Evaluation of Linezolid Pharmacokinetics in Critically Ill Obese Patients with Severe Skin and Soft Tissue Infections

Linezolid standard dosing is fixed at 600 mg every 12 h (q12h) for adults. Literature suggests critically ill, obese patients require higher doses.

ABSTRACT

Linezolid standard dosing is fixed at 600 mg every 12 h (q12h) for adults. Literature suggests critically ill, obese patients require higher doses. The study aim is 2-fold: (i) to describe linezolid pharmacokinetics (PK), and (ii) to evaluate if PK/pharmacodynamic (PD) target attainment is achieved with standard dosing in critically ill, obese patients with severe skin and soft tissue infections (SSTIs). Adult patients with a body mass index (BMI) of ≥30 kg/m² and receiving intravenous (i.v.) linezolid from August 2018 to April 2019 were eligible for consent in this prospective study. Severe SSTIs were defined as necrotizing fasciitis, myonecrosis, or SSTI with sepsis syndrome. Four blood samples were collected at steady state at 1, 3, 5 h postinfusion and as a trough. Target attainment was defined as achieving area under the concentration-time curve from 0 to 24 h to MIC (AUC_0–24h/MIC) of ≥100 h*mg/liter. Monte Carlo simulations were used to determine the probability of target attainment (PTA). Eleven patients were included in the study. The median BMI was 45.7 kg/m², and median total body weight (TBW) was 136.0 kg. Seven patients received standard linezolid doses, and four received 600 mg q8h. A one-compartment model described linezolid PK. Based on AUC_0–24h/MIC targets, for noncirrhotic patients at 140 kg, the PTA with standard linezolid doses was 100%, 98.8%, 34.1%, and 0% for MICs of 0.5, 1, 2, and 4 mg/liter, respectively. In conclusion, target attainment of ≥90% is not achieved with standard linezolid doses for noncirrhotic patients ≥140 kg with MICs of ≥2 mg/liter. This study adds to accumulating evidence that standard linezolid doses may not be adequate for all patients.

INTRODUCTION

Linezolid is a synthetic oxazolidinone antibiotic with broad activity against Gram-positive bacteria, including methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant Enterococcus (VRE) (1, 2). Linezolid is moderately lipophilic with penetration into soft tissues and has toxin-suppressive properties, making it a favorable option for treating skin and soft tissue infections (SSTIs).(1, 3) For adults, standard linezolid doses are fixed at 600 mg intravenously (i.v.) or orally (p.o.) given every 12 h (q12h) (4). Linezolid pharmacokinetics (PK) have been well characterized in noncritically ill, nonobese patients with clinical efficacy correlating with PK/pharmacodynamic (PD) targets of 24-h area under the concentration-time curve from 0 to 24 h to MIC (AUC_0–24h/MIC) of 80 to100 h*mg/liter or time above the MIC (%T > MIC) of ≥85% (1).

Obesity and critical illness are two patient characteristics known to alter PK of various medications (5–9). Both can increase medication volume of distribution (V) and clearance (CL) depending on drug-specific properties. This is of particular concern for antimicrobials, where subtherapeutic systemic exposure can lead to poor clinical outcomes, bacterial resistance, and increased mortality (5–9). Antibiotic PK in obese patients is of further concern in severe SSTIs, where antibiotic delivery to the site of infection is reduced due to poor peripheral blood flow (9).

Although conflicting, current literature suggests critically ill, obese patients may require higher linezolid doses to achieve PK/PD target attainment (10, 11); body mass index (BMI) in these studies ranges from 35 to 86 kg/m², with most effects of linezolid serum levels being observed in patients with total body weight (TBW) of >120 kg (10, 11). Alternately, other PK studies suggest that linezolid dosing is not dependent on body size descriptors, and standard doses achieve PK/PD targets for patients with TBW up to 150 kg (12, 13).

