Optimal Blood Sampling Time for Area under the Concentration–Time Curve Estimation of Vancomycin by Assessing the Accuracy of Four Bayesian Software

Abstract

Effective blood sampling times, beyond trough and peak levels, have not been determined for estimating vancomycin’s area under the concentration–time curve (AUC) using the Bayesian software. The aim of this study was to evaluate the accuracy of AUC estimation at different blood sampling times during the same dosing interval at steady state utilizing data from a prior phase I trial of vancomycin. Six healthy adult participants were sampled following intravenous administration of 1 g vancomycin for 1.5 h every 12 h. The AUC was estimated using four software packages and four population pharmacokinetic models. Accuracy was assessed using bias (difference between the estimated and reference AUC) and imprecision (absolute percentage difference between the estimated and reference AUC). The accuracy varied with the sampling time. The optimal two-point sampling times were determined to be 2.5 and 5.5 h post-injection using software packages for EasyTDM, Practical AUC-guided therapeutic drug monitoring (TDM), and Anti-MRSA Agents TDM Analysis Software (incorporating Rodvold, Yamamoto, and Yasuhara models). In these estimations, the mean bias (range, −1.7 to 9.5 µg·h/mL) was unbiased and the mean imprecision (range, −3.0% to 5.0%) was precise. The optimal one-point sampling time was 5.5 h post-injection for Anti-MRSA Agents TDM Analysis Software, which incorporated the Yamamoto and Yasuhara models. In conclusion, optimal blood sampling times may vary depending on the software and model used. Our findings suggest that identifying specific sampling times could improve the efficacy of TDM in clinical practice.

INTRODUCTION

Vancomycin, a glycopeptide antimicrobial agent, is the first-line treatment for methicillin-resistant Staphylococcus aureus (MRSA) infections.¹⁾ Therapeutic drug monitoring (TDM) utilizing the area under the concentration–time curve (AUC) for vancomycin improves therapeutic outcomes and mitigates the risk of renal impairment.^2,3) TDM guidelines recommend a daily AUC of 400–600 µg·h/mL (assuming a minimum inhibitory concentration [MIC] of 1 µg/mL in broth microdilution assay) as an indicator and target range for dosing.^2,3) In clinical TDM, the Bayesian software is used to estimate the AUC of vancomycin.^4,5) It uses Bayesian estimation to determine patient-specific pharmacokinetic (PK) parameters based on population PK models and patient clinical data.⁶⁾ The advantage of Bayesian estimation is its potential to estimate the AUC at any blood concentration during the dosing interval.

TDM guidelines recommend estimating the AUC using the Bayesian software with two sampling points (trough and peak) or one point of trough levels.^2,3) This recommendation is based on numerous Bayesian software accuracy assessment reports using trough and peak levels.^5,7–11) However, other sampling points for AUC estimation using the Bayesian software have not yet been established as applicable blood sampling time points.

The accuracy of vancomycin AUC estimation using the Bayesian software and population PK models fluctuates with the sampling time point.^5,7,8,12) When vancomycin concentration sampling deviates from the timing recommended in the TDM guidelines,¹³⁾ accurate AUC assessment-based TDM may not be feasible. However, among the population PK models in different Bayesian software used in Japan, only Yasuhara et al.’s¹⁴⁾ model has been employed to evaluate the precision of AUC estimation at various sampling points.¹²⁾

The aim of this study was to assess the accuracy of the Bayesian software and population PK models at various sampling time points. Additionally, the optimal blood sampling time was examined. Accuracy was evaluated using diverse blood concentration–time data within the same dosing interval at a steady state.

