Efficacy and Safety of COVID-19 Convalescent Plasma in Hospitalized Patients

Key Points Question Does COVID-19 convalescent plasma (CCP), compared with placebo, improve the clinical status of hospitalized patients with COVID-19 requiring noninvasive supplemental oxygen? Findings In this randomized clinical trial including 941 patients, based on the World Health Organization 11-point Ordinal Scale for Clinical Improvement, CCP did not benefit 468 participants randomized to CCP compared with 473 randomized to placebo from April 2020 to March 2021. However, in exploratory analyses, CCP appeared to benefit those enrolled from April to June 2020, the period when most participants received high-titer CCP and were not receiving remdesivir and corticosteroids at randomization. Meaning In this trial, CCP did not meet prespecified outcomes for efficacy, but high-titer CCP may have benefited hospitalized patients with COVID-19 early in the pandemic when other treatments were not in use, suggesting a heterogenous treatment effect over time.

F irst reported in December 2019, 1 the COVID-19 pandemic spread to the US, with an epicenter in New York City (NYC) resulting in 203 000 cases and 18 600 fatalities from March to June 2020. 2 The absence of effective therapies prompted COVID-19 convalescent plasma (CCP) use because of biological plausibility and historical success of convalescent plasma in prior pandemics [3][4][5] and randomized trials for diphtheria 6,7 and Argentine hemorrhagic fever. 8 Although early CCP treatment of hospitalized patients with COVID-19 reduced mortality in matched-control studies, [9][10][11][12] randomized clinical trials have yielded mixed results, reducing mortality in one study 13 but not others, [14][15][16][17][18][19] despite showing signals of efficacy in some subgroups.
On April 17, 2020, we initiated a randomized, doubleblind, placebo-controlled trial of CCP vs normal saline in hospitalized patients with COVID-19 in NYC and Long Island, New York, requiring noninvasive oxygen supplementation. When the spring 2020 COVID-19 wave abated in NYC sites, the trial expanded to other regions in the US and continued until March 15, 2021.

Trial Design and Oversight
CONTAIN COVID-19 was an investigator-initiated, multicenter, randomized, double-blind, placebo-controlled trial comparing CCP with normal saline in hospitalized patients with laboratory-confirmed COVID-19 who required noninvasive oxygen supplementation. Participants were enrolled from April 17, 2020, to March 15, 2021, at 21 hospitals at 7 centers in Manhattan, Bronx, Brooklyn, and Long Island, New York; New Haven, Connecticut; Miami, Florida; Houston and Tyler, Texas; Baltimore, Maryland; and Milwaukee, Wisconsin. The institutional review boards of each participating center approved the study. The New York University CONTAIN Coordinating Center and Data Safety Monitoring Board (DSMB) provided trial oversight. Patients or legally authorized representatives provided either written or witnessed oral informed consent for participation in accordance with institutional review boardapproved consent procedures. The trial protocol is available in Supplement 1.

Patient Population
Eligible patients were adults aged 18 years or older hospitalized for 3 days or less or with symptoms of respiratory illness for 7 days or less (to include patients with presumably early phases of disease) who required noninvasive oxygen supplementation and had a positive nasopharyngeal SARS-CoV-2 reverse-transcriptase polymerase-chain-reaction test. Exclusion criteria were receipt of pooled immunoglobulin in the preceding 30 days, contraindication to transfusion, invasive mechanical ventilation or extracorporeal membrane oxygenation, volume overload, considered unlikely to survive past 72 hours based on investigator assessment, and receipt of a COVID-19 vaccine (after vaccines were available). Patients whose clinical outcomes were deemed not assessable after hospital discharge were also excluded. Race and ethnicity data were obtained from entries in the medical record, as reported by the participants, using fixed categories. Race and ethnicity data were included to provide additional information about participants included in the study and the potential generalizability of the results.

Randomization and Risk Stratification
A centralized electronic system was used to randomly assign enrolled patients to receive CCP or placebo in a 1:1 ratio stratified by enrollment site and risk status using randomization block sizes of 4 and 6 to maintain balanced group sizes. Allocation was concealed. Patients, treating clinicians, trial personnel, and outcome assessors were blinded to group assignment. Patients were stratified as high or average risk for COVID-19 progression. High-risk participants were aged 60 years or older or younger than 60 years with at least 1 of the following criteria: chronic pulmonary or heart conditions, hypertension, chronic kidney disease, body mass index greater than or equal to 35 (calculated as weight in kilograms divided by height in meters squared), diabetes, or immunosuppression. 20 Average risk participants were younger than 60 years without any high-risk condition (Supplement 1 and eMethods in Supplement 2).

