Test design and randomization
In this trial, which was carried out in 62 hospitals in nine countries in Europe and North America (Canada, Denmark, France, Germany, Italy, the Netherlands, Spain, the United Kingdom and the United States), we enrolled adults (≥ 18 years of age) with severe Covid-19 pneumonia, confirmed by a positive polymerase chain reaction (PCR) test of any body fluid and evidenced by bilateral chest infiltrates on chest radiography or computed tomography. Eligible patients had a blood oxygen saturation of 93% or less or a ratio between partial oxygen pressure and the inspired oxygen fraction of less than 300 mm Hg. Patients were excluded if the attending physician determined that death was imminent and inevitable within 24 hours or if they had active tuberculosis or a bacterial, fungal or viral infection other than SARS-CoV-2. Standard care according to local practice (antiviral treatment, low-dose glucocorticoids, convalescent plasma and supportive care) was provided. However, concomitant treatment with another investigated agent (except antiviral drugs) or any immunomodulatory agent was prohibited. Informed written consent was obtained from all patients or, if written consent could not be provided, the patient’s legally authorized representative could provide oral consent with appropriate documentation by the investigator.
Eligible patients were randomly assigned a 2: 1 ratio to receive a single intravenous infusion of tocilizumab (at a dose of 8 mg per kilogram of body weight, with a maximum dose of 800 mg) or placebo plus standard treatment using an interactive voice response system or web-based and randomization of permuted blocks. Randomization was stratified according to geographic region (North America or Europe) and the use of mechanical ventilation (yes or no). If the clinical signs or symptoms have not improved or worsened (defined as sustained fever or worsening of the clinical condition on an ordinal scale), a second infusion of tocilizumab or placebo can be administered 8 to 24 hours after the first dose. The primary analysis was carried out on the 28th, and the final trial visit took place on the 60th. Additional details on the trial design are provided in the protocol document (which includes the statistical analysis plan), available with the full text of this article at NEJM .org.
Assessments
For the evaluation of patients in this trial, the baseline was defined as the last observation before administration of tocilizumab or placebo on day 1. The clinical status of patients was assessed on an ordinal scale according to the following categories: 1, with high or ready for discharge; 2, admission to a non-intensive care unit (ICU) without supplemental oxygen; 3, hospitalization outside the ICU with supplemental oxygen; 4, admission to the ICU or not with non-invasive ventilation or high flow oxygen; 5, admission to the ICU with intubation and mechanical ventilation; 6, admission to the ICU with oxygenation by extracorporeal membrane or mechanical ventilation and additional organ support; and 7, death. The clinical status was recorded at the beginning of the study and every day during hospitalization.
Patients were also assessed according to the level of clinical severity in the National Early Warning Score 2, which is a standardized assessment to identify patients with acute illnesses based on respiratory rate, oxygen saturation, systolic blood pressure, pulse, level of consciousness , and temperature; the values in this instrument range from 0 to 20, with higher scores indicating greater clinical risk.
Outcome measures
The primary efficacy result was clinical status on day 28, as assessed on the ordinal scale of seven categories. The main results of secondary efficacy were clinical status on day 14 on the ordinal scale, mortality on day 28, number of days without a ventilator on day 28, time to improvement from baseline in at least two categories on the ordinal scale, and the time for hospital discharge or readiness for discharge; the latter was defined as normal body temperature and respiratory rate and stable oxygen saturation when breathing ambient air or 2 liters or less of supplemental oxygen. Other secondary outcomes were the time to clinical failure, which was defined as death, interruption of study participation during hospitalization, initiation of mechanical ventilation or transfer to the ICU, or worsening in 1 category of clinical status in patients who were receiving mechanical ventilation. or who were in the ICU at the beginning of the study; the beginning of mechanical ventilation among patients who were not receiving mechanical ventilation at randomization; the incidence of ICU transfer among patients who were not in an ICU at the beginning of the study; and the length of stay in the ICU. Adverse events were recorded according to the organ system class and preferred terms in the Medical Dictionary for Regulatory Activities, version 23.0.
Test supervision
The study was conducted in accordance with the International Harmonization Council E6 Good Clinical Practice guidelines and the principles of the Declaration of Helsinki or local regulations, which provide greater protection to the patient. The protocol was reviewed by the institutional review board or ethics committee at each location.
The first draft of the manuscript was written by the penultimate author, with writing support provided by ApotheCom and funded by the sponsor, F. Hoffmann – La Roche. The data were analyzed by the sponsor. The authors had access to all data on patients enrolled in their place of study. All authors made the decision to submit the manuscript for publication and attest to the integrity and accuracy of the data and the adherence of the trial to the protocol.
Statistical analysis
We conducted efficacy assessments of primary and secondary outcomes in the modified intent-to-treat population, which included all patients who underwent randomization and received a dose of tocilizumab or placebo. We calculated that a sample size of 450 patients would provide a 90% power to determine a difference between groups in the primary outcome (clinical status on day 28), assuming a distribution on the ordinal scale that corresponded to an odds ratio of 2.0 . If significance was reached, we tested mortality on day 28 at the 5% level using a hierarchical approach, but no further adjustments for multiple comparisons were planned. In the statistical analysis plan, up to three interim efficacy analyzes were specified, but were not performed due to the rapid recruitment.
The analyzes were stratified according to the region and the status of mechanical ventilation at randomization, except for some subgroup analyzes, as pre-specified. For the primary endpoint of clinical status on day 28, we compared the distribution on the ordinal scale using a non-parametric Van Elteren test. We used a proportional probability model to calculate the odds ratios and 95% confidence intervals to determine the chance of being in a better clinical category in the tocilizumab group than in the placebo group. A multiple imputation approach was used to deal with missing data and was implemented using a bootstrap. This approach assumed that data were missing at random within the strata and test group. (Details on these methods are provided in the Methods section of the Supplementary Appendix, available at NEJM.org.)
We used the Cochran – Mantel – Haenszel test to analyze differences in mortality and incidence of mechanical ventilation and ICU transfer, the van Elteren test to assess differences in the number of days without ventilation and a log-rank test and graphs of Kaplan – Meier to evaluate the secondary results in time analysis until the event. Data on deaths were censored on the 28th for all time analyzes until the event involving clinical improvement. Patients who died on the 28th were considered to have no days without ventilation.24 Patients who died or discontinued participation in the study before discharge on day 28 were considered to be in need of mechanical ventilation or transfer to the ICU for their respective incidence analyzes. Cumulative incidence plots were generated using the non-parametric Aalen-Johansen estimator, in which death is a competitive risk, and an additional cause-specific Cox regression was performed.
Safety was assessed in the population that included all patients who received a dose of tocilizumab or placebo, according to the study agent who was received first. Patients who received tocilizumab or placebo by mistake were included in the safety analysis.