Inter- and Intraquinolone Predictors of Antimicrobial Effect in an In Vitro Dynamic Model: New Insight into a Widely Used Concept

Abstract

Earlier efforts to search for pharmacokinetic and bacteriological predictors of fluoroquinolone antimicrobial effects (AMEs) have resulted in conflicting findings. To elucidate whether these conflicts are real or apparent, several predictors of the AMEs of two pharmacokinetically different antibiotics, trovafloxacin (TRO) and ciprofloxacin (CIP), as well as different dosing regimens of CIP were examined. The AMEs of TRO given once daily (q.d.) and CIP given q.d. and twice daily (b.i.d.) against Escherichia coli, Pseudomonas aeruginosa, and Klebsiella pneumoniae were studied in an in vitro dynamic model. Different monoexponential pharmacokinetic profiles were simulated with a TRO half-life of 9.2 h and a CIP half-life of 4.0 h to provide similar eightfold ranges of the area under the concentration-time curve (AUC)-to-MIC ratios, from 54 to 432 and from 59 to 473 (microg x h/ml)/(microg/ml), respectively. In each case the observation periods were designed to incorporate full-term regrowth phases in the time-kill curves, and the AME was expressed by its intensity (IE; the area between the control growth and time-kill and regrowth curves up to the point at which the viable counts of regrowing bacteria are close to the maximum values observed without drug). Species-independent linear relationships were established between IE and log AUC/MIC, log AUC above MIC (log AUCeff), and time above the MIC (Teff). Specific and nonsuperimposed IE versus log AUC/MIC or log AUCeff relationships were inherent in each of the treatments: TRO given q.d. (r2 = 0.97 and 0.96), CIP given q.d. (r2 = 0.98 and 0.96), and CIP given b.i.d. (r2 = 0.95 and 0.93). This suggests that in order to combine data sets obtained with individual quinolones to examine potential predictors, one must be sure that these sets may be combined. Unlike AUC/MIC and AUCeff, the IE-Teff relationships plotted for the different quinolones and dosing regimens were nonspecific and virtually superimposed (r2 = 0.95). Hence, AUC/MIC, AUCeff and Teff were equally good predictors of the AME of each of the quinolones and each dosing regimen taken separately, whereas Teff was also a good predictor of the AMEs of the quinolones and their regimens taken together. However, neither the quinolones nor the dosing regimens could be distinguished solely on the basis of Teff whereas they could be distinguished on the basis of AUC/MIC or AUCeff. Thus, two types of predictors of the quinolone AME may be identified: intraquinolone and/or intraregimen predictors (AUC/MIC, AUCeff and Teff) and an interquinolone and interregimen predictor (Teff). Teff may be able to accurately predict the AME of one quinolone on the basis of the data obtained for another quinolone.

Full text

ANTIMICROBIAL AGENTS AND CHEMOTHERAPY,
0066-4804/98/$04.0010Mar. 1998, p. 659–665 Vol. 42, No. 3
Copyright © 1998, American Society for Microbiology
Inter- and Intraquinolone Predictors of Antimicrobial Effect in
an In Vitro Dynamic Model: New Insight into
a Widely Used Concept
ALEXANDER A. FIRSOV,
1
* ALEXANDER A. SHEVCHENKO,
1
SERGEY N. VOSTROV,
1
AND STEPHEN H. ZINNER
2
Department of Pharmacokinetics, Centre of Science & Technology LekBioTech, Moscow 117246, Russia,
1
and Division of Infectious Diseases, Roger Williams Medical Center, Rhode
Island Hospital, Brown University, Providence, Rhode Island
2
Received 15 May 1997/Returned for modification 5 October 1997/Accepted 20 December 1997
Earlier efforts to search for pharmacokinetic and bacteriological predictors of fluoroquinolone antimicrobial
effects (AMEs) have resulted in conflicting findings. To elucidate whether these conflicts are real or apparent,
several predictors of the AMEs of two pharmacokinetically different antibiotics, trovafloxacin (TRO) and cipro-
floxacin (CIP), as well as different dosing regimens of CIP were examined. The AMEs of TRO given once daily
(q.d.) and CIP given q.d. and twice daily (b.i.d.) against Escherichia coli,Pseudomonas aeruginosa, and Klebsiella
pneumoniae were studied in an in vitro dynamic model. Different monoexponential pharmacokinetic profiles
were simulated with a TRO half-life of 9.2 h and a CIP half-life of 4.0 h to provide similar eightfold ranges of
the area under the concentration-time curve (AUC)-to-MIC ratios, from 54 to 432 and from 59 to 473 (mgzh/ml)/
(mg/ml), respectively. In each case the observation periods were designed to incorporate full-term regrowth
phases in the time-kill curves, and the AME was expressed by its intensity (I
E
; the area between the control
growth and time-kill and regrowth curves up to the point at which the viable counts of regrowing bacteria are
close to the maximum values observed without drug). Species-independent linear relationships were estab-
lished between I
E
and log AUC/MIC, log AUC above MIC (log AUC
eff
), and time above the MIC (T
eff
). Specific
and nonsuperimposed I
E
versus log AUC/MIC or log AUC
eff
relationships were inherent in each of the treat-
ments: TRO given q.d. (r
2
50.97 and 0.96), CIP given q.d. (r
2
50.98 and 0.96), and CIP given b.i.d. (r
2
50.95
and 0.93). This suggests that in order to combine data sets obtained with individual quinolones to examine
potential predictors, one must be sure that these sets may be combined. Unlike AUC/MIC and AUC
eff
, the I
E
-
T
eff
relationships plotted for the different quinolones and dosing regimens were nonspecific and virtually super-
imposed (r
2
50.95). Hence, AUC/MIC, AUC
eff
, and T
eff
were equally good predictors of the AME of each of the
quinolones and each dosing regimen taken separately, whereas T
eff
was also a good predictor of the AMEs of
the quinolones and their regimens taken together. However, neither the quinolones nor the dosing regimens
could be distinguished solely on the basis of T
eff
, whereas they could be distinguished on the basis of AUC/MIC
or AUC
eff
. Thus, two types of predictors of the quinolone AME may be identified: intraquinolone and/or intra-
regimen predictors (AUC/MIC, AUC
eff
and T
eff
) and an interquinolone and interregimen predictor (T
eff
). T
eff
may
be able to accurately predict the AME of one quinolone on the basis of the data obtained for another quinolone.
