Wanted: Ebola test with perfect negative predictive value at time zero

The tragic death of Dr. Martin Salia has triggered a spate of news articles about the limitations of Ebola diagnostic testing. According to news reports, Dr. Salia tested negative twice early in his symptomatic period, and was already quite ill by the time he tested positive.

The fact that the most sensitive Ebola virus diagnostic test (PCR applied to a blood sample) may not be positive until 3 days after symptom onset is well known, however. The reason for this has to do with the pathogenesis of Ebola virus infection. The point of entry and initial replication for Ebola is the “dendritic” cell. Present in large numbers in the skin, mucosa and intestinal lining, dendritic cells are sentinels, guarding the interface between the human immune system and a hostile environment. They encounter invaders, engulf them, and present their antigens to cells of the adaptive immune system. Dendritic cells carry the virus to lymph nodes and other organs of the “reticuloendothelial system” (liver, spleen). So before the virus reaches detectable levels in the blood, there may be hours-to-days of replication in these cells and organs, along with symptoms of fever and fatigue. A test, even a very sensitive test, that is applied to blood will not detect the virus until it appears in the bloodstream in larger numbers. 

This lack of a rapid and accurate diagnostic test early in the Ebola disease course is a major problem—not only for early initiation of therapy for those infected, but also for the management of “persons under investigation” (who may require Ebola-level isolation precautions for several days while awaiting a negative test that has been taken at least 3 days after symptom onset). 

But this problem plagues infectious diseases diagnostics generally. As this recent Lancet ID article points out, the lack of early and sensitive diagnostic testing is one reason why so many patients in our hospitals receive days of unnecessary antimicrobials. More on this study later, perhaps from Eli, one of the authors!

Image by Sean Gallup, Getty Images

Should we go over (to) the Cliff?

There's been a lot of talk in the US about "the cliff," specifically the fiscal cliff. Should we go over the cliff? Is the cliff just a curb?  What should we do? Panic! In response, the GOP has started pushing "Plan B," which will apparently require all Americans to receive emergency contraception if they're pregnant. This has some merit since if we no longer exist, we can't really run up the budget deficit, now can we.

Anyway, this is ostensibly an infection prevention blog, so I better get back on topic.  In the annual Christmas issue of the BMJ, investigators from the Netherlands have reported a novel method for speeding up the diagnosis of Clostridium difficile infection. The name of their novel method?  Cliff.  Just as I expected, they call or email the CDC and ask Cliff McDonald what he thinks! NO? What?

It turns out that they've trained a beagle named Cliff to diagnosis C. difficile by smell (thank goodness it's not taste). Anyone who has done an ID fellowship or even a medical internship gets pretty good at recognizing the unique small of C. diff, so we know this could work. It turns out to work pretty well. Cliff's nose detected C. difficile positive clinical samples with a sensitivity of 83% and a specificity of 98%. Not too shaggy.

Addendum: A 2007 CID study reported self-selected nurses had a sensitivity of 55% and specificity of 83% in diagnosing C. diff, while an earlier study reported that nurses had a sensitivity and specificity of 84% and 77% for predicting C. diff using factors that included odor. I would like to see Cliff dual it out with these nurses in a future trial. Daniel Uslan suggested Cliff vs "Sniff": an RCT.

Multiplex PCR and diagnosis of sepsis

Just came across a really nice paper in the Journal of Clinical Microbiology by Ephraim Tsalik, Chris Woods and others at Duke and the Durham VA. They tested whether multiplex real-time PCR can be a useful addition to blood culture in patients presenting with suspected sepsis to the ED of their two hospitals. Over an almost 6-year period they enrolled 306 patients with suspected sepsis (43 were eventually excluded for non-infectious etiology). Patients had blood samples taken within ~2-3 hours and a questionnaire administered to determine recent exposures and symptoms. Most of the confirmed etiologies were S. aureus (34%) and E. coli (23%).

Results showed that blood culture had a sensitivity of 25% vs. 20% for PCR, a similar negative predictive value (18% vs. 17%), and an area under the curve of 0.63 vs 0.60. Using blood culture as the gold standard, PCR had a sensitivity of 61%. Both PCR and culture detected organisms in 40 patients (38 were the same organism). As far as individual methods, there were 24 organisms only detected by PCR (largely E. coli and Klebsiella) and 52 only detected by blood culture (E. coli, S. pneumo, S. aureus and a large number of CoNS considered contaminants). Six species detected by culture were not in the PCR menu (e.g Listeria, Salmonella).

I think these results speak for themselves. Neither method is perfectly sensitive at this point suggesting that both methods should be used if possible to improve diagnosis. PCR can be quicker and detect additional organisms, but still missed a significant number. Interestingly, blood culture was more sensitive in detecting organisms from patients who had previously received antibiotics (P = 0.06). This is counter to what I would've suspected.

What I really liked about this paper is that it was completed in a way that was clinically useful since they enrolled patients as they presented to the ED and asked the question in the way a clinician would. This was not some convenient sample study which we so frequently see in diagnostic test comparisons. I also liked how they presented the information, giving the sensitivity/specificity, predictive values and AUC. It will be interesting to see if future studies can document a clinical benefit in terms of reduced mortality and length of stay when PCR is added to our diagnostic battery. Finally, I was impressed that this study was completed at all. Enrolling so many septic patients with questionnaires over such a long period of time is no small feat. The authors should be congratulated.