Pharmacokinetics and pharmacodynamic target attainment of ceftriaxone in adult severely ill sub-Saharan African patients: a population pharmacokinetic modelling study

Pharmacokinetics and pharmacodynamic target attainment of ceftriaxone in adult severely ill... Abstract Background In sub-Saharan Africa (SSA), the highly albumin-bound β-lactam ceftriaxone is frequently used for the empirical treatment of severe bacterial infections. Systemic drug exposure of β-lactams can be altered in critically ill ICU patients, but pharmacokinetic and pharmacodynamic data for non-ICU SSA populations are lacking. Methods We performed a population pharmacokinetic study in an adult hospital population in Mozambique, treated with ceftriaxone for presumptive severe bacterial infection from October 2014 to November 2015. Four blood samples per patient were collected for total ceftriaxone (CEFt) and unbound ceftriaxone (CEFu) concentration measurement. We developed a population pharmacokinetic model through non-linear mixed effect analysis and performed simulations for different patient variable, dosing and pharmacodynamic target scenarios. Results Eighty-eight participants yielded 277 CEFt and 276 CEFu concentrations. The median BMI was 18.9 kg/m2 and the median albumin concentration was 29 g/L. In a one-compartment model with non-linear protein binding, creatinine clearance was positively correlated with CEFu clearance. For microorganisms with an MIC of 1 mg/L, simulations demonstrated that with a 1 g twice-daily regimen and a 2 g once-daily regimen, 95.1% and 74.8% would have a CEFu concentration > MIC during half of the dosing interval (fT>MIC = 50%), respectively, whereas this was only 58.2% and 16.5% for the fT>MIC = 100% target. Conclusions Severely ill adult non-ICU SSA patients may be at substantial risk for underexposure to CEFu during routine intermittent bolus dosing, especially when their renal function is intact. Introduction In sub-Saharan African (SSA), a region of the world with high HIV infection rates, 10%–34% of patients admitted with fever suffer from bacterial bloodstream infections or sepsis.1–3 A high proportion of these infections are caused by Streptococcus pneumoniae and non-typhoidal Salmonella serovars and mortality rates can be as high as 46%–72%.4,5 The third-generation cephalosporin ceftriaxone is among the most frequently used antibiotics for the empirical treatment of adults in this situation. Evidence from resource-rich ICU settings suggests that appropriateness of antibiotic treatment should not only be reviewed in terms of antibiotic drug choice but also in terms of systemic drug exposure.6,7 Pharmacokinetics (PK) of antibiotics in critically ill patients can be substantially altered. β-Lactam antibiotics are particularly vulnerable, as they are hydrophilic drugs with renal CL as the predominant route of elimination. Unlike most β-lactams, ceftriaxone is highly bound to albumin (85%–95%) at therapeutic concentrations. During sepsis and hypoalbuminaemia a lower protein-bound drug fraction, an increased volume of distribution (V) and increased CL may change the total as well as the unbound, active drug concentration.6,7 Such alterations may give rise to an inability to attain pharmacodynamic (PD) targets, including the increasingly promoted PD target of the unbound drug concentration remaining above the MIC throughout the dosing interval (fT>MIC = 100%).8 Ultimately, underexposure may lead to adverse clinical outcome.9,10 In SSA, highly prevalent chronic diseases such as TB, hepatitis B and hypertension also have the potential to influence the PK of antibiotics by means of cachexia and liver and kidney dysfunction.11–13 How coinciding acute and chronic conditions in severely ill (non-ICU) SSA patients influence the PK of β-lactam antibiotic drugs, including ceftriaxone, has been poorly investigated. What is clear is that underexposure to the unbound drug may not only pose a threat to an individual patient’s health, but also to public health, as it contributes to the emergence of antimicrobial resistance, a phenomenon that is already highly prevalent in SSA countries.14,15 In view of the above, we performed a population PK (PPK) study of ceftriaxone in a Mozambican, adult hospital medicine ward population. The specific aims of the study were to describe the PPK of unbound ceftriaxone (CEFu) and total ceftriaxone (CEFt) in order to identify sources of PK parameter variability. Additionally, we aimed to assess the probability of PK/PD target attainment (PTA) of CEFu for the treatment of bacterial pathogens commonly causing sepsis in SSA. Methods Setting The Beira Central Hospital (HCB) is a Mozambican governmental referral health facility. The proportion of patients infected with HIV on its medicine ward may be as high as 74% and up to 32% of patients may die during their hospital stay.16 Study design The current study was a prospective, observational PPK study of ceftriaxone, as part of a PPK study of commonly used antibiotics among patients admitted to the HCB medicine ward. In this study, PK data were collected from October 2014 until November 2015 from patients who were treated with intravenously administered ceftriaxone. The study was reviewed and approved by the Mozambican National Committee for Bio-ethics in Health (CNBS: study registration no. 118/CNBS/2013). Additionally, a letter of approval was obtained from the HCB director. Participants gave written informed consent. Those unable to read, write and/or understand Portuguese gave a thumbprint and an impartial, literate witness observed the entire informed consent process and subsequently co-signed the informed consent form. Recruitment and data collection Patients were eligible for study entry if they were hospitalized on the HCB medicine ward while being treated with ceftriaxone as prescribed by an HCB physician, and as documented in a patient’s medication record. Any ceftriaxone dosing regimen that was allowed for use by the Mozambican National Institute of Health (INS) was accepted.17 Inclusion criteria were age ≥18 years and being willing and able to give informed consent. Exclusion criteria were the use of drugs known to significantly affect the PK of the study antibiotics (probenecid, phenylbutazone, acetylsalicylic acid and indomethacin), a haemoglobin level ≤6 g/dL as measured by the HCB laboratory (haemoglobin safe limit adapted from the North Shore Human Subject Protection Program Guidance Document on maximum blood draw limits18), any condition necessitating a blood transfusion irrespective of haemoglobin level, and an altered level of consciousness. Two trained research nurses captured baseline characteristics and ceftriaxone dosing information, and measured the body weight and height of all study participants. Doses of 1 g of ceftriaxone powder for injection (Nirlife, Nirma Ltd, Gujarat, India) were dissolved in 10 mL of sterile water for injection and subsequently injected intravenously via a venous catheter in half a minute, according to the responsible physician’s prescription. Over a time period of two or more days, a maximum of four blood samples were collected for ceftriaxone concentration measurement. Sample times were pre-dose (trough level), 30–120 min after intravenous administration (peak level) and two random timepoints during the dosing interval (random levels). Ceftriaxone administration procedures were observed where possible and a maximum of 19 mL of blood was collected. One blood sample was also used for the measurement of albumin, AST, ALT, GGT and creatinine concentrations. Bilirubin levels were not measured for practical reasons in relation to bilirubin’s high photosensitivity. CLCR was estimated using the Cockcroft and Gault formula.19 Sample handling and drug assay EDTA anti-coagulated blood samples were refrigerated at 4–8°C immediately after collection until laboratory processing, which took place within two hours of collection. Samples were centrifuged and plasma was stored at −80°C in the local research laboratory until shipment on dry ice to the Netherlands for biochemical marker and drug concentration analysis. Plasma was ultrafiltrated (centrifugal filters: Millipore Amicon Ultra 0.5 mL/30K, Merck Millipore, Darmstadt, Germany) and the ultrafiltrated plasma was subsequently processed as a typical plasma sample to obtain the unbound concentration. CEFt and CEFu concentrations were measured using validated HPLC MS (LC: LC30 UPLC, Shimadzu, Kyoto, Japan; MS: QTRAP 5500 system, Sciex, Framingham, MA, USA). The lower limit of quantification (LLQ) was 1 mg/L and the higher limit of quantification was 40 mg/L. Concentrations higher than this were diluted and re-analysed. Within- and between-assay variability was lower than 5.6% and 6.8%, respectively. The accuracy of the assay was between 90% and 105%. PPK analysis Model development The PPK analysis was performed using the non-linear mixed-effect modelling software package NONMEM (7.1.2; Icon Development Solutions, Ellicott City, MD, USA). Model building was performed using a stepwise approach. During a first step, a structural, compartmental PPK model was developed in which the PK of CEFu and CEFt as well as protein binding of ceftriaxone were described, including their between-patient variability (BPV). In the second step, an attempt was made to explain BPV by building a covariate model in which patient demographics and pathophysiological factors were tested for their correlation with the estimated PK parameters. In the third and last step, the robustness and validity of the final model resulting from the second step were tested. Structural model One- and two-compartment models were tested. Estimated PK parameters were V, CL and, in the case of testing of a two-compartment model, peripheral V and intercompartmental CL. BPV of V and CL was estimated exponentially.20 Modelling of ceftriaxone protein binding and CEFt was done using linear and non-linear protein-binding models, according to the following equation: [CEFt] (mmol/L)=[CEFu]+[CEFu]×b Eqn 1 where b stands for ceftriaxone’s binding properties and [CEFu]×b stands for the bound ceftriaxone concentration. In the linear protein binding model, b is an estimated parameter, θbind, that correlates to the number of unoccupied binding places,21 and in the non-linear protein-binding model, b becomes Bmax/Km + [CEFu]. In this model, Bmax represents protein binding, defined as the maximum estimated concentration of ceftriaxone bound to albumin, and Km represents the CEFu concentration at which albumin binding is half maximal. The residual variability (i.e. the difference between measured CEFu and CEFt concentrations and the corresponding CEFu and CEFt concentrations predicted by the model) was modelled with additive or proportional models or a combination of both. The so-called M3 method was used for the handling of CEFt and CEFu concentrations below the LLQ.22 Covariate model Tested covariates included age, sex, weight, height, BMI, CLCR and haemoglobin, albumin, creatinine, GGT, ALT and AST concentrations. All covariates were screened for the significance of the correlation between the covariate and the PK parameter by univariate analysis, using a P value cut-off of 0.05. Furthermore, a reduction in BPV or residual variability, as well as biological plausibility of a covariate–PK parameter relationship was used as a criterion for covariate selection. All covariates selected during the univariate analysis subsequently entered an intermediate model for a backward elimination procedure (multivariate analysis) with a cut-off for statistical significance of 0.001, which yielded the final model. Potential improvement of the model by adding a compartment or by introducing a correlation between a covariate and a PK parameter was evaluated using the likelihood ratio test, in which the difference between the minimum objective function value (OFV) generated by NONMEM® for two hierarchical models is determined. Improvement in model fit was defined as an OFV decrease of ≥3.8 units while using a χ2 distribution with one degree of freedom, corresponding to a P value cut-off of 0.05. For the backward elimination covariate procedure an OFV decrease of ≥10.8 units was used. Model performance was also evaluated by visual