Statistical Analysis Plan

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 1 of 20

Version 1.0

Date: 09 June 2020

Protocol version: 6.0, 14 May 2020

IRAS no: 281712

REC ref: EE/20/0101

ISRCTN: 50189673

EudraCT: 2020-001113-21

RECOVERY SAP  Version date: 09 June 2020  Version number: 1.0 
 
RECOVERY SAP v1.0 09_06_20.DOCX    Page 2 of 20 
 
Table of Contents 
 
Abbreviations ......................................................................................................................... 4 
List of authors and reviewers................................................................................................. 5 
Roles and responsibilities ...................................................................................................... 6 
1  Introduction ....................................................................................................................... 7 
2  Background information .................................................................................................... 7 
2.1  Rationale ..................................................................................................................... 7 
2.2  Objectives of the trial .................................................................................................. 7 
  Primary objective ................................................................................................. 7 
  Secondary objectives ........................................................................................... 7 
2.3  Trial design .................................................................................................................. 7 
2.4  Eligibility ...................................................................................................................... 8 
  Inclusion criteria................................................................................................... 8 
  Exclusion criteria .................................................................................................. 8 
2.5  Treatments .................................................................................................................. 8 
  Main randomisation part A: ................................................................................. 8 
  Main randomisation part B: ................................................................................. 8 
  Second randomisation for patients with progressive COVID-19 ......................... 9 
2.6  Definitions of primary and secondary outcomes ........................................................ 9 
  Primary outcome ................................................................................................. 9 
  Secondary clinical outcomes ................................................................................ 9 
  Subsidiary clinical outcomes ................................................................................ 9 
  Detailed derivation of outcomes ....................................................................... 10 
2.9  Randomisation .......................................................................................................... 10 
  Main randomisation part A ................................................................................ 10 
  Main randomisation part B ................................................................................ 11 
  Second randomisation for patients with progressive COVID-19 ....................... 11 
2.10  Blinding .................................................................................................................. 11 
2.11  Data collection schedule........................................................................................ 11 
2.12  Data monitoring ..................................................................................................... 12 
2.13  Trial reporting ........................................................................................................ 12 
3  Analysis populations ........................................................................................................ 12 
3.1  Population definitions ............................................................................................... 12 
4  Descriptive analyses ......................................................................................................... 12 

RECOVERY SAP  Version date: 09 June 2020  Version number: 1.0 
 
RECOVERY SAP v1.0 09_06_20.DOCX    Page 3 of 20 
 
1  Participant throughput .............................................................................................. 12 
2  Baseline comparability of randomised groups ......................................................... 12 
  Main randomisation (part A and B) ................................................................... 13 
  Second randomisation ....................................................................................... 13 
3  Completeness of follow-up ....................................................................................... 13 
4  Adherence to treatment ........................................................................................... 13 
Comparative analyses ...................................................................................................... 14 
1  Main randomisation part A ....................................................................................... 14 
  Primary outcome ............................................................................................... 14 
  Secondary outcomes .......................................................................................... 14 
1.2.1  Time to discharge from hospital .................................................................... 14 
1.2.2  Use  of  mechanical  ventilation/ECMO  or  death  (among  those  not  on 
ventilation or ECMO at randomisation) ........................................................................... 14 
  Subsidiary clinical outcomes .............................................................................. 15 
1.3.1  Cause-specific mortality ................................................................................. 15 
1.3.2  Use of renal dialysis or haemofiltration ......................................................... 15 
1.3.3  Major cardiac arrhythmia............................................................................... 15 
1.3.4  Use of ventilation (overall and by type) ......................................................... 15 
1.3.5  Duration of ventilation (overall and by type) ................................................. 15 
2  Main randomisation part B ....................................................................................... 15 
3  Second randomisation .............................................................................................. 16 
4  Pre-specified subgroup analyses ............................................................................... 16 
5  Significance levels and adjustment of p-values for multiplicity ................................ 16 
6  Statistical software employed ................................................................................... 17 
7  Data standards and coding terminology ................................................................... 17 
Safety data ....................................................................................................................... 17 
Additional exploratory analysis ....................................................................................... 17 
DIFFERENCES FROM protocol V6.0 .................................................................................. 17 
References ....................................................................................................................... 17 
1  Trial documents ......................................................................................................... 17 
2  Other references ....................................................................................................... 17 
Approval ........................................................................................................................... 19 
Document history ............................................................................................................ 20 
   