MATERIALS AND METHODS

Study Setting

In this study, a phase I trial report was identified, and the accuracy of the Bayesian software for predicting intermittent vancomycin injections was evaluated based on baseline characteristics and blood concentration–time data for six subjects extracted from the report.¹⁵⁾ This trial reported intensive blood concentration data at a steady state following nine intravenously administered vancomycin (1 g) doses for 1.5 h every 12 h in six healthy subjects. The subjects were men, with a mean age of 34 (range, 24–48) years, weight 68 (56–79) kg, height 171 (164–179) cm, and BMI 23.1 (20.3–26.4) kg/m². The mean blood concentration at 0, 1.5, 2, 2.5, 3.5, 5.5, 7.5, and 12 h (C₀, C_1.5, …, C₁₂) post-injection at steady state was 4.3 (range, 3.5–5.8), 50.6 (44.7–58.8), 28.6 (20.0–32.2), 22.3 (17.4–25.7), 16.6 (13.8–20), 10.3 (8.2–13.5), 7.5 (5.2–9.7), and 3.7 (3.1–4.5) µg/mL, respectively. Based on these eight points, the 12-h AUC was calculated using the log-linear trapezoidal formula. The daily AUC, defined as the reference AUC, was calculated by doubling the 12-h AUC. The mean reference AUC was 320.8 (range, 258.4–369.1) µg·h/mL.

As we used anonymized data from scholarly articles, involved no treatment interventions, and did not entail the collection of human samples, the requirement for informed consent was waived. No ethical approval was required to conduct this study.

Bayesian Software

Table 1 enumerates the population PK models employed in the four Bayesian software. These software packages were sourced from Japan before data analysis and employed a two-compartment adult population PK model for vancomycin.¹⁶⁾ Each Bayesian software employed a nonlinear least-squares algorithm to maximize the posterior probabilities of the PK parameters, as follows: EasyTDM Ver3-6-1-1M (EasyTDM) (Kagawa Society of Hospital Pharmacists, Japan),¹⁷⁾ Vancomycin Meiji TDM Analysis Software Ver.1 (Meiji) (Meiji Seika Pharma Co., Ltd., Japan),¹⁸⁾ and Practical AUC-guided TDM ver. 2.1 (PAT) (Japanese Society of Chemotherapy, Japan)^5,14)—used the Levenberg–Marquardt method. Anti-MRSA Agents TDM Analysis Software Ver. 2 (Nichi-Iko) (Nichi-Iko Pharmaceutical Co., Ltd., Japan) used three algorithms (Simplex method, Damping Gauss-Newton method, and Modified Marquardt method [Fletcher’s Modified Marquardt Method]).^14,17,18) The accuracy of Nichi-Iko was assessed using its automatic algorithm selection function to select the algorithm that estimated blood concentrations closest to the measured values.

Table 1. Vancomycin Population Pharmacokinetic Models and the Bayesian Software

Model	Study	Parameters	Bayesian software
1	Rodvold et al.17)	CL (L/h)  =  0.003 × Wt + 0.045 × CCr (mL/min)	EasyTDM; Nichi-Iko
		V1 (L)  =  0.21 × Wt
		K12 (1/h) = 1.12
		K21 (1/h) = 0.48
2	Yamamoto et al.18)	CCr ≤ 85 mL/min: CL (L/h) = 0.0322 × CCr (mL/min) + 0.32	Meiji; Nichi-Iko
		CCr > 85 mL/min: CL (L/h) = 3.83
		Q (L/h) = 8.81
		Vss (L) = V1 + V2
		V1 (L)a) = 0.47792 × Wt
		V2 (L)a) = 60.5578
		K12 (1/h) = Q/V1
		K21 (1/h) = Q/V2
3	Yasuhara et al.14)	CL (L/h) = 0.04782 × CCr (mL/min)	PAT; Nichi-Iko
		Vss (L) = 60.7
		K12 (1/h) = 0.525
		K21 (1/h) = 0.213
4	Yasuhara et al.14)	CCr ≤ 85 mL/min: CL (L/h) = 0.04782 × CCr (mL/min)	Nichi-Iko
		CCr > 85 mL/min: CL (L/h) = 3.51
		Vss (L) = 60.7
		K12 (1/h) = 0.525
		K21 (1/h) = 0.213

CCr, creatinine clearance; CL, vancomycin clearance; Q, intercompartmental clearance; Vss, volume of distribution at steady state; V1, central volume of distribution; V2, peripheral volume of distribution; K12, transfer rate constant from the central compartment to the peripheral compartment; K21, transfer rate constant from the peripheral compartment to the central compartment; and Wt, weight. Nichi-Iko, Anti-MRSA Agents TDM Analysis Software; Meiji, Vancomycin Meiji TDM Analysis Software; and PAT, Practical AUC-guided therapeutic drug monitoring. a) Population pharmacokinetic parameters in patients with infectious diseases.

