Multicenter Population Pharmacokinetic Study of Unbound Ceftriaxone in Critically Ill Patients

ABSTRACT

The objective of this study was to describe the total and unbound population pharmacokinetics of ceftriaxone in critically ill adult patients and to define optimized dosing regimens. Total and unbound ceftriaxone concentrations were obtained from two pharmacokinetic studies and from a therapeutic drug monitoring (TDM) program at a tertiary hospital intensive care unit. Population pharmacokinetic analysis and Monte Carlo simulations were used to assess the probability of achieving a free trough concentration/MIC ratio of ≥1 using Pmetrics for R. A total of 474 samples (267 total and 207 unbound) were available from 36 patients. A two-compartment model describing ceftriaxone-albumin binding with both nonrenal and renal elimination incorporating creatinine clearance to explain the between-patient variability best described the data. An albumin concentration of ≤20 g/L decreased the probability of target attainment (PTA) by up to 20% across different dosing regimens and simulated creatinine clearances. A ceftriaxone dose of 1 g twice daily is likely therapeutic in patients with creatinine clearance of <100 mL/min infected with susceptible isolates (PTA, ~90%). Higher doses administered as a continuous infusion (4 g/day) are needed in patients with augmented renal clearance (creatinine clearance, >130 mL/min) who are infected by pathogens with a MIC of ≥0.5 mg/L. The ceftriaxone dose should be based on the patient’s renal function and albumin concentration, as well as the isolate MIC. Hypoalbuminemia decreases the PTA in patients receiving intermittent dosing by up to 20%.

INTRODUCTION

Antibiotic dose optimization in the critically ill patient is an important component of treatment that can influence patient outcomes (1, 2). For cephalosporins, concentrations that exceed the pathogen MIC throughout the dosing interval (i.e., minimum concentration at the end of the dosing interval [C_min/MIC of >1]) are associated with 3-fold and 2-fold increases in clinical cure rates and bacterial eradication rates, respectively, compared to patients with lower exposures (3–5). A separate study has shown that achieving a β-lactam C_min/MIC of ≥4.95 is associated with lower in-hospital mortality in patients with Gram-negative bacteremia (6). Such high blood exposures are consistent with the recommendations of some authors for use in patients with infections at sites where antibiotic penetration is low and/or variable, such as the epithelial lining fluid in pneumonia (7–9).

Achieving these pharmacokinetic/pharmacodynamic (PK/PD) targets can be particularly challenging in critically ill patients with acute physiological alterations that can manifest with hypoalbuminemia and augmented renal clearance (ARC), leading to altered antibiotic pharmacokinetics (10). Hypoalbuminemia can increase the unbound fraction of highly protein-bound drugs such as ceftriaxone, an antibiotic that is reported to have a variable unbound fraction of 5 to 17% (10–13). Although a higher unbound fraction increases the pharmacologically active antibiotic available for antibacterial activity, antibiotic elimination may also be enhanced (12–14). A serum albumin concentration of <25 g/L has previously been associated with a 25% increased ceftriaxone volume of distribution (V) compared with that of patients with a normal albumin concentration (40 g/L), affecting ≥40% of all patients admitted to the intensive care unit (ICU) (14, 15). An increased unbound fraction may also further increase ceftriaxone clearance in patients with ARC (creatinine clearance [CL_CR] of ≥130 mL/min/1.73 m²), which may occur in 65.1% of patients (10, 11, 16).

A previous pharmacokinetic study conducted in critically ill patients admitted to the ICU showed that a 1-g twice-daily dose would likely achieve therapeutic exposures in most patients (17). However, this study did not investigate the impact of albumin concentration or ARC on selecting the optimal ceftriaxone dose for individual patients. The results of this study are consistent with other previous studies in critically ill patients admitted to the ICU, whereby 1-g or 2-g once-daily dosing regimens are unlikely to achieve an unbound C_min/MIC of >1 (5, 18–20).

The current study aims to describe the population pharmacokinetics of total and unbound ceftriaxone, incorporating any effects of albumin and creatinine clearance. A nonparametric pharmacokinetic modeling approach was used given the potential for improved parameter outlier detection, which may be common in critically ill patients given the highly variable physiological processes altering antibiotic pharmacokinetics (21). We sought to use the model to perform dosing simulations to determine optimized ceftriaxone doses for various patient scenarios that reliably achieve unbound trough concentrations exceeding a C_min/MIC of >1 in adult patients, consistent with current therapeutic drug monitoring guidelines (2).

RESULTS

Study population.

A total of 474 (207 unbound paired samples and an additional 60 unpaired total concentration samples) ceftriaxone concentrations were available for 36 patients (Table 1). These patients had a mean age of 52 years, with hypoalbuminemia (mean, 23.9 g/L) and a normal creatinine clearance (median, 117.9 mL/min). Patients presented primarily with respiratory tract infections and received a 1-g twice-daily dose.

TABLE 1.

