key: cord-0690963-bupgqcye authors: Stadler, E.; Chai, K. L.; Schlub, T. E.; Cromer, D.; Polizzotto, M. N.; Kent, S. J.; Skoetz, N.; Estcourt, L. J.; McQuilten, Z. K.; Wood, E. M.; Khoury, D. S.; Davenport, M. P. title: Determinants of passive antibody effectiveness in SARS-CoV-2 infection date: 2022-03-22 journal: nan DOI: 10.1101/2022.03.21.22272672 sha: f93e807ae04acba1debf2ec95250e2eaaff97756 doc_id: 690963 cord_uid: bupgqcye Neutralising antibodies are an important correlate of protection from SARS-CoV-2 infection. Multiple studies have investigated the effectiveness of passively administered antibodies (either monoclonal antibodies, convalescent plasma or hyperimmune immunoglobulin) in preventing acquisition of or improving the outcome of infection. Comparing the results between studies is challenging due to different study characteristics including disease stage, trial enrolment and outcome criteria, and different product factors, including administration of polyclonal or monoclonal antibody, and antibody targets and doses. Here we integrate data from 37 randomised controlled trials to investigate how the timing and dose of passive antibodies predicts protection from SARS-CoV-2 infection. We find that both prophylactic and early therapeutic administration (to symptomatic ambulant subjects) have significant efficacy in preventing infection or progression to hospitalisation respectively. However, we find that effectiveness of passive antibody therapy in preventing clinical progression is significantly reduced with administration at later clinical stages (p<0.0001). To compare the dose-response relationship between different treatments, we normalise the administered antibody dose to the predicted neutralisation titre (after dilution) compared to the mean titre observed in early convalescent subjects. We use a logistic model to analyse the dose-response curve of passive antibody administration in preventing progression from symptomatic infection to hospitalisation. We estimate a maximal protection from progression to hospitalisation of 70.2% (95% CI: 62.1 - 78.3%). The dose required to achieve 50% of the maximal effect (EC-50) for prevention of progression to hospitalisation was 0.19-fold (95% CI: 0.087 - 0.395) of the mean early convalescent titre. This suggests that for current monoclonal antibody regimes, doses between 7- and >1000-fold lower than currently used could still achieve around 90% of the current effectiveness (depending on the variant) and allow much more widespread use at lower cost. For convalescent plasma, most current doses are lower than required for high levels of protection. This work provides a framework for the rational design of future passive antibody prophylaxis and treatment strategies for COVID-19. Summary of included studies. We identified 21 RCTs evaluating convalescent plasma, four RCTs evaluating hIVIG and 12 primary studies of monoclonal antibodies (including eight different monoclonal antibodies or combinations). The studies varied in the protocol design (summarised in Table S1 and Table S2 ), including stage of infection at the time of treatment, primary outcome measures, as well as the volume and antibody titres of the plasma administered. Thus, we grouped studies into broad categories by treatment (either CP/hIVIG or mAb), disease stage at enrolment, and outcomes reported (Fig. 1) . All studies evaluating hIVIG and the majority (17/21) of CP studies included hospitalised participants with moderate and/or severe disease, and analysed effectiveness in preventing invasive mechanical ventilation or death. Combining the results across CP and hIVIG studies the weighted mean effectiveness in preventing ventilation and death were 1.9% and 2.7% respectively ( Table S3) . Three CP studies recruited ambulatory patients with mild disease and studied the effectiveness of treatment in preventing progression to hospitalisation. These studies showed a higher weighted mean effectiveness after plasma administration of 31.4% (95% CI: 11.3 to 47.2%). Three studies evaluated monoclonal antibody treatment of hospitalised subjects for protection against progression to ventilation / death (again, using different definitions). None of these studies showed significant protection in the primary population, and the weighted mean protection from progression to ventilation was -51.7% and to death was 5.4% (see Table S3 ). However, one study reported significant efficacy (18.3% (95% CI: 8.3 -27.2)) for subjects who were seronegative at enrolment 16 . Five studies reported monoclonal antibody administration to ambulatory COVID-19 infected subjects and analysed efficacy in preventing progression to hospitalisation (using different enrolment criteria and outcome criteria, summarised in Table S1 ). Across these studies a mean protection of 70.0% (95% CI: 60.5 -77.5%) was observed. Interestingly, two studies of post-exposure prophylaxis administered to close contacts showed efficacies of 42.1% and 82.3%. One of these studies also reported an efficacy in preventing infection beyond one week after administration of antibodies of 92.8% (post-hoc analysis) 17 , which we assume was effectively true prophylaxis (since exposure most likely occurred after treatment). Finally, one study investigating true prophylaxis showed an efficacy of 92.4% in preventing symptomatic SARS-CoV-2 infection (preprint publication 18 ). Effects of timing on antibody effectiveness. Despite the significant heterogeneity in trial design, clear patterns of decreasing efficacy with disease stage at treatment emerge (see Fig. 1C for the efficacy of mAb treatment (orange) and CP or hIVIG treatment (blue) by treatment and outcome stage). We tested this relationship using a generalised linear mixedeffects