Association Between Administration of IL-6 Antagonists and Mortality Among Patients Hospitalized for COVID-19: A Meta-analysis | Critical Care Medicine | JAMA | JAMA Network
[Skip to Navigation]
Sign In
Figure 1.  Association Between IL-6 Antagonists vs Usual Care or Placebo and Primary Outcome of 28-Day All-Cause Mortality
Association Between IL-6 Antagonists vs Usual Care or Placebo and Primary Outcome of 28-Day All-Cause Mortality

The area of the data marker for each trial is proportional to its weight in the fixed-effects meta-analysis.

aCommon control group across both treatment comparisons of the COV-AID trial.

bNA indicates not available; there were insufficient data to estimate odds ratio.

cThe data for the PreToVid trial are based on events up until 30 days after randomization.

dThere were 21 patients in the control group for both treatment comparisons. The analyses have been adjusted to correct for this.

Figure 2.  Subgroup Analysis of 3 Outcomes by Treatment Group and Corticosteroid Use
Subgroup Analysis of 3 Outcomes by Treatment Group and Corticosteroid Use

For the outcome of progression to invasive mechanical ventilation (IMV), extracorporeal membrane oxygenation (ECMO), or death at 28 days, only patients not receiving IMV or ECMO at randomization were included in the analyses. All trials supplied data until 28 days after randomization, except for the PreToVid trial for which data are based on events up until 30 days after randomization. The ratios of odds ratios (ROR) compare the associations of anti–IL-6 agents with outcomes between patients receiving and not receiving corticosteroids within each trial. The displayed summary ROR for each comparison is based on a meta-analysis of trial-specific RORs. Only the trials that recruited patients receiving and not receiving corticosteroids at randomization contribute to these meta-analyses. The estimated RORs are not necessarily consistent with the ratios of the subgroup ORs (left panel), which were calculated from meta-analyses within subgroups defined by receipt or no receipt of corticosteroids at randomization.

aThe odds ratios are based on raw percentages while the ratios of odds ratios account for pooling trials of different sizes.

Figure 3.  Association Between IL-6 Antagonists vs Usual Care or Placebo and Secondary Outcome of Progression to Invasive Mechanical Ventilation, Extracorporeal Membrane Oxygenation, or Death
Association Between IL-6 Antagonists vs Usual Care or Placebo and Secondary Outcome of Progression to Invasive Mechanical Ventilation, Extracorporeal Membrane Oxygenation, or Death

The area of the data marker for each trial is proportional to its weight in the fixed-effects meta-analysis. Progression to requiring invasive mechanical ventilation or extracorporeal membrane oxygenation or death among patients not receiving invasive mechanical ventilation at randomization.

aNA indicates not available; there were insufficient data to estimate odds ratio or the trial did not supply data for this outcome.

bCommon control group across both treatment comparisons of the COV-AID trial.

cThe data for the PreToVid trial are based on events up until 30 days after randomization.

dThere were 21 patients in the control group for both treatment comparisons. The analyses have been adjusted to correct for this.

Figure 4.  Association Between IL-6 Antagonists vs Usual Care or Placebo and Secondary Infections
Association Between IL-6 Antagonists vs Usual Care or Placebo and Secondary Infections

The area of the data marker for each trial is proportional to its weight in the fixed-effects meta-analysis.

aCommon control group across both treatment comparisons of the COV-AID trial.

bNA indicates not available; the trial did not supply data for this outcome.

cThe data for the PreToVid trial are based on events up until 30 days after randomization.

Table 1.  Selected Characteristics of Included Trials
Selected Characteristics of Included Trials
Table 2.  Subgroup Analysis of 3 Outcomes by Treatment Group and Respiratory Support, Organ Support, Age, Sex, and C-Reactive Protein Level
Subgroup Analysis of 3 Outcomes by Treatment Group and Respiratory Support, Organ Support, Age, Sex, and C-Reactive Protein Level
Table 3.  Additional Outcomes With Analysis by Respiratory Support and Corticosteroid Use
Additional Outcomes With Analysis by Respiratory Support and Corticosteroid Use
1.
Leisman  DE, Ronner  L, Pinotti  R,  et al.  Cytokine elevation in severe and critical COVID-19: a rapid systematic review, meta-analysis, and comparison with other inflammatory syndromes.   Lancet Respir Med. 2020;8(12):1233-1244. doi:10.1016/S2213-2600(20)30404-5PubMedGoogle ScholarCrossref
2.
Laing  AG, Lorenc  A, Del Molino Del Barrio  I,  et al.  A dynamic COVID-19 immune signature includes associations with poor prognosis.   Nat Med. 2020;26(10):1623-1635. doi:10.1038/s41591-020-1038-6PubMedGoogle ScholarCrossref
3.
Mathew  D, Giles  JR, Baxter  AE,  et al; UPenn COVID Processing Unit.  Deep immune profiling of COVID-19 patients reveals distinct immunotypes with therapeutic implications.   Science. 2020;369(6508):eabc8511. doi:10.1126/science.abc8511PubMedGoogle Scholar
4.
Marshall  JC, Murthy  S, Diaz  J,  et al; WHO Working Group on the Clinical Characterisation and Management of COVID-19 Infection.  A minimal common outcome measure set for COVID-19 clinical research.   Lancet Infect Dis. 2020;20(8):e192-e197. doi:10.1016/S1473-3099(20)30483-7PubMedGoogle ScholarCrossref
5.
McElvaney  OJ, Curley  GF, Rose-John  S, McElvaney  NG.  Interleukin-6: obstacles to targeting a complex cytokine in critical illness.   Lancet Respir Med. 2021;9(6):643-654. doi:10.1016/S2213-2600(21)00103-XPubMedGoogle ScholarCrossref
6.
Tierney  JF, Fisher  DJ, Vale  CL,  et al.  A framework for prospective, adaptive meta-analysis (FAME) of aggregate data from randomised trials.   PLoS Med. 2021;18(5):e1003629. doi:10.1371/journal.pmed.1003629PubMedGoogle Scholar
7.
Sterne  JAC, Murthy  S, Diaz  JV,  et al; WHO Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group.  Association between administration of systemic corticosteroids and mortality among critically ill patients with COVID-19: a meta-analysis.   JAMA. 2020;324(13):1330-1341. doi:10.1001/jama.2020.17023PubMedGoogle Scholar
8.
Gordon  AC, Mouncey  PR, Al-Beidh  F,  et al; REMAP-CAP Investigators.  Interleukin-6 receptor antagonists in critically ill patients with Covid-19.   N Engl J Med. 2021;384(16):1491-1502. doi:10.1056/NEJMoa2100433PubMedGoogle Scholar
9.
Abani  O, Abbas  A, Abbas  F,  et al; RECOVERY Collaborative Group.  Tocilizumab in patients admitted to hospital with COVID-19 (RECOVERY): a randomised, controlled, open-label, platform trial.   Lancet. 2021;397(10285):1637-1645. doi:10.1016/S0140-6736(21)00676-0PubMedGoogle ScholarCrossref
10.
Hermine  O, Mariette  X, Tharaux  PL, Resche-Rigon  M, Porcher  R, Ravaud  P; CORIMUNO-19 Collaborative Group.  Effect of tocilizumab vs usual care in adults hospitalized with COVID-19 and moderate or severe pneumonia: a randomized clinical trial.   JAMA Intern Med. 2021;181(1):32-40. doi:10.1001/jamainternmed.2020.6820PubMedGoogle ScholarCrossref
11.
Rosas  IO, Bräu  N, Waters  M,  et al.  Tocilizumab in hospitalized patients with severe Covid-19 pneumonia.   N Engl J Med. 2021;384(16):1503-1516. doi:10.1056/NEJMoa2028700PubMedGoogle ScholarCrossref
12.
Salama  C, Han  J, Yau  L,  et al.  Tocilizumab in patients hospitalized with Covid-19 pneumonia.   N Engl J Med. 2021;384(1):20-30. doi:10.1056/NEJMoa2030340PubMedGoogle ScholarCrossref
13.
Stone  JH, Frigault  MJ, Serling-Boyd  NJ,  et al; BACC Bay Tocilizumab Trial Investigators.  Efficacy of tocilizumab in patients hospitalized with Covid-19.   N Engl J Med. 2020;383(24):2333-2344. doi:10.1056/NEJMoa2028836PubMedGoogle ScholarCrossref
14.
Veiga  VC, Prats  JAGG, Farias  DLC,  et al; Coalition COVID-19 Brazil VI Investigators.  Effect of tocilizumab on clinical outcomes at 15 days in patients with severe or critical coronavirus disease 2019: randomised controlled trial.   BMJ. 2021;372(n84):n84. doi:10.1136/bmj.n84PubMedGoogle Scholar
15.
Sterne  JAC, Savović  J, Page  MJ,  et al.  RoB 2: a revised tool for assessing risk of bias in randomised trials.   BMJ. 2019;366:l4898. doi:10.1136/bmj.l4898PubMedGoogle Scholar
16.
Langan  D, Higgins  JPT, Jackson  D,  et al.  A comparison of heterogeneity variance estimators in simulated random-effects meta-analyses.   Res Synth Methods. 2019;10(1):83-98. doi:10.1002/jrsm.1316PubMedGoogle ScholarCrossref
17.
Nikolakopoulou  A, Mavridis  D, Salanti  G.  Demystifying fixed and random effects meta-analysis.   Evid Based Ment Health. 2014;17(2):53-57. doi:10.1136/eb-2014-101795PubMedGoogle ScholarCrossref
18.
Serghiou  S, Goodman  SN.  Random-effects meta-analysis: summarizing evidence with caveats.   JAMA. 2019;321(3):301-302. doi:10.1001/jama.2018.19684PubMedGoogle ScholarCrossref
19.
Rücker  G, Cates  CJ, Schwarzer  G.  Methods for including information from multi-arm trials in pairwise meta-analysis.   Res Synth Methods. 2017;8(4):392-403. doi:10.1002/jrsm.1259PubMedGoogle ScholarCrossref
20.
Fisher  DJ, Carpenter  JR, Morris  TP, Freeman  SC, Tierney  JF.  Meta-analytical methods to identify who benefits most from treatments: daft, deluded, or deft approach?   BMJ. 2017;356:j573. doi:10.1136/bmj.j573PubMedGoogle Scholar
21.
White  IR.  Network meta-analysis.   Stata J. 2015;15(4):951-985. doi:10.1177/1536867X1501500403 Google ScholarCrossref
22.
Fisher  DJ.  Two-stage individual participant data meta-analysis and generalized forest plots.   Stata J. 2015;15(2):369-396. doi:10.1177/1536867X1501500203Google ScholarCrossref
23.
Lescure  FX, Honda  H, Fowler  RA,  et al; Sarilumab COVID-19 Global Study Group.  Sarilumab in patients admitted to hospital with severe or critical COVID-19: a randomised, double-blind, placebo-controlled, phase 3 trial.   Lancet Respir Med. 2021;9(5):522-532. doi:10.1016/S2213-2600(21)00099-0PubMedGoogle ScholarCrossref
24.
Soin  AS, Kumar  K, Choudhary  NS,  et al.  Tocilizumab plus standard care versus standard care in patients in India with moderate to severe COVID-19-associated cytokine release syndrome (COVINTOC): an open-label, multicentre, randomised, controlled, phase 3 trial.   Lancet Respir Med. 2021;9(5):511-521. doi:10.1016/S2213-2600(21)00081-3PubMedGoogle ScholarCrossref
25.
Xu  C, Rafique  A, Potocky  T,  et al.  Differential binding of sarilumab and tocilizumab to IL-6Rα and effects of receptor occupancy on clinical parameters.   J Clin Pharmacol. 2021;61(5):714-724. doi:10.1002/jcph.1795PubMedGoogle ScholarCrossref
26.
Rhen  T, Cidlowski  JA.  Antiinflammatory action of glucocorticoids—new mechanisms for old drugs.   N Engl J Med. 2005;353(16):1711-1723. doi:10.1056/NEJMra050541PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    4 Comments for this article
    EXPAND ALL
    IL-6 Antagonists Do Not Seem to Decrease Mortality in COVID-19
    Samer Singh, MSc, Ph.D. | Institute of Medical Sciences, Banaras Hindu University
    The study is a welcome addition to existing literature relevant to clinical practice.