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 4 of 20

Abbreviations

ADaM

Analysis Data Model

adverse event

CDISC

The Clinical Data Interchange Standards Consortium

confidence interval

CoV

Coronavirus

COVID

coronavirus-induced disease

CPAP

Continuous Positive Airway Pressure

CRP

C-reactive protein

CTU

clinical trials unit

CTSU

Clinical Trials Service Unit

DMC

Data Monitoring Committee

ECMO

Extra Corporeal Membrane Oxygenation

eCRF

Electronic case report form

FiO

fraction of inspired oxygen

ICD

International Classification of Diseases

IFN

interferon

ICNARC

Intensive Care National Audit and Research Centre

IQR

interquartile range

ITT

intention to treat

MedDRA

Medical Dictionary for Regulatory Activities

MERS

Middle East Respiratory Syndrome

NPEU

National Perinatal Epidemiology Unit

OPCS-4

NHS Classification of Interventions and Procedures

PaO

partial pressure of oxygen

risk ratio

SAE

serious adverse event

SARS

severe acute respiratory syndrome

SARS-CoV-2

virus causing COVID-19

SSAR

Suspected serious adverse reaction

SUSAR

Suspected unexpected serious adverse reaction

standard deviation

Steering Committee

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 5 of 20

List of authors and reviewers

Authors

Dr Louise Linsell, Lead Trial Statistician, Nuffield Department of Population Health (NDPH),

University of Oxford

Jennifer Bell, Trial Statistician, NDPH, University of Oxford

Reviewers

Professor Jonathan Emberson, Data Monitoring Committee (DMC) Statistician, NDPH,

University of Oxford (prior to unblinded interim analysis of trial outcomes)

Professor Richard Haynes, Steering Committee (SC) Member, NDPH, University of Oxford

Professor Peter Horby, Chief Investigator (CI), Epidemic Diseases Research Group, University

of Oxford

Professor Thomas Jaki, Steering Committee Member, Department of Mathematics and

Statistics, Lancaster University

Associate Professor Edmund Juszczak, SC Member, NDPH, University of Oxford

Professor Martin Landray, Deputy CI, NDPH, University of Oxford

Professor Alan Montgomery, SC Member, Nottingham Clinical Trials Unit, University of

Nottingham

Dr Natalie Staplin, DMC Statistician, NDPH, University of Oxford (prior to unblinded interim

analysis of trial outcomes)

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 6 of 20

Roles and responsibilities

Trial Statisticians

Dr Louise Linsell and Jennifer Bell

NDPH, University of Oxford

Role: To develop the statistical analysis plan and conduct the final comparative analysis.

Blinded to trial allocation.

Data Monitoring Committee (DMC) Statisticians

Professor Jonathan Emberson and Dr Natalie Staplin

NDPH, University of Oxford

Role: To conduct regular interim analyses for the DMC. Contribution restricted up until

unblinded to trial allocation.

Trial IT systems & Programmers

Andy King, David Murray, Richard Welsh

NDPH, University of Oxford

Role: To generate and prepare reports monitoring the randomisation schedule. To supply

data snapshots for interim and final analysis. Responsibility for randomisation system, clinical

databases and related activities.

Bob Goodenough

NDPH, University of Oxford

Role: Validation of IT systems

Dr Will Stevens

NDPH, University of Oxford

Role: To produce analysis-ready datasets according to CDISC standards.

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 7 of 20

1 INTRODUCTION

This document details the proposed presentation and analysis for the main paper(s) reporting

results from the multicentre randomised controlled trial RECOVERY (ISRCTN50189673) to

investigate multiple treatments on major outcomes in inpatients for COVID-19 (clinically

suspected or laboratory confirmed).

The results reported in these papers will follow the strategy set out here, which adheres to

the guidelines for the content of a statistical analysis plan.

Any subsequent analyses of a

more exploratory nature will not be bound by this strategy, and will be detailed in a separate

analysis plan.

Suggestions for subsequent analyses by oversight committees, journal editors or referees, will

be considered carefully in line with the principles of this analysis plan.

