Population Pharmacokinetics and Dosage Optimization of Linezolid in Patients with Liver Dysfunction

Linezolid is the first synthetic oxazolidone agent to treat infections caused by Gram-positive pathogens. Infected patients with liver dysfunction (LD) are more likely to suffer from adverse reactions, such as thrombocytopenia, when standard-dose linezolid is used than patients with LD who did not use linezolid. Currently, pharmacokinetics data of linezolid in patients with LD are limited. This study aimed to characterize pharmacokinetics parameters of linezolid in patients with LD, identify the factors influencing the pharmacokinetics, and propose an optimal dosage regimen.

ABSTRACT

Linezolid is the first synthetic oxazolidone agent to treat infections caused by Gram-positive pathogens. Infected patients with liver dysfunction (LD) are more likely to suffer from adverse reactions, such as thrombocytopenia, when standard-dose linezolid is used than patients with LD who did not use linezolid. Currently, pharmacokinetics data of linezolid in patients with LD are limited. This study aimed to characterize pharmacokinetics parameters of linezolid in patients with LD, identify the factors influencing the pharmacokinetics, and propose an optimal dosage regimen. We conducted a prospective study and established a population pharmacokinetics model with the Phoenix NLME software. The final model was evaluated by goodness-of-fit plots, bootstrap analysis, and prediction corrected-visual predictive check. A total of 163 concentration samples from 45 patients with LD were adequately described by a one-compartment model with first-order elimination along with prothrombin activity (PTA) and creatinine clearance as significant covariates. Linezolid clearance (CL) was 2.68 liters/h (95% confidence interval [CI], 2.34 to 3.03 liters/h); the volume of distribution (V) was 58.34 liters (95% CI, 48.00 to 68.68 liters). Model-based simulation indicated that the conventional dose was at risk for overexposure in patients with LD or severe renal dysfunction; reduced dosage (300 mg/12 h) would be appropriate to achieve safe (minimum steady-state concentration [C_min,ss] at 2 to 8 μg/ml) and effective targets (the ratio of area under the concentration-time curve from 0 to 24 h [AUC_0–24] at steady state to MIC, 80 to 100). In addition, for patients with severe LD (PTA, ≤20%), the dosage (400 mg/24 h) was sufficient at an MIC of ≤2 μg/ml. This study recommended therapeutic drug monitoring for patients with LD. (This study has been registered in the Chinese Clinical Trial Registry under no. ChiCTR1900022118.)

INTRODUCTION

Patients with liver dysfunction (LD) tend to suffer microecological imbalance due to changes in the intestinal microbial flora, increased intestinal permeability, and immune dysfunction (1, 2), so they are susceptible to infection by Gram-positive bacteria. This trend is complicated by the increasing prevalence of methicillin-resistant Staphylococcus aureus (MRSA) among nosocomial isolates. Linezolid is the first oxazolidinone antimicrobial developed against many clinically important Gram-positive bacteria, including MRSA, resistant enterococcus (VRE), and vancomycin-resistant staphylococcus (VRSA) (3, 4). It inhibits protein synthesis by binding to the bacterial 23S rRNA of the 50S subunit, hindering the connection between mRNA and ribosome, and blocking the formation of the functional 70S initiation complex in the initial stage of bacterial translation (5, 6). Previous studies have suggested that the efficacy pharmacokinetic/pharmacodynamic (PK/PD) indexes of linezolid therapy in adults are correlated with the ratio of the area under the concentration-time curve from 0 to 24 h (AUC_0–24) to the MIC (AUC_0–24/MIC) of between 80 and 120 (7–9). Linezolid is rapidly and extensively absorbed after oral dosing (both tablet and suspension formulations), with a bioavailability of approximately 100%. Linezolid is a moderate lipophilic drug and is widely distributed to well-perfused tissues and body fluids and exhibits a low protein binding capacity of 31% (4). Linezolid is primarily metabolized by oxidation of the morpholine ring, which results in two inactive ring-opened carboxylic acid metabolites: the aminoethoxyacetic acid metabolite (PNU-142300) and the hydroxyethyl glycine metabolite (PNU-142586). Nonrenal clearance accounts for 65% of the total clearance of linezolid. Approximately 30% of the dose is excreted unchanged in the urine, 40% as metabolite PNU-142586, and 10% as metabolite PNU-142300 under steady state (10). It is assumed that when the recommended dose of 600 mg every 12 h (q12 h) for all patients is used, sufficient blood concentration can be obtained most of the time. Therefore, adjustments in linezolid dosing regimens are generally not required in patients regardless of age or renal or hepatic function, and therapeutic drug monitoring (TDM) is considered unnecessary (11). However, this proposal is based on healthy volunteers or noncritically ill patients (10, 12). Furthermore, an association between increasing plasma exposure and drug toxicities has been described (13, 14). For instance, the prevalent hematological toxicity is known to be affected by the prolonged use and higher concentration of linezolid. Therefore, TDM of linezolid has been explored to achieve safe and effective drug concentrations (15–17).

