Program

An Interpretable Risk-Enrichment Model for Optimizing the Emergency Department Diagnostic Pathway for Sepsis Using Routine Laboratory and Physiological Data

Author: Szuren Javadi¹, Károly Gasteiger², Zsolt Hetesi³, István Papp⁴, Margit Tőkés-Füzesi⁴, Tihamér Molnár⁵

1 Medical School, University of Pécs, Pécs, Hungary
2 Faculty of Business and Economics, University of Pécs, Pécs, Hungary
3 Faculty of Sciences, University of Pécs, Pécs, Hungary
4 Department of Laboratory Medicine, Medical School, University of Pécs, Pécs, Hungary
5 Department of Anesthesiology and Intensive Therapy, Medical School, University of Pécs, Pécs, Hungary

Introduction

Early recognition of sepsis in the emergency department (ED) remains challenging: results for established inflammatory biomarkers (C-reactive protein, procalcitonin [PCT]) are unavailable for hours, and initial clinical signs are frequently nonspecific. Routine PCT testing in all patients is not cost-effective; a targeted, early trigger derived from complete blood count (CBC) and vital-sign data is needed to identify high-risk patients before biomarker results return.

Methods

We retrospectively analyzed 2,980 ED patients (Szigetvár, Hungary) to derive and validate a six-parameter additive risk score based on laboratory and physiological parameters available during the early phase of patient care; each threshold met added one point (range 0–6). Monocyte distribution width (MDW) was evaluated as an optional seventh parameter (range 0–7) where available. Discrimination was benchmarked against a cross-validated logistic-regression reference model built on the same parameter set.

Results

The deployed additive score achieved an area under the receiver-operating-characteristic curve (AUC) of 0.800 without MDW and 0.807 with MDW; the logistic-regression reference model reached AUC 0.798 and 0.849, representing the discriminative ceiling for this parameter set. A score ≥3 (without MDW) enriched sepsis prevalence from 9.9% to 25.2% (negative predictive value [NPV] 96.0%); a score ≥4 (with MDW) enriched prevalence from 9.3% to 18.4% (NPV 97.8%).

Conclusion

An analyzer-independent score based on routinely available CBC parameters and vital signs, without requiring specialized biomarkers, may enrich the prevalence of sepsis among flagged ED patients and could provide a low-cost trigger for targeted PCT testing. The optional addition of MDW may further improve discrimination and negative predictive value (NPV), although MDW is not required for the baseline model. The diagnostic utility and generalizability of this approach require prospective, multicentre validation before clinical implementation.

(Disclosure: artificial intelligence was used in preparing this abstract)

Explainable Deep Learning for Multi-Horizon Sepsis Prediction: A Multicentre Retrospective Validation Study

Authors: Ioannis Gkouzionis, Charikleia Gkrantouni, Carlos Ferrando, Guido Muñoz, Xavi Borrat, Ramses Marrero, Savvas Vlachos, Petros Mitseas, Sotirios Liliopoulos, Vasileios Dimakopoulos

Introduction

Sepsis accounts for more than 11 million deaths annually and remains a leading cause of preventable ICU mortality [1]. Although artificial intelligence (AI) approaches have shown promise in leveraging electronic health records for earlier prediction, many existing models are limited by poor generalizability, inadequate handling of temporal clinical data, and lack of interpretability [2]. Current clinical screening tools, including the National Early Warning Score 2 (NEWS2) and quick Sequential Organ Failure Assessment (qSOFA), demonstrate limited discriminative performance for early sepsis prediction [2, 3]. Furthermore, the heterogeneity and complexity of real-world clinical data across institutions pose significant challenges for reliable deployment, as models frequently underperform when applied outside their development setting, limiting their clinical adoption.

Objectives

This study aimed to develop and validate an explainable deep learning model for multi-horizon sepsis risk prediction using multimodal clinical data. Robustness and generalisability were assessed by training and evaluating the model on publicly available datasets and proprietary institutional data from Hospital Clínic de Barcelona (HCB). Model performance was benchmarked against standard-of-care scoring systems and published commercial AI approaches, with clinically stratified, interpretable risk outputs designed to support real-world clinical decision-making.

