Remdesivir for COVID-19: real-time meta analysis of 54 studies

3 In Vitro studies support the efficacy of remdesivir [De Forni, Delandre, Jeffreys].

An In Vivo animal study supports the efficacy of remdesivir [Vermillion].

[Vermillion] investigate a novel formulation of remdesivir that may be more effective for COVID-19.

Preclinical research is an important part of the development of treatments, however results may be very different in clinical trials. Preclinical results are not used in this paper.

Table 1 summarizes the results for all stages combined, with different exclusions, and for specific outcomes. Table 2 shows results by treatment stage. Figure 3, 4, 5, 6, 7, 8, 9, 10, and 11 show forest plots for random effects meta-analysis of all studies with pooled effects, mortality results, ventilation, ICU admission, hospitalization, progression, recovery, viral clearance, and peer reviewed studies.

To avoid bias in the selection of studies, we analyze all non-retracted studies. Here we show the results after excluding studies with major issues likely to alter results, non-standard studies, and studies where very minimal detail is currently available. Our bias evaluation is based on analysis of each study and identifying when there is a significant chance that limitations will substantially change the outcome of the study. We believe this can be more valuable than checklist-based approaches such as Cochrane GRADE, which may underemphasize serious issues not captured in the checklists, overemphasize issues unlikely to alter outcomes in specific cases (for example, lack of blinding for an objective mortality outcome, or certain specifics of randomization with a very large effect size), or be easily influenced by potential bias. However, they can also be very high quality.

The studies excluded are as below. Figure 15 shows a forest plot for random effects meta-analysis of all studies after exclusions.

[Arfijanto], unadjusted results with no group details.

[Elec], substantial confounding by time possible due to significant changes in SOC and treatment propensity during the study period.

[Elhadi], unadjusted results with no group details.

[Fried], excessive unadjusted differences between groups; substantial unadjusted confounding by indication likely.

[Kurniyanto], unadjusted results with no group details; substantial unadjusted confounding by indication likely.

[Madan], unadjusted results with no group details.

[Madan (B)], excessive unadjusted differences between groups.

[Malundo], unadjusted results with no group details.

[Mulhem], substantial unadjusted confounding by indication likely; substantial confounding by time possible due to significant changes in SOC and treatment propensity during the study period.

[Mustafa], unadjusted results with no group details.

[Oku], unadjusted results with no group details.

[Salehi], unadjusted results with no group details.

[Schmidt], confounding by indication is likely and adjustments do not consider COVID-19 severity at baseline.

[Solh], very late stage, >50% on oxygen/ventilation at baseline; substantial unadjusted confounding by indication likely.

Heterogeneity in COVID-19 studies arises from many factors including:

The time between infection or the onset of symptoms and treatment may critically affect how well a treatment works. For example an antiviral may be very effective when used early but may not be effective in late stage disease, and may even be harmful. Oseltamivir, for example, is generally only considered effective for influenza when used within 0-36 or 0-48 hours [McLean, Treanor]. Baloxavir studies for influenza also show that treatment delay is critical — [Ikematsu] report an 86% reduction in cases for post-exposure prophylaxis, [Hayden] show a 33 hour reduction in the time to alleviation of symptoms for treatment within 24 hours and a reduction of 13 hours for treatment within 24-48 hours, and [Kumar] report only 2.5 hours improvement for inpatient treatment.

Figure 16 shows a mixed-effects meta-regression for efficacy as a function of treatment delay in COVID-19 studies from 51 treatments, showing that efficacy declines rapidly with treatment delay. Early treatment is critical for COVID-19.

Details of the patient population including age and comorbidities may critically affect how well a treatment works. For example, many COVID-19 studies with relatively young low-comorbidity patients show all patients recovering quickly with or without treatment. In such cases, there is little room for an effective treatment to improve results (as in [López-Medina]).

Efficacy may differ significantly depending on the effect measured, for example a treatment may be very effective at reducing mortality, but less effective at minimizing cases or hospitalization. Or a treatment may have no effect on viral clearance while still being effective at reducing mortality.

There are many different variants of SARS-CoV-2 and efficacy may depend critically on the distribution of variants encountered by the patients in a study. For example, the Gamma variant shows significantly different characteristics [Faria, Karita, Nonaka, Zavascki]. Different mechanisms of action may be more or less effective depending on variants, for example the viral entry process for the omicron variant has moved towards TMPRSS2-independent fusion, suggesting that TMPRSS2 inhibitors may be less effective [Peacock, Willett].

Effectiveness may depend strongly on the dosage and treatment regimen.

The use of other treatments may significantly affect outcomes, including anything from supplements, other medications, or other kinds of treatment such as prone positioning.

