Temocillin dosing in haemodialysis patients based on population pharmacokinetics of total and unbound concentrations and Monte Carlo simulations

Temocillin dosing in haemodialysis patients based on population pharmacokinetics of total and... Abstract Objectives To develop a population model describing temocillin pharmacokinetics (PK) in patients undergoing haemodialysis and investigate how pharmacokinetic/pharmacodynamic (PD) targets can be met with different dosage regimens. Patients and methods Sixteen patients received the currently licenced dosing of 1, 2 or 3 g of temocillin (total of 61 doses) corresponding to an inter-dialytic period of 20, 44 or 68 h, respectively, and a dialysis period of 4 h. A non-linear mixed-effects model was developed jointly for total and unbound temocillin serum concentrations. The performance of clinically feasible dosing regimens was evaluated using a 5000-subject Monte Carlo (MC) simulation for determining the highest MIC for which the PK/PD target of 40%ƒT>MIC would be reached in 90% of patients [probability of target attainment (PTA)]. This PK study was registered at ClinicalTrials.gov (NCT02285075). Results Temocillin unbound and total serum concentrations (429 samples) were used to fit an open two-compartment model with non-linear albumin binding and first-order elimination. In addition to total body clearance, dialysis clearance was modelled using the Michaels function. The currently licenced dosing achieved a 90% PTA for an MIC up to 8 mg/L. A new temocillin dosage regimen was designed that would achieve a 90% PTA for an MIC of 16 mg/L (MIC90 of target organisms) adjusted to patient weight and inter-dialytic period. Conclusions Currently licensed dosage regimen is suboptimal for MICs >8 mg/L (frequently found in clinical isolates). Model-based simulations allowed suggestion of a new dosage regimen with improved probability of microbiological success, applicability in routine clinical practice and more appropriate for empirical therapy. Introduction Patients with end-stage renal disease (ESRD) suffer a very high morbidity and mortality, with infection being the second leading cause of mortality.1 Haemodialysis patients in particular have huge rates of bacteraemia.1,2 Although Gram-positive organism infections are predominant in haemodialysis populations, recent studies have highlighted an alarming increase in Gram-negative bacteraemia.3–5 In addition, multi-resistant organisms are on the rise in this setting, imposing a target-oriented use of antimicrobial drugs.3,5 Temocillin, a Gram-negative spectrum β-lactam antibiotic with a remarkable stability against most β-lactamases, including AmpC β-lactamases, ESBLs and most carbapenemases,6 could therefore fill a unmet medical need. Its renaissance in haemodialysis units was, however, challenged as most of its pharmacokinetic (PK) data dates back to the 1980s.7,8 This data no longer reflects the marked advances in dialysis equipment and changes in dosing schemes. We have conducted and have reported a PK study of temocillin in haemodialysis patients receiving this drug on dialysis days only using non-compartmental analysis.9 This study was initiated because dialysis can alter drug PK, influenced by several specific factors such as drug protein binding, dialysis system characteristics and geometry, and dialysis conditions.10 Based on this study, concerns about temocillin underdosing, if following the originally approved drug dosage recommendations, have been raised9 and acted upon by updating the Summary of Product Characteristics (effective in Belgium and France since June 2017).11 An additional difficulty, however, stems from the fact that the PK of temocillin are non-linear, due to saturable protein binding and extra-renal elimination.12,13 Hypoalbuminemia is frequent in ESRD patients10 and further adds to this complexity, as temocillin is highly protein bound (70%–85% in healthy subjects).6 As other β-lactams, temocillin exhibits time-dependent killing.6 Therefore, the most important PK/pharmacodynamic (PD) index predicting antimicrobial efficacy is the percentage of the dosing interval during which its unbound concentration is maintained above the MIC for the offending organism (ƒT>MIC).14 As treatment with temocillin is often empirical, the actual MIC of temocillin for the offending organism(s) is unknown. However, the temocillin MIC90 is 16 mg/L for Enterobacteriaceae.15,16 Monte Carlo (MC) simulations are a tool for estimating the probability of attaining a specific PK/PD target, accounting for PK data.17 They enable the optimization of dosage regimens and maximizing probability of target attainment (PTA), used as a surrogate for successful microbiological outcome.17 The purpose of the present study was to develop a mechanism-based population PK model of the concentration–time profile of temocillin unbound and total serum concentrations in haemodialysis patients, starting from the raw data of our previous study (see Vandecasteele et al.9). The present study also investigates the relationship between dosage regimens and achievement of PK/PD targets. Methods Study design and population This PK study was conducted and obtained ethics approval (unique Belgian no. B049201215528) at the AZ Sint-Jan Brugge Hospital in Belgium and was registered at ClinicalTrials.gov (NCT02285075). The study design as well as a non-compartmental analysis of temocillin serum concentration–time data has previously been published9 (see Figure S1, available as Supplementary data at JAC Online). This non-randomized, open-label, multiple-dose study reflects the current clinical practice at the study site with 4 h standard haemodialysis sessions. Patients were included if undergoing intermittent haemodialysis for ESRD and prescribed temocillin for documented or suspected infection caused by an organism susceptible to temocillin. Temocillin was administered in 1, 2 or 3 g doses on dialysis days only, for 24, 48 or 72 h inter-dialytic intervals, respectively. Treatment regimen and duration were determined by the attending physician. Exclusion criteria were age (<18 years), limited (<24 h) estimated life expectancy due to major comorbidity, pregnancy, IgE-mediated allergy to penicillins or withholding written informed consent. FX8 and FX10 dialysers (Fresenius Medical Care AG & Co., Bad Homburg, Germany) were used. Blood was sampled from an arterial or venous catheter as follows: pre-dose sample followed by samples at 0.5, 3, 6, 12, 24, 36, 48 and 68 h, truncated by the next dialysis session. During the 4 h dialysis session three additional hourly samples were collected (see Figure S2). Serum samples were obtained by centrifugation after blood clotting and frozen at −80°C until analysis. Validated HPLC methods, described in detail elsewhere18,19, with a lower limit of quantification of 5 and 0.5 mg/L for total and unbound concentrations (determined in ultrafiltrates), respectively, were used; the non-compartmental analysis of the data was previously reported.9 PK analysis PK data were analysed using non-linear mixed-effects models with NONMEM version 7.3 (ICON Development Solutions, Ellicott, MD, USA). The first-order conditional estimation method with interaction and subroutine ADVAN13 TRANS1 was used during model development.20 Automation and post-processing of results was with Perl-speaks-NONMEM (version 4.4.8; https://uupharmacometrics.github.io/PsN/?) and RStudio 0.99.467 with R 3.1.1 (http://www.r-project.org/). Pirana 2.9.2 (http://www.pirana-software.com/) organized the model development. The PK model was built to fit two types of data simultaneously: temocillin unbound and total serum concentrations. This required two principal model components: (i) a disposition model for unbound temocillin; and (ii) a model for the relationship between unbound drug and drug bound to serum proteins. A non-linear binding model21,22 was used to relate total serum concentrations (Ctot) and unbound concentrations (Cu): Ctot=Cu+[(Cu× Bmax)/(Cu+Kd)] (Eqn 1) where Bmax is the maximal binding capacity and Kd is the equilibrium dissociation constant. In addition to the intrinsic total body clearance (CL), temocillin dialysis clearance (CLdial) was implemented and set to zero only during inter-dialytic periods. Temocillin CLdial was estimated using Michaels equation (Eqn 2):23 CLdial=BFRexpKoABFR1-BFRDFR-1expKoADFR1-BFRDFR-BFRDFR (Eqn 2) where KoA is the mass transfer area coefficient of the dialysis filter for temocillin (L/h), BFR is the blood flow rate (specific to each session, with typical values of 300–400 mL/min) and DFR is the dialysate flow rate (specific to each session, with constant value in our study of 500 mL/min). It allows description of the haemodialysis operating conditions, by relating dialysis parameters (BFR, DFR and KoA) to the clearance of a solute. After evaluating several statistical models for variability, residual variability was best described by a proportional model and inter-individual variability and inter-occasion variability by a log-normal distribution. An ‘occasion’ was defined as a set of concentration–time data corresponding to a dosing interval. The covariates considered were age, dry body weight, serum albumin, urea concentration pre-dialysis, urea reduction rate, creatinine concentration pre-dialysis and creatinine reduction rate. Body weight, standardized to 70 kg, was included as an allometric function on all clearance (power coefficients = 3/4) and volume (power coefficient = 1) parameters for their strong theoretical and empirical evidence.24,25 The best-performing base model was used for covariate model building and empirical Bayes estimates plotted against covariate values to explore potential relationships. An observable trend between covariates and PK parameters led to consideration for inclusion in the population model. The influence of subject-specific covariates was assessed by the forward-inclusion and backward-deletion method, using significant levels of P < 0.01 and P < 0.005, respectively.26 Improvements in the fit obtained with each model were assessed in several ways. The NONMEM-generated objective function value (OFV) was used to perform the likelihood ratio test. When comparing nested models, a decrease in OFV of ≥3.84 was required to reach statistical significance (P < 0.05, 1 degree of freedom) and choosing the more complex model. Additional considerations were reductions in the inter-individual variability, inter-occasion variability, residual variability and precision of the parameter estimates as well as diagnostic plots and shrinkage. The final population model was evaluated using the sampling importance resampling (SIR) method27 and prediction-corrected visual predictive checks based on 1000 simulations.28 The SIR was deemed more appropriate than the bootstrap in this case because it is less sensitive to sample size. In addition, it allows fast run times as it does not require estimation steps. The model was also assessed by normalized prediction distribution errors (NPDE) metrics.29 For a model with good predictability, the NPDE should follow an N(0,1) distribution.30 PD simulations With model parameters determined, MC simulations were used to explore the probability of achieving a PD target for various MICs. Simulations generated unbound concentration–time profiles for 5000 subjects, with uniform weight distribution from 50 to 100 kg. From this, the ƒT>MIC was calculated for each subject over the treatment period.31 The PTA was defined as the probability of achieving the target of 40%ƒT>MIC, which is a commonly used target for bacteriological cure with β-lactams,32 over a plausible range of potential MICs of temocillin for Gram-negative organisms (1–128 mg/L). The main target organisms for temocillin are ESBL-producing Enterobacteriaceae, presenting an MIC90 of 16 mg/L.33 The PK/PD susceptibility breakpoint was defined as the highest MIC for which the PK/PD target of 40%ƒT>MIC is achieved in at least 90% of patients, which is the most commonly used acceptable level of