This PK/PD analysis included data from 20 elderly patients who underwent elective surgery from December 2024 to May 2025. The study was conducted after obtaining approval from the Ethics Committee of the General Hospital of Southern Theater Command of PLA (No. NZLLKZ2024149) and was prospectively registered at the Chinese Clinical Trial Registry (No. ChiCTR2400093796). All participants provided written informed consent prior to enrollment.

Exclusion criteria included: patients with severe cardiac, pulmonary, hepatic, or renal dysfunction, or coagulation disorders; those with a history of mental illness (e.g., dementia, schizophrenia), long-term use of psychotropic drugs or chronic analgesics, or a history of alcoholism; participants who had been involved in other drug clinical trials within the past 6 months; patients with a known allergy to the study drugs; and those with an expected intraoperative blood loss exceeding 800 mL.

All patients fasted from midnight on the day of surgery and did not receive any preoperatively prescribed medications. Upon entering the operating room, the following monitoring devices were connected: pulse oximeter, electrocardiograph, end-tidal carbon dioxide monitor, non-invasive blood pressure monitor, and bispectral index (BIS®) monitor (Covidien, Boulder, CO, United States). All monitoring data were continuously collected from the start of anesthesia induction until the termination of anesthesia.

The anesthesia implementation and administration regimen in this study strictly followed a standardized clinical pathway, and was designed based on the labeling information of ciprofol approved in China as well as the recommended usage for adult surgical patients reported in relevant clinical research literature (Liang et al., 2023). A 20G arterial catheter was placed in the patient’s radial artery to facilitate repeated arterial blood sampling. Prior to anesthesia induction, preoxygenation was performed by inhaling 100% oxygen via a face mask. Anesthesia induction was achieved with an intravenous bolus of ciprofol 0.4 mg kg^-1, followed by a continuous infusion of ciprofol at a rate of 0.8 mg kg^-1·h^-1 5 min later until the end of surgery. Meanwhile, remifentanil was administered via target-controlled infusion (TCI) of the effect site using the Minto model (Abad-Torrent et al., 2022).

During anesthesia maintenance, midazolam was intravenously administered as appropriate if the BIS value exceeded 60. The target effect-site concentration (Ce) of remifentanil was titrated according to the patient’s hemodynamic responses, maintained within the range of 2–20 ng ml^-1 to ensure hemodynamic stability. Cisatracurium besylate 2 mg kg^-1 was intravenously injected to assist with tracheal intubation. Intraoperative mechanical ventilation was initiated after intubation. The ventilator settings were titrated to maintain an end-tidal partial pressure of carbon dioxide (P_ETCO₂) within the target range of 35–45 mmHg for the duration of the procedure. If hemodynamic fluctuations occurred during surgery, vasoactive drugs were used for intervention as needed. The infusion of ciprofol and remifentanil was stopped when the surgical skin suturing began; neostigmine was intravenously injected at the end of surgery to reverse the neuromuscular blocking effect.

Blood samples (2 mL) were obtained via arterial puncture at scheduled times relative to ciprofol administration: at 0, 1, 3, and 5 min after a 1-min bolus injection; at 5, 10, 15, 30, 45, and 60 min after initiating continuous infusion; and at 3, 5, 10, 30 min, 1, 1.5, 2, 4, and 6 h after infusion cessation. Samples were immediately transferred to EDTA tubes, centrifuged (3,000×g, 10 min), and the resulting plasma was stored at −80 °C pending analysis.

Ciprofol plasma concentrations were quantified using a validated UPLC-MS/MS method (Zhu et al., 2025). Following protein precipitation with methanol using ciprofol-d6 as the internal standard, chromatographic separation was achieved on a Shimadzu Shim-pack GIST-HP C18 column (3 μm, 2.1 × 150 mm) maintained at 40 °C. The mobile phase, comprising 5 mmol L^-1 ammonium acetate (A) and methanol (B), was delivered isocratically at 0.4 mL min^-1. Detection employed electrospray ionization in negative mode with MRM. The monitored transitions were m/z 203.100 → 175.000 for ciprofol and m/z 209.100 → 181.100 for the internal standard. The method was linear from 5 to 5,000 ng mL^-1, with an LLOQ of 5 ng mL^-1.

Population pharmacokinetic nonlinear mixed-effects modeling (PopPK-NLME) was performed using NONMEM 7.2.0 software (ICON Development Solutions, Ellicott City, MD, United States) combined with the gfortran 9.0 compiler. During the modeling process, the Wings for NONMEM (WFN720; http://wfn.sourceforge.net) visualization platform was used to call NONMEM for calculations. The parameter estimation method adopted was First-Order Conditional Estimation with Interaction (FOCEI), which can improve the parameter estimation accuracy of data with significant individual differences by introducing interaction terms between inter-individual variability and fixed effects, making it particularly suitable for heterogeneous study populations such as the elderly.

