Identification of Factors Affecting In Vivo Aminoglycoside Activityin an Experimental Model of Gram-Negative Endocarditis
GILLES POTEL,* JOCELYNE CAILLON, FLORENCE LE GALLOU, DENIS BUGNON, PHILIPPE LE CONTE,JOCELYNE RAZA, JEAN-YVES LEPAGE, DENIS BARON, AND HENRI DRUGEON
Laboratoire dAntibiologie Clinique et Expehimentale, Facult de Mtdecine,1 rue Gaston-Veil, 44035 Nantes, France
Received 22 July 1991/Accepted 30 January 1992
Aminoglycoside bactericidal activity during the first 24 h of treatment probably is a determining parameterin the prognosis of severe gram-negative infections in immunocompromised patients. To identify the predictivefactors involved in the definition of the best therapeutic regimen for Enterobacter cloacae and Serratiamarcescens infections, we studied different gentamicin, tobramycin, and amikacin regimens by using an
experimental model of rabbit endocarditis. Two factors appear to play an important role in predicting in vivoefficacy: (i) the level of in vivo bactericidal activity, which can differ widely from one aminoglycoside to anotherfor the same bacterial strain and from one strain to another of the same species, and (ii) the critical serum drugconcentration (CSC, in milligrams per liter), defined as the lowest serum antibiotic concentration capable ofproducing a significant CFU reduction (P < 0.05) in endocarditis vegetations 24 h after the beginning of a
continuous infusion. Stepwise regression analysis showed that for gentamicin and S. marcescens, the area underthe concentration-time curve above the CSC and then the time above the CSC are the determining parametersfor efficacy (R = 0.69;F = 13.5; P = 0.001), whereas for amikacin and S. marcescens, the time above the CSCand then the area under the concentration-time curve above the CSC predict efficacy (R = 0.74; F = 24.0; P= 0.0001). The lowest CSC is that ofamikacin (about 8 mg/liter); those of gentamicin and tobramycin are about15 mg/liter. In severe S. marcescens infections, intermittent amikacin dosing offers excellent bactericidalactivity within the first 24 h.
It has been seen in numerous clinical and experimentalstudies that once-daily dosing of an aminoglycoside is atleast as effective as administration of the same quantity ofdrug as divided doses (1, 11, 13). However, most of thesestudies examined only a limited number of bacterial species(Escherichia coli, Kebsiellapneumoniae, and Pseudomonasaeruginosa), and all compared the effects of several treat-ment regimens (including once-daily dosing) with the sameaminoglycoside (4, 13). Among gram-negative bacteria, En-terobacter and Serratia species have assumed increasingimportance in the last few years (17, 26). In one Frenchseries, Enterobacter and Serratia species accounted for 21 of134 (15.6%) cases of gram-negative septicemia recorded in1989 (2).
Investigation of the in vivo bactericidal activities of anti-biotics in the first 24 h after administration is probably animportant measure, with potentially crucial implications forhuman use, principally in neutropenic (or aplastic) patients.
In a recent experimental study (22), we showed that thesame doses of gentamicin, tobramycin, and amikacin haddifferent effects in a model of Serratia marcescens en-docarditis, despite similar pharmacokinetics. In brief, thesame single dose administered as an intravenous (i.v.) bolusinjection exhibited the best efficacy for gentamicin, whereasthe same dose administered as a continuous i.v. infusionexhibited the best efficacy for amikacin. Furthermore, one invitro study (12) has suggested that higher initial peak con-centrations of aminoglycosides are necessary to achievebactericidal activity for S. marcescens and Enterobacter
* Corresponding author.
cloacae, compared with most susceptible species, such as E.coli. Thus, the recommended dosage regimens for aminogly-cosides are not necessarily identical from one species toanother.
Finally, although many authors (3, 4, 10, 18, 25) havedemonstrated a good correlation between some pharmaco-kinetic parameters and in vivo bactericidal activities ofaminoglycosides, these studies usually used an in vitroparameter (usually the MIC) to establish the correlation.
In an attempt to identify, at least in part, the factorsaffecting in vivo aminoglycoside activity, we chose to studythe in vivo effects of gentamicin, tobramycin, and amikacinat various dosage regimens on both E. cloacae and S.marcescens experimental endocarditis. We tried to confirmthe importance of the in vitro killing rate in the efficacy of asingle i.v. bolus dose, as previously shown in the samemodel of E. coli endocarditis (23). The second parameterwhich was investigated was the critical serum drug concen-tration (CSC, in milligrams per liter), defined as the lowestserum antibiotic concentration able to achieve significant invivo bactericidal activity 24 h after the beginning of acontinuous i.v. infusion in the S. marcescens endocarditismodel. Thus, we expected to explain some of the differencesin the results reported in the literature regarding the opti-mum doses of aminoglycosides (11, 15) by studying the earlybactericidal activities of three aminoglycosides in experi-mental gram-negative endocarditis, representing a model ofsevere intravascular infection without host defenses at thesite of infection.