    However, given the paucity of time, the conclusions drawn in the paper may be misinterpreted by clinicians in general.

    There is a high probability that the conclusions, as presented, could make clinicians err to construe IL-6 antagonist as helping in COVID-19 recovery (reducing mortality). It may cause overprescription of IL-6 antagonists. The effects observed are largely stemming from corticosteroids without any significant evidence for the IL-6 antagonist reducing COVID-19 mortality. The data presented rather seem to suggest IL-6 antagonists alone would be increasing the
    COVID-19 mortality. (Further extension of the logic could even suggest the possibility of IL-6 in combination with a suitable corticosteroid being more beneficial in COVID-19 treatment. However, only the controlled clinical trials may present a clear picture.)

    The effect of corticosteroids on certain categories of COVID-19 patients who also happen to be part of the current analysis is known to vary with corticosteroids used. The authors may like to present the data segregated with regard to the type of corticosteroids administered as the effect varies with corticosteroids used. Treating all corticosteroid treatments equally for the analysis would make the analysis inherently less reliable and introduce bias in the conclusion.




    CONFLICT OF INTEREST: None Reported
    READ MORE
    Statistical Analysis
    Ali Haider Bangash, MBBS student | STMU Shifa College of Medicine, Islamabad, Pakistan
    The authors have undergone a laborious data extraction and synthesis with subsequent in-depth analysis in a bid to cement and reinforce the clinical significance of IL-6 antagonists when it comes to severe COVID-19 management specifically vis a vis mortality reduction (1). The following is most humbly conveyed concerning the statistical approach:
    ------ The study's fixed-effect model does not take into account the variation instituted among the considered population of COVID-19 patients secondary to factors including but not limited to individual populations’ characteristics and institution of the intervention. It is a widely recognized practice adopted by the scientific community to implement
    a random-effects model for such cases (2).
    ------ The I2 statistic explores that proportion of the observed variance which arises secondary to such differences (“inconsistencies”) in true effect-sizes (odds ratios) that occur because of inherent divergence among the underlying respective study populations— divergence that cannot be accounted for by a fixed-effects model; hence, the almost completely across-the-board preference for a random-effects model.

    Therefore, the statistical approach might be revised to account for variance secondary to the underlying differences in the study population of the respective studies while calculating odds ratios. The highly respected authors, by adopting such an analysis approach that is forged with time-tested meta-analysis principles, would be in a better position to infer sound associations and subsequently inform the discourse around the respective research question with potential clinical implications aimed at decreasing morbidity and mortality associated with the global healthcare catastrophe which COVID-19 pandemic truly is.

    REFERENCES:

    1. WHO Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group, Shankar-Hari M, Vale CL, et al. Association Between Administration of IL-6 Antagonists and Mortality Among Patients Hospitalized for COVID-19: A Meta-analysis [published online ahead of print, 2021 Jul 6]. JAMA. 2021;e2111330. doi:10.1001/jama.2021.11330

    2. Spineli LM, Pandis N. The importance of careful selection between fixed-effect and random-effects models. Am J Orthod Dentofacial Orthop. 2020;157(3):432-433. doi:10.1016/j.ajodo.2019.12.003

    CONFLICT OF INTEREST: None Reported
    READ MORE
    Biased Inferences Due to Inappropriate Statistical Approaches
    Haluk Vahaboglu, Professor | Istanbul Medeniyet Universitesi, Tıp Fakültesi
    This comprehensive meta-analysis is undoubtedly valuable. However the conclusion that tocilizumab use was associated with better outcomes can mislead medical practice.

    Sensitivity analyses with placebo-controlled trials are presented as supplemental data. The effect direction from placebo-controlled trials showed no benefit for tocilizumab. In other words, this sensitivity analysis showed that open-label trials dominate the conclusion. Study type is a significant parameter in combining effect estimates. Therefore, a subgroup analysis using study types should be presented in the main document, and the results should be shown in the abstract.

    Secondly, we could not understand censoring data at 28-days. This
    meta-analysis combines effect estimates to obtain an odds ratio. It is not a time-to-event analysis with 28-days right-censored data. Keeping in mind the half-life and the long-lasting effect of tocilizumab, in-hospital mortality without time restriction should be accepted as the primary outcome in this meta-analysis.

    CONFLICT OF INTEREST: None Reported
    READ MORE
    What is the Cost Benefit Ratio?
    Alastair Miller, MA MB FRCP | Cumberland Infirmary Carlisle
    Useful analysis but given the huge cost of the pandemic especially in healthcare settings with limited resources, it would be helpful to know

    1. What is the number needed to treat and

    2. What is the cost per life saved?
    CONFLICT OF INTEREST: None Reported
    Original Investigation
    July 6, 2021

    Association Between Administration of IL-6 Antagonists and Mortality Among Patients Hospitalized for COVID-19: A Meta-analysis

    The WHO Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group
    JAMA. 2021;326(6):499-518. doi:10.1001/jama.2021.11330
    Key Points

    Question  Is administration of IL-6 antagonists associated with 28-day all-cause mortality in patients hospitalized for COVID-19?

    Findings  This prospective meta-analysis of 27 randomized trials included 10 930 patients, of whom 2565 died by 28 days. The 28-day all-cause mortality was lower among patients who received IL-6 antagonists compared with those who received usual care or placebo (summary odds ratio, 0.86). The summary odds ratios for the association of IL-6 antagonist treatment with 28-day all-cause mortality were 0.78 with concomitant administration of corticosteroids vs 1.09 without administration of corticosteroids.

    Meaning  Administration of IL-6 antagonists, compared with usual care or placebo, was associated with lower 28-day all-cause mortality in patients hospitalized for COVID-19.

    Abstract

    Importance  Clinical trials assessing the efficacy of IL-6 antagonists in patients hospitalized for COVID-19 have variously reported benefit, no effect, and harm.

    Objective  To estimate the association between administration of IL-6 antagonists compared with usual care or placebo and 28-day all-cause mortality and other outcomes.

    Data Sources  Trials were identified through systematic searches of electronic databases between October 2020 and January 2021. Searches were not restricted by trial status or language. Additional trials were identified through contact with experts.

    Study Selection  Eligible trials randomly assigned patients hospitalized for COVID-19 to a group in whom IL-6 antagonists were administered and to a group in whom neither IL-6 antagonists nor any other immunomodulators except corticosteroids were administered. Among 72 potentially eligible trials, 27 (37.5%) met study selection criteria.

    Data Extraction and Synthesis  In this prospective meta-analysis, risk of bias was assessed using the Cochrane Risk of Bias Assessment Tool. Inconsistency among trial results was assessed using the I2 statistic. The primary analysis was an inverse variance–weighted fixed-effects meta-analysis of odds ratios (ORs) for 28-day all-cause mortality.

    Main Outcomes and Measures  The primary outcome measure was all-cause mortality at 28 days after randomization. There were 9 secondary outcomes including progression to invasive mechanical ventilation or death and risk of secondary infection by 28 days.

    Results  A total of 10 930 patients (median age, 61 years [range of medians, 52-68 years]; 3560 [33%] were women) participating in 27 trials were included. By 28 days, there were 1407 deaths among 6449 patients randomized to IL-6 antagonists and 1158 deaths among 4481 patients randomized to usual care or placebo (summary OR, 0.86 [95% CI, 0.79-0.95]; P = .003 based on a fixed-effects meta-analysis). This corresponds to an absolute mortality risk of 22% for IL-6 antagonists compared with an assumed mortality risk of 25% for usual care or placebo. The corresponding summary ORs were 0.83 (95% CI, 0.74-0.92; P < .001) for tocilizumab and 1.08 (95% CI, 0.86-1.36; P = .52) for sarilumab. The summary ORs for the association with mortality compared with usual care or placebo in those receiving corticosteroids were 0.77 (95% CI, 0.68-0.87) for tocilizumab and 0.92 (95% CI, 0.61-1.38) for sarilumab. The ORs for the association with progression to invasive mechanical ventilation or death, compared with usual care or placebo, were 0.77 (95% CI, 0.70-0.85) for all IL-6 antagonists, 0.74 (95% CI, 0.66-0.82) for tocilizumab, and 1.00 (95% CI, 0.74-1.34) for sarilumab. Secondary infections by 28 days occurred in 21.9% of patients treated with IL-6 antagonists vs 17.6% of patients treated with usual care or placebo (OR accounting for trial sample sizes, 0.99; 95% CI, 0.85-1.16).

