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

Summary Background Aspirin has been proposed as a treatment for COVID-19 on the basis of its anti-thrombotic properties. We aimed to evaluate the efficacy and safety of aspirin in patients admitted to hospital with COVID-19. Methods In this randomised, controlled, open-label, platform trial, several possible treatments were compared with usual care in patients hospitalised with COVID-19. The trial took place at 177 hospitals in the UK, two hospitals in Indonesia, and two hospitals in Nepal. Eligible and consenting adults were randomly allocated in a 1:1 ratio to either usual standard of care plus 150 mg aspirin once per day until discharge or usual standard of care alone using web-based simple (unstratified) randomisation with allocation concealment. The primary outcome was 28 day mortality. All analyses were done by intention to treat. The trial is registered with ISRCTN (50189673) and ClinicalTrials.gov (NCT04381936). Findings Between Nov 1, 2020, and March 21, 2021, 14 892 (66%) of 22 560 patients enrolled into the RECOVERY trial were eligible to be randomly allocated to aspirin. 7351 patients were randomly allocated (1:1) to receive aspirin and 7541 patients to receive usual care alone. Overall, 1222 (17%) of 7351 patients allocated to aspirin and 1299 (17%) of 7541 patients allocated to usual care died within 28 days (rate ratio 0·96, 95% CI 0·89–1·04; p=0·35). Consistent results were seen in all prespecified subgroups of patients. Patients allocated to aspirin had a slightly shorter duration of hospitalisation (median 8 days, IQR 5 to >28, vs 9 days, IQR 5 to >28) and a higher proportion were discharged from hospital alive within 28 days (75% vs 74%; rate ratio 1·06, 95% CI 1·02–1·10; p=0·0062). Among patients not on invasive mechanical ventilation at baseline, there was no significant difference in the proportion meeting the composite endpoint of invasive mechanical ventilation or death (21% vs 22%; risk ratio 0·96, 95% CI 0·90–1·03; p=0·23). Aspirin use was associated with a reduction in thrombotic events (4·6% vs 5·3%; absolute reduction 0·6%, SE 0·4%) and an increase in major bleeding events (1·6% vs 1·0%; absolute increase 0·6%, SE 0·2%). Interpretation In patients hospitalised with COVID-19, aspirin was not associated with reductions in 28 day mortality or in the risk of progressing to invasive mechanical ventilation or death, but was associated with a small increase in the rate of being discharged alive within 28 days. Funding UK Research and Innovation (Medical Research Council), National Institute of Health Research, and the Wellcome Trust through the COVID-19 Therapeutics Accelerator.


INTRODUCTION 60
Thrombosis is a key feature of severe COVID-19, with 5-30% of hospitalised patients 61 (depending on illness severity) experiencing a major venous thromboembolic event 62 individuals involved in the trial were masked to aggregated outcome data during the 128 trial. 129

Procedures 130
A single online follow-up form was completed when participants were discharged, had 131 died or at 28 days after randomisation, whichever occurred earliest (appendix p 35-132 41). Information was recorded on adherence to allocated study treatment, receipt of 133 other COVID-19 treatments, duration of admission, receipt of respiratory or renal 134 support, and vital status (including cause of death). In addition, in the UK, routine 135 healthcare and registry data were obtained including information on vital status (with 136 date and cause of death), discharge from hospital, receipt of respiratory support, or 137 renal replacement therapy. 138

Outcomes 139
Outcomes were assessed at 28 days after randomisation, with further analyses 140 specified at 6 months. The primary outcome was all-cause mortality. Secondary 141 outcomes were time to discharge from hospital, and, among patients not on invasive 142 mechanical ventilation at randomisation, progression to invasive mechanical 143 ventilation (including extra-corporeal membrane oxygenation) or death. Prespecified 144 subsidiary clinical outcomes were use of non-invasive respiratory support, time to 145 successful cessation of invasive mechanical ventilation (defined as cessation of 146 invasive mechanical ventilation within, and survival to, 28 days), use of renal dialysis 147 or haemofiltration, cause-specific mortality, major bleeding events (defined as 148 intracranial bleeding or bleeding requiring transfusion, endoscopy, surgery or 149 vasoactive drugs), thrombotic events (defined as acute pulmonary embolism, deep 150 vein thrombosis, ischaemic stroke, myocardial infarction or systemic arterial 151 . 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 June 8, 2021. ; https://doi.org/10.1101/2021.06.08.21258132 doi: medRxiv preprint embolism) and major cardiac arrhythmias. Information on suspected serious adverse 152 reactions was collected in an expedited fashion to comply with regulatory 153 requirements. 154