Patient demographic details

Parametera	Result for parameter (n = 36)
Age, yr (IQR)	55.5 (39.0–61.3)
Female, n (%)	18 (48.7)
Wt, kg (IQR)	65 (59.5–80.0)
Albumin, g/L (IQR)	23 (20.0–28.3)
CLCR, mL/min (IQR)	104 (85–157.88)
Infection source, n (%)
Intra-abdominal	6 (16.7)
Respiratory	18 (50.0)
Urinary tract	2 (5.6)
Endovascular	2 (5.6)
Central nervous system	6 (16.7)
Prophylaxis extraventricular drain	2 (5.6)
Ceftriaxone dose administered
1 g once daily	1
2 g once daily	12
1 g twice daily	17
2 g twice daily	6

^aIQR, interquartile range.

Pharmacokinetic model.

The total and unbound ceftriaxone concentrations were adequately described using a two-compartment model with linear elimination and a complex protein binding model. Incorporating a complex ceftriaxone binding model improved the fit compared with a simple binding model when considering both a one-compartment model (Akaike information criterion [ΔAIC], −205; Bayesian information criterion [ΔBIC], −205) and two-compartment model (ΔAIC, −609; ΔBIC, −592) with linear elimination. A two-compartment complex binding model improved the model fit compared with a one-compartment complex binding model (ΔAIC, −378; ΔBIC, −369). The two-compartment model incorporating renal (TVR) and nonrenal (TVNR) ceftriaxone elimination improved the fit compared with renal elimination alone (ΔAIC, −55; ΔBIC, −51). Assuming that ceftriaxone may experience competitive binding with other molecules (i.e., the number of ceftriaxone molecules per albumin molecule may not equal 1), incorporating a variable binding number further improved the model fit (ΔAIC, −34; ΔBIC, −30). Incorporating creatinine clearance further improved the model fit (ΔAIC −42; ΔBIC, −41). The AIC and BIC values are included in Table 2. The incorporation of other covariates or power models did not improve the model fit. Therefore, the final clearance model is described by equation 1:

$$\text{Ceftriaxone clearance\hspace{0.17em}=\hspace{0.17em}TVNR}\hspace{0.17em}+\left(\text{TVR\hspace{0.17em}×}\frac{{\text{CL}}_{\text{CR}}}{\text{118}}\right)$$ (1)

TABLE 2.

Ceftriaxone pharmacokinetic model comparative log-likelihood in (2*LL), AIC, and BIC

Model description	2*LL	AIC	BIC
1 compartment
Simple binding	3,958	3,966	3,982
Complex binding	3,753	3,765	3,790
2 compartments
Simple binding	3,958	3,970	3,995
Complex binding	3,341	3,357	3,910
Complex binding incorporating creatinine clearance	3,313	3,330	3,363
Complex binding with 2 possible clearance routes	3,344	3,363	3,400
Complex binding incorporating creatine clearance and nonrenal clearance	3,278	3,296	3,333

TVNR and TVR are the mean nonrenal and renal ceftriaxone clearance population values, respectively, and CL_CR represents the creatinine clearance (in milliliters per minute). The final model differential equations are provided below, where equations 2, 3, and 4 describe the plasma unbound ceftriaxone concentration, the bound plasma ceftriaxone concentration, and the ceftriaxone concentration in the peripheral compartment, respectively:

$$\frac{dX1}{dt}=\text{dose}-\frac{\text{CL}}{V}\times X1-\frac{K_{\text{on}}}{V_{}}\times \left(B_{\text{max}}-X2\right)\times X1+K_{\text{off}}\times X2-K_{\text{cp}}\times X1+K_{\text{pc}}\times X3$$ (2)

$$\frac{dX2}{dt}=\frac{K_{\text{on}}}{V_{}}\times \left(B_{\text{max}}-X2\right)\times X1-K_{\text{off}}\times X2$$ (3)

$$\frac{dX3}{dt}=K_{\text{cp}}\times X1-K_{\text{pc}}\times X3$$ (4)

K_cp is the first-order constant for unbound ceftriaxone distribution from the central to peripheral compartment, K_pc is the first-order constant for unbound ceftriaxone distribution from the peripheral to central compartment. Other values have been defined in equations 5 to 7.

The final model was adequately fit as described by the visual predictive check for the unbound (Fig. 1) and total (Fig. 2) ceftriaxone concentrations. The unbound model population and individual predicted R² values were 0.7 and 0.96, respectively (Fig. 3A and B, respectively). The total concentration model population and individual predicted R² values were 0.83 and 0.97, respectively (Fig. 3C and D, respectively). The final model parameter estimates are described in Table 3. Model weighted residual error versus predicted and versus time are included in Fig. 4. Model parameter support points are detailed in Table S1 at https://doi.org/10.6084/m9.figshare.19614561.

FIG 1.

Observed unbound ceftriaxone concentration-time data and visual predictive check of the final model. Dots represent observed patient unbound ceftriaxone concentrations.

FIG 2.

Observed total concentration ceftriaxone concentration-time data and visual predictive check of the final model. Dots represent observed patient total ceftriaxone concentrations.

FIG 3.

Observed versus predicted goodness-of-fit plots for unbound and total ceftriaxone concentrations (open circles). (A) Population predicted unbound ceftriaxone concentration; (B) individual predicted unbound ceftriaxone concentration; (C) population predicted total ceftriaxone concentration; (D) individual predicted total ceftriaxone concentration. The ceftriaxone concentrations shown are milligrams per liter.

TABLE 3.