(This work was presented in part at the 28th InterscienceConference on Antimicrobial Agents and Chemotherapy,Los Angeles, Calif., 23 to 26 October 1988 [21a].)
Organisms. One strain of S. marcescens (HN229) and twostrains of E. cloacae (E474 and E475) were cultured from theurine (HN229), blood (E474), or lungs (E475) of hospitalizedpatients. All strains were resistant to rabbit serum.
Antibiotics. The three aminoglycosides studied were to-bramycin (Eli-Lilly), gentamicin (Schering-Plough), and ami-kacin (Bristol-Myers-Squibb).
In vitro studies. (i) Antibiotic susceptibility tests. The MICof each antibiotic was determined with Mueller-Hinton brothsupplemented with Ca2+ (50 ,ug/ml) and M 2+ (25 ,ug/ml) in200-,ul wells and with an inoculum of 10 CFU/ml in theexponential growth phase. The MIC was defined as thelowest concentration of antibiotic capable of inhibiting allvisible growth after 18 h of incubation. After 24 h, asubculture was transferred to Mueller-Hinton agar (DifcoLaboratories, Detroit, Mich.). The MBC corresponded tothe lowest concentration of drug permitting 0.1% of thebacteria to survive and was determined by placing 1-,uportions of cultures, by use of a Steers replicator, on agarplates with 3% polyanetholesulfonic acid sodium salt (SPS;Sigma) (19).
(ii) Killing curves. Time-kill curves were drawn for eachantibiotic at 12 concentrations in Mueller-Hinton broth:0.06, 0.12, 0.25, 0.5, 1, 2, 4, 8, 16, 32, 64, and 128 mg/liter.For each concentration, the antibiotics were incubated inmicrotubes (1-ml tubes; Macrowell; Skatron, Lier, Norway)with an inoculum of 107 S. marcescens or E. cloacaestationary-phase cells per ml. Surviving bacteria werecounted in each tube after 1.5, 3, 5, and 24 h of incubation bya semiautomatic dilution micromethod involving an auto-matic 96-well dispenser (Skatron) and a Steers replicatordistributing 2 ± 0.5 ,ul of each dilution onto an agar platesupplemented with 3% SPS, avoiding a carryover phenom-enon. After 24 h of incubation, the first dilution with 5 to 30colonies was read, and the colony count was multiplied bythe dilution factor. The standard error of this count was 0.2log1o CFU/ml. The sensitivity limit of detection was 2.4 log1oCFU/ml (7). Expression of the results was described in aprevious work (9). In brief, the number of surviving bacteriawas expressed as log1o CFU per milliliter after each incuba-tion period (1.5, 3, 5, and 24 h). Only the early phase of thekilling curve (i.e., the first S h) was plotted on a graph. Foreach concentration of antibiotic, the area under the curve forthe surviving bacteria between 0 and 5 h (AUC of survivingbacteria) was calculated, and a percentage of the hypothet-ical area for the reference inoculum if no bactericidal activityhad occurred was then determined. This parameter wascalled the index of surviving bacteria (ISB) and was calcu-lated as follows: ISB (percent) = (AUC of surviving bacte-ria/AUC of inoculum) x 100. This calculation allowed us torepresent the 5-h killing curves for all concentrations studiedon a two-dimensional graph.
Experimental endocarditis. In vivo studies were carriedout on New Zealand White female rabbits (age, 12 to 15weeks; weight, 2.5 to 3.5 kg). The animals were kept inindividual cages and allowed free access to food and waterthroughout the experiment. Left ventricular endocarditiswas induced as described previously (21, 22). At 24 h afterthe introduction of a polyethylene catheter through theaortic valve, each rabbit received 1 ml of a suspensioncontaining 107 organisms per ml, injected through the mar-ginal ear vein.
(i) Experimental design. (a) In vivo efficacy of a single i.v.bolus dose. In a previous work (23), we showed, by using E.
coli experimental endocarditis, that the in vitro killing ratewas predictive of the in vivo efficacy of a single i.v. dose,i.e., the higher the in vitro killing rate, the lower the singlei.v. dose required to achieve a significant reduction in theCFU in the vegetations 24 h after the single i.v. bolus dose.This dose was the minimal effective dose (MED, in milli-grams per kilogram). Since the two strains of E. cloacae(E474 and E475) differed in terms of in vitro killing ratesfollowing tobramycin and gentamicin, a single 48-mg/kg doseof each of these antibiotics was administered to two groupsof animals with E. cloacae E474 or E475 endocarditis to testhow well the observed difference in in vitro bacterial killingpredicted in vivo activity.