inspection of diagnostic ‘goodness-of-fit’ plots.23 These were generated using Pirana (version 2.9.0) and Xpose (version 4.3.2) software (Uppsala Pharmacometrics, Uppsala, Sweden).24,25 Model robustness and predictive performance The robustness of the parameter estimates from the final model resulting from the second step was tested using a bootstrap analysis. In this analysis, the dataset was resampled 1000 times. Based on 1000 simulations, visual predictive checks (VPCs) investigated whether the final model could adequately predict the observed time course of CEFu and CEFt, including the observed variability. Bootstrap and VPC analyses were performed using Perl-speaks-NONMEM (PsN) version 3.5.3 software (Uppsala Pharmacometrics, Uppsala, Sweden).26 Monte Carlo dosing simulations Using the final PPK model, CEFu concentration–time profiles were predicted based on Monte Carlo simulations following three dosing regimens: 1 g twice daily (q12h), 2 g once daily (q24h) and 2 g q12h. Each simulation generated concentration–time profiles for 1000 virtual patients per drug regimen. Based on these data, the PTA, being the percentage of patients with a CEFu remaining above a specified MIC during a specified percentage of time of the dosing interval (fT>MIC), was calculated for different fT>MIC targets. The primary target to be tested was fT>MIC = 50% and the secondary target was fT>MIC = 100%. The choice of PD targets was based on conclusions of recent reviews concerning β-lactam administration and PD targets in critically ill patients.6,8,10 The choice of the target MIC was based on the EUCAST MIC clinical breakpoints for susceptibility to ceftriaxone of Enterobacteriaceae (1 mg/L) and S. pneumoniae (0.5 mg/L).27 Results Patients and ceftriaxone concentrations We screened 762 patients for the larger PPK study and excluded 366 patients (Figure 1). The most common reason for exclusion was a haemoglobin level ≤6 g/dL. We included 98 patients in the current study on ceftriaxone and 88 participants remained for analysis. A large majority (93.1%) had a 1 g q12h ceftriaxone dosing schedule. Patient characteristics are presented in Table 1. Table 1. Baseline characteristics of study population (N = 88) Female, n (%) 48 (54.5) Age (years), median (range) 35 (18–76) Bodyweight (kg), median (range) 50 (29–70) BMI (kg/m2), median (range) 18.9 (11.3–26.4) Haemoglobin (g/dL), median (range) 9.4 (6.3–16.4) Albumin (g/L), median (range) 29 (11–44) GGT (U/L), median (range) 49 (6–1276) ALT (U/L), median (range) 25 (0–676) AST (U/L), median (range) 50 (11–594) Creatinine (μmol/L), median (range) 63 (19–971) CLCRa (mL/min), median (range) 91 (4–261) Ceftriaxone dose prescribed, n (%)  1 g q12h 82 (93.1)  2 g q24h 2 (2.3)  2 g q12h 4 (4.6) Female, n (%) 48 (54.5) Age (years), median (range) 35 (18–76) Bodyweight (kg), median (range) 50 (29–70) BMI (kg/m2), median (range) 18.9 (11.3–26.4) Haemoglobin (g/dL), median (range) 9.4 (6.3–16.4) Albumin (g/L), median (range) 29 (11–44) GGT (U/L), median (range) 49 (6–1276) ALT (U/L), median (range) 25 (0–676) AST (U/L), median (range) 50 (11–594) Creatinine (μmol/L), median (range) 63 (19–971) CLCRa (mL/min), median (range) 91 (4–261) Ceftriaxone dose prescribed, n (%)  1 g q12h 82 (93.1)  2 g q24h 2 (2.3)  2 g q12h 4 (4.6) a Estimated using the Cockcroft–Gault equation.19 Table 1. Baseline characteristics of study population (N = 88) Female, n (%) 48 (54.5) Age (years), median (range) 35 (18–76) Bodyweight (kg), median (range) 50 (29–70) BMI (kg/m2), median (range) 18.9 (11.3–26.4) Haemoglobin (g/dL), median (range) 9.4 (6.3–16.4) Albumin (g/L), median (range) 29 (11–44) GGT (U/L), median (range) 49 (6–1276) ALT (U/L), median (range) 25 (0–676) AST (U/L), median (range) 50 (11–594) Creatinine (μmol/L), median (range) 63 (19–971) CLCRa (mL/min), median (range) 91 (4–261) Ceftriaxone dose prescribed, n (%)  1 g q12h 82 (93.1)  2 g q24h 2 (2.3)  2 g q12h 4 (4.6) Female, n (%) 48 (54.5) Age (years), median (range) 35 (18–76) Bodyweight (kg), median (range) 50 (29–70) BMI (kg/m2), median (range) 18.9 (11.3–26.4) Haemoglobin (g/dL), median (range) 9.4 (6.3–16.4) Albumin (g/L), median (range) 29 (11–44) GGT (U/L), median (range) 49 (6–1276) ALT (U/L), median (range) 25 (0–676) AST (U/L), median (range) 50 (11–594) Creatinine (μmol/L), median (range) 63 (19–971) CLCRa (mL/min), median (range) 91 (4–261) Ceftriaxone dose prescribed, n (%)  1 g q12h 82 (93.1)  2 g q24h 2 (2.3)  2 g q12h 4 (4.6) a Estimated using the Cockcroft–Gault equation.19 Figure 1. View largeDownload slide Study profile. Hb, haemoglobin concentration. Figure 1. View largeDownload slide Study profile. Hb, haemoglobin concentration. A total of 277 plasma samples yielded 277 CEFt and 276 CEFu concentrations. Fewer than four plasma samples were available for 43/88 (48.9%) participants, with the most common explanation being a participant’s unforeseen discharge (Figure 1). Three or four plasma samples were available for analysis for 67/88 (76.1%) participants. There were 33/276 (12%) CEFu samples with a concentration below the LLQ. The median unbound fraction was 19% (IQR = 13–29). The observed CEFt and CEFu concentrations are shown in Figure 2. Figure 2. View largeDownload slide Observed ceftriaxone concentration–time data and VPC of the final model. The black dots are the observed concentrations. The solid black line is the observed median and the dashed lines are the 5th and 95th percentiles of the observed data. The red shaded area is the 95% CI of the model-predicted median and the blue shaded areas are the 95% CIs of the model-predicted 5th and 95th percentiles. (a) CEFt. (b) CEFu. The black solid and dashed lines run within their respective shaded areas, thereby demonstrating adequate fit of the model. Figure 2. View largeDownload slide Observed ceftriaxone concentration–time data and VPC of the final model. The black dots are the observed concentrations. The solid black line is the observed median and the dashed lines are the 5th and 95th percentiles of the observed data. The red shaded area is the 95% CI of the model-predicted median and the blue shaded areas are the 95% CIs of the model-predicted 5th and 95th percentiles. (a) CEFt. (b) CEFu. The black solid and dashed lines run within their respective shaded areas, thereby demonstrating adequate fit of the model. PPK analysis Structural model During step one of the analysis, CEFu data were fitted to several compartmental models and a one-compartmental model best fitted the data, with the difference in OFV between a one- and a two-compartmental model being only 0.21 units. The introduction of non-linear protein binding in the model resulted in an OFV decrease of 110 units compared with linear protein binding. Additionally, the residual error dropped from 13.1 mg/L to 9.8 mg/L and non-linear protein binding therefore best described CEFt concentration and protein binding (Figure 3). Km was estimated to be 0.01 mmol/L while Bmax was 0.11 mmol/L. The estimated BPV for Bmax was 31% (Table 2). Table 2. Parameter estimates of the different model building steps Parameter Model without covariates Model with covariates Bootstrap of model with covariates estimate RSE (%) estimate RSE (%) estimate 95% CI CEFu CL (L/h) 11 11 11 10 11 9.3–15 CEFuV (L) 48 10 48 10 49 35–60 Km (mmol/L) 0.010 28 0.0092 33 0.0096 0.004–0.020 Bmax (mmol/L) 0.11 12 0.12 11 0.12 0.094–0.15 BPV  CEFu CL (%CV) 93 15 77 14 73 37–108  CEFuV (%CV)a 33 a 37 a 38 7.9–112  Bmax (%CV)b 31 20 — — — — Residual variability  additive error CEFu (mmol/L) 0.015 12 0.015 12 0.015 0.011–0.017  additive error CEFt (mmol/L) 0.036 7 0.036 6 0.035 0.031–0.040 Covariate effects  CLCR on CEFu CLc — — 0.37 41 0.38 0.083–0.79  albumin on Bmaxc — — 1.30 41 1.3 0.93–1.70 Parameter Model without covariates Model with covariates Bootstrap of model with covariates estimate RSE (%) estimate RSE (%) estimate 95% CI CEFu CL (L/h) 11 11 11 10 11 9.3–15 CEFuV (L) 48 10 48 10 49 35–60 Km (mmol/L) 0.010 28 0.0092 33 0.0096 0.004–0.020 Bmax (mmol/L) 0.11 12 0.12 11 0.12 0.094–0.15 BPV  CEFu CL (%CV) 93 15 77 14 73 37–108  CEFuV (%CV)a 33 a 37 a 38 7.9–112  Bmax (%CV)b 31 20 — — — — Residual variability  additive error CEFu (mmol/L) 0.015 12 0.015 12 0.015 0.011–0.017  additive error CEFt (mmol/L) 0.036 7 0.036 6 0.035 0.031–0.040 Covariate effects  CLCR on CEFu CLc — — 0.37 41 0.38 0.083–0.79  albumin on Bmaxc — — 1.30 41 1.3 0.93–1.70 CV, coefficient of variation. For model-building purposes, observed concentrations of CEFu and CEFt were converted from mg/L into mmol/L by dividing observed concentrations by the molar mass of ceftriaxone of 661.6 g/mol.42 a The BPV of CEFuV was modelled as a function of the BPV of CEFu,20 leaving the BPV without an RSE. b The observation that BPV was not estimated for Bmax in the final model does not mean that the BPV of Bmax is zero, and as such fully explained by the association between albumin level and Bmax, but rather that there is not enough information in the data to support a reliable estimate of the BPV. c See Eqn 2 and Eqn 3. Table 2. Parameter estimates of the different model building steps Parameter Model without covariates Model with covariates Bootstrap of model with covariates estimate RSE (%) estimate RSE (%) estimate 95% CI CEFu CL (L/h) 11 11 11 10 11 9.3–15 CEFuV (L) 48 10 48 10 49 35–60 Km (mmol/L) 0.010 28 0.0092 33 0.0096 0.004–0.020 Bmax (mmol/L) 0.11 12 0.12 11 0.12 0.094–0.15 BPV  CEFu CL (%CV) 93 15 77 14 73 37–108  CEFuV (%CV)a 33 a 37 a 38 7.9–112  Bmax (%CV)b 31 20 — — — — Residual variability  additive error CEFu (mmol/L) 0.015 12 0.015 12 0.015 0.011–0.017  additive error CEFt (mmol/L) 0.036 7 0.036 6 0.035 0.031–0.040 Covariate effects  CLCR on CEFu CLc — — 0.37 41 0.38 0.083–0.79  albumin on Bmaxc — — 1.30 41 1.3 0.93–1.70 Parameter Model without covariates Model with covariates Bootstrap of model with covariates estimate RSE (%) estimate RSE (%) estimate 95% CI CEFu CL (L/h) 11 11 11 10 11 9.3–15 CEFuV (L) 48 10 48 10 49 35–60 Km (mmol/L) 0.010 28 0.0092 33 0.0096 0.004–0.020 Bmax (mmol/L) 0.11 12 0.12 11 0.12 0.094–0.15 BPV  CEFu CL (%CV) 93 15 77 14 73 37–108  CEFuV (%CV)a 33 a 37 a 38 7.9–112  Bmax (%CV)b 31 20 — — — — Residual variability  additive error CEFu (mmol/L) 0.015 12 0.015 12 0.015 0.011–0.017  additive error CEFt (mmol/L) 0.036 7 0.036 6 0.035 0.031–0.040 Covariate effects  CLCR on CEFu CLc — — 0.37 41 0.38 0.083–0.79  albumin on Bmaxc — — 1.30 41 1.3 0.93–1.70 CV, coefficient of variation. For model-building purposes, observed concentrations of CEFu and CEFt were converted from mg/L into mmol/L by dividing observed concentrations by the molar mass of ceftriaxone of 661.6 g/mol.42 a The BPV of CEFuV was modelled as a function of the BPV of CEFu,20 leaving the BPV without an RSE. b The observation that BPV was not estimated for Bmax in the final model does not mean that the BPV of Bmax is zero, and as such fully explained by the association between albumin level and Bmax, but rather that there is not enough information in the data to support a reliable estimate of the BPV. c See Eqn 2 and Eqn 3. Figure 3. View largeDownload slide Covariate relationship between albumin and the CEFu fraction. Plasma albumin concentration versus observed (red dots) and individual predicted (blue dots) CEFu fraction, with the model-predicted relationship (black line) for virtual patients with the median CEFu concentration of 11.8 mg/L. The unbound fraction increases in a non-linear fashion with a decreasing albumin concentration. Figure 3. View largeDownload slide Covariate relationship between albumin and the CEFu fraction. Plasma albumin concentration versus observed (red dots) and individual predicted (blue dots) CEFu fraction, with the model-predicted relationship (black line) for virtual patients with the median CEFu concentration of 11.8 mg/L. The unbound fraction increases in a non-linear fashion with a decreasing albumin concentration. Covariate model The multivariate covariate analysis, based on 100% availability of covariate results, yielded a model with significant associations between CEFu CL and CLCR as well as between Bmax and albumin concentration, as shown by Eqn 2 and Eqn 3: CEFu CL (L/h)=11×(CLCR/91)0.37 Eqn 2 Bmax=0.12×(albumin concentration/0.42)1.3 Eqn 3 With these associations