key: cord-0717661-im99d091
authors: Jiang, Yawen; Chen, Daqin; Cai, Dan; Yi, Yao; Jiang, Shan
title: Effectiveness of remdesivir for the treatment of hospitalized Covid‐19 persons: a network meta‐analysis
date: 2020-08-19
journal: J Med Virol
DOI: 10.1002/jmv.26443
sha: 078c6f316caf047f1c4b9560c1dc4c143107512f
doc_id: 717661
cord_uid: im99d091

INTRODUCTION: Several randomized clinical trials (RCTs) that investigated the effectiveness of remdesivir for the treatment of Covid‐19 have generated inconsistent evidence. The present study aimed to synthesize available RCT evidence using network meta‐analyses (NMAs). METHODS: Both blinded and open‐label RCTs in PubMed database from inception to June 7, 2020 that contained “remdesivir”, “Covid‐19”, and “trial” in the abstracts conducted on hospitalized Covid‐19 persons were identified and screened. The studies must have at least one remdesivir arm and evaluated one of the pre‐specified outcomes. The outcomes were clinical improvement between days 10‐15 after randomization and clinical recovery during the follow‐up period. The identified literature was supplemented with relatively recent studies that were known to the researchers if not already included. Frequentist NMAs with random effects were conducted. RESULTS: Both 10‐day and 5‐day remdesivir regimens were associated with higher odds of clinical improvement [odds ratio (OR) of 10‐day regimen: 1.35, 95% confidence interval (CI): 1.09 – 1.67); OR of 5‐day regimen: 1.81, CI: 1.32 – 2.45] and higher probabilities of clinical recovery [relative risk (RR) of 10‐day regimen: 1.24, CI: 1.07 – 1.43); RR of 5‐day regimen: 1.47, CI: 1.16 – 1.87] compared with placebo. CONCLUSIONS: Remdesivir may have clinical benefits among hospitalized Covid‐19 persons. This article is protected by copyright. All rights reserved.

The ongoing Covid-19 pandemic represents a major public health threat and healthcare burden globally [1] . Without approved pharmacological therapies [2] , several randomized clinical trials (RCTs) investigated the potential effectiveness This article is protected by copyright. All rights reserved.

of remdesivir for the treatment of Covid-19, which was already granted emergency use authorization for the disease by the United States Food and Drug Administration [2] . However, not all of these RCTs generated consistent evidence [3] [4] [5] . As such, the present study aimed to synthesize available RCT evidence using network meta-analyses (NMAs).

Two researchers (Dan.C and S.J.) searched the PubMed database from inception to June 7, 2020 for literature that contained "remdesivir", "Covid-19", and "trial" in the abstracts. Both blinded and open-label RCTs among hospitalized Covid-19 persons were included. To be eligible for inclusion, studies must have included at least one remdesivir group. We also required that the studies reported data on any of the pre-determined outcomes. The primary outcome was clinical improvement between days 10-15 after randomization. The secondary outcome was clinical recovery. For the secondary outcome, the longest follow up was used when studies reported different study durations. The identified literature was supplemented with relatively recent studies that were known to the researchers if not already included. The risk of bias of studies was assessed by two researchers (Y.J. and Daqin.C) independently using the Cochrane Collaboration risk of bias tool over the five domains of selection, performance, detection, attrition, and reporting [6] . This article is protected by copyright. All rights reserved.

Data were independently extracted by two investigators (Y.J. and Daqin.C.) using prespecified forms. The types of information extracted from the studies included participant characteristics, sample sizes, inclusion criteria, interventions, follow-up periods, and outcomes data. The relative effect was evaluated using odds ratio (OR) for clinical improvement and relative risk (RR) for clinical recovery to accommodate differential reporting routines of the endpoints. When not presented in the original reports, ORs and RRs were calculated using data of sample sizes and event frequencies. When data were inadequate to calculate RRs, rate ratios were used instead.

