Dosing regimen of meropenem for adults with severe burns: a 1 population PK study with Monte Carlo simulations 2 3 Amelia Ramon-Lopez1,2, Jane M Allen3*, Alison H Thomson1,4, Bajlit S. Dheansa5, S. Elizabeth 4 James6, Geoff W. Hanlon6, Bruce Stewart7, J. Graham Davies8 5 6 1 Strathclyde Institute of Pharmacy and Biomedical Sciences, University of Strathclyde, 7 Glasgow, UK 8 2 Division of Pharmacy and Pharmaceutics, Department of Engineering, 9 Miguel Hernandez University, Spain 10 3 Pharmacy Department, Queen Victoria Hospital NHS Foundation Trust, East Grinstead, UK 11 4 Pharmacy Department, Western Infirmary, Glasgow, UK 12 5Burns Centre, Queen Victoria Hospital NHS Foundation Trust, East Grinstead, UK. 13 6School of Pharmacy & Biomolecular Sciences, University of Brighton, UK. 14 7Department of Microbiology, Surrey and Sussex Healthcare NHS Trust, UK. 15 8 Institute of Pharmaceutical Science, Kings College London, UK. 16 *Corresponding author. Present address: Jane M Allen, Pharmacy Department, Surrey and 17 Sussex Healthcare NHS Trust, East Surrey Hospital, Canada Avenue, Redhill, Surrey RH1 18 5RH, UK 19 20 Fax: +44 1737 231684 21 Tel: +44 1737 768511 ext 6249 22 Email: [email protected] 23 Keywords: Pharmacodynamics, carbapenems, antibiotics, intensive care. 24 Running title: Population pk study of meropenem in burns 25 SYNOPSIS 26 27 Objectives: To develop a population model to describe the PK of intravenous meropenem in 28 adult patients with severe burns and investigate potential relationships between dosage 29 regimens and antimicrobial efficacy. 30 Patients and methods: A dose of 1 g every 8 h was administered to adult patients with total 31 body surface area burns of ≥15%. Doses for subsequent courses were determined using results 32 from the initial course and the patient’s clinical condition. Five plasma meropenem 33 concentrations were typically measured over the dosage interval on 1 – 4 occasions. An open 34 two-compartment PK model was fitted to the meropenem concentrations using NONMEM and 35 the effect of covariates on meropenem PK was investigated. Monte Carlo simulations 36 investigated dosage regimens to achieve a target T for at least 40%, 60% or 80% of the >MIC 37 dose interval. 38 Results: Data comprised 113 meropenem concentration measurements from 20 dosage 39 intervals in 12 patients. The parameters were CL (L/h) = 0.196 L/h/kg x (1-0.023 x (age - 46)) x 40 (1- 0.049 x (albumin-15)), V = 0.273 L/kg x (1 - 0.049 x (albumin-15)), Q = 0.199 L/h/kg and V 1 2 41 = 0.309 L/kg x (1 – 0.049 x (albumin-15)). For a target of 80% T , the breakpoint was 8 mg/L >MIC 42 for doses of 1 g every 4 h and 2 g every 8 h given over 3 h but only 4 mg/L if given over 5 43 minutes. 44 Conclusions: Although 1 g eight-hourly should be effective against E. coli and coagulase 45 negative Staphylococcus, higher doses, ideally with a longer infusion time, would be more 46 appropriate for empiric therapy, mixed infections and bacteria with MIC values  4 mg/L. 47 INTRODUCTION 48 49 Severely burned patients present several key challenges in their management, one being 50 infection, which is a major cause of illness and death.1 The earliest organisms isolated from 51 burn wounds tend to be Gram-positive organisms, such as Staphylococcus spp, but in the latter 52 part of the first post-burn week, Gram-negative organisms become dominant, with 53 Pseudomonas spp being the most common isolates.2,3 54 Meropenem is a broad-spectrum -lactam antibiotic commonly used to treat infections in 55 patients with burn injuries. A survey of UK hospitals which treat severely burned adults found 56 that thirteen of the sixteen respondents used meropenem as empiric therapy and / or if 57 susceptible organisms were identified (unpublished data). In most units, the standard adult dose 58 of 1g over 5 minutes every 8 h was used. However, it has been known since the 1970s that the 59 pathophysiological changes which follow a major burn injury may affect the pharmacokinetics 60 (PK) of drugs.4 These changes are influenced by a number of factors, including the presence of 61 sepsis, the area and depth of the burn, serum protein concentration, age, CL , degree of CR 62 hydration, presence of oedema and time after injury.5 As a result, several studies have 63 recommended using higher antibiotic doses than are given to patients without burn injuries.6-9 64 There is evidence to suggest that meropenem pharmacokinetics are also altered in severely 65 burned patients5,10,11and within our own unit, we previously reported the case of a 27 year old 66 man with a total body surface area (TBSA) burn of 52% in whom a dose increase to 1 g over 5 67 minutes every 4 h was needed to achieve target serum concentrations.12 No previous 68 population studies have examined intraindividual variability in pharmacokinetic parameters in 69 this patient group. 