The construction of the base model was performed by comparing the one-compartment, two-compartment, and three-compartment structures, which were screened based on the minimization of the objective function value (OFV), parameter precision, and goodness of fit. Statistical analysis showed that the three-compartment model could more accurately characterize the distribution and elimination profiles of ciprofol in elderly patients (Supplementary Material Table S1). These findings were consistent with those of previous studies (Hu et al., 2021; Liu et al., 2024; Teng et al., 2021), and the final model structure is presented in Figure 1.

The final model was assessed for accuracy and stability using diagnostic plots, bootstrap validation, and prediction-corrected visual predictive checks (pc-VPC) (Nguyen et al., 2017). Diagnostic evaluation included goodness-of-fit plots: observations (DV) versus population predictions (PRED) and individual predictions (IPRED), along with CWRES plotted against both PRED and time after dose (TAD). For bootstrap validation, 1,000 replicate datasets were generated by random sampling with replacement from the original data, and model parameters were re-estimated for each. The median parameter estimates from bootstrap replicates aligned closely with the original estimates, all of which fell within the 2.5th–97.5th percentile intervals, indicating robust model stability. Subsequently, pc-VPC was performed by simulating multiple datasets based on the final model. The 5th, 50th, and 95th percentiles of the simulated data at each time point were calculated to construct prediction intervals. Excellent predictive performance was confirmed by the close agreement between the percentiles of the observed data and the simulated prediction intervals across all time points.

To predict the target effects of anesthesia induction and maintenance, and provide a basis for dose optimization in future studies, Monte Carlo simulations were performed based on the final validated PK/PD model to estimate the plasma and effect-site concentrations of ciprofol under different dosing regimens. The simulation protocol was designed as follows: a loading dose of 0.4 mg kg^-1 was administered via intravenous infusion over 1 min, followed by continuous infusion of maintenance doses at 0.4, 0.6, 0.8, 1.0, 1.2, and 1.4 mg kg^-1·h^-1, respectively, with a total infusion duration of 2 h. Each simulation included 1,000 virtual subjects, whose PK/PD parameters and inter-individual variability were derived from random sampling of the normal or log-normal distributions of the final model parameters. Through comprehensive comparison of key efficacy indicators across different dosing regimens, the optimal administration strategy that best matched the target effects was finally identified.

A total of 20 elderly surgical patients (11 males and 9 females, aged 67–82 years, weighing 49–80 kg) were finally enrolled in this study. A total of 386 arterial plasma samples were collected, and 386 corresponding BIS values were recorded synchronously. Except for the pre-administration time points, the plasma concentrations of ciprofol in all samples were above the LLOQ, and no outliers were identified. The demographic characteristics, laboratory test results, and surgical types of the patients are detailed in Table 1; the medications administered during anesthesia maintenance are presented in Supplementary Material Table S3; and the time-dependent trends of ciprofol plasma concentrations and BIS values are illustrated in Figure 2.

A PopPK model describing the continuous infusion of ciprofol in elderly patients (aged ≥65 years) was successfully established, and the typical parameter estimates are summarized in Table 2. Pharmacokinetic characteristic analysis revealed that ciprofol exhibited rapid distribution and extensive tissue distribution in the elderly population. The model-estimated central compartment volume of distribution (V₁) was 2.95 L, while the volumes of distribution of the shallow peripheral compartment (V₂) and deep peripheral compartment (V₃) were 45.15 L and 76.79 L, respectively, indicating significant drug accumulation in deep tissues. The typical value of systemic clearance (CL) was 1.01 L min^-1. The inter-compartmental clearances Q₂ and Q₃ were 0.76 L min^-1 and 0.66 L min^-1, respectively.

Covariate analysis identified body weight (WT) and age as significant covariates affecting CL, and their relationships were described using power functions. The final model equation for clearance estimation is as follows: C L = 1.01 × W T 66 0.74 × A g e 72.5 − 0.21

This equation demonstrates that clearance increases with increasing body weight and decreases with advancing age, which is consistent with physiological expectations.

Model evaluation and validation results confirmed the reliability and robustness of the final model. The relative standard errors (RSE%) of all structural pharmacokinetic parameters were below 30%, and the RSE% of key parameters (CL、V₂、V₃、Q₂、Q₃) were all less than 10%, indicating high precision of parameter estimation. Internal validation via 1000-time Bootstrap resampling showed that the median values of Bootstrap-estimated parameters were highly consistent with those of the final model, with narrow 95% confidence intervals (CIs), further supporting the stability of parameter estimation.

