Pharmacokinetics and pharmacodynamics of fluoroquinolones
George Drusano l, Marie-Thhhe Labro2, Otto Cars3, Paul Mendes4, Prarnod Shah', Fritz Sorgel and Willi Weber7
'Division of Clinical Pharmacology, Department of Medicine, Albany Medical College, Albany, New York, USA; 21NSERM U479, Service d'Hematologie et d'Immunologie Biologiques, CHU Xavier BICHAT, Paris, France; 3Department of Infectious Diseases, Uppsala University Hospital, Uppsala, Sweden; 4Hoechst Marion Roussel Inc, Bridgewater, New Jersey, USA; 5Med Klinik 111, Frankfurt am Main, Germany; 61nstitute for Biomedical & Pharmaceutical Research, Nurnberg Herodsberg, Germany; 7Hoechst Marion Roussel AG, Clinical Development, Frankfurt am Main, Germany
INTRODUCTION antibacterial spectrum and are active mainly against the Enterobacteriaceae. These compounds are character- Since 1962, when the first quinolone derivative with ized by unfavorable pharmacokinetic properties and antibacterial activity, nalidixic acid, was synthesized, low in vivo potency. The newer derivatives have more there has been a concerted and systematic search for favorable pharmacokinetic profiles, resulting in higher compounds with enhanced antibacterial activity and/ serum concentrations, and have lower minimum or an improved pharmacokinetic profile. The potent inhibitory concentrations (MICs), making them appro- activity of nalidixic acid against Gram-negative bacteria priate for the treatment of systemic infections [l]. This has been greatly improved and expanded upon by paper focuses on the tissue penetration, intracellular combining a 7-piperazinyl ring and a 6-fluorine atom accumulation and pharmacodynamics of the fluoro- on the quinolone nucleus (as in the case of cipro- quinolones. floxacin, norfloxacin and pefloxacin) or on the 1,8- naphthyridine nucleus (as in enoxacin), or producing TISSUE PENETRATION fluorinated tricyclic derivatives (levofloxacin and ofloxacin). The fluoroquinolones can be divided into It is essential that antibacterial agents used to treat four groups on the basis of their chemical structure- infections caused by intracellular bacteria have favorable monocyclic, bicyclic, tricyclic and tetracyclic structures intracellular penetration, accumulation and disposition. linked to the pyridone-P-carboxylic acid nucleus-or Concentrations of antibiotics in various tissues have into four biological groups according to their anti- attracted considerable interest. The fluoroquinolones bacterial spectrum and extent of metabolic transform- are able to achieve high concentrations in cells, although ation. The earlier quinolone derivatives have a limited the exact mechanism by which this occurs is not known. Lipophilicity does not seem to play a significant role, since some of them are hydrophilic [2]. As with other antibiotics which exhibit intra- cellular accumulation (e.g. tetracyclines, macrolides), Corresponding author and reprint requests: the concentrations of fluoroquinolones in different George Drusano, Division of Clinical Pharmacology, tissues are often higher than the concurrent serum Department of Medicine, Albany Medical College, levels. One must, however, bear in mind that tissue Albany, NY 12208, USA concentrations are difficult to interpret. Since drugs Tel: +'I 518 262 6761 Fax: +1 518 262 6333 are not distributed evenly throughout the body or E-mail: within a specific tissue, concentrations obtained by
2S27 2528 Clinical Microbiology and Infection, Volume 4 Supplement 2
homogenizing tissue are inherently inaccurate and, as fluid levels of antibiotics do not always equilibrate, these are mean values of concentrations in different such as in large collections of inflammatory fluid, the tissue compartments, they cannot be used to estimate concentrations of free (non-protein-bound) antibiotic the drug concentration at the site of action [3,4]. This are the most useful surrogate markers for the concentra- is clearly evident from the studies of Baldwin et al, tions achieved at the site of infection for extracellular where concentrations of ciprofloxacin and lomefloxacin pathogens. However, for respiratory tract infections in bronchial biopsies, alveolar epithelial lining fluid caused by intracellular bacteria, such as Legionella spp. and alveolar macrophages were compared with simul- and Chlamydia pneumoniae, the high intracellular levels taneous serum levels [5]. The bronchial biopsy concen- of fluoroquinolones may be important. trations of ciprofloxacin and lomefloxacin were 1.6 and 1.7 times those of serum, respectively. The lntracellular accumulation of fluoroquinolones corresponding values for macrophage concentrations The cellular pharmacokinetics of antibacterial agents is were 11.8 and 20.1, respectively. Interestingly, the one of the parameters responsible for their intracellular concentrations of the two fluoroquinolones in epithelial activity (Figure 1). Fluoroquinolones are able to con- lining fluid were also higher than in serum (2.1 and 1.9 centrate within phagocytic and non-phagocytic cells times the serum levels, respectively), possibly indicating and remain active against different facultative, obligate, antibiotic delivery to the extracellular site from the intracellular pathogens. Many studies have dealt with intracellular antibiotic reservoir. Similar results have the characteristics of the cellular accumulation of been obtained with levofloxacin which showed higher fluoroquinolones in vitro, particularly in phagocytic concentrations in epithelial lining fluid and alveolar cells. macrophages than in serum [6]. The classical technique used to determine the The common bacteria causing acute respiratory various parameters of cellular pharntacokinetics (accunl- tract infections are extracellular pathogens. The ulation kinetics, cellular location and efflux) involves bactericidal action of the antibiotic is thus normally in vitro incubation of the test cells with the drug required in the interstitial fluid. In most tissues, the (either radiolabeled or unlabeled) followed by velocity antibiotic molecules diffuse freely between the vascular centrifugation through a water-impermeable oil cushion and interstitial fluid space. Even if serum and interstitial to separate the free antibiotic (aqueous phase) from the
