The Pharmacokinetics of Levetiracetam in Critically Ill Adult Patients: An Intensive Care Unit Clinical Study

Abstract

The aim of this study was to investigate levetiracetam pharmacokinetics in critically ill adult intensive care patients and to identify pathophysiological factors affecting its kinetics. Fourteen critically ill patients in an intensive care unit were enrolled in the study and received intravenous levetiracetam. Blood samples were collected at specific time points to determine the levetiracetam pharmacokinetics. Patient characteristics such as renal function, demographics, disease severity, organ dysfunction, and biochemical laboratory tests were evaluated for their influence on the kinetics of levetiracetam. Estimated glomerular filtration rate (eGFR) had a statistically significant (p = 0.001) effect on levetiracetam clearance. None of the other patient characteristics had a statistically significant effect on the pharmacokinetics. Simulations of dosing regimens revealed that even typically administered doses of levetiracetam may result in significantly increased concentrations and risk of drug toxicity in patients with impaired renal function. The Acute Physiology and Chronic Health Evaluation II (APACHE II) score differed significantly among the three groups with different epileptic activity (p = 0.034). The same groups also differed in terms of renal function (p = 0.031). Renal dysfunction should be considered when designing levetiracetam dosage. Patients with a low APACHE II score had the lowest risk of experiencing epileptic seizures.

1. Introduction

Levetiracetam, a second-generation antiepileptic drug (AED) with a unique mechanism of action, is approved by both the U.S. Food and Drug Administration and the European Medicines Agency (EMA) as an add-on therapy for the treatment of partial seizures, myoclonic seizures, and primary generalized tonic-clonic seizures. In Europe, it is also the only AED approved for the treatment of partial seizures in adults and adolescents aged 16 years and older [1].

In critical care, levetiracetam has been studied for the treatment of status epilepticus, tumor-related seizures, and seizures following subarachnoid or intracerebral hemorrhage, trauma, and stroke [2,3,4,5,6,7]. The pharmacokinetic and pharmacodynamic properties of this antiepileptic drug make it an advantageous option in the care of patients in the ICU. Compared to other AEDs, the pharmacokinetic profile of levetiracetam is favorable after both oral and intravenous administration (IV) [8,9]. After oral administration, there is rapid and nearly complete absorption (95%) with dose-proportional pharmacokinetics, low protein binding (10%), and low within-subject variability. It is excreted primarily renally by glomerular filtration with partial tubular reabsorption and has a plasma elimination half-life of approximately 6–8 h in adults. There are no known clinically significant drug interactions, and it is not associated with the common hemodynamic or cardiovascular side effects of other AEDs [8,10,11,12].

A clear association between levetiracetam serum levels and therapeutic or toxic effects has not been demonstrated [10,13]. In accordance with the International League Against Epilepsy (ILAE) guidelines for medication adherence monitoring, overdose and dose adjustment, a therapeutic range of 12–46 mg/L has been recommended [10,14,15]. However, different levetiracetam levels outside the reference range in different populations have been reported in the literature. Pregnant women, elderly and pediatric patients, and patients with end-stage renal failure or those requiring renal replacement therapy exhibit altered pharmacokinetics [16,17,18,19,20,21,22,23,24,25,26,27].

Relatively few studies in critically ill ICU patients have reported increased drug clearance (greater than 120–160 mL/min/1.73 m²) and treatment failure for seizures [21,28,29,30,31]. Based on empirical evidence, therapeutic drug monitoring for levetiracetam has been classified as ‘potentially useful’ but increasing evidence that altered pharmacokinetics of levetiracetam may be observed in critically ill patients has increased the need for therapeutic drug monitoring and pharmacokinetic modeling to determine optimal dosing regimens to maximize therapeutic effect and limit toxicity [22].

The aim of this study was to investigate the pharmacokinetics of levetiracetam in critically ill intensive care patients using nonlinear mixed-effect modeling approaches. The latter allow the identification of important pathophysiological factors (such as renal function) that could influence the kinetics of levetiracetam. Special emphasis was placed on the selection of the most suitable index for renal function and for this purpose several indices have been measured and investigated. Most of the patients participating in the study had from mild loss to severe loss of kidney function which allowed this exploration. Simulations were performed to assess the impact of important patient characteristics on the kinetics of levetiracetam. In addition, the potential influence of disease severity and renal function on epileptic status was investigated.

