Trial design
This is a two year multicentre (approximately 40 centres), double-blind pragmatic individually randomised controlled trial that will test the non-inferiority visual acuity from treatment with Bevacizumab and Aflibercept to Ranibizumab at 100 weeks in 459 adult participants with MO due to CRVO of less than 12 months duration.

Treatment arms
The trial is randomised with three arms and with equal allocation of participants in a 1:1:1 ratio to the three arms.
Arm B: Treatment: An intravitreal injection of Bevacizumab (Avastin, Roche) (1.25mg in 50ul) will be administered at baseline, 4, 8 and 12 weeks. After this the retreatment criteria is ascertained (see Protocol section 8.14).
Different labels can be used for the control arm, such as placebo, comparator, standard care, or control.

Type of RCT
This is a phase III, parallel groups' trial.

Frequency and duration of follow-up
Participants in all 3 study arms will be seen at weeks 0, 4, 8, 12, 16, 20 and 24. After this participants will potentially be seen every 4 weeks until week 96 if retreatment criteria are met. If retreatment criteria are not met at three successive visits from week 24 onwards the visit interval is increased to 8 weekly until week 96. They will also be finally seen at 100weeks.

Primary outcome
The primary outcome is Best Corrected Visual Acuity (BCVA) in the study eye measured in ETDRS letter score at 4 metres at 100 weeks. Measurements of BCVA at milestone visits are included in the analysis of the primary outcome. Any BCVA measurement will be excluded from the analysis if it is both more than 3 standard deviations below the mean at that timepoint (including all measurements) AND taken within 3 months of occurrence of a vitreous haemorrhage or another cause unrelated to maculopathy secondary to CRVO (such as neovascular glaucoma). Further colour fundus photographs and fluorescein angiography may be performed as per investigator discretion. Colour fundus photographs should be done if a patient converts from non-ischaemic to ischaemic CRVO. 3 To include review of screening assessment test results and confirmation of eligibility. 4 To be completed by participant, masked investigator, site optometrists. 5 To be performed (as required) if unscheduled visit is a milestone visit.

Participant duration in the study
Each study subject will participate in the trial from the day that the they give informed consent to their last final visit at 100 weeks.

Final assessment
The final study assessment is when the last study subject achieves their 100 weeks assessment.

Determination of the primary outcome effect size
Bevacizumab and aflibercept are defined to be substantially inferior to ranibizumab, if in each case, the mean of the primary outcome (change in best corrected ETDRS visual acuity letter score) is worse by a margin of five letters.
The two null hypotheses, that bevacizumab is substantially inferior to ranibizumab, and that aflibercept is substantially inferior to ranibizumab, will each be rejected if the estimated 95% confidence interval for the difference in treatment means lies wholly above the five letter margin in each case.
The choice of a five-letter margin is 32% higher than the available estimated 12-month placebo-controlled effect of 6.6 letters (15) for ranibizumab, the standard (comparator) treatment for LEAVO. This margin choice is therefore consistent with maintaining assay sensitivity sufficiently to be able to declare non-inferiority.

Determination of the primary outcome variability
For a similar trial on CRVO (15) the standard deviation reported in Ranibizumab arms at 12 months was 14.3. In the absence of 24-month data we have assumed a comparable SD of 14.3 at 24 months.

Clustering of outcomes from eyes within subjects effects
Only one eye per subject can be selected for the study. In 95% of cases, one eye will be affected by CRVO. As explained in the protocol, bilateral RVO is rare, but if it happens and both eyes are eligible, the eye included is the 'worst seeing eye'. However, participants will be given the choice if both eyes are found to be eligible. All observations are in this way able to be assumed to be independent in the sample size calculation and statistical analysis.

Power to detect effects
There is 80% power to detect non-inferiority using a two-sided 95% confidence interval from an analysis of covariance test with adjustment for baseline visual acuity and randomisation stratifiers.

Determination of the sample size based on the primary outcome
The sample size was set to be 459 participants, 153 per arm (1 eye per subject). The target of 390 subjects followed up with primary outcome required in the analysis involves a 15% allowance for dropout and protocol exclusions.
Sample size calculations were performed using nQuery Advisor 4.0 software.

Detectable effects sizes expressed in general standardised form
For a continuous secondary outcome, with 153 subjects per arm followed up we can detect effects of size 0.45 Standard Deviation's difference between means with 80% power using a two-sided t-test at the 5% significance level. For binary outcomes, we have at least 90% power to detect a difference in proportions of 0.2 using a chi-squared test at the 5% significance level.