in the model, the BPV of CEFu CL decreased from 93% to 77%, although a substantial part of the BPV in CEFu CL remained unexplained (Table 2). The BPV of Bmax could, however, no longer be reliably estimated. Using the Bmax and albumin concentration equation (Eqn 3), Eqn 1 for the calculation of CEFt is transformed into Eqn 4: [CEFt] (mmol/L)=[CEFu]+[CEFu]×[0.12×(albumin/0.42)1.3]/(0.0092+[CEFu]) Eqn 4 This relationship between CEFt concentrations and the observed and predicted CEFu concentrations is shown in Figure 4. Figure 4. View largeDownload slide Total versus unbound ceftriaxone concentrations. Black dots, observed concentrations; blue dots, individual predicted concentrations; black line, model-predicted relationship between total and unbound concentration for a patient with a median albumin concentration of 29 g/L. Figure 4. View largeDownload slide Total versus unbound ceftriaxone concentrations. Black dots, observed concentrations; blue dots, individual predicted concentrations; black line, model-predicted relationship between total and unbound concentration for a patient with a median albumin concentration of 29 g/L. Model robustness and predictive performance The final model had an adequate fit as shown by the VPC (Figure 2). The maximum relative standard error (RSE) with which parameters could be estimated was 41% and bootstrap estimations were similar to the ones from the final model (Table 2). The residual variability, an estimate of unexplained variability relating to measurement error, errors in data collection, intra-patient variability and model misspecification, was 9.9 mg/L (= 0.015 mmol/L; RSE 12%) for the CEFu concentration and 24 mg/L (= 0.036 mmol/L; RSE 6%) for the CEFt concentration (Table 2). Monte Carlo dosing simulations For the dosing regimens investigated, patients with a higher CLCR and lower albumin concentrations had lower CEFt concentrations (Figure 5a and c). As for CEFu, simulations indicated that only a higher CLCR was projected to be leading to lower CEFu concentrations (Figure 5b and d). Figure 5. View largeDownload slide Simulations of steady-state ceftriaxone concentration–time profiles of a virtual patient with all median characteristics of the population, but with three different plasma albumin concentrations (colours) and two CLCR values (line or dotted line). For albumin, the observed median (blue), 25th percentile (red) and 75th percentile (green) were used. For CLCR, the median (line) and 25th percentile (dotted line) were used. (a) CEFt following a 1 g q12h injection. (b) CEFu following a 1 g q12h injection. (c) CEFt following a 2 g q24h injection. (d) CEFu following a 2 g q12h injection. Figure 5. View largeDownload slide Simulations of steady-state ceftriaxone concentration–time profiles of a virtual patient with all median characteristics of the population, but with three different plasma albumin concentrations (colours) and two CLCR values (line or dotted line). For albumin, the observed median (blue), 25th percentile (red) and 75th percentile (green) were used. For CLCR, the median (line) and 25th percentile (dotted line) were used. (a) CEFt following a 1 g q12h injection. (b) CEFu following a 1 g q12h injection. (c) CEFt following a 2 g q24h injection. (d) CEFu following a 2 g q12h injection. For microorganisms with an MIC of 1 mg/L, the PTA of the 1 g q12h ceftriaxone regimen for patients with the median CLCR of 91 mL/min was 95.1% for the primary PD target of fT>MIC = 50% (Figure 6). The PTA of the secondary PD target of fT>MIC = 100% was 58.2%. In the case of a 2 g q24h ceftriaxone regimen, the PTA was 74.8% and 16.5% for the PD targets of fT>MIC = 50% and fT>MIC = 100%, respectively. When treating a pathogen with an MIC of 0.125 mg/L, the PTA of the fT>MIC = 50% PD target was 99.9% for a 1 g q12h regimen and 96.5% for a 2 g q24h ceftriaxone regimen in patients with a median albumin concentration of 29 g/L and median CLCR. The respective PTAs were 94.5% and 52.2% when the fT>MIC = 100% target was applied (Figure 6). Figure 6. View largeDownload slide Probability of target attainment with three different ceftriaxone dosing regimens. Percentages of 1000 simulated patients with a median albumin concentration (29 g/L) and a median estimated CLCR (91 mL/min) achieving a CEFu concentration above the MIC during half of the dosing interval (fT>MIC = 50%: a) and throughout the dosing interval (fT>MIC = 100%: b), for three different ceftriaxone dosing regimens and a range of MICs. Line with triangles, 2 g q24h; line with circles, 1 g q12h; line with squares, 2 g q12h. The clinical breakpoint for Enterobacteriaceae according to EUCAST is 1 mg/L.27 Figure 6. View largeDownload slide Probability of target attainment with three different ceftriaxone dosing regimens. Percentages of 1000 simulated patients with a median albumin concentration (29 g/L) and a median estimated CLCR (91 mL/min) achieving a CEFu concentration above the MIC during half of the dosing interval (fT>MIC = 50%: a) and throughout the dosing interval (fT>MIC = 100%: b), for three different ceftriaxone dosing regimens and a range of MICs. Line with triangles, 2 g q24h; line with circles, 1 g q12h; line with squares, 2 g q12h. The clinical breakpoint for Enterobacteriaceae according to EUCAST is 1 mg/L.27 Discussion The present study describes the development of a PPK model of ceftriaxone in an adult SSA hospital population with presumptive severe infection, based on observed CEFu and CEFt plasma concentrations. The study population was generally young and severely ill, as illustrated by its low median BMI and low median haemoglobin and albumin concentrations. A one-compartment model with non-linear protein binding best described CEFt/CEFu PK. Similar to other studies with ICU patients, the BPV of PK parameters was high, with the BPV of CEFu CL being as high as 77% in the final model. As can be expected with an antibiotic with renal CL as the predominant route of elimination, CEFu CL was significantly associated with CLCR and explained a substantial part of this BPV.28,29 Non-linear, saturable protein binding of ceftriaxone has been suggested by others as well.30,31 The relationship between the maximum ceftriaxone concentration bound to albumin (Bmax) and albumin concentration as found in our study implies that Bmax is lower at lower albumin concentrations. In agreement with this, the observed low median albumin concentration of 29 g/L corresponded to a median CEFu fraction of 19% which is high relative to what can be expected under normal albumin concentrations at therapeutic concentrations (5%–15%). Thus, with a higher CEFu fraction present at such lower albumin concentrations, more CEFu will be available for CL. At steady-state, such increased CL leads to a lower CEFt concentration that is also less representative of the CEFu concentration.30 Our findings match the results from earlier PK studies of ceftriaxone in non-SSA ICU patients that implied predominant dependency of CEFu exposure on CLCR.28,29,32,33 However, only one study suggested an inability to reach the fT>MIC = 100% target with a 2 g q12h prescription.32 This may be explained by the fact that not all studies based their conclusions on measured CEFu concentrations and that some may have been too small to draw meaningful conclusions. Although none of the measured liver enzymes proved to have a significant relationship with CEFu PK, other investigators have found bilirubin levels to be negatively correlated with albumin binding and non-renal CL of CEFu.28,34 Since we did not include bilirubin in our analysis, it cannot be excluded that part of the BPV may have resulted from a relationship between liver function and CEFu PK that was left unaccounted for. For a 2 g q24h dosing regimen, the PTA for the PD target of fT>MIC = 100% was just above 50% for pathogens with an MIC of 0.125 mg/L. Importantly, simulations also demonstrated that the use of the more commonly prescribed 1 g q12h ceftriaxone regimen may lead to CEFu underexposure as well, especially for patients with preserved kidney function. The risk for underexposure in patients with impaired renal function appears to occur mostly when stringent PD targets are applied. For pathogens with an MIC of 1 mg/L, both q24h and q12h dosing are likely to lead to CEFu underexposure. Although clinical and microbiological evidence is still limited, available data do seem to support the idea that in patients with severe disease, who may be at risk for impaired drug distribution and tissue penetration, the use of a more stringent PD target, such as fT>MIC = 100%, improves clinical outcome and reduces the emergence of antimicrobial resistance, compared with the most conservative target of fT>MIC = 50%.8,10,35,36 The CEFu concentration–time simulations could not be tested against the background of locally derived bacterial MICs, which means that the actual PTA could differ from the ones suggested on the basis of the MIC used for Enterobacteriaceae and S. pneumoniae (1 mg/L). Results from a surveillance study of global antimicrobial susceptibility showed that in Africa, 32% of S. pneumoniae isolates collected from 2009 to 2012 were resistant to penicillin (MIC >2 mg/L), and that in this group, roughly 25% of isolates were non-susceptible to ceftriaxone (MIC >1 mg/L).37 MDR of non-typhoidal Salmonella serovars, including resistance to third-generation cephalosporins, is emerging. Third-generation cephalosporin resistance still appears to be restricted to certain geographical areas, but very limited African susceptibility surveillance data are currently available.38 There are limitations to our study. Firstly, we did not collect information about a patient’s fluid balance and the presence of excess fluid in the interstitial and transcellular space, and are therefore not informed about how hydration status, oedema and ‘third spacing’ may have affected PK BPV. It is our experience that administration of sufficient amounts of fluids in adult hospitalized patients in resource-poor settings is rare and our observations are supported by reports about patients with presumptive sepsis from Uganda.39,40 An inadequate hydration status in a substantial amount of cases may imply that CEFu concentrations as well as PTAs could turn out to be lower than what was found in the current study once proper volume therapy is applied. Secondly, the Cockcroft and Gault formula’s predictive value is limited in critically ill patients without stable renal function. Although this formula has been shown to have relatively good accuracy for patients with BMI <18.5 kg/m2, CLCR estimation, and thus the estimation of the glomerular filtration rate, may therefore be biased.41 Thirdly, there was substantial residual PK variability as expressed by the CEFu and CEFt additive error. An explanation may lie in drug dispensing and administration procedures, even though these procedures were observed when possible. Although it is not unlikely that our study population’s risk of CEFu underexposure is indicative of what happens to PK/PD in critically ill patients in general, extrapolations to non-SSA severely ill (ICU) patient populations should be made with caution, as underlying disease, as well as treatment aspects, including antibiotic infusion time and volume therapy, are likely to be different. In conclusion, a one-compartment PPK model with non-linear protein binding adequately described the PK of CEFu and CEFt in a severely ill, non-ICU hospital population. Intermittent bolus dosing of the highly albumin-bound β-lactam ceftriaxone may lead to underexposure and can potentially pose a threat to a patient’s health, especially when a patient’s renal function is intact and when a q24h dosing regimen is used. Underexposure may also pose a threat to public health, as it is known to contribute to the emergence of antimicrobial resistance. These outcomes underscore a need for SSA antibiotic PK data against the background of locally derived MICs, so that well-known antibiotic drugs can be preserved for use in this region of the world. Acknowledgements In Mozambique, first of all, we want to thank the study participants for their participation and trust. We are indebted to the UCM Research Centre for Infectious Diseases (CIDI) for their