model with binomial error family and logarithmic link function and found that treatment at later disease stages (as a numerical variable) was significantly associated with decreasing efficacy at preventing progression to the next stage (Table S10 and Table S11 ), for both CP/hIVIG studies (for treated patients, the relative risk of progression increases by 1.42-fold per disease stage, 95% CI: 1.09-1.86, p=0.0089) and mAbs (relative risk 1.96, 95% CI: 1.70-2.28, p<0.0001). However, we note that this analysis aggregated studies with different disease outcomes, so we further tested this relationship with data stratified by outcome (i.e. for a given clinical outcome, testing whether earlier treatment was more effective). We found that for the mAb studies treatment at an earlier disease stage had significantly higher efficacy at preventing progression to all the disease outcomes tested All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in Table S10 ). For CP/hIVIG studies this association was not significant, likely due to small numbers (p=0.29, 5/330 and 9/328 ventilations in the treatment and control groups, respectively and p=0.45, 17/998 and 16/990 deaths in the treatment and control groups, respectively) ( Table S11 and Table S12 ). Thus, treatment either prophylactically or early in the course of infection is a major determinant to achieving protection with passive antibody administration in COVID-19. Dose-response for passive antibody administration. The dose of monoclonal or polyclonal antibody preparations that are required with passive antibody therapies or prophylaxis to prevent progression from mild to severe disease is not clear. This is largely because of the challenge of comparing the relative doses used in different studies, where the antibody products had different potencies and were given at different doses. Although one study analysed in vivo neutralisation capacity after treatment 19 , which provides a good surrogate measure of protection 1 , this was not routinely performed across the studies. We have previously compared vaccine-induced neutralising antibody titres by reference to the '(geometric) mean convalescent titre' seen in the first months after infection with the ancestral virus 1 . Therefore, we investigated whether a similar metric might be used to understand neutralising titres after passive antibody administration (see methods). On this 'convalescent equivalence' scale, a group of individuals with a mean neutralising titre of 1fold of convalescence has (on average) the same level of neutralising antibodies as the average convalescent individual (against ancestral virus after infection with ancestral virus). We find that most mAb studies administered doses of antibodies that would be equivalent to >100-fold the average neutralising titre observed in convalescent individuals ( Fig. 2 and Table S7 ). We have previously estimated that a mean level of neutralising antibody of around 20% of the convalescent titre provides 50% protection from acquisition of symptomatic SARS-CoV-2 infection (with the ancestral variant) 1 . Therefore, we first investigated whether the results of the two studies of prophylactic administration of monoclonal antibodies were consistent with this. Fig. 2A shows the relationship between neutralisation and protection following infection or vaccination with different agents (data and model from Khoury et al and Cromer et al 1, 2, 20 , in grey and black respectively), and the range of antibody levels expected after boosting with an mRNA vaccine 2 (shaded blue). The administered dose of monoclonal antibody is shown on the horizontal axis (in fold-convalescent) and it is clear that this dose is significantly higher than the levels achieved even after vaccination and boosting ( Fig. 2A) . However, the level of protection observed after high-dose passive antibody administration was similar to that seen after mRNA vaccination. This could be because the in vivo neutralising antibody titres were lower than might be predicted from the administered dose. Alternatively, this perhaps suggests a maximum level of protection from symptomatic SARS-CoV-2 infection that can be achieved by neutralising antibodies, which could not be estimated from the vaccine studies. To explore the antibody dose-response relationship for preventing progression from symptomatic infection to hospitalisation, we aggregated all six studies of monoclonal antibody and four studies of plasma administration to ambulant symptomatic subjects to study protection from progression to hospitalisation (Fig. 1) . Two studies were excluded from this analysis as they did not report hospitalisation as an outcome (one monoclonal antibody study and one convalescent plasma study that both treated symptomatic patients All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in and reported deaths as outcome). In Fig. 2 we plot the 'convalescent equivalent' of different administered doses of mAb and convalescent plasma against the observed protection. We fit the relationship between administered dose and protection from hospitalisation using a logistic model with a maximal efficacy (see methods and supplementary methods). We estimate a maximum protection of 70.2% (regardless of dose) and the administered dose to achieve 50% of this maximal effect (EC-50 for hospitalisation, equivalent to 35.1% protection overall) is 0.185-fold (95% CI: 0.087 -0.395) the mean convalescent serum titre (Extended Data Fig. 3 and Table S6 ). To ensure this relationship was not overly affected by the results of any single study, we performed a sensitivity analysis using a 'leave-one-out' approach, where we systematically refitted the curve omitting the results for one study (Extended Data Fig. 4 ). This