The concept of pharmacokinetic and bacteriological predic-
tors of the antimicrobial effect has been offered as an alterna-
tive to the traditional concentration-response relationships
usually exploited in pharmacology. Unlike many pharmacolog-
ical effects, the antimicrobial effect depends not only on the
drug concentration but also on the exposure time. Further-
more, the use of combined pharmacokinetic and bacteriolog-
ical predictors enhances their actual predictive potential, as-
suming that the predictor-response relationships are bacterial
species independent. Evaluation of a predictor(s) of the anti-
microbial effect and/or clinical outcome has generally been
accepted as a useful tool in the design of optimal dosing reg-
imens (4). For example, if the ratio of the peak antibiotic
concentration (C
max
) to MIC (C
max
/MIC) were shown to be
the best predictor, then intermittent drug administration at
relatively high doses and relatively long intervals might be
superior to drug administration at lower doses and shorter
intervals. On the other hand, if the antimicrobial effect corre-
lates better with the time at which the antibiotic concentration
exceeds the MIC (time above the MIC [T
eff
]), the opposite
dosing strategy would be preferable.
During the past decade the predictive potentials of param-
eters such as C
max
/MIC, T
eff
, the area under the concentration-
time curve (AUC) related to the MIC (AUC/MIC) or the
portion of AUC/MIC that reflects only the time at which the
concentrations are above the MIC (AUIC) (18), and the AUC
above the MIC (AUC
eff
) have been examined for various an-
timicrobial agents (5, 16, 17, 20). Due to considerable covari-
ance among these widely used predictors, the search for a
single optimal predictor was often futile, since one was usually
required to choose among equally good predictors, especially
for pharmacokinetically similar drugs (8).
This covariance has also been noted in in vitro and in vivo
studies with fluoroquinolones (2, 6). In this regard, the use of
only one predictor, for example, C
max
/MIC for enoxacin (2) or
AUIC for ciprofloxacin and ofloxacin (14), does not necessarily
exclude the appropriate selection of alternative predictors. Sim-
ilarly, the preference for AUC/MIC measured within 24 h (AUC/
MIC
24
) over C
max
/MIC as a potential predictor of the effects of
* Corresponding author. Mailing address: Department of Pharma-
cokinetics, Centre of Science & Technology LekBioTech, 8 Nauchny
proezd, Moscow 117246, Russia. Phone: 7 (095) 332-3435. Fax: 7 (095)
331-0101. E-mail: biotec@glas.apc.org.
659

ciprofloxacin and ofloxacin (15), without examination of AUC
eff
or T
eff
, might not be strictly appropriate, since AUC/MIC
24
represents the sum of AUCs produced by the administration of
repeated doses of the drug, while C
max
/MIC reflects only the
impact of the first dose administered in the dynamic model.
The infrequent attempts to directly compare several potential
predictors of quinolone antimicrobial effects have led to con-
flicting results. For example, no differences were reported among
C
max
/MIC, AUC/MIC, and T
eff
as predictors of the efficacy of
lomefloxacin against Pseudomonas sepsis in rats: the predictive
potential of each of these predictors was equally high (r
2
5
0.98 to 0.99) (6). In another study, C
max
/MIC, AUC/MIC, and
T
eff
did not predict the in vitro and in vivo effects produced
by seven antibiotics including ciprofloxacin and fleroxacin (3).
Similarly, AUIC did not predict the effects of four quinolones
in an in vitro dynamic model studied by Wiedemann et al. (25),
in contrast to the data presented by Madaras-Kelly et al. (14).
To elucidate whether these contradictions can be rectified,
several predictors of the antimicrobial effects of pharmacoki-
netically different quinolones, trovafloxacin and ciprofloxacin,
as well as those of two dosing regimens of ciprofloxacin were
examined on the basis of time-kill data obtained in an in vitro
dynamic model.
MATERIALS AND METHODS
Antimicrobial agents. Trovafloxacin mesylate and ciprofloxacin lactate pow-
ders, kindly provided by Roerig, a division of Pfizer, and by Bayer AG, respec-
tively, were used in the study. Stock solutions of the quinolones were prepared in
sterile distilled water.
