Discussion

Our study is one of only two survey studies of US institutions focused on antibiotic de-decisions during neutropenic fever and the first to focus on real-world management, including practice patterns in specific clinical situations. Our respondents represent a diverse group of institutions while still corresponding to the care of approximately one in six US hematologic malignancy and bone marrow transplant patients, with the infrastructure of SRN resulting in a relatively robust response rate.

Of respondents, most had guidelines for neutropenic fever prophylaxis, with the majority using fluoroquinolones, consistent with clinical guidelines. ^{Reference Baden, Swaminathan and Angarone4} Although there are limited randomized data about mortality benefit of antibacterial prophylaxis in neutropenic patients, the clinical benefits in preventing fevers, infections, bacteremia, and hospitalizations are well supported. ^{Reference Bucaneve, Micozzi and Menichetti28,Reference Cullen, Steven and Billingham29} In addition, a 2005 meta-analysis and 2012 Cochrane review, both done by the same group, show a reduction in all-cause mortality in the use of fluoroquinolones in this setting. ^{Reference Gafter-Gvili, Fraser, Paul and Leibovici30,Reference Gafter-Gvili, Fraser and Paul31} However, this benefit is balanced by the potential for resistance; a recent single-center pre/post retrospective study from Italy (in a self-described area of increasing fluoroquinolone resistance) showed a lower rate of fevers but a higher incidence of fluoroquinolone-resistant organisms in leukemia patients receiving prophylaxis, with no difference in blood stream infections or mortality. ^{Reference Urbino, Frairia and Busca32}

No respondents use carbapenems or aminoglycoside combinations as first-line therapy of febrile neutropenia, deviating from what respondents reported in other survey studies in prior years (Table 1). Only four institutions screen for resistant gram-negative bacterial colonization, in contrast to this practice being done more routinely outside of the United States based on increasing proportions of multidrug resistance in the Netherlands, ^{Reference De La Court, Woudt and Schoffelen33,Reference De La Court, Heijmans and Huynh34} Iran, ^{Reference Darakhshandeh, Fathi, Haji Gholami, Ashrafi, Mehrzad and Nasri35} Brazil, ^{Reference Schonardie, Beck and Rigatto36} Italy, ^{Reference Micozzi, Minotti and Capria37} Spain, ^{Reference Chumbita, Puerta-Alcalde and Yáñez38} and Japan. ^{Reference Hattori, Goto, Osaki, Ozawa and Miyamura39} A recent US single-center RCT showed benefit in using ceftolozane-tazobactam over standard-of-care agents for the management of febrile neutropenia. ^{Reference Chaftari, Hachem and Malek40} Despite these reports, a recent analysis of US neutropenic fever cases with bloodstream infections still showed high efficacy of cefepime or piperacillin-tazobactam, ^{Reference Zimmer, Stohs and Meza41} which all of our respondents used as first-line therapy.

We found significant heterogeneity in de-escalation of empiric antibiotics relative to ANC recovery, with varying willingness in doing so in “bland” neutropenic fever, identified clinical infectious source, or microbiologically identified infections. In addition to the previously discussed evidence for de-escalation of empiric antibiotics in this setting, ^{Reference Aguilar-Guisado, Espigado and Martín-Peña15,Reference De Jonge, Sikkens and Zweegman16} there are some additional limited data about the safety of de-escalation in bloodstream infections, whether to target the culture result ^{Reference Rainess, Campbell, Santamala, Kubin and Mehta42} or to treat with a short course of broad antibiotics for low-risk microbes. ^{Reference Metais, Torregrosa Diaz and Gallego Hernanz43} This heterogeneity is consistent with the uncertainty in de-escalation reported in the prior US survey study ^{Reference Barreto, Aitken and Krantz26} and highlights a need for more robust evidence for this practice, as well as its education and adoption across US institutions, in the context of local antibiograms and drug resistance.

