The pharmacokinetic–pharmacodynamic approach to a rational dosage regimen for antibiotics

Research in Veterinary Science 2002, 73, 105–114
doi:10.1016/S0034-5288(02)00039-5, available online at on

REVIEW

The pharmacokinetic–pharmacodynamic approach to a
rational dosage regimen for antibiotics
 LOU
P. L. TOUTAIN*, J. R. E. DEL CASTILLO, A. BOUSQUET-ME
UMR INRA de Physiopathologie et Toxicologie Exp
erimentales, Ecole Nationale V
et
erinaire de Toulouse,
23 Chemin des Capelles, 31076 Toulouse cedex 03, France
SUMMARY
Pharmacokinetic–pharmacodynamic (PK/PD) surrogate indices (AUIC, AUC/MIC, Cmax /MIC, T > MIC) for measuring
antibiotic efficacy are presented and reviewed. As clinical trials are not sufficiently sensitive to establish a dosage regimen
which guarantees total bacteriological cure (Pollyanna phenomenon), PK/PD indexes have been proposed from in vitro,
ex vivo, and in vivo infection models and subsequently validated in retrospective or prospective human clinical trials. The
target value for time-dependent antibiotics (b-lactams, macrolides) is a time above the MIC (T > MIC) of 50–80% of
the dosage interval, while for concentration-dependent antibiotics (quinolones and aminoglycosides), the area under the
inhibitory curve (AUIC, or more simply AUC/MIC of about 125 h) is the best surrogate indicator of activity. Using the
latter drugs, high concentrations achieved early during therapy are desirable to prevent the development of resistance. A
Cmax /MIC ratio greater than 10–12 seems to be an appropriate target for aminoglycosides. Ó 2002 Elsevier Science Ltd.
All rights reserved.

OVERUSE and misuse of antimicrobial drugs have
favoured the growth of resistant organisms and resistance can spread to other microbial populations, jeopardizing humans and animals, including those not
previously exposed to antimicrobial agents.
Among the documented misuses contributing to
drug resistance are inappropriate dosage regimens

(dose, dosage interval, duration of treatment, route and
conditions of administration) (Anonymous 1998). Rational antibiotic therapy requires dosage regimens to
be optimized, not only to guarantee clinical efficacy,
but also to minimize the selection and spread of resistant pathogens.
Pharmacokinetics (PK) is the tool used to describe
and predict drug concentration profiles in biological
fluids (usually plasma) and combining PK and pharmacodynamic (PD) information (i.e., bacterial susceptibility to antibiotics) constitutes the PK/PD modelling
approach to antibiotic efficacy. The goal of this approach is to describe, predict, and if possible understand the time course of the antibiotic effect as a
function of the drug dosage regimen. In addition, the
PK/PD approach addresses the two main sources (PK
and PD) of inter- and intra-individual variability in
therapy outcome and allows dual adaptation of the
antibiotic regimen (Schentag et al 1985).
The aim of this review is to indicate the limitations
of clinical endpoints for selection of the dosage regimen of antibiotics and to demonstrate the advantages

Corresponding author. Fax: +33-561-19-39-17;
E-mail:
0034-5288/02/$ - see front matter

of the PK/PD alternative approach. The potential
contribution of PK/PD to the minimization of drug
resistance development and determining breakpoints in
veterinary susceptibility tests will also be addressed.
This review focuses mainly on antibiotic therapy for
pathogens located in extracellular fluids where drug
exchange with the plasma is not impeded by a diffusion
barrier.

HOW TO DETERMINE A DOSAGE REGIMEN

FOR AN ANTIBIOTIC
As opposed to drugs acting on some physiological
system of the host, the action of antibacterial drugs can
only be investigated in spontaneously or experimentally infected and not in healthy animals. Ideally, dose
ranging and clinical trials in target species should explore responses to a variety of dosage regimens, including suboptimal schedules, but this is impractical for
economic and not possible for ethical reasons. In addition, the reliability of the results of clinical trials involving antibiotics is impaired by a large number of
host and bacterial factors that cannot be controlled in a
clinical setting.
Another problem with veterinary clinical trials is
that many suffer from one or several flaws which limit
their usefulness to evaluate the dosage regimen tested.
Van Donkersgoed (1992) performed a meta-analysis of
field trials of prophylactic mass medication for bovine
respiratory disease in feedlots. Among 107 trials considered, all but 10 were excluded on account of major
defects in experimental design or data analysis. A
Ó 2002 Elsevier Science Ltd. All rights reserved.


106

P. L. Toutain, J. R. E. del Castillo, A. Bousquet-Melou

similar finding was reported in pigs (Bording 1990) and
concerns over the informative value of published human clinical trials has led 13 leading medical journals to
express their concern about sponsorship, authorship,
and accountability of clinical trials and to revise their
editorial policy (Anonymous 2001).

CLINICAL OUTCOME AND THE POLLYANNA
PHENOMENON

In many infections, the ultimate goal of antibiotic
therapy is not simply to guarantee a clinical success but
to achieve it through a total bacteriological cure. If
bacterial eradication does not occur, less susceptible
bacteria are likely to head the recolonization process
after discontinuation of therapy and a more resistant
population will become predominant (Dagan et al
2001). Therefore, it is essential to establish whether
clinical success is a fully valid endpoint to compare the
efficacy of two antibiotics or to assess the value of a
dosage regimen for a bacteriological cure. Investigation
of this question has revealed the so-called ‘‘Pollyanna
phenomenon’’.
The Pollyanna phenomenon refers to the fact that if
antibiotic efficacy is measured by symptomatic responses, drugs or dosing strategies with excellent antibacterial activity will not be as efficacious as anticipated,
while the opposite will occur for antibiotics with poor
antibacterial activity. In otitis media in children, for
instance, it was calculated that the clinical success rate
will be high (89%) but not total when bacterial eradication is 100%, whereas a high clinical success (71%)
may still be expected for a bacteriological cure of 27%,
i.e., a probability of bacterial eradication which could be
achieved with no antibiotic at all (placebo effect)
(Marchant et al 1992). The lack of sensitivity of clinical
trials to discriminate a ‘‘good’’ from a ‘‘bad’’ antibiotic is
well illustrated by a study of bacteriological failures in
this disease. It was shown that two very different antibiotics in terms of their bacteriological failure rate,
Cefaclor (32%) and Cefuroxime axetil (15%), may be
expected to give rather similar clinical success (4% and
9%, respectively). Using the clinical outcome, 900 patients would be necessary to statistically discriminate
between these two antibiotics, while 200 would be sufficient when using the bacteriological cure itself as the

endpoint (Dagan et al 2001).
The Pollyanna phenomenon can be encountered in
veterinary therapeutics. Yancey et al (1990) tested the
efficacy of ceftiofur hydrochloride to treat experimental colibacillosis in neonatal swine in a large trial involving several hundred piglets. This study highlights
the possible discrepancy between dose-effect relationships based on bacteriological criteria (shedding of
bacteria), a clinical endpoint (abnormal stool) or
mortality as endpoints (Fig 1). Similarly, using the
Escherichia coli model described by Charleston et al
(1998) to test a quinolone in chickens, we computed
very different ED50 values for mortality (8 mg/kg) and
bacteriological cure (13 mg/kg) (Toutain, unpublished
results).

