SUPPLEMENT ARTICLE. Paul G. Ambrose

Transcription

1 SUPPLEMENT ARTICLE Use of Pharmacokinetics and Pharmacodynamics in a Failure Analysis of Community-Acquired Pneumonia: Implications for Future Clinical Trial Study Design Paul G. Ambrose Institute for Clinical Pharmacodynamics, Ordway Research Institute, Albany, and School of Pharmacy and Pharmaceutical Sciences, University at Buffalo, Buffalo, New York Regulatory uncertainty has led to the cessation of antimicrobial agent drug development for some communityacquired respiratory tract infections. This uncertainty stems from the fact that many people, including individuals within government, are unsure about the magnitude of drug effect relative to no treatment in patient populations perceived to be at low-to-moderate risk of mortality. Without such information, it is not possible to power noninferiority studies, which, in turn, necessitates the need for superiority study designs. Moreover, many believe that current categorical clinical trial end points (success or failure), determined 7 14 days after therapy, are insensitive measures of drug benefit and that new outcome measures are needed. To date, characterization of the magnitude of treatment effect relative to that of no treatment has not been accomplished through traditional examinations of existing clinical trial databases or evaluations of historical data. However, pharmacokinetic-pharmacodynamic analyses of existing clinical trial data may provide a new context to inform the debate. Herein, we examine pharmacokinetic-pharmacodynamic relationships for efficacy derived from recent clinical drug development programs involving agents for the treatment of community-acquired respiratory tract infections. Through a collective understanding of these data, it may be possible to estimate the no-treatment response rate without exposing patients to any risk incurred by conducting clinical trials with alternative designs (e.g., placebo-controlled studies or studies using suboptimal dose ranges or comparator regimens). Finally, the value and potential of adding continuous, numeric clinical trial end points to those currently used are discussed. For the past few years, a great debate has absorbed the attention and consumed the energies of academics, clinicians, drug developers, and drug regulators alike, regarding the proper study design for the evaluation of antimicrobial agents for the treatment of communityacquired pneumonia (CAP). The debate centers on the need for a reliable and reproducible estimate of the magnitude of drug effect, relative to that of no treat- Reprints or correspondence: Dr. Paul G. Ambrose, Institute for Clinical Pharmacodynamics, Ordway Research Institute, 150 New Scotland Ave., Albany, NY (PAmbrose-ICPD@OrdwayResearch.org). Clinical Infectious Diseases 2008; 47:S by the Infectious Diseases Society of America. All rights reserved /2008/4711S3-0021$15.00 DOI: / ment, especially for patient populations perceived to be at low-to-moderate risk of mortality [1]. Without such an estimate, some individuals think that it is not reasonable to conduct noninferiority studies, but that instead superiority clinical trials, or possibly placebo-controlled trials, should be conducted. Others think that the potential size and cost of superiority studies render them unfeasible and/or that designs using placebo-control arms, suboptimal comparator control arms, or suboptimal dose ranges are unethical, given the known mortality risk for this patient population. Many individuals on both sides of this debate agree that moresensitive, clinically relevant measures of drug benefit are needed to better distinguish drug effects. However, such measures have yet to be identified and fully dis- PK-PD Analyses as Guide to CAP Study Design CID 2008:47 (Suppl 3) S225

2 Table 1. Bacteriological response rate in children with acute otitis media, according to faropenem dose, in a double-tap study [3]. Faropenem dose, mg/kg Success rate, no. with bacteriological response/total treated (%) /29 (55.2) 15 42/60 (70.0) 30 37/43 (86.0) 40 46/49 (93.8) cussed by interested parties. The resultant regulatory uncertainty has led to a dramatic reduction in the development of antibacterial agents for treatment of community-acquired respiratory tract infections, which assuredly is not in the best interest of public health. Pharmacokinetic-pharmacodynamic (PK-PD) analyses have long served as a tool for preclinical drug evaluation, dose and dosing interval selection for early-stage clinical development, and confirmation of the appropriateness of late-stage doseregimen decisions [2]. The use of PK-PD analyses for these purposes has had an especially prominent role in drugs developed for the treatment of infectious diseases. Although a drug may fail in clinical trials or for an individual patient for many reasons, one obvious potential explanatory factor is suboptimal drug exposure. Given that suboptimal drug exposure is a PK-PD issue, PK- PD analyses may provide a new context to inform the debate that consumes us today. Herein, we examine PK-PD relationships derived from recent clinical drug-development programs involving agents for the treatment of