Pharmacokinetic profiling of imatinib in relation to CYP3A4 activity in leukaemia patients

Abstract

Aim

Imatinib pharmacokinetics exhibit large interindividual variability because of differences in CYP3A4 activity—the main imatinib‐metabolizing enzyme. While therapeutic drug monitoring is effective, it requires steady‐state conditions and frequent sampling. The endogenous biomarker 4β‐hydroxycholesterol (4‐β‐OHC) has been proposed to reflect CYP3A4 activity, enabling proactive dose optimization. This study investigated 4‐β‐OHC as a CYP3A4 biomarker in leukaemia patients and its relationship with imatinib pharmacokinetics

Methods

A pharmacokinetic study included 19 patients with chronic myeloid leukaemia or acute lymphoblastic leukaemia receiving imatinib. Blood samples were collected over 24 h. Imatinib, N‐desmethyl imatinib, 4‐β‐OHC and 4α‐hydroxycholesterol (4‐α‐OHC) plasma concentrations were quantified by liquid chromatography–tandem mass spectrometry (LC–MS/MS), while total cholesterol was measured by clinical chemistry autoanalyser. CYP3A4 activity was estimated using the 4‐β‐OHC/cholesterol ratio. Genotyping of relevant CYP3A4/5 gene variants was carried out. Pharmacokinetic parameters were estimated using PKanalix software. Simple and multiple linear regressions were performed

Results

Nineteen patients were included with median age of 66 years (range 44–77). Pharmacokinetic parameters were comparable to prior reporting. The 4‐β‐OHC/cholesterol ratio showed a significant positive association with imatinib clearance, explaining 30% of variance. No significant association was found between CYP3A4 activity and N‐desmethyl‐imatinib area under the concentration time–curve (AUC)/imatinib AUC ratio. Single‐point sampling of imatinib at 3–4 h post‐dose correlated better with AUC_{0 –24h} than trough concentrations. CYP3A4/5 genotype assessment was limited by low patient numbers

Conclusion

This is the first study investigating 4‐β‐OHC as a biomarker for CYP3A4 activity in imatinib‐treated leukaemia patients, demonstrating clinically meaningful associations with imatinib clearance and supporting further development of this personalized therapy approach.

What is already known about this subject

• Imatinib shows high variability in pharmacokinetics, largely driven by CYP3A4 activity.

• TDM improves treatment outcomes but requires steady state and frequent sampling, limiting its practicality.

• 4β‐hydroxycholesterol (4‐β‐OHC) reflects CYP3A4 activity, but its role in imatinib‐treated leukaemia patients has not been explored.

What this study adds

• The 4‐β‐OHC/cholesterol ratio explains 30% of imatinib clearance variability, confirming its biomarker potential.

• The biomarker is stable over 24 h, supporting feasibility for clinical use.

• Single samples at 3–4 h post‐dose outperform trough levels for estimating imatinib exposure.

1. INTRODUCTION

Imatinib is a selective tyrosine kinase inhibitor (TKI) used as treatment for chronic myeloid leukaemia (CML) and Philadelphia chromosome‐positive (Ph+)/‐like (Ph‐like, ABL‐class) acute lymphoblastic leukaemia (ALL). Its therapeutic efficacy and safety profile are influenced by its pharmacokinetics, which are largely determined by metabolic pathways involving cytochrome P450 enzymes, particularly CYP3A4 and CYP3A5. Imatinib is highly suitable for therapeutic drug monitoring (TDM) because of its pharmacokinetic properties, particularly a large inter‐individual but low intra‐individual pharmacokinetic variability. ¹ Besides high inter‐individual variability in CYP3A4 activity, variability in imatinib exposure is caused by a range of factors, including concomitant drugs, genetic polymorphisms and adherence patterns. ² , ³ Additionally, multiple TDM studies have demonstrated considerable clinical advantages by establishing the relationship between plasma imatinib trough concentrations >1000 ng/ml and improved treatment responses in adult CML patients. ⁴ , ⁵

However, TDM requires patients to reach steady‐state conditions, which are dependent on the half‐life of the drug and are typically achieved after approximately 4–5 half‐lives; for imatinib, steady state is usually reached within 1 week after treatment initiation or dose adjustments. ¹ , ⁶ A biomarker of CYP3A4 activity could provide immediate insights into individual metabolic capacity, enabling proactive dose optimization from treatment initiation rather than reactive adjustments based on TDM results. Such a biomarker would be particularly valuable for identifying patients at risk of suboptimal exposure or drug‐related toxicity, and for guiding dosing decisions in the presence of potential drug–drug interactions that could affect CYP3A4 activity.

To estimate an individual's CYP3A4 enzyme activity in vivo, a common approach is to administer a CYP3A4 probe drug and measure the resulting levels of the parent compound and its CYP3A4‐dependent metabolite in plasma. Midazolam is recommended as the gold standard for characterization of CYP3A phenotypes by the United States Food and Drug Administration as well as the European Medicines Agency. ⁷ To avoid the potential safety risk associated with administration of such probes, and for both economic and ethical reasons, using endogenous biomarkers of CYP3A4 activity would be of clinical value. ⁸ The cholesterol derivative 4β‐hydroxycholesterol (4‐β‐OHC), which is formed by CYP3A4‐mediated metabolism, appears to be stable with a half‐life of 17 days and exhibits low intra‐individual variability. ⁸ Baseline values of 4‐β‐OHC have been reported to range from 10 to 100 ng/ml and increase with increasing CYP3A4 activity. ⁸ , ⁹ To account for individual variations in cholesterol levels, the normalized ratio between 4‐β‐OHC and the total amount of cholesterol in plasma has been proposed to reflect CYP3A4 enzymatic activity and can be utilized to assess the extent of drug‐induced enzyme induction or inhibition. ⁸ The majority of clinical studies on 4‐β‐OHC have been conducted on healthy volunteers or with non‐antineoplastic therapeutics, primarily focussing on CYP3A‐inducing drugs, as presented in several reviews. ⁷ , ⁸ , ¹⁰

