72 hours to kill: Why can't we diagnose bacterial infections faster?
By Erich Scheller
Erich Scheller, PhD, is the Director of the TechAtlas division of RA Capital Management.
June 10, 2022
This article was inspired by conversations with my friend and colleague Parker Cassidy, a partner at Sands Capital Management. Sands and RA Capital are investors in both Selux and Qvella, two companies mentioned below.
*
We’re in year three of a global viral pandemic. And every time the world breathes a collective sigh of relief that science and public health measures have finally wrangled COVID-19 into submission, it throws us a new variant curveball. We’re practically running out of Greek letters to describe them. Meanwhile, in an epic example of failing to read the room, a growing monkeypox outbreak is grabbing headlines. All of which is to say, it’s easy to forget that viruses aren’t the only infectious game in town, and that bacteria continue to wreak havoc on humanity.
In 2019, antimicrobial-resistant (AMR) bacteria were the known cause of 1.27 million deaths, with another estimated unreported two to three million AMR deaths worldwide.
Let’s put that into perspective. The World Health Organization estimates that three million people died from COVID-19 in 2020. Influenza, on average, kills at least ~500,000 per year. So if you give bacteria conservative credit for causing three million worldwide deaths in 2019, that means bacteria kill at a rate that’s at least five times that of the flu, and on par with pre-vaccination COVID. This is in a world where we have over a hundred different antibiotics to combat bacterial infections, and before we start accounting for deteriorating quality of life among people who recover from these bugs. Bacteria have killed like this forever, even in the face of intense mitigation efforts over the last century (penicillin was first discovered in the 1920s).
There are substantial headwinds in making progress against this problem. The first is that bacteria, like viruses, evolve to avoid selective pressure against things trying to kill them. So in the same way that the flu evolves in response to the most recent vaccine, so too can bacteria evolve to survive antibiotics. To slow this kind of evolution, hospitals and healthcare providers attempt to practice “antimicrobial stewardship,” which, probably put too simply, is reserving the use of our best antibiotic weapons for when we truly need them, and only using antibiotics when necessary. If you’ve ever asked for an antibiotic when you had the flu, you were hopefully told “no,” because 1) they don’t work against viruses and 2) antibiotic overuse can lead to antibiotic resistance developing in the bacteria that live in and on you. (In case you weren’t aware of your status as a bacterial landlord: they outnumber your human cells 10:1 and make up about 1-3% of your total body mass!)
The second is that there is little incentive to develop new antibiotics. This is a complicated topic that’s been covered well elsewhere, but consider that the antimicrobial stewardship described above means these drugs should only sparingly be used and usually go generic well before they ever see the kind of use that would make them blockbusters.
The third barrier, and the focus of the rest of this piece, is that our ability to test for bacterial infections and identify which antibiotics to use is woefully slow, outdated, and relies on highly manual processes.
Hurry up and wait
If you are unlucky enough to have sepsis, a life-threatening disease caused by bacteria, fungi, or viruses, it might take, at a minimum, 72 hours to get you on an antibiotic that a doctor definitively knows will work. In the meantime, doctors will treat you with broad-spectrum antibiotics based on your doctor’s best guess as to what might work. However, population-level resistance is starting to develop to these drugs, they tend to be rather harsh on the body, and they can lead to other issues (like C. diff infections). There are plenty of reasons to want to get patients on the right drug as soon as possible.
Three or more days to find out the best answer is a long time to wait. But current medical practice requires physicians to identify the bacteria in question (Staph? E. coli?) and perform a phenotypic susceptibility test, which is essentially saying “The bacteria in this person’s bloodstream died in the presence of these antibiotics and we saw it happen!” With this information in hand, a doctor can then use the specific narrow-spectrum antibiotic that they are fairly confident will work.
