A commentary on: Personalised dosing of vancomycin: a
prospective and retrospective comparative quasi-experimental study by
Luqman Vali et al.
Vancomycin is frequently used to treat Staphylococcal infections
resistant to beta-lactam antibiotics. Overexposure to vancomycin
increases risk of nephrotoxicity while underexposure can lead to
therapeutic failure. This narrow therapeutic window makes it necessary
to measure vancomycin serum concentrations aiming for an effective and
justifiable dose. Often mathematical software is used to calculate a
pharmacokinetic (PK) / pharmacodynamic (PD) target and predict the right
vancomycin dosage.
In this journal, Vali et al. compared the prospective results of
a bedside Bayesian-guided software package (DoseMeRx) to guide
vancomycin dosing with retrospective results of vancomycin-dosing with a
standard algorithmic approach. Bedside Bayesian-guided dosing resulted
in significantly more area under the curves (AUC) measured within the
target range (350 – 450 mg/L*h) and in a significantly higher
percentage of time in the acceptable range in comparison with the
standard algorithmic approach.
Unfortunately, the authors were not able to relate their findings to
clinical outcomes such as clinical cure or mortality, possibly due to
insufficient power. Still, PK/PD modelling is the cornerstone in
calculating vancomycin dosages and user friendly bedside software can
increase the availability and integration of these models in clinical
practice. However, the precision of the dose calculations based on
pharmacokinetic and pharmacodynamic modelling can be misleading and
should not be used to determine the optimal dose without further
considerations.
Several factors impact the interpretation of individual dose
calculations of antibiotics based on PK/PD modelling: the PK/PD-index of
the antibiotic that is used to predict the efficacy; clinical and
pharmacological factors that influence PK and the accuracy of minimal
inhibitory concentration (MIC) of the causative pathogen. We will
discuss these three factors and elaborate on their influence on dose
calculations.
Several studies aimed to establish the dose-response relation of
vancomycin. Generally accepted vancomycin PK/PD targets are an AUC/MIC
> 400 or a trough serum or plasma level
(Ctrough)/MIC > 10-15. Using the AUC/MIC as
predicting parameter requires the use of specialized software to
calculate the AUC on intermittent dosing regimens, but is generally
accepted as the most appropriate PK/PD-index of vancomycin. A recent
meta-analysis of several cohort studies showed that AUC/MIC guided
dosing resulted in a lower incidence acute kidney injury (AKI) in
comparison to trough level dosing with an odds ratio of 0.68 (95% CI:
0.46 – 0.99). However, included studies were small, mostly
retrospective and the effect of PK/PD-based dosing on AKI and other
clinical outcomes therefore need to be confirmed in larger studies.
Experiments with novel in-vitro models, such as two-compartment
hollow fiber models, could add valuable information to the currently
available evidence on the optimal PK/PD index of vancomycin. However, as
far as we are aware, studies with these novel in-vitro models
with vancomycin have not yet been reported.
Tissue penetration is another major determinant in applying PK/PD
calculations in clinical practice. The extent of tissue penetration
depends on the solubility of the compound in fat (or octanol - thelog P ), the acid dissociation constant (pKa), the extent of
protein binding, the molecular mass and the affinity to certain
transmembrane transporters, such as P-glycoprotein. It could be
questionable if serum or plasma concentrations are a representative
factor in the treatment outcome when the tissue penetration of an
antibiotic is low, while the infection is mainly situated in tissue. A
good example is the high molecular mass of vancomycin and therefore low
penetration characteristics.
Microdialysis, has delivered promising results to quantify the tissue
penetration of antibiotics. With microdialysis, a small dialysis tube is
placed in the tissue of interest. Due to diffusion of the antibiotic
from the extracellular fluid into the dialysis fluid, the antibiotic
concentration can be measured in the dialysate. Microdialysis can help
to extrapolate the PK/PD index to infections in tissue either by
computing serum-tissue ratios or by building multiple-compartment PK
models. However, a complicating aspect is the large interindividual
variation in tissue concentrations. Studies on microdialysis with
vancomycin have shown large variation in interstitial fluid
concentrations which could not be explained by patient characteristics.
In addition, the penetration of vancomycin in the central nervous system
(CNS) is also highly variable. For CNS infections, one should consider
to administer vancomycin intrathecally or switch to a different
antibiotic with proven acceptable penetration in the CNS, such as
linezolid. Overall, vancomycin tissue penetration is relatively low but
some high unexplained variability is observed. Overall, microdialysis
and hollow fiber models are potentially useful techniques in gaining
more insight in the relation between serum concentrations, tissue
concentrations and efficacy. Yet, clinical evidence for their use is
still limited.
Beside the optimal PK/PD target and extent of tissue penetration, the
way of dosing is also of importance. Continuous vancomycin
administration is increasing in clinical practice, making sampling easy
as the concentration measured does not depend on the time of sampling.
Furthermore calculation of the AUC0-24h (mg/L*h) of
vancomycin on continuous infusion is simple: multiply the measured
concentration (in mg/L) by 24. In a meta-analysis the incidence of
nephrotoxicity was found lower in patients treated with continuous
vancomycin vs. patients treated with intermittent vancomycin, with no
difference in treatment failure or mortality. A limiting practical
factor in applying continuous vancomycin is the availability of one
dedicated intravenous (IV) catheter or lumen, since vancomycin is
incompatible with many other medicines and IV fluids. In the paper by
Vali et al. we missed vancomycin continuous infusion simulations
to collect evidence for this upcoming method of administration.
A final major determinant of PK/PD modelling is the MIC. The MIC is
defined as the lowest concentration that inhibits growth of the isolated
micro-organism. The golden standard for MIC testing is broth
microdilution. However, many clinical microbiological laboratories use
automated systems or antimicrobial gradient strips for their first-line
antimicrobial susceptibility testing (AST). These methods are rarely
precise enough to be used to calculate the AUC/MIC or
Ctrough/MIC. Therefore, Mouton et al. suggested
to include the epidemiological cut-off value (ECOFF) when interpreting
MIC results for target attainment calculations. The ECOFF is determined
as the highest MIC of the bacterial species of interest without acquired
resistance, i.e. the wildtype population. Using the ECOFF instead of the
MIC, however could add potential risks of overdosing of an antibiotic
for the wildtype population of a micro-organism with broad MIC ranges.
For example, the suggested ECOFF of S. aureus for vancomycin is 2
mg/L, which would indicate that the target AUC0-24hshould be above 800 mg/L*h (AUC/MIC > 400). Such high AUCs
increase the risk of nephrotoxicity and should not be applied in regular
care. PK/PD-based modelling with accurate MIC testing therefore seems of
importance to decrease risk of overdosing.
In conclusion, the use of software programs can facilitate the
measurement of these targets and the prediction of the right dose of
vancomycin. The prescriber and pharmacist should, however, always be
aware of the ’pseudo- precision’ of those calculations: many other
factors influence the efficacy and toxicity of vancomycin besides the
trough concentration or the AUC.