The University of Maryland Medical Center (UMMC) is equipped with a level I shock trauma center and cares for a large number of patients with severe SSTIs. The current institutional guidelines recommend use of linezolid over vancomycin if patients have risk factors for acute kidney injury, including obesity. Based on emerging linezolid PK data and uncertainty for appropriate dosing in obesity, institutional guidelines recommend 600 mg q8h for patients with TBW of >150 kg. The aim of this study is 2-fold: (i) to describe PK of linezolid in critically ill obese patients using population PK modeling, and (ii) to evaluate if PK/PD target attainment is achieved with standard linezolid dosing in this patient population.

RESULTS

Patient population.

A total of 20 patients were identified to meet criteria during the study period. Informed consent and linezolid levels were obtained from 11 patients. The median TBW was 141.3 kg, which ranged from 99.9 to 188.0 kg (Tables 1 and 2). The median BMI was 45.7 kg/m², which ranged from 34.6 to 72.6 kg/m². The median presenting sequential organ failure assessment (SOFA) score was 7, which ranged from 0 to 17, and the median Charlson comorbidity index (CCI) score was 4. The majority of patients met study criteria with the diagnosis of necrotizing fasciitis (81.8%). Streptococcus species other than Streptococcus pyogenes were the most common organisms identified from cultures (45.4%) (Table 1). Two patients were undergoing continuous veno-venous hemofiltration (CVVH) at the time of linezolid level draw (Table 1). One patient had liver cirrhosis. Four patients received a higher than standard linezolid dosing regimen at linezolid 600 mg i.v. q8h (Table 2). In 2018, the majority of the Staphylococcus aureus, Enterococcus faecalis, and Enterococcus faecium isolates cultured at UMMC, where an MIC was determined, had an MIC of 2 mg/liter. The institutional MIC distribution is summarized in Fig. 1.

TABLE 1.

Patient demographics^a

Patient demographic	Value
Total no.	11
Age (mean [±SD] [yrs])	59.6 (13.0)
Male (no. [%])	6 (54.5)
Caucasian (no. [%])	6 (54.5)
TBW (median [range] [kg])	141.3 (99.9–188.0)
BMI (median [range])	45.7 (34.6–72.7)
CCI score (median [range])	4 (2–6)
Presenting SOFA score (median [range])	7 (0–17)
Presenting LRINEC score (median [range])	10 (6–13)
Presenting from OSH	9 (81.8)
No. of serotonergic agents (median [range])	1 (0–2)
CVVH (no. [%])	2 (18.2)
Laboratory results
Baseline platelets (median [range] [109/L])	172 (75–300)
Baseline hemoglobin (median [range] [g/dl])	10.3 (6.4–13.1)
SCr at time of blood draw (median [range] [mg/liter])	1.65 (0.53–3.58)
AST at time of blood draw (median [range] [U/liter])	46 (19–239)
ALT at time of blood draw (median [range] [U/liter])	37 (10–74)
TBili at time of blood draw (median [range] [mg/liter])	0.7 (0.2–13.3)
SSTI characteristics (no. [%])
Severe cellulitis	1 (9.1)
Necrotizing fasciitis	9 (81.8)
Fournier’s gangrene	4 (44.4)
Necrotizing wound infection	1 (9.1)
Positive blood culture	1 (9.1)
Streptococcus spp.b	1 (9.1)
Positive tissue culture	9 (81.8)
CoNS	1 (9.1)
E. faecalis	1 (9.1)
MSSA	1 (9.1)
Streptococcus spp.c	5 (45.4)
MICd
E. faecalis	2
MSSA	2

^aTBW, total body weight; BMI, body mass index; CCI, Charlson comorbidity index; SOFA, sequential organ failure assessment score; LRINEC, laboratory risk indicator for necrotizing fasciitis; OSH, outside hospital; CVVH, continuous veno-venous hemofiltration; SCr, serum creatinine; AST, aspartate aminotransferase; ALT, alanine aminotransferase; TBili, total bilirubin; SSTI, skin and soft tissue infection; CoNS, coagulase-negative Staphylococcus aureus; MSSA, methicillin-sensitive Staphylococcus aureus.

^bGroup G Streptococcus.

^cGroup B Streptococcus (n = 3), group C Streptococcus (n = 1), and group G Streptococcus (n = 1).

^dOrganism-specific MIC for E. faecalis (n = 1) and MSSA (n = 1) grown from patient-specific cultures.

