Plasma and cerebrospinal fluid pharmacokinetics of ondansetron in humans

Abstract

Aims:

Changes in serotonergic sensory modulation associated with overexpression of 5-HT₃ receptors in the CNS have been implicated in the pathophysiology of neuropathic pain after peripheral nerve damage. 5-HT₃ receptor antagonists such as ondansetron can potentially alleviate neuropathic pain, but have limited effectiveness, due potentially to limited CNS access. However, there is currently limited information on CNS disposition of systemically-administered 5-HT₃ receptor antagonists. This study evaluated the cerebrospinal fluid (CSF) disposition of ondansetron, as a surrogate of CNS penetration.

Methods:

Fifteen patients were given a single 16 mg intravenous 15 min infusion of ondansetron, followed by serial blood and a single CSF sampling. Population PK modeling approach was implemented to describe the average and individual plasma and CSF profiles of ondansetron. A two-compartmental model was used to capture ondansetron plasma PK with a single CSF compartment to describe distribution to the CNS.

Results:

The individual model-estimated CSF to plasma partition coefficients of ondansetron were between 0.09 and 0.20. These values were mirrored in the calculated CSF penetration ratios, ranging from 0.08 to 0.26.

Conclusions:

After intravenous administration, CSF concentrations of ondansetron were approximately 7-fold lower than those observed in the plasma. A model could be developed to describe individual CSF concentration time profiles of ondansetron based on a single CSF data point. The low CSF penetration of ondansetron may explain its limited analgesic effectiveness, and affords an opportunity to explore enhancing its CNS penetration for targeting conditions such as neuropathic pain.

Introduction

Neuropathic pain affects 7–8% of the adult population worldwide and is particularly challenging to treat [1 2]. More than 30% of neuropathic pain patients continue to suffer despite treatment; and there is clearly an urgent need for new treatment approaches [3]. Serotonin is an important contributor to endogenous analgesic mechanisms. Under normal conditions, descending serotonergic neural control from the rostral ventromedial medulla to the spinal cord inhibits neuronal activity and hypersensitivity, and contributes to analgesia [4]. This is primarily a result of serotonin activity on 5-HT₁ and 5-HT₇ subtypes of serotonin receptors, which are all G-protein coupled receptors (GPCRs) [5]. However, it is suggested that after peripheral nerve damage, the character of serotonergic descending modulation changes from inhibitory to facilitatory through overexpression of 5-HT₃ receptors (5-HT₃Rs) in the spinal cord, as 5-HT₃Rs have excitatory properties, and are the only ion channels among the 5-HT receptor family [6 7]. These findings suggest that 5-HT₃Rs in the central nervous system (CNS), particularly in the spinal cord dorsal horn, could be a promising pharmacological target in neuropathic pain [8 9]. Clinical trials performed with systemically-administered 5-HT₃R antagonists for treating neuropathic pain have yielded mixed results [10 11]. On the other hand, preclinical literature suggests that intrathecal (IT, directly to the cerebrospinal fluid - CSF) administration of 5-HT₃R antagonists such as ondansetron alleviates mechanical and thermal hypersensitivity in animal models of nerve injury [8 12].

We hypothesized that one of the mechanisms leading to unsuccessful clinical translation of the promising animal data in neuropathic pain models is related to inability to achieve effective ondansetron concentration at the site of action following systemic administration. CNS distribution of 5-HT₃R antagonists such as ondansetron may depend on the expression level of efflux transporters at the capillaries of the blood-brain barrier (BBB), such as P-glycoprotein (Pgp), which limit CNS exposure [13 14]. The literature on the antiemetic effects of ondansetron (its primary clinical indication) indeed supports enhanced antiemetic effect in patients with single nucleotide polymorphisms (SNPs) in the ABCB1 gene that encodes the P-glycoprotein [15].

There are limited data on CNS penetration of 5-HT₃R antagonists. Direct sampling from the brain or spinal cord is rarely possible in humans, and assessment of the time-course of drug concentrations in the CSF may serve as a more practical approach [16]. Furthermore, serial CSF sampling requires either an intrathecal catheter, or serial subarachnoid punctures, the latter of which are undesirable and a safety concern. Considering the challenges associated with performing such studies, population pharmacokinetic (PK) modeling can be utilized for analysis of sparse data (such as a single CSF sample collected at different time points from different subjects) and provides both the mean population trend and between-subject variability. Therefore, the goal of the study was to investigate the CSF distribution of ondansetron after intravenous (iv) administration of clinically-relevant doses, and use population-based PK modeling approach to describe ondansetron distribution to the CNS.