FIG 1: Efficacy of ceftiofur hydrochloride for the treatment of experimental
colibacillosis in neonatal swine (Yancey et al (1990): response (mortality vs.
bacterial shedding) of E. coli-infected pigs treated orally with ceftiofur HCl (0–
64 mg/kg). The lowest dose tested (0.5 mg/kg) reduced mortality, but excluding the placebo group there was no significant difference in mortality between
doses (from 0.5 to 64 mg/kg). In contrast, bacterial shedding displayed a
significant dose–effect relationship. This illustrates the inability of a clinical
outcome (mortality) to discriminate doses having possible different efficacies
in terms of bacterial eradication.

This lack of sensitivity of clinical outcomes to find
the best dosage regimen in terms of bacteriological
cure, a prerequisite to minimize the risk of the emergence of resistance, opens the way to investigating the
efficacy of an antibiotic using PK/PD approaches and
surrogate indexes in healthy animals (Table 1).

EVIDENCE FOR PK/PD RELATIONSHIPS: IN
VITRO, EX VIVO AND IN VIVO MODELS IN

EXPERIMENTAL SPECIES
The relationships between antibiotic exposure, rate
of bacterial killing, and possible regrowth of bacteria
with increased MIC can be examined with in vitro
systems mimicking the expected in vivo antibiotic
concentration profiles in the target species (Murakawa
et al 1980). These models have been used to determine
the pharmacokinetic parameters (AUC, Cmax or times
> MIC . . .) which correlate best with antibacterial activity and are predictive of the emergence of resistance.
In veterinary medicine, the response of Staphylococcus
aureus to the simulated interstitial fluid pharmacokinetic profile of penicillin in sheep was obtained with an
in vitro PD model (Koritz et al 1994). The limitation of
in vitro models is that they simulate infection but do
not take into account the role of the immune system of
the host.
Using a tissue cage model, inflammatory (exudate)
and non-inflammatory (transudate) fluids can be collected and ex vivo antibacterial activity can be measured over a 24 h incubation period allowing the
computation of the AUIC24 h corresponding to bacteriostasis (no change in bacterial count), bactericidal
activity (99.9% reduction of bacterial count) or total
bacterial eradication (Lees and Aliabadi 2002). Table 2
gives ex vivo AUC24 h /MIC values for danofloxacin in
serum for ruminant species. As for in vitro investigation, these ex vivo AUICs24 h do not take into account


The pharmacokinetic–pharmacodynamic approach to antibiotics

TABLE 1:

107

PK/PD vs. dose titration or clinical trials

Features

Subjects
Endpoints
Validity (clinical relevance)
Sensitivity to dose ranging
Reliability
Application to drug discovery and
development
Extrapolation (from in vitro models
or other species)
Dual dosage individualization
Prediction of the emergence of resistance
Breakpoint setting
Population studies: PK or PD origin of
variability
Regulatory acceptance
Cost
Independent evaluation/objectivity

Approach
PK/PD

Dose titration or clinical trials

Healthy
Surrogates: T > MIC, Cmax /MIC, and
AUIC

Need to be validated (prospectively or
retrospectively)
Yes
High
Early screening

Infection models, patients
Clinical outcome (cure, failure) bacteriological outcome (eradication,
resistance)
Gold standard but many possible drawbacks and Pollyanna
phenomenon
No (difficult to perform dose ranging in ill patients)
Low
Later, confirmatory

Easy

Difficult

Yes
Possible
To be explored (promising)
Yes

No
Possible
Yes
No, if only clinical outcomes are measured

In progress

Low
Independent investigations possible

Pivotal, designed to satisfy authorities but not to optimize treatments
High
Requires commercial funding

TABLE 2: Critical ex vivo AUC2 4/MIC (h) values for Danofloxacin in serum to obtain a bacteriostasis, a bactericidal or a bacterial elimination in
different ruminant species (Aliabadi and Lees 2001; Lees and Aliabadi 2002)
AUC24 h/MIC

Bacteriostatic
Bactericidal
Elimination

Species
Calf

Sheep

Goat

Camel

15:9 Æ 2:0
18:1 Æ 1:9
33:5 Æ 3:5

17:8 Æ 1:7
20:2 Æ 1:7

28:7 Æ 1:8

22:6 Æ 1:7
29:6 Æ 2:5
52:4 Æ 8:1

17:2 Æ 3:6
21:2 Æ 3:7
68:7 Æ 15:6

Danofloxacin was administered intramuscularly to each species, at a dose rate of 1.25 mg/kg. Ex vivo antibacterial activity was evaluated by bacterial count after
24 h of incubation. The tested pathogens were Mannheimia haemolytica or E. coli; the relationship between ex vivo AUIC24 h in serum and the log10 difference in
bacterial count (CFU) was modelised by a Hill model; the AUIC24 h for bacteriostasis and bactericidal activity were defined as values that resulted in no change in
bacterial count and the value that resulted in 99.9% reduction in bacterial count respectively. The AUIC24 h for bacterial elimination was defined as the lowest
value that resulted in the maximum antibacterial effect (actually the limit of detection i.e., 10 CFU/mL).
Values are mean Æ SE of the mean (n = 6).

host defence mechanisms but this approach was validated for danofloxacin in calves by comparing ex vivo
data with findings obtained in vivo in a model of calf
pneumonia (Lees and Aliabadi 2002).
Murine thigh and lung infection models (e.g., with
Klebsiella pneumoniae) have been used to describe the
antibiotic dose–response curve and to generate and
discriminate quantitative parameters of the in vivo
antibiotic effect (Vogelman et al 1988; Leggett et al
1991). Neutropenic mice are inoculated and treated
with one of a wide range of dosage regimens (varying
dose and dosage interval) minimizing the interdependencies between the PD parameters tested (T > MIC,
AUC/MIC, and Cmax /MIC). The mice are then serially
killed and the numbers of bacterial cells remaining in

the infected tissue (and other outcomes) are correlated
with the different plasma drug profiles.
In these laboratory animal models, different PK/PD
indices have been proposed for time- and concentration-dependent antibiotics. The b-lactams exhibited a
short-lived post-antibiotic effect (PAE) and minimal
concentration-dependent killing, with optimal bactericidal action at a threshold of approximately 4 Â MIC,
suggesting that the duration of plasma concentration
exceeding the MIC (T > MIC) is the major PK/PD
parameter determining their in vivo efficacy (Craig