2. Materials and Methods

2.1. Patients and Data Collection

The study was conducted in the intensive care unit of a tertiary hospital (Latsio Burn Center, Thriassio General Hospital of Elefsina, Attica, Greece). Fourteen critically ill patients were enrolled in the study and all of them received intravenous (IV) levetiracetam (Keppra^®) either for prophylaxis or treatment of epilepsy. The study protocol was reviewed by the hospital Scientific Committee and written informed consent was obtained from all patients or their legal representatives before enrolment in the study. The study was conducted in accordance with the International Conference on Harmonization guidelines for good clinical practice and the Declaration of Helsinki, while all patient data were collected and processed in full compliance with the EU General Data Protection Regulation 2016/679 and EU Directive 2016/680 of the European Parliament [32,33,34,35].

Patient data included information on levetiracetam administration and ICU admission, demographics, biochemical parameters (e.g., liver enzymes, urea, albumin), fluid balance (in mL) and organ function expressed by the Sequential Organ Failure Assessment Score (SOFA) and the Acute Physiology and Chronic Health Evaluation II (APACHE II) score [36]. From patients’ weight and height, the body mass index (BMI) and body surface area (BSA) were also calculated [37]. The SOFA score is used to track the status of an ICU patient and determine the extent of their organ function or rate of failure. In this study, several expressions of this score were calculated, such as SOFA hemodynamic, SOFA liver, SOFA kidneys, and overall SOFA. The APACHE II is calculated within 24 h of a patient’s admission to the ICU using various measurements such as heart rate, respiratory rate, blood pH, and hematocrit. Taking all these factors into account, an integer value from 0 to 71 is calculated, with higher values indicating a more severe condition and a higher risk of death. Epileptic activity was also monitored and classified as negative, possible, and positive. The “positive” group included patients who had epileptic seizures, while the “negative” group included those who had no epileptic activity and levetiracetam was administered for prophylactic reasons only. The “possible” category included patients who had a clinical profile of epileptic activity, but this was not verified by electroencephalogram.

In this study, particular attention was paid to the role of renal function, as many of the patients in the ICU had impaired renal function. For this reason, various measures of renal function were calculated, such as creatinine clearance over time, Cockcroft–Gault creatinine clearance, and measures of estimated renal function (eGFR) such as MDRD eGFR (Modification of Diet in Renal Disease study group), CKD-EPI eGFR (Chronic Kidney Disease epidemiology collaboration), Jelliffe eGFR in mL/min/1.73 m² [38,39,40]. Patients in this study were divided into two subgroups according to their renal function; patients with stable renal function (group A) and patients with acute renal impairment (group B) (

Table 1. Patients’ demographic characteristics, levetiracetam dosing regimen, epileptic activity, and subgroup classification.

ID	Subgroup	Renal Function	Sex	Age (Years)	Weight (Kg)	Height (cm)	Dosing Regimen	Epileptic Activity
1	A	Stable renal function	Female	62	84	165	1000 mg BID	±
2			Male	26	70	180	1000 mg BID	±
3			Female	72	68	160	1500 mg BID	+
4			Female	59	75	165	1500 mg BID	+
5			Male	47	80	170	1000 mg BID	-
6			Male	75	110	170	1000 mg BID	-
7			Female	48	58	160	1500 mg BID	±
8			Male	33	80	170	1000 mg BID	+
9			Male	33	80	170	1500 mg BID	+
10			Male	61	65	160	1000 mg BID	-
11			Male	83	65	170	1000 mg BID	-
12	B	Acute kidney injury	Female	45	75	170	1000 mg BID	-
13			Male	73	65	170	1500 mg BID	-
14			Female	82	50	150	1000 mg BID	-

Key: Epileptic activity: - no activity, ± possible, + epileptic activity; BID, twice daily; eGFR, estimated glomerular filtration rate; Stable renal function, when eGFR > 60 mL/min/1.73 m².

). Levetiracetam treatment data included dosing regimen, cause of administration, chronic treatment, and loading dose administration.

