|Home | About | Journals | Submit | Contact Us | Français|
Ertapenem is a broad-spectrum carbapenem antibiotic whose activity against Mycobacterium tuberculosis is being explored. Carbapenems have antibacterial activity when the plasma concentration exceeds the MIC at least 40% of the time (40% TMIC). To assess the 40% TMIC in multidrug-resistant tuberculosis (MDR-TB) patients, a limited sampling strategy was developed using a population pharmacokinetic model based on data for healthy volunteers. A two-compartment population pharmacokinetic model was developed with data for 42 healthy volunteers using an iterative two-stage Bayesian method. External validation was performed by Bayesian fitting of the model developed with data for volunteers to the data for individual MDR-TB patients (in which the fitted values of the area under the concentration-time curve from 0 to 24 h [AUC0–24, fit values] were used) using the population model developed for volunteers as a prior. A Monte Carlo simulation (n = 1,000) was used to evaluate limited sampling strategies. Additionally, the 40% TMIC with the free fraction (f 40% TMIC) of ertapenem in MDR-TB patients was estimated with the population pharmacokinetic model. The population pharmacokinetic model that was developed was shown to overestimate the area under the concentration-time curve from 0 to 24 h (AUC0–24) in MDR-TB patients by 6.8% (range, −17.2 to 30.7%). The best-performing limited sampling strategy, which had a time restriction of 0 to 6 h, was found to be sampling at 1 and 5 h (r2 = 0.78, mean prediction error = −0.33%, root mean square error = 5.5%). Drug exposure was overestimated by a mean percentage of 4.2% (range, −15.2 to 23.6%). When a free fraction of 5% was considered and the MIC was set at 0.5 mg/liter, the minimum f 40% TMIC would have been exceeded in 9 out of 12 patients. A population pharmacokinetic model and limited sampling strategy, developed using data from healthy volunteers, were shown to be adequate to predict ertapenem exposure in MDR-TB patients.
Ertapenem (ETP) is a broad-spectrum carbapenem antibiotic used against a range of infectious diseases (1). Like for all other beta-lactam antimicrobial products, the efficacy of ertapenem is characterized by time-dependent killing. Carbapenems have antibacterial activity when the plasma concentration exceeds the MIC at least 40% of the time (40% TMIC) (1, 2). Although it has not yet been studied in tuberculosis (TB) patients, the 40% TMIC with the free fraction (f 40% TMIC) of drug is expected to be an important pharmacodynamic parameter (3). Interest in the use of carbapenems in combination with clavulanic acid was created when it was shown that they have activity in a murine model of TB (3). Additionally, a recent study showed that carbapenems efficiently inactivated peptidoglycan cross-linking in Mycobacterium tuberculosis (3, 4), and a recent study of the early bactericidal activity of meropenem and amoxicillin-clavulanic acid showed that carbapenems have activity in patients with TB (5). A new susceptibility testing method to estimate the MIC of ertapenem was recently introduced (6), and it was shown that ertapenem might be more potent in vitro than was previously thought because its chemical degradation had never been taken into account (3). To date, only a limited number of multidrug-resistant tuberculosis (MDR-TB) patients have been treated with ertapenem as part of a multidrug regimen. On the basis of these data, the drug appeared to be well tolerated during prolonged treatment (7, 8). However, ertapenem has not yet been added to the World Health Organization (WHO) list of anti-TB drugs, in contrast to imipenem and meropenem.
The pharmacokinetics of ertapenem have typically been studied in healthy volunteers (9), people with obesity (10, 11), patients with renal failure (12,–14), and critically ill patients with various pathologies (15,–17). Lower levels of drug exposure were observed in obese individuals (11), and an increase in the dosing interval was needed in patients with renal insufficiency with an estimated glomerular filtration rate (eGFR) of less than 30 ml/min/1.73 m2 (13), suggesting that the optimal dose of ertapenem is different in patients with different health conditions. A recent study on the pharmacokinetics of drugs in MDR-TB patients suggested that there was substantial variability in these patients (7).
