Convalescent plasma in patients admitted to hospital with COVID-19 (RECOVERY): a randomised, controlled, open-label, platform trial

Background: Treatment of COVID-19 patients with plasma containing anti-SARS-CoV-2 antibodies may have a beneficial effect on clinical outcomes. We aimed to evaluate the safety and efficacy of convalescent plasma in patients admitted to hospital with COVID-19. Methods: In this randomised, controlled, open-label, platform trial (Randomised Evaluation of COVID-19 Therapy [RECOVERY]) several possible treatments are being compared with usual care in patients hospitalised with COVID-19 in the UK. Eligible and consenting patients were randomly allocated to receive either usual care plus high titre convalescent plasma or usual care alone. The primary outcome was 28-day mortality. Findings: Between 28 May 2020 and 15 January 2021, 5795 patients were randomly allocated to receive convalescent plasma and 5763 to usual care alone. There was no significant difference in 28-day mortality between the two groups: 1398 (24%) of 5795 patients allocated convalescent plasma and 1408 (24%) of 5763 patients allocated usual care died within 28 days (rate ratio [RR] 1.00; 95% confidence interval [CI] 0.93 to 1.07; p=0.93). The 28-day mortality rate ratio was similar in all prespecified subgroups of patients, including in those patients without detectable SARS-CoV-2 antibodies at randomisation. Allocation to convalescent plasma had no significant effect on the proportion of patients discharged from hospital within 28 days (66% vs. 67%; rate ratio 0.98; 95% CI 0.94-1.03, p=0.50). Among those not on invasive mechanical ventilation at baseline, there was no significant difference in the proportion meeting the composite endpoint of progression to invasive mechanical ventilation or death (28% vs. 29%; rate ratio 0.99; 95% CI 0.93-1.05, p=0.79). Interpretation: Among patients hospitalised with COVID-19, high-titre convalescent plasma did not improve survival or other prespecified clinical outcomes.


INTRODUCTION 52
A substantial proportion of individuals infected with severe acute respiratory syndrome 53 coronavirus 2 (SARS-CoV-2) require hospital care, which can progress to critical illness 54 with hypoxic respiratory failure. In those with severe Coronavirus , immunomodulation with corticosteroids and interleukin-6 receptor antagonists has 56 been shown to improve survival. 1,2 Treatments that effectively inhibit viral replication 57 may reduce tissue damage and allow time for the host to develop an adaptive immune 58 response that will clear the infection. To date, however, no treatment directed against 59 the virus has been shown to reduce mortality (although remdesivir may shorten the 60 duration of hospital stay). 3 61 Humoral immunity is a key component of the immune response to SARS-CoV-2 and 62 matures over several weeks following infection. Anti-SARS-CoV-2 antibodies are 63 detectable at a mean of 13 days after symptom onset, but neutralising titres do not peak 64 until day 23 and there is wide variation in both the timing of seroconversion and peak 65 antibody levels between infected individuals. 4 While patients with severe COVID-19 66 generally have higher final antibody concentrations than those with mild disease, their 67 antibody responses are delayed. 5 Antibodies may modulate acute viral disease either 68 by a direct antiviral effect, binding and neutralizing free virus, or indirectly by activating 69 antiviral pathways such as the complement cascade, phagocytosis and cellular 70 cytotoxicity. Conversely, there is also a possibility that antibodies may enhance disease, 71 randomisation for patients with progressive COVID-19 (evidence of hypoxia and a 137 hyper-inflammatory state) to tocilizumab versus usual care. Participants and local study 138 staff were not masked to the allocated treatment. Several of these treatment arms were 139 added to or removed from the protocol over the period that convalescent plasma was 140 evaluated (appendix pp 29-34). The trial steering committee, investigators, and all other 141 individuals involved in the trial were masked to outcome data during the trial. Authorization (EUA). 19 For those allocated convalescent plasma, two units (275mls ± 155 75mls) were given intravenously, the first as soon as possible after randomisation and 156 the second (from a different donor) the following day and at least 12 hours after the first. 157 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 10, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 Early safety outcomes were recorded using an online form 72 hours following 158 randomisation (appendix pp [38][39][40][41][42]). An online follow-up form was completed when 159 patients were discharged, had died, or at 28 days after randomisation, whichever 160 occurred earlier (appendix pp 43-49). Information was recorded on adherence to 161 allocated trial treatment, receipt of other COVID-19 treatments, duration of admission, 162 receipt of respiratory or renal support, and vital status (including cause of death). In 163 addition, routine health care and registry data were obtained including information on 164 vital status at day 28 (with date and cause of death); discharge from hospital; and 165 receipt of respiratory support or renal replacement therapy. 166

