EUVIMEDEUVIMED entdecken
LeserbriefOpen Access · CC BY-NC 3.0

Fragility of Observed Associations in Selection of Empirical Antifungal Therapy for Critically Ill Non-Neutropenic Patients [Letter]

Saba Manzoor, Daniele Roberto Giacobbe iD, Ronald Chow iD

Infection and Drug Resistance · 2026-06 · Volume 19 · 1-3

DOI: 10.2147/idr.s630961 ↗

Kurz gefasst

Leserbrief in der Zeitschrift Infection and Drug Resistance, 2026-06. Verfasst von Saba Manzoor und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of Medicine, University of Toronto, Toronto, Ontario, Canada. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 3.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In a recent multicenter observational study, Albakri et al evaluated empirical antifungal therapy (EAFT) in 324 critically ill, non-neutropenic ICU patients.1 The authors report a statistically significant association between antifungal selection and in-hospital mortality on unadjusted analysis (p = 0.028), with lower mortality in patients receiving fluconazole (63.2% vs. 76.8% for caspofungin, 75.0% for liposomal amphotericin B, and 84.8%for anidulafungin), along with differences in clinical improvement, hepatotoxicity, and nephrotoxicity apparently favoring fluconazole.Importantly, as the authors acknowledge, the mortality association did not persist after multivariable adjustment, underscoring the complexity of interpreting observational data in this setting.Conversely, no multivariable analyses were conducted for adjusting associations of empirical antifungal selection with clinical improvement, hepatotoxicity, and nephrotoxicity.Fragility analysis is a complementary methodological tool that evaluates the stability of statistically significant results.The FI estimates the minimum number of outcome reclassifications required to render a significant result nonsignificant, providing an intuitive measure of how sensitive an observed association is to potential misclassification or bias. 2 In observational studies of critically ill patients, where confirmed infection rates are low and subgroup sizes are small, results may be especially fragile.We therefore applied fragility analysis to the key significant findings reported by Albakri et al regarding the favorable association of fluconazole selection with endpoints assessed only in univariable comparisons (clinical improvement, hepatotoxicity, and nephrotoxicity), to better assess the generalization potential of their conclusions.We extracted raw event counts and sample sizes for each statistically significant comparison from the published tables of Albakri et al 1 We constructed 2×2 contingency tables of events and non-events for each antifungal agent pair.The FI was calculated by iteratively simulating changes from non-events to events in the comparator arm until a nonsignificant result was obtained using Fisher's exact test, with type I error set at 0.05.Comparisons that did not achieve statistical significance in the original study were not amenable to fragility analysis.Our analysis yielded an FI of 6 for clinical improvement (fluconazole vs. anidulafungin), 3 for hepatotoxicity (fluconazole vs. anidulafungin), and 2 for both nephrotoxicity comparisons (fluconazole vs. caspofungin; fluconazole vs. anidulafungin).In practical terms, even the most robust finding, clinical improvement with an FI of 6, could be overturned by reclassifying the outcome of just six patients.The safety outcomes were more fragile still: the hepatotoxicity and nephrotoxicity differences could be reversed by changing the outcome status of 3 and 2 patients, respectively.2There are multiple complementary interpretations of these fragile findings.From an epidemiological perspective, low FI values reflect the challenges of detecting between-agent differences in a population where confirmed fungal infection occurred in only 7.41% of patients and subgroup sizes were small, with the fluconazole arm comprising just 38 individuals.Even modest imbalances in baseline severity, plausible in a retrospective non-randomized design, could Infection and Drug Resistance 2026:19 630961 1

Autor:innen und Institutionen

  • Department of Medicine, University of Toronto, Toronto, Ontario, Canada.
  • Department of Health Sciences (DISSAL), University of Genoa, Genoa, Italy.
  • Clinica Malattie Infettive, Azienda Ospedaliera Metropolitana, Genoa, Italy.
  • Centre for Evidence-Based Medicine, University of Oxford, Oxford, UK.

ORCID-Profile: Daniele Roberto Giacobbe, Ronald Chow

Zitieren

APA

Manzoor, S., Giacobbe, D. R., & Chow, R. (2026). Fragility of Observed Associations in Selection of Empirical Antifungal Therapy for Critically Ill Non-Neutropenic Patients [Letter]. Infection and Drug Resistance, Volume 19, 1-3. https://doi.org/10.2147/idr.s630961

Vancouver

Manzoor S, Giacobbe DR, Chow R. Fragility of Observed Associations in Selection of Empirical Antifungal Therapy for Critically Ill Non-Neutropenic Patients [Letter]. Infection and Drug Resistance. 2026;Volume 19:1-3. doi:10.2147/idr.s630961

BibTeX
@article{Manzoor2026,
  author = {Manzoor, Saba and Giacobbe, Daniele Roberto and Chow, Ronald},
  title = {Fragility of Observed Associations in Selection of Empirical Antifungal Therapy for Critically Ill Non-Neutropenic Patients [Letter]},
  journal = {Infection and Drug Resistance},
  year = {2026},
  volume = {Volume 19},
  pages = {1--3},
  doi = {10.2147/idr.s630961}
}
Kostenlos starten →