PTA.17 Initially, the dosage regimens used in this study (1 g q24h, 2 g q48h and 3 g q72h), which correspond to the recently updated licensed dosage regimens,11 were evaluated against the above criteria. As these regimens fell short of achieving the treatment target for an MIC and PTA cut-off of 16 mg/L and 90%, a new dosage regimen was developed for 24, 48 and 72 h inter-dialytic periods. Simplicity and ease of use were key criteria for the proposed regimen, while not exceeding single dose administrations of 3 g. In the absence of maximum tolerated doses, current clinical practice has shown that single doses of 3 g lead to well tolerated drug concentrations in the studied population.9 Restricting maximum doses to this value further ensures drug concentrations remain in the range observed during model development, hence maintaining the model’s validity. The new regimen was obtained by iterative dose optimization for all weights from 50 to 100 kg, evaluated in steps of 2.5 kg. The cumulative fraction of response (CFR) was calculated according to Mouton et al.31 to estimate the expected overall response of Enterobacteriaceae to temocillin for different dosage regimens. MIC distributions reported for clinical specimens, including strains with unusual resistance as reported by Woodford et al.,34 were used (Table S1). MIC distributions of isolates producing OXA, VIM, IMP or NDM carbapenemases were, however, excluded from the dataset for their known high-level resistance to temocillin, leading to a total of 1920 strains of Escherichia coli, Enterobacter spp. and Enterobacter spp. considered. Results Study population The study included 16 patients of which 14 were male (88%), the median age was 68.5 years (range 24–91 years) and the median dry body weight was 73.1 kg (range 41.5–104 kg). All patients required intermittent haemodialysis and received temocillin as part of regular therapy for suspected or documented Gram-negative infection. Their key characteristics have been published elsewhere.7 Median albumin, pre-dialysis creatinine and blood urea nitrogen concentrations were 3.3 g/dL (range 2.1–4.4 g/dL), 6.5 mg/dL (1.26–10.29 mg/dL) and 107 mg/dL (10–169 mg/dL), respectively. Seven patients (44%) were diagnosed with sepsis. Gram-negative bacteria from the Enterobacteriaceae family were identified as the causative organism in 11 (69%) patients: E. coli (36.6%; n = 4), Klebsiella oxytoca (27.3%; n = 3), Enterobacter cloacae (27.3%; n = 3) and Klebsiella pneumoniae (9.1%; n = 1). Serum concentration–time profiles Four hundred and twenty-nine serum temocillin concentration measurements were available, from 48 dosing cycles, one cycle starting with the administration of temocillin and ending after dialysis. They include 4 cycles for 1 g q24h, 31 for 2 g q48h and 13 for 3 g q72h, with a median of 3 (range 1–6) dosing cycles per patient. Haemodialysis duration was constant (4 h), while the inter-dialytic interval was 24, 48 or 72 h depending on the patient’s clinical condition. Follow-up was for a median of 5.5 days (range 2–9 days). Non-compartmental analysis of serum concentration–time data for temocillin was published previously.9 PK analysis Serum unbound concentration–time profiles were best described by a two-compartment model with zero-order input and linear distribution from the central and peripheral compartment, after intravenous administration. The final model also included a first-order total body clearance that was estimated at 1.43 L/h. Assuming a typical DFR of 500 mL/min and a BFR of 300 mL/min, the CLdial calculated from the parameter KoA (estimated at 7.83 L/h) was 7.67 L/h. The total temocillin clearance is the sum of CL and CLdial. The OFV did not decrease significantly (ΔOFV = −2.6) compared with the model with constant dialysis clearance, but the Michaels equation was selected due to its mechanistic and physiological rationale.23 The final base structural model describing unbound drug and bindings in serum is shown in Figure 1. The addition of inter-occasion variability between dose administrations produced a further improvement in the fit (ΔOFV = −38). Dry body weight, allometrically scaled on clearance and volume parameters, resulted in an improved model (ΔOFV = −9.2). The addition of any other covariates in the model could not be statistically supported. The median elimination t½ of temocillin for all subjects during the inter-dialytic period, estimated using empirical Bayes estimates of the final model parameters, was 22.8 h (range 11–48.7 h). The final population model parameters are presented in Table 1. The values of Bmax and Kd describing the protein binding are within the non-linear range of Eqn 1 and as a consequence the median bound fraction experienced across the population is between 32% and 67%. This is expectedly lower than for healthy subjects. Goodness-of-fit plots and correlation of random effects for the final model were evaluated and did not show any model misspecification (see Figures S3 and S4). The prediction-corrected visual predictive check plot (Figure 2) showed a good predictive power until ∼130 h, as few patients were followed up for longer than that amount of time. The estimated NPDE values followed an N(0,1) distribution (Figure S5). Table 1. Parameter estimates of the final population PK model for unbound temocillin in patients undergoing haemodialysis Parameter Population estimate (%RSE) SIR 95% CI Structural model parameters  V1 (L/70 kg)a 22.7 (9.6) 19.3–26.4  V2 (L/70 kg)a 18.6 (9.7) 15.8–21.2  Q (L/h/70 kg)a 3.99 (8.0) 3.4–4.7  KoA (L/h)b 7.83 (17.1) 6.7–9.1  CL (L/h/70 kg)a 1.43 (12.5) 1.2–1.7  Kd (mg/L) 34.3 (21.7) 24.1–46.4  Bmax (mg/L) 117 (12.7) 96.3–142.3 Inter-individual variabilityc (CV%)  V1 38.2 (16.6) 31.7–49.3  V2 32.6 (25.3) 21.2–49.4  CL 50.2 (18.2) 39.0–66.5  Kd 82.1 (23.5) 60.8–106.7  Bmax 42.2 (25.8) 32.2–58.4 Inter-occasion variability (CV%)  CL 16.8 (25.7) 12.5–22.6 Proportional residual variabilityc (CV%)  unbound 23.0 (18.9) 21.7–24.6  total 18.0 (18.7) 16.6–19.3 Parameter Population estimate (%RSE) SIR 95% CI Structural model parameters  V1 (L/70 kg)a 22.7 (9.6) 19.3–26.4  V2 (L/70 kg)a 18.6 (9.7) 15.8–21.2  Q (L/h/70 kg)a 3.99 (8.0) 3.4–4.7  KoA (L/h)b 7.83 (17.1) 6.7–9.1  CL (L/h/70 kg)a 1.43 (12.5) 1.2–1.7  Kd (mg/L) 34.3 (21.7) 24.1–46.4  Bmax (mg/L) 117 (12.7) 96.3–142.3 Inter-individual variabilityc (CV%)  V1 38.2 (16.6) 31.7–49.3  V2 32.6 (25.3) 21.2–49.4  CL 50.2 (18.2) 39.0–66.5  Kd 82.1 (23.5) 60.8–106.7  Bmax 42.2 (25.8) 32.2–58.4 Inter-occasion variability (CV%)  CL 16.8 (25.7) 12.5–22.6 Proportional residual variabilityc (CV%)  unbound 23.0 (18.9) 21.7–24.6  total 18.0 (18.7) 16.6–19.3 RSE, relative standard error (standard error of the estimate/estimate × 100); CV, coefficient of variation; Q, inter-compartmental clearance; CL, total body clearance. a Allometric model with a standard body weight of 70 kg and an exponent of 1 (to scale the volume of both central and peripheral compartments) or 0.75 (to scale Q and CL). b CLDial = BFR×{EXP[KoA/BFR×(1 − BFR/DFR)]−1}/{EXP[KoA/BFR×(1 − BFR/DFR)]−BFR/DFR}.23 c The η-shrinkage for inter-individual variability was <19%; ɛ-shrinkage was 4.8%. Table 1. Parameter estimates of the final population PK model for unbound temocillin in patients undergoing haemodialysis Parameter Population estimate (%RSE) SIR 95% CI Structural model parameters  V1 (L/70 kg)a 22.7 (9.6) 19.3–26.4  V2 (L/70 kg)a 18.6 (9.7) 15.8–21.2  Q (L/h/70 kg)a 3.99 (8.0) 3.4–4.7  KoA (L/h)b 7.83 (17.1) 6.7–9.1  CL (L/h/70 kg)a 1.43 (12.5) 1.2–1.7  Kd (mg/L) 34.3 (21.7) 24.1–46.4  Bmax (mg/L) 117 (12.7) 96.3–142.3 Inter-individual variabilityc (CV%)  V1 38.2 (16.6) 31.7–49.3  V2 32.6 (25.3) 21.2–49.4  CL 50.2 (18.2) 39.0–66.5  Kd 82.1 (23.5) 60.8–106.7  Bmax 42.2 (25.8) 32.2–58.4 Inter-occasion variability (CV%)  CL 16.8 (25.7) 12.5–22.6 Proportional residual variabilityc (CV%)  unbound 23.0 (18.9) 21.7–24.6  total 18.0 (18.7) 16.6–19.3 Parameter Population estimate (%RSE) SIR 95% CI Structural model parameters  V1 (L/70 kg)a 22.7 (9.6) 19.3–26.4  V2 (L/70 kg)a 18.6 (9.7) 15.8–21.2  Q (L/h/70 kg)a 3.99 (8.0) 3.4–4.7  KoA (L/h)b 7.83 (17.1) 6.7–9.1  CL (L/h/70 kg)a 1.43 (12.5) 1.2–1.7  Kd (mg/L) 34.3 (21.7) 24.1–46.4  Bmax (mg/L) 117 (12.7) 96.3–142.3 Inter-individual variabilityc (CV%)  V1 38.2 (16.6) 31.7–49.3  V2 32.6 (25.3) 21.2–49.4  CL 50.2 (18.2) 39.0–66.5  Kd 82.1 (23.5) 60.8–106.7  Bmax 42.2 (25.8) 32.2–58.4 Inter-occasion variability (CV%)  CL 16.8 (25.7) 12.5–22.6 Proportional residual variabilityc (CV%)  unbound 23.0 (18.9) 21.7–24.6  total 18.0 (18.7) 16.6–19.3 RSE, relative standard error (standard error of the estimate/estimate × 100); CV, coefficient of variation; Q, inter-compartmental clearance; CL, total body clearance. a Allometric model with a standard body weight of 70 kg and an exponent of 1 (to scale the volume of both central and peripheral compartments) or 0.75 (to scale Q and CL). b CLDial = BFR×{EXP[KoA/BFR×(1 − BFR/DFR)]−1}/{EXP[KoA/BFR×(1 − BFR/DFR)]−BFR/DFR}.23 c The η-shrinkage for inter-individual variability was <19%; ɛ-shrinkage was 4.8%. Figure 1. View largeDownload slide Schematic diagram of the final PK model for temocillin after intravenous (iv) administration. The model parameters are defined in Table 1. Figure 1. View largeDownload slide Schematic diagram of the final PK model for temocillin after intravenous (iv) administration. The model parameters are defined in Table 1. Figure 2. View largeDownload slide Prediction-corrected visual predictive check of the final model describing the PK of total and unbound temocillin concentrations, in patients undergoing intermittent haemodialysis. The continuous line represents the median observed serum concentrations. The observed 5th and 95th percentiles are represented by broken lines. The area shaded dark grey represents the 95% CI for the simulated-based median and the areas shaded light grey represent the 95% CIs for the simulated-based 5th and 95th percentiles. The prediction-corrected concentrations are plotted as dots. Figure 2. View largeDownload slide Prediction-corrected visual predictive check of the final model describing the PK of total and unbound temocillin concentrations, in patients undergoing intermittent haemodialysis. The continuous line represents the median observed serum concentrations. The observed 5th and 95th percentiles are represented by broken lines. The area shaded dark grey represents the 95% CI for the simulated-based median and the areas shaded light grey represent the 95% CIs for the simulated-based 5th and 95th percentiles. The prediction-corrected concentrations are plotted as dots. PD analysis Figure 3 shows the PTA for 40%ƒT>MIC versus MIC, for various simulated temocillin dosage regimens as well as their proposed PK/PD-based temocillin breakpoints. The recently updated dosage regimens (1 g q24h, 2 g q48h and 3 g q72h),11 which also correspond to the dosage regimens currently used in the clinical centre treating the patients included in this study, reveal a probability of successful treatment (PTA ≥90%) for MICs ≤8 mg/L. Of note, the doses used to calculate this PTA were already twice the original licensed doses (0.5 g q24h and 1 g q48h), but led to an increase of only one log2 (one dilution) in the attainable MIC. For bacteria with an MIC of 16 mg/L, the administered doses of 1 g q24h, 2 g q48h and 3 g q72h would achieve a PTA of only 60%, 72% and 71%, respectively. Because of the poor performance of these modalities of treatment, a new regimen is proposed based on the model built and using MC simulations. Table 2 shows the dose needed to achieve a PK/PD target of 40%ƒT>MIC for an MIC of 16 mg/L in at least 90% of patients for 24, 48 and 72 h inter-dialytic periods. Simulations suggest that for the two extremes of patient weight (50 and 100 kg), this new proposed dosage regimen, when used in an intermittent haemodialysis schedule (Figure 4), should be safe for most patients, as they do not exceed the maximal unbound concentrations simulated for currently used doses. During