Discussion

Among hospitalized patients with cancer, we observed that despite internal guidance to order GIPs primarily for community-onset diarrhea (on or before HD 3), approximately one-third of GIPs were ordered after HD 3. Prevalence of detection of ≥1 non-C. difficile target after HD 3 was lower compared to GIPs ordered on HD 1–3. These findings suggest that GIP yield for the detection of non-C. difficile pathogens in patients with cancer and hospital-onset diarrhea is limited, similar to another study in oncology patients ^{Reference Laracy, Yan and Kodama11} as well as studies evaluating broader hospitalized populations. ^{Reference Hitchcock, Gomez and Banaei3,Reference Campodonico, Hanlon and Mikula4,Reference Alhothli, Payne and Stefanovic6}

In our study, BF-GIP yield for detection of ≥1 non-C. difficile target after HD 3 was 10%, which is nearly identical to the overall BF-GIP positivity rate reported by Laracy et al among inpatients at a cancer center after the third HD. ^{Reference Laracy, Yan and Kodama11} Hitchcock et al reported a 6.0% BF-GIP yield in all hospitalized adults >3 days after admission (6.6% yield in the subset who were immunosuppressed); however, targets frequently detected in asymptomatic individuals (enteropathogenic, enteroaggregative, and enterotoxigenic E. coli) were excluded from their analyses. ^{Reference Hitchcock, Gomez and Banaei3} Ford et al found that only 3% of BF-GIP tests detected non-C. difficile targets in hospitalized patients with acute leukemia and allogeneic HCT recipients with hospital-onset diarrhea. ^{Reference Ford, Lopansri, Hunter, Asch and Hoda9,Reference Ford, Lopansri, Hunter, Wilkes, Asch and Hoda10} Differences in BF-GIP yield among studies could be explained by variations in stool testing practices and pathogen prevalence across centers.

Aside from C. difficile, norovirus was the second most common target detected by GIP both before and after HD 3, which is similar to other studies evaluating the BF-GIP. ^{Reference Rogers, Westblade and Soave2,Reference Hitchcock, Gomez and Banaei3,Reference Ford, Lopansri, Hunter, Asch and Hoda9,Reference Ford, Lopansri, Hunter, Wilkes, Asch and Hoda10,Reference Shu, Wang and Misra22} Following the 2024 FDA recall of the BF-GIP due to an increased risk of false positive norovirus results, ²⁰ several studies have reported that >25% of norovirus detections by BF-GIP were likely false positives based on alternative molecular testing and/or melting curve analysis. ^{Reference Caza, Kuchinski and Locher23–Reference Matic, Lawson and Young25} Additionally, recent data from our microbiology laboratory demonstrated that the BF-GIP Noro-1 assay has a false discovery rate as high as 74%. ^{Reference Reed, Mauer and Mack21} It is therefore likely that a significant portion of norovirus detections in the present study may represent false positive results. In a sensitivity analysis, we found a significantly lower prevalence of detecting ≥1 non-C. difficile, non-norovirus target after HD 3, representing a 38% reduction compared to HD 1–3. We speculate that false positive norovirus results may have inflated the yield of the GIP for detection of ≥1 non-C. difficile target in our primary analysis, thereby over-estimating GIP yield after HD 3. Understanding the clinical implications of false positive norovirus results on the BF-GIP is an important area for future study.

In the present study, detection of a non-C. difficile target in the context of hospital-onset diarrhea infrequently led to changes in clinical management, consistent with a prior evaluation of GIPs in hospitalized patients with leukemia ^{Reference Ford, Lopansri, Hunter, Wilkes, Asch and Hoda10} and another report in a general acute care population. ^{Reference Alhothli, Payne and Stefanovic6} In our study, positive guideline-discordant GIPs (tests detecting a non-C. difficile target that were ordered for new-onset diarrhea after HD 3) led to intentional antimicrobial therapy in only 5 patients, 2 of whom may have received unnecessary treatment for EPEC and norovirus, where antimicrobials have little or uncertain benefit. ^{Reference Olvera, Carter and Rajan26–Reference Hu and Torres28}