Ceftriaxone pharmacokinetic model parameter estimates^a

Parameter	Mean (SD)	Median	95% CI	% RSE	CV%	Shrinkage (%)
TVNR (L/h)	7.27 (4.89)	6.86	5.73–8.77	11.21	67.29	7.79
TVR (L/h)	4.12 (5.47)	1.98	2.25–5.67	22.13	132.75	4.30
V (L)	7.43 (3.25)	7.25	6.45–8.33	7.29	43.66	18.50
Kon (L/mg/h)	1,290.19 (472.19)	1,312.04	1,169–1,412	6.10	36.60	27.56
Koff (h−1)	18,537.17 (4507.28)	18,961.57	17,177–19,916	4.05	24.32	17.95
Kcp (h−1)	4.00 (3.20)	3.04	3.15–4.89	13.33	80.01	24.15
Kpc (h−1)	0.65 (0.53)	0.48	0.49–0.80	13.59	81.23	16.97
N	0.82 (0.43)	0.69	0.69–0.94	8.74	51.71	21.84

^a95% CI, 95% bootstrapped confidence interval; TVNR, nonrenal ceftriaxone clearance; TVR, renal ceftriaxone clearance; V, volume of distribution; K_on, second-order association rate constant; K_off, first-order dissociation rate constant; K_cp, rate transfer constant from the central to peripheral compartment; K_pc, rate transfer constant from the peripheral to central compartment; N, number of ceftriaxone binding sites per albumin molecule; SD, standard deviation; RSE, residual squared error; CV, coefficient of variation.

FIG 4.

Residual diagnostic plots. (A) Weighted residual error versus predicted concentration; (B) weighted residual error versus time; (C) weighted residual error frequency histogram.

Dosing simulations.

The probability of target attainment (PTA) for the various ceftriaxone dosing regimens is presented in Table 4. Across the different dosing regimens and renal functions, lower albumin concentrations were associated with a reduced PTA. For each albumin reduction of 10 g/L, the PTA decreases at a given renal function by up to 20% (Table 4).

TABLE 4.

Probability of target attainment for an unbound ceftriaxone concentration C_min/MIC of >1 over the first 24 h of therapy^a

Dose	CLCR (mL/min)	Albumin (g/L)	Probability of achieving Cmin/MIC of >1
			0.032	0.064	0.125	0.25	0.5	1	2	4	8
1 g q24h	40	20	1	1	1	0.98	0.86	0.61	0.33	0.11	0
	40	30	1	1	1	1	0.91	0.70	0.39	0.12	0
	40	40	1	1	1	1	0.94	0.77	0.42	0.12	0
	100	20	1	1	0.97	0.88	0.67	0.32	0.10	0.03	0
	100	30	1	1	1	0.94	0.79	0.50	0.13	0.03	0
	100	40	1	1	1	0.98	0.86	0.58	0.17	0.03	0
	160	20	0.99	0.97	0.9	0.73	0.36	0.13	0.07	0.02	0
	160	30	1	0.99	0.96	0.84	0.59	0.20	0.08	0.02	0
	160	40	1	1	0.98	0.91	0.72	0.31	0.09	0.02	0
2 g q24h	40	20	1	1	1	1	0.95	0.82	0.53	0.28	0.10
	40	30	1	1	1	1	0.99	0.88	0.65	0.32	0.10
	40	40	1	1	1	1	1	0.92	0.72	0.35	0.10
	100	20	1	1	1	0.95	0.86	0.60	0.25	0.08	0.03
	100	30	1	1	1	1	0.91	0.75	0.37	0.10	0.03
	100	40	1	1	1	1	0.96	0.81	0.50	0.12	0.03
	160	20	1	0.99	0.96	0.88	0.65	0.27	0.11	0.06	0.02
	160	30	1	1	1	0.95	0.81	0.49	0.14	0.06	0.02
	160	40	1	1	1	0.98	0.88	0.65	0.21	0.07	0.02
1 g q12h	40	20	1	1	1	1	0.99	0.94	0.76	0.40	0.07
	40	30	1	1	1	1	0.99	0.97	0.78	0.39	0.05
	40	40	1	1	1	1	1	0.98	0.79	0.35	0.03
	100	20	1	1	1	1	0.97	0.86	0.62	0.15	0.01
	100	30	1	1	1	1	0.99	0.88	0.68	0.17	0.01
	100	40	1	1	1	1	1	0.93	0.71	0.15	0.01
	160	20	1	1	1	0.99	0.92	0.76	0.33	0.07	0.01
	160	30	1	1	1	1	0.96	0.80	0.46	0.07	0.01
	160	40	1	1	1	1	0.97	0.83	0.52	0.07	0.01
2 g q12h	40	20	1	1	1	1	1	0.99	0.90	0.73	0.35
	40	30	1	1	1	1	1	0.99	0.94	0.75	0.36
	40	40	1	1	1	1	1	1	0.97	0.76	0.34
	100	20	1	1	1	1	0.99	0.95	0.82	0.52	0.13
	100	30	1	1	1	1	0.99	0.98	0.86	0.61	0.13
	100	40	1	1	1	1	1	0.99	0.88	0.66	0.13
	160	20	1	1	1	1	0.98	0.91	0.72	0.24	0.07
	160	30	1	1	1	1	0.99	0.93	0.77	0.33	0.06
	160	40	1	1	1	1	1	0.97	0.78	0.33	0.06
1 g q8h	40	20	1	1	1	1	1	0.97	0.85	0.61	0.17
	40	30	1	1	1	1	1	0.98	0.88	0.61	0.12
	40	40	1	1	1	1	1	0.99	0.89	0.55	0.09
	100	20	1	1	1	1	0.99	0.94	0.78	0.41	0.05
	100	30	1	1	1	1	1	0.96	0.80	0.41	0.04
	100	40	1	1	1	1	1	0.97	0.81	0.35	0.04
	160	20	1	1	1	1	0.98	0.88	0.71	0.17	0.03
	160	30	1	1	1	1	0.99	0.92	0.75	0.19	0.03
	160	40	1	1	1	1	1	0.95	0.77	0.19	0.02
2 g q8h	40	20	1	1	1	1	1	0.99	0.96	0.81	0.57
	40	30	1	1	1	1	1	1	0.98	0.85	0.57
	40	40	1	1	1	1	1	1	0.98	0.86	0.54
	100	20	1	1	1	1	1	0.99	0.91	0.45	0.32
	100	30	1	1	1	1	1	0.99	0.94	0.77	0.34
	100	40	1	1	1	1	1	1	0.96	0.79	0.34
	160	20	1	1	1	1	0.99	0.96	0.86	0.63	0.14
	160	30	1	1	1	1	1	0.98	0.89	0.71	0.15
	160	40	1	1	1	1	1	0.99	0.92	0.74	0.16
1-g LD + 2-g CI	40	20	1	1	1	1	1	1	1	0.96	0.72
	40	30	1	1	1	1	1	1	1	0.96	0.66
	40	40	1	1	1	1	1	1	1	0.96	0.52
	100	20	1	1	1	1	1	1	1	0.89	0.52
	100	30	1	1	1	1	1	1	1	0.88	0.49
	100	40	1	1	1	1	1	1	1	0.87	0.39
	160	20	1	1	1	1	1	1	0.98	0.82	0.26
	160	30	1	1	1	1	1	1	0.98	0.82	0.26
	160	40	1	1	1	1	1	1	0.98	0.82	0.23
1-g LD + 4-g CI	40	20	1	1	1	1	1	1	1	1	0.91
	40	30	1	1	1	1	1	1	1	1	0.85
	40	40	1	1	1	1	1	1	1	1	0.73
	100	20	1	1	1	1	1	1	1	0.99	0.82
	100	30	1	1	1	1	1	1	1	0.99	0.78
	100	40	1	1	1	1	1	1	1	0.98	0.69
	160	20	1	1	1	1	1	1	1	0.98	0.78
	160	30	1	1	1	1	1	1	1	0.98	0.74
	160	40	1	1	1	1	1	1	1	0.98	0.63