the haemodialysis session, concentrations can drop to ∼7 mg/L, calling for the need of a replacement dose after each dialysis session. Table 2. Dosing table: proposed temocillin dosage regimen for haemodialysis patients, according to their weight and the inter-dialytic period Single or first doses should be administered immediately after dialysis. Inter-dialytic periods can be considered independent from each other. Dosage regimens were rounded up to the closest 0.5 g. Table 2. Dosing table: proposed temocillin dosage regimen for haemodialysis patients, according to their weight and the inter-dialytic period Single or first doses should be administered immediately after dialysis. Inter-dialytic periods can be considered independent from each other. Dosage regimens were rounded up to the closest 0.5 g. Figure 3. View largeDownload slide PTA of 40%ƒT>MIC versus MIC, for simulated temocillin dosage regimens (0.5 g q24h, 1 g q24h, dosing table 24 h; 1 g q48h, 2 g q48h, dosing table 48 h; 3 g q72h, dosing table 72 h) in four consecutive dosing cycles. All regimens assume a uniform weight distribution. Displayed values are the 95% CI lower bound on the binomial estimate of the PTA derived from 5000 simulated patients per regimen. Grey shading corresponds to the area in which treatment target succeeded, assuming an MIC and PTA cut-off of 16 mg/L and 90%, respectively. The solid vertical lines indicate the PK/PD breakpoint for each dosage regimen, i.e. the highest MIC for which the PK/PD target of 40%ƒT>MIC is achieved in at least 90% of patients. Figure 3. View largeDownload slide PTA of 40%ƒT>MIC versus MIC, for simulated temocillin dosage regimens (0.5 g q24h, 1 g q24h, dosing table 24 h; 1 g q48h, 2 g q48h, dosing table 48 h; 3 g q72h, dosing table 72 h) in four consecutive dosing cycles. All regimens assume a uniform weight distribution. Displayed values are the 95% CI lower bound on the binomial estimate of the PTA derived from 5000 simulated patients per regimen. Grey shading corresponds to the area in which treatment target succeeded, assuming an MIC and PTA cut-off of 16 mg/L and 90%, respectively. The solid vertical lines indicate the PK/PD breakpoint for each dosage regimen, i.e. the highest MIC for which the PK/PD target of 40%ƒT>MIC is achieved in at least 90% of patients. Figure 4. View largeDownload slide Concentration–time profiles based on 5000 MC simulations of unbound temocillin serum concentrations in haemodialysis patients. A typical thrice-weekly haemodialysis schedule was simulated for two patient weights (50 and 100 kg) when administered at the dosage regimens used in this study (2 g q48h and 3 g q72h) (a and b) and with the new proposed regimen, as per Table 2 (c and d). Dark grey shading corresponds to the 95% CI and the continuous black line corresponds to the median. The times for the haemodialysis session were 44–48 h, 92–96 h and 164–168 h (indicated by light grey shading). Figure 4. View largeDownload slide Concentration–time profiles based on 5000 MC simulations of unbound temocillin serum concentrations in haemodialysis patients. A typical thrice-weekly haemodialysis schedule was simulated for two patient weights (50 and 100 kg) when administered at the dosage regimens used in this study (2 g q48h and 3 g q72h) (a and b) and with the new proposed regimen, as per Table 2 (c and d). Dark grey shading corresponds to the 95% CI and the continuous black line corresponds to the median. The times for the haemodialysis session were 44–48 h, 92–96 h and 164–168 h (indicated by light grey shading). The CFRs against E. coli, Klebsiella spp. and Enterobacter spp. are summarized in Table 3. They were all <46% at the original licensed dosing, increased to ∼60% with the updated licensed dosing and were >80% with the new proposed dosing. Table 3. Cumulative fraction of response (%) to achieve the target of >40%ƒT>MIC for E. coli, Klebsiella spp. and Enterobacter spp., with different temocillin dosage regimens, based on published temocillin MIC distributions34 Temocillin dosage regimen Major β-lactamase isolates KPC (MIC50/90: 8/32 mg/L), n = 669 other (MIC50/90: 16/32 mg/L), n = 1251 all (MIC50/90: 16/32 mg/L), n = 1920 Former licensed: 0.5 g q24h 38.7 37.5 37.9 Current licensed: 1 g q24h 67.6 59.7 62.4 Proposed: dosing table 24 h 85.9 78.0 80.7 Former licensed: 1 g q48h 47.6 44.1 45.3 Current licensed: 2 g q48h 73.3 65.5 68.2 Proposed: dosing table 48 h 87.0 79.5 82.1 Current licensed: 3 g q72h 73.1 66.0 68.5 Proposed: dosing table 72 h 87.2 80.0 82.5 Temocillin dosage regimen Major β-lactamase isolates KPC (MIC50/90: 8/32 mg/L), n = 669 other (MIC50/90: 16/32 mg/L), n = 1251 all (MIC50/90: 16/32 mg/L), n = 1920 Former licensed: 0.5 g q24h 38.7 37.5 37.9 Current licensed: 1 g q24h 67.6 59.7 62.4 Proposed: dosing table 24 h 85.9 78.0 80.7 Former licensed: 1 g q48h 47.6 44.1 45.3 Current licensed: 2 g q48h 73.3 65.5 68.2 Proposed: dosing table 48 h 87.0 79.5 82.1 Current licensed: 3 g q72h 73.1 66.0 68.5 Proposed: dosing table 72 h 87.2 80.0 82.5 KPC, organisms expressing KPC-type carbapenemases. Proposed dosage regimens are described in Table 2. Table 3. Cumulative fraction of response (%) to achieve the target of >40%ƒT>MIC for E. coli, Klebsiella spp. and Enterobacter spp., with different temocillin dosage regimens, based on published temocillin MIC distributions34 Temocillin dosage regimen Major β-lactamase isolates KPC (MIC50/90: 8/32 mg/L), n = 669 other (MIC50/90: 16/32 mg/L), n = 1251 all (MIC50/90: 16/32 mg/L), n = 1920 Former licensed: 0.5 g q24h 38.7 37.5 37.9 Current licensed: 1 g q24h 67.6 59.7 62.4 Proposed: dosing table 24 h 85.9 78.0 80.7 Former licensed: 1 g q48h 47.6 44.1 45.3 Current licensed: 2 g q48h 73.3 65.5 68.2 Proposed: dosing table 48 h 87.0 79.5 82.1 Current licensed: 3 g q72h 73.1 66.0 68.5 Proposed: dosing table 72 h 87.2 80.0 82.5 Temocillin dosage regimen Major β-lactamase isolates KPC (MIC50/90: 8/32 mg/L), n = 669 other (MIC50/90: 16/32 mg/L), n = 1251 all (MIC50/90: 16/32 mg/L), n = 1920 Former licensed: 0.5 g q24h 38.7 37.5 37.9 Current licensed: 1 g q24h 67.6 59.7 62.4 Proposed: dosing table 24 h 85.9 78.0 80.7 Former licensed: 1 g q48h 47.6 44.1 45.3 Current licensed: 2 g q48h 73.3 65.5 68.2 Proposed: dosing table 48 h 87.0 79.5 82.1 Current licensed: 3 g q72h 73.1 66.0 68.5 Proposed: dosing table 72 h 87.2 80.0 82.5 KPC, organisms expressing KPC-type carbapenemases. Proposed dosage regimens are described in Table 2. Discussion This article distinguishes itself for being the first known study to use a model-based approach to determine the optimal dosing of temocillin in patients undergoing intermittent haemodialysis with contemporary haemodialysers. In this work, the PK of temocillin were characterized using non-linear mixed-effects modelling to estimate population PK parameters, to evaluate the performance of different dosage regimens and to propose new dosage guidelines. The best-performing disposition model consisted of two compartments and was in line with other studies.12,35–37 A number of covariates were available in the dataset; however, only dry body weight was found valid for inclusion in the final model. The relatively small patient cohort and large heterogeneity within this population probably explains that many of the covariates did not produce statistically significant improvements in the model objective function. Without this, empirical evidence was considered insufficient to include the respective covariate. Interestingly enough, there was no significant correlation between Bmax and albumin concentrations, implying that there may be other molecules involved in the binding process.9,38 It has been suggested for temocillin and other β-lactams (e.g. piperacillin)39 that an extra-renal elimination pathway (like biliary secretion) assumes a more important role on drug clearance in case of renal dysfunction.9,12,13 Thus, during model development, a mixed-order (capacity-limited) elimination process was also evaluated. Although leading to a decrease in the OFV, it was insufficient to reach statistical significance, probably due to the limited sample size. It was, therefore, not included in the final model. The data in the present study was previously used in a non-compartmental analysis.9 Other than this, previous research on temocillin in a haemodialysis setting is limited to two small studies conducted in the 1980s, using haemodialysers which no longer reflect the performance of the contemporary dialysers.7,8 Therefore, comparison with this data is of limited value. Temocillin population PKs have only been reported twice and both times in intensive care patients.36,40 De Jongh et al.40 assumed a constant unbound to total ratio of 25% and modelled only total concentration. This linear relationship was not confirmed by the data in the present study, where the non-linear relationship proposed finds its biological explanation in saturable protein binding. Laterre et al.36 modelled the unbound concentration on a cohort of non-renally impaired intensive care patients. Thus, only some parameters can be compared with this study.36 The volume of distribution of the central compartment found in the present study is nevertheless in line with the one estimated in that study. The strength of the present study is the joint model of both unbound and total concentrations collected in a routine clinical setting. Such an integrated model has not been reported previously, to the best of our knowledge. The unbound and total samples are processed differently and hence have different assay error distributions. By integrating both into the same model, one can reduce the variability in the estimated model parameters. This model also offers improved flexibility for inferring and deriving data useful for clinical application. By working with both unbound and total concentrations, the model can be provided with only the readily obtainable total concentration data and provide estimates of the more clinically relevant (pharmacologically active) unbound concentration expected in patients. Moreover, this model can be used to tailor dosage regimens for a specific target ƒT>MIC, given a bacterium with a known MIC. Based on the developed model, the MC simulations determined the PK/PD breakpoints for different temocillin regimens. Temocillin is likely to be used against bacteria with MICs up to 16 mg/L, which implies that the updated licensed regimens of 1 g q24h, 2 g q48h and 3 g q72h are insufficient. This updated regimen is an improvement on the former one, which was only effective, on a PK/PD standpoint, against highly susceptible bacteria (MIC <4 mg/L). However, the updated licensed regimen still delivers an unacceptably low overall CFR (∼60%). Furthermore, while it improves the PTA, it is still insufficient to meet the more clinically effective criteria of attaining a PTA for 40%ƒT>MIC = 16 mg/L in 90% of the patient population. The currently licensed dosage regimens produced a PK/PD breakpoint of 8 mg/L, which corresponds to the temocillin clinical breakpoint for systemic infections from BSAC.41 The corresponding CFRs were <70% and therefore also unacceptably low. In contrast, the regimen proposed in the present study produced better results, offering the possibility to cover organisms with reported MICs ≤16 mg/L (calculated value <18 mg/L) and reliably produce a CFR of ∼80%. This study has limitations that deserve mentioning. First, the relatively small sample size, due to practical considerations, and the performance of the study in a single clinical centre, may limit the applicability of the conclusions. This is, however, mitigated by the wealth of available data (in terms of assays collected) used to describe the temocillin concentration–time profiles. Second, we did not evaluate the performance of the dosage regimens for more aggressive PK/PD targets (e.g. 80%–100%ƒT>MIC or even 4–5× MIC for 100%ƒT>MIC). For the target of this study (40%ƒT>MIC), which is also that considered by EUCAST to establish susceptibility breakpoints for penicillins,42 