Remdesivir and three other drugs for hospitalised patients with COVID-19: final results of the WHO Solidarity randomised trial and updated meta-analyses

Abstract

Background: The Solidarity trial among COVID-19 inpatients has previously reported interim mortality analyses for four repurposed antiviral drugs. Lopinavir, hydroxychloroquine, and interferon (IFN)-β1a were discontinued for futility but randomisation to remdesivir continued. Here, we report the final results of Solidarity and meta-analyses of mortality in all relevant trials to date.

Methods: Solidarity enrolled consenting adults (aged ≥18 years) recently hospitalised with, in the view of their doctor, definite COVID-19 and no contraindication to any of the study drugs, regardless of any other patient characteristics. Participants were randomly allocated, in equal proportions between the locally available options, to receive whichever of the four study drugs (lopinavir, hydroxychloroquine, IFN-β1a, or remdesivir) were locally available at that time or no study drug (controls). All patients also received the local standard of care. No placebos were given. The protocol-specified primary endpoint was in-hospital mortality, subdivided by disease severity. Secondary endpoints were progression to ventilation if not already ventilated, and time-to-discharge from hospital. Final log-rank and Kaplan-Meier analyses are presented for remdesivir, and are appended for all four study drugs. Meta-analyses give weighted averages of the mortality findings in this and all other randomised trials of these drugs among hospital inpatients. Solidarity is registered with ISRCTN, ISRCTN83971151, and ClinicalTrials.gov, NCT04315948.

Findings: Between March 22, 2020, and Jan 29, 2021, 14 304 potentially eligible patients were recruited from 454 hospitals in 35 countries in all six WHO regions. After the exclusion of 83 (0·6%) patients with a refuted COVID-19 diagnosis or encrypted consent not entered into the database, Solidarity enrolled 14 221 patients, including 8275 randomly allocated (1:1) either to remdesivir (ten daily infusions, unless discharged earlier) or to its control (allocated no study drug although remdesivir was locally available). Compliance was high in both groups. Overall, 602 (14·5%) of 4146 patients assigned to remdesivir died versus 643 (15·6%) of 4129 assigned to control (mortality rate ratio [RR] 0·91 [95% CI 0·82-1·02], p=0·12). Of those already ventilated, 151 (42·1%) of 359 assigned to remdesivir died versus 134 (38·6%) of 347 assigned to control (RR 1·13 [0·89-1·42], p=0·32). Of those not ventilated but on oxygen, 14·6% assigned to remdesivir died versus 16·3% assigned to control (RR 0·87 [0·76-0·99], p=0·03). Of 1730 not on oxygen initially, 2·9% assigned to remdesivir died versus 3·8% assigned to control (RR 0·76 [0·46-1·28], p=0·30). Combining all those not ventilated initially, 11·9% assigned to remdesivir died versus 13·5% assigned to control (RR 0·86 [0·76-0·98], p=0·02) and 14·1% versus 15·7% progressed to ventilation (RR 0·88 [0·77-1·00], p=0·04). The non-prespecified composite outcome of death or progression to ventilation occurred in 19·6% assigned to remdesivir versus 22·5% assigned to control (RR 0·84 [0·75-0·93], p=0·001). Allocation to daily remdesivir infusions (vs open-label control) delayed discharge by about 1 day during the 10-day treatment period. A meta-analysis of mortality in all randomised trials of remdesivir versus no remdesivir yielded similar findings.

Interpretation: Remdesivir has no significant effect on patients with COVID-19 who are already being ventilated. Among other hospitalised patients, it has a small effect against death or progression to ventilation (or both).

Funding: WHO.

Figure 1

Trial profile 14 304 hospital inpatients were randomly allocated (with equal probability) between the local standard of care (control group) and whichever of the four study drugs (active group) were locally available. 83 patients with a refuted COVID-19 diagnosis (all of whom survived) or with no encrypted image of their signed consent forwarded to the database were excluded, leaving 14 221 patients included. For each study drug, the control participants for that drug were those who could have been randomly allocated to receive it but were, by chance, randomly allocated to receive the same management without it. IFN=interferon. *Entry ended on Jan 29, 2021. †Entry ended on June 19, 2020. ‡Entry ended on July 4, 2020. §Entry ended on Oct 16, 2020.

Figure 2

Primary outcome of in-hospital mortality for remdesivir vs its control, by respiratory support at study entry Kaplan-Meier graphs to day 28, then total in-hospital mortality after day 28 (dashed lines); all known deaths were before day 150. Kaplan-Meier denominators include all patients except those who had already died in hospital and the few already lost to follow-up. The log-rank mortality rate ratio is standardised for age and respiratory support, and uses all in-hospital deaths, before or after day 28.

Figure 3

Secondary outcome of ventilation initiation for remdesivir vs its control in patients not already ventilated at study entry 2 X 2 analyses of ventilation, log-rank analyses of death, and combined analyses of death or ventilation. Analyses are stratified by age and by respiratory support, so each total is stratified for both factors. O − E=observed minus expected number of events. RR=rate ratio. *High-flow and low-flow oxygen were not recorded separately at entry into Solidarity.

Figure 4

Secondary outcome of time-to-discharge alive from hospital, subdivided by respiratory support at study entry All enrolled patients with outcomes reported are included. Vertical red line shows the end of scheduled treatment duration (10 days) if still in hospital. Horizontal lines at 50% crossed graphs at median time to discharge.

Figure 5

Comparison between the effects of random allocation to remdesivir on the daily discharge rate in Solidarity and in two placebo-controlled trials (A) Solidarity data are shown as treatment effects during different time periods. (B) ACTT-1 data are shown as treatment effects, split by initial respiratory support. (C) Wuhan data are shown as treatment effects among all patients. NA=not applicable. ND=not done. NR=median not reached. RR=rate ratio. *In ACTT-1, there was a chance imbalance favouring remdesivir in the initial proportions at higher risk, as defined in this figure (remdesivir: 226 [42%] of 533 vs placebo: 252 [49%] of 518, p=0·02). So, any findings combining low-risk and high-risk patients in that trial depend on whether this imbalance is allowed for. Cited ACTT-1 analyses are from the published text and its appendix.

Figure 6

Meta-analysis of the effects of remdesivir vs control on mortality in Solidarity and other trials, by respiratory support at study entry High-flow and low-flow oxygen were not recorded separately at entry into Solidarity. Ventilation includes non-invasive ventilation. Full details of these meta-analyses are given in the appendix (p 53). Solidarity data are from figure 2 and table 2, and other data are published (supplementary table 10). O − E=observed minus expected number of deaths. RR=rate ratio. *If V is the variance of the log-rank statistic O − E then RR is obtained by taking log_e RR to be (O − E) / V with normal variance 1 / V. Summation of (O − E) and of V yields the stratified total (providing the inverse-variance-weighted average of the separate log_e RR values).