Levetiracetam was administered diluted in 100 mL N/S 0.9% w/v by IV infusion over 15 min, while the loading dose was administered in 250 mL N/S 0.9% w/v by IV infusion over 30 min. Blood samples for all participants were taken two days after the start of dosing to ascertain whether the serum levels had achieved a steady state. The specific time points for blood collection were determined according to the pharmacokinetics of levetiracetam to obtain as much information as possible. The schedule for blood collection was 0 h (trough level), 20 min, and 3 h after steady-state administration of levetiracetam. The actual blood sampling times were recorded.

For blood collection and storage of serum or plasma samples the following procedure was applied: collection of 5 mL of whole blood in a coated Vacutainer tube with a red tip (Becton Dickinson). After blood collection, it was allowed to clot undisturbed at room temperature (usually 15–30 min). The clot was removed by centrifugation at 1000–2000 rpm for 10 min. After centrifugation, the serum was immediately divided into two fractions >200 μL and immediately transferred to polypropylene tubes (not gel tubes) where they were stored in a freezer −70 °C. For pharmacokinetic analysis, serum concentrations of levetiracetam were determined by high performance liquid chromatography using a UV detector at 205 nm after sample preparation by solid phase extraction using ClinRep^® HPLC Complete Kit, levetiracetam (Keppra^®) in serum/plasma (order number 15500) (RECIPE Chemicals & Instruments GmbH, Munich, Germany). The lower limit of detection was 0.14 mg/L, the lower limit of quantification was 0.46 mg/L, the upper quantification limit was 104 mg/L, and the recovery was 97–105%. The chromatographic test was performed with the ClinTest^® standard solution, calibration with the ClinCal^® serum calibrator and quality control with the ClinChek^® serum controls level I & II test solutions (order number 15582). The precision of the method was determined by calculating the relative standard deviation at three plasma concentrations during the same analysis (within-day precision) and in triplicate over six analyses (between-day precision). For the entire concentration range, the relative standard errors for both within- and between-assay precision were less than 7.2%.

2.2. Statistical Analysis

All statistical analyses of dosing schedules and clinical data were performed using IBM SPSS Statistics version 25 (IBM Corporation, Armonk, NY, USA). Statistical comparisons were performed at 5% significance level. For scale variables such as SOFA and eGFR measurements, a normality test was first performed (using the Shapiro–Wilk test) to determine whether the variables followed a normal distribution. In the case of normally distributed data, the independent t-test was performed to compare these variables between the two groups of patients. For comparison of more than two groups, the one-way method ANOVA was used with the post-hoc criteria of least significant difference and Tukey. The non-parametric analogue, Mann–Whitney, was used for variables that deviated from the normal distribution. The relationship between two nominal (or ordinal) measures (e.g., patient group and epilepsy status) was examined using the chi-square test. APACHE II score is actually an ordinal scale, but with many values. In other words, the APACHE II score can be considered as a Likert scale and therefore parametric statistical methods can be applied.

2.3. Population Pharmacokinetic Analysis

2.3.1. Non-Linear Mixed Effect Modeling

Population pharmacokinetic analysis was implemented in Monolix^TM 2020R1 (Lixoft, Orsay, France, Simulation Plus) where individual serum concentration profile data were analyzed using the expectation maximization algorithm of stochastic approximation for nonlinear mixed effects followed by importance test methods. The value of the objective function was calculated using the Monte Carlo method for the final population parameter values.

2.3.2. Structural Models, Error Models, and Covariates

One- and two-compartment models for intravenous infusion with first-order elimination and initial estimates for the parameters were examined [23,25,26,32,33,34]. A lognormal distribution of pharmacokinetic parameters was assumed, while several residual error models were tested, including constant, proportional, and combined. The parameters were estimated using the stochastic approximation estimation method (SAEM), and the objective function value was determined using the importance sampling Monte Carlo approach at the final population parameter values.

