FormalPara Key Points

Clearance of acetaminophen in critically ill patients is reduced.

Dosing regimen modifications are suggested for intravenous acetaminophen in this patient population.

1 Introduction

Critically ill patients are often have a need for multiple drug administration owing to the severity of their primary illness and the associated co-morbidities. A recent study estimated a median (range) of 11 (8–16) drugs administered in critically ill adults [1]. Factors altering the pharmacokinetics of drugs in critically ill patients include changes in the gastric emptying time, variations in the plasma protein binding and hepatic blood flow, and alterations in renal clearance [2]. Generally, an increased volume of distribution (Vd) is observed for hydrophilic drugs, necessitating loading doses [3]. Regarding changes in the renal clearance, a reduction is often observed owing to acute kidney injury [4]. However, recently augmented renal clearance (ARC) has been observed in critically ill patients (in particular with widespread systemic inflammation) to an extent of around 52.5% [5]. Together, these critical factors carry a greater potential to result in altered pharmacokinetics and consequently pharmacodynamics of drugs in this sub-population.

Acetaminophen is a hydrophilic drug, and the most used antipyretic and analgesic drug worldwide [6]. Several routes of administration of acetaminophen are used in clinical practice, namely, oral, rectal, and intravenous. In critically ill patients, oral bioavailability of paracetamol may be compromised due to reduced gastric motility as well as altered intestinal blood flow, and so an intravenous route with a better predictable pharmacokinetics (and accordingly pharmacodynamics) is preferred [7]. In adult healthy humans, clearance (CL) and volume of distribution (Vd) were 24.0 L/70 kg, while 43.5 L/70 kg was observed for acetaminophen [8]. Critically ill patients on ventilator support demonstrated an elimination half-life of 2.2 h, Vd of 1.03 L/kg, and CL of 0.33 L/kg/h [9]. A recent study in critically ill patients with traumatic brain injury observed a greater acetaminophen CL and a wide range of Vd of acetaminophen following intravenous administration, and recommended higher doses for the optimal control of pyrexia [10]. Considering a therapeutic reference range of 10–20 µg/ml for the antipyretic and analgesic effects, a significant number of critically ill patients (8 of 21) were observed with below-detectable limits of serum acetaminophen levels [11]. Despite a uniform recommendation that acetaminophen at 1 g every 6 h is ineffective and higher doses must be considered for the critically ill, none of these studies evaluated the impact of various dosing regimens on achieving therapeutic concentrations. Acetaminophen is primarily metabolized to its sulfate and glucuronide metabolites, while a small fraction did to an alkylating metabolite, N-acetyl-p-benzoquinone imine [12]. Nearly a fifth of critically ill patients have been observed with acquired liver injury and failure due to cholestatic/hypoxia/mixed etiology [13]. Critically ill patients often need administration of acetaminophen as a part of their pain or fever control; yet they are at increased risk of hepatotoxicity due to the severity of their illness and the toxic metabolite from acetaminophen. There is an absence of any published literature on the estimation of acetaminophen metabolites in critically ill adult patients. Hence, the present study was undertaken to evaluate the pharmacokinetics of acetaminophen and the principal metabolites following intravenous administration, using a population pharmacokinetic model approach, and to find out the optimized dosing regimen for obtaining therapeutic trough concentrations.

2 Methods

2.1 Study Design and Ethics

This was a cross-sectional study carried out following Institutional Ethics Committee approval and written consent from patients or their legally acceptable representatives. The study was carried out between November 2021 and April 2022 in the Salmaniya Medical Complex, the largest intensive care unit in the Kingdom of Bahrain. We adhered to the latest Declaration of Helsinki guidelines.