Trial Interventions
One unit of CCP (approximately 250 mL) was infused within 24 hours of randomization at a rate of less than or equal to 500 mL/h. From April 2020 to January 2021, participants at Montefiore Medical Center received CCP from donors who participated in the Montefiore COVID-19 convalescent plasma donor program. 21-23 Because CCP could not be transferred between institutions, all other sites used CCP from New York Blood Center donors with a reactive anti-SARS-CoV-2 antibody test on the SARS-CoV-2 Microsphere Immunoassay. 24 Criteria for high-titer CCP were not available in April 2020. From January 2021 onward, all sites used CCP qualified by the New York Blood Center as high titer by a signal to cutoff value greater than or equal to 12 on the Ortho-Clinical Diagnostics VITROS Anti-SARS-CoV-2 immunoglobulin G (IgG) platform. 25 Placebo recipients received normal saline of equivalent volume. The trial product was masked with an opaque covering to ensure blinding of treating clinicians, research staff, and participants. The CCP SARS-CoV-2 spike protein IgG titers were determined retrospectively (eMethods in Supplement 2).

Outcomes
The primary outcome was clinical status based on the participant scores on the 11-point WHO Ordinal Scale for Clinical Improvement (WHO scale) 26 14 days after randomization; the secondary outcome was clinical status on the scale 28 days after randomization. WHO scale scores range from 0 to 10, with 0 indicating uninfected and no viral RNA detected and 10 indicating dead. Mortality at 14 and 28 days after randomization was a tertiary outcome (eMethods in Supplement 2).

Subgroup Analyses
The following exploratory analyses were proposed in the protocol: (1) CCP and participant plasma SARS-CoV-2 Spike Protein binding antibody titer and neutralizing titer, (2) CCP and participant SARS-CoV-2 antibody profiles and functional assays, (3) rates, levels and duration of SARS-CoV-2 RNA in nasopharyngeal swabs, (4) SARS-CoV-2 variants, (5) clinical status at other visit days, mortality and rates of discharge, (6) lymphocytes, neutrophils, and cytokines, and (7) moderating effect of concomitant medications-including corticosteroids, remdesivir, and anticoagulants on CCP effects. Studies 2 through 6 are not reported because they have not been completed. Analyses 1 and 7 were prespecified as exploratory. We report CCP and participant plasma antibody titers (analysis 1) and effects of corticosteroids and remdesivir, which became standard of care during the study (analysis 7), because of their explanatory power and the insights they provide into the primary outcome. Prespecified subgroup analyses were conducted for the following characteristics at randomization: age, WHO score, symptom duration, concomitant medications, CCP SARS-CoV-2 titer, and pretransfusion plasma SARS-CoV-2 IgG serostatus. Post hoc analysis was conducted to evaluate treatment effects over time.
Adverse events were systematically collected between randomization and study end point, including occurrence of transfusion-related acute lung injury, transfusion-associated circulatory overload, and other allergic reactions.

Stopping the Trial
The DSMB conducted interim analyses every 2 to 4 weeks. The statistical analysis plan specified that the DSMB consider stopping the trial for success with P(cumulative adjusted odds ratio [cOR]<1) greater than or equal to 95% and P(cOR<0.8) greater than or equal to 50% (statistical analysis plan in Supplement 3). The stopping rules for harm and safety were defined, respectively, as P(OR>1) greater than or equal to 80% and P(OR adverse event >1) greater than or equal to 75% (statistical analysis plan in Supplement 3). There were no prespecified stopping criteria for futility. However, after reviewing data on 920 participants on March 12, 2021, the DSMB recommended ceasing enrollment on March 15, 2021, based on slowing recruitment, the need for rapid reporting, and a 0.2% probability that the study would meet criteria for success if enrollment continued to 1000 participants.

