Optimal Vancomycin AUC Target in Pediatric MRSA Bacteremia: A Bayesian-guided Approach

Abstract

Background:

The optimal vancomycin area under the concentration-time curve (AUC) target for pediatric methicillin-resistant Staphylococcus aures (MRSA) bacteremia remains unclear. This study aimed to determine the optimal AUC target using Bayesian software.

Methods:

A retrospective analysis was conducted on patients 3 months to 18 years of age diagnosed with MRSA bacteremia at Asan Medical Center Children`s Hospital between September 2013 and December 2021. The vancomycin AUC was estimated using Bayesian software, and the relationship between AUC_24–48 and outcomes, including persistent bacteremia ≥48 hours, acute kidney injury (AKI), 30-day all-cause mortality and recurrence, was analyzed.

Results:

Fifty-six cases were included, with a median age of 2.4 years. Most cases were healthcare-associated infections (96.4%) and occurred in patients with underlying conditions (92.9%). Persistent bacteremia, recurrent bacteremia, 30-day all-cause mortality and AKI were observed in 17.9%, 14.8%, 3.7% and 7.1%, respectively. Although an AUC_24–48 ≥400 mg·h/L did not demonstrate clinical benefit, the receiver operating characteristic curve analysis identified 530 mg·h/L as the appropriate AUC_24–48 threshold for predicting persistent bacteremia and AKI. Persistent bacteremia and AKI were more frequent in patients with AUC_24–48 >530 mg·h/L (62.5% vs. 10.4%, P < 0.01; 37.5% vs. 2.1%, P < 0.01). No significant differences in 30-day mortality or recurrence were observed between the groups with AUC_24–48 above and below this threshold.

Conclusions:

AUC_24–48 ≤530 mg·h/L was associated with reduced persistent bacteremia and AKI in pediatric MRSA bacteremia, without significant disadvantages in mortality and recurrence. Future research should explore the lower limit of AUC targets to optimize vancomycin therapy in pediatric patients.

Vancomycin is the most used drug for treating methicillin-resistant Staphylococcus aures (MRSA) infections. The parameter associated with its therapeutic efficacy is the area under the concentration-time curve (AUC) divided by the minimum inhibitory concentration (MIC).¹ However, because of difficulties in directly measuring AUC, trough concentration (C_trough) monitoring has been employed as a surrogate marker since 2009.² Nonetheless, several studies have indicated that C_trough does not consistently reflect AUC values and lacks a significant correlation with outcomes.^3–5 Moreover, a meta-analysis suggested that AUC monitoring may reduce the incidence of acute kidney injury (AKI) compared with C_trough measurement.⁶ The 2020 revised vancomycin therapeutic drug monitoring (TDM) guidelines recommend AUC monitoring and suggest using Bayesian software to estimate AUC.⁷ Bayesian-derived TDM can easily estimate AUC using the vancomycin population pharmacokinetic model and individual patient data.^7,8

Previous studies on adults with severe MRSA infections have recommended an AUC/MIC ≥400 mg·h/L as a therapeutic target, despite variations in AUC estimation methods and the inherent limitations of retrospective studies.^9–11 Given the differences in clinical severity, prognosis and physiological responses to vancomycin in pediatric patients with MRSA bacteremia compared with adults, it remains uncertain whether the AUC/MIC target recommended for adults can be directly extrapolated to the pediatric population.¹² Two retrospective pediatric studies, in which the AUC was calculated using Bayesian estimation or equations, have investigated clinical outcomes with an AUC/MIC ratio of 400 mg·h/L as the cutoff point.^13,14 However, neither study found significant clinical benefits at an AUC/MIC ≥400 mg·h/L. Another retrospective pediatric study observed a statistically significant increase in persistent bacteremia at 48–72 hours at an AUC/MIC <300 mg·h/L.¹⁵ Furthermore, although the guidelines recommend measuring MIC values using the broth microdilution (BMD) method, the significant variability in MIC results across different measurement methods poses challenges for the practical clinical application of AUC/MIC.^5,16,17

Therefore, we conducted a study to determine the optimal pharmacokinetic/pharmacodynamic (PK/PD) target for achieving the best outcomes in pediatric patients with MRSA bacteremia, using AUC as a parameter instead of AUC/MIC while considering clinical outcomes such as persistent bacteremia, 30-day all-cause mortality, recurrence and AKI.

METHODS

Study Population and Design

This retrospective study reviewed the medical records of all patients 3 months to 18 years old diagnosed with MRSA bacteremia at Asan Medical Center Children’s Hospital, a tertiary referral center in Seoul, between September 2013 and December 2021. Patients meeting the following criteria were excluded: those who received vancomycin within 72 hours before the onset of MRSA bacteremia, those with a duration of vancomycin treatment <72 hours, and those who received renal replacement therapy at baseline.