Materials and Methods

Study Design

The study was approved by the Washington University IRB and was registered on clinicaltrials.org (NCT02901054) prior to participant enrollment. Candidates for a surgical total knee or total hip replacement were invited to participate in the study, where a single dose of iv ondansetron was administered in the pre-operative waiting room, followed by serial blood and a single CSF sampling. Patients from this surgical population were selected (in lieu of healthy volunteers), as these patients receive intrathecal anesthesia as a standard of care, allowing CSF access. Ondansetron was administered at varying times before the anticipated lumbar puncture, to achieve a reasonable distribution of sampling intervals to allow a population PK modeling approach.

Inclusion and exclusion criteria

An overview of the patient participant flow chart is shown in Figure 1. Patient inclusion criteria were: 1) 18 and 70 years old; 2) elective hip or knee arthroplasty with spinal anesthesia; 3) ability to provide informed consent. Exclusion criteria were: 1) history of or current hepatic or renal insufficiency; 2) BMI ≥ 33; 3) heart failure or active arrhythmias; 5) patients with severe systemic disease that is a constant threat to life; 6) contraindication or allergy to ondansetron ; 7) concurrent use of drugs known to prolong the QT-interval (such as thioridazine or quetiapine), and strong inhibitors of CYP450 enzymes (such as fluconazole or erythromycin); 8) patients who are pregnant or lactating.

Figure 1.

Participant flow chart

Study drug administration

An intravenous catheter was inserted in an arm for drug administration. Ondansetron was administered in a single 16 mg dose as a 15 min intravenous infusion. A 5-lead continuous electrocardiogram (ECG) monitoring was performed throughout the infusion and for approximately 30 min after the end of the infusion. Intraoperative and postoperative monitoring was performed per standard of care. No changes were made to the routine intraoperative management and hemodynamic physiological monitoring.

Data collection

An intravenous catheter was inserted in the arm opposite to the arm used for drug administration for obtaining blood samples. At baseline, a 5 mL blood sample was collected for pharmacokinetics (PK) and another 5 mL sample for genetic analysis. Six 5 mL serial venous blood samples were obtained from all subjects at 0 (pre-treatment), 15 (end-infusion), 30, 60, 120, and 180 min from the beginning of ondansetron infusion in each patient. The samples were collected into heparinized tubes, put on ice, and centrifuged for 10 min at 5000 RPM. Plasma was then transferred to two 1.5 mL vials and stored at −80°C until analysis by HPLC-UV. A single CSF sample of 4mL was collected from each patient immediately following spinal needle insertion and before administration of spinal anesthesia. The timing of ondansetron infusions was scheduled to allow for obtaining a single CSF sample 30–90 minutes after the beginning of the infusion. The CSF samples were transferred to 1.5 mL vials and stored at −80°C until analysis.

Analysis

Plasma and CSF sample preparation and analysis

The plasma and cerebrospinal fluid samples were assayed for ondansetron by high-performance liquid chromatography using Agilent Technologies 1260 Infinity HPLC-UV system by slight modification of reported methods [17 18]. In brief, 100 μL aliquot of plasma sample was used and 10 μL of a 100 μg/mL antipyrine solution added as internal standard. 200 μL of NaOH 1M solution was added as an alkalizing agent for plasma, and 200 μL of cold acetonitrile was added for protein precipitation. Solutions were vortexed at high-speed and extracted with 3mL of ethyl acetate. Samples were then centrifuged at 4°C 3900 RPM for 7 min, the organic phase was transferred to a fresh tube and evaporated under nitrogen gas. The assay for ondansetron concentration in CSF samples follows the same procedure, however 200 μL saturated sodium carbonate solution was used as the alkalizing agent. The samples were then reconstituted with 70:30 water:methanol mixture and a 50 μL aliquot was injected into the HPLC system. Separation was achieved using a Poroshell EC-C₁₈ column (Agilent Technologies, 4.6 × 100 mm, 2.7 μm). The mobile phase, consisted of 10 mM ammonium acetate (pH adjusted to 3.5 with glacial acetic acid):methanol (80:20), and the flow rate was set to 1.5 mL/min. The detection wavelength was 310 nm. The retention times of ondansetron and antipyrine were approximately 11 min and 4.6 min respectively. The detection limit of ondansetron in plasma and CSF samples were 10 and 5 ng/mL.