Once the structural model was determined, several covariates were tested. The covariates examined in this study were related to subject-specific characteristics, including sex, age, body weight, height, APACHE II score, timed creatinine clearance, Cockcroft–Gault eGFR, Jelliffe eGFR, CKD-EPI eGFR, MDRD eGFR, serum creatinine, urea, SGOT, and total SOFA. Three expressions of body weight were used: typical body weight, ideal body weight, and adjusted body weight. Linear and lognormal (allometric) models were assessed. Analyses of covariates were performed using stepwise forward selection and backward elimination. Continuous covariates were examined either untransformed or centered around the mean or median value of the covariate. The categorical covariate examined was sex, while SOFA hemodynamic, SOFA renal, and SOFA liver were ordinal variables. The total SOFA score, which comes from the sum of all individual SOFA scores, can be considered a continuous variable

The Pearson correlation test and one-sided ANOVA were used for continuous and categorical covariates, respectively. The Wald test was used to check whether the covariates could explain the variability of the parameters in the final model. For all pharmacokinetic analyses, the significance threshold was set at 5%.

2.3.3. Model Evaluation

Model selection was based on goodness-of-fit criteria, visual inspection of diagnostic plots, comparison of relative standard deviations of estimated parameters, precision of estimates, and changes in Akaike and Bayesian information criteria and log-likelihood. Visual inspection of goodness-of-fit was performed using plots of observed values versus predicted values for the population and individual weighted residuals versus time. Visual predictive checks (VPCs) were used to assess the predictive performance, stability, and robustness of the model. VPCs were generated from 1000 Monte Carlo simulations and 90% prediction intervals. Normalized prediction distribution errors versus concentration were also used.

2.3.4. Simulations

The final pharmacokinetic model was used to simulate the concentration–time profiles of levetiracetam at different levels of renal function. Three simulated subject groups of 50 subjects each were formed: (a) subjects with normal renal function using the model parameters reported in the literature [23], (b) subjects with 50% clearance limitation, and (c) subjects with the pharmacokinetic characteristics determined in the fourteen patients in this study. The dosing regimen for each subject was identical to that actually used in the ICU (see Table 1). All simulations were performed in Simulx^® (Monolix^TM 2020R1).

3. Results

Eleven patients with stable renal function in the ICU had an eGFR estimate greater than 60 mL/min/1.73 m² and were classified in subgroup “A”. The other three patients had acute renal failure and were classified in subgroup “B” (Table 1).

Regarding the reasons for administration of levetiracetam, most patients (71.4%) received levetiracetam for prophylactic purposes and only 28.6% for therapy. Only one patient (i.e., 7.1%) received levetiracetam for chronic treatment. The proportion of patients receiving a loading dose of levetiracetam was 42.9%, while the higher proportion (57.1%) initiated levetiracetam without a loading dose. The reason for levetiracetam administration was either prophylaxis (71.4% of patients) or treatment of seizures (28.6%). The mean eGFR of patients was 94.3 mL/min/1.73 m² (or equivalent to 5.66 L/h/1.73 m²) and the mean APACHE II score was 17.4.

3.1. Epileptic Activity

The possible influence of disease severity (expressed by the APACHE II score or the SOFA score) and renal function (expressed by the CKD-EPI eGFR) on epileptic status was investigated. The Shapiro–Wilk test indicated normal distribution of the data and therefore one-way ANOVA was performed to detect a difference in the values of APACHE II and eGFR between the three levels of epileptic activity (positive, negative, possible). It was found that the APACHE II score was significantly different between the three groups (p = 0.034). A statistically significant difference was also found for the CKD-EPI eGFR (p = 0.031). For the APACHE II score, the absence of epileptic activity (i.e., the “negative” group) differed from the “possible” (p = 0.033) and the “positive” epileptic activity (p = 0.027). The patients with the lowest APACHE II score belonged to the group without epileptic seizures.

Regarding the CKD-EPI eGFR values, the three epilepsy groups showed significant differences (p = 0.031). Similarly, the group of patients without seizures was statistically different from the “possible” (p = 0.034) and the “positive” group (p = 0.022). Statistical comparisons were made for all other characteristics, but no significant differences were found.

It should be clarified that the abovementioned analysis was not the primary purpose of the study and for this reason no power assessment was made focusing on this. In addition, this assessment only attempted to link disease severity with renal function and not to explore their relationship with levetiracetam pharmacokinetics. In any case, it should be underlined that due to the limited sample size and therefore the number of patients within each group, these findings should be considered as exploratory and further adequately powered studies are necessary to confirm the results.