The drug product instruction suggested that linezolid in 7 patients with mild to moderate liver diseases (Child-Pugh class A or B) showed no significant differences in drug clearance values compared with healthy volunteers, and the pharmacokinetics properties were not evaluated in patients with severe LD. To the best of our knowledge, the liver plays a crucial role in drug disposition because it is a major site for drug metabolism and clearance and the hepatic blood flow is one of the factors that affect the liver’s ability to metabolize and eliminate drugs (18). Consequently, liver function changes cause changes in drug disposition. At present, there are no endogenous liver clearance markers or a scoring system that has a good correlation with liver clearance and drug metabolism in LD (19). Therefore, it may be difficult to determine the effect of LD on drug disposition. However, the known pharmacokinetics changes have occurred in patients with LD, especially during end-stage disease. Although the metabolic pathway of linezolid is not fully understood and in vitro studies have demonstrated that linezolid is minimally metabolized by human cytochrome P450, some studies have shown increased risk of linezolid overexposure and adverse events in patients with LD. Nowadays, very few regimens have been recommended for linezolid dose adjustments in patients with LD. Therefore, patients with LD using linezolid is an important group who have rarely been studied previously. According to the early reports (20), a high frequency of thrombocytopenia in patients with acute-on-chronic liver failure treated with linezolid has been observed. A few studies explored the contribution of liver function to the pharmacokinetics of linezolid. Sasaki et al. (21) performed a population pharmacokinetics analysis among Japanese patients and revealed that renal function and severe liver cirrhosis significantly affected the pharmacokinetics of linezolid, although only 4 liver cirrhosis patients were included. Luque et al. (22) demonstrated that patients with liver cirrhosis were more likely to be exposed to supratherapeutic concentrations and suffer adverse events when they received standard-dose linezolid (600 mg/12 h) than patients with LD who did not use linezolid.

Although several studies have assessed the pharmacokinetics of linezolid in patients with hepatic dysfunction, its pharmacokinetics profile has not been fully evaluated and the dose regimen to achieve a PK/PD target has yet to be established. Improper dosing in patients with LD may cause treatment failure, antibiotic resistance, or increase toxicity. The clinical application of linezolid has increased progressively over the years. Data on the linezolid pharmacokinetics profile and TDM-guided individualized dosing regimens in hepatic dysfunction patients are still lacking but are of utmost importance.

Population pharmacokinetics (PPK) (23) is a combination of classical pharmacokinetics and statistical methods to explore, describe, and predict the pharmacokinetics and pharmacodynamics of drugs in specific groups. PPK seeks to identify the factors of patient-related and clinical-related variability that cause changes in the drug concentration and the degree of these changes. Therefore, the aims of the present study are to characterize linezolid pharmacokinetics parameters in intravenous linezolid patients with LD, to identify the factors influencing the pharmacokinetics of linezolid, and to propose an evidence-based optimal dosage regimen for this vulnerable population with Monte Carlo Simulation (MCS).

RESULTS

Demographics and characteristics of patients.

A total of 51 patients were collected and 45 were included in the study. The other 6 patients were excluded due to missing data. Ultimately, among all enrolled patients, one had reduced dose (600 mg, q24 h) at the beginning and a second sampling was taken from two patients after the dose was reduced (600 mg, q24 h). The latter two patients were not regarded as new cases. Patient demographics and clinical characteristics were summarized in Table 1. Thirty patients (66.67%) were diagnosed with liver failure or cirrhosis (Child‐Pugh C), 13 patients (28.89%) had moderate LD (Child‐Pugh B), and 2 (4.44%) had mild LD (Child‐Pugh A).

TABLE 1.

Demographics and clinical characteristics of patients

Characteristica	No.	Median (range)
Gender
Male	39
Female	6
Sample (μg/ml)	163	17.12 (2.24–59.84)
Age (yr)		47 (28–68)
Height (cm)		168 (154–176)
Body weight (kg)		65.5 (45.5–95)
BMI (kg/m2)		23.03 (16.12–36.98)
MELD score		21.7 (5.10–41.37)
HGB (g/liter)		89.8 (54–131)
PLT count (109/liter)		94 (19.5–262.5)
ALT (U/liter)		33.8 (2.77–501.03)
AST (U/liter)		53.23 (12.93–967.88)
ALB (g/liter)		30.3 (17.40–52.03)
TBIL (μmol/liter)		100.43 (4.15–630.95)
DBIL (μmol/liter)		65.3 (1.90–493.75)
TBA (μmol/liter)		102.6 (4.20–398.50)
GGT (U/liter)		96.3 (8.2–712.40)
ALP (U/liter)		129.3 (44–933.10)
INR		1.65 (1.09–4.65)
PTA (%)		48.07 (17.20–88.95)
BUN (mmol/liter)		6.02 (1.64–54.81)
UA (μmol/liter)		244.7 (50.20–764.05)
CLcr (ml/min)		99.3 (13.98–257.51)

^aBMI, body mass index; MELD score, model for end-stage liver disease; HGB, hemoglobin; PLT count, platelet count; ALB, albumin; ALT, alanine aminotransferase; AST, aspartate aminotransferase; TBIL, total bilirubin; DBIL, direct bilirubin; TBA, total bile acid; GGT, gamma-glutamyl transpeptidase; ALP, alkaline phosphatase; INR, international normalized ratio; PTA, prothrombin activity; BUN, blood urea nitrogen; UA, uric acid; CLcr, creatinine clearance rate.

Development of population pharmacokinetics model.

