By Leo Goldstein – Re-Blogged From WUWT
- The use of Remdesivir for COVID-19 was authorized by the FDA based on a single RCT, conducted by NIAID with the participation of Gilead Sciences, the exclusive manufacturer of Remdesivir. A final report from this study was published on October 8, five months after the drug’s authorization.
- The final report shows that at least 35% of the patients were treated with Hydroxychloroquine, probably with Azithromycin. The data in the final report suggests that Hydroxychloroquine, not Remdesivir, was the main factor benefitting the patients in this study.
- Nothing in the study supports the hypothesis that Remdesivir is an effective antiviral for SARS-COV-19
- The study’s own numbers show an association between RDV and increased mortality in the most severe patients. It is also possible to conclude that RDV is net harmful for most hospitalized patients.
- The trial was conducted and reported with multiple defects, including:
- The study was not double blind, but was reported as such
- The pre-registered protocol was changed multiple times over the course of the trial
- The primary outcome was changed in the middle of the trial, apparently because the researchers noticed a lack of effectiveness of their drug as tried
- The outcome measures were subjective and not reliable
- The study was marred with conflicts of interest, aggravated by the design giving NIAID and Gilead leverage over the hospitals and physicians treating patients
- At least three of the study researchers-authors failed to report grants and/or personal fees received from Gilead recently.
This study, also known as Adaptive COVID-19 Treatment Trial (ACTT-1), was registered as NCT04280705, and conducted by the National Institute of Allergy and Infectious Diseases (NIAID), headed by Dr. Anthony Fauci. The study started as “A Multicenter, Adaptive, Randomized Blinded Controlled Trial of the Safety and Efficacy of Investigational Therapeutics for the Treatment of COVID-19 in Hospitalized Adults,” but quickly became a Phase 3 trial for Remdesivir. It was reported under the title “Remdesivir for the Treatment of Covid-19 — Preliminary Report,” (Beigel – Lane PR, 2020). The most important author is H. Clifford Lane, a Co-Chair of the NIH COVID-19 Treatment Guidelines Panel, and the deputy director of the NIAID.
This study was the only trial that supported the emergency authorization of Remdesivir (RDV) for COVID-19 treatment. Its updated version “Remdesivir for the Treatment of Covid-19 — Final Report” (Beigel – Lane FR, 2020) (also, “the paper”) was published on October 8, and is reviewed here.
The assumption that Remdesivir is effective against SARS-COV-2 is probably based on the similarity of its in-vitro effect with that of chloroquine, without taking into account that chloroquine and hydroxychloroquine accumulate in lungs, while RDV does not (Goldstein, 2020).
Unless stated otherwise, table and figure numbers refer to (Beigel – Lane FR, 2020) Supplementary Appendix, downloaded from https://www.nejm.org/doi/suppl/10.1056/NEJMoa2007764/suppl_file/nejmoa2007764_appendix.pdf on October 11, 2020.
Fatal Methodological Defects
Each of the several study defects described below invalidate this study’s results.
The paper’s Abstract incorrectly claims that the study was double blind. In actuality, the study was not double blind. Doctors in the European site and “some” of the other sites were unblinded and knew what they were administering.
Too Many Sites
Contrary to common sense and best methodological practices (Kraemer, 2000), the trial was conducted in 60 sites, plus 13 sub-sites. Such a multi-site design is a potentially misleading “centralized multicenter collaborative RCT,” per (Kraemer & Robinson, 2005).
Notice that less than 600 courses of Remdesivir were distributed among 73 sites, with an average 9 RDV course per site – a small sample, sufficient to tease the appetite, but not sufficient for hospitals to draw their own conclusions about its efficiency.
Taking into account the intentional unblinding of the treating physicians and the limited drug supply, the sites could have been incentivized to compete against each other for the best results from RDV, and be rewarded with a prioritized supply of the drug in the future. RDV was believed to be a miracle cure, and the access to it was priceless.
Arbitrary Removal of Sites from Final Statistics
Two sites were removed from the Registry in the May 6 update:
Rocky Mountain Regional Veteran Affairs Medical Center – Department of Infectious Diseases, Aurora, Colorado, United States, 80045
University of Florida Health – Shands Hospital – Division of Infectious Diseases and Global Medicine, Gainesville, Florida, United States, 32610
These two sites started enrolling patients between April 2 and April 15, as indicated by the status Recruiting on April 16. Enrollment in all sites ended on April 19. Neither site was mentioned in the Beigel – Lane (both versions, including appendices), and their removal was not explained. This raises suspicions that these sites were removed because of undesirable results.
Administration of HCQ before and/or after RDV treatment was permitted. Co-administration of RDV with HCQ was also permitted in some sites (which had a written policy of HCQ use for COVID-19) but forbidden in other sites.