Bacterial strains. The clinical isolates Escherichia coli 224, Pseudomonas
aeruginosa 48, and Klebsiella pneumoniae 121 were used in the study. Suscepti-
bility testing was performed in triplicate at 24 h postexposure with organisms
grown in Ca
21
- and Mg
21
-supplemented Mueller-Hinton broth; the inoculum
size was 10
6
CFU/ml. The MICs of trovafloxacin for E. coli,P. aeruginosa, and
K. pneumoniae (0.25, 0.3, and 0.25 mg/ml, respectively) were found to be com-
parable to those of ciprofloxacin (0.12, 0.15, and 0.12 mg/ml, respectively).
Simulated pharmacokinetic profiles. A series of monoexponential profiles for
trovafloxacin and ciprofloxacin were simulated. The simulated half-lives (t
1/2
s;
9.25 h for trovafloxacin and 4.0 h for ciprofloxacin) were consistent with the
values reported in humans: 7.2 to 9.9 h (19, 27) and 3.2 to 5.0 h (1, 13, 26),
respectively. Regimens of trovafloxacin given once daily (q.d.) and ciprofloxacin
given q.d. and twice daily (b.i.d.) were used in experiments with both E. coli and
P. aeruginosa. For studies with K. pneumoniae only regimens of trovafloxacin
given q.d. and ciprofloxacin given b.i.d. were simulated. Regardless of the anti-
biotic or bacterial strain, the simulated AUCs and the respective amounts of the
drugs actually administered in the model were chosen to provide similar eight-
fold ranges of the AUC/MIC. These ratios averaged from 54 to 432 (mgzh/ml)/
(mg/ml) for trovafloxacin and from 59 to 473 (mgzh/ml)/(mg/ml) for ciprofloxa-
cin. For regimens of ciprofloxacin given b.i.d., the AUC/MICs presented reflect
the sum of two AUC/MICs provided by the two doses of the quinolone admin-
istered at 12-h intervals but with the residual concentrations at the end of the first
interval taken into account. The simulated time courses of the trovafloxacin and
ciprofloxacin concentrations related to the MIC are presented in Fig. 1.
In vitro dynamic model and operating procedure. A previously described dy-
namic model (11) was used in the study. Briefly, the model consisted of two con-
nected flasks, one containing fresh Mueller-Hinton broth and the other, the
central unit, containing the same type of broth plus a bacterial culture (control
growth experiments) or a bacterial culture plus antibiotic (killing and regrowth
experiments). The central unit was incubated at 37°C in a shaking water bath.
Peristaltic pumps (Minipuls 2; Gilson) circulated fresh nutrient medium to the
bacterium- or bacterium- and antibiotic-containing medium from the central 40-
ml unit at a flow rate of 3 or 7 ml/h to simulate trovafloxacin or ciprofloxacin
pharmacokinetics, respectively. Hence, the clearances provided by the designed
flow rates plus the volume of the central unit ensure the monoexponential elim-
ination of both trovafloxacin or ciprofloxacin and bacteria from the system, with
elimination rate constants of 0.075 h
21
(t
1/2
59.25 h) and 0.170 h
21
(t
1/2
54.0 h),
respectively. Accurate simulations of the desired pharmacokinetic profiles are pro-
vided by maintaining constant flow rates and a constant volume in the central unit.
The system is filled with sterile Mueller-Hinton broth and is placed in a
temperature-regulated incubator at 37°C. The central unit was inoculated with
18-h cultures of E. coli,P. aeruginosa,orK. pneumoniae, and after a further 2 h
of incubation, trovafloxacin or ciprofloxacin was injected into the central unit.
The resulting counts of the organisms in the exponentially growing cultures
approached approximately 10
6
CFU/ml. The duration of the experiments was
defined in each case as the time until the antibiotic-exposed bacteria (N
A
)
reached the maximum numbers observed in the absence of antibiotic (control
growth [N
C
]), i.e., the time when N
A
becomes equal to N
C
. In all cases the
experiments were stopped when N
A
reached $10
11
CFU/ml. Since the experi-
ments that simulated low AUC/MIC ratios met this requirement earlier than
those that simulated high AUC/MIC ratios, the duration of the former experi-
ments was shorter than that of the latter experiments. As illustrated in Fig. 1, the
lower the AUC/MIC ratio, the shorter the requisite observation period.
Validation of the model. To validate the dynamic model, trovafloxacin or
ciprofloxacin concentrations in samples of Mueller-Hinton broth withdrawn
from the central unit were determined in duplicate by high-performance liquid
chromatography (HPLC). To precipitate the proteins from the broth, 200 mlof
acetonitrile (in the presence of trovafloxacin) or methanol (with ciprofloxacin)
was added to a 100-ml sample. The mixture was vortexed and centrifuged at
2,000 3gfor 10 min. A total of 25 ml of the upper aqueous layer was injected into
the HPLC system. Chromatography was carried out on a reversed-phase column
(Silasorb C
8
;5mm; 250 by 4.6 mm [internal diameter]). The mobile phase con-
sisted of acetonitrile and 0.02 M KH
2
PO
4
solution (30:70 [vol/vol]) for trova-
floxacin and an acetonitrile, ethyl alcohol, and 0.02 M KH
2
PO
4
solution (10:20:70
[vol/vol]) for ciprofloxacin; the mobile phase was delivered through a Waters chro-
matographic pump (model 501) at flow rates of 1.7 and 1.3 ml/min, respectively.