In the S. marcescens endocarditis model, various doses ofgentamicin, tobramycin, and amikacin were tested as bolusinjections to achieve the MED. The doses of amikacin andtobramycin tested were 48 and 72 mg/kg. Since, in the caseof gentamicin, a dose of 48 mg/kg had produced a maximumin vivo antibacterial effect at 24 h (22), animals treated with24 or 48 mg/kg were sacrificed 6 h after the i.v. bolus dose toascertain how early bacterial killing occurred in vivo.
(b) Determination of the CSC. To explain the differencespreviously observed in the model of S. marcescens en-docarditis during continuous 24-h infusions of identicaldoses of tobramycin, amikacin, and gentamicin (22), wedetermined the in vivo CSC of each antibiotic by using thesame model. The doses were chosen in light of previousresults showing that a dose of 48 mg/kg was ineffective fortobramycin and gentamicin, while a maximum effect wasobserved with amikacin. This dose corresponded to asteady-state concentration of about 10 mg/liter. At 48 h afterinoculation, the animals were randomized to one of thefollowing groups, each antibiotic being administered as acontinuous infusion over 24 h: amikacin at 24 mg/kg orgentamicin or tobramycin at 72 mg/kg.
(c) In vivo effect of multiple doses of amikacin on S.marcescens endocarditis. Since the continuous infusion ofamikacin over 24 h appeared more effective in the S.marcescens endocarditis model than the same dose admin-istered as a single bolus injection (22), various split-doseregimens (6 mg/kg every 6 h, 12 mg/kg every 6 h, 12 mg/kgevery 12 h, and 24 mg/kg every 12 h) were tested in someanimals to verify that a 24-h continuous i.v. infusion waspredictive of the efficacy of a split-dose regimen.
(ii) Evaluation of treatment. The effect of each treatmentwas evaluated 24 h after the therapeutic regimen assignedwas begun. The animals were sacrificed with an i.v. bolusinjection of thiopental. The heart was removed, and vegeta-tions were excised and rapidly rinsed in sterile saline. Someof the vegetations were weighed and homogenized in aThomas Teflon pestle tissue homogenizer with 0.5 ml ofsterile saline. Serial dilutions of 0.05 ml of the homogenatewere spread by use of a Spiral System (Interscience) andquantitatively cultured for 24 h at 37°C on Trypticase soyagar plates containing 3% SPS, avoiding a carryover phe-nomenon. Bacterial titers were expressed as log1o CFU pergram of vegetation. We were able to detect quantities assmall as 20 CFU/ml of homogenate, and the value integratedfor the calculation of the mean bacterial titer took intoaccount the weights of vegetations. Part of each vegetationwas frozen prior to the antibiotic assays.
(iii) Pharmacokinetic studies. Blood samples were takenfrom three animals in each group to determine the plasmaconcentration-time curve for each therapeutic regimen. An-imals from the following different bolus groups were studied:24 mg/kg for gentamicin; 6, 12, 24, and 72 mg/kg for
amikacin; and 72 mg/kg for tobramycin. For the continuousinfusion groups, blood samples were taken after 6 h (steadystate) from animals receiving 24 mg of amikacin per kg or 72mg of tobramycin or gentamicin per kg.The blood samples were obtained by placement under
local anesthesia of a catheter in the left femoral artery inrabbits receiving a bolus injection and by acute femoralpuncture in animals receiving a continuous infusion. Thesamples were immediately centrifuged, and the serum wasfrozen until the time of the assays.The various pharmacokinetic parameters (half-life, total
AUC, time or AUC above the CSC, and MIC) were calcu-lated by use of a monocompartmental model with the aid ofour own computer program.
(iv) Antibiotic assays. Serum antibiotic concentrations foreach antibiotic regimen were determined by use of a micro-biological assay with Bacillus subtilis ATCC 6633. The rangeof measurable concentrations with this strain was 0.06 to 1,ug/ml for all three antibiotics.After being weighed and homogenized with 0.3 ml of 0.1 M
phosphate buffer, the vegetations were centrifuged and thesupernatant fluid was sampled for a microbiological assay.The same strain of B. subtilis as that listed above was used.
(v) Statistical evaluation. An analysis of variance (SuperANOVA; Abacus Concepts, Inc.) and Scheffe's S test wereperformed to compare the bacterial titers measured intreated animals versus controls. A normality test (Stat-works; Cricket Software, Inc.) was performed first to estab-lish the feasibility of the statistical procedure.The number of sterile vegetations in each experimental
group was compared with that in the control group by a 2test with Yates' continuity correction. Linear and thenstepwise regression analyses were performed to study therelationship between the various pharmacokinetic parame-ters and in vivo antibacterial effects, expressed as Alog1oCFU per gram of vegetation in experimental groups versuscontrols (Statview; Abacus Concepts, Inc.). The analysisfocused first on the mean results for each group (mean Alog1oCFU per gram of vegetation in experimental groups versuscontrols). Gentamicin and amikacin were subjected to aseparate analysis by examining the relationship between theindividual results (A&log1o CFU per gram of vegetation) andthe pharmacokinetic parameters.
