PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of aacPermissionsJournals.ASM.orgJournalAAC ArticleJournal InfoAuthorsReviewers
 
Antimicrob Agents Chemother. 2017 April; 61(4): e02490-16.
Published online 2017 March 24. Prepublished online 2017 January 17. doi:  10.1128/AAC.02490-16
PMCID: PMC5365684

Peak Measurement for Vancomycin AUC Estimation in Obese Adults Improves Precision and Lowers Bias

ABSTRACT

Vancomycin area under the curve (AUC) estimates may be skewed in obese adults due to weight-dependent pharmacokinetic parameters. We demonstrate that peak and trough measurements reduce bias and improve the precision of vancomycin AUC estimates in obese adults (n = 75) and validate this in an independent cohort (n = 31). The precision and mean percent bias of Bayesian vancomycin AUC estimates are comparable between covariate-dependent (R2 = 0.774, 3.55%) and covariate-independent (R2 = 0.804, 3.28%) models when peaks and troughs are measured but not when measurements are restricted to troughs only (R2 = 0.557, 15.5%).

KEYWORDS: MRSA, obese, obesity, peak, pharmacodynamics, pharmacokinetics, population pharmacokinetics, trough, vancomycin

TEXT

Vancomycin is an essential methicillin-resistant Staphylococcus aureus (MRSA) active antimicrobial dosed on the basis of total body weight (TBW) and kidney function, with maintenance dose adjustments guided by therapeutic drug monitoring (TDM) (1). The recommended TDM approach includes measurement of a trough concentration with dosage adjustments to a target value based on the seriousness of the MRSA infection (1). This trough target range serves as a surrogate for the vancomycin area under the concentration-time curve (AUC) that, coupled with the MIC (AUC/MIC), best predicts clinical success (1). As expected, the translation of single trough concentration measurements into an integrated concentration-time profile for AUC estimation is achievable but requires parameter assumptions and mathematical modeling (2). These assumptions include a population estimate of a weight-based volume of distribution (V) and clearance (CL), which is based on creatinine clearance (CLcr) (3).

A major assumption, and the focus of the manuscript, is the widely utilized vancomycin V population value estimate of 0.5 to 1.0 liters/kg of body weight and the use of CLcr to estimate CL (2). Several studies have benchmarked the relationship between vancomycin V and TBW but have also suggested that the vancomycin V estimate per kilogram diminishes with increasing TBW (3,5). Bauer et al. demonstrated that the mean vancomycin V values were 0.32 liters/kg and 0.68 liters/kg in morbidly obese (165 kg) and normal-weight (68 kg) subjects, respectively (4). The mean absolute vancomycin V values were 52.8 liters (165-kg subject) and 46.2 liters (68-kg subject), which represent marginal differences relative to the average TBW of their two cohorts (4). Likewise, multiple approaches to estimate CLcr in obese patients exist, which can confuse dose selection (6, 7). As a consequence, the usefulness of TBW and CLcr as covariates to aid computation of vancomycin AUC values is not well characterized in the setting of single trough measurements.

We implemented a peak and trough measurement clinical protocol in obese patients and sought to compare the precision and bias of vancomycin AUC estimations using our obese-population-derived parameter estimates (covariate independent) compared to values from the standard-weight and CLcr-dependent model (covariate dependent) (5). Previous studies have shown that a 2-compartment model more optimally defines the vancomycin concentration-time profile; however, a 1-compartment model is often used in practice, as suggested by Web-based dosing calculators (2, 3). Consequently, we used the Maximum-Likelihood Expectation Maximization algorithm in ADAPT 5 software (BMSR, Los Angeles, CA) to compute the population estimate and variance of system parameters to serve as the Bayesian prior for individual estimation of AUC values in a sample of 75 obese adult subjects. The 2-compartment model (covariate-independent)-derived estimates served as the reference standard.

A (covariate-independent) 1-compartment system served as a potential model, while mean and coefficient of variation (CV) values of V = 0.7 liters/kg (50%) and CL = (0.79 · CLcr + 15.7) · 0.06 for CL in liters/h (50%) served as typical values for the standard covariate-dependent 1-compartment model (8). In the latter equation, the CLcr value was estimated using the Cockcroft-Gault equation and normalized to a body surface area of 1.73 m2 (8). The AUC24 values were computed for each individual using the full complement of data, and the computations were repeated for the 1-compartment models, including only the trough data. The precision was defined by linear regression, and bias was computed as percent difference in the AUC24 values from those of the reference model. For external validation, the analyses were repeated using an additional group of 31 subjects whose data were not used to generate the original reference 2-compartment model.

The median (range) age, TBW, and CLcr for the original data set were 58 (19 to 86) years, 108 (71.4 to 236) kg, and 93.3 (36.6 to 326) ml/min/1.73 m2, respectively. Details regarding the subject demographics and data source, including the number and timing of samples per subject, have been reported previously (5). The vancomycin concentration distribution in relation to the time since last dose shows a biphasic decline (Fig. 1A). The population model explained 37% of the interindividual variability, and the individual model predictions were effectively fitted (R2 = 0.95) by a 2-compartment system (Fig. 1B). The mean (standard deviation [SD]) V, CL, volume of the peripheral compartment (Vp), and intercompartmental clearance (Q) estimates for this reference 2-compartment model are 48.0 (14.6) liters, 4.69 (1.97) liters/h, 38.5 (19.3) liters, and 7.42 (2.62) liters/h, respectively. The mean (SD) V and CL for the 1-compartment (covariate-independent) model were 71.9 (17.5) liters and 5.18 (20.9) liters/h, respectively. Table 1 summarizes the Bayesian estimates of AUC24 by the various models with comparisons to the reference model. Use of peak and trough data yielded similar values for precision and bias of AUC24 irrespective of the model. However, when trough-only data were analyzed, lower precision and higher bias were observed with the use of the standard covariate-dependent model. The external validation data set (n = 31) had similar demographics, with median (range) age, TBW, and CLcr of 63 (24 to 88) years, 108 (70 to 152) kg, and 87.7 (46.3 to 240) ml/min/1.73 m2, respectively. Similar findings were noted with the standard covariate-dependent model and the use of trough-only data (Table 1). The individual model fits for one subject are shown in Fig. 2, which helps to illustrate the bias associated with trough-only analyses when a covariate-dependent model is used.