Demographic and clinical characteristics were collected, including underlying diseases, primary infection source, history of infection source control, details of vancomycin administration (timing, dose and interval), use of inotropes, ventilator support, extracorporeal membrane oxygenation, renal replacement therapy and intensive care unit (ICU) stay. Laboratory data, such as blood culture results, vancomycin MIC, vancomycin C_trough and serum creatinine levels, were extracted. Estimated glomerular filtration rate (eGFR) was calculated using the Schwartz equation.¹⁸ Serum vancomycin levels were measured using the Cobas Integra 800 (Roche, Basel, Switzerland), and vancomycin C_trough was generally sampled before the 4^th or 5^th dose within the 30 minutes before the next dose of vancomycin.

The study protocol was approved by the Institutional Review Board of Asan Medical Center (IRB No. 2021-1667).

Outcomes

The primary outcome measures were microbiological and clinical outcomes. As a microbiological outcome, persistent bacteremia was evaluated, which was defined as a positive blood culture persisting for ≥48 hours after the initiation of vancomycin following the onset of MRSA bacteremia. For clinical outcomes, 30-day all-cause mortality, referred to as death within 30 days after the onset of MRSA bacteremia, and recurrence of MRSA bacteremia were assessed. Recurrence was defined as a positive MRSA blood culture at least 7 days after 2 consecutive negative blood cultures or within 30 days after the discontinuation of anti-MRSA therapy. Recurrence cases were not counted separately; however, instances of MRSA bacteremia occurring more than 30 days after the discontinuation of anti-MRSA therapy were classified as separate cases. Cases with 30-day mortality or recurrence were considered to have clinical failure.

The secondary outcome included the occurrence of AKI, which was defined as an increase in serum creatinine by 50% or 0.5 mg/dL (whichever is greater) from baseline,² or the need for renal replacement therapy within 14 days after vancomycin administration.

Definitions

An immunocompromised state was defined as primary immune deficiency, a history of hematopoietic stem cell transplantation within 1 year, and the use of immunosuppressants or corticosteroid therapy with prednisone or its equivalent at doses of ≥20 mg/day (≥2 mg/kg/day for patients weighing less than 10 kg) for at least 14 days within 30 days before MRSA bacteremia. Severe presentation at the onset of MRSA bacteremia was defined as cases requiring new ICU care for the management of sepsis, mechanical ventilation, extracorporeal membrane oxygenation or inotropic drug use within 24 hours before or after the onset of MRSA bacteremia. Polymicrobial bloodstream infection (BSI) was defined as the detection of common commensal microorganisms in 2 or more consecutive blood cultures from separate blood draws, or at least one pathogenic microorganism, in addition to MRSA, in a blood culture. The definitions of healthcare-associated infection, central line-associated bloodstream infection (CLABSI), specific primary site definitions of secondary bacteremia and catheter-related bloodstream infection (CRBSI) were based on respective references.^19–21

The category of “uncertain focus” included cases where the definition of CRBSI was not met within CLABSI, as well as instances where no identifiable focus was present. The “eradicable focus” category included CRBSI, peripheral catheter infection, surgically removable infections and drainable abscesses. The term “eradicated focus” refers to the removal of an infected device or the completion of incision and drainage within 3 days of onset in cases classified as eradicable focus. The “noneradicable focus” included infection sources that are difficult to eliminate, such as pneumonia, infective endocarditis and osteomyelitis without abscess.²²

Microbiologic Data

Bacterial cultures were performed using the BACTEC FX automated incubation system (Becton Dickinson, NJ). Automated susceptibility testing was conducted using MicroScan WalkAway (Siemens, CA), and methicillin resistance was defined as oxacillin MIC ≥4 mg/L.²³ In cases where multiple MRSA strains with different vancomycin MICs were detected in a single episode of MRSA bacteremia, the highest MIC value was selected for analysis.

Estimation of VANCOMYCIN AUC

Estimation of vancomycin AUC was conducted using the Bayesian-guided dosing software, PrecisePK (PrecisePK, CA). The software uses different prior population models to make predictions based on each individual’s PK parameters (age, sex, height, weight and serum creatinine), automatically selecting the model for pediatric patients with serum creatinine levels <0.9 mg/dL according to Le et al,²⁴ and for patients with renal insufficiency and serum creatinine levels ≥0.9 mg/dL according to Le et al.²⁵ Afterward, individual adjustments were made based on the posterior probabilities derived from the obtained vancomycin concentrations to estimate the AUCs.