RESULTS
In vitro studies. (i) Antibiotic susceptibility tests. The MICs(and MBCs) for the S. marcescens strain studied were 0.5 ,gof gentamicin per ml and 1 ,ug of tobramycin or amikacin perml. The MICs (and MBCs) for both of the E. cloacae strainsstudied were 0.5 ,ug of gentamicin or tobramycin per ml and1 ,ug of amikacin per ml.
(ii) Killing curves. The 5-h bacterial killing curves obtainedafter exposure of the S. marcescens strain to the threeaminoglycosides are shown in Fig. 1 for each concentrationstudied. Each concentration at which the ISB was greaterthan 100% corresponds to regrowth. At all concentrationsabove 1 ,ug/ml, the percentage of surviving bacteria at 5 h(expressed as the ISB) was consistently lower for gentamicinthan for amikacin or tobramycin.
Figure 2 shows the differences observed with tobramycinand gentamicin for the two strains of E. cloacae, strain E475being killed more rapidly than strain E474 at concentrationsbetween 2 and 32 ,ug/ml. These results represent the mean ofthree consecutive experiments. The differences observedwere far above the standard deviation of the method of
iSB (XAUC)1 40-
120-
100-
80-
60-
40
__e. Tobammycin-.- Gentauicina Amikacin
0 0.12 0.5 2 8 32 128CONCEnITRAIONS (C g/ml)
FIG. 1. In vitro killing curves for S. marcescens and gentamicin,amikacin, and tobramycin after 5 h of incubation. Each pointrepresents the bactericidal effect of 1 of the 12 concentrations tested(0.06 to 128 ,ug/ml). Each result is expressed as the ISB (see thetext).
counting (see Materials and Methods) and appeared highlysignificant.
In each experiment, it was possible to confirm that theMICs for the surviving bacteria were the same as those forthe parental strain with respect to the tested antibiotics.
In vivo studies. (i) In vivo efficacy of a single i.v. bolus dose.For E. cloacae endocarditis, the same bolus dose of genta-micin or tobramycin (48 mg/kg) produced a significant reduc-tion in the CFU of strain E475 but not strain E474 (Table 1).For S. marcescens endocarditis, we showed in an earlier
study (22) that the same bolus dose of an aminoglycoside (48mg/kg) had different effects, depending on the compoundused, gentamicin producing the maximum effect (2.7 ± 0.2
AISB (ZAUC)
140120 E 474100 - E 47580-
60 -
40
20-
0.06 0.2S 1 4 16 64COCENTRATIONS (W i)
BISB (ZAMC)
140-
120-
100-
80-
60-40-
20-
.w E 474-_- E 475
0 0.06 0.2s 1 4 16 64CONCENTRATIONS (gmi)
FIG. 2. In vitro killing curves for gentamicin (A) and tobramycin(B) and E. cloacae E474 and E475 after 5 h of incubation. Each pointrepresents the ISB (see the text and the legend to Fig. 1).
TABLE 1. In vivo effect of the same dose of tobramycin orgentamicin (48 mg/kg) 24 h after i.v. bolus administration
on E. cloacae E474 or E475 endocarditis
Mean ± SD vegetation No. of sterile vege-Treatment Strain titer (loglo CFU/g of tations/total no. ofgroup vegetation) vegetations
Control E474 8.5 + 0.7 0/7E475 8.9 ± 0.5 0/8
Gentamicin E474 7.8 ± 0.8 0/7E475 6.1 ± 1.4a 1/9
Tobramycin E474 7.3 + 1.6 0/10E475 5.6 + 2.0" 0/9
aP = 0.008 versus control E475 (Sheffe's S test).h P = 0.001 versus control E475 (Sheffe's S test).
loglo CFU/g of vegetation; 6 of 9 sterile vegetations) andtobramycin producing practically no effect (6.8 + 1.9 log1oCFU/g of vegetation; 0 of 10 sterile vegetations), like ami-kacin (7.5 + 1.3 log1o CFU/g of vegetation; 0 of 10 sterilevegetations). In the present study, we were able to demon-strate that an antibacterial effect could be achieved in vivowith a higher dose (72 mg/kg) of tobramycin or amikacin: 4.3± 1.7 log1o CFU/g of vegetation (P = 0.056 versus controls)for tobramycin and 4.4 ± 2.1 log1o CFU/g of vegetation (P =0.03 versus controls) for amikacin; this result revealed adose-related killing effect of these two drugs on the S.marcescens strain.
Since the effect of 48 mg of gentamicin per kg was virtuallymaximal after 24 h, a number of animals were sacrificed 6 hafter the same dose was administered as a bolus injection toevaluate in vivo bacterial killing by this antibiotic. Bacteri-cidal activity was already very marked 6 h after a dose of 48mg/kg (3.5 ± 1.9 log1o CFU/g of vegetation; P = 0.003),whereas it was not significant after a dose of 24 mg/kg (5.1 ±2.5 log1o CFU/g of vegetation).
