How low-cost is low-cost?

Ok, there are times when marketing hype becomes flat out ridiculous. Last month, Cepheid announced a new low-cost PCR based test for Tuberculosis. Several challenges with performing PCR on mycobacteria aside (its actually extremely difficult to get DNA out of the buggers) this would be a great breakthrough. PCR has the potential to be an extremely specific, extremely sensitive diagnostic test for… well, pretty much any infectious disease. Its not without its problems, however.  I had a chance to hear Dr. Jason Stout (director for TB control in NC) talk about some of the issues they’ve been having with their PCR. These problems have been enough of a showstopper than they still rely primarily on liquid culture which beats their PCR solution in sensitivity, specificity and cost.

Still, a low-cost PCR test would be a big deal, the problem is that Cepheid’s test is most definitely not a low-cost test. In fact, the cheapest PCR readers that Cepheid sells will cost you about $27,000 a unit. You could buy a flow cytometer for that much and you’d save a lot more lives using that kind of cash by doing CD4 counts than TB diagnostics when there are simpler, cheaper solutions for TB already out there.

Interestingly enough, this was developed in collaboration with FIND, a Gates-funded non-profit working on diagnostic technologies for developing countries. This really casts a lot of doubt in my mind about FIND’s priorities when it comes to developing world diagnostics. FIND’s product pipeline for TB diagnostics limits itself to three approaches: squeezing incremental improvements out of microscopy, making better and cheaper PCR/NAAT based tests and developing antibody-based rapid tests.

I think there’s a huge gap between the first category and the second two in terms of their “appropriateness” for the countries they’re trying to help. Microscopy is dirt cheap, its a relatively poor diagnostic test but it scales very easily. Incremental improvements will help, but really, as a diagnostic test its in bad need of an overhaul. PCR/NAATs and antibody tests suffer from a lot of draw-backs, mostly centered around their costly, proprietary, short-lived and temperature sensitive reagents (I shudder to think of how long PCR probes would last at 40C).

As second-line tests, these approaches may bear significant fruit, but the lack of effort in developing the $1 screening test that’s badly needed to replace the aging microscope is extremely troubling. Without an easy-to-use, scalable and cost-effective first-line diagnostic, we will never be able to bring the TB epidemic under control. No amount of amazing species differentiating, drug-resistance detecting, super sensitive diagnostic tests will be able to make a significant, sustainable impact in these markets if their prices can’t drop an order of magnitude or more over the next few years.

Testing the test: Tuberculosis Diagnostics

Source

Naman pointed me to a really interesting study out of Hopkins on evaluating the cost-effectiveness of potential new TB diagnostics tests. Since I’m currently developing one of said potential new TB diagnostic tests, this was of particular interest.

Global Health is really an engineering field: its all about optimization. With a limited set of resources and finite demand and a limited set of tools available to us, what is the most efficient way to spend our resources to maximize impact? Granted this depends on your definition of impact, but thankfully the DALY is a pretty well accepted metric when it comes to infectious diseases.

In the case of TB, we have a fairly narrow assortment of tests: sputum smear microscopy, mycobacterial culture and PCR. Some folks (I’m looking at you, FIND!) are working on antibody based rapid tests, but these aren’t due out for 2-3 years, optimistically speaking. When you add latent TB to the mix, a couple other techniques come into play including products based on ELISA and ELISPOT, but we’ll leave out latent TB for now.

The three cornerstones of TB diagnostics (yes, yes, I’m leaving out chest X-rays, but that’s screening not diagnostics. Any… stuff… in your lungs will show up right away, but without any of these other techniques, you’d never know if it were TB or nearly any other pulmonary disease) each have their own challenges:

Microscopy is, well, microscopy. Its manual, time-consuming and error-prone. There’s some push to phase out bright field microscopy for fluorescence, but the improvement is relatively incremental. You’re still fundamentally limited by the human operator and how much time he or she is willing to put into scanning a smear at 1,000x (bright field) or 450x (fluorescence) magnification. The up-side? Its the cheapest method we have at around $1.45 a test and even though its pretty hard to become a master microscopist, it doesn’t take much to become proficient.

PCR is God’s Gift to the Geneticists. Its also an amazingly powerful diagnostic tool. The problem? Its way too expensive- if the developing world were to perform as many PCR tests as they do sputum smears it would cost over $4 Billion at $50 a pop… 4x current global TB control expenditures. Infrastructure is also a bit of a problem since reagents are typically temperature sensitive and the assay’s need pretty clean water (heaven forbid any DNase gets in there!).

The basic premise behind culture is fairly straightforward: if you don’t have enough bugs in the sputum to see easily, give em some food and let them grow till there are enough! Unfortunately, M. TB happens to be one of the slowest growing microorganisms out there with a life cycle of about 15-20 hours so it can take up to 6 weeks for enough to grow out… yikes. The good thing is that TB doesn’t kill you very fast and while culture takes a pretty good amount of infrastructure to support, its a lot less than PCR and each test weighs in at around $3-4 per test. On the other hand, every day that a patient walks around un-diagnosed could mean dozens of future cases, not the ideal situation.

