A Faster, Cheaper Test -- Will It Really Democratize Coronavirus Testing?

Hi, everyone. Welcome to this episode of 16 Minutes. I'm Sonal, your host, and this is our show where we discuss recent headlines, put them in context of broader tech trends, and more. This week, we're covering the news that Abbott Labs got emergency use authorization, or an EUA, for a rapid antigen test that detects COVID-19 in 15 minutes and will cost $5. So we quickly cover how, where the test fits in the taxonomy of coronavirus testing with super quick explainers on key concepts that everyone needs to know, dig into what the data does and doesn't tell us, including doing some quick math with implications for widespread detection and testing, and briefly touch on practical considerations from a clinical perspective. Our expert in this episode is A6NZ Biogeneral partner Vanita Agarwala, who is a practicing physician, former director of product management at Flatiron Health, and former researcher who did graduate work in computational biology and human genetics. To start us off, Vanita, this latest test was described as, quote, a game changer by the assistant secretary in the Department of Health and Human Services. So that phrase dominated the headlines. And we've also been hearing a lot of discussion about what's working, not working when it comes to testing. Since the premise of this show is to help tease apart what's hype, what's real, would you say this news is real and what is or isn't the significance of it from your vantage point?

Yes, there are kind of 3 salient features that make it real. One is that Abbott shared the price point explicitly, not the cost, actual price that they will charge for the test, which is only $5 a test, which is meaningful. The second is the data that they presented and that they shared admittedly in a very small sample size on sensitivity and specificity of the test. Was generated actually from an anterior nasal swab, not the very painful, harder to obtain deep swab. And then the third, and actually the most important, is that this is an antigen test that's built using lateral flow technology and looks a lot more in form factor and equipment required to do the test like a pregnancy test. So, or like a Yaks stool test for those of you who've done those in the hospital. This is a test that is a credit card-sized little device, and that's the only thing that you need to actually run the test. There's not even a benchtop machine.

So let's talk specifically about the type of test and this— the fact that it is this type of viral antigen test. So first of all, in the taxonomy of all the different types of tests out there, in the early days of the coronavirus pandemic, everyone was very excited about the fact that we sequenced things so quickly and the significance of the RT-PCR tests. Can you talk about where this fits and against the taxonomy of broader testing options out there?

Yeah, so the RT-PCR test amplifies short portions of the virus's RNA genome that are specific to the COVID-19 virus to detect whether or not that virus is present in a patient specimen. That was against the backdrop of a broader kind of research sequencing effort that was ongoing to understand how is this virus moving, mutating, what strains exist, and so on, in which you have to sequence a much larger part of the viral genome to really understand that. Those were not diagnostic tests. Those were sort of research-based sequencing assays. In the diagnostic world, you're absolutely right. The RT-PCR whole wave of multiple test options were, and still are, the sort of gold standard for detecting whether or not a patient is infected. That's in contrast to antibody tests, which detect proteins called antibodies, which are suggestive of prior infection with the coronavirus. And we don't really know yet, frankly, exactly how the antibody testing is going to be clinically useful or whether it will really be an indication of ongoing immunity against the virus. So, so far, the tests that we had were primarily either RT-PCR tests, which amplify viral RNA, versus protein antigen tests, which look for proteins encoded by that RNA genome that are also specific to the coronavirus. This is an example of the latter, a protein antigen-based assay. It's not the first.

3 other antigen tests that the FDA previously gave authorization for, all of them required some component of a machine in which to run each test through.

This is potentially the most attractive form factor, price, as well as performance data.

So if the, if the RT-PCR is a gold standard, why wouldn't the rapid RT-PCR type tests be just as much of a candidate if they could get the price point down? I'm basically trying to understand, like, of all the testing options out there, why is this one so interesting and what about it specifically makes it that interesting?

Yeah, absolutely. Terrific question. So Abbott's prior test, which also generated a whole news cycle, called ID NOW, was also a 15-minute assay. The initial data from the urgent care setting for that test suggested roughly 95% sensitivity and over 98% specificity. Subsequently, as that test was deployed in community settings and nursing home settings and hospital settings, especially where sometimes symptom onset was, was delayed, the numbers fell a little bit and the sensitivity came down to closer to 80%. And in other studies and places, people are reporting even less. So that was kind of one issue. The second big difference between ID NOW and BinaxNOW is that ID NOW still requires an instrument. So you still need a machine. This test is like a pregnancy test. So it's hypothetically a lot easier to deploy and hypothetically could make testing more accessible than it is today.

Also, Abbott is describing just a more significant investment in their ability to manufacture this test at pretty extraordinary scale. So the number that they came out with is this 50 million tests per month could be done through this platform. 50 million capacity relative to 20 million being done starts to feel pretty interesting. Because this test alone could help us get to maybe triple the volume of testing that we're doing now.