community-acquired respiratory tract infections. Special attention is paid to one program, in which the challenge drug did not meet the criteria for noninferiority relative to the active comparator. Through a collective understanding of these data, it is possible to estimate the no-treatment response rate in the context of low or no drug exposure. Moreover, such an approach avoids the risk to patients incurred by conducting clinical trials with alternative designs (e.g., placebo-controlled studies or studies purposely using suboptimal dose ranges or comparator regimens). DOSE- AND EXPOSURE-RESPONSE RELATIONSHIPS Phase 2 dose-ranging studies evaluating antimicrobial agents for the treatment of community-acquired respiratory tract infections rarely identify efficacy differences between dose cohorts. This is primarily because of 3 reasons. The first and second reasons relate to the relatively small numbers of patients enrolled in such trials and the even smaller numbers of patients who meet the protocol-defined criteria for treatment failure. The third reason is that dose is an extremely insensitive measure of an individual patient s drug exposure, because it does not account for interindividual PK-PD variability. Despite these limitations, there are examples in which it has been possible to discern a dose-response relationship for antimicrobial agents evaluated for the treatment of communityacquired respiratory tract infections [3, 4]. Tables 1 and 2 show the response rates by dose cohorts receiving faropenem and grepafloxacin for the treatment of acute otitis media and acute exacerbations of chronic bronchitis, respectively. These data are important, because they provide a strong signal that the magnitude of drug exposure is, not unexpectedly, related to therapeutic response. Unfortunately, even when a dose-response relationship for efficacy is identified, the inferences that can be made are limited. For instance, dose-response relationships may not identify the maximally effective dose but instead may provide efficacy information only for the dose levels studied. However, when PK samples are collected during the course of a clinical trial, it is possible to estimate individual patient drug exposures, which, in turn, can be used to evaluate exposure-response relationships, resulting in a greater sensitivity to detect between-regimen differences, compared with that of dose-response analyses. In fact, exposure-response analyses often detect drug effects in circumstance for which no clear dose-response effect can be discerned. The identification of exposure-response relationships allows one to predict drug effect for regimens not studied within the dose range evaluated. Moreover, asymptotes (sometimes extrapolated) for such relationships provide estimates of drug effect as exposure increases toward the maximum treatment effect or as exposure decreases toward the effect of no treatment. Figure 1 shows the relationship between clinical response and drug exposure, as measured by the ratio of the area under the serum concentration-time curve to the MIC of the drug to the pathogen (AUC:MIC ratio) for 76 patients with acute exacerbations of chronic bronchitis who received treatment with grepafloxacin [5]. In many ways, these data are an exten- Table 2. Clinical response rate in adult patients with acute exacerbations of chronic bronchitis associated with Streptococcus pneumoniae, according to treatment [4]. Treatment Success rate, no. with clinical response/total treated (%) Grepafloxacin 400 mg daily 29/40 (72) 600 mg daily 35/41 (85) Comparator 38/44 (86) S226 CID 2008:47 (Suppl 3) Ambrose

3 Figure 1. Plot of a Hill-type pharmacokinetic-pharmacodynamic model for the percentage probability of clinical success versus the ratio of the area under the serum concentration-time curve (AUC) to MIC (AUIC mp )in patients with acute exacerbations of chronic bronchitis. AUC:MIC values for individual patients are plotted on the X-axis, and the group s percentage probability of clinical success is plotted on the Y-axis, with random jitter shown for graphical clarity. Reprinted from [5], with permission from J. J. Schentag and Oxford University Press. sion of the dose-response data presented in table 2. Of particular relevance to this discussion are the response rates associated with the upper and lower asymptotes. Note that, as drug exposure approaches zero, the model-predicted response rate is 69%, which suggests that 7 of 10 patients would show a favorable clinical response at the test-of-cure visit without receiving of antimicrobial therapy. Conversely, the model-predicted maximal response rate at clinically achievable drug exposures is 96%. The difference between these 2 response rates, 27% ( 96% 69% ), can be thought of as an estimate of the maximum treatment effect that is, the maximal benefit of grepafloxacin therapy for this particular patient population. Such an estimate, if reliable and reproducible, may provide the missing critical data piece for powering noninferiority studies and thus obviate the need for superiority designs. To this end, several published exposure-response analyses involving patients with CAP are particularly instructive [6 11]. For instance, Preston et al. [6] evaluated the relationship between different PK-PD measures with regard to both clinical and microbiological outcomes for patients given treatment with levofloxacin for urinary tract, pulmonary, and skin and soft tissue infections. Patients categorized as having pulmonary infections included patients with CAP, acute exacerbations of chronic bronchitis, or acute bacterial sinusitis. Results of this evaluation yielded final logistic regression models for clinical and microbiological outcome in which the ratio of the peak plasma concentration to the MIC of the drug to the pathogen, evaluated as a continuous variable, was predictive of outcome. In the case of clinical outcome, the final logistic regression model also contained site of infection, which translated to exposure-response relationships with y-intercepts (i.e., the extrapolated model-predicted effect in the absence of drug exposure) that differed by disease state. As shown in figure 2, the model-predicted probability of clinical success was 72.5% as drug exposure approached zero for patients with pulmonary infections, which is concordant with that for patients with acute exacerbations of chronic bronchitis who received treatment with grepafloxacin [5]. Given that the exposure-response analyses are typically based on 1 or 2 studies in which the collective number of patients who experience treatment failure is limited [7 10], greater insight can be gained by pooling data from clinical trials for the same indication. The power that is gained by pooling such data is particularly useful when exposure-response relationships are likely to differ by the genus of the pathogen (as is often the case). In these circumstances, pooling of data provides the best opportunity to evaluate relatively homogeneous data in which Figure 2. Plot of a logistic regression pharmacokinetic-pharmacodynamic model for the probability of clinical success versus the ratio of peak plasma concentration to MIC (peak/mic ratio). Breakpoints for the exposure measures indicate the value for which there is a significantly increased probability of successful outcome, as determined by classification and regression tree analysis. Shown are the probability curves for successful clinical response, including peak/mic ratio and 3 infection sites. The pulmonary infection site included patients with communityacquired pneumonia, acute exacerbations of chronic bronchitis, and acute bacterial sinusitis. Reprinted from [6], with permission from the American Medical Association. Copyright 1998, American Medical Association. All rights reserved. PK-PD Analyses as Guide to CAP Study Design CID 2008:47 (Suppl 3) S227

4 the infections are attributable to a single pathogen or pathogen group. Given that, for CAP, patient morbidity and mortality are primarily associated with Streptococcus pneumoniae, exposure-response relationships that focus on this pathogen are of particular interest. In a recent analysis [11], exposure-response data were collected for 249 patients with pneumococcal infection from 14 clinical trials for community-acquired respiratory tract infections, in which patients received standard daily doses of ciprofloxacin, gatifloxacin, garenoxacin, gemifloxacin, grepafloxacin, or levofloxacin. The percentage of patients with each disease was 37% with acute bacterial sinusitis, 36% with CAP, and 27% with acute exacerbation of chronic bronchitis. Data that were gathered from 10 clinical trials in which patients with CAP received daily doses of gatifloxacin (400 mg; 12 patients), garenoxacin (400 mg; 7 patients), gemifloxacin (320 mg; 34 patients), or levofloxacin (500 mg; 36 patients), clinical and microbiological success was observed for 93% of patients. Classification and regression tree analysis demonstrated that a freedrug AUC:MIC ratio of 33.8 was predictive of a successful response. The probability of a successful clinical response was 95% for values at and above this breakpoint and was 67% for values below this breakpoint. Univariate logistic regression models demonstrated that the free-drug AUC:MIC ratio categorized using this breakpoint was a statistically significant predictor of clinical response (OR, 9.00; 95% CI, ; P p.03) and microbiological response (OR, 9.38; 95% CI, ; P p.03). As shown in figure 3A and 3B, when the free-drug AUC:MIC ratio was evaluated as a continuous variable by using a Hill-type PK-PD model for this subset of patients, a steep exposure-response relationship was evident, with a y-intercept of 69% and a maximal model-predicted effect of 95% for both clinical and microbiological response. These pooled data, gathered across multiple trials, provided greater power to detect exposure-response relationships of efficacy for an indication and pathogen of interest, relationships that ultimately proved to be consistent with those found on the basis of nonclinical infection models [12]. It is important to note that not all analyses involving patients with CAP have identified relationships between drug exposure and response, categorized as success or failure at 7 14 days after therapy [8, 10]. The lack of such a finding could be because of a variety of