^aLD, loading dose; CI, continuous infusion; CL_CR, creatinine clearance; q8h, q12h, and q24h, administered every 8, 12, and 24 h, respectively.

A 1-g once-daily dose should only be used if the MIC is low (≤0.125 mg/L) in patients with impaired renal function (creatinine clearance, <40 mL/min). In this population of patients with renal impairment (creatinine clearance, <40 mL/min), doses of 1 g twice-daily or 2 g once-daily are needed to ensure a PTA of >90% in patients infected with susceptible bacteria (MIC, ≤1 mg/L) (Table 4). Toxicity is unlikely (<1%) at this dose (Table 5). Patients with a creatinine clearance of 100 mL/min require a dose 1 g three times daily to ensure a PTA of >90% in susceptible bacteria (MIC, ≤1 mg/L), which is also associated with a probability of toxicity <1% (Table 5). Only high-dose (4 g/day) therapy administered in divided doses or as a continuous infusion has the potential to achieve the target fC_min/MIC >1 for susceptible isolates in >90% of patients with augmented renal clearance (22). Toxicity is unlikely in patients with augmented renal clearance (<1%), despite the high doses used in patients with augmented renal clearance. High-dose therapy (4 g/day) may result in toxicity for approximately 1.4% of patients with a creatinine clearance of <40 mL/min (Table 5).

TABLE 5.

Probability of toxicity within a 7-day treatment course^a

Dose	CLCR (mL/min)	Albumin (g/L)	Probability of achieving total Cmin of >100 mg/L
1 g q24h	20	20	0
	20	30	0
	20	40	0
	40	20	0
	40	30	0
	40	40	0
	100	20	0
	100	30	0
	100	40	0
	160	20	0
	160	30	0
	160	40	0
2 g q24h	20	20	0
	20	30	0
	20	40	0
	40	20	0
	40	30	0
	40	40	0
	100	20	0
	100	30	0
	100	40	0
	160	20	0
	160	30	0
	160	40	0
1 g q12h	20	20	0
	20	30	0
	20	40	0
	40	20	0
	40	30	0
	40	40	0
	100	20	0
	100	30	0
	100	40	0
	160	20	0
	160	30	0
	160	40	0
2 g q12h	20	20	0.01
	20	30	0.01
	20	40	0.01
	40	20	0.01
	40	30	0.01
	40	40	0.01
	100	20	0
	100	30	0
	100	40	0
	160	20	0
	160	30	0
	160	40	0
1 g q8h	20	20	0.01
	20	30	0.01
	20	40	0.01
	40	20	0
	40	30	0
	40	40	0
	100	20	0
	100	30	0
	100	40	0
	160	20	0
	160	30	0
	160	40	0
2 g q8h	20	20	0.06
	20	30	0.06
	20	40	0.06
	40	20	0.01
	40	30	0.01
	40	40	0.01
	100	20	0.01
	100	30	0.01
	100	40	0.01
	160	20	0
	160	30	0
	160	40	0
1-g LD + 2-g CI	20	20	0
	20	30	0
	20	40	0
	40	20	0
	40	30	0
	40	40	0
	100	20	0
	100	30	0
	100	40	0
	160	20	0
	160	30	0
	160	40	0
1-g LD + 4-g CI	20	20	0.03
	20	30	0.03
	20	40	0
	40	20	0.01
	40	30	0.01
	40	40	0.01
	100	20	0.01
	100	30	0.01
	100	40	0.01
	160	20	0
	160	30	0
	160	40	0