DRUG MICROORGANISM CELL HOST Microbiologi@ -Susceptibility -Bactericidal - Cytokines activity (MIC, MBC) 4- - - - Resistance - - - - + activity
Location -Orige PWPD w Extracellular Phagocytic Absorption non-phagocytic Distribution Metabolism --Species I Elimination State: I Mediators -\.Clinical status Resting, activated elicited T Cellular Pharmacokinetics lntracellular Accumulation (cytosol, phagosome, Efflux phagotysome)
Metabolism Quiescence, growth
Figure 1 Parameters of antibacterial bioactivlty. Drusano et al: Fluoroquinolone pharmacokinetics and pharmacodynamics 2S29 cell-associated drug. Drug measurement may be done extracellular concentration. Otherwise, the C/E ratio either by a fluorometric assay, owing to the fluorescent may just characterize the overall amount of cell- property of most fluoroquinolones, or by radioactivity associated drug, including the membranes and counting. The amount of cell-associated drug (ng/x associated ‘true’ intracellular compartments. Drug cells or ng/mg protein) is determined by standard drug efflux is measured by similar techniques, using drug- dilutions. After determination of the intracellular water loaded cells isolated and placed in drug-free medium. content (by isotopic techniques, cytometiy, etc.), it is At repeated time intervals, aliquots are centrifuged and possible to calculate the intracellular concentration the drug measured in both the supernatant and the cell (mg/L) and, from that, the parameter most frequently pellet to calculate the percentage of drug released. used for comparison, the cellular/extracellular concen- Variations in the experimental medium (such as pH and tration ratio (C/E). inhibitors) and other conditions (such as temperature It is important to determine the location of and killed cells) are used to investigate the mechanisms the drug (by cell fragmentation techniques), e.g. in underlying drug uptake and efAux. cytosol, granules or lysosomes, to ensure that the C/E The main features of fluoroquinolone cellular ratio represents the true ratio between intra- and pharmacokinetics are shown in Table 1 [7-161. The
Table 1 Comparative cellular pharmacokinetics of fluoroquinolones
1)rug C/E ratio Ernux Activity (technique) Cell (tinie) (time) (bacteria) Reference
Ciprofloxacin (bioassay) MPH (m) 2.7 (30 min) ND + (S.a.) 7 (HPLC) PMN 4 (30 min to 2 h) to 8 8 (5 inin) (fluorometry) PMN 4.7 (20 min) 30% (5 min) + (Pa.) 9 8.6 (30 imn) 10 J774 4.4 ( 5 min) + (S.a.) 11
Levofloxacin (fluorometry) PMN 6.1 (20 min) to + (S.a.) 9 >8.3 (10 min) 12 Hep 2 2.1 (10 inin) 12 McCoy 3.9 (10 min) 12 Vero 2.1 (10 inin) 12
Ofloxacin (fluoromrtry) PMN 4.9 (5 min) to 65588% (5 min) + (P.a., Sa.) 9 7.6 (30 min) 10 Hep 2 2.8 (10 min) 13 McCoy 4.7 (10 min) 13 Vero 2.4 (10 min) 13 J774 5.3 (5 mill) + (S.a.) 11 (HPLC) WI-38 8.6 (30 min) + (S.e.) 14
Spadoxacin (fluorometry) PMN 6.5 (20 min) to 60% (5 min) + (S.a.) 9 >13.2 (5 inin) 11 15 Hep 2 7 (20 mm) 15 McCoy 10 (20 inin) 15 J774 6.4 (30 min) + (S.3.) 11
Trovafloxaciii (I-adioartivity) PMN 11 (20 min) 80% (i min) + (La.) 16 MPH (h) 10 (20 mill) 65% (5 inin) McCoy 9.6 (20 niin)
11, human: 1x1, IIIOUFC: MPH, macrophage: ND, not determined; Pa., Pwudomorras aerrrqqirruru; PMN, polymorphonuclear neutrophil; S.a., Stuphylococcus atrmu; S.e., Snlrnoridln enreriridis. 2S30 Clinical Microbiology and Infection, Volume 4 Supplement 2
upt'ike of most fluoroquinolones is rapid, followed by decreased, and the drug was located in the same a long-lasting plateau. Whatever the cell species, the conipartnient as the bacteria [I 81. However, the C/E ratio is ~i~~iallymoderate (i.c. < 10) 1171. The niechaiiisms still remain unkno~vn. cellular location. riiaiiil>-cytosolic, has been determined In vivo. in cxpcririierital models arid Various only for fleroxacin, loniefloxacin and pefloxacin [ 181. infectious diseases. fluoi-oquinolones have proved effec- Effl~ixis also rapid, with 60-90%1 of the intracellular tive against infections caused by intracellular pathogens. drug being released within the first 5 min. Since they are frequently used in iinmunocomproniised The mechanism underlying the accumulation of individuals or as combination therapy for mixed fluoroquinolones has not been clarified. Fluoroquino- infections, it is interesting to note that their cellular lone uptake does not 'ippear to he saturable; most uptake (at least in vitro) i\ not niodified by vxions studies denionstrate that decreasing the incubation antimicrobial agents and cheiiiotlierapeutic drugs teniperaturr to 1°C results in decreased uptake except [9,24.25]. Targeted delivery of fluoroquinolones by hi- trov'ifloxacin. ~hichshows increased uptake Table 2 Phartiiacohnetic parameters F (96) CI, (mL/min) C1, (mL/niin) Cl,, (niL/min) V~ii(L/kg) f,, ("h of dose") f'ke\ ("9 of dosr") Ciprofloxacin 60-70 550-750 300-400 280-320 25-35 5 65 15.2 1)ifloxacin NA 41 3b 43 37 0 97.5' 96 NA €nosacin 7(&90 500-650 260-350 220-300 2.5-3.0 5 60 NA Fleroxacin 95-100 100-120 7(HO 25-40 0.8-1 .2<' 60-75 1.5 Levoflosac1n 100 140-200 100-1 40 4(MO 1.1-1.3 60-90 12-15 Lonieflosacin 9.5-100 200-250 - 180 70 1.5-2. I' -70 NA Norflosacin <35' NA -250 NA NA < 30 NA Oflosacin 95-100 200-250 160-200 40-50 1.2-1.4 8G90 NA Pefloxacin 95-100 100-13l~ 7-12 100-120 1.1-1.7 6- 10 60' Sparflosacin 60' 160-2001' 15-20 130-170 1.6-1.Q' 33-50 NA Temafloxacin - 95 100-250 120-1 60 6(WO 1.7-2.1 60-70 NA Tosufloxacin <30" NA 90-120 NA NA 5 401' NA Adapted with permission from the American Journal of Medicine [XI. F = fraction of dose absorbed, C1,= total clearance; C1, = renal clcarancc; CI,,,= non-renal clearance; Vn,! = volume of distribution; f,, = fraction of dose excreted unchanged in urine: it,,,, = fraction of dose excreted unchanged in feces; NA = data not available " Excreted as unchanged compound. '' CVF. ' Vnll/F. I' Vd,, (Vd at steady state). ' Percentage may be sniallrr, sincc no intravenous dosagc can bc studicd. ' Only data on excretion of total ''C radioactivity were used; no discrimination was made benveen parent drug and metabolite. a Numbers are estimates of the niniinial dose absorbed. The actual nuniber may be subrtantially higher. I' Dependent on food intake (without food <30; with food 540). and norfloxacin [26]. The extent of absorption of basic group, which affect pH-dependent