some doses were spread across two administrations to avoid potential safety issues. More aggressive targets would probably imply using frequent administration regimens that are impractical in a clinical setting. Furthermore, differences in MIC distributions in various locations were not taken into account, which are needed to accurately assess the probability of treatment success. Also, the study used serum concentrations as a proxy for blood concentrations in the estimation of dialysis clearance for practical reasons and due to lack of information on temocillin blood cell partitioning. Prospective evaluation of the proposed dosage regimen is therefore necessary. Modelling, which is a cost-effective and quantified precursor step in the development of improved dosing recommendations, informs the design of appropriate future clinical studies.43 In conclusion, a detailed population PK model of temocillin with saturable protein binding is reported. Dry body weight was found to influence temocillin clearance and volume of distribution. Although recently updated, the current licensed dosage regimen was found to be adequate only for MICs ≤8 mg/L and, therefore, suboptimal for higher MICs that are reported in around half of the clinical isolates. Model-based simulations suggest a new dosage regimen with improved probability of treatment success, applicability in routine clinical practice and improved suitability for empirical therapy. Acknowledgements We are grateful to patients and the on-site nursing/medical team for their participation. We also thank Professor Nick Holford (University of Auckland, Auckland, New Zealand) for valuable advice during model development. Funding This work was supported in part by the Région Wallonne (project TEMOEXPAND 1217668). Transparency declarations A. C. M. B. and A. S. are employees of the Université catholique de Louvain. S. J. V. works at the AZ Sint-Jan Brugge-Oostende AV. P. M. T. was unpaid. P. M. T. is an unpaid advisor to Eumedica, the registration holder of temocillin. F. V. B. is Research Director of the Fonds de la Recherche Scientifique (F.R.S.-FNRS). Supplementary data Figures S1 to S5 and Table S1 are available as Supplementary data at JAC Online. References 1 Dalrymple LS , Johansen KL , Chertow GM et al. Infection-related hospitalizations in older patients with ESRD . Am J Kidney Dis 2010 ; 56 : 522 – 30 . Google Scholar CrossRef Search ADS PubMed 2 Vandecasteele SJ , Kurella TM. A patient-centered vision of care for ESRD: dialysis as a bridging treatment or as a final destination? J Am Soc Nephrol 2014 ; 25 : 1647 – 51 . Google Scholar CrossRef Search ADS PubMed 3 Girndt M. Bacteraemia in haemodialysis patients–not always Staphylococcus aureus . Nephrol Dial Transplant 2015 ; 30 : 1055 – 7 . Google Scholar CrossRef Search ADS PubMed 4 Murray EC , Marek A , Thomson PC et al. Gram-negative bacteraemia in haemodialysis . Nephrol Dial Transplant 2015 ; 30 : 1202 – 8 . Google Scholar CrossRef Search ADS PubMed 5 Pop-Vicas A , Strom J , Stanley K et al. Multidrug-resistant gram-negative bacteria among patients who require chronic hemodialysis . Clin J Am Soc Nephrol 2008 ; 3 : 752 – 8 . Google Scholar CrossRef Search ADS PubMed 6 Livermore DM , Tulkens PM. Temocillin revived . J Antimicrob Chemother 2009 ; 63 : 243 – 5 . Google Scholar CrossRef Search ADS PubMed 7 Boelaert J , Daneels R , Schurgers M et al. Effect of renal function and dialysis on temocillin pharmacokinetics . Drugs 1985 ; 29 Suppl 5: 109 – 13 . Google Scholar CrossRef Search ADS PubMed 8 Leroy A , Humbert G , Fillastre JP et al. Pharmacokinetics of temocillin (BRL 17421) in subjects with normal and impaired renal function . J Antimicrob Chemother 1983 ; 12 : 47 – 58 . Google Scholar CrossRef Search ADS PubMed 9 Vandecasteele SJ , Miranda Bastos AC , Capron A et al. Thrice-weekly temocillin administered after each dialysis session is appropriate for the treatment of serious Gram-negative infections in haemodialysis patients . Int J Antimicrob Agents 2015 ; 46 : 660 – 5 . Google Scholar CrossRef Search ADS PubMed 10 Verbeeck RK , Musuamba FT. Pharmacokinetics and dosage adjustment in patients with renal dysfunction . Eur J Clin Pharmacol 2009 ; 65 : 757 – 73 . Google Scholar CrossRef Search ADS PubMed 11 Agence fédérale des médicaments des produits de santé RCP online (http://www.fagg-afmps.be/fr). Temocillin Summary of Product Characteristics (available in French or Dutch under the trade name of NEGABAN). 12 Boelaert J , Daneels R , Schurgers M et al. The pharmacokinetics of temocillin in patients with normal and impaired renal function . J Antimicrob Chemother 1983 ; 11 : 349 – 56 . Google Scholar CrossRef Search ADS PubMed 13 Hoffler D , Koeppe P. Temocillin pharmacokinetics in normal and impaired renal function . Drugs 1985 ; 29 Suppl 5: 135 – 9 . Google Scholar CrossRef Search ADS PubMed 14 Craig WA. Basic pharmacodynamics of antibacterials with clinical applications to the use of β-lactams, glycopeptides, and linezolid . Infect Dis Clin North Am 2003 ; 17 : 479 – 501 . Google Scholar CrossRef Search ADS PubMed 15 Glupczynski Y , Huang TD , Berhin C et al. In vitro activity of temocillin against prevalent extended-spectrum β-lactamases producing Enterobacteriaceae from Belgian intensive care units . Eur J Clin Microbiol Infect Dis 2007 ; 26 : 777 – 83 . Google Scholar CrossRef Search ADS PubMed 16 Vanhoof R , Berin C , Carpentier M et al. Comparative in vitro activity of temocillin and other antimicrobial agents against Enterobacteriaceae isolated from patients admitted to five Belgian hospitals . Acta Clin Belg 2001 ; 56 : 354 – 9 . Google Scholar CrossRef Search ADS PubMed 17 Mouton JW , Setting clinical MIC breakpoints from a PK/PD point of view: it is dose that matters. In: Vincks A , Derendorf H , Mouton JW , eds. Fundamentals of Antimicrobial Pharmacokinetics and Pharmacodynamics . New York, NY, USA : Springer , 2014 : 45 – 59 . Google Scholar CrossRef Search ADS 18 Miranda Bastos AC , Vandecasteele SJ , Tulkens PM et al. Development and validation of a high performance liquid chromatography assay for the determination of temocillin in serum of haemodialysis patients . J Pharm Biomed Anal 2014 ; 90 : 192 – 7 . Google Scholar CrossRef Search ADS PubMed 19 Ngougni Pokem P , Miranda Bastos AC , Tulkens PM et al. Validation of a HPLC-MS/MS assay for the determination of total and unbound concentration of temocillin in human serum . Clin Biochem 2015 ; 48 : 542 – 5 . Google Scholar CrossRef Search ADS PubMed 20 Beal SL , Sheiner LB , Boeckmann AJ. NONMEM Users Guides . Ellicott City, MD, USA : Icon Development Solutions , 2006 . 21 Gonzalez D , Schmidt S , Derendorf H. Importance of relating efficacy measures to unbound drug concentrations for anti-infective agents . Clin Microbiol Rev 2013 ; 26 : 274 – 88 . Google Scholar CrossRef Search ADS PubMed 22 Toutain PL , Bousquet-Melou A. Free drug fraction vs free drug concentration: a matter of frequent confusion . J Vet Pharmacol Ther 2002 ; 25 : 460 – 3 . Google Scholar CrossRef Search ADS PubMed 23 Michaels AS. Operating parameters and performance criteria for hemodialyzers and other membrane-separation devices . Trans Am Soc Artif Intern Organs 1966 ; 12 : 387 – 92 . Google Scholar PubMed 24 Holford NH. A size standard for pharmacokinetics . Clin Pharmacokinet 1996 ; 30 : 329 – 32 . Google Scholar CrossRef Search ADS PubMed 25 Holford NHG , Anderson BJ. Allometric size: the scientific theory and extension to normal fat mass . Eur J Pharm Sci 2017 ; 109S : S59 – 64 . Google Scholar CrossRef Search ADS PubMed 26 Wahlby U , Jonsson EN , Karlsson MO. Comparison of stepwise covariate model building strategies in population pharmacokinetic-pharmacodynamic analysis . AAPS PharmSci 2002 ; 4 : E27 . Google Scholar CrossRef Search ADS PubMed 27 Dosne AG , Bergstrand M , Harling K et al. Improving the estimation of parameter uncertainty distributions in nonlinear mixed effects models using sampling importance resampling . J Pharmacokinet Pharmacodyn 2016 ; 43 : 583 – 96 . Google Scholar CrossRef Search ADS PubMed 28 Bergstrand M , Hooker AC , Wallin JE et al. Prediction-corrected visual predictive checks for diagnosing nonlinear mixed-effects models . AAPS J 2011 ; 13 : 143 – 51 . Google Scholar CrossRef Search ADS PubMed 29 Comets E , Brendel K , Mentre F. Computing normalised prediction distribution errors to evaluate nonlinear mixed-effect models: the npde add-on package for R . Comput Methods Programs Biomed 2008 ; 90 : 154 – 66 . Google Scholar CrossRef Search ADS PubMed 30 Brendel K , Comets E , Laffont C et al. Metrics for external model evaluation with an application to the population pharmacokinetics of gliclazide . Pharm Res 2006 ; 23 : 2036 – 49 . Google Scholar CrossRef Search ADS PubMed 31 Mouton JW , Dudley MN , Cars O et al. Standardization of pharmacokinetic/pharmacodynamic (PK/PD) terminology for anti-infective drugs: an update . J Antimicrob Chemother 2005 ; 55 : 601 – 7 . Google Scholar CrossRef Search ADS PubMed 32 Craig WA. Pharmacokinetic/pharmacodynamic parameters: rationale for antibacterial dosing of mice and men . Clin Infect Dis 1998 ; 26 : 1 – 10 . Google Scholar CrossRef Search ADS PubMed 33 Rodriguez-Villalobos H , Bogaerts P , Berhin C et al. Trends in production of extended-spectrum β-lactamases among Enterobacteriaceae of clinical interest: results of a nationwide survey in Belgian hospitals . J Antimicrob Chemother 2011 ; 66 : 37 – 47 . Google Scholar CrossRef Search ADS PubMed 34 Woodford N , Pike R , Meunier D et al. In vitro activity of temocillin against multidrug-resistant clinical isolates of Escherichia coli, Klebsiella spp. and Enterobacter spp., and evaluation of high-level temocillin resistance as a diagnostic marker for OXA-48 carbapenemase . J Antimicrob Chemother 2014 ; 69 : 564 – 7 . Google Scholar CrossRef Search ADS PubMed 35 Brown RM , Wise R , Andrews JM. Temocillin, in-vitro activity and the pharmacokinetics and tissue penetration in healthy volunteers . J Antimicrob Chemother 1982 ; 10 : 295 – 302 . Google Scholar CrossRef Search ADS PubMed 36 Laterre PF , Wittebole X , Van de Velde S et al. Temocillin (6 g daily) in critically ill patients: continuous infusion versus three times daily administration . J Antimicrob Chemother 2015 ; 70 : 891 – 8 . Google Scholar CrossRef Search ADS PubMed 37 Lockley MR , Brown RM , Wise R. Pharmacokinetics and tissue penetration of temocillin . Drugs 1985 ; 29 Suppl 5: 106 – 8 . Google Scholar CrossRef Search ADS PubMed 38 Liebchen U , Kratzer A , Wicha SG et al. Unbound fraction of ertapenem in intensive care unit patients . J Antimicrob Chemother 2014 ; 69 : 3108 – 11 . Google Scholar CrossRef Search ADS PubMed 39 Hayashi Y , Roberts JA , Paterson DL et al. Pharmacokinetic evaluation of piperacillin-tazobactam . Expert Opin Drug Metab Toxicol 2010 ; 6 : 1017 – 31 . Google Scholar CrossRef Search ADS PubMed 40 De Jongh R , Hens R , Basma V et al. Continuous versus intermittent infusion of temocillin, a directed spectrum penicillin for intensive care patients with nosocomial pneumonia: stability, compatibility, population pharmacokinetic studies and breakpoint selection . J Antimicrob Chemother 2008 ; 61 : 382 – 8 . Google Scholar CrossRef Search ADS PubMed 41 Andrews JM , Jevons G , Walker R et al. Temocillin susceptibility by BSAC methodology . J Antimicrob Chemother 2007 ; 60 : 185 – 7 . Google Scholar CrossRef Search ADS PubMed 42 EUCAST . Rationale Documents for Antibacterial Agents. http://www.eucast.org/documents/rd. 43 Ambrose PG. Antibacterial drug development program successes and failures: a pharmacometric explanation . Curr Opin Pharmacol 2017 ; 36 : 1 – 7 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy. All rights reserved. For Permissions, please email: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Antimicrobial Chemotherapy Oxford University Press

Temocillin dosing in haemodialysis patients based on population pharmacokinetics of total and unbound concentrations and Monte Carlo simulations

Loading next page...