70 Since meropenem demonstrates time-dependent killing at clinically relevant 71 concentrations,13the most important pharmacodynamic (PD) index to predict antimicrobial 72 efficacy is the percentage of the dosing interval that the antibiotic concentrations remain above 73 the MIC of the pathogenic organism (T ). Many PD studies have selected a T for at least >MIC >MIC 74 40% of the dose interval as the target.14-19 However, as treatment with meropenem is often 75 empiric, the MIC is not known. Whilst the EUCAST 201320 susceptibility breakpoint of 2 mg/L 76 could be selected as the target MIC, a 2009 study of meropenem activity against nosocomial 77 isolates across Europe found 79% of Pseudomonas aeruginosa isolates susceptible at a 78 breakpoint of 4 mg/L21 suggesting this might be a more suitable target. However, such 79 considerations should always be based on local epidemiology, where it is available. In this 80 context, dosage regimens can be optimised through integration of PK-PD targets, derived from 81 both PK data and exposure-response data, with Monte Carlo simulation to predict the probability 82 of attaining a specific PD target at various dosage regimens.15,22 83 The aim of this study was to determine the PK profile of intravenous meropenem given at an 84 initial dose of 1g over 5 minutes every 8 h to adult patients admitted to hospital with severe 85 burns, to develop a population model to describe the PK of meropenem in this patient group, 86 and to use Monte Carlo simulation techniques to investigate potential relationships between 87 dosage regimens and the achievement of PK/PD targets. 88 89 PATIENTS AND METHODS 90 Patients 91 Adults admitted to a Regional Burns Centre with a major burn (defined as a TBSA burn of at 92 least 15%), receiving meropenem, were eligible for inclusion in the study. Consent was 93 obtained from patients who were deemed fit to give it. For incapacitated adults, assent was 94 obtained from the next of kin, and consent to use the data sought retrospectively from those 95 patients who survived their injury. The study was approved by the Trust Research and 96 Development Committee, the National Ethics Research Committee 3/3/045 and the MHRA 97 (Reference 21310/0001/001-002). 98 Patient demographics (gender, age, weight and height), burn details (TBSA burn, full and partial 99 thickness burn surface area and percentage burn remaining at time of diagnosis of infection), 100 routine clinical data (e.g. serum creatinine and serum albumin) and antibiotic prescribing 101 information were collected for each patient. In addition, the following data were recorded: post- 102 burn day when blood samples were taken; length of stay in the Intensive Therapy Unit (ITU); 103 Abbreviated Burn Severity Index (ABSI) Score23 and patient outcome. 104 Study protocol 105 Initial courses of meropenem commenced at a standard dose of 1 g over 5 minutes every 8 h, 106 as per Trust antimicrobial guidelines. After at least 24 h of therapy, blood samples were taken at 107 the following times: predose; 30 minutes, 1, 2 and 4 h post dose; and immediately before the 108 next dose. Blood samples (3 mL) were collected using serum gel tubes, centrifuged at 4,500 109 rpm, then the resulting serum was separated into plain 2 mL plastic tubes, stored and 110 transported at 40C for analysis within 24 hours, in line with previous published stability data.24 111 Samples were analysed by HPLC at the National Antimicrobial Reference Laboratory (approved 112 by Clinical Pathology Accreditation Ltd (UK)) using a previously reported method.25 This has a 113 lower limit of detection of 0.3 mg/L and a limit of quantification of 1 mg/L, where the intra and 114 inter assay coefficient of variation (CV)% were both less than 10%. The results were reported 115 within 24 hours and the dosage regimen was then modified if necessary and when the length of 116 course allowed to maintain concentrations above 4 mg/L for at least 40% of the dose interval. If 117 a patient required a second or third course of meropenem, the decision of what starting dose to 118 use was influenced by results from the initial course and the patient’s clinical condition. Serum 119 concentrations were measured and doses amended as described for initial courses. 