lipophilicity. tiorfloxacin niay also be affected, but that of Many important properties of distribution may be ciprofloxacin is not [26]. Total absorption of ofloxacin explained by interactions between the pH of these and levofloxacin is unaltered when given with groups and the local pH in the body. These physico- iiornial or fat-rich food [26]. An unchanged AUC chemical properties may also partly explain the for fleroxacin, ofloxacin, levofloxacin, pefloxacin and elimination properties of fluoroquinolones. They all rufloxacin suggests that N;-methylation protects undergo tubular secretion as either acids or bases and against the effects of food. Food does not alter the some are also significantly reabsorbed. Hepatic handling overall absorption but prolongs the T,,,,, slightly but and resultant metabolite excretion is also influenced by significantly [26]. Delay in gastric emptying may cause K;-niethylation. the increase in T,,,,, of fluoroquinolones ingested with fat-rich food. Elevated gastric pH may enhance or delay PHARMACODVNAMICS absorption by affecting dissolution and may shift the equilibriuni between ionized (less absorbed) and non- Pharmacodynaniics deals with the relationship between ionized (better absorbed) forms. Antacids may increase exposure to a drug and a resulting effect, either gastric emptying time and thus reduce tlie C,,,.,, and therapeutic or toxic. Through delineation of such delay the T,,,,,.The chelating effect of antacids on exposure-response relationships, therapy for the patient absorption depends on the interval between fluoro- can be optimized and toxicity minimized while the quiriolone and antacid intake. Therefore, the dose of therapeutic response is maximized. There have been antacid should not be given earlier than 2 h after the many studies of anti-infective agents in which drugs of fluoroquinolone do5e. different classes have had their pharmacodynaniics Most fluoroquinolones are excreted primarily by determined for bacteria, as well as for other pathogens. the kidney but the gastrointestinal route also plays a For the fluoroquinolones in particular, a considerable significant role in the overall elimination of-these agents amount of pharrnacodynamic data has been recorded. 1261. Hepatic elimination plays a role only for certain These data stem froin in vitro studies, animal models of agents such as pefloxacin [26]. The physicochemical infection and clinical studies. In the case of bactericidal properties of fluoroquinolones are greatly influenced by agents, the rate of kill can be thought of as the their two functional groups, the carboxyl group and pharniacodynaiiiic variable most closely linked to 2532 Clinical Microbiology and Infection, Volume 4 Supplement 2 organism kill at the primary infection site. To gain an insight into bacterial killing dynamics, two pieces of information are critical. The first is a measure of drug exposure and the second is a measure of drug potency for the pathogen in question. Multiple measures of drug exposure can be posited. The first issue concerns whether drug con- centrations should be directly measured at the primary infection site or whether serum exposure measures will suffice. In the majority of cases, the pharrnacokinetics are linear. Therefore, first-order rate constants will link drug transfer from serum to the site of infection. Consequently, while the actual concentration-time profiles will differ slightly between two sites, linkage 0 4 8 12 16 20 24 to the effect will not be impeded, as different para- Time (h) meters will be obtained for whatever function is Figure 2 Pharmacokinetic surrogate relationyhips. AU<: = employed for the analysis. Consequently, either central area under thc scrum concentration time curve; C,,,,,= or peripheral compartment concentrations may be maximum serum concentration; MIC: = minimum inhibitory used. For ease of data collection, this will usually result concentration: T>MIC = time that the serum concentration in serum data being chosen for the pharmacodynamic exceeds the MIC. andysis. Reducing the number of pathogens located at the yite of infection is the primary target of an antibiotic. which keep trough levels high would produce the best Only the drug concentration at the site of infection results. is relevant for its degree of efficacy. However, it is Fluoroquinolones not only have activiq against not possible to determine the exact location of the extracellular pathogens but are also active against infection where the concentration of interest should be intracellular pathogens such AS A4yroylarriio and meawred. In general, therefore, the seruni concentra- Qzluriiydia, amongst others. Their ability to enter cells tion is used as a surrogate marker of the infection is probably the key to their activity against these compartment concentration. After administration of an pathogens. Generally, they penetrate well into most antibiotic, the infection site concentration will follow cellular compartments where these organisms reside. In the changes in serum concentration. This delay is addition, the uptake into cells may explain some of the characterized by an equilibration half-time. When findings of the concentration-time curve in the lung for pharmacokinetics are linear and a steady-state situation fluoroquinolones when measured by the technique is established, the infection site concentration is directly described by Baldwin et a1 1271. Concentrations of proportional to the serum concentration, which reflects fluoroquinolones in extracellular fluid often consider- exposure of the pathogens to the drug and, therefore, ably exceed those seen in the serum. A partial explan- should be a good surrogate marker (Figure 2). ation for this is efflux of drug from lung cells and lung For 5ituations where different organisms will be niacrophages, which concentrates fluoroquinolones part of the analysis (as in the analysis of clinical trial intracellularly because of ion-trapping effects. When data), it is important to noririalize the drug exposure to serum co~iceiitratio~isdecline, this could ‘illow efRux of the nieasui-e of potency of the drug for the pathogen drug from cellular reservoirs which would maintain (in most instances, this will be the MIC). This results drug concentrations locally