Genotyping

The DNA isolation and genotyping included analysis of single nucleotide polymorphisms (SNPs) in P-glycoprotein transporters and Organic Cation Transporter 1 (OCT1), which have been associated with altered response to ondansetron [15 19 20]. The following five P-glycoprotein (ABCB1) SNPs: C3435T, C1236T, G2677T, G1199A, and T129C, and four OCT1 (SLC22A1) SNPs: R61C, G456R, G401S, and C88R that have been reported to affect ondansetron pharmacokinetics or clinical antiemetic effect were evaluated. The genotyping was performed at Washington University Genome Technology Access Center. Whole blood samples were extracted using QiIAamp DNA mini Blood kit (QIAGEN). The DNA was quantified and quality controlled using nanodrop and gel readings. The SNPs were interrogated using Taqman probes (Applied Biosystems). Each SNP had its own 20 μl reaction well with a final concentration of 1X Taqman probe mix, 1X Taqman PCR master mix, and 20–40 ng of DNA. The samples were cycled and analyzed on CFX96 Real-time PCR Detection System (Bio-Rad). The collected blood and DNA samples were de-identified and coded to ensure patient confidentiality and HIPPA compliance.

Pharmacokinetic analysis

Given that the CSF samples largely were not obtained at the same time as the plasma sample, an estimated plasma concentration at CSF sampling time was determined. This estimated value was obtained by selecting two observed plasma concentrations: one immediately before the time of CSF sample and one immediately following the time of CSF sample. An exponential regression equation was used to calculate plasma concentration at the time of CSF sample.

A noncompartmental analysis was conducted on the plasma concentration-time profiles for each patient using Phoenix WinNonlin (v7, Certara, Princeton, NJ); volume of distribution at steady state (V_d,ss), total systemic clearance (CL_T), mean residence time (MRT), the half-life (t_1/2), and the area under the plasma concentration-time curve (AUC_0−∞, from time zero to infinity, using linear up log down approach) were calculated.

Population pharmacokinetic modeling

The plasma and CSF data were analyzed using NONMEM Version VII (ICON Development Solutions, Ellicott City, Maryland, USA). First-order conditional estimation method (FOCE) with interaction and ADVAN6 subroutine were utilized for all model runs. Post-processing was conducted with Pirana and R (R-project, www.rproject.org, version 3.3.1). Standard step-wise approach was used for population PK model building that included 1) construction of a base model for plasma PK based on plasma data only, 2) addition of a CSF compartment and CSF data, 3) assessment of inter-individual variability (IIV) in PK parameters, and 4) assessment of covariate effects for PK parameters. The model fitting process was guided by multiple criteria including successful model convergence, visual inspection of the model fits and standard diagnostic plots, and assessment of precision and accuracy of estimated model parameters.

