Abstract The predictive performance of the calculation methods for estimating vancomycin trough concentrations (C_trough) in patients with chronic kidney disease (CKD) remains uncertain. This study evaluated the predictive performance of a priori population-based prediction (P_pop) and linear pharmacokinetic (PK) calculation methods (L_pop). Patient characteristics and clinical data on the vancomycin initiation date were used to predict C_trough values. Predictive performance was assessed using median prediction error (Median PE), median percentage prediction error (MDPE), median absolute percentage prediction error (MDAPE), and root mean square prediction error (RMSE). Clinical decision agreement between predicted and measured C_trough values was also assessed. Of 96 participants, the median [interquartile range] age was 67.0 [61.0 to 79.0] years, and the median actual body weight was 55.0 [50.0 to 62.0] kg. The median PE and MDPE were -2.11 µg/mL and -13.54% for P_pop and -0.77 µg/mL and -4.61% for L_pop, respectively. The MDAPE and RMSE were 21.82% and 8.79 µg/mL for P_pop, and 32.17% and 10.23 µg/mL for L_pop, respectively. Clinical decision agreement with measured C_trough was 50.0% for P_pop and 43.8% for L_pop. Both methods underestimated C_trough values in approximately one-third of cases. These findings underscore the need for therapeutic drug monitoring, particularly during the initial days of vancomycin treatment in the CKD population.

Keywords: Vancomycin, Therapeutic drug monitoring, A priori population-based prediction, Linear pharmacokinetic calculation, Chronic kidney disease

Funding: This work was partially supported by Chiang Mai University.

Citation:  Singhan, W., Penthinapong, T., Na Takuathung, M., Vadcharavivad, S., Tidwong, N. 2026. Predictive performance of a priori population-based prediction and linear pharmacokinetic calculation for vancomycin monitoring in patients with chronic kidney disease. Natural and Life Sciences Communications. 25(4): e2026090.

Graphical Abstract:

INTRODUCTION

Vancomycin is commonly used to treat methicillin-resistant Staphylococcus aureus infections. Maintaining vancomycin serum concentration within the therapeutic range is crucial, especially during the early days after treatment initiation (Rybak et al., 2020). An optimal dosage regimen is necessary to avoid overdosage, which can lead to supratherapeutic concentrations and is linked to an elevated risk of acute kidney injury (Kim et al., 2022). Conversely, underdosage resulting in insufficient drug exposure has been associated with increased treatment failure and mortality (Song et al., 2015). Patients with chronic kidney disease (CKD) have decreased renal clearance of the drug (Panwar et al., 2013). Given the susceptibility of patients with CKD to drug toxicity, individualized adjustment of maintenance dosages should be thoroughly deliberated within this population (Bosso et al., 2011). Therefore, therapeutic drug monitoring (TDM) is recommended as a standard practice to guide optimal dosing, with a target area under the concentration-time curve (AUC) of 400 to 600 µg*h/mL (Rybak et al., 2020). Steady-state concentration monitoring is typically performed after the third dose of vancomycin. However, the prolonged elimination of half-life observed among patients with CKD (Šíma et al., 2021) implies that achieving a steady state might not occur during the early days of treatment. Thus, considering concentration monitoring after the first dose would be reasonable.

The published guidelines endorse a Bayesian-derived method incorporating one or two vancomycin concentrations for AUC estimation (Rybak et al., 2020). PrecisePK^®, a widely used Bayesian software, has been reported to be the least biased and most accurate for monitoring critically ill populations compared with other Bayesian software (Turner et al., 2018). Alternatively, a linear pharmacokinetic (PK) calculation using two-point sampling, which has demonstrated comparable AUC values to those of the Bayesian-derived method (Pai et al., 2014), could also be employed when the Bayesian approach is inapplicable (Meng et al., 2019). Certain institutions have adopted Bayesian approaches requiring only a single trough concentration (C_trough) for monitoring (Alzahrani et al., 2023). However, implementation of Bayesian approaches for AUC estimation remains limited in resource-limited settings due to the need for trained personnel and the associated subscription costs. In this context, C_trough may serve as a reasonable alternative marker for therapeutic monitoring (Nix et al., 2022), and linear PK calculation can be used to estimate C_trough. A target range of 15 to 20 µg/mL should be maintained within the first few days following treatment initiation (Rybak et al., 2020).

Although Bayesian-derived calculation methods for predicting C_trough and determining optimal maintenance dosages have demonstrated accuracy and precision in clinical studies, these investigations have primarily involved individuals without renal impairment (Cunio et al., 2020; Shingde et al., 2020; Chen et al., 2022). Linear PK calculations using population PK parameters remain widely employed in clinical practice (Kufel et al., 2019), especially during the early days of vancomycin therapy in patients with CKD, when TDM is typically not yet performed. Bayesian estimation initially uses population pharmacokinetic parameters and patient-specific characteristics to generate a priori prediction. When measured vancomycin concentrations become available, the model is updated to provide an a posteriori estimate. However, in resource-limited settings where vancomycin concentration is unavailable, a priori population-based prediction may still support initial maintenance dosing. This study therefore evaluates the predictive performance of a priori population-based prediction and linear PK calculation methods, using clinical data available at vancomycin initiation, to predict C_trough levels among patients with CKD receiving intermittent intravenous vancomycin.