Measurement of participant baseline SARS-CoV-2 serostatus 167
Baseline SARS-CoV-2 serostatus for each participant was determined using serum 168 samples taken at the time of randomisation. Analysis was performed at a central 169 laboratory using a validated 384-well plate indirect ELISA (appendix p 29). 20 170 Participants were categorised as seropositive or seronegative using a predefined assay 171 threshold that has ≥99% sensitivity and specificity in detecting individuals with SARs-172 CoV-2 infection at least 20 days previously. 20 173

Outcomes 174
Outcomes were assessed at 28 days after randomisation, with further analyses 175 specified at six months. The primary outcome was all-cause mortality. Secondary 176 outcomes were time to discharge from hospital and, among patients not receiving 177 invasive mechanical ventilation at randomisation, subsequent receipt of invasive 178 mechanical ventilation (including extra-corporeal membrane oxygenation) or death. 179 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 10, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 Prespecified, subsidiary clinical outcomes included receipt of ventilation, time to 180 successful cessation of invasive mechanical ventilation (defined as removal of invasive 181 mechanical ventilation within, and survival to, 28 days), and use of renal dialysis or 182 haemofiltration. 183 Prespecified safety outcomes were transfusion related adverse events at 72 hours 184 following randomisation (worsening respiratory status, suspected transfusion reaction, 185 fever, hypotension, haemolysis, and thrombotic events), cause-specific mortality, and 186 major cardiac arrhythmia. Information on serious adverse reactions to convalescent 187 plasma was collected in an expedited fashion via the existing NHS Serious Hazards Of 188 Tranfusion (SHOT) haemovigilence scheme. 189

Statistical Analysis 190
In accordance with the statistical analysis plan, an intention-to-treat comparison was 191 conducted between patients randomised to convalescent plasma and patients 192 randomised to usual care in those for whom convalescent plasma was both available 193 and suitable as a treatment. For the primary outcome of 28-day mortality, the log-rank 194 observed minus expected statistic and its variance were used both to test the null 195 hypothesis of equal survival curves (i.e. the log-rank test) and to calculate the one-step 196 estimate of the average mortality rate ratio. We constructed  curves to display cumulative mortality over the 28-day period. We used similar methods 198 to analyse time to hospital discharge and successful cessation of invasive mechanical 199 ventilation, with those patients who died in hospital right-censored on day 29. Median 200 time to discharge was derived from Kaplan-Meier estimates. For the prespecified, 201 composite, secondary outcome of progression to invasive mechanical ventilation or 202 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 10, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021 death within 28 days (among those not receiving invasive mechanical ventilation at 203 randomisation) and the subsidiary clinical outcomes of receipt of ventilation and use of 204 haemodialysis or haemofiltration, the precise dates were not available and so the risk 205 ratio was estimated instead. (Through the play of chance, a slightly lower proportion of 206 males were allocated convalescent plasma than usual care; analyses adjusted for sex 207 are provided in the appendix [webtable 7] and are virtually identical to the main results 208 shown.) Sensitivity analyses of the primary and secondary outcomes were conducted 209 among those patients with a positive PCR test for SARS-COV-2. 210 Prespecified analyses of the primary outcome were performed in seven subgroups 211 defined by characteristics at randomisation: age, sex, ethnicity, level of respiratory 212 support received, days since symptom onset, use of systemic corticosteroids, and 213 presence of anti-SARS-CoV-2 antibody. Observed effects within these subgroup 214 categories were compared using a chi-squared test for heterogeneity or trend. Post-hoc 215 exploratory analyses included further examination by days since symptom according to 216 four rather than two levels and by level of respiratory support by sub-dividing the 217 'oxygen only' group into three sub-categories. In late 2020, a new SARS-CoV-2 variant, 218 difference in the effectiveness of convalescent plasma before and after the emergence 225 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 10, 2021. ; https://doi.org/10.1101/2021.03.09.21252736 doi: medRxiv preprint of B.1.1.7, a further post-hoc exploratory analysis was done of the primary outcome 226 comparing effects in those randomised before 1 December 2020 with those randomised 227 from 1 December 2020 onwards. 21 228 Estimates of rate and risk ratios are shown with 95% confidence intervals. All p-values 229 are 2-sided and are shown without adjustment for multiple testing. The full database is 230 held by the trial team who pooled the data from trial sites and performed the analyses at 231 the Nuffield Department of Population Health, University of Oxford. 232 Analyses were performed using SAS version 9.4 and R version 3.4. The trial is 233 registered with ISRCTN (50189673) and clinicaltrials.gov (NCT04381936). 234