reasons, including (1) there was no exposureresponse relationship (i.e., the drug lacks benefit), (2) the exposures observed for patients are in the upper, flat portion of an exposure-response relationship curve, and (3) the end point, which is typically categorized as success or failure at 7 14 days after the end of therapy, was an end point too insensitive to capture the drug s benefit. Which reason or reasons are explanatory is difficult to discern. It is recognized that pharmaceutical companies endeavor to select dose regimens with high probabilities of clinical success, and, to quote Shakespeare s Hamlet, ay, there s the rub. That is, if dose regimens are selected so that few patients experience treatment failure, how can one deduce which of the aforementioned reasons are likely explanations for the failure to identify an exposure-response relationship? In such a circumstance, 2 additional types of data are important to consider. First, dosing regimens are often based on prior knowledge gained from animal infection models. If the exposures observed in humans are far greater than those required for efficacy in animals, it supports the notion that the human exposures lie on the upper, flat portion of an exposure-response curve. For example, no exposure-response relationship was identified for Figure 3. Plots of Hill-type pharmacokinetic-pharmacodynamic models for the probability of clinical response (A) and microbiological success (B) versus the free-drug ratio of area under the serum concentration-time curve to MIC (AUC:MIC ratio) in patients with community-acquired pneumonia. AUC:MIC ratios for individual patients are plotted on the X-axis, and the group s probability of clinical success is plotted on the Y-axis, with random jitter shown for graphical clarity [11]. S228 CID 2008:47 (Suppl 3) Ambrose

5 Table 3. Clinical response success rate, according to prior effective antimicrobial therapy in hospitalized patients with community-acquired pneumonia given treatment with either daptomycin or ceftriaxone. Prior effective therapy Daptomycin No. with clinical response/ total treated Success rate, % Ceftriaxone No. with clinical response/ total treated Success rate, % 95% CI a Yes 88/ / to11.5 No 205/ / to 6.0 NOTE. Note that the success rate difference between the daptomycin-treated cohorts is 15.3% ( 90.7% 74.4% ), and the maximum difference between the ceftriaxone-treated cohort and the daptomycin-treated cohort is 12.4% ( 87.8% 75.4% ). Data were pooled from 2 randomized, double-blind studies and represent the clinically evaluable population in a post hoc analysis. Reprinted from [13], withpermission from the University of Chicago Press. a The 95% CI for the difference in cure rates between daptomycin and ceftriaxone. garenoxacin treatment among patients with community-acquired respiratory tract infections [8]. When one considers that, for S. pneumoniae and fluoroquinolones, a 99% reduction in bacterial burden and 100% animal survival have both been demonstrated to be associated with free-drug AUC:MIC ratios of [12] and that only a single patient in the garenoxacin clinical program had a free-drug AUC:MIC ratio for pneumococci!200 [8], it is not surprising that no obvious exposureresponse relationship was identified. In such a circumstance, the consideration of continuous, numeric end points may provide a better indication of drug effect. For instance, as with the above-described example of garenoxacin, tigecycline exposures among patients with CAP far exceeded the thresholds necessary for efficacy in pneumococcal animal infection models, and, in data from clinical trials, no relationship between drug exposure and response, categorized as success or failure at 7 14 days after therapy, was detected [10]. However, a signal of drug effect was observed when time-to fever resolution, stratified by drug exposure, was evaluated. Fever in patients with AUC:MIC ratios tended to resolve faster than that in patients with lower exposures (median difference, 12 h; P p.05). Data such as these further support the conclusion that the failure to distinguish a relationship between tigecycline exposure and response, defined categorically as success or failure at days after therapy, was a result of both the selection of a dose regimen with nearoptimal efficacy and the use of relatively insensitive primary end points, rather than a lack of drug effect. Clearly, the exposure-response analyses presented here are based on a small number of treatment failures, which results in relatively wide 95% CIs around the model-predicted y-intercepts. Although one may feel buoyed by the concordance of the collective evidence presented here, additional analyses are possible. Several analyses should be considered, including the pooling of data across New Drug Applications in which patient PK samples were collected to derive a more robust sample size for analysis; the use of demographic models to predict drug exposures in patients from whom PK samples were not collected, when appropriate; and the use of surrogates for exposure, such as dose/weight/mic, when PK data for patients are not available. It is important to note that these data already exist, because, during