The isomer of 4‐β‐OHC, 4α‐hydroxycholesterol (4‐α‐OHC), is formed by auto‐oxidation of cholesterol and not by enzyme activity. Increasing levels of 4‐α‐OHC have formerly been suggested to reveal uncontrolled handling of sample or storage conditions, making it a marker of stability of clinical samples ⁹ ; however, reporting of 4‐α‐OHC is not standard in 4‐β‐OHC studies. ⁸

The aim of this study was to investigate the utility of 4‐β‐OHC as a biomarker of CYP3A4 activity in leukaemia patients and to examine its relationship with imatinib pharmacokinetics. Additionally, we aimed to assess CYP3A4 enzyme activity associated with different CYP3A4 and CYP3A5 gene variants and evaluate the optimal sampling time points for TDM in this population.

2. MATERIALS AND METHODS

The study was approved by the Scientific Ethics Committee, Capital Region of Denmark (H‐22057870). Written informed consent was obtained from all patients prior to enrolment. The study was conducted from June 2024 to August 2024 at Rigshospitalet, Copenhagen University Hospital, Copenhagen, Denmark, in accordance with the Helsinki Declaration (2024).

Drug and molecular target nomenclature follows the International Union of Basic and Clinical Pharmacology (IUPHAR)/British Pharmacological Society (BPS) Guide to PHARMACOLOGY recommendations. ¹¹ Compounds not listed in the database are reported using nomenclature consistent with established literature usage. Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, and are permanently archived in the Concise Guide to PHARMACOLOGY 2021/2022. ¹² Key targets discussed in the study belong to the enzyme family (cytochrome P450 enzymes) and the transporter family. ¹³ , ¹⁴

2.1. Patients

The patients were selected from a larger cohort of CML and ALL patients enrolled in an ongoing longitudinal pharmacokinetic/dynamic study of imatinib conducted at the same department. Eligibility for the present sub‐study was restricted to adult patients, and an attempt was made to achieve a balanced representation across sex and age groups. The patients were identified from the pharmacokinetic/dynamic study inclusion list in sequential order and approached consecutively until approximately 20 consenting participants were obtained. A total of 27 patients were invited to participate; eight patients declined because of practical issues or needle anxiety. Nineteen adult patients were ultimately enrolled (18 CML and 1 Ph+ ALL). The patient with Ph+ ALL had completed primary chemotherapy blocks years prior and was receiving imatinib monotherapy at the time of study enrolment.

The patients' medication list was verified directly with each patient to ensure that only actively used medications were recorded. As needed (P.N.) medications were included only, if they had been taken within the week preceding study entry. All concomitant medications were reviewed and classified according to their CYP3A4 interaction profile using the Flockhart interaction table. ¹⁵

To ensure steady‐state conditions at study Day 1, all patients were contacted by phone at least 5 days prior to confirm consistent daily imatinib intake upon study initiation. Some patients needed to modify their usual dosing schedule to accommodate the practical requirements of the study design. When necessary, the patients received daily SMS reminders regarding these dosing time adjustments.

The authors confirm that the PI for this paper is Christen Lykkegaard Andersen and that he had direct clinical responsibility for patients.

2.2. Sample collection and storage

Each patient's regularly prescribed dose of imatinib was administered orally at study initiation. Blood samples were collected according to the following schedule: Half of the patients (n = 10) had samples obtained at 0.5 and 1.0 h post‐dose, while the other half had samples collected at 1.0 and 1.5 h post‐dose to explore any differences in absorption. All patients provided samples at 2, 3, 4 and 24 h post‐dose. Additionally, as patients could choose their preferred time for starting the dosing and sampling protocol, six patients had a 5‐h sample, one patient had a 6‐h sample, and one patient had a 16.5‐h sample. Analysis of 4‐α‐OHC and 4‐β‐OHC was performed on the first and last samples from each patient. Blood samples for covariates (liver enzymes, kidney function and so forth) were obtained at the time of the first sample for each patient. All samples were taken under steady‐state conditions.

Blood samples (6 ml) were collected by either vein puncture or via an indwelling cannula placed in a forearm vein using blood sampling tubes containing K2‐ethylenediaminetetraacetic acid (EDTA) anticoagulant. Samples were centrifuged at 3220g (4000 rpm) in 10 min at room temperature (20–23°C). After centrifugation, plasma was transferred into polypropylene tubes before storage at −80°C until analysis. One additional sample of 9 ml whole blood per patient was kept in the freezer of −20°C for subsequent analysis of gene polymorphisms.

2.3. Analysis of imatinib and metabolite

The plasma concentrations of imatinib and metabolite N‐desmethyl‐imatinib were analysed at the Pediatric Oncology Research Laboratory (Bonkolab), Rigshospitalet, Copenhagen University Hospital. After thawing, the samples were added stable isotope‐labelled internal standards before extracting the analytes using ethyl acetate. The organic fraction was then transferred to a fresh vial and evaporated. The dried samples were redissolved, after which concentrations of imatinib and N‐desmethyl‐imatinib were measured by reversed phase liquid chromatography–tandem mass spectrometry (LC–MS/MS).