^aLD, loading dose; CI, continuous infusion; CL_CR, creatinine clearance; q8h, q12h, and q24h, administered every 8, 12, and 24 h, respectively.

When considering the fractional target attainment (FTA) for specific bacterial species, a 1-g twice-daily ceftriaxone dose is likely adequate in most patients who are infected with isolates other than Staphylococcus aureus (Table 6). Only high-dose continuous infusions of ceftriaxone (4 g/day) are likely to achieve a target exposure against S. aureus with a MIC of ≤4 mg/L for most patients.

TABLE 6.

Fractional target attainment for different ceftriaxone dosing regimens and bacterial species^a

Dose	CLCR (mL/min)	Albumin (g/L)	Fractional target attainment in:
			E. coli	H. influenzae	K. pneumoniae	S. pneumoniae	S. aureus
1 g q24h	40	20	1	0.99	1	0.96	0.20
	40	30	1	1	1	0.97	0.23
	40	40	1	1	1	0.98	0.24
	100	20	1	0.98	0.98	0.92	0.06
	100	30	1	0.99	0.99	0.95	0.08
	100	40	1	1	1	0.96	0.09
	160	20	0.98	0.95	0.94	0.88	0.04
	160	30	0.99	0.98	0.97	0.91	0.05
	160	40	1	0.99	0.98	0.93	0.05
2 g q24 h	40	20	1	1	1	0.98	0.38
	40	30	1	1	1	0.99	0.46
	40	40	1	1	1	0.99	0.50
	100	20	1	0.99	0.99	0.96	0.16
	100	30	1	1	1	0.97	0.21
	100	40	1	1	1	0.98	0.28
	160	20	0.99	0.98	0.97	0.92	0.08
	160	30	1	0.99	0.99	0.95	0.10
	160	40	1	1	1	0.96	0.13
1 g q12h	40	20	1	1	1	0.99	0.54
	40	30	1	1	1	0.99	0.55
	40	40	1	1	1	1	0.52
	100	20	1	1	1	0.99	0.34
	100	30	1	1	1	0.99	0.38
	100	40	1	1	1	0.99	0.38
	160	20	1	1	1	0.97	0.18
	160	30	1	1	1	0.98	0.23
	160	40	1	1	1	0.98	0.26
2 g q12h	40	20	1	1	1	1	0.79
	40	30	1	1	1	1	0.82
	40	40	1	1	1	1	0.84
	100	20	1	1	1	0.99	0.64
	100	30	1	1	1	1	0.70
	100	40	1	1	1	1	0.74
	160	20	1	1	1	0.99	0.44
	160	30	1	1	1	0.99	0.51
	160	40	1	1	1	1	0.57
1 g q8h	40	20	1	1	1	1	0.70
	40	30	1	1	1	1	0.71
	40	40	1	1	1	1	0.68
	100	20	1	1	1	0.99	0.55
	100	30	1	1	1	1	0.56
	100	40	1	1	1	1	0.54
	160	20	1	1	1	0.99	0.39
	160	30	1	1	1	0.99	0.42
	160	40	1	1	1	1	0.43
2 g q8h	40	20	1	1	1	1	0.87
	40	30	1	1	1	1	0.89
	40	40	1	1	1	1	0.90
	100	20	1	1	1	1	0.80
	100	30	1	1	1	1	0.83
	100	40	1	1	1	1	0.85
	160	20	1	1	1	1	0.72
	160	30	1	1	1	1	0.77
	160	40	1	1	1	1	0.80
1-g LD + 2-g CI	40	20	1	1	1	1	0.97
	40	30	1	1	1	1	0.98
	40	40	1	1	1	1	0.97
	100	20	1	1	1	1	0.92
	100	30	1	1	1	1	0.92
	100	40	1	1	1	1	0.91
	160	20	1	1	1	1	0.87
	160	30	1	1	1	1	0.87
	160	40	1	1	1	1	0.87
1-g LD + 4-g CI	40	20	1	1	1	1	1
	40	30	1	1	1	1	1
	40	40	1	1	1	1	1
	100	20	1	1	1	1	0.99
	100	30	1	1	1	1	0.99
	100	40	1	1	1	1	0.98
	160	20	1	1	1	1	0.98
	160	30	1	1	1	1	0.98
	160	40	1	1	1	1	0.98

^aLD, loading dose; CI, continuous infusion; CL_CR, creatinine clearance; q8, q12, and q24h, administered every 8, 12, and 24 h, respectively.

