CAUTION: The evidence available regarding COVID-19 treatments (even limited to the randomized trial evidence) is very limited and potentially unstable. Much of the available evidence is not published in peer-reviewed final form, and most treatments are currently considered experimental.
Outcome | Key Findings | Sample Size | Certainty/Quality of Evidence | What this means |
---|---|---|---|---|
Mortality within 14 days | Relative risk 0.65 (95% CI 0.45 to 0.94) Absolute risk difference -4% (95% CI -1% to -8%) | 2 trials (1,290 patients; unclear whether some patients from ACTT-1 had not reached 14-day point yet) | Low – analysis limited by preliminary nature of ACTT-1 report and potential inconsistency1,2 | Remdesivir may reduce 14-day mortality, but more data needed from the ACTT-1. |
Mortality within 28 days | 13.9% vs. 12.8% in Chinese trial | 1 trial (236 patients) | Very low – analysis limited by early termination and very serious imprecision3,4,5 | Remdesivir might or might not reduce 28-day mortality. |
Mechanical ventilation, ECMO or death at 2 weeks | Relative risk 0.69 (95% CI 0.56 to 0.86) Absolute risk difference -9% (95% CI -4% to -14%) | 2 trials (1,075 patients; 215 additional patients not included in preliminary analysis of ACTT-1) | Moderate – analysis limited by preliminary nature of ACTT-1 report1 | Remdesivir may reduce the chance of being dead, on a ventilator, or on ECMO at 2 weeks, but more data needed from the ACTT-1. |
Time to recovery | Median 11 days vs. 15 days (hazard ratio 1.32, 95% CI 1.12 to 1.55, p < 0.001) | 1 trial had 1,063 patients (but preliminary analysis with 301 patients [28.3%] not yet analyzed) | Moderate – analysis limited by preliminary nature of ACTT-1 report1 | Remdesivir may reduce the time to recovery, but more data needed from ACTT-1. |
Time to clinical improvement | Median 21 days vs. 23 days (hazard ratio 1.23, 95% CI 0.87 to 1.75) | 1 trial (236 patients), 148 had recovered by 28 days | Low – analysis limited by early termination and imprecision3,4,5 | Remdesivir might reduce the time to clinical improvement |
ECMO = extracorporeal membrane oxygenation
Outcome | 1,063 patients hospitalized with COVID-19 (pneumonia or need for oxygen support), 73 sites in North America/Europe/Asia (IV remdesivir 200 mg then 100 mg once daily for 9 days or until hospital discharge) ACTT-1 | 236 patients hospitalized with COVID-19 in China (IV remdesivir 200 mg then 100 mg once daily for 9 days) Chinese trial |
---|---|---|
Mortality within 14 days | 32/538 (5.9%) vs. 54/521 (10.4%); Risk difference -4.4% (95% CI -1.1% to -7.8%) Kaplan-Meier estimate 7.1% (95% CI 5.0% to 9.9%) vs. 11.9% (9.2% to 15.4%); hazard ratio 0.70 (95% CI 0.47 to 1.04)1,5 |
15/153 (9.8%) vs. 7/78 (9%); Risk difference 0.8% (95% CI -8.4% to 8.2%)3,4,5 |
Mortality within 28 days | Data not yet available | 22/158 (13.9%) vs. 10/78 (12.8%); Risk difference 1.1% (95% CI -8.1% to 10.3%)3,4,5 |
Time to recovery | Median 11 days (95% CI 9 to 12 days) vs. 15 days (95% CI 13 to 19 days) Median difference -4 days (hazard ratio 1.32, 95% CI 1.12 to 1.55, p < 0.001)1 |
|
Mechanical ventilation, ECMO or death at day 14 (Chinese trial) or 15 (ACTT-1) | 93/434 (21.4%) vs. 127/410 (31.0%); Risk difference -9.5% (95% CI -3.6% to -15.4%)1 | 19/153 (12.4%) vs. 14/78 (17.9%); Risk difference -5.5% (95% CI -16.4% to 3.8%)3,4,5 |
Time to clinical improvement | Median 21 days (IQR 13 to 28 days) vs. 23 days (IQR 15 to 28 days) Median difference -2 days (hazard ratio 1.23, 95% CI 0.87 to 1.75)3,4,5 |
ECMO = extracorporeal membrane oxygenation
Meta-analytic results for the relative risk and risk difference are both from a fixed-effects model for the outcomes of 14-day mortality and death or need for invasive mechanical ventilation or ECMO at 14 or 15 days. Additionally, we note the risk difference is meta-analyzed directly (as opposed to using the relative risk in conjunction with a presumed baseline risk to estimate absolute treatment effects). Results for a random-effects model with a Paule-Mandel estimator yielded comparable findings, though we note that for the outcome of 14-day mortality, the findings no longer reached conventional statistical significance:
Although some might lean toward a random-effects model given the values of I2 and τ2, we note that choosing between a fixed-effects and random-effects model based solely on these values is an ill-advised practice; additionally, it can be difficult to obtain a robust estimate of variance with only two trials. Considering more broadly the totality of the available evidence, the particularities of the trials included in these meta-analyses, and the consistency of the results between fixed- and random-effects models (other than wider confidence intervals for the random-effects model for 14-day mortality), we consider it reasonable to present the fixed-effects model as the primary results.