Subject Data Input into Bayesian Software

Subject data used for Bayesian estimation included age (years), sex, height (cm), and weight (kg). As the phase I trial report did not list the subjects’ serum creatinine concentration and creatinine clearance (CCr),¹⁵⁾ a CCr of 120 mL/min was assumed. This assumption was based on the mean vancomycin clearance of 97.6 mL/min for six subjects,¹⁵⁾ the reported ratio (0.80) between vancomycin clearance and CCr in 23–48-year-old subjects with stable renal function,¹⁹⁾ and a previous accuracy study.⁵⁾ Utilizing this ratio, the CCr for individual subjects was assumed based on the subject’s vancomycin clearance in the phase I trial. To estimate the daily AUC at a steady state, the number of doses was set to 10 in the Bayesian software.

Estimation of AUC and Blood Concentration Using the Bayesian Software

The daily AUC at steady state was estimated using the Bayesian software based on two-, one- (C₀, C_1.5, C_2.5, C_3.5, C_5.5, and C_7.5), and five-point (C₀, C_1.5, C_2.5, C_5.5, and C_7.5) subsets for each subject. Two-point subsets were generated using six concentration data points (C₀, C_1.5, C_2.5, C_3.5, C_5.5, and C_7.5). The number of rich blood concentration–time data points used in the Bayesian software was limited to five, which was the maximum number of PAT inputs.

Using the EasyTDM and Nichi-Iko models (1, 2, and 3), the mean concentrations at eight time points, estimated based on two- (C_0,2.5 and C_2.5,5.5) and one-point (C₀ and C_5.5) subsets, were plotted. Of the eight time points, the mean estimated and measured concentrations were compared at 0 and 1.5 h post-injection. The comparison excluded Meiji, PAT, and Nichi-Iko Model 4. The exclusion was because of reduced accuracy in the 5 points-based AUC estimation for Meiji and Nichi-Iko Model 4, and the inability of PAT to display estimated blood concentrations at user-specified dates and times.

Study Endpoint

Accurate estimation of the AUC using the Bayesian software was quantified by calculating the bias and imprecision, and estimating their 95% confidence intervals (CIs).^8,20) Bias (µg·h/mL) was defined as the difference between the estimated and reference AUC (bias = estimated AUC − reference AUC). The mean bias was deemed unbiased when the 95% CI included zero. Imprecision (%) was defined as the absolute value of the percentage difference between the estimated and reference AUC (imprecision = 100 × |estimated AUC − reference AUC|/reference AUC). An upper confidence limit of 95% CI for the mean imprecision of <10 and 20% was deemed as precise and moderately precise, respectively.^8,21) The threshold for imprecision was set because the target range of the daily AUC (400–600 µg·h/mL) was ±20% of 500 µg·h/mL.^2,3)

Statistical Analysis

Mean bias and imprecision based on blood concentration data for each Bayesian software program were compared using a one-way repeated-measures ANOVA. A paired t-test was used to compare the means of the estimated and measured concentrations at 0 and 1.5 h post-injection. For all statistical analyses, statistical significance was set at p < 0.05. Statistical analyses were performed using EZR version 1.61.²²⁾

RESULTS

Accuracy of AUC Estimation at Various Time Points and Optimal Sampling Time Points for the Bayesian Software and Population PK Models

Supplementary Tables 1 and 2 list the mean bias and imprecision of the AUC estimation based on the blood concentration data in the Bayesian software and population PK models. Table 2 presents the optimal sampling times.

Table 2. Optimal Blood Sampling Time of Four Bayesian Software for AUC Estimation of Vancomycin

The check mark indicates that the 95% confidence interval for the mean bias (µg·h/mL) includes zero. The gray and light-gray sampling times indicate the upper limits of the 95% confidence interval for the mean imprecision (%) of <10 and <20%, respectively. Nichi-Iko, Anti-MRSA Agents TDM Analysis Software; Meiji, Vancomycin Meiji TDM Analysis Software; and PAT, Practical AUC-guided therapeutic drug monitoring.