Statistical Analysis 155
An intention-to-treat comparison was conducted between patients randomised to 156 aspirin and patients randomised to usual care but for whom aspirin was both available 157 and suitable as a treatment. For the primary outcome of 28-day mortality, the log-rank 158 observed minus expected statistic and its variance were used to both test the null 159 hypothesis of equal survival curves (i.e., the log-rank test) and to calculate the one-160 step estimate of the average mortality rate ratio. We constructed Kaplan-Meier survival 161 curves to display cumulative mortality over the 28-day period. We used the same 162 method to analyse time to hospital discharge and successful cessation of invasive 163 mechanical ventilation, with patients who died in hospital right-censored on day 29. 164 Median time to discharge was derived from Kaplan-Meier estimates. For the pre-165 specified composite secondary outcome of progression to invasive mechanical 166 ventilation or death within 28 days (among those not receiving invasive mechanical 167 ventilation at randomisation), and the subsidiary clinical outcomes of receipt of 168 ventilation and use of haemodialysis or haemofiltration, the precise dates were not 169 available and so the risk ratio was estimated instead. 170 Prespecified subgroup analyses (defined by characteristics at randomisation: age, 171 sex, ethnicity, level of respiratory support, days since symptom onset, and use of 172 corticosteroids) were performed for the primary outcome using the statistical test of 173 interaction (test for heterogeneity or trend), in accordance with the prespecified 174 analysis plan (appendix p 113). A sensitivity analysis restricting analysis of the primary 175 . 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 June 8, 2021. ; https://doi.org/10.1101/2021.06.08.21258132 doi: medRxiv preprint outcome to patients with a positive PCR test for SARS-COV-2 was conducted. In 176 addition, post-hoc exploratory analyses of the primary and secondary outcomes by 177 venous thromboprophylaxis treatment at randomisation was conducted. Observed 178 effects within subgroup categories were compared using a chi-squared test for 179 heterogeneity or trend, in accordance with the prespecified analysis plan. 180 Estimates of rate and risk ratios are shown with 95% confidence intervals. All p-values 181 are 2-sided and are shown without adjustment for multiple testing. The full database 182 is held by the study team which collected the data from study sites and performed the 183 analyses at the Nuffield Department of Population Health, University of Oxford 184 (Oxford, UK). 185 As stated in the protocol, appropriate sample sizes could not be estimated when the 186 trial was being planned at the start of the COVID-19 pandemic (appendix p 53). As the 187 trial progressed, the trial steering committee, whose members were unaware of the 188 results of the trial comparisons, determined that sufficient patients should be enrolled 189 to provide at least 90% power at a two-sided significance level of 1% to detect a 190 clinically relevant proportional reduction in 28-day mortality of 12.5% between the two 191 groups. Consequently, on 21 March, 2021, the steering committee, masked to the 192 results, closed recruitment to the aspirin comparison as sufficient patients had been 193

Role of the funding source 197
. 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 June 8, 2021. ; https://doi.org/10.1101/2021.06.08.21258132 doi: medRxiv preprint The funder of the study had no role in study design, data collection, data analysis, data 198 interpretation, or writing of the report. The corresponding authors had full access to all 199 the data in the study and had final responsibility for the decision to submit for enrolled into the RECOVERY trial were eligible to be randomly allocated to aspirin (i.e. 205 aspirin was available in the hospital at the time and the attending clinician was of the 206 opinion that the patient had no known indication for or contraindication to aspirin, figure  207 1). 7351 patients were randomly allocated to usual care plus aspirin and 7541 were 208 randomly allocated to usual care alone. The mean age of study participants in this 209 comparison was 59.2 years (SD 14.2) and the median time since symptom onset was 210 9 days (IQR 6 to 12 days) (webtable 1). At randomisation, 5035 patients (34%) were 211 receiving thromboprophylaxis with higher dose low molecular weight heparin (LMWH), 212 8878 (60%) with standard dose LMWH, and 979 (7%) were not receiving 213