FIG 1
(A) Scatter plot of the vancomycin concentration profile over time since the last dose. (B) Predicted population model/predicted individual model over observed concentration scatter plot with linear fit.
TABLE 1
Comparisons of bias and precision of the vancomycin AUC24 estimates based on the reference model using data from two independent sourcesa
FIG 2
Observed individual (subject 16) concentration-time profile and model-predicted concentration-time profile of vancomycin based on compartment (comp) model, reliance on covariates (Cov), and use of peak (P) and/or trough (T) data with vancomycin area under ...

The current investigation reaffirms the importance of two-concentration measurements to aid vancomycin dosing in obese adults. As shown by these analyses, use of a simpler 1-compartment model can adequately explain the vancomycin concentration-time data when peak and trough concentrations are measured. In contrast, reliance on measurements based on the use of trough data only with covariate dependency lowers the precision and overestimates the vancomycin AUC24 values as well as the probability of target attainment (Table 1). These findings are consistent with the detailed Bayesian analyses performed by Neely and colleagues, who demonstrated that measurements using both peak and trough data were more precise and less biased than measurements using trough data only (9). Furthermore, our analyses extend their observations showing that a covariate-independent model was sufficient to estimate AUC in obese individuals and was most relevant when only a trough was measured. These findings are relevant because the prevalence of obesity remains high and strategies to optimize the dose of vancomycin in this population have been unclear (5). The challenge of dose optimization is compounded by the confusion in the selection of the optimal body size descriptor or equation to estimate kidney function in the obese population (7).

A criticism of this investigation could be that a sparse data set of sampled patients was used to derive the reference model. However, this limitation is mitigated by the similarity of our population parameter estimates to those by Neely and colleagues, individual Bayesian estimate comparisons, and the use of 1-compartment models that reflect the clinical scenario of limited individual sampling (9). In conclusion, the potential bias introduced by covariates can be avoided for individualized vancomycin maintenance dosing by measurement of both peak and trough concentrations. Improving the accuracy of vancomycin AUC estimation can serve to deliver more-precise dosing of this high-use agent.

REFERENCES

1. Rybak M, Lomaestro B, Rotschafer JC, Moellering R, Craig W, Billeter M, Dalovisio JR, Levine DP 2009. Therapeutic monitoring of vancomycin in adult patients: a consensus review of the American Society of Health-System Pharmacists, the Infectious Diseases Society of America, and the Society of Infectious Diseases Pharmacists. Am J Health Syst Pharm 66:82–98. doi:.10.2146/ajhp080434 [PubMed] [Cross Ref]
2. Pai MP, Neely M, Rodvold KA, Lodise TP 2014. Innovative approaches to optimizing the delivery of vancomycin in individual patients. Adv Drug Deliv Rev 77:50–57. doi:.10.1016/j.addr.2014.05.016 [PubMed] [Cross Ref]
3. Marsot A, Boulamery A, Bruguerolle B, Simon N 2012. Vancomycin: a review of population pharmacokinetic analyses. Clin Pharmacokinet 51:1–13. doi:.10.2165/11596390-000000000-00000 [PubMed] [Cross Ref]
4. Bauer LA, Black DJ, Lill JS 1998. Vancomycin dosing in morbidly obese patients. Eur J Clin Pharmacol 54:621–625. doi:.10.1007/s002280050524 [PubMed] [Cross Ref]
5. Hong J, Krop LC, Johns T, Pai MP 2015. Individualized vancomycin dosing in obese patients: a two-sample measurement approach improves target attainment. Pharmacotherapy 35:455–463. doi:.10.1002/phar.1588 [PubMed] [Cross Ref]
6. Pai MP. 2010. Estimating the glomerular filtration rate in obese adult patients for drug dosing. Adv Chronic Kidney Dis 17:e53–e62. doi:.10.1053/j.ackd.2010.05.010 [PubMed] [Cross Ref]
7. Wilhelm SM, Kale-Pradhan PB 2011. Estimating creatinine clearance: a meta-analysis. Pharmacotherapy 31:658–664. doi:.10.1592/phco.31.7.658 [PubMed] [Cross Ref]
8. Rodvold KA, Blum RA, Fischer JH, Zokufa HZ, Rotschafer JC, Crossley KB, Riff LJ 1988. Vancomycin pharmacokinetics in patients with various degrees of renal function. Antimicrob Agents Chemother 32:848–852. doi:.10.1128/AAC.32.6.848 [PMC free article] [PubMed] [Cross Ref]
9. Neely MN, Youn G, Jones B, Jelliffe RW, Drusano GL, Rodvold KA, Lodise TP 2014. Are vancomycin trough concentrations adequate for optimal dosing? Antimicrob Agents Chemother 58:309–316. doi:.10.1128/AAC.01653-13 [PMC free article] [PubMed] [Cross Ref]

Articles from Antimicrobial Agents and Chemotherapy are provided here courtesy of American Society for Microbiology (ASM)