Any deviations from the statistical analysis plan will be described and justified in the final

report to the funder. The analysis will be carried out by an identified, appropriately qualified

and experienced statisticians, who will ensure the integrity of the data during their processing

e.g. by parallel programming.

This statistical analysis plan is based on the latest version of the protocol. A record of

amendments to the protocol can be found in the RECOVERY trial directory:

https://www.recoverytrial.net/for-site-staff/site-set-up-1.

2 BACKGROUND INFORMATION

2.1 Rationale

In early 2020, as the protocol was being developed, there were no approved treatments for

COVID-19. The aim of the trial is to provide reliable evidence on the efficacy of candidate

therapies (including re-purposed and novel drugs) for suspected or confirmed COVID-19

infection on major outcomes in hospitalised adult patients receiving standard care.

2.2 Objectives of the trial

Primary objective

To provide reliable estimates of the effect of study treatments on all-cause mortality within

28 days of randomisation.

Secondary objectives

To investigate the effect of study treatments on the duration of hospital stay, the need for

(and duration of) ventilation, and the need for renal replacement therapy.

2.3 Trial design

This is a multi-centre, multi-arm, adaptive, open-label, randomised controlled trial with three

possible stages of randomisation. In the main randomisation patients are allocated to no

additional treatment or one of 4 anti-viral or host-directed treatments. In addition, in a

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 8 of 20

factorial design, eligible patients can also be allocated simultaneously to no additional

treatment or convalescent plasma. Patients who deteriorate according to predefined criteria

can be further randomised to no additional treatment or an immunomodulatory treatment.

The trial is designed with streamlined processes in order to facilitate rapid large-scale

recruitment with minimal data collection.

2.4 Eligibility

Inclusion criteria

Patients are eligible for the trial if all of the following are true:

 Hospitalised

 SARS-Cov-2 infection (clinically suspected or laboratory confirmed)

 No medical history that might, in the opinion of the attending clinician, put the patient

at significant risk if they were to participate in the trial.

Exclusion criteria

If one or more of the active drug treatments is not available at the hospital or is believed, by

the attending clinician, to be contraindicated (or definitely indicated) for the specific patient,

then this fact will be recorded via the web-based form prior to randomisation; random

allocation will then be between the remaining (or indicated) arms.

2.5 Treatments

All patients will receive standard management for the participating hospital. The main

randomisation will be between the following treatment arms (although not all arms may be

available at any one time). The doses listed are for adults; paediatric dosing is described in

the protocol.

Main randomisation part A:

 No additional treatment

 Lopinavir 400mg-Ritonavir 100mg by mouth (or nasogastric tube) every 12 hours for

10 days.

 Corticosteroid in the form of dexamethasone, administered as an oral liquid or

intravenous preparation 6 mg once daily for 10 days. In pregnancy, prednisolone 40

mg administered by mouth (or intravenous hydrocortisone 80 mg twice daily) should

be used instead.

 Hydroxychloroquine by mouth for 10 days (4 doses in first 24 hours and 1 dose every

12 hours for 9 days).

 Azithromycin 500mg by mouth (or nasogastric tube) or intravenously once daily for a

total of 10 days.

Main randomisation part B:

In a factorial design, eligible patients may be randomised to the arms below. The doses listed

are for adults; paediatric dosing is described in the protocol.

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 9 of 20

 No additional treatment

 Convalescent plasma Single unit of ABO compatible convalescent plasma (275mls +/-

75 mls) intravenous per day on study days 1 (as soon as possible after randomisation)

and 2 (with a minimum of 12 hour interval between 1st and 2nd units). ABO identical

plasma is preferred if available. The second transfusion should not be given if patient

has a suspected serious adverse reaction during or after the first transfusion.

Second randomisation for patients with progressive COVID-19

Patients enrolled in the main RECOVERY trial and with clinical evidence of a hyper-

inflammatory state may be considered for a second randomisation if they meet the following

criteria:

 Randomised into the main RECOVERY trial no more than 21 days ago

 Clinical evidence of progressive COVID-19:

- oxygen saturation <92% on room air or requiring oxygen (or in children, significant

systemic disease with persistent pyrexia, with or without evidence of respiratory

involvement); and

- C-reactive protein (CRP) ≥75 mg/L

 No medical history that might, in the opinion of the attending clinician, put the patient

at significant risk if they were to participate in this aspect of the RECOVERY trial

Eligible participants may be randomised between the following treatment arms:

 No additional treatment

 Tocilizumab by intravenous infusion with the dose determined by body weight.

