SED Pack: Safety and Performance of a Medical Device Software for Predictive Monitoring During Sedation

Overview

The SED-PACK study is a prospective, observational, single-center clinical investigation evaluating the clinical performance and safety of SED Pack, a Medical Device Software (MDSW) developed by Predictheon Medical S.L., in a fully blinded, non-interventional setting during routine procedural sedation in adult patients at Hospital Clínic de Barcelona, Spain. Procedural sedation consists in the administration of sedative and analgesic drugs to allow patients to undergo routine diagnostic or therapeutic procedures comfortably. Despite advances in monitoring technology, adverse events such as arterial hypotension, respiratory depression, or inadequate sedation remain frequent due to marked interindividual variability in drug response. Currently, clinicians rely on standard physiological monitoring that measures what is happening in real time but does not anticipate what is about to happen. SED Pack is designed to address this gap by continuously processing physiological monitoring data and drug-infusion information to generate four real-time predictive indices (updated once per second): pSED (sedation depth), pNOC (likelihood of response to a noxious stimulus), pHYPO (likelihood of arterial hypotension), and pDEP (likelihood of respiratory depression). In this clinical investigation, one hundred and twenty adult patients (≥18 years, ASA I-III) will be enrolled. Predictive performance is evaluated using the c-index for pSED, pHYPO, and pDEP, and AUC-ROC for pNOC. The primary hypothesis is that each index achieves a discrimination performance ≥0.70 (lower bound of the 95% CI also ≥0.70). The software runs in fully blinded mode: its predictions are not visible to clinicians and do not influence patient management. No additional procedures beyond standard clinical care are introduced. The study complies with MDR 2017/745 and ISO 14155, with AEMPS authorization and CEIm approval (reference: HCB/2026/0307). Expected outcomes include confirmation that each SED Pack predictive index achieves clinically acceptable discrimination performance, supporting the clinical use of this investigational software as a tool to anticipate and prevent sedation-related complications.

Investigational device: SED Pack is a Medical Device Software (MDSW) that continuously processes physiological monitoring data and drug-infusion parameters to generate four real-time predictive indices (scale 0-100, updated once per second): pSED (probability of sedation depth from light sedation to general anesthesia), pNOC (probability of motor response to mild noxious stimuli), pHYPO (probability of clinically relevant arterial hypotension), and pDEP (probability of clinically relevant respiratory depression). The software incorporates a predefined individualization step that adjusts patient-specific parameters within a fixed model structure based on observed physiological signals. This process is deterministic and does not involve adaptive learning, model retraining, or self-modification. The algorithmic structure, logic, and parameters are fully locked throughout the investigation. Study design: Prospective, observational, single-arm, single-center clinical investigation conducted under Article 62 of Regulation (EU) 2017/745 at Hospital Clínic de Barcelona, Spain. The investigation is non-interventional: no additional procedures beyond standard clinical care are introduced. The SED Pack software is installed on a dedicated PC connected via wired, unidirectional, read-only serial interfaces to standard clinical monitoring systems; the PC is not connected to any wireless or hospital network during the investigation. The software operates in fully blinded mode throughout: its predictions are not visible to the attending anesthesiologist, endoscopist, nursing staff, or investigators performing clinical reference assessments, and there is no break-glass mechanism for real-time visualization. All clinical decisions rely exclusively on standard-of-care CE-marked monitoring systems. Patient enrollment: Participants are recruited among adult patients scheduled for elective gastrointestinal endoscopic procedures requiring procedural sedation at Hospital Clínic de Barcelona. Potentially eligible patients are identified during routine pre-procedural evaluation, provided with a patient information sheet and an informed consent form, and enrolled only after written informed consent is obtained. Participation is entirely voluntary and refusal or withdrawal does not affect standard medical care. Because the investigation does not deviate from routine clinical practice and introduces no additional procedures or risks, no study-specific patient training is required. Performance evaluation and statistical analysis: The primary objective is to evaluate the predictive performance of the four SED Pack indices (pSED, pNOC, pHYPO, and pDEP) against predefined clinical and physiological reference standards, using the concordance statistic (c-index) for pSED, pHYPO, and pDEP, and the AUC-ROC for pNOC, complemented by sensitivity and specificity analyses at predefined operational thresholds. Secondary objectives include evaluation of predictive performance at short-term prediction horizons (+5 and +10 minutes), threshold-based AUC-ROC analyses against predefined binary clinical thresholds, and exploratory subgroup analyses by ASA physical status, age group, and sex. One pre-planned interim analysis will be conducted after enrollment of approximately 50% of the target sample for data quality assessment purposes only. The anticipated outcome of this investigation is the confirmation that the SED Pack predictive indices achieve clinically acceptable discrimination performance against predefined reference standards, supporting the clinical performance claims of this investigational software. Ultimately, SED Pack has the potential to transform the monitoring paradigm during procedural sedation by enabling clinicians to anticipate adverse events before they occur, improving patient safety and supporting more individualized drug management.