professional collaboration with regard to the implementation of the study. Within the CIDI, we thank Ms Ivete Meque and Ms Kajal Chhaganlal, subsequent CIDI managers, as well as CIDI support staff for their share in ethics, good clinical and laboratory practice training, finances and daily transport of blood samples. We want to thank Professor Janneke van de Wijgert of the University of Liverpool for her advice with regard to the preparation of local contracts and agreements as well as Susan Mason of the United States Military HIV Research Program (MHRP), for the programming of the CIDI laboratory’s study database. We are grateful to the executive board of the HCB and the late Carlos de Oliveira, internist and former head of the HCB medicine department, for making research office space available on the HCB wards, as well as to the entire nursing staff from the HCB medicine wards for their collaboration. In the Netherlands, we thank Ms Femke Schrauwen, trial coordinator of the AMC biochemistry laboratory, and Ms Marloes van der Meer and Mr Marcel Pistorius, research analysts of the AMC pharmacology laboratory, for their useful input in the interpretation of biochemistry and ceftriaxone concentration results. Funding This work was funded with internal funding from the Academic Medical Centre (AMC). Additionally, it was indirectly supported by a grant from the Gilead Foundation (14 July 2014; IA 356007) and a grant from a Dutch private donor who wants to stay anonymous (CA 356001), but whose professional activities do not create conflict of interest for one of the authors. Both external funding parties supported the local presence of J. C. Bos in Mozambique in the context of a long-running medical educational capacity building project with the Faculty of Health Sciences of the Catholic University of Mozambique (FCS-UCM). Transparency declarations R. M. v. H. reports having received personal fees from Nordic Pharma. All other authors: none to declare. Author contributions J. C. Bos, R. A. A. M. and J. M. P. designed the study. J. C. Bos performed the literature search. J. C. Bos obtained ethical approval. M. C. M., G. N., C. N. L. and J. C. Beirão implemented the study and J. C. B. supervised data collection and study progress on a daily basis. J. C. Bos and R. M. v. H. analysed the data. J. C. Bos and R. M. v. H. drafted the manuscript. J. M. P. and R. A. A. M. critically examined the analysis and findings, and all authors critically read and commented on draft versions of the report. All authors approved the final version. References 1 Ford N , Shubber Z , Meintjes G et al. Causes of hospital admission among people living with HIV worldwide: a systematic review and meta-analysis . Lancet HIV 2015 ; 2 : e438 – 44 . Google Scholar CrossRef Search ADS PubMed 2 Meintjes G , Kerkhoff AD , Burton R et al. HIV-related medical admissions to a South African district hospital remain frequent despite effective antiretroviral therapy scale-up . Medicine (Baltimore) 2015 ; 94 : e2269. Google Scholar CrossRef Search ADS PubMed 3 Huson MAM , Kalkman R , Stolp SM et al. The impact of HIV on presentation and outcome of bacterial sepsis and other causes of acute febrile illness in Gabon . Infection 2015 ; 43 : 443 – 51 . Google Scholar CrossRef Search ADS PubMed 4 Prasad N , Murdoch DR , Reyburn H et al. Etiology of severe febrile illness in low- and middle-income countries: a systematic review . PLoS One 2014 ; 9 : e92131 . Google Scholar CrossRef Search ADS PubMed 5 Uche IV , MacLennan CA , Saul A. A systematic review of the incidence, risk factors and case fatality rates of invasive nontyphoidal Salmonella (iNTS) disease in Africa (1966 to 2014) . PLoS Negl Trop Dis 2016 ; 11 : e0005118 . Google Scholar CrossRef Search ADS 6 Roberts JA , Abdul-Haziz M , Lipman J et al. Individualized antibiotic dosing for patients who are critically ill: challenges and potential solutions . Lancet Infect Dis 2014 ; 14 : 498 – 509 . Google Scholar CrossRef Search ADS PubMed 7 Jager NGL , van Hest RM , Lipman J et al. Therapeutic drug monitoring of anti-infective agents in critically ill patients . Expert Rev Clin Pharmacol 2016 ; 9 : 961 – 79 . Google Scholar CrossRef Search ADS PubMed 8 Osthoff M , Siegemund M , Balestra G et al. Prolonged administration of β-lactam antibiotics—a comprehensive review and critical appraisal . Swiss Med Wkly 2016 ; 146 : w14368 . Google Scholar PubMed 9 Abdul-Aziz MH , Sulaiman H , Mat-Nor MB et al. β-Lactam infusion in severe sepsis (BLISS): a prospective, two-centre, open-labelled randomised controlled trial of continuous versus intermittent β-lactam infusion in critically ill patients with severe sepsis . Intensive Care Med 2016 ; 42 : 1535 – 45 . Google Scholar CrossRef Search ADS PubMed 10 Roberts JA , Paul SK , Akova M et al. DALI: defining antibiotic levels in intensive care unit patients: are current β-lactam antibiotic doses sufficient for critically ill patients? Clin Infect Dis 2014 ; 8 : 1072 – 83 . Google Scholar CrossRef Search ADS 11 Trobec K , Kos MK , von Haehling S et al. Pharmacokinetics of drugs in cachectic patients: a systematic review . PLoS One 2013 ; 8 : e79603. Google Scholar CrossRef Search ADS PubMed 12 Stanifer JW , Jing B , Tolan S et al. The epidemiology of chronic kidney disease in sub-Saharan Africa: a systematic review and meta-analysis . Lancet Glob Health 2014 ; 2 : e174 – 81 . Google Scholar CrossRef Search ADS PubMed 13 Spearman CW , Sonderup MW. Health disparities in liver disease in sub-Saharan Africa . Liver Int 2015 ; 35 : 263 – 71 . Google Scholar CrossRef Search ADS PubMed 14 Seboxa T , Amogne W , Abebe W et al. High mortality from blood stream infection in Addis Ababa, Ethiopia, is due to antimicrobial resistance . PLoS One 2015 ; 10 : e0144944 . Google Scholar CrossRef Search ADS PubMed 15 Leopold SJ , van Leth F , Tarekegn H et al. Antimicrobial drug resistance among clinically relevant bacterial isolates in sub-Saharan Africa: a systematic review . J Antimicrob Chemother 2014 ; 69 : 2337 – 53 . Google Scholar CrossRef Search ADS PubMed 16 Bos JC , Smalbraak L , Macome C et al. TB diagnostic process management of patients in a referral hospital in Mozambique in comparison with the 2007 WHO recommendations for the diagnosis of smear-negative pulmonary TB and extrapulmonary TB . Int Health 2013 ; 5 : 302 – 8 . Google Scholar CrossRef Search ADS PubMed 17 Ministério da Saúde da Republica de Moçambique . Formulário Nacional de Medicamentos. 5a edição. 2007. http://apps.who.int/medicinedocs/en/d/Js19267pt/. 18 North Shore LIJ. Human Subject Protection Program Guidance Document: Maximum Blood Draw Limits. 2014 . https://www.feinsteininstitute.org/professionals/resources-for-investigators/administrative-services/human-research-protection-program/tools-and-guidance/. 19 Cockcroft DW , Gault MH. Prediction of creatinine clearance from serum creatinine . Nephron 1976 ; 16 : 31 – 41 . Google Scholar CrossRef Search ADS PubMed 20 Bonate PL. Pharmacokinetic-Pharmacodynamic Modelling and Simulation, 2nd edn . New York, NY, USA : Springer , 2011 . Google Scholar CrossRef Search ADS 21 van Hest RM , van Gelder T , Vulto AG et al. Pharmacokinetic modelling of the plasma protein binding of mycophenolic acid in renal transplant recipients . Clin Pharmacokinet 2009 ; 48 : 463 – 76 . Google Scholar CrossRef Search ADS PubMed 22 Ahn JE , Karlsson MO , Dunne A et al. Likelihood based approaches to handling data below the quantification limit using NONMEM VI . J Pharmacokinet Pharmacodyn 2008 ; 35 : 401 – 21 . Google Scholar CrossRef Search ADS PubMed 23 Ette EI , Ludden TM. Population pharmacokinetic modelling: the importance of informative graphics . Pharm Res 1995 ; 12 : 1845 – 55 . Google Scholar CrossRef Search ADS PubMed 24 Keizer RJ , Karlsson MO , Hooker A. Modeling and simulation workbench for NONMEM: tutorial on Pirana, PsN, and Xpose . CPT Pharmacometrics Syst Pharmacol 2013 ; 2 : e50. Google Scholar CrossRef Search ADS PubMed 25 Jonsson EN , Karlsson MO. Xpose–an S-PLUS based population pharmacokinetic/pharmacodynamic model building aid for NONMEM . Comput Methods Programs Biomed 1999 ; 58 : 51 – 64 . Google Scholar CrossRef Search ADS PubMed 26 Lindbom L , Pihlgren P , Jonsson EN. PsN-Toolkit–a collection of computer intensive statistical methods for non-linear mixed effect modeling using NONMEM . Comput Methods Programs Biomed 2005 ; 79 : 241 – 57 . Google Scholar CrossRef Search ADS PubMed 27 EUCAST . Breakpoint Tables for Interpretation of MICs and Zone Diameters, Version 3.1. http://www.eucast.or/clinical_breakpoints/. 28 Schleibinger M , Steinbach CL , Töpper C et al. Protein binding characteristics and pharmacokinetics of ceftriaxone in intensive care unit patients . Br J Clin Pharmacol 2015 ; 80 : 525 – 33 . Google Scholar CrossRef Search ADS PubMed 29 Garot D , Respaud R , Lanotte P et al. Population pharmacokinetics of ceftriaxone in critically ill septic patients: a reappraisal . Br J Clin Pharmacol 2011 ; 72 : 758 – 67 . Google Scholar CrossRef Search ADS PubMed 30 Roberts JA , Pea F , Lipman J. The clinical relevance of plasma protein binding changes . Clin Pharmacokinet 2013 ; 52 : 1 – 8 . Google Scholar CrossRef Search ADS PubMed 31 Stoeckel K , McNamara PJ , Brandt R et al. Effects of concentration-dependent plasma protein binding on ceftriaxone kinetics . Clin Pharmacol Ther 1981 ; 29 : 650 – 7 . Google Scholar CrossRef Search ADS PubMed 32 Joynt GM , Lipman J , Gomersall CD et al. The pharmacokinetics of once-daily dosing of ceftriaxone in critically ill patients . J Antimicrob Chemother 2001 ; 47 : 421 – 9 . Google Scholar CrossRef Search ADS PubMed 33 Tsaj D , Stewart P , Goud R et al. Total and unbound ceftriaxone pharmacokinetics in critically ill Australian indigenous patients with severe sepsis . Int J Antimicrob Agents 2016 ; 48 : 748 – 52 . Google Scholar CrossRef Search ADS PubMed 34 Stoeckel K , Tuerk H , Trueb V et al. Single-dose ceftriaxone kinetics in liver insufficiency . Clin Pharmacol Ther 1984 ; 36 : 500 – 9 . Google Scholar CrossRef Search ADS PubMed 35 Sime FB , Roberts MS , Peake SL et al. Does pharmacokinetic variability in critically ill patients justify therapeutic drug monitoring? A systematic review . Ann Intensive Care 2012 ; 2 : 35. Google Scholar CrossRef Search ADS PubMed 36 Feng Y , Hodiamont CJ , van Hest RM et al. Development of antibiotic resistance during simulated treatment of Pseudomonas aeruginosa in chemostats . PLoS One 2016 ; 11 : e0149310 . Google Scholar CrossRef Search ADS PubMed 37 Tomic V , Dowzicky MJ. Regional and global antimicrobial susceptibility among isolates of Streptococcus pneumoniae and Haemophilus influenzae collected as part of the tigecycline evaluation and surveillance trial (T.E.S.T.) from 2009 to 2012 and comparison with previous years of T.E.S.T. (2004-2008) . Ann Clin Microbiol Antimicrob 2014 ; 13 : 52. Google Scholar CrossRef Search ADS PubMed 38 Marks F , von Kalckreuth V , Aaby P et al. Incidence of invasive salmonella disease in sub-Saharan Africa: a multicentre population-based surveillance study . Lancet Glob Health 2017 ; 5 : e310 – 23 . Google Scholar CrossRef Search ADS PubMed 39 Jacob ST , Moore CC , Banura P et al. Severe sepsis in two Ugandan hospitals: a prospective observational study of management and outcomes in a predominantly HIV-1 infected population . PLoS One 2009 ; 4 : e7782. Google Scholar CrossRef Search ADS PubMed 40 Jacob ST , Banura P , Baeten JM et al. The impact of early monitored management on survival in hospitalized adult Ugandan patients with severe sepsis: a prospective intervention study . Crit Care Med 2012 ; 40 : 2050 – 8 . Google Scholar CrossRef Search ADS PubMed 41 Michels WM , Grootendorst DC , Verduijn M et al. Performance of the Cockcroft-Gault, MDRD, and new CKD-EPI formulas in relation to GFR, age and body size . Clin J Am Soc Nephrol 2010 ; 5 : 1003 – 9 . Google Scholar CrossRef Search ADS PubMed 42 Reynolds JEF (ed). Martindale: The Extra Pharmacopoeia, 13th edn . London, UK : The Pharmaceutical Press , 1993 . © 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

Pharmacokinetics and pharmacodynamic target attainment of ceftriaxone in adult severely ill sub-Saharan African patients: a population pharmacokinetic modelling study

Loading next page...
