What We Learned About HCQ and Remdesivir This Week

Editor's Note: The Lancet study discussed below and another COVID-19 study in the New England Journal of Medicine were both retracted on June 4, after authors said they could no longer vouch for the data. Click here to read Milton Packer's discussion of these retractions and his call for a reinvention of the peer-review process.

Several weeks ago in this blog, I reviewed the available data on the treatments for COVID-19. My essay focused on chloroquine (CQ) and hydroxychloroquine (HCQ), since some had hyped these drugs as being effective in preventing or treating COVID-19, in the absence of any good supportive evidence.

In my blog post, I reached two conclusions. First, there was no reliable evidence that these antimalarial drugs favorably influence the clinical course of COVID-19. Second, there was good evidence that these drugs are capable of causing serious cardiac toxicity. I concluded that the use of these drugs outside of a clinical trial was practicing politics rather than medicine.

My blog received a near-record number of comments -- 127! A large number of the comments were highly critical of my conclusions. Some argued that I have been ignoring essential evidence and that I was being too harsh on drugs that had been used for decades for the treatment of malaria. I responded by noting that CQ and HCQ are known to cause cardiotoxic effects, and they were now being prescribed for a novel disease at much higher doses. And they were now being combined with azithromycin, another drug that can potentiate the adverse cardiac effects of CQ and HCQ. As expected, many readers presented political arguments to counter my medical facts, but political thinking is a poor counter to scientific evidence.

Last week, the largest observational study on the efficacy and safety of CQ and HCQ was published in The Lancet. This study evaluated the in-hospital outcomes of 96,032 patients with COVID-19 who were treated at 671 hospitals across 6 continents. The patients received CQ or HCQ (alone or in combination with azithromycin) as compared with none of these treatments. Patients on mechanical ventilation or who received remdesivir were not included in the analysis. The mortality was 9.3% in patients receiving neither drug, 16.4%-18.0% in those receiving CQ or HCQ, and 22.2-23.8% in patients receiving CQ or HCQ in combination with azithromycin. Use of CQ and HCQ did not prevent death in COVID-19 patients, but these drugs were independently associated with increased risk of death and increased risk of cardiac arrhythmia.

On Monday, the World Health Organization decided to pause its ongoing randomized controlled trial of HCQ for COVID-19 due to safety concerns.

Does anyone still advocate for the use of CQ and HCQ for COVID-19? Apparently, as of May 20, Jair Bolsonaro, president of Brazil, wants everyone with COVID-19 in Brazil to receive CQ or HCQ. Bolsonaro has been widely criticized for his denial of the seriousness of COVID-19 as well as his leadership of the crisis. As the disease spreads rapidly in Brazil, calls for his impeachment grow louder. Anyone who wants to take medical advice from Bolsonaro is free to do so.

Yet, the most important medical news regarding the treatment of COVID-19 last week was the publication of the NIH trial of remdesivir, a drug that has been shown to have antiviral activity against SARS-CoV-2. Uncontrolled case series in patients with COVID-19 had previously yielded uninterpretable findings. In contrast, the new trial (ACTT-1) was placebo-controlled and double-blind, with time-to-recovery as the primary endpoint.

On April 29, NIAID Director Anthony Fauci, MD, announced that the ACTT-1 trial had achieved success on its primary endpoint. His announcement was criticized in the press for reasons that made no sense to me. Some wondered about the fact that the trial had changed its primary endpoint shortly after the start of recruitment. However, a change in the primary endpoint performed in a blinded manner is not uncommon in clinical trials and should not have raised any concerns. Furthermore, in this case, the change was proposed after only 72 of 1107 patients had been enrolled. I tried to assuage concerns about the endpoint change in my comments to a Washington Post reporter.

The results of the trial were posted online by the New England Journal of Medicine on May 22. A total of 1063 patients with COVID-19 were randomized (double-blind) to placebo or remdesivir after a median of 9 days after symptom onset. Remdesivir shortened the median time to recovery from 15 to 11 days (a 27% reduction), P<0.001. The magnitude of the benefit did not depend on whether patients had been randomized early or late after the onset of symptoms. However, patients in the trial who had advancing respiratory insufficiency -- which is believed to be related to cytokine storm rather than due to viral replication -- did not respond favorably to remdesivir.

Importantly, the risk of death was 30% lower in the remdesivir group than the placebo group at 14 days after randomization. The hazard ratio for death was 0.70 (95% CI 0.47-1.04). As per New England Journal of Medicine policy, no P value was provided, but it can be estimated at 0.06-0.08. The largest subgroup of the patients in the trial (comprising 40% of the study population) were those receiving supplemental oxygen without ventilatory support. The hazard ratio for mortality in this subgroup was 0.22 (95% CI 0.08-0.58), which was nominally significant. In this subgroup, 19 patients died in the placebo group and four died in the group receiving remdesivir. Although this reduction in mortality is impressive, it is based on a sparse number of events.

In the overall trial, there were 28 serious respiratory failure adverse events in the remdesivir group and 42 in the placebo group, a 35% risk reduction. The P value for this difference can also be estimated at 0.06-0.08.

How should one interpret borderline P values for differences between the two groups in the risk of death and in the risk of respiratory failure? I have two responses.

First, the occurrence of death and respiratory failure are extremely important events, and we have not had any treatment that favorably influences the clinical course of COVID-19. In light of these features, a 30% reduction in risk (assuming replicability) is exceptionally meaningful.

Second, I have written about the interpretation of borderline P values in an essay published in September 2019. Here are two excerpts from the paper:

"Statistical methods were applied to clinical trials to allow us to quantify uncertainty; they were not developed as a decision tool... . So the next time you hear investigators report definitive conclusions based on a P value being greater or less than some prespecified threshold, it might be time to remember the advice recently issued by 800 statisticians who begged clinical investigators to abandon P values for decision-making."

My main take-home point: P values depict probabilities of a false positive result. A P value less than or greater than 0.05 is not intended as a decision point for determining if a particular finding is "real" or "not real".

So anyone who labels the between-group differences with respect to mortality and serious respiratory failure events seen in the ACTT-1 trial as being "not significant" does not understand the meaning of a P value. Such a characterization assumes that a P value of 0.06 has the same meaning as a P value of 0.99 -- when in fact, these two P values depict two very different sets of conditions with widely different implications.

During the past several months, the FDA granted emergency authorization use for CQ, HCQ and remdesivir for use in patients with COVID-19. Its actions with respect to CQ and HCQ were not based on evidence but were related to political pressures.

In contrast, its actions with respect to remdesivir were based on knowledge of the results of ACTT-1 and were evidence-based.

Given these new results, anyone who is still advocating for CQ and HCQ while denigrating the evidence for remdesivir is following a conscious or subconscious political agenda that is not informed by evidence. Politics is certainly capable of distorting one's ability to think clearly and make useful judgments. And everyone is entitled to their opinions, regardless of their validity. As then-Secretary of State John Kerry once said, in America, "you have a right to be stupid if you want to be." This description presumably also applies to Brazil.

Last week, we learned a great deal about chloroquine, hydroxychloroquine and remdesivir for the treatment of COVID-19. Ignore the new evidence at your own peril.

Comment