2.6 Definitions of primary and secondary outcomes

Outcomes will be assessed at 28 days and then 6 months after randomisation. Analysis of

longer-term outcomes collected beyond this will be described in a separate Statistical Analysis

Plan.

Primary outcome

Mortality (all-cause)

Secondary clinical outcomes

 Time to discharge from hospital

 Use of mechanical ventilation/Extra Corporal Membrane Oxygenation (ECMO) or

death (among patients not on ventilation or ECMO at baseline)

Subsidiary clinical outcomes

 Cause-specific mortality (COVID-19; cardiovascular; non-vascular; other)

 Use of renal dialysis or haemofiltration

 Serious cardiac arrhythmia (recorded in a subset)

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 10 of 20

 Use of ventilation (overall and by type)

 Duration of ventilation (overall and by type)

Detailed derivation of outcomes

The detailed derivation of outcomes included in statistical analysis will be described

separately in a data derivation document and included in the Study Data Reviewer’s Guide.

2.7 Hypothesis framework

For each of the primary, secondary and subsidiary outcomes, the null hypothesis will be that

there is no true difference in effect between any of the treatment arms.

2.8 Sample size

The larger the number randomised, the more accurate the results will be, but the numbers

that can be randomised will depend critically on how large the epidemic becomes. If

substantial numbers are hospitalised in the participating centres then it may be possible to

randomise several thousand with moderate disease and a few thousand with severe disease.

Some indicative sample sizes and projected recruitment will be estimated using emerging

data for several different scenarios. Sample size and recruitment will be monitored by the

Steering Committee (SC) throughout the trial.

2.9 Randomisation

Eligible patients will be randomised using a 24/7 secure central web-based randomisation

system, developed and hosted within NDPH, University of Oxford. Users of the system will

have no insight into the next allocation, given that simple randomisation is being used. In the

event that a patient is randomised inadvertently more than once during the same hospital

admission, the first allocation will be used.

The implementation of the randomisation procedure will be monitored by the Senior Trials

Programmer, and the SC notified if an error in the randomisation process is identified.

Main randomisation part A

Simple randomisation will be used with a 2:1:1:1:1 allocation ratio to one of the following

treatment arms (in addition to usual care), which is subject to change:

 No additional treatment

 Lopinavir-Ritonavir

 Corticosteroid

 Hydroxychloroquine

 Azithromycin

The randomisation programme will allocate patients in a ratio of 2:1 between the no

additional treatment arm and each of the other arms that are not contra-indicated and

available. Hence if all 4 active treatment arms are available, then the randomisation will be in

the ratio 2:1:1:1:1. If one or more of the active drug treatments is not available at the hospital

or is believed, by the attending clinician, to be contraindicated (or definitely indicated) for the

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 11 of 20

specific patient, then this fact will be recorded via the web-based form prior to randomisation;

random allocation will then be between the remaining arms (in a 2:1:1:1, 2:1:1 or 2:1 ratio).

Main randomisation part B

In a factorial design, eligible patients will be randomised simultaneously using simple

randomisation with allocation ratio 1:1 to one of the following arms:

 No additional treatment

 Convalescent plasma

Second randomisation for patients with progressive COVID-19

Eligible participants may be randomised using simple randomisation with an allocation ratio

1:1 between the following arms:

 No additional treatment

 Tocilizumab

2.10 Blinding

This is an open-label study. However, while the study is in progress, access to tabular results

of study outcomes by treatment allocation will not be available to the research team, CIs, trial

statisticians, clinical teams, or members of the SC (unless the DMC advises otherwise). The

DMC and DMC statisticians will be unblinded.

2.11 Data collection schedule

Baseline and outcome information will be collected on trial-specific electronic case report

forms (eCRFs) and entered into a web-based IT system by a member of the hospital or

research staff. Follow-up information will be collected on all study participants, irrespective

of whether or not they complete the scheduled course of allocated study treatment. Study

staff will seek follow-up information through various means, including routine healthcare

systems and registries.