 
/lp/ou_press/pharmacokinetics-and-pharmacodynamic-target-attainment-of-ceftriaxone-55EIlmStpp
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/dky071
Publisher site
See Article on Publisher Site

Abstract

Abstract Background In sub-Saharan Africa (SSA), the highly albumin-bound β-lactam ceftriaxone is frequently used for the empirical treatment of severe bacterial infections. Systemic drug exposure of β-lactams can be altered in critically ill ICU patients, but pharmacokinetic and pharmacodynamic data for non-ICU SSA populations are lacking. Methods We performed a population pharmacokinetic study in an adult hospital population in Mozambique, treated with ceftriaxone for presumptive severe bacterial infection from October 2014 to November 2015. Four blood samples per patient were collected for total ceftriaxone (CEFt) and unbound ceftriaxone (CEFu) concentration measurement. We developed a population pharmacokinetic model through non-linear mixed effect analysis and performed simulations for different patient variable, dosing and pharmacodynamic target scenarios. Results Eighty-eight participants yielded 277 CEFt and 276 CEFu concentrations. The median BMI was 18.9 kg/m2 and the median albumin concentration was 29 g/L. In a one-compartment model with non-linear protein binding, creatinine clearance was positively correlated with CEFu clearance. For microorganisms with an MIC of 1 mg/L, simulations demonstrated that with a 1 g twice-daily regimen and a 2 g once-daily regimen, 95.1% and 74.8% would have a CEFu concentration > MIC during half of the dosing interval (fT>MIC = 50%), respectively, whereas this was only 58.2% and 16.5% for the fT>MIC = 100% target. Conclusions Severely ill adult non-ICU SSA patients may be at substantial risk for underexposure to CEFu during routine intermittent bolus dosing, especially when their renal function is intact. Introduction In sub-Saharan African (SSA), a region of the world with high HIV infection rates, 10%–34% of patients admitted with fever suffer from bacterial bloodstream infections or sepsis.1–3 A high proportion of these infections are caused by Streptococcus pneumoniae and non-typhoidal Salmonella serovars and mortality rates can be as high as 46%–72%.4,5 The third-generation cephalosporin ceftriaxone is among the most frequently used antibiotics for the empirical treatment of adults in this situation. Evidence from resource-rich ICU settings suggests that appropriateness of antibiotic treatment should not only be reviewed in terms of antibiotic drug choice but also in terms of systemic drug exposure.6,7 Pharmacokinetics (PK) of antibiotics in critically ill patients can be substantially altered. β-Lactam antibiotics are particularly vulnerable, as they are hydrophilic drugs with renal CL as the predominant route of elimination. Unlike most β-lactams, ceftriaxone is highly bound to albumin (85%–95%) at therapeutic concentrations. During sepsis and hypoalbuminaemia a lower protein-bound drug fraction, an increased volume of distribution (V) and increased CL may change the total as well as the unbound, active drug concentration.6,7 Such alterations may give rise to an inability to attain pharmacodynamic (PD) targets, including the increasingly promoted PD target of the unbound drug concentration remaining above the MIC throughout the dosing interval (fT>MIC = 100%).8 Ultimately, underexposure may lead to adverse clinical outcome.9,10 In SSA, highly prevalent chronic diseases such as TB, hepatitis B and hypertension also have the potential to influence the PK of antibiotics by means of cachexia and liver and kidney dysfunction.11–13 How coinciding acute and chronic conditions in severely ill (non-ICU) SSA patients influence the PK of β-lactam antibiotic drugs, including ceftriaxone, has been poorly investigated. What is clear is that underexposure to the unbound drug may not only pose a threat to an individual patient’s health, but also to public health, as it contributes to the emergence of antimicrobial resistance, a phenomenon that is already highly prevalent in SSA countries.14,15 In view of the above, we performed a population PK (PPK) study of ceftriaxone in a Mozambican, adult hospital medicine ward population. The specific aims of the study were to describe the PPK of unbound ceftriaxone (CEFu) and total ceftriaxone (CEFt) in order to identify sources of PK parameter variability. Additionally, we aimed to assess the probability of PK/PD target attainment (PTA) of CEFu for the treatment of bacterial pathogens commonly causing sepsis in SSA. Methods Setting The Beira Central Hospital (HCB) is a Mozambican governmental referral health facility. The proportion of patients infected with HIV on its medicine ward may be as high as 74% and up to 32% of patients may die during their hospital stay.16 Study design The current study was a prospective, observational PPK study of ceftriaxone, as part of a PPK study of commonly used antibiotics among patients admitted to the HCB medicine ward. In this study, PK data were collected from October 2014 until November 2015 from patients who were treated with intravenously administered ceftriaxone. The study was reviewed and approved by the Mozambican National Committee for Bio-ethics in Health (CNBS: study registration no. 118/CNBS/2013). Additionally, a letter of approval was obtained from the HCB director. Participants gave written informed consent. Those unable to read, write and/or understand Portuguese gave a thumbprint and an impartial, literate witness observed the entire informed consent process and subsequently co-signed the informed consent form. Recruitment and data collection Patients were eligible for study entry if they were hospitalized on the HCB medicine ward while being treated with ceftriaxone as prescribed by an HCB physician, and as documented in a patient’s medication record. Any ceftriaxone dosing regimen that was allowed for use by the Mozambican National Institute of Health (INS) was accepted.17 Inclusion criteria were age ≥18 years and being willing and able to give informed consent. Exclusion criteria were the use of drugs known to significantly affect the PK of the study antibiotics (probenecid, phenylbutazone, acetylsalicylic acid and indomethacin), a haemoglobin level ≤6 g/dL as measured by the HCB laboratory (haemoglobin safe limit adapted from the North Shore Human Subject Protection Program Guidance Document on maximum blood draw limits18), any condition necessitating a blood transfusion irrespective of haemoglobin level, and an altered level of consciousness. Two trained research nurses captured baseline characteristics and ceftriaxone dosing information, and measured the body weight and height of all study participants. Doses of 1 g of ceftriaxone powder for injection (Nirlife, Nirma Ltd, Gujarat, India) were dissolved in 10 mL of sterile water for injection and subsequently injected intravenously via a venous catheter in half a minute, according to the responsible physician’s prescription. Over a time period of two or more days, a maximum of four blood samples were collected for ceftriaxone concentration measurement. Sample times were pre-dose (trough level), 30–120 min after intravenous administration (peak level) and two random timepoints during the dosing interval (random levels). Ceftriaxone administration procedures were observed where possible and a maximum of 19 mL of blood was collected. One blood sample was also used for the measurement of albumin, AST, ALT, GGT and creatinine concentrations. Bilirubin levels were not measured for practical reasons in relation to bilirubin’s high photosensitivity. CLCR was estimated using the Cockcroft and Gault formula.19 Sample handling and drug assay EDTA anti-coagulated blood samples were refrigerated at 4–8°C immediately after collection until laboratory processing, which took place within two hours of collection. Samples were centrifuged and plasma was stored at −80°C in the local research laboratory until shipment on dry ice to the Netherlands for biochemical marker and drug concentration analysis. Plasma was ultrafiltrated (centrifugal filters: Millipore Amicon Ultra 0.5 mL/30K, Merck Millipore, Darmstadt, Germany) and the ultrafiltrated plasma was subsequently processed as a typical plasma sample to obtain the unbound concentration. CEFt and CEFu concentrations were measured using validated HPLC MS (LC: LC30 UPLC, Shimadzu, Kyoto, Japan; MS: QTRAP 5500 system, Sciex, Framingham, MA, USA). The lower limit of quantification (LLQ) was 1 mg/L and the higher limit of quantification was 40 mg/L. Concentrations higher than this were diluted and re-analysed. Within- and between-assay variability was lower than 5.6% and 6.8%, respectively. The accuracy of the assay was between 90% and 105%. PPK analysis Model development The PPK analysis was performed using the non-linear mixed-effect modelling software package NONMEM (7.1.2; Icon Development Solutions, Ellicott City, MD, USA). Model building was performed using a stepwise approach. During a first step, a structural, compartmental PPK model was developed in which the PK of CEFu and CEFt as well as protein binding of ceftriaxone were described, including their between-patient variability (BPV). In the second step, an attempt was made to explain BPV by building a covariate model in which patient demographics and pathophysiological factors were tested for their correlation with the estimated PK parameters. In the third and last step, the robustness and validity of the final model resulting from the second step were tested. Structural model One- and two-compartment models were tested. Estimated PK parameters were V, CL and, in the case of testing of a two-compartment model, peripheral V and intercompartmental CL. BPV of V and CL was estimated exponentially.20 Modelling of ceftriaxone protein binding and CEFt was done using linear and non-linear protein-binding models, according to the following equation: [CEFt] (mmol/L)=[CEFu]+[CEFu]×b Eqn 1 where b stands for ceftriaxone’s binding properties and [CEFu]×b stands for the bound ceftriaxone concentration. In the linear protein binding model, b is an estimated parameter, θbind, that correlates to the number of unoccupied binding places,21 and in the non-linear protein-binding model, b becomes Bmax/Km + [CEFu]. In this model, Bmax represents protein binding, defined as the maximum estimated concentration of ceftriaxone bound to albumin, and Km represents the CEFu concentration at which albumin binding is half maximal. The residual variability (i.e. the difference between measured CEFu and CEFt concentrations and the corresponding CEFu and CEFt concentrations predicted by the model) was modelled with additive or proportional models or a combination of both. The so-called M3 method was used for the handling of CEFt and CEFu concentrations below the LLQ.22 Covariate model Tested covariates included age, sex, weight, height, BMI, CLCR and haemoglobin, albumin, creatinine, GGT, ALT and AST concentrations. All covariates were screened for the significance of the correlation between the covariate and the PK parameter by univariate analysis, using a P value cut-off of 0.05. Furthermore, a reduction in BPV or residual variability, as well as biological plausibility of a covariate–PK parameter relationship was used as a criterion for covariate selection. All covariates selected during the univariate analysis subsequently entered an intermediate model for a backward elimination procedure (multivariate analysis) with a cut-off for statistical significance of 0.001, which yielded the final model. Potential improvement of the model by adding a compartment or by introducing a correlation between a covariate and a PK parameter was evaluated using the likelihood ratio test, in which the difference between the minimum objective function value (OFV) generated by NONMEM® for two hierarchical models is determined. Improvement in model fit was defined as an OFV decrease of ≥3.8 units while using a χ2 distribution with one degree of freedom, corresponding to a P value cut-off of 0.05. For the backward elimination covariate procedure an OFV decrease of ≥10.8 units was used. Model performance was also evaluated by visual inspection