For studies exploring the use of ertapenem against M. tuberculosis, it would be valuable to assess the f 40% TMIC of ertapenem in patients. To calculate the f 40% TMIC, a good indication of the plasma concentration profile is mandatory. However, measurement of the plasma concentration over the entire 24-h dosing interval is time-consuming, expensive, and burdensome for the patients. A limited sampling strategy can be used to predict this plasma concentration profile through the use of a population pharmacokinetic model, as has been done for other anti-TB drugs (18,–21).
The aim of this study was to develop a population pharmacokinetic model and a limited sampling strategy on the basis of data from healthy volunteers, in order to estimate ertapenem exposure in MDR-TB patients.
Data for 42 healthy volunteers were used to develop the population pharmacokinetic model. Since blood samples were collected from MDR-TB patients for another purpose, no data from between 5 and 8 h after drug administration were available.
All baseline characteristics of the healthy volunteers and the MDR-TB patients except for age were shown to differ significantly (P < 0.05) (Table 1). The median age of the volunteers was 31 years (range, 23 to 38 years), and the body mass index was 24.5 kg/m2 (range, 23.6 to 26.2 kg/m2).
The selection of the two-compartmental model was based on Akaike information criterion (AIC) values for one-compartment (AIC = 1,280) and two-compartment (AIC = −1,073) models (22). The final population pharmacokinetic model parameters developed with data for healthy volunteers (n = 42) are shown in Table 2.
The area under the concentration-time curve from 0 to 24 h (AUC0–24) for all except three healthy volunteers (AUC0–24, n − 3) estimated with the population pharmacokinetic model values in the internal validation was compared with the area under the concentration-time curve from 0 to 24 h used as a reference value (AUC0–24, ref). The results, shown in Fig. 1, were underestimated by a mean value of 0.3% (range, −8.1 to 7.6%). The observed AUC0–24, ref and the model-calculated AUC0–24, n − 3 for ertapenem were assessed for agreement using Passing and Bablok regression (Fig. 2).
The pharmacokinetic parameters for 18 healthy volunteers, who received 1 g of ertapenem, compared to those for TB patients are shown in Table 3. The fitted AUC0–24 (AUC0–24, fit) values for the MDR-TB patient data were underestimated by a mean of 6.8% (range, −17.2 to 30.7%) when the values were compared with the AUC0–24, ref values. The values of AUC0–24, fit correlated well with the values of AUC0–24, ref, as determined by a Bland-Altman analysis. The results are shown in Fig. 3.
Using the population pharmacokinetic model, limited sampling strategies were evaluated for restrictions of the dosing interval of from 0 to 6 h, 0 to 12 h, and 0 to 24 h. The r2, bias, and root mean square error (RMSE) values were subsequently determined. For each dosing interval and for one, two, or three sampling time points, the limited sampling strategies providing the best performance according to the values of RMSE and bias are shown in Table 4. All limited sampling strategies met the bias and RMSE criteria. The use of three sampling time points, at 1, 4 and 9 h, enabled the best prediction of ertapenem exposure, reflected by the AUC0–24 obtained by the limited sampling strategy (AUC0–24, LSS) when the values for bias, RMSE, and r2 were considered (r2 = 0.92, mean prediction error [MPE] = −0.46%, RMSE = 4.7%). However, due to the lack of clinical data from within these time intervals and the clinical relevance of the sampling times within a certain amount of time, it would be preferred to use sampling time points of 1, 3, and 5 h (r2 = 0.83, RMSE = 4.7%, MPE = −0.39%).
On the basis of clinical suitability, the use of two sampling time points with a time restriction of 0 to 6 h, a limited sampling strategy at 1 and 5 h showed the lowest RMSE (5.5%) and a low MPE (−0.33%). AUC0–24, LSS values, estimated by applying this two-sampling-time-point limited sampling strategy, were compared with AUC0–24, ref values using Bland-Altman analysis, and a bias in AUC0–24, LSS of 4.2% (range, −15.2 to 23.6%) was shown (Fig. 4).
When a free fraction of 5% was considered and the MIC was set at 0.5 mg/liter, the minimum f 40% TMIC (range, 6.8 h to 19.7 h) would have been exceeded in 9 out of 12 patients; thereby, ertapenem would have a sufficient therapeutic effect in MDR-TB patients with once daily dosing.