Trovafloxacin was detected with a Waters Lambda-Max model 481 absorbance
detector at 275 nm, and ciprofloxacin was detected with a Waters model 420-AC
fluorescence detector; the excitation wavelength was set at 274 nm, and the
emission wavelength was set at 418 nm. The detection limit was 0.1 mg/ml for
trovafloxacin and 0.05 mg/ml for ciprofloxacin, and the linearity ranged from 0.25
to 10 and from 0.1 to 4 mg/ml, respectively. The interday coefficient of variation
was close to 8% for fluoroquinolone concentrations of both 2 and 0.5 mg/ml. The
trovafloxacin and ciprofloxacin concentrations in the central compartment of the
model determined by the HPLC method were close to the designed values, with
no systematic deviation from the expected values (Fig. 2). Hence, the model
provided reasonably accurate simulations of the pharmacokinetic profiles.
FIG. 1. In vitro simulated pharmacokinetic profiles of trovafloxacin (curves
labeled 1) and ciprofloxacin given q.d. (curves labeled 2) and b.i.d. (curves
labeled 3). The numbers in the upper right corner of each plot are the average
values of the simulated AUC/MICs [in (mgzh/ml)/(mg/ml)] of trovafloxacin/
simulated AUC/MICs of ciprofloxacin.
FIG. 2. Observed (open symbols) and designed (lines) concentrations of tro-
vafloxacin and ciprofloxacin in the central compartment of the dynamic model
when simulating the same initial concentrations (2.35 mg/ml) but different t
1/2
(9.25 and 4.0 h, respectively).
660 FIRSOV ET AL. ANTIMICROB.AGENTS CHEMOTHER.

Quantitation of bacterial growth and killing. In each experiment 0.1-ml sam-
ples were withdrawn from the bacterium-containing medium in the central unit
throughout the observation period, at first every 30 min, later hourly, then every
3 h, and during the last 6 to 7 h, again hourly. These samples were subjected to
serial 10-fold dilutions with chilled, sterile 0.9% NaCl and were plated in dupli-
cate onto Mueller-Hinton agar. Antibiotic carryover at low counts was avoided
by washing the bacteria with 0.9% NaCl. After overnight incubation at 37°C the
resulting bacterial colonies were counted, and the numbers of CFU per milliliter
were calculated. The limit of detection was 2 310
2
CFU/ml.
Preliminary experiments performed in duplicate showed good within-day and
day-by-day reproducibilities of the results. The respective pairs of representative
time-kill curves observed in repeated experiments with E. coli exposed to trova-
floxacin are shown in Fig. 3. As seen in Fig. 3, the data obtained from each of the
paired runs were virtually superimposed.
To reveal possible changes in susceptibility, the quinolone concentrations
corresponding to the time when the numbers of surviving organisms in the
regrowth curves reached the level of the initial inoculum (C
regrowth
) were deter-
mined for each run (10). No AUC/MIC-induced systematic differences in the
C
regrowth
s were documented for any of the regimens; moreover, the appearance
of bacterial regrowth was associated with values of unity for ratios of quinolone
concentrations to MICs.
Quantitative evaluation of the antimicrobial effect and comparison of its
predictors. The antimicrobial effect was expressed by the intensity I
E
, which
describes the area between the control growth and bacterial killing and regrowth
curves from time zero (the moment of drug input into the model) to the time
when viable counts on the regrowth curve are close to the maximum values
observed without drug (9). The upper limit of bacterial numbers, i.e., the cutoff
level in the regrowth and control growth curves, used to determine the I
E
was
10
11
CFU/ml (Fig. 4).
To compare the predictive potentials of AUC/MIC, AUC
eff
, and T
eff
, the
antimicrobial effects expressed by I
E
were related to each predictor for each of
the treatment regimens, i.e., trovafloxacin given q.d. and ciprofloxacin given q.d.
and b.i.d. Correlation and regression analyses of the relationships between I
E
and log AUC/MIC, log AUC
eff
,orT
eff
were performed by using STATISTICA
software (version 4.3; StatSoft, Inc.). Statistical comparison of the regressions
was performed at P,0.05, as described elsewhere (22).
RESULTS
The time courses of viable counts that reflect killing and
regrowth of E. coli,P. aeruginosa, and K. pneumoniae exposed
to monoexponentially decreasing concentrations of trovafloxa-
cin given q.d. and ciprofloxacin given q.d. and b.i.d. and the
respective control growth curves are presented in Fig. 5 to 7.
As seen in Fig. 5 to 7, at all the AUC/MIC ratios studied,
regrowth occurred following a considerable reduction in bac-
terial numbers. Unlike the rate of killing or the minimum
bacterial numbers achieved, the time shift of the regrowth
phase to the right along the time axis was distinctly dependent
on the simulated AUC/MIC: the higher the AUC/MIC, the
later the regrowth. Furthermore, at every simulated AUC/MIC
the time-kill curves observed for each of the quinolones and
regimens were similar for the different bacteria, whereas quin-
olone-induced and regimen-induced (q.d. versus b.i.d. for cip-
rofloxacin) differences in the appearance of bacterial regrowth
were evident. For all three bacterial species exposed to trova-
floxacin, at each AUC/MIC, regrowth was observed later than
that with the regimens of ciprofloxacin given b.i.d. and espe-
cially ciprofloxacin given q.d. Since no substantial species-de-
pendent effects were established, subsequent comparison of
the predictors allows the data obtained with different micro-
organisms in each of the experiments with a given quinolone or
regimen to be combined.