Receiver Operating Characteristic Analysis for the Optimal Vancomycin AUC Target

In line with recently published guidelines,⁷ an initial analysis was conducted to assess whether an AUC ≥400 mg·h/L (assuming an MIC of 1.0 mg/L) could serve as a surrogate marker for predicting clinical or microbiological outcomes in pediatric patients with MRSA bacteremia. The AUC during the second 24 hours (AUC_24–48), which corresponds to the period when vancomycin concentrations are predominantly measured, was chosen for analysis.^7,26 If an AUC_24–48 ≥400 mg·h/L was determined to be ineffective as a surrogate marker, receiver operating characteristic (ROC) analysis was performed to identify alternative AUC values that could optimize primary and secondary outcomes. To evaluate the predictive performance of vancomycin AUC for the specified outcomes and determine optimal threshold values, the maximized Youden index (J = sensitivity + specificity – 1) was calculated for vancomycin AUC values measured during 2 distinct periods: the first 24 hours (AUC_0–24) and the second 24 hours (AUC_24–48).

Statistical Analysis

Pearson`s correlation coefficient was used to assess the linear relationship between vancomycin C_trough and AUC. For comparisons between the 2 groups, Fisher’s exact test was applied to categorical data, and the Mann-Whitney U test or independent sample t-test was used for numerical data. Logistic regression analysis was performed to identify risk factors associated with outcomes, incorporating the following covariates: age, initial eGFR, immunocompromised state, ICU stay at baseline, polymicrobial BSI, infection source control, severe presentation at onset and vancomycin MIC.

Variables with a P-value <0.1 in the univariate analysis were included in the adjustment to identify risk factors for persistent bacteremia and clinical failure. All analyses were 2-tailed, with P <0.05 considered statistically significant. Statistical analyses were conducted using IBM SPSS Statistics version 23.

RESULT

Study Population

Between September 1, 2013, and December 31, 2021, a total of 79 cases of MRSA bacteremia were observed. Among these, 23 cases were excluded for the following reasons: (1) administration of vancomycin within 72 hours before the onset of MRSA bacteremia (n=5); (2) duration of vancomycin treatment <72 hours (n=10); and (3) receiving renal replacement therapy at baseline (n=8).

Consequently, 56 patients with MRSA bacteremia were included in this study. The study population had a median age of 2.4 years (interquartile range [IQR]: 0.7–6.5), categorized into the following age groups: ≤2 years (n = 33), 3–4 years (n = 5), 5–9 years (n = 9) and ≥10 years (n = 9) (Table 1). Most cases occurred in patients with underlying diseases (92.9%) and were healthcare-associated infections (96.4%). Among these, 35.7% occurred in ICU settings, and 19.6% initially presented with severe conditions, including the new application of mechanical ventilation, inotropic drugs and ICU care. A total of 78.6% (44/56) of cases were vascular catheter-associated BSI, including CLABSI, CRBSI or peripheral phlebitis. Persistent bacteremia lasting ≥48 hours after the initiation of vancomycin was observed in 17.9% (n=10). Clinical failure occurred in 17.9% (n=10), while recurrent bacteremia and the 30-day all-cause mortality rate were 14.8% (n=8) and 3.7% (n=2), respectively. AKI occurred in 7.1% (4/56) of cases, with a median time from the initiation of vancomycin to AKI occurrence of 64.8 hours (IQR: 56.8–73.4).

TABLE 1. .

Demographics, Clinical Characteristics and Outcomes of MRSA Bacteremia

Characteristics	Overall (n = 56)
Age, years	2.4 (0.7–6.5)
Male	29 (51.8%)
Body weight, kg	9.8 (6.9–18.0)
Baseline serum creatinine, mg/dL	0.37 (0.21–0.46)
Initial eGFR, mL/min/1.732 m2	154.3 (106.1–211.4)
Presence of underlying disease	52 (92.9%)
Hemato-oncologic diseases	11 (19.6%)
Congenital heart diseases	12 (21.4%)
Chronic gastrointestinal diseases	10 (17.9%)
Chronic lung diseases	9 (16.1%)
Neuromuscular diseases	3 (5.4%)
Other genetic diseases*	6 (10.7%)
Primary immunodeficiency	1 (1.8%)
Immunocompromised state	17 (30.4%)
Healthcare-associated infection	54 (96.4%)
ICU stay at baseline	20 (35.7%)
Severe presentation at onset	11 (19.6%)
Polymicrobial BSI†	12 (21.4%)
Primary focus of bacteremia
Without a focus	1 (1.8%)
Definite CRBSI	20 (35.7%)
CLABSI except CRBSI	20 (35.7%)
Peripheral phlebitis	4 (7.1%)
Deep-seated infection‡	6 (10.7%)
Pneumonia	4 (7.1%)
Osteomyelitis	1 (1.8%)
Control of infection sources
Uncertain focus	21 (37.5%)
Eradicated focus§	15 (26.8%)
Noneradicable/not eradicated focus¶	20 (35.7%)
Outcomes
Persistent bacteremia ≥48 hours	10 (17.9%)
Recurrence#	8 (14.8%)
30-day all-cause mortality#	2 (3.7%)
Acute kidney injury	4 (7.1%)