of diagnostic ‘goodness-of-fit’ plots.23 These were generated using Pirana (version 2.9.0) and Xpose (version 4.3.2) software (Uppsala Pharmacometrics, Uppsala, Sweden).24,25 Model robustness and predictive performance The robustness of the parameter estimates from the final model resulting from the second step was tested using a bootstrap analysis. In this analysis, the dataset was resampled 1000 times. Based on 1000 simulations, visual predictive checks (VPCs) investigated whether the final model could adequately predict the observed time course of CEFu and CEFt, including the observed variability. Bootstrap and VPC analyses were performed using Perl-speaks-NONMEM (PsN) version 3.5.3 software (Uppsala Pharmacometrics, Uppsala, Sweden).26 Monte Carlo dosing simulations Using the final PPK model, CEFu concentration–time profiles were predicted based on Monte Carlo simulations following three dosing regimens: 1 g twice daily (q12h), 2 g once daily (q24h) and 2 g q12h. Each simulation generated concentration–time profiles for 1000 virtual patients per drug regimen. Based on these data, the PTA, being the percentage of patients with a CEFu remaining above a specified MIC during a specified percentage of time of the dosing interval (fT>MIC), was calculated for different fT>MIC targets. The primary target to be tested was fT>MIC = 50% and the secondary target was fT>MIC = 100%. The choice of PD targets was based on conclusions of recent reviews concerning β-lactam administration and PD targets in critically ill patients.6,8,10 The choice of the target MIC was based on the EUCAST MIC clinical breakpoints for susceptibility to ceftriaxone of Enterobacteriaceae (1 mg/L) and S. pneumoniae (0.5 mg/L).27 Results Patients and ceftriaxone concentrations We screened 762 patients for the larger PPK study and excluded 366 patients (Figure 1). The most common reason for exclusion was a haemoglobin level ≤6 g/dL. We included 98 patients in the current study on ceftriaxone and 88 participants remained for analysis. A large majority (93.1%) had a 1 g q12h ceftriaxone dosing schedule. Patient characteristics are presented in Table 1. Table 1. Baseline characteristics of study population (N = 88) Female, n (%) 48 (54.5) Age (years), median (range) 35 (18–76) Bodyweight (kg), median (range) 50 (29–70) BMI (kg/m2), median (range) 18.9 (11.3–26.4) Haemoglobin (g/dL), median (range) 9.4 (6.3–16.4) Albumin (g/L), median (range) 29 (11–44) GGT (U/L), median (range) 49 (6–1276) ALT (U/L), median (range) 25 (0–676) AST (U/L), median (range) 50 (11–594) Creatinine (μmol/L), median (range) 63 (19–971) CLCRa (mL/min), median (range) 91 (4–261) Ceftriaxone dose prescribed, n (%)  1 g q12h 82 (93.1)  2 g q24h 2 (2.3)  2 g q12h 4 (4.6) Female, n (%) 48 (54.5) Age (years), median (range) 35 (18–76) Bodyweight (kg), median (range) 50 (29–70) BMI (kg/m2), median (range) 18.9 (11.3–26.4) Haemoglobin (g/dL), median (range) 9.4 (6.3–16.4) Albumin (g/L), median (range) 29 (11–44) GGT (U/L), median (range) 49 (6–1276) ALT (U/L), median (range) 25 (0–676) AST (U/L), median (range) 50 (11–594) Creatinine (μmol/L), median (range) 63 (19–971) CLCRa (mL/min), median (range) 91 (4–261) Ceftriaxone dose prescribed, n (%)  1 g q12h 82 (93.1)  2 g q24h 2 (2.3)  2 g q12h 4 (4.6) a Estimated using the Cockcroft–Gault equation.19 Table 1. Baseline characteristics of study population (N = 88) Female, n (%) 48 (54.5) Age (years), median (range) 35 (18–76) Bodyweight (kg), median (range) 50 (29–70) BMI (kg/m2), median (range) 18.9 (11.3–26.4) Haemoglobin (g/dL), median (range) 9.4 (6.3–16.4) Albumin (g/L), median (range) 29 (11–44) GGT (U/L), median (range) 49 (6–1276) ALT (U/L), median (range) 25 (0–676) AST (U/L), median (range) 50 (11–594) Creatinine (μmol/L), median (range) 63 (19–971) CLCRa (mL/min), median (range) 91 (4–261) Ceftriaxone dose prescribed, n (%)  1 g q12h 82 (93.1)  2 g q24h 2 (2.3)  2 g q12h 4 (4.6) Female, n (%) 48 (54.5) Age (years), median (range) 35 (18–76) Bodyweight (kg), median (range) 50 (29–70) BMI (kg/m2), median (range) 18.9 (11.3–26.4) Haemoglobin (g/dL), median (range) 9.4 (6.3–16.4) Albumin (g/L), median (range) 29 (11–44) GGT (U/L), median (range) 49 (6–1276) ALT (U/L), median (range) 25 (0–676) AST (U/L), median (range) 50 (11–594) Creatinine (μmol/L), median (range) 63 (19–971) CLCRa (mL/min), median (range) 91 (4–261) Ceftriaxone dose prescribed, n (%)  1 g q12h 82 (93.1)  2 g q24h 2 (2.3)  2 g q12h 4 (4.6) a Estimated using the Cockcroft–Gault equation.19 Figure 1. View largeDownload slide Study profile. Hb, haemoglobin concentration. Figure 1. View largeDownload slide Study profile. Hb, haemoglobin concentration. A total of 277 plasma samples yielded 277 CEFt and 276 CEFu concentrations. Fewer than four plasma samples were available for 43/88 (48.9%) participants, with the most common explanation being a participant’s unforeseen discharge (Figure 1). Three or four plasma samples were available for analysis for 67/88 (76.1%) participants. There were 33/276 (12%) CEFu samples with a concentration below the LLQ. The median unbound fraction was 19% (IQR = 13–29). The observed CEFt and CEFu concentrations are shown in Figure 2. Figure 2. View largeDownload slide Observed ceftriaxone concentration–time data and VPC of the final model. The black dots are the observed concentrations. The solid black line is the observed median and the dashed lines are the 5th and 95th percentiles of the observed data. The red shaded area is the 95% CI of the model-predicted median and the blue shaded areas are the 95% CIs of the model-predicted 5th and 95th percentiles. (a) CEFt. (b) CEFu. The black solid and dashed lines run within their respective shaded areas, thereby demonstrating adequate fit of the model. Figure 2. View largeDownload slide Observed ceftriaxone concentration–time data and VPC of the final model. The black dots are the observed concentrations. The solid black line is the observed median and the dashed lines are the 5th and 95th percentiles of the observed data. The red shaded area is the 95% CI of the model-predicted median and the blue shaded areas are the 95% CIs of the model-predicted 5th and 95th percentiles. (a) CEFt. (b) CEFu. The black solid and dashed lines run within their respective shaded areas, thereby demonstrating adequate fit of the model. PPK analysis Structural model During step one of the analysis, CEFu data were fitted to several compartmental models and a one-compartmental model best fitted the data, with the difference in OFV between a one- and a two-compartmental model being only 0.21 units. The introduction of non-linear protein binding in the model resulted in an OFV decrease of 110 units compared with linear protein binding. Additionally, the residual error dropped from 13.1 mg/L to 9.8 mg/L and non-linear protein binding therefore best described CEFt concentration and protein binding (Figure 3). Km was estimated to be 0.01 mmol/L while Bmax was 0.11 mmol/L. The estimated BPV for Bmax was 31% (Table 2). Table 2. Parameter estimates of the different model building steps Parameter Model without covariates Model with covariates Bootstrap of model with covariates estimate RSE (%) estimate RSE (%) estimate 95% CI CEFu CL (L/h) 11 11 11 10 11 9.3–15 CEFuV (L) 48 10 48 10 49 35–60 Km (mmol/L) 0.010 28 0.0092 33 0.0096 0.004–0.020 Bmax (mmol/L) 0.11 12 0.12 11 0.12 0.094–0.15 BPV  CEFu CL (%CV) 93 15 77 14 73 37–108  CEFuV (%CV)a 33 a 37 a 38 7.9–112  Bmax (%CV)b 31 20 — — — — Residual variability  additive error CEFu (mmol/L) 0.015 12 0.015 12 0.015 0.011–0.017  additive error CEFt (mmol/L) 0.036 7 0.036 6 0.035 0.031–0.040 Covariate effects  CLCR on CEFu CLc — — 0.37 41 0.38 0.083–0.79  albumin on Bmaxc — — 1.30 41 1.3 0.93–1.70 Parameter Model without covariates Model with covariates Bootstrap of model with covariates estimate RSE (%) estimate RSE (%) estimate 95% CI CEFu CL (L/h) 11 11 11 10 11 9.3–15 CEFuV (L) 48 10 48 10 49 35–60 Km (mmol/L) 0.010 28 0.0092 33 0.0096 0.004–0.020 Bmax (mmol/L) 0.11 12 0.12 11 0.12 0.094–0.15 BPV  CEFu CL (%CV) 93 15 77 14 73 37–108  CEFuV (%CV)a 33 a 37 a 38 7.9–112  Bmax (%CV)b 31 20 — — — — Residual variability  additive error CEFu (mmol/L) 0.015 12 0.015 12 0.015 0.011–0.017  additive error CEFt (mmol/L) 0.036 7 0.036 6 0.035 0.031–0.040 Covariate effects  CLCR on CEFu CLc — — 0.37 41 0.38 0.083–0.79  albumin on Bmaxc — — 1.30 41 1.3 0.93–1.70 CV, coefficient of variation. For model-building purposes, observed concentrations of CEFu and CEFt were converted from mg/L into mmol/L by dividing observed concentrations by the molar mass of ceftriaxone of 661.6 g/mol.42 a The BPV of CEFuV was modelled as a function of the BPV of CEFu,20 leaving the BPV without an RSE. b The observation that BPV was not estimated for Bmax in the final model does not mean that the BPV of Bmax is zero, and as such fully explained by the association between albumin level and Bmax, but rather that there is not enough information in the data to support a reliable estimate of the BPV. c See Eqn 2 and Eqn 3. Table 2. Parameter estimates of the different model building steps Parameter Model without covariates Model with covariates Bootstrap of model with covariates estimate RSE (%) estimate RSE (%) estimate 95% CI CEFu CL (L/h) 11 11 11 10 11 9.3–15 CEFuV (L) 48 10 48 10 49 35–60 Km (mmol/L) 0.010 28 0.0092 33 0.0096 0.004–0.020 Bmax (mmol/L) 0.11 12 0.12 11 0.12 0.094–0.15 BPV  CEFu CL (%CV) 93 15 77 14 73 37–108  CEFuV (%CV)a 33 a 37 a 38 7.9–112  Bmax (%CV)b 31 20 — — — — Residual variability  additive error CEFu (mmol/L) 0.015 12 0.015 12 0.015 0.011–0.017  additive error CEFt (mmol/L) 0.036 7 0.036 6 0.035 0.031–0.040 Covariate effects  CLCR on CEFu CLc — — 0.37 41 0.38 0.083–0.79  albumin on Bmaxc — — 1.30 41 1.3 0.93–1.70 Parameter Model without covariates Model with covariates Bootstrap of model with covariates estimate RSE (%) estimate RSE (%) estimate 95% CI CEFu CL (L/h) 11 11 11 10 11 9.3–15 CEFuV (L) 48 10 48 10 49 35–60 Km (mmol/L) 0.010 28 0.0092 33 0.0096 0.004–0.020 Bmax (mmol/L) 0.11 12 0.12 11 0.12 0.094–0.15 BPV  CEFu CL (%CV) 93 15 77 14 73 37–108  CEFuV (%CV)a 33 a 37 a 38 7.9–112  Bmax (%CV)b 31 20 — — — — Residual variability  additive error CEFu (mmol/L) 0.015 12 0.015 12 0.015 0.011–0.017  additive error CEFt (mmol/L) 0.036 7 0.036 6 0.035 0.031–0.040 Covariate effects  CLCR on CEFu CLc — — 0.37 41 0.38 0.083–0.79  albumin on Bmaxc — — 1.30 41 1.3 0.93–1.70 CV, coefficient of variation. For model-building purposes, observed concentrations of CEFu and CEFt were converted from mg/L into mmol/L by dividing observed concentrations by the molar mass of ceftriaxone of 661.6 g/mol.42 a The BPV of CEFuV was modelled as a function of the BPV of CEFu,20 leaving the BPV without an RSE. b The observation that BPV was not estimated for Bmax in the final model does not mean that the BPV of Bmax is zero, and as such fully explained by the association between albumin level and Bmax, but rather that there is not enough information in the data to support a reliable estimate of the BPV. c See Eqn 2 and Eqn 3. Figure 3. View largeDownload slide Covariate relationship between albumin and the CEFu fraction. Plasma albumin concentration versus observed (red dots) and individual predicted (blue dots) CEFu fraction, with the model-predicted relationship (black line) for virtual patients with the median CEFu concentration of 11.8 mg/L. The unbound fraction increases in a non-linear fashion with a decreasing albumin concentration. Figure 3. View largeDownload slide Covariate relationship between albumin and the CEFu fraction. Plasma albumin concentration versus observed (red dots) and individual predicted (blue dots) CEFu fraction, with the model-predicted relationship (black line) for virtual patients with the median CEFu concentration of 11.8 mg/L. The unbound fraction increases in a non-linear fashion with a decreasing albumin concentration. Covariate model The multivariate covariate analysis, based on 100% availability of covariate results, yielded a model with significant associations between CEFu CL and CLCR as well as between Bmax and albumin concentration, as shown by Eqn 2 and Eqn 3: CEFu CL (L/h)=11×(CLCR/91)0.37 Eqn 2 Bmax=0.12×(albumin