1998).Thus, the daily dose of ceftazidime required to
prevent death in 50% of the animals was 1.5 mg/kg for
continuous infusion but 24.4 mg/kg for 6 h intermittent
injections (Roosendaal et al 1985).
Moreover, T > MIC is the parameter which correlates best with efficacy for clindamycin and macrolides.
In the case of the macrolide spiramycin, PK/PD modelling of staphylococcal infections of the mammary
gland in cows was used to predict efficacy (Renard et al
1996) and an optimal dosage regimen for the treatment
of mastitis, showing the feasibility of this approach as
proposed by Koritz and Bevill (1991).
The dosage interval appears to be less important for
the efficacy of aminoglycosides and quinolones. In
laboratory animal models, these antibiotics displayed
major concentration-dependent killing, such that dosage regimens should aim at the highest (safe) plasma
concentration. The Cmax /MIC and/or AUC/MIC ratios
are the main PK/PD parameters correlating with efficacy, although Cmax /MIC ratios may be more relevant
in humans for infections where the risk of resistance
development is significant (Craig 1998; Drusano et al
1993). The efficiency of spaced administration of large
doses of quinolones or aminoglycosides is related to

their prolonged and concentration-dependent PAE,
which prevents bacterial regrowth when serum levels


108

P. L. Toutain, J. R. E. del Castillo, A. Bousquet-Melou

fall below the MIC. In veterinary medicine, the PK/PD
relationships for danofloxacin and marbofloxacin have
been investigated in ruminants using a tissue cage
model allowing computation of the AUIC values producing bacteriostasis and bactericidal action (Aliabadi
and Lees 1997).
Although tetracyclines do not exhibit concentrationdependent killing, the AUC/MIC ratio is the major PK/
PD parameter correlating with the therapeutic efficacy
of these drugs (Craig 1998).

PK/PD PREDICTIVE INDICES OF IN VIVO
EFFICACY ARE BUILT ON (FREE) PLASMA
ANTIBIOTIC CONCENTRATIONS, NOT
TOTAL TISSUE ANTIBIOTIC LEVELS
Most pathogens of clinical interest are located extracellularly and the biophase for antibiotics is the extracellular fluid (Schentag 1990). Except for plasma,
extracellular fluids are difficult to sample, but if there is
no barrier to impede drug diffusion, the concentration
of free (unbound) antibiotic in plasma approximates its
free concentration in the extracellular space. The extravascular fluid penetration of free drug is complete,

regardless of the extent of plasma protein binding
(Schentag et al 1985), which makes it the best surrogate
to free antibiotic in the biophase. Therefore, the free

drug concentration in plasma is the best drug-related
predictor of clinical success, even for tissue infection
(Schentag 1989; Cars 1997). In contrast, where there is
a barrier to drug diffusion (central nervous system, eye,
prostate. . .), the plasma concentration may be less
useful to predict concentrations at the infection site.
Similarly, a discrepancy may exist between antibiotic
concentrations in plasma and in a biophase if the normal rapid equilibration between plasma and infected
site is impaired by a reduced blood supply (abscess,
inflammatory debris, shock syndrome, sequestered
bone fragments, tissue cage. . .) (Fig 2).
The plasma binding of some classes of antibiotics
(aminoglycosides, several fluoroquinolones) is low and
the measured plasma concentration may be considered
to be similar to the free concentration in the biophase.
Conversely, if drug binding is important (e.g., free
fraction less than 20% of the total plasma concentration), a correction for binding is warranted. It is noteworthy that the free concentration is controlled only by
the intrinsic drug clearance. Hence, even when the free
fraction increases (e.g., displacement, low protein lev-

FIG 2: Antibiotic access to the bacterial biophase. Most bacteria (B) are located in extracellular fluid (plasma and interstitial fluids). Unbound (F) drug circulating
in plasma is the only fraction which can gain access to the interstitial fluid through porous capillaries to combat infection due to an extracellular pathogen and a
drug will develop antibacterial action if the free drug concentration exceeds the MIC. Some tissues have permeability limitations at the capillary level and/or
possess an efflux pump. This impedes accumulation of drugs in the tissue (e.g., blood–brain barrier) and only lipophilic drugs can cross such barriers (e.g.,
quinolones). The blood perfusion rate can also be a limiting factor (clot, abscess). Some bacteria are located within cells (facultative or obligatory intracellular
pathogens). Inside a cell (e.g., polymorphonuclear neutrophils), different locations are possible (cytosol, phagosome, and phagolysosome), where the antibiotic
concentrations can be very different. Macrolides for example are trapped in phagolysosomes which have a low pH (about 4–5) and this gives a ‘‘high total cell’’
concentration. However, as the antibacterial potency of macrolides is pH dependent (low or no activity at acidic pH), a high local concentration is not synonymous
with high activity.

The pharmacokinetic–pharmacodynamic approach to antibiotics

109

FIG 3: Differential influence of protein binding on the free concentration (Cfree ) and free fraction (fu) for a drug having a low extraction ratio. The impact of
protein binding on antibacterial activity causes confusion because the relationships between Cfree and the total antibiotic concentration (Ctot ) are fundamentally
different in an in vitro (closed) and an in vivo (open) system. Cfree is the only active fraction under in vivo or in vitro conditions. In vitro, when an MIC is measured in
broth (no binding), the antibiotic is entirely in its active form. If the MIC is measured in a matrix able to bind antibiotics (e.g., serum), only free fraction is active and
effects (e.g., inhibition diameter) decrease as the binding increases. In vitro, a decrease in fu is synonymous with a decrease in Cfree . In vivo, the situation is
different and when discussing relationships between fu, Cfree and Ctot , the structural equation fu ¼ Cfree =Ctot (right, in vivo) should not be rearranged into
Cfree ¼ fu à Ctot (left, in vitro). In vivo, an increase in fu is not synonymous to an increase in Cfree but rather to a decrease in Ctot . Cfree is actually the independent
variable which controls Ctot through Bmax and Kd, the maximum binding capacity and equilibrium association constant. Ctot is only a dependent variable and
competitive interaction decreases Ctot without increasing Cfree . This is of relevance for drug monitoring as Ctot values are measured by analytical techniques. In
the case of competitive displacement (arrow), there is only a transient increase in Cfree , the small amount of antibiotic displaced from the transport protein being
rapidly (within a few seconds) redistributed and eliminated whereas fu increases (e.g., from 0.2 to 0.4).

els,. . .), this does not mean that more free drug is
available for tissue distribution and drug action (see
Fig 3 for explanation).
There is a persistent inclination in veterinary medicine to report ‘‘total tissue concentrations’’ for some
antibiotics (especially macrolides) and to argue that
this ‘‘tissue level’’ is better related to efficacy than the
plasma concentration. However, this point has been
challenged because the total tissue concentration determined after homogenization may be very different
from the biophase concentration whatever its location
(intra- or extracellular). For further information about
the irrelevance of tissue concentrations to predict antibiotic efficacy, see Barza (1994), Carbon (1990),
Kneer (1993), Schentag (1989), Schentag (1990) and
EMEA Points to consider (Anonymous 2000).