Let's talk really quickly about the significance of these types of immunoassays. What is the significance of lateral flow type testing versus other types of testing?

So it's a simple-to-use, typically diagnostic device that, that we can use to confirm the presence or the absence of a target analyte, such as a protein that is virus-specific or a biomarker in your urine. It can typically be performed by either a health professional, but doesn't require a ton of training to do, or potentially even a patient. We use lateral flow assays in many industries, not just healthcare, but environmental testing, animal health testing, plant crop health. The test is basically a membrane, colored nanoparticles or labels that, that show up, that, you know, that reveal the presence or the absence of the analyte, and immobilized antibodies through which when the sample flows, antigens in the sample bind the antibodies and produce the signal of the test.

The simplest version of that is the antibodies and antigens interact and they create color along this kind of line, this lateral, literal linear array in order to show the presence or not presence of something.

That's right.

And it's just a testing modality and platform that we use very broadly already.

You know, we hear a lot about this question of, oh my God, tests are taking so long. So how does this all play into that whole discussion and debate and question?

You know, tests don't have to take that long. So even though you hear those examples of people's RT-PCR tests taking 2 weeks to come back, many of you will also have heard an example of somebody who got their test back in 2 hours. The RT-PCR reaction itself does not take that long. The issue is around the supply chain of testing. So how do you get the specimen to the testing site? How do you staff the lab with enough people to process enough tests per day. And so this sort of rapid test, it's not a property of the antigen test per se, like an RT-PCR test can also be quite fast. What's interesting about this test is that it doesn't necessarily require that staffing. It has the properties of a test that seem attractive enough to be able to have anybody run it. And if that's what's causing the 2-week test turnarounds, maybe this is a solution.

So does this mean this can be done at home? You actually mentioned the pregnancy test a couple of times, and I think that's super interesting.

So that is exactly why there has been a lot of excitement about this test is this idea that, well, maybe my flight attendant could run it. Maybe I could run it in my bathroom in the morning. I feel like I have a light fever. In fact, the test also comes with a mobile app that shows your test results and sort of a safe pass that you might be able to take. The vision is that you might be able to actually show somebody your test in a sort of uniform recognized way? And the answer is no, that's not the way this test was approved today. Today, it has to be ordered by a doctor. The sample has to be placed on the test by a healthcare professional. The test is supposed to be run only in a CLIA lab or in a CLIA-waived facility. This includes a lot. You know, it includes potentially your drive-through testing site, which is administered by your local hospital. It includes thousands of pharmacies across the country that are also CLIA-waived facilities. So it could still be a force function to democratize access to testing, but this ultimate vision that it's a pregnancy test-like COVID test is a future potential.

Got it. And so if we were to think about it in the long arc of where we are right now, it's on the road to being continually cheaper, deliverable, etc. Can we talk about what is the hype in the narrative in the sense of what isn't working? Like, this sounds very promising, but—

Yeah.

What are the questions on the table?

Yeah, so the big ones are that the test really isn't democratized today. It's not really this everywhere available to everyone test yet. And so in some ways, as you asked earlier, in that sense, it's not that different than ID NOW potentially. Maybe the performance is better, maybe it's not. We don't really know for sure yet in a large sample, but even that test is a 15-minute test. It has to be done in the CLIA lab. So practically it's not, it's not yet a huge step forward. And then the second biggest issue is that the performance is just measured in a very limited sample of 35 patients who are symptomatic, whose physician believes that they might have COVID-19. That's not the same as a test whose performance we understand in an asymptomatic population.

So these settings that seem like sort of this holy grail that I'm getting a test before I walk into a classroom, before I walk into a hospital, before I get onto a flight, that setting is typically a setting of an asymptomatic person, right?

But the data that Abbott reported, they reported this sort of positive agreement rate with gold standard PCR in 7 days after symptom onset. And then they also actually reported that the one patient in whom they missed a positive test, that patient had a lower viral titer.

What does that mean?

In a quantitative RT-PCR test, that patient had less virus. So the individual that was missed was somebody who had a lower viral titer. And so that is pointing us to the possibility— it's, you know, it's a small data point— but that this test is likely to have a potentially quite significant drop in sensitivity and performance as we move into the asymptomatic population where viral titer is even lower or not detectable. So we don't know yet about how this test performance will extend to the future where maybe I could do this test at home. One, when I do the test at home, I am probably not as sick as the patients who got this test and probably not as symptomatic as the patient who got this test. That means that the test in my Saturday morning setting may just be less sensitive.

Yes.

We talk a lot about the trade-offs between sensitivity and specificity.