concentration/0.42)1.3 Eqn 3 With these associations in the model, the BPV of CEFu CL decreased from 93% to 77%, although a substantial part of the BPV in CEFu CL remained unexplained (Table 2). The BPV of Bmax could, however, no longer be reliably estimated. Using the Bmax and albumin concentration equation (Eqn 3), Eqn 1 for the calculation of CEFt is transformed into Eqn 4: [CEFt] (mmol/L)=[CEFu]+[CEFu]×[0.12×(albumin/0.42)1.3]/(0.0092+[CEFu]) Eqn 4 This relationship between CEFt concentrations and the observed and predicted CEFu concentrations is shown in Figure 4. Figure 4. View largeDownload slide Total versus unbound ceftriaxone concentrations. Black dots, observed concentrations; blue dots, individual predicted concentrations; black line, model-predicted relationship between total and unbound concentration for a patient with a median albumin concentration of 29 g/L. Figure 4. View largeDownload slide Total versus unbound ceftriaxone concentrations. Black dots, observed concentrations; blue dots, individual predicted concentrations; black line, model-predicted relationship between total and unbound concentration for a patient with a median albumin concentration of 29 g/L. Model robustness and predictive performance The final model had an adequate fit as shown by the VPC (Figure 2). The maximum relative standard error (RSE) with which parameters could be estimated was 41% and bootstrap estimations were similar to the ones from the final model (Table 2). The residual variability, an estimate of unexplained variability relating to measurement error, errors in data collection, intra-patient variability and model misspecification, was 9.9 mg/L (= 0.015 mmol/L; RSE 12%) for the CEFu concentration and 24 mg/L (= 0.036 mmol/L; RSE 6%) for the CEFt concentration (Table 2). Monte Carlo dosing simulations For the dosing regimens investigated, patients with a higher CLCR and lower albumin concentrations had lower CEFt concentrations (Figure 5a and c). As for CEFu, simulations indicated that only a higher CLCR was projected to be leading to lower CEFu concentrations (Figure 5b and d). Figure 5. View largeDownload slide Simulations of steady-state ceftriaxone concentration–time profiles of a virtual patient with all median characteristics of the population, but with three different plasma albumin concentrations (colours) and two CLCR values (line or dotted line). For albumin, the observed median (blue), 25th percentile (red) and 75th percentile (green) were used. For CLCR, the median (line) and 25th percentile (dotted line) were used. (a) CEFt following a 1 g q12h injection. (b) CEFu following a 1 g q12h injection. (c) CEFt following a 2 g q24h injection. (d) CEFu following a 2 g q12h injection. Figure 5. View largeDownload slide Simulations of steady-state ceftriaxone concentration–time profiles of a virtual patient with all median characteristics of the population, but with three different plasma albumin concentrations (colours) and two CLCR values (line or dotted line). For albumin, the observed median (blue), 25th percentile (red) and 75th percentile (green) were used. For CLCR, the median (line) and 25th percentile (dotted line) were used. (a) CEFt following a 1 g q12h injection. (b) CEFu following a 1 g q12h injection. (c) CEFt following a 2 g q24h injection. (d) CEFu following a 2 g q12h injection. For microorganisms with an MIC of 1 mg/L, the PTA of the 1 g q12h ceftriaxone regimen for patients with the median CLCR of 91 mL/min was 95.1% for the primary PD target of fT>MIC = 50% (Figure 6). The PTA of the secondary PD target of fT>MIC = 100% was 58.2%. In the case of a 2 g q24h ceftriaxone regimen, the PTA was 74.8% and 16.5% for the PD targets of fT>MIC = 50% and fT>MIC = 100%, respectively. When treating a pathogen with an MIC of 0.125 mg/L, the PTA of the fT>MIC = 50% PD target was 99.9% for a 1 g q12h regimen and 96.5% for a 2 g q24h ceftriaxone regimen in patients with a median albumin concentration of 29 g/L and median CLCR. The respective PTAs were 94.5% and 52.2% when the fT>MIC = 100% target was applied (Figure 6). Figure 6. View largeDownload slide Probability of target attainment with three different ceftriaxone dosing regimens. Percentages of 1000 simulated patients with a median albumin concentration (29 g/L) and a median estimated CLCR (91 mL/min) achieving a CEFu concentration above the MIC during half of the dosing interval (fT>MIC = 50%: a) and throughout the dosing interval (fT>MIC = 100%: b), for three different ceftriaxone dosing regimens and a range of MICs. Line with triangles, 2 g q24h; line with circles, 1 g q12h; line with squares, 2 g q12h. The clinical breakpoint for Enterobacteriaceae according to EUCAST is 1 mg/L.27 Figure 6. View largeDownload slide Probability of target attainment with three different ceftriaxone dosing regimens. Percentages of 1000 simulated patients with a median albumin concentration (29 g/L) and a median estimated CLCR (91 mL/min) achieving a CEFu concentration above the MIC during half of the dosing interval (fT>MIC = 50%: a) and throughout the dosing interval (fT>MIC = 100%: b), for three different ceftriaxone dosing regimens and a range of MICs. Line with triangles, 2 g q24h; line with circles, 1 g q12h; line with squares, 2 g q12h. The clinical breakpoint for Enterobacteriaceae according to EUCAST is 1 mg/L.27 Discussion The present study describes the development of a PPK model of ceftriaxone in an adult SSA hospital population with presumptive severe infection, based on observed CEFu and CEFt plasma concentrations. The study population was generally young and severely ill, as illustrated by its low median BMI and low median haemoglobin and albumin concentrations. A one-compartment model with non-linear protein binding best described CEFt/CEFu PK. Similar to other studies with ICU patients, the BPV of PK parameters was high, with the BPV of CEFu CL being as high as 77% in the final model. As can be expected with an antibiotic with renal CL as the predominant route of elimination, CEFu CL was significantly associated with CLCR and explained a substantial part of this BPV.28,29 Non-linear, saturable protein binding of ceftriaxone has been suggested by others as well.30,31 The relationship between the maximum ceftriaxone concentration bound to albumin (Bmax) and albumin concentration as found in our study implies that Bmax is lower at lower albumin concentrations. In agreement with this, the observed low median albumin concentration of 29 g/L corresponded to a median CEFu fraction of 19% which is high relative to what can be expected under normal albumin concentrations at therapeutic concentrations (5%–15%). Thus, with a higher CEFu fraction present at such lower albumin concentrations, more CEFu will be available for CL. At steady-state, such increased CL leads to a lower CEFt concentration that is also less representative of the CEFu concentration.30 Our findings match the results from earlier PK studies of ceftriaxone in non-SSA ICU patients that implied predominant dependency of CEFu exposure on CLCR.28,29,32,33 However, only one study suggested an inability to reach the fT>MIC = 100% target with a 2 g q12h prescription.32 This may be explained by the fact that not all studies based their conclusions on measured CEFu concentrations and that some may have been too small to draw meaningful conclusions. Although none of the measured liver enzymes proved to have a significant relationship with CEFu PK, other investigators have found bilirubin levels to be negatively correlated with albumin binding and non-renal CL of CEFu.28,34 Since we did not include bilirubin in our analysis, it cannot be excluded that part of the BPV may have resulted from a relationship between liver function and CEFu PK that was left unaccounted for. For a 2 g q24h dosing regimen, the PTA for the PD target of fT>MIC = 100% was just above 50% for pathogens with an MIC of 0.125 mg/L. Importantly, simulations also demonstrated that the use of the more commonly prescribed 1 g q12h ceftriaxone regimen may lead to CEFu underexposure as well, especially for patients with preserved kidney function. The risk for underexposure in patients with impaired renal function appears to occur mostly when stringent PD targets are applied. For pathogens with an MIC of 1 mg/L, both q24h and q12h dosing are likely to lead to CEFu underexposure. Although clinical and microbiological evidence is still limited, available data do seem to support the idea that in patients with severe disease, who may be at risk for impaired drug distribution and tissue penetration, the use of a more stringent PD target, such as fT>MIC = 100%, improves clinical outcome and reduces the emergence of antimicrobial resistance, compared with the most conservative target of fT>MIC = 50%.8,10,35,36 The CEFu concentration–time simulations could not be tested against the background of locally derived bacterial MICs, which means that the actual PTA could differ from the ones suggested on the basis of the MIC used for Enterobacteriaceae and S. pneumoniae (1 mg/L). Results from a surveillance study of global antimicrobial susceptibility showed that in Africa, 32% of S. pneumoniae isolates collected from 2009 to 2012 were resistant to penicillin (MIC >2 mg/L), and that in this group, roughly 25% of isolates were non-susceptible to ceftriaxone (MIC >1 mg/L).37 MDR of non-typhoidal Salmonella serovars, including resistance to third-generation cephalosporins, is emerging. Third-generation cephalosporin resistance still appears to be restricted to certain geographical areas, but very limited African susceptibility surveillance data are currently available.38 There are limitations to our study. Firstly, we did not collect information about a patient’s fluid balance and the presence of excess fluid in the interstitial and transcellular space, and are therefore not informed about how hydration status, oedema and ‘third spacing’ may have affected PK BPV. It is our experience that administration of sufficient amounts of fluids in adult hospitalized patients in resource-poor settings is rare and our observations are supported by reports about patients with presumptive sepsis from Uganda.39,40 An inadequate hydration status in a substantial amount of cases may imply that CEFu concentrations as well as PTAs could turn out to be lower than what was found in the current study once proper volume therapy is applied. Secondly, the Cockcroft and Gault formula’s predictive value is limited in critically ill patients without stable renal function. Although this formula has been shown to have relatively good accuracy for patients with BMI <18.5 kg/m2, CLCR estimation, and thus the estimation of the glomerular filtration rate, may therefore be biased.41 Thirdly, there was substantial residual PK variability as expressed by the CEFu and CEFt additive error. An explanation may lie in drug dispensing and administration procedures, even though these procedures were observed when possible. Although it is not unlikely that our study population’s risk of CEFu underexposure is indicative of what happens to PK/PD in critically ill patients in general, extrapolations to non-SSA severely ill (ICU) patient populations should be made with caution, as underlying disease, as well as treatment aspects, including antibiotic infusion time and volume therapy, are likely to be different. In conclusion, a one-compartment PPK model with non-linear protein binding adequately described the PK of CEFu and CEFt in a severely ill, non-ICU hospital population. Intermittent bolus dosing of the highly albumin-bound β-lactam ceftriaxone may lead to underexposure and can potentially pose a threat to a patient’s health, especially when a patient’s renal function is intact and when a q24h dosing regimen is used. Underexposure may also pose a threat to public health, as it is known to contribute to the emergence of antimicrobial resistance. These outcomes underscore a need for SSA antibiotic PK data against the background of locally derived MICs, so that well-known antibiotic drugs can be preserved for use in this region of the world. Acknowledgements In Mozambique, first of all, we want to thank the study participants for their participation and trust. We