showed that the estimates of the EC-50 ranged between 0.069and 0.38-fold convalescent which agrees well with the 95% CI of the fit to all data (0.087 -0.395). Using this curve, we can also estimate the minimum dose required to achieve at least 90% of the maximum efficacy in preventing progression from ambulatory COVID-19 to hospitalisation (EC-90, noting that the estimated maximum protection is only 70.2%, so the EC-90 is equivalent to 63.2% protection). The estimated EC-90 is 0.904-fold (95% CI: 0.208 -17.803) of the mean convalescent titre, which is considerably lower than the level administered in most monoclonal antibody studies (see Table S7 with conversion to dose in milligrams for the relevant antibodies). Visual inspection of the dose-response curve for monoclonal antibodies in preventing progression from symptomatic disease to hospitalisation (Fig. 2B) suggests a potential decrease in efficacy at higher doses. Indeed, a Spearman correlation of the log10transformed dose versus efficacy suggests a significant association between increasing administered monoclonal antibody dose and reduced efficacy ( = -0.95, p < 0.0001). However, the data contains considerable uncertainty in the estimated efficacy (as indicated by the vertical 95% CIs for the estimated efficacy in Fig. 2B) , as well as the fact that several points come from the same study with antibodies at different doses, and some studies enrolled either high-risk or low-risk subjects (different shape and line sizes in Fig. 2B ). Using a generalised linear mixed-effects model (with binomial error family and logarithmic link function) incorporating these factors, we find a strong trend towards decreasing efficacy with higher administered doses, but this is not significant (p=0.054, Table S13). As described above, studies of passive antibody administration to hospitalised patients showed low or no efficacy overall. This was the case even with administration of high doses of monoclonal antibodies 16, 21, 22 , in a similar range of doses to those which were effective at earlier stages of infection ( Fig. 2B and Extended Data Fig. 5 ). Using lower doses of mAbs can avert more hospitalisations when products are limited. To date, the availability of mAbs has been limited, and thus we explored the predicted hospitalisations that can be averted with lower doses of mAbs than those used in the studies so far. One strategy to allow treatment of more people is to give less compound per person, and this may be a viable strategy given that most studies of mAbs have used doses much higher than the EC-90 ( Fig. 2B and Table S7 ). However, at lower doses the antibodies will be less effective for each treated individual. Thus, for each antibody included in the trials analysed in this study, we analysed the lowest dose used and predicted the foldincrease in the hospitalisations averted by giving less of the compound per person and distributing the total dose amongst more individuals (supplementary methods). We find All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted March 22, 2022. ; https://doi.org/10.1101/2022.03.21.22272672 doi: medRxiv preprint that decreasing the dose of each antibody to the EC-90 level, would avert between 7.1-and 710.1-fold more hospitalisations ( Fig. 3 and Table S8 ). Lower doses could avert even more cases with a maximum effect at a dose of 0.094-fold convalescent (less than 7mg for each considered mAb, Table S8 ), but such a low dose is estimated to provide only 19.6% protection for treated individuals and thus would only be optimal when product availability is exceptionally low compared with the number of cases requiring treatment. Reducing administered doses below 0.094-fold of the mean convalescent titre equivalent allows distribution to more individuals but is predicted to result in more hospitalisations due to the low efficacy beginning to be a limiting factor. Passive antibody therapy in the context of SARS-CoV-2 variants. The analysis above compares antibodies by reference to their neutralisation titre against ancestral (Wuhan-like) virus. However, both convalescent serum and monoclonal antibodies show a significant loss of recognition of different SARS-CoV-2 variants, especially the Omicron variant 2,23,24 . From our analysis above of the relationship between dose on the 'convalescent equivalence' scale and efficacy for preventing disease progression, we can translate the loss of neutralisation to predict the dose required for each antibody to achieve the EC-90 against Omicron (both for BA.1 and BA.2 subvariants, Table S9 ). We identified three studies that examined the loss of neutralisation of the relevant monoclonal antibodies against Omicron BA.1 [24] [25] [26] . All mAbs used in the RCTs included in this analysis lost all detectable neutralisation to Omicron BA.1 (experiencing a drop in neutralisation titres that translated to at least a 1000-fold increase in IC-50, Table S9 ), except for sotrovimab which was estimated to have only a 7.8-fold increase in IC-50 against Omicron BA.1 compared with ancestral virus (geometric mean of three studies). This means that while other mAbs are unlikely to confer prophylactic or therapeutic protection against Omicron BA.1, as the EC-50 for protection against BA.1 is dramatically higher than the currently administered doses, sotrovimab is likely to remain protective at its current dose. Sotrovimab is predicted to give 87.8% protection from symptomatic Omicron BA.1 infection and 63.2% protection against progression from symptomatic disease to hospitalisation (at 500 mg, the dose used in the COMET-ICE 27 and ACTIV-3 21 trials) assuming the relationship between neutralisation and protection for mAbs follows the same trend as in vaccination, which may