Staphylococcus aureus: Here, there and everywhere

To spend any time on our infectious diseases consult service is to be knee-deep in invasive, difficult-to-treat S. aureus disease. Most S. aureus (including MRSA) disease is caused by a strain previously colonizing the host, and given that up to a third of the human population carries S. aureus it is easy to understand why the disease is so common. Exposure to healthcare is a major risk factor for invasive S. aureus disease, simply because so many healthcare interventions (surgery, device use, antibiotic exposure) provide opportunities for the organism to invade. 

Nonetheless, the conventional wisdom still holds that a large proportion of healthcare-associated S. aureus disease results from patient-to-patient transmission events—the corollary being that prevention of S. aureus disease should focus primarily on preventing transmission (including active detection and isolation). However, a careful assessment may demonstrate that interventions solely designed to interrupt transmission are responsible for only a small portion of disease reduction in observational studies (for an example, see my previous post on the VA MRSA directive). This is important, as it should focus our attention on preventing disease among those at risk for colonization (everyone), via such horizontal measures as device-associated infection prevention bundles, chlorhexidine bathing, and suppression/eradication of the carrier state during high-risk intervals. 

A study by UK investigators published in Clinical Infectious Diseases provides further evidence that patient-to-patient S. aureus transmission is a relatively uncommon event, even as an explanation for S. aureus “acquisition” events in the ICU. Using whole genome sequencing (WGS), the investigators found that only 7 of 37 ICU patients who “newly acquired” S. aureus were colonized with strains that were closely related to other patients who had an overlapping ICU stay. There are several limitations to the work, most of which are outlined by the authors in their discussion and by the excellent accompanying editorial by David and Daum. The limitation most concerning to me is the assumption that a single nares + perineum culture plated directly to solid agar media (chromogenic agar and Columbia CNA) is a sensitive method for detection of S. aureus carriers. It isn’t. Failure to perform a throat culture or to use broth enrichment probably reduced sensitivity by 30-50% (explaining their overall carriage rate of only 16.7%, when most published studies demonstrate S. aureus colonization rates of closer to 30%). We recently performed a study wherein we cultured 500 pregnant women at 5 body sites—the table below is taken from our presentation at the Decennial meeting in 2010 in Atlanta. The bottom line? Using only nares and perineal cultures directly plated to solid agar media would have missed more than half of our S. aureus carriers. Achieving 90% sensitivity (using a positive culture at any of the five body sites as the gold standard) required sampling both the nares and the throat and using overnight broth enrichment.

So what about all those “acquirers” carrying isolates that didn’t match other ICU patients? In addition to implicating other potential reservoirs (personnel, visitors, etc.), I’d wager that some were prior carriers who were newly detected due to sampling issues, increase in CFU associated with healthcare exposure, etc. Intermittent detection of S. aureus carriage is well described, and would be magnified using the microbiological techniques in this study.

Finally, while I agree with the authors that WGS is the new gold standard for assessing genetic relatedness, the use of spa typing as the “conventional method” comparator is too easy. As we found in our recently published study, roughly half of all MRSA isolated from epidemiologically unrelated clinical infections in 43 US hospitals were from a single spa type! (How’s that for discriminatory power?) If you’re going to write off the conventional methods, at least use a method with better discrimination, such as PFGE!

Staphylococcus aureus continues to evolve: MRSA without mec edition

I just returned from a wonderful MRSA conference this past week where I enjoyed learning about how European countries are handling antimicrobial use in livestock and the emergence of potential human pathogens like livestock-associated MRSA. Perhaps I will post further thoughts on the conference later in the week.  However, I just saw this new report in JAC on a novel resistance mechanism in S. aureus, so first things first.

It is known that hyper beta-lactamase producing strains of S. aureus exist and that some strains have chromosomal mutations and resultant modified penicillin binding proteins. These strains can have phenotypic methicillin resistance while lacking the mecA (or mec C) gene. Thus, they can evade detection by genotypic methods and the underlying mechanisms can be missed by phenotypic methods, as Dan nicely described here.

There is now a report by Xiaoliang Ba and colleagues that describes four clinical S. aureus isolates that were found to be MRSA but lacked mec A or mec C. The strains were from clinical wound infections and deposited in the Scottish MRSA Reference Lab. They belonged to sequence types 1, 15 (two isolates) and 8. Three strains were resistant to oxacillin, cefoxitin and PCN by ETest and disc diffusion and the fourth isolate was at the oxacillin Etest breakpoint, while three were beta-lactamase producers. Whole genome sequencing confirmed that none contained a mec-like sequence and that beta-lactamase production was not mediating the resistance. Interestingly they found similar single amino acid substitutions across sequence types in PBP1, PBP2 and PBP3 suggesting independent evolution of the same trait (homoplasy).

The authors acknowledge that their targeted search of the genomes could have missed other possible mechanisms of resistance in these isolates. I would add that this was in four isolates out of an unreported denominator, so we don't know the magnitude of the clinical impact yet. What we can say is that our surveillance techniques had better keep up with S. aureus. Unfortunately, it seems to be just as good at evolving away from our prevention methods as it does our antibiotics.

Obligatory S. aureus image courtesy of wikipedia

Addendum: For further insight, scroll down to read Dan's comment on this study

Everything old is new again, with WGS!

As we’ve pointed out, whole genome sequencing (WGS) is the hottest new tool to help us decipher the epidemiology of healthcare-associated pathogens. Last week’s NEJM included a study using WGS to investigate the molecular epidemiology of C. difficile disease (CDD) in Oxfordshire, UK. In a 3.6 year study that included 1223 CDD patient isolates, the investigators found that only 333 were genetically related to at least one previously obtained isolate. Of those 333, only 126 (38%) had nosocomial exposure to the earlier patient. And the finding receiving the most attention: 45% of strains isolated were genetically distinct from all previous isolates.