Data presented as median (interquartile range) or as n (%). P value is by Fisher exact test, Mann-Whitney U test, or independent samples t-test.

^*Includes mitochondrial disease (n = 2), congenital insensitivity to pain with anhidrosis (n = 1), SLC7a7 mutation (n = 1), arthrogryposis multiplex congenita (n = 1) and Miller-Dieker syndrome (n = 1).

^†Includes Enterococcus faecalis (n = 4), Enterococcus faecium (n = 2), Streptococcus pneumoniae (n = 2), Staphylococcus epidermidis (n = 2), Candida albicans (n = 2), Burkholderia spp. (n = 1), Candida glabrata (n = 1), Acinetobacter baumannii (n = 1) and Micrococcus spp. (n = 1).

^‡Includes mediastinitis after cardio-surgery (n = 3), mediastinitis associated with tracheostomy site infection (n = 1), retropharyngeal abscess with mediastinitis (n = 1) and osteomyelitis associated with coccyx sores (n = 1).

^§Eradicated focus included CRBSI (n = 9), peripheral phlebitis (n = 3) and deep-seated infection (n = 3).

^¶Not eradicated focus included CRBSI (n = 11), deep-seated infection (n = 3) and thrombophlebitis with skin and subcutaneous abscess (n = 1).

#Data are missing for 2 cases due to loss of follow-up.

Vancomycin PK/PD Parameters

The initial vancomycin dose, initial C_trough, AUC_0–24 and AUC_24–48 are shown in Table 2. The median time of measuring the initial C_trough was 30.8 hours (IQR: 19.8–50.3) after the initiation of vancomycin. Among the MRSA isolates, 76.8% had a vancomycin MIC of ≤1.0 mg/L, with 42 isolates (75%) showing an MIC of 1.0 mg/L and 1 isolate (1.8%) showing an MIC of 0.5 mg/L. Meanwhile, 13 isolates (23.2%) had an MIC of 2.0 mg/L. A moderate positive correlation (r = 0.646; P < 0.01) was observed between the average C_trough within the first 72 hours and AUC_24–48.

TABLE 2. .

Vancomycin Pharmacokinetic/Pharmacodynamic Parameters

Characteristics	Overall (n = 56)
Initial vancomycin dose, mg/kg/day	41.1 (40.0–59.2)
Vancomycin dosing interval
q 6 hours	50 (89.3)
q 8–12 hours	6 (10.7)
Adjustment within 72 hours	14 (25.0)
Initial Ctrough, mg/L	7.5 (5.2–12.2)
AUC0–24, mg·h/L	323.1 (256.7–408.1)
AUC24–48, mg·h/L	382.2 (294.0–479.1)
Vancomycin MIC
0.5 mg/L	1 (1.8)
1.0 mg/L	42 (75.0)
2.0 mg/L	13 (23.2)

Data presented as median (interquartile range) or as n (%).

Significance of Vancomycin AUC_24–48 of 400 mg·h/L

Patients with an AUC_24–48 ≥400 mg·h/L had similar demographic and clinical characteristics to those with an AUC_24–48 <400 mg·h/L, except for a higher median age (2.8 vs 1.0 year, P < 0.01) (Table 3). Primary outcomes, including clinical failure and persistent bacteremia, did not differ significantly between the 2 groups. However, AKI occurred exclusively in patients with an AUC_24–48 ≥400 mg·h/L. Among patients with AKI, AUC_24–48 values were significantly higher (572.1 mg·h/L for AKI vs. 374.1 mg·h/L for non-AKI, P = 0.01), despite these patients receiving slightly lower initial vancomycin doses (30.6 vs. 42.6 mg/kg/day, P = 0.06).

TABLE 3. .

Comparison of Baseline Characteristics and Outcomes of MRSA Bacteremia Stratified by AUC_24–48 of 400 mg·h/L and 530 mg·h/L.