not be the case if factors other than antibodies are important for protection 1 . More recently the Omicron BA.2 variant has increased in prevalence. Initial studies show that the cross-reactivity of the various monoclonals to this sublineage differs to that observed for BA.1 (Table S9) . For example, sotrovimab has been reported to have 77-fold increase in its IC-50 against Omicron BA.2 25,28 (instead of 7.8-fold for Omicron BA.1). This means that to provide an equivalent level of protection against Omicron BA.2, an approximately ten times higher dose of sotrovimab would be needed compared to what was required to protect against Omicron BA.1. The current dose of sotrovimab (500 mg) is predicted to provide only 19.3% protection against progression from symptomatic Omicron BA.2 infection to hospitalisation. In addition, our modelling predicts that a high dose of imdevimab could have some clinical utility against Omicron BA.2. Since imdevimab was administered at high doses in the trials, the highest dose of imdevimab tested (4000 mg) is still predicted to provide 60.1% protection against progression from symptomatic infection to hospitalisation for Omicron BA.2, however clinical trials would be required to confirm its efficacy in clinical settings. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in For convalescent or post-vaccination serum, the in vitro neutralisation titre for either subvariant of Omicron (compared to the ancestral virus) has been estimated to be around 10-to 20-fold 23,29 higher than for ancestral virus. Thus, for historical convalescent plasma samples collected after infection with ancestral virus, high plasma volumes or high plasma titres would be necessary. However, vaccination of recovered individuals has been shown to raise antibody levels to around 10-fold higher than infection alone 2 . In addition, convalescent plasma from individuals infected with Omicron (or vaccinated with a variantspecific vaccine) may be expected to be more specific to the Omicron variants. Thus, studies are needed to assess the variant-specific titre of different CP or hIVIG pools against future SARS-CoV-2 variants to predict their effectiveness. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in Neutralising antibodies have been shown to be a correlate of vaccine-mediated protection from acquisition of symptomatic SARS-CoV-2 infection, and passive antibody studies suggest they can also play a major role in reducing the severity of disease 1, 3, 6, 17, 18 . Comparing the available data from RCTs of passive antibody therapy, we find that both prophylactic therapy and treatment in the early stages of symptomatic infection can achieve significant protection from infection or hospitalisation, respectively, and that the antibody doses required to achieve this protection are up to 1000-fold lower than many current treatment regimes. An important prediction of our analysis is that the dose of monoclonal antibody required to prevent progression to severe SARS-CoV-2 infection is likely significantly lower than that administered in most current treatment schedules (Table S7 ). This result is robust to the exclusion of data from any one particular study (Extended Data Fig. 4) and suggests that studies of lower dose administration may be valuable. If the efficacy of low-dose antibodies can be demonstrated, this would significantly reduce costs and may facilitate alternative modes of administration. In addition, it suggests that higher doses of convalescent plasma (or selection of higher titre plasma) may provide additional benefit if administered to ambulant subjects, especially if mAb neutralising activity is reduced with the emergence of new variants. We also find that the timing of antibody administration is a major factor in determining the success of treatment. For example, for the same antibody doses we see 92.4% protection from infection when given as prophylaxis 18 , but only 73.5% protection from progression to hospitalisation when given to ambulant COVID-19 patients 6 , and negligible protection when given to hospitalised subjects 16 . One explanation for reduced effectiveness with later disease stage might be that higher antibody levels are required for protection in later disease. However, our analysis suggests that the neutralising antibody titre for 50% of the maximal protection appears very similar for preventing acquisition of infection and preventing progression from ambulant COVID-19 to hospitalisation. Moreover, it appears there is a 'ceiling' of effectiveness for therapeutic treatment of COVID-19, where the maximum effectiveness of passive antibody administration appears to decrease with disease stage. This ceiling suggests that as time progresses there is a reduced proportion of individuals in which antibodies can affect subsequent outcome. This seems likely to arise from two potential factors. Firstly, it may be that passive antibodies appear less effective with time simply because endogenous antibody responses are already present. The level of antibodies required for 90% protection from progression is around 0.90-fold of the mean convalescent titre. Thus, as endogenous antibody responses rise over time during infection, they may reach protective levels where the addition of passive antibodies may provide minimal extra protection. This interpretation is supported by studies showing higher treatment effectiveness in seronegative subjects 16 . Alternatively, it may be that as the pathological process of infection proceeds it reaches a stage where antibodies can no longer affect outcome. This interpretation is supported by studies of early antiviral treatment 30,31 , despite a very different mode of action to passive antibodies. However, since the progression of