Pharmacokinetic Investigation of Imatinib Using Accelerator Mass Spectrometry in Patients with Chronic Myeloid Leukemia

Imatinib, an inhibitor of the Bcr-Abl tyrosine kinase, is an effective treatment for chronic myeloid leukemia (CML) and has revolutionized the treatment of this disease (1). However, some patients fail to achieve a complete hematologic or molecular response, and others relapse with disease that is then resistant to treatment with imatinib (2, 3). A number of theories have been advanced to account for CML resistance to imatinib, including innate or acquired mutations in the Bcr-Abl gene (4, 5), imatinib binding to α₁-acid glycoprotein (6, 7), and altered imatinib pharmacokinetics (8–10). The possibility that imatinib resistance is due to initial or acquired differences in drug absorption, distribution, or metabolism has been explored to some extent in conventional pharmacokinetic studies (9, 11–13). Using liquid chromatography/mass spectrometry (LCMS) to measure drug and metabolites in plasma (12, 13) or using ¹⁴C-imatinib to detect total drug-derived species in mass-balance studies (14), a number of metabolites of imatinib have been identified. The major metabolite is CGP74588, which accounts for ∼10% of an oral dose of imatinib and retains inhibitory activity towards the Bcr-Abl kinase (13).

The bioavailability of imatinib is close to 100% from either tablets or capsules and does not seem to vary significantly among individuals, during chronic therapy or with the dose administered (15, 16). Therefore, the most likely changes in pharmacokinetics that may influence therapeutic efficacy are in metabolism or distribution, given that only a small fraction of a dose of imatinib is excreted as parent drug in the urine. Analysis based on LCMS may not be sufficiently sensitive in all cases to detect altered metabolism, binding to plasma proteins, or distribution to the target tissue in the relevant patient population.

Accelerator mass spectrometry (AMS) is a relatively novel technique that can be used to measure extremely low concentrations of drug and metabolites in a variety of tissues (17, 18). AMS has previously been used in mass-balance studies (19) and in micro-dosing studies in early drug development (17, 20). Although AMS requires the drug molecule to be labeled with ¹⁴C, the method of detection does not depend on radioactive decay. Rather, the ratio of ¹⁴C to ¹²C in samples taken after drug administration is compared with that in pretreatment samples, and in the dose administered, to derive drug concentrations (20).

To assess the suitability of AMS as a technique to evaluate the pharmacology of imatinib in patients on long-term therapy for CML, a pilot study in six patients was done. Plasma and peripheral blood lymphocyte samples were obtained during a 72-h period after administration of a ¹⁴C-labeled dose of imatinib. Data from the AMS analysis were compared with conventional LCMS data on imatinib and its major metabolite.

Capsules containing 100 mg of ¹⁴C-imatinib as the mesylate salt at a specific activity of 135.7 Bq/mg were supplied by Novartis AG. The structure of imatinib, indicating the site of the ¹⁴C label, is shown in Fig. 1. Imatinib tablets (100 mg) were supplied by the Pharmacy Department, Royal Victoria Infirmary. Imatinib, CGP74588, and deuterated (D8) imatinib, the internal standard for LCMS assays, were supplied by Novartis.

Six patients were studied, all of whom were on continuous therapy with 400 mg/d imatinib for chronic myeloid leukemia. Details of the patients, time since diagnosis, duration of imatinib therapy, and PCR-based response are given in Table 1. All patients gave written informed consent and the study was approved by the Newcastle Hospitals Healthcare Trust Ethics committee. All patients were studied in the Clinical Research Facility, Newcastle University.

On the day of study, patients were asked not to take their regular dose of imatinib. Pretreatment plasma and peripheral blood lymphocyte (PBL) samples were collected before administration of the study treatment. Each patient then took 100-mg imatinib containing 13.6 kBq (368 nCi) ¹⁴C in a capsule, together with three 100-mg tablets of Glivec to make up their regular dose of imatinib. Dosing and sampling procedures were done in separate rooms to reduce the risk of contamination. Following administration of the ¹⁴C dose, blood samples were obtained in heparin tubes at the following times for separation of plasma (800 × g, 10 min at 4°C): 0.5, 1, 1.5, 2, 3, 4, 6, 8, 12, 24, 48, and 72 h. At 24 h, a blood sample was taken for the isolation of PBL using a Ficoll gradient (Lymphoprep). All samples were frozen immediately and stored at −20°C for subsequent analysis. Patients received their normal 400-mg dose of Glivec as a single tablet 24 and 48 h after the test dose, and thereafter continuously.

