Pharmacokinetic–pharmacodynamic target attainment analysis of doripenem in infected patients

International Journal of Antimicrobial Agents 33 (2009) 276–279

Contents lists available at ScienceDirect

International Journal of Antimicrobial Agents journal homepage: http://www.elsevier.com/locate/ijantimicag

Short communication

Pharmacokinetic–pharmacodynamic target attainment analysis of doripenem in infected patients Kazuro Ikawa a,∗ , Norifumi Morikawa a , Shinya Uehara b , Koichi Monden b , Yoshiaki Yamada c , Nobuaki Honda c , Hiromi Kumon b a

Department of Clinical Pharmacotherapy, Graduate School of Biomedical Sciences, Hiroshima University, 1-2-3 Kasumi, Minami-ku, Hiroshima 734-8551, Japan Department of Urology, Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama University, 2-5-1 Shikata-cho, Okayama 700-8558, Japan c Department of Urology, Aichi Medical University School of Medicine, Nagakute, Aichi 480-1195, Japan b

a r t i c l e

i n f o

Article history: Received 14 August 2008 Accepted 29 August 2008 Keywords: Carbapenem Population pharmacokinetics Monte Carlo simulation Probability of target attainment Breakpoint

a b s t r a c t This study was a pharmacokinetic (PK)–pharmacodynamic (PD) target attainment analysis of doripenem. Drug concentration data in plasma (115 samples) and urine (61 samples) from 18 infected patients were concurrently analysed to develop a more accurate population PK model for doripenem. In the final PK model, creatinine clearance (CLCr ) was the most significant covariate: CLr (L/h) = 0.137 × CLCr ; CLnr (L/h) = 2.49; V1 (L) = 8.29; Q (L/h) = 8.10; and V2 (L) = 9.37, where CLr and CLnr are the renal and non-renal clearances, V1 and V2 are the volumes of distribution of the central and peripheral compartments, and Q is the intercompartmental (central–peripheral) clearance. Based on the PK model, a Monte Carlo simulation predicted the probabilities of attaining the bactericidal exposure target (40% of the time above the minimum inhibitory concentration (MIC)) in plasma and defined the PK–PD breakpoints (the highest MIC values at which the target attainment probabilities were ≥90%). The breakpoint for 500 mg every 8 h (q8h) (1-h infusion) with a CLCr of 80 mL/min (1 ␮g/mL) corresponded to those for 250 mg q8h with a CLCr of 40 mL/min and 250 mg every 12 h with a CLCr of 20 mL/min. Prolonging the infusion time was a more effective strategy than dose escalation to increase the breakpoint. These results provide guidance for constructing a PK–PD-based strategy for dosing guidance for tailoring doripenem regimens. © 2008 Elsevier B.V. and the International Society of Chemotherapy. All rights reserved.

1. Introduction Doripenem is a new carbapenem with a broad spectrum of activity against Gram-positive and Gram-negative bacteria. Approved in Japan (July 2005) and the USA (October 2007) for treating urinary tract infections (UTIs) and intra-abdominal infections in adult patients [1], the dosing regimen for doripenem is usually 500–1500 mg daily in two to three divided doses by 0.5-h or 1-h infusion. Antibacterial regimens should be tailored and optimised on a case-by-case basis, but information is very limited on choosing the appropriate doripenem dosage for individual situations. This choice entails actively considering such factors as the pathophysiological state of the patient and the doripenem susceptibility of the particular pathogen involved.

∗ Corresponding author. Tel.: +81 82 257 5296; fax: +81 82 257 5320. E-mail address: [email protected] (K. Ikawa).

Integration of pharmacokinetic (PK) and pharmacodynamic (PD) targets derived from PK and exposure–response data can be utilised to support antibacterial dosage choice and optimise the dosing regimen. This PK–PD target attainment analysis, particularly using a population PK modelling and Monte Carlo simulation, has also been applied to doripenem. However, in earlier analyses, drug concentration data were obtained from healthy subjects [2–4], the PKs of whom are often different from those of infected patients. The population PK modelling [2,5] did not investigate any covariates that affected PKs in their populations, although doripenem was reported to primarily be eliminated renally [6]. Therefore, for an appropriate PK–PD target attainment analysis, characterising the population PKs of doripenem in infected patients, particularly focusing on the drug urinary excretion process, is important. The current study first developed a population PK model in infected patients using both plasma and urine concentration data. Based on the developed model, the study then predicted the probability of target attainment (PTA) profiles for various doripenem regimens in typical patient populations.