the last decade, pharmaceutical companies have conducted numerous studies involving patients with CAP. DAPTOMYCIN VIGNETTE Daptomycin and ceftriaxone were evaluated in 2 phase 3, randomized, double-blind clinical trials that enrolled adult hospitalized patients with CAP. When the first of these 2 studies was completed, it became apparent that daptomycin did not meet the predetermined criteria for noninferiority relative to ceftriaxone, and the enrollment of patients for the second study was terminated [13]. Subsequently, results from a pneumococcal murine pneumonia model study confirmed that daptomycin efficacy was poor relative to that of ceftriaxone [14]. Ultimately, daptomycin was shown to interact with and suffer inactivation by pulmonary surfactant, which resulted in daptomycin in vitro MIC values increasing 100-fold [14]. Table 3 shows the clinical cure rates in the 2 studies combined, stratified by the receipt of prior effective antimicrobial therapy [14]. Of particular relevance to this discussion is the difference in response rates (12.4% 15.3%) among patients receiving daptomycin who had not received prior effective therapy, relative to those among all patients receiving ceftriaxone or daptomycin who did receive prior effective antimicrobial therapy. It is also important to note (data not shown) that the persistance of S. pneumoniae in sputum samples was longer for patients receiving daptomycin who had not received prior effective antimicrobial therapy than for patients receiving ceftriaxone ( P.023) [14]. When one remembers that the AUC: MIC ratio is the PK-PD measure associated with daptomycin efficacy and that the daptomycin MIC value increases 100-fold in the presence of pulmonary surfactants, it becomes apparent that drug exposure in the daptomycin cohort without prior effective antimicrobial therapy functionally approached zero. PK-PD Analyses as Guide to CAP Study Design CID 2008:47 (Suppl 3) S229

6 Given that it is possible that daptomycin therapy may have provided some benefit (such as prevention of bacteremia), one might consider the difference in response rates, 12.4% 15.3%, as a conservative estimate of treatment effect in this patient population. That is, if daptomycin did provide some benefit, then the magnitude of the treatment effect would be 112.4% 15.4%. CONTINUOUS, NUMERIC END POINTS Even if the above-suggested approach of using data from exposure-response analyses to power noninferiority studies proves fruitful, it is likely that we still have a problem with clinical trial end point. Perhaps we need better outcome measures to capture specific response elements, rather than composite measures of success or failure at 7 14 days after therapy. Studies in the 1950s often evaluated drug concentrations, appetite, pain, cough, fever, pulse rate, WBC counts, radiographic findings, and/or the patient s sense of well being over time. These continuous, numeric end points are more sensitive than are categorical end points, which results in better power to distinguish regimen differences. In the current paradigm (figure 4), an event (e.g., cure) occurring 2 weeks after therapy is treated in the same manner as that event occurring 2 days after the start of therapy. The loss of such fundamental timerelated information regarding between-regimen differences is critical to the patient, physician, and society. These differences are not theoretical but are real, because between-regimen and between-exposure differences have been demonstrated in a variety of PK-PD analyses of clinical data from studies of community-acquired respiratory tract infections [5, 10]. Using continuous, numeric end points, we can evaluate the impact of drug exposure on time to event, as illustrated by the above-mentioned example involving tigecycline and time to fever resolution [10]. Moreover, we can have an impact on the number of patients required to detect between-regimen differences. The sample size required to detect a 24-h difference in the median time to an event with an a of 0.05 and 90% power would be 50 evaluable patients per group [15]. Finally, timeto-event analyses may finally help define the optimal length of antimicrobial therapy and, thus, provide much more informative data from phase 2/3 clinical trials. COLLECTIVE MEANING AND COMMON SENSE In aggregate, the data presented above provide evidence that there is a drug benefit for patients with community-acquired respiratory tract infections, including pneumonia, and that the use of PK-PD analyses can provide a paradigm for determining the magnitude of treatment effect without the use of alternative designs (e.g., placebo-controlled studies or studies using suboptimal dose-ranges or comparator regimens). On the basis of Figure 4. Schematic depicting the current paradigm of clinical trial design for community-acquired respiratory tract infections. Note that the differences between treatment regimens are not captured during the testof-cure visit interval. the data presented, the treatment effect ranged from 12.4 to 27%. Under the assumption that, in noninferiority studies, we want a high probability of maintaining not less than one-half of the midpoint of this range of