C. difficile Toxin Protein Assay Application & Misconception
An important question in the field of infectious disease diagnostic testing is whether or not Clostridium difficile (C. difficile) toxin protein assays (hereafter, toxin assays) differentiate C. difficile infection (CDI) from mere carriage of toxigenic C. difficile in the GI tract. A considerable number (~3% to 16%) of hospitalized patients not being evaluated for CDI, even without GI symptoms at all, can test positive for toxigenic C. difficile by sensitive nucleic acid amplification tests (NAAT, e.g. polymerase chain reaction or PCR). NAAT tests can therefore be positive when the organism is not contributing to disease and does not need to be treated, as well as when it is contributing to cases of real CDI.
It is important to keep in mind that there is robust literature supporting that patients who are NAAT positive but toxin negative (NPTN) are considerably less likely to have severe outcomes than those who are toxin positive (NPTP). For example, in a recent study published in JAMA Internal Medicine, researchers observed 10 CDI related complications (colectomy, megacolon, ICU admission) and 11 deaths in the NPTP group (n=131) but no complications and only 1 death in the NPTN group (n=162). In another study, published in the Journal of Clinical Microbiology, researchers performed a quasi-experimental non-inferiority study that supported these findings.
What remains unclear, and is regularly (and at times heatedly) put to debate, is whether less sensitive toxin assays, which detect the actual C. difficile toxin proteins (A and B) that cause CDI, are effective at winnowing out carriage cases and identifying CDI when it really needs to be treated. The above studies suggest that toxin assays can be used to determine which patients have CDI and require treatment and which don't. However, data from 2 recently published papers, both of which utilize a new ultrasensitive assay (Simoa) for detecting C. difficile toxins A and B with sub-picogram limits of detection, strongly suggest the opposite.
The first study, published in Clinical Infectious Diseases, evaluated patients in 2 groups: CDI-NAAT (PCR positive, and with clinically significant diarrhea) and Carrier-NAAT (PCR positive, without clinically significant diarrhea). Notably, the second group consisted of patients who were already hospitalized, and then were subsequently recruited to the control arm of the study. Two important concepts emerged from the data: high toxin concentrations correlated with CDI, but there was no threshold of toxin concentration for which one could expect good sensitivity for CDI, without poor specificity, compared to the carriers. Toxin assay fans could find solace in the exceptionally low limit of detection for toxin, as well as the possibility that perhaps the CDI-NAAT group was too inclusive of patients with low concentrations of toxin that didn't have as many (or any) bad CDI-related outcomes, and thus, were less important to detect.
Since the ultrasensitive Simoa assay isn't currently in clinical use, it is valuable to look at these data through the lens of the most commonly used clinical C. difficile toxin assay in the U.S. (among labs with proficiency testing results reported through the College of American Pathologists), the TechLab C. DIFF QUIK CHEK COMPLETE®. The limit of detection for this assay is provided in the manufacturer's instructions for use and represented by the red lines in the images below (0.63 ng/mL for toxin A and 0.16 ng/mL for toxin B). Line placements are approximate, and the 2 tests were not directly compared. Although there is a correlation between CDI and very high toxin concentrations, results for the 2 groups of patients are extensively overlapping, and one can see how many patients designated as CDI would be missed by the less sensitive assay.
The authors of the first Simoa paper came back with another beautiful study that could be interpreted as closing the case of whether toxin assays can be used to differentiate CDI from carriage. This study took a deep clinical dive, categorizing patients for severity of disease, bad outcomes and whether or not those outcomes were attributed to CDI or CDI contributed. Among 615 patients in the study, 19 had a severe outcome primarily attributed to CDI, and among these 18/19 had at least 20 pg/mL of toxin detected.
Yes, there is correlation between toxin concentration and disease severity and outcomes, but it is clearly correlation without differentiation. Panels A and B from figure 1 of this study are shown below. This is remarkably damning data for the diagnostic utility of C. difficile toxin protein assays. Using the proposed clinical cutoff of 20 pg/mL would cause 1 of the patients with a severe outcome attributed to CDI to be missed entirely, as well as more than 1/4 of the patients with severe outcomes for which CDI was deemed a contributing factor. Furthermore, more than 1/2 of the cases that had no severe outcome at all would be detected. Drawing from the findings of the first study, one can conclude that more than 1/2 of the carriers would also be detected.
Looking again at these data with the overlay of the commonly used TechLab C. DIFF QUIK CHEK COMPLETE® limits of detection, while acknowledging that this is only an approximation without comparing the 2 tests directly, confirms that some important cases are missing.
At this point it appears that C. difficile toxin concentration has almost no utility, if any at all, for differentiating important CDI cases from carriage. Why, then, do we see relatively few bad outcomes in patients that are NAAT positive and toxin negative, when only the toxin result is reported? It may relate to a major confounder in interpreting these studies—many patients that test toxin negative are treated for CDI anyway. Physicians are using clinical judgment to treat a subset of these patients and preventing additional bad outcomes. In the Polage et al. study cited above, among NPTP patients, 2.3% were treated with metronidazole or oral vancomycin within 48 hours prior to toxin testing, whereas among NPTN patients, this figure was 19.8%. Within 14 days of result reporting 40.7% of NPTN were treated, despite the negative test results.
As mentioned previously, there is no question that NAAT testing detects a considerable number of patients that simply have toxigenic C. difficile as part of their GI flora and do not require treatment. However, NAAT testing elucidates the only really reliable piece of information the lab can provide for the clinical consideration of CDI—excellent negative predictive value, with the added benefit of detecting carriage cases for hospital epidemiological purposes to implement precautions and modestly reduce the spread of this organism in the hospital environment.
In light of the new data generated using an ultrasensitive toxin assay, it is difficult to suggest that toxin tests will ever have any ability to separate patients with carriage from patients with CDI. Thus we continue, perhaps in vain, to seek a C. difficile testing option that provides reliable differentiation rather than merely correlation.