Visualization of plasma concentration-time profiles for all patients indicated that a multicompartment model was needed to describe plasma data. Two-compartment model with linear elimination provided a good description of the data (and a three-compartmental model did not significantly improve model fit); and the model was parameterized in clearance and volume terms with systemic clearance from the central compartment (CL), inter-compartmental clearance Q, volume of the central compartment (V_C) and volume of the peripheral compartment (V_T). Model schematic and equations (1 and 2) used to describe the model are outlined in Figure 2. Upon establishing the base model for plasma pharmacokinetics, a CSF compartment was included to capture the disposition of ondansetron to the CNS. Two forms of the CSF model were explored: 1) using one or two first-order rate constants to describe the rates with which drug is entering and exiting the CSF compartment, and 2) using an equilibrium partition coefficient (K_P) for modeling distribution of ondansetron to the CSF. The final model used the K_P approach (Figure 2 and Equation 3). An exponential form was used for estimating inter-individual variability (IIV) on the PK parameters (Equation 4); and the final model included IIV for CL, V_C, and K_P terms. Covariance of inter-individual variability between systemic clearance and central volume was estimated. A constant coefficient of variation model was used to describe the residual random error (Equation 5). Since there was only a single CSF sample obtained from each patient, and the same bioanalytical assay was used for quantifying CSF and plasma concentrations, the residual variability for CSF samples was fixed to the value estimated from plasma model. Several demographic and clinical covariates were available, including total body weight, height, body mass index (BMI), sex, age, and creatinine clearance (calculated by Cockcroft-Gault equation). A stepwise covariate modeling (SCM) method was executed using Perl Speaks NONMEM (PsN) to identify covariates. A covariate was considered significant during a forward addition step if the objective value function was reduced by at least 3.84 (p-value < 0.05, degree of freedom (df) =1). In the backward elimination step an increase greater than 6.35 (p-value < 0.01) or more was required for a covariate to be retained in the model. The covariates were evaluated on parameters containing IIV terms (CL, V_C, K_P). The relationship between the parameter and continuous covariates were evaluated using linear, exponential, and power functions, and the categorical covariates were evaluated using linear parameterization. The final equation used to describe the covariate that was included into the model is shown in Equation 6.

Figure 2.

Schematic of the population PK model used to capture plasma pharmacokinetics and CSF distribution. Equations 1–3 present a two compartment model that was used to describe systemic disposition, where C₁ and C₂ are drug concentrations in the central and peripheral compartment; and the concentration in the C_CSF is described using an equilibrium partition coefficient (K_P). Other structural model parameters are systemic clearance CL, inter-compartmental clearance Q, volume of the central compartment V_C, and volume of the peripheral compartment V_T was used for modeling distribution of ondansetron to the CSF. Equation 4 exemplifies inclusion of inter-individual variability on pharmacokinetic parameters, where CL_i is the value of CL in ith subject, CL_TV is a typical value of CL in the population, and η_i is the inter-individual random effects for individual i for this parameter. Equation 5 describes the residual random error, where C_ij and C_pred,ij represent the jth observed and predicted concentrations for the ith individual respectively and ε_ij is the residual random effect. Equation 6 exemplifies inclusion of a covariate (COV) effect for a PK parameter, where the coefficient A and exponent B are estimated.

Results

Overall, 19 subjects met the study inclusion criteria and signed informed consent (Figure 1 - CONSORT flow diagram). Two subjects were further excluded due to abnormal baseline ECG, and one subject withdrew consent before the study day. In an additional subject, the surgery was cancelled about 30 minutes prior to the scheduled time, due to a previously undetected infected wound. The subject did receive the intravenous ondansetron infusion, but no CSF sample had been obtained as spinal anesthesia was not performed, and this subject was excluded. Participant demographics, including age, sex, BMI, height, weight, and creatinine clearance are depicted in Table 1. Serial plasma samples and a single CSF sample (timing for CSF sample is shown in Table 2) were collected from 15 subjects. Due to technical issues, the samples from one subject (#14) could not be reliably quantified, and were therefore excluded from pharmacokinetic analysis. Genotyping data of subjects for P-gp and OCT1 transporters, as well as the individual CSF to plasma concentration ratios of ondansetron are presented in Table 2.

Table 1.

Subject demographics

Patient ID	Age	Sex	BMI (kg/m2)	Height (cm)	Total BW (kg)	CLCR (mL/min)
1	54	F	31.6	168	89.1	98.2
3	59	M	27.4	178	86.8	132.0
5	56	M	27.2	180	88.2	84.4
8	63	F	21.4	183	71.8	98.9
9	66	F	27.4	168	77.3	110.7
10	70	M	24.4	183	81.8	74.3
11	59	F	28.1	173	84.1	-
12	60	F	28.5	170	82.3	-
13	45	M	26.2	185	90.0	127.7
14a	62	F	26.9	157	66.8	64.0
15	64	M	26.9	175	82.7	87.4
16	49	M	29.1	170	84.1	103.2
17	51	F	32.3	173	96.8	147.4
18	49	M	24.2	178	76.6	115.4
19	58	F	24.5	170	70.9	99.5
Mean (±SD)	57 (7)	M (50%)	27.1 (2.9)	175.3 (5.9)	83.1 (7.2)	103 (23)
Median	57		27.3	174	83.4	99.5
Mean (±SD) b	57 (7)	M (50%)	27 (2.8)	175.3 (5.6)	83.0 (6.9)	106 (20)
Median b	58		27.3	174	83.4	101

BMI = body mass index, BW = body weight, CL_CR = creatinine clearance, calculated

^{a -}Subject 14 was excluded from population pharmacokinetic analysis due to irregular plasma concentration profile

^{b –}Value excluding Subject 14

Table 2.