The one-way repeated-measures ANOVA revealed significant differences in mean bias (p < 0.05) and no significant differences in mean imprecision (p = 0.13) in the 5-point-based AUC estimation across various Bayesian software programs. In the AUC estimation using EasyTDM, PAT, and Nichi-Iko (1, 2, and 3), the mean bias ranged from −7.7 to 3.4 µg·h/mL and mean imprecision ranged from 1.8 to 3.0%, demonstrating unbiased and precise estimations. In contrast, the AUC estimations by Meiji and Nichi-Iko Model 4 were overbiased and moderately precise.

The one-way repeated-measures ANOVA demonstrated significant differences in the mean bias and mean imprecision of the respective AUC estimates for one- or two-point subsets in each Bayesian software program (p < 0.05). The mean bias and mean imprecision of two-point AUC estimates at C_2.5,5.5 using EasyTDM, PAT, and Nichi-Iko (1, 2, and 3) ranged from −1.7 to 9.5 µg·h/mL and from 3.0 to 5.0%, respectively, and were unbiased and precise. C_2.5 and C_5.5 were the optimal two points in these software programs. AUC estimation using the Nichi-Iko models (2 and 3) at one point of C_5.5 demonstrated that the mean bias (−1.5 and 1.1 µg·h/mL, respectively) and mean imprecision (5.3 and 3.6%, respectively) were unbiased and precise. C_5.5 of the Nichi-Iko models (2 and 3) was the optimal one point.

Supplementary Tables 3, 4 list the mean bias and imprecision of the AUC estimation based on the blood concentration data using the Bayesian software and population PK models when the CCr was assumed for individual subjects, and Supplementary Table 5 presents the optimal sampling times.

Comparison of Blood Vancomycin Concentrations Estimated Using EasyTDM and Nichi-Iko Models (1, 2, and 3)

Figure 1 illustrates the mean estimated concentrations based on the two- (C_0,2.5 and C_2.5,5.5) and one-point (C₀ and C_5.5) subsets in the EasyTDM and Nichi-Iko models (1, 2, and 3) and the mean measured concentrations.

Fig. 1. Observed and Estimated Vancomycin Concentrations through Bayesian Software Based on Various Blood Concentration–Time Data

Bayesian estimation using the EasyTDM and Nichi-Iko models (1, 2, and 3) was performed based on subsets of two points (C_0,2.5, C_2.5,5.5) and one point (C₀, C_5.5) for six subjects. A, B, C, and D in Fig. 1 show the plots of eight time points of Bayesian software-estimated vancomycin concentrations based on subsets of C_0,2.5, C_2.5,5.5, C₀, and C_5.5, respectively. Of the eight time points, the mean estimated concentrations at 0 and 1.5 h post-injection were compared with the measured concentrations. The X-axis indicates the time post-injection at the steady state. Data are presented as mean and standard deviation. Black circles, observed concentrations; white circles, EasyTDM estimations; crosses, Nichi-Iko Model 1 estimations; squares, Nichi-Iko Model 2 estimations; triangles, Nichi-Iko Model 3 estimations.

The mean estimated concentration (standard deviation (S.D.)) at 0 h post-injection ranged from 4.47 (0.788) (estimated at C₀ using EasyTDM) to 7.27 (1.42) µg/mL (estimated at C_5.5 using Nichi-Iko Model 2), which was significantly higher than that of the mean measured concentration (S.D.) of 4.33 (0.845) µg/mL (p < 0.05). However, the estimations at C_2.5,5.5 and C₀ using Nichi-Iko models 1 and 3, respectively, were not significantly different from the mean measured concentration (p = 0.063 and 0.10, respectively).

The mean estimated concentrations (S.D.) at 1.5 h post-injection ranged from 27.8 (2.74) (estimated at C₀ by Nichi-Iko Model 2) to 40.5 (3.47) µg/mL (estimated at C_0,2.5 by Nichi-Iko Model 1), which were significantly lower than that of the mean measured concentration (S.D.) of 50.6 (5.46) µg/mL (p < 0.01).