0924-8579/$ – see front matter © 2008 Elsevier B.V. and the International Society of Chemotherapy. All rights reserved. doi:10.1016/j.ijantimicag.2008.08.031

K. Ikawa et al. / International Journal of Antimicrobial Agents 33 (2009) 276–279

2. Materials and methods 2.1. Study patients Doripenem PK data were obtained from the Department of Urology, Okayama University Hospital, Okayama, Japan, and the Department of Urology, Aichi Medical University School of Medicine, Nagakute, Japan. The two study institutions reviewed and approved the protocol and informed consent form. Adult patients with UTIs and prostatitis received a 0.5-h infusion of 250 mg doripenem. Venous blood samples were drawn in heparinised tubes 0.5–12 h after the start of the infusion, and urine samples were obtained with both their volume and collection interval measured. 2.2. Doripenem assay The concentrations of doripenem in plasma and urine were determined by high-performance liquid chromatography (HPLC) as reported previously [7]. In brief, plasma and urine samples (400 ␮L) were transferred to an ultrafiltration device and centrifuged. Then, 20 ␮L of the filtered solution was injected onto a chromatograph with a C18 column and an ultraviolet absorbance detector at 300 nm. A mixture of 0.1 M sodium acetate buffer (pH 4.6) and acetonitrile (95:5) was used as the mobile phase at a flow rate of 1 mL/min. The lower limit of quantification was 0.05 ␮g/mL and the coefficients of variation were within 6%. 2.3. Population PK modelling Population PK modelling was conducted using the NONMEM program version 6.2.0 (ICON Development Solutions, Ellicott, MD). Based on a preliminary analysis, a multicompartment model (Fig. 1) was selected to describe both the plasma and urine data as follows: dX1 = Rin − dt

 CL

r

V1

+

CLnr Q + V1 V1

dX2 Q Q × X1 − × X2 = V1 V2 dt



× X1 +

Q × X2 V2 (1)

CLr dX3 = × X1 V1 dt where X1 , X2 and X3 are the amounts of doripenem in the central, peripheral and urine compartments (mg), Rin is the drug

277

infusion rate (mg/h), CLr and CLnr are the renal and non-renal clearances (L/h), Q is the intercompartmental (central–peripheral) clearance (L/h), and V1 and V2 are the volumes of distribution of the central and peripheral compartments (L). The interindividual variability was modelled exponentially:  i =  × exp(i ), where  i is the fixed-effects parameter for the ith subject,  is the mean value of the fixed-effects parameter in the population and  is a random interindividual variable, which is normally distributed with mean 0 and variance ω2 . The residual (intraindividual) variability was modelled using an additive error model: Cobs, ij = Cpred, ij + εij , where Cobs, ij and Cpred, ij denote the jth observed and predicted concentrations for the ith subject, and ε is a random intraindividual error, which is normally distributed with mean 0 and variance  2 . In the modelling, the first-order conditional estimation (FOCE) method was used. The influence of patient characteristics on the individual PK parameters obtained from the basic model was graphically explored. The covariates showing a correlation with the PK parameters were introduced into the basic model (Eq. (2)) one at a time, as expressed by Eqs. (3) and (4): P = k ,

(2)

P = k × (Cov) + k+1 ,

(3)

P = k × (Cov)k+1

(4)

where P is the population PK parameter, ␪k and ␪k + 1 are the mean values of the kth and (k + 1)th fixed-effects parameters in the population, and Cov is the value of the examined covariate. The significance of the influence of the examined covariate was evaluated by changes in −2 log likelihood (the minimum value of the objective function, OBJ). An OBJ decrease of more than 3.84 from the basic model (P < 0.05, 2 test) was considered to be statistically significant during the forward inclusion process. The full model was built by incorporating the significant covariates, and the final model was developed by a backward deletion method. The covariates in the full model were excluded from the model one at a time, and an OBJ increase of more than 7.88 from the full model (P < 0.005, 2 test) was considered to be statistically significant. A re-sampling technique called the bootstrap method was performed to validate the reliability and stability of the population PK model developed. The Wings program for NONMEM (version 612; University of Auckland, Auckland, New Zealand) was used to create new re-sampled data sets. The 95% confidence intervals (CIs) of PK parameters from 1000 bootstrap replicates were compared with the final model estimates. 2.4. Monte Carlo simulation

Fig. 1. Schematic diagram of the basic pharmacokinetic model. X1 , X2 and X3 , amounts of doripenem (mg) in the central, peripheral and urine compartments; Rin , drug infusion rate (mg/h); CLr and CLnr , renal and non-renal clearances (L/h); Q, intercompartmental clearance (L/h); and V1 and V2 , volumes of distribution of the central and peripheral compartments (L).

