What change looks like

The practice of medicine changes slowly, at least more slowly than science changes our understanding of disease.  There are some good reasons for this, and some bad.  The good reasons are that science does not always proceed smoothly from lesser to greater understanding.  Partial understanding of complex processes - and few are more complex than health and disease - can lead to actions and recommendations that are more harmful than helpful.  One does not have to look far for examples: the classification of homosexuality as a disease, the use of general anesthesia during childbirth (my mother had to physically fight this off), and any number of misguided dietary recommendations.  In many cases, doctors and their patients are well-served by a conservative and skeptical attitude toward modernization of medical practices.  If a change is not clearly going to lead to better outcomes, then perhaps it is best to wait before adopting it.  Scientists are free to proclaim the beauty and elegance of their new findings without having to deal with any messy unexpected consequences.  Doctors have no such luxury.

Bad reasons to resist change can be described in many ways, but mostly comes down to ego and money.  No one like to be told that they way they have been doing things is wrong.  No one likes it, and some like it so little that they refuse to see the obvious.  This has been the state of play in antibiotic prescribing practices for the last decade or so.  Doctors, for the most part, have not treated patients; they have not treated specific pathogens; instead they have treated signs and symptoms.  Urinary tract infection? Prescribe ciprofloxacin - even though 30% of E coli UTIs are resistant to fluoroquinolines.  Child with a fever? Time for Augmentin.  Respiratory tract infection?  A Z-pack should take care of that.  The names of the antibiotics have changed over the decades, but the approach has not: a certain set of symptoms triggers a corresponding prescription.  No microbiological data are involved.

This system was unavoidable when it took 2-3 days to get a microbiology result.  No one, doctor or patient, should wait that long for a treatment that is well-tolerated and highly likely to be effective.  But rapid tests for microbiology have been available for close to a decade now.  These tests certainly have their limitations, and they are more pricey than streaking bugs on a Petri dish and waiting.  However, most doctors are very comfortable with the practice of evidence-free (empiric) prescription, and are not demanding data from rapid tests.  If the doctors are not demanding rapid test results, the clinical microbiology lab is very unlikely to expend the time and money needed to provide them.  The weak sales of existing rapid tests starves their developers of the capital needed to create better tests, creating a death spiral for their companies.

There are signs that this invidious dynamic may be beginning to change.  Cohen et al recently published “A multifaceted ‘omics’ approach for addressing the challenge of antimicrobial resistance”, describing some of the changes that new technology is poised to make in the treatment of infectious disease.  They include a wonderful infographic that succinctly summarizes the causes, challenges and costs of resistance:

Some of their predictions and recommendations for the use of “omics” technology are overkill: phenotypic testing of susceptibility and resistance will always be more reliable than molecular methods such as proteomic profiling.  But that’s fine.  These issues will work them selves out in due time.

The important point of the paper is not any specific recommendation or prediction.  The important part is its emphasis on data-driven treatment of infectious diseases, as part of the “TAILORED-treatment” research program.  Ultimately it doesn’t matter whether individualized treatment data comes from PCR, or mass-spec, or accelerated culture methods.   All of these data can be used to improve public health by enabling the prudent use of antibiotics, and to improve patient health by providing the most effective treatment.

These data are sometimes available now, and will become more and more available as the technology gets better.  The question is, how hard will it be to get doctors to use the data?  Let’s hope that this generation of doctors is more open to positive change than their antecedents.

Rapid diagnostics - ID is not enough

In a previous post I discussed some of the hurdles - mainly economic - to developing rapid diagnostics for bacterial infectious disease. Despite these problems, several new technologies have been developed and are finding their way, however slowly, into clinical microbiology labs.

The problem with all of them is that they are great at identifying bacteria, but are limited in their abilities to determine resistance and susceptibility. In a low-resistance environment, this is not a significant limitation. Historically, doctors would consult a reference that matches clinical presentation and bacterial species with preferred antibiotics, and prescribe accordingly. They might also consult their hospitals' antibiogram, a tally of observed antibiotic susceptibilities at their hospital.

But what do you do when the antibiogram looks like the one below (which is very typical)?