Dosing recommendations are provided in Table 7.

TABLE 7.

Dosing recommendations for susceptible isolates

CLCR (mL/min)	Dosing recommendationa
≤60	1 g twice daily
61–139	1 g 3 times daily
≥140	1 g LD then 2–4 g CI or 2 g twice daily

^aAssuming a MIC of 1 mg/L and a PTA of ≥90%. Lower doses may be suitable for isolates with a MIC of <1 mg/L, depending on the patient’s clinical context.

DISCUSSION

The present study has demonstrated that the presence of ARC and hypoalbuminemia affects the ceftriaxone dosing requirements for critically ill patients. Overall, a 1-g twice-daily dose is likely to provide a target ceftriaxone exposure in most patients with a normal renal function and a highly susceptible isolate (MIC, ≤0.25 mg/L). Such a clinical scenario is likely to be commonly encountered based on the high FTA observed at lower ceftriaxone doses (1 g twice daily) for common bacterial species, even in patients with ARC. However, the dose may need to be increased to at least 3 g administered in divided intermittent doses (i.e., 1 g three times daily) or as a continuous infusion in patients infected with an isolate that has a MIC of ≥0.5 mg/L and/or in patients with ARC. A continuous infusion dosing regimen may improve the PTA in patients with hypoalbuminemia compared with intermittent dosing regimens. The PTA for ceftriaxone against methicillin-susceptible S. aureus (MSSA) is unfavorable due to the high clinical MIC susceptible breakpoint (≤8 mg/L, based on a wild-type distribution) compared with Enterobacterales (≤1 mg/L), and alternative treatment options should be considered.

Our volume of distribution estimates did not differ appreciably between critically ill patients and healthy study participants (19). Conversely, other studies in critically ill patients have identified an increased volume of distribution in critically ill patients compared with healthy participants (17–19, 23, 24). The likely cause for such a discrepancy is the different modeling methods. Our study incorporated a ceftriaxone-albumin binding model, whereas previous studies have either only modeled the unbound ceftriaxone or extrapolated an unbound concentration from the total values, such as those by Schleibinger et al. (11) and Grégoire et al. (9). This is an important consideration as the unbound antibiotic concentration may not be reliably estimated using published values (12). A previous study has shown that such an approach is likely to overpredict the unbound ceftriaxone concentration by up to 83% (12). Furthermore, a calculated unbound concentration is unable to incorporate the inter- and intrapatient variability that may occur due to dynamic fluctuations in ceftriaxone and albumin concentrations. On the other hand, the total clearance in our study was comparable to those in studies previously conducted in patients admitted to the ICU (17, 18, 23, 24).

Despite the differences in the observed volume of distribution and clearance, our dosing recommendations are similar to those of previous studies. Schleibinger et al. recommended the use of a 2-g once-daily dose in patients with normal or impaired renal function, which may be appropriate for patients infected with highly susceptible organisms (MIC ≤0.125 mg/L) (11). On the other hand, patients with ARC require higher doses. Ollivier et al. identified that once-daily dosing is unlikely to achieve a C_min/MIC of >1 in patients with ARC, who require a dose of 2 g twice daily, a finding similar to ours (23). Moreover, our simulations describe a similar dosing requirement to that previously proposed, despite a different modeling approach to describe the bound ceftriaxone concentration and the maximum ceftriaxone binding concentration (B_max) (25). This pharmacokinetic study, similar to that conducted by Leegwater et al. (25), supports the use of administration by a continuous infusion to improve the PTA. Grégoire et al. recommend a ceftriaxone dose based on a nomogram considering both weight and renal function, which would also be approximately 2 g twice daily for a 70-kg patient with a creatinine clearance of 100 mL/min/1.73 m² (9). Furthermore, a small clinical study of patients randomized to receive a 2-g once-daily dose or a 2-g continuous infusion observed an increased clinical cure rate in continuous infusion recipients controlling for confounding baseline factors (odds ratio, 22.8; 95% confidence interval, 2.24 to 232.3) (26). The use of a continuous infusion may be acceptable as ceftriaxone is stable in aqueous solutions at room temperature up to 48 h (27).

Clinicians must consider the infecting pathogen when selecting both the antibiotic agent and dose. Our pharmacokinetic study describing a low PTA of ceftriaxone against MSSA aligns with small clinical studies, whereby ceftriaxone therapy is associated with an increased risk of treatment failure for MSSA bacteremia compared with other therapy, such as cefazolin (54.5% failure for ceftriaxone versus 28.9% failure for cefazolin, P = 0.029) (28). Given the appreciable ceftriaxone concentration intra- and interpatient variability following standard dosing, therapeutic drug monitoring (TDM) may be appropriate to provide individualized dosing recommendations to improve the probability that target exposures are met and potentially improve patient outcomes (1, 2).