3.2. Population Pharmacokinetic Model

Population pharmacokinetic analysis was initially performed separately for each patient group. However, due to the limited number of ICU patients in group B, no model could be derived. Therefore, in order not to waste the information of group B patients, all individuals were pooled into one group to increase the ability to develop a robust pharmacokinetic model, but each patient’s individual characteristics (e.g., renal function) were assessed and all available data were used. Furthermore, this pooling was possible because there was no statistically significant difference (p-value > 0.05) in levetiracetam concentration values between the two groups.

A one-compartment model with first order elimination was finally selected as the best model for the description of the levetiracetam concentration—time profiles. The estimates of the pharmacokinetic parameters of this best model are given in

Table 2. Population parameters of the final pharmacokinetic model of levetiracetam.

Parameters (Units)	Estimate	Standard Error	Relative Standard Error (%)	p-Value
Fixed effects
V (L)	47.19	5.096	10.8	-
Cl (L/h)	1.45	0.221	15.2	-
beta_eGFR	0.25	0.034	13.6	0.00164
Random effects
ω_V	0.184	0.037	20.1	-
ω_Cl	0.251	0.0454	18.1	-
Error model parameters
b	0.099	0.0124	12.5	-

Key: V, volume of distribution; Cl, clearance; ω_V, between-subject variability value for V; ω_Cl; between-subject variability value for Cl; beta_eGFR, allometric scaling factor for CKD-EPI eGFR (centered around median) on Cl; b, proportional component of the error model.

. It is of note that the volume of distribution of the central compartment (V) was 47.19 L and the clearance (CL) was 1.45 L/h. Of all the covariates examined for their influence on the kinetics of levetiracetam, only CKD-EPI eGFR was found to be significant for the clearance of levetiracetam. The coefficient of CKD-EPI eGFR with respect to clearance was 0.25, implying that better renal function (in terms of eGFR) leads to higher clearance of levetiracetam. The residual error model that led to the best results was the proportional error model with b = 0.099. The percentage relative standard errors for all estimated parameters had relatively low values. The significant covariate in the final model was the CKD-EPI eGFR (Wald test p = 0.00164), which is the Chronic Kidney Disease Epidemiology Collaboration equation, developed in an effort to provide a more precise formula for estimating glomerular filtration rate using serum creatinine, age, sex, and race data.

Figure 1 shows the graphical evaluation of the final model. The visual representation of the predictive power shows that the prediction interval of the developed model includes the experimental concentration data in all cases. Similarly, the good predictive ability of the model is also illustrated in Figure 2, where the population predicted versus observed concentration values are almost linearly correlated (Figure 2a) and the NPDE versus time (Figure 2b) and concentration (Figure 2c) plots show an adequate performance. Additional goodness-of-fit plots are shown in the supplementary material (Figure S1a–c).

The shrinkage of the volume and clearance estimates (parameter values were randomly sampled from the conditional distribution) was adequate and equal to 18.3% and −3.23%, respectively. The relationship between CKD-EPI eGFR and levetiracetam clearance is shown in Figure 3. An almost linear relationship (levetiracetam clearance = 0.01534·(CKD-EPI eGFR) − 0.21031) is observed with a correlation coefficient of 0.76. This finding is further evidence that an increase in CKD-EPI eGFR is associated with a subsequent increase in levetiracetam clearance.

3.3. Simulations

To determine the effects of renal function (expressed by the significant covariate CKD-EPI eGFR) on levetiracetam levels, simulations were performed based on the developed pharmacokinetic model. The utilized model parameters were those listed in Table 2, while two physiological/pathological situations were further simulated: Subjects with normal renal function [23] and subjects with 50% clearance restriction. It can be seen from Figure 4 that increased levetiracetam concentrations are observed with increasing renal function impairment. In healthy subjects, the mean maximum concentration was 31.1 mg/L (Figure 4a), which increased to 37.67 mg/L and 44.12 mg/L in patients with a 50% reduction in eGFR (Figure 4b) and patients with identical physiological characteristics to those in the study (Figure 4c), respectively. This means that the dosing regimens administered resulted in concentrations above the upper value of the reference range of 12–46 mg/L in almost half of the study patients [14].