Characteristics	AUC24–48, mg·h/L*			AUC24–48, mg·h/L
	<400 (n = 31)	≥400 (n = 25)	P	≤530 (n = 48)	>530 (n = 8)	P
Age, years	1.0 (0.5–5.7)	2.8 (2.2–10.0)	<0.01	2.0 (0.6–6.4)	3.1 (2.5–7.8)	0.06
Male	19 (61.3%)	10 (40.0%)	0.18	26 (54.2%)	3 (37.5%)	0.46
Body weight, kg	8.4 (6.3–12.3)	12.0 (8.0–30.7)	0.03	9.1 (6.8–16.1)	12.7 (9.1–26.4)	0.22
Baseline serum creatinine, mg/dL	0.35 (0.21–0.41)	0.40 (0.27–0.50)	0.27	0.36 (0.21–0.42)	0.48 (0.32–0.68)	0.07
Initial eGFR, mL/min/1.732 m2	137.1 (105.7–180.0)	163.0 (107.4–229.2)	0.19	156.9 (109.8–210.3)	122.9 (69.9–221.3)	0.50
Presence of underlying disease	29 (93.5%)	23 (92.0%)	>0.99	44 (91.7%)	8 (100.0%)	>0.99
Hemato-oncologic diseases	3 (9.7%)	8 (32.0%)	0.05	9 (18.8%)	2 (25.0%)	0.65
Congenital heart diseases	7 (22.6%)	5 (20.0%)	>0.99	9 (18.8%)	3 (37.5%)	0.35
Chronic gastrointestinal diseases	6 (19.4%)	4 (16.0%)	>0.99	9 (18.8%)	1 (12.5%)	>0.99
Chronic lung diseases	7 (22.6%)	2 (8.0%)	0.17	8 (16.7%)	1 (12.5%)	>0.99
Neuromuscular diseases	1 (3.2%)	2 (8.0%)	0.58	3 (6.3%)	0	>0.99
Other genetic diseases	5 (16.1%)	1 (4.0%)	0.21	5 (10.4%)	1 (12.5%)	>0.99
Primary immunodeficiency	0	1 (4.0%)	0.45	1 (2.1%)	0	>0.99
Immunocompromised state	6 (19.4%)	11 (44.0%)	0.08	14 (29.2%)	3 (37.5%)	0.69
Healthcare-associated infection	30 (96.8%)	24 (96.0%)	>0.99	46 (95.8%)	8 (100.0%)	>0.99
ICU stay at baseline	13 (41.9%)	7 (28.0%)	0.40	17 (35.4%)	3 (37.5%)	>0.99
Severe presentation at onset	5 (16.1%)	6 (24.0%)	0.51	9 (18.8%)	2 (25.0%)	0.65
Polymicrobial BSI	7 (22.6%)	5 (20.0%)	>0.99	11 (22.9%)	1 (12.5%)	0.67
Primary focus of bacteremia
Without a focus	0	1 (4.0%)	0.45	1 (2.1%)	0	>0.99
Definite CRBSI	10 (32.3%)	10 (40.0%)	0.59	15 (31.3%)	5 (62.5%)	0.12
CLABSI except CRBSI	11 (35.5%)	9 (36.0%)	>0.99	19 (39.6%)	1 (12.5%)	0.24
Peripheral phlebitis	3 (9.7%)	1 (4.0%)	0.62	4 (8.3%)	0	>0.99
Deep-seated infection	4 (12.9%)	2 (8.0%)	0.68	6 (12.5%)	0	0.58
Pneumonia	2 (6.5%)	2 (8.0%)	>0.99	2 (4.2%)	2 (25.0%)	0.09
Osteomyelitis	1 (3.2%)	0	>0.99	1 (2.1%)	0	>0.99
Control of infection sources
Uncertain focus	11 (35.5%)	10 (40.0%)	0.79	20 (41.7%)	1 (12.5%)	0.24
Eradicated focus	8 (25.8%)	7 (28.0%)	>0.99	13 (27.1%)	2 (25.0%)	>0.99
Noneradicable/not eradicated focus	12 (38.7%)	8 (32.0%)	0.78	15 (31.3%)	5 (62.5%)	0.12
Vancomycin MIC of 2.0 mg/L	8 (25.8%)	5 (20.0%)		11 (22.9%)	2 (25.0%)	>0.99
Outcomes
Persistent bacteremia ≥48 hours	3 (9.7%)	7 (28.0%)	0.09	5 (10.4%)	5 (62.5%)	<0.01
Recurrence†	4 (13.8%)	4 (16.0%)	>0.99	7 (15.2%)	1 (12.5%)	>0.99
30-day all-cause mortality†	0	2 (8.0%)	0.21	1 (2.2%)	1 (12.5%)	0.28
Acute kidney injury	0	4 (16.0%)	0.03	1 (2.1%)	3 (37.5%)	<0.01

Data presented as median (interquartile range) or as n (%). P value is by Fisher exact test or Mann-Whitney U test.