A total of 163 linezolid concentrations ranging from 2.04 to 59.84 μg/ml were obtained for PPK modeling (2 patients had only a single sample). The linezolid concentration-versus-time profile is shown in Fig. 1. Most of the time points fell within 12 h. A second sampling (2 to 3 blood samples) was done when the initial dose changed in two patients, so the time axis extended to 24 h. The one-compartment model with first-order elimination adequately described the PK characteristics of linezolid. The population base model was parameterized in terms of clearance (CL) and the volume of distribution (V). The interindividual variability and residual variability were illustrated by the exponential model and the proportional error model, respectively.

FIG 1.

Linezolid concentrations versus time.

We performed a correlation analysis of covariates to avoid the simultaneous incorporation of correlated or colinear variables before the screening of covariates, and then the examined covariates were added to the base model with a stepwise method to create the full covariate model. Among all the 19 covariates tested, PTA and creatinine clearance rate (CLcr) exhibited a significant effect on linezolid CL and were positively correlated with CL. When only PTA was enrolled into the final model, the interindividual variability of CL was reduced from 36.47% to 20.55%. After including PTA and CLcr into the PPK model, the interindividual variability of CL decreased from 36.47% to 15.47%, suggesting PTA would better decrease interindividual variability. After combining these covariables, no other covariables were found to significantly affect the PK of linezolid. In the backward elimination procedure, PTA and CLcr were retained in the final model. The relationship of covariates in developing the full covariate model was shown in Table 2.

TABLE 2.

Covariate hypothesis testing in the model development^a

Model description	OFV	ΔOFV	P value
Basic model	1,009.90
Full covariate model (V-WT Cl-PTA Cl-CLcr)	976.127
Backward elimination
Removing Cl-PTA Cl-CLcr	980.868	4.741	<0.05
Removing Cl-CLcr	1,005.726	24.858	<0.01
Removing CL-PTA	992.308	11.44	<0.01

^aOFV, objective function value; V, the volume of distribution; CL, clearance; WT, body weight; PTA, prothrombin activity; CLcr, creatinine clearance rate; ΔOFV, change in the OFV compared to reference model.

The final model is as follows:

$$\text{CL}(\text{liter}/\text{h})=2.68\times {(\text{PTA}/48.07)}^{0.84}\times {(\text{CLcr}/99.30)}^{0.36}\times \text{exp}(\eta \text{CL})$$

$$V\hspace{0.17em}(\text{liter})=58.34\times \text{exp}(\eta V)$$

The population parameters of V and CL estimated in the final model were 58.34 liter and 2.68 liters/h, respectively. The details are displayed in Table 3.

TABLE 3.

Linezolid PPK parameter estimates from the final model and bootstrap analysis^a

Parameter	Final model		Bootstrap analysis (successful running times = 1,000)
	Estimate (shrinkage %)	CV (%)	Median estimate	95% CIs
TVV (liter)	58.34	8.97	57.77	48.41–69.91
TVCL (liter/h)	2.68	6.52	2.67	2.32–3.04
ΘPTA	0.84	21.01	0.85	0.39–1.21
ΘCLcr	0.36	21.63	0.37	0.20–0.60
Interindividual variability
ω2V (%)	8.89 (47.83)	29.82	7.74
ω2CL (%)	15.47 (7.23)	39.33	15.10
Residual variability
σ (%)	18.50	10.80	18.30	0.14–0.23

^aCV, coefficient of variation; CIs, confidence intervals; TV_V, typical value of apparent distribution; TV_CL, typical value of apparent volume clearance; Θ_PTA exponent for PTA as a covariate for CL; Θ_CLcr, exponent for CLcr as a covariate for CL; ω²_V, variance of interindividual variability for V; ω²_CL, variance of interindividual variability for V; σ, square root of residual variability for the final model.

Model validation.

The diagnostic goodness-of-fit plots obtained from the basic and final PPK model are presented in Fig. 2 and 3. The plots of PRED (B1) and IPRED (A1) versus DV showed no structural deviations in terms of visual biases, and the fit of the final model was improved compared with the basic model. In the plots of conditional weighted residuals (CWRES) versus PRED and time, most concentration data were randomly distributed around 0 and within −2 to +2, which indicated no significant systematic deviations in the model fit. The final model (C1 and D1) showed an improvement of fit over that of the basic model (C and D).

FIG 2.

Goodness-of-fit plots for the basic model (A and B) and the final model (A1 and B1). (A and A1) The observed concentrations (DV) versus individual-predicted concentrations (IPRED); (B and B1) the DV versus population-predicted concentrations (PRED).

FIG 3.

Goodness-of-fit plots for the basic model (C and D) and the final model (C1 and D1). (C and C1) The conditional weighted residuals versus population-predicted concentrations; (D and D1) the conditional weighted residuals versus time.

The obtained medians and 95% confidence intervals (CIs) of parameter estimates from a 1,000-run bootstrap sets analysis are shown in Table 3. The parameter estimates from the original data lay within the 95% CIs resulted from the nonparametric bootstrap method, and the 95% CIs did not include zero. The biases between the final model estimates and the bootstrapped median parameter estimates were less than 10% for all parameters, suggesting good stability and robustness of the final model.

Figure 4 showed a prediction corrected-visual predictive check (pc-VPC) of concentration versus time after the last dose. Most of the observed 5th, 50th, and 95th quantiles were within the 90% CIs of the predicted corresponding quantiles, indicating an acceptable consistency between the observed and simulated concentrations. Overall, the evaluation of the linezolid PPK model demonstrated that the final model provided sufficient description of the data.

FIG 4.