The quality of medications may vary significantly between manufacturers and production batches, which may significantly affect efficacy and safety. [Williams] analyze ivermectin from 11 different sources, showing highly variable antiparasitic efficacy across different manufacturers. [Xu] analyze a treatment from two different manufacturers, showing 9 different impurities, with significantly different concentrations for each manufacturer.

We present both pooled analyses and specific outcome analyses. Notably, pooled analysis often results in earlier detection of efficacy as shown in Figure 17. For many COVID-19 treatments, a reduction in mortality logically follows from a reduction in hospitalization, which follows from a reduction in symptomatic cases, etc. An antiviral tested with a low-risk population may report zero mortality in both arms, however a reduction in severity and improved viral clearance may translate into lower mortality among a high-risk population, and including these results in pooled analysis allows faster detection of efficacy. Trials with high-risk patients may also be restricted due to ethical concerns for treatments that are known or expected to be effective.

Pooled analysis enables using more of the available information. While there is much more information available, for example dose-response relationships, the advantage of the method used here is simplicity and transparency. Note that pooled analysis could hide efficacy, for example a treatment that is beneficial for late stage patients but has no effect on viral replication or early stage disease could show no efficacy in pooled analysis if most studies only examine viral clearance. While we present pooled results, we also present individual outcome analyses, which may be more informative for specific use cases.

Currently, 37 of the treatments we analyze show statistically significant efficacy or harm, defined as ≥10% decreased risk or >0% increased risk from ≥3 studies. 94% of treatments showing statistically significant efficacy/harm with pooled effects have been confirmed with one or more specific outcomes, with a mean delay of 3.1 months. When restricting to RCTs only, 55% of treatments showing statistically significant efficacy/harm with pooled effects have been confirmed with one or more specific outcomes, with a mean delay of 2.9 months.

The distribution of studies will alter the outcome of a meta analysis. Consider a simplified example where everything is equal except for the treatment delay, and effectiveness decreases to zero or below with increasing delay. If there are many studies using very late treatment, the outcome may be negative, even though early treatment is very effective. This may have a greater effect than pooling different outcomes such as mortality and hospitalization. For example a treatment may have 50% efficacy for mortality but only 40% for hospitalization when used within 48 hours. However efficacy could be 0% when used late.

All meta analyses combine heterogeneous studies, varying in population, variants, and potentially all factors above, and therefore may obscure efficacy by including studies where treatment is less effective. Generally, we expect the estimated effect size from meta analysis to be less than that for the optimal case. Looking at all studies is valuable for providing an overview of all research, important to avoid cherry-picking, and informative when a positive result is found despite combining less-optimal situations. However, the resulting estimate does not apply to specific cases such as early treatment in high-risk populations. While we present results for all studies, we also present treatment time and individual outcome analyses, which may be more informative for specific use cases.

Figure 18 shows a mixed-effects meta-regression of efficacy as a function of followup duration, which shows decreasing efficacy with longer followup. This may reflect antiviral efficacy being offset by side effects of treatment.

Publishing is often biased towards positive results. Trials with patented drugs may have a financial conflict of interest that results in positive studies being more likely to be published, or bias towards more positive results. For example with molnupiravir, trials with negative results remain unpublished to date (CTRI/2021/05/033864 and CTRI/2021/08/0354242).

One method to evaluate bias is to compare prospective vs. retrospective studies. Prospective studies are more likely to be published regardless of the result, while retrospective studies are more likely to exhibit bias. For example, researchers may perform preliminary analysis with minimal effort and the results may influence their decision to continue. Retrospective studies also provide more opportunities for the specifics of data extraction and adjustments to influence results.

Figure 19 shows a scatter plot of results for prospective and retrospective studies. 38% of retrospective studies report a statistically significant positive effect for one or more outcomes, compared to 31% of prospective studies, consistent with a bias toward publishing positive results. The median effect size for retrospective studies is 12% improvement, compared to 5% for prospective studies, suggesting a potential bias towards publishing results showing higher efficacy.

Funnel plots have traditionally been used for analyzing publication bias. This is invalid for COVID-19 acute treatment trials — the underlying assumptions are invalid, which we can demonstrate with a simple example. Consider a set of hypothetical perfect trials with no bias. Figure 20 plot A shows a funnel plot for a simulation of 80 perfect trials, with random group sizes, and each patient's outcome randomly sampled (10% control event probability, and a 30% effect size for treatment). Analysis shows no asymmetry (p > 0.05). In plot B, we add a single typical variation in COVID-19 treatment trials — treatment delay. Consider that efficacy varies from 90% for treatment within 24 hours, reducing to 10% when treatment is delayed 3 days. In plot B, each trial's treatment delay is randomly selected. Analysis now shows highly significant asymmetry, p < 0.0001, with six variants of Egger's test all showing p < 0.05 [Egger, Harbord, Macaskill, Moreno, Peters, Rothstein, Rücker, Stanley]. Note that these tests fail even though treatment delay is uniformly distributed. In reality treatment delay is more complex — each trial has a different distribution of delays across patients, and the distribution across trials may be biased (e.g., late treatment trials may be more common). Similarly, many other variations in trials may produce asymmetry, including dose, administration, duration of treatment, differences in SOC, comorbidities, age, variants, and bias in design, implementation, analysis, and reporting.