The LC–MS/MS analysis employed a slightly modified version of the method developed by Bouchet et al. ¹⁶ The system consisted of a Sciex Qtrap 6500+ mass spectrometer integrated with a Sciex ExionLC ultra‐high‐performance liquid chromatography (UHPLC) system (Sciex, Framingham, MA). Separation was performed using an Acquity UPLC™ HSS T3 column (2.1 × 50 mm, 1.8 μ) with an Acquity UPLC™ HSS T3 VanGuard pre‐column (2.1 × 5 mm, 1.8 μ) for protection, both sourced from Waters (Wexford, Ireland). Lower limit of quantification (LLOQ) for imatinib was 94 ng/ml, determined based on the lowest calibration level fulfilling accepted bioanalytical criteria for accuracy (±20%) and precision (≤20% coefficient of variation [CV]) across two independent analytical days. To limit carryover effects, a blank sample was injected between all patient samples and calibrators. The LLOQ for N‐desmethyl‐imatinib was 50 ng/ml, determined using the same criteria. The limits of detection (LOD), determined empirically using a signal‐to‐noise ratio >3, were ≤74.1 ng/ml for imatinib and ≤50 ng/ml for N‐desmethyl‐imatinib. All plasma samples were analysed in duplicate, and reported concentrations represent the mean of the two measurements.

2.4. Analysis of CYP3A4 activity

4‐α‐OHC and 4‐β‐OHC concentrations were quantified in plasma samples at the Department of Public Health, University of Southern Denmark, by liquid chromatography and high‐resolution mass spectrometry (LC/HR‐MS) according to the method described by Hautajärvi et al. ⁹ with minor modifications. The utilized LC/HR‐MS system consisted of a Thermo Vanquish UHPLC system (a Dionex Ultimate 3000 RS pump, a Split Sampler FT autosampler and a column compartment) connected to a Q‐Exactive Orbitrap high‐resolution mass spectrometer with a heated electrospray ionization (H‐ESI) interface (Thermo Fisher Scientific Inc., Waltham, MA). Both analytes and the internal standard (d₇‐4β‐hydroxycholesterol) were quantified from the exact mass of the sodium adduct [M + Na⁺]. The calibration curve ranged from 0.2 to 500 ng/ml. More details are described in the Supporting Information.

The within‐batch and between‐batch variation (CV%) of the QC samples was less than 8.1% for 4‐α‐OHC and less than 5.4% for 4‐β‐OHC. The accuracy determined as bias was less than 10.5% for 4‐α‐OHC and less than 9.5% for 4‐β‐OHC. LLOQ was 1.0 ng/ml for 4‐α‐OHC and 4‐β‐OHC.

The cholesterol concentration in plasma was determined using a Thermo Scientific cholesterol kit on a Konelab 20 Clinical Chemistry Autoanalyser (Thermo Scientific, Vantaa, Finland). Calibrators, blanks and QC samples for normal values and abnormal (high) levels (Nortrol and Abtrol) were included in the batch of samples analysed. The within‐ and between‐batch variation (CV%) of the QC samples was less than 1%. The accuracy determined as bias was −2% for both types of QC samples. LLOQ was 0.1 mmol/l (40 μg/ml).

To account for variations in plasma cholesterol concentrations, CYP3A4 activity was expressed as the ratio of 4‐β‐OHC to cholesterol. Cholesterol concentrations were converted from mg/ml to ng/ml, and the ratio was calculated as 4‐β‐OHC (ng/ml)/cholesterol (ng/ml). For consistency with previously published work, the ratio was multiplied by 10⁴ and reported as 4‐β‐OHC/cholesterol (ng/ng × 10⁴).

2.5. Pharmacogenetic analysis

CYP3A4 and CYP3A5 genotyping was performed to investigate potential associations with imatinib pharmacokinetics and CYP3A4 activity. Genotyping was performed at the Department of Clinical Genetics, Rigshospitalet, Copenhagen University Hospital.

DNA was extracted from whole blood using the QIAsymphony® DSP DNA Midi Kit (96) (Qiagen, cat.no.: 937255) according to manufacturer's instructions. Samples were quantified using a NanoDrop 1000 Spectrophotometer (Thermo Fisher).

Genotyping was performed using the TaqMan® OpenArray® PGx (Pharmacogenomics) Express Panel for QuantStudio™ 12K Flex (cat.no.: 4488847, Thermo Fisher) according to manufacturer's instructions. The OpenArray® chips were loaded using the Accufill System (Thermo Fisher). Samples were run in duplicates as were a no‐template control (NTC). The output was analysed using the QuantStudio® 12K Flex Software v.1.2.4 (Applied Biosystems) and TaqMan® Genotyper v.1.5.0. The genotyping assay detected the following CYP3A4 variants: *1B, *2, *3, *12, *17 and *22; and CYP3A5 variants: *2, *3/*10, *3B, *6, *7, *8 and *9. Detailed variant annotations are provided in Tables S2 and S3).

2.6. Pharmacokinetic analysis

Individual pharmacokinetic parameters (area under the concentration–time curve [AUC_0–24h], apparent oral clearance at steady‐state [CLss/F], maximum concentration [C _max], time to maximum concentration [T _max], and terminal half‐life [T _½]) were estimated for both imatinib and the major metabolite using the software PKanalix (SimulationsPlus 2025).

AUC from time zero to the last measurable concentration time point (AUC_0‐tlast) was calculated using PKanalix's standard linear‐up/log‐down trapezoidal method. For each time interval, the linear trapezoidal rule was applied during increasing concentrations, and the logarithmic trapezoidal rule during decreasing concentrations.

Because samples were collected over a full 24‐h dosing interval at steady state, CLss/F was calculated as dose/AUC_0–24h, consistent with non‐compartmental steady‐state conditions.