So what are Hopkins’ thoughts on this? Basically, if you were to try to introduce a high specificity (95%) test with a fairly high per-test cost ($20) that is somewhat more sensitive than microscopy (70%) its not going to be very cost-effective. In fact, they found that specificity coupled with price was the single most important factor in a new diagnostic test being able to beat the cost-effectiveness of microscopy. Find the people that are very sick very quickly, get them out of the public and on treatment before they spread more and you’re going to get the most bang for your diagnostic buck. Good advice that FIND, the WHO and the Gates Foundation need to hear more of: PCR and antibody tests may be extremely impressive, but until they can become drastically cheaper they’re still likely to be beat out by a cool blue LED and a $20 camera sensor.

How do you use a tool you don’t know you can trust? Issues with diagnosing latent TB

I attended an excellent talk a while back by Dr. Jason Stout, the TB medical director for the state of NC. The talk was entitled TB Diagnostics: New Tools, New Dilemmas, and dealt primarily with the challenges we see in attempting to diagnose and treat latent mycobacterial infections. While active TB infections are still a primary health concern for much of the world (roughly 90% of active TB cases occur in the developing world), the biggest challenge facing the developed world in TB control is going to be the risk posed by latent infections.

The latent infection itself is not directly a health concern- it is the risk of conversion, which occurs in an estimated 5% of patients, from a latent to an active infection that worries most health professionals in the developed world. With an estimated prevalence of about 10-15 million individuals in the US alone, this represents over 500,000 potential future active cases even without direct contact or transmission. These numbers are pretty rough estimates though, when it comes to diagnosing latent TB we’re pretty much flying blind. Almost everyone is familiar with the Tuberculin Skin Test (TST), the standard for diagnosing latent TB. By injecting a cocktail of proteins from sterilized cultures just under the skin, we expect to see a patient that has been exposed to TB to mount an immune response and cause swelling at the site of the injection. Given the complexity of the human immune system and how little we really know about it, it isn’t surprising that the test isn’t very reliable. Everything from the BCG vaccine to exposure to non-TB mycobacteria can throw off the results of the test.

Unfortunately, most modern latent tests haven’t improved significantly over the TST like how PCR is revolutionizing active TB diagnosis in the US and other developed countries. Some technologies based on using ELISA to monitor production of the cytokine interferon gamma (IFN-γ) by immune cells in human blood have been brought to market and approved for use. The problem is an extremely complex one, however, and breakthroughs are likely to hinge on a better quantitative understanding of the immune system itself. According to Dr. Stout, while these tests do seem to fare better in patients that have had a BCG vaccination, the root problem still exists.

We don’t know how many people have latent TB. We don’t know how good our tests are because we have no way of knowing how many patients our tests miss. Our tests rely on a somewhat arbitrary cut-off that patients often fluctuate around. In the end, developing a sane policy for dealing with the danger that latent TB poses is going to depend heavily on developing innovative and effective new diagnostics for the infection.

Shadows and daggers

Sources:
Ozcan A., Demirci, U. Ultra wide-field lens-free monitoring of cells on-chip
Lab on a Chip, 2008, 8, 98 – 106.
Seo S., Su T., Erlinger A., Ozcan A. Multi-color LUCAS : Lensfree On-chip Cytometry Using Tunable Monochromatic Illumination and Digital Noise Reduction
Cellular and Molecular Bioengineering, 2008, Vol. 1, #2-3, 146-156.

There’s been a lot of coverage of a technology out of UCLA called the Lensless Ultra-wide-field Cell monitoring Array based on Shadow imaging (LUCAS) in the media. Wired’s coverage say’s the UCLA group has “modded a cellphone into a portable blood tester capable of monitoring HIV, malaria, leukemia and detecting diseases.” With any technology, especially medical technologies, it is important to understand really what has been achieved rather than the (often misguided) media coverage. LUCAS is essentially a microfluidic chip built directly onto an image sensor (the marketing material shows a cell-phone camera, but the papers describe a higher end CCD array) that images the shadows of cells passing over it.

The first paper describing LUCAS intended it to be a method of counting a homogeneous population of cells with the idea being that some sort of affinity column would be used to sort, say, CD4+ T-cells from blood for HIV/AIDS monitoring. There’s a number of potential problems with chromatographic separation of CD4 cells, although the concept does sound promising. The key point here is that there are a number of simpler, cheaper, and just as effective for counting cells in a homogeneous solution. A Coulter counter, perhaps? A pair of electrodes is a great deal more robust and cost-effective than a CCD array and a simple comparator is a great deal easier to implement than a live-video blob tracking algorithm.

The second key paper, perhaps following the realization that using a CCD array and an image processing algorithm to count particles might have been a bit overkill, sought perhaps to overcome this limitation by looking at diffraction patterns from various particles. While interesting at first, there’s still some glaring issues with this not the least of which is that you’re only going to see morphological differences between cells. In the case of CD4 T-cells where the only difference between your cell of interest and other T-cells is a single protein, you can see why that might be a problem. It seems Ozcan’s group noticed it was a problem too since their proof-of-concept was that they could differentiate between yeast cells, red blood cells, and polymer microbeads.

I don’t mean to disparage the engineering that’s been done here; the technology is nifty and I’ll be looking forward to seeing some cool applications. I just wish the UCLA press folks and other science journalists would do some more homework on what it actually takes to monitor diseases like HIV/AIDS and malaria before proclaiming every single piece of hardware designed for a biological application to be a magical tool that will diagnose every exotic disease we’re afraid of.