PK/PD ENDPOINTS: ANALYTICAL
PRESENTATION
Various empirical PK/PD indices have been proposed (Aliabadi and Lees 1997, 2000; Hyatt et al 1995;
Sanchez-Navarro and Sanchez Recio 1999; Schentag
et al 1991) to predict the success or failure of therapy.
Three appear to be sufficient to predict drug effectiveness: T > MIC (therapeutic time) when antibiotics are
time-dependent, AUIC (but also AUC/MIC), and Cmax /
MIC (inhibitory ratio) when antibiotics are concentra-

tion-dependent. Fig 4 shows how to compute these indices, while Figs 5 and 6 illustrate some of the
difficulties encountered (discontinuity and non-discrimination between different pharmacokinetic profiles).
AUIC, AUC/MIC, T > MIC, and Cmax /MIC are said
to be PK/PD indices of efficacy because they comprise
a PK parameter (AUC, T > MIC, Cmax . . .) and a
common PD parameter, the MIC. Hence they allow
dual dosage individualization, based on a point consideration of the microbiological susceptibility and on
the variation of the disposition kinetics.
One of the disadvantages of the MIC is that it is
established over a 1:2 dilution scheme which has an
inherent inaccuracy of 100%. Forrest et al (1997) introduced the MIC mid-point, the mid-point between
the recorded MIC, and the next lower value in the dilution series, to replace the conventional MIC when
calculating the efficacy index. Although MIC90 is the
standard value, MIC50 is more precisely computable
and has been suggested as a reasonable target to avoid
administration of excessively high doses of antibiotics
(Schentag 2000).
AUIC is the partial AUC for the period of time
during which concentrations are above the MIC divided by the MIC (Schentag et al 1991) and thus considers only the period when inhibitory activity is
present. AUIC should be calculated at the steady-state

and over 24 h (Schentag et al 1996). It is very frequently


110

P. L. Toutain, J. R. E. del Castillo, A. Bousquet-Melou

FIG 4: Computation of the main PK/PD indices for an MIC of 0.6 lg/mL.
AUIC considers the AUC between t1 and t2 (hatched area). Units should be in
hours and these indexes should preferably be established under steady state
conditions and for a 24 h dosing interval. Data were simulated with a monocompartmental model (Vc ¼ 1.5 L/kg), for a rate constant of absorption (Ka) of
0:006 hÀ1 , a rate constant of elimination (K10 ) of 0:004 hÀ1 and a lag time of
60 min. The dose was 5 mg/kg.

FIG 6: Lack of discrimination observed with AUIC. The curves were simulated with the same total dose (5 mg/kg) as either a single IV bolus (5 mg/kg),
or an IV loading bolus (0.45 mg/kg) followed by a 1011 min infusion of 4.5 lg/
kg/min. MIC was 0.3 lg/mL. The AUIC values are very similar for both dosage
regimens despite very dissimilar plasma profiles, while for a concentrationdependent antibiotic Cmax /MIC can discriminate between the two regimens. At
an MIC of over 0.30 lg/mL, AUIC becomes null during the infusion.

24 h dosing interval should be about five times the MIC
(actually 125/24 h).
If it is acknowledged that AUC is determined only by
plasma clearance and bioavailability and that free (not
total) concentrations should be considered, a maintenance dose achieving a given AUIC (or AUC/MIC) is
easily estimated with the following general equation:
AUIC Â MIC Â Clðper dayÞ
;
ð1Þ
fu  F %  24 h

where AUIC (or AUC/MIC) is the targeted endpoint
in hours (e.g., 125 h), the MIC is the targeted pathogen,
Cl the plasma (total) clearance in days, fu the free
fraction of the drug in plasma (from 0 to 1) and F% the
bioavailability factor (from 0 to 1). In Eq. (1), AUIC/
24 h may be viewed as the desired multiplicative factor
for MIC.
Eq. (1) can be simplified by ignoring fu when the
free fraction is dominant (e.g., for aminoglycosides)
and also F% for the IV route F ẳ 1ị. Conversely, for
drugs extensively bound in plasma fu should be taken
into account (Hyatt et al 1995). The classical values
reported for AUIC (125 and 250 h) were obtained for
quinolones with low plasma binding. However, for a
quinolone extensively bound to plasma proteins, it is
preferable to introduce fu into Eq. (1) rather than to
increase the targeted AUIC. This parameter should be
considered as a target for free, not total plasma AUC,
as MICs are homogeneous to free, not total concentrations, and only free concentrations are microbiologically active (Cars 1997). The advantage of this
approach is to use a single targeted AUIC value for all
antibiotics of a given class, whatever the extent of the
binding to plasma proteins.
Cmax /MIC (inhibitory ratio) is another PK/PD index,
Cmax being a hybrid parameter influenced by plasma

Doseper dayị ẳ

FIG 5: Discontinuity of PK/PD indices. The curves were simulated with a
monocompartmental model for two close dosage regimens: a total dose of
1.28 (dose 1) or 0.92 mg/kg (dose 2), i.e., a loading dose of 0.525 (dose 1) or

0.375 mg/kg (dose 2), followed by a 720 min infusion of 0.00105 lg/kg/min
(dose 1) or 0.000750 lg/kg/min (dose 2). At an MIC of 0.3 lg/mL, the differences between the two regimens in terms of T > MIC and AUIC are very large
and unlikely to reflect the actual clinical difference, whereas at a slightly lower
MIC (0.25 lg/mL) these differences become minimal. AUC/MIC is not subject
to this discontinuity and may be preferred to AUIC.

reported in the literature as a dimensionless number
(e.g., 125, and 250), but AUIC (or AUC/MIC) actually
has a time dimension and saying that the AUIC should
be 125 h to optimize efficacy is in practice equivalent to
saying that the average plasma concentration over a


The pharmacokinetic–pharmacodynamic approach to antibiotics

clearance and bioavailability, and by the rate constants
of absorption and elimination as well. Thus Cmax /MIC
reflects better than AUC/MIC the initial concentration
build-up in plasma, which can be relevant if rapid attainment of a high concentration is desirable to optimize drug efficacy and minimize the emergence of
resistance.
T > MIC (therapeutic time) is obtained by simple
inspection of the simulated curve and generally expressed as a percentage of the dosage interval. It is
kinetically more complicated and is largely controlled
by the terminal half-life, which is a hybrid process involving both plasma clearance and drug distribution or
the rate constant of absorption (for long acting drug
formulations undergoing a flip-flop process).
These indices have been validated by various approaches ranging from in silico (computer) simulations
to meta-analyses of clinical trials. All methods found
the different indices to be highly correlated (Preston
et al 1998; Sanchez-Recio et al 2000) and it was difficult

to determine which PK/PD parameter was most informative, as all three (T > MIC, AUIC, and Cmax /
MIC) increased with increasing dose.
To circumvent this difficulty, Corvaisier et al (1998)
using in silico simulations proposed a new composite
index for the first 24 h, the weighted AUC (WAUC),
which is the AUC/MIC weighted by the percentage of
the total time for which plasma drug levels are above
the MIC:
T > MIChị
AUC
h
;
2ị
WAUChị ẳ
MIC
T > MICÞ maxðhÞ
where (T > MICÞ max is equal to 24 h. This index can
be used for both concentration- and time-dependent
antibiotics and has a high sensitivity to changes in MICs.