Arms
Each participant will be equally randomised to one of three arms: bevacizumab, aflibercept or ranibizumab.

Relative timing of randomisation
Randomisation will be via a bespoke web based randomisation system hosted at the King's CTU on a secure server. Once a participant enters the study and their data is entered into the
These are based on the fact that visual gain in the low vision group may be higher than that achieved by the high vision group and this effect may be different between arms. Patients with ischaemic CRVO may not have similar visual acuity gains to those with no ischaemia and this effect may be different between treatment arms. The shorter the duration of disease, the better the visual acuity outcomes but this may vary between treatment arms.

Blinding
The trial will be double masked. Study participants, clinicians and members of the research team who will undertake key measurements (visual acuity, morphology) will be masked to group allocation. The clinician administering the drug injected into the vitreous will not be masked. This will ensure that the study has a high level of both treatment integrity (delivery of the treatment as intended) and treatment differentiation (treatment conditions differed from one another in the intended manner).
The trial statistician will have access to the accumulating outcome data that is required for reporting to the DMC.Both the trial statisticians will attend both the open and closed DMC meetings.

Data decisions made
The data manager will make limited decisions about data variables and values so that issues such as missing data can be comprehensively handled by the trial statistician. Decisions which impact on the analysis will be recorded in an appendix of this statistical analysis plan.

Outcomes requiring derivation
List of outcomes with source of derivation code: 1) VFQ-25 (16): a validated tool for vision related quality of life. It consists of a base set of 25 vision targeted questions representing 11 vision-related sub-scales, plus an additional single-item general health rating question. The overall composite score is computed as the simple average of the vision-targeted sub-scale scores, excluding the general health rating question. The overall score can range from 0 (worst possible score) to 100 (best).
EQ-5D (17)(18)(19) with and without vision bolt-on: The EQ-5D is a generic instrument for describing and valuing health. It is based on a descriptive system that defines health in terms of 5 dimensions (Mobility, Self-care, Usual activities, Pain/Discomfort, Anxiety/Depression. Each dimension has 5 response categories (EQ-5D-5L) corresponding to eg. "no problems","slight problems", "moderate problems", "severe problems", and "unable to/extreme problems". A preferencebased score ranges from states worse than dead (<0) to 1 (full health), anchoring dead at 0. In addition, the EQ-5D includes a visual analogue scale (EQ-VAS), which records the respondent's self-rated health on a vertical scale where the endpoints are labelled 'Best imaginable health state' (marked as 100) and 'Worst imaginable health state' (marked as 0). The EQ-5D with bolt-on is similar to the EQ-5D-5L but another dimension was added (vision) in order to overcome perceived inadequacies in a particular population. The corresponding scoring system for the EQ-5D vision 'bolt-on' on has not been finalised yet. Further details are developed in the HEDMAP.

Procedure for deriving variables
If there is existing syntax code to derive a variable within the King's Clinical Trials Unit then this will be used. Otherwise new code will be developed by the trial statistician and verified by the senior statistician.

Missing items in scale and subscales
The number (%) of patients with complete data for each scale will be reported. If scales provide missing value guidance then this will be used.

Use of data transformation
It is not anticipated that any continuous outcomes will need to be considered for transformation, because the sample size is reasonably large for group comparisons in the main trial analyses. Assumptions of normality and constant variance required by the models will be examined using residual and other diagnostic plots. If it is relevant, and necessary, where sample size is reduced, a log transformation will be considered, because this retains a sensible interpretation for inferences; in relative terms between arms. If an absolute interpretation is needed, then data transformation may not be undertaken, but a nonparametric Bootstrap method for obtaining confidence intervals may be considered (20).

Defining Outliers
Outliers are observations that have extreme values relative to other observations observed under the same conditions. An outlier will be defined here as a data-point being at least four standard deviations from the mean of its distribution of values observed across other patients. This definition will apply to the transformed scale for those outcomes that have been log transformed.
A "bivariate outlier" for checking will be defined here as a pair of successive serial datapoints of the same measure for a participant whose difference is at least four standard deviations from the mean of all patients' such differences. Simple plots of successive pairs of serial measures will be used through the 24-month period to assist in identifying outliers for data checking. Outliers will be identified for further investigation by looking at the distributions of the data through histograms, scatter plots or box-plots. Univariate tests for the compatibility of the distribution with a normal distribution will not be undertaken since they can be too sensitive to departures that are often not relevant for the comparison of means (Central Limit Theorem).