are indebted to the UCM Research Centre for Infectious Diseases (CIDI) for their professional collaboration with regard to the implementation of the study. Within the CIDI, we thank Ms Ivete Meque and Ms Kajal Chhaganlal, subsequent CIDI managers, as well as CIDI support staff for their share in ethics, good clinical and laboratory practice training, finances and daily transport of blood samples. We want to thank Professor Janneke van de Wijgert of the University of Liverpool for her advice with regard to the preparation of local contracts and agreements as well as Susan Mason of the United States Military HIV Research Program (MHRP), for the programming of the CIDI laboratory’s study database. We are grateful to the executive board of the HCB and the late Carlos de Oliveira, internist and former head of the HCB medicine department, for making research office space available on the HCB wards, as well as to the entire nursing staff from the HCB medicine wards for their collaboration. In the Netherlands, we thank Ms Femke Schrauwen, trial coordinator of the AMC biochemistry laboratory, and Ms Marloes van der Meer and Mr Marcel Pistorius, research analysts of the AMC pharmacology laboratory, for their useful input in the interpretation of biochemistry and ceftriaxone concentration results. Funding This work was funded with internal funding from the Academic Medical Centre (AMC). Additionally, it was indirectly supported by a grant from the Gilead Foundation (14 July 2014; IA 356007) and a grant from a Dutch private donor who wants to stay anonymous (CA 356001), but whose professional activities do not create conflict of interest for one of the authors. Both external funding parties supported the local presence of J. C. Bos in Mozambique in the context of a long-running medical educational capacity building project with the Faculty of Health Sciences of the Catholic University of Mozambique (FCS-UCM). Transparency declarations R. M. v. H. reports having received personal fees from Nordic Pharma. All other authors: none to declare. Author contributions J. C. Bos, R. A. A. M. and J. M. P. designed the study. J. C. Bos performed the literature search. J. C. Bos obtained ethical approval. M. C. M., G. N., C. N. L. and J. C. Beirão implemented the study and J. C. B. supervised data collection and study progress on a daily basis. J. C. Bos and R. M. v. H. analysed the data. J. C. Bos and R. M. v. H. drafted the manuscript. J. M. P. and R. A. A. M. critically examined the analysis and findings, and all authors critically read and commented on draft versions of the report. All authors approved the final version. References 1 Ford N , Shubber Z , Meintjes G et al. Causes of hospital admission among people living with HIV worldwide: a systematic review and meta-analysis . Lancet HIV 2015 ; 2 : e438 – 44 . Google Scholar CrossRef Search ADS PubMed 2 Meintjes G , Kerkhoff AD , Burton R et al. HIV-related medical admissions to a South African district hospital remain frequent despite effective antiretroviral therapy scale-up . Medicine (Baltimore) 2015 ; 94 : e2269. Google Scholar CrossRef Search ADS PubMed 3 Huson MAM , Kalkman R , Stolp SM et al. The impact of HIV on presentation and outcome of bacterial sepsis and other causes of acute febrile illness in Gabon . Infection 2015 ; 43 : 443 – 51 . Google Scholar CrossRef Search ADS PubMed 4 Prasad N , Murdoch DR , Reyburn H et al. Etiology of severe febrile illness in low- and middle-income countries: a systematic review . PLoS One 2014 ; 9 : e92131 . Google Scholar CrossRef Search ADS PubMed 5 Uche IV , MacLennan CA , Saul A. A systematic review of the incidence, risk factors and case fatality rates of invasive nontyphoidal Salmonella (iNTS) disease in Africa (1966 to 2014) . PLoS Negl Trop Dis 2016 ; 11 : e0005118 . Google Scholar CrossRef Search ADS 6 Roberts JA , Abdul-Haziz M , Lipman J et al. Individualized antibiotic dosing for patients who are critically ill: challenges and potential solutions . Lancet Infect Dis 2014 ; 14 : 498 – 509 . Google Scholar CrossRef Search ADS PubMed 7 Jager NGL , van Hest RM , Lipman J et al. Therapeutic drug monitoring of anti-infective agents in critically ill patients . Expert Rev Clin Pharmacol 2016 ; 9 : 961 – 79 . Google Scholar CrossRef Search ADS PubMed 8 Osthoff M , Siegemund M , Balestra G et al. Prolonged administration of β-lactam antibiotics—a comprehensive review and critical appraisal . Swiss Med Wkly 2016 ; 146 : w14368 . Google Scholar PubMed 9 Abdul-Aziz MH , Sulaiman H , Mat-Nor MB et al. β-Lactam infusion in severe sepsis (BLISS): a prospective, two-centre, open-labelled randomised controlled trial of continuous versus intermittent β-lactam infusion in critically ill patients with severe sepsis . Intensive Care Med 2016 ; 42 : 1535 – 45 . Google Scholar CrossRef Search ADS PubMed 10 Roberts JA , Paul SK , Akova M et al. DALI: defining antibiotic levels in intensive care unit patients: are current β-lactam antibiotic doses sufficient for critically ill patients? Clin Infect Dis 2014 ; 8 : 1072 – 83 . Google Scholar CrossRef Search ADS 11 Trobec K , Kos MK , von Haehling S et al. Pharmacokinetics of drugs in cachectic patients: a systematic review . PLoS One 2013 ; 8 : e79603. Google Scholar CrossRef Search ADS PubMed 12 Stanifer JW , Jing B , Tolan S et al. The epidemiology of chronic kidney disease in sub-Saharan Africa: a systematic review and meta-analysis . Lancet Glob Health 2014 ; 2 : e174 – 81 . Google Scholar CrossRef Search ADS PubMed 13 Spearman CW , Sonderup MW. Health disparities in liver disease in sub-Saharan Africa . Liver Int 2015 ; 35 : 263 – 71 . Google Scholar CrossRef Search ADS PubMed 14 Seboxa T , Amogne W , Abebe W et al. High mortality from blood stream infection in Addis Ababa, Ethiopia, is due to antimicrobial resistance . PLoS One 2015 ; 10 : e0144944 . Google Scholar CrossRef Search ADS PubMed 15 Leopold SJ , van Leth F , Tarekegn H et al. Antimicrobial drug resistance among clinically relevant bacterial isolates in sub-Saharan Africa: a systematic review . J Antimicrob Chemother 2014 ; 69 : 2337 – 53 . Google Scholar CrossRef Search ADS PubMed 16 Bos JC , Smalbraak L , Macome C et al. TB diagnostic process management of patients in a referral hospital in Mozambique in comparison with the 2007 WHO recommendations for the diagnosis of smear-negative pulmonary TB and extrapulmonary TB . Int Health 2013 ; 5 : 302 – 8 . Google Scholar CrossRef Search ADS PubMed 17 Ministério da Saúde da Republica de Moçambique . Formulário Nacional de Medicamentos. 5a edição. 2007. http://apps.who.int/medicinedocs/en/d/Js19267pt/. 18 North Shore LIJ. Human Subject Protection Program Guidance Document: Maximum Blood Draw Limits. 2014 . https://www.feinsteininstitute.org/professionals/resources-for-investigators/administrative-services/human-research-protection-program/tools-and-guidance/. 19 Cockcroft DW , Gault MH. Prediction of creatinine clearance from serum creatinine . Nephron 1976 ; 16 : 31 – 41 . Google Scholar CrossRef Search ADS PubMed 20 Bonate PL. Pharmacokinetic-Pharmacodynamic Modelling and Simulation, 2nd edn . New York, NY, USA : Springer , 2011 . Google Scholar CrossRef Search ADS 21 van Hest RM , van Gelder T , Vulto AG et al. Pharmacokinetic modelling of the plasma protein binding of mycophenolic acid in renal transplant recipients . Clin Pharmacokinet 2009 ; 48 : 463 – 76 . Google Scholar CrossRef Search ADS PubMed 22 Ahn JE , Karlsson MO , Dunne A et al. Likelihood based approaches to handling data below the quantification limit using NONMEM VI . J Pharmacokinet Pharmacodyn 2008 ; 35 : 401 – 21 . Google Scholar CrossRef Search ADS PubMed 23 Ette EI , Ludden TM. Population pharmacokinetic modelling: the importance of informative graphics . Pharm Res 1995 ; 12 : 1845 – 55 . Google Scholar CrossRef Search ADS PubMed 24 Keizer RJ , Karlsson MO , Hooker A. Modeling and simulation workbench for NONMEM: tutorial on Pirana, PsN, and Xpose . CPT Pharmacometrics Syst Pharmacol 2013 ; 2 : e50. Google Scholar CrossRef Search ADS PubMed 25 Jonsson EN , Karlsson MO. Xpose–an S-PLUS based population pharmacokinetic/pharmacodynamic model building aid for NONMEM . Comput Methods Programs Biomed 1999 ; 58 : 51 – 64 . Google Scholar CrossRef Search ADS PubMed 26 Lindbom L , Pihlgren P , Jonsson EN. PsN-Toolkit–a collection of computer intensive statistical methods for non-linear mixed effect modeling using NONMEM . Comput Methods Programs Biomed 2005 ; 79 : 241 – 57 . Google Scholar CrossRef Search ADS PubMed 27 EUCAST . Breakpoint Tables for Interpretation of MICs and Zone Diameters, Version 3.1. http://www.eucast.or/clinical_breakpoints/. 28 Schleibinger M , Steinbach CL , Töpper C et al. Protein binding characteristics and pharmacokinetics of ceftriaxone in intensive care unit patients . Br J Clin Pharmacol 2015 ; 80 : 525 – 33 . Google Scholar CrossRef Search ADS PubMed 29 Garot D , Respaud R , Lanotte P et al. Population pharmacokinetics of ceftriaxone in critically ill septic patients: a reappraisal . Br J Clin Pharmacol 2011 ; 72 : 758 – 67 . Google Scholar CrossRef Search ADS PubMed 30 Roberts JA , Pea F , Lipman J. The clinical relevance of plasma protein binding changes . Clin Pharmacokinet 2013 ; 52 : 1 – 8 . Google Scholar CrossRef Search ADS PubMed 31 Stoeckel K , McNamara PJ , Brandt R et al. Effects of concentration-dependent plasma protein binding on ceftriaxone kinetics . Clin Pharmacol Ther 1981 ; 29 : 650 – 7 . Google Scholar CrossRef Search ADS PubMed 32 Joynt GM , Lipman J , Gomersall CD et al. The pharmacokinetics of once-daily dosing of ceftriaxone in critically ill patients . J Antimicrob Chemother 2001 ; 47 : 421 – 9 . Google Scholar CrossRef Search ADS PubMed 33 Tsaj D , Stewart P , Goud R et al. Total and unbound ceftriaxone pharmacokinetics in critically ill Australian indigenous patients with severe sepsis . Int J Antimicrob Agents 2016 ; 48 : 748 – 52 . Google Scholar CrossRef Search ADS PubMed 34 Stoeckel K , Tuerk H , Trueb V et al. Single-dose ceftriaxone kinetics in liver insufficiency . Clin Pharmacol Ther 1984 ; 36 : 500 – 9 . Google Scholar CrossRef Search ADS PubMed 35 Sime FB , Roberts MS , Peake SL et al. Does pharmacokinetic variability in critically ill patients justify therapeutic drug monitoring? A systematic review . Ann Intensive Care 2012 ; 2 : 35. Google Scholar CrossRef Search ADS PubMed 36 Feng Y , Hodiamont CJ , van Hest RM et al. Development of antibiotic resistance during simulated treatment of Pseudomonas aeruginosa in chemostats . PLoS One 2016 ; 11 : e0149310 . Google Scholar CrossRef Search ADS PubMed 37 Tomic V , Dowzicky MJ. Regional and global antimicrobial susceptibility among isolates of Streptococcus pneumoniae and Haemophilus influenzae collected as part of the tigecycline evaluation and surveillance trial (T.E.S.T.) from 2009 to 2012 and comparison with previous years of T.E.S.T. (2004-2008) . Ann Clin Microbiol Antimicrob 2014 ; 13 : 52. Google Scholar CrossRef Search ADS PubMed 38 Marks F , von Kalckreuth V , Aaby P et al. Incidence of invasive salmonella disease in sub-Saharan Africa: a multicentre population-based surveillance study . Lancet Glob Health 2017 ; 5 : e310 – 23 . Google Scholar CrossRef Search ADS PubMed 39 Jacob ST , Moore CC , Banura P et al. Severe sepsis in two Ugandan hospitals: a prospective observational study of management and outcomes in a predominantly HIV-1 infected population . PLoS One 2009 ; 4 : e7782. Google Scholar CrossRef Search ADS PubMed 40 Jacob ST , Banura P , Baeten JM et al. The impact of early monitored management on survival in hospitalized adult Ugandan patients with severe sepsis: a prospective intervention study . Crit Care Med 2012 ; 40 : 2050 – 8 . Google Scholar CrossRef Search ADS PubMed 41 Michels WM , Grootendorst DC , Verduijn M et al. Performance of the Cockcroft-Gault, MDRD, and new CKD-EPI formulas in relation to GFR, age and body size . Clin J Am Soc Nephrol 2010 ; 5 : 1003 – 9 . Google Scholar CrossRef Search ADS PubMed 42 Reynolds JEF (ed). Martindale: The Extra Pharmacopoeia, 13th edn . London, UK : The Pharmaceutical Press , 1993 . © 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 7, 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