^*Because the broth microdilution method was not used for MIC measurement, AUC was used instead of AUC/MIC (assuming an MIC of 1 mg/L).

^†Data are missing for 2 cases due to loss of follow-up.

Determining Vancomycin AUC Thresholds Using ROC Analysis

Compared with AUC_0–24, AUC_24–48 demonstrated a significantly higher area under the ROC curve for predicting persistent bacteremia (0.735; P = 0.02) and AKI (0.861; P = 0.02) (Table 4). However, neither AUC_0–24 nor AUC_24–48 showed statistically significant performance as classifiers for predicting clinical failure.

TABLE 4. .

ROC Analysis for the Determination of Vancomycin AUC for Predicting Outcomes

Parameters	Persistent Bacteremia*		Clinical Failure†		Acute Kidney Injury
	Vancomycin AUC (J)‡	Area Under the ROC Curve (P)	Vancomycin AUC (J)‡	Area Under the ROC Curve (P)	Vancomycin AUC (J)‡	Area Under the ROC Curve (P)
AUC0–24	478.2 (0.378)	0.659 (0.12)	N.A.§	0.443 (0.58)	417.3 (0.558)	0.784 (0.06)
AUC24–48	528.2 (0.535)	0.735 (0.02)	458.8 (0.282)	0.507 (0.95)	536.8 (0.654)	0.861 (0.02)

^*Persistent bacteremia was evaluated as a microbiological outcome with a positive blood culture persisting for ≥48 hours after the initiation of vancomycin.

^†Cases with 30-day mortality or recurrence were considered to have clinical failure.

^‡Vancomycin AUC thresholds and their corresponding maximized Youden index (J) for predicting each specific outcome.

^§The area under the ROC curve for clinical failure in AUC_0–24 was found to be below 0.5 (0.443), indicating that it is not statistically valid.

N.A. indicates not applicable.

ROC analysis identified an AUC_24–48 of 528.2 mg·h/L as a threshold for predicting persistent bacteremia, resulting in a maximized Youden index of 0.535, with a sensitivity of 60% and specificity of 93.5%. The incidence of persistent bacteremia was significantly higher in the group with an AUC_24–48 ≥528.2 mg·h/L compared with the group with an AUC_24–48 <528.2 mg·h/L (66.7% vs. 8.5%, P < 0.01).

For predicting AKI, the Youden index was maximized at an AUC_24–48 of 536.8 mg·h/L (0.654), with a sensitivity of 75% and specificity of 90.4%. The incidence of AKI was significantly higher in the group with an AUC_24–48 ≥536.8 mg·h/L compared with the group with an AUC_24–48 <536.8 mg·h/L (37.5% vs. 2.1%, respectively, P < 0.01).

In the ROC curve analysis, no statistically significant lower bound was identified for predicting persistent bacteremia, AKI, or clinical failure within the range of AUC_24–48 <528.2 mg·h/L. Given these circumstances, the approximate median value of 530 mg·h/L for AUC_24–48, considered the target upper bound for predicting persistent bacteremia and AKI, was used as the reference point for subsequent analyses.

Comparison Between the 2 Groups Stratified by an AUC_24–48 of 530 mg·h/L

When comparing the 2 groups based on an AUC_24–48 of 530 mg·h/L, the demographics and clinical characteristics were similar (Table 3). Persistent bacteremia and AKI were observed more frequently in the group with an AUC_24–48 >530 mg·h/L (62.5% vs. 10.4%, P < 0.01; and 37.5% vs 2.1%, P < 0.01, respectively). However, clinical failure, including 30-day mortality and recurrent bacteremia, did not show statistically significant differences between the 2 groups (25.0% vs. 17.4%, P = 0.63).

When classified by an age threshold of 10 years, within the younger group (<10 years), persistent bacteremia (50.0% vs. 7.3%, P = 0.02) and AKI (33.3% vs. 2.4%, P = 0.04) were markedly higher in the higher vancomycin exposure group (Table, Supplemental Digital Content 1, https://links.lww.com/INF/G209). However, in the group ≥10 years of age, a similar trend was observed, but statistical significance was not found for either persistent bacteremia (100% vs. 28.6%, P = 0.17) or AKI (50% vs. 0%, P = 0.42).

Risk Factors Associated With Microbiological and Clinical Failure

The group with persistent bacteremia showed a trend, though not statistically significant, toward initiating a dosing interval of 8–12 hours rather than 6 hours, compared with the nonpersistent bacteremia group (Table, Supplemental Digital Content 2, https://links.lww.com/INF/G209). However, the initial vancomycin dose and the proportion of severe presentations were similar between the 2 groups. In addition, the group with persistent bacteremia had a higher proportion of noneradicable or noneradicated foci (70.0% vs. 28.3%, P = 0.03).