All randomised participants will be followed up until death or 6 months post-randomisation

to the main trial (whichever is sooner). NHS Digital and equivalent organisations in the

devolved nations will supply data fields relevant to trial baseline and outcome measures to

NDPH, University of Oxford on a regular basis, for participants enrolled into the trial. This will

be combined with the trial-specific data collected via the web-based IT system and

adjudicated internally.

Longer term (up to 10 years) follow-up will be sought through linkage to electronic healthcare

records and medical databases including those held by NHS Digital, Public Health England and

equivalent bodies, and to relevant research databases (e.g. UK Biobank, Genomics England).

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 12 of 20

2.12 Data monitoring

During the study all study data will be supplied in strict confidence to the independent DMC

for independent assessment and evaluation. The DMC will request such analyses at a

frequency relevant to the emerging data from this and other studies.

The DMC has been requested to determine if, in their view, the randomised comparisons in

the study have provided evidence on mortality that is strong enough (with a range of

uncertainty around the results that is narrow enough) to affect national and global treatment

strategies. Hence, multiple reviews by the Data Monitoring Committee have no material

impact on the final analysis. In such a circumstance, the DMC will inform the SC who will make

the results available to the public and amend the trial arms accordingly.

2.13 Trial reporting

The trial will be reported according to the principles of the CONSORT statements.

2, 3, 4

The

exact composition of the trial publication(s) depends on the size of the epidemic, the

availability of drugs, and the findings from the various pairwise comparative analyses (with

the no additional treatment arm) in the main trial.

3 ANALYSIS POPULATIONS

3.1 Population definitions

The intention to treat (ITT) population will be all participants randomised, irrespective of

treatment received. This ITT population will be used for analysis of efficacy and safety data.

For interim analyses, baseline data will be reported for all participants with data available and

outcome data will be reported for all participants who have died, been discharged from

hospital, or reached day 28 after the first randomisation.

4 DESCRIPTIVE ANALYSES

4.1 Participant throughput

The flow of participants through the trial will be summarised for each separate pairwise

comparison using a CONSORT diagram, for the main and second randomisation separately.

The flow diagram for the nested factorial design (main randomisation part B to convalescent

plasma) will be stratified by the 5 arms included in the main randomisation part A. The flow

diagrams will describe the numbers of participants randomly allocated, who received

allocation, withdrew consent, and included in the ITT analysis population. The flow diagrams

for arms in the main randomisation will also report the number of participants who

underwent the second randomisation.

4.2 Baseline comparability of randomised groups

The following characteristics will be described separately for patients randomised to each

main comparison (for each separate pairwise comparison of active treatment with the no

additional treatment arm), and separately for the first and second randomisation.

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 13 of 20

Main randomisation (part A and B)

 Age at randomisation

 Sex

 Ethnicity

 Time since COVID-19 symptoms onset

 Time since hospitalisation

 Current respiratory support requirement

 Currently requiring renal dialysis or haemofiltration

 Comorbidities (diabetes, heart disease, chronic lung disease, tuberculosis, human

immunodeficiency virus, severe liver disease, severe kidney impairment)

 If female, known to be pregnant

Second randomisation

In addition to the above:

 Type of ventilation support currently required (none, CPAP alone, non-invasive

ventilation, high-flow nasal oxygen, mechanical ventilation, ECMO)

 Latest oxygen saturation measurement (%)

 Latest CRP measurement (mg/L)

 Latest ferritin measurement (ng/mL)

 Latest creatinine measurement (µmol/L)

 Allocation in first randomisation

 Interval between first and second randomisation

The number and percentage will be presented for binary and categorical variables. The mean

and standard deviation or the median and the interquartile range will be presented for

continuous variables, or the range if appropriate. There will be no tests of statistical

significance performed nor confidence intervals calculated for differences between

randomised groups on any baseline variable.

4.3 Completeness of follow-up

All reasonable efforts will be taken to minimise loss to follow-up, which is expected to be

minimal as data collection for primary and secondary outcomes using trial-specific eCRFs is

combined with linkage to routine clinical data on study outcomes from NHS Digital, ICNARC,

and similar organisations in the devolved nations.