Statistical Analysis
The trial design used a bayesian approach based on continuous monitoring, allowing real-time decisions given the urgency to find effective treatment. There was no maximum sample size, but enrollment of 1000 participants was anticipated. We used a skeptical prior distribution, N(mean, 0; SD, 0.354) for the treatment effect to ensure a type I error rate less than 5% and conducted regular monitoring using bayesian techniques. Simulations based on prespecified criteria and found the type I error rate was less than 5%. Convergence of the bayesian models was confirmed through inspection of trace plots (eFigure 1 in Supplement 2). 29 COVID-19 convalescent plasma and placebo recipient WHO scores were compared, with the placebo group as the reference arm. Primary and secondary outcomes were analyzed with a bayesian proportional cumulative odds model with adjustment for the following prespecified covariates: age, sex, prerandomization WHO score, symptom duration, and the stratification variables: risk status (high vs average) and study site. We examined goodness-of-fit of the model and confirmed the proportional odds assumption (eTable 1, eFigure 2 in Supplement 2).
For the primary outcome, CCP efficacy was defined as a cOR less than 1 and clinically meaningful effects were defined as cORs less than 0.8. Trial success was defined by posterior probability distributions of the cOR (P[cOR]): high, greater than or equal to 95% for effectiveness, and moderate, greater than or equal to 50% for clinical meaningfulness. Between-group differences were reported using point estimates based on median, 95% credible intervals (CrI), and posterior probabilities drawn from the estimated posterior distribution.
Analyses were performed using R, version 4.0.3 (R Foundation for Statistical Computing) (statistical analysis plan in Supplement 3).

Primary and Secondary Outcomes
The primary (WHO scores on day 14) and secondary (WHO scores on day 28) outcomes, adjusted for prespecified covariates, did not meet prespecified definitions of efficacy ( Figure 2 and

Exploratory and Post Hoc Subgroup Analyses
As the trial neared completion, it was apparent there were differences in participant characteristics over time. Between April-June and July-September 2020, median participant age decreased (from 70 to 59 years), while increases were noted in symptom duration less than 7 days (from 43.5% to 73.5%), highrisk status (from 62% to 90%), remdesivir use (from 1% to 47%), and corticosteroid use (from 24% to 85%) (eTable 3 in Supplement 2). Thus, we conducted a post hoc analysis to assess heterogeneous treatment effects across time, analyzing the data by enrollment quarter (Q): Q2, April-June 2020; Q3, July-September 2020; Q4, October-December 2020; and Q5, January-March 2021. At day 28 We assessed heterogeneity in treatment effects based on remdesivir and/or corticosteroid use at randomization (eTable 8 and eTable 9 in Supplement 2). At day 14, the cOR      . Mortality appeared to be lower for CCP recipients who received high EC 50 CCP than placebo in Q2 (eFigure 8 in Supplement 2), but there were no significant associations between CCP EC 50 or neutralizing titer and clinical outcome after adjustment for covariates.

Pretreatment Participant Plasma SARS-CoV-2 Spike Protein IgG
Plasma SARS-CoV-2 IgG was present before randomization in 486 (66.8%) of 728 participants from whom samples were available. At day 28, mortality (WHO score of 10) was lower in 486-seropositive than 242-seronegative participants irrespective of treatment arm, and in seronegative CCP (14.4%) than placebo (17.9%) recipients, which did not meet the definition of efficacy (eTable 11, eFigure 7 in Supplement 2), but analysis was restricted by sample availability, particularly for Q2 (62 samples available, 170 randomized).