Supplementary Fig. 1 illustrates the mean estimated concentrations using the software and models when assuming the CCr for individual subjects.

DISCUSSION

This study demonstrated variations in the optimal sampling points for estimating the AUC of vancomycin in subjects using the Bayesian software and population PK models (Table 2). Our results indicate that sampling at two points of C_2.5,5.5 and one point of C_5.5 at the steady state may accurately estimate AUC on the day of blood sampling in the software packages under study.

The results of this study indicated that the accuracy of AUC estimation using the population PK models varies based on the incorporated Bayesian software. While Oda et al.⁵⁾ reported accuracy variations in Yasuhara et al.’s¹⁴⁾ model, we observed accuracy variations in Yamamoto et al.’s¹⁸⁾ and Rodvold et al.’s¹⁷⁾ models (Table 2). Specifically, Meiji, which incorporated Yamamoto et al.’s¹⁸⁾ model, demonstrated a decrease in accuracy across the entire sampling period. This variation in accuracy could be attributed to the differences in the nonlinear least-squares calculation algorithm for each Bayesian software. In the accuracy assessment of Nichi-Iko, specific algorithms failed to calculate the subjects’ PK parameters from the blood concentration data. Hence, the algorithms automatically selected by the software were evaluated. Further research is needed to evaluate the effect of algorithm selection on AUC estimation accuracy.

Our findings suggest that sampling at steady-state C_2.5,5.5 may offer an accurate AUC estimation in the EasyTDM, PAT, and Nichi-Iko models (1, 2, and 3). A simulation study on the accuracy of the non-steady-state AUC_24–48 h estimation was reported for Yasuhara et al.’s¹⁴⁾ model.¹²⁾ Oda et al.¹²⁾ reported that for individuals with normal renal function (CCr > 90 mL/min), the accuracy of the two measurement points at 1 and 5 h after the end of injection on the second day of intermittent vancomycin administration was equivalent to the trough and peak levels. The alignment between the results of this study and those of previous studies highlights the possibility that C_2.5,5.5 is a novel sampling point suitable for AUC estimation. However, clinicians should note that Bayesian estimation using C_2.5,5.5 may overestimate trough levels (Fig. 1B). Furthermore, one-point blood sampling at C_5.5 may also overestimate trough levels (Fig. 1D). Therefore, in the treatment of osteomyelitis and Gram-positive bacterial infections with the exception of MRSA infection, vancomycin dosing based on trough level measurements is recommended owing to insufficient evidence supporting the efficacy of AUC-guided dosing.^2,3) Two-point sampling at C_2.5,5.5 may have exhibited higher accuracy than one-point sampling at C_5.5 because it is visually better suited for fitting during the distribution and elimination phase (Figs. 1B, D). Therefore, in patients at risk of acute renal dysfunction (severe infection and reduced renal function), an accurate AUC assessment with two-point sampling at C_2.5,5.5 is recommended rather than one-point sampling at C_5.5.³⁾

This study indicated that the AUC estimations using the EasyTDM, PAT, and Nichi-Iko models (1, 2, and 3) may be accurate when sampled at steady-state C_5.5 (Table 2). This may be because Bayesian estimation using C_5.5 underestimated the AUC from 0 to 5.5 h post-injection and overestimated the AUC from 5.5 to 12 h (Fig. 1D). The optimal sampling time for AUC_0-∞ estimation in the InsightRx Bayesian software that incorporated Rodvold et al.’s¹⁷⁾ model was 4–6 h following a single vancomycin injection.⁸⁾ In the model by Yasuhara et al.,¹⁴⁾ the optimal sampling time point on day 3 of intermittent vancomycin administration was the trough level, and the subsequent best sampling time point was between post-peak and pre-trough.¹²⁾ The findings of this study underscore these results. Consequently, employing TDM that evaluates AUC with a single-point blood sampling at C_5.5 might help avoid the risk of vancomycin-induced renal injury. This may also help reduce puncture site pain and the workload associated with blood collection for healthcare workers compared with two-point blood sampling.²³⁾ Thus, in patients with normal renal function and a low risk of acute kidney injury, one-point blood sampling at C_5.5 is recommended rather than two-point sampling at C_2.5,5.5.

