AMICASim: Adaptive Multidrug Control System of Major Surgery Featuring Human-in-the-Loop Capabilities for Balanced Anesthesia Care

The development of automated anesthesia-delivery systems requires reproducible simulation environments in which control strategies can be evaluated before clinical testing. This paper presents AMICASim v01.2026, a control-oriented virtual patient simulator developed for benchmarking multidrug anesthesia and hemodynamic control. The framework combines established pharmacokinetic–pharmacodynamic models, nonlinear drug–drug interactions, coupled anesthesia–hemodynamic dynamics, procedure-informed disturbance profiles, covariate-defined patient variability, and a standardized controller interface. These elements support systematic evaluation of set-point tracking, disturbance rejection, interaction management, and robustness across heterogeneous virtual scenarios. The platform was assessed through open-loop and closed-loop simulations and through retrospective comparison with 335 VitalDB cases from four surgical categories. The results show that AMICASim generates scenario-dependent dynamics and disturbance conditions that allow control performance to be differentiated quantitatively across surgical profiles. A retrospective matched-case assessment showed qualitative and cohort-level consistency with selected clinical trajectories, while also revealing residual errors associated with unmodelled physiology, measurement noise, clinical interventions, and incomplete event information. These findings do not establish accurate patient-specific trajectory prediction. A Titration Paradox Index is additionally introduced to characterize the apparent relationship between drug administration and measured effect during closed-loop regulation. Within the defined simulation benchmark, the index complements conventional tracking and variability metrics when assessing disturbance compensation. AMICASim v01.2026 is therefore positioned as a clinically motivated and reproducible benchmark simulator for controller development and robustness testing. It is not presented as a prospectively validated patient-specific digital twin or clinical decision-support system. Patient-level predictive validation, explicit clinician-intervention experiments, and prospective clinical studies remain necessary before translational use.