An aliquot of each plasma sample was analyzed using AMS (see below). In addition, parent drug was chromatographically separated from other ¹⁴C-labeled material before AMS analysis. Briefly, to one volume of plasma, an equal volume of chilled acetonitrile was added and vortex mixed. The sample was centrifuged for 5 min at 2,500 × g at +4°C. The resulting supernatant layer was removed into a separate vial and the residue further extracted with one volume of acetonitrile/water (1:1, v/v). The residue was macerated with a clean spatula before vortexing for 1 min and centrifuging for 5 min at 2,500 × g at 4°C. The resulting supernatants were pooled and reduced to almost dryness under a gentle stream of nitrogen. The extracts were reconstituted in mobile phase and added to 10 μL of a 50:50 mixed standard solution of imatinib and the major metabolite (CGP74588), both at a concentration of 0.5 mg/mL.

Chromatographic separation of imatinib from CGP74588 and other ¹⁴C-labeled material was done using a Waters Symmetry Shield RP8 column (150 × 4.6 mm, 3.5 μm). A gradient elution system was used varying from 25% acetonitrile (A), 75% 0.05 mol/L ammonium acetate buffer pH 6.38 (B), to 31% A after 2 min, then increasing to 50% A between 13 and 16 min, decreasing to 20% A after 18 min, and returning to the starting conditions between 25 and 40 min. Flow rate was 1 mL/min and the eluate was monitored at 254 nm. The column eluant was collected as a series of 500-μL fractions (at ∼30-s intervals) throughout the 40-min high-performance liquid chromatography run. Fractions corresponding to the parent compound were pooled and subsequently analyzed by AMS.

WBC samples were treated with HCl, sonicated for 30 min to disrupt the cells, and centrifuged for 10 min at 2,500 × g at +4°C. The final supernatant was removed and 200 μL of a 0.2 mg/mL RNase solution were added to the residues. Samples were sonicated for 80 min before AMS analysis.

Full details of the AMS analysis procedure are described by Garner (19). In brief, duplicate aliquots of plasma and high-performance liquid chromatography fractions were graphitized according to the method described previously (21). The resulting graphite/cobalt mix was pressed into aluminum cathodes and each sample analyzed using a 5-MV 15SDH-2 Pelletron AMS system (National Electrostatics Corp.). Samples containing minimal amounts of carbon were bulked up using tributyrin. Each sample was counted in the AMS instrument for a minimum of 100 s, and this was repeated for each sample at least thrice. AMS data were converted to disintegrations per minute per milliliter, taking into account the carbon content of the sample and any added carrier. Data in disintegrations per minute per milliliter were converted to nanograms per milliliter by dividing by the specific activity of ¹⁴C. Carbon contents were measured using a carbon, hydrogen, nitrogen analyzer (Elemental Microanalysis Ltd.). The limit of quantitation for unchanged imatinib detected by AMS following high-performance liquid chromatography separation was 380 pg/mL (0.003 dpm/mL), and that for total ¹⁴C was equivalent to 6 ng/mL imatinib (0.06 dpm/mL). The coefficient of variation for the AMS assay was >5% across the range of concentrations analyzed.

A separate aliquot of plasma was analyzed by a previously published LCMS method for imatinib and the major metabolite CGP74588 (22). To 200 μL of plasma, 50 μL of methanol and 50 μL of internal standard solution (540 ng/mL D8-imatinib) were added, followed by 250 μL of acetonitrile and vortex mixing for ∼20 s. The samples were centrifuged at 15,000 × g for 5 min in a microfuge and 150 μL of the supernatant were transferred to a limited volume high-performance liquid chromatography insert. Twenty microliters were injected onto the LCMS. The mobile phase consisted of 60% methanol, 40% 0.02 mol/L ammonium acetate at a flow rate of 300 μL/min. Analysis was done on an Applied Biosystems API2000 in MRM mode, with parent/daughter ion combinations of m/z+ 495/395, 481/395, and 503/395 for imatinib, CGP74588, and the D8 internal standard, respectively. The lower limits of quantitation for parent drug and metabolite were 1 and 2 ng/mL, respectively. Plasma concentrations of α₁-acid glycoprotein were determined according to a previously published method (23).

Pharmacokinetic analysis. To compare the imatinib plasma concentration data obtained by the two different methods (AMS and LCMS), the contribution from the previous doses of imatinib was subtracted from the LCMS data. This was achieved by determining the slope of terminal exponential decay of imatinib in each patient (from the imatinib LCMS data) and simulating the decline of the pretreatment concentration (i.e., the imatinib concentration predicted if the test dose had not been administered). These exponentially declining concentrations were then subtracted from the observed plasma imatinib concentration, as measured by LMCS, at each time point. An analogous correction was applied to the sum of imatinib and CGP74588 as measured by LCMS.