2.2 Study Procedure

We included critically ill adults (aged 21 years) receiving intravenous paracetamol for any indications. Those admitted only for observatory purposes were excluded. The following details were captured: age, sex, body weight, APACHE II score (acute physiology and chronic health evaluation II score) within a day of admission, time-point of intravenous paracetamol administration, liver function tests [serum alkaline phosphatase (ALP), alanine aminotransferase (ALT), gamma glutamyl transpeptidase (GGT), and total proteins], and serum creatinine. Creatinine clearance (CLCR) was estimated using the Cockcroft–Gault equation [14]. Those with CLCR exceeding 130 ml/min were categorized as having ARC [15]. The standard of care in our intensive care unit is to administer intravenous acetaminophen as infusion in 100 ml normal saline at 1 g over 30 min, irrespective of indications. One to three blood samples (5 ml) were conveniently withdrawn between 0.5 and 6 h following acetaminophen administration per study participant for the estimation of acetaminophen, and its metabolites (acetaminophen glucuronide and acetaminophen sulfate). The whole blood sample was kept at room temperature for 30 min and then centrifuged at 2000 rpm for 10 min. The obtained serum was stored at – 80 °C pending analysis. The therapeutic range for serum acetaminophen concentration was considered to be 66–132 μM (10–20 µg/ml), while 990 μM (150 µg/ml) was considered as the threshold for toxic concentrations [16].

2.3 Bioanalytical Method for Estimation of Serum Paracetamol and its Metabolites

High-performance liquid chromatography (HPLC) was used to measure serum acetaminophen, acetaminophen glucuronide, and acetaminophen sulfate, as described previously [17, 18]. A Waters USA UV/visible detector, consisting of an e2695 pump, and an autosampler,,was used to detect the compounds at the wavelength of 254 nm. All the reagents (acetaminophen, acetaminophen glucuronide, acetaminophen sulfate, potassium dihydrogen orthophosphate, and isopropanol) were obtained from Sigma-Aldrich. Tetrahydrofuran (THF) was purchased from Fluka. Methanol and water at the ratio of 3: 1 was used as the isocratic mobile phase that was passed on through the C18 analytical column with 1.2 ml/min flow rate for acetaminophen estimation. For the estimation of acetaminophen metabolites, the mobile phase consisted of potassium dihydrogen orthophosphate–isopropanol–THF combination that was passed on through C18 column at a flow rate of 1 ml/min. Serum acetaminophen was estimated with the limit of detection of 16 μM, and a range between 16 and 1324 μM was the linearity. Similarly, acetaminophen metabolites were estimated with the limit of detection of 0.1 μM with 0.1 and 10 μM as the linearity range. The percent coefficient of variations for all the analytes were less than 5%, thus establishing the accuracy and precision of the assays. Only one concentration exceeded the limit of quantification for serum acetaminophen and was included in the analysis, while none of the measured concentrations for the metabolites were out of the quantification limits.

2.4 Pharmacokinetic Modeling

We used a nonlinear mixed-effects approach for joint modeling of the pharmacokinetics of acetaminophen and its metabolites (acetaminophen glucuronide and acetaminophen sulfate). Estimation of pharmacokinetic parameters was achieved using the stochastic approximation expectation maximization algorithm implemented in Monolix 2021R2 software [19]. Since the data were very sparse, a Bayesian approach was implemented by considering prior pharmacokinetic parameters from the literature [8]. We used a two-compartment model to model the pharmacokinetics of acetaminophen. A one-compartment model with linear elimination was used to model acetaminophen glucuronide and the sulfate metabolites. Since the Vd of the two metabolites is not identifiable, it was set to the central volume of distribution (V1) of acetaminophen. The CL for the two metabolites was estimated from the data. The concentration–time profiles of acetaminophen, acetaminophen glucuronide, and acetaminophen sulfate were jointly modeled according to the series of differential equations (Eqs. 1–4) as illustrated in Fig. 1. The amount of acetaminophen in the central and peripheral compartments is denoted by Am and Ap, respectively. The acetaminophen central and peripheral volume of distributions and inter-compartmental clearance were represented by V1, V2, and Q, respectively. The rate constant associated with the formation of acetaminophen glucuronide is denoted at kpm1, while that of acetaminophen sulfate as kpm2. Am1 is the amount of acetaminophen glucuronide and Am2 the amount of acetaminophen sulfate.

