Questioning the accuracy of trough concentrations as surrogates for area under the curve in determining vancomycin safety

The glycopeptide antibiotic, vancomycin, has been in use for the treatment of serious Gram-positive infections such as methicillin-resistant Staphylococcus aureus (MRSA) infection for nearly 60 years, and therapeutic monitoring for drug safety is acknowledged as an important part of the management strategy when treating patients with this agent [Moellering, 2006; Levine, 2006]. Vancomycin is still the treatment of choice for treatment of MRSA infection, however some authors comment that its days of utility are in the twilight due to rising minimum inhibitory concentrations (MICs) of MRSA isolates, which necessitates such high dosing and drug exposure that the probability of toxicity becomes too great to use it [Van Hal et al. 2013; Patel et al. 2011].

Historically, both peak and trough concentrations have been used at times to monitor the safety of vancomycin [Rybak, 2006]. Peak concentrations, however, have been discarded as a monitoring method for vancomycin as they do not correlate with therapeutic efficacy or key safety concerns of nephrotoxicity [Saunders, 1994; Suzuki et al. 2012]. Trough concentrations have become the dominant method for monitoring vancomycin toxicity and efficacy in clinical practice today; however, the pharmacokinetic–pharmacodynamic (PKPD) index of the area under the concentration curve (AUC) to MIC ratio of the organism being treated is recommended for monitoring treatment efficacy in serious infection [Kullar et al. 2011; Li et al. 2012]. An AUC/MIC ratio of over 400 has been associated with greater clinical success and more rapid bacterial eradication compared with lower AUC/MIC ratios [Moise-Broder et al. 2004]. Trough concentrations up to 15 mg/liter have been reported to correlate with an AUC/MIC ratio of 400 or greater when the organism MIC is up to 1 mg/liter [Giuliano et al. 2010].

The US guidelines on the therapeutic monitoring of vancomycin, released some 5 years ago, were an important step forward in garnishing consensus in guiding practice for the efficacious and safe use of this antibiotic. These guidelines promote aggressive dosing of vancomycin to attain a trough target in the range of up to 15–20 mg/liter, concurring that concentrations greater than 15 mg/liter will achieve an AUC/MIC ratio greater than 400 when the organism MIC is up to 1 mg/liter [Rybak et al. 2009].

The advent of more aggressive vancomycin dosing has led to increased concerns about the safety of dosing regimens which aim to achieve trough concentrations greater than 15 mg/liter. A systematic review of higher dosing to achieve troughs of 15–20 mg/liter reported greater nephrotoxicity [odds ratio 2.67; 95% confidence interval (CI) 1.95–3.65] compared with trough concentrations less than 15 mg/liter [Van Hal et al. 2013]. Curiously, however, a systematic review and meta-analysis of vancomycin continuous infusions achieving targets of 20–30 mg/liter resulted in significantly less nephrotoxicity compared with dosing by intermittent infusion [Cataldo et al. 2012]. This anomaly poses the question as to whether trough concentrations really are a good surrogate marker for AUC and vancomycin exposure.

A recently published paper by Neely and colleagues set out to determine this very question: Are trough concentrations an appropriate surrogate for AUC? Briefly, they performed PK/PD modelling on data from 47 adults with 569 vancomycin concentrations. AUCs were determined using Pmeterics (version 1.1.1, Laboratory of Applied Pharmacokinetics, University of Southern California; www.lapk.org), which employs a two-compartmental nonparametric model for vancomycin modelling and simulation. They found the AUC was underestimated by a mean of 23% (CI 11–33%, p = 0.0001), calculated from a model using trough concentration data alone compared with a full model incorporating peak and trough concentrations to determine AUC. When modelling was performed using the full model as a Bayesian prior and only incorporating trough concentrations, a 97% accuracy of AUC estimation was achieved (CI 93–102%, p = 0.23). Using Bayesian modelling they simulated over 5000 concentration–time profiles and found that in adult patients with normal renal function and an AUC of at least 400 mg⋅h/liter with an organism MIC of 1 mg/liter, some 60% are expected to have trough concentration below 15 mg/liter. Thus, dosing to achieve a trough over 15 mg/liter may lead to excessive vancomycin exposure and unnecessary risk of toxicity. Further, the authors propose that a 24 h AUC of 700 mg⋅h/liter represents a conservative upper limit of vancomycin exposure that is safe with minimal risk of nephrotoxity [Neely et al. 2014].

This assertion of a 24 h AUC 700 mg⋅h/liter as an upper limit for safe vancomycin exposure is supported by the US consensus guidelines [Rybak et al. 2009], and others [Patel et al. 2011; Avent et al. 2013], in that if the organism being treated has a higher MIC (i.e. ≥2 mg), maintaining an AUC/MIC ratio of at least 400 in patients with normal renal function will require such high drug exposure that treatment will not be viable due to the increased risk of nephrotoxicity.

The study by Neely and colleagues suggests vancomycin trough concentrations alone are not a good surrogate for AUC. If computerized Bayesian modelling of vancomycin is performed to determine AUC, we may be able to achieve the desired therapeutic AUC/MIC targets with less aggressive dosing and reduced risk of toxicity. Arguments against computerized estimation of AUC are that it requires additional software, trained staff, new processes and logistics; however, these barriers have not been insurmountable as some hospitals have adopted computerized AUC monitoring of other agents such as aminoglycosides into routine clinical practice [Baysari et al. 2012]. Computerized software for AUC determination of vancomycin is freely available.

A considerable amount of time and money is wasted in routine care with vancomycin treatment when inappropriately timed blood samples are collected. Morrison and colleagues analyzed nearly 2600 vancomycin serum samples over a 13-month period (Boston, MA, USA) and found that 41% of samples were drawn too early, which resulted in clinicians reducing, withholding or ceasing patients’ vancomycin doses in addition to reordering concentrations (29.2% versus 20.0%, p < 0.001) compared with concentrations taken at the appropriate time, that is, 1 h predose and at steady state [Morrison et al. 2012]. Such wasted resources could be redirected to education with regard to appropriate timing of sample collection and implementation of vancomycin monitoring using AUC computation.

Monitoring of vancomycin concentrations has been demonstrated to reduce nephrotoxicity (odds ratio 0.25, 95% CI 0.13–0.48, p < 0.0001), defined as a rise in serum creatinine greater than 44 µmol/liter (5 mg/liter) or 50% increase in serum creatinine from baseline during vancomycin therapy compared with patients who do not receive vancomycin monitoring [Ye et al. 2013]. However, strategies to improve vancomycin dosing and attainment of target concentrations, such as implementation of clinical practice guidelines, are important [Phillips et al. 2013], as are electronic decision support tools [McCluggage et al. 2010]. Determining vancomycin exposure by computation of AUC may help us to truly individualize therapy and better manage toxicity risk, rather than relying on trough concentrations alone. Such a strategy may help to prolong the utility of vancomycin in the era of increasing antibiotic resistance.