Compared with the base model, the OFV of the final model decreased significantly (OFV = −24.99), and this difference was statistically significant, justifying the inclusion of body weight and age as covariates. IIV was described using an exponential model, and its variance (ω) was within the acceptable range. Notably, the IIV of CL decreased after the introduction of covariates, indicating that the model successfully explained this part of population heterogeneity. In addition, the intra-individual residual variability remained stable during model optimization.

A sigmoidal E_max model was successfully established to characterise the relationship between ciprofol plasma concentration and its pharmacodynamic effects. The final model parameter estimates, parameter precision, and bootstrap validation results are summarised in Table 3. The effect-site equilibrium half-life (K_e0) was approximately 1.09 min^-1, indicating that the distribution equilibrium of ciprofol in the effect site is achieved very rapidly. The baseline BIS value (E₀) was estimated to be 93.4, which is close to the fully awake state. The maximum effect (Emax) was 45.77, the median effective concentration (EC₅₀) was 233.91 ng mL^-1, and the Hill coefficient (γ) was 3.00.

Covariate analysis results showed that common physiological covariates such as age, body weight, height, and gender did not significantly improve model fitting. Therefore, no covariates were incorporated into the final PD model. The RSE% of all structural parameters was below 40%, and the RSE% of most key parameters (E₀, Emax, EC₅₀) was less than 10%, indicating good estimation precision. The median values from 1000-time bootstrap validation were highly consistent with the final estimates, further supporting the stability of the model.

The scatter plots depicting DV against PRED, DV against IPRED, CWRES against PRED, and CWRES against TAD for the final model are illustrated in Figures 3, 4. The results indicated that all data points were randomly distributed around the line of identity (y = x), demonstrating a strong correlation between PRED/IPRED and observed values, as well as adequate model fitting performance. The CWRES versus PRED and CWRES vs. TAD diagnostic plots exhibited a symmetrical distribution of CWRES around the reference line (y = 0), with the majority of values residing within the 2 range. No discernible trends of CWRES with respect to concentration or duration were noted, indicating effective model fitting.

In addition, individual concentration-time profiles (Supplementary Material Figure S1) provided a visual evaluation of the individual predictive ability of the model. The plots showed that the model-predicted concentration trajectories could well capture the overall trends and inter-individual variability of observed concentrations in most patients, further confirming the reliability of the model in describing the complex pharmacokinetic behavior of ciprofol in the elderly population.

Figure 5 displays the pc-VPC charts. The median, 5th, and 95th percentiles of the prediction-corrected observed ciprofol concentrations and BIS values significantly coincided with the 95% confidence ranges of the corresponding prediction-corrected predicted values derived from simulated data. The results demonstrated that the developed population PK/PD model can accurately characterize the pharmacokinetic and pharmacodynamic characteristics of ciprofol.

Based on the experimental observations, a loading dose of 0.4 mg kg^-1 administered intravenously over 1 min was first determined to rapidly achieve the target effect of anaesthesia induction. On this basis, simulation analyses of different maintenance dose regimens were performed using the PK/PD model. The optimal regimen was defined as the one that could stably maintain the BIS value within the range of 40–60 for anaesthesia induction and maintenance.

Figure 6 shows the Monte Carlo simulations of BIS changes over time under different administration regimens based on the final PK/PD model. The simulation results indicated that during general anaesthesia in elderly patients, the regimen consisting of a loading dose of 0.4 mg kg^-1 (infused over 1 min) combined with an initial maintenance dose of 0.6 mg kg·h^-1, with continuous infusion for 2 h, could sustain the BIS score within the ideal range of 40–60.