The plots of I
E
versus log AUC/MIC, log AUC
eff
, and T
eff
are presented in Fig. 8. As seen in Fig. 8, a specific linear
relationship between I
E
and log AUC/MIC is inherent for each
of the three treatments. Moreover, the correlation coefficients
have similar high values (r
2
50.95 to 0.98), although the slopes
of the I
E
-log AUC/MIC plots differed (Table 1). The slope for
trovafloxacin [276 log (CFU/ml) zh] is 1.8-fold higher than that
for ciprofloxacin given b.i.d. [151 log (CFU/ml) zh], which is in
turn 2.4-fold higher than that for ciprofloxacin given q.d. [113
log (CFU/ml) zh] (P,0.05).
Similar considerations apply to a comparison of the relation-
ships between I
E
and AUC
eff
. As seen in Fig. 8, the I
E
-log
AUC
eff
plots obtained for trovafloxacin and ciprofloxacin given
q.d. and b.i.d. differed. The slopes of the trovafloxacin plot [250
log (CFU/ml) zh] are 1.8- and 2.3-fold higher than those for
the plots of ciprofloxacin given b.i.d. and q.d. [141 and 108 log
(CFU/ml) zh, respectively], (P,0.05). Again, these three plots
cannot be superimposed, although I
E
is highly correlated with
log AUC
eff
for each treatment (r
2
50.96, 0.96, and 0.93,
respectively; Table 1).
FIG. 3. Time courses of killing and regrowth of E. coli exposed to trovafloxacin observed in parallel runs performed on the same day [AUC/MIC 543 (mgzh/ml)/
(mg/ml) (A)] and on different days [AUC/MIC 592 (mgzh/ml)/(mg/ml) (B)]. The data obtained in the respective paired experiments are indicated by different symbols.
FIG. 4. Schematic presentation of the I
E
determination applied to the kinet-
ics of killing and regrowth of K. pneumoniae when mimicking twice-daily cipro-
floxacin administration [AUC/MIC 559 (mgzh/ml)/(mg/ml)]. I
E
describes the
dashed area between the control growth (empty triangles) and the killing and
regrowth (filled triangles) curves at a cutoff level of 10
11
CFU/ml.
VOL. 42, 1998 PREDICTORS OF QUINOLONE ANTIMICROBIAL EFFECTS 661

The plots of I
E
versus T
eff
for trovafloxacin and for both
ciprofloxacin regimens are linear with the single exception of
the regimen of ciprofloxacin given b.i.d., in which the points
corresponding to the lowest T
eff
departed from the straight line
that fits the points corresponding to the three higher values of
T
eff
. For this reason, only the linear portion of the I
E
-versus-
T
eff
plots for ciprofloxacin given b.i.d. was used for further
analysis. The I
E
-T
eff
plots for the two quinolones and the two
ciprofloxacin dosing regimens are practically superimposed
and the respective slopes are similar, with no statistically sig-
nificant differences (Table 1). As seen in Fig. 8, the I
E
-T
eff
sets
combined for all three treatments showed a very good corre-
lation between the effect and its predictor (r
2
50.95) that is
comparable to the respective correlations found for the regi-
mens of trovafloxacin and ciprofloxacin given q.d. and b.i.d.
taken separately (r
2
50.96, 0.98, and 0.94, respectively; Ta-
ble 1).
DISCUSSION
This study suggests that the antimicrobial effect of an indi-
vidual quinolone (trovafloxacin or ciprofloxacin) or a dosing
FIG. 5. The kinetics of killing and regrowth of gram-negative bacteria when mimicking trovafloxacin administration q.d. for E. coli (Œ), P. aeruginosa (■), and
K. pneumoniae (F) with (filled symbols) and without (empty symbols) quinolones. The numbers in the bottom right corner of each plot are the simulated AUC/MICs
[in (mgzh/ml)/(mg/ml)].
FIG. 6. The kinetics of killing and regrowth of gram-negative bacteria when mimicking ciprofloxacin administration q.d. for E. coli (Œ), P. aeruginosa (■), and
K. pneumoniae (F) with (filled symbols) and without (empty symbols) quinolones. The numbers in the bottom right corner of each plot are the simulated AUC/MICs
[in (mgzh/ml)/(mg/ml)].
662 FIRSOV ET AL. ANTIMICROB.AGENTS CHEMOTHER.

regimen (ciprofloxacin given q.d. or b.i.d.) as expressed by its
intensity (I
E
) correlates equally well with each of the three
predictors (AUC/MIC, AUC
eff
, and T
eff
). Thus, no preferences
for any of them may be supported by these data. This is con-
sistent with the reported similar predictive potentials of AUC/
MIC and T
eff
applied to a single quinolone (6, 21) and is quite
expected, since each of the three predictors examined in our
study covaried strongly (r
2
.0.98 for each treatment). Similar
covariance was reported previously in studies with enoxacin (2)
and lomefloxacin (6).
Although all three predictors could be accurately related to
the effect in a similar bacterial species-independent fashion,
the log AUC/MIC-response and log AUC
eff
-response rela-
tionships were specific for each drug (trovafloxacin and cip-
rofloxacin) and for each of the ciprofloxacin dosing regi-
mens. This circumstance allows quantitative comparisons of
the effects of the quinolones, as reported recently (7, 12).
