Now this is a covid antibody test! Not only looks good on its face w/ 100% sensitivity (likelihood of seeing antibodies if they are there) & 99.8% specificity (not seeing them if they aren’t there).... roche.com/media/releases…
...but b/c Roche validated on over 5000 samples, they know specificity to within 0.2%. That matters!!! Had Stanford used this test (and could maybe still re-analyze their samples), it would have been able to reach valid conclusions about SARS2 lethality.
Here’s what it means for you that Roche is launching such an accurate and well characterized antibody (serology) test. Let’s say that in your region 2% of people have been infected. You wonder whether you have been. You remember feeling off a few weeks back.
So you go to a clinic and get your blood draw. Roche’s instruments are high throughout (300/hour), so your results could come back quickly. Out of 1000 people tested in your region, of whom 20 are positive (and you want to know if you’re one of them)...
...this test would probably detect all 20 of them (maybe 19 to be conservative). It would also incorrectly tell 2 but maybe as many as 4 other people that they are positive for antibodies. That’s what 99.8% specificity with a confidence interval of 99.65-99.91% means.
The test’s false positive rate is most likely between 0.1-0.35%. So worst case, it will tell 23-24 people out of the 1000 that they’re positive & 19-20 of them actually would be. That’s means that if the test told you that you were positive, you would be 83% sure that you were.
We’re talking positive for SARS2 antibodies, which means you were infected, not that you are. That might make you believe you are immune, but we can’t yet be certain of that. On the one hand, you could only be 83% sure you even have antibodies. On other hand, even if you do...
...those antibodies might not protect you from reinfection. Odds are they would protect you from getting a serious case of covid, but you might still get a very mild, superficial infection that makes you a danger to others at risk of serious covid. So you still should...
...wear a mask and social distance. To be more confident in a positive antibody result, it’s possible that there would be benefit from rerunning the same test or a different antibody test. We don’t know yet, but will be great to find out.
What if you are negative? Might that be wrong? There is one way this test could miss that you’ve been infected: if you get tested too soon after your infection. It takes your immune system a couple of weeks to develop antibodies after an infection. Roche is clear that...
...their test is as accurate at detecting antibodies in recovered patients as they say but only for people who are at least 2 weeks post infection. So what you shouldn’t do is take this test as soon as you feel symptoms or right after you feel better. Wait 2 weeks.
And this test is even more reliable in an area with higher rates of infection. If 10% of a 1000 people are infected, the test would tell 103 people that they are positive and 100 would be. If you got a positive result, you would be 97% sure it was right.
So thank you @RocheDia for finally developing an industrial strength test and validating it on a ton of samples so we actually know how good it is. Let’s now rerun samples from all those surveillance studies like Stanford’s to see how many people were really infected.
With those data, we can then truly estimate how lethal SARS2 is, not just in general but for people with various risk factors such as age, copd, diabetes, coronary disease, and obesity.
If we’re lucky enough to discover that there are people who have a much lower risk of serious disease, we can consider data-drive risk-based lockdown strategies to ease psychological strain; I’ve written about that here: link.medium.com/ctFIsVLTf6
By the way, tests are essential to proper healthcare. They are grossly underappreciated, & America’s insurance plans are brutally myopic in how hard they strive to deny paying for new tests. As a biotech investor, I can tell you that it’s way harder to justify investing in Dx...
...because I know that even if a company launches a good one, insurance plans will stop at nothing to deny coverage until the company has spent a ton of money marketing the test to docs to get them to pressure insurance to pay for it. That friction is discouraging.
Recall how insurance plans refused to cover covid tests for the first Americans who went in for testing. Truly, America’s insurance system is just plain stupid- its bureaucracy has trapped even its best people into limply supporting its stupid policies. It has to change.
Because of how hard insurance tries to not pay for new diagnostic tests, investors have learned that investing in diagnostic innovation is less rewarding than developing new drugs (except antibiotics- America doesn’t pay much for those either- yes, stupid).
And yet, you can’t treat a patient properly until you know what their problem is. Think of every episode of House; they spent 95% of the show searching for the diagnosis and the last few minutes administering the right treatment. That’s real medicine. images.app.goo.gl/Hq7pobgBoQG6qr…
Congress should pass reforms to increase Dx reimbursement to create incentives for much more Dx innovation. We could be diagnosing cancer earlier, when it’s more treatable, even curable. We can catch bacterial infections earlier & select right antibiotic from the start.
All of healthcare could be revolutionized with better tests. The public’s and Washington’s enthusiasm for better covid tests (and willingness to pay more for them) should encompass better diagnostics & screening tests across all diseases & disorders, from autism to Alzheimer’s.
It’s not that dx innovation is not happening. There is a diagnostics industry. But it’s at the stage that biotech drug development sector was at in the 1990s, a mere glimpse of what it could be, and could blossom if we gave it the attention and resources it deserves.