the pathological process and the rise in endogenous antibodies may be linked in time, this does not separate these alternatives. A major question is whether treatment earlier after symptom onset might produce better outcomes. For example, would treatment on day 1 versus day 5 post onset of symptoms have higher effectiveness? Although the five studies of monoclonal antibody treatment of All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in Fig. 2) . If the absence of an endogenous antibody response is important to passive antibody efficacy, it may be that the level of endogenous antibody response rather than time since symptom onset is more predictive of outcome. Studies are urgently needed to identify the optimal time for passive antibody treatment in symptomatic subjects. Previous studies have identified the antibody levels providing protection from COVID-19 after vaccination and suggest that a titre equivalent to 0.2-fold of the average early convalescent titre provides 50% protection from acquiring symptomatic SARS-CoV-2 1 . Here we estimate the level of antibody required to give 50% of the maximal protection from progression from symptomatic to severe infection and find an administered dose equivalent to 0.19-fold the mean convalescent titre. This is remarkably similar to the neutralisation level associated with protection from symptomatic infection in the vaccine studies and suggests that the major difference between prophylaxis and early therapeutic intervention is not in the level of antibody required, but rather the maximum protection that can be achieved. This study has a number of limitations. Firstly, it aggregates studies using different therapeutics and with different enrolment and outcome criteria. Secondly, it tries to equate administered doses of convalescent plasma and monoclonal antibodies based on a single study comparing in vitro neutralisation of pseudovirus 32 . In the case of convalescent plasma, we consider mean titre of donor plasma and mean plasma volume for dilution, which does not reflect the considerable variability that exists between individual donor plasma neutralisation titres and individual recipients' plasma volumes. In addition, we consider only the 'administered dose', as the studies did not directly measure plasma neutralisation titres in recipients after administration. Gordon et al 33 studied the pharmacokinetics of convalescent plasma after administration of a volume of 5 ml/kg to infants (leading to an estimated dilution of approximately 10-fold 34 ). Direct measurement of recipient titres 30 minutes after infusion found a mean of 6.2% of donor titres, suggesting a decrease in titre of around 40% compared to what would be predicted by dilution alone. Thus, there may be a tendency for the 'administered dose' to be higher than a serum neutralisation level that might be directly measured in vivo after treatment. Despite these limitations, a major strength of our approach is to apply a rigorous quantitative analysis to the available data on passive antibody treatment for SARS-CoV-2. A major factor that will affect future use of antibody therapies is the recognition of SARS-CoV-2 variants with different degrees of immune escape from neutralising antibodies. Studies of immune protection from vaccination suggest that neutralising antibody levels remain predictive of immune protection once the drop in titre to the variants is taken into account 2,23 . Recent studies have shown that many of the monoclonal antibodies show a >1000 fold drop in titre to the Omicron variant 26 , and similarly neutralising antibody titres from subjects that have been previously vaccinated or infected and were boosted by vaccination also drop by at least 10-fold (responses in non-boosted individuals typically drop below detection) 23, 29 . This drop in neutralising titre would be expected to have a major effect on the potency and effectiveness of passive antibody therapy. A number of other antibodies have shown smaller drops in neutralisation titre (10-to 30-fold) to the Omicron BA.1 and BA.2 variants 26 . Consideration of the dose-response curve suggests that these may still retain potency at current dosing levels ( Fig. 2 and Table S9) . Clinical studies to confirm that the dose-response curve is predictive of protection against VOC are urgently required. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in However, it is hoped that, with appropriate consideration of the neutralisation potency of different antibody-VOC combinations, effective dosing regimens may be found that can provide protection from SARS-CoV-2 variants. Interestingly, the two monoclonal antibodies in our analysis that were predicted to have preserved effectiveness against either BA.1 or BA.2, i.e. sotrovimab and imdevimab, were the two least potent antibodies with the highest IC-50 against ancestral virus 32 (Table S4) . This result may reflect that high avidity is often accompanied by high specificity 35 , and thus the most potent antibodies against ancestral virus may have reduced recognition of variants. It is interesting to consider that most existing monoclonals in clinical use have been selected because of their high avidity, but some existing monoclonals with lower avidity and alternative specificities may be more broadly cross-reactive. Despite the availability of effective vaccines for COVID-19, a significant population of elderly or immunosuppressed individuals may be unable to fully benefit from existing vaccines 36 . Passive antibody therapy has the potential to be used either prophylactically or therapeutically in this population. Using current dosing strategies, the costs and difficulties in administering large doses of monoclonal antibodies may severely limit the use of these agents. This work provides quantitative and testable predictions of how passive antibody therapy may be optimally deployed to benefit a larger number of subjects. Further work is clearly required to better understand the relationship between administered dose and in vivo neutralising antibody titres, the impact of different SARS-CoV-2 variants on antibody efficacy, and to directly test the efficacy of early, low-dose passive antibody administration. This work provides a quantitative framework to help guide the rational testing and deployment of this important therapeutic modality. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in (Table S8) . All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in Search strategy and selection criteria. We performed a systematic search of the literature to identify randomised controlled trials of passive antibody therapy (including SARS-CoV-2 neutralising mAb, CP and hIVIG) for prevention and treatment of COVID-19. We included studies of individuals with a confirmed diagnosis of COVID-19 and studies of individuals without confirmed diagnosis of COVID-19, including pre-exposure, and peri-(post)-exposure stages. We included the following interventions: • SARS-CoV-2 neutralising mAb • Convalescent plasma • HIVIG We excluded studies evaluating standard immunoglobulin and mAbs that were not specifically designed to target SARS-CoV-2. We included the following outcomes in all participants: • All-cause mortality at day 30 or last follow up • Need for invasive mechanical ventilation We included the additional outcomes in participants with asymptomatic or mild disease: • Admission to hospital We included the additional outcomes in participants at the pre-exposure, peri-exposure or post-exposure stage without confirmed diagnosis of COVID-19: • Infection • Symptomatic infection • Admission to hospital We explored subgroups, including low vs. high titre CP and serostatus of patients at baseline (seronegative vs. seropositive). We identified studies from published search strategies conducted by the Cochrane Haematology living systematic review teams which searched the following databases -MEDLINE, Embase, the Cochrane COVID-19 Study Register, Pubmed, Epistemonikos L*OVE List Coronavirus disease, World Health Organization COVID-19 Global literature on coronavirus disease and trial registry platforms, including ClinicalTrials.gov, WHO International Clinical Trials Registry Platform (ICTRP) and medRxiv (see references 8, 10 ) . Detailed information about the search strategies used are available in published versions of the Cochrane living systematic review appendices. Additional recent studies were identified from review articles, monitoring of ongoing trials and ongoing updated searches until 7 January 2022. Data collection. Data for analysis was extracted from included studies into a spreadsheet, including dose and antibody levels of antibody administered, disease stage according to the latest WHO clinical progression scale, and timing of administration, primary outcome of study and whether they reported on our prespecified outcomes of interest, which include protection against symptomatic infection, hospitalisation, need for invasive mechanical ventilation (IMV) and death (all-cause mortality at 30 days) (summarised in Table S1 and Table S2 ). We classified disease stage into the following categories -uninfected (preexposure, peri-(post-)exposure) or infected (symptomatic infection or hospitalised with moderate or severe disease). We also collected if studies reported results for seronegativity/seropositivity of recipients at baseline. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted March 22, 2022. ; https://doi.org/10.1101/2022.03.21.22272672 doi: medRxiv preprint Data analysis. Efficacy and 95% confidence interval. Studies were grouped according to product used and stage at enrolment. We calculated the efficacy of preventing progression from one infection stage (including pre-and peri-(post-) exposure) to another stage using reported numbers of patients with disease progression and total numbers of patients in the following way: 1 − number events in treatment group/number of patients in treatment group number events in control group/number of patients in control group . Thus, efficacy is 1-"relative risk of progression". We transformed the efficacy to percent and computed the corresponding 95% CI for efficacy using where and denote the number of events and total number of patients in the treatment group and and denote the number of events and total number of patients in the control group, respectively. Statistical analysis. We used a generalised linear mixed-effects model with binomial error family and logarithmic link function (using R, version 3.6.0 37 , and the glmer function from the lme4 package 38 ). The model includes random intercepts for different trials to account for variability between different trials. We used this model to pool data from different studies and find the average efficacy of mAb or CP/hIVIG treatment for various stage transitions ( Fig. 1C and Table S3 , see supplementary methods for more details). We also used this model to understand the relationship between infection stage at treatment and efficacy in preventing progression to a subsequent disease stage (including prophylactic treatment and prevention of symptomatic infection) and the influence of the infection stage at treatment on the efficacy of treatment for each of the reported outcomes symptomatic infection, hospitalisation, ventilation, and death separately (Table S10, Table S11 , and supplementary methods). Moreover, we investigated whether the efficacy of preventing hospitalisations decreases by administered mAb dose (Fig. 2B) using the