Unlike AUC/MIC and AUC
eff
, the I
E
-T
eff
plots obtained with
trovafloxacin given q.d. and ciprofloxacin given q.d. and b.i.d.
were virtually superimposed and therefore are not specific.
This means that the effect of one quinolone may be predicted
by the I
E
-T
eff
relationship established for another quinolone.
Thus, unlike AUC/MIC and AUC
eff
, which may be referred to
as intraquinolone and intraregimen predictors only, T
eff
may
also be considered the best interquinolone and interregimen
predictor of the antimicrobial effect. However, the I
E
-T
eff
re-
lationships do not reveal obvious differences between the quin-
olones and/or dosing regimens of ciprofloxacin, whereas the
I
E
-AUC/MIC and I
E
-AUC
eff
relationships do reveal such dif-
ferences.
As already mentioned, a specific I
E
-log AUC/MIC or I
E
-log
AUC
eff
relationship was inherent for each of the treatments.
Since these plots were not superimposed and the data could
not be considered a homogeneous set, combining them would
FIG. 7. The kinetics of killing and regrowth of gram-negative bacteria when mimicking ciprofloxacin administration b.i.d. for E. coli (Œ), P. aeruginosa (■), and
K. pneumoniae (F) with (filled symbols) and without (empty symbols) quinolones. The numbers in the bottom right corner of each plot are the simulated AUC/MICs
[in (mgzh/ml)/(mg/ml)].
FIG. 8. Antimicrobial effects of trovafloxacin given q.d. (——) and ciprofloxacin given b.i.d. (– – –) and q.d. (....)related to the different predictors. The three
points that departed from linear I
E
-T
eff
plot are enclosed in a circle (for more detailed discussion, see the text). , all treatments.
VOL. 42, 1998 PREDICTORS OF QUINOLONE ANTIMICROBIAL EFFECTS 663

be incorrect. For example, if the I
E
-AUC/MIC data from the
regimens of trovafloxacin given q.d. and ciprofloxacin given
q.d. and b.i.d. were combined, only a loose correlation between
the effect and its predictor would be established (r
2
50.46).
Therefore, neither AUC/MIC nor AUC
eff
may be considered
quinolone- or regimen-independent predictors of the antimi-
crobial effect produced by trovafloxacin and ciprofloxacin.
This conclusion is consistent with the lack of predictability of
the effects of different quinolones taken together by using
AUIC (25) or of the effects of different antibiotics including
ciprofloxacin and fleroxacin by using AUC/MIC (3). At the
same time, our data do not support recent reports of successful
prediction of the effects of two different quinolones by AUIC
(14) or the statement that AUC/MIC or AUC/MIC
24
were
general predictors of antimicrobial effects of the fluoroquino-
lones (15). At least in part, these contradictions are probably
less than they appear, because neither AUIC (14) nor AUC/
MIC (15) was compared to alternative predictors such
as AUC
eff
and T
eff
. Moreover, this statement was based on a
scattered AUC/MIC
24
-response curve (r
2
50.58) obtained
with combined data for ciprofloxacin and ofloxacin as well as
different dosing regimens. Perhaps these data (15) should be
converted into a family of more accurate plots for each drug
and regimen taken separately.
The comparative single-dose study of Wiedemann et al. (25),
which has been performed with four gram-positive and gram-
negative bacteria exposed to biexponentially decreasing con-
centrations of ciprofloxacin, sparfloxacin, fleroxacin, and oflox-
acin, indirectly supports this hypothesis. Those investigators
stated “no clear-cut relationship between AUIC and killing
activity” was found when data for four different quinolones
were combined. However, the differences between the loga-
rithm of the initial inoculum and the logarithm of the mini-
mum bacterial numbers achieved (Dlog N
min
) could be related
to AUIC if the data for the drugs were considered sepa-
rately. For example, Dlog N
min
correlates with the AUIC of
fleroxacin and ofloxacin taken separately (Fig. 9), although
the log AUIC-response plots associated with the individual
quinolones are quite different and these data do not belong to
the same homogeneous set. Thus, the conclusion that similar
effects are produced by the same AUC/MICs of different fluo-
roquinolones (15) was not confirmed by our current findings or
by the results reported by Wiedemann et al. (25). This analysis
also suggests that before data sets obtained with pharmacoki-
netically different drugs and different dosing regimens may be
combined, it is necessary to test whether these sets are homo-
geneous.
One more reason for apparently conflicting results from
studies of different predictors of the antimicrobial effect is the
use of different parameters to quantitate the effect. The pa-
rameters ln 1/AUEC (AUEC may be referred to as the anti-
logarithm of the area under the bacterial count-time curve
[AUBC] [23]), the area above the time-kill curve (AAC [24]),
and Dlog N
min
used in the previously cited studies (15, 25) were
shown to be insufficiently sensitive to the AUC/MIC of cipro-
floxacin (11). Moreover, the use of AUBC and AAC might
result in degenerative AUC-response relationships. These rea-
sons also might contribute to the scattered plot relating ln
1/AUEC and AUC/MIC referred to above (15) and to the
uncertain relationships between AAC and AUIC which were
reconstructed for individual quinolones with data from Wiede-
mann et al. (25) (data not shown). In the present study an
integral parameter of the antimicrobial effect, its intensity (I
E
),
was related to AUC/MIC, AUC
eff
, and T
eff
, and logical rela-
tionships between each of them and I
E
were established. By its
very definition, I
E
includes the evaluation of full-term killing
and regrowth curves from the onset to the end of the antimi-
crobial effect (9). The impact of recording the entire regrowth
phase on the evaluation of the antimicrobial effects of quino-
lones has been reported recently (11).