120 Pharmacokinetic analysis 121 A population PK modelling approach was applied to the data using NONMEM Version 7.2.26 122 (ICON Development Solutions, Ellicott City, MD, USA) with first order conditional estimation and 123 interaction (FOCEI). Post-processing of the NONMEM results was performed with R 2.1.4.027 124 and diagnostic plots were performed with Xpose version 4 programmed in R 2.1.4.0.28 125 Based on a graphical exploratory analysis, an open, two-compartment model with zero order 126 input and linear elimination and linear distribution from the central to peripheral compartment 127 was selected to describe the meropenem plasma concentrations after intravenous 128 administration. This model was parameterized in terms of CL, central volume of distribution (V ), 1 129 intercompartmental clearance (Q) and volume of distribution of the peripheral compartment (V ). 2 130 Observed C was defined as the measured serum concentration at 30 minutes in each max 131 patient. Individual parameter estimates were obtained from the Empirical Bayes Estimates 132 (EBEs) and were used to calculate half-lives; AUC was calculated from the total daily dose 0-24 133 and individual estimates of CL. 134 Log-normal distributions were assumed for between-subject variability (BSV) and between 135 occasion variability (BOV) in the PK parameters; an “occasion” was defined as a set of 136 concentration-time data collected over one day. A proportional model was used to describe the 137 residual error. The shrinkage of the EBE of the BSV were calculated as previously suggested.29 138 Once the base model had been identified and, in the absence of significant shrinkage, EBE of 139 the BSV were used to identify potential relationships between individual PK estimates and the 140 clinical covariates gender, age, weight (using linear and allometric relationships), serum 141 creatinine concentration, measured CL , serum albumin, percentage of TBSA burn, CR 142 percentage of full and partial thickness burn surface area, percentage burn remaining at time of 143 diagnosis of infection and post-burn day. These covariates were first examined using scatter 144 plots then added to and removed from the population model in a stepwise manner.30 145 Improvements in the fit obtained with each model were assessed in several ways. First, the 146 NONMEM generated objective function value (OFV) was used to perform the likelihood ratio 147 test. A decrease in OFV of ≥10.83 was required to reach statistical significance (p = 0.001) for 148 the addition of one fixed effect in a hierarchical model. In addition, improvement in the fit was 149 assessed by reductions in the BSV, BOV, residual variability and standard errors of the 150 parameter estimates. Diagnostic plots and shrinkage were also examined.29 151 The final population model was evaluated in three ways: a non-parametric bootstrap sampling 152 procedure with 1,000 samples was conducted using PsN toolkit31 and a prediction-corrected 153 visual predictive check (pcVPC) was based on 1,000 simulations.32 Finally, normalised 154 prediction distribution errors (NPDE) obtained from 10,000 simulations were computed using the 155 software developed by Brendel et al.33 156 Pharmacodynamic simulations 157 The final PK model was used for simulations that were undertaken to explore the role of the 158 dosage regimen on the probability of target attainment (PTA). The final parameters of the 159 population PK model were used to generate individual total drug concentration profiles for each 160 of the 1,000 simulated patients using NONMEM. The clinical characteristics of the simulated 161 patients mirrored those of the original patient group. Simulations were performed for four steady 162 state dosage regimens given by bolus injection over 5 minutes: 1 g every 8 h; 2 g every 8 h; 1 g 163 every 6 h; 1 g every 4 h. In addition, three 3 hour infusion regimens: 1 g every 8 h; 2 g every 8 h 164 and 1 g every 6 h and steady state concentrations arising from three continuous infusions: 3, 4 165 and 6 g over 24 h were simulated. For evaluation of these dosage regimens, MIC values were 166 chosen across the range 0.125-128 mg/L. In each patient, the time that the drug concentration 167 remained above the MIC was calculated as the cumulative percentage of the dosage period.34 168 For each MIC and dosing regimen, PTA was defined as the probability of 1000 simulated 169 patients achieving the target T for at least 40%, 60% or 80% of the dose interval. For each >MIC 170 meropenem regimen, the highest MIC at which the PTA was ≥ 90% was defined as the PK-PD 171 susceptible breakpoint. 