In terms of PopPK, this study confirmed that the three-compartment model could optimally characterize the PK behavior of ciprofol in elderly patients, which is consistent with previously reported model structures (Hu et al., 2021; Liu et al., 2024; Teng et al., 2021). This indicates that the distribution and elimination patterns of ciprofol in the elderly population have not changed fundamentally, but the parameter values exhibit population-specific characteristics. The typical value of CL (1.01 L min^-1) was slightly higher than that reported in previous studies on the elderly population (Guo et al., 2025). This difference may be attributed to variations in study population characteristics (e.g., comorbidities, baseline liver and kidney function) and surgical types. Moreover, the inclusion of body weight as a significant covariate increased the population typical value of CL. The estimated volume was greater for the deep peripheral compartment (V₃, 76.79 L) than for the shallow compartment (V₂, 45.15 L), which may be related to changes in body composition of elderly patients (increased fat proportion and decreased muscle mass). This makes ciprofol, a lipophilic drug, more prone to accumulate in deep tissues such as fat, thus rendering the deep peripheral compartment the main drug reservoir (Thürmann, 2020). This is consistent with the high lipophilicity of ciprofol and its extensive distribution and accumulation in adipose tissues, verifying the physiological rationality of the model structure. The differences in inter-compartmental clearances further reflect the unbalanced transfer rates of the drug between the shallow and deep peripheral compartments, suggesting that clinical attention should be paid to the risk of delayed emergence caused by the slow release of drugs from deep tissues.

Covariate analysis revealed the key sources of inter-individual variability in ciprofol PK among elderly patients. WT and age had significant effects on CL, which is consistent with the general PK rules in the elderly population: increased body weight can enhance clearance capacity, while the physiological decline of liver and kidney function with aging leads to a slight decrease in clearance (Thürmann, 2020). Based on the final model, the influence exponent of body weight on CL was 0.74, which is within the range reported for propofol (Puri, A., Medhi, 2012; Sahinovic et al., 2018; Van Kralingen et al., 2011). Notably, this study found no statistically significant effects of other covariates (including gender, BMI, aspartate aminotransferase, alanine aminotransferase, total bilirubin, total protein, and creatinine clearance) on ciprofol PK. This result is different from some studies on propofol (Araújo et al., 2020; Schüttler and Ihmsen, 2000), which may be due to the masking effect of the general decline in physiological function of the elderly population on inter-individual variability.

The relationship between ciprofol plasma concentration and BIS value conforms to the sigmoidal Emax model, which is consistent with literature reports (Hu et al., 2021; Liu et al., 2024; Teng et al., 2021). The key PD parameters highlight the advantages of ciprofol in elderly anesthesia. The K_e0 was 1.09 min^-1, indicating that the drug can rapidly reach equilibrium between plasma and the central nervous system, which is consistent with the clinically observed characteristics of rapid onset (1–2 min) and rapid offset of sedation. This parameter is lower than that of propofol (K_e0: 0.57 min^-1) (Chen et al., 2024) and superior to that of remimazolam (K_e0: 1.38 min^-1) (Lim et al., 2006), making ciprofol suitable for rapid sequence induction and titratable anesthesia. The E₀ is close to the normal awake level, confirming the validity of the baseline measurement. The Emax indicates that ciprofol can reduce the BIS value by approximately 46 units compared with the baseline, which is sufficient to produce deep sedation/anesthesia without causing excessive suppression. The EC₅₀ was 233.91 ng mL^-1, reflecting the overall sensitivity of this elderly cohort to ciprofol; this value is lower than that reported in young adults (284 ng mL^-1) (Teng et al., 2021). The γ indicates a relatively steep concentration-effect curve, meaning that minor changes in plasma concentration near EC₅₀ can cause significant changes in BIS value. This characteristic facilitates precise titration of anesthesia depth and reduces the risk of inadequate or excessive sedation, which is particularly important for elderly patients with limited physiological reserve.

Covariate analysis showed that factors such as age, body weight, and gender had no significant effects on PD parameters, indicating that the pharmacodynamic responses of elderly surgical patients to ciprofol are relatively homogeneous. This suggests that the inter-individual variability in anesthetic effect is mainly derived from PK factors rather than differences in central nervous system sensitivity.

The final PopPK model yielded precise estimates of all pharmacokinetic parameters, which align with the typical properties of GABAA receptor agonists (Schüttler and Ihmsen, 2000). The thorough model validation affirmed the resilience and superior predictive capability of the final PK/PD model. Internal validation with 1,000 bootstrap resampling demonstrated that the median parameter values were remarkably consistent with the final model, and the narrow 95% confidence intervals indicated a high precision of parameter estimation. Goodness-of-fit plots indicated that both individual and population predicted concentrations were closely aligned with the line of identity relative to observed concentrations, and conditional weighted residuals were randomly and symmetrically distributed around zero, devoid of time- or concentration-dependent trends, thereby confirming the absence of systematic bias in the model. The prediction-corrected visual predictive assessment plots further substantiated the model’s predictive capability: the observed median, 5th, and 95th percentiles of ciprofol concentrations and BIS values significantly coincided with the 95% confidence intervals of the simulated data, demonstrating that the model effectively encapsulates the central tendency and inter-individual variability of the data. The PK/PD model simulation indicated that a loading dose of 0.4 mg kg^-1 (administered over 1 min), followed by an initial maintenance dose of 0.6 mg kg^-1·h^-1, can maintain the intraoperative BIS value of older patients within the optimal sedation range of 40–60. Consistent anesthetic depth can be sustained with continuous infusion for over 2 h. This optimal regimen differs from This optimized regimen diverges from the previously recommended starting maintenance dose of 0.8 mg kg·h^-1 in adult trials (Liu et al., 2024), which primarily considers the diminished clearance and heightened risk of drug buildup in older patients. The maintenance dose was suitably decreased to equilibrate anesthetic efficacy and safety.