This is the first study showing that a population pharmacokinetic model of ertapenem based on data for healthy volunteers can predict the pharmacokinetics of ertapenem in patients with MDR-TB, even though the baseline characteristics of both healthy volunteers and MDR-TB patients differed significantly (Table 1). We showed that the AUC0–24 for MDR-TB patients can be estimated with this population pharmacokinetic model with a mean overestimation of 6.8% (range, −17.2 to 30.7%).
The robustness of this population pharmacokinetic model was validated using an n − 3 cross-validation, which showed an underestimation of 0.3%. The limited sampling strategy that we present here can be used to assess the level of exposure to ertapenem in individual TB patients with limited treatment options. Moreover, the model and the limited sampling strategy can be used to evaluate the level of exposure to ertapenem in phase II studies evaluating the early bactericidal activity of ertapenem in TB patients. Such a study is urgently needed to provide data on the efficacy of this potentially attractive drug with activity against M. tuberculosis.
In the population pharmacokinetic model, multiple doses were treated as single doses on day 1 to avoid duplication, as an earlier study found that there was no accumulation of ertapenem following dosing over 8 days, and the mean plasma concentrations were found to be very similar on day 1 as well as on day 8 (8).
Pharmacokinetic modeling of ertapenem has been performed in previous studies (10,–17, 23,–25), but it has never been performed for application to the treatment of MDR-TB. Comparing healthy volunteers and TB patients, we found that there was a low level of variability in the pharmacokinetic parameters between the two groups; in contrast, the pharmacokinetic parameters of other antimicrobial drugs show high levels of variability (18, 19). This might be explained by the parenteral route of administration of ertapenem, in which there is no loss of ertapenem due to absorption. Several studies looking at the level of exposure to ertapenem have shown that it varies greatly among patients (13, 14). There was little variability in the values of the pharmacokinetic parameters among the MDR-TB patients (Table 3).
A limited sampling strategy using two sampling time points is favored, as it would place the smallest burden on patients because it is minimally invasive and the least time-consuming. Additionally, less time between sampling time points is more feasible in clinical practice, since it makes the collection of samples at incorrect times less likely. Moreover, a limitation of this study is that after 6 h, sparse data were available; therefore, the results obtained after 6 h were less well substantiated by clinical data.
As there are limited options for treating MDR-TB and resistance to antibiotics is an emerging problem, we think that it is time to start assessing the efficacy of ertapenem in MDR-TB patients in a phase II clinical trial testing early bactericidal activity. The limited sampling strategy developed here can be used to evaluate drug exposure and thereby reduce study costs and the burdens for study subjects.
A pharmacokinetic model and a limited sampling strategy based on data for healthy volunteers were able to predict the AUC0–24 and f 40% TMIC in MDR-TB patients. This model can be used in phase II studies.
This study was based on two data sets. The first data set was comprised of data for 42 healthy volunteers receiving 0.25- to 2-g intravenous doses of ertapenem in five clinical studies (9). For comparison of the population pharmacokinetic model for healthy volunteers with that for MDR-TB patients, we used only the data for healthy volunteers receiving ertapenem at 1 g. The second data set comprised a retrospectively collected set of data for patients with MDR-TB receiving 1 g of ertapenem administered once daily via a 30-min infusion at the Tuberculosis Center Beatrixoord, University Medical Center Groningen, The Netherlands, between 1 December 2010 and 1 March 2013 (7). Samples for determination of plasma ertapenem concentrations were collected at steady state before administration and at 1, 2, 3, 4, 5, 6, 8, and 12 h after administration. Plasma ertapenem concentrations were analyzed by a validated liquid chromatography-tandem mass spectrometry (LC-MS/MS) method (26). Both data sets included demographic and medical data, such as age at the start of treatment, height, body weight, and the serum creatinine concentration at the time of pharmacokinetic assessment. (The study was evaluated by the Medical Ethical Review Board of the University Medical Center Groningen [METc 2013-492]. The need for written informed consent was waived for the retrospective collection and analysis of anonymous data because it was not required under Dutch Law [WMO].)