~35% of the patients in both Remdesivir and placebo groups also received Hydroxychloroquine (Table S3). ~80% of the patients received antibiotics, but the paper does not specify which patients. 93% of the patients were recruited on March 22 or later, after President Trump tweeted about HCQ & Azithromycin. In late March – early April, HCQ + AZ was the standard of care in some countries (Sermo, April 15). New York state required patients to be hospitalized in order to receive HCQ based treatment. Even before the President’s tweet, it was commonly known that Azithromycin has some effect against the coronavirus, and thus, its purchases had sharply increased at the expense of other antibiotics (Vaduganathan et al., 2020).
The paper gives no information regarding the use of Zinc and vitamin C, both of which were frequently used in COVID-19 treatments (Sermo 2020, April 9).
Substantial Changes of the Registered Protocol
The protocol was changed many times during the study. The primary outcome was changed on April 8 from “Percentage of subjects reporting each severity rating on an 8-point ordinal scale” to “Time to recovery”.
The relevant part of the ordinal scale is:
8 – death (appropriately removed from the scale and counted separately)
7 – hospitalized, receiving invasive mechanical ventilation or extracorporeal membrane oxygenation (ECMO)
6 – hospitalized, requiring noninvasive ventilation or use of high-flow oxygen devices
5 – hospitalized, requiring any supplemental oxygen
4 – hospitalized, not requiring supplemental oxygen but requiring ongoing medical care
“Recovery” is defined as getting a score below 4, usually discharged from the hospital.
Defective Outcome Measure
There is a problem with the outcome measurements on this scale. The scores reflect subjective decisions by the doctor, such as “receiving invasive mechanical ventilation,” “requiring noninvasive ventilation” (here, requiring is the equivalent of receiving) and similar. These scores are not objectively measurable, like blood oxygen level or heart rate. And the doctor is likely unblinded and incentivized to evaluate RDV as superior to a placebo. Most secondary outcomes were also defined by the scores on this scale.
Lack of Baseline Information
The paper did not report statistics of baseline conditions of patients. Instead, it reported subjective scores on the same ordinal scale. It also reported selected comorbidities.
Unexplained Data Mismatch between the Preliminary and Final Reports
Referring to the data supplement, the numbers in the Final Report are substantially different from the numbers in the Preliminary Report.
No Benefits of RDV were Shown
The average mortality in the RDV arm appeared much lower than in the placebo group only on the first measurement date – 15 days after the start of treatment. The mortality rates reversed after that. By the end of the study (day 29), the total reported mortality was 11% in the RDV group vs 15% in the placebo arm (per table S12). This difference can be attributed to the differences in the initial conditions and data manipulation by the authors. Tables S13 and S15 support this conclusion.
Table S15 displays the differences in scores between the Remdesivir and placebo arms on pre-selected days. Here, a small difference in favor of RDV develops immediately after randomization and stays almost the same for some time. This suggests not patients’ improvement due to RDV, but initial selection or scoring of patients, manipulated favorably to RDV. After changing only 0.1 points over eight days of the treatment (RDV treatment is given up to 10 days), the difference jumps 0.3 points (from day 11 to day 15) within four days without treatment – just in time to measure the outcome on Day 15!
This table does not consider deaths, discharges, and withdrawals.
Table 15 shows results according to the objective National Early Warning Score. According to it, at the beginning of the treatment, symptoms of the RDV arm are better than the placebo arm: 5.7 : 6.1 (the higher score, the worse the symptoms). Similarly to the previous table, the difference increases in the first few days, mostly due to worsening in the placebo arm, becoming 5.6 : 6.6. Then the difference starts to decrease, becoming 6.3 : 6.6 (worse for RDV than at Baseline) on day 11, then increases by day 15. This table does not consider deaths, discharges, and withdrawals.
It is possible that RDV does have measurable antiviral effect against SARS-COV-2 in humans when administered early, but that was not demonstrated in this trial.
RDV Increased Deaths among the Sickest Patients
Even the manipulated data cannot conceal the harm that RDV has inflicted on the sickest patients. Unfortunately, Remdesivir was authorized and recommended for treatment of the sickest patients before the completion of the trial.
The following is the reported mortality by the Day 29 (from Table S5):
Score 7 21% 19%
Score 6 20% 20%
The actual odds are even worse for RDV. The following picture is Graph E from Figure S5. It is a Kaplan–Meier Estimate of Survival in the Score 7 group (the most severe group), by day.
The trick is easy to see. RDV patients appear to have slightly better chances of survival by Day 15, the first reporting measurement day. After day 15, their chances drop sharply and become much worse than the placebo arm. The difference narrows again on Day 26, just before the second measurement day.