MATERIALS AND METHODS

Study design and patients

This retrospective study was performed among hospitalized patients receiving intermittent intravenous vancomycin infusions at Maharaj Nakorn Chiang Mai Hospital, a 1,400-bed university-affiliated hospital in Chiang Mai, Thailand. All procedures adhered to the ethical standards of the Helsinki Declaration. Ethics approval for the study was granted through an expedited review by the Research Ethics Committee of the Faculty of Medicine, Chiang Mai University (Approval No: NONE-2565-09084), including a waiver of informed consent. The present study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.

Eligible participants were those with (i) age ≥18 years, (ii) hospital admission between February 2021 and April 2022, (iii) a diagnosis of CKD stages G3 to G5, (iv) treatment with intermittent intravenous vancomycin infusion and (v) at least one serum vancomycin concentration measurement within the first seven days of treatment initiation. Patients with the following conditions were excluded: (i) presented acute kidney injury (AKI) before initiating vancomycin, (ii) received peritoneal dialysis (PD) or continuous renal replacement therapy (CRRT) during the course of vancomycin or (iii) had blood samples for C_trough measurement drawn after a hemodialysis (HD) session.

The first dose of vancomycin ranged from 20 to 35 mg/kg, followed by maintenance doses of 15 to 20 mg/kg administered every 12 to 48 hours, tailored to renal function as determined by creatinine clearance (Rybak et al., 2020). Each dose of vancomycin infusion was given slowly over 1 to 1.5 hours. For C_trough measurement, blood samples were collected immediately before the subsequent vancomycin infusion or prior to hemodialysis. Throughout the study, vancomycin monitoring was conducted using the one-compartment linear PK equation as the standard calculation method.

Data collection 

Patient demographics and clinically relevant information, including dosage regimens and measured C_trough, along with the date and time of blood sample collection for measurement, were extracted from the electronic medical records of Maharaj Nakorn Chiang Mai Hospital. The severity of illness was assessed using the Acute Physiology and Chronic Health Evaluation II score (Knaus et al., 1985) and the Charlson Comorbidity Index (Charlson et al., 1987). Data on the first measured C_trough following vancomycin treatment (C_Measured) including the timing of drug administration and blood sample collection were recorded for each participant. Additional collected data included age, sex, actual body weight, height, serum creatinine levels, details of the vancomycin dosage regimen (the first dose, maintenance dose(s), dosing interval, infusion time and administration time), and the starting date of the vancomycin treatment. This information was used to predict C_trough by employing the two different methods of interest: linear PK calculation and a priori population-based prediction methods.

Calculation of predicted vancomycin trough concentration

The predicted trough concentration using the linear PK method (C_Linear) was calculated based on a one-compartment intravenous PK model (Rushing and Ambrose, 2001; Winter et al., 2010), with all calculations performed in Microsoft Excel (Redmond, WA, USA).

For the first administered vancomycin dose, the volume of distribution (V_d) and maximum concentration after the first dose (C_max,1) were estimated as follows:

where ABW represents the actual body weight in kilograms (kg), and Dose₁ is the first vancomycin dose in milligrams (mg). Next, the elimination rate constant (k_e) was subsequently computed using the following equations:

where CrCl denotes estimated creatinine clearance (L/h) as calculated using the Cockcroft-Gault equation (Cockcroft and Gault, 1976), and CL_VAN is vancomycin clearance (L/h). Finally, these PK parameters were applied to determine the predicted trough concentration after the first dose (C_trough,1) using the following equation:

where t is the time difference between the time to maximum drug concentration (T_max) and the time of concentration measurement (h).

For patients receiving multiple vancomycin doses, C_trough,1 was initially calculated using Equations (1)-(5). This value was then used to estimate the predicted maximum concentration after each subsequent dose as follows:

where C_max,n is the predicted maximum concentration after the n^th dose, C_trough,n-1 is the predicted trough concentration immediately before the n^th dose, and Dose_n is the n^th vancomycin dose in milligrams (mg). The predicted C_max,n was then used to estimate the subsequent trough concentration as follows:

This iterative process was repeated for all subsequent doses administered up to the index date to account for drug accumulation, particularly in patients with CKD.