In the univariate logistic regression analysis, persistent bacteremia was associated with a noneradicable or not eradicated focus [odds ratio (OR): 5.92; 95% confidence interval (CI): 1.33–26.47; P = 0.02] and an AUC_24–48 >530 mg·h/L (OR: 14.33; 95% CI: 2.61–78.84; P < 0.01) (Table 5). After adjusting for variables with P <0.1 in the univariate analysis, only an AUC_24–48 >530 mg·h/L was independently associated with persistent bacteremia (adjusted OR: 11.63; 95% CI: 1.91–70.82; P < 0.01).

TABLE 5. .

Risk Factors Associated With Microbiological and Clinical Failure of MRSA Bacteremia

Variables	Persistent Bacteremia		Clinical Failure†
	OR (95% CI)	Adjusted OR (95% CI)*	OR (95% CI)
Age <12 months	0.52 (0.10–2.74)	N.A.	1.02 (0.23–4.58)
Initial eGFR, mL/min/1.732 m2	1.00 (0.99–1.01)	N.A.	1.01 (1.00–1.01)
Immunocompromised state	0.21 (0.02–1.79)	N.A.	0.92 (0.21–4.09)
ICU stay at baseline	0.73 (0.17–3.21)	N.A.	0.36 (0.07–1.90)
Polymicrobial BSI	0.35 (0.04–3.11)	N.A.	0.85 (0.16–4.66)
Noneradicable/not eradicated focus	5.92 (1.33–26.47)	4.72 (0.91–24.46)	3.21 (0.78–13.24)
Severe presentation at onset	1.03 (0.19–5.70)	N.A.	1.93 (0.41–9.13)
Vancomycin MIC of 2.0 mg/L	2.74 (0.64–11.80)	N.A.	3.00 (0.68–13.17)
Vancomycin AUC24–48 >530 mg·h/L	14.33 (2.61–78.84)	11.63 (1.91–70.82)	1.32 (0.23–7.59)

^*Variables associated with outcomes with P <0.1 in univariate analysis, including noneradicable/not eradicated focus and vancomycin AUC_24–48 >530 mg·h/L, were included in a multivariate model for persistent bacteremia. Hosmer-Lemeshow goodness-of-fit test: P = 0.232.

^†Since no variables showed a P-value <0.1 in the univariate analysis for clinical failure, a logis.tic regression analysis including appropriate variables for adjustment could not be performed.

N.A. indicates not applicable.

No statistically significant risk factors were identified for clinical failure, including 30-day all-cause mortality and recurrence.

DISCUSSION

In this retrospective study involving pediatric MRSA bacteremia cases with a 30-day all-cause mortality rate of 3.7%, we suggest an optimal target for vancomycin AUC simulated using Bayesian dosing software. An AUC_24–48 ≤530 mg·h/L may serve as a threshold for improved outcomes, reducing the risk of persistent bacteremia ≥48 hours and AKI.

Bayesian-derived TDM is more adaptive to the patient’s dynamic physiological changes and has the advantage of requiring fewer constraints on the timing or frequency of samplings, even before steady-state is achieved.^7,8,27 Though still evolving, this approach shows promise in optimizing dosing strategies for patient outcomes and may be cost-effective, depending on the institution’s size.^28,29 In pediatric patients with MRSA bacteremia, no study has proposed an applicable Bayesian-derived AUC target.¹³ One study utilizing an equation model indicated an association between an initial AUC/MIC <300 mg·h/L and persistent bacteremia at 48–72 hours.¹⁵ However, the study relied on equations using only C_trough to estimate the AUC, and as MIC values were identified using MicroScan, the interpretation of AUC/MIC values cannot be directly compared with those derived from the BMD method. For MIC determination of S. aureus isolates, this study adopted the MicroScan method, which tends to overcall MIC compared with the BMD method.^16,17 These discrepancies in MIC results depending on measurement methods raise challenges in establishing interinstitutional targets of AUC/MIC as parameters for appropriate PK/PD targets.^5,16,26 Consequently, AUC itself may be a more viable monitoring target. In addition, the 48-hour time point following antibiotic administration serves as a critical juncture, marking both the identification of the causative pathogen and the assessment of treatment response. This time window is corroborated by a recent multicenter prospective study, which indicated that persistence at day 3 marked an inflection point for increased mortality.³⁰ Therefore, the timeframes of AUC_24–48 and persistent bacteremia ≥48 hours utilized in this study are expected to be clinically practical.