This study, performed with bacterial strains with similar sus-
ceptibilities, allowed the selection of the best interquinolone
predictor of the antimicrobial effect (T
eff
), but it did not sup-
port the choice of the best intraquinolone predictor among
AUC/MIC, AUC
eff
, and T
eff
. Recently, such a selection was
proven to be possible on the basis of data obtained with one
quinolone (ciprofloxacin) for organisms with different suscep-
tibilities (21). A specific I
E
-log AUC
eff
plot was inherent for
each of four strains of gram-negative and gram-positive bacte-
ria (MICs, 0.013 to 0.60 mg/ml), whereas the respective I
E
-log
AUC/MIC and I
E
-T
eff
relationships were bacterial species in-
dependent. From this point of view, AUC/MIC and T
eff
may be
preferred to AUC
eff
as intraquinolone predictors.
Overall, these and other recently published (21) data suggest
that the optimal interquinolone predictors of the antimicrobial
effect may be selected from studies with pharmacokinetically
different drugs, and intraquinolone predictors might be se-
lected from studies with bacteria with different susceptibilities.
The concept of inter- and intraquinolone predictors might be
useful for the in vitro evaluation of future quinolone com-
pounds. Additional studies with other fluoroquinolones are
needed to further support this concept.
FIG. 9. AUIC-dependent antimicrobial effects of two fluoroquinolones against
different bacterial strains as expressed by Dlog N
min
. The figure is reconstructed
from Wiedemann et al. (25).
TABLE 1. Slopes of the regressions of I
E
on the
predictors and the correlation coefficients
Antibiotic and
regimen
AUC/MIC AUC
eff
T
eff
Slope
[log (CFU/
ml) zh] r
2
Slope
[log (CFU/
ml) zh] r
2
Slope
(log CFU/ml) r
2
Trovafloxacin q.d. 276
a
0.97 250
a
0.96 9.0 0.96
Ciprofloxacin q.d. 151
b
0.95 141
b
0.93 9.7 0.94
Ciprofloxacin b.i.d. 113 0.98 108 0.96 8.6 0.98
a
Statistically significant difference between the drugs.
b
Statistically significant difference between the regimens of ciprofloxacin.
664 FIRSOV ET AL. ANTIMICROB.AGENTS CHEMOTHER.

ACKNOWLEDGMENTS
This study was supported by Roerig, a division of Pfizer.
We are thankful to Yury Portnoy for assistance in computer analysis
and graphic presentation of the data.
REFERENCES
1. Bergan, T., and S. B. Thorsteinsson. 1986. Pharmacokinetics and bioavail-
ability of ciprofloxacin, p. 111–121. In H. C. Neu and H. Weuta (ed.).
Proceedings of the 1st International Ciprofloxacin Workshop, Leverkusen,
Germany, 6 to 8 November 1985. Current clinical practice series 34. Elsevier
Science Publishers B. V. (Excerpta Medica), Amsterdam, The Netherlands.
2. Blaser, J., B. B. Stone, M. C. Groner, and S. H. Zinner. 1987. Comparative
study with enoxacin and netilmicin in a pharmacodynamic model to deter-
mine importance of ratio of antibiotic peak concentration to MIC for bac-
tericidal activity and emergence of resistance. Antimicrob. Agents Chemo-
ther. 31:1054–1060.
3. Blaser, J., P. Verge`res, A. F. Widmer, and W. Zimmerli. 1995. In vivo
verification of in vitro model of antibiotic treatment of device-related infec-
tion. Antimicrob. Agents Chemother. 39:1134–1139.
4. Craig, W. A. 1995. Interrelationship between pharmacokinetics and pharma-
codynamics in determining dosage regimens for broad-spectrum cephalo-
sporins. Diagn. Microbiol. Infect. Dis. 22:89–96.
5. Drusano, G. L. 1991. Human pharmacodynamics of beta-lactams, aminogly-
cosides and their combinations. Scand. J. Infect. Dis. 74(Suppl.):235–248.
6. Drusano, G. L., D. E. Johnson, M. Rosen, and H. C. Standiford. 1993.
Pharmacodynamics of a fluoroquinolone antimicrobial agent in a neutro-
penic rat model of Pseudomonas sepsis. Antimicrob. Agents Chemother.
37:483–490.
7. Firsov, A., S. Vostrov, A. Shevchenko, and S. Zinner. 1997. Trovafloxacin vs.
ciprofloxacin against bacteria of similar susceptibility to both drugs: a study
with Pseudomonas aeruginosa in an in vitro dynamic model. Clin. Microbiol.
Infect. 3(Suppl. 2):85. (Abstract P382.)
8. Firsov, A. A. 1993. Critical reappraisal of modern approaches to search
determinants of efficacy of antimicrobials. Eur. Bull. Drug Res. 2(Suppl. 1):
33–38.