Prediction corrected-visual predictive check of the final model. The observed linezolid concentrations are shown as blue circles. Red solid and dashed lines represent the 5th, 50th, and 95th percentiles of the observed concentrations; the 3 shaded areas represent the 90% CIs of the 5th, 50th, and 95th percentiles of the simulated concentrations.

Model-based simulation and dosage optimization.

Table 4 and 5 displayed the results of MCS based on the final PPK model. Dosage was adjusted according to different grades of liver function (PTA) and renal function (CLcr), the target probability (percent), and the distribution of C_min,ss. Simulations demonstrated that the conventionally approved dosage at 600 mg, q12 h would lead to a high risk of overexposure for patients with liver or renal dysfunction. As shown in Table 4, at an MIC of 2 μg/ml, 600 mg/12 h would be sufficient to obtain an AUC_0–24/MIC of >100 in most scenarios, and when patients with PTA of ≤20% or CLcr of ≤10 ml/min, the probabilities of C_min,ss of >8 μg/ml were 64.4% and 34.1%, respectively. The probabilities of AUC_0–24/MIC of >80 (92.9%, 100%, and 96.9%) were desirable at 600 mg/day in patients with PTA of ≤40%, PTA of ≤20%, and CLcr of ≤10 ml/min. In addition, the regime 300 mg/12 h was more likely to obtain desirable C_min,ss than 600 mg/24 h. For patients with PTA of ≤40% or CLcr of ≤10 ml/min, 300 mg/12 h was superior to 600 mg/24 h (Table 5). Furthermore, for patients with LD with PTA of ≤20%, a dose of 400 mg/24 h might still reach the desirable probability, indicating the need for reduced dosage for patients with LD. When MIC was 4 μg μg/ml, with a dosage of 300 mg/12 h, the probability reached 91.9% for a AUC_0–24/MIC ratio of ≥80, and a dosage of 400 mg, q24 h made the probability reach 59.1% for patients with PTA of ≤20%.

TABLE 4.

Percentage of patients with distribution probability of AUC_0–24/MIC and C_min,ss in each simulated population^a

Dosage (mg/12 h)	PTA (%)	CLcr (ml/min)	% of patients by Cmin,ss (μg/ml) category			% of patients by distribution probability of AUC0–24/MIC
						MIC of 2 μg/ml			MIC of 4 μg/ml
			<2	2–8	>8	<80	80–100	>100	<80	80–100	>100
600	80	99.3b	3.7	95.4	0.9	6.0	12.0	82.0	64.3	19.7	16.0
	40	99.3b	0.3	74.2	25.5	0	0.3	99.7	7.1	12.8	80.1
	20	99.3b	0	35.6	64.4	0	0	100	0	0.3	99.7
	80	50	0.6	94.6	4.8	1.8	2.9	95.3	23.2	38.9	37.9
	80	30	0.2	89.2	10.6	0.3	2	97.7	19.4	21.7	58.9
	80	10	0	65.9	34.1	0	0	100	3.1	5.0	91.9
300	80	99.3b	37.3	62.7	0	64.3	19.7	16.0	98.8	0.9	0.3
	40	99.3b	2.6	97.3	0.1	7.1	12.8	80.1	66.9	17.9	15.2
	20	99.3b	0	99.8	0.2	0	0.3	99.7	8.1	14.8	77.1
	80	50	13.6	86.4	0	37.9	23.2	38.9	93.8	4.4	1.8
	80	30	5.2	94.8	0	19.2	21.8	59.0	84.3	9.6	6.1
	80	10	0.5	99.4	0.1	3.1	4.7	92.2	29.1	23.8	47.1
200	80	99.3b	84.1	15.9	0	91.3	6.6	2.1	100	0	0
	40	99.3b	32.6	67.4	0	34.5	21.9	43.6	93.3	4.2	2.5
	20	99.3b	3.4	96.6	0	2.1	3.3	94.6	40.1	19.1	40.8
	80	50	67.2	32.8	0	77.3	13.7	9.0	99.6	0.4	0
	80	30	52.9	42.1	0	56.7	21.0	22.3	98.2	0.6	1.2
	80	10	20.1	79.9	0	10.7	25.0	64.3	84.1	9.3	6.6

^an = 1,000. PTA, prothrombin activity; CLcr, creatinine clearance rate; C_min,ss, minimum steady-state concentration.

^bMedian value.

TABLE 5.

Percentage of patients with distribution probability of C_min,ss in each simulated population^a

Dosage (mg/24 h)	PTA (%)	CLcr (ml/min)	% of patients by distribution of Cmin,ss
			<2 μg/ml	2–8 μg/ml	>8 μg/ml
600	80	99.3b	58.7	41.3	0
	40	99.3b	7.2	92.2	0.6
	20	99.3b	0.1	83.9	16
	80	50	31.7	68.3	0
	80	30	19	80.6	0.4
	80	10	3.2	95.2	2
400	80	99.3b	21.7	78.3	0
	40	99.3b	10.4	89.6	0
	20	99.3b	1.3	98.7	0
	80	50	61.3	38.6	0
	80	30	44.6	55.4	0
	80	10	12.8	87.2	0

^an = 1,000. PTA, prothrombin activity; CLcr, creatinine clearance rate; C_min,ss, minimum steady-state concentration.

^bMedian value.