All pharmacokinetic calculations were performed using the MW/Pharm software package (version 3.82; Mediware, Zuidhorn, The Netherlands). On the basis of the findings presented in previous reports and the findings of recent pharmacokinetic studies of ertapenem (10,–17, 23,–25), concentration-time curves were evaluated in one-compartment and two-compartment models. The final model was selected on the basis of the AIC (27). The plasma drug concentrations for the 42 healthy volunteers were used to develop a two-compartment population pharmacokinetic model using an iterative two-stage Bayesian (ITSB) procedure (the KinPop module of the MW/Pharm software package) (22). Clearance (CL) was calculated using the equation (CLm · BSA)/(1.85 + fr · CLCR), where CLm is metabolic clearance (in liters per hour per 1.85 m2), BSA is the body surface area (in square meters), fr is the drug clearance/creatinine clearance ratio, and CLCR is creatinine clearance (in liters per hour) (23). Pharmacokinetic parameters were assumed to be log normally distributed, and the residual error was assumed to be normally distributed. The standard deviation (SD) was equal to 0.1 + (0.1 · C), where C is the observed plasma ertapenem concentration. The nonparametric 95% confidence intervals of the population parameters and their interindividual standard deviations were estimated by bootstrap analysis (n = 1,000). The area under the plasma concentration-time curve from 0 to 24 h used as a reference value (AUC0–24, ref) was calculated using the log trapezoidal rule (in the KinFit module of the MW/Pharm software package).
Internal validation was performed by leaving the data for three healthy volunteers, obtained by randomization using Microsoft Excel 2010 software, out of pharmacokinetic model development, creating 14 n − 3 submodels (14 submodels without the data for three healthy volunteers). The AUC0–24 estimated from these data (AUC0–24, n − 3) was obtained by Bayesian fitting using the data for the three volunteers left out of the corresponding n − 3 submodels. The agreement between AUC0–24, n − 3 and AUC0–24, ref was assessed by use of Bland-Altman analysis, Passing and Bablok regression, and a subsequent residual plot. External validation was performed by Bayesian fitting of the model developed with the data for the volunteers to the data for individual MDR-TB patients (AUC0–24, fit), using the population model developed with data from the volunteers as a prior. For comparison of the pharmacokinetics between MDR-TB patients and volunteers, a similar analysis was performed with the data for the 18 volunteers who received 1 g of ertapenem. Bland-Altman analysis was also used to assess the agreement between AUC0–24, fit and AUC0–24, ref for the MDR-TB patients.
A Monte Carlo simulation was used to calculate the values of AUC0–24 estimated by the limited sampling strategies (AUC0–24, LSS values), as implemented in the MW/Pharm software package. This stochastic simulation consisted of data for 1,000 random patients drawn from the population pharmacokinetic model. For each patient, limited sampling strategies were calculated by the use of 1 to 3 sampling time points and a Bayesian maximum a posteriori procedure. We evaluated limited sampling strategies on the basis of separate calculations with time restrictions of 0 to 6 h, 0 to 12 h, and 0 to 24 h. Performance was considered suitable for application in prospective studies if the adjusted r squared (r2) value was >0.95, the root mean square error (RMSE) value was <15%, and the mean prediction error (MPE) was <5%. The prediction errors were calculated as [(AUC0–24, LSS − AUC0–24, ref)/AUC0–24, ref] · 100.
The ertapenem concentration-time curve for each patient was used to establish whether the f 40% TMIC was reached. For this purpose, the time that the concentration in the concentration-time curve was above the MIC was assessed by use of the MW/Pharm software package. The percentage of ertapenem unbound to protein used for the assessment was 5 (3, 4). The European Committee on Antimicrobial Susceptibility Testing (EUCAST) MIC value for ertapenem (non-species related) of 0.5 mg · liter−1 was used to calculate f 40% TMIC (3, 6). Exposure was considered adequate if the concentration was above the MIC 40% of the time. This value corresponds to 9.6 h in each 24-h interval, as shown in Fig. 5.
Differences between the population characteristics and pharmacokinetic parameters of healthy volunteers and TB patients were calculated using the Mann-Whitney U test. All statistics were calculated with Analyze-it for Microsoft Excel software (version 2.30).
We thank Merck & Co., Inc., USA for providing data from earlier pharmacokinetic studies of ertapenem.
We have no conflict of interest to declare.