Of even greater concern is that this graph is very different from the graph published in the preliminary report, (on May 22 and corrected on May 26), which depicted the RDV arm performing even worse.
This graph shows not only the sharp reversal of mortality odds after Day 15, favoring the placebo arm, but also a sudden narrowing of the gap on Day 26, three days before the final reporting Day 29. No explanation to this post-study data changes was given.
Hydroxychloroquine + Azithromycin
Table S8 purports to analyze the patients that did not take Hydroxychloroquine. Only 46% (486 out of 1062, both arms) of the patients recovered by Day 29 without the use of Hydroxychloroquine. This percentage is much more significant than any Remdesivir percentages.
This table shows median “time to recovery” (MTTR) organized by how soon RDV or placebo was administered after symptom onset. The First Quartile (“Q-I”) of patients were randomized and treated 6 days or less after the onset of symptoms, the Fourth Quartile (“Q-IV”) were randomized and treated 13 days or later. Only patients, who recovered by Day 29 are included. The rest (~30%) are those who died, quit, or not recovered by Day 29.
The Q-III (10 to ≤ 12 Days) had the shortest recovery time – 7 days. If RDV were an effective antiviral for SARS-COV-2, the Q-I would have the quickest recovery because they got it early. The recovery would be significantly longer in the Q-III, when the viral stage is over in most patients. However, the Q-III has the shortest MTTR for RDV. This contradiction alone refutes the hypothesis that RDV is an effective antiviral.
Another remarkable thing is that the MTTR for the placebo group in the Q-I is 24 days, which is 1.5-2x longer than for other quartiles. For placebo, one would expect similar times to recovery, or a monotonous change.
Both contradictions are resolved by accepting that many patients received effective anti-viral treatment, other than RDV.
- Most patients who were randomized in Q-I did not receive this other antiviral treatment and thus, had the longest average recovery time.
- Patients who were randomized in Q-II and Q-III were more likely to receive the other antiviral treatment early, upon symptom onset, and before receiving Remdesivir. Therefore, both RDV and placebo arms in Q-II and Q-III recovered the quickest.
- The RDV arm of Q-III had the shortest MTTR of only 7 days, this means that most of this group’s patients did not receive the full 10-day course of toxic RDV. This effect, stemming from receiving less Remdesivir, created the gap between it and the other groups.
- Q-IV was less likely to receive the other antiviral treatment, or received it too late, and had longer MTTR
- In each quartile, RDV arm reported results than “placebo” arm because of patients’ selection or another manipulation
Notice that somebody, convinced that HCQ has no effect, might misinterpret this statistic as evidence that HCQ interferes with the action of RDV.
HCQ and RDV
Gilead claimed incorrectly and dishonestly that HCQ interferes with RDV antiviral effect, and the FDA slavishly repeated those claims. It is explained above how the statistics of trials using both HCQ and RDV might be misinterpreted this way.
In addition, Gilead presented results of a lab trial in a cell culture, in which chloroquine phosphate interfered with conversion of RDV into what Gilead called “Remdesivir triphosphate” (GS-441524 triphosphate, or GS-443902). This result is not related to HCQ, which is taken as hydroxychloroquine sulfate.
WHO has just published a preprint of the preliminary report (WHO, 2020) from the Solidarity trial of four antiviral drugs for COVID-19. It confirmed the lack of efficacy of RDV, which was administered as recommended by Gilead. This is, even though Gilead is well connected with WHO and makes substantial contributions to it.
The Solidarity report has also alleged a lack of efficacy of HCQ, but HCQ was administered in toxic doses, using the same regimen as in the RECOVERY trial – 2,400 mg in the first 24 hours (6x recommended dose), 800 mg for the next 9 days (2x recommended dose). These doses are 4-5x higher than the recommended doses for COVID-19 (and even for malaria).
Disclosed and Undisclosed Conflicts of Interest, related to Gilead
- Thomas F. Patterson, M.D. – Consultant to Gilead (Personal Fees) in 2019
- William R. Short, M.D., M.P.H., – Consultant to Gilead (Personal Fees) (reported in 2019 and 2018)
- Norio Ohmagari, M.D., Ph.D., – One of the researchers in another Gilead-sponsored study of RDV for COVID-19 (Grein et al.), which completed in early March. This counts as a research Grant.
The following grants and personal fees, received from Gilead over the prior 36 months, have been disclosed by the following report authors:
- Anu Osinusi – Employee of Gilead Sciences, Inc.
- Thomas Benfield – Grants and Personal Fees
- Gerd Fätkenheuer – Grants and Personal Fees
- Roger Paredes – Grants and Personal Fees
- Anne Luetkemeyer – Grant(s)
- Sarah Pett – Grant
- Giota Touloumi – Grant
Unexplained Late Start of Treatment
It is surprising that a drug that is meant to act as an antiviral was given much too late. Only 25% of patients started the RDV treatment within 6 days from the onset of symptoms. 75% of patients started it between 7 and 34 days from the onset of symptoms, too late for antiviral.