DISCUSSION

Although the clinical application of linezolid has increased progressively over the years, there are few pharmacokinetics data of linezolid in patients with LD. Data on the linezolid pharmacokinetics profiles and therapeutic drug monitoring-guided individualized dosing regimens in patients with LD are very limited. To the best of our knowledge, the present work is the first prospective study to analyze and report a PPK model, with dosage optimization of intravenous linezolid in patients with LD. The PPK model was established and validated to determine linezolid pharmacokinetics parameters in patients with LD and to identify the impact of demographics and clinical factors on linezolid pharmacokinetics. Ultimately, the dosage was optimized in patients with LD with MCS based on the final model.

The one-compartment model with first-order elimination, along with PTA and CLcr as significant covariates, was appropriate for the final PPK model. In the final model, the population estimate CL of linezolid in LD was 2.68 liters/h (95% CI, 2.34 to 3.03 liters/h), which is significantly lower than the findings of previous studies with healthy volunteers and other patients (4.7 to 8.3 liters/h) (7, 8, 12, 24). Nevertheless, the result is similar to that of the study of Sasaki et al. (21), in which a significant decrease in linezolid clearance (CL about 1.35 liter/h) was observed in patients with severe liver cirrhosis. In addition, this finding is consistent with a report of an approximately 60% reduction in the CL in patients after the liver transplantation (25). The reason for the lower linezolid CL in patients with liver cirrhosis may be the numerous pathophysiological changes in the liver, with oxidative stress being the likely possible underlying mechanism (17). Wynalda et al. proposed that the major metabolite PNU-142586 in human liver microsomes was formed through reactive oxygen species and was not produced by cytochrome P450 enzymes (26). It was reported that the downregulation of proteins related to reactive oxygen species occurred in the liver during advanced fibrosis (27). In addition, cirrhosis is characterized by the destruction of the lobular structure. Therefore, liver fibrosis changed the enzyme expression, reduced the blood flow, and ultimately resulted in a reduction in the transport of the drug and oxygen to hepatocytes (28).

Previous studies have also identified the presence of moderate or severe liver diseases as a risk factor for developing hematological toxicity (20, 29–31). In the same way, lactic acidosis has proven to be more common in patients with liver impairment (32). Liver diseases are also associated with increased linezolid trough concentrations. In a Japanese study (21), patients with severe cirrhosis all showed high trough linezolid concentrations (32.5, 36.4, 40.8, and 45.4 μg/ml). The study also showed that liver cirrhosis was an independent risk factor for thrombocytopenia. Luque et al. (22) investigated the relationship between linezolid standard dose and risky toxicity in patients with liver cirrhosis and found that patients were more likely to achieve supratherapeutic concentrations and to suffer severe thrombocytopenia. The reason may be a reduced nonrenal clearance (26, 27). In a study (33) supporting these results, Wicha et al. explored the value of the maximal liver function capacity in linezolid clearance, which showed that LD was closely related to reduced nonrenal clearance of linezolid, and therapeutic drug monitoring was recommended for those patients. Therefore, to reduce adverse events caused by an overexposure of linezolid, a lower drug dose or longer intervals should be clinically considered. Many studies (34–38) have proposed that therapeutic drugs should be monitored and individualized dosing regimen adjusted based on linezolid concentrations and disease progression.

Meanwhile, many studies (21, 39–42) demonstrated that the clearance of linezolid was low in patients with renal insufficiency. The metabolic pathway of linezolid has not yet been fully elucidated. It is known that linezolid is mainly cleared in a nonrenal way (65%) into two metabolites (PNU142300 and PNU-142586), and 30% of the parent compound is excreted via the kidney (4). Therefore, candidates of covariates in the present study consisted of blood biochemical indexes of renal function, which were accurately collected and appropriately analyzed. In the present study, 4 patients had mild renal insufficiency (CLcr, 30 ml/min to 50 ml/min), 4 patients had moderate renal insufficiency (CLcr, <30 ml/min), and no patient had severe renal insufficiency (CLcr, <10 ml/min). PTA and CLcr were found associated with the CL of linezolid in this study. The final model demonstrated that a decrease in linezolid CL was closely related to low PTA values. PTA is an indicator of the coagulation function. In patients with LD, due to the decreased synthesis of coagulation factors, coagulation time was prolonged to various degrees, and the PTA decreased greatly. Therefore, decreased PTA is commonly observed in patients with LD as a consequence of the impaired synthetic capacity of the liver, which is also regarded as one of the important indicators in the diagnosis of cirrhosis and liver failure (43–45). In our study, most patients had different degrees of decrease in PTA, which should be taken into account to adjust linezolid dosage clinically. This study showed that CL was low in patients with LD when receiving standard-dose linezolid. In addition, CL of linezolid was influenced by CLcr, indicating that deteriorations in the liver function and severe renal dysfunction may affect the pharmacokinetics of linezolid. Therefore, linezolid therapeutic drug monitoring that helps individualize dosing for the vulnerable population to avoid potential overexposure and subsequent toxicity is required.

According to previous research, actual body weight (WT) is known to have an obvious impact on drug disposition. However, linezolid blood samples were not collected at the distribution phase of the current study. Therefore, it is possible that a search of covariates for V has not been adequately performed. Thus, WT was not a significant covariate for V. The final PPK model indicated that the population typical value of V was 58.34 liters (95% CI, 48.00 to 68.68 liters), which was similar to that (40 to 70 liters) reported in previous studies (21, 42, 45), suggesting that the distribution of linezolid may not be affected by LD. From the perspective of pharmacokinetics, medicine, and physiology, the PPK parameters obtained in this study were considered clinically valid.