(ii) Determination of the CSC for the S. marcescens strain.The results obtained for determination of the CSC for the S.marcescens strain at the various concentrations studied areshown in Table 2. The lowest CSC was that of amikacin, atabout 8 mg/liter. The lowest concentration of this antibioticstudied (5.3 mg/liter) produced a nonsignificant (P = 0.11versus controls) antibacterial effect (4.8 ± 2.2 log1o CFU/g of
TABLE 2. In vivo effect of various steady-state concentrationsof gentamicin, tobramycin, or amikacin maintained in serum
for 24 h by a continuous infusion of 24, 48, or 72 mg/kgon S. marcescens endocarditis
Steady-state Mean + SDconcn in se- v No. of sterileTreatment group rum (mg/ veg vegetations/total(mg/kg) liter, mean (go CFU/g no. of rabbits
q, every.b P < 0.01 versus control (Scheffe's S test).Pp < 0.05 versus control (X2 test with Yates' correction).
d p = 0.05 versus control (X2 test with Yates' correction).
vegetation). On the other hand, the CSCs of tobramycin andgentamicin were 15 and 16.8 mg/liter, respectively. Noantibacterial effect was observed at lower concentrations. Acomparison of the percentages of sterile vegetations (versuscontrols) yielded the same results.
In each treated group, the residual concentrations ofantibiotics in the vegetations at the time of sacrifice wereabove the MICs, excluding the possibility of regrowth or apostantibiotic effect. Most of these data confirmed earlierresults (22).
(iii) In vivo effect of multiple doses of amikacin on S.marcescens endocarditis. The in vivo antibacterial effect ofthe various multiple doses of amikacin is shown in Table 3.The lowest bactericidal cumulative dose over 24 h decreasedwith the frequency of injections, being 24 mg/kg for theregimen with four injections (one injection every 6 h), 48mg/kg for the regimen with two injections (one injectionevery 12 h), and 72 mg/kg for the regimen with a singleinjection.For both S. marcescens and E. cloacae, it was verified
that the bacterial counts in vegetations at the start oftreatment were not different from those in control groups.Thus, the initial bacterial counts were in the range of thoseused in the killing curves.
Pharmacokinetics. The various pharmacokinetic parame-ters measured with each i.v. dose are shown in Table 4.
Predictive value of pharmacokinetic parameters. Simplelinear regression analysis of the mean Alog1o CFU per gramof vegetation in each of the test groups against the variouspharmacokinetic parameters studied (AUC, AUC above theCSC, AUC above the MIC or MBC, time above the CSC,and time above the MIC or MBC) revealed that only theparameter time above the CSC was significantly correlatedwith in vivo bactericidal activity (R = 0.55; P = 0.027). Ananalysis of the individual values for Alog1o CFU per gram ofvegetation separately for gentamicin and amikacin proved tobe much more discriminating. In the case of gentamicin,simple regression analysis revealed that the parameter withthe best correlation was AUC above the CSC (R = 0.61; P =0.0002); this was followed by AUC above the MIC (R =0.45; P = 0.008), time above the CSC (R = 0.39; P = 0.027),and total AUC (R = 0.38; P = 0.032). Time above the MICwas not significantly correlated (R = 0.20; P = 0.25). Foramikacin, the best-correlated parameters were time abovethe CSC (R = 0.41; P = 0.006) and time above the MIC (R= 0.64; P = 0.0001); these were followed by AUC (R = 0.30;
a Calculated from time zero to the last measurable point and then extrap-olated to infinity. For animals receiving an injection every 6 h (or every 12 h),the AUC was calculated from 0 to 6 h (or 12 h) and then multiplied by 4 (or 2)to obtain the AUC over 24 h.
P = 0.03). Stepwise regression analysis revealed that thebest correlation was shown byAUC above the CSC and thentime above the CSC for gentamicin (R = 0.69; F = 13.54; P= 0.001), compared with time above the CSC and then AUCabove the CSC for amikacin (R = 0.74; F = 24.07; P =0.0001).
DISCUSSION
The mode of administration of aminoglycosides has been asubject of debate for many years. A number of clinicalstudies in the late 1970s (8, 14) encouraged the use ofcontinuous infusion in combination with a beta-lactam anti-biotic for neutropenic patients. However, these studies werenot comparative. Nevertheless, in the study of Keating et al.(14), the combination of carbenicillin and amikacin as acontinuous infusion proved more effective than that ofcarbenicillin and gentamicin as a continuous infusion for thetreatment of septicemia in neutropenic patients, although itwas not possible to establish the respective roles of theaminoglycoside concentrations or that gram-negative bacilliare less susceptible to gentamicin (in terms of the MIC). Ourresults seem to show that the in vivo activities of aminogly-cosides (and the best dosage schedules) depend on manyfactors, including the bacterial species, the strain studied,the aminoglycoside used, and probably the site of infection.To identify the factors predictive of the best dosage regimenfor a severe gram-negative intravascular infection, we deter-mined in vitro (killing rate) and in vivo (CSC) parameters toexplain the differences in the in vivo activities of threeaminoglycosides.
