Tocilizumab in patients admitted to hospital with COVID-19 (RECOVERY): a randomised, controlled, open-label, platform trial

Randomisation and masking

Data were collected at study entry using a web-based case report form that included demographics and major comorbidities (appendix p 32). All eligible and consenting patients received usual standard of care and underwent an initial (main) randomisation comprising up to three parts in a factorial design (appendix pp 29–30): part 1, no additional treatment versus either dexamethasone, lopinavir–ritonavir, hydroxychloroquine, azithromycin, or colchicine; part 2, no additional treatment versus either convalescent plasma or REGN-COV2 (a combination of two monoclonal antibodies directed against SARS-CoV-2 spike protein); and part 3, no additional treatment versus aspirin. Over time, treatment groups were added to and removed from the protocol (appendix pp 26–29), and not all treatments were available at every hospital. Similarly, not all treatments were suitable for some patients (eg, owing to comorbid conditions or concomitant medication). In any of these cases, randomisation was between fewer groups.

Up to 21 days after the main randomisation and regardless of treatment allocation, RECOVERY trial participants with clinical evidence of progressive COVID-19 (defined as oxygen saturation <92% on room air or receiving oxygen therapy, and CRP ≥75 mg/L) could be considered for randomisation to tocilizumab versus usual care alone. Baseline data collected for this randomisation included amount of respiratory support, markers of progressive COVID-19 (including most recent oxygen saturation, CRP, ferritin, and creatinine result before second randomisation), suitability for the study treatment, and treatment availability at the site (appendix pp 33–34). For some patients, tocilizumab was unavailable at the hospital at the time of enrolment or was considered by the managing physician to be either definitely indicated or definitely contraindicated. In such cases, the patients were not eligible for the tocilizumab randomisation. Patients with known hypersensitivity to tocilizumab, evidence of active tuberculosis infection or clear evidence of active bacterial, fungal, viral, or other infection (besides COVID-19) were not eligible for randomisation to tocilizumab.

Patients who were eligible for randomisation to tocilizumab were assigned to either usual standard of care or usual standard of care plus tocilizumab in a 1:1 ratio by means of web-based simple (unstratified) randomisation with allocation concealed until after randomisation. Allocated treatment was prescribed by the managing doctor. Roche Products (Welwyn Garden City, UK) supported the trial through provision of tocilizumab. Participants and local study staff were not masked to the allocated treatment. The steering committee, investigators, and all others involved in the trial were masked to the outcome data during the trial.

Procedures

Patients allocated to tocilizumab were to receive tocilizumab as a single intravenous infusion over 60 min. The dose of tocilizumab was established by bodyweight (800 mg if weight >90 kg; 600 mg if weight >65 and ≤90 kg; 400 mg if weight >40 and ≤65 kg; and 8 mg/kg if weight ≤40 kg). A second dose could be given 12–24 h later if, in the opinion of the attending clinician, the patient’s condition had not improved.

A single online follow-up form was completed when participants were discharged, had died, or at 28 days after the initial randomisation, whichever occurred earliest (appendix pp 35–41). Information was recorded on adherence to allocated study treatment, receipt of other COVID-19 treatments, duration of admission, receipt of respiratory or renal support, and vital status (including cause of death). In addition, routine health-care and registry data were obtained for the full follow-up period, including information on vital status (with date and cause of death), discharge from hospital, receipt of respiratory support, or renal replacement therapy.

Outcomes

Outcomes were assessed at 28 days after randomisation to tocilizumab versus usual care alone, with further analyses specified at 6 months. The primary outcome was all-cause mortality. Secondary outcomes were time to discharge from hospital, and, among patients not receiving invasive mechanical ventilation at randomisation, receipt of invasive mechanical ventilation (including extracorporeal membrane oxygenation) or death. Prespecified subsidiary clinical outcomes were use of non-invasive respiratory support (defined as high-flow nasal oxygen, continuous positive airway pressure, or non-invasive ventilation), time to successful cessation of invasive mechanical ventilation (defined as cessation of invasive mechanical ventilation within, and survival to, 28 days), and use of renal dialysis or haemofiltration. Prespecified safety outcomes included cause-specific mortality and major cardiac arrhythmia.

Information on suspected serious adverse reactions was collected in an expedited fashion to comply with regulatory requirements.

Statistical analysis

In accordance with the statistical analysis plan (version 2.1, appendix pp 93–117), an intention-to-treat comparison was done between patients who entered the randomised comparison of tocilizumab versus usual care. For the primary outcome of 28-day mortality, the log-rank observed minus expected statistic and its variance were used to test the null hypothesis of equal survival curves (ie, the log-rank test) and to calculate the one-step estimate of the average mortality rate ratio. We constructed Kaplan-Meier survival curves to display cumulative mortality over the 28-day period. We used the same method to analyse time to hospital discharge and successful cessation of invasive mechanical ventilation, with patients who died in hospital right-censored on day 29. For the prespecified composite secondary outcome of invasive mechanical ventilation or death within 28 days (among those not receiving invasive mechanical ventilation at randomisation) and the subsidiary clinical outcomes of receipt of ventilation and receipt of haemodialysis or haemofiltration, the precise dates were not available and so the risk ratio was estimated instead.

Prespecified analyses of the primary outcome were done in subgroups defined by six characteristics at the time of randomisation: age, sex, ethnicity, amount of respiratory support, days since symptom onset, and use of systemic corticosteroids (including dexamethasone). Observed effects within subgroup categories were compared by means of a χ² test for heterogeneity or trend, in accordance with the prespecified analysis plan.

Estimates of rate and risk ratios are shown with 95% CIs. All p values are two-sided and are shown without adjustment for multiple testing. The full database is held by the study team which collected the data from study sites and did the analyses at the Nuffield Department of Population Health, University of Oxford (Oxford, UK).

Before commencement of the randomisation to tocilizumab versus usual care, the trial steering committee determined that if 28-day mortality in the usual care group was above 25% then recruitment of around 4000 patients to this comparison would provide 90% power at two-sided p=0·01 to detect a proportional reduction in 28-day mortality of one-fifth. Consequently, Roche Products provided sufficient treatment for 2000 patients to receive tocilizumab. The trial steering committee, masked to the results, closed recruitment to the tocilizumab comparison at the end of Jan 24, 2021, as over 4000 patients had been randomly assigned.

For the primary outcome of 28-day mortality, the results from RECOVERY were subsequently included in a meta-analysis of results from all previous randomised trials of tocilizumb versus usual care in patients with COVID-19. For each trial, we compared the observed number of deaths among patients allocated tocilizumab with the expected number if all patients were at equal risk (ie, we calculated the observed minus expected statistic [o–e], and its variance v). For RECOVERY, these were taken as the log-rank observed minus expected statistic and its variance but for other trials, where the exact timing of each death was not available, these were calculated from standard formulae for 2 × 2 contingency tables. We then combined trial results using the log of the mortality rate ratio calculated as the inverse-variance weighted average S/V with variance 1/V (and hence with 95% CI S/V ±1·96/√V), where S is the sum over all trials of (O–E) and V is the sum over all trials of v.

Analyses were done by means of SAS version 9.4 and R version 3.4. The trial is registered with ISRCTN (50189673) and ClinicalTrials.gov (NCT04381936).

Role of the funding source

Neither the funders of the study nor Roche Products had any role in study design, data collection, data analysis, data interpretation, or writing of the report. Roche Products supported the study through the supply of tocilizumab and reviewed the draft publication for factual accuracy relating to tocilizumab.