KQ1. Diagnostic Tests

The literature has shown a strong shift from immunoassays to nucleic acid amplification tests, mirroring the evolution of clinical practice for diagnosis of CDI. Given the greatly increased published literature of diagnostic studies, we were disappointed at the lack of eligible studies of the impact of diagnostic tests on patient or health system outcomes; that is, does more accurate and expeditious diagnosis of CDI lead to improved outcomes for patients or measured improvement for health systems (with respect to cost of care, length of stay, or rates of CDI). Although some retrospective studies described changes in incidence and prevalence in an institution before and after implementation of a new testing strategy, these generally did not include verification of the reported incidence and prevalence with an acceptable reference standard and are thus difficult to interpret. One study that did not meet our inclusion criteria (due to an inadequate reference standard) showed that a health system's change from a testing strategy based on culture and cytotoxicity assay to PCR or algorithmic approaches decreased time to results, decreased vancomycin and metronidazole use in patients without CDI, and decreased time to appropriate therapy in patients with CDI.⁶⁷

Interpreting the findings of the diagnostic testing evidence requires a nuanced approach. First, the reference standard used to define the presence or absence of disease and the implications of that reference standard must be considered. We opted to use toxigenic culture or CCNA performed on loose stool as the reference standard; this is not a clinical reference standard that includes clinical information such as severity of disease and antibiotic exposure.

Further, the pretest probability of CDI (and the severity of the CDI, if present) varies with patient characteristics as well as clinical setting, with inpatient populations having higher prevalence and severity of disease. Determining the presence or absence of a given disease is not simply the obverse and reverse of each other, respectively. In the inpatient setting—from which the vast majority of patients in the studies included in this report were drawn—a clinician's priority is on ruling out disease (and not treating for CDI), and thus a higher false positive rate is likely acceptable. Since specificity was uniformly quite good across test classes, the difference in performance between classes of tests for CDI appears to be derived mostly from differences in sensitivity between classes. This has significant implications for testing strategy selection, since ensuring a negative test is accurately ruling out disease is likely of more clinical importance than ensuring that a positive test (in a patient with signs and symptoms consistent with CDI) represents true CDI. Thus, we believe the differences in sensitivity and negative likelihood ratios to have clinical importance.

There was substantial heterogeneity in the studies of diagnostic accuracy from both measured and unmeasured (and thus undescribed) sources. The prevalence of CDI in the study population is one of the few consistently published population characteristics. The prevalence of CDI in the examined studies varied widely, between 6 and 48 percent. While sensitivity and specificity should theoretically not vary with prevalence, they often do vary in studies of diagnostic testing, and future work should examine how the prevalence of CDI (likely a reflection of the population in a study and local testing behavior) influences the measured operating characteristics. There are many other undescribed clinical variables that may differ between populations and lead to heterogeneity. For example, training and implementation procedures generally were not described in detail. Lastly, the reference standard for each study was performed according to local protocols and, while used as the definition for the presence or absence of disease in the studies, may vary substantially, leading to different prevalence and operating characteristics. Unless a large, prospective trial is performed in which all reference standards (toxigenic culture and/or CCNA) are performed centrally to avoid local variation, these limitations are unlikely to be avoided in future studies.

The diagnostic studies included in this report included patients only with suspected CDI and thus the operating characteristics (that is, sensitivity and specificity) are defined in patients with suspected CDI, not general patients with diarrhea or healthy patients. Thus, these tests should only be interpreted in patients similar to patients enrolled in the included studies and clinicians must be aware of the prevalence of CDI in their own local population in determining whether they choose to employ a more sensitive or specific testing strategy. Further, the reference standard used (toxigenic culture or CCNA) does not include clinical information.

For this update we used a different approach to examine diagnostic tests, pooling studies by test class, since the selection of tests from within a test class for use at a certain institution will likely depend on both the operating characteristics of the test class and individual test as well other factors including cost and vendor preference. We found moderate to high evidence that NAAT tests (one LAMP, 10 PCR) are highly sensitive and specific. All other diagnostic tests (eight Toxin A/B immunoassays, four GDH immunoassays, and 11 test algorithms) were high in sensitivity and/or specificity but when compared with NAAT tests, lack the same combination of high sensitivity and specificity.

Test algorithms, intended to make the best of individual test strengths performed in series, did not perform as a class as well as NAAT tests. Clinical interest in test strategies has declined because of the issue of what to do when a positive initial test is followed by a negative test. Many clinicians will continue treating based on the first test because of the uncertainty (increased probability of CDI after the test), and because the test is usually ordered based on the clinician's pretest assessment of the patient's probability of CDI. Dichotomous, positive/negative results are easier for clinicians to interpret and require less laboratory followup.

NAAT tests come with a different set of concerns, including whether switching to NAAT will falsely inflate nosocomial CDI rates; these highly sensitive tests may identify people who are asymptomatic carriers or patients with diarrhea from a cause other than CDI. Since NAAT tests nearly approximate toxigenic culture in sensitivity and specificity, implementation of a NAAT-based testing strategy may lead to a higher observed prevalence/incidence compared with other testing strategies. Further research is required to determine if NAAT-based testing strategies lead to overtreatment for CDI in patients who are asymptomatic carriers or have diarrhea from another cause. It is likely that NAATs will be used as the reference standard in future studies and the implications of this change must be considered in interpretation of these studies.