T _½ was calculated from the elimination rate constant (Kel), which was estimated by log‐linear regression of the terminal elimination phase using ≥3 quantifiable data points and PKanalix's automatic R ²‐based selection of terminal points. T _½ was then calculated as ln2/Kel. For three patients for whom Kel could not be estimated because of insufficient terminal data, T _½ was instead estimated using the relationship T _½ = ln2 × Vd/CL, where clearance (CL) was available from pharmacokinetic analysis and volume of distribution (Vd) was obtained from published literature (imatinib Vd = 2.4–4.2 l/kg ¹⁷ ). This approach resulted in a range of T _½ values reflecting the reported Vd variability. For the metabolite, T _½ could not be estimated for two patients because of the absence of literature‐reported Vd values combined with insufficient elimination phase data.

Summary statistics (median values) were calculated for the study population.

2.7. Statistical analysis

No formal sample size calculation was conducted given the exploratory design of this study. Demographic parameters are presented by the median, as well as minimum and maximum ranges and interquartile ranges, or as counts with percentages. Patient age is reported at the time of study initiation.

CYP3A4 activity was estimated using the 4‐β‐OHC/cholesterol ratio. Given the reported half‐life of 4‐β‐OHC of 17 days, the 24‐h sample was used for all regression analyses to maintain consistency in sampling timing across patients. As a supplementary assessment, stability of the 4‐β‐OHC/cholesterol ratio over the 24‐h study period was evaluated using Wilcoxon signed‐rank test comparing first and last samples.

Primary outcome was the metabolite AUC/imatinib AUC ratio, calculated as N‐desmethyl‐imatinib AUC_{0– 24h}/imatinib AUC_{0 – 24h}. Simple linear regression was performed to assess univariate associations between metabolite AUC/imatinib AUC ratio and the following covariates: 4‐β‐OHC/cholesterol ratio, leukocyte count, haemoglobin, albumin, age, body weight, and concomitant moderate CYP3A4 inhibitor use (binary variable). Multiple linear regression was then performed including all seven covariates simultaneously. Results are presented with regression coefficients, corresponding standard errors and adjusted R ² statistics. The covariates were chosen to reflect disease status, hepatic function, elimination capacity and potential drug–drug interactions known to influence imatinib pharmacokinetics. Because of the limited sample size, the number of covariates was restricted to minimize overfitting.

Secondary analyses included (1) exploratory assessment of the relationship between 4‐β‐OHC/cholesterol ratio and imatinib clearance using simple linear regression, and (2) evaluation of imatinib concentrations at individual time points for predicting AUC_{0 – 24h} using simple linear regression at each sampling time. Results of the time point regression modelling are presented with regression coefficients, corresponding 95% confidence intervals and adjusted R ² statistics.

To assess any difference in 4‐β‐OHC, 4‐β‐OHC/cholesterol ratio and total cholesterol between patients receiving statin treatment and those not receiving statin treatment, Wilcoxon rank‐sum tests were performed, including stratified analysis of linear regression.

A p < .05 was considered statistically significant. All statistical analyses were carried out using RStudio version 2024.04.2 with packages dplyr and ggplot2.

3. RESULTS

A total number of 19 patients (aged 43–77 years; 58% female) were enrolled in the study. The patients were diagnosed with either CML (n = 18) or Ph+ ALL (n = 1) and were treated at the Department of Haematology, Rigshospitalet, Copenhagen University Hospital. Median time since diagnosis was 2.3 years with 1 year since the latest dose change, indicating steady‐state conditions and stable disease. More than 70% of the patients had a BMI > 25, therefore categorized as overweight or obese. Biochemistry values, including kidney and liver function and total cholesterol levels, were within normal ranges. Baseline demographics and disease characteristics are presented in Table 1.

TABLE 1.

Baseline demographics and disease characteristics.

Characteristics	N = 19
Age at study initiation, years in median (min, Q1, Q3, max)	66 (44, 55.5, 67, 77)
Time since diagnosis, years	2.3 (0.5, 1.4, 9.4, 21.5)
Time since latest dose change, years	1 (0.1, 0.4, 1.3, 5.2)
Sex, N (%)
Female	11 (57.9)
Male	8 (42.1)
Diagnosis, N (%)
Chronic myeloid leukaemia	18 (94.7)
Acute lymphoblastic leukaemia	1 (5.3)
BMI (kg/m2), N (%)
Normal weight (BMI 18.5–24.9)	5 (26.3)
Overweight (BMI 25–29.9)	9 (47.4)
Obese (BMI ≥ 30)	5 (26.3)
Imatinib dose (mg), median (min, Q1, Q3, max)	300 (100, 200, 400, 600)
Haemoglobin (mmol/l)	7.9 (6.4, 7.6, 8.6, 9.2)
Leukocytes (×109/l)	5.8 (3, 5, 7.6, 9.6)
Thrombocytes (×109/l)	229 (150, 190, 253, 385)
Creatinine (micromol/l)	88 (60, 78, 104, 158)
eGFR/1.73 m2 (ml/min)	64.5 (39, 56.5, 80.5, 90)
Albumin (g/l)	39 (35, 36.5, 41.5, 44)
ALT (U/L) a	22 (9, 16.5, 29.5, 59)
Total cholesterol (mmol/l)	4.8 (2.8, 4.3, 5.6, 8)

Abbreviations: ALT: Alanine aminotransferase, BMI: body mass index, eGFR: estimated glomerular filtration rate.

^a One sample haemolysed but the same patient's previous ALT from 2 weeks prior was within the normal range.

3.1. Imatinib pharmacokinetics

Individual and median pharmacokinetic parameters are presented in Table 2. Median C _max of imatinib was 1661 ng/ml with median T _max of 2 h, while median metabolite C _max concentration was approximately 25% of the concentration of the mother drug and with corresponding half‐life of 31.5 h. Half‐life values for three patients were calculated using an alternative method because of insufficient sampling points for standard elimination rate analysis (T½ range: 6.7–23.3 h, estimated using literature Vd values of 2.4–4.2 l/kg).