This study determined the optimal sampling time in young subjects with normal renal function and no infections. However, vancomycin is typically administered to elderly patients with MRSA infection and reduced renal function. Yamamoto et al. reported a population PK analysis involving healthy adults (mean age 21.7 years, CCr 89.3 mL/min) and patients with Gram-positive bacterial infection (65.4 years, 79.6 mL/min).¹⁸⁾ The population mean parameters for central volume of distribution (V1) and peripheral volume of distribution (V2) were estimated as V1 (L) = 0.206 × weight, V2 (L) = 39.4, and V1 (L) = 0.478×weight, V2 (L) = 60.6, respectively. Ducharme et al. also reported an increase in the volume of distribution in elderly infected patients with reduced renal function.²⁴⁾ The increase in the volume of distribution due to infection has been simulated to decrease the maximum blood concentration and increase the trough concentration.¹⁸⁾ The accuracy of Bayesian estimation may be improved by reducing variability in blood concentrations. Therefore, the optimal sampling time points for AUC estimation using the Bayesian software and population PK models in actual patients may occur more frequently than those determined in this study. Further prospective clinical studies in elderly patients with MRSA infection and impaired renal function are needed to validate this hypothesis.

Several issues can be anticipated in incorporating the blood collection times recommended in this study into clinical practice. First, it is imperative that pharmacists and healthcare professionals receive education regarding novel and unfamiliar blood collection timing protocols.^4,23) Second, TDM at C_2.5,5.5 diminishes the available time for simulating and determining the optimal vancomycin dosage in comparison to TDM at C_0,2.5. Finally, while AUC-guided dosing using trough and peak levels is established, there is insufficient clinical evidence for the efficacy and safety of AUC-guided dosing using blood collection at C_5.5.^2,3)

The primary strength of the present study is the assessment of AUC estimation accuracy at various time points, not just the trough and peak levels, using several Bayesian estimation software packages available in Japan. Although the Japanese TDM guidelines for vancomycin indicate the feasibility of Bayesian estimation of the AUC through random time sampling, the scientific rationale for this approach is limited.^3,12) Another strength of this study was the assessment of AUC estimation accuracy at the sampling date following multiple doses of vancomycin. This study may offer valuable data for the analysis of blood vancomycin levels at a steady state.

Nevertheless, this study had some limitations. First, the accuracy of the population PK model, which was constructed based on data from infected patients, was tested in non-infected subjects. As illustrated in Fig. 1, the mean estimated concentration at 1.5 h post-injection was significantly lower than that of the mean measured concentration (p < 0.01). This difference may be associated with the increased distribution volume and vancomycin clearance in infected patients compared with those in healthy individuals.^25,26) Second, the subjects’ CCr was assumed to be 120 mL/min. This limitation may be acceptable because CCr and vancomycin clearance do not linearly correlate in patients with CCr > 85 mL/min.¹⁴⁾ Even in the validation based on the assumption of subject-specific CCr, the optimal sampling times were determined to be the C_2.5,5.5 and C_5.5 blood sampling points. However, when comparing the validation results based on the two different CCr assumptions, variations in the optimal sampling times were observed (Table 2, Supplementary Table 5). The variation in model 3 can be attributed to the fact that its population PK parameters depend solely on CCr. This variation indicates the need to examine in detail the effect of CCr on the accuracy of the population PK models in future studies. Third, although the AUC estimation accuracy in Yasuhara et al.’s¹⁴⁾ model was reduced in patients with impaired renal function,¹²⁾ the accuracy of once-daily dosing in patients with renal failure could not be assessed in this study. Future studies should assess the accuracy of various Bayesian software and population PK models in patients with varying levels of renal function. Finally, we assumed only a fixed maintenance dose of vancomycin (1 g for 1.5 h every 12 h) as the administration method. Increasing the single dose, decreasing the number of doses, and shortening the infusion time increase the concentration difference between the maximum blood concentration and the trough concentration. With such dosing methods that increase blood concentration variability, the accuracy of the validated Bayesian software and population PK model may be reduced compared to the results of this study. The accuracy of the model for various dosing regimens in actual clinical practice necessitates further validation in subsequent studies. Additionally, clinical studies are required to assess the generalizability of these results by evaluating the AUC estimation accuracy, clinical efficacy, and nephrotoxicity of vancomycin.

