Doctors (well, some doctors) are being cautious here, for good reason. After the fiasco of that last big hydroxychloroquine study, every splashy new finding about the coronavirus and drug efficacy is suspect. Researchers are paddling around in a toxic brew of political narratives, miracle-cure claims, and media glory for whoever makes a “breakthrough.” The temptation towards confirmation bias is enormous.

This would be wonderful news if it bears out, though. Until this morning there was no known drug that successfully reduced deaths from COVID-19. The one drug proven to help patients recover faster, remdesivir, didn’t reduce mortality rates in a statistically significant way in Gilead’s trials. A drug that helps people who aren’t severely ill get better sooner is good; a drug that stops severely ill people from dying is better. Oxford’s team says they have it — dexamethasone.

The best part: Unlike remdesivir, it’s not an obscure, expensive medication. It’s a common steroid. Cheap and widely available.

The study randomly assigned 2,104 patients to receive six milligrams of dexamethasone once a day, by mouth or intravenous injection. These were compared to 4,321 patients assigned to receive usual care alone.

In patients who needed to be on a ventilator, dexamethasone reduced the death rate by 35%, meaning that doctors would prevent one death by treating eight ventilated patients. In those who needed oxygen but were not ventilated, the death rate was reduced 20%, meaning doctors would need to treat 25 patients to save one life. Both results were statistically significant.

There was no benefit in patients who didn’t require any oxygen.

One in 25 sounds underwhelming in the abstract sense but it amounts to thousands of saved lives given the magnitude of the death toll in the U.S. (The IHME model now predicts 200,000 dead here — by October, which is right around when scientists expect the second wave to begin.) Outside scientists are suspicious that today’s results were reported via press release rather than a peer-reviewed paper whose methodology could be closely scrutinized…

…but the researchers behind the study are ecstatic:

“This is the only drug so far that has been shown to reduce mortality, and it reduces it significantly. It’s a major breakthrough,” Dr. Peter Horby, chief investigator on the study, said in a statement. Head researcher Dr. Martin Landray added that the drug’s low cost would lead to an even greater impact.

“There is a clear, clear benefit. The treatment is up to 10 days of dexamethasone and it costs about £5 per patient. So essentially it costs £35 to save a life. This is a drug that is globally available,” he said, emphasizing that for every eight patients on ventilators that are treated with the drug, one life could be saved.

The British government’s chief science advisor called it “a ground-breaking development in our fight against the disease.” Why would a steroid work on a virus like SARS-Cov-2? It might not, actually; rather, the steroid may be mitigating one of the most notorious complications of COVID-19, the “cytokine storm” that happens in some patients when their immune system goes haywire in trying to fight the virus off and begins attacking the body itself. Steroids reduce inflammation. It may be that dexamethasone is suppressing the inflammation in some far-gone patients just enough to help them beat back the virus and eventually recover.

That would also explain why it has no benefit on patients who are infected but don’t require oxygen. In a mild-ish case like that, when the immune system hasn’t turned hyperactive, a steroid logically might not do much of anything. If that’s what’s happening here, that the drug is essentially mitigating fatal side effects of COVID-19 rather than the virus itself, then it’s of no use as a prophylactic. It’s for people who are at death’s door.

Speaking of higher survival rates, a mystery has begun to brew here at home. While it’s true that some states like Texas and Arizona are seeing worrisome spikes, the national picture overall continues to look a bit brighter. Yesterday just 375 deaths were recorded, pushing the seven-day average in the U.S. down to 700 or so after it routinely topped 2,000 daily on weekdays in April. Check out the trend via Worldometer:

Good stuff. But here’s the mysterious part: The trend in confirmed cases doesn’t resemble it. It’s basically flat, not downward.

If roughly the same number of people are getting sick each day as were getting sick weeks ago, and yet the number of deaths is declining, why is that? The case fatality rate is clearly dropping — but how come? Pundits are kicking around theories:

All good theories. Maybe the weather plus better treatment at hospitals plus older people being stricter about self-isolating plus the “dry tinder” effect Foster mentions means that we’ve reached a phase where infections are more common among Americans who are better able to fight them off.

But here’s another possibility, more prosaic: Maybe the trend in confirmed cases captured in the graph above is simply bogus. Our testing capacity continues to improve from its woeful levels earlier in the pandemic. It could be that there were many, many more infections in March and April than were known because testing was largely unavailable at the time and that there are many fewer now. In that case, the reason the graph of confirmed cases looks “flat” isn’t because infection levels are steady, it’s that we’re simply detecting more infections even though there are less of them out there. This isn’t true everywhere — cases really are going up in Texas and Arizona. But if we had had the same testing capacity on March 15, say, as we have now, the confirmed-cases graph might look very much like the fatality graph. A huge spike, followed by a (more or less) continually descending downward slope.