same type of generalised linear mixed-effects model and taking variability in patients' risk for progression to severe disease into account (Table S13) . Normalising plasma/mAb dose to convalescence titre equivalent. Previous analyses have compared neutralisation titre after vaccination by normalising to the 'mean early convalescent titre' (after COVID-19) 1,39 . For convalescent plasma treatment, if unselected convalescent plasma is used, we might consider the product itself to have a mean titre of '1fold' of the average convalescent serum titre, and the average 'titre administered' to reflect simply the plasma dilution upon administration (e.g., if a volume of 250mL of unselected convalescent plasma were administered to a recipient with 3L plasma volume, we could estimate that the equivalent of 0.083-fold the mean convalescent titre was administered). Importantly we considered the mean administered dose, not implying the neutralisation titre of recipient plasma (if measured directly) would necessarily reflect this, since this estimate ignores any potential loss of product or loss of potency after administration. Similarly, some convalescent plasma studies specified that high titre plasma were selected; where this was the case, we normalised using the available data (Table S5 and supplementary methods) . For this analysis, we focused on trials that treated outpatients All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in 32 . This provides a metric for equating convalescent serum and the equivalent in vitro concentration of a monoclonal antibody required to achieve the same neutralisation. If we again assume a mean plasma volume of 3 litres, we can estimate the 'convalescent neutralisation equivalent' for different administered doses of monoclonal antibody ( Table S4 , see supplementary methods). Importantly for both convalescent plasma and monoclonal antibodies we assume 100% distribution of the antibody in plasma, ignoring any potential loss during injection or thereafter. If there is a loss of antibody during infusion or thereafter this would lead to a tendency for reported administered titres to be higher than the serum titre that might be directly measured in vivo after infusion. Dose-response curve fitting. We fitted a logistic efficacy function to both the efficacy for preventing hospitalisation after treatment of symptomatic patients ( Fig. 2B and Extended Data Fig. 3 ) and to the efficacy for preventing deaths after mAb treatment of hospitalised patients (Extended Data Fig. 5 ) by administered dose in fold convalescent. The parameters of the efficacy functions were the maximal efficacy, the dose at which the half-maximal efficacy is achieved (EC-50 dose), and a slope parameter. We used a maximum likelihood approach with a likelihood function based on count data from individual studies to estimate these parameters (Table S6, Table S14 , and supplementary methods). The 95% confidence region for the model fit and the 95% confidence intervals for the EC-50 and the EC-90 dose were calculated using parametric bootstrapping in R (version 3.6.0) 37 with the rmvnorm function from the mvtnorm package 42,43 . The sensitivity of the dose-response curve for prevention of hospitalisations was investigated by a leave-one-out analysis (Extended Data Fig. 4 and supplementary methods) . All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in All data and code will be made available on GitHub upon publication. The funding source had no role in the writing of the manuscript or the decision to submit it for publication, nor in data collection, analysis, or interpretation; or any aspect pertinent to the study. Authors had full access to data in the study, and they accept responsibility to submit for publication. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in Fig. 1 Studies searched, selected, and included in the analysis. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted March 22, 2022. ; https://doi.org/10.1101/2022.03.21.22272672 doi: medRxiv preprint Extended Data Fig. 2 Efficacy of therapeutic treatment by time from symptom onset to treatment. For therapeutic treatment, i.e., treatment of symptomatic patients and prevention of hospitalisations, the efficacy of treatment is shown for different median times between symptom onset and treatment. Vertical bars denote 95% CIs for the efficacy of preventing hospitalisations and horizontal lines indicate the range of treatment days (for "BLAZE-1 (bamlanivimab + etesevimab)", "COMET-ICE (sotrovimab)", and "Sullivan") or interquartile range of treatment days (for "Eom (regdanvimab)" and "Weinreich (casirivimab + imdevimab)") timed from symptom onset. Note that "COMET-ICE (sotrovimab)" did not report a median time since symptom onset but a range of 3 to 5 days, we thus used a median of 4 days for visualisation. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted March 22, 2022. ; https://doi.org/10.1101/2022.03.21.22272672 doi: medRxiv preprint Extended Data Fig. 3 The dose response curve for treatment of symptomatic ambulant subjects and efficacy in preventing progression to hospitalisation with maximal efficacy, EC-50, and EC-90 indicated. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in Fig. 4 Leave-one-out analysis for the dose-response curve for preventing hospitalisation. We investigated the sensitivity of the three parameters for the logistic efficacy function, the EC50 dose (A), maximal efficacy (B), and slope parameter (C), by fitting the dose-response curve to the data excluding one study at a time. The parameter estimates with one omitted study agree well with the 95% CIs from the fit to all data, with only the EC-50 dose after omitting the Korley study 40 and the slope parameter after omitting the Libster study 19 as outliers. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted March 22, 2022. ; https://doi.org/10.1101/2022.03.21.22272672 doi: medRxiv preprint Extended Data Fig. 5 Dose-response curve for treatment of hospitalised subjects and prevention of death. No studies show significant efficacy for the treated population despite high antibody levels. Only the subgroup of sero-negative patients in the RECOVERY trial showed significant efficacy (18.3%, 95% CI: 8.3-27.2%). Apart from the sero-negative patients in the RECOVERY trial, only the patients with unknown sero-status at baseline in the RECOVERY had an estimated efficacy that was positive (1.1%, 95% CI: -19.7 to 18.3%). For all other subgroups and studies, efficacy was negative and thus only the 95% CI for efficacy is shown. The shaded area indicates the 95% confidence region for the fitted dose-response curve. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in Neutralizing antibody levels are highly predictive of immune protection from symptomatic SARS-CoV-2 infection Neutralising antibody titres as predictors of protection against SARS-CoV-2 variants and the impact of boosting: a meta-analysis Immune correlates analysis of the mRNA-1273 COVID-19 vaccine efficacy clinical trial Convalescent plasma in patients admitted to hospital with COVID-19 (RECOVERY): a randomised controlled, open-label, platform trial Writing Committee for the REMAP-CAP Investigators, et al. Effect of Convalescent Plasma on Organ Support-Free Days in Critically Ill Patients With COVID-19: A Randomized Clinical Trial REGEN-COV Antibody Combination and Outcomes in Outpatients with Covid-19 Bamlanivimab plus Etesevimab in Mild or Moderate Covid-19 Convalescent plasma or hyperimmune immunoglobulin for people with COVID-19: a living systematic review How should we use convalescent plasma therapies for the management of COVID-19? SARS-CoV-2-neutralising monoclonal antibodies for treatment of COVID-19 Association of Convalescent Plasma Treatment With Clinical Outcomes in Patients With COVID-19: A Systematic Review and Meta-analysis The Effect of Convalescent Plasma Therapy on Mortality Among Patients With COVID-19: Systematic Review and Meta-analysis Association between convalescent plasma treatment and mortality in COVID-19: a collaborative systematic review and meta-analysis of randomized clinical trials Antibody and cellular therapies for treatment of covid-19: a living systematic review and network meta-analysis National COVID-19 Clinical Evidence Taskforce Casirivimab and imdevimab in patients admitted to hospital with COVID-19 (RECOVERY): a randomised, controlled, open-label, platform trial Subcutaneous REGEN-COV Antibody Combination to Prevent Covid-19 Repeat Subcutaneous Administration of REGEN-COV in Adults is Well-Tolerated and Prevents the Occurrence of COVID-19 Early High-Titer Plasma Therapy to Prevent Severe Covid-19 in Older Adults Relating in vitro neutralisation level and protection in the CVnCoV (CUREVAC) trial Efficacy and safety of two neutralising monoclonal antibody therapies, sotrovimab and BRII-196 plus BRII-198, for adults hospitalised with COVID-19 (TICO): a randomised controlled trial. The Lancet Infectious Diseases ACTIV-3/TICO LY-CoV555 Study Group. A Neutralizing Monoclonal Antibody for Hospitalized Patients with Covid-19 A meta-analysis of Early Results to predict Vaccine efficacy against Omicron An infectious SARS-CoV-2 B.1.1.529 Omicron virus escapes neutralization by therapeutic monoclonal antibodies Neutralization of SARS-CoV-2 Omicron BA.2 by Therapeutic Monoclonal Antibodies Antibody Evasion Properties of SARS-CoV-2 Omicron Sublineages Early Treatment for Covid-19 with SARS-CoV-2 Neutralizing Antibody Sotrovimab Efficacy of Antiviral Agents against the SARS-CoV-2 Omicron Subvariant BA.2 Omicron extensively but incompletely escapes Pfizer BNT162b2 neutralization Oral Nirmatrelvir for High-Risk, Nonhospitalized Adults with Covid-19 Molnupiravir for Oral Treatment of Covid-19 in Nonhospitalized Patients SARS-CoV-2 immune evasion by the B.1.427/B.1.429 variant of concern Pharmacokinetics of high-titer anti-SARS-CoV-2 human convalescent plasma in high-risk children Blood sample volumes in child health research: review of safe limits The Concept of Crossreactivity and Specificity in Immunology Efficacy of covid-19 vaccines in immunocompromised patients: systematic review and meta-analysis R: A language and environment for statistical computing. R Foundation for Statistical Computing Fitting Linear Mixed-Effects Models Using lme4 Evidence for antibody as a protective correlate for COVID-19 vaccines The authors thank all participants and original study teams of the studies analysed. This work would not have been possible without the data from these studies. We also thank David Sullivan for his help in understanding antibody titres measured in Sullivan et al 41 .We also thank the authors from the Cochrane living systematic review teams of 'Convalescent plasma or hyperimmune immunoglobulin for people with COVID-19: a living systematic review' and 'SARS-CoV-2-neutralising monoclonal antibodies for treatment of COVID-19', for their contribution in developing the search strategies and identification of the included trials. ES, KLC, DC, SJK, EMW, ZKM, DSK and MPD contributed to conceptualization, supervision, and resources. ES, KLC, DC, NS, LE, EMW, ZKM, DSK and MPD contributed to data curation, methodology, formal analysis, and visualization. All authors contributed to the writing and reviewed and approved the final report. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication. MNP declares receiving provision of drug for clinical trials from CSL Behring, Takeda, Grifols, Emergent Biosciences, and Gilead.