Methods

We conducted a retrospective, multicenter study using de-identified patient data from two publicly available clinical databases, MIMIC-IV and AmsterdamUMCdb (AUMC), and a proprietary institutional repository at Hospital Clínic de Barcelona (HCB). Sepsis was defined per Sepsis-3 criteria [1]. Time-series vital signs, laboratory measurements, and static patient characteristics, including demographics and comorbidities, were segmented into fixed-length temporal windows using a sliding window approach, with prediction targets defined at 3, 6, 12, 24, and 48 hours horizons. The model (Sepsis Prediction Index, SPI®) was developed using a hybrid convolutional-transformer architecture, in which convolutional layers capture local temporal features and self-attention mechanisms model long-range temporal dependencies across the clinical time-series. Model outputs were stratified into four clinically actionable risk categories (low, moderate, high, very high). Model development was assessed via 5-fold patient-level cross-validation; The final performance evaluation was conducted on a held-out dataset using bootstrapped confidence intervals (200 runs, 30% subsampling) to assess generalizability. Subgroup analyses were conducted across age, sex, comorbidity and institution groups. Performance was benchmarked against qSOFA, NEWS2, and published AI models. SHapley Additive exPlanations (SHAP) feature attribution provided global and patient-level interpretability.

Results

The development cohort comprised 60,429 total patients (34,996 male, 7,468 septic, median age = 65 years, sepsis prevalence = 12.3%) across 3 institutions, entailing patient admissions in emergency department, ward, and ICU. The held-out validation cohort comprised 15,120 patients (8,747 male, median age = 65 years, sepsis prevalence = 11.1%). On the combined held-out validation cohort, the model achieved AUROC 89.7% (95% CI [88.2 – 91.2%]). On the HCB dataset, AUROC was 81.7%, PPV 78%, specificity 82.0% and sensitivity 65.0%. SPI® risk bands were associated with a monotonic increase in sepsis prevalence (p-value < 0.001, Cohran-Armitage test). The SPI® demonstrated a median lead time of 16.4 hours (mean 20.6 hours) prior to clinical sepsis recognition, substantially exceeding qSOFA (AUROC 54.8%; median and mean lead time 0.0 hours) and NEWS2 (AUROC 62.3%; median lead time 0.0 hours and mean lead time 8.1 hours). SPI® outperformed all published state-of-the-art AI comparators across all prediction horizons.

Conclusion

Trained and validated across heterogeneous clinical datasets spanning three international institutions, this hybrid convolutional-transformer model demonstrates robust sepsis prediction up to 48 hours before clinical recognition, substantially outperforming standard-of- care scoring systems and published AI models. Clinically stratified, SHAP-interpretable risk outputs are designed to support, rather than replace, bedside clinical judgement. Prospective evaluation and clinical workflow integration are underway.

Blood Culture Positivity and ICU Mortality in Septic Shock: A Multivariable Analysis

Authors: Pieter Vanwalleghem, L. De Bus, P. Depuydt

Introduction

Blood cultures are positive in a substantial proportion of patients with sepsis, but the prognostic significance of culture positivity in septic shock remains uncertain. We investigated the association between blood culture positivity and ICU mortality in patients with septic shock.

Methods

This single-center retrospective study used the COSARA dataset, comprising all patients admitted to the intensive care department of a tertiary university hospital in Ghent, Belgium, between 2013 and 2022. Adult patients with an episode of septic shock were identified and categorized according to blood culture status (positive vs negative blood culture). The primary outcome was ICU mortality. Multivariable logistic regression was performed to assess the independent association between blood culture status and ICU mortality, adjusting for relevant patient and disease characteristics.