    Conclusions and Relevance  In this prospective meta-analysis of clinical trials of patients hospitalized for COVID-19, administration of IL-6 antagonists, compared with usual care or placebo, was associated with lower 28-day all-cause mortality.

    Trial Registration  PROSPERO Identifier: CRD42021230155

    Introduction

    Excessive systemic inflammation and raised IL-6 levels resulting from dysregulated host immune responses1-3 are associated with adverse clinical outcomes in patients hospitalized with COVID-19.4 This led to the design of several randomized clinical trials assessing the efficacy of IL-6 antagonists in patients with COVID-19. The IL-6 antagonists commonly investigated were monoclonal antibodies that bind either to membrane-bound and soluble IL-6 receptors (eg, tocilizumab and sarilumab) or directly to IL-6 (eg, siltuximab).5

    To address the need for reliable efficacy data to guide clinical management, the World Health Organization (WHO) Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group developed a prospective meta-analysis protocol to perform a prospective meta-analysis of IL-6 antagonists in patients hospitalized for COVID-19. This approach was recently used6 to evaluate the use of corticosteroids in patients with COVID-19.7 During this initiative, trials variously reported potential clinical benefit,8-10 no benefit,11-13 and potential harm14 with IL-6 antagonists in patients hospitalized for COVID-19.

    The primary objective of this prospective meta-analysis of randomized trials6 was to estimate the association between administration of IL-6 antagonists, compared with usual care or placebo, and mortality at 28 days after randomization in patients hospitalized for COVID-19. The secondary objectives were to estimate associations within subgroups relating to disease severity (eg, level of respiratory support), treatments at randomization (eg, receipt of corticosteroids), patient characteristics (eg, age), and risk of bias15 overall and separately for tocilizumab and sarilumab.

    Methods
    Identification and Eligibility of Trials

    Trials were identified through systematic searches of ClinicalTrials.gov, the EU Clinical Trials Register, and the WHO International Clinical Trials Registry Platform from October 7, 2020, to January 11, 2021. The search terms used were random* AND COVID in the title or abstract, along with terms for common IL-6 antagonists individually (tocilizumab, sarilumab, clazakizumab, siltuximab, olokizumab) and the term interleukin 6. Individual searches were then combined. Searches were not restricted by trial status (ongoing or completed), publication status, or language. Additional trials were identified through contact with experts from the REACT Working Group. Queries regarding eligibility for inclusion were resolved by consensus. Eligible trials randomly assigned patients hospitalized for COVID-19 to IL-6 antagonists vs usual care or placebo. Trials in which anti–IL-6 therapies were combined with other immunomodulatory agents or with active comparators other than systemic corticosteroids were excluded.

    Development of Prospective Meta-analysis Protocol

    The WHO chief scientist invited investigators of eligible trials to participate in this prospective meta-analysis. Representative investigators and sponsors of potentially eligible trials were asked to participate in weekly development calls for the prospective meta-analysis protocol starting on November 23, 2020. The prospective meta-analysis protocol was registered on the PROSPERO database on January 14, 2021, and regularly updated. The PICO (patient problem or population, intervention, comparison or control, and outcome) framework, definitions of outcomes, and subgroups of interest were agreed upon prior to collection of outcome data.6 The final version of the prospective meta-analysis protocol was registered before analyses started on March 29, 2021.

    Trial-level aggregate data sharing agreements were established. All trials had secured institutional review board approval, but approval was not required for secondary analyses. Informed consent for participation in each trial was obtained, consistent with local institutional review board requirements. Trial investigators were asked to complete baseline and outcome data collection forms that were subsequently verified by trial teams. Finalized data sets from contributing trials were received by May 11, 2021.

    Outcomes and Comparisons

    The primary outcome measure was all-cause mortality at 28 days after randomization. Two comparisons were specified a priori. The primary comparison investigated the class effect of IL-6 antagonists vs usual care or placebo and tocilizumab and sarilumab were examined separately. The second comparison was of IL-6 antagonists vs corticosteroids.

    The secondary outcomes included: (1) invasive mechanical ventilation (IMV), extracorporeal membrane oxygenation (ECMO), or death by 28 days in patients not receiving IMV at randomization (this is the most important secondary outcome for which data on all subgroups were collected); (2) cardiovascular system support (defined as receipt of vasopressors) or death by 28 days in patients not receiving cardiovascular system support at randomization; (3) secondary infections by 28 days (this is the most important safety outcome); (4) in-hospital mortality; (5) kidney replacement therapy (KRT) or death by 28 days in patients not receiving KRT at randomization (excluding patients with underlying dialysis dependence or ≥stage III chronic kidney disease); (6) discharged alive from the hospital by 28 days; (7) mortality by 90 days; (8) duration of IMV up to 28 days (in those receiving IMV at randomization, with duration coded as 28 days for patients who died); and (9) secondary infections by 90 days. Data on serious adverse events or serious adverse reactions (as defined in each trial) were collected; however, no meta-analysis was planned because diverse definitions were used by different trials.

    Subgroup Analyses

    Trial investigators supplied summary data for all outcomes according to intervention group, overall, and in subgroups based on: (1) degree of respiratory support at randomization (patient not receiving supplemental oxygen therapy, patient receiving supplemental oxygen therapy [defined as oxygen flow rate ≤15 L/min by face mask or nasal cannula], patient receiving noninvasive ventilation [defined as oxygen flow rate >15 L/min, high-flow nasal cannula, continuous positive airway pressure], or patient receiving IMV or ECMO) and (2) receipt of systemic corticosteroids at randomization. In addition, the following subgroups were used for the outcomes of 28-day all-cause mortality and progression to IMV or death: (1) patients receiving acute organ support therapy at randomization (vasopressors or KRT) among those receiving noninvasive ventilation, IMV, or ECMO; (2) age (<70 years or ≥70 years); (3) sex (female or male); (4) race/ethnicity (collected by investigators in each individual trial); and (5) C-reactive protein level at baseline (categorized as <75, 75-<150, ≥150 μg/mL). The assigned dose of IL-6 antagonists was classified as low (4 mg/kg of tocilizumab; 200 mg of sarilumab) or high (>4 mg/kg of tocilizumab or multiple doses; >400 mg of sarilumab or multiple doses).

    Risk of Bias Assessment

    For each trial, the risk of bias (low risk, some concerns, or high risk) was assessed using version 2 of the Cochrane Risk of Bias Assessment Tool.15 Risk of bias assessments were based on the trial protocols and flowcharts following the Consolidated Standards of Reporting Trials together with information supplied by the investigators for each trial in a standard format. Risk of bias assessments were done independently by 3 of the investigators (J.P.T.H., F.S., J.S.) with disagreements resolved through discussion. The Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach was used to assess the certainty of the evidence.

    Data Analyses

    The primary analysis was an inverse variance–weighted fixed-effects meta-analysis of odds ratios (ORs). For the duration of IMV therapy, investigators supplied the mean difference and associated 95% CIs in days comparing the treatment and control groups. For the 90-day outcomes, the trial investigators were asked to estimate hazard ratios and 95% CIs (or log hazard ratios and associated standard errors) using Cox regression. Inconsistency in associations among the trials was quantified using the I2 statistic. P values for heterogeneity were derived using the Cochran Q statistic. Precise P values were reported; however, the prospective meta-analysis protocol specified that a threshold for statistical significance would not be used. As a sensitivity analysis for the primary outcome of 28-day all-cause mortality, overall associations also were estimated using random-effects meta-analyses with a restricted maximum likelihood estimate of heterogeneity16 and Hartung-Knapp adjustment17,18 to account for uncertainty in the estimation of between-study variance. To obtain illustrative absolute risk estimates for patients not receiving treatment with IL-6 antagonists, a mortality risk of 25% and a progression risk of 33% to IMV or death were assumed (the approximate risks among all eligible patients allocated to usual care or placebo). Meta-analytic ORs were then applied to obtain the corresponding risk with IL-6 antagonists. Because outcome data were generally complete or nearly complete across trials, we restricted the analyses to trial participants with outcomes recorded.

    Differences in associations between the subgroups were quantified by calculating ratios of ORs (or analogous statistics for other outcome types) to compare the effects in the subgroups along with corresponding P values for interaction. If the ratio of ORs was equal to 1, the estimated associations in the 2 subgroups were the same. The further the ratio of ORs was from 1, the greater was the difference between the estimated associations in the 2 subgroups. Comparisons between subgroups defined by trial characteristics were made using random-effects meta-regression and appropriately accounted for common controls19 in trials with 3 treatment groups. Comparisons between subgroups defined by patient characteristics were done by estimating trial-specific ratios of ORs comparing associations between subgroups and then combining these in meta-analyses.20 The ORs in patients not receiving corticosteroids were compared with patients receiving corticosteroids at randomization within the respiratory support subgroups. Subgroup-specific estimates adjusted to correspond with the ratios of ORs that were derived from the within-trial approach were also estimated.

    In the sensitivity analyses, associations were estimated that (1) excluded the Randomised Evaluation of COVID-19 Therapy (RECOVERY) trial; (2) restricted the analyses to trial results at low risk of bias; (3) restricted the analyses to trials published in peer-reviewed journals; (4) restricted the analyses to placebo-controlled trials; and (5) restricted the analyses to open-label trials. The first and third of these were post hoc sensitivity analyses. All analyses were conducted using Stata version 16 (StataCorp) and new Stata commands to conduct and graph the results of the meta-analyses.21,22

    Results

    A total of 72 potentially eligible trials were identified. After screening these trials, 38 ineligible trials, 3 duplicated records, and 2 trials directly comparing IL-6 antagonists with corticosteroids (NCT04329650 [n = 158 patients] and NCT04345445 [n = 59 patients]) were excluded. Of 29 eligible trials that randomized patients to receive IL-6 antagonists vs usual care or placebo, 1 trial (n = 50 patients) was unable to supply data in a timely manner and 1 trial (n = 295 patients) was still following up patients for the primary outcome.