TABLE 2.

Patient-specific pharmacokinetic parameters^a

Patient no.	TBW (kg)	BMI (kg/m2)	Linezolid dosing regimen	NCA CL (liter/h)	NCA AUC0–24h (mg*h/liter)
1	136.0	38.5	600 mg q12h	3.5	345.8
2	132.0	41.8	600 mg q12h	8.7	137.9
3	113.0	50.2	600 mg q12h	5.1	233.6
4	113.6	35.9	600 mg q12h	8.5	142.0
5	159.0	56.6	600 mg q12h	8.3	144.0
6	99.9	34.6	600 mg q12h	3.7	321.9
7	188.0	64.9	600 mg q8h	9.4	191.6
8	131.7	42.9	600 mg q12h	6.9	175.0
9	167.8	72.2	600 mg q8h	12.6	142.6
10	166.0	45.7	600 mg q8h	14.3	126.3
11	147.0	58.9	600 mg q8h	5.5	328.3
Mean	141.2	49.3		7.9	208.1
SD	27.1	12.4		3.4	75.8
Median	136.0	45.7		8.3	175.0

^aTBW, total body weight; BMI, body mass index; NCA, noncompartmental analysis; CL, clearance; AUC, area under the concentration-time curve from 0 to 24 h; PK, pharmacokinetics.

FIG 1.

2018 University of Maryland Medical Center linezolid MIC distribution.

Pharmacokinetic analysis.

Using 44 linezolid concentrations obtained from 11 patients, a one-compartment model with linear elimination best described the linezolid PK. A proportional error model best explained the residual variability. The linezolid population PK model is summarized in the following equations:

$$\mathrm{CL}=\left[\mathrm{nonrenal}\hspace{0.17em}\mathrm{CL}\hspace{0.17em}\times \hspace{0.17em}{\left(0.066\right)}^{\mathrm{cirrhosis}}\hspace{0.17em}+\hspace{0.17em}\mathrm{renal}\hspace{0.17em}\mathrm{CL}\right]\times {\left(\frac{\mathrm{ABW}}{140}\right)}^{1.12}$$

$$V=64.3\hspace{0.17em}\times {\left(\frac{\mathrm{ABW}}{140}\right)}^{1.67}$$

where CL is the typical linezolid clearance, which was parameterized as renal and nonrenal clearance; V is the typical volume of distribution of linezolid; and TBW and cirrhosis were covariates found to significantly affect linezolid PK. Including these covariates in the model accounted for 23% of the between-subject variability on CL and 14% of the between-subject variability on V. The inclusion of cirrhotic status as a covariate on CL resulted in an objective function value (OFV) drop from 199.06 for the base model to 190.19. The inclusion of TBW on V resulted in a further drop in OFV to 183.82. The final covariate model, upon adding TBW on CL, had an OFV of 178.54. The parameter estimates and their relative standard errors for the final model are provided in Table 3. The diagnostic plots of observed concentrations versus predicted concentrations, conditional weighted residuals versus time, and conditional weighted residuals versus population-predicted concentrations and individual observed and predicted concentrations versus time indicated that the model was able to describe the data adequately (Fig. 2 and 3). The visual process check (VPC) and quantitative predictive check (QPC) plots indicate adequate model performance and no bias. For the VPC, the 2.5th percentile, median, and 97.5th percentile predicted concentrations overlap the 2.5th percentile, median, and 97.5th percentile observed concentrations, respectively. For the QPC, though the simulated median AUC is slightly lower than the observed AUC (−7.2% deviation), the observed median AUC is within the 95% confidence interval of the simulated median AUCs, suggesting that the model is predicting the median AUC adequately (Fig. 4). The observed 2.5th and 97.5th AUC percentiles are adequately captured as well (10% deviation and −4.5% deviation, respectively). The Monte Carlo simulations were only performed in noncirrhotic patients by fixing the cirrhosis covariate on clearance to zero.

TABLE 3.