Handling outliers
Once an outlier is found, a blinded member of the team with sufficient clinical experience will be involved in the decisions as to whether a data value is impossible versus implausible versus plausible. If the outlier is impossible, then it will be set to missing, and a list of these occurrences will be appended to this SAP. If an outlier is clinically plausible, the outlier will remain. If an outlier is clinically implausible (but possible), it will not be ignored or deleted but will be retained for ITT analysis.
If outliers remain in the distribution of a variable, then data transformations or nonparametric methods of analysis may be considered.
Sensitivity analysis will be undertaken to check whether the outlier is influential by obtaining results with and then without inclusion of the outlier. If the conclusions are changed, then this will be noted. The flow diagram of the study is the one below. This will include the number randomised, who comprise the intention to treat and per protocol population, and the numbers followedup to be in the analyses of the primary outcome as well as the main reasons for missing data by stages of the trial.

Baseline comparability of randomised groups
Baseline descriptions of participants by treatment and overall will be summarised (into Table  1 of the report). No significance testing will be carried out as any differences found may be chance-generated and not for hypothesised reasons.
Continuous variables such as OCT central subfield thickness and VFQ-25 will be summarised using means and standard deviations (SD) and/or medians and interquartile range (IQR) for variables presenting a skewed distribution. Categorical variables such as proportion of patients gaining ≥15 BCVA or participants with OCT CST < 320μm will be described using numbers and percentages.

Comparison of rates of adherence and follow-up
High compliance and low attrition rates are anticipated for this study according to previous clinical trial experience (91.6% of subjects completed the active treatment arms in the 12 month CRUISE (CRVO) study and withdrawals were mainly due to physician and patients decisions (see Protocol). A cumulative drop-out of approximately 15% by year 2 was predicted and reflected in the sample size calculations. Nevertheless, compliance rates and attrition rates will be compared and reported by arm using Fisher's exact test.

Stratifiers
It is important to consider which, if any, covariates are to be adjusted for in the analyses. The ICH E9 guideline (1) recommends that consideration be given to accounting for randomisation stratifiers by adjusting for them as covariates in linear model. This tends to improve the precision of estimated treatment effects. Therefore, for continuous outcomes, the analysis will include adjustment for the randomisation stratifiers of screening BCVA letter score (3 levels) and disease duration (2 levels).

Baseline
The corresponding baseline measure for a continuous outcome is also often predictive of the outcome at follow-up. Therefore "baseline", if collected, will be an additional covariate when modelling continuous outcomes. This will be the case for visual acuity and macular volume. The continuous baseline will have precedence for inclusion in the model over the corresponding categorical randomisation stratifier, where applicable.

Statistical Model
The following description of the statistical analysis applies to each of the two investigational treatments, bevacizumab and afilbercept and the standard treatment, ranibizumab. The primary efficacy measure is the change from baseline in refracted best corrected visual acuity (BCVA) in the study eye, using the ETDRS letter score at 100 weeks. As the analysis approach for continuous outcomes below makes advantage of covariate-adjustment for the baseline of the outcome, the primary endpoint can equivalently be regarded to be each participant's 100-week measurement. This is convenient because then those with a 100-week outcome, but whose baseline measurement is missing, are not regarded to be missing the endpoint. The primary outcome may therefore be referred to below as the 100-week visual acuity, rather than the change in this from baseline to 100 weeks.
The primary outcome will be analysed using a linear mixed effects (LME) model incorporating the 5 post-baseline measurements of the refracted BCVA outcome " (12, 24, 52, 76 and 100 weeks). This mixed model will have, by definition a mix of random and fixed effect terms. The random effect in the model will be participant, represented as a random intercept at each follow-up timepoint, with allowance for within-participant correlation in the adjusted post-baseline outcomes. The fixed effects in the model will be the main effect terms for arm, the two stratifiers: visual acuityand disease duration, "time", the baseline of the outcome and its missing indicator required for the missing indicator method (21). The other fixed effects to be included in the model will be the interactions between "time" and each of the other fixed effects in the model. This model allows the treatment effect to be formally tested at 52 weeks, at the primary timepoint of 100 weeks, and estimated at 24 and 76 weeks.

Intention to Treat Strategy
Outcome data will be valid and included if the BCVA measure is refracted. All randomised subjects who provide at least one post-baseline valid measurement will be included.

Per Protocol analysis
For the analysis of the primary outcome, the mixed effects model will be re-fitted in a reduced per protocol (PP) population already described in section 3.3.2.. Only valid (refracted) measurements will be included, and so the per protocol analysis will be a subset of the outcome measurements in the 52 and 100-week ITT analysis LME model.