4. Discussion

The aim of this study was to investigate the pharmacokinetics of levetiracetam in critically ill intensive care adult patients and to identify possible pathophysiological factors affecting the kinetics of levetiracetam and, consequently, clinical outcome. A clinical study was conducted on 14 patients from an intensive care unit of a tertiary care hospital in whom levetiracetam was administered for either prophylactic or therapeutic reasons. Eleven patients had stable renal function according to the physicians’ assessment, while the remaining three had acute renal injury. Having stable renal function does not mean that their renal function was normal, i.e., a glomerular filtration rate greater than 90 mL/min. In these 11 patients, the median eGFR was 97.7 mL/min and three patients (3 of 11) had an eGFR of less than 70 mL/min. In the patient group with acute kidney injury, the median eGFR was 45.3 mL/min. Overall, only five patients had an eGFR greater than 90 mL/min (i.e., normal renal function), and the remaining nine patients ranged from mild to severe loss of renal function.

Using measurements of serum concentrations in these patients, a population pharmacokinetic model for levetiracetam was developed. The final model for levetiracetam thus derived, referred to a linear one-compartment model with intravenous infusion administration and first-order elimination. This one-compartment model was consistent with other published levetiracetam studies conducted in children, adults, and elderly patients (Table S1) [23,24,25,26,27,41,42,43]. It should be mentioned that loading doses of levetiracetam were used. However, in the model development and the simulations (Figure 4) performed in this study, loading doses were not considered because they do not alter steady-state plasma levels, but only shorten the time required to reach these plateau levels.

Among the covariates tested, only measures of renal function showed a significant effect on clearance. Among the different eGFR measures explored, CKD-EPI eGFR was selected because it resulted in the lowest p-value (p = 0.00164, Table 2) and is one of the most commonly used eGFR measures. No significant relationships were found for other scores describing disease severity (such as APACHE II or SOFA), likely due to the small number of patients participating in the study. Levetiracetam clearance in this study (1.45 L/h, Table 2) was lower than the values reported in the literature (from 2.17 L/h to 6.87 L/h, Table S1). This finding may be attributed to the fact that our study included patients with impaired renal function. Moreover, the clearance estimates would also increase with increasing eGFR of our patients, as beta_eGFR had a positive sign (equal to 0.25, Table 2) [44]. In the study by Sime et al., which also recruited critically ill adult patients, the estimated clearance was 2.51 L/h [7]. Higher levetiracetam clearance estimates (e.g., 5.9 L/h) are reported in studies of epilepsy patients who are not critically ill, such as the large study by Ito et al. [41]. In that study the median eGFR estimate was 97.6 (mL/min)/1.73m², while in our study the median eGFR estimate of all participants was much lower 81.62 (mL/min)/1.73 m². The range of eGFR values in our study was from 18.8 mL/min to 123.3 mL/min, whereas for the study of Ito et al. the reported eGFR was as high as 189 mL/min [41]. The issue that our patients had poorer renal function compared with other studies reporting higher levetiracetam clearance studies, may explain the low clearance values found in our study.

The individual fitting plots for the 14 patients are shown in Figure S2. These fitting results demonstrate the adequacy of the descriptiveness of the developed model. Using only three blood samples per patient and data from only 14 subjects, it was possible to develop a model that describes the pharmacokinetics of levetiracetam. For comparison purposes, simulations were performed for the other relevant models in the literature. The model estimates (structural model, mean model parameter, between-subject variabilities, error model) were those reported in the literature (summarized in Table S1) and were kept fixed, while the covariates were related to our study patients. Then, it was attempted to examine the adequacy of fitting of these literature models to the concentration–time data of the 14 patients in this study. VPCs and normalized prediction distribution errors (NPDEs) plots were constructed (Figures S3 and S4 in the Supplementary Material). Figures S3 and S4 reveal that some literature models adequately fit the C-t data of our study, The best overall performance was observed for the models of Karatza et al., Pigeolet et al., Ito et al., and Chhun et al. [23,26,41,43]. However, for all these models, the performance was not better than for our model.