Final model parameter estimates^a

Population PK parameters	Final model estimates	Bootstrap estimates	Bootstrap RSE (%)
Nonrenal CL (liter/h)	5.1	5.1	11.8
Renal CL (liter/h)	1.6	1.7	10.7
Total CL (liter/h)	6.7	6.8
V (liter)	64.3	65	9.6
Between-subject variability CL (%)	25	22	42
Between-subject variability V (%)	21	15	72
Proportional error	0.14	0.14	20
Effect of TBW on V	1.67	1.57	45.4
Effect of TBW on CL	1.12	1.11	53.4
Effect of cirrhosis on nonrenal CL	0.066	0.052	81.6

^aCL, clearance; V, volume of distribution; TBW, total body weight; RSE, relative standard error.

FIG 2.

Goodness-of-fit plots. (A) Observed concentrations versus individual predicted concentrations. (B) Observed concentrations versus population predicted concentrations. (C) Conditional weighted residuals versus time. (D) Conditional weighted residuals versus population predicted concentrations.

FIG 3.

Individual goodness-of-fit plots. White dots are the observed concentrations, and black lines are individual predicted concentrations. The numbers at the top of each panel refer to the individual patient number.

FIG 4.

(A) Visual predictive check with 1,000 replicates. The solid black line is the median predicted concentration curve. The dashed black lines are the 2.5th and 97.5th percentile predicted concentration curves. The solid red line is the observed concentration curve. The shaded region is the observed 95% confidence interval. (B) Quantitative predictive check plot using AUC. The dashed red line represents the observed 2.5th, 50th, or 97.5th percentile AUC in the left, middle, and right-side plots, respectively. The solid black line represents the simulated 2.5th, 50th, or 97.5th percentile AUC in the left, middle, and right-side plots, respectively. Lastly, the dashed black lines represent the 95% confidence interval of the predicted AUCs in each percentile group.

Pharmacodynamic profiling.

The Monte Carlo simulations were only performed in noncirrhotic patients by fixing the cirrhosis covariate on clearance to zero. Based on Monte Carlo simulations following the defined PK/PD index of AUC_0–24h/MIC for efficacy and minimum concentration of drug in serum (C_min) of <8.06 mg/liter for safety, at the median TBW of 140 kg, for patients who received standard linezolid doses, the PTAs were 100%, 98.8%, 34.1%, and 0% for organisms with MICs of 0.5, 1, 2, and 4 mg/liter, respectively (Fig. 5). Patients who received linezolid 600 mg q8h had higher PTAs of 100%, 100%, 86.8%, and 6.1% for organisms with those respective MICs. For organisms with MICs of 1, a PTA of ≥90% was achieved for linezolid doses of 600 mg q12h at TBW 100 to 170 kg, 600 mg q8h at TBW 100 kg to 190 kg, and 600 mg q6h at TBW 140 kg to 190 kg (Fig. 6). For organisms with MICs of 2, PTA of ≥90% was achieved for linezolid 600 mg q8h for TBW of 100 kg and 600 mg q6h at TBW 140 kg to 175 kg. Standard linezolid doses did not achieve PTAs of ≥90% for TBW tested for organisms at MICs of 2.

FIG 5.

Percent target attainment for noncirrhotic patients based on TBW and MICs for various linezolid dosing regimens. Target attainment is defined based on AUC/MIC_0–24h ≥ 100 h*mg/liter; C_min < 8.06 mg/liter.

FIG 6.

Percent target attainment based on AUC/MIC_0–24h of ≥100 h*mg/liter and C_min of <8.06 mg/liter for noncirrhotic, obese patients based on TBW for various linezolid dosing regimens.

As a secondary measure, when using %T > MIC > 85% as the efficacy target, the PTAs for the standard linezolid dose, in patients with TBW of 140 kg, were 100%, 99.8%, 92.1%, and 16.9% for organisms with MICs of 0.5, 1, 2, and 4 mg/liter, respectively (Fig. 7). Linezolid doses at 600 mg q8h had higher PTA of 100%, 100%, 99.8%, and 70.7%, respectively, for organisms with those respective MICs.

FIG 7.

Percent target attainment for noncirrhotic patients based on TBW and MICs for various linezolid dosing regimens. Target attainment is defined based on %T > MIC ≥ 85% and C_min < 86 mg/liter.

