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

Findings: Between 23 April 2020 and 25 January 2021, 4116 adults were included in the assessment of tocilizumab, including 562 (14%) patients receiving invasive mechanical ventilation, 1686 (41%) receiving non-invasive respiratory support, and. 1868 (45%) receiving no respiratory support other than oxygen. Median CRP was 143 [IQR 107-205] mg/L and 3385 (82%) patients were receiving systemic corticosteroids at randomisation. Overall, 596 (29%) of the 2022 patients allocated tocilizumab and 694 (33%) of the 2094 patients allocated to usual care died within 28 days (rate ratio 0.86; 95% confidence interval [CI] 0.77-0.96; p=0.007). Consistent results were seen in all pre-specified subgroups of patients, including those receiving systemic corticosteroids. Patients allocated to tocilizumab were more likely to be discharged from hospital alive within 28 days (54% vs. 47%; rate ratio 1.23; 95% CI 1.12-1.34; p<0.0001). Among those not receiving invasive mechanical ventilation at baseline, patients allocated tocilizumab were less likely to reach the composite endpoint of invasive mechanical ventilation or death (33% vs. 38%; risk ratio 0.85; 95% CI 0.78-0.93; p=0.0005). Interpretation: In hospitalised COVID-19 patients with hypoxia and systemic inflammation, tocilizumab improved survival and other clinical outcomes regardless of the level of respiratory support received and in addition to the use of systemic corticosteroids.


INTRODUCTION 61
The majority of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) 62 infections are either asymptomatic or result in only mild disease. 1 However, a substantial 63 proportion of infected individuals develop a respiratory illness requiring hospital care, 64

Statistical Analysis 171
In accordance with the statistical analysis plan (version 2.1 appendix pp 93-117), an 172 intention-to-treat comparison was conducted between patients who entered the 173 randomised comparison of tocilizumab vs. usual care. For the primary outcome of 28-day 174 mortality, the log-rank observed minus expected statistic and its variance were used to 175 test the null hypothesis of equal survival curves (i.e., the log-rank test) and to calculate 176 the one-step estimate of the average mortality rate ratio. We constructed Kaplan-Meier 177 survival curves to display cumulative mortality over the 28-day period. For this preliminary 178 report, information on the primary outcome is available for 92% of randomised patients. 179 Those patients who had not been followed for 28 days and were not known to have died 180 by the time of the data cut for this preliminary analysis (8 February 2021) were either 181 censored on 8 February 2021 or, if they had already been discharged alive, were right-182 censored for mortality at day 29 (that is, in the absence of any information to the contrary 183 they were assumed to have survived 28 days). [Note: This censoring rule will not be 184 necessary when all patients have completed the 28 day follow-up period on 25 February 185 2021.] We used the same method to analyse time to hospital discharge and successful 186 cessation of invasive mechanical ventilation, with patients who died in hospital right-187 censored on day 29. For the pre-specified composite secondary outcome of invasive 188 mechanical ventilation or death within 28 days (among those not receiving invasive 189 mechanical ventilation at randomisation) and the subsidiary clinical outcomes of receipt 190 of ventilation and receipt of haemodialysis or haemofiltration, the precise dates were not 191 available and so the risk ratio was estimated instead. 192 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 11, 2021. ; https://doi.org/10.1101/2021.02.11.21249258 doi: medRxiv preprint Prior to commencement of the randomisation to tocilizumab vs. usual care, the trial 204 steering committee determined that if 28-day mortality in the usual care group was 205 above 25% then recruitment of around 4000 patients to this comparison would provide 206 90% power at two-sided P=0.01 to detect a proportional reduction in 28-day mortality of 207 one-fifth. Consequently, Roche Products Ltd provided sufficient treatment for 2000 208 patients to receive tocilizumab. The trial steering committee, masked to the results, 209 closed recruitment to the tocilizumab comparison at the end of January 24, 2021 as 210 over 4000 patients had been randomised. 211 Analyses were performed using SAS version 9.4 and R version 3.4. The trial is registered 212 with ISRCTN (50189673) and clinicaltrials.gov (NCT04381936). 213 214 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted February 11, 2021. ;