In the E. cloacae endocarditis model, despite the similarMICs of the two compounds used (gentamicin and tobramy-cin), the slight difference in the in vivo activity of a single i.v.bolus injection (48 mg/kg) on the two strains (E474 and E475)could have been related to the higher killing rate in vitro(expressed as the ISB) of these two drugs for strain E475than for strain E474. This result is in agreement with that ofa previous study (23), in which a good correlation wasdemonstrated between the in vitro killing rate and the in vivoeffect of a single dose of antibiotic in an E. coli endocarditismodel. Similarly, the better in vitro activity of gentamicinthan of tobramycin or amikacin on the strain of S. marces-
cens was probably responsible for the low MED of theformer compound (48 mg/kg for gentamicin versus 72 mg/kgfor amikacin and tobramycin). Nevertheless, determinationof the CSC seemed to be necessaxy to establish a valuable invivo dose-effect relationship, allowing recommendation ofthe best dosage regimen for an aminoglycoside in a givenbacteriological and clinical situation. In fact, as has beenreported in several studies of thigh infection in neutropenicmice by Craig and colleagues (3, 16, 25), time above the MICis most often the best predictor of the in vivo efficacies ofaminoglycosides for the majority of tested strains, as long asthe dosage interval is over 6 h. Under these conditions, inwhich the administered substance is rapidly eliminated bysmall animals, high, extremely bactericidal aminoglycosideconcentrations cannot be maintained for a sufficient time toproduce an in vivo bactericidal effect. In our study, AUCabove the CSC was still important for the three antibioticstested but was more important for gentamicin than foramikacin. The fact that gentamicin had the lowest MED (48mg/kg) for S. marcescens is explained by the much higherbacterial killing rate with this drug than with amikacin ortobramycin. Sacrifice of the animals treated with gentamicinafter 6 h confirmed that most of the bactericidal activity tookplace in the first 6 h, during which concentrations in excessof the CSC were observed in the serum for about 3 h. In thecase of amikacin, the same dose of antibiotic resulted in nobactericidal activity at the end of 24 h since, although theconcentrations in serum exceeded the CSC for 6 h, thisperiod was inadequate to allow the expression of a bacteri-cidal effect slower than that of gentamicin. The results forthe multiple doses of amikacin were in agreement with thoseobtained by Craig et al. for mice with S. marcescens thighinfections (4). The cumulative 24-h dose required to achievea bactericidal effect in vivo increased with the length of thedosage intervals (24, 48, and 72 mg/kg for intervals of 6, 12,and 24 h, respectively). It should be noted that a possiblepostantibiotic effect was not investigated in our model, sincethe residual concentrations in the vegetations were alwaysequal or superior to the MIC, thus excluding the possibilityof regrowth as well.
Studies comparing the relative efficacies of several amino-glycosides in the same model of gram-negative infection arerare, while experimental or clinical studies comparing theefficacy of a once-daily injection with that of intermittentdosing yield divergent results (1, 13, 20, 24, 25). One of thereasons for these differences could be a failure to take intoaccount the in vitro bacterial killing rate, which seemed toplay a major predictive role in the present study for both E.cloacae and S. marcescens. For three aminoglycosides withcomparable pharmacokinetics, it was found that the higherthe killing rate, the lower the MED. Finally, in the absenceof a comparative study, we would have shown that, in asevere S. marcescens infection, amikacin was more effectiveas a continuous infusion than as a bolus injection, thatgentamicin was more effective as a bolus injection than as acontinuous infusion, and that tobramycin was equally effec-tive in both modes of administration. By taking into accountCSC and in vitro bacterial killing rate, the observed differ-ences could be explained; a valuable dose-effect relationshipwas based on different predictive factors for the in vivoefficacies of gentamicin and amikacin in this model (AUCabove the CSC for gentamicin and time above the CSC foramikacin). Measurement of these two parameters (in vitrokilling rate expressed as ISB and CSC) should, in ouropinion, be included in experimental studies of the pharma-codynamics of antibiotics. Nevertheless, we are not sure
whether a CSC defined for one infected site is predictive foranother one.