The predicted trough concentration using a priori population-based prediction method (C_Priori) was calculated using the commercially available PrecisePK^® web-based software (Healthware Inc., San Diego, CA, USA). Patient characteristics at the time of vancomycin initiation—including age, sex, actual body weight, height, and serum creatinine—were entered into the software. The vancomycin dosage regimen, including the first dose, maintenance dose(s), dosing interval, infusion time, and administration time, was also entered into the software, notably without incorporating the measured C_trough. Using this information, PrecisePK^® computed individual PK parameters based on the selected PK models from Rodvold et al. (1988), Adane et al. (2015), or Crass et al. (2018), as appropriate. These parameters were then used to predict the C_trough and area under the concentration-time curve for 0-24 hours (AUC₂₄) for the specified vancomycin regimen.

Vancomycin concentration measurement

Serum concentrations were measured using ARCHITECT iVancomycin 1P30 (Abbott GmbH & Co., Wiesbaden, Germany) on the ARCHITECT i1000SR analyzer (Abbot, IL, USA). This assay, a chemiluminescent microparticle immunoassay, is designed to quantify vancomycin levels in human serum or plasma. Calibration of the assay was performed using six standards, covering a nominal range from 0.0 to 100.0 µg/mL. In this study, the limit of blank was established at 0.27 µg/mL; the limit of detection was 0.42 µg/mL, and the limit of quantitation was 2.50 µg/mL, yielding a validated quantitative range of 2.50 to 100.00 µg/mL. All measured vancomycin concentrations in this study fell within this quantifiable range, with no value observed below the limit of quantitation. The assay demonstrated precision with a total coefficient of variation of ≤10%. Data reduction for calibration curve generation was accomplished using a 4-parameter logistic curve fit data method (4PLC, Y-weighted).

Predictive performance evaluation

The index date was defined as the date of each patient’s most recent vancomycin dose administered prior to concentration monitoring. Measured concentrations were compared with the corresponding predicted concentrations obtained at the same sampling time. The association between measured and predicted trough concentrations was evaluated using Pearson’s correlation coefficient (r). Agreement between measured and predicted concentrations was further assessed using Bland-Altman analysis by calculating the mean difference (bias) and the 95% limits of agreement (mean difference ± 1.96 standard deviations).

Predictive performance was evaluated using median-based error metrics because the prediction-error distributions were non-normal and potentially influenced by extreme values. Predictive bias was assessed using the median prediction error (Median PE) and the median percentage prediction error (MDPE), whereas precision was assessed using the median absolute percentage prediction error (MDAPE). Median PE, MDPE, and MDAPE were calculated using Equations (8)-(10) and presented with their interquartile ranges. Root mean square error (RMSE) was calculated using Equation (11) as an additional measure of the overall magnitude of prediction error.

where n is the total number of vancomycin concentrations.

In this study, the acceptable bias was defined as one half the width of the therapeutic range for vancomycin. Given vancomycin’s therapeutic trough concentration range of 15 to 20 µg/mL, the Median PE value of less than ±2.5 µg/mL and MDPE value of ±15% were deemed acceptable (Sheiner et al., 1979). These cut-points were also utilized in a previous study on the predictive performance of population PK models of vancomycin (Heus et al., 2022). While a specific precision cut-off point was not established, lower MDAPE and RMSE values close to zero indicated precise prediction.

C_Measured, C_Linear, and C_Priori were categorized into three clinically relevant concentration ranges: subtherapeutic (<15 µg/mL), therapeutic (15 to 20 µg/mL) and supratherapeutic (>20 µg/mL). Matched values were subsequently identified when both measured and predicted concentrations aligned within the same concentration category. As indicated by matched percentages, categorical matching demonstrated the impact of different methods on decision-making regarding the increase, decrease, or maintenance of vancomycin dosage in clinical practice.

Statistical analysis

All data analysis was conducted using STATA 14 (Stata Corp LLC, College Station, TX, USA). Continuous variables were expressed either as mean ± standard deviation (SD) or median [interquartile range, IQR] as appropriate. Shapiro-Wilk test was used to assess the normality of data.

RESULTS

Patient demographics and vancomycin administration

A total of 978 patients received intermittent intravenous vancomycin during the study period; 201 met the inclusion criteria. However, 105 patients were excluded due to the following reasons: 61 had C_trough levels measured after HD, 22 had AKI on top of CKD before initiating vancomycin, 17 received PD and 5 received CRRT. Therefore, the study included 96 patients with 96 vancomycin concentrations who were eligible for outcome analysis (Figure 1). The demographic characteristics of the patients included are summarized in Table 1. The majority of the patients were male (55.2%). The median [IQR] age was 67.0 [61.0 to 79.0] years, and the median actual body weight was 55.0 [50.0 to 62.0] kg. Among patients with CKD who were not receiving chronic HD before admission, the median estimated glomerular filtration rate (eGFR), calculated using the CKD-EPI equation (Inker et al., 2021), was 27 [13-38] mL/min/1.73m². The most common first dose administered to the study patients was 1,000 mg (20 mg/kg). Approximately 65% of blood samples for C_trough monitoring were drawn just before the second dose, and the median sampling time after the first vancomycin dose was 45.5 [32.0 to 67.8] hours.