MCS based on the final model demonstrated that standard dose is appropriate in patients with normal liver function and without insufficient renal function when the infection was caused by pathogens with an MIC of 2 μg/ml. However, whether the MIC was 2 or 4 μg/ml in patients with LD (PTA, ≤40%; PTA, ≤20%) or sever renal impairment (CLcr, ≤10 ml/min), standard dose could make the C_min,ss exceed the toxicity threshold (the probabilities of C_min,ss of >8 μg μg/ml were 25.5%, 64.4% and 34.1%, respectively). For patients with 20% ≤ PTA ≤ 40% or CLcr of ≤10 ml/min, the dose of 300 mg/12 h was considered safe and effective based on the AUC_0–24/MIC ratio of ≥80 (at an MIC of 2 μg/ml, the probabilities were 92.9% and 96.9%) and toxicity threshold (the probabilities of C_min,ss within 2 to 8 μg μg/ml were 97.3% and 99.4%). This is consistent with the findings of a previous PPK/PD study of Japanese patients (21), which included 4 patients with severe cirrhosis. Another study (22) about the risk of standard dose of linezolid for patients with liver cirrhosis suggested that a reduced dose at 600 mg/24 h was preferred. What is more, 400 mg, q24 h was sufficient to achieve the target (AUC_0–24/MIC ≥80) (probability, 97.9%) against pathogens with an MIC of 2 μg/ml for patients with LD with a PTA of ≤20%. When the MIC was 4 μg/ml, the probabilities for AUC_0–24/MIC of ≥80, C_min,ss of >8 μg/ml, and C_min,ss within 2 to 8 μg μg/ml were 91.9%, 0.2%, and 99.8% at 300 mg/q12 h for patients with PTA of ≤20%. Therefore, individualized administration with therapeutic drug monitoring is recommended for these patients.

There are some limitations in this study. First, this was a single-center study with a limited sample size. External validation was not performed in the final model, and the results need further cross validation. Second, the results from MCS and the relationships between pharmacodynamics (efficacy and safety) and pharmacokinetics were not evaluated in the present study. A multicenter study with a sufficient number of patients is needed in this particular population to verify the current results to ensure more effective clinic application of linezolid in the future.

In this study, we successfully developed the PPK model of linezolid in hospitalized patients with LD. In the final model, PTA and CLcr are positively correlated with linezolid CL, confirming that LD is an important factor for the pharmacokinetics and dosing optimization of linezolid. Simulations based on the final model reveal that a standard dose (600 mg/12 h) is more likely to cause overexposure than a reduced dose (300 mg/12 h) for patients with liver cirrhosis (PTA, ≤40%) or severe insufficient renal function (CLcr, ≤10 ml/min). For patients with liver cirrhosis (20% ≤ PTA ≤ 40%), renal dysfunction (CLcr, ≤10 ml/min), and severe liver cirrhosis (PTA, ≤20%), the dosage of 300 mg/12 h and 400 mg/24 h would be sufficient at MICs of ≤2  μg/ml. For infected with pathogens with an MIC of ≥2 μg/ml, the adequate dose regimen is 600 mg/12 h and 300 mg/12 h, respectively. The present study demonstrates the need for the dosage optimization of linezolid based on the PPK model, the need to perform therapeutic drug monitoring, and individualized administration for patients with LD or sever insufficient renal function. In conclusion, we identify the optimal dosage of linezolid that may be safely used to effectively treat Gram-positive infections in populations with LD.

MATERIALS AND METHODS

Patients and study design.

A prospective, single-center study was performed in the Department of Infectious Diseases, Second Xiangya Hospital of Central South University, from November 2018 to December 2019. This study was conducted in accordance with the legal requirements and the Declaration of Helsinki. It was approved by the Medical Ethics Committee of Second Xiangya Hospital of Central South University, and the clinical trial was registered on the Chinese Clinical Trail Registry (no. ChiCTR1900022118). Written informed consent was obtained from either the patient or their designated decision-makers. Patient privacy was protected, and personal information was processed so that patients could not be identified.

Patients with proven or probable Gram-positive infection who received linezolid were eligible based on the following inclusion criteria: (i) 18 years or older; (ii) diagnosis with LD, such as liver cirrhosis or liver failure based on the Child‐Pugh grade or end-stage liver disease score; and (iii) treated with at least one linezolid concentration during treatment. Patients were excluded when they (i) were younger than 18 years; (ii) were pregnant; (iii) had known hypersensitivity to linezolid; (iv) had a planned hospitalization of <3 days; (v) had linezolid administration 72 h prior to the study; (vi) had concomitant use of inducers that might significantly affect linezolid pharmacokinetics, such as rifampin; (vii) was undergoing renal replacement therapies; and (viii) missed providing clinical data, such as weight. The dosing regimen for patients was based on the linezolid manufacture package insert (600 mg, q12 h) with intravenous infusion between 1 and 2 h. Dosage adjustments were recommended by the pharmacy department when linezolid minimum steady-state concentration (C_min,ss) fell outside the desirable range of 2 to 7 μg/ml (34). However, the physician finally decided whether to adopt the recommendation after considering the patient’s condition. In some cases, a second sampling occurred after the initial dose was changed.