Patient outcomes.

The mean length of linezolid therapy was 4.6 days (standard deviation [SD], ±2.8) (Table 4). Two patients (patients 1 and 8) developed thrombocytopenia, and no patients developed anemia during the study period. The mean length of hospital stay was 12.2 days (SD, ±14.4). Nine patients met definitions for complete resolution. Two patients (patients 9 and 11) met definitions for clinical failure. The majority of patients were discharged to a facility at the end of their hospital stay (72.7%).

TABLE 4.

Patient outcomes

Patient outcome	Value
Total no.	11
Length of linezolid therapy (mean [±SD] [days])	4.6 (2.8)
Thrombocytopenia(no. [%])	2 (18.2)
Anemia (no. [%])	0 (0)
ICU length of stay (mean [±SD] [days])a	17.9 (12.2)
Hospital length of stay (mean [±SD] [days])	12.2 (14.4)
Complete clinical cure (no. [%])	9 (81.8)
Patient disposition (no. [%])
Death	2 (18.2)
Facility	8 (72.7)
Home	1 (9.1)

^aICU, intensive care unit.

DISCUSSION

Linezolid’s activity against Gram-positive bacteria, lipophilic characteristics, and toxin-suppressive properties make it an important treatment option for severe SSTIs. However, literature is controversial as to whether standard linezolid doses of 600 mg q12h are adequate for critically ill, obese patients. This study aimed to describe linezolid PK and to assess if standard linezolid doses achieve PK/PD targets for this patient population. In the present study, Monte Carlo simulations based on our population PK model suggested that standard linezolid doses did not achieve target attainment of ≥90% for noncirrhotic patients with TBW of ≥140 kg with MICs ≥2 mg/liter based on AUC_0–24h/MIC efficacy targets and C_min of <8.06 mg/liter safety targets (Fig. 5). Increased linezolid doses of 600 mg q8h achieved AUC/MIC efficacy and safety targets for patients with TBW of 100 kg to 190 kg and MICs ≤ 1 mg/liter. For MICs of 2 mg/liter, the MIC observed most often at UMMC in 2018, increased linezolid doses of 600 mg q6h were required to achieve AUC/MIC efficacy and safety targets for TBW ranging from 140 to 175 kg (Fig. 5).

A one-compartment model with linear elimination best described the PK of linezolid in this study population, with TBW and cirrhosis being covariates observed to significantly affect linezolid PK. In the literature, linezolid PK has often been described as exhibiting two-compartment kinetics (10, 13, 14), although one-compartment models of linezolid have been previously described (15–17). The earliest concentration obtained in the present study was 1 h postinfusion. Therefore, the distribution phase of linezolid was not adequately captured to use a two-compartment model. Literature that previously described linezolid PK as exhibiting two-compartment kinetics had concentrations taken within 1 h postinfusion (10, 13). The mean estimated total CL was 6.7 liters/h, which is similar to the mean total CL of 7.33 liters/h reported by Cojutti et al. and 7.61 liters/h reported by Bhalodi et al. (12, 13). Approximately 65% of linezolid clearance is nonrenal. From our analysis, the estimated nonrenal CL was 76%. Cirrhosis was found to have a significant effect on nonrenal CL, which was previously found to be a significant covariate in Japanese patients by Sasaki et al. (16). The extent of the cirrhotic effect on clearance was found to be higher in our model with a 93.2% reduction in nonrenal CL. This is a 71% reduction of clearance compared to 52.8% described by Sasaki et al. (16). This difference could be attributed to there being only one cirrhotic patient in the present study. The mean estimated V was 64.3 liters in this study, which is similar to 62.2 liters reported by Bhalodi et al. (12). TBW was found to be a covariate of both V and CL. The increase observed in V with increasing TBW is justified by the fact that linezolid is a lipophilic drug, and obese patients have an increased amount of adipose tissue. The increase observed in CL with increasing TBW may be attributed to the predominantly nonenzymatic oxidative metabolism of linezolid mediated by reactive oxygen species, which has been shown to increase in obesity (18, 19). For these reasons, we estimated the exponents of weight on CL and V.