Concluding non-inferiority
Non-inferiority will only be concluded if this is declared by both the ITT analysis and the PP analysis at 100 weeks. Non-inferiority will also be assessed secondarily in ITT and PP populations at 52 weeks from the same models. Non-inferiority will be declared if the estimated 95% confidence interval for the difference in means lies wholly above the margin of -5 letters in both ITT and PP analysis models primarily at 100 weeks and secondarily at 52 weeks.

Superiority
If non-inferiority is concluded, superiority will be assessed from the ITT LME model by reporting the p-value from the two-sided test of the hypothesis of a zero difference in population means using a 5% significance level without need for correction for multiple testing.
In addition, if both investigative treatments were considered non-inferior to the standard treatment at 100 weeks then superiority of the investigative treatments will be assessed to each other.

Subgroup analysis
The threesubgroup variables will be assessed by extending the primary outcome model to have an interaction between arm and each categorical subgroup variable. Subgroup variables with more than two categories that are ordinal will be entered as linear in the interaction. The treatment effects will also be presented within each subgroup category with a 95% confidence interval.

Sensitivity to missing data
An expert missing-data group concluded that rather than statisticians reacting to missing data at the end of a trial, there should be comprehensive, proactive planning for handling missing data at the stage of designing trials (22). The group recommended there should be consideration of missing data mechanisms (e.g Missing At Random), and, if the missing data may be informative that appropriate sensitivity analyses should be undertaken to investigate the robustness of the inferences to the different assumptions made by the main analysis. It has also been recommended that analyses allowing for non-response and low intervention uptake (or compliance) are best specified in advance and included in the analysis plan (23).
As it is expected that compliance will be high from the fear of loss of sight, and as noninferiority is concluded only when declared in both a compliant PP population and a less compliant ITT population, the focus is on handling of missing data.
A sensitivity analysis will be undertaken to assess the possibility of alternative plausible values of treatment effect arising from potential mishandling of missing data in the primary analysis model.
The LME model for the primary outcome analysis described above is the first of a two-part approach called the Intention to Treat Strategy (13) in which a second analysis examines the sensitivity of the results to missing data in the full randomised, Intention to Treat, population. This meets the ideal of ITT. The approach to missing data taken for Leavo follows the recently published implementation paper of the ITT strategy (24). This is then also applied again to the PP population so that the non-inferiority conclusion can be reassessed under the sensitivity analysis.
For the sensitivity analysis, we pre-specify a range for best visual acuity from -20 letters to +20 letters over which the mean of the "unobserved outcome data" might depart (or be different) from the mean of the "observed outcome data" (24). In other words, this range can be thought of as how much a typical subject with missing data may on average have had a different estimated treatment effect compared to the corresponding subject with the outcome data observed (given the same baseline covariates and follow-up data in the LME model).
The range (-20 to +20) is chosen to represent both negative and positive departures that could potentially arise as the "net effect" of alternative reasons which may be unknown; such as dropout due to no anticipated further improvement, or dropout due to no improvement so far together with no anticipated achievable improvement.
This range of 40 letters (from -20 to +20) is generously wide for exploring sensitivity of the main results to departures from the MAR assumption, because 20 letters (as the maximum departure in either direction) is larger than the detectable between-arm treatment effect of 3 lines (15 letters) seen in superiority trials (difference in means) which is a sizeable shift in the mean of the distribution for dropouts compared to completers.
At the end of the trial, the fractions of individuals with missing data for visual acuity at 100 weeks will be available in each arm f i (for intervention) and f c (for control). The parameter representing excess visual acuity in those missing compared to those observed, δ, will take values by passing across the range -20 to +20. Three scenarios will be undertaken within the sensitivity analysis (23,24). These reflect whether departures from the MAR assumption apply within the intervention arms only (aflibercept and bevacizumab), within the control arm only (ranibizumab), or within both arms equally and in the same direction (thereby potentially cancelling out across the sensitivity range, if the dropout rate were to be the same in both arms).
Scenario 1: the treatment effect from the LME model will be increased by f i δ Scenario 2: the treatment effect from the LME model will be increased by -f c δ Scenario 3: the treatment effect from the LME model will be increased by (f i -f c )δ

Sensitivity analysis to use of concomitant treatments
The use of concomitant treatments will be monitored by the DMC. If necessary, a sensitivity analysis will be undertaken to examine the robustness of the 100-week per protocol analysis to the use of concomitant treatments.