Sample size and decision to stop enrolment 235
As stated in the protocol, appropriate sample sizes could not be estimated when the trial 236 was being planned at the start of the COVID-19 pandemic. During the trial, external 237 data suggested that any benefits of antibody-based therapies may be greater among 238 those patients who had not raised an adequate antibody response of their own. 12 239 Consequently, while still blind to the results of the trial, the RECOVERY steering 240 committee determined that the trial should enrol sufficient patients to provide at least 241 90% power at a two-sided p-value of 0.01 to detect a proportional reduction in 28-day 242 mortality of one-fifth among those patients with and, separately, without detectable 243 SARS-CoV-2 antibodies at randomisation (appendix p 34). 244 On 7 th January 2021, the independent data monitoring committee (DMC) conducted a 245 routine review of the data and recommended that the chief investigators pause the 246 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 10, 2021. On 14 th January 2021, the DMC conducted another routine review of the data and 251 notified the chief investigators that there was no convincing evidence that further 252 recruitment would provide conclusive proof of worthwhile mortality benefit either overall 253 or in any pre-specified subgroup. The DMC therefore recommended that recruitment to 254 the convalescent plasma portion of the study should cease and follow-up be completed. 255 Enrolment of patients to the convalescent plasma group was closed on 15 th January 256 2021 and the preliminary result for the primary outcome was made public. 257

Role of the funding source 258
The funders of the trial had no role in trial design, data collection, data analysis, data 259 interpretation, or writing of the report. The corresponding authors had full access to all 260 the data in the study and had final responsibility for the decision to submit for 261 publication. 262 263 264 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