    Among the 27 trials included in the meta-analyses, 9 were published8-14,23,24 and the remaining 18 were unpublished or were reported as preprints (NCT04412772, NCT04331808 [there were 2 separate trials conducted under a common protocol], NCT04330638, NCT04479358, NCT04577534, NCT04435717, NCT04377750, NCT04409262, EU-CTR 2020-001748-24, EU-CTR 2020-001375-32, EU-CTR 2020-001442-19, NCT04324073 [there were 2 separate trials conducted under a common protocol], NCT04315298, NCT04357808, EU-CTR 2020-001531-27, and EU-CTR 2020-002037-15; Table 1 and eTables 1-3 in Supplement 1). Outcome data were supplied for 10 930 patients, representing 95.4% of all patients randomized in eligible trials (eFigure 1 in Supplement 1). Patients were recruited from 28 countries from February 26, 2020.

    The IL-6 antagonists assessed were tocilizumab (19 trials allocating 4299 patients to tocilizumab and 3749 patients to usual care or placebo), sarilumab (9 trials allocating 2073 patients to sarilumab and 753 patients to usual care or placebo), and siltuximab (1 trial allocating 77 patients to siltuximab and 72 patients to usual care or placebo). The Randomized, Embedded, Multifactorial Adaptive Platform Trial for Community-Acquired Pneumonia8 (REMAP-CAP) and COV-AID (NCT04330638) trials randomized patients to more than 1 IL-6 antagonist. Due to limited data (including outcome events), associations for siltuximab within predefined subgroups were not estimated. Similarly, due to limited data, associations were not estimated in the predefined low-dose strata (2 trials for tocilizumab [27 patients] and in 2 trials for sarilumab [307 patients]), in the no oxygen respiratory support subgroup (27 patients and 4 deaths in 3 trials), and for trials reporting secondary infections at 90 days. Because of the diversity of classification of race/ethnicity among different trials, the subgroup analyses according to race/ethnicity are not reported. Because not all trials estimated hazard ratios for 90-day mortality, the event numbers were also analyzed to estimate the ORs.

    The median age across the trials was 61 years (range of medians, 52-68 years) and 3560 patients (33%) were women. Concurrent treatments at randomization varied substantially among the trials. Most patients received respiratory support at randomization. A greater proportion of patients in the sarilumab trials received IMV (31% [873/3136 patients]) compared with patients in the tocilizumab trials (15% [1211/8134 patients]) and a smaller proportion received corticosteroids (35% [890/3136 patients] vs 66% [5317/8134 patients], respectively; Table 1). The primary outcome was missing for 183 patients (1.6%). Three trials recorded no deaths by 28 days (COVID-19: Salvage Tocilizumab as a Rescue Measure [COVIDSTORM {NCT04577534}; n = 39]; Efficacy of Tocilizumab in Modifying the Inflammatory Parameters of Patients With COVID-19 [COVITOZ-01 {NCT04435717}; n = 26]; and Clinical Trial of the Use of Tocilizumab for Treatment of SARS-CoV-2 Infection [COVID-19; TOCOVID] {NCT04332094} [n = 270]).

    Risk of bias was assessed to be low in 22 of the trials contributing to the meta-analysis of 28-day all-cause mortality, comprising 78% of the weight in the analysis. Six trials were judged to have some concerns, mainly due to small numbers of patients being excluded from the data set because they did not receive their assigned intervention. In 1 trial judged as high risk, comprising 0.65% of the weight, the usual procedures were not in place to ensure that the allocation sequence was concealed; however, there was no reason to suspect that the concealed allocation was not implemented as intended. Risk of bias assessments were similar for progression to IMV or death. For secondary infections, results from open-label trials were judged to have some concerns over bias in determining whether such infections had occurred due to the subjective nature of the decision (eFigure 2 and eTable 4 in Supplement 1).

    Association Between IL-6 Antagonists and 28-Day All-Cause Mortality

    By 28 days after randomization, there were 1407 deaths among 6449 patients randomized to IL-6 antagonists and 1158 deaths among 4481 patients randomized to usual care or placebo. Using a fixed-effects meta-analysis, the summary OR was 0.86 (95% CI, 0.79-0.95; P = .003). This corresponds to an absolute mortality risk of 22% for IL-6 antagonists compared with an assumed mortality risk of 25% for usual care or placebo. The summary OR was 0.89 (95% CI, 0.76-1.05; P = .16) in a sensitivity analysis using random-effects meta-analysis (eFigure 3 in Supplement 1). The certainty in this result was assessed to be high in the GRADE assessment.

    In 19 trials that randomized 4299 patients to tocilizumab (960 deaths) and 3749 patients to usual care or placebo (1023 deaths), the summary OR was 0.83 (95% CI, 0.74-0.92; P < .001). This corresponds to an absolute mortality risk of 22% for tocilizumab compared with an assumed mortality risk of 25% for usual care or placebo. In 9 trials that randomized 2073 patients to sarilumab (473 deaths) and 753 patients to usual care or placebo (139 deaths), the summary OR was 1.08 (95% CI, 0.86-1.36; P = .52). This corresponds to an absolute mortality risk of 26% for sarilumab compared with an assumed mortality risk of 25% for usual care or placebo. There was little inconsistency between the trial results (I2 = 18% overall, I2 = 3% for tocilizumab, and I2 = 0% for sarilumab). The inverse association with 28-day all-cause mortality appeared more marked for tocilizumab than for sarilumab (ratio of ORs, 0.76 [95% CI, 0.59-0.98], P = .04 for interaction; Figure 1 and Table 2).

    Data on receipt of corticosteroids at randomization were available in 22 trials (9953 patients and 2495 deaths). The summary ORs for 28-day all-cause mortality comparing IL-6 antagonists with usual care or placebo were 1.09 (95% CI, 0.91-1.30) for 3637 patients (830 deaths) not receiving corticosteroids and 0.78 (95% CI, 0.69-0.88) for 6316 patients (1665 deaths) receiving corticosteroids (Figure 2). The corresponding absolute mortality risk in patients receiving corticosteroids was 21% for IL-6 antagonists compared with an assumed mortality risk of 25% for usual care or placebo. Based on within-trial estimates combined across 17 trials that included patients receiving and not receiving corticosteroids, the inverse association between IL-6 antagonists and mortality was more marked in patients receiving corticosteroids (ratio of ORs, 0.72 [95% CI, 0.56-0.92]; P = .008 for interaction). The summary OR for the association with mortality for tocilizumab (15 trials, 7490 patients, and 1951 deaths) was 1.06 (95% CI, 0.85-1.33) in patients not receiving corticosteroids at randomization and was 0.77 (95% CI, 0.68-0.87) in patients receiving corticosteroids at randomization. The summary ratio of ORs (based on within-trial comparisons) was 0.69 (95% CI, 0.52-0.91; P = .008 for interaction). The corresponding summary ORs for sarilumab (8 trials, 2406 patients, and 538 deaths) were 1.18 (95% CI, 0.88-1.58) and 0.92 (95% CI, 0.61-1.38), respectively. The summary ratio of ORs (based on within-trial comparisons) was 0.77 (95% CI, 0.44-1.33; P = .34 for interaction). The corresponding absolute mortality risks in patients receiving corticosteroids were 20% for tocilizumab and 23% for sarilumab compared with an assumed mortality risk of 25% for usual care or placebo. In additional analyses, associations were compared in patients not receiving and receiving corticosteroids at randomization within the respiratory support subgroups. The tendency for more marked inverse associations among patients receiving corticosteroids appeared broadly consistent across respiratory support subgroups; however, the associations were not estimated precisely.

    Detailed results, forest plots, and comparisons between subgroups for 28-day all-cause mortality appear in Supplements 1 and 2. Data on respiratory support at randomization were available in 21 trials (9835 patients and 2493 deaths). The summary ORs for 28-day all-cause mortality comparing IL-6 antagonists with usual care or placebo were 0.81 (95% CI, 0.67-0.98) in 3954 patients (560 deaths) receiving supplemental oxygen at randomization, 0.83 (95% CI, 0.72-0.96) in 3864 patients (1132 deaths) receiving noninvasive ventilation or high-flow nasal cannula at randomization, and 0.95 (95% CI, 0.78-1.16) in 2017 patients (801 deaths) receiving IMV or ECMO at randomization (P = .71 for the differences between associations across these subgroups; Table 2). The corresponding summary ORs for tocilizumab were 0.82 (95% CI, 0.67-1.00), 0.80 (95% CI, 0.68-0.93), and 0.92 (95% CI, 0.72-1.17), respectively (P = .43 for differences between subgroups) and the corresponding summary ORs for sarilumab were 0.74 (95% CI, 0.42-1.30), 1.20 (95% CI, 0.78-1.84), and 1.05 (95% CI, 0.74-1.50), respectively (P = .65 for differences between subgroups).

    The associations between IL-6 antagonists and 28-day all-cause mortality within other subgroups defined by patient characteristics at randomization appeared consistent across all these subgroups (all P values for comparisons between subgroups were greater than .11; Table 2 and Supplement 2).

    Association Between IL-6 Antagonists and Progression to IMV or Death

    Among patients not requiring IMV at randomization (24 trials), 1236 of 4650 randomized to IL-6 antagonists and 1220 of 3609 randomized to usual care or placebo progressed to requiring IMV or ECMO or died within 28 days. Most of the data (87%) were from trials assessing tocilizumab. The summary ORs compared with usual care or placebo were 0.77 (95% CI, 0.70-0.85; P < .001) for all IL-6 antagonists, 0.74 (95% CI, 0.66-0.82) for tocilizumab, and 1.00 (95% CI, 0.74-1.35) for sarilumab (Figure 3). The corresponding absolute risks of progression to IMV or death were 28% for all IL-6 antagonists, 27% for tocilizumab, and 33% for sarilumab compared with an assumed risk of 33% for usual care or placebo. There was little inconsistency between the trial results (I2 = 0% for each meta-analysis). The certainty in the overall result was assessed to be high in the GRADE assessment. The ratio of ORs comparing the associations for tocilizumab and sarilumab was 0.74 (95% CI, 0.54-1.01; P = .06 for interaction).

