Anesthesiology, October 20 edition

This systematic review and meta-analysis included 12 studies, with nine contributing 22 validation datasets. Deep-learning models achieved pooled sensitivity of 0.78, specificity of 0.88, and SROC-AUC of 0.87 for predicting intraoperative hypotension; hybrid convolutional neural network–recurrent neural network architectures and 5-minute prediction windows performed most favorably. However, the absence of formal calibration assessment, limited external validation, and geographic concentration of studies constrain immediate clinical implementation.

Impact: This paper provides the first unified diagnostic-accuracy synthesis of deep-learning models for intraoperative hypotension prediction and identifies the gap between discrimination performance and clinical readiness. It is particularly valuable because it explicitly incorporates methodological quality and certainty-of-evidence assessment rather than relying solely on pooled accuracy.

Clinical Implications: Deep-learning systems may support earlier recognition and treatment of intraoperative hypotension, but they should currently be regarded as decision-support tools rather than autonomous alarms. Prospective, geographically diverse validation with calibration, clinically meaningful thresholds, workflow integration, and patient-outcome assessment is needed before guideline-level adoption.

Future Directions: Future studies should use prospective multicenter designs, preregistered analytical protocols, independent external validation, calibration assessment, fairness analyses across patient subgroups, and trials evaluating whether model-guided intervention improves organ injury and mortality outcomes.

This retrospective observational cohort study evaluated older adults undergoing elective or emergency surgery under general anesthesia from 2022 to 2025. After 1:2 propensity-score matching, the cohort included 165 patients with GLIM-defined malnutrition and 330 without malnutrition; postoperative delirium occurred in 25.5% versus 11.2%, respectively. The findings support GLIM-based nutritional assessment as a clinically useful component of perioperative delirium risk stratification.

Impact: The study connects a globally standardized malnutrition framework with a major anesthetic and geriatric outcome, postoperative delirium. Its findings suggest that nutritional assessment can be incorporated into routine preoperative risk assessment and may identify a modifiable vulnerability domain.

Clinical Implications: Older surgical patients meeting GLIM malnutrition criteria may warrant intensified delirium prevention, medication review, early mobilization, sleep optimization, sensory support, and nutritional intervention. The association does not establish that nutritional treatment prevents delirium, so GLIM-guided management should be tested prospectively.

Future Directions: Prospective multicenter studies should validate GLIM-based risk prediction, evaluate serial nutritional trajectories, and test bundled interventions that combine nutritional optimization with established delirium-prevention strategies.

In this prospective randomized controlled trial, 68 patients undergoing video-assisted thoracoscopic surgery were assigned to dexmedetomidine-based or conventional fentanyl-based anesthesia. Rescue fentanyl during the maintenance phase was lower with dexmedetomidine than with control anesthesia, 18.7 versus 33.2 micrograms, but postoperative pain, rescue analgesia, postoperative nausea and vomiting, and respiratory complications were similar. Dexmedetomidine increased intraoperative bradycardia and post-anesthesia care unit hypotension.

Impact: This trial directly tests the trade-off between modest opioid reduction and hemodynamic harm in a common thoracic anesthetic setting. Its negative or mixed postoperative findings are clinically important because they challenge the assumption that reduced intraoperative opioid use automatically improves recovery.

Clinical Implications: Dexmedetomidine may be considered as part of multimodal anesthesia for VATS when opioid reduction is specifically desired, but routine use should account for bradycardia and hypotension risk. Hemodynamic monitoring, dose individualization, and careful patient selection are necessary, particularly in patients with limited cardiovascular reserve.

Future Directions: Larger multicenter trials should compare dose-stratified dexmedetomidine regimens, assess patient-centered recovery outcomes, and identify cardiovascular risk groups in whom opioid reduction outweighs hemodynamic adverse effects.