1104. Comparison of Antibiotic Sampling Techniques: Predicting Plasma Vancomycin Concentrations Using Volumetric Absorptive Microsampling (VAMS) from Capillary and Venous/Arterial Whole Blood

Abstract Background Therapeutic drug monitoring (TDM) is paramount to optimize the safety and efficacy of vancomycin (VAN). In children, TDM is challenged by difficulty in obtaining venous samples, impeding timely sampling. We assessed the ability of volumetric absorptive microsampling (VAMS) as a novel, whole blood sampling technique to predict plasma VAN concentrations in plasma. Methods We conducted a prospective pilot study among critically ill children prescribed VAN for clinical care. Coincident with VAN TDM in plasma (P), we collected 20 µL of capillary whole blood (C) and venous/arterial whole blood (V) using VAMS. Paired VAMS-P samples drawn >5 mins apart and VAMS samples with over- or under-loaded filter tip on visual inspection were excluded. Plasma concentrations were measured via chemiluminescent immunoassay in the Chemistry Laboratory. VAMS C and V concentrations were measured using LC/MS in the Bioanalytic Core Laboratory. Plasma concentrations were predicted from whole blood VAMS with Passing-Bablok regression using 3 methods: 1) uncorrected VAMS measures, 2) hematocrit-corrected VAMS, and 3) lab-corrected VAMS (Figure 1). We then assessed bias, imprecision, and accuracy of plasma predictions from VAMS (C and V) as compared to coincident P concentrations for each technique (Figure 1). Figure 1. Methods for relating whole blood vancomycin concentrations collected via VAMS to plasma concentrations and measure to evaluate predictive performance. Results Paired samples were collected from 31 enrolled subjects (Figure 2), with a median age of 3.3 years (range 0.1-17.9). Measured P concentrations ranged from 4.6 - 54.9 mg/L. 11 C samples (29%) and 3 V samples (10%) were excluded due to collection issues. Prediction results are shown in Figure 3. The 3 prediction techniques had similar performance characteristics, with each method displaying minimal bias (-0.4-2.0%) and reasonable imprecision (13.7-20.2%). The accuracy of prediction of P concentrations using VAMS was better for V than C samples. Figure 2. Flow diagram from sample collection to evaluation. Abbreviations: C-P, capillary VAMS-plasma; V-P, venous/arterial VAMS-plasma; VAMS, volumetric absorptive microsampling. Figure 3. Performance of 3 techniques to predict plasma vancomycin concentrations using whole blood collected via VAMS. Conclusion Our pilot highlights the challenges of using VAMS for TDM. Sample collection issues were common. When VAMS is used, education on collection techniques is imperative. The predictive performance of VAMS was modest and V sampling had higher accuracy than C, although our sample size was small. Larger studies will be needed to further evaluate the predictive performance of the regression equations derived by our study. Disclosures Kevin J. Downes, MD, Merck, Inc. (Grant/Research Support)


Session: P-62. PK/PD Studies
Background. Therapeutic drug monitoring (TDM) is paramount to optimize the safety and efficacy of vancomycin (VAN). In children, TDM is challenged by difficulty in obtaining venous samples, impeding timely sampling. We assessed the ability of volumetric absorptive microsampling (VAMS) as a novel, whole blood sampling technique to predict plasma VAN concentrations in plasma.
Methods. We conducted a prospective pilot study among critically ill children prescribed VAN for clinical care. Coincident with VAN TDM in plasma (P), we collected 20 µL of capillary whole blood (C) and venous/arterial whole blood (V) using VAMS. Paired VAMS-P samples drawn >5 mins apart and VAMS samples with over-or under-loaded filter tip on visual inspection were excluded. Plasma concentrations were measured via chemiluminescent immunoassay in the Chemistry Laboratory. VAMS C and V concentrations were measured using LC/MS in the Bioanalytic Core Laboratory. Plasma concentrations were predicted from whole blood VAMS with Passing-Bablok regression using 3 methods: 1) uncorrected VAMS measures, 2) hematocrit-corrected VAMS, and 3) lab-corrected VAMS (Figure 1). We then assessed bias, imprecision, and accuracy of plasma predictions from VAMS (C and V) as compared to coincident P concentrations for each technique (Figure 1). Figure 1. Methods for relating whole blood vancomycin concentrations collected via VAMS to plasma concentrations and measure to evaluate predictive performance.
Results. Paired samples were collected from 31 enrolled subjects (Figure 2), with a median age of 3.3 years (range 0.1-17.9). Measured P concentrations ranged from 4.6 -54.9 mg/L. 11 C samples (29%) and 3 V samples (10%) were excluded due to collection issues. Prediction results are shown in Figure 3. The 3 prediction techniques had similar performance characteristics, with each method displaying minimal bias (-0.4-2.0%) and reasonable imprecision (13.7-20.2%). The accuracy of prediction of P concentrations using VAMS was better for V than C samples.  Conclusion. Our pilot highlights the challenges of using VAMS for TDM. Sample collection issues were common. When VAMS is used, education on collection techniques is imperative. The predictive performance of VAMS was modest and V sampling had higher accuracy than C, although our sample size was small. Larger studies will be needed to further evaluate the predictive performance of the regression equations derived by our study. Background. Tebipenem pivoxil hydrobromide (TBP-PI-HBr) is an oral prodrug that is converted to tebipenem (TBP), the active moiety. TBP is a carbapenem with activity against multidrug-resistant Gram-negative pathogens, including extended-spectrum-β-lactamase-producing Enterobacterales and is being developed for treating complicated urinary tract infections (cUTI) and acute pyelonephritis (AP). Data from three Phase 1 studies and one Phase 3 study in patients with cUTI/AP were used to develop a population pharmacokinetic (PPK) model for TBP and identify covariates that described the variability in TBP pharmacokinetics (PK).
Methods. The PPK model was developed using TBP plasma and urine concentration-time data from the above-described Phase 1 and 3 studies. TBPPI-HBr doses, which ranged from 100 to 900 mg, were administered as single or multiple doses every 8 hours. After development of the structural model, stepwise forward and backward selection procedures were used to identify significant covariate relationships. The robustness of the final PPK model was assessed using a prediction-corrected visual predictive check (PC-VPC).
Results. The final dataset included 3448 plasma concentrations from 99 Phase 1 subjects and 647 Phase 3 patients and urine concentrations from 128 Phase 1 subjects. A two-compartment model with linear, first-order elimination and transit compartments to describe the rate of drug absorption after oral administration of TBP-PI-HBr best described TBP PK. The most clinically significant covariate effect, which would warrant dose adjustment, was the relationship between apparent oral clearance and creatinine clearance. In contrast, age, body size, sex, and fed status each had a minimal impact on TBP exposure. The PC-VPC showed good agreement between median simulated plasma concentrations based on the final PPK model and the median observed plasma concentrations for the pooled dataset (Figure 1).