TABLE 2.

Pharmacokinetic parameters for imatinib (I) and the active metabolite desmethyl‐imatinib (DI).

Patient	Dose (mg)	AUC0–24h (ng/ml⋅h)		CLss/F (ml/h)		C max (ng/ml)		T max (h)		T ½ (h)
		I	DI	I	DI	I	DI	I	DI	I	DI
1	400	20 249	6584	19 762	60 780	1661	475	1	2	12.6	17.3
2	400	34 186	10 892	11 701	36 723	2292	594	2	2	18.2	27.7
3	300	23 510	6765	12 754	44 313	1576	413	2.1	2.1	11.2	16.1
4	300	25 531	8758	11 750	34 254	1676	482	4	4	13.3–23.3 a	‐
5	300	22 803	6382	13 146	46 957	1321	360	2	2	19.8	40.8
6	400	22 238	9989	17 988	40 045	1226	493	4	4	6.7–11.7 a	‐
7	300	32 483	6640	9208	44 997	1958	396	2	1.5	18.2	40.8
8	200	11 307	4389	17 674	45 518	741	256	1	1	20.4	46.2
9	600	42 471	9993	14 145	60 129	2595	598	4	3.1	11.4–20 a	22.4
10	100	19 756	4012	5056	24 886	1266	212	2	2	16.1	33.0
11	200	13 673	4344	14 534	45 611	1029	253	1.5	1.5	16.5	40.8
12	400	18 374	7183	21 643	55 264	1054	354	3	3	18.2	46.2
13	200	29 959	12 758	6659	15 628	1748	679	1.1	1.5	18.7	31.5
14	300	36 956	7658	8111	39 134	2270	420	2	2	23.1	40.8
15	500	30 144	12 385	16 504	40 173	1654	710	3	2	23.9	23.1
16	300	35 412	12 075	8458	24 792	2427	673	2	2	15.4	26.7
17	200	23 849	7534	8445	26 769	1676	500	1	2	14.7	19.3
18	400	24 752	8948	16 168	44 732	1624	513	3	3	14.7	30.1
19	300	24 848	6019	12 160	50 329	1760	386	1	1	17.3	36.5
Median	300	24 752	7534	12 754	44 313	1661	475	2	2	17.8	31.5

Abbreviations: AUC: area under the concentration–time curve, CLss/F: apparent oral clearance at steady state, C _max: maximum concentration, DI: N‐desmethyl–imatinib, I: imatinib, T _max: time to maximum concentration, T _½: half‐life. Median of T _½ is without patient 4, 6 and 9.

^a T _½ for patient 4, 6 and 9 is presented in range.

Individual plasma imatinib concentration curves are presented in Figure 1. Each colour represents a different dosing level. Peak level occurs around 2–4 h after dosing (reproducing data from Table 1). The dose–concentration relationship is evident revealing higher doses (red and orange lines) generally produce higher plasma concentrations. The plot demonstrates the individual variability between patients within the same dose group (e.g. green group dosed with 300 mg) both in terms of C _max and trough concentration. In Figure 1B, the data are presented on a semilogarithmic scale revealing similar elimination rates among the different patients. The elimination curves indicate that imatinib is distributed by a one‐compartment model. Figure 1C shows the dose–concentration curves for N‐desmethyl‐imatinib.

FIGURE 1.

Plasma concentration‐time profiles of imatinib and N‐desmethyl imatinib during 24 h at steady state. (A) Imatinib on linear scale. (B) Semilogarithmic plot of imatinib concentrations. (C) N‐desmethyl imatinib on linear scale. Each line represents one patient's concentration‐time profile. Colours indicate different dosing levels.

In four patients sampled approximately 1 h after dosing, the first measurement already represented the highest concentration observed over 24 h, indicating that the absorption phase was not captured. This pattern is visible in Figure 1. In these individuals, the true C _max was likely higher than the measured value, which also contributes to lower AUC_0–24h estimates and correspondingly higher calculated clearance. This included the patient with the lowest reported C _max, indicating that the reduced value reflects sampling rather than unusually low exposure.

Neither univariate nor multivariable linear regression analyses demonstrated significant associations between the metabolite AUC/imatinib AUC ratio and any of the covariates 4‐β‐OHC/cholesterol ratio, leukocytes, haemoglobin, albumin, age, body weight, or use of concomitant moderate CYP inhibitors (Tables S1, S2 and Figure S1). Sensitivity analyses excluding the ALL patient did not alter the results (Table S3).

3.2. CYP3A4 activity

Median of 4‐α‐OHC and 4‐β‐OHC concentrations were 11.3 and 74.6 ng/ml (first sample) and 10.9 and 70.3 ng/ml (24‐h sample), respectively. Observed levels of 4‐α‐OHC and 4‐β‐OHC in plasma are presented in Table S4.

Linear regression analysis examined the relationship between imatinib clearance and CYP3A4 activity (4‐β‐OHC/cholesterol ratio). A significant positive association was observed: imatinib clearance increased with increasing 4‐β‐OHC/cholesterol ratio (adjusted R ² = 0.32, p = .007). Figure 2 illustrates this positive linear relationship between imatinib clearance and the 4‐β‐OHC/cholesterol ratio with individual variation around the fitted regression line.

FIGURE 2.

Relationship between clearance of imatinib and 4β‐hydroxycholesterol/cholesterol ratio.

A Wilcoxon signed‐rank test comparing the first and last 4‐β‐OHC/cholesterol ratios found no significant difference between samples (median difference = 0.007, p = .59), indicating stable CYP3A4 activity during the 24‐h study period.