Figure 1. Flow chart outlining patient inclusion and exclusion criteria for the study.

Table 1. Demographic characteristics of study patients. Data are presented as mean ± standard deviation, median [interquartile range] or frequency (percentage).

Note: APACHE, Acute Physiology and Chronic Health Evaluation II; GFR, glomerular filtration rate; IQR, interquartile range; SD, standard deviation; ^a Chronic kidney disease staging was determined based on eGFR calculated by CKD-EPI equation as follows: eGFR 45-59 mL/min/1.73 m², CKD G3a; eGFR 30-44 mL/min/1.73 m², CKD G3b; eGFR 15-29 mL/min/1.73 m², CKD G4; eGFR <15 mL/min/1.73 m², CKD G5; ^bCKD patients without chronic hemodialysis or peritoneal dialysis before admission (n=43).

Correlation between predicted and measured trough concentrations

The median C_Measured was 18.6 [13.8 to 25.0] µg/mL. Pearson correlation analysis demonstrated weak positive correlations between C_Measured and the predicted C_trough values from both methods (Figure 2). Notably, the correlation coefficient for the predicted C_trough from a priori population-based prediction was slightly stronger than that of the linear PK calculation, with Pearson r values of 0.430 and 0.382, respectively.

Figure 2. Correlation between measured and predicted vancomycin trough concentrations. Scatter plots illustrate the relationship between measured and predicted vancomycin trough concentrations using (A) linear pharmacokinetic calculation and (B) a priori population-based prediction methods. The solid line represents the fitted linear regression with its 95% confidence interval (gray shading). The dashed line represents the line of identity.

Predictive performance of calculation methods

The predictive performance of the two prediction methods is summarized in
Table 2. The bias in the predicted C_trough for both methods was within acceptable limits of ±2.5 µg/mL for Median PE and ±15% for MDPE, respectively (Table 2). A priori population-based prediction demonstrated a higher precision, indicated by lower MDAPE and RMSE values (Table 2). Bland–Altman plots showed wide limits of agreement for both methods. Mean bias was closer to zero for the linear pharmacokinetic method, whereas a priori population-based prediction method demonstrated slightly narrower limits of agreement (Figure 3).

Table 2. Predictive performance of the calculation methods for predicting trough concentrations.

Note: PE, prediction error; MDPE, median percentage prediction error; MDAPE, median absolute percentage prediction error; RMSE, root mean square error; CV, coefficient of variation. Data for Median PE, MDPE, and MDAPE are presented as median (interquartile range).

Figure 3. Bland–Altman plots comparing predicted and measured vancomycin trough concentrations using (A) linear pharmacokinetic calculation and (B) a priori population-based prediction methods. The solid red lines represent the mean biases, the dashed lines indicate the 95% limits of agreement (mean bias ± 1.96 SD), and the dotted horizontal lines indicate zero difference.

Clinical decision agreement

The categorical matching between the measured and predicted C_trough values revealed that a priori population-based prediction method had a slightly higher proportion of matched values (50.0%) than the linear PK calculation method (43.8%). Both methods showed a substantially high proportion of underestimation (around 33%). Notably, a priori population-based prediction method had a lower proportion of overestimation than the linear PK calculation method (16.7 vs. 25.0%). Further details of these clinical decision agreements are provided in Table 3.

Table 3. Clinical decision agreement of the methods; n (%).

1. Linear pharmacokinetics calculation.

    B. priori population-based prediction.

DISCUSSION

This study evaluated the predictive performance of the vancomycin calculation methods, specifically a priori population-based prediction and linear PK calculation, in a retrospective cohort of 96 adult patients with CKD. Our findings indicated that both calculation methods demonstrated acceptable levels of bias, albeit with suboptimal precision. Each method's clinical decision agreement rate hovered at approximately 50%, with a priori population-based prediction showing a slightly higher clinical decision agreement rate. Notably, both methods tended to underestimate concentrations in about one third of cases.