Although there are potential benefits of dose optimization, excessive ceftriaxone exposure may increase the risk of dose-related adverse events. In one study of patients receiving high-dose ceftriaxone for central nervous system infections, the risk of a suspected ceftriaxone adverse reaction was increased with total trough concentrations of ≥100 mg/L (odds ratio, 4.06; 95% confidence interval, 1.31 to 12.60), greatly exceeding the target exposure used for dosing simulations in our study (data not shown) (29). Furthermore, despite the use of such high doses, adverse events were uncommon, affecting 2.6% of patients. Such high exposures are unlikely to be encountered in patients when our dosing recommendations are followed. However, clinicians should monitor for changes in renal function that may necessitate dose reduction consistent with our recommendations.

Our study has several limitations. First, different ceftriaxone assays were used for the separate studies. Nonetheless, each assay methodology adhered strictly to regulatory standards, and meaningful systematic measurement errors are unlikely to have occurred. Second, other patient factors, such as illness severity or bilirubin concentrations, were not available for all patients to ascertain any impact these covariates may have on ceftriaxone pharmacokinetics (11). Third, the MIC distribution for specific isolates was obtained from the EUCAST MIC database, which may not generalize to all institutions. Fourth, the ceftriaxone concentrations were obtained in blood and not from the site of infection. Although this will describe the doses required to achieve the desired PK/PD target for patients with bacteremia, the doses described in this study may not provide a sufficient exposure in patients with other infection sites, such as the epithelial lining fluid for pneumonia (7). Nonetheless, the selection of a serum target of 100% time above the MIC (T_MIC) may achieve the lowest likely effective exposure at the site of infection (60% fT_MIC). Fifth, a calculated creatinine clearance measurement was available and is considered to be less accurate than a measured urinary creatinine clearance (30). However, this is also labor intensive, and the calculated creatinine clearance affords a practical bedside approach to improve dosing in critically ill patients. Sixth, there was a minority of patients with renal dysfunction, limiting the applicability of our findings in this patient population. Last, patient-centered outcomes such as mortality or toxicity were not assessed in this study—an important focus for future studies assessing ceftriaxone dosing regimens. Importantly, doses of >2 g daily should be used with caution in patients with renal dysfunction given the potential for adverse reactions such as seizures.

Conclusions.

Our study has highlighted the need for increased ceftriaxone doses to achieve a C_min/MIC of >1 in critically ill patients with reduced albumin concentrations and/or increasing creatinine clearance. In patients with a normal creatinine clearance (~100 mL/min), a 1-g twice-daily dose is likely sufficient to achieve appropriate exposures in highly susceptible isolates (MIC, ≤0.25 mg/L). However, patients with ARC or those infected with a higher-MIC isolate (>0.25 mg/L) require a 3- to 4-g dose per day, either in divided doses or administered as a continuous infusion to achieve a C_min/MIC of >1. Ceftriaxone should not be routinely used for the treatment of MSSA infections. Patients with ARC may benefit from TDM and early implementation of higher dosing regimens.

MATERIALS AND METHODS

Setting.

This study analyzed total and unbound ceftriaxone plasma concentrations obtained in three separate studies involving critically ill patients admitted to the ICU in two tertiary university hospitals—Royal Brisbane and Womens’ Hospital, University of Queensland, and Prince of Wales Hospital, Chinese University of Hong Kong. Data from a previous pharmacokinetic study with ceftriaxone conducted by Joynt et al. (17) (10 patients, Prince of Wales Hospital, Hong Kong) was combined with data obtained from a single-center therapeutic drug monitoring (TDM) program (31) conducted at the Royal Brisbane and Womens’ Hospital (17 patients) and a ceftriaxone pharmacokinetic study (nine patients) that included patients from both of the tertiary centers (Optimizing Antibiotic Cerebrospinal Fluid Exposures Study [OACES]). An ethics waiver was granted to use the TDM data that was performed as part of standard care. Ethical clearance was obtained for the OACES study (HREC/16/QRBW/157) and for the study by Joynt et al. as previously described (17).

Study population.

The study population and relevant inclusion/exclusion criteria for the patients included in the study by Joynt et al. have been previously described (17). For the OACES study, adult patients (age ≥18 years) admitted to the ICU receiving ceftriaxone intravenously for the treatment of sepsis with an extraventricular drain were eligible. Patients who were pregnant or receiving either intermittent or continuous hemodialysis were excluded. Clinical and anthropometric data, including the patient’s age (years), weight (kilograms), suspected infection source, microbiological results, albumin (grams per liter), and creatinine clearance (milliliters per minute) were extracted from the patient’s medical records, where available.

Sample handing, storage, and measurement.

Ceftriaxone concentration analysis was conducted as per the specific study site, although the methods were similar (17). The following unpublished method was performed for patients enrolled in the OACES study.

Unbound ceftriaxone was separated from plasma using a Centrifree device at 37°C prior to concurrent analysis as for a total ceftriaxone concentration. Samples were spiked with deuterated ceftriaxone as the internal standard, and proteins were precipitated using acetonitrile. Total and unbound ceftriaxone plasma concentrations were quantified using ultrahigh-pressure liquid chromatography with tandem mass spectrometry (UHPLC-MS/MS) on a Shimadzu Nexera device connected to a Shimadzu 8030+ triple quadrupole mass spectrometer. Briefly, chromatographic separation was achieved with an Xbridge C₁₈ column (Waters, Milford, MA, USA) with a C₁₈ guard column (Phenomenex, Torrance, CA, USA) using a mobile phase A consisting of ammonium formate (10 mM) at pH 1.5 adjusted with formic acid and mobile phase B of acetonitrile and 0.2% (vol/vol) formic acid. Ceftriaxone and the internal standard were detected using an electrospray source in positive mode optimized with multiple-reaction monitoring at fragmentation ions of 554.7→396.1 (sample ceftriaxone) and 557.7→399.10 (internal standard, deuterated ceftriaxone). Mean intrabatch accuracy and precision values were 4.4% and 9.4%.