Studies that evaluate linezolid PK in critically ill, obese patients are sparse and conflicting as to whether dose adjustments are required based on body size descriptors. A linezolid PK study, including obese, critically ill patients, evaluated 15 patients (BMI > 35 kg/m²) with MRSA pneumonia. This study found that TBW and age significantly affect linezolid PK, specifically finding that linezolid standard doses for TBW extremes in the study of 85 kg and 160 kg for an MIC of 4 mg/liter had poor target attainment of 27.3% and 19.2%, respectively (10). Findings from our study also show poor target attainment for linezolid standard doses for an MIC of 4 mg/liter for patients with TBW of 100 kg and 190 kg with PTA of 56.6% and 0.1%, respectively. Another linezolid PK study evaluated 9 obese, critically ill patients (>120 kg) receiving i.v. linezolid for undefined infections. These patients were morbidly obese, with a median TBW of 174 kg, ranging from 134 to 195 kg. Monte Carlo simulations found that linezolid 600 mg i.v. q8h provided adequate exposure for pathogens with MICs ≤ 1 mg/liter; however, larger doses or alternative therapies were suggested to be needed for pathogens with MICs of >1, which is similar to our findings (11). Conversely, Cojutti et al. found no effect of body size descriptors on linezolid PK in overweight patients (n = 112) (13). Only one-third of patients in this study were septic, and the degree of critical illness was not described. In addition, the 13 patients who met class III obesity (≥40 kg/m²) had a lower median TBW (interquartile range [IQR]) of 123.0 kg (119.0 to 130.0 kg) compared to the present study of 153.0 kg (131.9 to 166.4 kg). Specifically, the study found that CrCl based on the Chronic Kidney Diseases Epidemiology formula (CKD-EPI) was a significant covariate in the study population and recommended dose reductions. In our study, CrCl was not observed to be a covariate, and there were 2 patients receiving CVVH at the time of linezolid level draw (patients 8 and 9). Observed noncompartmental analysis (NCA) AUC_0–24h for these patients was comparable to non-CVVH patients (Table 1). In addition, the half-lives derived from the individual post hoc parameter estimates for these two patients were 7.85 h and 5.48 h, respectively, which are within the interquartile range of the study (5.38 h to 8.17 h). Case reports (n = 6) describing linezolid PK in obese patients where TBW ranged from 102 to 286 kg and BMI ranged from 35 to 86 kg/m² noted a reduction in serum levels compared to nonobese patients from other PK studies (14, 20–23).

While PK studies provide important insights relative to drug exposures, connecting drug exposure to outcome is critical to optimize use and minimize toxicity in the infected population. Fleming et al. conducted a retrospective review of 35 critically ill patients and found that obese patients (>120 kg) receiving higher doses of linezolid had similar outcomes as cross-matched nonobese patients (<120 kg) receiving standard linezolid dosing. The incidence of thrombocytopenia was similar in both treatment arms when excluding patients with concurrent sepsis (24). In the present study, the observed in-hospital mortality rate was 18.2%, which is lower than mortality rates reported in the literature for necrotizing fasciitis, the most common indication for study inclusion, of around 30% (25). Two patients met the definition for clinical failure. Both patients received 600 mg i.v. q8h and met AUC efficacy targets with an NCA AUC of 142.6 mg*h/liter and 328.3 mg*h/liter. Two patients developed thrombocytopenia during the study period, both of whom were receiving standard linezolid doses. One patient had liver cirrhosis and the highest linezolid exposure observed in an individual study patient (NCA AUC of 345.8 mg*h/liter; C_min of 15.88 mg/liter). The other patient’s linezolid exposure was in line with other patients (NCA AUC of 175.0 mg*h/liter; C_min of 5.81 mg/liter). Recent studies have suggested that therapeutic drug monitoring (TDM) may be a potential solution for evaluating linezolid exposure (26, 27). A prospective study that included 61 nonobese (median TBW of 75 kg) patients receiving linezolid for >10 days obtained weekly linezolid troughs (target, 2 to 8 mg/liter) and platelet counts; 54.1% of patients experienced overexposure, 45.9% always had desired troughs, and no patients experienced underexposure. Incidence of thrombocytopenia was independently associated with baseline platelet count and median linezolid troughs (27). The utility of linezolid TDM needs to be further investigated in this patient population.