3.3. Concomitant medication

Seventeen of the included patients received concomitant medication (in total 64 different drugs). Of these, three patients were treated with amlodipine, a moderate CYP3A4 inhibitor; no other systemically administered CYP3A4 inhibitors/inducers were identified. Five patients received long‐term statin therapy at the time of sampling (four treated with rosuvastatin 5–20 mg daily and one treated with atorvastatin 80 mg daily), with no recent dose changes.

Plasma concentrations of 4‐β‐OHC were significantly lower in statin‐treated patients compared with non‐treated patients (median 34.8 vs. 84.3 ng/ml, p < .05). Total cholesterol concentrations were also significantly lower in statin‐treated patients (median 1.5 vs. 1.98 mg/ml, p = .01). The 4‐β‐OHC/cholesterol ratio was likewise significantly lower in statin‐treated patients than in non‐treated patients (median 0.25 vs. 0.41 ng/ng × 10,⁴ p = .002). Univariate linear regression analyses were conducted to assess the association between the 4‐β‐OHC/cholesterol ratio and the metabolite AUC/imatinib AUC ratio stratified by statin use, and no evidence of effect modification by statin use was observed (Table S5).

3.4. TDM sampling

Linear regression analysis evaluated the ability of imatinib concentrations at individual time points to predict AUC_0–24 (Table 3). All time points showed significant associations with AUC_0–24 (p ≤ .003). Prediction performance varied by sampling time. Early sampling at 1 h showed moderate association, while concentrations at 2–5 h demonstrated strong predictive ability (R ² = 0.79–0.97). The highest associations occurred at 5 h post‐dose with 97% of AUC_0–24 variability explained by concentrations at this time point. Trough sampling (24 h) showed a strong association as well (R ² = 0.85).

TABLE 3.

Results of linear regression of imatinib AUC_0–24h and concentrations at different time points.

Time point (h)	No. of observations	Regression coefficient	95% confidence interval	Adjusted R 2	p‐value
1	19	9.8	3.9–15.7	0.39	.003
2	19	13.7	10.2–17.2	0.79	<.001
3	19	16.2	14.3–18	0.95	<.001
4	19	16.7	14.6–18.7	0.94	<.001
5	7	18.4	14.9–21.9	0.97	<.001
24	19	32.2	25.7–38.8	0.85	<.001

Note: Regression coefficient is calculated as change in AUC_0–24h (ng·h/ml) per unit change in concentration (ng/ml).

3.5. Genotyping

Genotyping of CYP3A4 and CYP3A5 variants revealed limited genetic variability in our cohort. All patients were homozygous wildtype for the analysed variants, with one exception. One patient carried the heterozygous C/T genotype for CYP3A4*22 (g.15389C > T), while the remaining 18 patients had the wildtype C/C genotype. Pharmacogenetic association analysis was not deemed feasible for this variant, because this patient was the only individual receiving 600 mg imatinib, which could have been a potential confounder in the assessment of genotype‐dependent pharmacokinetic differences. All analysed genotypes are available in Tables S6 and S7).

4. DISCUSSION

This is the first study to investigate 4‐β‐OHC as a biomarker of CYP3A4 activity in leukaemia patients receiving imatinib therapy. The individual pharmacokinetic parameters presented in our study are comparable with prior reportings. ²

Using an endogenous biomarker of CYP3A4 activity, such as 4‐β‐OHC, has obvious advantages compared to using traditional CYP3A4 probe drugs. These advantages include lower costs, fewer ethical considerations, for example, avoiding additional medication burden and possible side effects, and minimal sampling requirements because of its stability—provided it demonstrates similar performance to validated CYP3A4 probes.

Importantly, the 4‐β‐OHC/cholesterol ratio demonstrated a significant positive association with imatinib clearance, explaining approximately 30% of the variance—a clinically meaningful relationship for an endogenous biomarker. However, it was not possible to significantly predict the metabolite AUC/imatinib AUC ratio from the 4‐β‐OHC/cholesterol ratio in our cohort. One likely explanation is that imatinib is not cleared solely by CYP3A4 but is also a substrate of ATP‐binding cassette (ABC) transporters, including P‐glycoprotein (ABCB1) and breast cancer resistance protein (BCRP/ABCG2), which contribute to variability in absorption and distribution and may limit the predictive value of CYP3A4 biomarkers alone. While our results are insufficient to establish the 4‐β‐OHC/cholesterol ratio as a reliable predictor of imatinib metabolic activity, the substantial clearance correlation supports the potential utility of this biomarker in a clinical context. This relationship suggests that CYP3A4 activity measurements could potentially inform imatinib starting dose selection, with patients showing higher enzyme activity requiring higher starting doses to achieve optimal exposure. Further analysis in a larger cohort with additional covariates is needed to fully explore this relationship.

The comparison of 4‐β‐OHC/cholesterol ratios demonstrated no difference between the first and last sample, which is consistent with half‐life of 4‐β‐OHC of 17 days. Measurements of 4‐α‐OHC confirmed appropriate sample handling throughout the study. Because imatinib is metabolized by CYP3A4 and simultaneously inhibits CYP3A4 activity, measuring both 4‐α‐OHC and 4‐β‐OHC prior to imatinib initiation would be valuable to establish baseline CYP3A4 activity levels before potential drug‐induced CYP3A4 inhibition occurs. This would require pre‐treatment sample collection in future studies to better understand the impact of imatinib on endogenous CYP3A4 biomarkers.