A 10 000-subject Monte Carlo simulation was conducted using the final population PK model to predict the PTA profile for each combination of doripenem regimen (1-h or 4-h infusion) and minimum inhibitory concentration (MIC). The following process was iterated from the 1st to the 10 000th subject using Crystal Ball 2000 software (Decisioneering, Denver, CO). A set of fixed-effects parameters (CLr , CLnr , V1 , Q and V2 ) was generated randomly according to each mean estimate () and the interindividual variance (ω) of the final population PK model. The steady-state unbound drug concentration versus time curve was simulated using the fixed-effects parameters, where a value of 8.5% protein binding was employed [2]. The time point at which the free drug concentration coincided with a two-fold diluted MIC (0.125–64 ␮g/mL) was determined, and the exposure time for which the free drug concentration remained at the MIC (fT > MIC) was finally calculated as the cumulative percentage of a 24-h period [8]. The PTA (%) was determined as the

278

K. Ikawa et al. / International Journal of Antimicrobial Agents 33 (2009) 276–279

proportion that achieved at least 40% fT > MIC (bactericidal exposure target [9]) of 10 000 estimates. 2.5. Determination of the PK–PD breakpoint For each doripenem regimen, the PK–PD breakpoint was determined as the highest MIC at which the PTA in plasma was ≥90% [10]. 3. Results and discussion The demographic and pathophysiological parameters of the study patients (n = 18) were as follows: gender, 16 males and 2 females; age, 63.4 ± 11.9 years (mean ± standard deviation) (range 42–79 years); body weight, 57.2 ± 8.4 kg (41.5–73.3 kg); serum creatinine, 1.4 ± 0.9 mg/dL (0.7–3.9 mg/dL); and creatinine clearance (CLCr ), 48.9 ± 19.4 mL/min (17.8–101.8 mL/min). A total of 115 plasma samples (Fig. 2a) and 61 urine samples (Fig. 2b) were used for the population PK modelling. The mean parameters for the basic model were: CLr (L/h) = 6.61; CLnr (L/h) = 2.38; V1 (L) = 8.63; Q (L/h) = 8.19; and V2 (L) = 9.61. During the forward inclusion process to build the covariate model, multiplicative incorporation of CLCr into CLr (Eq. (3)) caused the largest OBJ change (OBJ, −61.5), although CLCr , age and body weight each also significantly affected CLr . Because age and body weight each showed a high correlation with CLCr , they were not additionally incorporated into CLr to avoid a collinearity effect. None of the examined covariates significantly effected CLnr , V1 , Q or V2 . Then, during the backward elimination process, the intercept for CLr was excluded from the full model owing to insignificant change in OBJ (OBJ, +1.1).

Fig. 2. (a) Observed plasma concentrations (䊉, 115 samples) and (b) cumulative urinary excretions (, 61 samples) after 0.5-h infusion of 250 mg doripenem to 18 patients. The lower limit of quantification was 0.05 ␮g/mL. The simulation curves are illustrated using the mean fixed-effects parameters in the cases of creatinine clearance of 80 mL/min (—) and 20 mL/min (. . . . . . .).

Table 1 Estimates of population pharmacokinetic parameters in the final model for doripenem. Parameter

Population estimate

Fixed-effects parameter () CLr (L/h) =  1 × CLCr (mL/min) 1 0.137 CLnr (L/h) =  2 2

2.49

V1 (L) =  3 3

8.29

Q (L/h) =  4 4

8.10

V2 (L) =  5 5

9.37

SE (%)a

8.03 11.6 4.60 15.2 8.07

95% CIb

0.113–0.160 1.92–3.04 7.51–8.99 5.73–10.6 7.75–10.8

Interindividual variability () 0.051 (ωCLr = 0.227) CLr 0.403 (ωCLnr = 0.704) CLnr V1 0.548 (ωV1 = 0.854) 0.158 (ωQ = 0.413) Q V2 0.292 (ωV2 = 0.582)