thromboprophylaxis. 214
The follow-up form was completed for 7290 (99%) participants in the aspirin group and 215 7457 (99%) participants in the usual care group. Among participants with a completed 216 follow-up form, 6587 (90%) allocated to aspirin received at least one dose and 210 217 (3%) allocated to usual care received at least one dose of aspirin (figure 1; webtable 218 2). Of the 6587 participants allocated to aspirin that received at least one dose of 219 aspirin, 5040 (77%) received aspirin on most days following randomisation (≥90% of 220 the days from randomisation to time to discharge or 28 days after randomisation, 221 . 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 June 8, 2021. ; https://doi.org/10.1101/2021.06.08.21258132 doi: medRxiv preprint whichever was earlier). Use of other treatments for COVID-19 was similar among 222 participants allocated aspirin and among those allocated usual care, with nearly 90% 223 receiving a corticosteroid, about one-quarter receiving remdesivir, and one-eighth 224 receiving tocilizumab (webtable 2). 225 Primary and secondary outcome data are known for 99% of randomly assigned 226 patients. We observed no significant difference in the proportion of patients who met 227 the primary outcome of 28-day mortality between the two randomised groups (1222 228 [17%] patients in the aspirin group vs. 1299 (17%) patients in the usual care group; 229 rate ratio 0·96; 95% confidence interval [CI], 0·89 to 1·04; p=0·35; figure 2, table 2). 230 The rate ratio was similar across all pre-specified sub-groups (figure 3). In an 231 exploratory analysis restricted to the 14467 (97%) patients with a positive SARS-CoV-232 2 test result, the result was virtually identical (rate ratio 0.96, 95% CI 0·89 to 1·04; 233 p=0·31). 234 Allocation to aspirin was associated with a reduction of 1 day in median time until 235 discharge alive from hospital compared to usual care (median 8 days vs. 9 days [IQR 236 for each 5 to >28 days]) and an increased rate of discharge alive within 28 days (75% 237 vs. 74%, rate ratio 1·06, 95% CI 1·02 to 1·10, p=0·0062) (table 2). Among those not 238 on invasive mechanical ventilation at baseline, the number of patients progressing to 239 the pre-specified composite secondary outcome of invasive mechanical ventilation or 240 death among those allocated to aspirin was similar to that among those allocated to 241 usual care (21% vs. 22%, risk ratio 0·96, 95% CI 0·90 to 1·03, p=0·23). There was no 242 evidence that the effect of allocation to aspirin vs. usual care on time until discharge 243 alive from hospital or on invasive mechanical ventilation or death differed between the 244 pre-specified subgroups of patients (webfigure 1, webfigure 2). In a post-hoc 245 . 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 June 8, 2021. ; https://doi.org/10.1101/2021.06.08.21258132 doi: medRxiv preprint exploratory analysis there was no evidence that the effect of allocation to aspirin vs. 246 usual care on the primary and secondary outcomes differed by use of LMWH use at 247 randomisation (webfigure 3). 248 We found no significant differences in the prespecified subsidiary clinical outcomes of 249 cause-specific mortality (webtable 3), use of ventilation, successful cessation of 250 invasive mechanical ventilation, or receipt of renal dialysis or haemofiltration (table 2). 251 As expected with the use of aspirin, the incidence of thrombotic events was lower 252 (4.6% vs. 5.3%; absolute difference 0.6%, SE 0.4%) and the incidence of major 253 bleeding events was higher (1.6% vs. 1.0%; absolute difference 0.6%, SE 0.2%) in 254 the aspirin group (webtable 4). The incidence of new cardiac arrhythmias was similar 255 in the two groups (webtable 5). There were 18 reports of a serious adverse event 256 believed related to aspirin, all of which were due to haemorrhagic events (webtable 6). 257