in excess of the serum levels. in three pharmacodynaniic variables which are drawn The concentration dependence of the bacterial kill from various aspects of the shape of the concentratioii- rate will deteriiiine which of these dynamic variables is time curve: the peak/MIC ratio, the AUC/MIC ratio most closely linked to organism kill. If the drug’ 1s’ not (AUIC) and the time>MIC, which may all be linked very concentration-depeiideiit in kill rate (as is the case to outconie. If the peak/MIC ratio is linked to with p-lactani agents), the rate of kill cngendered by outconie, this indicates that relatively infrequent dosing high concentrations around the peak will not differ intervals xvould lead to the best results, with production cubstantially from the kill rate engendered by concen- of a very peaked shape for the concentratiori-time trations doivn near the MIC. Consequently, the curve. The AUC/MIC ratio would indicate that the number of organisms killed by a dose of drug will be therapeutic outcome is independent of curve shape, relatively constant, i.e. the kill rate multiplied by the aid tinie>MIC would mean that flat curve shapes time that the co~iceiitratioiiexceeds the MIC. In this Drusano et al: Fluoroquinolone pharmacokinetics and pharmacodynamics 2S33 case, tinie>MIC would be the pharniacodynamic MIC) may be suppressed. Consequently, while variable of interest. AUC/MIC can be thought of as being linked to cell For drugs which are very concentration-dependent kill for the parent strain, for peak/MIC suppression of in kill rate (e.g. fluoroquinolones and aminoglycosides), resistant subpopulations should be considered as well. the kill rate changes continuously with concentration, and the total number of organisms killed is provided by In vitro data the path integral of kill rate (which is concentration- Studies which have shed considerable light on fluoro- dependent) over time. That is, the kill rate near the quinolone pharmacodynamics are those by Blaser et a1 peak multiplied by the duration over which it is and Dudley et al, employing the hollow-fiber pharma- appropriate (i.e. until the concentration is reached codynamic model [28,29]. This model is particularly which produces a noticeably changed kill rate) plus the instructive, as human serum pharmacokinetic-time next kill rate multiplied by the time over which it is profiles can be simulated and effects on organism load effective, and so on. With this situation, the AUC/MIC ascertained. Blaser et al examined enoxacin against ratio should be most closely linked to outcome, as the a strain of Klebsiella pnecrmoniae [28]. In a series of AUC is, itself, the integral of the concentration-time experiments, once-daily dosing with enoxacin was curve. compared with half the dose administered twice daily The peak/MIC ratio should never be linked to (same total daily dose). In one of three experiments outcome in the scenario created. Were this so, there with twice-daily dosing, there was clearcut break- would be an identifiable dose of drug which had through growth which was not seen with once-daily demonstrable activity, whereas half this dose would dosing (Figure 3) [28]. They then performed a large have markedly reduced activity to alniost nil. In number of experiments and plotted residual organisms general, antibiotics do not behave in this manner determined at the end of the experiment. As can be because the scenario, as generated, is incomplete and seen in Figure 4 [28], when an organism-dosing examines the sensitivity of the organism to the drug as regimen pair had a resultant peak/MIC ratio of 10/1 a static phenomenon. However, it is known that dense or greater, the unit was always completely sterilized. bacterial populations, such as are seen in serious This is important, as it lends credence to the idea of infections, are composed of distributions of organisms, suppressing resistant mutants by employment of specific some of which may be spontaneously mutated to dosing regimens. altered drug susceptibility. Frequently, there will be a This idea was directly addressed by Dudley et a1 number of resistant mutants in large populations of with I? aertrginosa [29]. They simulated an intravenous organisms. In the case of nosocomial pneumonia, dose of ciprofloxacin 200 mg on a 12-hourly schedule where Enterobacter cloacae and l? aeruginosa are common (a regimen then in frequent clinical use). The baseline pathogens, bacterial densities are high and there is bacterial population was examined on agar plates considerable pus in the chest, giving a total population containing differing concentrations of ciprofloxacin. of 10'"-10". Mutational frequencies for resistance This was then repeated prior to the 12-h simulated dose differ, and for these pathogens, mutational frequencies and again at 24 h. The administration of ciprofloxacin are in the order of 10*-10' and 106-107, respectively. at a suboptimal dose rapidly selected the resistant part Consequently, there are 10'-10' resistant organisms of the initial bacterial population (Figure 5) [29]. As present before any selective pressure (antibiotic admin- well as amplifying the resistant mutant subpopulation, istration) has been applied. this model enables the effects of altering dose and With fluoroquinolones, resistant Gram-negative schedule on bacterial cell kill to be seen. It also organisms are frequently mutated in the gyrA gene, and provided the first clue that dosing could be modified to Gram-positive bacteria in the parC gene, with single obtain a specific end: suppression of the resistant part point mutations most frequently causing a four- to of the bacterial population. eight-fold shift in the MIC. The size of the MIC change is important, because larger changes (normally Animal models 16- to 128-fold), as are seen with the selection of Leggett et a1 and Fantin et a1 provided considerable mutants stably derepressed for Bush class I enzyme insight into the pharmacodynamics of fluoroquino- production by some p-lactams, could not be counter- lones in an animal infection model [30,31]. This was selected with a greater dosing intensity. It may be that the standardized neutropenic mouse thigh infection peak/MIC plays a role here. If a high enough peak/ model employed by Craig's laboratory in