Discussion

Diarrhea is frequently encountered in patients undergoing therapy for acute leukemia and expeditiously arriving at a diagnosis and initiating treatment, when necessary, may significantly contribute to maximizing desired outcomes and minimizing costs. As in our program, MGPP is likely frequently employed at other institutions in the investigation of leukemia-associated diarrhea, resulting in increased odds of identifying a potentially pathologic organism. However, the clinical significance of these identifications is not always clear, and there is a need for better definitions of the judicious use of MGPP in these patients.

Our data show that the most efficient use of MGPP in patients undergoing therapy for acute leukemia is in the evaluation of outpatient diarrhea. This is not surprising as a new diarrhea in these instances is more likely to be due to community acquired pathogens, and all targets in the MGPP except for C. difficile are most commonly acquired in the community setting. Conversely, hospitalized patients would be expected to more frequently have diarrhea caused by the multiple potential noninfectious etiologies seen in the acute leukemia population. In addition, the lack of actionable diarrheagenic organisms other than C. difficile (with the rare exception of Giardia lamblia) identified in our cohort with admission and hospital-onset diarrheas suggests that testing for C. difficile, perhaps including an assay for C. difficile toxin, may be the more efficient approach in these situations.

A potentially significant problem with the use of PCR methodologies is overdiagnosis and overtreatment. Although detecting more potential pathogens because of greater analytic sensitivity, such tests do not distinguish between non-viable forms, colonization, and clinically insignificant infections, and can lead to delays in identifying the actual etiology of the diarrhea. In patient populations such as our leukemia cohort, a higher prevalence of noninfectious diarrhea, especially in the inpatient setting, combined with increased colonization rates may contribute to a significant lowering of MGPP testing specificity.

The problem of overdiagnosis is especially concerning for C. difficile, the most frequently and consistently identified organism in MGPP testing. Our data show that diarrhea is common in patients undergoing therapy for acute leukemia with the majority of instances due to causes other than C. difficile infection. Also, we have previously reported a C. difficile colonization rate at admission of 13% in our acute leukemic population. ^{Reference Ford, Lopansri and Webb5} The risk for overdiagnosis may be greater when the PCR positivity is accompanied by a negative toxin assay. Prior studies in large general inpatient populations have shown that only toxin-positive stools are associated with significant morbidity and mortality and that reliance on molecular testing alone may lead to overdiagnosis and overtreatment. ^{Reference Planche, Davies and Coen6–Reference Polage, Gyorke and Kennedy8} Similar results may apply to our acute leukemia cohort, although more information is needed. It was with these challenges in mind that our new guidelines for C. difficile diagnostic and antibiotic stewardship were adopted in our unit in 2018. ^{Reference Ford, Hoda, Lopansri, Parra, Sharma and Asch4}

Similar issues of overdiagnosis/treatment are encountered with the detection of diarrheagenic E. coli species, which are commonly reported in MGPP series. ^{Reference Cybulski, Bateman and Bourassa1–Reference Axelrad, Freedberg, Whittier, Greendyke, Lebwohl and Green3,Reference Khare, Espy and Cebelinshi9,Reference Spina, Kerr and Cermican10} These seemed to be less common in our acute leukemia cohort and our hematopoietic stem-cell transplant patients, which is not surprising as many of these targets (EPEC, EAEC, ETEC) are less common causes of gastroenteritis in the United States. In addition, the lack of reproducibility for certain MGPP rare targets ^{Reference Hitchcock, Hogan, Budvitiene and Banaei12} and the inability to confirm the presence of EPEC, EAEC, and ETEC may create other challenges in a cohort of patients such as our leukemia population with numerous potential causes of diarrhea, especially since these targets are not infrequently detected with other targets. These observations point to important questions for future study such as how to establish the clinical significance, need for treatment, and impact of colonization of these organisms in the acute leukemia population.

Limitations of our study include a retrospective design, a single institution experience that may reflect local practice preferences and dependence on chart reviews for clinical data. Given the long duration of the study data, it is likely that changes in policies during the interval, such as introduction of the C. difficile guideline, may have affected clinician ordering and treatment patterns.

In summary, we conclude that when used to evaluate for an infectious cause of diarrhea in patients with acute leukemia, BFA is most useful in the outpatient setting. For hospital-acquired diarrhea, targeted testing for C. difficile toxin is a more efficient, cost-effective approach, and routine testing with MGPP, which consists mostly of pathogens that are transmitted in the community, is not warranted. The clinical significance of diarrheagenic E. coli, other than STEC, and MGPP rare targets needs further evaluation.