Can you just quickly like explain the difference?

Yeah.

Yeah, the sensitivity is out of all the people who are actually sick or who actually have the virus, how many did I find? The specificity is out of all the people who actually don't have the virus and who are well, how many did I correctly classify in that group? In this case, Abbott reported right at the outset the data that resulted from these 35 patient tests that they did where they had side-by-side RT-PCR testing done and they reported a sensitivity of 97.1%. And a specificity of 98.5%.

So it actually sounds really good on both dimensions.

It does. And compared to the RT-PCR tests that were already available, it is good. And that's why this test is getting attention because the hypothesis had been that antigen-based tests might have lower performance, although the other ones also had similar performance. They just had, you know, less attractive form factor. But another useful metric to think about are what is the positive predictive value of the test? So if I get a test result that's positive, how often will that actually be true. I encourage everybody to check out this JAMA paper written by Arjun Manrai, Gaurav Bhatia, Zak Kohane, and Sachin Jain a couple years ago called "Medicine's Uncomfortable Relationship with Math: Calculating the PPV, Positive Predictive Value." And in there, you'll see this kind of example, and we could do the same if we take a little detour on this in the COVID context and ask, okay, we have a test. That has a 97.1% sensitivity and a 98.5% specificity, what is the chance that a positive test result is actually a COVID-positive patient? It sounds like that number should be something in the 90s. We just heard two really big numbers, right?

Yeah, exactly.

So now let's imagine that the prevalence of COVID-19 in a particular setting is actually only 2%. Okay, which is actually true in a lot of the places where COVID tests are being done. The test positivity rate is turning out to be on that order. 2 out of 100 people or 20 out of 1,000 people actually have COVID-19. So out of those 20 who are sick, this 97.1% sensitivity is telling us we're going to find 19 of them. There's only 1 we missed, and that is a good feature of the test. But in that 1,000 people, 980 will actually be COVID negative. And in that group, we will also report that 15 are COVID positive by the test.

Ah, I see.

And so actually in that 1,000 population, we're going to report 19 out of 20 who are really positive, but 15 out of 980 who are testing positive, but not really positive. So out of all the tests we do, only about half are true.

So this is a super interesting thing. And first of all, I will link to that paper in the show notes for the listeners. This is not just an academic question. And I remember this from my stats days. This is literally the trade-off between like type 1 and type 2 errors and like false positives and false negatives. Does it really matter? I mean, I hate to ask this question.

Sure.

It might be torturous mentally, psychologically to think you're COVID positive and you're not. Does that latter thing really even matter as long as you're catching the majority of COVID-positive patients?

It's a good question. I think the positive predictive value and the flip side, the negative predictive value, do matter because we're trying to take interventions based on them, right? So if in a low-prevalence environment we're running this test and 1 in 2 of the tests that result positive are actually somebody who's not positive, It might not matter in a setting where patients are already sick, they're already hospitalized, you're going to do some extra quarantining, but in the asymptomatic scenario where you might stop somebody from taking a flight, you might stop somebody from going to work, it could be a much more significant impact.

Right. And that actually has impacts thinking about even longer term on this whole question of reigniting the economy and driving people back to work in certain contexts too. And by the way, I was even thinking of the early days of the pandemic when we had shortages. The impact of actually saying these people are positive and not giving rooms or equipment to people who need it. So bottom line for me, Vanitha, how should we think about this news about Abbott's new test and more broadly?

Just the process of kind of thinking through what makes this test special makes you think through what are the properties we care about in a diagnostic, any diagnostic test in medicine. We care about ease of use. We care about speed of test result. We care about accuracy. We care about what to do if the test is wrong and how often it will be wrong. It's sort of a good framework to think in general about both diagnostic testing and in particular point-of-care diagnostic testing. And it's pretty impressive and exciting to see that R&D teams are coming out with these, you know, albeit piecemeal steps that really feel like they can change our ability to get a handle on this pandemic. But most emergency rooms and urgent care clinics are today often running side by side with whatever their COVID test of choice is. Ideally, the same swab and the same specimen is being tested at the same time for a whole panel. As we head into the winter flu season, when everybody's gonna have common cold coronavirus or influenza or rhinovirus, it's even less the case that, you know, coronavirus is frankly top of the differential.

It's not.

And so I really like the idea of a single panel or a single assay that reports out on a whole set of respiratory viruses. And, you know, it's on our healthcare systems and payment systems to some extent to figure out how to fully operationalize all of this. But I'm excited that we have sort of more tools in the toolkit to bring to the fight.

Great. Thank you so much for joining this episode, Vanitha. Listeners, as a reminder, you can find the show notes, other related pieces, and more at a6nz.com/16minutes.