Determinants for antibiotic therapy less or more than 7 days in pediatric oncology patients with bacteremia: a global multicenter retrospective study

Abstract

Randomized trials support the non-inferiority of short-course (≤7 days) antibiotic therapy for bacteremia; however, these studies commonly exclude immunocompromised patients. Shorter regimens may reduce drug toxicity while optimizing resource utilization and patient quality of life. Evidence to guide treatment duration in immunocompromised children remains limited. We conducted a global, multicenter retrospective study of bacteremia in children with cancer ≤18 years old treated between April 1, 2022, and March 31, 2023. Episodes were categorized based on treatment duration (≤7 vs. >7 days). Data collected included oncologic diagnosis, treatment phase, isolated pathogens, management strategies, antimicrobial susceptibility, and outcomes, including all-cause and infection-attributable mortality. Data were captured in REDCap Cloud and analyzed using descriptive statistics and mixed-effects logistic regression. Eighteen centers contributed 1,123 bacteremia episodes from 1,006 patients. Most patients had hematologic malignancies (n=743, 74%). Median absolute neutrophil count (ANC) at onset was 133 cells/mm3; 33% of episodes occurred with ANC >500 cells/mm3. Gram-negative organisms caused 75% of episodes; carbapenem resistance was detected in 23.3% of these. Escherichia coli (n=289), Klebsiella pneumoniae (n=174), Pseudomonas aeruginosa (n=124) and Staphylococcus aureus (n=95) were the most commonly isolated pathogens. Median antibiotic duration was 10 days, and 359 episodes (32%) received ≤7 days of therapy. The isolated organism was susceptible to empiric therapy in 55% of episodes, and 14% (n=144) of patients died. Infection contributed to 81% (n=117) of deaths. In the adjusted mixed-effects logistic regression model, episodes in hematopoietic stem cell transplant (HSCT) recipients were more likely to be treated >7 days compared with non-HSCT patients (OR 2.27, 95% CI 1.09–4.72). Compared with episodes beginning at ANC <100 cells/mm3, those with ANC 100–500 and >500 cells/mm3 had significantly reduced odds of therapy >7 days. Gram-negative infections (OR 1.92, 95% CI 1.38–2.67) were more likely than Gram-positive infections to be treated >7 days. Episodes where the patient died had 71% lower odds of receiving >7 days of therapy (OR 0.29, 95% CI 0.19–0.45). Carbapenem-resistant infections were strongly associated with mortality (OR 3.03, 95% CI 2.04–4.35; p < 0.001). In this large global cohort of immunocompromised children with bacteremia, antibiotic duration was strongly associated by neutrophil count and pathogen type. Pathogens were often resistant to empiric therapy, and carbapenem resistance strongly predicted death. These findings highlight the need to improve antimicrobial resistance detection and management, and to identify patients who may safely receive shorter treatments.