Investigation of bactericidal activity during a period asshort as 24 h offers several advantages. (i) The in vivoconcentration-effect relationship can be studied with greaterprecision than during longer observation periods (24), sincepossible initial differences between two antibiotic regimensmay vanish during a treatment lasting several days. (ii) It isprobable that the activities of antibiotics (especially amino-glycosides) in the first hours of a severe gram-negativeinfection in an immunocompromised patient constitute animportant prognostic factor, justifying the search for thebest, well-adapted dosing regimen.The limitations of animal models in the comparative study
of once-daily versus intermittent dosing regimens have re-cently been underlined by Craig and coworkers (4) andinvolve the rapid clearance of antibiotics in small animals,resulting in a half-life shorter than that in humans. Thisparameter probably plays an important role in the relation-ship between the 24-h cumulative dose and efficacy, and thepresent study confirms the high predictive value of timeabove the CSC for the level of in vivo bactericidal activity.In our study, the most rapid bactericidal effect of gentamicinadministered as a single i.v. bolus injection was achievedwith doses producing concentrations not recommended forhuman treatment. It is thus by no means certain that afavorable conclusion can be drawn from this study withregard to once-daily dosing of gentamicin. Similarly, it ispossible that with a longer half-life, once-daily amikacincan achieve efficacy in the model of S. marcescens en-docarditis equivalent to that observed with a continuousinfusion or intermittent administration. Nevertheless, itmust be stressed that the CSC of amikacin (5.3 to 8 ,ug/ml) isconsistent with the concentrations achieved in humans un-der normal dosage schedules. In fact, a continuous infusionof 15 mg of amikacin per kg in humans produces steady-stateserum drug concentrations of about 10 ,ug/ml, comparable tothe concentrations that we observed in rabbits with 48 mg/kg(6). Thus, a possible conclusion that can be extrapolated tohumans with severe S. marcescens infections is that amika-cin, administered as three or four daily doses or as acontinuous infusion, displays outstanding efficacy within 24h of administration. Shorter intervals are probably notsuitable, because of the possibility of a down-regulationphenomenon, responsible for adaptive resistance in the 6 hfollowing a brief exposure to an aminoglycoside (5).
In conclusion, two factors affect the in vivo efficacies ofthe aminoglycosides and probably explain, at least in part,the discrepancies reported in the literature for the optimaldosing regimens of these drugs: the in vitro killing rate,which may be measured at several concentrations, and the invivo CSC, which can be much higher than the MIC. Furtherstudies are needed to examine whether there is a relationshipbetween these early bactericidal effects and the ultimatetreatment outcome of severe gram-negative sepsis.
ACKNOWLEDGMENTS
We thank M. 0. Hervy for helpful secretarial assistance.This work was supported in part by grants from Faculte de
Medecine de Nantes, Nantes, France.
REFERENCES1. Beaucaire, G., 0. Leroy, C. Beuscart, P. Karp, C. Chidiac, M.
Caillaux, and the Study Group. 1991. Clinical and bacteriologicalefficacy, and practical aspects of amikacin given once daily for
severe infections. J. Antimicrob. Chemother. 27(Suppl. C):91-103.
2. Billian, V., C. Beuscart, 0. Leroy, P. J. Dujardin, G. Beaucaire,C. Chidiac, Y. Mouton, and the SES Group. 1989. Septicemies aEnterobacter, Serratia, Acinetobacter ou Pseudomonas. Pathol.Biol. 37:39-42.
3. Craig, W. A., J. Legett, K. Totsaka, and B. Vogelman. 1988.Key pharmacokinetics of antibiotic efficacy in experimentalanimal infections. J. Drug Dev. 1(Suppl. 3):7-15.
4. Craig, W. A., J. Redington, and S. C. Ebert. 1991. Pharmaco-dynamics of amikacin in vitro and in mouse thigh and lunginfections. J. Antimicrob. Chemother. 27(Suppl. C):29-40.
5. Daikos, G. L., G. G. Jackson, V. T. Lolans, and D. M.Livermore. 1990. Adaptive resistance to aminoglycoside antibi-otics from first-exposure down-regulation. J. Infect. Dis. 162:414-420.
6. De Broe, M. E., L. Verbist, and G. A. Verpooten. 1991. Influenceof dosage schedule on renal cortical accumulation of amikacinand tobramycin in man. J. Antimicrob. Chemother. 27(Suppl.C):41-47.
7. Drugeon, H. B., J. Caillon, M. E. Juvin, and J. L. Pirault. 1987.Dynamics of ceftazidime-pefloxacin interaction shown by a newkilling curve-chequerboard method. J. Antimicrob. Chemother.19:197-203.
8. Feld, R., P. G. Tuffnell, J. E. Curtis, H. A. Messner, and R.Hadelback. 1979. Empiric therapy for infections in granulocy-topenic cancer patients: continuous infusion of amikacin pluscephalothin. Arch. Intern. Med. 139:310-314.
9. Garraffo, R., H. B. Drugeon, P. Dellamonica, E. Bernard, and P.Lapalus. 1990. Determination of optimal dosage regimen foramikacin in healthy volunteers by study of pharmacokineticsand bactericidal activity. Antimicrob. Agents Chemother. 34:614-621.