MAGNITUDE OF THE PK/PD PARAMETER
REQUIRED FOR EFFICACY
No PK/PD indices have yet been firmly validated in
veterinary medicine, but as the differences can only
reflect variations in species PK and MIC, it is reasonable to assume that the critical (breakpoint) values of
these parameters to achieve efficacy will be similar in
different animal species (Craig 1998). Thus, the results
obtained in animal infection models or clinical human
trials should be good starting points to design dosage
regimens for a new antibacterial or a new species.

Studies of b-lactams in animal infection models have
demonstrated that T > MIC does not need to be 100%
of the dosage interval to develop a significant antibacterial effect. In patients with otitis media (Streptococcus
pneumoniae, Haemophilus influenza), a T > MIC of
over 40% was required to achieve an 85–100% bacteriological cure rate with different b-lactams.
When mortality was selected as an endpoint for animals infected with S. pneumoniae and treated for several days with penicillins or cephalosporins, mortality
was close to 100% if T > MIC was 620% of the dosage
interval, but 90–100% survival was reached when
T > MIC was P40–50% of this interval (Craig 1998).

111

Finally, it can be recommended that T > MIC
should be at least 50% and preferably P80% of the
dosage interval to achieve an optimal bactericidal effect. If the drug is extensively bound to plasma proteins, this recommendation holds for free, not total
concentrations.
Using fluoroquinolones, in different models of infection with various species of gram-positive and gramnegative bacteria, AUC/MIC ratios of < 30 h were associated with >50% mortality but AUC/MIC values of
P100 h with almost no mortality (Craig 1998).
In seriously ill patients, a 24 h AUC/MIC value
of P125 h for ciprofloxacin achieved a satisfactory
outcome whereas lower values resulted in clinical
and bacteriological cure rates of <50% (Forrest et al
1993).
In treatment with quinolones, for an AUIC of
>250 h, 60% of patients became culture negative within
one day. When AUIC lay between 125 and 250 h, negative cultures were generally not achieved until the
sixth day, while in patients whose AUIC was < 125 h, a
second antibiotic was required (Schentag 2000).
Preston et al (1998) showed that for levofloxacin the
Cmax /MIC, AUC/MIC and T > MIC ratios were indistinguishable to predict a successful clinical outcome. In

contrast, Cmax /MIC was discriminant for microbiological outcomes and patients achieving a Cmax /MIC of
P12:2 displayed 100% microbiological eradication.
In the case of aminoglycosides, in animal infection
models the 24 h AUC/MIC ratio was a better predictor
of therapeutic efficacy than the Cmax /MIC ratio,
whereas the reverse was true in human clinical trials
(Craig 1998). To obtain a clinical response of P90%
and reduce the risk of emergence of resistance, Cmax /
MIC needs to be 8–10 (Moore et al 1984). This is easily
achieved with a single daily large dose of aminoglycosides, which also minimizes the consequences of
adaptive resistance (Daikos et al 1991). Adaptive resistance is a phenotypic and reversible increase in MIC
associated with a temporary lack of drug transport into
the bacterial cells. Its dissipation and restoration of
microbial susceptibility requires a drug-free period,
easily obtained with a once daily dosage regimen because the half-lives of aminoglycosides are short (about
2 h). Once daily dosing also limits the incidence of
nephrotoxicity and ototoxicity, as the tissular accumulation of aminoglycosides is saturable at clinically
meaningful concentrations.
Additional work will be required to establish the
magnitudes of the PK/PD parameters correlating with
the efficacy of macrolides, azalides, clindamycin, tetracyclines, and glycopeptides (Craig 1998). For further
information, see the comprehensive reviews of Hyatt
et al (1995) and Craig (1998).

PK/PD INDICES AND THE RISK OF RESISTANCE
Study of the emergence of resistance is an integral
part of the PK/PD approach aimed at limiting antimicrobial resistance (Anonymous 2000) and, according to
Schentag et al (1996), the design of appropriate dosage

112

P. L. Toutain, J. R. E. del Castillo, A. Bousquet-Melou

regimens may be the single most important contribution of clinical pharmacology to the resistance problem.
Resistance mechanisms can arise as the result of a
single point mutation. Since the frequency of occurrence is relatively high, the bacterial population is not
homogeneous and behaves as a mixture of distinct
populations having their own antibiotic susceptibility.
In this situation, exposure to antibiotics does not induce
but selects resistance. The emergence of resistance is
only the predictable overgrowth of a pre-existing subpopulation with an initially lower level of susceptibility
(Schentag et al 1998) and dosage regimens designed to
rapidly eradicate this less susceptible subpopulation
limit the risk of resistance (Schentag 2000).
The most important risk factor for emergence of
resistance is repeated exposure to suboptimal concentrations of antibiotics (Burgess 1999). In in vitro models
simulating human PK of ciprofloxacin and sparfloxacin,
high Cmax /MIC ratios were associated with a lower incidence of bacterial resistance for S. pneumoniae
(Thorburn and Edwards 2001). Similarly, in a mouse
peritonitis model, less Pseudomonas aeruginosa resistance was observed for ciprofloxacin when Cmax /MIC
was % 20 as compared to 10 (Michae-Hamzehpour et al
1987). Using ciprofloxacin against P. aeruginosa in
man, a single daily dose of 1200 mg triggered less resistance than 600 mg twice or 400 mg three times daily
(Marchbanks et al 1993).
In pneumococcal infection, Thomas et al (1998) investigated the probability of the development of resistant organisms in relation to the antibiotic dose. After 5
days treatment, approximately 50% of patients who
had AUIC < 100 h developed resistance and this
increased to 93% after 3 weeks treatment. On the
contrary, among patients who had AUIC > 100 h, resistance developed in only 8%.