Individual CSF/plasma concentration ratio data, and P-gp and OCT1 genotype

Patient ID	CSF sampling after ondansetron (min)	CSF/Plasma ratio (at sampling time)	P-glycoprotein SNPs					OCT-1 SNPs
			C3435T	C1236T	G2677T	G1199A	T129C	R61C	G465R	G401S	C88R
1	47	0.077	GA	GA	CC	CC	AA	TC	GG	CC	TT
3	80	0.147	AA	AA	N/A	CC	AA	CC	GG	CC	TT
5	65	0.132	GG	GG	CC	CC	AA	CC	GG	CC	TT
8	56	0.170	GA	GA	CC	CC	AA	CC	GG	CC	TT
9	54	0.167	AA	AA	N/A	CC	AA	CC	GG	CC	TT
10	47	0.077	GA	GA	CC	CC	AA	CC	GG	CC	TT
11	58	0.128	GA	GA	CC	CC	AA	CC	GG	CT	TT
12	76	0.138	GA	GA	CC	CC	AA	CC	GG	CC	TT
13	39	0.144	GA	GA	CC	TC	AA	CC	GG	CC	TT
15	83	0.124	GA	GG	TC	CC	AA	CC	GG	CC	TT
16	75	0.212	GA	GA	CC	CC	AA	CC	GG	CC	TT
17	37	0.264	AA	AA	N/A	CC	AA	CC	GG	CC	TT
18	54	0.131	GA	AA	CC	CC	AA	TC	GA	CC	TT
19	59	0.227	GA	GA	CC	CC	AA	CC	GG	CC	TT

Population Pharmacokinetic Model

For initial data assessment, noncompartmental analysis of individual plasma concentration-time profiles in all subjects was conducted and the results are presented in Table 3. Calculated parameters in this study are similar to published previously values [22].

Table 3.

Noncompartmental analysis of human plasma pharmacokinetics for ondansetron (n=14)

Parameter	Mean ±SD
t1/2 (h)	4.9 ± 2.3
MRT (h)	6.9 ± 3.3
AUC0−∞ (h·μg/L)	625 ± 296
Vd,ss (L)	185 ± 50
CL (L/h)	33 ± 20

SD – standard deviation; t_1/2 – half-life; MRT – mean residence time; AUC0_−∞- AUC from time 0 to infinity;

V_d,ss – volume of distribution at steady-state, CL – total clearance

Pharmacokinetic data were further analyzed using mixed-effect (population) modeling approach. Two-compartment distribution model with linear elimination was selected to describe plasma concentration-time profiles from 14 subjects. Good description of the experimental data was obtained, and all parameters were estimated with sufficient precision. Expanding the model to three compartments did not improve model fits. Individual and population model fits along with observed plasma data are shown in Figure 3. Final model estimated parameters are shown in Table 4. Inclusion of inter-individual variability improved model fits; and it was estimated to be 49% for CL and 42% for V_C. Inter-individual variability on peripheral volume and intercompartmental clearance could not be estimated with sufficient precision and were not included in the final model. Multiple covariates were evaluated on various model parameters based on criteria specified in the Methods. Inclusion of subject’s age as a covariate on V_C was found to significantly improve the model. It was determined that with increasing age, the central volume term decreases, with an estimated exponent of −4.91 (Table 4).

Figure 3.

Concentration-time profiles of ondansetron in plasma for 14 subjects included in the population pharmacokinetic analysis. The model provided a good description of the experimental data.

Table 4.