Of note, even among those institutions with “early” de-escalation protocols, only 1 respondent was comfortable doing since before a period of apyrexia. This practice was consistent with the de-escalation protocol of the How Long trial, ^{Reference Aguilar-Guisado, Espigado and Martín-Peña15} which used a 72-hour period of apyrexia, but was not consistent with the protocol used by de Jonge et al. ^{Reference De Jonge, Sikkens and Zweegman16} The latter randomized one group to de-escalation after 72 hours of carbapenem treatment, regardless of febrile status at the time, compared to a control group of 9+ days of treatment and 5+ days of apyrexia. However, these results did not meet non-inferiority criterion in the per-protocol treatment failure comparison (23% vs 16% in treatment failure, 12% vs 7% in re-admission, and 3% vs 1% in 30-day mortality), with three patients noted to have gram-negative bacillus bacteremia. These findings may explain why our respondent institutions continue to be cautious with de-escalation in patients with continued fevers, despite the trial’s top line findings, as the ongoing fevers likely represent a propensity for treatment failure with recurrent fever, whether from untreated infection or non-infectious fevers from malignant tumor.

Our study provides several notable differences compared to the initial US-based survey reported by Barretto et al. ^{Reference Barreto, Aitken and Krantz26} First, as discussed, we focused on real-world management rather than institutional guidelines, with a focus on specific patient scenarios that capture a diverse number of clinical presentations. Second, although our survey similarly targeted infectious disease physicians and antimicrobial stewardship pharmacists, it was conducted through the SRN, which includes both tertiary referral centers and community-based institutions providing cancer care, rather than focusing solely on cancer centers. Indeed, our response rate compares favorably and captures clinical settings with various levels of academic affiliation and clinical volume. Third, by virtue of its timing, our findings reflect the current state of practice, including any potential impact of the more recent RCT, ^{Reference De Jonge, Sikkens and Zweegman16} as well as the COVID-19 pandemic.

In contrast to these strengths, our study has several limitations as well. Although the use of SRN allowed us to capture a wide variety of healthcare institutions and settings, survey respondents remain limited to SRN members, who have a stated commitment to promote antimicrobial stewardship. This raises the limitation of generalizability to the care of cancer patients in non-SRN institutions, though our response rate was reasonable and compares favorably to the prior US-based study (54.8% vs 30%). Next, the respondents consist of SRN representatives at their institutions, which typically consistent of infectious disease and stewardship providers and pharmacists, rather than hematologists, oncologists, or hospitalists that may be those carrying out direct clinical care for neutropenic fever patients. The decision regarding de-escalation prior to ANC recovery can be complex and nuanced on the “front lines” of direct patient care. Thus, responses about de-escalation may represent a mix of personal clinical judgment, aspirational or tangible institutional guidelines, and a true reflection of clinical decision-making. This is a limitation of most survey studies about stewardship but nonetheless raises challenges with interpretation and comparison between studies. We sought to mitigate these limitations by using concrete clinical scenarios with various situations, which hopefully captures real-world decision-making across a range of potential patient presentations. Furthermore, our findings of the variability of clinical practices, even as reported by stewardship-focused practitioners, reinforces the lack of guideline-directed information to support evidence based uniform practice, rather than simply an implantation issue or differences in cautiousness in clinical practice. Finally, our survey only captured a small number of pediatric healthcare institutions, where guidelines regarding de-escalation are more well established, limiting our ability to draw conclusions about this patient population.

In conclusion, our study provides valuable insight into real-world practice patterns in the management of neutropenic fever at US healthcare institutions, with a robust survey sample accounting for a sizable proportion of oncologic care nationwide. There was relative agreement in the practice of prophylaxis and initial management of neutropenic fever, though with some variation in empiric coverage of non-gram-negative organisms, both upfront and in the case of prolonged fevers. Most notably, we found significant heterogeneity in the de-escalation of empiric gram-negative coverage, representing a potential opportunity for antibiotic stewardship. Though there has been increasing evidence for the safety of this practice, randomized data remain relatively scarce, highlighting the need for more research regarding this question to convince and guide both clinical guidelines and individual practitioners.