The number and percentage of participants with follow-up information at day 28 and at 6

months after the main randomisation will be reported. Data will be shown for each of the

following: all-cause mortality, hospital discharge status, ventilation status, and will be shown

for each randomised group for the main and second randomisation separately.

4.4 Adherence to treatment

The number and proportion of patients who did not receive the treatment they were

allocated to will be reported. If any other trial treatment options were known to be received,

instead of or in addition to, the allocated treatment during the 28 day follow-up period after

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 14 of 20

the first randomisation, these will be collected and reported. Details on the number of days

(or doses) of treatment received will be reported for all trial treatments received where

available.

5 COMPARATIVE ANALYSES

For all outcomes, the primary analysis will be performed on the intention to treat (ITT)

population at 28 days after the main randomisation. An ITT analysis of all outcomes at 6

months post-randomisation will also be conducted.

Pairwise comparisons will be made between each treatment arm and the no additional

treatment arm (reference group) in that particular randomisation (main randomisation part

A, main randomisation part B, and second randomisation). Since not all treatments may be

available or suitable for all patients, those in the no additional treatment arm will only be

included in a given comparison if, at the point of their randomisation, they could alternatively

have been randomised to the active treatment of interest (i.e. the active treatment was

available at the time and it was not contra-indicated). The same applies to treatment arms

added at a later stage; they will only be compared to those patients recruited concurrently.

5.1 Main randomisation part A

Primary outcome

Mortality (all-cause) will be summarised with counts and percentages by randomised

comparison group. A time-to-event analysis will be conducted using the log-rank test, with

the p-value reported. Kaplan-Meier estimates for the time to event will also be plotted (with

associated log-rank p-values). The log-rank ‘observed minus expected’ statistic (and its

variance) will be used to estimate the average event rate ratio and confidence interval for

each treatment group versus the no additional treatment group.

For the primary outcome,

discharge alive before the relevant time period (28 days) will be assumed as absence of the

event (unless there is additional data confirming otherwise).

Secondary outcomes

5.1.2.1 Time to discharge from hospital

A time-to-event analysis will be used to compare each treatment group with the no additional

treatment group using Kaplan-Meier and the log-rank test, as described above. Patients who

die in hospital will be censored after 28 days. This gives an unbiased estimate of the recovery

rate and comparable estimates to the competing risks approach in the absence of other

censoring (which is expected to be very minimal).

5.1.2.2 Use of mechanical ventilation/ECMO or death (among those not on ventilation or

ECMO at randomisation)

Counts and percentages will be presented by randomised group and the risk ratio will be

calculated for each pairwise comparison with the no additional treatment arm, with

confidence intervals and p-values reported. The absolute risk difference will also be presented

with confidence intervals. Patients who were already on ventilation at randomisation will be

excluded from the denominator.

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 15 of 20

Subsidiary clinical outcomes

5.1.3.1 Cause-specific mortality

Cause-specific mortality will be analysed in a similar manner to the primary outcome. Deaths

from other causes will be censored at the date of death and a separate survival curve will be

presented for each cause of death (COVID-19, other infection, cardiovascular and other).

5.1.3.2 Use of renal dialysis or haemofiltration

Counts and percentages will be presented by randomised group and the risk ratio will be

calculated for each pairwise comparison with the no additional treatment arm, with

confidence intervals and p-values reported. The absolute risk difference will also be presented

with confidence intervals. Patients who were already on renal dialysis or haemofiltration at

randomisation will be excluded from the denominator.

5.1.3.3 Major cardiac arrhythmia

Counts and percentages will be presented by randomised group and the risk ratio for any

major cardiac arrhythmia will be calculated for each pairwise comparison with the no

additional treatment arm, with confidence intervals and p-values reported. The absolute risk

difference will also be presented with confidence intervals. Type of arrhythmia will also be

described: (i) atrial flutter or fibrillation; (ii) supraventricular tachycardia; (iii) ventricular

tachycardia; (iv) ventricular fibrillation; (v) atrioventricular block requiring intervention, with

subtotals for (i)-(ii) and (iii)-(iv).