which nlice MIC ratio can be achieved, not only may the parent are made neutropenic with cyclophosphamide [32]. organism be killed, but also the gp4 mutant (or any The pathogen is injected into the mouse thigh and other type of mutant with a relatively low change in allowed to multiply for several hours. Antimicrobial 2S34 Clinical Microbiology and Infection. Volume 4 Supplement 2 /U-? / \ I I I I I I I 1 0 4 8 12 16 20 24 28 Time (h) Figure 3 BactericicM ‘ictivity of two enoxacin regimens dgainst Klchsirlin ptiu~~~imiimin a pliariiiacokinetic niodcl. Cultures xvcre eradicated in three of three experiments when the total daily dose was given as one single dose but in only two of three c~periiiient\when the saiiic total ctdy dose was giveii as two equal doses [ 381. Reproduced with permission froiii Antimicrobial Agent, and Chciiiotherapy. >3 - - >3 2 NETlLMlClN 4 h 0 -2 00 O 0 O -,. +-- +-- = - I- <-4 - >3 2 0 NETlLMlClN 24 h 0 -2 <-4 - b I-<-4 O+ 10 100 O++ 1 10 100 Ratio of peak concentration to MIC Figure 4 Antibactcrinl- effect ofiiiultiple-dose reginicns ofcnoz.lcm and netilmicin against tive organisins. Change 10 - 9- 8- 7- 6- I 5- I I I I I I 4- I I I I I I 3- I I I I I I I I 2- I I I I I I I I I 1- I ITotal CFU/mL I I 1 I I I I Resistant to: I I 1E 0- I I I I I 00.05mg/L I I I I I I 0.10 mg/L I I I I I I 0.50 mg/L I I I I I 1 mg/L I I I2 mg/L I I I I I4 mg/L I 7 I 6 5 4 3 2 1 0 0 4 8 12 16 20 24 1 Time (h) Doset Dose Figure 5 Bactericidal activity of ciprofloxaciti and subpopulation analy5i5 of Pseudoitionas acrqinoso 810 in an in vtvo model following two simulated 200-mg doses. Values are mean k SD [29]. Reproduced with permission from the American Journal of Medicine. regimens are then initiated and the mice sacrificed at then challenged with a dense inoculuni intraperitone- various time points, and the number of organisnis in ally. As in the Craig model, the organisms are allowed the thigh determined. Large numbers of doses and to multiply before therapy is initiated. However, in this schedules are examined and the linkage of one of the model, there is very reliable bactereniia and survival is three above mentioned pharniacodynaniic variables used as the endpoint. An important point to emphasize determined by multivariate regression analysis. For is that the inoculum in the Craig model is of the order fluoroquinolones, it was clear that the AUC/MIC ratio of 10" nlicroorganisms, while in the Johnson niodel (at was the dynamic variable most clearly linked to least in this series of experiments) an inoculum of 10" outcome. This supports the theory that drugs which are I? aevugirzosn bacteria (1 00 x LDjo) was employed. As concentration-dependent in kill rate will have an might be expected, there is a relatively low probability AUC/MIC ratio linked to outcome. However, it seems that a resistant mutant will be included in the inoculuni to contradict the data of Blaser et a1 and Dudley et a1 in the Craig model, whereas the larger inoculuni in the 128,291. Johnson model provides a much higher probability of Some insight into this discrepancy was provided having multiple resistant mutants in the population by Drusano et a1 employing a neutropenic rat model (mutational frequency for fluoroquinolone resistance first described by Johnson [33,34]. Rats are made in Pseudomonas spp. is frequently in the range of profoundly neutropenic with cyclophosphamide and 1 0"-I 07). 2S36 Clinical Microbiology and Infection, Volume 4 Supplement 2 A series of dose-finding experiments identified and 20 rng/kg four times a day groups), the dynaniic- 80 nig/kg of lomefloxacin as the dose which would ally linked variable is the AUC/MIC ratio. When this generate approximately 70% survival in this model [33]. ratio is achieved, the peak/MIC ratio explains outcome This is still on the relatively steep part of the exposure- better, most likely for the reasons evident in the Blaser response curve. An experiment was then performed et a1 and Dudley et a1 in vitro experiments, i.e. suppres- with 200 animals (50 animals per group). One group sion of emergence of resistance [28,29]. The second received saline placebo, while the other three all experiment examined three dosing groups of 20 received 80 rng/kg per day, but on three different animals each (control, 40 nig/kg o.d. and 20 mg/kg schedules of administration (80 nig/kg o.d., 40 mg/kg twice a day). The results are shown in Figure 7 [33]. bid. and 20 nig/kg q.i.d.). A pharniacokinetic study Clearly, when none of the regimens could produce a documented that the AUCs were not statistically peak/MIC ratio of >10:1, AUC/MIC was seen to different for the three regimens and, thus, all had the be linked to outcome (no difference between o.d. and same AUC/MIC ratio. As expected, the 0.d. dosing twice a day dosing). group had the highest peak/MIC ratio, while the The experiments above examined the issue of 20 mg/kg four times a day group had the longest response with changing drug exposure, with the MIC tinie>MIC. If AUUMIC were linked to outcome, all constant. The third set of experiments addressed three regimens should provide the same outcome. If holding the exposure constant and examining outcome tinie>MIC were linked to outcome, four times a day as a function of changing MIC. T~vomutant organisms dosing should be best, and if peak/MIC ratio is the were selected and found to have stable MICs of four linked dynamic variable, o.d. dosing should provide the 2nd eight times that of the parent strain (MICs of 1, 4 greatest survival. The results are shown in Figure 6 [33]. and 8 mg/L). Consequently, challenges were used with Two things are evident. The first is that o.d. dosing the same genetic background and virulence, but provided a significantly better outcome (y<0.001). differing in MIC. Four dosing groups (plus three The second is that the survival curves for the other control groups not displayed) were examined. A groups are, in essence, identical. This caused us to reginien of 80 nig/kg o.d. was used for each of the hypothesize that, when peak/MIC ratio does not three organisms with the different MICs, producing exceed 1O:l (as it did not in the 40 mg/kg twice a day both peak/MIC and AUC/MIC ratios for the three M 80 mg/kg every 24 h * 40 mg/kg every 12 h M 20 mg/kg every 6 h 100 H Control animals received a saline placebo injection 75 \ \\ L h vs .