 
/lp/ou_press/temocillin-dosing-in-haemodialysis-patients-based-on-population-s79fUaDAc6
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy. All rights reserved. For Permissions, please email: journals.permissions@oup.com.
ISSN
0305-7453
eISSN
1460-2091
D.O.I.
10.1093/jac/dky078
Publisher site
See Article on Publisher Site

Abstract

Abstract Objectives To develop a population model describing temocillin pharmacokinetics (PK) in patients undergoing haemodialysis and investigate how pharmacokinetic/pharmacodynamic (PD) targets can be met with different dosage regimens. Patients and methods Sixteen patients received the currently licenced dosing of 1, 2 or 3 g of temocillin (total of 61 doses) corresponding to an inter-dialytic period of 20, 44 or 68 h, respectively, and a dialysis period of 4 h. A non-linear mixed-effects model was developed jointly for total and unbound temocillin serum concentrations. The performance of clinically feasible dosing regimens was evaluated using a 5000-subject Monte Carlo (MC) simulation for determining the highest MIC for which the PK/PD target of 40%ƒT>MIC would be reached in 90% of patients [probability of target attainment (PTA)]. This PK study was registered at ClinicalTrials.gov (NCT02285075). Results Temocillin unbound and total serum concentrations (429 samples) were used to fit an open two-compartment model with non-linear albumin binding and first-order elimination. In addition to total body clearance, dialysis clearance was modelled using the Michaels function. The currently licenced dosing achieved a 90% PTA for an MIC up to 8 mg/L. A new temocillin dosage regimen was designed that would achieve a 90% PTA for an MIC of 16 mg/L (MIC90 of target organisms) adjusted to patient weight and inter-dialytic period. Conclusions Currently licensed dosage regimen is suboptimal for MICs >8 mg/L (frequently found in clinical isolates). Model-based simulations allowed suggestion of a new dosage regimen with improved probability of microbiological success, applicability in routine clinical practice and more appropriate for empirical therapy. Introduction Patients with end-stage renal disease (ESRD) suffer a very high morbidity and mortality, with infection being the second leading cause of mortality.1 Haemodialysis patients in particular have huge rates of bacteraemia.1,2 Although Gram-positive organism infections are predominant in haemodialysis populations, recent studies have highlighted an alarming increase in Gram-negative bacteraemia.3–5 In addition, multi-resistant organisms are on the rise in this setting, imposing a target-oriented use of antimicrobial drugs.3,5 Temocillin, a Gram-negative spectrum β-lactam antibiotic with a remarkable stability against most β-lactamases, including AmpC β-lactamases, ESBLs and most carbapenemases,6 could therefore fill a unmet medical need. Its renaissance in haemodialysis units was, however, challenged as most of its pharmacokinetic (PK) data dates back to the 1980s.7,8 This data no longer reflects the marked advances in dialysis equipment and changes in dosing schemes. We have conducted and have reported a PK study of temocillin in haemodialysis patients receiving this drug on dialysis days only using non-compartmental analysis.9 This study was initiated because dialysis can alter drug PK, influenced by several specific factors such as drug protein binding, dialysis system characteristics and geometry, and dialysis conditions.10 Based on this study, concerns about temocillin underdosing, if following the originally approved drug dosage recommendations, have been raised9 and acted upon by updating the Summary of Product Characteristics (effective in Belgium and France since June 2017).11 An additional difficulty, however, stems from the fact that the PK of temocillin are non-linear, due to saturable protein binding and extra-renal elimination.12,13 Hypoalbuminemia is frequent in ESRD patients10 and further adds to this complexity, as temocillin is highly protein bound (70%–85% in healthy subjects).6 As other β-lactams, temocillin exhibits time-dependent killing.6 Therefore, the most important PK/pharmacodynamic (PD) index predicting antimicrobial efficacy is the percentage of the dosing interval during which its unbound concentration is maintained above the MIC for the offending organism (ƒT>MIC).14 As treatment with temocillin is often empirical, the actual MIC of temocillin for the offending organism(s) is unknown. However, the temocillin MIC90 is 16 mg/L for Enterobacteriaceae.15,16 Monte Carlo (MC) simulations are a tool for estimating the probability of attaining a specific PK/PD target, accounting for PK data.17 They enable the optimization of dosage regimens and maximizing probability of target attainment (PTA), used as a surrogate for successful microbiological outcome.17 The purpose of the present study was to develop a mechanism-based population PK model of the concentration–time profile of temocillin unbound and total serum concentrations in haemodialysis patients, starting from the raw data of our previous study (see Vandecasteele et al.9). The present study also investigates the relationship between dosage regimens and achievement of PK/PD targets. Methods Study design and population This PK study was conducted and obtained ethics approval (unique Belgian no. B049201215528) at the AZ Sint-Jan Brugge Hospital in Belgium and was registered at ClinicalTrials.gov (NCT02285075). The study design as well as a non-compartmental analysis of temocillin serum concentration–time data has previously been published9 (see Figure S1, available as Supplementary data at JAC Online). This non-randomized, open-label, multiple-dose study reflects the current clinical practice at the study site with 4 h standard haemodialysis sessions. Patients were included if undergoing intermittent haemodialysis for ESRD and prescribed temocillin for documented or suspected infection caused by an organism susceptible to temocillin. Temocillin was administered in 1, 2 or 3 g doses on dialysis days only, for 24, 48 or 72 h inter-dialytic intervals, respectively. Treatment regimen and duration were determined by the attending physician. Exclusion criteria were age (<18 years), limited (<24 h) estimated life expectancy due to major comorbidity, pregnancy, IgE-mediated allergy to penicillins or withholding written informed consent. FX8 and FX10 dialysers (Fresenius Medical Care AG & Co., Bad Homburg, Germany) were used. Blood was sampled from an arterial or venous catheter as follows: pre-dose sample followed by samples at 0.5, 3, 6, 12, 24, 36, 48 and 68 h, truncated by the next dialysis session. During the 4 h dialysis session three additional hourly samples were collected (see Figure S2). Serum samples were obtained by centrifugation after blood clotting and frozen at −80°C until analysis. Validated HPLC methods, described in detail elsewhere18,19, with a lower limit of quantification of 5 and 0.5 mg/L for total and unbound concentrations (determined in ultrafiltrates), respectively, were used; the non-compartmental analysis of the data was previously reported.9 PK analysis PK data were analysed using non-linear mixed-effects models with NONMEM version 7.3 (ICON Development Solutions, Ellicott, MD, USA). The first-order conditional estimation method with interaction and subroutine ADVAN13 TRANS1 was used during model development.20 Automation and post-processing of results was with Perl-speaks-NONMEM (version 4.4.8; https://uupharmacometrics.github.io/PsN/?) and RStudio 0.99.467 with R 3.1.1 (http://www.r-project.org/). Pirana 2.9.2 (http://www.pirana-software.com/) organized the model development. The PK model was built to fit two types of data simultaneously: temocillin unbound and total serum concentrations. This required two principal model components: (i) a disposition model for unbound temocillin; and (ii) a model for the relationship between unbound drug and drug bound to serum proteins. A non-linear binding model21,22 was used to relate total serum concentrations (Ctot) and unbound concentrations (Cu): Ctot=Cu+[(Cu× Bmax)/(Cu+Kd)] (Eqn 1) where Bmax is the maximal binding capacity and Kd is the equilibrium dissociation constant. In addition to the intrinsic total body clearance (CL), temocillin dialysis clearance (CLdial) was implemented and set to zero only during inter-dialytic periods. Temocillin CLdial was estimated using Michaels equation (Eqn 2):23 CLdial=BFRexpKoABFR1-BFRDFR-1expKoADFR1-BFRDFR-BFRDFR (Eqn 2) where KoA is the mass transfer area coefficient of the dialysis filter for temocillin (L/h), BFR is the blood flow rate (specific to each session, with typical values of 300–400 mL/min) and DFR is the dialysate flow rate (specific to each session, with constant value in our study of 500 mL/min). It allows description of the haemodialysis operating conditions, by relating dialysis parameters (BFR, DFR and KoA) to the clearance of a solute. After evaluating several statistical models for variability, residual variability was best described by a proportional model and inter-individual variability and inter-occasion variability by a log-normal distribution. An ‘occasion’ was defined as a set of concentration–time data corresponding to a dosing interval. The covariates considered were age, dry body weight, serum albumin, urea concentration pre-dialysis, urea reduction rate, creatinine concentration pre-dialysis and creatinine reduction rate. Body weight, standardized to 70 kg, was included as an allometric function on all clearance (power coefficients = 3/4) and volume (power coefficient = 1) parameters for their strong theoretical and empirical evidence.24,25 The best-performing base model was used for covariate model building and empirical Bayes estimates plotted against covariate values to explore potential relationships. An observable trend between covariates and PK parameters led to consideration for inclusion in the population model. The influence of subject-specific covariates was assessed by the forward-inclusion and backward-deletion method, using significant levels of P < 0.01 and P < 0.005, respectively.26 Improvements in the fit obtained with each model were assessed in several ways. The NONMEM-generated objective function value (OFV) was used to perform the likelihood ratio test. When comparing nested models, a decrease in OFV of ≥3.84 was required to reach statistical significance (P < 0.05, 1 degree of freedom) and choosing the more complex model. Additional considerations were reductions in the inter-individual variability, inter-occasion variability, residual variability and precision of the parameter estimates as well as diagnostic plots and shrinkage. The final population model was evaluated using the sampling importance resampling (SIR) method27 and prediction-corrected visual predictive checks based on 1000 simulations.28 The SIR was deemed more appropriate than the bootstrap in this case because it is less sensitive to sample size. In addition, it allows fast run times as it does not require estimation steps. The model was also assessed by normalized prediction distribution errors (NPDE) metrics.29 For a model with good predictability, the NPDE should follow an N(0,1) distribution.30 PD simulations With model parameters determined, MC simulations were used to explore the probability of achieving a PD target for various MICs. Simulations generated unbound concentration–time profiles for 5000 subjects, with uniform weight distribution from 50 to 100 kg. From this, the ƒT>MIC was calculated for each subject over the treatment period.31 The PTA was defined as the probability of achieving the target of 40%ƒT>MIC, which is a commonly used target for bacteriological cure with β-lactams,32 over a plausible range of potential MICs of temocillin for Gram-negative organisms (1–128 mg/L). The main target organisms for temocillin are ESBL-producing Enterobacteriaceae, presenting an MIC90 of 16 mg/L.33 The PK/PD susceptibility breakpoint was defined as the highest MIC for which the PK/PD target of 40%ƒT>MIC