In this prospective cohort study, we evaluated vancomycin TDM using trough concentration-based and AUC24/MIC-based approaches, compared pharmacokinetic equation-based and Bayesian-based methods for AUC24/MIC calculation, and investigated the association between the rs2789047 genetic variant and relevant parameters. Although trough concentrations were strongly correlated with AUC24/MIC values, substantial discordance in therapeutic classification was observed. Higher vancomycin exposure, as reflected by elevated trough concentrations and AUC24/MIC values, was associated with VA-AKI. Additionally, pharmacokinetic differences observed among carriers of the rs2789047 A allele suggest a potential pharmacogenetic influence on vancomycin disposition.

These findings support the transition toward AUC24/MIC–guided monitoring, as recommended in the 2020 consensus guideline, by highlighting the limitations of trough concentration-based monitoring in accurately classifying vancomycin exposure. While trough concentrations were correlated with AUC24/MIC values, correlation alone did not ensure concordant therapeutic categorization, underscoring the importance of assessing agreement in therapeutic classification [1].

Earlier guidelines recommended trough concentrations of 15–20 mg/L as a surrogate for achieving an AUC24/MIC ≥ 400 in MRSA infections. However, the updated 2020 guideline advocates targeting an AUC24/MIC range of 400–600 (MIC = 1 mg/L), which more effectively accounts for interindividual pharmacokinetic variability and is associated with improved safety outcomes. In clinical practice, both first-order pharmacokinetic equations and Bayesian modeling are widely used to estimate AUC24/MIC, yet methodological differences between these approaches may still affect therapeutic classification [1, 4, 13].

Firstly, we compared pharmacokinetic equation-based AUC24/MIC monitoring with trough concentration-based monitoring. A very strong positive correlation was observed between the two methods (r = 0.866, p < 0.001). Previous studies comparing these approaches have reported correlation coefficients ranging from 0.51 to 0.75 [5, 13,14,15]. The variability in correlation strength across studies suggests that trough concentration-based monitoring provides only a partial representation of true AUC24/MIC exposure. This heterogeneity may be attributed to differences in patient populations, pharmacokinetic modeling approaches, methods for MIC determination, as well as the timing and accuracy of sampling strategies.

In our study, the agreement in therapeutic classification between pharmacokinetic equation-based AUC24/MIC monitoring and trough concentration–based monitoring was 63.9%; in other words, the two methods resulted in discordant therapeutic classifications in 36.1% of patients. This finding indicates that, despite a high correlation coefficient, clinically meaningful discrepancies may occur between the two monitoring approaches. Previous studies have reported discordance rates ranging from 21% to 75% [5, 11, 13]. Collectively, these findings indicate that high correlation coefficients do not necessarily translate into concordance in clinical classification.

Secondly, we evaluated trough concentration-based monitoring against Bayesian-based AUC24/MIC monitoring. A very strong correlation was observed between the two methods (r = 0.843; p < 0.001); however, the rate of therapeutic classification disagreement was 33.3%. In line with our results, a large multicenter study analyzing 26,769 dosing records reported a r² value of 0.77 between trough concentrations and Bayesian-based AUC24/MIC, along with a therapeutic classification disagreement rate of 34.3% [16]. Earlier studies reported an r² value of 0.51 [17], while another study identified an r value of 0.427 [18]. These findings indicate that trough concentration-based monitoring demonstrates only limited concordance with AUC24/MIC-based approaches whether derived from pharmacokinetic equations or Bayesian models thereby limiting the reliability of dosing decisions in clinical practice. Notably, the therapeutic classification discordance observed despite strong correlation coefficients indicates that trough levels do not provide adequate accuracy for determining individual exposure. A trough concentration within the therapeutic range may mask supratherapeutic AUC24/MIC exposure, whereas subtherapeutic trough levels may prompt unnecessary dose escalation in patients who already achieve adequate exposure. These differences may have important implications for both treatment failure and toxicity risk.