Patient clinical data were recorded on a data collection template by reviewing electronic medical records, including linezolid dosage, frequency, and duration; demographics, such as age, gender, height, and actual body weight (WT); and routine blood and biochemical indexes, such as hemoglobin (HGB), platelet count (PLT), alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (TBIL), direct bilirubin (DBIL), albumin (ALB), gamma-glutamyl transpeptidase (GGT), alkaline phosphatase (ALP), prothrombin activity (PTA), international normalized ratio (INR), serum creatinine (CREA), blood urea nitrogen (BUN), and uric acid (UA). Data are shown as median (range) for continuous variables.

Blood sampling.

A predetermined sampling scheme was performed when the steady-state concentration (Css) of linezolid was achieved (at least 3 days after the treatment initiation). From each patient, 1 to 4 blood samples were randomly collected at the time points of 0.5, 1.0, 1.5, 2, 4, 6, and 8 h after the administration and before the next round of administration. Blood samples were immediately placed on ice and centrifuged within 60 min at 3,000 rpm for 10 min before being stored at –80°C until the analysis.

Determination of linezolid concentrations.

The powder of a linezolid standard (purity of 98%, lot number 1-MLM-3-1) for high-performance liquid chromatography (HPLC) was purchased from Toronto Research Chemicals (Canada). All other reagents were of analytical or HPLC grade and were commercially available. The linezolid plasma concentrations were analyzed by automatic two‐dimensional liquid chromatography (2D‐HPLC; Demeter Instrument Co., Ltd., Hunan, China). The two‐dimensional separation conditions were as follows: the first‐dimensional chromatographic column was Aston SNCB (4.6 by 50 mm, 5 μm), and the mobile phase (pH adjusted to 8.0 by diethylamine) consisted of a mixture of methanol and a solution of 10 mmol/liter formic acid (38:62 [vol/vol]) delivered isostatically at 1.0 ml/min. The second‐dimensional chromatographic column was Aston SC5 (4.6 by 275 mm, 5 μm); the mobile phase included an acetonitrile:isopropanol:10 mmol/liter formic acid solution (diethylamine was used to adjust pH to 3.3) (72:24:4 [vol/vol/vol]) and the flow rate at 1.2 ml/min. For each sample, the plasma was deproteinized with methanol; 200 μl plasma was mixed with 500 μl methanol and was vortexed for 1 minute. The mixture was then centrifuged for 8 minutes at 14,000 × g in a refrigerated laboratory centrifuge before 20 μl of the supernatant was injected into the system for the assay. The chromatogram running time was 14 minutes; the column temperature was maintained at 40°C and the UV detector wavelength was set at 254 nm.

Good linearity was found over the full range of the assay concentrations (0.50 to 50.57 μg/ml in the plasma). The limit of quantification of linezolid in the plasma was 0.22 μg/ml (signal-to-noise ratio >10). The regression equation was Y = 1,498,355X + 41,557, R = 0.999, and the accuracy was within 85% to 115% for all concentrations. The method was sensitive and specific to measure linezolid in the plasma. The concentrations of low-, mid-, and high-quality controls were 1.76 μg/ml, 19.99 μg/ml, and 37.63 μg/ml, respectively. The intraday and interday precisions were 2.09% to 3.21% and 2.49% to 4.21%, respectively. The extraction and method recovery ranged from 95.84% to 100.07% and 93.48% to 96.98% for all concentrations, respectively. The 24-h precisions of storage in an autosampler were within 1.81%. The benchtop stabilities at room temperature for 48 h and 5 repeated freeze‐thaw cycles were within 2.89% and 2.09%, respectively. The precisions of stock solution and long-term stabilities were within 2.48% and 2.37%. The accuracy and precision of dilution were 99.29% and 1.93%, respectively, indicating linezolid was stable after dilution.

Population pharmacokinetic analysis.

The concentration‐time data of linezolid were obtained using a nonlinear mixed‐effects population approach with the Phoenix NLME computer program (version 8.1; Pharsight, A Certara Company, USA). A first-order conditional estimation-extended least-squares method (FOCE-ELS) was used throughout the development of population pharmacokinetics model.

Structural model.

One-and two-compartment pharmacokinetics models with first-order elimination were tested in the structure model screening. The optimal structure model was evaluated by goodness-of-fit plots criterion, accuracy and reliability of the parameter estimation, and improvement of objective function value (OFV). In the maximum likelihood approach, OFV is defined as the −2 × log likelihood value.

Statistical model.

For the linezolid population pharmacokinetics model, the interindividual variability was described by an exponential error model:

$$P_i=\text{TV}(P)\times \text{exp}({\eta }_i)$$

where P_i is the pharmacokinetic parameter estimation of the i-th individual, TV (P) is the typical value, and η_i is a random variable for individual i-th, which is a normally distributed random variable with mean zero and variance ω².

The residual variability models were performed as follows:

$$\text{Additive error model:}{C}_{\text{obs}}=C_{\text{pred}}+\epsilon$$

$$\text{Proportional error model:}{C}_{\text{obs}}=C_{\text{pred}}\times (1+\epsilon )$$

$$\text{Combined error model:}{C}_{\text{obs}}=C_{\text{pred}}\times (1+\epsilon )+{\epsilon }_1$$

C_obs is the measured concentration and C_pred is the predicted concentration based on the model, and ε and ε₁ are variables with mean zero and variance σ² and are associated with additive and proportional residual random errors, respectively. The model error; intraindividual variability; and errors in dosage, collection, processing, and bioassay of samples were all included in the residual variability.