9. Firsov, A. A., V. M. Chernykh, and S. M. Navashin. 1991. Quantitative
analysis of antimicrobial effect kinetics in an in vitro dynamic model. Anti-
microb. Agents Chemother. 34:1312–1317.
10. Firsov, A. A., A. D. Nazarov, and V. M. Chernykh. 1989. Pharmacokinetic
approaches to rational antibiotic therapy. In Series: Advances in Science and
Engineering, vol. 17. VINITI Publishers, Moscow, Russia. (In Russian.)
11. Firsov, A. A., S. N. Vostrov, A. A. Shevchenko, and G. Cornaglia. 1997.
Parameters of bacterial killing and regrowth kinetics and antimicrobial effect
examined in terms of area under the concentration-time curve relationships:
action of ciprofloxacin against Escherichia coli in an in vitro dynamic model.
Antimicrob. Agents Chemother. 41:1281–1287.
12. Firsov, A. A., S. N. Vostrov, A. A. Shevchenko, and S. H. Zinner. 1997.
Trovafloxain vs. ciprofloxacin against Klebsiella pneumoniae in an in vitro
dynamic model, abstr. 2205. In Final Programme and Book of Abstracts of
the 20th International Congress of Chemotherapy.
13. Hoffken, G., H. Lode, C. Prinzing, K. Borner, and P. Koeppe. 1985. Phar-
macokinetics of ciprofloxacin after oral and parenteral administration. Anti-
microb. Agents Chemother. 27:375–379.
14. Madaras-Kelly, K. J., A. J. Larsson, and J. C. Rotschafer. 1996. A pharma-
codynamic evaluation of ciprofloxacin and ofloxacin against two strains of
Pseudomonas aeruginosa. J. Antimicrob. Chemother. 37:703–710.
15. Madaras-Kelly, K. J., B. E. Ostergaard, L. Baeker Hovde, and J. C. Rotscha-
fer. 1996. Twenty-four-hour area under the concentration-time curve/MIC
ratio as a generic predictor of fluoroquinolone antimicrobial effect by using
three strains of Pseudomonas aeruginosa and an in vitro pharmacodynamic
model. Antimicrob. Agents Chemother. 40:627–632.
16. Mattie, H., W. A. Craig, and J. C. Peche`re. 1989. Determinants of efficacy
and toxicity of aminoglycosides. J. Antimicrob. Chemother. 24:281–293.
17. Schentag, J. J. 1991. Correlation of pharmacokinetic parameters to efficacy
of antibiotics: relationships between serum concentrations, MIC values, and
bacterial eradication in patients with gram-negative pneumonia. Scand. J. In-
fect. Dis. 74(Suppl.):218–234.
18. Schentag, J. J., D. E. Nix, and M. H. Adelman. 1991. Mathematical exami-
nation of dual individualization principles. Relationships between AUC
above MIC and area under the inhibitory curve (AUC/MIC) for cefmenox-
ime, ciprofloxacin, and tobramycin. DICP Ann. Pharmacother. 25:1050–1057.
19. Teng, R., S. C. Harris, D. E. Nix, J. J. Schentag, G. Foulds, and T. E. Liston.
1995. Pharmacokinetics and safety of trovafloxacin (CP-99,219), a new quin-
olone antibiotic, following administration of single oral doses to healthy male
volunteers. J. Antimicrob. Chemother. 36:385–394.
20. Vogelman, B., S. Gudmundsson, J. Leggett, J. Turnidge, S. Ebert, and W. A.
Craig. 1988. Correlation of antimicrobial pharmacokinetic parameters with
therapeutic efficacy in an animal model. J. Infect. Dis. 158:831–847.
21. Vostrov, S., A. Shevchenko, A. Firsov, and S. Zinner. 1997. MIC-based
interspecies prediction of the antimicrobial effect: ciprofloxacin against dif-
ferent bacteria in an in vitro dynamic model. Clin. Microbiol. Infect. 3(Suppl.
2):84. (Abstract P375.)
22. Weber, E. 1964. Grundriss der biologischen Statistik, 5e auflage, p. 247–248.
VEB Gustav Fischer Verlag, Jena, Germany.
23. White, C. A., and R. G. Toothaker. 1985. Influence of ampicillin elimination
half-life on in-vitro bactericidal effect. J. Antimicrob. Chemother. 15(Suppl.
A):257–260.
24. Wiedemann, B., and A. Jansen. 1990. Antibacterial activity of cefpodoxime
proxetil in a pharmacokinetic in-vitro model. J. Antimicrob. Chemother. 26:
71–79.
25. Wiedemann, B., C. Rustige-Wiedemann, and B. Kratz. 1995. Comparison of
the pharmacodynamic properties of quinolones. Drugs 49(Suppl. 2):269–271.
26. Wise, R., D. Lister, C. A. M. McNulty, D. Griggs, and J. M. Andrews. 1986.
The comparative pharmacokinetics of five quinolones. J. Antimicrob. Che-
mother. 18(Suppl. D):71–81.
27. Wise, R., D. Mortiboy, G. Child, and G. M. Andrews. 1996. Pharmacokinetics
and penetration into inflammatory fluid of trovafloxacin (CP-99,219). Anti-
microb. Agents Chemother. 40:47–49.
VOL. 42, 1998 PREDICTORS OF QUINOLONE ANTIMICROBIAL EFFECTS 665