May 29, 2020 This Week in Cardiology Podcast

John M. Mandrola, MD


May 29, 2020

Please note that the text below is not a full transcript and has not been copyedited. For more insight and commentary on these stories, subscribe to the This Week in Cardiology podcast.

In This Week’s Podcast

For the week ending May 29, 2020, John Mandrola, MD comments on the following news and features stories.

Brief COVID-19 Numbers Update

When I recorded last week, there were 1.6 million cases in the United States. Today there are roughly 1.75 million. The rate of growth of 1.08x was not changed much from the last few weeks. US deaths last week were 96,000 and today about 102,000; the rate of rise in deaths declined from 1.12x to 1.06x this week.

That’s now about 6 weeks with rates of cases and rates of death on a plateau. The question for the next month or so is whether opening up will lead to change in slope for death rates.

A quick global note: there are now almost 6 million COVID-19 cases with 2.5 million recovered. Brazil continues to look scary, with its rate of rise of cases and deaths. I still don’t know what to make of India, with its exceedingly low rate of death of only 3 per million. How is this country, with its massively populated cities, faring so well?

Nursing Homes

We had a small surge of COVID-19 patients in recent days here in our medium-sized city of Louisville, Kentucky. It was easily handled by the hospitals, which are completely ready with plenty of excess capacity. The notable feature of this surge: it was all from a handful of nursing homes. I looked into this a bit and learned how hard it is for nursing facilities to protect against pathogens.

Remember, healthy people don’t get admitted to the nursing home. These are older patients with multiple conditions, who are often quite close to end of life. A stat from the CDC: In 2016, well before the pandemic, nursing-home residents made up just 0.4% of the U.S. population, but accounted for 19% of deaths. I also learned that more than half of residents in nursing homes need up to 4 hours of hands-on-care daily, for things like bladder incontinence, mobility assistance, and feeding. Another CDC nugget: pre-pandemic, as many as 380,000 people die of infections in longterm care facilities every year.

Recently, I was talking with a patient who has a brother and sister in a nursing home and he said that both of them are locked in their rooms with no visitors. My patient said both of his siblings were of sound minds and were pleading to get out even at the risk of getting coronavirus.

I don’t have a policy answer for the problem of COVID-19 in nursing homes but I do know that these are highly vulnerable patients who are susceptible to any insult, be it infectious or non-infectious; due to their frailty, these patients need lots of human contact; solitary confinement is pretty darn inhumane. So, when we implement strategies to mitigate infection-related death in this cohort, we should be careful about our metrics for success.

Testing Basics

Special Shout Out this week to Dr. Venk Murthy who, in a Twitter thread, using 2x2 contingency tables, explained a huge hurdle for widespread testing of asymptomatic people in areas with low virus prevalence.

Norway published a story of how they engaged in screening asymptomatic people for the virus. They used the PCR test because of its super high specificity. Yet, Norway estimates a viral prevalence of 1 out of 10,000. Norway also told us if you test a town with 12,000 people, one person has COVID-19 and has a positive test. But for every true positive, you get 12.5 false positives. That means the probability of having the disease if you have a positive test is only 7%.

Venk used these numbers to fill in the 2x2 table and back calculated a PCR test specificity of 99.9%. You see the problem: even with an amazingly specific test, if you have a low disease prevalence, then for every one positive, you have 12 false positives. This has huge implications for businesses, professional sports teams, or other large organizations that think testing is the answer.

COVID-19 and Cardiology

The big news this week on both fronts came from an observational study published in the Lancet looking at the use of hydroxychloroquine (HCQ), chloroquine (CQ), or a macrolide, primarily azithromycin. The study reported that the drugs were associated with significant harm, including VT.

Recall that HCQ and AZ use was codified into COVID-19 protocols in some of the most respected programs in the world. This was shocking to me because of the utter weakness of the evidence for efficacy. Recall that the original study from southern France had no randomization and the authors simply removed patients who did worse on the drug from their analysis. Since that initial study, two small RCTs and numerous other observational studies have been published and there has been no signal that these drugs benefit patients with coronavirus infection.