Results

A total of 1,740 patients were included with the clinical suspicion of septic shock, of whom 541 (31%) had positive and 1,199 (69%) had negative blood cultures. Baseline characteristics were largely comparable, although APACHE II scores were higher in culture-negative patients (28.0 vs 26.0; p = 0.005), and infection foci differed significantly between groups (p < 0.001). Respiratory and fungal infections were more frequent among culture-negative patients (39% versus 14% and 5.3% versus 0.2% respectively), whereas intravascular/catheter-related and urogenital infections were more common among culture-positive patients (10% vs 0.6% and 14% vs 5.6% respectively). Hospital and ICU mortality (45% and 36%) did not differ significantly between groups. Culture-negative patients had a significant longer ICU stay (7 vs 5 days), a significant higher use of mechanical ventilation (78% vs 62%) and a significant greater length of vasopressor therapy (2-7.5 days vs 2-7 days). In multivariable analysis, blood culture status was not independently associated with ICU mortality (OR 0.90, 95% CI 0.70–1.17). In contrast, increasing age (OR 1.02 per year, 95% CI 1.01–1.03; p < 0.001), higher APACHE II score (OR 1.08 per point, 95% CI 1.07–1.10; p < 0.001), and infection focus (p < 0.001) were independently associated with ICU mortality. Urogenital, abdominal, and skin/soft tissue infections were associated with lower odds of mortality compared with respiratory infections (OR 0.17, 0.59, and 0.57, respectively).

Conclusion

Blood culture positivity was not independently associated with ICU mortality in patients with septic shock. Instead, mortality was associated with age, severity of illness, and infection focus. These findings suggest that blood culture positivity should not be interpreted as a prognostic marker in septic shock.

Invasive hemodynamic monitoring-guided resuscitation improves survival in shock: A systematic review and meta-analysis

Authors: Liliána Nagy^1,2,, Petra Réka Tóth^1,2, Caner Turan^1,2, Dávid Laczkó^1,3, Lili Légár¹, Basak Danisan¹, Zoltán Sipos^1,4, Péter Hegyi^1,4,5, Zsolt Molnár^1,4,6, László Zubek^1,2

1 Centre for Translational Medicine, Semmelweis University, Budapest, Hungary
2 Department of Anesthesiology and Intensive Therapy, Semmelweis University, Budapest, Hungary
3 Department of Interventional Radiology, Heart and Vascular Centre, Semmelweis University, Budapest, Hungary
4 Institute for Translational Medicine, Medical School, University of Pécs, Pécs, Hungary
5 Institute of Pancreatic Diseases, Semmelweis University, Budapest, Hungary
6 Department of Anaesthesiology and Intensive Therapy, Poznan University of Medical Sciences, Faculty of Medicine, Poznan, Poland

Background

Mortality in shock may reach 60%; hence, immediate, adequate resuscitation has a crucial role in improving outcomes. Detailed hemodynamic monitoring is desirable, but evidence on its outcome benefits is limited. Therefore, this study aimed to compare advanced hemodynamic monitoring (AHDM)-guided clinical decision-making and treatment with conventional ones in terms of outcomes in shock.

Methods

A systematic search was performed in three databases (PubMed, EMBASE, and Cochrane Library) until 9 November 2024. Randomised controlled trials, non-randomised and observational studies involving adult shock patients were eligible for inclusion. Main outcomes were in-hospital and 30-day mortality, and secondary outcomes included length of stay, need for and duration of organ support, and amount of fluid administered. Meta-analyses were performed using a random-effects model, with heterogeneity and risk of bias assessed. The review protocol was registered in PROSPERO (ID: CRD42024607758).

Results

A total of 34 studies, including seven RCTs and 636,441 shock patients, were analysed to compare AHDM with conventional monitoring. The use of any type of AHDM was associated with a significantly lower in-hospital mortality for any type of shock (OR: 0.66; 95% CI: [0.48; 0.91], p = 0.014), especially in cardiogenic shock patients, managed with pulmonary artery catheter-guided therapy (OR: 0.68, 95% CI: [0.60; 0.78], p < 0.001). This contrasts with the significantly higher odds of requiring organ support treatment, including inotropes (OR: 2.32, CI: [1.29; 4.19], p = 0.012), vasopressors (OR: 1.46, CI: [1.05; 2.04], p=0.030), mechanical circulatory support (OR: 2.85, CI: [1.62; 5.02], p=0.002), renal replacement therapy or mechanical ventilation, although the duration of mechanical ventilation was shorter. Heterogeneity was predictably high due to the variety of shock types and monitoring methods. Risk of bias was predominantly low in RCTs and serious in observational studies.

Conclusion

AHDM use is associated with a significant reduction in mortality in shock patients, with the greatest benefit observed in cardiogenic shock. The observed outcomes suggest that AHDM may facilitate qualitative changes in decision-making, consistent with precision-guided resuscitation.

Exploring the Impact of Tissue Perfusion Guided Resuscitation in Shock: A Systematic Review and Meta-Analysis

Authors: Tamás Tóth^2,7, Patricia Schneidereit², Julia Hollosi², Dávid Lackó^2,8, Bence Szabó², Daniel Louis Albert², Caner Turan^2,5, László Zubek^2,5, Péter Hegyi^1,2,3,4, Zsolt Molnar^2,5,6

1 Institute for Translational Medicine, Medical School, University of Pécs, Pécs, Hungary
2 Centre for Translational Medicine, Semmelweis University, Budapest, Hungary
3 Institute of Pancreatic Diseases, Semmelweis University, Budapest, Hungary
4 Division of Gastroenterology, First Department of Medicine, Medical School, University of Pécs, Pécs, Hungary
5 Department of Anesthesiology and Intensive Therapy, Semmelweis University, Budapest, Hungary
6 Department of Anesthesiology and Intensive Therapy, Poznan University of Medical Sciences, Poznan, Poland
7 Department of Anaesthesiology and Intensive Therapy, Bajcsy-Zsilinszky Hospital, Budapest, Hungary
8 Department of Interventional Radiology, Heart and Vascular Centre, Semmelweis University, Budapest, Hungary

Introduction

Resuscitation from shock generally targets macrohemodynamic parameters. However, this approach may not improve tissue perfusion and may increase the risk of fluid overload. We aimed to evaluate the efficacy of tissue perfusion guided therapy (TP-GT) compared with standard care in adult patients treated with shock.

Methods

Following the PRISMA 2020 guidelines, we searched MEDLINE, Embase, and CENTRAL on October 26, 2025, for randomized controlled trials (RCTs) comparing TP-GT to standard therapy. The protocol was prospectively registered on PROSPERO (CRD420251163043). The primary outcomes were 30-day and 90-day mortality rates. Secondary outcomes included intensive care unit length of stay (ICU LOS), hospital length of stay (HLOS), fluid administration volumes, fluid balance, and organ support requirements. Meta-analyses were conducted using random-effects model, and the risk of bias was assessed using the RoB2 tool.

Results

Eight RCTs comprising 2,394 patients were analysed. TP-GT did not significantly reduce 30-day (Risk Ratio [RR] 0.96; 95% confidence interval [CI] 0.83–1.10; P = 0.49) or 90-day mortality (RR 0.94; 95% CI 0.82–1.07; P = 0.17) compared with standard care. However, patients managed with TP-GT had a significantly shorter ICU LOS (Mean Difference [MD] = -0.72 days; 95% CI = -1.42 to -0.01; P = 0.048). Furthermore, TP-GT significantly reduced the volume of fluid administered until the end of the first 6–8 hours of resuscitation (MD -466.95 mL; 95% CI -917.17 to -16.72; P=0.046).

Conclusion

TP-GT supports personalised fluid stewardship by significantly reducing early fluid administration and accelerating clinical stabilisation, potentially resulting in a significantly shorter ICU length of stay.

Reference:

Tóth T, Schneidereit P, Hollosi J, Lackó D, Szabó B, Albert DL, Turan C, Zubek L, Hegyi P,

Molnar Z. The effectiveness of tissue-perfusion-guided resuscitation in shock: A systematic

review and meta-analysis. Ann Intensive Care. 2026 Jun 25;16:100106. doi:10.1016/j.aicoj.2026.100106

Blood purification is associated with improved organ function in sepsis: a systematic review and meta-analysis

Authors: Julia Hollosi¹, Caner Turan^1,2, Anna Walter³, Péter Hegyi^1,3,4, Péter Bakos^1,5, László Zubek^1,2, Dávid Laczkó¹, Cem Alkan¹, Karin May Koch¹, Momen Elkheshen¹, Mustafa Al Majali¹, Zsolt Molnár^1,2,6

1 Centre for Translational Medicine, Semmelweis University, Budapest, Hungary
2 Department of Anesthesiology and Intensive Therapy, Semmelweis University, Budapest, Hungary
3 Institute for Translational Medicine, Medical School, University of Pécs, Pécs, Hungary
4 Institute of Pancreatic Diseases, Semmelweis University, Budapest, Hungary
5 Department of Anesthesiology and Intensive Therapy, Csolnoky Ferenc Hospital, Veszprém, Hungary
6 Department of Anesthesiology and Intensive Therapy, Poznan University for Medical Sciences, Poznan, Poland

Objective:

Sepsis remains a leading cause of mortality in intensive care units (ICUs), hence blood purification therapies such as plasmapheresis, hemadsorption, or polymyxin B adsorption have been proposed as adjunctive treatments to standard care, but clear evidence on their effects is missing. Our aim was to compare these different therapies in sepsis, with a network meta-analysis.

Data Sources:

A systematic search was performed on June 7, 2026 in three major databases: PubMed, EMBASE, and Cochrane.

Study Selection:

Only randomized controlled trials (RCTs) of adult septic patients receiving blood purification therapies reporting changes in organ dysfunction scores or mortality were included. 

Data Extraction:

Out of 1665 identified articles, 30 RCTs met inclusion criteria.

Data Synthesis:

Comparisons were performed using a frequentist network meta-analysis approach, applying a random effects model to account for the considerable between-study heterogeneity observed among the included studies. All analyses were conducted using the netmeta (v2.9-0) and dmetar (v0.1.0) packages in R (v4.4.2), with the confidence level set at 95%. Results were visualized via forest plot and surface under the cumulative ranking curve (SUCRA) plots. All the blood purification therapies showed a significant improvement in the SOFA score compared to standard therapy, and reached statistical significance as well. In addition, plasmapheresis showed a significant odds ratio of 0.49 (95% CI: 0.25 to 0.96) in 28-day mortality, indicating it to be a beneficial treatment compared to standard therapy, while other therapies showed a tendency toward improved mortality odds as well.

Conclusions:

Our findings suggest possible advantages of adjunctive blood purification therapies in sepsis management. Therefore, although more trials are required to determine the optimal patient population and timing-dosage strategy, doctors should consider adding these adjunctive therapies to standard therapies for septic patients.

Funding:

Supported by the Semmelweis University Research, Development and Innovation Fund.

Beyond Haemoglobin: Red Blood Cell Transfusion Practices and Physiological Transfusion Triggers in Hungarian Intensive Care Units

Authors: Szilvia Mukk¹, Csanád Geréd¹, Orsolya Czuth¹, Dorottya Csillag¹, Angéla Mikó¹, Zoltán Ruszkai^1,2

1 Department of Anaesthesiology and Intensive Therapy, Flór Ferenc Hospital Kistarcsa, Kistarcsa, Hungary
2 Department of Anaesthesiology and Intensive Therapy, Semmelweis University, Budapest, Hungary

Introduction

Red blood cell (RBC) transfusion remains a common supportive therapy in critically ill patients. Although haemoglobin concentration traditionally guides transfusion, physiological parameters reflecting oxygen delivery and tissue perfusion may support more individualized decision-making.

Methods

We conducted a nationwide questionnaire-based survey among Hungarian intensive care unit (ICU) physicians between July and September 2025 to characterize current RBC transfusion practices, with particular emphasis on physiological transfusion triggers and patient blood management (PBM).

Results

Ninety-seven ICU physicians completed the survey. Physiological parameters beyond haemoglobin were used to support transfusion decisions by 93.8% of respondents. The most frequently reported indications were active bleeding (88.7%), low haemoglobin concentration (83.5%), haemodynamic instability (75.3%), and postoperative anaemia (75.3%). The preferred haemoglobin threshold was 70-90 g/L in 45.4%, while 38.1% reported a threshold below 70 g/L. Post-transfusion targets frequently included lactate <2 mmol/L (52.8%) and central venous oxygen saturation of 70–72% (51.7%). A single-unit transfusion strategy was preferred by 35.1%, whereas 52.6% routinely administered two units. Only 33.0% reported the availability of written institutional PBM protocol.

Conclusion

Hungarian intensivists frequently incorporate physiological parameters into RBC transfusion decision-making, indicating a shift beyond haemoglobin alone. However, heterogeneity in transfusion thresholds, physiological targets, and unit-based strategies, together with the limited implementation of structured PBM programmes, highlights the need for further education and research to promote a more individualized, physiology-guided approach to transfusion practice.

Diaphragm-Targeted Neuromuscular Modulation in Mechanically Ventilated ICU Patients: A Pairwise Meta-Analysis of Randomized Controlled Trials

Authors: Patricia Schneidereit, Tamás Tóth, Julia Hollosi, Dávid Laczkó, Anna Walter, Péter Hegyi, László Zubek, Zsolt Molnár

Background

Ventilator-induced diaphragmatic dysfunction (VIDD) is associated with weaning difficulty, increased ICU morbidity and mortality. Although neuromuscular modulation (NMM) has been introduced to preserve respiratory muscle function, research has concentrated predominantly on peripheral muscle stimulation, leaving evidence on diaphragm-targeted NMM scarce. This study evaluates the effectiveness of diaphragm-focused NMM in reducing the duration of mechanical ventilation (DMV) and ICU length of stay (ILOS).

Methods

A systematic search of PubMed, Embase, and Cochrane Central Register of Controlled Trials was conducted from inception to November 2, 2025, and updated on September 4, 2026, to identify randomized controlled trials (RCTs) comparing any technique of NMM with standard of care in invasively mechanically ventilated adult patients. Primary outcome was DMV, with ILOS assessed as a secondary outcome. Risk of bias was assessed using the Cochrane RoB 2 tool and the certainty of evidence was evaluated with the GRADE approach. A meta-analysis was conducted using a random-effects model. The review was prospectively registered in PROSPERO (CRD420251164793).

Results

The search identified 740 records, of which 8 RCTs (n=634) met the eligibility criteria; 7 RCTs comparing NMM with standard of care were included in the meta-analysis. Diaphragm-targeted NMM was associated with a significant reduction in DMV (MD -1.52 days [95% CI: -2.52; -0.52], p = 0.011, I2 = 29%) and ILOS (MD -1.36 days [95% CI: -1.93; -0.79], p < 0.001, I2 = 0%) compared with standard of care.

Conclusion

Current very low-certainty evidence suggests that diaphragm-targeted neuromuscular modulation may reduce the duration of mechanical ventilation and ICU length of stay. However, heterogeneity in stimulation approaches, methodological limitations of the available trials, and uncertainty regarding the optimal population and stimulation protocol mean that routine clinical use cannot yet be recommended, and larger, high-quality trials are needed to confirm these findings.

This study was supported by the Semmelweis University Research, Development and Innovation Fund.

ElecTrical Impedance TomogRaphy vs dynamic compliANce guided positive end-expiratory pressure Titration during laparoscopic gynaecological surgery (TITRANT)

Authors: Csanád Geréd¹, Csenge Szigetváry², Angéla Mikó¹, Gergely Péter Bokrétás¹, András Lovas³, Robin Forgács³, Mónika Nemes³, Zsolt Molnár², Zoltán Ruszkai^1,2

1 Flór Ferenc Hospital Kistarcsa, Department of Anaesthesiology and Intensive Therapy
2 Semmelweis University, Department of Anaesthesiology and Intensive Therapy, Budapest
3 Kiskunhalas Semmelweis Hospital the Teaching Hospital of the University of Szeged

Introduction

Intraoperative mechanical ventilation can contribute to ventilator-induced lung injury and postoperative pulmonary complications. Mechanical power integrates the principal ventilatory variables determining energy delivery to the respiratory system and may therefore provide a comprehensive measure of potentially injurious ventilation. Individualized positive end-expiratory pressure (PEEP) may reduce mechanical power by improving respiratory mechanics; however, the optimal method for PEEP titration remains uncertain. Electrical impedance tomography (EIT) enables continuous assessment of regional ventilation and simultaneous estimation of alveolar overdistension and collapse, whereas dynamic respiratory system compliance (Cdyn) provides a simpler bedside approach. The TITRANT trial will investigate whether EIT-guided PEEP titration reduces mechanical power more effectively than Cdyn-guided titration during laparoscopic gynaecological surgery.

