No news is bad news
We all hope that effective drugs to treat COVID-19 disease can be found. The number of lives at stake is enormous. It’s not just deaths from infection, but also from the ensuing economic disruption. We learned in the Great Recession of 2008 that long-term unemployment results in a loss of 1 to 1.5 years of life expectancy. Ending the economic shutdown is indeed a matter of life and death.
But antiviral drug development is a tough game. It is much harder than developing antibiotics to treat bacterial infections (see box). And new drug development of any kind takes years. Our only hope in the near future is to find an existing drug that also has anti-coronavirus activity.
Chloroquine (an anti-arthritis and anti-malarial drug) and remdesivir (an anti-viral developed to treat Ebola) are undergoing clinical trials. The planned closing dates for those trials are in April and May. But if the drugs are highly effective — if they can stop the pandemic and return us to normal life — we will know much sooner.
Cognitive computing - a skill-set widely considered to be the most vital manifestation of…
As its users, we have grown to take technology for granted. Hardly anything these days is as commonplace and…
It is unethical to withhold treatments known to be effective. The current trials are placebo-controlled, meaning that half the patients get the antiviral candidate and the other half get only a sugar pill (for chloroquine) or saline (for remdesivir). Placebo-controlled trials are ethical only when there is no known effective treatment. That is the current situation with COVID-19.
Highly effective drugs reveal themselves during the course of a trial. When this happens, the only ethical course is to terminate the placebo arm and treat all patients with the drug. Here is one example from a heart disease trial.
The statistical criteria for early termination are set out in any well-designed trial protocol. These criteria vary according to the best judgement of the researchers. But it won’t surprise you to learn that the stronger the positive effect, the fewer patients needed to demonstrate the superiority of treatment over placebo.
The graph below is an illustration. Let’s say you decide that a 20% reduction in deaths (or any clinical indication) among COVID-19 patients is meaningful. The graph shows the number of patients you would need in each arm to have 95% confidence that the 20% threshold was met.
The exact numbers here depend on the thresholds chosen beforehand. These are judgement calls made by experienced clinicians. But the principle is clear: big effects are detected with few patients. They will be seen early in the trial, within the first few hundred patients.
For perspective, it will take a 10-fold reduction in death rates to make COVID-19 no more deadly than seasonal influenza. Detecting an effect of this size would require only 167 patients each in the treatment and placebo arms.
Given the number of COVID-19 patients, it won’t take long to accrue enough to see an effect like this. Conversely, detection of small effects will require thousands of patients and many weeks or months.
The NIH remdesivir trial started on Feb 25. A first evaluation of data was scheduled for day 15 of the trial, and another evaluation after the first 100 patients. Day 15 has come and gone with no announcement of results. This could mean either too small an effect or too few patients. But as the graph above shows, detecting big effects does not require many patients. We are now (Mar 26) on day 30 of the trial, and no word has yet been forthcoming.
Every day that passes decreases the likelihood of there being a game-changing drug out there. This means that case fatality rates are not going to plunge, ICU stays will not be shortened, life will not be returning to normal. At least not soon, and not because of remdesivir.
Some possibly encouraging results for chloroquine have been announced, but as this review notes “…no data has been provided …results were produced in ten different hospitals and possibly from a number of different clinical protocols … it is difficult to reach any firm conclusion.” Big, strong effects don’t have this kind of ambiguity, or require subtle statistical analyses to detect. They hit you over the head.
That said, drugs don’t have to be game-changers to be beneficial. Even if they have little effect on patients with advanced disease (the most likely outcome), they still could be of use as prophylactics. Giving them to health care workers would have tremendous value in keeping hospitals open and running.
But getting the economy back up and running? The window for the drugs to do that is rapidly closing. Public health measures are our surest bet, and we should not expect too much from drug therapy.
Update 6 Apr: Still no word from the remdesivir trial. The likelihood that it is any kind of “cure” for COVID-19 is close to zero, although it still may be shown to have some clinical benefit.
A French group led by Didier Raoult has published a couple of studies showing encouraging results for hydroxychloroquine plus azithromycin. The first study had a decent design (a placebo-controlled randomized trial). But it has many flaws. The biggest one is that it excluded from analysis those patients who got the treatment but then got so sick that they died or went to the ICU before completing the trial protocol. When you count those patients (as you must), most of the reported effect goes away. There are many other problems with this work and with the second study as well. You can read about them in detail here. Bottom line is that these studies don’t show what some politicians claim they do. The jury is still out on chloroquine.
Disclosure: I have held stock in Gilead, maker of remdesivir, since 1999 when they acquired my employer (NeXstar Pharma). I never worked for Gilead (they drew a line across the org chart and fired everyone above it) and don’t know anyone there, and have no inside information.