Summary statistics from meta analysis necessarily lose information. As with all meta analyses, studies are heterogeneous, with differences in treatment delay, treatment regimen, patient demographics, variants, conflicts of interest, standard of care, and other factors. We provide analyses by specific outcomes and by treatment delay, and we aim to identify key characteristics in the forest plots and summaries. Results should be viewed in the context of study characteristics.

Some analyses classify treatment based on early or late administration, as done here, while others distinguish between mild, moderate, and severe cases. Viral load does not indicate degree of symptoms — for example patients may have a high viral load while being asymptomatic. With regard to treatments that have antiviral properties, timing of treatment is critical — late administration may be less helpful regardless of severity.

Details of treatment delay per patient is often not available. For example, a study may treat 90% of patients relatively early, but the events driving the outcome may come from 10% of patients treated very late. Our 5 day cutoff for early treatment may be too conservative, 5 days may be too late in many cases.

Comparison across treatments is confounded by differences in the studies performed, for example dose, variants, and conflicts of interest. Trials affiliated with special interests may use designs better suited to the preferred outcome.

In some cases, the most serious outcome has very few events, resulting in lower confidence results being used in pooled analysis, however the method is simpler and more transparent. This is less critical as the number of studies increases. Restriction to outcomes with sufficient power may be beneficial in pooled analysis and improve accuracy when there are few studies, however we maintain our pre-specified method to avoid any retrospective changes.

Studies show that combinations of treatments can be highly synergistic and may result in many times greater efficacy than individual treatments alone [Alsaidi, Andreani, Biancatelli, De Forni (B), Gasmi, Jeffreys (B), Jitobaom, Jitobaom (B), Ostrov, Thairu]. Therefore standard of care may be critical and benefits may diminish or disappear if standard of care does not include certain treatments.

This real-time analysis is constantly updated based on submissions. Accuracy benefits from widespread review and submission of updates and corrections from reviewers. Less popular treatments may receive fewer reviews.

No treatment, vaccine, or intervention is 100% available and effective for all current and future variants. Efficacy may vary significantly with different variants and within different populations. All treatments have potential side effects. Propensity to experience side effects may be predicted in advance by qualified physicians. We do not provide medical advice. Before taking any medication, consult a qualified physician who can compare all options, provide personalized advice, and provide details of risks and benefits based on individual medical history and situations.

Meta analysis shows 12% [4‑20%] lower mortality, and pooled analysis using the most serious outcome reported shows 11% [4‑17%] improvement. Results are similar for Randomized Controlled Trials, similar after exclusions, and worse for peer-reviewed studies. While studies to date show a small mortality improvement, meta regression with followup duration shows decreasing efficacy. This may reflect antiviral efficacy being offset by side effects of treatment.

Studies show significantly increased risk of acute kidney injury [Gérard, Wu, Zhou].

[Ader] RCT 857 hospitalized patients, showing no significant differences with remdesivir treatment. EudraCT2020-000936-23.

Submit Corrections or Updates.

[Aghajani] Retrospective 991 hospitalized patients in Iran focusing on aspirin use but also showing results for HCQ, remdesivir, and favipiravir.

Submit Corrections or Updates.

[Ali] RCT 1,282 hospitalized patients in Canada showing lower mechanical ventilation with remdesivir treatment, but no significant difference for mortality.

Submit Corrections or Updates.

[Alshamrani] PSM retrospective 29 hospitals in Saudi Arabia, showing lower mortality with remdesivir treatment.

Submit Corrections or Updates.

[Arch] Prospective PSM analysis of remdesivir use in the UK showing statistically significantly lower mortality at 28 days. For unspecified reasons, the study prioritized short-term outcomes. Mortality at 14 days was also lower but not statistically significant. Confounding by indication is likely and may only be partially addressed by the variables included in the PSM.

Submit Corrections or Updates.

[Arfijanto] Retrospective 162 hospitalized COVID-19 patients in Indonesia, showing no significant difference in delayed viral clearance with remdesivir treatment in unadjusted results.

Submit Corrections or Updates.

[Aweimer] Retrospective 149 patients under invasive mechanical ventilation in Germany showing no significant difference in mortality with remdesivir in unadjusted results.

Submit Corrections or Updates.