Frequentist NMAs with random effects were conducted [7]. I 2 was used to quantify cross-study heterogeneity and NMAs using full design-by-treatment interaction specifications were used to test network inconsistency [8, 9] . Sensitivity analyses were conducted by restricting the primary outcome to a minimum 2-score improvement in ordinal scales and excluding studies that did not contain patients with severe conditions. All analyses were performed using Stata 15 (Stata Corp, College Station, Texas) and R 3.6.1. The study protocol was registered and available at International Prospective Register of Systematic Reviews (CRD42020190560).

This article is protected by copyright. All rights reserved.

The initial search identified 13 publications. After screening for randomized clinical trials, two studies remained eligible [3, 4] . The list was manually supplemented with two relatively recent studies that were not indexed in PubMed by the time of searching [5, 10, 11] . The four studies covered three comparators, namely remdesivir 10-day treatment, remdesivir 5-day treatment, and placebo.

The baseline characteristics of the included studies are listed in Table 1 .

The four studies provided data from 2,049 individuals on the primary outcome. A network graph is presented in Fig S1. The follow-up period ranged from 11 days to 15 days. Two of the studies exclusively recruited severe patients (defined in the footnote of Table 1 ), one study dominantly included severe patients Data on the same three treatments from three studies were included to analyze clinical recovery. For the ACCT-1 trial, the rate ratio was used to approximate the RR because the latter was not available. The results (Fig. S3) showed that both 10-day (RR: 1.24, CI: 1.07 -1.43) and 5-day (RR: 1.47, CI: 1.16 -1.87) remdesivir regimens were associated with greater probabilities of clinical recovery.

Results related to I 2 , inconsistency tests, and risk of bias are provided in Table   S1 . We did not spot substantial heterogeneity or significant inconsistency in any analyses. The I 2 was were 0.0% and 19.2% in the analyses of the two outcomes, respectively. The test of inconsistency in the analysis of the primary outcome was statistically insignificant (p=0.528), whereas inconsistency was not applicable to This article is protected by copyright. All rights reserved.

the analysis of the secondary outcome because direct and indirect comparison data were not available for any pairs of contrast.

In the present analysis, the effectiveness of remdesivir regimens in relation to placebo were evaluated using NMAs of RCTs. Our findings showed that both The individual studies provided mixed results of remdesivir effectiveness. In particular, at least one trial showed an absence of statistically significant clinical benefit of remdesivir [3] . In the meantime, other trials suggested no difference between 10-day and 5-day remdesivir treatments [5, 10] . Such differential results might have been driven by heterogeneity in study design and sample sizes. The present study provided relatively comprehensive evidence to document the This article is protected by copyright. All rights reserved.

benefits of remdesivir among hospitalized Covid-19 persons as well as differential effects between 10-day and 5-day regimens. These results indicate that it may be advisable to prescribe the 5-day remdesivir regimen for the treatment of Covid-19.

The result of the sensitivity analysis when the study that only included moderate patients was excluded suggested that the point estimates of effects on clinical improvement associated with remdesivir regimens were higher than those in the base case. A potential cause was that the effect of remdesivir may be stronger among severe patients than in moderate patients.

Several limitations must be noted when interpreting the results. First, there was discrepancy in the definition of endpoints and reporting of effects.

Specifically, the studies engaged scales using different numbers of points to gauge the scales and considered either 1-point or 2-point alleviation as clinical improvement. Such differential scales would impact the estimated effect sizes.

Second, the baseline severity of the participants varied across trials, which might compound the heterogeneity across trials. Although sensitivity analyses were conducted to mitigate the impacts of the first and the second limitations on the results, remaining consequences due to such complexity may exist. Finally, it has been suspected that remdesivir might demonstrate better efficacy if administered within 10 days of symptom onset compared with beyond that [3] . However, data from the studies were not sufficient to conduct subgroup NMAs by timeliness of 

Potential association between COVID-19 mortality and health-care resource availability. The Lancet Global Health

Emergency Use Authorization of Remdesivir: The Need for a Transparent Distribution Process

Remdesivir in adults with severe COVID-19: a randomised, double-blind, placebo-controlled, multicentre trial

Remdesivir for the Treatment of Covid-19 -Preliminary Report

Remdesivir for 5 or 10 Days in Patients with Severe Covid-19

The Cochrane Collaboration's tool for assessing risk of bias in randomised trials