A priori population-based prediction method in our study exhibited relatively small median PE and MDPE, suggesting acceptable levels of bias (Table 2). This finding aligns with that of Ohnishi et al. (2001), who evaluated the predictability of the Bayesian-derived calculation method across individuals with varying degrees of renal function. Their approach utilized population PK parameters from Yasuhara et al. (1998) and incorporated selected vancomycin concentration data from each individual for the calculations. Their results demonstrated remarkable predictability with minimal bias (MPE <5 µg/mL), regardless of renal function. However, our study observed a substantially high RMSE, which may be partly explained by the absence of measured vancomycin concentrations for individualized posterior estimation and by differences in the PK models used for the calculations. Unlike Bayesian posterior estimation, a priori population-based prediction method used in our study relied solely on patient covariates and dosing information without incorporating measured C_trough values to update individual PK parameters. In addition, the software primarily employed the PK model proposed by Rodvold et al. (1988) for the calculations. Although this model included patients with a wide range of renal impairments, it had limited representation of CKD patients with severe renal impairment. This limitation could have affected the precision of C_trough predictions in our population, where the majority of patients were classified as CKD stage G5 (eGFR <15 mL/min/1.73 m²). Beyond the method’s predictive performance, real-life clinical decisions regarding maintenance dosage adjustments are influenced mainly by whether C_trough levels fall within the therapeutic range. This study used the clinical decision agreement as a metric to evaluate the concordance between predicted and measured C_trough levels in clinical decision-making. A priori population-based prediction method in our study showed moderate clinical decision agreement, with 50% concordance (Table 3B). This rate was slightly lower than that reported in a related study (Narayan et al., 2021), where various Bayesian-derived calculations yielded between 50 and 60% concordance. The inclusion of critically ill patients in the earlier study, whose PK parameters differed significantly from those of patients with CKD with varying degrees of residual renal function (Ghasemiyeh et al., 2022), could primarily account for this observed discrepancy.

The cost of Bayesian software constrains its accessibility for clinical implementation, particularly in resource-limited settings (Lee et al., 2021). Consequently, one-compartment linear PK calculation has been alternatively applied for vancomycin monitoring among patients with CKD, notwithstanding the paucity of validation (Kufel et al., 2019). In this study, we evaluated the predictive performance of such a method. Since validated one-compartment equations for calculating V_d in CKD patients are limited, we utilized the equation applicable to the general population in this study. Precision was relatively inferior to a priori population-based prediction, as indicated by its higher MDAPE and RMSE values (Table 2). The observed suboptimal predictive performance could be attributed to the use of one-compartment linear PK equations. Although vancomycin has been reported to exhibit two-compartment PK in renal impairment populations (Zaric et al., 2018; Jaisue et al., 2020), employing such model requires more complex calculations involving multiple PK parameters, demanding specialized expertise. A previous study suggested that one-compartment model incorporating a trough-only dataset adequately predicted vancomycin AUC with no significant deviation from the reference values (Maung et al., 2022). Hence, one-compartment PK calculations were employed in our study owing to their simplicity, offering a more pragmatic representation of routine clinical practices. We found that calculation using linear PK equations showed only 44% concordance regarding clinical decision agreement. With this method, approximately one third of the patients predicted as subtherapeutic were actually in the supratherapeutic range when C_trough levels were measured (Table 3A). This discrepancy could lead to a dose increment; thereby, increasing the risk of nephrotoxicity, particularly among individuals with renal impairment (Bosso et al., 2011). Thus, caution should be exercised regarding potential underestimation when applying this method to patients with CKD.

Despite dosing adjustment based on renal function as recommended by guideline (Rybak et al., 2020), a substantial proportion of patients in our cohort (72/96; 75%) exhibited undesirable measured C_trough levels. Concurrently, software-based calculations also indicated undesirable AUC₂₄ levels in the majority of patients (61/96; 63.5%). These deviations from the desired therapeutic range for both C_trough and AUC₂₄ highlight a potential increase in the risk of clinical failure or nephrotoxicity (Kullar et al., 2011; Johnston et al., 2021). This finding strongly supports the guideline recommendation to implement TDM during the initial days of treatment to ensure both efficacy and safety (Rybak et al., 2020), especially among patients with CKD. Nevertheless, in resource-limited settings where the TDM may not be fully implemented due to resource constraints, predicting C_trough using various calculation methods could be considered as an alternative. Notably, our study demonstrated that a priori population-based prediction and linear PK calculation methods tended to underestimate issues in approximately one third of cases. Incorporating individual C_trough values for a posteriori estimation may be a strategy to enhance prediction accuracy (Narayan et al., 2021; Bai et al., 2025), although further research and validation in a CKD population are necessary.

Several strengths have been addressed in this study. To our knowledge, this is the first study to evaluate the predictive performance of a priori population-based prediction and linear PK calculation methods specifically in patients with CKD with moderate to severe renal impairment. This population has often been overlooked in related studies, which primarily focused on critically ill patients. Second, we conducted a comprehensive evaluation of the prediction performance and the clinical decision agreement, ensuring relevance and applicability in real-world clinical practice. This approach effectively demonstrates the influence of predicted C_trough on dosage adjustment decision-making, potentially affecting clinical outcomes. Nonetheless, the study encountered limitations. First, the AUC₂₄ estimation could not be performed using the linear PK calculation method because a single C_trough value was retrospectively collected in each patient.  Although a correlation between C_trough and AUC₂₄ has been evaluated in other population, data specifically in CKD patients remain limited. The suboptimal predictability of C_trough observed in our study could also suggest potential limitations in AUC₂₄ predictability in the CKD population, although further studies are warranted. Second, both calculation methods in this study relied solely on patient-specific characteristics available at the time of vancomycin initiation, without incorporating any measured vancomycin concentrations. This limited data input may have contributed to the observed underprediction, particularly with a priori population-based prediction method. Nonetheless, this approach aligns with real-world TDM practice in CKD population within resource-limited settings, where typically only one vancomycin concentration is obtained during the initial days of treatment. To improve predictive performance, incorporating at least one measured vancomycin concentration as a posteriori dataset should be considered.

