Tocilizumab, sarilumab and anakinra in critically ill patients with COVID-19: a randomised, controlled, open-label, adaptive platform trial

Abstract

Introduction: Tocilizumab improves outcomes in critically ill patients with COVID-19. Whether other immune-modulator strategies are equally effective or better is unknown.

Methods: We investigated treatment with tocilizumab, sarilumab, anakinra and no immune modulator in these patients. In this ongoing, adaptive platform trial in 133 sites in 9 countries, we randomly assigned patients with allocation ratios dependent on the number of interventions available at each site. The primary outcome was an ordinal scale combining in-hospital mortality (assigned -1) and days free of organ support to day 21 in survivors. The trial used a Bayesian statistical model with predefined triggers for superiority, inferiority, efficacy, equivalence or futility.

Results: Of 2274 critically ill participants enrolled between 25 March 2020 and 10 April 2021, 972 were assigned to tocilizumab, 485 to sarilumab, 378 to anakinra and 418 to control. Median organ support-free days were 7 (IQR -1, 16), 9 (IQR -1, 17), 0 (IQR -1, 15) and 0 (IQR -1, 15) for tocilizumab, sarilumab, anakinra and control, respectively. Median adjusted ORs were 1.46 (95% credible intervals (CrI) 1.13, 1.87), 1.50 (95% CrI 1.13, 2.00) and 0.99 (95% CrI 0.74, 1.35) for tocilizumab, sarilumab and anakinra relative to control, yielding 99.8%, 99.8% and 46.6% posterior probabilities of superiority, respectively, compared with control. All treatments appeared safe.

Conclusions: In critically ill patients with COVID-19, tocilizumab and sarilumab have equivalent effectiveness at reducing duration of organ support and death. Anakinra is not effective in this population.

Trial registration number: NCT02735707.

Keywords: COVID-19; Critical Care; Pneumonia.

Figure 1. Screening, enrolment, randomisation and inclusion in analysis. ^aPatients could meet more than one ineligibility criterion. Full details are provided in online supplement 1. ^bContraindications include known hypersensitivity to an agent specified as an intervention; known or suspected pregnancy; known hypersensitivity to proteins produced by Escherichia coli (exclusion from receiving anakinra); baseline alanine aminotransferase or aspartate aminotransferase that is more than five times the upper limit of normal (exclusion from receiving tocilizumab or sarilumab); baseline platelet count <50×10⁹/L (exclusion from receiving tocilizumab or sarilumab). *Commencement of organ support in ICU was used instead of date and time of ICU admission for patients who had already received an allocation in the Moderate State. ICU, intensive care unit; REMAP-CAP, Randomised, Embedded, Multifactorial Adaptive Platform Trial for Community-Acquired Pneumonia.

Figure 2. Distributions of organ support-free days. (A) The cumulative proportion (y-axis) for each intervention group by day (x-axis), with death listed first. Curves that rise more slowly indicate a more favourable distribution in the number of days alive and free of organ support. The height of each curve at ‘–1’ indicates the in-hospital mortality rate for each intervention. The height of each curve at any point, for example, at day=10, indicates the proportion of patients with organ support-free days (OSFD) of 10 or lower (ie, 10 or worse). The difference in height of the two curves at any point represents the difference in the percentile in the distribution of OSFDs associated with that number of days alive and free of organ support. (B) Organ support-free days as horizontally stacked proportions by intervention group. Red represents worse outcomes and blue represents better outcomes. The median adjusted ORs from the primary analysis, using a Bayesian cumulative logistic model, were 1.46 (95% CrI 1.13, 1.87) for tocilizumab, 1.50 (95% CrI 1.13, 2.00) for sarilumab and 0.99 (95% CrI 0.74, 1.35) for anakinra, yielding 99.8%, 99.8% and 46.5% posterior probabilities of superiority, respectively, compared with control. CrI, credible intervals.

Figure 3. Time-to-event analysis. This plot is restricted to participants in the severe state randomised to the immune modulation therapy domain or another unblinded domain. The Kaplan-Meier curve for survival up to 90 days according to individual interventions is shown. There were 323, 161, 160 and 151 deaths in the tocilizumab, sarilumab, anakinra and control groups, respectively. This resulted in an HR of 1.39 (95%CrI 1.11 to 1.74) for tocilizumab, 1.44 (95% CrI 1.11 to 1.89) for sarilumab and 1.13 (95% CrI 0.87 to 1.49) for anakinra, yielding 99.9%, 99.6% and 82.3% respective posterior probabilities of superiority to control. These ‘survival HRs’ are defined as the reciprocal of the mortality HR to be consistent with the convention that ORs andHRs >1 imply benefit.