43.1 29.0 41.2 30.4 45.2

0.015–0.132 0.107–0.887 0.263–0.731 0.031–0.251 0.030–0.555

Residual variability ε 0.449 ( = 0.67 ␮g/mL)

13.8

0.315–0.561

SE, standard error; CI, confidence interval; CLr and CLnr , renal and non-renal clearances; CLCr , creatinine clearance; V1 and V2 , volumes of distribution of the central and peripheral compartments; Q, intercompartmental (central–peripheral) clearance; ω and , variance. a SE determined via the covariance step in the NONMEM modelling. b 95% CI determined from 1000 bootstrap replicates using the Wings program for NONMEM.

The parameter estimates for the final model are shown in Table 1. All standard errors via the covariate step in the NONMEM modelling were <46%, and all parameter estimates were in the range of the 95% CI using the bootstrap method. Based on the validated population PK model and the range of the study patients’ data for CLCr (7.8–101.8 mL/min), three patient populations (CLCr = 80, 40 and 20 mL/min) were supposed. In the cases of CLCr = 80 mL/min and CLCr = 20 mL/min, the simulated plasma concentrations at 12 h after doses of 250 mg were <0.05 ␮g/mL and 1.2 ␮g/mL, respectively (Fig. 2a), and the corresponding cumulative urinary excretions were 78.4% and 42.2% (Fig. 2b). Consequently, the PK–PD breakpoints for the various doripenem regimens were 1–3 MIC dilutions (i.e. 2–8 times) higher with a CLCr of 20 mL/min than a CLCr of 80 mL/min (Table 2). For every patient population, the PK–PD breakpoint increased in the following order: 250 mg every 12 h (q12h) < 250 mg every 8 h (q8h) < 500 mg q8h < 1000 mg q8h (1-h infusions) ≤500 mg q8h < 1000 mg q8h (4-h infusions) (Table 2), indicating that prolonging the infusion time was a more effective strategy than dose escalation to increase the breakpoint. The doripenem package insert [11] recommends the following dosages (1-h infusion) for patients with renal dysfunction: 500 mg q8h for CLCr > 50 mL/min (no dosage adjustment necessary); 250 mg q8h for CLCr ≥ 30 mL/min to ≤50 mL/min; and 250 mg q12h for CLCr > 10 mL/min to <30 mL/min. However, the reasoning and validity of this recommendation have not been clearly explained. The current analysis showed the breakpoint for 500 mg q8h with a CLCr of 80 mL/min (1 ␮g/mL; Table 2) corresponded to the values for 250 mg q8h with a CLCr of 40 mL/min and 250 mg q12h with a CLCr of 20 mL/min. This provided a PK–PD rationale for the recommended dosage adjustment and a limitation that these regimens cannot treat a bacterium with a MIC ≥ 2 ␮g/mL. The in vitro activities (MIC for 90% of the organisms (MIC90 )) of doripenem against common bacteria in UTIs were 0.06 ␮g/mL for Escherichia coli, 8 ␮g/mL for Enterococcus faecalis, 0.06 ␮g/mL for

K. Ikawa et al. / International Journal of Antimicrobial Agents 33 (2009) 276–279

279

Table 2 Pharmacokinetic (PK)–pharmacodynamic (PD) breakpoints for doripenem regimens in patient populations with various degrees of creatinine clearance (CLCr ). Doripenem regimen

PK–PD breakpoint (␮g/mL)a CLCr = 80 mL/min

CLCr = 40 mL/min

1-h infusion 250 mg q12h (500 mg/day) 250 mg q8h (750 mg/day) 500 mg q8h (1500 mg/day) 1000 mg q8h (3000 mg/day)

0.125 0.5 1 2

0.5 1 2 4

4-h infusion 500 mg q8h (1500 mg/day) 1000 mg q8h (3000 mg/day)

4 8

4 8

CLCr = 20 mL/min 1 2 4 8 8 16

a Defined as the highest minimum inhibitory concentration (MIC) at which the probability of 40% fT > MIC (time for which the free drug concentration remained at the MIC) attainment in plasma was ≥90%.