CONCLUSION

A priori population-based prediction and linear PK calculation methods demonstrated a lack of precision in predicting vancomycin concentrations in a CKD population. This imprecision could lead to significant discrepancies between predicted and actual measured concentrations. Implementing TDM during the initial days of treatment is advisable in this population.

ACKNOWLEDGEMENTS

The authors would like to thank the staff from Maharaj Nakorn Chiang Mai Hospital and the Faculty of Medicine, Chiang Mai University, for supporting the dataset.

AUTHOR CONTRIBUTIONS

Wanchana Singhan: Conceptualization (Lead), Methodology (Equal), Formal Analysis (Equal), Investigation (Equal), Data Curation (Equal), Writing – Original Draft (Equal), Writing – Review & Editing (Equal), Visualization (Equal), Funding Acquisition (Lead); Thitichaya Penthinapong: Conceptualization (Equal), Methodology (Equal), Formal Analysis (Lead), Investigation (Equal), Data Curation (Equal), Writing – Original Draft (Equal), Writing – Review & Editing (Equal), Visualization (Lead); Mingkwan Na Takuathung: Methodology (Supporting), Resources (Lead), Writing – Review & Editing (Supporting); Somratai Vadcharavivad: Conceptualization (Equal), Methodology (Equal), Writing – Review & Editing (Equal), Supervision (Lead); Nattapong Tidwong: Conceptualization (Equal), Methodology (Lead), Formal Analysis (Equal), Investigation (Equal), Data Curation (Lead), Writing – Original Draft (Lead), Writing – Review & Editing (Lead), Project Administration (Lead).

CONFLICT OF INTEREST

The authors declare no competing interests. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

REFERENCES

Adane, E.D., Herald, M., and Koura, F. 2015. Pharmacokinetics of vancomycin in extremely obese patients with suspected or confirmed Staphylococcus aureus infections. Pharmacotherapy: The Journal of Human Pharmacology and Drug Therapy. 35(2): 127-139. https://doi.org/10.1002/phar.1531

Alzahrani, A.M., Naeem, A., Alzhrani, R.M., Harbi, M.A., Alghamdi, S.A., Karim, S., Ali, A.S., Alsenaini, G., Hasan, H., Alkatheeri, A.A., et al. 2023. The Bayesian-based area under the curve of vancomycin by using a single trough level: An evaluation of accuracy and discordance at tertiary care hospital in KSA. Healthcare. 11(3): 362. https://doi.org/10.3390/healthcare11030362

Bai, G., Qi, H., Huang, Y., Zhang, J., Zhao, H., Wen, R., and Zhang, X. 2025. Predictive performance of Bayesian dosing software for vancomycin in intensive care unit patients. Therapeutic Drug Monitoring. 47(5): 594-602. https://doi.org/10.1097/ftd.0000000000001310

Bosso, J.A., Nappi, J., Rudisill, C., Wellein, M., Bookstaver, P.B., Swindler, J., and Mauldin, P.D. 2011. Relationship between vancomycin trough concentrations and nephrotoxicity: A prospective multicenter trial. Antimicrobial Agents and Chemotherapy. 55(12): 5475-5479. https://doi.org/10.1128/aac.00168-11

Charlson, M.E., Pompei, P., Ales, K.L., and MacKenzie, C.R. 1987. A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. Journal of Chronic Diseases. 40(5): 373-383. https://doi.org/10.1016/0021-9681(87)90171-8

Chen, A., Gupta, A., Do, D.H., and Nazer, L.H. 2022. Bayesian method application: Integrating mathematical modeling into clinical pharmacy through vancomycin therapeutic monitoring. Pharmacology Research & Perspectives. 10(6): e01026. https://doi.org/10.1002/prp2.1026

Cockcroft, D.W. and Gault, M.H. 1976. Prediction of creatinine clearance from serum creatinine. Nephron. 16(1): 31-41. https://doi.org/10.1159/000180580

Crass, R.L., Dunn, R., Hong, J., Krop, L.C., and Pai, M.P. 2018. Dosing vancomycin in the super obese: Less is more. Journal of Antimicrobial Chemotherapy. 73(11): 3081-3086. https://doi.org/10.1093/jac/dky310