Discussion
The CONTAIN COVID-19 trial was initiated in April 2020 during the first pandemic wave in NYC and Long Island, expanded to other US sites in August 2020, and continued until March 2021, spanning 11 months during which COVID-19 care changed substantially. The primary outcome did not meet the prespecified definition for CCP efficacy. However, exploratory subgroup analyses revealed a possible benefit of CCP in Q2 (April-June 2020), when all participants were enrolled in NYC and Long Island, most received hightiter CCP, and most did not receive remdesivir and/or corticosteroids. These medications were incorporated into COVID-19 care after the corticosteroid results from the RECOVERY trial were reported in July 2020 30 and the US Food and Drug Administration issued an emergency use authorization for remdesivir in May 2020 followed by approval in October 2020. 31,32 Consistent with the ACTT-I 32 and RECOVERY 30 trial results, remdesivir and corticosteroids appeared to improve clinical status irrespective of treatment arm. However, in the CONTAIN COVID-19 trial, use of these medications at randomization resulted in heterogeneous treatment effects. At day 14, CCP use appeared to improve clinical status when only corticosteroids were in use, but there was no evidence of CCP benefit when remdesivir and corticosteroids were both in use, and those who received both may have done worse. Our trial can-not establish the effect of these medications on CCP efficacy; they were not randomized, the trial was not designed to investigate their effects, and the analyses were exploratory. Nonetheless, based on other trial results, interactions between CCP, corticosteroids, and remdesivir warrant further investigation. 13,17 A randomized clinical trial in which 81% of 223 participants received corticosteroids and 6% received remdesivir found a CCP mortality benefit. 13 However, the 11 558-participant RECOVERY trial, in which 93% of 5795 recipients of CCP received corticosteroids and 32% received remdesivir, did not find a CCP mortality benefit, although CCP recipients not receiving corticosteroids appeared less likely to be intubated or die than controls (18% vs 24%; P = .07). 17 Data for remdesivir were not reported. Further studies are needed to understand interactions between CCP, corticosteroids, and remdesivir.
We found no associations between clinical outcome and CCP EC 50 or neutralizing titer, or participant SARS-CoV-2 serostatus. Less than 15% of our cohort had cancer or other immunosuppressing conditions that are associated with an impaired SARS-CoV-2 antibody response. A benefit of CCP has been shown in these patients. 33 The largest CCP effect was in Q2, particularly at day 28, when its effect (P[cOR<1] = 93%) approached the prespecified bayesian definition of efficacy. Retrospective analysis showed the median Q2 CCP-neutralizing titer was greater than 1:160, which likely fulfilled criteria for high-titer CCP, 34,35 whereas the CCP that was used during Q3 to Q5 was likely not high titer. Recently aggregated randomized clinical trial data suggest high-titer CCP is necessary, although it may not be sufficient, to benefit hospitalized patients with COVID-19. 36 Clearly, there is a need for standardized platforms and thresholds to qualify CCP for use. Nonetheless, CCP may have had heterogeneous effects over time as viral variants changed in this population. The effect of SARS-CoV-2 variants on our results is unknown, but 60% of Q5 enrollments were at NYC sites when the alpha and iota variants predominated, 37 and surveillance data identified alpha and beta variants in Miami and alpha in Houston. 38 Although exploratory subgroup analyses suggested CCP may be beneficial in participants aged 65 years or older and those with less severe disease (WHO 5), the posterior probabilities of these findings exhibited considerable uncertainty. Nonetheless, consistent with these findings, other hospitalized patient studies identified a possible CCP benefit in older patients 13-15,17 and those with less severe disease. 9,14,17 Given the absence of overall CCP benefit in our trial and randomized clinical trials of hospitalized patients with severe to life-threatening disease, [14][15][16][17][18]36 it is possible that patients with less severe disease could benefit the most from CCP therapy. Further insight may come from the COMPILE cohort, which included patients not requiring oxygen (WHO 4). 39,40

Strengths and Limitations
Strengths of the trial include its multicenter, blinded nature and use of a placebo control; an 11-month enrollment period that provided insights into CCP efficacy as COVID-19 treatments were being developed; a highly diverse population that  allows for generalizability; and use of a bayesian statistical approach that allowed near real-time monitoring of accruing data. Limitations of the trial include that the primary outcome at day 14 was likely too early for a disease now known to have a prolonged course. Therefore, day 28 findings may be more important clinically. In addition, there were heterogeneous treatment effects over time, perhaps related to changing patient characteristics, treatment options, and other factors. Compared with Q3 to Q5, Q2 participants were older, most received CCP with a median neutralizing titer greater than 1:160 and were not receiving remdesivir or corticosteroids. COVID-19 convalescent plasma obtained in the NYC area was used in non-NY sites and may not have matched local viral species, 38,41 and emergence of SARS-CoV-2 variants, which were not studied, may have reduced CCP efficacy over time. Because most Q3 to Q5 participants received CCP with a neutralizing titer less than 1:160, more than 1 unit may have been beneficial. 18 Participants with shorter symptom duration had higher mortality and we may have inadvertently enrolled patients with more severe disease by using symptom duration as an inclusion criterion. Analysis of the association between serostatus and CCP efficacy, as done by others 42 was restricted by sample availability.

Conclusions
This placebo-controlled double-blind randomized clinical trial of use of CCP in hospitalized patients with COVID-19 requiring noninvasive oxygen supplementation did not meet the prespecified definition of CCP efficacy. However, a possible benefit of CCP was observed early in the pandemic when high-titer CCP was used and corticosteroids and remdesivir were not in use. This supports the concept that convalescent plasma may be a feasible treatment option at the beginning of a pandemic or when other therapies are not in use or available. Further investigation is needed to understand the effects of corticosteroids and remdesivir on CCP efficacy and establish thresholds for antibody quantity and function that are most likely to confer a benefit.