There are limitations to the present study. The small number of patients enrolled with a wide range of characteristics, including TBW, renal function, and liver function, is a limitation. However, the total number of samples collected (n = 44) was sufficient to describe a one-compartment model with good precision, and broad characteristics allow for generalizability of the study population. In addition, the study population largely underwent surgical intervention, resulting in blood loss and fluid administration, which could alter V. Although this can affect antibiotic PK, the results of this study are externally valid for this patient population. Authors also attempted to control for this by drawing linezolid levels at steady state. Ideally, more linezolid levels would have been drawn within the dosing interval. Four levels were selected based on resources available and to ensure consistency in timing of level draws among patients given the high acuity study population. Also, the linezolid duration of therapy was low, with the average duration of therapy being 4.6 days. Although the incidence of thrombocytopenia and anemia was low, this could have been due to the shorter duration of therapy. Cirrhosis status was included as a covariate in the population PK model, but this was based on one cirrhotic patient in the study. Therefore, the extent to which liver cirrhosis affects linezolid PK may not be accurately reflected by the current PK model. Nevertheless, the results of this study indicate that the effect of liver cirrhosis on linezolid PK needs to be considered and further explored.

Overall, standard linezolid doses may not achieve PK/PD targets for critically ill, noncirrhotic, obese patients with higher TBW and MICs of ≥1 mg/liter. This study adds to accumulating evidence that a fixed dose may not be adequate for all patients, including critically ill, obese patients. Additional research is needed to evaluate long-term outcomes and safety of increased doses for this patient population.

MATERIALS AND METHODS

Study design and patient population.

This was a prospective pharmacokinetic study conducted at the University of Maryland Medical Center, Baltimore, Maryland, from August 2018 to April 2019. The study protocol was reviewed and approved by the University of Maryland, Baltimore, and the R. Adams Cowley Shock Trauma Center Institutional Review Board. Written informed consent was obtained from all patients or legal representatives enrolled in the study.

Patients were eligible for study inclusion if they met the following criteria: (i) ≥18 years of age, (ii) BMI ≥ 30 kg/m², and (iii) receiving i.v. linezolid for a severe SSTI indication. A severe SSTI was defined as one of the following: necrotizing fasciitis, myonecrosis, or SSTI presenting with sepsis based on quick sequential organ failure assessment (qSOFA) score or the systemic inflammatory response syndrome (SIRs) criteria (28). Patients were excluded from enrollment in the study if they were pregnant.

Drug administration, sample collection, and serum analysis.

All linezolid dosing regimens were at the discretion of the primary care team. All doses were administered intravenously as a 30-minute infusion. Four blood samples were collected within the same dosing interval at 1, 3, 5 h postinfusion and as a trough immediately prior to the next dose. Blood levels were drawn after patients received at least 3 consecutive linezolid doses. Linezolid doses could not have been interrupted by intermittent hemodialysis or surgical intervention with documented blood loss of ≥1,000 ml.

Blood samples were centrifuged for 10 min at 2,500 rpm within 30 min of collection to obtain plasma. Samples were then frozen at −80°C until shipment for analysis. Samples were analyzed at the Center for Anti-Infective Research and Development at Hartford Hospital (Hartford, CT). Concentrations of linezolid (total drug) were quantified by high-pressure liquid chromatography (HPLC) using a validated assay method published by Tobin et al. (29).

Pharmacokinetic analysis.