Figure 1
figure 1

Schematic of the parent-metabolite pharmacokinetic model of acetaminophen. Ac and Ap are the amounts of acetaminophen in the central and peripheral compartments, respectively, V1 is the central volume of distribution for acetaminophen which is assumed to be equal to acetaminophen sulfate and acetaminophen glucuronide volume of distribution, and Am1 and Am2 are the amounts of glucuronide and sulfate metabolite of acetaminophen, respectively

Full size image
$$\frac{dAc}{dt}=-\frac{CL}{V_1}\times Ac-kpm1\times Ac-kpm2\times Ac-\frac{Q}{V_1}\times Ac+\frac{Q}{V_2}\times Ap$$
(1)
$$\frac{dAp}{dt}=\frac{Q}{V_1}\times Ac-Ap\times \frac{Q}{V_2}$$
(2)
$$\frac{dAm1}{dt}=kpm1\times Ac-Am1\times \frac{CLm1}{V_1}$$
(3)
$$\frac{dAm2}{dt}=kmp2\times Ac-Am2\times \frac{CLm2}{V_1}$$
(4)

The residual unexplained variability was explored with proportional, combined (additive and proportional), and exponential models. The variability in the pharmacokinetic parameters was described by the interindividual variability model.

The effect of covariates on the pharmacokinetic parameters was investigated, to account for body size effect (weight), allometric scaling by fixing exponents of the volume of distribution and clearance to 1 and 0.75, respectively. The covariates tested were, age, sex, APACHE II score, serum total protein, creatinine clearance, alanine transaminase, alkaline phosphatase, and gamma-glutamyl transferase. All covariates were introduced in the model one at a time. If the introduction of the covariate resulted in a statistically significant change in interindividual variability and objective function value (OFV) of a minimum 3.84 points, it was retained in the model. This was followed by a backward deletion step of covariates in the full model. Deletion of the covariate was carried out if it led to the decrease in the OFV by at least 6.63 points and the Wald test was significant (p < 0.05).

2.5 Model Evaluation

The final pharmacokinetic model was assessed for possible misspecification using visual diagnostic plots such model-predicted concentration versus observed concentration, individual weighted residuals (IWRES) versus time/predicted concentration, and visual predicted checks (VPC). Numerical diagnoses were also performed to assess how well the observed data fit the final pharmacokinetic model. These included the decrease in the OFV defined as \(-2*loglikelihood\) precision in the estimated pharmacokinetic parameters described as the percentage relative standard error (%RSE). The robustness and stability of the final pharmacokinetic model were evaluated by performing a non-parametric bootstrap (500 bootstrap runs) resampling procedure. This was achieved using Rsmlx ® Speaks ‘Monolix’), which is a Monolix R package [20].

2.6 Assessment of Current Dose and Dosing Regimen Optimization

The minimum acetaminophen serum concentration needed for analgesia and antipyresis is 66 μM (10 μg/mL) [16] while a serum concentration of 990 μM (150 μg/mL) is considered the threshold for hepatotoxicity [21]. Therefore, the target for the pharmacokinetics simulation for the assessment of the present dose and dosing regimen was the acetaminophen serum concentration between ≥ 66 and < 990 μM in 24 h. Similarly, hepatoxicity was assessed by determining the time the acetaminophen plasma concentration was above 990 μM (T ≥ 990 μM) in 24 h. Monte Carlo simulations were performed in Simulx 2021 to determine the optimum dosing regimen.

3 Results

3.1 Demographic Characteristics

Eighty-seven participants were recruited, and Table 1 provides a summary of their characteristics. The following diagnoses were observed amongst the study participants: post-operative surgery (n = 31), sickle cell disease with vaso-occlusive crisis (n = 16), cerebro-vascular accidents (n = 12), sepsis (n = 11), road-traffic accident with head injury (n = 7), burns (n = 2), and one each with diabetic ketoacidosis, immune thrombocytopenic purpura, epilepsy, renal vein thrombosis, beta thalassemia, Hodgkin’s lymphoma, end-stage renal dysfunction, and acute pneumonia with cystic fibrosis. Following were the concomitant disorders observed: systemic hypertension (n = 24), diabetes mellitus (n = 19), and systemic lupus erythematosus (n = 2).