A recent multicenter observational study on hospitalized adults with MRSA bacteremia treated with vancomycin reported that higher vancomycin exposure with an AUC >515 mg·h/L did not provide advantages, nor did it act as a worsening factor in terms of treatment failure, defined as 30-day mortality or persistent bacteremia lasting ≥7 days.²⁶ However, a study by McNeil et al³¹ on pediatric healthcare-associated S. aureus bacteremia observed a trend toward a longer duration of bacteremia and higher mortality in the higher vancomycin exposure group (vancomycin C_trough > 15 mg/L). Similarly, in our study, the higher vancomycin exposure group (AUC_24–48 > 530 mg·h/L) showed increased rates of AKI, along with trends toward higher 30-day mortality and greater rates of persistent bacteremia at 48 hours. These findings suggest that higher vancomycin exposure is associated not only with increased AKI but also with poorer clinical and microbiological outcomes in pediatric MRSA bacteremia. The exact explanation for these findings remains unclear. While there were no significant differences in baseline glomerular filtration rate between the higher and lower vancomycin exposure groups, patients in the higher exposure group tended to present with more severe conditions at baseline (although this trend was not statistically significant), which may have led to more aggressive dosing. Moreover, consistently worse clinical and microbiological outcomes were observed in the higher vancomycin exposure group, regardless of age group. This suggests that the observed outcomes are unlikely to be attributed to age-related differences. However, these findings should be interpreted cautiously due to the small sample size, and the role of noneradicable/not eradicated foci in persistent bacteremia cannot be conclusively excluded. Further large-scale studies on vancomycin exposure in pediatric bacteremia will be necessary to elucidate these findings more clearly.

This study could not evaluate the direct effect of AUC on mortality, as only 2 fatal cases were observed during the study period. In addition, no patients were diagnosed with infective endocarditis. The rates of infective endocarditis in pediatric patients with S. aureus bacteremia range from 1.7% to 2.8%,^32,33 while the incidence of infective endocarditis in adult is as high as 29%.^26,34 The lack of a discernible association between AUC and clinical outcomes may be influenced by the relatively low mortality and severity rate among pediatric patients with MRSA bacteremia compared with adults.^26,33,34

This study had some limitations. First, the single-center, small sample size and retrospective nature of the study may limit the generalizability of the findings. This study primarily focused on nondialysis patients with mostly healthcare-associated infections and relatively few musculoskeletal infections, making it difficult to represent patients with community-acquired infections or those receiving renal replacement therapy. In addition, since this study did not include cases of infective endocarditis or central nervous system infections, the proportion of high-burden diseases may be relatively low. Therefore, caution is warranted when applying a target AUC ≤530 mg·h/L based solely on the results of this study to patients with these diseases, which are known to have high mortality and treatment failure rates.^26,32,35 Second, differences in disease patterns across age groups may exist. Identifying trends from younger children to older children and adults would be particularly helpful in guiding treatment strategies, especially given the current lack of pediatric research. Regardless of the age classification used—whether <10 years or ≥10 years—participants in the higher vancomycin exposure group showed higher rates of persistent bacteremia and AKI, with the statistical limitations imposed by the small sample size in the ≥10 years age group. Patients younger than 3 months had to be excluded because of the inconsistency of the AUC estimation method, although a higher mortality rate in this age group among pediatric patients with MRSA bacteremia has been reported.³⁶ However, since this age group, which can include many preterm infants in the neonatal ICU, may have very different characteristics from the pediatric population over that age, it would be more appropriate to investigate this as a separate study. Third, prior studies adopted by the Bayesian software for estimating AUC primarily comprised Hispanic and Caucasian races/ethnicities, with Asians being in the minority (3%–4%).^24,25 As most of the patients in this study were East Asians, there may have been some potential for error in the AUC estimation. Finally, in this study, a vancomycin MIC of 2.0 mg/L, as determined by the Microscan method, was not associated with persistent bacteremia. However, this finding may be attributed to the small sample size, and different MIC measurement methods could yield varying results. We believe that a larger, standardized study will be essential to address these issues comprehensively.

In conclusion, an AUC_24–48 ≤530 mg·h/L was associated with reduced persistent bacteremia and AKI without significant disadvantages in mortality and recurrence, suggesting it as the upper bound of the optimal target for AUC in children older than 3 months. However, we did not find evidence in this study to support that an AUC_24–48 ≥400 mg·h/L serves as the lower bound or to identify an alternative value. Future research should focus on validating these findings in larger, multicenter studies and exploring the potential benefits of more precise lower AUC targets to further optimize vancomycin therapy in pediatric patients.