-Q c 50 .- s> L_ 3 v) 25 0 0 12 24 36 48 60 72 Figure 6 1)ore fractionation esperinicnt 1. The MIC of lomefoxacin For the challense organisni \vas 1 mg/L. There were i0 minials evduated per group (331. Reproduced with perniission from Antimicrobial Azents and Chemotherapy. Drusano et al: Fluoroquinolone pharmacokinetics and pharmacodynamics 2S37 100 40 mg/kg every 24 h o----o 20 mg/kg every 12 h H Control animals received a saline placebo injection 75 h .-Q 2= 50 .-9 > L 3 v) 25 0 0 8 16 24 32 40 48 56 64 72 Time (h) Figure 7 Dose fractionation experiment 2. The MIC of lomefloxacin for the challenge organism was 1 mg/L. There were 20 animals evaluated per group [33].Reproduced with permission from Antimicrobial Agents and Chemotherapy. groups in ratios of 8:4:1. The fourth group employed Further patients were studied and the data re- 2 dose of 20 mg/kg with the MIC= 1 mg/L challenge analyzed by Forrest et al [36,37]. Patients were classified organism, producing peak/MIC and AUC:/MIC ratios as having succeeded or failed clinically and niicro- equivalent to the group with 80 mg/kg 0.d. dosing and biologically but, importantly, the lower respiratory J MIC=4 mg/L challenge organism. The results are tract flora were sampled daily and the time to initial shown in Figure 8 [33]. pathogen eradication determined. The first two end- Changing organism MIC had, as expected, a pro- points were subjected to logistic regression analysis, and found effect on the outcome seen. Most importantly, the time to eradication endpoint was examined by Cox outcomes were essentially identical for the two groups proportional hazards modeling. with identical AUC/MIC and peak/MIC: ratios. This The logistic regression analysis produced statistic- indicates that it does not matter whether the MIC (a ally significant linkages between the AUC/MIC ratio more or less sensitive organism is being treated) or the and both the probability of clinical success and that of exposure (larger or smaller doses are employed for microbiological success. It should be noted, however, treatment) are changed; the outcome is driven by the that the authors also indicated that the correlation ratio of exposure to MIC (peak/MIC or AUC/MIC), between AUC/MIC ratio and peak/MIC ratio in their however it is achieved. population was 0.9. The Cox proportional hazards modeling again showed a significant linkage between Clinical studies AUC/MIC ratio and the time to bacterial eradication. The first clinical data regarding the pharmacodynamics Indeed, they also showed differences by AUC/MIC of fluoroquinolones were published by Peloquin et a1 thresholds. A patient with an AUC/MIC ratio 9125 [35]. They examined ciprofloxacin in a study ofpatients had significantly shorter times to eradication than a with lower respiratory tract infection caused by Gram- patient with a smaller AUC/MIC ratio. Further, a negative bacilli and concluded that time>MIC was patient with an AUC/MIC ratio 2250 had a time to the pharniacodynamic variable most closely linked to eradication significantly shorter than a patient with a outcome. This somewhat discordant result was modified ratio between 125 and 249 (Figure 9) [37]. These data by their finding that the emergence of resistance among are an elegant validation of the in vitro and animal I? aevuginosa was, indeed, driven by having a peak/MIC model data. However, it should be recognized that the ratio which did not exceed 10:l. analysis was retrospective. Further, it should be empha- 2S38 Clinical Microbiology and Infection, Volume 4 Supplement 2 5--0 MIC 1 mg/L 100 MIC 4 mg/L MIC 8 mg/L M MIC 1 mg/L 75 8h .-a 2= 50 .-9 > L 3 v) 25 0 a 16 24 32 40 48 56 64 72 Time (h) Figure 8 Effect of kcred MIC upon survivorship. Three isogenic organisms for which the loiiiefloxicin MICs were difii.rent served as the bacterial challenge. Thc three groups (MICs of 1. 4 and 8 iiig/L) received 80 nig/kg every 24 11. A fourth group (0) had the strain for which the MIC \\as 1 mg/L used as the challenge organisin and a dosing regimen of 20 nig/kg every 24 11. This provided the smie pedk/MIC ratio as the challenge organisin for \vhich the MI(; \\'as 4 mg/L used with dnimals treated with XO mg/kg every 24 h. There were saline-treated controls for each challenge organism (all died). T\vcnt\- animals Lvere evaluated [XI.Reproduced with permission from Antimicrobial Agents md Chcmotherapy. -=?---- AUIC 425 AUIC 125-250 AUK >250 (D E 80 I I I I I I 0 2 4 6 8 10 12 14 Days of therapy Figure 9 Time (day\ oftherapy) to bacterial eradication versus AUIC (AUC/MIC ratio) illustrated by a time-to -event (survival) plot. Therapy versus the percentage of patients remaining culture positive on that day i\ shown. The three AUIC #roup\ differed \ignificantlv (p<0.005)[37]. Reproduced with permission from Antimicrobial Agents and C:heniotherapy. Drusano et al: Fluoroquinolone pharrnacokinetics and pharrnacodynarnics 2539 sized that the outcome is specific to the population Otherwise, peak/MIC will appear to be linked to from which the patients were drawn: mainly elderly outcome, probably because of suppression of emerg- patients with very severe hospital-acquired lower ence of resistant mutants. In the Forrest study, it is clear respiratory tract infection (the majority due to Gram- that approximately 50% of the patients developed negative pathogens) on mechanical ventilation. The peak/MIC ratios of <10 [36,37]. This is because of the findings should not be extrapolated beyond this group. large mixture of patients in this study with Staphy- Most recently, Preston et a1 reported the first lococcus atireus, I! aeruginosa and Enterobacter cloacae, all prospective, multicenter trial for the explicit purpose of pathogens with higher MICs for fluoroquinolones, determining the pharmacodynamics of an anti- including ciprofloxacin. In the studies by Preston et al, microbial agent [38,39]. Levofloxacin 500 mg 0.d. was 80% of patients developed peak/MIC ratios of B 10, employed for the treatment of community-acquired almost certainly because of the common pathogens lower respiratory tract infections, complicated and encountered in the community [38,39]. Consequently, uncomplicated skin and soft tissue infections and it is not surprising that the studies of Forrest et a1 uncomplicated urinary tract infections. An optimal identified AUC/MIC ratio as the primary pharmaco- sampling analysis was performed to guide clinical dynamically linked variable for fluoroquinolones, while investigators in the acquisition of serum samples for the studies of Preston et a1 identified peak/MIC ratio later analysis. This sample design was employed at all as the dynamically linked variable [36-391. 