10. Gerber, A. U., H. P. Brugger, C. Feller, T. Stritzko, and B.Stalder. 1986. Antibiotic therapy of infection due to Pseudomo-nas aeruginosa in normal and granulocytopenic mice: compari-son of murine and human pharmacokinetics. J. Infect. Dis.153:90-97.
11. Gilbert, D. N. 1991. Once-daily aminoglycoside therapy. Anti-microb. Agents Chemother. 35:399-405.
12. Jackson, G. G., V. T. Lolans, and G. L. Daikos. 1990. Theinductive role of ionic binding in the bactericidal and post-exposure effects of aminoglycoside antibiotics with implicationsfor dosing. J. Infect. Dis. 162:408-413.
13. Kapusnik, J. E., C. J. Hackbarth, H. F. Chambers, T. Carpen-ter, and M. A. Sande. 1988. Single, large, daily dosing versusintermittent dosing of tobramycin for treating experimentalPseudomonas pneumonia. J. Infect. Dis. 158:7-12.
14. Keating, M. J., G. P. Bodey, M. Valdivieso, and V. Rodriguez.1979. A randomized comparative trial of three aminoglycosides.Comparison of continuous infusions of gentamicin, amikacinand sisomicin combined with carbenicillin in the treatment ofinfections in neutropenic patients with malignancies. Medicine(Baltimore) 58:159-170.
15. Le Bel, M., and M. Spino. 1988. Pulse dosing versus continuousinfusion of antibiotics. Pharmacokinetic-pharmacodynamic con-siderations. Clin. Pharmacokinet. 14:71-95.
16. Legett, J., B. Fantin, S. Ebert, K. Totsuka, B. Vogelman, W.Calame, and W. Craig. 1989. Comparative antibiotic dose-effectrelations at several dosing intervals in murine pneumonitis andthigh-infection models. J. Infect. Dis. 159:281-292.
17. McConkey, S. J., D. C. Coleman, F. R. Falkiner, S. R. McCann,and P. A. Daly. 1989. Enterobacter cloacae in a haematology/oncology ward. First impressions. J. Hosp. Infect. 14:277-284.
18. Moore, R. D., P. S. Lietman, and C. R. Smith. 1987. Clinicalresponse to aminoglycoside therapy: importance of the ratio ofpeak concentration to minimal inhibitory concentration. J. In-fect. Dis. 155:93-99.
19. Pearson, R. D., R. T. Steibigel, H. T. Davis, and S. W. Chapman.1980. Method of reliable determination of minimal lethal antibi-otic concentrations. Antimicrob. Agents Chemother. 18:699-708.
20. Pechere, M., R. Letarte, and J. C. Pechere. 1987. Efficacy of
different dosing schedules of tobramycin for treating a murineKlebsiella pneumoniae bronchopneumonia. J. Antimicrob. Che-mother. 19:487-491.
21. Perlman, B. B., and L. R. Freedman. 1971. Experimentalendocarditis. II. Staphylococcal infection of the aortic valvefollowing placement of a polyethylene catheter in the left side ofthe heart. Yale J. Biol. Med. 44:206-213.
21a.Potel, G., J. Caillon, D. Bugnon, H. B. Drugeon, and D. Baron.1988. Program Abstr. 28th Intersci. Antimicrob. Agents Che-mother., abstr. 373.
22. Potel, G., J. Caillon, B. Fantin, J. Raza, F. Le Gallou, J. Y.Lepage, P. Le Conte, D. Bugnon, D. Baron, and H. Drugeon.1991. Impact of dosage schedule on the efficacy of gentamicin,tobramycin, or amikacin in an experimental model of Serratiamarcescens endocarditis: in vitro-in vivo correlation. Antimi-crob. Agents Chemother. 35:111-116.
23. Potel, G., N. P. Chau, B. Pangon, B. Fantin, J. M. Vallois, F.
Faurisson, and C. Carbon. 1991. Single daily dosing of antibiot-ics: importance of killing rate, serum half-life, and proteinbinding. Antimicrob. Agents Chemother. 35:2085-2090.
24. Powell, S. H., W. L. Thompson, M. A. Luthe, R. C. Stern, D. A.Grossniklaus, D. D. Bloxham, D. L. Groden, M. R. Jacobs, A. 0.Di Scenna, H. A. Cash, and J. D. Klinger. 1988. Once-dailyversus continuous aminoglycoside dosing: efficacy and toxicityin animal and clinical studies of gentamicin, netilmicin, andtobramycin. J. Infect. Dis. 147:918-932.
25. Vogelman, B., S. Gudmundsson, J. Tarnidge, J. Legett, andW. A. Craig. 1988. Correlation of antimicrobial pharmacoki-netic parameters with therapeutic efficacy in an animal model. J.Infect. Dis. 158:831-847.
26. Weinstein, L. 1985. Gram negative bacterial infections. A lookat the past, a view of the present, and a glance at the future.Rev. Infect. Dis. 7(Suppl. 4):538-544.