5.1.3.4 Use of ventilation (overall and by type)

Counts and percentages will be presented by randomised group for patients who received

any assisted ventilation. Patients who were already on assisted ventilation at randomisation

will be excluded from the denominator. The number of patients receiving the different types

of ventilation will also be reported: (i) CPAP; (ii) other non-invasive; (iii) high-flow nasal

oxygen; (iv) mechanical; (v) ECMO, with subtotals for (i)-(iii) (non-invasive) and (iv)-(v)

(invasive).

5.1.3.5 Duration of ventilation (overall and by type)

The mean (SD) duration of ventilation will be calculated in days from the main randomisation

for each randomised group in those who received ventilation, separately for survivors and

non-survivors. This will be reported overall for any assisted ventilation and separately for

mechanical ventilation or ECMO. The mean difference and confidence intervals will be

presented for each pairwise comparison with the no additional treatment arm.

5.2 Main randomisation part B

For the evaluation of treatment effect in the factorial design, the main effect of convalescent

plasma across all arms in main randomisation part A combined, will be presented and tested,

as described in 5.1. Data stratified by allocation in part A will also be reported to aid

interpretation, but no tests for statistical interaction will be performed.

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 16 of 20

Additional safety data will be collected in a subset of patients randomised to part B. These

will be tabulated separately by allocation (convalescent plasma versus no additional

treatment): (i) sudden worsening in respiratory status; (ii) severe allergic reaction; (iii)

temperature >39°C or ≥2°C rise since randomisation; (iv) sudden hypotension, clinical

haemolysis and thrombotic event.

5.3 Second randomisation

Evaluation of treatment effects in the main randomisation and the second randomisation will

be conducted independently, as described in 5.1. In addition to the overall comparison for

Tocilizumab vs no additional treatment, results will be stratified according to allocation in the

main randomisation (part A and part B), however no interaction tests will be performed

between the allocations in the two stages.

5.4 Pre-specified subgroup analyses

Pre-specified subgroup analyses will be conducted for the main randomisation (part A and

part B) and the second randomisation, for the following outcomes:

 Mortality (all-cause)

 Time to discharge from hospital

 Use of mechanical ventilation/ECMO or death

The analyses will be conducted using a test for heterogeneity (or test for trend for 3 or more

ordered groups). Results will be presented on forest plots as event rate ratios (or risk ratios)

with confidence intervals. The following subgroups will be examined:

 Risk group (three risk groups with approximately equal number of deaths based on

factors recorded at randomisation)

 Requirement for respiratory support at randomisation (None; Oxygen only;

Ventilation or ECMO)

 Time since illness onset (≤7 days; >7 days)

 Age (<70; 70-79; 80+ years)

 Sex (Male; Female)

 Ethnicity (White; Black, Asian or Minority Ethnic; Unknown)

Additional analyses will set the results for children (<18 years) and pregnant women in the

context of the overall results.

5.5 Significance levels and adjustment of p-values for multiplicity

Evaluation of the primary trial (main randomisation) and secondary randomisation will be

conducted independently and no adjustment be made for these. Formal adjustment will not

be made for multiple treatment comparisons, the testing of secondary and subsidiary

outcomes, or subgroup analyses. 95% confidence intervals will be presented for estimates of

between-group effects throughout.

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 17 of 20

5.6 Statistical software employed

The statistical software SAS version 9.4, R Studio 3.6.2 and Stata/SE version 15 (or later) for

Windows will be used for the interim and final analyses.

5.7 Data standards and coding terminology

Datasets for analysis will be prepared using CDISC standards for SDTM and ADaM. Wherever

possible, clinical outcomes (which may be obtained in a variety of standards, including ICD10

and OPCS-4) will be coded using MedDRA version 20.1.

6 SAFETY DATA

Suspected serious adverse reactions (SSARs) and suspected unexpected serious adverse

reactions (SUSARs) will be listed by trial allocation.

7 ADDITIONAL EXPLORATORY ANALYSIS

Any post-hoc analysis requested by the oversight committees, a journal editor or referees will

be labelled explicitly as such. Any further future analyses not specified in the analysis protocol

will be exploratory in nature and will be documented in a separate statistical analysis plan.