Patients included from the TDM study had ceftriaxone concentrations analyzed using a previously validated and published methodology (32).

Population pharmacokinetic modeling.

Total and unbound ceftriaxone concentrations were comodeled using the Pmetrics version 1.5.2 software package for R (21). One- and two-compartment pharmacokinetic models were assessed incorporating linear elimination from the central compartment, as well as a linear intercompartmental distribution process for two-compartment models. Simple and complex ceftriaxone protein binding models were assessed as previously described by Byrne et al. (33). A simple binding model related the unbound ceftriaxone concentration to the total ceftriaxone concentration by equation 5:

$$C_{\text{free}}=C_{\text{bound}}\times \text{FF}\times \frac{23}{\text{Alb}}$$ (5)

Alb is the albumin concentration (grams per liter), FF is the unbound free fraction, 23 is the mean albumin concentration (grams per liter), and C_free and C_bound represent the unbound and bound ceftriaxone concentrations, respectively (milligrams per liter).

The complex binding model assumes ceftriaxone is solely bound to albumin and is described by equations 6, 7, and 8:

$$C_{\text{bound}}=\frac{B_{\text{max}}\times C_{\text{free}}}{K_D+C_{\text{free}}}$$ (6)

$$B_{\text{max}}=\text{Alb}\times N\times \frac{M_{\text{CRO}}}{M_{\text{Alb}}}\times 1,000$$ (7)

$$K_D=\frac{1}{K_A}=\frac{K_{\text{off}}}{K_{\text{on}}}$$ (8)

C_bound and C_free represent the bound and unbound ceftriaxone concentrations (milligrams per liter), B_max is the maximum ceftriaxone binding concentration (milligrams per liter), Alb is the concentration of albumin (grams per liter), N is the number of binding sites per albumin molecule, M_CRO is the molecular weight of ceftriaxone, M_Alb is the molecular weight of albumin, K_D is the equilibrium dissociation rate constant (milligrams per liter), K_A is the equilibrium affinity constant (liters per milligram), K_off is the first-order dissociation rate constant (per hour), and K_on is the second-order association rate constant (liters per milligram per hour).

For each model, the additive and multiplicative error models incorporating a linear residual error estimate within Pmetrics were assessed and optimized. Diagnostic plots, including the observed versus predicted concentrations, residuals versus predicted concentrations, and concentrations versus time, and the visual predictive plots, as well as the Akaike information criterion (AIC), Bayesian information criterion (BIC), and log-likelihood ratio (LLR), were considered to compare different models. Any reduction in the AIC or BIC was considered a potentially improved model, while a reduction in the LLR of ≥3.84 U was considered statistically significant when comparing nested models.

Patient variables were selected a priori for potential inclusion in the model to account for interpatient variability based on physiological plausibility. Variables assessed against the volume of distribution, intercompartmental clearance, nonrenal clearance, and renal clearance included age, sex, weight, creatinine clearance (determined using the Cockcroft-Gault equation using total body weight) (34), and serum albumin concentration. The incorporation of different patient variables into the model that reduced the LLR, AIC, or BIC and/or improved the goodness-of-fit plots were included.

Probability of target attainment and fractional target attainment.

Monte Carlo simulations (n = 1,000) were performed using ceftriaxone doses of 1 or 2 g administered every 8, 12, or 24 h (q8h, q12h, or q24h, respectively) as an intermittent dose or as a 1-g loading dose followed by a continuous infusion (2 and 4 g/day) in patients with a creatinine clearance ranging from 40 to 160 mL/min, with serum albumin concentrations of 20, 30, or 40 g/L. The probability of these regimens achieving an unbound ceftriaxone C_min/MIC of >1 for MICs ranging from 0.032 to 8 mg/L was determined (22).

The fractional target attainment (FTA) represents the probability that patients would achieve the ceftriaxone PK/PD target against a specific pathogen assuming a wild-type distribution of MICs. The FTA, assuming an unbound ceftriaxone target of a C_min/MIC of >1, was calculated for the above ceftriaxone dosing regimens for Streptococcus pneumoniae, Haemophilus influenzae, Escherichia coli, Klebsiella pneumoniae, and methicillin-susceptible Staphylococcus aureus (MSSA) isolates, which are commonly implicated in severe community-acquired pneumonia and urosepsis infections. The MIC distribution for the bacterial species was obtained from the European Committee on Antimicrobial Susceptibility Testing (EUCAST) database (www.eucast.org), assuming 1 mg/L as the upper limit of susceptibility for isolates, with the exception of S. aureus, where the EUCAST epidemiological cutoff of 8 mg/L was used as the upper limit of susceptibility.

Dosing regimens were considered successful if the PTA was ≥90% of the simulated patient population. A dosing regimen was considered successful in treating a particular bacterial species if the FTA was ≥95%.