A population PK analysis was performed to describe the plasma pharmacokinetics of linezolid and to explain the variability in the PK parameters by patient-specific covariates (30). Noncompartmental analysis was performed to use as initial estimates for the population PK analysis. Both one-compartment and two-compartment models were considered to describe the disposition of linezolid. Between-subject variability was modeled on the PK parameters with the assumption of a log-normal distribution as shown in the following equation:

$${\mathrm{\theta }}_i={\mathrm{\theta }}_{\mathit{pop}}\hspace{0.17em}·\hspace{0.17em}{e}^{{\mathrm{\eta }}_i}$$

where θ_i is the individual PK parameter for individual i, θ_pop is the population mean of the PK parameter, and η_i is the between-subject random effect on individual i. η_i is assumed to be normally distributed with a mean of 0 and a variance of ω². Proportional error and additive/proportional combined error models were evaluated as the residual error model. The effect of covariates on the PK parameters was assessed by examining the residual between-subject variability plots. Covariates considered included patient age, BMI, TBW, serum creatinine (SCr), alanine aminotransferase (ALT), aspartate aminotransferase (AST), and total bilirubin (TBili). Covariates were then included in the model if the residual plot revealed a significant relationship and satisfied standard statistical model selection criterion (3.84-point decrease in the objective function value as assessed by log-likelihood ratio test [P < 0.05; degrees of freedom, 1]). The inclusion of a covariate in the population PK model was based on valid physiological reasoning and if the covariate was able to explain the variability of a PK parameter. The adequacy of model predictions was examined by standard goodness-of-fit plots and quantitative predictive check (QPC) (31). Visual process check (VPC) was performed with 1,000 replicates. AUC_0–24h was used as the PK metric of interest to perform QPC. The median observed AUC_0–24h was calculated from noncompartmental analysis (NCA). We then simulated 1,000 replicates with the same dosing and covariates as the original data using parameter estimates from the final model. The AUC_0–24h of the 1,000 replicates was derived using noncompartmental analysis. The median observed AUC_0–24h was then overlaid on the distribution of median simulated AUC_0–24h from each of the 1,000 simulated replicates in a plot to perform QPC. The 2.5th percentile observed AUC_0–24h and 97.5th percentile observed AUC_0–24h were also overlaid on the distribution of 2.5th percentile AUC_0–24h and 97.5th percentile AUC_0–24h from each of the 1,000 simulated replicates, respectively. The population PK analysis was performed using the first-order conditional estimation (FOCE) method in pumas v0.10.0. Pumas is a new pharmacometrics package in the Julia programming language which could be used for nonlinear mixed-effects modeling and simulation.

Individual PK parameters were estimated from post hoc empirical Bayesian estimates. As a primary outcome, target attainment was defined as achieving AUC_0–24h/MIC of ≥100 with C_min of <8.06 mg/liter for estimation of toxicity (32). As a secondary outcome, target attainment was defined as achieving time above the MIC of ≥85% (1).

Pharmacodynamic profiling.

Monte Carlo simulations were performed to assess the probability of target attainment (PTA) for different linezolid doses using the developed population PK model. Patients with 3 possible TBWs were simulated based on the lowest observed, median, and highest observed TBW from the data. Three linezolid dosing regimens (600 mg q12h, 600 mg q8h, or 600 mg q6h) were evaluated using these simulations. Dual parameterization was employed for patients having to meet AUC_0–24h/MIC of ≥100 with C_min of <8.06 mg/liter. Each scenario listed above was simulated 4,000 times to evaluate the PTA. A range of MICs from 0.5 to 4 mg/liter were utilized based on breakpoint susceptibilities of staphylococci and enterococci.

Patient outcomes.

The incidence of thrombocytopenia and anemia was evaluated. Thrombocytopenia and anemia were defined as platelet count and hemoglobin, respectively, being <75% of the lower limit of normal for patients with normal values at baseline. For patients with abnormal values at baseline, thrombocytopenia and anemia were defined as platelet count and hemoglobin, respectively, being <50% of baseline (33). Clinical response was defined as clinical resolution, indeterminate, and failure within 30 days of linezolid start. Clinical resolution was defined as complete signs and symptoms of SSTI resolution or surgical wound closure without prior failure. Clinical failure was defined as the need for addition or change in antibiotic treatment due to SSTI discontinuation due to lack of efficacy. Indeterminate was defined as either persistence of symptoms without objective evidence of infection or any extenuating circumstances precluding a classification of clinical resolution or failure (34). Evaluation clinical response was adjudicated by two study team members, an infectious diseases pharmacist, and an infectious diseases physician.