The new Lancet study was a large multinational registry analysis of the use of HCQ or CQ with or without a macrolide for treatment of COVID-19 from 670 hospitals on 6 continents. Four treatment groups plus a control; patients who received none of these treatments formed the control group. Patients on ventilators and patients who received remdesivir were excluded. The primary endpoint was in-hospital death or VT.

The study included 96,000 patients from a Surgisphere database. About 15,000 pts were in the 4 treatment groups and about 82,000 were in the control arm. Like all observational studies the authors attempted to adjust for baseline differences that occur because the groups were not randomized.

The results were huge effect sizes: a 33-45% higher rate of death and a 4-5-fold higher rate of ventricular arrhythmia for patients taking the drugs alone or in combination. The authors conclude “We were unable to confirm a benefit of HCQ or CQ, when used alone or with a macrolide, on in-hospital outcomes for COVID-19. Each of these drug regimens was associated with decreased in-hospital survival and an increased frequency of ventricular arrhythmias when used for treatment of COVID-19.

My friends, this sort of analysis is deeply flawed. Here are some comments from famous people I found on Twitter when the study was published.

  • Prof Rod Hayward from Michigan called it “non-evidence” that should not have been published.

  • Drs David Brown and Mike Johansen said it was unreliable data because you can’t adjust away the differences in selection bias.

  • Dr Robert Yeh wrote that the large sample size does nothing to overcome confounding and in fact, the lack of granular data can make confounding even more likely.

  • Dr Robert Califf: I'm disturbed by the number of highly health literate people (including some regulators) who look at observational studies and pronounce that the evidence in already in.

Columbia statistician professor Andrew Gelman has been chronicling other problems with the study, including inconsistencies in death numbers from Australia, dubious data from Japan, and inconsistencies in drug dosing. Then this morning, Dr Allen Cheng, an ID doc and epidemiologist from Australia, tweeted an open letter of concern to the Lancet about many more inconsistencies in the data. These inconsistencies do not necessarily imply fraud, they simply express concern that the data is inaccurate, which is hardly implausible because the database extracts information from electronic health records.

There are ongoing RCTs of these agents and the answer will come from proper trials.


Just a couple of hours after this podcast posted last Friday, the New England Journal of Medicine published preliminary results of the double-blind placebo-controlled ACTT-1 trial of remdesivir in COVID-19.

On May 16, the Lancet published an RCT performed in Hubei, China, with 237 patients with COVID-19. Note: 2:1 randomization with 158 in the remdesivir group vs 79 in the placebo group. Remdesivir use was not associated with a difference in time to clinical improvement. Remdesivir also did not result in significant reductions in viral RNA loads or detectability in upper respiratory tract or sputum specimens despite showing strong antiviral effects in preclinical models of infection with coronaviruses. Also, mortality was not significantly different.

Then comes the ACTT-1 trial published on Friday evening before a holiday weekend. The trial has roughly 1000 patients equally distributed in remdesivir vs placebo groups. Patients in the remdesivir group had a shorter time to recovery than patients in the placebo group (median, 11 days, as compared with 15 –this was highly significant result with a very low p-value). The tough signal was mortality. Mortality by 14 days were 7.1% and 11.9% in the remdesivir and placebo groups.

An important sentence—one not emphasized in the mostly celebratory coverage in the mainstream media: The Kaplan–Meier estimates of mortality by 28 days are not reported in this preliminary analysis, given the large number of patients that had yet to complete day 29 visits.  The biggest talking point on this trial was the decision faced by the independent Data and Safety Monitoring Board who reviewed interim results. This review included a total of 482 recoveries (exceeding the estimated number of recoveries needed for the trial) and 81 deaths had been entered in the database. Here was the dilemma – they had a clear signal of benefit in time to improvement but the absolute risk reduction of 3.8%, though quite large, did not meet statistical significance.

If the mortality signal was real, it is clearly worth giving this drug. But due to the small numbers of deaths, you can’t be sure it wasn’t due to chance.

The trial was unblinded and patients in the placebo group were allowed to take remdesivir. And now, going forward in this adaptive RCT design, the new control group is remdesivir. This means we will never know for certain whether this drug actually reduces death. Not only do we have uncertainty on the mortality effect in this trial, we have the Lancet trial, with similar patients, showing no effect.

I consider the Lancet HCQ observational study and the remdesivir RCT analyses good practice for evaluating cardiology studies.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: