Daily Anesthesiology Research Analysis
Three perioperative studies stand out today: a randomized trial shows that adding a five-item care bundle to modern standard care did not reduce surgical site infections after abdominal surgery; a prospective multicenter study using gastric ultrasound in patients on GLP-1 receptor agonists identifies practical withholding and fasting thresholds linked to high residual gastric content; and a multicenter, interpretable multitask model accurately predicts postoperative AKI, respiratory failure, and
Summary
Three perioperative studies stand out today: a randomized trial shows that adding a five-item care bundle to modern standard care did not reduce surgical site infections after abdominal surgery; a prospective multicenter study using gastric ultrasound in patients on GLP-1 receptor agonists identifies practical withholding and fasting thresholds linked to high residual gastric content; and a multicenter, interpretable multitask model accurately predicts postoperative AKI, respiratory failure, and in-hospital mortality using 16 preoperative EHR features.
Research Themes
- Perioperative infection prevention effectiveness
- Aspiration risk management with GLP-1 receptor agonists
- Interpretable AI for perioperative risk stratification
Selected Articles
1. A care bundle added to standard care versus standard care for the prevention of surgical site infections after abdominal surgery (EPO
In a randomized comparison of 1,777 abdominal surgery patients, adding a five-element care bundle to contemporary standard care did not reduce surgical site infections (18.4% vs 18.9%; RR 0.98). Serious adverse events were also similar, suggesting limited incremental benefit of bundles in high-performing settings.
Impact: This high-quality RCT challenges the assumption that care bundles always add value beyond optimized standard SSI prevention, informing resource allocation and quality initiatives.
Clinical Implications: In settings with robust antibiotic prophylaxis and alcohol-based skin prep, additional bundles may not reduce SSI; focus may shift to adherence, context-specific gaps, or novel strategies.
Key Findings
- SSI incidence was 18.4% with the care bundle vs 18.9% with standard care (RR 0.98, 95% CI 0.81–1.18).
- Per-protocol analysis also showed no benefit (RR 0.91, 95% CI 0.60–1.37).
- Serious adverse events were similar between groups (33.3% vs 33.5%; RR 0.99, 95% CI 0.87–1.23).
Methodological Strengths
- Randomized controlled design with large sample (n=1777)
- Clear outcome definitions and intention-to-treat and per-protocol analyses
Limitations
- Conducted in high-income settings with optimized standard care, potentially limiting detectable incremental effects
- The abstract does not specify intervention adherence or implementation fidelity
Future Directions: Targeted interventions addressing context-specific risk factors, auditing adherence, and testing novel SSI prevention strategies (e.g., decolonization, tailored antibiotic protocols) are warranted.
BACKGROUND: Surgical site infections (SSI) are common. We selected five interventions from recent SSI prevention guidelines, to form the Enhanced PeriOperative Care and Health program (EPO METHODS: The EPO FINDINGS: Between March 1st, 2016, and March 26th, 2020, 1777 patients were included. The intervention group included 869 patients (mean age 63.1, 467 female and 402 male) versus 908 in the control group (mean age 64.0, 530 female and 378 male). The incidence of SSI was 18.4% (160/869) in the intervention and 18.9% (172/908) in the control group; relative risk 0.98 (95% CI: 0.81-1.18) in the intention-to-treat analysis and 0.91 (95% CI: 0.60-1.37) in the per-protocol analysis. The percentage of patients with a serious adverse event was 33.3% (289/869) versus 33.5% (304/908), RR 0.99, 95% CI 0.87-1.23. INTERPRETATION: In a high-income health care setting, a care bundle did not lead to a lower incidence of surgical site infections when added to standard care including preoperative systemic antibiotic prophylaxis and alcohol-based surgical skin preparation. Considering the persistent high risk of SSI, research into interventions that may help to reduce this risk remains urgently needed. FUNDING: The Netherlands Organisation for Health Research and Development (ZonMW), and co-financed by Innovatiefonds Zorgverzekeraars, and Ethicon.
2. Assessment of Gastric Content Using Gastric Ultrasound in Patients on Glucagon-Like Peptide-1 Receptor Agonists Before Anesthesia.
In 316 adults taking GLP-1 receptor agonists, 35.8% had high residual gastric content on preoperative gastric ultrasound. For weekly injections, withholding ≤7.5 days and fasting solids ≤21.3 hours were associated with high residual content, providing actionable thresholds to mitigate aspiration risk.
Impact: Provides the first multicenter, prospectively collected thresholds that directly inform fasting/withholding strategies for the rapidly growing population on GLP-1 agonists undergoing anesthesia.
Clinical Implications: Consider extended withholding (beyond ~7.5 days for weekly injections) and longer solid fasting (>~21 hours) or manage as a 'full stomach' if thresholds cannot be achieved; preoperative gastric ultrasound can individualize risk.
Key Findings
- High residual gastric content (RGC) was present in 35.8% (113/316) of GLP-1 RA users.
- For weekly injections, withholding was shorter in high RGC vs low RGC (6 vs 8 days; P=0.003); ROC cutoff ≤7.5 days.
- Shorter solid fasting was associated with high RGC (15.0 vs 20.0 hours; P<0.001); ROC cutoff ≤21.3 hours.
Methodological Strengths
- Prospective, multicenter assessment using standardized point-of-care gastric ultrasound
- Objective primary outcome with ROC-derived thresholds
Limitations
- Observational design limits causal inference
- Potential residual confounding despite assessment of multiple preoperative factors
Future Directions: Prospective interventional trials testing withholding/fasting protocols guided by gastric ultrasound, and validation across different GLP-1 formulations and surgical populations.
BACKGROUND: The use of glucagon-like peptide-1 receptor agonists (GLP-1 RAs) has significantly increased in recent years. GLP-1 RAs delay gastric emptying, resulting in early satiety and weight loss. This may increase the risk of pulmonary aspiration of residual gastric contents (RGC) during anesthesia delivery. Evidence is urgently needed to guide perioperative anesthesia management for patients taking GLP-1 RAs. This study evaluated preoperative factors that may be associated with high RGC. METHODS: Adult patients who were taking GLP-1 RAs and scheduled to receive anesthesia at 3 hospitals between June 30, 2023, and August 15, 2024, were evaluated via preoperative point-of-care gastric ultrasonography (GUS). The primary outcome was high RGC, defined by the presence of solids food or >1.5 mL/kg of clear liquids on GUS, and its association with preoperative factors (eg, existing medical conditions, indication and route of taking GLP-1 RAs, length of taking GLP-1 RAs, days of withholding GLP-1 RAs before surgery, and preoperative fasting periods). Data are presented as median (interquartile range [IQR]). RESULTS: Among the 316 patients (60.9 years [52.1-68.9] of age; 167 [52.8%] females) included in the study, 113 (35.8%) had high RGC. A higher percentage (5.3%; 6/113) of patients in the high RGC group had an opioid prescription for pain management within 3 months of the GUS assessment compared to the low RGC group (1.0%; 2/203; P = .027). No statistical difference was found between the groups in other existing medical conditions, indication and route of taking GLP-1 RAs, and length of taking GLP-1 RAs. Of the 294 patients taking weekly injections, there were 187 (63.6%) with low RGC and 107 (36.4%) with high RGC. Patients with low RGC withheld their GLP-1 RAs for 8 days [5-10], while patients with high RGC withheld for 6 days [3-9] (P = .003). Receiver operating characteristic (ROC) analysis found ≤7.5 days of withholding the medication as cutoff for increased prevalence of high RGC in patients taking GLP-1 RA injections. Patients with low RGC fasted from solid food for 20.0 hours [14.8-40.8], and patients with high RGC fasted from solid food for 15.0 hours [12.8-19.0] (P < .001). ROC found ≤21.3 hours of fasting from solid food as the cutoff for increased prevalence of high RGC. CONCLUSIONS: GLP-1 RA usage may delay gastric emptying. In preoperatively fasting adults, ≤7.5 days of withholding injections and ≤21.3 hours of fasting from solid food are associated with high RGC.
3. Multicenter validation of a scalable, interpretable, multitask prediction model for multiple clinical outcomes.
An interpretable, tree-based multitask model using 16 preoperative EHR features predicted AKI, postoperative respiratory failure, and in-hospital mortality with strong AUROCs across derivation and two external cohorts. Feature contributions were elucidated, supporting transparent, scalable preoperative risk stratification.
Impact: Demonstrates externally validated, interpretable multitask learning for key postoperative outcomes using minimal routinely available data, enabling practical clinical integration.
Clinical Implications: Hospitals can implement transparent preoperative risk tools to identify patients at risk for AKI, PRF, and mortality, enabling targeted optimization and resource planning without complex data pipelines.
Key Findings
- Multitask model achieved AUROCs up to 0.805 (AKI), 0.925 (PRF), and 0.913 (mortality) across derivation and two external cohorts.
- Used only 16 preoperative EHR features, facilitating scalability and implementation.
- Provided interpretability with variable contribution profiles for each predicted outcome.
Methodological Strengths
- External validation across multiple cohorts
- Interpretable tree-based multitask learning with limited, routinely available inputs
Limitations
- Retrospective design; prospective impact on clinical decision-making not assessed
- The abstract does not report cohort sizes or calibration metrics
Future Directions: Prospective implementation studies to assess clinical utility, calibration, and net benefit; expansion to additional outcomes and heterogeneous health systems.
Predicting multiple postoperative complications remains challenging in perioperative care. Current approaches often address complications individually, limiting the potential for integrated risk assessment. We developed and externally validated a scalable, interpretable, tree-based multitask learning model to predict three critical postoperative outcomes-acute kidney injury (AKI), postoperative respiratory failure (PRF), and in-hospital mortality-using 16 preoperative features generally available in electronic health records. Our model achieved AUROCs of 0.805, 0.789, and 0.863 for AKI; 0.886, 0.925, and 0.911 for PRF; and 0.907, 0.913, and 0.849 for mortality in the derivation cohort and external validation cohorts A and B, respectively (all p < 0.001, except for AKI in derivation and PRF in cohort B). We also elucidated the contribution of each input variable to predictions among different outcomes. Our findings highlight the potential of multitask learning to streamline preoperative risk assessment and present a scalable, interpretable, and generalizable framework for improving perioperative care.