is achieved in at least 90% of patients, which is the most commonly used acceptable level of PTA.17 Initially, the dosage regimens used in this study (1 g q24h, 2 g q48h and 3 g q72h), which correspond to the recently updated licensed dosage regimens,11 were evaluated against the above criteria. As these regimens fell short of achieving the treatment target for an MIC and PTA cut-off of 16 mg/L and 90%, a new dosage regimen was developed for 24, 48 and 72 h inter-dialytic periods. Simplicity and ease of use were key criteria for the proposed regimen, while not exceeding single dose administrations of 3 g. In the absence of maximum tolerated doses, current clinical practice has shown that single doses of 3 g lead to well tolerated drug concentrations in the studied population.9 Restricting maximum doses to this value further ensures drug concentrations remain in the range observed during model development, hence maintaining the model’s validity. The new regimen was obtained by iterative dose optimization for all weights from 50 to 100 kg, evaluated in steps of 2.5 kg. The cumulative fraction of response (CFR) was calculated according to Mouton et al.31 to estimate the expected overall response of Enterobacteriaceae to temocillin for different dosage regimens. MIC distributions reported for clinical specimens, including strains with unusual resistance as reported by Woodford et al.,34 were used (Table S1). MIC distributions of isolates producing OXA, VIM, IMP or NDM carbapenemases were, however, excluded from the dataset for their known high-level resistance to temocillin, leading to a total of 1920 strains of Escherichia coli, Enterobacter spp. and Enterobacter spp. considered. Results Study population The study included 16 patients of which 14 were male (88%), the median age was 68.5 years (range 24–91 years) and the median dry body weight was 73.1 kg (range 41.5–104 kg). All patients required intermittent haemodialysis and received temocillin as part of regular therapy for suspected or documented Gram-negative infection. Their key characteristics have been published elsewhere.7 Median albumin, pre-dialysis creatinine and blood urea nitrogen concentrations were 3.3 g/dL (range 2.1–4.4 g/dL), 6.5 mg/dL (1.26–10.29 mg/dL) and 107 mg/dL (10–169 mg/dL), respectively. Seven patients (44%) were diagnosed with sepsis. Gram-negative bacteria from the Enterobacteriaceae family were identified as the causative organism in 11 (69%) patients: E. coli (36.6%; n = 4), Klebsiella oxytoca (27.3%; n = 3), Enterobacter cloacae (27.3%; n = 3) and Klebsiella pneumoniae (9.1%; n = 1). Serum concentration–time profiles Four hundred and twenty-nine serum temocillin concentration measurements were available, from 48 dosing cycles, one cycle starting with the administration of temocillin and ending after dialysis. They include 4 cycles for 1 g q24h, 31 for 2 g q48h and 13 for 3 g q72h, with a median of 3 (range 1–6) dosing cycles per patient. Haemodialysis duration was constant (4 h), while the inter-dialytic interval was 24, 48 or 72 h depending on the patient’s clinical condition. Follow-up was for a median of 5.5 days (range 2–9 days). Non-compartmental analysis of serum concentration–time data for temocillin was published previously.9 PK analysis Serum unbound concentration–time profiles were best described by a two-compartment model with zero-order input and linear distribution from the central and peripheral compartment, after intravenous administration. The final model also included a first-order total body clearance that was estimated at 1.43 L/h. Assuming a typical DFR of 500 mL/min and a BFR of 300 mL/min, the CLdial calculated from the parameter KoA (estimated at 7.83 L/h) was 7.67 L/h. The total temocillin clearance is the sum of CL and CLdial. The OFV did not decrease significantly (ΔOFV = −2.6) compared with the model with constant dialysis clearance, but the Michaels equation was selected due to its mechanistic and physiological rationale.23 The final base structural model describing unbound drug and bindings in serum is shown in Figure 1. The addition of inter-occasion variability between dose administrations produced a further improvement in the fit (ΔOFV = −38). Dry body weight, allometrically scaled on clearance and volume parameters, resulted in an improved model (ΔOFV = −9.2). The addition of any other covariates in the model could not be statistically supported. The median elimination t½ of temocillin for all subjects during the inter-dialytic period, estimated using empirical Bayes estimates of the final model parameters, was 22.8 h (range 11–48.7 h). The final population model parameters are presented in Table 1. The values of Bmax and Kd describing the protein binding are within the non-linear range of Eqn 1 and as a consequence the median bound fraction experienced across the population is between 32% and 67%. This is expectedly lower than for healthy subjects. Goodness-of-fit plots and correlation of random effects for the final model were evaluated and did not show any model misspecification (see Figures S3 and S4). The prediction-corrected visual predictive check plot (Figure 2) showed a good predictive power until ∼130 h, as few patients were followed up for longer than that amount of time. The estimated NPDE values followed an N(0,1) distribution (Figure S5). Table 1. Parameter estimates of the final population PK model for unbound temocillin in patients undergoing haemodialysis Parameter Population estimate (%RSE) SIR 95% CI Structural model parameters  V1 (L/70 kg)a 22.7 (9.6) 19.3–26.4  V2 (L/70 kg)a 18.6 (9.7) 15.8–21.2  Q (L/h/70 kg)a 3.99 (8.0) 3.4–4.7  KoA (L/h)b 7.83 (17.1) 6.7–9.1  CL (L/h/70 kg)a 1.43 (12.5) 1.2–1.7  Kd (mg/L) 34.3 (21.7) 24.1–46.4  Bmax (mg/L) 117 (12.7) 96.3–142.3 Inter-individual variabilityc (CV%)  V1 38.2 (16.6) 31.7–49.3  V2 32.6 (25.3) 21.2–49.4  CL 50.2 (18.2) 39.0–66.5  Kd 82.1 (23.5) 60.8–106.7  Bmax 42.2 (25.8) 32.2–58.4 Inter-occasion variability (CV%)  CL 16.8 (25.7) 12.5–22.6 Proportional residual variabilityc (CV%)  unbound 23.0 (18.9) 21.7–24.6  total 18.0 (18.7) 16.6–19.3 Parameter Population estimate (%RSE) SIR 95% CI Structural model parameters  V1 (L/70 kg)a 22.7 (9.6) 19.3–26.4  V2 (L/70 kg)a 18.6 (9.7) 15.8–21.2  Q (L/h/70 kg)a 3.99 (8.0) 3.4–4.7  KoA (L/h)b 7.83 (17.1) 6.7–9.1  CL (L/h/70 kg)a 1.43 (12.5) 1.2–1.7  Kd (mg/L) 34.3 (21.7) 24.1–46.4  Bmax (mg/L) 117 (12.7) 96.3–142.3 Inter-individual variabilityc (CV%)  V1 38.2 (16.6) 31.7–49.3  V2 32.6 (25.3) 21.2–49.4  CL 50.2 (18.2) 39.0–66.5  Kd 82.1 (23.5) 60.8–106.7  Bmax 42.2 (25.8) 32.2–58.4 Inter-occasion variability (CV%)  CL 16.8 (25.7) 12.5–22.6 Proportional residual variabilityc (CV%)  unbound 23.0 (18.9) 21.7–24.6  total 18.0 (18.7) 16.6–19.3 RSE, relative standard error (standard error of the estimate/estimate × 100); CV, coefficient of variation; Q, inter-compartmental clearance; CL, total body clearance. a Allometric model with a standard body weight of 70 kg and an exponent of 1 (to scale the volume of both central and peripheral compartments) or 0.75 (to scale Q and CL). b CLDial = BFR×{EXP[KoA/BFR×(1 − BFR/DFR)]−1}/{EXP[KoA/BFR×(1 − BFR/DFR)]−BFR/DFR}.23 c The η-shrinkage for inter-individual variability was <19%; ɛ-shrinkage was 4.8%. Table 1. Parameter estimates of the final population PK model for unbound temocillin in patients undergoing haemodialysis Parameter Population estimate (%RSE) SIR 95% CI Structural model parameters  V1 (L/70 kg)a 22.7 (9.6) 19.3–26.4  V2 (L/70 kg)a 18.6 (9.7) 15.8–21.2  Q (L/h/70 kg)a 3.99 (8.0) 3.4–4.7  KoA (L/h)b 7.83 (17.1) 6.7–9.1  CL (L/h/70 kg)a 1.43 (12.5) 1.2–1.7  Kd (mg/L) 34.3 (21.7) 24.1–46.4  Bmax (mg/L) 117 (12.7) 96.3–142.3 Inter-individual variabilityc (CV%)  V1 38.2 (16.6) 31.7–49.3  V2 32.6 (25.3) 21.2–49.4  CL 50.2 (18.2) 39.0–66.5  Kd 82.1 (23.5) 60.8–106.7  Bmax 42.2 (25.8) 32.2–58.4 Inter-occasion variability (CV%)  CL 16.8 (25.7) 12.5–22.6 Proportional residual variabilityc (CV%)  unbound 23.0 (18.9) 21.7–24.6  total 18.0 (18.7) 16.6–19.3 Parameter Population estimate (%RSE) SIR 95% CI Structural model parameters  V1 (L/70 kg)a 22.7 (9.6) 19.3–26.4  V2 (L/70 kg)a 18.6 (9.7) 15.8–21.2  Q (L/h/70 kg)a 3.99 (8.0) 3.4–4.7  KoA (L/h)b 7.83 (17.1) 6.7–9.1  CL (L/h/70 kg)a 1.43 (12.5) 1.2–1.7  Kd (mg/L) 34.3 (21.7) 24.1–46.4  Bmax (mg/L) 117 (12.7) 96.3–142.3 Inter-individual variabilityc (CV%)  V1 38.2 (16.6) 31.7–49.3  V2 32.6 (25.3) 21.2–49.4  CL 50.2 (18.2) 39.0–66.5  Kd 82.1 (23.5) 60.8–106.7  Bmax 42.2 (25.8) 32.2–58.4 Inter-occasion variability (CV%)  CL 16.8 (25.7) 12.5–22.6 Proportional residual variabilityc (CV%)  unbound 23.0 (18.9) 21.7–24.6  total 18.0 (18.7) 16.6–19.3 RSE, relative standard error (standard error of the estimate/estimate × 100); CV, coefficient of variation; Q, inter-compartmental clearance; CL, total body clearance. a Allometric model with a standard body weight of 70 kg and an exponent of 1 (to scale the volume of both central and peripheral compartments) or 0.75 (to scale Q and CL). b CLDial = BFR×{EXP[KoA/BFR×(1 − BFR/DFR)]−1}/{EXP[KoA/BFR×(1 − BFR/DFR)]−BFR/DFR}.23 c The η-shrinkage for inter-individual variability was <19%; ɛ-shrinkage was 4.8%. Figure 1. View largeDownload slide Schematic diagram of the final PK model for temocillin after intravenous (iv) administration. The model parameters are defined in Table 1. Figure 1. View largeDownload slide Schematic diagram of the final PK model for temocillin after intravenous (iv) administration. The model parameters are defined in Table 1. Figure 2. View largeDownload slide Prediction-corrected visual predictive check of the final model describing the PK of total and unbound temocillin concentrations, in patients undergoing intermittent haemodialysis. The continuous line represents the median observed serum concentrations. The observed 5th and 95th percentiles are represented by broken lines. The area shaded dark grey represents the 95% CI for the simulated-based median and the areas shaded light grey represent the 95% CIs for the simulated-based 5th and 95th percentiles. The prediction-corrected concentrations are plotted as dots. Figure 2. View largeDownload slide Prediction-corrected visual predictive check of the final model describing the PK of total and unbound temocillin concentrations, in patients undergoing intermittent haemodialysis. The continuous line represents the median observed serum concentrations. The observed 5th and 95th percentiles are represented by broken lines. The area shaded dark grey represents the 95% CI for the simulated-based median and the areas shaded light grey represent the 95% CIs for the simulated-based 5th and 95th percentiles. The prediction-corrected concentrations are plotted as dots. PD analysis Figure 3 shows the PTA for 40%ƒT>MIC versus MIC, for various simulated temocillin dosage regimens as well as their proposed PK/PD-based temocillin breakpoints. The recently updated dosage regimens (1 g q24h, 2 g q48h and 3 g q72h),11 which also correspond to the dosage regimens currently used in the clinical centre treating the patients included in this study, reveal a probability of successful treatment (PTA ≥90%) for MICs ≤8 mg/L. Of note, the doses used to calculate this PTA were already twice the original licensed doses (0.5 g q24h and 1 g q48h), but led to an increase of only one log2 (one dilution) in the attainable MIC. For bacteria with an MIC of 16 mg/L, the administered doses of 1 g q24h, 2 g q48h and 3 g q72h would achieve a PTA of only 60%, 72% and 71%, respectively. Because of the poor performance of these modalities of treatment, a new regimen is proposed based on the model built and using MC simulations. Table 2 shows the dose needed to achieve a PK/PD target of 40%ƒT>MIC for an MIC of 16 mg/L in at least 90% of patients for 24, 48 and 72 h inter-dialytic periods. Simulations suggest that for the two extremes of patient weight (50 and 100 kg), this new proposed dosage regimen, when used in an intermittent haemodialysis schedule (Figure 4), should be safe for most patients, as they