DISCUSSION 259
In this large, randomised trial involving over 14,000 patients and over 2000 deaths, 260 allocation to aspirin was not associated with reductions in mortality or, among those 261 not on invasive mechanical ventilation at baseline, the risk of progressing to the 262 composite endpoint of invasive mechanical ventilation or death. Allocation to aspirin 263 was, however, associated with a small increase in the rate of being discharged from 264 hospital alive within 28 days. These results were consistent across the prespecified 265 subgroups of age, sex, ethnicity, duration of symptoms prior to randomisation, level of 266 respiratory support at randomisation, and use of corticosteroids. 267 . 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 June 8, 2021. ; https://doi.org/10.1101/2021.06.08.21258132 doi: medRxiv preprint As expected, allocation to aspirin was associated with an increased risk of major 268 bleeding and a decreased risk of thromboembolic complications, such that for every 269 1000 patients treated with aspirin, approximately 6 more would experience a major 270 bleeding event and approximately 6 fewer would experience a thromboembolic event. 271 The rate of reported thromboembolic events in our study population was low (5. The RECOVERY trial only studied hospitalised COVID-19 patients and, therefore, is 336 not able to provide evidence on the safety and efficacy of aspirin used in other patient 337 groups. Further studies to identify the safety and efficacy of aspirin in non-hospitalised 338 patients are needed and are ongoing. 339 . 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 June 8, 2021. ; https://doi.org/10.1101/2021.06.08.21258132 doi: medRxiv preprint In summary, the results of this large, randomised trial do not support the addition of 340 aspirin to standard thromboprophylaxis or therapeutic anticoagulation in patients 341 hospitalised with COVID-19. 342 343 344 . 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.

Contributors 345
This manuscript was initially drafted by PWH and MJL, further developed by the 346 Writing Committee, and approved by all members of the trial steering committee. PWH 347 and MJL vouch for the data and analyses, and for the fidelity of this report to the study 348 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.

Data sharing 381
The protocol, consent form, statistical analysis plan, definition & derivation of clinical 382 characteristics & outcomes, training materials, regulatory documents, and other 383 relevant study materials are available online at www.recoverytrial.net. As described in 384 the protocol, the trial Steering Committee will facilitate the use of the study data and 385 approval will not be unreasonably withheld. Deidentified participant data will be made 386 available to bona fide researchers registered with an appropriate institution within 3 387 months of publication. However, the Steering Committee will need to be satisfied that 388 any proposed publication is of high quality, honours the commitments made to the 389 study participants in the consent documentation and ethical approvals, and is 390 compliant with relevant legal and regulatory requirements (e.g. relating to data 391 protection and privacy). The Steering Committee will have the right to review and 392 comment on any draft manuscripts prior to publication. Data will be made available in 393 line with the policy and procedures described at: https://www.ndph.ox.ac.uk/data-394 access. Those wishing to request access should complete the form at Above all, we would like to thank the thousands of patients who participated in this 400 trial. We would also like to thank the many doctors, nurses, pharmacists, other allied 401 . 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. 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 June 8, 2021. ; https://doi.org/10.1101/2021.06.08.21258132 doi: medRxiv preprint The views expressed in this publication are those of the authors and not necessarily 426 those of the NHS or the NIHR. 427

Conflicts of interest 428
The authors have no conflict of interest or financial relationships relevant to the 429 submitted work to disclose. No form of payment was given to anyone to produce the 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 June 8, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 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 June 8, 2021. ; https://doi.org/10.1101/2021.06.08.21258132 doi: medRxiv preprint 552 . 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 June 8, 2021. ; https://doi.org/10.1101/2021.06.08.21258132 doi: medRxiv preprint 0.93 RR=Rate Ratio for the outcomes of 28-day mortality and hospital discharge, and risk ratio for the outcome of receipt of invasive mechanical ventilation or death (and its subcomponents). CI=confidence interval. *Analyses exclude those on invasive mechanical ventilation at randomization. 554 555 . 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 June 8, 2021. ; https://doi.org/10.1101/2021.06.08.21258132 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 June 8, 2021.  Figure 2: Effect of allocation to aspirin on 28−day mortality . 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 June 8, 2021.