The concept of a mutant prevention concentration
(MPC) is a possible application of the PK/PD approach
for fluoroquinolones. Briefly, two successive mutations
(e.g., on gyrase and then on topoisomerase IV) result in
mutant strains of high resistance. In this framework of
sequential mutations, there exists a concentration window lying between the MIC of wild bacteria (no mutation) and the MPC, a concentration which blocks the
growth of first step mutants. In this mutant selective
window, the first step mutant population has an advantage over fully susceptible bacteria and increasing its
population size increases the probability of having
double mutants. In contrast, above the MPC the probability that a wild bacterium will undergo the two resistance mutations is very low. A practical strategy is to
reduce the size of the mutant selective window, which
can be achieved in different ways including by adjustment of the dosage regimen (Zhao and Drlica 2001).
Large initial doses of quinolones and aminoglycosides are recommended to eradicate the resistant subpopulations. Thus, for aminoglycosides a Cmax /MIC
ratio of 10–12 and for quinolones an AUIC of >125–
250 h are desirable to minimize the survival and overgrowth of resistant strains. Finally, for b-lactams increasing the duration of T > MIC should help to
prevent the emergence of resistance.

PK/PD SUSCEPTIBILITY TESTING AND DOSAGE
REGIMEN INDIVIDUALIZATION
The objective of dual dosage regimen individualization is to adapt the antibiotic dosage regimen for the
bacterial susceptibility (PD) and for the effect of the
disease state (or other co-variables) on the antibiotic
availability (PK). In the future, inexpensive methods of
performing quantitative in vitro susceptibility tests
would allow the practitioner to adapt the dosage regimen to the pathogen susceptibility as given by its
measured MIC value. The practitioner would then be
in a position to determine himself the best dosage
regimen (dose, interval of administration) to reach a
given target endpoint. Such an approach would require
a knowledge of the relevant population kinetics, computation assistance and regulations allowing flexible

labelling (i.e., antibiotics with a marketing authorization for a range of doses) (Martinez et al 1995).

PK/PD AND VALIDATED CLINICAL
BREAKPOINTS FOR VETERINARY MEDICINE
The results of antimicrobial susceptibility tests are
generally reported qualitatively, the isolate being designated as susceptible, intermediate or resistant. This
classification is based on breakpoint values, i.e., specific
MICs allowing one to predict clinical efficacy or failure
on the basis of an in vitro susceptibility test. The clinical value of these tests for the guiding of an individual
animal therapy remains unclear because most breakpoints were determined on the basis of human microbiological, pharmacological and clinical outcomes.
Recently, interpretive criteria for bacterial pathogens
isolated from animals and breakpoints for several
pathogen drug combinations for swine and cattle have
been developed (for more information see nccls.org).
Though performance standards and testing criteria of
veterinary antimicrobial susceptibility tests have developed (NCCLS 1999a,b), true assessment of their
predictive value of clinical outcome has seldom been
addressed. When such assessment has been carried out
a posteriori, it appeared that the predictive value of
susceptibility tests was less than ideal (see Shpigel et al
1998). Therefore, it is time to examine if this method
for laboratory detection of resistance is as good for
guiding an individual patient therapy than for providing resistance surveillance data (Gould 2000).
Currently, approved interpretive criteria do not
formally take into account, in either human or veterinary medicine, the population concepts of PK/PD, i.e.,
they do not combine kinetic variability in the animal
population (including diseased animals) with what is
known about the population distribution of MIC values
for the target pathogen (not a single MIC value as
currently done). A proposal for the determination of

breakpoints having a clinical value should be carried
out within the framework of population PK/PD as recently outlined by Ambrose and Grasela (2000).
Briefly, a rational approach would consist of: (i) generating by simulation all possible drug exposures


The pharmacokinetic–pharmacodynamic approach to antibiotics

113

get bacteria, i.e., without direct measurement of the
antibiotic effect (in infection models) or efficacy (in
clinical trials). The main advantages of this approach
are summarized in Table 1.
In the case of a new antibiotic, a knowledge of the
expected MIC for the target pathogens and pharmacokinetic parameters in healthy animals can give very
early an order of magnitude of the future dosage regimen, without recourse to clinical trials or infection
models which can be associated with difficulties in
terms of validity (Pollyanna effect).
The influence of antibiotic exposure on the bacterial
efficacy of an antibiotic can be evaluated in in vitro
kinetic models simulating in vivo situations and data
are readily extrapolated from these models by means of
PK/PD indices.
In a clinical setting, the PK/PD paradigm offers a
rational approach for dual dosage adaptation, i.e., adjustment for variations in both antibiotic availability
(PK) and bacterial susceptibility (PD). Consideration
of PK and PD variability should also in the future be
the best way to select appropriate breakpoints for
susceptibility tests, whereby population studies will
have a major influence on the prudent and rational use

of antibiotics.
Finally, PK/PD for a given antibiotic can easily be
determined by different independent groups, thus
limiting the risk of conflicts of interest in commercial
clinical drug trials.
FIG 7: Population AUIC breakpoint for marbofloxacin in dog. The percentage of dogs having a serum area under the inhibitory curve (AUIC) > 125 h (A)
and 48 h (B) for the selected MIC (lg/mL) are shown after a single marbofloxacin dose of 1, 2, 3, and 4 mg/kg (curves 1, 2, 3, and 4, respectively).
Dashed lines indicate critical MIC values to guarantee an AUIC of 125 or 48 h
in 90% of a dog population. Kinetic parameters were determined in a population of 63 dogs given a single intravenous dose of marbofloxacin (2 mg/kg).
Serum data were analysed using a non-linear mixed effect regression model
allowing to compute mean population parameters and the variance–covariance matrix of the individual parameters measuring the dispersion of the individual parameters in the population (Regnier et al unpublished data).

(AUC, AUIC, Cmax /MIC, T > MIC) for the standard
dosage regimen, which requires a knowledge of population parameters with typical (mean) values and their
variance, (ii) establishing MIC distributions for clinically relevant pathogens, and (iii) generating random
values across pharmacokinetic (e.g., AUC) and MIC
distributions conform to their probabilities.
The resultant AUC/MIC probability distribution
would allow one to examine the entire range of possible AUC/MIC ratios and the probability of achieving
each ratio (Ambrose and Quintiliani 2000). Such an
approach was used for marbofloxacin in dog (Fig 7).

CONCLUSION
The contribution of the PK/PD approach to the
determination of an antibiotic dosage regimen relies on
the paradigm that the dosage regimen can be approximated from plasma levels during antibiotic exposure in
healthy animals scaled by the susceptibility of the tar-

ACKNOWLEDGMENTS
The authors wish to thank Prof. Peter Lees for his

valuable critical comments. The work was supported by
a grant from INRA (Project entitled ‘‘Transversalit
e:
utilisation raisonnee des antibiotiquesa visee th
erapeutique en elevage’’).