With a couple exceptions, the named NIAID/Gilead researchers who conducted this study appear to have most of their experience in HIV, and in attempts of using RDV for Hepatitis C or Ebola, rather than acute respiratory tract infections. Thus, some of the decision makers could be not fully aware of the two-stage dynamics of COVID-19. The viral phase is followed by the immune response phase, and the immune response overreaction is more dangerous than the virus. Immune response decreases the viral load and impact, so these phases overlap. Their overlap is frequently classified as its own phase (Siddiqi – Mehra, 2020, Fig. 1).
It should be noted that early administration of RDV cannot be done routinely because RDV is administered by an infusion (not an injection), thus requiring a hospitalization. RDV is not safe, and its dangers are unknown, thus justifying its use only when the diagnosis is certain, and the patient is sick enough. Finally, one cannot charge $3,000 for treating cough and fever.
In COVID-19, the phase and severity of the disease are routinely confused. Patients in the early (viral) phase are routinely classified as mild. Patients with a strong immune overreaction in the late (immune response) phase are routinely classified as severe.
Antivirals are useless after the viral phase. Hydroxychloroquine is useful in all phases of COVID-19 because it is both antiviral against SARS-COV-2 and immunomodulator.
Even if RDV were an efficient antiviral, its only effect in the immune response phase is toxicity (Zampino et al., 2020). It is expected to be especially dangerous to severe patients. Even the manipulated data shows higher mortality rates in the RDV arm of the most severe subgroup.
Suspicion of Differential Care
The placebo arm was reported slightly worse than the RDV arm at the baseline: 30% vs 24% patients on invasive mechanical ventilation (IVM, also called intubation) or ECMO, per Table S1.
Table S3 shows that the percentage of placebo vs RDV patients that received ECMO or intubation in the process of treatment was 45.5% : 34.6% (as-treated vs intent-to-treat population; slightly smaller). This is an increase of 52% in the placebo vs 44% in the RDV group, which cannot be explained by properties of RDV in this trial.
Most of the increase of ECMO patients was due to intubation. Intubation is a controversial procedure in the COVID-19 treatment. It is believed to have been used excessively in the pandemic and has sometimes caused avoidable deaths. The percentage of intubated patients (“need for invasive mechanical ventilation”) is part of the measured outcomes. Given lack of positive effect from RDV and the incentives of the sites, disproportional intubation of placebo patients raises the suspicion that some of them were intubated unnecessarily to improve results for Remdesivir and Gilead. Some of the patients might have died as a result.
- Many hospitals have concluded that RDV is not what was promised, according to Reuters.
- This study was published in NEJM on October 8, the same day as another paper reporting old news on the tragicomic RECOVERY trial, and a shrieking anti-Trump editorial “Dying in a Leadership Vacuum”. This is not an accident, but a pattern. The PR version of this paper was published on May 22 – the same day as the infamous (Mehra et al., 2020) was published in The Lancet. Of note, Mehra was affiliated with Brigham and Women’s Hospital, which was contracted by Gilead for an RDV for COVID-19 clinical trial, possibly another one.
- It is strange that the research of a potential treatment for pandemic disease started at Phase 3, which is needed for FDA approval of a novel drug. Phase 3 trials do not allow flexibility necessary for research, especially in time of emergency. It seems like the pandemic was used as an opportunity to shortcut the drug’s approval, which could not pass scrutiny under normal circumstances.
- Contrary to the principles of scientific archiving, the Final Report re-uses the DOI number and the URL (doi: 10.1056/NEJMoa2007764, https://www.nejm.org/doi/full/10.1056/NEJMoa2007764) of the Preliminary Report, despite massive changes in the text and appendices.
- The Preliminary Report claimed that the study outcome measure was changed on April 2, 2020 “without any knowledge of outcome data from the trial and before any interim data were available.“ This subsentence is removed in the Final Report, essentially admitting that it was changed with the knowledge of the outcome data.
- The unusually large number of sites and their selection by Gilead might indicate an attempt to co-opt influential institutions. (Roussel & Raoult, 2020) found a nearly perfect correlation between the amounts received from Gilead and public opposition to hydroxychloroquine in France.
- If anybody wants to test the hypothesis that HCQ and HCQ + AZ are effective for COVID-19 treatment, re-analysis of the raw data from this study would work.
- This analysis does not cover reporting of adverse events from the study
No Competing Interest
The author declares no competing interest.
No funding was provided for this work.
All relevant ethical guidelines have been followed.