22 centers, and 1524 serum samples were obtained It should be noted that the guidance for the from 274 patients (of 313 entered into the study). interpretation of the clinical data came from the in vitro These were analyzed for levofloxacin, and the resultant and animal model systems. In part this is because of the concentrations were subjected to a population pharma- clinical reality since it is almost unheard of for different cokinetic analysis. The population means and the schedules of administration to be tested in clinical trials Associated co-variance matrix were used to estimate the with adequate numbers of patients. Consequently, in pharmacokinetic parameter values for each patient large trials, a drug is administered on a fixed dose and through the use of maximum a posteriori probability schedule and thus there is a large degree of co-variance (MAP) Bayesian estimation. between pharniacodynamic variables. In this situation, A total of 134 patients had a microbiologically the peak/MIC ratio cannot be increased without documented infection with an MIC determined for also increasing the AUC/MIC ratio and time>MIC. levofloxacin. Logistic regression analysis was performed Therefore, it should not be surprising that different for clinical and microbiological outcome, employing trials can find different dynamically linked variables, pharmacodynamic co-variates as well as other demo- depending on the populations and pathogens studied. graphic and clinical co-variates. The result was clearcut, These differences, it should be noted, are readily with the peak/MIC ratio being significantly linked to explained by the lessons learned from the in vitro and both clinical and microbiological outcome. As in the animal model systems. Forrest study, there was a high degree of correlation between peak/MIC and AUC/MIC, with a correlation coefficient value exceeding 0.9. Further, a classification CONCLUSIONS and regression tree analysis was performed to see if a breakpoint could be ascertained. It was found that a The pharniacodynamic effect of fluoroquinolones can peak/MIC ratio of > 12 was associated with both the only be understood in a rational manner if the best clinical and the best microbiological outcomes (it information regarding them is integrated from at least should also be recognized that the response rate below four sources. Obviously, some measure of potency of the ratio of 12 still exceeded 80%). This finding agrees the drug for the organisms in question is a critical piece remarkably well with the data of Blaser et al where a of information, and most clinicians obtain this from a peak/MIC ratio of 10 was identified as producing knowledge of the drug’s MIC for clinically important optimal bacteriologic outcomes [28]. The same caveats pathogens. More than for most anti-infectives, the apply to this study as to that of Forrest et a1 [36,37]. pharmacokinetic profile of the fluoroquinolones also This was a study of community-acquired infections plays a major role in determining the observed pharma- in moderately ill patients. Its results should not be codynamics. As has been demonstrated in an animal extrapolated beyond the population studied. model of neutropenic infection, it is actually the ratio How then, can these two studies be reconciled? of these two things (some measure of drug exposure, The answer comes from the in vitro and animal model either peak concentration or AUC, and organism studies. An analysis will find AUC/MIC linked to MIC) which mostly determines the observed pharma- outcome if, on average, the peak/MIC ratio is However, it should be realized that the differences of paclitaxel alonr or 111 combination on the intracellular in pharmacokinetic profile among the fluoroquino- penetration and activity of quinolones in human neutrophils. lones are close to the differences in their niicrobio- J Antiniici-oh Chemother 1996; 38: 859-63. logical activities. For most of the fluoroquinolones, 10. Ozaki M, Komori K, Matsuda M, ct al. Uptake and iiitra- protein binding is low and almost negligible (-30%). cellular activity of NM 394, a nen. quinolone, in human ~~o1~tiioi-plioniicle~rleukocytes. Antiniicrob Agents Chenio- However, for some newer drugs, the binding is ther 1996; 4(i: 739-42. somewhat greater (-70%) and may adversely impact 11. Yanianioto T, Kucajnn~H, Hosaka M, Fukuda H, Oon~ori upon their microbiological activity. This needs to be Y, Shin~od~iH. Uptake and intracellular activity ofAM-l 155 taken into account in any comparative analysis. Finally, in phagocytic cellc. Antiniicrob Agents Cheniothcr 1996; the ability of these drugs to penetrate cells adds many 40: 2756-9. intracellular pathogens to their spectrum of activity. 12. I’ascual A, Garcia I, l’zi-e~ EJ. Uptake md intracellulx Insight into the physicochemical properties ofthe drugs activity of an optically active ofloxaciti isonier in huiiiaii which promote this cellular penetration is key for neutrophils and ticsue culture cells. Antimicrob Agentc coiiiparing these agents. Che~liothrr1990: 34: 277-80. Fluoroquinolones are the first new class of antibac- 3. I’ascual A, Garcia I, Pcrc‘i EJ. Fluoronietric iiiedsui-enirnt of terials for which a detailed pharniacodynaniic under- oflovacin uptake by human polyniorphonuclear leukocytec. Antiniicrob Agentc Cheniother 1989; 33: 653-6. standing of their action has been gained early in the 4. Nounii T, Nishida N, Minami S, Watanabe Y, Yasuda T. course of development. This understanding comes from Intracellular activity of- totuflouacni (T-3262) dgmict Sol- in vitro data, animal models of infection and, uniquely, iitii~ic/lacntrritidi$ aid abiliv to penetrate into tissue culture clinical trial data. This knowledge has led to the cells of human origin. Aiitimicrob Agents Chemothei- 1990; optimization of dose schedule for the newer drugs (e.g. 34: 949-53. demonstrating once-daily dosing is supportable), the 5. Garcia I, Pascual A, Guznian MC. I’erza EJ. Uptdke .lnd ability to compare different drugs within the class on a inti-acellular activity of spaifloxacin in human polvnior- rational basis and, perhaps most importantly, the ability phonucleai- leukocytes and tiscue culture cells. Aiitiniicrob to optimize the clinical and microbiological outcome for Agetits Chriiiothei- 1992; 36: 1053-6. seriously ill patients on the bask of organism MIC and 6. I’ascual A, Garcia I, Uallesta S, I’erea EJ. Uptake and observed or expected serum drug concentrations. inti-acellular activity of trovaflosacin in hunian phagocytec and tissue-cultured epitheli.11 cell\. Antimicroh Agentc Che~liother1997; 41: 274-7. 