REFERENCES
ALIABADI, F. S. & LEES, P. (1997) Pharmacodynamic and pharmacokinetic
inter-relationships of antibacterial drugs. Journal of Veterinary Pharmacology and Therapeutics 20, 14–17
ALIABADI, F. S. & LEES, P. (2000) Antibiotic treatment for animals: effect
on bacterial population and dosage regimen optimisation. International
Journal of Antimicrobial Agents 14, 307–313
ALIABADI, F. S. & LEES, P. (2001) Pharmacokinetics and pharmacodynamics of danofloxacin in serum and tissue fluids of goats following
intravenous and intramuscular administration. American Journal of Veterinary Research 62, 1979–1989
AMBROSE, P. G. & GRASELA, D. M. (2000) The use of Monte Carlo
simulation to examine pharmacodynamic variance of drugs: fluoroquinolone pharmacodynamics against Streptococcus pneumoniae. Diagnostic
Microbiology and Infectious Disease 38, 151–157
AMBROSE, P. G. & QUINTILIANI, R. (2000) Limitations of single point
pharmacodynamic analysis. Pediatric Infectious Disease Journal 19, 769
ANONYMOUS. (1998) The Copenhagen Recommendations. In The Microbial Threat. Eds: Ministry of Health, Ministry of Food & Agriculture and
Fisheries Denmark. Copenhagen, Denmark
ANONYMOUS. (2000) Points to consider on pharmacokinetics and pharmacodynamics in the development of antibacterial medicinal products, In:
EMEA European Agency for the Evaluation of Medicinal Products. Eds
CPMP. London, pp 1–7
ANONYMOUS. (2001) Sponsorship, authorship, and accountability. Medical
Journal of Australia 175, 294–296
BARZA, M. (1994) Challenges to antibiotic activity in tissue. Clinical
Infectious Diseases 19, 910–915

114

P. L. Toutain, J. R. E. del Castillo, A. Bousquet-Melou

BORDING, A. (1990) The clinical trial in veterinary practice. In 11th IPVS
Congress. Eds. Lausanne. p. 413
BURGESS, D. S. (1999) Pharmacodynamic principles of antimicrobial therapy
in the prevention of resistance. Chest 115, 19S–23S
CARBON, C. (1990) Significance of tissue levels for prediction of antibiotic
efficacy and determination of dosage. European Journal of Clinical
Microbiology & Infectious Disease 9, 510–516
CARS, O. (1997) Efficacy of beta-lactam antibiotics: integration of pharmacokinetics and pharmacodynamics. Diagnostic Microbiology and Infectious
Disease 27, 29–33
CHARLESTON, B., GATE, J. J., AITKEN, I. A., STEPHAN, B. &
FROYMAN, R. (1998) Comparison of the efficacies of three fluoroquinolone antimicrobial agents, given as continuous or pulsed-water medication,
against Escherichia coli infection in chickens. Antimicrobial Agents and
Chemotherapy 42, 83–87
CORVAISIER, S., MAIRE, P. H., BOUVIER D’YVOIRE, M. Y., BARBAUT, X., BLEYZAC, N. & JELLIFFE, R. W. (1998) Comparisons
between antimicrobial pharmacodynamic indices and bacterial killing as
described by using the Zhi model. Antimicrobial Agents and Chemotherapy
42, 1731–1737
CRAIG, W. A. (1998) Pharmacokinetic/pharmacodynamic parameters: rationale for antibacterial dosing of mice and men. Clinical Infectious Diseases
26, 1–10, quiz 11–12
DAGAN, R., KLUGMAN, K. P., CRAIG, W. A. & BAQUERO, F. (2001)
Evidence to support the rationale that bacterial eradication in respiratory
tract infection is an important aim of antimicrobial therapy. Journal of
Antimicrobial Chemotherapy 47, 129–140
DAIKOS, G. L., LOLANS, V. T. & JACKSON, G. G. (1991) First-exposure
adaptive resistance to aminoglycoside antibiotics in vivo with meaning for
optimal clinical use. Antimicrobial Agents and Chemotherapy 35, 117–123

DRUSANO, G. L., JOHNSON, D. E., ROSEN, M. & STANDIFORD, H. C.
(1993) Pharmacodynamics of a fluoroquinolone antimicrobial agent in a
neutropenic rat model of Pseudomonas sepsis. Antimicrobial Agents and
Chemotherapy 37, 483–490
FORREST, A., NIX, D. E., BALLOW, C. H., GOSS, T. F., BIRMINGHAM,
M. C. & SCHENTAG, J. J. (1993) Pharmacodynamics of intravenous
ciprofloxacin in seriously ill patients. Antimicrobial Agents and Chemotherapy 37, 1073–1081
FORREST, A., CHODOSH, S., AMANTEA, M. A., COLLINS, D. A. &
SCHENTAG, J. J. (1997) Pharmacokinetics and pharmacodynamics of oral
grepafloxacin in patients with acute bacterial exacerbations of chronic
bronchitis. Journal of Antimicrobial Chemotherapy 40, 45–57
GOULD, I. M. (2000) Towards a common susceptibility testing method?
Journal of Antimicrobial Chemotherapy 45, 757–762
HYATT, J. M., MCKINNON, P. S., ZIMMER, G. S. & SCHENTAG, J. J.
(1995) The importance of pharmacokinetic/pharmacodynamic surrogate
markers to outcome. Focus on antibacterial agents. Clinical Pharmacokinetics 28, 143–160
KNEER, J. (1993) Relevance of antibiotic tissue penetration in treating
respiratory tract infections. Respiration 60, 32–37
KORITZ, G. D. & BEVILL, R. F. (1991) Role of pharmacokinetics in the
outcome of infections. In 5th Congress of EAVPT Copenhagen, Acta
Veterinaria Scandinavia, pp. 18–27
KORITZ, G. D., KILROY, C. R. & BEVILL, R. F. (1994) Pharmacokineticspharmacodynamic modelling of antibacterial therapy in vitro. In 6th
International Congress of the European Association fro Veterinary Pharmacology and Therapeutics Edinburgh, UK, Blackwell Scientific Publications
LEES, P. & ALIABADI, F. S. (2002) Rational dosing of antimicrobial drugs:
animals versus humans. International Journal of Antimicrobial Agents 19,
269–284
LEGGETT, J. E., EBERT, S., FANTIN, B. & CRAIG, W. A. (1991)
Comparative dose-effect relations at several dosing intervals for betalactam, aminoglycoside and quinolone antibiotics against gram-negative
bacilli in murine thigh-infection and pneumonitis models. Scandinavian
Journal of Infectious Diseases 74(Suppl.), 179–184

MARCHANT, C. D., CARLIN, S. A., JOHNSON, C. E. & SHURIN, P. A.
(1992) Measuring the comparative efficacy of antibacterial agents for acute
otitis media: the ‘‘Pollyanna phenomenon’’. Journal of Pediatrics 120, 72–
77
MARCHBANKS, C. R., MCKIEL, J. R., GILBERT, D. H., ROBILLARD,
N. J., PAINTER, B., ZINNER, S. H. & DUDLEY, M. N. (1993) Dose
ranging and fractionation of intravenous ciprofloxacin against Pseudomonas aeruginosa and Staphylococcus aureus in an in vitro model of
infection. Antimicrobial Agents and Chemotherapy 37, 1756–1763
MARTINEZ, M. N., RIVIERE, J. E. & KORITZ, G. D. (1995) Review of the
first interactive workshop on professional flexible labeling. Journal of the
American Veterinary Medical Association 207, 865
MICHAE-HAMZEHPOUR, M., AUCKENTHALER, R., REGAMEY, P. &
PECHERE, J. C. (1987) Resistance occurring after fluoroquinolone
therapy of experimental Pseudomonas aeruginosa peritonitis. Antimicrobial Agents and Chemotherapy 31, 1803–1808
MOORE, R. D., SMITH, C. R. & LIETMAN, P. S. (1984) Association of
aminoglycoside plasma levels with therapeutic outcome in gram-negative
pneumonia. American Journal of Medicine 77, 657–662