Role of the funding source 215
Neither the funders of the study nor Roche Products Ltd had any role in study design, 216 data collection, data analysis, data interpretation, or writing of the report. Roche Products 217 Ltd supported the study through the supply of tocilizumab and reviewed the draft 218 publication for factual accuracy relating to tocilizumab. The corresponding authors had 219 full access to all the data in the study and had final responsibility for the decision to submit 220 for publication. 221

RESULTS 223
Between 14 April 2020 and 24 January 2021, 4116 (19%) of 21550 patients enrolled into 224 the RECOVERY trial at one of the 131 sites in the UK participating in the tocilizumab 225 comparison were eligible for randomisation. 2022 patients were randomly allocated to 226 tocilizumab and 2094 were randomly allocated to usual care. The mean age of these 227 participants was 63·6 years (SD 13·7). At randomisation, 562 (14%) patients were 228 receiving invasive mechanical ventilation, 1686 (41%) were receiving non-invasive 229 respiratory support (including high-flow nasal oxygen, continuous positive airway 230 pressure, and non-invasive ventilation), and 1868 (45%) were receiving no respiratory 231 support other than simple oxygen therapy (9 of these patients were reportedly not 232 receiving oxygen at randomisation) ( Allocation to tocilizumab was associated with a greater probability of discharge from 253 hospital alive within 28 days (54% vs. 47%; rate ratio 1·22, 95% CI 1·12 to 1·34, p<0·0001; 254 figure 2b and table 2). Among those not on invasive mechanical ventilation at baseline, 255 allocation to tocilizumab was associated with a reduction in the risk of progressing to the 256 pre-specified composite secondary outcome of invasive mechanical ventilation or death 257 when compare to usual care alone (33% vs. 38%, risk ratio 0·85, 95% CI 0·78 to 0·93, 258 p=0·0005; table 2). 259 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted February 11, 2021. ; https://doi.org/10.1101/2021.02.11.21249258 doi: medRxiv preprint Tocilizumab for COVID-19

13
We observed similar results across all pre-specified sub-groups (figure 3 and webfigures 260 2 and 3), including the level of respiratory support at randomisation (figure 3). Given the 261 number of hypothesis tests conducted, the suggestion of a larger proportional mortality 262 reduction among those receiving a corticosteroid compared with those not [interaction 263 p=0.01] may reflect the play of chance. An exploratory analysis showed that the effects 264 of tocilizumab on 28-day mortality were similar for those randomised ≤2 or >2 days since 265 hospitalisation (interaction p=0.86). 266 In pre-specified subsidiary analyses, we found no significant effect of tocilizumab on 267 subsequent receipt of non-invasive respiratory support or invasive mechanical ventilation 268 among those on no respiratory support at randomisation (table 2, webfigure 1). Nor was 269 there a significant effect on the rate of successful cessation of invasive mechanical 270 ventilation among those on invasive mechanical ventilation at randomisation. However, 271 allocation to tocilizumab reduced the use of haemodialysis or haemofiltration (5% vs. 7%, 272 risk ratio 0·75, 95% CI 0·59 to 0·96, p=0·02; table 2). Preliminary information on cause-273 specific mortality shows no evidence of excess deaths from other infections (webtable 2). 274 We observed no significant differences in the frequency of new cardiac arrhythmias 275 CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