Methods and analysis

TITRANT is an investigator-initiated, multicentre, two-arm, parallel-group, patient-blinded, superiority randomized controlled trial. Two hundred adults undergoing elective laparoscopic gynaecological surgery will be randomly allocated 1:1 to EIT-guided or Cdyn-guided decremental PEEP titration after induction of anaesthesia, establishment of pneumoperitoneum, and 35° Trendelenburg positioning. In both groups, PEEP will be decreased from 18 to 6 cmH2O in 2 cmH2O steps. In the EIT group, optimal PEEP will be defined as the level providing the best balance between regional overdistension and alveolar collapse; in the Cdyn group, it will be defined as the PEEP level associated with the highest dynamic respiratory system compliance. All participants will otherwise receive standardized lung-protective ventilation. The primary outcome is intraoperative mechanical power, assessed after PEEP titration and at the end of surgery. Secondary outcomes include oxygenation, respiratory mechanics, EIT-derived global inhomogeneity and regional overdistension and collapse, postoperative pulmonary complications, length of hospital stay, adverse events, and in-hospital and 28-day mortality. Analyses will follow the intention-to-treat principle.

Ethics and dissemination

The study was approved by the Hungarian Scientific and Medical Research Council Ethics Committee (NNGYK/21074-6/2025). Written informed consent will be obtained from all participants. Findings will be submitted for publication in a peer-reviewed journal and presented at international scientific meetings.

Epidural Analgesia in Severe Acute Pancreatitis: Effects on Opioid Use, Intra- abdominal Pressure and Early ICU Outcomes

Authors: Dilán Márk Karim¹, F. K. Gaál¹, N. Szabo¹

1 Department of Anaesthesiology and Intensive Therapy, Semmelweis University, Budapest, Hungary

Introduction

Severe acute pancreatitis remains a challenging critical illness with limited treatment options beyond supportive care. Epidural analgesia may improve pain control, intra-abdominal pressure (IAP) and gastrointestinal function. We assessed its association with systemic opioid use and physiological outcomes.

Methods

This retrospective cohort included patients with severe acute pancreatitis admitted to the ICU of Semmelweis University between January 2023 and December 2025. Patients receiving epidural analgesia (EDA+) were compared with those managed without EDA. First-week data were collected in REDCap.

Results

Fifty-one patients were included: 31 EDA+ and 20 EDA-. Opioid use did not change significantly around EDA initiation, with median morphine-equivalent doses of 13.25 mg before, 15.00 mg on the first EDA day and 24.42 mg the following day (p=0.667). EDA was not associated with improved mobilisation, bowel function or lower IAP. Higher IAP was weakly associated with lower daily urine output (Spearman rho=-0.192, p=0.0019) while no association between EDA and IAP was identified. Sedation was lighter in the EDA+ group by day 7 (median RASS 0 vs -2, p=0.034). Bupivacaine exceeded the recommended daily dose limit on 47.9% of treatment days whereas this did not occur with ropivacaine.

Conclusion

Epidural analgesia was not associated with lower opioid use or improvement in IAP, mobilisation or bowel function during the first ICU week. The association between higher IAP and lower urine output supports its physiological relevance but was not related to EDA use. Lighter day 7 sedation may indicate a modest favourable analgesia-sedation effect. Interpretation is limited by the retrospective design and likely differences in illness severity. Ropivacaine appeared safer than bupivacaine regarding recommended daily dose limits.

AI disclosure: Grammarly assisted language editing was used in preparing the abstract.

Procalcitonin Has Fair Accuracy in Diagnosing Postoperative Infections in Pediatric Congenital Heart Surgery – A Systematic Review and Meta-Analysis

Authors: Ildiko Szomoru, Balazs Fule, Maren Romundstad, Krisztina Csőke-Kabai, Dávid Laczkó, Bence Szabo, Péter Hegyi, Caner Turan, László Zubek, Zsolt Molnar

Introduction

Children undergoing surgery for congenital heart disease are particularly vulnerable to postoperative infections that is one of the leading causes for adverse outcome. Procalcitonin  (PCT) is one of the most frequently used biomarkers in intensive care unit (ICU) patients but its practicality in this population is still undetermined. Thus, the aim of our study was to investigate the accuracy of PCT in diagnosing postoperative infections in children post cardiac surgery.

Methods

We performed our search on November 24, 2025 in Medline, Embase and CENTRAL, following PROSPERO registration (CRD420251231843) and included articles on pediatric patients undergoing cardiac surgery with the use of cardiopulmonary bypass (CPB) providing data on diagnostic accuracy of PCT in indicating bacterial culture proven infection and comparative PCT values of infected and non-infected patient groups. Sensitivity and specificity were analyzed using a bivariate binomial generalized linear mixed model, while AUCs were pooled using generic inverse-variance method and for absolute values median difference (MedD) was the effect measure. Risk of bias was assessed using QUADAS-2.

Results

21 prospective and retrospective observational studies were included in the systematic review and 5570 patients in the meta-analysis. The pooled sensitivity of 0.67 (0.57-0.76) and specificity of 0.85 (0.79-0.89) for the third postoperative day were established for substantially different cutoff values between studies, while the combined AUC was 0.79 (0.70-0.88). Regarding MedD analysis, we found a significant difference between PCT values obtained in the late postoperative period from patients clinically suspected of infection and those confirmed by bacterial cultures.

Conclusion

Procalcitonin indicated the presence of infection with fair accuracy in postoperative CHD patients, thus on its own cannot be recommended to diagnose infection in children after congenital heart surgery.

Timing of hypothermic temperature control does not affect neurological outcomes after cardiac arrest: a systematic review and meta-analysis

Authors: Krisztina Csőke-Kabai^1,2, Zsolt Molnár^1,3,4, László Zubek^1,3, Dávid Laczkó^1,5, Caner Turan^1,3, Péter Hegyi^1,6,7, Emőke Henrietta Kovács¹, Gábor Nagy^1,8, Sude Yilmaz¹, Zoltán Sipos¹, Krisztián Tánczos^1,3

1 Centre for Translational Medicine, Semmelweis University, Budapest, Hungary
2 Emergency Department, University of Szeged, Szeged, Hungary
3 Department of Anaesthesiology and Intensive Therapy, Semmelweis University,Budapest, Hungary
4 Department of Anaesthesiology and Intensive Therapy, Poznan University for Medical Sciences, Poznan, Poland
5 Department of Interventional Radiology, Heart and Vascular Centre, Semmelweis University, Budapest, Hungary
6 Institute for Translational Medicine, Medical School, University of Pécs, Pécs, Hungary
7 Institute of Pancreatic Diseases, Semmelweis University, Budapest, Hungary
8 Department of Emergency Medicine, Semmelweis University, Budapest, Hungary

Introduction

Out-of-hospital cardiac arrest is associated with a low rate of favourable neurological recovery, estimated at 8%. Lowering the cerebral metabolic rate with hypothermia is hypothesized to improve this outcome. Since in-hospital temperature control is recommended to start as early as possible, we investigated whether initiating earlier hypothermic temperature control in the prehospital setting could further enhance neurological recovery.

Methods

A pre-registered systematic search (PROSPERO: CRD42024601051) was run in PubMed, Embase and Cochrane Central on November 3, 2024, to identify randomised controlled trials that compare prehospital vs. in-hospital initiation of hypothermic temperature control of out-of-hospital cardiac arrest patients. The Cochrane Handbook was followed for data extraction and synthesis. The RoB 2 tool was used for risk of bias assessment. The meta-analyses were conducted using a random-effects model.

Results

A total of 7117 articles were screened, and 7 studies were selected for the analysis containing the data of 3288 patients. The intervention group achieved the target temperature of 33 °C significantly earlier (MD: -97.27 min; 95% CI: -139.46 to -55.08; p=0.010). In the intervention group, no significant difference was observed in the odds for favourable neurology (OR: 0.98; 95% CI: 0.80-1.22, p=0.855). The odds for recurrent cardiac arrest and pulmonary oedema were significantly higher in the intervention group (OR: 1.33; 95% CI: 1.20-1.47; p=0.009 and OR: 1.66; 95% CI: 1.24-2.23; p=0.001).

Conclusion

Despite reaching the target temperature significantly earlier, prehospital initiation of hypothermic temperature control did not improve favourable neurological recovery compared to in-hospital initiation. Furthermore, early initiation also significantly increased the odds for recurrent cardiac arrest and pulmonary oedema. Therefore, these findings do not support the implementation of prehospital hypothermic temperature control in out-of-hospital cardiac arrest, however, findings indicate that adverse events may depend on cooling method.