Plasma 4‐β‐OHC concentrations in this cohort were higher than those reported in Scandinavian reference populations, where typical mean or median values around 25–30 ng/ml have been reported. ¹⁸ , ¹⁹ In our study, median 4‐β‐OHC was 74.6 ng/ml (IQR 47.2–98 ng/ml), which, although elevated relative to population averages, remains within the previously reported interindividual range. While imatinib is both a CYP3A4 substrate and inhibitor, 4‐β‐OHC appears to be more responsive to CYP3A4 induction than to moderate inhibition, which generally causes only modest reductions in concentration. ¹⁸ Moreover, given the long elimination half‐life of 4‐β‐OHC and its high intraindividual stability, absolute levels largely reflect baseline CYP3A4 activity. Because pre‐treatment values were unavailable, it is not possible to determine the extent to which imatinib reduced 4‐β‐OHC in individual patients. Consistent with this, steady‐state data indicate that long‐term imatinib therapy may not markedly suppress overall CYP3A4 function; coadministration with the potent inhibitor ritonavir did not alter imatinib exposure, suggesting a shift towards alternative elimination pathways and auto‐inhibition at steady state. ²⁰

Our finding of significantly lower plasma 4‐β‐OHC concentrations in statin‐treated patients is consistent with previous reports demonstrating dose‐dependent reductions in 4‐β‐OHC during short‐term fluvastatin and atorvastatin treatment. ²¹ In contrast to that study, which reported an unchanged or slightly increased 4‐β‐OHC/cholesterol ratio after 4 weeks of statin treatment, we additionally observed a significant reduction in the ratio in our statin‐treated patients. This discrepancy is likely explained by differences in treatment duration and pharmacokinetics. Because 4‐β‐OHC has a long half‐life, steady state is unlikely to be achieved within 4 weeks, whereas cholesterol levels decline rapidly after statin initiation. Our findings therefore reflect long‐term steady‐state adaptation of oxysterol metabolism rather than transient effects of early cholesterol lowering. Differences in statin type, including the use of rosuvastatin (which is not primarily metabolized by CYP3A4), may further contribute to the observed differences between studies.

Few studies have been published focusing on endogenous CYP3A4 activity in haematological malignancies. Phase I/II studies of enasidenib and ivosidenib for acute myeloid leukaemia demonstrated significant increases in 4‐β‐OHC plasma concentrations from baseline to steady‐state conditions, suggesting 4‐β‐OHC as an adequate biomarker of strong CYP3A induction. ²² , ²³ Notably, in the AML study, baseline 4‐β‐OHC/cholesterol ratios were comparable to values previously reported in a human immunodeficiency virus patient population, suggesting that baseline levels of this biomarker may be relatively stable across disease populations. ²³

Conflicting results are observed regarding CYP3A4 activity and imatinib treatment response. One study of 14 CML patients using quinine as a probe drug found that patients achieving complete molecular response (undetectable level of the BCR‐ABL1 transcript) after 12 months of therapy had significantly higher CYP3A activity compared to patients who failed to achieve complete molecular response. ²⁴ However, this finding could not be validated in a subsequent larger study (n = 43) showing no difference in CYP3A4 activity between patients with and without optimal response after 12 months. ²⁵

While our cohort included one patient with Ph+ ALL receiving imatinib monotherapy, the pharmacokinetic behaviour of imatinib in this clinical context is expected to be comparable to that of CML patients, as both involve the same molecular target (BCR‐ABL1) and identical monotherapy without concurrent chemotherapy that could alter drug metabolism. The pharmacokinetic properties of imatinib are primarily determined by CYP3A4 activity and physiological factors influencing drug absorption, distribution and elimination rather than the underlying haematological malignancy. Therefore, we considered this patient's pharmacokinetic profile to be representative of the broader imatinib‐treated population and included the data in our primary analysis. A sensitivity analysis excluding the single ALL patient did not materially alter the findings, supporting inclusion of this individual in the final analyses.

Our results suggest that single‐point sampling at 3–4 h post‐dose reliably could estimate AUC_{0 – 24h} in clinical practice and even performs better than trough concentrations, which are normally chosen in TDM of imatinib. The high correlation between these time points and AUC is expected given the incorporation of concentrations at different time points into the AUC calculation. This finding could have practical implications for the TDM strategy of imatinib but awaits validation.

Steady‐state pharmacokinetic studies critically depend on patients maintaining consistent medication timing intake on the days preceding study initiation. Some participants in our cohort had established evening or nighttime dosing routines but were required to switch to a daytime schedule to align with our study protocol. Daily SMS reminders were sent at the new medication time throughout the pre‐study period, and we received confirmation feedback from the relevant patients, supporting that steady‐state conditions were likely achieved at study initiation.

Surprisingly, no apparent pharmacokinetic interaction was observed between imatinib and amlodipine despite amlodipine's moderate CYP3A4 inhibitory properties. This may reflect a diminished role of CYP3A4 in imatinib clearance under steady‐state conditions. ²⁰ However, the absence of baseline imatinib AUC measurements prior to amlodipine initiation means that any potential effect of amlodipine on imatinib exposure cannot be determined from our data. Although medication lists were carefully reviewed with patients, suboptimal adherence to amlodipine therapy cannot be excluded and could have contributed to the lack of observed interaction.

This study has several limitations that should be considered when interpreting the results. The 24‐h study design limited our investigation to pharmacokinetic parameters. Pharmacodynamic outcomes, including safety assessments and therapeutic response indicators such as BCR‐ABL1 transcript measurements, were not evaluated, as these endpoints require extended follow‐up periods beyond the scope of this short‐term pharmacokinetic study. While additional sampling points would have improved pharmacokinetic characterization, ethical considerations regarding patient burden limited our sampling schedule and excluded nighttime collection. As sampling was performed via repeated venipuncture in all but one patient, several participants were reluctant to accept additional blood draws, particularly during the terminal elimination phase. This further reduced the number of feasible sampling time points. Consequently, sparse sampling in the terminal elimination phase limited reliable half‐life determination in three patients and constrained characterization of metabolite pharmacokinetics. Although ≥3 quantifiable data points were available for most patients, the terminal points selected for regression often occurred relatively early after T _max, where distribution processes may still contribute. Thus, estimation of Kel and t _½ is inherently less robust across the cohort and not only in the three patients where regression was not possible. In contrast, clearance estimates are considered robust, as CLss/F was based on the observed AUC_0–24h, which is mainly determined by concentrations collected during the absorption and early distribution phases and did not require extrapolation.

