The Remdesivir Results, Visualized

3 minutes, 3 charts to understand the recent study results

Tom Seufert, MD
4 min readApr 30, 2020
Image by Myriam Zilles at Pixabay

The National Institutes of Health (NIH) recently announced preliminary results of a randomized controlled trial comparing Gilead’s experimental anti-viral drug remdesivir to placebo, finding the drug beneficial for recovery time and with a statistically insignificant trend toward a mortality benefit.

“Patients who received remdesivir had a 31% faster time to recovery than those who received placebo.”

Confusingly, a study published in the Lancet the same day showed no benefit from the drug. Both were randomized, double-blind, placebo-controlled trials: the gold standard.

“Remdesivir use was not associated with a difference in time to clinical improvement.”

So…Which is right?

Time will tell. The Lancet study was stopped early due to a drop-off in available patients as COVID-19 was brought under control in Wuhan. The NIH claims sound promising, but have yet to be subjected to peer review.

In the meantime I thought it would be helpful to illustrate exactly what the NIH is claiming.

Shortened Recovery Time

The NIH stated that “the median time to recovery was 11 days for patients treated with remdesivir compared with 15 days for those who received placebo.”

Image comparing recovery times
Image by Author; Icons by Vectors Market at flaticon.com; Calendars by freelatestcalendar.com

As shown above, a study patient hospitalized on May 1st and receiving standard care might expect to be discharged on May 16th; if s/he were given remdesivir they would go home 4 days earlier. Of course, the NIH values are medians, not exact predictions — so, 50% of patients would go home earlier, and 50% would stay longer.

This represents a clear benefit, and undoubtedly heralds wide use of remdesivir among hospitalized patients as soon as it’s available. However, as anyone who has been hospitalized (or works in medicine) knows, an 11-day hospital stay is still a very long one! Further, the drug must be administered intravenously, which will largely limit its delivery to hospital settings.

The results suggest that use of remdesivir is by no means a panacea for COVID-19. The disease’s strain on the healthcare system will remain substantial if it is not well controlled.

Trend Toward Reduced Mortality

Now, in order to make it home, you have to survive, right? So, what about mortality?

“Results also suggested a survival benefit, with a mortality rate of 8.0% for the group receiving remdesivir versus 11.6% for the placebo group (p=0.059).”

A picture’s worth a thousand words here:

Pie chart showing difference in mortality
Image by Author

For this sort of outcome, I prefer a measure called Number Needed to Treat (NNT): the number of patients who would have to receive an intervention in order for a single patient to benefit (survive, in this case).

The beauty in NNT is that it’s conceptually quite simple: You don’t need to know the math (though if you must, sigh: NNT=1/(11.6%-8%)=27.8). For the NIH study, remdesivir has a NNT of 28 for preventing mortality.

So, if you administered remdesivir to 28 patients similar to those studied, you would expect to save 1 life. If this sounds less than earth-shaking, that’s because it is.

Still, it’s better than what we’ve had up to date (nothing). And a surprising number of widely used medical interventions must be used on many patients for one to benefit. For example, there’s an NNT of about 50 for using aspirin to prevent cardiovascular disease.

Finally, with massive worldwide concern over COVID-19’s death toll, why isn’t the NIH study’s survival benefit its main headline? The answer lies at the end of the quote above: (p=0.059).

Image by Author

In English, “p=0.059” means that the NIH believes, based on their data, that there is about a 94% chance that remdesivir improves mortality for patients similar to those studied. However, they acknowledge a 6% chance the effect they observed was due to chance.

It’s arbitrary, but the generally accepted threshold for accepting that a study finding is not due to random chance is less than 5%, or p<0.05. There are no guarantees in medicine, but one would hope and expect that a larger trial, with more patients, would confirm the trend.

Overall, the NIH study is extremely good news. But the scientific and medical communities, to say nothing of the public at large, still have a long road ahead.

--

--

Tom Seufert, MD
Tom Seufert, MD

Written by Tom Seufert, MD

Emergency physician, clinical informaticist, software engineer, author, dad.