The findings from the PK/PD model-based Monte Carlo simulations in this study confirmed that the administration regimen consisting of a loading dose of 0.4 mg kg^-1 (infused over 1 min) combined with an initial maintenance dose of 0.6 mg kg^-1·h^-1can stably maintain the intraoperative BIS values of elderly patients within the optimal target range of 40–60. This regimen is significantly lower than the recommended initial maintenance dose of 0.8 mg kg^-1·h^-1 for general adult populations (Liu et al., 2024), reflecting the academic rationale of this study: by fully accounting for the pathophysiological characteristics of elderly patients, such as reduced drug clearance and elevated risk of drug accumulation, an appropriate dose reduction was implemented to balance anesthetic efficacy and safety.

Given the current clinical practice where ciprofol administration for elderly patients largely relies on empirical approaches or passive titration, which is prone to causing delayed emergence or circulatory depression, this study adhered to the principle of “minimum effective dose”. Through the integration of population pharmacokinetic parameters and simulation verification, we confirmed the superiority of the 0.6 mg kg^-1·h^-1 infusion rate in achieving the optimal benefit-risk balance, and revealed the critical role of incorporating individual covariates in reducing drug accumulation and accelerating postoperative emergence. This paradigm shift from an experience-driven model to an evidence-based precision medicine model holds profound clinical guiding significance for optimizing the perioperative safety of elderly patients and improving the quality of postoperative recovery.

In clinical anesthesia practice, ciprofol is frequently administered in combination with a variety of adjunctive drugs, and its potential drug-drug interactions (DDIs) warrant close attention. From a PK perspective, studies have demonstrated that ciprofol is primarily eliminated via UGT1A9-mediated glucuronidation and CYP2B6-mediated oxidation in the liver (Zhou et al., 2021). In contrast, remifentanil undergoes rapid hydrolysis by non-specific plasma and tissue esterases, and cisatracurium is metabolized through Hofmann elimination; neither of these two drugs relies on the conventional hepatic enzyme system. Midazolam, on the other hand, is predominantly metabolized via CYP3A4. Owing to the heterogeneity of the metabolic pathways among these drugs, the likelihood of competitive PK interactions occurring at the clinical doses employed in this study is extremely low. From a PD perspective, ciprofol, remifentanil, and midazolam all exert their effects on the central nervous system, suggesting that significant pharmacodynamic synergism may exist among them. Existing studies have indicated that opioids can significantly reduce the concentration of sedatives required to achieve the desired depth of anesthesia through pharmacodynamic synergistic effects. This phenomenon has been well-documented in the clinical application of propofol. As a novel sedative with a mechanism of action similar to that of propofol, ciprofol may exhibit analogous synergistic effects (Bouillon et al., 2004). In addition, cisatracurium, as a peripheral neuromuscular blocker, does not have direct PD interactions with ciprofol.

To minimize the potential bias of DDIs on the estimation of model parameters, a standardized combined administration regimen was implemented in this study. The aim was to ensure that the derived population parameters could objectively characterize the pharmacological properties of elderly patients under the standard clinical pathway.

Despite its strengths, this study has several limitations that need to be addressed. First, the sample size of this study (n = 20) and the single-center study design impose certain limitations on the extrapolation of its conclusions to more diverse surgical populations. Therefore, large-sample, multi-center studies are urgently needed in the future to conduct more extensive external validation of the dosing regimens and model parameters derived from this research. Second, potential drug-drug interactions with concomitant medications (e.g., opioids, benzodiazepines) were not considered in the construction of the ciprofol PK/PD model. In addition, the effects of surgery-related factors (e.g., intraoperative fluid infusion volume, blood transfusion volume, body temperature) or genetic polymorphisms on ciprofol PK/PD were not evaluated. Finally, the PD model was established solely based on BIS values; integrating additional endpoint indicators (e.g., Modified Observer’s Assessment of Alertness/Sedation scale, electroencephalogram spectral entropy) would help improve the comprehensiveness of the model.