MURAKAWA, T., SAKAMOTO, H., HIROSE, T. & NISHIDA, M. (1980)
New in vitro kinetic model for evaluating bactericidal efficacy of antibiotics. Antimicrobial Agents and Chemotherapy 18, 377–381
NCCLS (1999a) Performance Standards for Antimicrobial Disk and Dilution
Susceptibility Tests for Bacteria Isolated from Animals; Approved Standard
NCCLS Document M31-A. The National Committee for Clinical Laboratory Standards, Wayne, PA, USA
NCCLS (1999b) Development of In Vitro Susceptibility Testing Criteria and
Quality Control Parameters for Veterinary Antimicrobial Agents; Approved
Guideline NCCLS Document M37-A. The National Committee for Clinical
Laboratory Standards, Wayne, PA, USA
PRESTON, S. L., DRUSANO, G. L., BERMAN, A. L., FOWLER, C. L.,
CHOW, A. T., DORNSEIF, B., REICHL, V., NATARAJAN, J. &
CORRADO, M. (1998) Pharmacodynamics of levofloxacin: a new paradigm for early clinical trials. Journal of American Medical Association 279,

125–129
RENARD, L., SANDERS, P., LAURENTIE, L. & DELMAS, J. M. (1996)
Pharmacokinetic–pharmacodynamic model for spiramycin in staphylococcal mastitis. Journal of Veterinary Pharmacology and Therapeutics 19, 95–
103
ROOSENDAAL, R., BAKKER-WOUDENBERG, I. A., VAN DEN BERG,
J. C. & MICHEL, M. F. (1985) Therapeutic efficacy of continuous versus
intermittent administration of ceftazidime in an experimental Klebsiella
pneumoniae pneumonia in rats. Journal of Infectious Disease 152, 373–378
SANCHEZ-NAVARRO, A. & SANCHEZ RECIO, M. M. (1999) Basis of
anti-infective therapy: pharmacokinetic–pharmacodynamic criteria and
methodology for dual dosage individualisation. Clinical Pharmacokinetics
37, 289–304
SANCHEZ-RECIO, M. M., COLINO, C. I. & SANCHEZ-NAVARRO, A.
(2000) A retrospective analysis of pharmacokinetic/pharmacodynamic
indices as indicators of the clinical efficacy of ciprofloxacin. Journal of
Antimicrobial Chemotherapy 45, 321–328
SCHENTAG, J. J., SWANSON, D. J. & SMITH, I. L. (1985) Dual
individualization: antibiotic dosage calculation from the integration of invitro pharmacodynamics and in-vivo pharmacokinetics. Journal of Antimicrobial Chemotherapy 15(Suppl. A), 47–57
SCHENTAG, J. J. (1989) Clinical significance of antibiotic tissue penetration.
Clinical Pharmacokinetics 16, 25–31
SCHENTAG, J. J. (1990) The significance of the relationship between tissue:
serum ratios, tissue concentrations and the location of microorganisms.
Research and Clinical Forums 12, 23–27
SCHENTAG, J. J., NIX, D. E. & ADELMAN, M. H. (1991) Mathematical
examination of dual individualization principles (I): relationships between
AUC above MIC and area under the inhibitory curve for cefmenoxime,
ciprofloxacin, and tobramycin. Drugs Intelligence and Clinical Pharmacy
25, 1050–1057
SCHENTAG, J. J., NIX, D. E., FORREST, A. & ADELMAN, M. H. (1996)
AUIC—the universal parameter within the constraint of a reasonable

dosing interval. Annals of Pharmacotherapy 30, 1029–1031
SCHENTAG, J. J., HYATT, J. M., CARR, J. R., PALADINO, J. A.,
BIRMINGHAM, M. C., ZIMMER, G. S. & CUMBO, T. J. (1998) Genesis
of methicillin-resistant Staphylococcus aureus (MRSA), how treatment of
MRSA infections has selected for vancomycin-resistant Enterococcus
faecium, and the importance of antibiotic management and infection
control. Clinical Infectious Diseases 26, 1204–1214
SCHENTAG, J. J. (2000) Clinical pharmacology of the fluoroquinolones:
studies in human dynamic/kinetic models. Clinical Infectious Diseases
31(Suppl. 2), S40–S44
SHPIGEL, N. Y., WINKLER, M., ZIV, G. & SARAN, A. (1998) Relationship
between in vitro sensitivity of coliform pathogens in the udder and the
outcome of treatment for clinical mastitis. Veterinary Record 142, 135–137
THOMAS, J. K., FORREST, A., BHAVNANI, S. M., HYATT, J. M.,
CHENG, A., BALLOW, C. H. & SCHENTAG, J. J. (1998) Pharmacodynamic evaluation of factors associated with the development of bacterial
resistance in acutely ill patients during therapy. Antimicrobial Agents and
Chemotherapy 42, 521–527
THORBURN, C. E. & EDWARDS, D. I. (2001) The effect of pharmacokinetics on the bactericidal activity of ciprofloxacin and sparfloxacin against
Streptococcus pneumoniae and the emergence of resistance. Journal of
Antimicrobial Chemotherapy 48, 15–22
VAN DONKERSGOED, J. (1992) Meta-analysis of field trials of antimicrobial mass medication for prophylaxis of bovine respiratory disease in
feedlot cattle. Canadian Veterianry Journal 33, 786–795
VOGELMAN, B., GUDMUNDSSON, S., LEGGETT, J., TURNIDGE, J.,
EBERT, S. & CRAIG, W. A. (1988) Correlation of antimicrobial
pharmacokinetic parameters with therapeutic efficacy in an animal model.
Journal of Infectious Disease 158, 831–847
YANCEY, Jr.,, R. J., EVANS, R. A., KRATZER, D. D., PAULISSEN, J. B.
& CARMER, S. G. (1990) Efficacy of ceftiofur hydrochloride for
treatment of experimentally induced colibacillosis in neonatal swine.
American Journal of Veterinary Research 51, 349–353

ZHAO, X. & DRLICA, K. (2001) Restricting the selection of antibioticresistant mutants: a general strategy derived from fluoroquinolones studies.
Clinical Infectious Diseases 33, S147–S156