Three patients with hospital-onset diarrhea were treated for Campylobacter, Yersinia enterocolitica, and Giardia, and these detection events are more likely clinically significant infections. However, highly sensitive stool PCR panels cannot distinguish active infection from colonization or prolonged pathogen shedding following a previous infection. ^{Reference Castillo Almeida, Cichon and Gomez8–Reference Ford, Lopansri, Hunter, Wilkes, Asch and Hoda10,Reference Liesman and Binnicker12,Reference Berdal, Follin-Arbelet and Bjornholt29,Reference Liu, Rosen and Stohs30} This challenge commonly arises in oncology patients, who often have multiple concurrent causes of diarrhea and frequent exposure to antibiotics for prophylaxis or treatment. ^{Reference Liu, Rosen and Stohs30,Reference Gottesdiener and Satlin31} While GIPs may enable early, targeted therapy or antibiotic de-escalation in some cases of acute gastroenteritis, ^{Reference Cybulski, Bateman and Bourassa1} use of GIPs has also been associated with increased or unnecessary antibiotic use. ^{Reference Brendish, Beard and Malachira32,Reference Robilotti, Powell, Usiak, Taur, Babady and Kamboj33} Further studies are needed to understand how GIP results impact clinical outcomes and appropriate versus inappropriate antibiotic use.

Additionally, false positive results or detection of targets with unclear clinical significance by GIP, such as non-Shiga toxin-producing E. coli, ^{Reference Shu, Wang and Misra22,Reference Hu and Torres28,Reference Berdal, Follin-Arbelet and Bjornholt29} could lead to premature diagnostic closure or unnecessary isolation and infection prevention resource utilization. ^{Reference Robilotti, Powell, Usiak, Taur, Babady and Kamboj33} Misdiagnosis may be of particular concern in HCT recipients, who are at risk for several diarrheal syndromes associated with significant morbidity and mortality, including gastrointestinal graft-versus-host disease, cytomegalovirus disease, and adenovirus. ^{Reference Castillo Almeida, Cichon and Gomez8} Furthermore, GIPs that detect only adenovirus serotypes F40/41 have demonstrated low yield for diagnosis of adenovirus infections in patients with hematologic malignancy and HCT recipients. ^{Reference McMillen, Lee, Kamboj and Babady34}

Limitations of this study include the potential for confounding due to the retrospective design and lack of prior antibiotic exposure data for inclusion as a covariate in the GEE model. Additionally, GIPs ordered after HD 3 did not occur randomly, as our institutional recommendations and a provider nudge discourage use of GIPs for hospital-onset diarrhea. Because our guidelines also allow GIPs to be considered after HD 3 for prolonged, unexplained, or bloody diarrhea, not all GIPs ordered after HD 3 necessarily represent true, hospital-onset diarrhea. Furthermore, since we only reviewed order indication and intentional treatment for the subset of GIPs that detected a non-C. difficile target after HD 3, we could not directly compare these characteristics between GIPs ordered on HD 1–3 versus after HD 3. Finally, this is a single-center study thus results may not be generalizable to other centers or clinical settings where pathogen prevalence and stool testing practices may differ.

Overall, this study builds on existing literature demonstrating low yield of GIPs after HD 3 and indicates that GIPs provide limited additional value beyond SACD testing for evaluation of hospital-onset diarrhea in patients with cancer. Our findings support our center’s guideline to reserve GIPs primarily for patients with community-onset diarrhea, even among those with cancer and HCT recipients. At our cancer center, future GIP diagnostic stewardship efforts may have the most potential for impact in the hematologic malignancy and HCT populations, where GIP utilization is highest and positive results after HD 3 rarely affect antimicrobial management.

Additional diagnostic stewardship interventions, such as requiring providers to specify a clinical indication for GIPs ordered after HD 3, could be employed to limit ordering of low-value GIPs for hospital-onset diarrhea. Evaluation of the efficacy and safety of more restrictive approaches, such as implementation of a “hard stop” with laboratory approval for any GIP orders after HD 3, could be investigated with future prospective studies. Though yield after HD 3 may be low, further studies are needed to better define which subsets of oncology patients are most likely to have a clinically significant GIP result after HD 3 and how these tests impact clinical outcomes.