In conclusion, the results of this study indicate that the optimal sampling times for AUC estimation in vancomycin TDM in healthy adults with normal renal function depend on the Bayesian software and population PK model used. This study indicates that at the steady state with intermittent vancomycin administration for 1.5 h every 12 h, the assessed Bayesian software may perform TDM at two time points (2.5 and 5.5 h) and one point (5.5 h) post-injection. Our results suggest that scheduling blood sampling time points according to the accuracy characteristics of the software may enhance the feasibility of blood sampling for TDM in clinical settings.

Acknowledgments

This work was supported by JSPS KAKENHI Grant Number: 24K18311.

Conflict of Interest

K.Y. was involved in the development of EasyTDM software. The other authors declare no conflict of interest.

Supplementary Materials

This article contains supplementary materials.

REFERENCES

• 1) Liu C, Bayer A, Cosgrove SE, Daum RS, Fridkin SK, Gorwitz RJ, Kaplan SL, Karchmer AW, Levine DP, Murray BE, Rybak MJ, Talan DA, Chambers HF. Clinical practice guidelines by the Infectious Diseases Society of America for the treatment of methicillin-resistant Staphylococcus aureus infections in adults and children. Clin. Infect. Dis., 52, e18–e55 (2011).
• 2) Rybak MJ, Le J, Lodise TP, Levine DP, Bradley JS, Liu C, Mueller BA, Pai MP, Wong-Beringer A, Rotschafer JC, Rodvold KA, Maples HD, Lomaestro B. Therapeutic monitoring of vancomycin for serious methicillin-resistant Staphylococcus aureus infections: a revised consensus guideline and review by the American Society of Health-System Pharmacists, the Infectious Diseases Society of America, the Pediatric Infectious Diseases Society, and the Society of Infectious Diseases Pharmacists. Am. J. Health Syst. Pharm., 77, 835–864 (2020).
• 3) Matsumoto K, Oda K, Shoji K, Hanai Y, Takahashi Y, Fujii S, Hamada Y, Kimura T, Mayumi T, Ueda T, Nakajima K, Takesue Y. Clinical practice guidelines for therapeutic drug monitoring of vancomycin in the framework of model-informed precision dosing: a consensus review by the Japanese Society of Chemotherapy and the Japanese Society of Therapeutic Drug Monitoring. Pharmaceutics, 14, 489 (2022).
• 4) Bradley N, Lee Y, Sadeia M. Assessment of the implementation of AUC dosing and monitoring practices with vancomycin at hospitals across the United States. J. Pharm. Pract., 35, 864–869 (2022).
• 5) Oda K, Hashiguchi Y, Kimura T, Tsuji Y, Shoji K, Takahashi Y, Matsumoto K, Kawamura H, Saito H, Takesue Y. Performance of area under the concentration–time curve estimations of vancomycin with limited sampling by a newly developed web application. Pharm. Res., 38, 637–646 (2021).
• 6) Donagher J, Martin JH, Barras MA. Individualised medicine: why we need Bayesian dosing. Intern. Med. J., 47, 593–600 (2017).
• 7) Turner RB, Kojiro K, Shephard EA, Won R, Chang E, Chan D, Elbarbry F. Review and validation of Bayesian dose-optimizing software and equations for calculation of the vancomycin area under the curve in critically ill patients. Pharmacotherapy, 38, 1174–1183 (2018).
• 8) Shingde RV, Reuter SE, Graham GG, Carland JE, Williams KM, Day RO, Stocker SL. Assessing the accuracy of two Bayesian forecasting programs in estimating vancomycin drug exposure. J. Antimicrob. Chemother., 75, 3293–3302 (2020).
• 9) Neely MN, Youn G, Jones B, Jelliffe RW, Drusano GL, Rodvold KA, Lodise TP. Are vancomycin trough concentrations adequate for optimal dosing? Antimicrob. Agents Chemother., 58, 309–316 (2014).