do not exceed the maximal unbound concentrations simulated for currently used doses. During the haemodialysis session, concentrations can drop to ∼7 mg/L, calling for the need of a replacement dose after each dialysis session. Table 2. Dosing table: proposed temocillin dosage regimen for haemodialysis patients, according to their weight and the inter-dialytic period Single or first doses should be administered immediately after dialysis. Inter-dialytic periods can be considered independent from each other. Dosage regimens were rounded up to the closest 0.5 g. Table 2. Dosing table: proposed temocillin dosage regimen for haemodialysis patients, according to their weight and the inter-dialytic period Single or first doses should be administered immediately after dialysis. Inter-dialytic periods can be considered independent from each other. Dosage regimens were rounded up to the closest 0.5 g. Figure 3. View largeDownload slide PTA of 40%ƒT>MIC versus MIC, for simulated temocillin dosage regimens (0.5 g q24h, 1 g q24h, dosing table 24 h; 1 g q48h, 2 g q48h, dosing table 48 h; 3 g q72h, dosing table 72 h) in four consecutive dosing cycles. All regimens assume a uniform weight distribution. Displayed values are the 95% CI lower bound on the binomial estimate of the PTA derived from 5000 simulated patients per regimen. Grey shading corresponds to the area in which treatment target succeeded, assuming an MIC and PTA cut-off of 16 mg/L and 90%, respectively. The solid vertical lines indicate the PK/PD breakpoint for each dosage regimen, i.e. the highest MIC for which the PK/PD target of 40%ƒT>MIC is achieved in at least 90% of patients. Figure 3. View largeDownload slide PTA of 40%ƒT>MIC versus MIC, for simulated temocillin dosage regimens (0.5 g q24h, 1 g q24h, dosing table 24 h; 1 g q48h, 2 g q48h, dosing table 48 h; 3 g q72h, dosing table 72 h) in four consecutive dosing cycles. All regimens assume a uniform weight distribution. Displayed values are the 95% CI lower bound on the binomial estimate of the PTA derived from 5000 simulated patients per regimen. Grey shading corresponds to the area in which treatment target succeeded, assuming an MIC and PTA cut-off of 16 mg/L and 90%, respectively. The solid vertical lines indicate the PK/PD breakpoint for each dosage regimen, i.e. the highest MIC for which the PK/PD target of 40%ƒT>MIC is achieved in at least 90% of patients. Figure 4. View largeDownload slide Concentration–time profiles based on 5000 MC simulations of unbound temocillin serum concentrations in haemodialysis patients. A typical thrice-weekly haemodialysis schedule was simulated for two patient weights (50 and 100 kg) when administered at the dosage regimens used in this study (2 g q48h and 3 g q72h) (a and b) and with the new proposed regimen, as per Table 2 (c and d). Dark grey shading corresponds to the 95% CI and the continuous black line corresponds to the median. The times for the haemodialysis session were 44–48 h, 92–96 h and 164–168 h (indicated by light grey shading). Figure 4. View largeDownload slide Concentration–time profiles based on 5000 MC simulations of unbound temocillin serum concentrations in haemodialysis patients. A typical thrice-weekly haemodialysis schedule was simulated for two patient weights (50 and 100 kg) when administered at the dosage regimens used in this study (2 g q48h and 3 g q72h) (a and b) and with the new proposed regimen, as per Table 2 (c and d). Dark grey shading corresponds to the 95% CI and the continuous black line corresponds to the median. The times for the haemodialysis session were 44–48 h, 92–96 h and 164–168 h (indicated by light grey shading). The CFRs against E. coli, Klebsiella spp. and Enterobacter spp. are summarized in Table 3. They were all <46% at the original licensed dosing, increased to ∼60% with the updated licensed dosing and were >80% with the new proposed dosing. Table 3. Cumulative fraction of response (%) to achieve the target of >40%ƒT>MIC for E. coli, Klebsiella spp. and Enterobacter spp., with different temocillin dosage regimens, based on published temocillin MIC distributions34 Temocillin dosage regimen Major β-lactamase isolates KPC (MIC50/90: 8/32 mg/L), n = 669 other (MIC50/90: 16/32 mg/L), n = 1251 all (MIC50/90: 16/32 mg/L), n = 1920 Former licensed: 0.5 g q24h 38.7 37.5 37.9 Current licensed: 1 g q24h 67.6 59.7 62.4 Proposed: dosing table 24 h 85.9 78.0 80.7 Former licensed: 1 g q48h 47.6 44.1 45.3 Current licensed: 2 g q48h 73.3 65.5 68.2 Proposed: dosing table 48 h 87.0 79.5 82.1 Current licensed: 3 g q72h 73.1 66.0 68.5 Proposed: dosing table 72 h 87.2 80.0 82.5 Temocillin dosage regimen Major β-lactamase isolates KPC (MIC50/90: 8/32 mg/L), n = 669 other (MIC50/90: 16/32 mg/L), n = 1251 all (MIC50/90: 16/32 mg/L), n = 1920 Former licensed: 0.5 g q24h 38.7 37.5 37.9 Current licensed: 1 g q24h 67.6 59.7 62.4 Proposed: dosing table 24 h 85.9 78.0 80.7 Former licensed: 1 g q48h 47.6 44.1 45.3 Current licensed: 2 g q48h 73.3 65.5 68.2 Proposed: dosing table 48 h 87.0 79.5 82.1 Current licensed: 3 g q72h 73.1 66.0 68.5 Proposed: dosing table 72 h 87.2 80.0 82.5 KPC, organisms expressing KPC-type carbapenemases. Proposed dosage regimens are described in Table 2. Table 3. Cumulative fraction of response (%) to achieve the target of >40%ƒT>MIC for E. coli, Klebsiella spp. and Enterobacter spp., with different temocillin dosage regimens, based on published temocillin MIC distributions34 Temocillin dosage regimen Major β-lactamase isolates KPC (MIC50/90: 8/32 mg/L), n = 669 other (MIC50/90: 16/32 mg/L), n = 1251 all (MIC50/90: 16/32 mg/L), n = 1920 Former licensed: 0.5 g q24h 38.7 37.5 37.9 Current licensed: 1 g q24h 67.6 59.7 62.4 Proposed: dosing table 24 h 85.9 78.0 80.7 Former licensed: 1 g q48h 47.6 44.1 45.3 Current licensed: 2 g q48h 73.3 65.5 68.2 Proposed: dosing table 48 h 87.0 79.5 82.1 Current licensed: 3 g q72h 73.1 66.0 68.5 Proposed: dosing table 72 h 87.2 80.0 82.5 Temocillin dosage regimen Major β-lactamase isolates KPC (MIC50/90: 8/32 mg/L), n = 669 other (MIC50/90: 16/32 mg/L), n = 1251 all (MIC50/90: 16/32 mg/L), n = 1920 Former licensed: 0.5 g q24h 38.7 37.5 37.9 Current licensed: 1 g q24h 67.6 59.7 62.4 Proposed: dosing table 24 h 85.9 78.0 80.7 Former licensed: 1 g q48h 47.6 44.1 45.3 Current licensed: 2 g q48h 73.3 65.5 68.2 Proposed: dosing table 48 h 87.0 79.5 82.1 Current licensed: 3 g q72h 73.1 66.0 68.5 Proposed: dosing table 72 h 87.2 80.0 82.5 KPC, organisms expressing KPC-type carbapenemases. Proposed dosage regimens are described in Table 2. Discussion This article distinguishes itself for being the first known study to use a model-based approach to determine the optimal dosing of temocillin in patients undergoing intermittent haemodialysis with contemporary haemodialysers. In this work, the PK of temocillin were characterized using non-linear mixed-effects modelling to estimate population PK parameters, to evaluate the performance of different dosage regimens and to propose new dosage guidelines. The best-performing disposition model consisted of two compartments and was in line with other studies.12,35–37 A number of covariates were available in the dataset; however, only dry body weight was found valid for inclusion in the final model. The relatively small patient cohort and large heterogeneity within this population probably explains that many of the covariates did not produce statistically significant improvements in the model objective function. Without this, empirical evidence was considered insufficient to include the respective covariate. Interestingly enough, there was no significant correlation between Bmax and albumin concentrations, implying that there may be other molecules involved in the binding process.9,38 It has been suggested for temocillin and other β-lactams (e.g. piperacillin)39 that an extra-renal elimination pathway (like biliary secretion) assumes a more important role on drug clearance in case of renal dysfunction.9,12,13 Thus, during model development, a mixed-order (capacity-limited) elimination process was also evaluated. Although leading to a decrease in the OFV, it was insufficient to reach statistical significance, probably due to the limited sample size. It was, therefore, not included in the final model. The data in the present study was previously used in a non-compartmental analysis.9 Other than this, previous research on temocillin in a haemodialysis setting is limited to two small studies conducted in the 1980s, using haemodialysers which no longer reflect the performance of the contemporary dialysers.7,8 Therefore, comparison with this data is of limited value. Temocillin population PKs have only been reported twice and both times in intensive care patients.36,40 De Jongh et al.40 assumed a constant unbound to total ratio of 25% and modelled only total concentration. This linear relationship was not confirmed by the data in the present study, where the non-linear relationship proposed finds its biological explanation in saturable protein binding. Laterre et al.36 modelled the unbound concentration on a cohort of non-renally impaired intensive care patients. Thus, only some parameters can be compared with this study.36 The volume of distribution of the central compartment found in the present study is nevertheless in line with the one estimated in that study. The strength of the present study is the joint model of both unbound and total concentrations collected in a routine clinical setting. Such an integrated model has not been reported previously, to the best of our knowledge. The unbound and total samples are processed differently and hence have different assay error distributions. By integrating both into the same model, one can reduce the variability in the estimated model parameters. This model also offers improved flexibility for inferring and deriving data useful for clinical application. By working with both unbound and total concentrations, the model can be provided with only the readily obtainable total concentration data and provide estimates of the more clinically relevant (pharmacologically active) unbound concentration expected in patients. Moreover, this model can be used to tailor dosage regimens for a specific target ƒT>MIC, given a bacterium with a known MIC. Based on the developed model, the MC simulations determined the PK/PD breakpoints for different temocillin regimens. Temocillin is likely to be used against bacteria with MICs up to 16 mg/L, which implies that the updated licensed regimens of 1 g q24h, 2 g q48h and 3 g q72h are insufficient. This updated regimen is an improvement on the former one, which was only effective, on a PK/PD standpoint, against highly susceptible bacteria (MIC <4 mg/L). However, the updated licensed regimen still delivers an unacceptably low overall CFR (∼60%). Furthermore, while it improves the PTA, it is still insufficient to meet the more clinically effective criteria of attaining a PTA for 40%ƒT>MIC = 16 mg/L in 90% of the patient population. The currently licensed dosage regimens produced a PK/PD breakpoint of 8 mg/L, which corresponds to the temocillin clinical breakpoint for systemic infections from BSAC.41 The corresponding CFRs were <70% and therefore also unacceptably low. In contrast, the regimen proposed in the present study produced better results, offering the possibility to cover organisms with reported MICs ≤16 mg/L (calculated value <18 mg/L) and reliably produce a CFR of ∼80%. This study has limitations that deserve mentioning. First, the relatively small sample size, due to practical considerations, and the performance of the study in a single clinical centre, may limit the applicability of the conclusions. This is, however, mitigated by the wealth of available data (in terms of assays collected) used to describe the temocillin concentration–time profiles. Second, we did not evaluate the performance of the dosage regimens for more aggressive PK/PD targets (e.g. 80%–100%ƒT>MIC or even 4–5× MIC for 100%ƒT>MIC). For the target of this study (40%ƒT>MIC), which is also