    The summary ORs for progression to IMV or death were 0.96 (95% CI, 0.79-1.17) in 2545 patients (707 progressed) not receiving corticosteroids and 0.71 (95% CI, 0.63-0.80) in 5482 patients (1715 progressed) receiving corticosteroids (Figure 2). The corresponding absolute risk for progression to IMV or death in patients receiving corticosteroids was 26% for IL-6 antagonists compared with an assumed risk of 33% for usual care or placebo. The ratio of ORs comparing the associations in those receiving and not receiving corticosteroids was 0.78 (95% CI, 0.59-1.02; P = .07 for interaction based on within-trial estimates combined across trials). The corresponding summary ORs for tocilizumab (17 trials, 6608 patients, and 2104 progressed) were 0.95 (95% CI, 0.76-1.20) and 0.69 (95% CI, 0.61-0.78), respectively, and the corresponding ratio of ORs was 0.70 (95% CI, 0.52-0.94; P = .02 for interaction). The corresponding summary ORs for sarilumab (7 trials, 1362 patients, and 298 progressed) were 0.98 (95% CI, 0.67-1.44) and 1.08 (95% CI, 0.67-1.75), respectively, and the corresponding ratio of ORs was 1.41 (95% CI, 0.65-3.07; P = .38 for interaction). The corresponding absolute risks for progression to IMV or death in patients receiving corticosteroids were 25% for tocilizumab and 35% for sarilumab compared with an assumed risk of 33% for progression to IMV or death for usual care or placebo.

    The summary ORs for progression to IMV or death comparing IL-6 antagonists with usual care or placebo were 0.75 (95% CI, 0.64-0.89) in 4044 patients (758 progressed) receiving supplemental oxygen at randomization and 0.77 (95% CI, 0.68-0.89) in 3765 patients (1661 progressed) receiving noninvasive ventilation or high-flow nasal cannula (P = .67 for differences between these associations; Table 2). The corresponding summary ORs for tocilizumab were 0.72 (95% CI, 0.60-0.86) and 0.74 (95% CI, 0.64-0.85), respectively (P = .92 for differences between subgroups) and the corresponding summary ORs for sarilumab were 0.96 (95% CI, 0.60-1.53) and 1.06 (95% CI, 0.71-1.57), respectively (P = .31 for differences between subgroups). The corresponding absolute risks for progression to IMV or death were 27% for all IL-6 antagonists, 27% for tocilizumab, and 33% for sarilumab compared with an assumed risk for progression to IMV or death of 33% for usual care or placebo.

    The associations between IL-6 antagonists and progression to IMV or death within other subgroups defined by patient characteristics at randomization appeared consistent across all other subgroups (all P values for comparisons between subgroups were greater than .28; Table 2 and Supplement 3).

    Association Between IL-6 Antagonists and Infections by 28 Days

    Among the 22 trials that reported 28-day infection outcomes, 750 events occurred among 3428 patients randomized to IL-6 antagonists and 330 events occurred among 1787 patients randomized to usual care or placebo. The fixed-effect summary OR was 0.99 (95% CI, 0.85-1.16) and there was little inconsistency between the trial results (I2 = 0%, P = .49 for heterogeneity; Figure 3). The certainty in this result was assessed to be moderate in the GRADE assessment due to minor concerns over risk of bias (because of subjectivity in the outcome assessment) and minor concerns over imprecision (because of the result being compatible with a slightly lower or higher risk among those receiving IL-6 antagonists). The ORs were 0.95 (95% CI, 0.77-1.16) for tocilizumab and 1.03 (95% CI, 0.80-1.32) for sarilumab (Figure 4 and Table 2). The summary ORs within subgroups were close to 1. Data on 28-day secondary infections appear in Supplement 4.

    Association Between IL-6 Antagonists and Other Secondary Outcomes

    Data on in-hospital mortality were available from 19 trials. The summary ORs for in-hospital mortality comparing IL-6 antagonists with usual care or placebo were 0.80 (95% CI, 0.71-0.89) in 7261 patients (1848 deaths), with little inconsistency between trials (I2 = 0%) (Table 3). Most of the data (90.7%) were from 14 trials (6587 patients and 1741 deaths) assessing tocilizumab and the summary OR was 0.80 (95% CI, 0.71-0.90).

    Among patients not requiring cardiovascular system support at randomization (15 trials), 344 of 1587 patients randomized to IL-6 antagonists and 343 of 1199 patients randomized to usual care or placebo progressed to requiring cardiovascular system support or death within 28 days. Most of the data (2553/2786 patients; 91.1%) were from 13 trials assessing tocilizumab. The summary ORs were 0.71 (95% CI, 0.59-0.86) for IL-6 antagonists and 0.70 (95% CI, 0.57-0.85) for tocilizumab. Among patients not requiring KRT at randomization (13 trials), 935 of 3653 patients randomized to IL-6 antagonists and 1069 of 3351 patients randomized to usual care progressed to requiring KRT or died within 28 days. The summary OR was 0.79 (95% CI, 0.71-0.88); most of the data (6884/7004 patients; 98.2%) were from 12 trials assessing tocilizumab.

    Among 10 904 patients recruited to participate in 26 trials, 6609 were discharged alive by 28 days. The summary OR comparing IL-6 antagonists with usual care or placebo was 1.22 (95% CI, 1.12-1.33), favoring IL-6 antagonists. The corresponding ORs were 1.30 (95% CI, 1.18-1.43) for tocilizumab and 0.95 (95% CI, 0.79-1.15) for sarilumab.

    Data were available for all-cause mortality at 90 days in 13 trials and at 60 days in 4 trials (1104 deaths among 4651 patients). Two trials reported no events. The summary OR comparing IL-6 antagonists with usual care or placebo was 0.89 (95% CI, 0.76-1.04). The corresponding ORs were 0.85 (95% CI, 0.69-1.05) for tocilizumab and 0.92 (95% CI, 0.74-1.16) for sarilumab. Additional survival analyses for all-cause mortality at 90 days are reported in eTable 5 in Supplement 1.

    Among 1171 patients who were receiving IMV at randomization and were recruited to 9 trials, the weighted mean difference comparing IL-6 antagonists with usual care or placebo in the duration of IMV was −0.84 (95% CI, −1.82 to 0.13), favoring IL-6 antagonists. Most of the data were from 8 trials assessing tocilizumab (1101/1171 patients; 94.0%).

    Table 3 and Supplements 5-10 provide detailed analyses comparing IL-6 antagonists with usual care or placebo overall and in predefined subgroups for all of the secondary outcomes. Although the associations appeared broadly consistent across subgroups, many were not estimated precisely.

    Serious Adverse Events or Reactions

    Data describing serious adverse events were supplied by 23 trials. Risks of serious adverse events were broadly similar for patients randomized to IL-6 antagonists and to usual care or placebo across all trials. Data on secondary infections at 90 days after randomization were limited (11 trials and 310 events) (eTable 6 in Supplement 1), but the risk of secondary infections by 28 days was similar in patients treated with IL-6 antagonists (750/3428; 21.9%) and in those treated with usual care or placebo (330/1787; 17.6%) (OR accounting for trial sample sizes, 0.99 [95% CI, 0.85-1.16]).

    Additional Analyses

    The results of the prespecified and post hoc sensitivity analyses for the outcomes of 28-day all-cause mortality; progression to IMV, ECMO, or death by 28 days; and secondary infections by 28 days appear in eTable 7 in Supplement 1. After excluding the large RECOVERY trial, the ORs comparing tocilizumab with usual care or placebo were 0.82 (95% CI, 0.68-0.98) for 28-day all-cause mortality and 0.71 (95% CI, 0.59-0.84) for progression to IMV or death within 28 days (consistent with the primary analyses). The ORs for the trials at low risk of bias were similar to the overall ORs. The ORs restricted to trials published in peer-reviewed journals were consistent with the overall ORs for tocilizumab but were imprecisely estimated for sarilumab because of exclusion of the largest trial. The ORs were similar for open-label and placebo-controlled trials; however, the association of sarilumab compared with usual care for secondary infections appeared more marked in open-label trials (1.97 [95% CI, 0.89-4.34]) than in placebo-controlled trials (0.96 [95% CI, 0.74-1.24]). Supplement 1 includes summary details for all of the sensitivity analyses. Supplements 2-10 include details of prespecified sensitivity analyses by risk of bias and blinding status. Further additional analyses for all outcomes within patients receiving and not receiving corticosteroids at randomization appear in Supplement 11. The baseline and outcome data collection forms appear in Supplement 12. The prospective meta-analysis protocol appears in Supplement 13.

    Discussion

    In this prospective meta-analysis based on 10 930 patients hospitalized for COVID-19 from 27 randomized clinical trials, administration of IL-6 antagonists was associated with lower all-cause mortality 28 days after randomization. Administration of IL-6 antagonists also was associated with lower progression to IMV or death, cardiovascular support or death, and KRT or death in patients not receiving support for the corresponding organ at randomization and with a greater probability of being discharged alive by 28 days. Administration of IL-6 antagonists was not associated with an increased risk of 28-day infection compared with usual care or placebo. There was no clear association between administration of IL-6 antagonists and all-cause mortality at 90 days or in the duration of IMV among patients who required IMV at randomization; however, the data were limited.

    Among the a priori–defined subgroups, the association of IL-6 antagonists with improved outcomes appeared more marked among patients who were receiving corticosteroids at randomization compared with those who were not. The association of IL-6 antagonists with lower 28-day all-cause mortality was more marked among patients who did not require IMV at randomization, consistent with the inverse association of progression to IMV or death among these patients. However, these differences between subgroups may have arisen due to sampling variation. Associations appeared broadly consistent across patient subgroups according to levels of cardiovascular support, C-reactive protein level, age, and sex.

