therapeutic outcome are difficult to assess in the absence of actual clinical data.
Furthermore, plasma drug concentrations, the primary component for building the
PK portion of the PK/PD index, do not necessarily reflect in a predictable manner a
compound’s ability to diffuse first into the site of infection and then into the
bacterial cell.
It is clear that the PK/PD variable providing the most appropriate surrogate for
drug effectiveness is dependent upon several factors. These include the drug’s
mechanism of action, whether its effects are time or concentration-dependent,
and the duration of its PAE. Examples of targets proposed for a range of compounds
and pathogens have been reviewed by Gunderson et al. ( 2001 ).
For drugs exhibiting concentration-dependent killing,Cmax/MIC ratios may be
particularly important when the pathogen has a high MIC value or is proliferating
rapidly. Such conditions lead to a greater risk of mutational events that could lead to
a resistant subpopulation (Craig and Dalhoff 1998 ). In infectious diseases, a high
bacterial burden (inoculum effect) can also increase the risk of a mutational event
due simply to the laws of probability (Drusano et al. 1993 ; Craig and Dalhoff 1998 ).
Therefore, in the presence of a high bio-burden, for drugs exhibiting concentration-
dependent killing, the goal is to achieve high drug concentrations and therefore
rapid killing (Drusano et al. 1993 ; Preston et al. 1998 ). The therapeutic objective is
to reduce bacterial numbers to a level at which the host can destroy those bacteria
that are not killed directly by the antimicrobial agent (and to avoid the double step
mutation).
Monte Carlo simulations (a method for generating a pseudo-random set of
parameter values that conform to some a priori probability distribution, such as
normal, log-normal, uniform etc.) allows for the prediction of antimicrobial effec-
tiveness in patients whose physiological attributes are likely to be encountered
under clinical conditions (Zelenitsky et al. 2005 ; Drusano 2007 ). Once a numerical
value for a PK/PD target has been defined (e.g. AUC/MIC¼100 h), the goal is to
estimate the dose needed to achieve that target in the patient population. The
proportion of the population for which the target is achieved (e.g. the proportion
of subjects in the intended patient population that are expected to achieve an AUC/
MIC value of 100 h) is termed the target attainment rate (TAR). The TAR is usually
set at 90% (i.e. that 90% of the patient population will achieve the targeted PK/PD
target). An important component of strategic dosing paradigms is an appreciation of
the population variability that exists, not only in terms of bacteria but also in terms
of the host. Within veterinary medicine, the latter consideration is often ignored
and PK characterisation is usually based upon data generated in normal healthy
animals. The use of such data results in estimates that normally and possibly
markedly under-estimate the variability likely to occur in the true patient popula-
tion (Martinez and Modric 2010 ). As a result, dosage regimens may fail to achieve
the intended TAR.
To illustrate this point, nine AUC datasets (n¼1000 iterations per set) were
generated (using a Ln-normal distribution), each described by the same mean value
(125mgh/mL), but with the percentage coefficient of variation (%CV) varying
from 15% to 100%. This situation is comparable to the difference between studies
242 M. Martinez and P. Silley