8 DIFFERENCES FROM PROTOCOL V6.0

Use and duration of ventilation are described as secondary objectives in the protocol, and

listed as subsidiary outcomes in the statistical analysis plan. The testing of multiple treatment

arms will not formally be adjusted for, but given the number of comparisons, due allowance

will be made in their interpretation. Formal methods of adjustment for multiplicity were not

adopted because of treatment arms being added over time (including the factorial

convalescent plasma comparison), unequal recruitment into each arm, and the ultimate

number of treatments under evaluation not known in advance. While methods for these

situations exist it was felt that the resulting change in level of significance was not

appropriate.

9 REFERENCES

9.1 Trial documents

Dummy tables and the data derivation document can be found in the RECOVERY trial

directory and will be published with this SAP on the trial website.

9.2 Other references

1. Gamble C, Krishan A, Stocken D, Lewis S, Juszczak E, Doré C, Williamson PR, Altman

DG, Montgomery A, Lim P, Berlin J, Senn S, Day S, Barbachano Y, Loder E. Guidelines for the

Content of Statistical Analysis Plans in Clinical Trials. JAMA 2017;318(23):2337-2343.

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 18 of 20

2. Schulz KF, Altman DG, Moher D for the CONSORT Group. CONSORT 2010 Statement:

updated guidelines for reporting parallel group randomised trials. BMJ 2010;340:698-702.

3. Juszczak E, Altman DG, Hopewell S, Schulz KF. Reporting of multi-arm parallel-group

randomized trials: extension of the CONSORT 2010 statement. JAMA 2019;321(16):1610-

1620.

4. Dimairo M, Coates E, Pallmann P, Todd S, Julious SA, Jaki T, Wason J, Mander AP,

Weir CJ, Koenig F, Walton MK. Development process of a consensus-driven CONSORT

extension for randomised trials using an adaptive design. BMC medicine 2018;16(1):210.

5. Peto R, Pike MC, Armitage P, Breslow NE, Cox DR, Howard SV, Mantel N, McPherson

K, Peto J, Smith PG. Design and analysis of randomized clinical trials requiring prolonged

observation of each patient. Part II: analysis and examples. Br J Cancer 1977;35:1-39.

6. Betensky RA and Schoenfeld DA. Nonparametric Estimation in a Cure Model with

Random Cure Times. Biometrics 2001;57:282-286.

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 19 of 20

10 APPROVAL

Lead Trial

Statistician

Name: Dr Louise Linsell

Signature:

Date:

Chief

Investigator

Name: Professor Peter Horby

Signature:

Date:

Deputy Chief

Investigator

Name: Professor Martin Landray

Signature:

Date:

Steering

Committee

Statistician

Name: Associate Professor Ed Juszczak

Signature:

Date:

Steering

Committee

Statistician

Name: Professor Alan Montgomery

Signature:

Date:

Steering

Committee

Statistician

Name: Professor Thomas Jaki

Signature:

Date:

RECOVERY SAP Version date: 09 June 2020 Version number: 1.0

RECOVERY SAP v1.0 09_06_20.DOCX Page 20 of 20

11 DOCUMENT HISTORY

Version

Date

Edited

Comments/Justification

Timing in

relation to

unblinded

interim

monitoring

Timing in

relation to

unblinding of

Trial

Statisticians

0.1

20/03/20

LL/JB

First draft.

Prior

0.2

01/04/20

LL/JB

Comments and amendments

from Martin Landray, Jonathan

Emberson & Natalie Staplin.

Also aligned with updated

protocol and CRFs.

Prior

0.3

01/04/20

EJ/LL

Further edits and comments.

Prior

0.4

07/04/20

JB/EJ/LL

Following statistics group

meeting on 02/04/20.

Prior

0.5

22/04/20

JB/LL/EJ

Following statistics group

meeting on 09/04/20 and

further protocol update.

After

Prior

0.6

24/04/20

Following statistics group

meeting on 23/04/20.

After

Prior

0.7

10/05/20

Protocol update.

After

Prior

0.8

15/05/20

Following statistics group

meeting on 15/05/20.

After

Prior

0.9

27/05/20

Further comments from SC

members prior to interim

analysis on 28/05/20.

After

Prior

1.0

09/06/20

Revised following the stopping

of the hydrochloroquine arm,

and prior to the trial statisticians

receiving unblinded data for this

arm.

After

Prior