DISCUSSION 281
The results of this large, randomised trial indicate that tocilizumab is an effective 282 treatment for hospitalised COVID-19 patients who have hypoxia and evidence of 283 inflammation (CRP ≥75 mg/L). Treatment with tocilizumab improved survival and the 284 chances of discharge from hospital alive by 28 days, and reduced the chances of 285 progressing to require invasive mechanical ventilation. These benefits were consistent 286 across all patient groups studied, including those receiving invasive mechanical 287 ventilation, non-invasive respiratory support, or no respiratory support other than simple 288 oxygen. The benefits of tocilizumab were clearly seen among those also receiving 289 treatment with a systemic corticosteroid such as dexamethasone. There was also a 290 significant reduction in the need for haemodialysis or haemofiltration. Although we did not 291 see any effect on the duration of invasive mechanical ventilation, only 185 patients had 292 such ventilation successfully removed within the first 28 days, so power to detect any 293 benefit was low. 294 Since mid-2020, seven randomised controlled trials of tocilizumab for the treatment of 295 COVID-19 have reported. These include six small trials (fewer than 100 deaths in each) 296 and the somewhat larger REMAP-CAP trial, which recruited critically ill patients with 297 COVID-19, over 99% of whom required non-invasive respiratory support or invasive 298 mechanical ventilation. 8-14 Taken together, these previous trials did not show a significant 299 mortality benefit for treatment with tocilizumab (death rate ratio 0·91, 95% CI 0·72-1·14, 300 figure 4). The RECOVERY trial contains over three times as many deaths as all the 301 previous trials combined. When all 8 trials are considered together, allocation to 302 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 11, 2021. ; https://doi.org/10.1101/2021.02.11.21249258 doi: medRxiv preprint tocilizumab is associated with a 13% proportional reduction in 28-day mortality (death 303 rate ratio 0·87, 95% CI 0·79-0·96, p=0·005, figure 4). 304 These benefits are in addition to those previously reported for corticosteroids, which is 305 now usual standard of care for COVID-19 patients requiring treatment with oxygen. 6,18 306 Our data suggest that in COVID-19 patients who are hypoxic and have evidence of 307 systematic inflammation, treatment with a combination of a systemic corticosteroid plus 308 tocilizumab would be expected to reduce mortality by about one-third for patients 309 receiving simple oxygen and nearly one-half for those receiving invasive mechanical 310