But still, you might ask, why does that take three days? I can know if I have COVID in about 15 minutes! Two reasons: working with bacteria is tricky (they grow at different rates, it’s easy to contaminate samples with the bacteria on your skin, sometimes they can look like they are growing when they are in fact dying) and you need a lot of them to draw conclusions (between about 100 million and a billion). A 10 mL blood draw might only contain a handful of bacteria (fewer than 100), so you need to grow more. Those first 24 hours, if not longer, are spent growing a bunch of bacteria in blood culture bottles. This is also a screening step, since only about 10% of patients with suspected bacterial sepsis actually have it, and this helps weed out the patients that might have fungal or viral sepsis (or something else entirely).
Let’s say the blood culture comes back positive. Now what? The lab will run a Gram stain, which informs doctors whether the bacteria are gram positive or negative. Knowing that can help inform broad spectrum antibiotic use – if the bacteria are gram negative, you’d stop using a broad-spectrum antibiotic meant to kill gram positives. Labs can also run a molecular panel, like bioMérieux’s BioFire, which can tell you if the bacteria have specific resistance genes (will they be resistant to vancomycin, or not?) and further inform what drugs not to use. We are also starting to see some approaches for bacterial identification at this step, though they are far from widely accepted. There is even one FDA-approved product, Accelerate Diagnostics’ Pheno, that can determine susceptibility at this step, but it’s yet to see much commercial traction since launching a few years back. (More on Accelerate later.)
But in standard practice, after 24 hours of blood culture and initial lab tests, your doctor knows 1) whether this is likely a bacterial or fungal infection (and sometimes fungi can take even longer) and 2) whether the bacteria are gram positive or negative (ruling out some broad spectrum antibiotics). The next step is subculturing, which is taking bacteria from your blood culture bottle and growing them on agar plates. You might have done this yourself in high school or college biology class, and this continues to be part of the SOP for identification and antibiotic susceptibility testing (ID/AST). Again, this step can take upwards of 24 hours and is often performed overnight.
Why do we have to grow the bacteria AGAIN, you might ask? There are a few reasons. For example, we might need to make sure that the blood culture isn’t contaminated or doesn’t contain two or more species of different bacteria. Another reason is that there really just aren’t any other great ways to recover viable, happy bacteria from blood culture for ID/AST. There are some new technologies on the horizon, and Accelerate has a few drug-bug combinations approved for their Pheno, but generally, bacteria cultured from blood samples are unhappy and require another night to grow.
Now we’re 48 hours in and have a good amount of bacterial growth on our plates. We can identify them pretty quickly - in just a few hours using MALDI-TOF mass spectrometry. But we still need to perform a susceptibility test, and if you’ve been following along so far you know that in order to do that, we indeed need to grow them for another 24 hours in the presence of a variety of antibiotics.
The old-school way to do this (which is still used quite frequently) is called a disk diffusion assay. Basically, you streak another agar plate with bacteria and place little disks soaked in antibiotics on top. You can then observe bacterial growth and determine resistance. The newer method for this (and by “newer” I mean technology from the 1980s developed by NASA and commercialized by Vitek, which was also acquired by bioMérieux), is a large, automated platform that lets labs run multiple samples at once on cards or cartridges that contain antibiotics. But platforms like this, while automated, remain limited because 1) you still need to wait long enough for the bacteria to grow and 2) the cards only have so many spots available for antibiotics, so labs either will run multiple cartridges or run disk diffusion assays in parallel.
Little practical innovation
Exhausted yet? Me too. And so are the lab techs that need to run this highly manual, time-intensive process in order to match the right antibiotic to the right patient - lab techs that are also busy running COVID tests, strep tests, and influenza tests. And for all these reasons, the ID/AST market hasn’t seen substantial innovation in decades. It relies on antiquated approaches despite the fact that every hour matters to a patient with sepsis.
Why has so little changed in the last few decades? It’s difficult to say definitively, but there are a few factors that I believe contribute.