RESULTS 265
Patients 266 Between 28 May 2020 and 15 January 2021, 13127 (81%) of 16287 patients enroled 267 into the RECOVERY trial, were eligible to be randomised to convalescent plasma (that 268 is, convalescent plasma was available in the hospital at the time and the patient had no 269 known contraindication to convalescent plasma (figure 1). Of these, 5795 were 270 randomised to convalescent plasma plus usual care and 5763 were randomised to 271 usual care alone (figure 1), with the remainder being randomised to receive REGN-272 COV2. The mean age of trial patients in this comparison was 63.5 (SD 14.7) years and 273 the median time from symptom onset to randomization was 9 days (IQR 6 -12) (table  274 1, webtable 1). At randomisation, 617 (5%) were receiving invasive mechanical 275 ventilation, 10044 (87%) were receiving oxygen only (with or without non-invasive 276 respiratory support), and 897 (8%) were receiving no oxygen therapy (webtable 1). 92% 277 of patients were receiving corticosteroids at time of randomisation. 278 Of the 9385 (81%) patients for whom a baseline serology result was available, 5774 279 (62%) were SARS-CoV-2 antibody seropositive (webtable 1). Patients were more likely 280 to be seronegative if they were older, female, white, had shorter duration of symptoms, 281 were receiving less intensive respiratory support, or were SARS-CoV-2 RNA negative 282 by PCR (webtable 2). There was an imbalance in the availability of a baseline serology 283 sample, with more missing samples in the usual care arm (table1). (This likely reflects a 284 mistaken belief by some trial staff that a serology sample was only required in patients 285 allocated to convalescent plasma.) 286 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 10, 2021. ; https://doi.org/10. 1101/2021 The median time to discharge was 11 days in both those allocated convalescent plasma 308 and those allocted usual care, and allocation to convalescent plasma was associated 309 with a similar probability of discharge alive within 28 days compared to usual care (66% 310 vs. 67%, rate ratio 0·98 95% CI 0·94 to 1·03, p=0·50) (table 2). Among those not 311 receiving invasive mechanical ventilation at baseline, the number of patients 312 progressing to the prespecified composite secondary outcome of invasive mechanical 313 ventilation or death was similar for those allocated to convalescent plasma or usual care 314 (28% versus 29%, risk ratio 0·99 95% CI 0·93 to 1·05, p=0·79) (table 2). For both of 315 these secondary outcomes, there was some evidence of heterogeneity by patient 316 SARS-CoV-2 antibody test result, with slightly more favourable outcomes with 317 convalescent plasma seen among seronegative than among seropositive patients 318 (webfigures 3 and 4). Results were consistent across all other pre-specified subgroups 319 of patients. 320 We observed no significant differences in the prespecified subsidiary clinical outcomes 321 of use of ventilation, successful cessation of invasive mechanical ventilation, or 322 progression to use of renal replacement therapy (table 2). 323 We observed no significant differences in cause-specific mortality (webtable 4). Within 324 the first 72 hours after randomisation, severe allergic reactions were reported for 16 325 patients in the convalescent plasma group vs. 2 patients in the usual care group. The 326 frequency of sudden worsening in respiratory status, temperature >39 o C or ≥2 o C rise 327 above baseline, sudden hypotension, clinical haemolysis, and thrombotic events were 328 broadly similar in the two groups (webtable 5). We also observed no significant 329 differences in the frequency of major cardiac arrhythmia (webtable 6). There were 13 330 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 10, 2021. ; https://doi.org/10.1101/2021.03.09.21252736 doi: medRxiv preprint serious adverse reactions reported to SHOT: 9 patients with pulmonary reactions 331 (including 3 deaths possibly related to transfusion), and 4 patients with serious febrile, 332 allergic or hypotensive reactions (all of whom recovered). 333