Covariate model.

The covariates investigated in the study included gender, age, WT, height, BMI, HGB, PLT count, ALT, AST, ALB, TBIL, DBIL, TBA, ALP, GGT, INR, BUN, UA, and creatinine clearance rate (CLcr). CLcr was calculated with the Cockcroft and Gault equation (46). Before the screening of covariates, correlation analysis was performed to ensure that the final model would not simultaneously include the effects of correlated or colinear variables. Covariates were screened with a stepwise method. During the forward stepwise process, covariates were added to the basic model and considered statistically significant if the OFV decreased (ΔOFV) by more than 3.84 (P < 0.05 with 1 degree of freedom). Then, a backwards elimination process was conducted, and covariates were removed from the full model if the increase of the OFV (ΔOFV) was <6.635 (P < 0.01 with 1 degree of freedom).

For continuous covariates, the exponential model was performed according to the following equation:

$$P_i=\text{TV}(P)\times {(\text{co}{\text{v}}_j/\text{co}{\text{v}}_{\text{median}})}^{\theta }$$

Covariates were normalized by the median, where P_i is the value of parameter P for the i-th individual, TV (P) is the typical value of P, cov_j represents the j-th covariate, cov_median is the median of the covariate, and θ is the estimated parameter describing the fixed effect of covariates on the pharmacokinetics parameters.

For categorical covariates, such as gender, the effect on parameter P was modeled as follows:

$$P_i=\text{TV}(P)\text{gender}=0$$

$$P_i=\text{TV}\hspace{0.17em}(P)\times \theta \text{gender}=1$$

Model validation.

Goodness-of-fit plots were used to evaluate the adequacy of the final model, which included population-predicted concentrations and individual-predicted concentrations (PRED and IPRED, respectively) versus observed concentrations (DV) and conditional weighted residuals (CWRES) versus population prediction concentration (PRED) and time. The nonparametric bootstrap method was conducted to assess the robustness and stability of the final model. One thousand data sets were generated by random sampling with replacement from the original data. The final model parameters were all estimated, and their median and 95% CIs were also calculated when the bootstrap validation was performed. A bootstrap was considered successful if the 95% CIs for each parameter encompassed the initial estimate parameter and met the prespecified convergence rate (95% CIs did not include zero). A prediction corrected-visual predictive check (pc-VPC) was performed to evaluate the final model and parameter estimates. For pc-VPC method evaluation, 1,000 times simulation replicates of the original data set were performed with the final model. The 5th, 50th, and 95th percentiles and the 90% CIs of 5th, 50th, and 95th quantiles were calculated. Then, they were plotted against time after dose and the observed concentrations were compared with the distribution of simulated data.

Monte Carlo simulation.

With the final PPK model, a set of plasma concentrations and clearance rates of patients with PTA%s of 80, 40, or 20, and CLcrs of 50, 30, or 10 ml/min (1,000 individuals in each clinical scenario) was simulated. The interindividual variability and residual variability were both included during the simulation. The dosing schemes were set at 600, 300, or 200 mg twice daily and 600 or 400 mg once daily for 2-h intravenous infusion.

Efficacy threshold.

An AUC_0–24/MIC ratio within 80 to 100 (9, 13) was selected as the pharmacodynamic target of efficacy, and the MIC was set at 2.0 and 4.0 μg/ml. A probability of ≥80% was defined as acceptable and a probability of ≥90% as desirable. The probability of an AUC_0–24/MIC of <80 or >100 was also calculated. The rationale behind these choices derives from the data European Committee on Antimicrobial Susceptibility Testing (EUCAST) (47) and the fact that linezolid retained potent activity against S. aureus (including MRSA), VRE, and streptococci (MIC₉₀, ≤2 μg/ml) (48). For vancomycin-resistant enterococci (up to 30.0%), linezolid inhibited >99% of strains at ≤2 μg/ml. Considering the higher MICs of a few strains (for instance, sometimes the MIC reached 4 μg/ml for Enterococcus faecalis and Enterococcus faecium), we explored the AUC_0-24/MIC ratio at an MIC of 4 μg/ml.

Toxicity threshold.

Some articles associated a higher frequency of toxicity with higher linezolid concentrations. The good linear relationship between trough concentration and estimated AUC_0–24 suggested that C_min,ss might represent a useful predictor of total body exposure of linezolid in clinical practice (14). A toxicodynamic model had shown that C_min,ss of >8.06 μg/ml may result in thrombocytopenia by inhibiting platelet synthesis by 50% (49). Matsumoto et al. (50). set the C_min,ss threshold as 8 μg/ml to minimize linezolid-induced thrombocytopenia. Recently, Pea et al. (34) conducted a cohort study of 1,049 patients for whom linezolid was monitored and concluded that maintaining C_min,ss within a range of 2 to 7 μg/ml might be a valuable tool to optimize drug exposure and to reduce toxicity. Therefore, we recommend maintaining linezolid C_min,ss within a desirable range of 2 to 8 μg/ml to ensure optimal drug exposure and reduce related adverse events, and the threshold of potential overexposure was defined as C_min,ss of >8 μg/ml. The distribution of C_min,ss (<2, 2 to 8, and >8 μg/ml) of different dosage regimens was evaluated.