172 A second analysis was conducted using MIC distributions of Escherichia coli, coagulase 173 negative Staphylococcus, P. aeruginosa and Enterococcus faecalis derived from the EUCAST 174 database.20 These MIC distributions were extracted from 8005 strains of E. coli, 143 strains of 175 coagulase negative Staphylococcus, 57505 strains of P. aeruginosa and 12369 strains of E. 176 faecalis. The cumulative fraction of response (CFR) was used to estimate the overall response 177 of pathogens to meropenem for each of the ten dosage regimens, subdivided according to CL. 178 This estimate accounts for the variability of drug exposure in the population and the variability in 179 the MIC combined with the distributions of MICs for the pathogens. For each MIC, the fraction of 180 simulated patients who met the PD target was multiplied by the fraction of the distribution of 181 microorganisms for each MIC. The CFR was calculated as the sum of fraction products over all 182 MICs. 183 RESULTS 184 185 Patient Demographics 186 Twelve patients (7 male) were recruited to the study with a mean age at the time of the first 187 course of 46 years (range 27 to 73).The median percentage of TBSA burn was 41% (range 20 188 to 80) and the median ABSI Score was 10 (range 5 to 12). Most burns (n = 10) resulted from 189 flame injuries; inhalation injury was present in 7 cases. All patients were mechanically 190 ventilated, spending a median of 40.5 days in intensive care (range 19 to 119 days). Five did 191 not survive their injury. The following pathogenic bacteria were isolated: coagulase negative 192 Staphylococcus in 9 patients; P. aeruginosa in 4 patients, mixed coliforms and Enterococcus 193 spp in 4 patients, E. coli, Stenotrophomonas maltophilia and Enterobacter cloacae in 3 patients. 194 Other microorganisms found were E. faecalis, Bacillus cereus, Staphylococcus aureus, 195 Acinetobacter baumannii, Haemophilus influenzae, Klebsiella spp and Proteus mirabilis. 196 In general, renal function was not impaired at the time of recruitment into the study and none of 197 the patients required renal replacement therapy. The median (range) of serum creatinine was 198 41 mol/L (22 to 112) and of measured CL was 136.5 ml/min (60 to 217). Measured CL CR CR 199 was only available for 8 of the 12 patients. 200 Serum Concentration-Time Profiles 201 A total of 113 plasma meropenem concentration measurements were available, with a median 202 of 9 (range 4-24) measurements per patient. One high trough concentration that was 203 inconsistent with all other data from the same patient was removed from the analysis. Overall, 204 there were 20 sets of measurements (occasions); 7 patients had one occasion, 3 patients had 205 two occasions, 1 patient had three occasions and 1 patient had four occasions. Individual 206 concentration-time profiles are presented in Figure 1. 207 Patients initially received a standard intravenous infusion of meropenem over 5 minutes at 208 doses of 1 g every 8 h for 3-5 consecutive days. In seven patients, sub-optimal serum 209 concentrations were reported, which required an increase in the frequency of administration in 210 three patients to 1 g every 6 h, in one patient to 2 g every 8 h and to 1 g every 4 h in one 211 patient. Observed C ranged from 9.2 to 79.2 mg/L with a mean (SD) of 28.4 (16.1) mg/L max 212 while the pre-dose trough ranged from 0.3 to 19.2 mg/L with a mean (SD) of 2.8 ±4.2 mg/L. 213 Pharmacokinetic Analysis 214 An open two compartment disposition model with zero order input and linear elimination and 215 distribution adequately described the time course of plasma concentration following meropenem 216 administration. 217 All parameters were linearly related to body weight. Scatterplots of individual estimates of the 218 parameters against the measured and derived clinical and demographic data identified 219 additional potential relationships between CL and age, measured CL , serum albumin, TBSA CR 220 burn, full thickness burn surface area and percentage burn remaining at time of diagnosis of 221 infection. Relationships were identified between V and V and serum albumin; only the inclusion 1 2 222 of age on CL and serum albumin on CL, V and V achieved statistically significant reductions in 1 2 223 the OFV when included individually in the population model. A further improvement in the fit was 224 achieved by including BOV in CL in the model. The final population model reduced the OFV 225 from 385.5 (base model) to 276.0 and had the following structure: 226 CL = 0.196 L/h/kg x (1-0.023 x (age - 46)) x (1- 0.049 x (albumin-15)) 227 V = 0.273 L/kg x (1 - 0.049 x (albumin-15)) 1 228 Q = 0.199 L/h/kg