• 10) Carreno JJ, Lomaestro B, Tietjan J, Lodise TP. Pilot study of a Bayesian approach to estimate vancomycin exposure in obese patients with limited pharmacokinetic sampling. Antimicrob. Agents Chemother., 61, e02478-16 (2017).
• 11) Pai MP, Hong J, Krop L. Peak measurement for vancomycin AUC estimation in obese adults improves precision and lowers bias. Antimicrob. Agents Chemother., 61, e02490-16 (2017).
• 12) Oda K, Yamada T, Matsumoto K, Hanai Y, Ueda T, Samura M, Shigemi A, Jono H, Saito H, Kimura T. Model-informed precision dosing of vancomycin for rapid achievement of target area under the concentration–time curve: a simulation study. Clin. Transl. Sci., 16, 2265–2275 (2023).
• 13) Neely MN, Kato L, Youn G, Kraler L, Bayard D, van Guilder M, Schumitzky A, Yamada W, Jones B, Minejima E. Prospective trial on the use of trough concentration versus area under the curve to determine therapeutic vancomycin dosing. Antimicrob. Agents Chemother., 62, e02042-17 (2018).
• 14) Yasuhara M, Iga T, Zenda H, Okumura K, Oguma T, Yano Y, Hori R. Population pharmacokinetics of vancomycin in Japanese adult patients. Ther. Drug Monit., 20, 139–148 (1998).
• 15) Nakashima M, Katagiri K, Oguma T. Phase I studies on vancomycin hydrochloride for infection. Chemotherapy, 40, 210–214 (1992).
• 16) Rybak MJ. The pharmacokinetic and pharmacodynamic properties of vancomycin. Clin. Infect. Dis., 42 (Suppl. 1), S35–S39 (2006).
• 17) Rodvold KA, Pryka RD, Garrison M, Rotschafer JC. Evaluation of a two-compartment Bayesian forecasting program for predicting vancomycin concentrations. Ther. Drug Monit., 11, 269–275 (1989).
• 18) Yamamoto M, Kuzuya T, Baba H, Yamada K, Nabeshima T. Population pharmacokinetic analysis of vancomycin in patients with gram-positive infections and the influence of infectious disease type. J. Clin. Pharm. Ther., 34, 473–483 (2009).
• 19) Matzke GR, Mcgory RW, Halstenson CE, Keane WF. Pharmacokinetics of vancomycin in patients with various degrees of renal function. Antimicrob. Agents Chemother., 25, 433–437 (1984).
• 20) Sheiner LB, Beal SL. Some suggestions for measuring predictive performance. J. Pharmacokinet. Biopharm., 9, 503–512 (1981).
• 21) Sheiner LB, Beal S, Rosenberg B, Marathe VV. Forecasting individual pharmacokinetics. Clin. Pharmacol. Ther., 26, 294–305 (1979).
• 22) Kanda Y. Investigation of the freely available easy-to-use software ‘EZR’ for medical statistics. Bone Marrow Transplant., 48, 452–458 (2013).
• 23) Kufel WD, Seabury RW, Mogle BT, Beccari MV, Probst LA, Steele JM. Readiness to implement vancomycin monitoring based on area under the concentration–time curve: A cross-sectional survey of a national health consortium. Am. J. Health Syst. Pharm., 76, 889–894 (2019).
• 24) Ducharme MP, Slaughter RL, Edwards DJ. Vancomycin pharmacokinetics in a patient population: effect of age, gender, and body weight. Ther. Drug Monit., 16, 513–518 (1994).
• 25) Katip W, Jaruratanasirikul S, Pattharachayakul S, Wongpoowarak W, Jitsurong A, Lucksiri A. The pharmacokinetics of vancomycin during the initial loading dose in patients with septic shock. Infect. Drug Resist., 9, 253–260 (2016).
• 26) Hochart C, Berthon C, Corm S, Gay J, Cliquennois M, Tricot S, Alfandari S. Vancomycin serum concentration during febrile neutropenia in patients with acute myeloid leukemia. Med. Mal. Infect., 41, 652–656 (2011).