Table 1 Demographic characteristics of the study participants (n = 87).
Full size table

3.2 Pharmacokinetic Modeling

A two-compartment pharmacokinetic model has been built starting from the prior pharmacokinetic model [8]. The residual variability was best explained by the exponential error model. The interindividual variability and residual error values of the base model are displayed in Table 2. Inclusion of body weight through allometric scaling improved the model (ΔOFV, − 8.49), which explained 15.8% and 10% of the variation in clearance and Vd, respectively. Parameters were estimated with better precision when allometric exponents were fixed than when estimated. The clearance or Vd was not significantly affected by sex, age, serum total protein, alanine transaminase, alkaline phosphatase, and gamma-glutamyl transferase. The final pharmacokinetic model parameters are displayed in Table 3. Equations (5–10) represent the individual models for Vd and clearance:

$${V}_{1,i}=7.87\times {\left(\frac{weigh{t}_{i}}{70}\right)}^{1}\times {e}^{{\eta }_{i}}$$
(5)
$$C{L}_{i}=0.58\times {\left(\frac{weigh{t}_{i}}{70}\right)}^{0.75}\times {\left(\frac{CL_{CR,i}}{96.8}\right)}^{0.56}\times {e}^{{\eta }_{i}}$$
(6)
$$V{}_{2,i}=8.87\times {\left(\frac{weigh{t}_{i}}{70}\right)}^{1}\times {\left(\frac{CL_{CR,i}}{96.8}\right)}^{1.02}\times {e}^{{\eta }_{i}}$$
(7)
$${Q}_{I}=44.2\times {\left(\frac{weigh{t}_{i}}{70}\right)}^{0.75}\times {e}^{{\eta }_{I}}$$
(8)
$$Clm{1}_{I}=22\times {\left(\frac{weigh{t}_{i}}{70}\right)}^{0.75}\times {\left(\frac{APACH{E}_{i}}{11.5}\right)}^{-0.48}\times {e}^{{\eta }_{I}}$$
(9)
$$CLm{2}_{i}=94.7\times {\left(\frac{weigh{t}_{i}}{70}\right)}^{0.75}\times {e}^{{\eta }_{i}}$$
(10)

where acetaminophen central and peripheral individual volumes of distribution models are denoted by V1,i and V2,i, respectively, individual clearance and intercompartmental clearance are denoted by CLi and Qi, respectively, CLm1i is the individual clearance of acetaminophen glucuronide, CLm2i denotes the individual clearance of acetaminophen sulfate, and the random effect in the individual parameter is denoted by ηi. APACHE and CLCR are covariates predictive of the pharmacokinetic parameters. All pharmacokinetic parameters have been normalized to the typical individual in the population weighing 70 kg.

Table 2 Interindividual variability and residual error values of the base model
Full size table
Table 3 Population pharmacokinetics parameters of acetaminophen and its metabolites
Full size table

A good agreement was observed in the diagnostic plots between the predicted and observed concentration data points (Fig. 2). However, there was a slight underprediction of higher concentrations for both metabolites. The VPC (Fig. 3) for acetaminophen and the two metabolites showed that the observed data points were within the model-predicted range. The plots of the IWRES (Fig. 4) versus time or predicted concentration did not show any significant trends in the splines. The bootstrap parameters were comparable with the estimated parameters. There was high BSV in the CL, Q, and V2. Additionally, Q and kpm2 were not estimated with good precision as indicated by the high %RSE of 99 and 88.8%, respectively (Table 3).

Figure 2
figure 2

Observed acetaminophen, acetaminophen glucuronide (aceta.gluc), and acetaminophen sulfate (aceta.sulp) concentrations versus population or individual predictions; the red line is the spline and the black line is the line of unity

Full size image
Figure 3
figure 3

Visual predictive checks of 1000 simulations for acetaminophen, acetaminophen glucuronide, and acetaminophen sulfate; the upper pink shaded area the 75th percentile, the middle green shaded area is the 50th percentile, and the lower pink shaded area is the 5th percentile

Full size image
Figure 4
figure 4

Individual weighted residuals (IWRES) versus time or predicted concentration of acetaminophen (aceta.), acetaminophen glucuronide (aceta.gluc) and acetaminophen sulfate (aceta.sulp); the red line is the spline and the dotted line is the mean of the residuals