Data from AMS and LCMS assays were analyzed using noncompartmental methods in WinNonlin, version 1.3 (Pharsight Corp.). The AUC_0-24 for imatinib at steady-state, determined by LCMS, was also calculated. Under the conditions of linear, time-invariant pharmacokinetics, this AUC_0-24 should be identical to the corrected AUC_0-∞.

Patients' ages ranged from 35 to 54 years; duration of imatinib treatment was 20 to 63 months; and time since diagnosis ranged from 31 to 71 months (Table 1). There were no adverse effects observed in patients participating in this investigation. Data from patient 1 were excluded from some elements of the summary in Table 2 because he had taken his subsequent dose of imatinib before the 24-h sample was drawn. This dose did not contain ¹⁴C-imatinib and, hence, did not affect the AMS analysis.

A plot of the uncorrected LCMS and AMS data for imatinib, CGP74588, and total ¹⁴C material by AMS is shown for patient 5 in Fig. 2. Apart from patient 1, in whom the concentrations of imatinib and CGP74588 as measured by LCMS increased between 8 and 24 h, and patient 6, who had a long half-life (33 h) for ¹⁴C-imatinib, the estimates of terminal half-life from either LCMS or AMS data were within 1 h of each other in all patients. Figure 3 shows the corrected data for patient 5 after subtracting the imatinib concentrations predicted to be present from previous imatinib doses, based on the LCMS data. In this patient, there was good agreement between the imatinib concentrations as determined by LCMS and AMS, with the imatinib AUC_0-∞ for the AMS data being within 2% of the value generated from the LCMS data.

To compare the known imatinib-derived materials to the total ¹⁴C material measured by AMS, concentrations of imatinib and CGP74588, as determined by LCMS, were combined. The contribution from previous doses of imatinib to this composite measure was subtracted for comparison with the AMS data. No correction for the change in molecular weight was applied because the ¹⁴C measure represents a mixture of contributions from species of varying molecular weight. A comparison of total ¹⁴C material by AMS and the sum of imatinib and CGP74588 is shown for patient 5 in Fig. 3.

Applying noncompartmental analysis to the AMS and LCMS data, Fig. 4 shows a comparison of AUC_0-∞ for the single dose of imatinib measured by two different methods (LCMS and AMS) and AUC_0-24 by the LCMS method for all six patients.

There was poor agreement between AUC_0-24 (41 ± 13 μg/mL·h) for the uncorrected LCMS data and AUC_0-∞ after subtracting the contribution from previous doses (27 ± 11 μg/mL·h). On average, there was reasonable agreement in AUC_0-∞ between the two analytic methods (AMS, 26 ± 3 μg/mL·h; LCMS, 27 ± 11 μg/mL·h), with considerably more variability in the LCMS data. This greater variability (CV = 41%) is not solely due to increased variation introduced in correcting for the contribution of previous doses because a similar coefficient of variation was seen in the AUC_0-∞ for the uncorrected LCMS data (CV = 40%).

For three patients, there was good agreement between AMS and LCMS (patients 3, 5, and 6; AUC_0-∞ values all within 20%; Fig. 4). In patients 1 and 2, the LCMS value exceeded that from AMS by up to 2-fold. Conversely, in patient 4, the AUC_0-∞ by AMS was nearly 2-fold higher than the LCMS value. Comparing the total ¹⁴C-labeled material measured by AMS with the sum of imatinib and CGP74588 concentrations (Fig. 5), up to 65% of the drug-derived material in the plasma is not accounted for by the parent drug and the major known metabolite.

In the WBC sample taken at 24 h, imatinib was detectable (5.2-28 ng/10⁷ cells) in three of six patients. There was no correlation between WBC concentrations and corresponding plasma concentrations in the same patient. Average plasma concentrations of α₁-acid glycoprotein ranged from 273 to 399 μg/mL, with little variation in any individual patient over the 72-h period of study (CV ≈ 10%).

The emergence of resistance to imatinib, particularly in CML, may be seen as inevitable, and the identification of mutations in Bcr-Abl (4) fits the model of resistance due to selective pressure on a genetically unstable or heterogeneous population of tumor cells. However, there are a number of studies that have suggested that pharmacokinetic explanations may underlie innate or acquired resistance to imatinib in some patients (7–10). One approach to overcoming such variability is to use doses higher than 400 mg/d (24, 25).