Interim analysis
Formal interim analysis of the primary outcome for early stopping is not planned for this study. Regular interim reports will be prepared as needed for DMEC meetings.

Analysis of continuous outcomes
As for the primary outcome, the analysis of continuous secondary outcomes will be compared between arms at 100-weeks using linear mixed effect model adjusting for all randomisation stratifiers and where collected, the baseline of the outcome with the associated missing indicator. Time will be represented as categorical contrasts in main effect form and in interaction with all other fixed effects.

Analysis of binary outcomes
For the binary outcomes, such as the proportion of participants with ≥15 ETDRS letter improvement, chi-squared tests will be used. Safety outcomes will be reported as unadjusted patient proportions and rates within and between arms with 95% confidence intervals using exact methods where appropriate.

Analysis methods for secondary outcomes
All study analyses will be based on tests that are two-sided, including the two-sided 95% confidence intervals.
For the secondary outcomes mentioned in section 6.2, the following analysis will be used:

Handling multiple comparisons
Significance tests will be used sparingly and restricted where possible to addressing stated hypotheses. Secondary outcomes, as well as the primary outcome, will be summarised using an effect size with a 95% confidence interval. Interpretation for those secondary outcomes that do not directly address the stated study hypotheses will be more cautious. An online data collection system for clinical trials (MACRO; InferMed Ltd) will be used. This is hosted on a dedicated server at KCL and managed by the MH&N CTU. The MH&N CTU Data Manager will extract data periodically as needed and provide these in comma sepa (.csv) format.

Statistical analysis:
The principal software package will be IBM SPSS Statistics 23 and R software will be available.

DMC monitoring
We expect the DMEC would want to monitor the non-inferiority of the investigational treatments in relation to the standard treatment and we would regularly provide information such as non-compliance and withdrawal and other information listed on appendix 1.

Acknowledgments
In translating the study protocol into this statistical analysis plan, we are grateful to explanations from the study team including Philip Hykin and Sobha Sivaprasad. Further versions of the plan will be commented on by members of the Data Monitoring and Trial Steering Committees. Version 5 was produced on 30 th September 2014 and updated on 26 th November 2014 to take into account LEAVO protocol v3.0. This will be the final version approved by the independent TSC after their comments.

Amendments to Versions
Amendments to versions will be listed here.
Version 5.1 was amended to Version 5.2 as a result of the DMC meeting held on 11 th December, in open session, and the DMC recommendation to the TSC, discussed at the TSC meeting on 8 th January 2016.
The DMC discussed the circumstances under which a BCVA score at 100 weeks, or other timepoints, would not reflect the underlying visual status of a participant. In particular, recent vitreous haemorrhages may cause low BCVA scores which would then return to normal for the patient, either spontaneously or through appropriate clinical management (vitrectomy). The challenge is that any such measurements could artificially induce very large negative changes in BCVA which would have enormous influence in statistical analysisspecifically by leading to very large inflations in the standard deviation for the change from baseline. This could have profound implications for the ability of this noninferiority trial to achieve its objectives, which rely on the 95% confidence interval for the difference between randomised groups in the change from baseline falling within prespecified bounds (the non-inferiority margin). As such values intrinsically do not reflect the underlying visual status of the patient, the DMC proposed that the TSC consider amending the primary analysis population measurements to exclude from analysis any refracted BCVA measurement which is both >3 SD below the mean at that timepoint (including all measurements) and taken within 3 months of occurrences of a vitreous haemorrhage. The TSC also considered a proposal from the PI that Visual Acuity loss due to other casues unrelated to maculopathy secondary to CRVO be included. The absolute number of measurements excluded across the timepoints of measurement of refracted visual acuity is expected to be small. The TSC requested confirmation that these occurrences (number and nature) will be transparently reported by arm, and this has been included in this SAP.
Version 5.2 was amended to Version 5.3 as a result of the DMC meeting held on 1 st November 2016. The inclusion/exclusion criteria and the Per Protocol definition was updated to be in conformity with the Protocol version 4.0 as well as the wording of the secondary objectives. The randomisation stratifier 'previous treatment' was removed from the outcome analysis models as a covariate as well as a variable in subgroup analysis due to the very small number of patients having had previous treatment in the trial. Also the categories 3-6 months and >6months of the disease duration stratifier will be merged for the same reasons and will be analysed as such in the models and subgroup analysis. Finally, the method of randomisation had been mis-typed in on section 8.2 in the SAP as being stratified, whereas it has all along been minimisation. This wording has been corrected.