The take home point? In current hospital settings (where we isolate every known CDD patient and use enhanced environmental measures to try to eradicate their C. difficile spores), symptomatic CDD cases are no longer the major reservoir for C. difficile acquisition. Focusing only on transmission prevention, then, will have a limited impact (antimicrobial stewardship, anyone?). Most obviously, further work is clearly needed to identify other sources of exposure and acquisition of C. difficile.

This may come as news to many, but probably not to Matt Samore, who made a similar observation….in 1994.

Riding the epidemic curve to glory, WGS edition

One reassuring lesson all healthcare epidemiologists learn is that every outbreak will, eventually, come to an end. The trick is to prevent outbreaks in the first place, or to recognize them early enough to intervene effectively. Many outbreaks, though, are recognized as they peak and are entering the “downhill” part of the epidemic curve. Any interventions prescribed by erstwhile epidemiologists are then, in retrospect, credited with helping to contain the outbreak (even if said interventions were completely idiotic). This phenomenon was described by Dr. Alexander Langmuir, the father of the CDC’s Epidemic Intelligence Service, as “riding to glory on the downhill slope of the epidemic curve”.

Now we can add the performance of whole-genome sequencing (WGS) to the list of activities that epidemiologists can “ride to glory” as the key to outbreak control. As WGS becomes faster and more affordable, reports have been published in NEJM, Science Translational Medicine and Lancet Infectious Diseases suggesting that WGS can be the key to real-time or “actionable” information to help contain outbreaks (due to MRSA, KPC-producing K. pneumoniae, and MRSA, respectively). As we’ve pointed out previously, though, it isn’t at all clear that WGS was important to real-time outbreak management, or that WGS is ready for prime-time and coming soon to a hospital near you.

Why? For any technology to see widespread adoption in clinical diagnostic laboratories, there must be sufficient automation (including of the analysis and interpretation of the massive amounts of WGS data), and it must provide a substantial advantage over existing testing approaches.

Which brings us to the more important question: given our crude approaches to outbreak control, how does WGS provide any immediate advantage over other same-day typing methods? Does the added discrimination really make a difference in whether we decide to isolate or cohort patient X, decolonize healthcare worker Y, or close unit Z to new admissions? These questions are particularly pertinent when we don’t yet understand the “within host” variation in genotype and how “within” versus “between” host variation can be applied to determine direction of transmission (or even the fact of transmission). Eli recently sent me a link to this interesting discussion of the Lancet ID report, which addresses some of these issues. 

The bottom line is that we have a relatively few crude tools for outbreak response: enhanced basic practices (e.g. hand hygiene, environmental disinfection), active surveillance, isolation, cohorting (of patients and/or HCWs), decolonization of carriers (both HCWs and patients), removal (temporarily or permanently) of HCWs implicated in transmission, closure of units, mitigation of any identified common sources, etc. The level of discrimination provided by WGS is a great research tool, and will undoubtedly help in retrospectively piecing together the most likely outbreak scenarios--but for now I’m not convinced it has much advantage over other typing methods for guiding real-time outbreak investigation and management.

Nothing to see here, please move along

It's been a quiet week out here on the edge of the blogging prairie. Some of us are recertifying, some are in the middle of huge grant deadlines and some are chairing a giant meeting planning committee with the meeting imminent. But, we still think about you every second and we miss providing you with up-to-date infection prevention information...

In the interim, we have created a little poll off to the right, which you can use to let us know how you like to read or follow the blog.  Vote early and often.

From the nothing to see here column: We've heard new reports that the NIH KPC outbreak that was halted by whole-genome sequencing is back. On September 7th there was a new case of the KPC strain, the first since January and 19th overall. The blood stream infection resulted in the unfortunate death of boy from Minnesota, the seventh fatality attributable to the strain.

NOW SEE THIS: Registration for ScienceOnline2013 is now officially open. It's the seventh annual un-conference exploring science on the Web and takes place Jan. 30-Feb. 2, 2013, in Raleigh, NC. Registration for the first round of 100 slots is closed for today, but there are two more opportunities to register: Thursday, Sept 20, 2012 at 2:00 PM (EDT) and Friday, Sept 21, 2012 at 11:00 PM (EDT). By rumor I heard that these sessions last only minutes, so log on near those start times and keep refreshing your browser. All seems pretty exciting.

Whole genome sequencing and infection prevention

While attending a HICPAC meeting this week, I had a chance to chat with David Henderson about the use of whole genome sequencing (WGS) to help investigate an Acinetobacter outbreak at the NIH Clinical Center. I’m sure you’ll hear more about this at IDWeek (yet another reason to attend!). In the meantime you can check out the report in this week’s NEJM describing use of rapid WGS to investigate a NICU MRSA outbreak. My Journal Watch summary is here, if you want an abbreviated version. I’m a little surprised the paper made it into the NEJM, as it isn’t clear how the WGS contributed much to control of this particular outbreak. The work is most useful as an example of how WGS is now rapid and inexpensive enough to contribute to patient care and infection prevention work in real time (rather than only in retrospect, or as a research tool).

However, there are a couple reasons that you won’t see WGS coming to many labs in the near future. First, analysis and interpretation of the enormous amount of resulting data (the entire bacterial genome sequence) is not simple, requiring substantial local expertise. Second, we have a very incomplete understanding of which genetic differences and mutations are clinically important. Until we expand our knowledge base and bring some automation to the analysis and interpretation, bacterial WGS to inform outbreak and transmission investigations will remain limited to a few large centers that have in-house capacity.