    In general, associations with improved outcomes were more marked for tocilizumab than for sarilumab, although comparisons between tocilizumab and sarilumab were indirect (made between trials). However, the trials of sarilumab were generally done earlier in the pandemic than those of tocilizumab and before corticosteroids became the standard of care.7 The majority of patients in trials of sarilumab were not receiving corticosteroids at randomization, whereas the majority of patients in trials of tocilizumab were receiving corticosteroids at randomization. When comparisons were made within groups defined by receipt of corticosteroids at randomization, the differences between associations for these 2 IL-6 antagonists were less marked. Nearly 3 times as many patients were randomized to trials comparing tocilizumab with usual care or placebo compared with trials comparing sarilumab with usual care or placebo. For this reason, associations were estimated more precisely for tocilizumab than for sarilumab. Both drugs were IL-6 receptor antagonists, but there may be differences between tocilizumab and sarilumab in receptor binding or lung concentrations.25 Concurrent administration of IL-6 antagonists5 and corticosteroids,26 which both have anti-inflammatory effects, may provide greater improvement than either type of drug given individually.8,9

    This prospective meta-analysis included an estimated 97% of patients randomized to IL-6 receptor antagonists vs usual care worldwide. Because data were shared based on standardized definitions of outcomes and subgroups agreed upon in advance, these aggregate data meta-analyses had many of the advantages of individual-patient data meta-analyses while avoiding the need to establish formal data sharing agreements. The methods used in this meta-analysis limit bias in the selection and appraisal of trials with prespecified subgroup analyses based on clinically relevant questions. For tocilizumab, the results from other trials were similar to those from the large RECOVERY trial, supporting generalizability of the findings across settings.

    Limitations

    This study has several limitations. First, some of the included trials are ongoing and have not been published in peer-reviewed journals. It is possible that lack of participation or participation by some of the ongoing trials may be based on knowledge of their interim results. This limitation was addressed in the sensitivity analyses and the results were consistent with the primary analyses.

    Second, there were limited data for some comparisons and questions of interest such as IL-6 antagonists vs corticosteroids and the effect of siltuximab. Third, potential differences in treatment effect by differences in the baseline risk of death (eg, that arose either from trial-specific eligibility criteria, geographic differences, or improving trends in the outcomes of patients with COVID-19 during the pandemic) could not be accounted for.

    Fourth, the definitions and reporting of serious adverse events were not consistent across the trials and therefore a meta-analysis for this secondary end point was not conducted. Fifth, larger trials were mainly conducted in high-income settings; 65.9% of the tocilizumab data were provided by participants in the RECOVERY trial9 and 71.0% of the sarilumab data were provided by participants in the Regeneron trial (NCT04315298).

    Conclusions

    In this prospective meta-analysis of clinical trials of patients hospitalized for COVID-19, administration of IL-6 antagonists, compared with usual care or placebo, was associated with lower 28-day all-cause mortality.

    Back to top
    Article Information

    Corresponding Author: Manu Shankar-Hari, MSc, PhD, St Thomas’ Hospital, Guy’s and St Thomas’ NHS Foundation Trust London, East Wing, London SE17EH, England (manu.shankar-hari@kcl.ac.uk).

    Accepted for Publication: June 24, 2021.

    Published Online: July 6, 2021. doi:10.1001/jama.2021.11330

    The WHO Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group authors and collaborators: Manu Shankar-Hari, PhD, MSc; Claire L. Vale, PhD; Peter J. Godolphin, PhD; David Fisher, MSc; Julian P. T. Higgins, PhD; Francesca Spiga, PhD; Jelena Savović, PhD; Jayne Tierney, PhD; Gabriel Baron, PhD; Julie S. Benbenishty, PhD; Lindsay R. Berry, PhD; Niklas Broman, MD; Alexandre Biasi Cavalcanti, MD; Roos Colman, MSc; Stefanie L. De Buyser, PhD; Lennie P. G. Derde, MD, PhD; Pere Domingo, PhD; Sharifah Faridah Omar, MD; Ana Fernandez-Cruz, MD, PhD; Thijs Feuth, MD, PhD; Felipe Garcia, MD, PhD; Rosario Garcia-Vicuna, PhD; Isidoro Gonzalez-Alvaro, PhD; Anthony C. Gordon, MBBS, MD; Richard Haynes, DM; Olivier Hermine, MD, PhD; Peter W. Horby, MBBS; Nora K. Horick, MS; Kuldeep Kumar, MSc; Bart N. Lambrecht, MD, PhD; Martin J. Landray, PhD; Lorna Leal, MD, PhD; David J. Lederer, MD; Elizabeth Lorenzi, PhD; Xavier Mariette, MD, PhD; Nicolas Merchante, MD, PhD; Nor Arisah Misnan, MMed; Shalini V. Mohan, MD; Michael C. Nivens, PhD; Jarmo Oksi, MD, PhD; Jose A. Perez-Molina, PhD; Reuven Pizov, MD; Raphael Porcher, PhD; Simone Postma, MD; Reena Rajasuriar, PhD; Athimalaipet V. Ramanan, MD; Philippe Ravaud, MD, PhD; Pankti D. Reid, MD, MPH; Abraham Rutgers, MD, PhD; Aranzazu Sancho-Lopez, MD, PhD; Todd B. Seto, MD; Sumathi Sivapalasingam, MD; Arvinder Singh Soin, MS; Natalie Staplin, PhD; John H. Stone, MD, MPH; Garth W. Strohbehn, MD; Jonas Sunden-Cullberg, MD, PhD; Julian Torre-Cisneros, PhD; Larry W. Tsai, MD; Hubert van Hoogstraten, MD, PhD; Tom van Meerten, MD, PhD; Viviane Cordeiro Veiga, PhD; Peter E. Westerweel, MD, PhD; Srinivas Murthy, MD, CM, MHSc; Janet V. Diaz, MD; John C. Marshall, MD; Jonathan A. C. Sterne, PhD.

    Affiliations of The WHO Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group authors and collaborators: Guy’s and St Thomas’ NHS Foundation Trust, ICU Support Offices, St Thomas’ Hospital, London, England (Shankar-Hari); School of Immunology and Microbial Sciences, Kings College London, London, England (Shankar-Hari); University College London, MRC Clinical Trials Unit at UCL, London, England (Vale, Godolphin, Fisher, Tierney); Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, England (Higgins, Savović, Sterne); NIHR Bristol Biomedical Research Centre, Bristol, England (Higgins, Sterne); National Institute for Health Research Applied Research Collaboration West at University Hospitals Bristol and Weston NHS Foundation Trust, Bristol, England (Higgins, Savović); University of Bristol, Bristol, England (Spiga); Assistance Publique–Hôpitaux de Paris, Centre for Clinical Epidemiology, Hôpital Hôtel-Dieu, Paris, France (Baron, Porcher, Ravaud); INSERM UMRS-1153, Centre de Recherche Epidémiologie et Statistique Université de Paris, METHODS Team, Paris, France (Baron, Porcher, Ravaud); Department of Nursing, Hadassah Hebrew University Medical Center, Jerusalem, Israel (Benbenishty); Berry Consultants, Austin, Texas (Berry, Lorenzi); Turku University Hospital, Department of Infectious Diseases, Turku, Finland (Broman, Oksi); BP-A Beneficência Portuguesa de São Paulo, Rua Maestro Cardim, São Paulo, Brazil (Cavalcanti, Veiga); Department of Public Health and Primary Care, Ghent University, Ghent, Belgium (Colman, De Buyser); Department of Intensive Care Medicine, University Medical Center Utrecht, Utrecht, the Netherlands (Derde); Department of Infectious Diseases, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain (Domingo); Department of Medicine, University of Malaya, Kuala Lumpur, Malaysia (Omar, Rajasuriar); Infectious Diseases Unit, Internal Medicine Department, Hospital Universitario Puerta de Hierro-Majadahonda, Madrid, Spain (Fernandez-Cruz); Department of Pulmonary Diseases, Turku University Hospital, Turku, Finland (Feuth); Infectious Diseases Department, Hospital Clinic Barcelona-IDIBAPS, Barcelona, Spain (Garcia, Leal); Rheumatology, Hospital Universitario La Princesa IIS-IP, Madrid, Spain (Garcia-Vicuna, Gonzalez-Alvaro); Division of Anaesthetics, Pain Medicine and Intensive Care, Imperial College London, London, England (Gordon); Nuffield Department of Population Health, University of Oxford, Oxford, England (Haynes, Landray, Staplin); MRC Population Health Research Unit, University of Oxford, Oxford, England (Haynes, Landray, Staplin); Department of Hematology, Necker Hospital, Paris, France (Hermine); Imagine Institute, University of Paris, INSERM U1153, Paris, France (Hermine); Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, England (Horby); International Severe Acute Respiratory and Emerging Infections Consortium, University of Oxford, Oxford, England (Horby); Pandemic Sciences Centre, University of Oxford, Oxford, England (Horby); Department of Medicine, Massachusetts General Hospital, Boston (Horick); Medanta-The Medicity, Institute of Liver Transplantation and Regenerative Medicine, Gurugram, India (Kumar, Soin); Research Department, Medanta Institute of Education and Research, Gurugram, India (Kumar); VIB Center for Inflammation Research, Ghent University, Ghent, Belgium (Lambrecht); Department of Internal Medicine and Pediatrics, Ghent University, Ghent, Belgium (Lambrecht); Regeneron Pharmaceuticals Inc, Tarrytown, New York (Lederer, Nivens, Sivapalasingam); Centre for Immunology of Viral Infections and Autoimmune Diseases, Université Paris-Saclay, INSERM UMR1184, Le Kremlin-Bicêtre, Paris, France (Mariette); Department of Rheumatology, Assistance Publique–Hôpitaux de Paris, Le Le Kremlin-Bicêtre, Paris, France (Mariette); Unit of Infectious Diseases and Microbiology, Valme University Hospital, Institute of Biomedicine of Sevilla, Seville, Spain (Merchante); Hospital Sungai Buloh, Ministry of Health, Buloh, Malaysia (Misnan); Genentech, South San Francisco, California (Mohan, Tsai); Hospital Universitario Ramón y Cajal IRYCIS, Infectious Diseases Department, Madrid, Spain (Perez-Molina); Department of Anesthesilogy Critical Care and Pain Medicine, Hadassah Hebrew University Hospital, Jerusalem, Israel (Pizov); University de Paris, CRESS UMR1153, INSERM, INRA, Paris, France (Porcher, Ravaud); Department of Rheumatology and Clinical Immunology, University Hospital Groningen, University Medical Center Groningen, Groningen, the Netherlands (Postma, Rutgers); Peter Doherty Institute for Infection and Immunity, University of Melbourne, Melbourne, Australia (Rajasuriar); Department of Paediatric Rheumatology, University Hospitals Bristol, NHS Foundation Trust, Bristol, England (Ramanan); Department of Medicine (Rheumatology), University of Chicago Medical Center, Chicago, Illinois (Reid); Department of Clinical Pharmacology, Hospital Universitario Puerta de Hierro Majadahonda, Madrid, Spain (Sancho-Lopez); Center for Outcomes Research and Evaluation, Queen’s Medical Center, Honolulu, Hawaii (Seto); Division of Rheumatology, Allergy, and Immunology, Massachusetts General Hospital, Boston (Stone); Department of Medicine (Rheumatology), Massachusetts General Hospital, Boston (Stone); VA Ann Arbor, Center for Clinical Management and Research, Ann Arbor, Michigan (Strohbehn); Department of Infectious Diseases, Karolinska Institute at Karolinska University Hospital Huddinge, Stockholm, Sweden (Sunden-Cullberg); Maimonides Institute for Biomedical Research of Cordoba/Reina Sofia University Hospital/University of Córdoba, Córdoba, Spain (Torre-Cisneros); Global Medical Affairs, Sanofi-Genzyme, Bridgewater, New Jersey (van Hoogstraten); Department of Hematology, University Medical Center Groningen, Groningen, the Netherlands (van Meerten); Department of Internal Medicine, Albert Schweitzer Hospital, Dordrecht, the Netherlands (Westerweel); Department of Pediatrics, University of British Columbia, Vancouver, Canada (Murthy); Clinical Unit, Health Emergencies Programme, World Health Organization, Geneva, Switzerland (Diaz); Li Ka Shing Knowledge Institute, St Michael’s Hospital, University of Toronto, Toronto, Ontario, Canada (Marshall); Health Data Research UK South-West, Bristol, England (Sterne).