that considered by EUCAST to establish susceptibility breakpoints for penicillins,42 some doses were spread across two administrations to avoid potential safety issues. More aggressive targets would probably imply using frequent administration regimens that are impractical in a clinical setting. Furthermore, differences in MIC distributions in various locations were not taken into account, which are needed to accurately assess the probability of treatment success. Also, the study used serum concentrations as a proxy for blood concentrations in the estimation of dialysis clearance for practical reasons and due to lack of information on temocillin blood cell partitioning. Prospective evaluation of the proposed dosage regimen is therefore necessary. Modelling, which is a cost-effective and quantified precursor step in the development of improved dosing recommendations, informs the design of appropriate future clinical studies.43 In conclusion, a detailed population PK model of temocillin with saturable protein binding is reported. Dry body weight was found to influence temocillin clearance and volume of distribution. Although recently updated, the current licensed dosage regimen was found to be adequate only for MICs ≤8 mg/L and, therefore, suboptimal for higher MICs that are reported in around half of the clinical isolates. Model-based simulations suggest a new dosage regimen with improved probability of treatment success, applicability in routine clinical practice and improved suitability for empirical therapy. Acknowledgements We are grateful to patients and the on-site nursing/medical team for their participation. We also thank Professor Nick Holford (University of Auckland, Auckland, New Zealand) for valuable advice during model development. Funding This work was supported in part by the Région Wallonne (project TEMOEXPAND 1217668). Transparency declarations A. C. M. B. and A. S. are employees of the Université catholique de Louvain. S. J. V. works at the AZ Sint-Jan Brugge-Oostende AV. P. M. T. was unpaid. P. M. T. is an unpaid advisor to Eumedica, the registration holder of temocillin. F. V. B. is Research Director of the Fonds de la Recherche Scientifique (F.R.S.-FNRS). Supplementary data Figures S1 to S5 and Table S1 are available as Supplementary data at JAC Online. References 1 Dalrymple LS , Johansen KL , Chertow GM et al. Infection-related hospitalizations in older patients with ESRD . Am J Kidney Dis 2010 ; 56 : 522 – 30 . Google Scholar CrossRef Search ADS PubMed 2 Vandecasteele SJ , Kurella TM. A patient-centered vision of care for ESRD: dialysis as a bridging treatment or as a final destination? J Am Soc Nephrol 2014 ; 25 : 1647 – 51 . Google Scholar CrossRef Search ADS PubMed 3 Girndt M. Bacteraemia in haemodialysis patients–not always Staphylococcus aureus . Nephrol Dial Transplant 2015 ; 30 : 1055 – 7 . Google Scholar CrossRef Search ADS PubMed 4 Murray EC , Marek A , Thomson PC et al. Gram-negative bacteraemia in haemodialysis . Nephrol Dial Transplant 2015 ; 30 : 1202 – 8 . Google Scholar CrossRef Search ADS PubMed 5 Pop-Vicas A , Strom J , Stanley K et al. Multidrug-resistant gram-negative bacteria among patients who require chronic hemodialysis . Clin J Am Soc Nephrol 2008 ; 3 : 752 – 8 . Google Scholar CrossRef Search ADS PubMed 6 Livermore DM , Tulkens PM. Temocillin revived . J Antimicrob Chemother 2009 ; 63 : 243 – 5 . Google Scholar CrossRef Search ADS PubMed 7 Boelaert J , Daneels R , Schurgers M et al. Effect of renal function and dialysis on temocillin pharmacokinetics . Drugs 1985 ; 29 Suppl 5: 109 – 13 . Google Scholar CrossRef Search ADS PubMed 8 Leroy A , Humbert G , Fillastre JP et al. Pharmacokinetics of temocillin (BRL 17421) in subjects with normal and impaired renal function . J Antimicrob Chemother 1983 ; 12 : 47 – 58 . Google Scholar CrossRef Search ADS PubMed 9 Vandecasteele SJ , Miranda Bastos AC , Capron A et al. Thrice-weekly temocillin administered after each dialysis session is appropriate for the treatment of serious Gram-negative infections in haemodialysis patients . Int J Antimicrob Agents 2015 ; 46 : 660 – 5 . Google Scholar CrossRef Search ADS PubMed 10 Verbeeck RK , Musuamba FT. Pharmacokinetics and dosage adjustment in patients with renal dysfunction . Eur J Clin Pharmacol 2009 ; 65 : 757 – 73 . Google Scholar CrossRef Search ADS PubMed 11 Agence fédérale des médicaments des produits de santé RCP online (http://www.fagg-afmps.be/fr). Temocillin Summary of Product Characteristics (available in French or Dutch under the trade name of NEGABAN). 12 Boelaert J , Daneels R , Schurgers M et al. The pharmacokinetics of temocillin in patients with normal and impaired renal function . J Antimicrob Chemother 1983 ; 11 : 349 – 56 . Google Scholar CrossRef Search ADS PubMed 13 Hoffler D , Koeppe P. Temocillin pharmacokinetics in normal and impaired renal function . Drugs 1985 ; 29 Suppl 5: 135 – 9 . Google Scholar CrossRef Search ADS PubMed 14 Craig WA. Basic pharmacodynamics of antibacterials with clinical applications to the use of β-lactams, glycopeptides, and linezolid . Infect Dis Clin North Am 2003 ; 17 : 479 – 501 . Google Scholar CrossRef Search ADS PubMed 15 Glupczynski Y , Huang TD , Berhin C et al. In vitro activity of temocillin against prevalent extended-spectrum β-lactamases producing Enterobacteriaceae from Belgian intensive care units . Eur J Clin Microbiol Infect Dis 2007 ; 26 : 777 – 83 . Google Scholar CrossRef Search ADS PubMed 16 Vanhoof R , Berin C , Carpentier M et al. Comparative in vitro activity of temocillin and other antimicrobial agents against Enterobacteriaceae isolated from patients admitted to five Belgian hospitals . Acta Clin Belg 2001 ; 56 : 354 – 9 . Google Scholar CrossRef Search ADS PubMed 17 Mouton JW , Setting clinical MIC breakpoints from a PK/PD point of view: it is dose that matters. In: Vincks A , Derendorf H , Mouton JW , eds. Fundamentals of Antimicrobial Pharmacokinetics and Pharmacodynamics . New York, NY, USA : Springer , 2014 : 45 – 59 . Google Scholar CrossRef Search ADS 18 Miranda Bastos AC , Vandecasteele SJ , Tulkens PM et al. Development and validation of a high performance liquid chromatography assay for the determination of temocillin in serum of haemodialysis patients . J Pharm Biomed Anal 2014 ; 90 : 192 – 7 . Google Scholar CrossRef Search ADS PubMed 19 Ngougni Pokem P , Miranda Bastos AC , Tulkens PM et al. Validation of a HPLC-MS/MS assay for the determination of total and unbound concentration of temocillin in human serum . Clin Biochem 2015 ; 48 : 542 – 5 . Google Scholar CrossRef Search ADS PubMed 20 Beal SL , Sheiner LB , Boeckmann AJ. NONMEM Users Guides . Ellicott City, MD, USA : Icon Development Solutions , 2006 . 21 Gonzalez D , Schmidt S , Derendorf H. Importance of relating efficacy measures to unbound drug concentrations for anti-infective agents . Clin Microbiol Rev 2013 ; 26 : 274 – 88 . Google Scholar CrossRef Search ADS PubMed 22 Toutain PL , Bousquet-Melou A. Free drug fraction vs free drug concentration: a matter of frequent confusion . J Vet Pharmacol Ther 2002 ; 25 : 460 – 3 . Google Scholar CrossRef Search ADS PubMed 23 Michaels AS. Operating parameters and performance criteria for hemodialyzers and other membrane-separation devices . Trans Am Soc Artif Intern Organs 1966 ; 12 : 387 – 92 . Google Scholar PubMed 24 Holford NH. A size standard for pharmacokinetics . Clin Pharmacokinet 1996 ; 30 : 329 – 32 . Google Scholar CrossRef Search ADS PubMed 25 Holford NHG , Anderson BJ. Allometric size: the scientific theory and extension to normal fat mass . Eur J Pharm Sci 2017 ; 109S : S59 – 64 . Google Scholar CrossRef Search ADS PubMed 26 Wahlby U , Jonsson EN , Karlsson MO. Comparison of stepwise covariate model building strategies in population pharmacokinetic-pharmacodynamic analysis . AAPS PharmSci 2002 ; 4 : E27 . Google Scholar CrossRef Search ADS PubMed 27 Dosne AG , Bergstrand M , Harling K et al. Improving the estimation of parameter uncertainty distributions in nonlinear mixed effects models using sampling importance resampling . J Pharmacokinet Pharmacodyn 2016 ; 43 : 583 – 96 . Google Scholar CrossRef Search ADS PubMed 28 Bergstrand M , Hooker AC , Wallin JE et al. Prediction-corrected visual predictive checks for diagnosing nonlinear mixed-effects models . AAPS J 2011 ; 13 : 143 – 51 . Google Scholar CrossRef Search ADS PubMed 29 Comets E , Brendel K , Mentre F. Computing normalised prediction distribution errors to evaluate nonlinear mixed-effect models: the npde add-on package for R . Comput Methods Programs Biomed 2008 ; 90 : 154 – 66 . Google Scholar CrossRef Search ADS PubMed 30 Brendel K , Comets E , Laffont C et al. Metrics for external model evaluation with an application to the population pharmacokinetics of gliclazide . Pharm Res 2006 ; 23 : 2036 – 49 . Google Scholar CrossRef Search ADS PubMed 31 Mouton JW , Dudley MN , Cars O et al. Standardization of pharmacokinetic/pharmacodynamic (PK/PD) terminology for anti-infective drugs: an update . J Antimicrob Chemother 2005 ; 55 : 601 – 7 . Google Scholar CrossRef Search ADS PubMed 32 Craig WA. Pharmacokinetic/pharmacodynamic parameters: rationale for antibacterial dosing of mice and men . Clin Infect Dis 1998 ; 26 : 1 – 10 . Google Scholar CrossRef Search ADS PubMed 33 Rodriguez-Villalobos H , Bogaerts P , Berhin C et al. Trends in production of extended-spectrum β-lactamases among Enterobacteriaceae of clinical interest: results of a nationwide survey in Belgian hospitals . J Antimicrob Chemother 2011 ; 66 : 37 – 47 . Google Scholar CrossRef Search ADS PubMed 34 Woodford N , Pike R , Meunier D et al. In vitro activity of temocillin against multidrug-resistant clinical isolates of Escherichia coli, Klebsiella spp. and Enterobacter spp., and evaluation of high-level temocillin resistance as a diagnostic marker for OXA-48 carbapenemase . J Antimicrob Chemother 2014 ; 69 : 564 – 7 . Google Scholar CrossRef Search ADS PubMed 35 Brown RM , Wise R , Andrews JM. Temocillin, in-vitro activity and the pharmacokinetics and tissue penetration in healthy volunteers . J Antimicrob Chemother 1982 ; 10 : 295 – 302 . Google Scholar CrossRef Search ADS PubMed 36 Laterre PF , Wittebole X , Van de Velde S et al. Temocillin (6 g daily) in critically ill patients: continuous infusion versus three times daily administration . J Antimicrob Chemother 2015 ; 70 : 891 – 8 . Google Scholar CrossRef Search ADS PubMed 37 Lockley MR , Brown RM , Wise R. Pharmacokinetics and tissue penetration of temocillin . Drugs 1985 ; 29 Suppl 5: 106 – 8 . Google Scholar CrossRef Search ADS PubMed 38 Liebchen U , Kratzer A , Wicha SG et al. Unbound fraction of ertapenem in intensive care unit patients . J Antimicrob Chemother 2014 ; 69 : 3108 – 11 . Google Scholar CrossRef Search ADS PubMed 39 Hayashi Y , Roberts JA , Paterson DL et al. Pharmacokinetic evaluation of piperacillin-tazobactam . Expert Opin Drug Metab Toxicol 2010 ; 6 : 1017 – 31 . Google Scholar CrossRef Search ADS PubMed 40 De Jongh R , Hens R , Basma V et al. Continuous versus intermittent infusion of temocillin, a directed spectrum penicillin for intensive care patients with nosocomial pneumonia: stability, compatibility, population pharmacokinetic studies and breakpoint selection . J Antimicrob Chemother 2008 ; 61 : 382 – 8 . Google Scholar CrossRef Search ADS PubMed 41 Andrews JM , Jevons G , Walker R et al. Temocillin susceptibility by BSAC methodology . J Antimicrob Chemother 2007 ; 60 : 185 – 7 . Google Scholar CrossRef Search ADS PubMed 42 EUCAST . Rationale Documents for Antibacterial Agents. http://www.eucast.org/documents/rd. 43 Ambrose PG. Antibacterial drug development program successes and failures: a pharmacometric explanation . Curr Opin Pharmacol 2017 ; 36 : 1 – 7 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy. All rights reserved. For Permissions, please email: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Journal of Antimicrobial ChemotherapyOxford University Press

Published: Mar 22, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off