Cunio, C.B., Uster, D.W., Carland, J.E., Buscher, H., Liu, Z., Brett, J., Stefani, M., Jones, G.R.D., Day, R.O., Wicha, S.G., et al. 2020. Towards precision dosing of vancomycin in critically ill patients: An evaluation of the predictive performance of pharmacometric models in ICU patients. Clinical Microbiology and Infection. 27(5): 783.e7-783.e14. https://doi.org/10.1016/j.cmi.2020.07.005

Ghasemiyeh, P., Vazin, A., Zand, F., Haem, E., Karimzadeh, I., Azadi, A., Masjedi, M., Sabetian, G., Nikandish, R., and Mohammadi-Samani, S. 2022. Pharmacokinetic assessment of vancomycin in critically ill patients and nephrotoxicity prediction using individualized pharmacokinetic parameters. Frontiers in Pharmacology. 13: 912202. https://doi.org/10.3389/fphar.2022.912202

Heus, A., Uster, D.W., Grootaert, V., Vermeulen, N., Somers, A., In't Veld, D.H., Wicha, S.G., and De Cock, P.A. 2022. Model-informed precision dosing of vancomycin via continuous infusion: A clinical fit-for-purpose evaluation of published PK models. International Journal of Antimicrobial Agents. 59(5): 106579. https://doi.org/10.1016/j.ijantimicag.2022.106579

Inker, L.A., Eneanya, N.D., Coresh, J., Tighiouart, H., Wang, D., Sang, Y., Crews, D.C., Doria, A., Estrella, M.M., Froissart, M., et al. 2021. New creatinine- and cystatin C-based equations to estimate GFR without race. New England Journal of Medicine. 385(19): 1737-1749. https://doi.org/10.1056/nejmoa2102953

Jaisue, S., Pongsakul, C., D'Argenio, D.Z., and Sermsappasuk, P. 2020. Population pharmacokinetic modeling of vancomycin in Thai patients with heterogeneous and unstable renal function. Therapeutic Drug Monitoring. 42(6): 856-865. https://doi.org/10.1097/ftd.0000000000000801

Johnston, M.M., Huang, V., Hall, S.T., Buckley, M.S., Bikin, D., and Barletta, J.F. 2021. Optimizing outcomes using vancomycin therapeutic drug monitoring in patients with MRSA bacteremia: Trough concentrations or area under the curve? Diagnostic Microbiology and Infectious Disease. 101(2): 115442. https://doi.org/10.1016/j.diagmicrobio.2021.115442

Kim, J.Y., Yee, J., Yoon, H.Y., Han, J.M., and Gwak, H.S. 2022. Risk factors for vancomycin-associated acute kidney injury: A systematic review and meta-analysis. British Journal of Clinical Pharmacology. 88(9): 3977-3989. https://doi.org/10.1111/bcp.15429

Knaus, W.A., Draper, E.A., Wagner, D.P., and Zimmerman, J.E. 1985. APACHE II: A severity of disease classification system. Critical Care Medicine. 13(10): 818-829.

Kufel, W.D., Seabury, R.W., Mogle, B.T., Beccari, M.V., Probst, L.A., and Steele, J.M. 2019. Readiness to implement vancomycin monitoring based on area under the concentration-time curve: A cross-sectional survey of a national health consortium. American Journal of Health-System Pharmacy. 76(12): 889-894. https://doi.org/10.1093/ajhp/zxz070

Kullar, R., Davis, S.L., Levine, D.P., and Rybak, M.J. 2011. Impact of vancomycin exposure on outcomes in patients with methicillin-resistant Staphylococcus aureus bacteremia: Support for consensus guidelines suggested targets. Clinical Infectious Diseases. 52(8): 975-981. https://doi.org/10.1093/cid/cir124

Lee, B.V., Fong, G., Bolaris, M., Neely, M., Minejima, E., Kang, A., Lee, G., and Gong, C.L. 2021. Cost-benefit analysis comparing trough, two-level AUC and Bayesian AUC dosing for vancomycin. Clinical Microbiology and Infection. 27(9): 1346.e1-1346.e7. https://doi.org/10.1016/j.cmi.2020.11.008

Maung, N.H., Methaneethorn, J., Wattanavijitkul, T., and Sriboonruang, T. 2022. Comparison of area under the curve for vancomycin from one- and two-compartment models using sparse data. European Journal of Hospital Pharmacy. 29(e1): e57-e62. https://doi.org/10.1136/ejhpharm-2020-002637

Meng, L., Wong, T., Huang, S., Mui, E., Nguyen, V., Espinosa, G., Desai, J., Holubar, M., and Deresinski, S. 2019. Conversion from vancomycin trough concentration-guided dosing to area under the curve-guided dosing using two sample measurements in adults: Implementation at an academic medical center. Pharmacotherapy. 39(4): 433-442. https://doi.org/10.1002/phar.2234

Narayan, S.W., Thoma, Y., Drennan, P.G., Yejin Kim, H., Alffenaar, J.W., Van Hal, S., and Patanwala, A.E. 2021. Predictive performance of Bayesian vancomycin monitoring in the critically ill. Critical Care Medicine. 49(10): e952-e960. https://doi.org/10.1097/ccm.0000000000005062