    Author Contributions: Dr Vale had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs Shankar-Hari and Vale contributed equally.

    Concept and design: Shankar-Hari, Vale, Godolphin, Tierney, Misnan, Benbenishty, Omar, Garcia, Garcia-Vicuna, Gordon, Hermine, Perez-Molina, Pizov, Rajasuriar, Reid, Sancho-Lopez, Stone, Tsai, Veiga, Murthy, Diaz, Marshall, Sterne.

    Acquisition, analysis, or interpretation of data: Shankar-Hari, Vale, Godolphin, Fisher, Higgins, Spiga, Savovic, Misnan, Baron, Benbenishty, Berry, Broman, Biasi Cavalcanti, Colman, De Buyser, Domingo, Derde, Omar, Fernández-Cruz, Feuth, Garcia, Garcia-Vicuna, Gonzalez-Alvaro, Gordon, Haynes, Hermine, Horby, Horick, Kumar, Lambrecht, Landray, Leal, Lederer, Lorenzi, Mariette, Merchante, Mohan, Nivens, Oksi, Perez-Molina, Pizov, Porcher, Postma, Rajasuriar, Ramanan, Ravaud, Reid, Rutgers, Seto, Sancho-Lopez, Sivapalasingam, Soin, Staplin, Strohbehn, Sunden-Cullberg, Torre-Cisneros, Tsai, van Hoogstraten, van Meerten, Veiga, Westerweel, Murthy, Marshall, Sterne.

    Drafting of the manuscript: Shankar-Hari, Vale, Higgins, Baron, Benbenishty, Omar, Kumar, Pizov, Postma, Rajasuriar, Reid, Murthy, Sterne.

    Critical revision of the manuscript for important intellectual content: Shankar-Hari, Vale, Godolphin, Fisher, Higgins, Spiga, Savovic, Tierney, Misnan, Baron, Berry, Broman, Biasi Cavalcanti, Colman, De Buyser, Domingo, Derde, Fernández-Cruz, Feuth, Garcia, Garcia-Vicuna, Gonzalez-Alvaro, Gordon, Haynes, Hermine, Horby, Horick, Lambrecht, Landray, Leal, Lederer, Lorenzi, Mariette, Merchante, Mohan, Nivens, Oksi, Perez-Molina, Porcher, Ramanan, Ravaud, Reid, Rutgers, Seto, Sancho-Lopez, Sivapalasingam, Soin, Staplin, Strohbehn, Stone, Sunden-Cullberg, Torre-Cisneros, Tsai, van Hoogstraten, van Meerten, Veiga, Westerweel, Murthy, Diaz, Marshall, Sterne.

    Statistical analysis: Shankar-Hari, Vale, Godolphin, Fisher, Spiga, De Buyser, Feuth, Gonzalez-Alvaro, Horick, Kumar, Lorenzi, Merchante, Nivens, Oksi, Postma, Sancho-Lopez, Strohbehn, Murthy, Sterne.

    Obtained funding: Derde, Gordon, Hermine, Lambrecht, Lederer, Ravaud, Stone.

    Administrative, technical, or material support: Shankar-Hari, Vale, Misnan, Benbenishty, Derde, Gordon, Horby, Leal, Lederer, Mariette, Mohan, Nivens, Rajasuriar, Ramanan, Ravaud, Sancho-Lopez, Tsai, Veiga, Westerweel, Murthy, Diaz.

    Supervision: Shankar-Hari, Higgins, Benbenishty, Derde, Garcia, Gordon, Hermine, Lambrecht, Nivens, Oksi, Pizov, Sancho-Lopez, Stone, Torre-Cisneros, Westerweel, Murthy, Diaz, Marshall, Sterne.

    Conflict of Interest Disclosures: Dr Shankar-Hari reported being supported by clinician scientist award NIHR-CS-2016-16-011 from the UK National Institute for Health Research. Dr Vale, Mr Fisher, and Dr Tierney reported being supported by grant MC_UU_12023/24 from the UK Medical Research Council. Dr Godolphin reported being fully supported and Mr Fisher was partially supported by grant RIA 16-ST2-020 from Prostate Cancer UK. Dr Higgins reported being supported by senior investigator award NF-SI-0617-10145 from the UK National Institute for Health Research. Drs Higgins and Savović were supported by grants from the UK National Institute for Health Research Applied Research Collaboration West. Drs Higgins and Sterne reported being supported by grants from the UK National Institute for Health Research Bristol Biomedical Research Centre, Weston NHS Foundation Trust, and the University of Bristol. Dr Berry reported receiving grants from Berry Consultants. Dr Derde reported being a member of the COVID-19 guideline committee for the Society of Critical Care Medicine/European Society of Intensive Care Medicine/Surviving Sepsis Campaign. Drs De Buyser and Lambrecht reported being supported by grants from Belgian Health Care Knowledge Centre. Dr Domingo reported receiving support from the General Subdirectorate of Networks and Cooperative Research Centres, Ministry of Science and Innovation, Spanish Network for Research in Infectious Diseases; being co-financed by the European Regional Development Fund; receiving grant support from the Instituto de Salud Carlos III for the TOCOVID clinical trial; and receiving honoraria from Merck Sharp & Dohme, Gilead Sciences, ViiV Healthcare, Janssen, Cilag, Theratechnologies, and Roche. Dr Omar reported being supported by grants from the University of Malaya. Dr Garcia-Vicuna reported receiving grants and personal fees from Sanofi and Lilly. Dr Gonzalez-Alvaro reported receiving grants from Sanofi, Biohope, and Gebro; serving on advisory boards for Lilly and Sanofi; receiving personal fees from Lilly, Sanofi, Roche, Bristol Myers Squibb, Merck Sharp & Dohme, Abbvie, Pfizer, and Novartis; and owning stock in PharmaMar. Dr Gordon reported being supported by grants from the UK National Institute for Health Research and the European Union and receiving personal fees from Thirty Respiratory Ltd and GlaxoSmithKline. Dr Haynes reported receiving study drugs from Roche and receiving grants from Novartis and Boehringer Ingelheim. Dr Hermine reported receiving grants from Celgene, Bristol Myers Squibb, Alexion, Inatherys, and AB Science. Dr Horby reported receiving study drugs from Roche. Dr Horby reported being supported by grants from UK Research and Innovation–National Institute for Health Research. Dr Lambrecht reported receiving consultancy fees from GlaxoSmithKline, Sanofi, Argenx, Oncoarendi, and Novartis. Dr Landray reported receiving nonfinancial support from Roche and Regeneron and receiving grants from Boehringer Ingelheim, Novartis, and Janssen. Drs Lederer, Nivens, and Sivapalasingam reported being employees of and owning stock in Regeneron. Dr Lorenzi reported receiving personal fees from Berry Consultants. Dr Mariette reported receiving personal fees from Bristol Myers Squibb, Gilead, GlaxoSmithKline, Janssen, Novartis, Pfizer, UCB, and Sanofi and receiving grants from Biogen, Ose Pharmaceutical, and Pfizer. Dr Merchante reported receiving grants from Merck Sharp & Dome and personal fees from Gilead, Merck Sharp & Dome, and Shionogi for providing expert testimony. Drs Mohan and Tsai reported being employees of Genetech. Dr Nivens also reported receiving grants from the Biomedical Advanced Research and Development Authority. Dr Perez-Molina reported receiving grants from Roche Spain. Dr Ramanan reported receiving personal fees from Roche, Abbvie, Eli Lilly, Novartis, UCB, and the Swedish Orphan Biovitrum AB. Dr Ravaud reported being the primary investigator of the Corimmuno platform funded by the Ministry of Health in France. Drs Reid and Strohbehn reported being co-inventors of a filed patent (held by the University of Chicago) covering the use of low-dose tocilizumab for treatment of viral infections. Dr Staplin reported receiving study drugs from Roche and Regeneron and receiving grants from Boehringer Ingelheim. Dr Stone reported receiving grants from Roche/Genentech. Dr Strohbehn also reported being an employee of the US government. Dr Sunden-Cullberg reported receiving grants from the Swedish Research Council and the Center for Innovative Medicine. Dr Torre-Cisneros reported being supported by General Sub-Directorate of Networks and Cooperative Research Centres, Ministry of Science and Innovation, Spanish Network for Research in Infectious Diseases; and being co-financed by the European Regional Development Fund. Dr Tsai also reported being involved with 2 patents pending that were filed and are owned by Genentech/Roche (one for a method to treating pneumonia, including COVID-19 pneumonia with an IL-6 antagonist, and another for tocilizumab and remdesivir combination therapy for COVID-19 pneumonia). Dr van Hoogstraten reported being an employee of and owning stock in Sanofi Genzyme. Dr Veiga reported receiving personal fees from Aspen Pharmacare, Cristália, and Pfizer for speaking and serving on advisory boards. Dr Murthy reported receiving grants from the Canadian Institutes of Health Research, Innovative Medicines Canada, and the Canadian Health Research Foundation. Dr Marshall reported receiving personal fees from AM Pharma (for serving as the chair of a data and safety monitoring board), Gilead (for serving as a consultant), and Critical Care Medicine (for serving as associate editor). Dr Sterne also reported being supported by grants from Health Data Research UK. No other disclosures were reported.