ventilation. 311
Previous trials have provided weak evidence that tocilizumab may shorten time to 312 discharge or reduce progression to invasive mechanical ventilation or death. 10,14 Our 313 results show that in a broad spectrum of patients, tocilizumab increases the chances of 314 being discharged alive within 28 days and reduces the chance of progression to receiving 315 invasive mechanical ventilation. As with the mortality benefit, these effects are consistent 316 regardless of the level of respiratory support at the time of enrolment. 317 Strengths of this trial included that it was randomised, had a large sample size, and 318 included patients requiring various levels of respiratory support (from simple oxygen 319 through to invasive mechanical ventilation). There are some limitations: For this 320 preliminary report, information on the primary outcome is available for 92% of patients. 321 This should increase to >99% by early March when all patients have passed the 28-day 322 follow-up period. Following random assignment, 17% of patients in the tocilizumab group 323 did not receive this treatment. The reasons for this were not recorded. The size of the 324 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 11, 2021. ; https://doi.org/10.1101/2021.02.11.21249258 doi: medRxiv preprint effects of tocilizumab reported in this paper are therefore an underestimate of the true 325 effects of actually using the treatment. Hospital stay is very long for many of these patients 326 (median >28 days); the pre-planned analyses at 6 months will provide additional 327 information on the full effects of tocilizumab on clinical outcomes. 328 The RECOVERY results support the use of tocilizumab, but other IL-6 antagonists are 329 available. Although the effects of sarilumab in REMAP-CAP were similar to tocilizumab, 330 only 48 participants received sarilumab. 8 Two larger trials of sarilumab have completed 331 but have not reported any results (NCT044327388, NTC044315298). Publication of 332 results from those trials is now essential to assess whether alternative IL-6 antagonists 333 to tocilizumab are effective. 334 Guidelines on the use of IL-6 antagonists for patients with severe COVID-19 vary. For 335 example the US National Institutes for Health conclusion is that there are insufficient data 336 to recommend either for or against the use of tocilizumab or sarilumab, a view shared by 337 some commentators. 19,20 By contrast, interim guidance in the NHS states that IL-6 338 antagonists should be considered for patients within 24 hours of starting non-invasive 339 respiratory support or invasive mechanical ventilation. 21 Our results show that the benefits 340 of tocilizumab extend to a broader group of patients receiving oxygen with or without other 341 forms of respiratory support, and that those benefits include a reduction in the need for 342 invasive mechanical ventilation and other organ support such as renal replacement 343 therapy. Since complicating bacterial infections are infrequent in the early hospitalisation 344 period of COVID-19, this recognised concern in relation to the use of tocilizumab would 345 be lessened with earlier use. 22 346 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. study materials are available online at www.recoverytrial.net. As described in the protocol, 445 the trial steering committee will facilitate the use of the study data and approval will not 446 be unreasonably withheld. Deidentified participant data will be made available to bona 447 fide researchers registered with an appropriate institution within 3 months of publication. 448 However, the steering committee will need to be satisfied that any proposed publication 449 is of high quality, honours the commitments made to the study participants in the consent 450 documentation and ethical approvals, and is compliant with relevant legal and regulatory 451 requirements (e.g. relating to data protection and privacy). The steering committee will 452 have the right to review and comment on any draft manuscripts prior to publication. Data 453 will be made available in line with the policy and procedures described at: Above all, we would like to thank the thousands of patients who participated in this study. 461 We would also like to thank the many doctors, nurses, pharmacists, other allied health is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 11, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 11, 2021. ; https://doi.org/10.1101/2021.02.11.21249258 doi: medRxiv preprint CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 11, 2021. ; https://doi.org/10.1101/2021.02.11.21249258 doi: medRxiv preprint First randomisation$ Number of days since first randomisation 0 (0-1) 0 (0-1) . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 11, 2021. 1 (<1%) 6 (<1%) Data are mean (SD), n(%), or median (IQR). Information on sex, ethnicity, and SARS-CoV-2 test result were recorded on the main randomisation form when patients first entered the study. All other information was recorded on the second randomisation form (when patients were randomly assigned to tocilizumab vs. usual care alone). * includes 3 pregnant women. † Includes 9 patients not receiving any oxygen and 1859 patients receiving low-flow oxygen. ‡ includes patients receiving high-flow nasal oxygen, continuous positive airway pressure, or other noninvasive ventilation). § Includes patients receiving invasive mechanical ventilation or extra-corporeal membranous oxygenation. ¶ Defined as requiring ongoing specialist care. || Defined as estimated glomerular filtration rate <30 mL/min/1.73m² $ 2631 and 1615 participants were randomised into parts B and C of the first randomisation respectively. ^Information on use of corticosteroids was collected from 18 June 2020 onwards following announcement of the results of the dexamethasone comparison from the RECOVERY trial. Participants undergoing first randomisation prior to this date (and who were not allocated to dexamethasone) are assumed not to be receiving systemic corticosteroids. 582 583 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 11, 2021. ; https://doi.org/10.1101/2021.02.11.21249258 doi: medRxiv preprint Tocilizumab for COVID-19 30 584 .02 Data are n(%), n/N (%), or median (interquartile range). RR=rate ratio for the outcomes of 28-day mortality, hospital discharge and successful cessation of invasive mechanical ventilation, and risk ratio for other outcomes. * Analyses include only those on no ventilator support or non-invasive ventilation at second randomisation. † Analyses include only those on no ventilator support at second randomisation. ‡ Analyses restricted to those on invasive mechanical ventilation at second randomisation. § Analyses exclude those on haemodialysis or haemofiltration at second randomisation.    Subgroup−specific rate ratio estimates are represented by squares (with areas of the squares proportional to the amount of statistical information) and the lines through them correspond to the 95% CIs. The ethnicity and days since onset subgroups exclude those with missing data, but these patients are included in the overall summary diamond. * Includes 9 patients not receiving any oxygen and 1859 patients receiving simple oxygen only. † Includes patients receiving high−flow nasal oxygen, continuous positive airway pressure ventilation, other non−invasive ventilation. ‡ Includes patients receiving invasive mechanical ventilation and extra−corporeal membranous oxygenation. $ Information on use of corticosteroids was collected from 18 June 2020 onwards following announcement of the results of the dexamethasone comparison from the RECOVERY trial. Participants undergoing first randomisation prior to this date (and who were not allocated to dexamethasone) are assumed not to be receiving systemic corticosteroids.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted February 11, 2021. ; https://doi.org/10.1101/2021.02.11.21249258 doi: medRxiv preprint