DISCUSSION 334
The results of this large, randomised trial show that convalescent plasma did not including RECOVERY (which is more than ten times larger than all other trials 343 combined) allocation to convalescent plasma does not improve mortality (mortality RR 344 0·99, 95% CI 0·92-1·06, p=0·77 (figure 4). 345 It has been suggested that the benefits of convalescent plasma may depend on the 346 transfused neutralising titre, and that using plasma with lower titres could explain 347 negative results from previous randomised trials. In RECOVERY, all convalescent 348 plasma was supplied via the UK National Blood Services using standardised laboratory 349 processing. Convalescent donors were chosen based on high anti-spike IgG levels, 350 using an ELISA that has been shown to correlate well with neutralising antibody. [23][24][25] We 351 used a EUROIMMUN S/CO ratio of ≥6 for plasma to qualify for use in this trial, which is 352 substantially above the level of ≥3.5 that the US FDA recognises as high titre. 19 353 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 10, 2021. ; https://doi.org/10.1101/2021.03.09.21252736 doi: medRxiv preprint Recipients received plasma from two different donors to increase the chance that at 354 least one contained higher levels of neutralising antibodies. 355 The presence of anti-SARS-CoV-2 antibodies in recipients prior to transfusion with 356 convalescent plasma has also been cited as a possible reason for a lack of effect of 357 convalescent plasma. 12 In this trial we found that around 38% of patients were 358 seronegative at randomisation and, although they had a markedly higher 28-day 359 mortality risk than seropositive patients, we did not observe a survival benefit from 360 convalescent plasma in these seronegative patients. There was, however, a suggestion 361 of small improvements (proportional risk reduction of about one tenth) in the probability 362 of successful discharge from hospital by day 28 and of progressing to invasive 363 mechanical ventilation or death in seronegative patients allocated to convalescent 364 plasma. The apparent heterogeneity in these secondary outcomes according to 365 serostatus should be interpreted with a great deal of caution however, not least because 366 (perhaps by chance or perhaps as a result of conscious or unconscious decisions about 367 who to collect a serological sample from) the seronegative convalescent plasma 368 recipients were slightly younger than the seronegative usual care group, whereas 369 seropositive convalescent plasma recipients were slightly older than the seropositive 370 usual care group (webtable 2). Due to the known strong effects of age on mortality risk, 371 even these minor age imbalances could have led to a spuriously lower relative relative 372 risk of death in the seronegative convalescent plasma recipients and a spuriously higher 373 relative risk of death in the seropositive convalescent plasma recipients. 374 It has also been suggested that antibody based therapies are likely to be most effective 375 in the early stages of COVID-19, when viral replication dominates. 26, We did not identify 376 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. convalescent plasma have been reported. 28 Consistent with this, we did not identify any 398 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 10, 2021. ; https://doi.org/10.1101/2021.03.09.21252736 doi: medRxiv preprint evidence of a differential effect of convalescent plasma prior to and after the emergence 399 of B.1.1.7 in the UK. 22 400 During an epidemic caused by a novel virus, convalescent plasma is an appealing 401 treatment as it may be available within weeks of the outbreak, long before other 402 targeted therapies are available. Consequenty, convalescent plasma has been widely 403 used for COVID-19 outside of clinical trials but, until now, there has been insufficient 404 evidence from randomised trials to reliably assess its safety and efficacy. 8 In 405

RECOVERY, the largest clinical trial of convalescent plasma for any infectious 406
indication, high-titre convalescent plasma did not improve survival or other prespecified 407 clinical outcomes. 408 409 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 10, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 10, 2021.

Declaration of interests 485
The authors have no conflict of interest or financial relationships relevant to the 486 submitted work to disclose. No form of payment was given to anyone to produce the trial materials are available online at www.recoverytrial.net. As described in the protocol, 496 the trial steering committee will facilitate the use of the trial data and approval will not be 497 unreasonably withheld. Deidentified participant data will be made available to bona fide 498 researchers registered with an appropriate institution within 3 months of publication. 499 However, the steering committee will need to be satisfied that any proposed publication 500 is of high quality, honours the commitments made to the trial patients in the consent 501 documentation and ethical approvals, and is compliant with relevant legal and 502 regulatory requirements (e.g. relating to data protection and privacy). The steering 503 committee will have the right to review and comment on any draft manuscripts prior to 504 publication. Data will be made available in line with the policy and procedures described 505 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. label randomised control trial. medRxiv 2020: 2020.11.25.20237883. 585 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 10, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 10, 2021. ;https://doi.org/10.1101https://doi.org/10. /2021   Data are n (%) or n/N (%). RR=rate ratio for the outcomes of 28-day mortality, hospital discharge, and successful cessation of invasive mechanical ventilation, and risk ratio for other outcomes. * Analyses exclude those on invasive mechanical ventilation at randomisation. † Analyses exclude those on invasive or non-invasive ventilation at randomisation. ‡ Analyses exclude those not receiving invasive mechanical ventilation at randomisation. § Analyses exclude those on renal replacement therapy at randomisation. 631 632 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 10, 2021. ; https://doi.org/10.1101/2021.03.09.21252736 doi: medRxiv preprint CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 10, 2021. ; https://doi.org/10.1101/2021.03.09.21252736 doi: medRxiv preprint Number eligible for randomisation to convalescent plasma n=13127 (81%)

Convalescent plasma unavailable (n=965 [6%]) and/or considered unsuitable (n=2764 [17%])
Total recruited * n=16287 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 10, 2021.  Figure 2: Effect of allocation to convalescent plasma on 28−day mortality . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 10, 2021. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 10, 2021. ; https://doi.org/10.1101/2021.03.09.21252736 doi: medRxiv preprint