    Funding/Support: Funding for administrative and communications support was provided by the World Health Organization.

    Role of the Funder/Sponsor: The World Health Organization (WHO) had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The WHO contributed to the design and conduct of the study by convening the WHO COVID-19 Clinical Management and Characterization Working Group. This group assembled information on ongoing trials and invited trial investigators to participate in this prospective meta-analysis. The WHO chief scientist invited trial investigators to participate and provided a secure portal for submission of data. Other than the contributions of Dr Diaz as a co-author, the WHO had no role in the preparation, review, or approval of the manuscript. The WHO had no role in the decision to submit the manuscript for publication.

    Group Information: The WHO Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group nonauthor collaborators are listed in Supplement 14.

    Disclaimer: The views expressed in this article are those of the authors and do not necessarily reflect the opinions of the UK National Health Service, the UK National Institute for Health Research, UK Medical Research Council, the UK Department of Health and Social Care, or the US government.

    References
    1.
    Leisman  DE, Ronner  L, Pinotti  R,  et al.  Cytokine elevation in severe and critical COVID-19: a rapid systematic review, meta-analysis, and comparison with other inflammatory syndromes.   Lancet Respir Med. 2020;8(12):1233-1244. doi:10.1016/S2213-2600(20)30404-5PubMedGoogle ScholarCrossref
    2.
    Laing  AG, Lorenc  A, Del Molino Del Barrio  I,  et al.  A dynamic COVID-19 immune signature includes associations with poor prognosis.   Nat Med. 2020;26(10):1623-1635. doi:10.1038/s41591-020-1038-6PubMedGoogle ScholarCrossref
    3.
    Mathew  D, Giles  JR, Baxter  AE,  et al; UPenn COVID Processing Unit.  Deep immune profiling of COVID-19 patients reveals distinct immunotypes with therapeutic implications.   Science. 2020;369(6508):eabc8511. doi:10.1126/science.abc8511PubMedGoogle Scholar
    4.
    Marshall  JC, Murthy  S, Diaz  J,  et al; WHO Working Group on the Clinical Characterisation and Management of COVID-19 Infection.  A minimal common outcome measure set for COVID-19 clinical research.   Lancet Infect Dis. 2020;20(8):e192-e197. doi:10.1016/S1473-3099(20)30483-7PubMedGoogle ScholarCrossref
    5.
    McElvaney  OJ, Curley  GF, Rose-John  S, McElvaney  NG.  Interleukin-6: obstacles to targeting a complex cytokine in critical illness.   Lancet Respir Med. 2021;9(6):643-654. doi:10.1016/S2213-2600(21)00103-XPubMedGoogle ScholarCrossref
    6.
    Tierney  JF, Fisher  DJ, Vale  CL,  et al.  A framework for prospective, adaptive meta-analysis (FAME) of aggregate data from randomised trials.   PLoS Med. 2021;18(5):e1003629. doi:10.1371/journal.pmed.1003629PubMedGoogle Scholar
    7.
    Sterne  JAC, Murthy  S, Diaz  JV,  et al; WHO Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group.  Association between administration of systemic corticosteroids and mortality among critically ill patients with COVID-19: a meta-analysis.   JAMA. 2020;324(13):1330-1341. doi:10.1001/jama.2020.17023PubMedGoogle Scholar
    8.
    Gordon  AC, Mouncey  PR, Al-Beidh  F,  et al; REMAP-CAP Investigators.  Interleukin-6 receptor antagonists in critically ill patients with Covid-19.   N Engl J Med. 2021;384(16):1491-1502. doi:10.1056/NEJMoa2100433PubMedGoogle Scholar
    9.
    Abani  O, Abbas  A, Abbas  F,  et al; RECOVERY Collaborative Group.  Tocilizumab in patients admitted to hospital with COVID-19 (RECOVERY): a randomised, controlled, open-label, platform trial.   Lancet. 2021;397(10285):1637-1645. doi:10.1016/S0140-6736(21)00676-0PubMedGoogle ScholarCrossref
    10.
    Hermine  O, Mariette  X, Tharaux  PL, Resche-Rigon  M, Porcher  R, Ravaud  P; CORIMUNO-19 Collaborative Group.  Effect of tocilizumab vs usual care in adults hospitalized with COVID-19 and moderate or severe pneumonia: a randomized clinical trial.   JAMA Intern Med. 2021;181(1):32-40. doi:10.1001/jamainternmed.2020.6820PubMedGoogle ScholarCrossref
    11.
    Rosas  IO, Bräu  N, Waters  M,  et al.  Tocilizumab in hospitalized patients with severe Covid-19 pneumonia.   N Engl J Med. 2021;384(16):1503-1516. doi:10.1056/NEJMoa2028700PubMedGoogle ScholarCrossref
    12.
    Salama  C, Han  J, Yau  L,  et al.  Tocilizumab in patients hospitalized with Covid-19 pneumonia.   N Engl J Med. 2021;384(1):20-30. doi:10.1056/NEJMoa2030340PubMedGoogle ScholarCrossref
    13.
    Stone  JH, Frigault  MJ, Serling-Boyd  NJ,  et al; BACC Bay Tocilizumab Trial Investigators.  Efficacy of tocilizumab in patients hospitalized with Covid-19.   N Engl J Med. 2020;383(24):2333-2344. doi:10.1056/NEJMoa2028836PubMedGoogle ScholarCrossref
    14.
    Veiga  VC, Prats  JAGG, Farias  DLC,  et al; Coalition COVID-19 Brazil VI Investigators.  Effect of tocilizumab on clinical outcomes at 15 days in patients with severe or critical coronavirus disease 2019: randomised controlled trial.   BMJ. 2021;372(n84):n84. doi:10.1136/bmj.n84PubMedGoogle Scholar
    15.
    Sterne  JAC, Savović  J, Page  MJ,  et al.  RoB 2: a revised tool for assessing risk of bias in randomised trials.   BMJ. 2019;366:l4898. doi:10.1136/bmj.l4898PubMedGoogle Scholar
    16.
    Langan  D, Higgins  JPT, Jackson  D,  et al.  A comparison of heterogeneity variance estimators in simulated random-effects meta-analyses.   Res Synth Methods. 2019;10(1):83-98. doi:10.1002/jrsm.1316PubMedGoogle ScholarCrossref
    17.
    Nikolakopoulou  A, Mavridis  D, Salanti  G.  Demystifying fixed and random effects meta-analysis.   Evid Based Ment Health. 2014;17(2):53-57. doi:10.1136/eb-2014-101795PubMedGoogle ScholarCrossref
    18.
    Serghiou  S, Goodman  SN.  Random-effects meta-analysis: summarizing evidence with caveats.   JAMA. 2019;321(3):301-302. doi:10.1001/jama.2018.19684PubMedGoogle ScholarCrossref
    19.
    Rücker  G, Cates  CJ, Schwarzer  G.  Methods for including information from multi-arm trials in pairwise meta-analysis.   Res Synth Methods. 2017;8(4):392-403. doi:10.1002/jrsm.1259PubMedGoogle ScholarCrossref
    20.
    Fisher  DJ, Carpenter  JR, Morris  TP, Freeman  SC, Tierney  JF.  Meta-analytical methods to identify who benefits most from treatments: daft, deluded, or deft approach?   BMJ. 2017;356:j573. doi:10.1136/bmj.j573PubMedGoogle Scholar
    21.
    White  IR.  Network meta-analysis.   Stata J. 2015;15(4):951-985. doi:10.1177/1536867X1501500403 Google ScholarCrossref
    22.
    Fisher  DJ.  Two-stage individual participant data meta-analysis and generalized forest plots.   Stata J. 2015;15(2):369-396. doi:10.1177/1536867X1501500203Google ScholarCrossref
    23.
    Lescure  FX, Honda  H, Fowler  RA,  et al; Sarilumab COVID-19 Global Study Group.  Sarilumab in patients admitted to hospital with severe or critical COVID-19: a randomised, double-blind, placebo-controlled, phase 3 trial.   Lancet Respir Med. 2021;9(5):522-532. doi:10.1016/S2213-2600(21)00099-0PubMedGoogle ScholarCrossref
    24.
    Soin  AS, Kumar  K, Choudhary  NS,  et al.  Tocilizumab plus standard care versus standard care in patients in India with moderate to severe COVID-19-associated cytokine release syndrome (COVINTOC): an open-label, multicentre, randomised, controlled, phase 3 trial.   Lancet Respir Med. 2021;9(5):511-521. doi:10.1016/S2213-2600(21)00081-3PubMedGoogle ScholarCrossref
    25.
    Xu  C, Rafique  A, Potocky  T,  et al.  Differential binding of sarilumab and tocilizumab to IL-6Rα and effects of receptor occupancy on clinical parameters.   J Clin Pharmacol. 2021;61(5):714-724. doi:10.1002/jcph.1795PubMedGoogle ScholarCrossref
    26.
    Rhen  T, Cidlowski  JA.  Antiinflammatory action of glucocorticoids—new mechanisms for old drugs.   N Engl J Med. 2005;353(16):1711-1723. doi:10.1056/NEJMra050541PubMedGoogle ScholarCrossref
    ×