This study evaluated the correlation and agreement in therapeutic classification between the two AUC24/MIC calculation methods recommended by the 2020 vancomycin consensus guideline: pharmacokinetic equation–based and Bayesian-based approaches. A strong correlation was observed between the two methods (r = 0.934; p < 0.001), with an agreement rate of 86.1%, consistent with previous research reporting a correlation coefficient of r = 0.963 and an agreement rate of 87.4% [12]. Nevertheless, because the pharmacokinetic equation-based approach relies on a one-compartment structural approximation, its performance may be more limited in patients with more complex or unstable pharmacokinetic profiles. In our analysis, 3 of 23 patients (13.0%) classified as supratherapeutic according to the pharmacokinetic equation were categorized within the therapeutic range by the Bayesian model. Conversely, 2 of 13 patients (15.4%) classified as therapeutic by the pharmacokinetic equation were assessed as supratherapeutic using the Bayesian approach. In the referenced study, these proportions were reported as 21.4% and 2.23%, respectively [12]. These results indicate that when AUC24/MIC values are close to therapeutic thresholds, method-dependent differences may still affect classification despite strong overall agreement.

This study found a statistically significant association between the development of VA-AKI and number of concurrent nephrotoxic medications per patient (p = 0.029). Previous studies have demonstrated that the concomitant use of aminoglycosides, amphotericin B, acyclovir, loop diuretics, piperacillin–tazobactam, cephalosporins, and carbapenems with vancomycin substantially increases the risk of VA-AKI [19,20,21]. These findings highlight the need for careful clinical assessment of concurrent nephrotoxic drug use during vancomycin therapy. Owing to the limited sample size with low and heterogeneous frequency of exposure to individual nephrotoxic agents in our cohort, the individual association of each agent with VA-AKI could not be reliably assessed. Accordingly, our findings regarding nephrotoxic co-medication should be interpreted as reflecting cumulative nephrotoxic burden rather than as evidence for the effect of any specific drug combination.

Our research identified a statistically significant correlation between the development of VA-AKI and both baseline creatinine clearance (p = 0.049) and the percentage change in serum creatinine (p < 0.001). In a multicenter retrospective study, the incidence of VA-AKI was significantly higher among patients with a baseline creatinine clearance < 80 mL/min (OR = 7.73, 95% CI: 1.20–49.71) [22]. These results indicate baseline renal function and acute renal changes during treatment could be important factors in VA-AKI. This shows how important it is to thoroughly evaluate renal function before starting treatment.

Our study’s results showed significant associations between the development of VA-AKI and vancomycin trough concentrations (p < 0.001), peak concentrations (p < 0.001), and AUC24/MIC values calculated using both pharmacokinetic equation-based and Bayesian methods (p < 0.001). A systematic review identified trough concentrations > 15 µg/mL (OR: 2.10; 95% CI: 1.43–3.07) and > 20 µg/mL (OR: 2.84; 95% CI: 1.48–5.44) as significant risk factors for nephrotoxicity [19]. Another study reported that patients with trough concentrations of ≥ 20 mg/L had a markedly increased risk of VA-AKI (OR: 3.450; 95% CI: 1.146–10.390) [23]. In addition, a meta-analysis demonstrated that lower AUC24 exposure was associated with a significantly reduced risk of AKI compared with higher AUC24 levels (> 650 mg*h/L) (OR: 0.36; 95% CI: 0.23–0.56) [24]. In line with these findings, our study found that AUC24/MIC values were considerably higher in patients who developed VA-AKI; in other words, total vancomycin exposure is a crucial predictor of nephrotoxicity.

Because of the limited number of patients, genotype groups were analyzed under a dominant model (AC/AA vs. CC) to improve statistical stability. Importantly, the genotype distribution was consistent with HWE (χ² = 0.719, p = 0.396), supporting the reliability of the genotyping data. In our analysis, carriers of the rs2789047 variant exhibited significantly higher trough concentrations and lower Kel, the primary elimination metric, than non-carriers, suggesting a possible effect of this variant on vancomycin elimination. Although the percentage change in serum creatinine was higher in the variant group, the difference was not statistically significant (p = 0.317). In a GWAS of 489 individuals of European American ancestry receiving vancomycin, rs2789047 at chromosome 6q22.31 was significantly associated with increases in serum creatinine during treatment (β = −0.06; p = 1.1 × 10⁻⁷), with the A allele identified as the risk allele [9], although no significant association was reported with vancomycin trough concentrations or Kel. Taken together, these findings suggest that rs2789047 may contribute to interindividual variability in vancomycin pharmacokinetics and renal response, but larger studies are needed to clarify its clinical relevance.