coagulase-negative staphylococci, 0.12 ␮g/mL for Klebsiella spp., 8 ␮g/mL for Pseudomonas aeruginosa and 4 ␮g/mL for Acinetobacter spp. in the latest global surveillance [12]. Therefore, against E. faecalis, P. aeruginosa or a bacterium with a MIC ≥ 8 ␮g/mL, prolongation of infusion time and/or escalation of the dose is required, even for a CLCr of 20 mL/min. In conclusion, this study developed a more accurate population PK model for doripenem by using both plasma and urine concentration data in infected patients, showing that CLCr was the most significant factor affecting PKs. The PK–PD breakpoint (the highest MIC at which the PTA in plasma was ≥90%) varied with CLCr and doripenem regimen, and the value for 500 mg q8h (1-h infusion) with a CLCr of 80 mL/min corresponded to those for 250 mg q8h with a CLCr of 40 mL/min and 250 mg q12h with a CLCr of 20 mL/min. Prolonging the infusion time was a more effective strategy than dose escalation to increase the breakpoint. These results could provide a PK–PD-based strategy for tailoring a doripenem regimen according to an infected patient’s CLCr and bacterial susceptibility (if MIC data for the patient are available before or after therapy), although further studies are needed to confirm the clinical implications of our findings and proposals. Funding: No funding sources. Competing interests: None declared. Ethical approval: The Institutional Ethical Review Board of Okayama University Hospital, Okayama, Japan, and the Ethics Committee of Aichi Medical University School of Medicine, Nagakute, Japan. References [1] Greer ND. Doripenem (Doribax): the newest addition to the carbapenems. Proc (Bayl Univ Med Cent) 2008;21:337–41.

[2] Bhavnani SM, Hammel JP, Cirincione BB, Wikler MA, Ambrose PG. Use of pharmacokinetic–pharmacodynamic target attainment analyses to support phase 2 and 3 dosing strategies for doripenem. Antimicrob Agents Chemother 2005;49:3944–7. [3] Watanabe A, Fujimura S, Kikuchi T, Gomi K, Fuse K, Nukiwa T. Evaluation of dosing designs of carbapenems for severe respiratory infection using Monte Carlo simulation. J Infect Chemother 2007;13:332–40. [4] Nagasawa Z, Kusaba K, Aoki Y. Susceptibility of clinical isolates of Pseudomonas aeruginosa in the Northern Kyushu district of Japan to carbapenem antibiotics, determined by an integrated concentration method: evaluation of the method based on Monte Carlo simulation. J Infect Chemother 2008;14: 238–43. [5] Ikawa K, Morikawa N, Ikeda K, Ohge H, Sueda T. Pharmacodynamic assessment of doripenem in peritoneal fluid against Gram-negative organisms: use of population pharmacokinetic modelling and Monte Carlo simulation. Diagn Microbiol Infect Dis 2008;62:292–7. [6] Nakashima M, Oguma T. Phase I study of doripenem, a new carbapenem antibiotic for injection in healthy volunteers. Jpn J Chemother 2005;53(Suppl.):104–23. [7] Ikeda K, Ikawa K, Morikawa N, Kameda K, Urakawa N, Ohge H, et al. Quantification of doripenem in human plasma and peritoneal fluid by high-performance liquid chromatography with ultraviolet detection. J Chromatogr B Anal Technol Biomed Life Sci 2008;867:20–5. [8] Mouton JW, Dudley MN, Cars O, Derendorf H, Drusano GL. Standardization of pharmacokinetic/pharmacodynamic (PK/PD) terminology for anti-infective drugs: an update. J Antimicrob Chemother 2005;55: 601–7. [9] Kim A, Banevicius MA, Nicolau DP. In vivo pharmacodynamic profiling of doripenem against Pseudomonas aeruginosa by simulating human exposures. Antimicrob Agents Chemother 2008;52:2497–502. [10] Frei CR, Wiederhold NP, Burgess DS. Antimicrobial breakpoints for Gramnegative aerobic bacteria based on pharmacokinetic–pharmacodynamic models with Monte Carlo simulation. J Antimicrob Chemother 2008;61: 621–8. [11] Ortho-McNeil-Janssen Pharmaceuticals. DORIBAXTM (doripenem for injection) package insert. Titusville, NJ: Ortho-McNeil-Janssen Pharmaceuticals; 2008. http://www.doribax.com/doribax/interactive pi.html [accessed 26 September 2008]. [12] Fritsche TR, Stilwell MG, Jones RN. Antimicrobial activity of doripenem (S-4661): a global surveillance report (2003). Clin Microbiol Infect 2005;11:974–84.