Baseline 4‐β‐OHC concentrations prior to imatinib initiation were not available, limiting our ability to determine the isolated effect of imatinib on endogenous CYP3A4 activity. Although medication histories were carefully reviewed, pre‐imatinib treatments could potentially have influenced CYP3A4 activity. Six patients briefly received cytoreductive therapy with hydroxyurea or anagrelide for 1–3 weeks shortly after diagnosis, followed by long‐term imatinib treatment. The patient with Ph+ ALL had previously completed Hyper‐CVAD chemotherapy, with maintenance therapy ending more than 1 year before study inclusion. None of the treatments are known strong inducers or inhibitors of CYP3A4, with the exception of prednisolone, a weak CYP3A4 inducer used as part of the ALL maintenance therapy. Given that prednisolone therapy ended more than 1 year prior to sampling, any inducing effect would be expected to have resolved. Therefore, an impact on 4‐β‐OHC concentrations is considered unlikely but cannot be fully excluded because of the absence of baseline measurements.

We initially planned to investigate potential associations between CYP3A4 and CYP3A5 genotypes and imatinib pharmacokinetics, as well as their relationship to CYP3A4 activity. However, a limitation of this study was the homogeneous genetic background of our cohort regarding CYP3A4 and CYP3A5 variants. Allele frequencies of the CYP3A4 and CYP3A5 genotypes investigated in this study exhibit low genetic variability in European populations. For instance, the CYP3A4*1B g.‐392A > G variant has a minor allele frequency of approximately 4%, with the vast majority (~96%) of Europeans carrying the wildtype (A/A) genotype. Given our limited sample size (n = 19) and the fact that all patients included in the study were Caucasian, the probability of detecting such rare variant genotypes in our population was inherently low. This expectation was confirmed by our genotyping results, which revealed that our cohort lacked sufficient genetic variability for meaningful analysis of the investigated variants, with one exception: a single patient carried the C/T variant genotype for CYP3A4*22 (g.15389C > T). The remaining cohort had the C/C wild‐type genotype. The lack of genetic diversity, combined with the possible confounding effect of dosing differences in the single variant carrier, prevented pharmacogenetic association analysis. Future studies with larger, more genetically diverse populations would be needed to adequately assess the impact of these variants on imatinib pharmacokinetics, especially for low‐frequency variants.

5. CONCLUSION

4‐β‐OHC remains promising as an endogenous CYP3A4 biomarker, and our findings suggest it may have utility for predicting individual imatinib clearance in leukaemia patients in clinical practice. Larger studies are needed to determine whether this biomarker could be valuable for identifying leukaemia patients at risk for drug interactions or for guiding initial dosing strategies. Given the clinically relevant association with imatinib clearance demonstrated in this proof‐of‐concept study, further investigation in expanded cohorts with greater genetic diversity and longer follow‐up periods is warranted to assess clinical outcomes.

AUTHOR CONTRIBUTIONS

All authors wrote the manuscript; Anna Sofie Buhl Rasmussen, Mathilde Rønne Koch, Cecilie Utke Rank, Christen Lykkegaard Andersen, Mette Klarskov Andersen, Kjeld Schmiegelow and Kim Dalhoff designed the research; Anna Sofie Buhl Rasmussen, Mathilde Rønne Koch, Allan Weimann, Flemming Nielsen and Kim Dalhoff performed the research; Anna Sofie Buhl Rasmussen, Cecilie Utke Rank, Christen Lykkegaard Andersen, Allan Weimann, Tianwu Yang, Trine Meldgaard Lund, Ib Jarle Christensen, Tore Bjerregaard Stage, Sara Bisgaard Jensen, Mette Klarskov Andersen, Kjeld Schmiegelow and Kim Dalhoff analysed the data; Allan Weimann and Flemming Nielsen contributed new reagents/analytical tools.

CONFLICT OF INTEREST STATEMENT

The authors declared no competing interests for this work.

Supporting information

Table S1. Univariate linear regression analysis of covariates associated with the metabolite AUC/imatinib AUC ratio.

Figure S1. Multipanel plots showing univariate linear regressions of covariates with the metabolite AUC/imatinib AUC ratio.

Table S2. Multivariable linear regression analysis of covariates associated with the metabolite AUC/imatinib AUC ratio.

Table S3. Univariate and multivariable regression analyses of the association between the 4‐β‐OHC/cholesterol ratio and the metabolite AUC/imatinib AUC ratio in CML patients (sensitivity analysis).

Table S4. 4α‐Hydroxykolesterol (4‐α‐OHC), 4β‐Hydroxykolesterol (4‐β‐OHC) and cholesterol levels at different time points.

Table S5. Univariate linear regression of metabolite AUC/imatinib AUC ratio on 4‐β‐OHC stratified by statin use.

Table S6. Individual genotypes.

Table S7. Common allele name, sequence, assay name, rs number for analysed genotypes.

Rasmussen ASB, Koch MR, Rank CU, et al. Pharmacokinetic profiling of imatinib in relation to CYP3A4 activity in leukaemia patients. Br J Clin Pharmacol. 2026;92(10):3375‐3386. doi: 10.1002/bcp.70622

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.