The positive correlation (correlation coefficient equal to 0.76) between eGFR and levetiracetam clearance, shown in Figure 3, indicates a lower elimination capacity in patients with renal dysfunction. This result is consistent with the prescribing information for levetiracetam, which recommends adjusting the levetiracetam dose based on renal function [24,41,42]. At this point, it should be mentioned that several other covariates related to patient disease severity have been tested for their potential role in influencing levetiracetam pharmacokinetics. However, none of these covariates were found to have a significant effect on any of the model parameters of levetiracetam (i.e., p > 0.05).

In a further step, the developed population pharmacokinetic model was used to evaluate the effects of impaired renal function on the kinetics of levetiracetam (Figure 4). In the simulated “healthy” subjects, the dosing regimens resulted in levetiracetam levels that were within the therapeutic range (Figure 4a), which is consistent with the literature [40]. The same levetiracetam doses resulted in significantly increased concentration levels in patients with impaired renal function, such as a 50% decrease in clearance and in the study patients (Figure 4b,c), as well as a higher risk of drug toxicity. It is worth noting that this observation is consistent with recommendations in the literature regarding treatment with levetiracetam. It is of note that if seizures cannot be controlled even with high concentrations of levetiracetam, increasing the dose above 2000 mg/day will not benefit the patient; instead, clinicians should consider switching to an antiepileptic drug with a different mechanism of action [23,42]. These findings highlight the importance of dose adjustment in patients depending on their eGFR [45,46]. In general, monitoring levetiracetam serum concentrations as part of therapeutic drug monitoring, ensures more effective and safer dosing regimens.

The relationship between patients’ epileptic activity (group 1: no epileptic activity, group 2: possible epileptic activity, group 3: confirmed epileptic activity), APACHE II score, and eGFR levels has also been investigated. Despite the fact that only a few subjects were included in the study, statistically significant (p = 0.034) differences in the APACHE II score between the three groups of patients (no epileptic activity, possible, and confirmed epileptic activity) were found. The same groups also differed statistically significantly in terms of Cockcroft–Gault eGFR (p = 0.031).

The study had some limitations, one of which was the small number of patients who participated. Due to the small sample size, correlations between variables and their influence on pharmacokinetics may not have been detected (e.g., body weight) [45,46]. Other studies have been published with large numbers of subjects of epileptic adult patients receiving levetiracetam [25,26,40,41]. However, in this study special emphasis was based on exploring the most appropriate index for renal function in the case of critically ill patients. For this reason, this study explored seven indices for glomerular filtration rate (urea, creatinine, renal impairment stage, Cockroft–Gault_eGFR, MDRD_eGFR, CKD-EPI_eGFR, and Jeliffe_eGFR). Even using the data from this small sample, one covariate (the eGFR values collected from CKD-EPI) was found to perform best compared with all other indices and to play a significant role in the pharmacokinetics of levetiracetam. To increase the convergence of the model, five Markov chains were run in parallel. All available data were replicated, resulting in a number of 50 patients and improving the precision of the estimates.

5. Conclusions

A clinical trial was conducted on fourteen critically ill patients in an intensive care unit. Levetiracetam was administered intravenously either for prophylaxis or treatment of seizures. Blood samples were collected and plasma levels of levetiracetam were quantified. A population pharmacokinetic analysis was performed that included a number of patient-related somatometric characteristics and pathophysiological variables. A one-compartment model with intravenous infusion and first-order elimination kinetics was found to best describe the concentration–time data. Several renal function measures were estimated and evaluated in the modeling. Among them, CKD-EPI eGFR showed a statistically significant effect on levetiracetam clearance. With impaired renal function (especially at lower CKD-EPI eGFR values), levetiracetam clearance decreased. Simulations of dosing regimens revealed that even typically administered doses of levetiracetam may result in significantly increased concentrations and a higher risk of drug toxicity in patients with impaired renal function. Therefore, renal function should be considered when developing levetiracetam dosing regimens. Other measures describing disease severity, such as SOFA and APACHE II, had no effect on levetiracetam kinetics. Finally, when analyzing the association between APACHE II score and epileptic activity, it was found that patients with the lowest APACHE II score had a lower risk of experiencing epileptic seizures.