treatment effect, these data suggest a noninferiority margin of 10%. Clearly, additional exposure-response analyses are possible, and they would provide greater precision in the point estimate of the treatment effect and would allow a better estimate of the noninferiority margin. Finally, the addition of continuous, numeric end points to future clinical trials holds promise for an evaluation of the impact of drug exposure on time to event, thereby reducing the number of patients required to detect between-regimen differences and, finally, defining the optimal length of therapy. Given the improved sensitivity and power associated with such end points, it is likely that, after experience and knowledge are gained, we will be able to use these as primary end points in superiority studies. In the meantime, common sense would seem to dictate the continued use of noninferiority designs with the inclusion of continuous, numeric end points as secondary end points. It is time to plan and conduct the aforementioned exposure-response analyses, with input from statistical, pharmacometric, and clinical specialists, so that a new era of drug development for treatment of community-acquired respiratory tract infections can begin. Acknowledgments I acknowledge the contributions of my colleagues to the ideas presented herein, including Drs. Sujata M. Bhavnani, George L. Drusano, and Alan Forrest. Supplement sponsorship. This article was published as part of a supplement entitled Workshop on Issues in the Design and Conduct of Clinical Trials of Antibacterial Drugs for the Treatment of Community-Acquired Pneumonia, sponsored by the US Food and Drug Administration and the Infectious Diseases Society of America. S230 CID 2008:47 (Suppl 3) Ambrose

7 Potential conflicts of interest. References P.G.A.: no conflicts. 1. US Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research. Guidance for industry. Antibacterial drug products: use of noninferiority studies to support approval. Draft guidance. October Available at: Accessed 7 October Ambrose PG, Bhavnani SM, Rubino CM, et al. Pharmacokineticspharmacodynamics of antimicrobial therapy: it s not just for mice anymore. Clin Infect Dis 2007; 44: Arguedas A, Dagan R, Wang E, et al. Dose-efficacy of faropenem in the treatment of acute otitis media using double tympanocentesis in obtained middle ear fluid [abstract G-987]. In: Program and abstracts of the 47th Interscience Conference on Antimicrobial Agents and Chemotherapy (Chicago). Washington, DC: American Society for Microbiology, 2007: Raxar (grepafloxacin hydrochloride tablets) clinical pharmacology. Available at: Accessed 7 October Forrest A, Chodash S, Amantea MA, Collins DA, Schentag JJ. Pharmacokinetics and pharmacodynamics of oral grepafloxacin in patients with acute bacterial exacerbations of chronic bronchitis. J Antimicrob Chemother 1997; 40(Suppl A): Preston SL, Drusano GL, Berman AL, et al. Pharmacodynamics of levofloxacin: a new paradigm for early clinical trials. JAMA 1998; 279: Ambrose PG, Grasela DM, Grasela TH, Passarell J, Mayer HB, Pierce PF. Pharmacodynamics of fluoroquinolones against Streptococcus pneumoniae in patients with community-acquired respiratory tract infection. Antimicrob Agents Chemother 2001; 45: Van Wart S, Phillips L, Ludwig EA, et al. Population pharmacokinetics and pharmacodynamics of garenoxacin in patients with communityacquired respiratory tract infections. Antimicrob Agents Chemother 2004; 48: Lodise TP, Preston S, Bhargava V, et al. Pharmacodynamics of an 800- mg dose of telithromycin in patients with community-acquired pneumonia caused by extracellular pathogens. Diagn Microbiol Infect Dis 2005; 52: Rubino CM, Bhavnani SM, Forrest A, Korth-Bradley J, Ambrose PG. Pharmacokinetic-pharmacodynamic analysis for efficacy of tigecycline in patients with hospital- or community-acquired pneumonia [abstract A-584]. In: Program and abstracts of the 47th Interscience Conference on Antimicrobial Agents and Chemotherapy (Chicago). Washington, DC: American Society for Microbiology, Bhavnani SM, Forrest A, Hammel JP, Drusano GL, Rubino CM, Ambrose PG. Pharmacokinetics-pharmacodynamics of quinolones against Streptococcus pneumoniae in patients with community-acquired respiratory tract infections. Diagn Microbiol Infect Dis 2008; 62: Craig WA, Andes DR. Correlation of the magnitude of the AUC 24 /MIC for 6 fluoroquinolones against Streptococcus pneumoniae with survival and bactericidal activity in an animal model [abstract A-289]. In: Program and abstracts of the 40th Interscience Conference on Antimicrobial Agents and Chemotherapy (Toronto). Washington, DC: American Society for Microbiology, Pertel PE, Bernardo P, Fogarty C, et al. Effects of prior effective therapy on the efficacy of daptomycin and ceftriaxone for the treatment of community-acquired pneumonia. Clin Infect Dis 2008; 46: Silverman JA, Mortin LI, VanPraagh AD, Li T, Alder J. Inhibition of daptomycin by pulmonary surfactant. J Infect Dis 2005; 191: Ambrose PG, Anon J, Owen JS, et al. Use of pharmacodynamic end points for the treatment of acute maxillary sinusitis. Clin Infect Dis 2004; 38: PK-PD Analyses as Guide to CAP Study Design CID 2008:47 (Suppl 3) S231