Nix, D.E., Davis, L.E., and Matthias, K.R. 2022. The relationship of vancomycin 24-hour AUC and trough concentration. American Journal of Health-System Pharmacy. 79(7): 534-539. https://doi.org/10.1093/ajhp/zxab457

Ohnishi, A., Yano, Y., Shimamura, K., and Oguma, T. 2001. Evaluation of Bayesian predictability of vancomycin concentration in patients with various degrees of renal function. Biological and Pharmaceutical Bulletin. 24(12): 1446-1450. https://doi.org/10.1248/bpb.24.1446

Pai, M.P., Neely, M., Rodvold, K.A., and Lodise, T.P. 2014. Innovative approaches to optimizing the delivery of vancomycin in individual patients. Advanced Drug Delivery Reviews. 77: 50-57. https://doi.org/10.1016/j.addr.2014.05.016

Panwar, B., Johnson, V.A., Patel, M., and Balkovetz, D.F. 2013. Risk of vancomycin-induced nephrotoxicity in the population with chronic kidney disease. American Journal of the Medical Sciences. 345(5): 396-399. https://doi.org/10.1097/maj.0b013e318268023d

Rodvold, K.A., Blum, R.A., Fischer, J.H., Zokufa, H.Z., Rotschafer, J.C., Crossley, K.B., and Riff, L.J. 1988. Vancomycin pharmacokinetics in patients with various degrees of renal function. Antimicrobial Agents and Chemotherapy. 32(6): 848-852. https://doi.org/10.1128/aac.32.6.848

Rushing, T.A. and Ambrose, P.J. 2001. Clinical application and evaluation of vancomycin dosing in adults. Journal of Pharmacy Technology. 17(2): 33-38. https://doi.org/10.1177/875512250101700201

Rybak, M.J., Le, J., Lodise, T.P., Levine, D.P., Bradley, J.S., Liu, C., Mueller, B.A., Pai, M.P., Wong-Beringer, A., Rotschafer, J.C., et al. 2020. 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. American Journal of Health-System Pharmacy. 77(11): 835–864. https://doi.org/10.1093/ajhp/zxaa036

Sheiner, L.B., Beal, S., Rosenberg, B., and Marathe, V.V. 1979. Forecasting individual pharmacokinetics. Clinical Pharmacology and Therapeutics. 26(3): 294-305. https://doi.org/10.1002/cpt1979263294

Shingde, R.V., Reuter, S.E., Graham, G.G., Carland, J.E., Williams, K.M., Day, R.O., and Stocker, S.L. 2020. Assessing the accuracy of two Bayesian forecasting programs in estimating vancomycin drug exposure. Journal of Antimicrobial Chemotherapy. 75(11): 3293-3302. https://doi.org/10.1093/jac/dkaa320

Šíma, M., Hartinger, J., Grus, T., and Slanař, O. 2021. Initial dosing of intermittent vancomycin in adults: Estimation of dosing interval in relation to dose and renal function. European Journal of Hospital Pharmacy. 28(5): 276-279. https://doi.org/10.1136/ejhpharm-2019-002013

Song, K.H., Kim, H.B., Kim, H.S., Lee, M.J., Jung, Y., Kim, G., Hwang, J.H., Kim, N.H., Kim, M., Kim, C.J., et al. 2015. Impact of area under the concentration-time curve to minimum inhibitory concentration ratio on vancomycin treatment outcomes in methicillin-resistant Staphylococcus aureus bacteraemia. International Journal of Antimicrobial Agents. 46(6): 689-695. https://doi.org/10.1016/j.ijantimicag.2015.09.010

Turner, R.B., Kojiro, K., Shephard, E.A., Won, R., Chang, E., Chan, D., and Elbarbry, F. 2018. 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(12): 1174-1183. https://doi.org/10.1002/phar.2191

Winter, M.E. 2010. Basic clinical pharmacokinetics. Wolters Kluwer Health, Philadelphia.

Yasuhara, M., Iga, T., Zenda, H., Okumura, K., Oguma, T., Yano, Y., and Hori, R. 1998. Population pharmacokinetics of vancomycin in Japanese adult patients. Therapeutic Drug Monitoring. 20(2): 139-148. https://doi.org/10.1097/00007691-199804000-00003

Zaric, R.Z., Milovanovic, J., Rosic, N., Milovanovic, D., Zecevic, D.R., Folic, M., and Jankovic, S. 2018. Pharmacokinetics of vancomycin in patients with different renal function levels. Open Medicine. 13: 512-519. https://doi.org/10.1515/med-2018-0068      

OPEN access freely available online

Natural and Life Sciences Communications

Chiang Mai University, Thailand. https://cmuj.cmu.ac.th