Febrile neutropenia (FN) is a common oncologic emergency in patients with hematologic malignancies, associated with high morbidity and mortality. Early identification of patients at higher risk of complications such as sepsis or septic shock is critical to optimize antimicrobial management.

This study aims to characterize the human and microbial plasma proteome using high-resolution mass spectrometry to identify biomarker combinations ("combitypes") capable of predicting complications in oncohematologic patients with FN.

A cohort of 350 adult patients with high-risk FN and initially uncomplicated clinical presentation will be enrolled across three tertiary hospitals. Plasma samples will be collected at fever onset (before antibiotic initiation) and after 48 hours. Proteomic data will be integrated with clinical information using multivariate and machine learning models to develop a predictive model for complications.

This multicenter, prospective, observational study will evaluate whether combined proteomic profiles of host and microbial origin can predict complications in patients with hematologic malignancies presenting with high-risk febrile neutropenia (FN).

FN is defined as an oral temperature ≥38.3 °C once or ≥38.0 °C for ≥1 hour in patients with an absolute neutrophil count (ANC) <500 cells/mm³ or expected to decrease below that threshold within 48 hours. Despite empirical broad-spectrum antibiotics, up to 50% of these patients develop sepsis, and 10% progress to septic shock.

Current biomarkers such as C-reactive protein (CRP) or procalcitonin (PCT) have limited specificity in this immunocompromised population. This study proposes a novel integrative proteomic approach based on mass spectrometry to simultaneously quantify host and microbial proteins in plasma, identifying molecular patterns associated with poor outcomes.

Plasma samples (10 mL, EDTA) will be obtained at two time points: the first febrile episode (prior to antibiotic administration) and 48 hours later. Proteins will be processed using PreOmics® ENRICHplus technology and analyzed via LC-MS/MS on an Evosep One-timsTOF Pro2 platform. Differentially expressed proteins will be identified using a data-independent acquisition (DIA-PASEF) workflow and validated in a subset of 200 patients through targeted mass spectrometry.

Clinical, analytical, and microbiological data will be collected via the REDCap platform. Machine learning models (XGBoost, SHAP interpretability) will be used to generate a predictive risk model for complications, integrating proteomic and clinical data.

This study is expected to establish a new decision-support tool for early identification of high-risk FN patients, facilitating personalized antimicrobial strategies and improved prognosis.

Inclusion Criteria:

• Adults (≥18 years).

• Written informed consent provided by patient or legal representative.

• Diagnosis of hematologic malignancy under induction chemotherapy, post-allogeneic hematopoietic stem cell transplantation, or CAR-T therapy.

• High-risk febrile neutropenia (ANC ≤ 100 cells/mm³, expected duration ≥ 7 days, or significant comorbidities).

• Fever defined as oral temperature ≥38.3 °C once or ≥38.0 °C for ≥1 hour.

• Hospitalized or requiring immediate admission at the time of FN diagnosis.

  ´- Initial uncomplicated clinical presentation, with no previous infection or colonization by multidrug-resistant bacteria.

• Eligible for initial monotherapy with broad-spectrum empirical antibiotic.

• Availability for serial plasma sampling and clinical follow-up.

Exclusion Criteria:

• Age <18 years.
• Low-risk FN according to MASCC/CISNE criteria.
• Initial sample collected after antibiotic administration.
• Decline or inability to provide informed consent.
• Any condition preventing safe participation or reliable sample collection.
• Fever induced by noninfectious causes (considered as adjustment factor, not exclusion).

Oncohematologic Patients With Febrile Neutropenia

Following anonymization, individual participant-level data supporting the publications will be made available. These data include demographic and clinical variables collected in REDCap. A data dictionary, eCRF, and-when possible-analysis scripts (R/Python) and FAIR-compliant metadata will also be provided.

Data will be hosted within the controlled infrastructure of the project (XNAT/Core-lab) and, for controlled dissemination, within the Zenodo community of IBSAL, with DOI assignment and regulated access, in accordance with the GDPR/LOPDGDD and the study protocol as approved by the Research Ethics Committee.

Data will be available starting 12 months after publication of the primary results and will remain accessible for at least 5 years thereafter.

Qualified researchers with a methodologically sound proposal may request access. Requests will be reviewed by the study's Data Committee/Steering Committee. Applications should be addressed to the Principal Investigator (Jesús Francisco Bermejo Martín). A Data Use Agreement (DUA) will be required, including commitments to non-reidentification, appropriate data security, and mandatory citation of both the source and dataset DOI. Applicants may also be required to share their analytical code and a publication plan prior to authorization.