To assume that one “best” test exists for all healthcare purposes is an oversimplification, especially with respect to populations with different pretest probabilities, as previously discussed. The need to understand the pretest probability of CDI extends to the stool samples for which laboratories will perform testing for C. difficile and ensuring that only unformed specimens from patients at risk for CDI are tested to avoid false positive tests that lack clinical significance and may lead to overtreatment. In addition, optimal testing strategies between health systems may differ on factors other than test analytics, including start-up costs, expertise, incremental (per-assay) cost, and other factors. Further, clear delineation must be made between the most effective test characteristics for at-risk individuals who may benefit from CDI treatment versus for population surveillance or epidemiologic evaluation.

KQ2. Prevention

The prevention literature remained generally low-strength, and little evidence connects prevention strategies directly to patient-related outcomes such as CDI incidence. Studies of transmission interruption techniques were often excluded due to lack of patient-related outcomes (they used swabbing and culturing to assess the presence of C. difficile organisms or spores). However, we did identify some small updates to the original review. Low-strength evidence supports handwashing campaigns. Low-strength evidence also suggests that prevention programs are sustainable in the long-term. However, it remains difficult from a research perspective to definitively state that bundled, multicomponent interventions are effective, as each remains relatively unique to the specific location and the components included in that bundle. The information is still insufficient to answer which components are essential or what might be added.

Low-strength evidence continues to support antibiotic prescribing practices. Again, none of the studies explicitly addressed potential harms of changes in antibiotic use policy, such as the possibility that preferred drugs will be less effective than the drugs physicians are discouraged from using, or that preferred antimicrobials might have greater costs or greater toxicities unrelated to CDI.

KQ3. Standard Treatment

Three new studies of standard treatment raised confidence in several findings from the original review. We increased strength of evidence from moderate to high for vancomycin as a more effective agent than metronidazole for CDI, with moderate-strength evidence of the effect regardless of severity. Current treatment guidelines from the Infectious Diseases Society of America (IDSA) support vancomycin as the drug of choice for severe CDI, and metronidazole as the drug of choice for mild to moderate CDI.⁶⁸ This review's finding is consistent with reconsidering the preferred agent for mild to moderate CDI, although the long-term effects of increased vancomycin use are unknown. This is especially true in light of scant evidence to suggest that vancomycin promotes the emergence of vancomycin-resistant enterococci more so than other agents and a decrease in the price differential between metronidazole and vancomycin.

A second important finding is continuing moderate-strength evidence that fidaxomicin is similar to vancomycin for the initial cure of CDI, and increased strength of evidence for fidaxomicin is superior for the prevention of recurrent CDI. Since the desired outcome with CDI treatment is cure of the initial illness without subsequent recurrence, this finding ought to prompt consideration of fidaxomicin for the initial treatment of CDI. This is especially relevant to the treatment of CDI since each episode of recurrence increases the likelihood of further episodes. Since fidaxomicin was licensed after publication of the most recent IDSA guidelines, they include no mention of fidaxomicin. Accordingly, its role in treating CDI has been a topic of considerable discussion. A recent cost-benefit analysis concluded that the per-course price of fidaxomicin would need to decrease by more than 10-fold in order to make such use cost-effective.⁶⁹ The current high cost of fidaxomicin prompted its manufacturer to seek and obtain a new technology add-on payment from the Centers for Medicare and Medicaid Services. This add-on provides hospitals additional payment to offset fidaxomicin's high cost. Future guidelines will hopefully give clinicians guidance as to how to best use this agent to maximize the value seen in terms of reduced episodes of recurrent CDI.

A final updated finding is that in the observational study of intravenous metronidazole verses oral metronidazole and vancomycin, intravenous metronidazole performed significantly worse than either oral drug. This finding should be interpreted with caution given the observational nature of the study and the significant possibility of confounding. Since this finding largely confirms current clinical practice, it will not likely have a major impact on the treatment of patients with CDI.

The findings in this review remain applicable for the general adult CDI patient population. Given the paucity of the literature, we were unable to assess findings for important subgroups of interest.

KQ4. Other Treatments

Adjunctive treatments in the updated literature have largely focused on restoring the colonic microbiome for the prevention of subsequent CDI, although a few explored different mechanisms such as toxin-binding (tolevamer, cholestyramine) and direct antimicrobial properties (lactoferrin). The diverse bacterial species residing in the human gut, commonly referred to as the colonic micobiome, provide host resistance to infection by C. difficile. Several factors, such as antimicrobial use and chemotherapy, disrupt the diversity of the colonic microbiome and lower the resistance to CDI. Antimicrobials are effective in treating CDI, but also disrupt the colonic biodiversity and do not address the necessary repopulation of these organisms. These changes make the host susceptible to recurrent episodes of CDI. Probiotics aim to recolonize the intestinal flora with nonpathogenic bacteria, while FMT involves the transfer of the entire microbiome from one individual or a pool of donors to the host.