Final parameter estimates from population pharmacokinetic model for ondansetron human pharmacokinetics

Population Pharmacokinetic Parameters
Parameter	Description	Estimate	RSE%	Shrinkage (%)
CL (L/h)	Systemic clearance	24.6	17
VC (L)	Volume of the central compartment	63.3	14
Q (L/h)	Inter-compartmental distribution clearance	211	7
VT (L)	Volume of the peripheral tissues compartment	107	9
KP	CSF/Plasma partition coefficient	0.145	9
B for (VC, AGE)	Exponent for covariate effect of age on VC	−4.91	19
Inter-individual Variability (CV%)
IIV CL		49.7%	30	2
IIV VC		41.9%	23	2
IIV KP		23.7%	33	18
Residual Variability (CV%)
RVplasma	Residual error for plasma concentration	19.1%	1
RVCSF	Residual error for CSF concentration	-a	-a

RSE% - relative standard error in %; CV% - coefficient of variation in %; IIV – estimate for the variance of the inter-individual variability.

^{a -}residual variability value was fixed in the final model run to allow for estimation of inter-individual variability for K_P

Initially, a model that included a differential equation with one or two rate constants was attempted to describe ondansetron disposition into the CSF. Reasonable description of experimental observations in the CNS was obtained as shown in Figure 4. However, due to a limited number and time frame of CSF observations, inter-individual variability could not be reliably estimated in such a model. Current data suggest a fast equilibration between plasma and CSF; therefore, it was assumed that concentrations in the CSF could be described using an equilibrium partition coefficient K_P. The population typical value was estimated as 0.145 for K_P (with %RSE of 9%), and the inter-individual variability for K_P was estimated to be 23.7% (Table 4). The average CSF:plasma ratio of ondansetron was 0.15 (range 0.08 and 0.26), with individual values presented in Table 2. Exploratory data analysis did not reveal any effect of individual genotype for P-gp or OCT-1 on partition of ondansetron to the CSF; however the number of study participants was small, therefore, genotype data were not included as covariates for CSF:plasma partition coefficient Kp. The Individual CSF:plasma concentration ratios of the study subjects as a function of P-gp and OCT-1 polymorphisms are presented in Appendix Figure 1.

Figure 4.

Concentration-time profiles of ondansetron in cerebrospinal fluid for 14 subjects included in the population pharmacokinetic analysis. The model provided a good description of the experimental data.

No adverse effects were reported during ondansetron infusion or post-administration monitoring. No ECG or vital sign changes had been observed with the continuous monitoring during, and for 30-minutes after the ondansetron infusion. The surgical and postoperative course was normal in all study subjects.

Discussion

Our experimental approach in combination with population PK modeling provided important quantitative information for ondansetron distribution into the CNS using a single CSF sample at variable times from each participant. The study demonstrated that after iv administration, the mean CSF to plasma partition coefficient of ondansetron is approximately 0.15; and individual plasma concentration-time profiles of ondansetron and partition to the CSF were captured well by the developed model.

While intrathecally administered ondansetron has shown antinociceptive effects in animal models of neuropathic pain due to its pharmacologic activity as a 5-HT₃R antagonist, clinical studies have shown mixed results with systemic ondansetron alleviating neuropathic pain in patients [10 11]. Ondansetron is capable of blocking 5-HT₃ receptors in vitro in low nanomolar concentrations [23]. Most rodent studies have used the administration of 1–2 mg/mL ondansetron solution intrathecally in 25–100 μg doses, potentially creating micromolar local concentration of ondansetron around the spinal cord [6 24]. We hypothesized that the lack of clinical translation from preclinical studies can be, at least in part, attributable to the challenge in achieving target drug concentrations within the CNS following systemic administration to humans.

Ondansetron plasma pharmacokinetics have been documented in the literature, however very limited information has been published describing CNS disposition, and the quantitative relationship between plasma and CNS disposition. Understanding the required therapeutic concentrations to achieve effects within the CNS is challenging, especially given the difficulty of obtaining direct samples for analysis. CSF has been proposed as a surrogate sample to represent the disposition of drugs in the CNS to build a quantitative understanding of drug concentrations in the brain [26 27]. Our measured and modeled results demonstrated approximately 7-fold lower concentration of ondansetron in the CSF compared to plasma, suggesting relatively poor CSF penetration. The CSF concentrations of ondansetron, 1–2 hours after IV administration were around 0.01–0.02 mg/L (i.e. 35–70 nM). If achieving effective analgesic concentrations of ondansetron in the spinal cord require CSF concentrations in excess of the ~50 nM concentrations observed in the current study (compared with ~350 nM peak plasma concentrations), then the systemic administration of ondansetron might not allow for appropriate testing of its analgesic effects in neuropathic pain.