 

Your rapid test may have identified the infecting bug as Acinetobacter, but then what? None of the antibiotics on the list are highly likely to be effective for any particular strain, but there is a pretty good chance that one or two might be. You just don't know without a susceptibility report on the bug that has been isolated from the patient - and that still takes 2-3 days. The rapid ID test has told you what the patient is infected with, but not what treatments are likely to work.

This limitation is true whether the rapid-test technology is PCR, PNA-FISH, or the new favorite of many, mass spectrometry. It's true that PCR tests can detect the presence of some resistance genes. But the prevalence of these genes (e.g., mecA in S. aureus, vanA in E. faecium) is so high that doctors usually assume resistance and treat accordingly regardless of the test result. A negative result for a resistance gene does not necessarily mean that the strain is susceptible, as there can be new variants and new pathways for resistance that gene tests may not detect. Accordingly, no gene tests have been cleared by the FDA for determination of antibiotic susceptibility. The distinction between negative and susceptible is not always obvious to doctors, and the FDA has required the Cepheid MRSA PCR test to carry a label to that effect.

The basic problem is that antibiotic susceptibility and resistance are complex phenotypes that can't be reduced to a problem of detecting a few molecules or gene sequences. Resistance can occur through thickening or modification of the cell wall, chemical modification or mutation of ribosomes, or production of degradative enzymes and efflux pumps. Each of these mechanisms can have many variants and pathways, and new ones are always evolving. Even if (actually, when) it becomes possible to get an entire genome sequenced within a few hours, it will not be possible to read out a susceptible phenotype with high confidence. There are just too many possible genetic variants that would have to be validated first.

So for now and the indefinite future, ID-only rapid tests will continue to have limited clinical impact, and susceptibility testing will have to proceed by traditional methods: expose a bacterial strain to the antibiotics of interest, and observe the response. This process typically takes 2-3 days at present. Some methods to reduce this time are being developed, and I will write about them in a later post.

 

Posted with Blogsy

WGS in the micro lab

Nucleic acid technologies are making steady inroads into the clinical microbiology lab as they get faster, better and cheaper. The logical endpoint of this development is the use of microbial genome sequencing as a diagnostic method. Given the pace of improvement in sequencing technology, the feasibility of routine sequencing of clinical isolates is inevitable. The question is whether this technology will add any value.

Writing in the Journal of Antimicrobial Chemotherapy, Torok and Peacock say the answer is yes: "...we believe that rapid whole bacterial genome sequencing has the potential to transform diagnostic clinical and public health microbiology in the not too distant future."

I am more skeptical and think the value of whole genome sequencing will be limited:

ID/Speciation: Genome sequencing is overkill. All the DNA information needed to speciate an isolate is contained in its ribosomal RNA genes. PCR and oligonucleotide hybridization methods return gold standard speciation results now. Additional sequence information will just be noise.

Epidemiology: Probably the best use of WGS information, as the additional sequence data allow the spread and evolution of strains to be tracked. This has already happened in the case of the KPC outbreak at NIH, and will become more common.

Resistance and susceptibility testing: A seductive but terrible idea. The response of bacteria to antibiotics is a complex phenotype, involving many genes. Reliably predicting this response would require a thorough understanding of the actions of all these genes. Even more difficult, it would require us to be able to predict the effect of various mutations in these genes on their activity. For example, a point mutation that changes the amino acid sequence of a metallo beta lactamase could make it more active, less active, or have no effect at all. There are billions of possible mutations in hundreds of genes that would have to be accounted for in order to reliably predict antibiotic response. This is not going to happen soon, or probably ever. More to the point, what actionable information would WGS provide that phenotypic susceptibility testing does not?

Don't get me wrong, I love this technology. I wish it had been around when I was in grad school - synthesizing oligos manually and then sequencing them by the Maxam-Gilbert method was some of the most tedious lab work I've ever done. But the application of WGS to clinical microbiology is likely to be much more hype than substance. Even the epidemiological applications could potentially be done just as well by much more old-fashioned methods such as phage typing (if anyone still knew how to do this). Just because it's new doesn't mean it's better.

 

Posted with Blogsy