Low-strength evidence supports FMT as a promising therapy for recurrent CDI. Our findings are consistent with another recent systematic review which provided greater detail regarding method and route of FMT, as well as donor characteristics, but did not include six recent studies included in this report.⁷⁰ Since our original review, numerous studies have addressed FMT for the treatment of recurrent CDI, including two small unblinded RCTs comparing FMT to vancomycin-based control groups, one small RCT comparing two different modalities of administration for FMT, and numerous case series. The case series ranged from small to medium size and provide a cumulative experience with FMT of 751 individuals, with reported success rates from 48 – 100 percent. However, the high probability of publication bias and the lack of control groups are major limitations. The data from the RCTs comparing FMT to vancomycin are encouraging, demonstrating a significant benefit for FMT, although the study risk of bias is high. Specifically, participants and providers were unblinded, both trials were stopped early, and the control groups in both of the trials had success rates from 23 to 31 percent, far lower than the 55-60 percent rates expected based on the sample size calculations published in the study protocols. Additionally, followup was limited in most studies; thus, the long-term consequences of FMT treatment are unknown.

Insufficient evidence exists for FMT for refractory CDI. In contrast to FMT for recurrent CDI—which is administered after a course of antimicrobial therapy has eliminated or greatly reduced symptoms of CDI, and whose main aim is to prevent subsequent recurrences—FMT for refractory CDI is administered to patients with ongoing symptoms of CDI despite antimicrobial therapy. Since the great majority of patients with CDI respond to initial antimicrobial treatment, studies of refractory CDI are inherently difficult.

The scientific and regulatory issues for FMT pose unique challenges, as there are no standard formulations, methods of quantifying, or assessing safety of stool. The composition of stool, and which constituents may be active in reducing recurrent CDI, is also currently unknown. Initial FDA guidance required an Investigational New Drug (IND) for any use of FMT. After significant public input, current FDA policy is to exercise enforcement discretion regarding IND requirements for FMT in specific situations. To proceed under enforcement discretion, FMT must be used to treat CDI not responding to standard therapies, and the treating physician must obtain an informed consent including, at a minimum, discussion of both the investigational nature of FMT and its potential risks (www.fda.gov/BiologicsBloodVaccines/GuidanceComplianceRegulatoryInformation/Guidances/Vaccines/ucm361379.htm). This guidance is subject to change as more evidence and experience accumulates on FMT, including the optimal route of FMT delivery. The sources of material for FMT are also variable and include use of unrelated or related donors. No standard criteria exist for screening donors, but several guideline documents have been developed. The optimal team to administer and oversee an FMT program is uncertain but may include gastroenterology, infectious diseases, pharmacy, infection control, nursing, and facility management.

The low-strength evidence supporting probiotics for the prevention of CDI is mixed. Preparations containing S. boulardii alone did not seem to significantly affect subsequent rates of CDI, whereas preparations containing lactobacillus strains or multiorganism mixes did significantly reduce rates of CDI. Notably, the studies aimed to examine probiotics for primary prevention of CDI among patients without a prior episode of CDI. Whether the findings apply to patients with a history of CDI (that is, to prevent recurrence of CDI) is unknown. Our findings are generally consistent with another systematic review, although we differed on the S. boulardii finding due to our review including one additional study.⁷¹ Administering a course of probiotics to every patient taking antimicrobials would also be a rather substantial change in medical practice. The cost/benefit ratio of such a policy is unclear, based on the mixed and low-strength findings of this report, as is whether benefits could be conferred by ingesting probiotics in the form of yogurt, kefir, and other similar foods. Given the multitude of such foods available to consumers, the prospect of obtaining rigorous data on each seems unlikely.

Finally, rifaximin and lactoferrin were studied in separate small placebo controlled trials and oral cholestyramine in a case series of 46 patients regarding their ability to prevent subsequent episodes of CDI. Rifaximin was given after a course of standard antimicrobial therapy for CDI in hopes of preventing recurrent CDI. In contrast, lactoferrin and cholestyramine were given concomitantly to antibiotics prescribed for non-CDI indications in the hopes of preventing an initial episode of CDI. In the controlled trials the investigational agent reduced the incidence of subsequent diarrhea, but not confirmed CDI. CDI was confirmed in the cholestyramine study.

The bulk of the new studies of adjunctive treatment for the prevention of subsequent CDI involve efforts to reconstitute the colonic microbiome with either FMT or probiotics. The supporting evidence is low-strength. The FMT studies show a large treatment effect but are limited by methodological weaknesses. In contrast, the studies supporting probiotics demonstrate a less-impressive treatment effect. The FMT studies included patients with at least a single prior episode of CDI and, in many cases, multiple prior episodes. The probiotic trials, on the other hand, are examining primary prevention of CDI. Both primary and secondary prevention of CDI are important, particularly since the burden of CDI has significantly increased over the past 15 years.