To quantitatively describe the time-course of drug concentrations in the CNS we developed a population PK model that simultaneously described systemic disposition and CSF distribution of ondansetron. A classical PK analysis would have required multiple CSF samples from each subject, which is hardly feasible during surgery. Furthermore, to perform a standard statistical analysis of population variability samples had to be taken at the exact same times from all patients. The advantage of the population modeling approach is in simultaneous description of the mean population tendency in pharmacokinetic parameters, inter-individual variability, and effect of covariates. It also allows for conducting simulations (and extrapolations) beyond experimental conditions of the study (as long as model assumptions are believed to hold). Our study demonstrates that the population modeling is a useful approach, and allows for reliable estimation of individual and population averaged CNS partition based on sparse sampling data informed by plasma pharmacokinetics as shown in Figure 5. The model indicated that partition of ondansetron to the CSF is relatively rapid. These results are supported by our preclinical study that showed that the highest concentration of ondansetron in rat CNS after intravenous dosing was achieved at the first sampling point – 10 minutes (unpublished data).

Figure 5.

Concentration-time profile of ondansetron in plasma for all 14 subjects (open circles) are overlaid with model-based predictions for each individual predictions (black dotted line) and plotted against population predicted (solid red line).

In a previously published population PK model, both weight and age were identified as significant covariates improving the description of clinical data for 99 patients [25]. Our work also identified age as an important model covariate. This difference in parameters may be attributed to the limited sample size of our study compared to the previous report, which included a large population and a wider range of patients for each covariate screened. An important advancement in the current population model is linking plasma disposition to a CNS disposition using a single partition-coefficient term (K_P). This will allow for guiding the design of future studies assessing the effect of ondansetron on neuropathic pain.

The study has potential limitations. Our cohort was relatively small, and only a single CSF sample was obtained per subject to measure ondansetron concentration. A larger patient cohort would enable a more robust understanding for the CSF distribution and elimination and assessment of transporters polymorphism on PK parameters. Furthermore, only a limited time range of CSF sampling was included in the study, and a longer sampling would be required to fully understand the CSF distribution and elimination of ondansetron. Sampling of the CSF at earlier time points may help identify the initial distribution rates and exposure. Only subjects 45–70 years of age and BMI of 21–33 kg/m² were enrolled, which may limit the extrapolation of the results to populations with age and BMI outside these ranges. Our study has notable strengths, including rigorous screening criteria to ensure patient safety, CSF concentration measurement, and population PK modeling. Patients from this surgical population were selected (in lieu of healthy volunteers), as these patients receive intrathecal anesthesia as a standard of care, allowing CSF access, which eliminates any potential risks associated with performing a lumbar puncture for CSF collection in healthy volunteers.

Conclusion

The results suggest relatively limited CSF penetration of ondansetron after a single 16 mg intravenous dose. The population PK modelling is a valuable approach for studying CNS disposition of drugs using sparse experimental sampling. A more detailed investigation with repetitive CSF collection and pharmacodynamics endpoints will be required for further understanding of CNS distribution kinetics of ondansetron and its potential analgesic effects in humans.

What is already known:

• The overexpression of serotonergic receptors in the central nervous system has been implicated in neuropathic pain.

• Ondansetron, a 5-HT₃ receptor antagonist, has demonstrated pharmacological potential in alleviating thermal and mechanical hypersensitivity in preclinical animal models of nerve injury.

• Central nervous system distribution of 5-HT₃ receptor antagonist may be dependent on and limited by expression levels of efflux transporters at the blood-brain barrier, such as P-glycoprotein.

What this study adds:

• Ondansetron concentrations in the CSF were approximately 7-fold lower than in plasma

• Population pharmacokinetic successfully captured individual CSF and plasma ondansetron concentration time profiles

Data Availability Statement:

The data that support the findings of this study are available on request from the corresponding author, upon submission of a methodologically sound research proposal. No patient-identifying fields (including dates) will be included in the shared dataset. The data are not publicly available due to privacy or ethical restrictions.