Whereas differences in pharmacokinetic variables following a single dose and at steady-state seem to be modest for the parent drug (26), it has not previously been possible to investigate the pharmacokinetics of an individual dose under steady-state conditions. The AUC values in the present study may involve an inaccuracy in the calculation of half-life from the LCMS data, which may account in part for the discrepancy between single-dose and steady-state AUCs. Previously published data on the use of radiolabeled drug to investigate the pharmacokinetics of a single dose at steady-state are very limited (27). It may be that the pharmacokinetics of the parent drug alone does not give sufficient information to identify the effects of enzyme induction or qualitative changes in drug metabolism.

Previous pharmacokinetic studies have focused largely on the pharmacokinetics of imatinib and CGP74588 after single oral doses (26). Where comparisons have been made with steady-state pharmacokinetics, a modest degree of accumulation has been observed, together with a slightly longer half-life at steady-state (28). Pharmacokinetic variables in the current study, including C_max (2.7 ± 0.9 μg/mL), T_max (3 ± 2 h), and half-life (21 ± 6 h), derived from the LCMS analysis of imatinib (Table 2) were similar to those previously reported (11, 26, 28), although the half-life value is slightly higher than that reported by Peng et al. of 19.3 h.

In a study using ¹⁴C-imatinib (14), healthy volunteers received a single oral dose of 200 mg of imatinib with a total radioactive dose of 1.18 MBq (32 μCi), 80-fold higher than the dose of radioactivity administered in the current study. In the latter, the AUC_0-∞ was 11 ± 1 μg/mL·h and that for CGP74588 was 1.7 ± 0.2 μg/mL·h. The AUC_0-∞ for total radioactivity was 26 ± 1 μg/mL·h in plasma. The comparison between imatinib AUC and that for total radioactivity in the current study gives AUC_0-∞ values of 26 ± 3 and 76 ± 19 μg/mL·h. Thus, the ratio between imatinib and total radioactivity is 41% in the previous single-dose study and 34% in the current study at steady state. Albeit in a study with a small number of patients, these data may suggest that the greater proportion of ¹⁴C label present as non–parent drug material in the steady-state setting represents a shift towards the formation of other unknown metabolites. By sampling over 1 week following administration, the previous investigation identified a terminal half-life for total ¹⁴C material of 57 ± 12 h, compared with a half-life of 32 ± 7 h observed here with sampling out to only 72 h. Previous studies (28) at steady state have reported an AUC_0-∞ of 82 ± 45 μg/mL·h, comparable to that of 72 ± 29 μg/mL·h, as measured by LCMS in the current study. Indeed, the values for t_1/2 and AUC_0-24 (Table 2) are almost identical to those previously reported (28).

A potential role for pharmacokinetic variability in resistance to imatinib is supported by evidence from drug binding and transport studies. Binding of imatinib to α₁-acid glycoprotein has been shown to limit the antitumor effect of imatinib, an effect which is reversible by displacing imatinib from α₁-acid glycoprotein (6). Imatinib has also been identified as a substrate for transport proteins such as ABCB1 (29) and ABCG2 (8). The influence of these proteins on the intestinal absorption and subsequent tissue distribution of imatinib has been suggested to play a role in pharmacologic resistance to the drug (8, 9). The development of resistance during chronic therapy is possibly linked to induction of α₁-acid glycoprotein (6) or to changes in imatinib pharmacokinetics over time (30). With regard to binding to α₁-acid glycoprotein, it is possible that imatinib and CGP74588 compete for binding sites, such that the fraction unbound in plasma will be affected both by changing concentrations of α₁-acid glycoprotein in plasma and by changes in the relative concentrations of drug and metabolite (31). In the patients studied here, levels of α₁-acid glycoprotein were not markedly elevated, with little variation among patients. Variation in distribution to cellular components of the blood may also affect drug action, but in the current study the ¹⁴C label was detectable in WBC at concentrations close to the limit of detection, and in only three of six patients.

This study has shown that AMS can be applied to investigate the pharmacokinetics of imatinib during chronic therapy. There were some discrepancies between the LCMS and AMS data for parent drug, but only a small number of patients were investigated in this feasibility study. Further analyses, including those of unbound drug and of other potential metabolites, may yield additional insights into the underlying mechanisms responsible for the discrepancies observed in some patients between imatinib concentrations measured by LCMS and AMS. Although there was not a clear association between imatinib pharmacokinetics and clinical response in this small patient group, application of AMS technology in a larger study would be of value in defining the role of pharmacologic variation in resistance to imatinib therapy.