17. Taira K, Koga H, Kohiio 5. Accnmulation of ‘1 newly References developed fluoroquinolonr, OPC-I 71 16, by human poly- 1. Bryskier A, Chantot J-F. Classificatmn ‘and ctructure-activity morphonuclear leukocyte?. Antiniicrob Agents Chcniother relrltioiiships of fluoroquinoloner. I)rugs 1995; 49(euppl 2): 1993; 37: 1877-X1. I(,-28. 18. Carlier ME, Scorneaux B. Zciizbergh A, Ikcnottes JF, 2 I )ecrC I), Ikrgogie-BCrCziii E Pharrnacokinetics of quino- Tulkens PM. Cellular uptake, localization and activity of lones \vith special referencz to the respiratory tree. Antiniicrob fluoroyuinolories in uninfected and infected macrophagee. Cheniother 1993; 31: 331-43. J Antiinicrob Chemothci- 1990; 26(\uppl H): 27-39. 3. C.irc 0. I’harnincokineticc of mtiblotlcs 111 tissue and ticcue 19. Cao CX, Silverstein SC, Neu HC, Steinberg TH. J77-1 fluids: a review. Scand J Infect l)i, 1991: 74(suppl): 23-33. macrophagec secrete antibiotics \’n organic anion trans- 4. Kneer J. Kelevmce of antibiotic ticwe penetration in treat- porters. J IiifDis 1992; 165: 322-8. ing recpiratory tract infections. Respiration 199.3; hO(cuppl): 20. Kudiii LIE, Gao PX, Cao CX, Neu HC, Silver\tein SC. 32-7. C;emfibroril enhances the listerimdal effects of fluoro- 5. Uald~vinLIR, Wise K, Andrews JM, Gill M, Honeybourne quiiiolorie mtibioticc in JT-4 iiiacrophagcs. J Exp Mcd 1 992; D. Compar~itivebronchocilveolar concentrationc of cipro- 176: 1439-47. floxacin arid lonicflosacin following oral adniini~ation. 21. Van Kensburg CEJ, Jooiie G. Andereon R Iiiteractionr of Kespir Med 1993: X7: 59.5-601 the oxygen-dependent antimicrobial system of the human 6. Honcybouriie L3, Andrews JM. Jevon) G,Brenward NP, ~ieutrophil\\-ith diflox 24. Garcia I, Pascual A, Perea EJ. Effect of several antimicrobial rieutropenic rat model of Psercdornorms sepsis. Antimicrob agentr on ciprofloxacin uptake by human neutrophils. Eur J Agents Chemother 1993; 37(3): 483-90. Clin Microbiol Infect Dis 1992; 11: 260-2. 34. Johnson D. Use of discriminative models of Pseudonionns 25. Pascual A, Garcia I, Conejo MC, Perea EJ. Effects of aenqinosa bacteremia in granulocytopenic rats for testing antimicrobial and antineoplastic drugs on the uptake of spar- antimicrobial efficacy. Eur J Clin Microbiol 1985; 4(2): floxacin by human neutrophils. J Antiniicrob Chemother 207-12. 1994; 34: 171-4. 35. Peloquin CA, Cunibo TJ, Nix DE, Sands MF, Schentag 26. Sorgel F, Kinzig M. Pharmacokinetics of gyrave inhibitorr, JJ. Evaluation of intravenous ciprofloxacin in patients with Parts 1 & 2. Am J Med 1993; 94(3A): 44-69. nosocomial lower respiratory tract infections: impact of 27. Baldwin DR, Honeybourne D, Wise R. Pulmonary plasma concentrations, organism, minimum inhibitory con- disposition of antimicrobial agents: in tho observations and centration, and clinical condition on bacterial eradication. clinical relevance. Antimicrob Agents Chemother 1992; 36: Arch Intern Med 1989; 149(10): 2269-73. 1176-80. 36. Forrest A, Ballow CH, Nix DE, Birmingham MC, Schentag 28. Blaser J, Stone BB, Groner MC, Zinner SH. Comparative JJ. Development ofa population pharmacokinetic model and study with enoxacin and netilmicin in a pharmacodynamic optimal sampling strategies for intravenous ciprofloxaciii. model to determine importance of ratio of antibiotic peak Antimicrob Agents Chemother 1993; 37(5): 1065-72. concentration to MIC for bactericidal activity and emer- 37. Forrest A, Nix DE, Ballow CH, Goss TF, Birmingham MC, gence of reyistance. Antimicrob Agents Chemother 1987; Schentag JJ. Pharmacodynamics of intravenous ciprofloxacin 31(7): 1054-60. in seriously ill patients. Antiniicrob Agents Chemother 29. Dudley MN, Mandler HI], Gilbert 11, Ericson J, Mayer KH, 1993; 37(5): 1073-81. Zinrier SH. Pharniacokinetics and pharmacodynamics of 38. Preston SL, Drusano GL, Berman AL, et al. Levofloxacin (L) intravenous ciprofloxacin: studm in vivo and in an in vitro population pharmacokinetics (PK) and creation of a demo- dynamic model. Am J Med 1987; 82(suppl 4A): 363-8. graphics model for prediction of individual drug clearance 30. Leggett JE, Ebert S, Fantin B, Craig WA. Comparative in patients with serious community acquired infection. dose-effect relations at several dosing intervals for beta- [abstract A121. In: 36th Interscience Conference on lactam, aminoglycoside and quinolone antibiotics against Antimicrobial Agents and Chemotherapy, New Orleans, Gram-negative bacilli in murine thigh-infection and pneu- Louisiana. Wa5hington DC: American Society for Micro- monitis models. Scand J Infect Dis 1991; 74: 179-84. biology, 1996: 3. 31. Fantin B, Leggett J, Ebert S, Craig WA. Correlation between 39. Preston SL, Drusano GL, Berman AL, et al. Prospective in vitro and in vivo activity of antimicrobial agents against development of pharmacodynamic relationships between Gram-negative bacdli in a murine infection model. Anti- measures of levofloxacin (L) exposure and measures of patient microb Agent5 Chemother 1991; 35(7): 1413-22. outcome [abstract A98). In: 36th Inter~ienceConference 32. Vogelman B, Gudmundsson S, Leggett J, Turnidge J, Ebert on Antimicrobial Agents and Chemotherapy, New Orleans, S, Craig WA. Correlation of antimicrobial pharmacokinetic Louisiana. Washington, DC: American Society for Micro- parameters with therapeutic efficacy in an animal model. biology, 1996: 19. J Infect Dis 1988; 158: 831-47. 33. Drusano GL, Johnson DE, Rosen M, et al. Pharmaco- dynamics of a fluoroquinolone antimicrobial agent in a