Full size image

3.3 Dosing Regimen Optimization

The original study design was a combination of 1000 mg BID (twice a day), TDS (three times a day) and QID (four times a day) dose regimens. Simulations of a BID dose regimen showed that 47% of the population attained the desired target concentration range within an average time of 15 h. Simulation of the TDS and QID dose regimes led to 40% of the population attaining the target concentration range within the time range of 13–14 h. The BID dose had the lowest average time in which plasma concentration was above 990 μM compared to the TDS and QID dose regimens. In terms of toxicity in the BID group, the acetaminophen serum concentration in 20% of the population was in the toxic range for 5 h. The plasma concentration in 34% of the population for both TDS and QID was in the toxic range for an average of 8 h.

4 Discussion

We evaluated the pharmacokinetic parameters of acetaminophen (parent drug and sulfate and glucuronide metabolites) following intravenous administration in 87 critically ill patients. The following were the primary pharmacokinetic parameters observed: Vd (7.87 L/70 kg) and CL (0.58 L/h/70 kg) of the parent drug. Dose optimization suggested twice-daily administration of acetaminophen at 1000 mg to ensure that a relatively higher proportion of the population retained their serum concentrations in the therapeutic range with reduced risk of their concentrations remaining in the toxic range.

Studies evaluating pharmacokinetics of intravenous acetaminophen in the critically ill are limited. In healthy humans, a mean (SD) CL following intravenous acetaminophen was 352 (40) ml/min [22], while in the present study (among the critically ill), it was 9.7 ml/min (0.58 L/h). Even the summary of product characteristics for intravenous acetaminophen recommends a mean CL of 18.9 L/h (0.27 L/h/kg) in general, which is nearly 32-fold higher than the critically ill [23]. A recent pooled analysis from seven studies and 2755 concentrations revealed a mean CL of 16.7 L/h/70 kg [24]. Morse et al., in 116 healthy volunteers, also observed a CL of 24 L/h/70 kg [8]. de Maat et al. evaluated the pharmacokinetics of intravenous acetaminophen among patients admitted in medium and intensive care units [25]. The authors used non-compartment analysis using formulae-based estimation of the pharmacokinetic parameters and observed a median CL of 23.65 L/h with the median body weight of 78 kg. They also observed a higher Vd (52 L) compared to the present study. Samson et al. [9] used population pharmacokinetic modeling in 19 critically ill patients following the first dose of intravenous acetaminophen (dosing regimen: 1 g every 6 h) and observed a clearance of 0.33 L/kg/h. The authors also observed that almost all patients had their serum concentrations below the therapeutic range. Acetaminophen is perceived as a relatively safer drug, ubiquitously compared to other non-steroidal anti-inflammatory drugs. However, we observed that a 1-g dose administered three or four times daily is associated with increased risk of toxicity (due to toxic drug concentrations). Mild to moderate hepatic dysfunction is commonly observed in the critically ill, attributed either to ischemic hepatitis manifesting as diffuse hepatic necrosis or to cholestatic liver disease [26]. This explains the reduced clearance of the parent drug and the simulation results pointing towards the administration of a twice-daily regimen which is associated with the least risk of toxicity in the present study.

The present study is the first to evaluate the pharmacokinetics of intravenous acetaminophen and its principal metabolites in critically ill patients through a population pharmacokinetic approach. However, we did not evaluate the hypotensive effect of intravenous acetaminophen and its relationship to the dose and pharmacokinetic parameters. Due to feasibility concerns, we did not collect blood samples for analysis in steady state as a steady state was not achieved based on the required doses. Concomitant drugs with potential interaction with paracetamol were not considered as there are hardly any well-established pharmacokinetic interactions with the drug and a small sample size. Pharmacodynamic studies evaluating the therapeutic response to various dosing regimens in addition to PK parameters would help in a better understanding of the therapeutic implications of various dosing regimens in this vulnerable population.

5 Conclusion

Primary pharmacokinetic parameters for intravenous acetaminophen in critically ill populations have been established along with their principal metabolites. A reduced clearance of the parent drug was observed, and a reduction in the frequency of administration should be considered to reduce the risk of supra-therapeutic drug concentrations in this population.