The first is that, as we said before, bacteria are tricky. There are many different kinds of them with many different idiosyncrasies, and so far any technology that tries to compensate for that is some combination of 1) too expensive, 2) not comprehensive, and 3) disruptive to lab workflow.
For example, one bacterial trick is that early on in exposure to certain antibiotics, some bacteria can look like they are growing, even though they’ll eventually all die if you just wait long enough. So one solution that Accelerate tried in order to capture this phenomenon in action (called filamentous growth) was to train a high-powered microscope to recognize it. However, this solution had its own shortcomings. It cost $100-200, compared to $5-10 for a Vitek card. And Accelerate did not earn approval for all bacteria, or all antibiotics. Plus, the new microscope didn’t simplify workflow by replacing existing machines, but required an additional machine that needed to be added to the lab. I’d argue the end result was an impressive engineering solution that failed to take off because its cost, lack of coverage, and clunky workflow requirements outweighed the few times it generated an actionable decision ahead of standard AST.
Investors also haven’t seen much in the way of attractive returns in the antibiotic technology space. Companies like Accelerate and T2, which attempted to identify bacteria and fungi from that initial blood draw, never found commercial traction due to their shortcomings. The one success in the field, BioFire, which was acquired by bioMérieux for $450M in the early 2010s, is looking like a major coup for bioMérieux, as bioMérieux has seen consistent $1B revenues in its molecular biology division (which includes BioFire). BioFire (and those that came after it, like GenMark’s ePlex) have seen success because they can provide molecular identification and information on three resistance genes in little over an hour across a variety of sample types, including blood. While BioFire cannot tell you which drug to use, it can inform which not to use. So imagine the value of a platform that can definitely tell you what you should use!
Better, faster solutions
Luckily there is hope on the horizon as newer technologies get closer to approval. Within the realm of what’s publicly known, Specific Diagnostics (SpecificDX), recently acquired by (you guessed it) bioMérieux, is looking to one-up Accelerate Pheno with a potentially more comprehensive panel to perform AST from a positive blood culture bottle. The jury is still out on coverage as we await news on FDA approval, price, and impact to workflow, but it looks to be a step towards innovation.
Another company, Selux Diagnostics (which, for full disclosure, is in RA’s portfolio), is aiming to replace legacy machines (such as Vitek) and might revolutionize the entire microbiology lab workflow. Their initial commercial launch (pending FDA approval) promises to run samples from culture plates with results in about five hours, versus needing to run samples overnight on the legacy machines. And if Selux can achieve its most expansive vision, the company could deliver the first truly disruptive impact on lab workflow by replacing existing legacy machines and allowing labs to run rapid AST from positive blood culture bottles as well as traditional agar plates – enabling all work to be done on one machine (rather than many). And Qvella (another RA portfolio company) has recently launched a product that cuts one day from the standard workflow by harvesting bacteria from positive culture bottles. These and other efforts, including Accelerate’s new offering (ARC), are making vital progress on understanding how to keep bacteria happier when grown from positive blood, which should also shorten timelines.
For the first time in what feels like a very long time, there is plenty of promise on the horizon. Here’s where I think our focus should be in the near future:
Incentivizing new antibiotics. Because there is little financial incentive to develop novel antibiotics, government contracting and more incentive programs like BARDA’s CARB-X can ensure that we continue adding to our armamentarium while rewarding the innovators that develop these drugs.
Redoubling antibiotic stewardship efforts to reduce emerging resistance to the drugs we have. This includes better monitoring and communication between hospitals across the country and the world, as well as utilizing older antibiotics that have fallen out of conventional use (which Selux’s platform, for example, would enable).
Further innovating to improve the speed and accuracy of diagnostics. Getting answers (both ID and phenotypic AST) directly from a patient’s blood draw is the holy grail, but there are also opportunities to innovate further up the workstream, such as reducing the time to culture a positive blood bottle.
If we can make real progress here, today’s 72-hour lag may become tomorrow’s lunch break.
Please click here for important RA Capital disclosures.
Related reading
