FormalPara Key Points

A retrospective observational cohort study was conducted to assess underexposure in imatinib-treated gastrointestinal stromal tumour (GIST) patients in routine clinical care.

In routine clinical care, 32.4 % of imatinib-treated GIST patients were systematically underexposed, with trough plasma concentration (C min) values <1000 µg/L.

Further prospective clinical studies are needed to investigate the value of C min-guided imatinib dosing in GIST patients.

1 Introduction

Gastrointestinal stromal tumours (GISTs) are the most common mesenchymal malignancies arising from the gastrointestinal tract. Activating mutations in KIT proto-oncogene receptor tyrosine kinase (KIT) or platelet-derived growth factor receptor (PDGFR), resulting in activation of the tyrosine kinase signalling pathway, are considered to be the main molecular drivers in GIST. Imatinib is a tyrosine kinase inhibitor (TKI), which targets protein kinases such as Bcr-Abl, KIT and PDGFR-A and -B [1]. Since the introduction of imatinib, survival has improved spectacularly in advanced GIST patients, and recurrence-free survival has improved in the adjuvant setting. The recommended dose of imatinib is 400 mg, based on previous phase III studies [2, 3]. However, a large variability in plasma imatinib concentrations is observed during treatment [4, 5]. This variability may be caused by a range of factors. Imatinib is metabolized by cytochrome P450 (CYP) 3A4 and CYP3A5, and is also a substrate for drug transporters such as P-glycoprotein (P-gp; ATP-binding cassette sub-family B member 1 [ABCB1]) and breast cancer resistance protein (BCRP; ATP-binding cassette sub-family G member 2 [ABCG2]). Exposure may therefore be influenced by genetic polymorphisms and co-administered drugs [6, 7]. In addition, patients undergoing a major gastrectomy have been shown to have significantly lower C min values than other patients [8], and one study has reported a significant decrease in exposure to imatinib over time [9].

Several trials have found a correlation between higher plasma imatinib concentrations and better response to treatment in GIST [4, 10–12] and chronic myeloid leukaemia (CML) [13–15]. Given the increasing evidence that exposure is relevant to clinical outcomes and the large variability in pharmacokinetics, which may be even larger in routine clinical care than in clinical trials, measurement of plasma imatinib concentrations may be useful to guide treatment with this drug. Over the last 3 years, plasma samples have been drawn from GIST patients during routine outpatient visits at our institute. This study describes the pharmacokinetics and occurrence of underexposure to imatinib in a large observational cohort of GIST patients, with over 400 concentrations measured in more than 100 patients during routine outpatient care.

2 Methods

2.1 Patients

All GIST patients treated with imatinib at the outpatient clinic of the Netherlands Cancer Institute (NKI) were identified retrospectively and included in this study. Identification was done through a search in the database of the Dutch GIST Registry, containing all patients diagnosed with GIST from 2009 to 2014 and treated at five GIST centres in the Netherlands. Only patients treated at the NKI were included. Patients who were diagnosed before 2009 and had one or more plasma imatinib concentrations measured were identified separately, and their data were added manually.

2.2 Variables

Patient characteristics (sex and ethnicity) and tumour characteristics (location, size, mitotic index and mutation status) were extracted from the Dutch GIST Registry. The mutation analysis protocol included analysis of KIT (exons 9, 11, 13 and 17) and PDGFRA (exons 12, 14 and 18) by Sanger sequencing. Sequencing was performed on a capillary sequencer (ABI 3730 DNA Analyzer; Life Technologies, USA), and mutation analysis was performed using specific software (MutationSurveyer; Softgenetics, USA).

Also, the treatment objective [palliative or (neo-)adjuvant], imatinib dose, dosing schedule and adverse events were included in the analysis. Past surgeries for GIST and surgery results were entered, as were concomitant medication and medical history. For patients diagnosed before 2009, patient files were used for extracting the aforementioned variables. Response evaluations were derived from regularly performed computed tomography (CT) scans and were performed according to the Response Evaluation Criteria In Solid Tumors (RECIST) version 1.1. The best overall response was defined as the best response recorded from the start of imatinib treatment until disease progression/recurrence. Patients were classified as responders if their best response was found to be a complete response or a partial response. Patients were classified as non-responders if stable disease or progressive disease was their best response.

2.3 Pharmacokinetics

Blood samples were drawn during regularly scheduled visits at the outpatient clinic. The time of the last intake of imatinib and the time of the blood sampling were recorded. Plasma imatinib concentrations were determined using a validated liquid chromatography assay with tandem mass spectrometry [16]. An estimate of the imatinib C min was calculated on the basis of the measured concentration and the interval between the last ingested dose and the sampling time, using the algorithm developed by Wang et al. [17]. Adequate plasma imatinib concentrations were defined as imatinib ≥1000 µg/L, as described in previous studies [13, 15, 18]. For the analysis, the first steady-state imatinib C min was used. A representative C min was defined as the first representative sample at least 2 weeks after the start of imatinib treatment.

2.4 Statistical Analysis

Statistical analyses were executed using IBM SPSS Statistics 20 and R version 3.2.2 software [19]. Univariate and multivariate Cox regression, using relevant characteristics such as the KIT mutational status and the imatinib dose, were used for assessing the correlations of exposure to imatinib and time on imatinib treatment with the time to progression (TTP). Also, exploratory analyses using non-linear mixed-effects modelling were conducted to evaluate changes in the imatinib C min over time. Inter- and intrapatient variabilities were calculated using coefficients of variation. The association between imatinib C min values and clinical and demographic variables—such as age, sex, tumour site, surgery and tumour characteristics—was assessed using independent Mann–Whitney U tests. All tests were two sided, and a P value of <0.05 was considered significant.

3 Results

Between January 2009 and May 2014, 111 patients who received imatinib therapy were identified from the Dutch GIST Registry database. Not all patients had known imatinib C min values. From August 2012 to December 2014, 582 plasma imatinib concentrations were measured in 123 GIST patients. An additional 33 patients who started imatinib treatment before 2009 and had imatinib drug concentrations measured were identified at the outpatient clinic. All samples below the lower limit of quantification were excluded, in case this was due to a planned end of treatment or interruption due to adverse events. Also, samples with a missing time of the last dose or of sampling and samples drawn within 2 weeks after the start of imatinib treatment were excluded. This resulted in 421 representative plasma imatinib concentrations from 108 patients included in the analysis. The median sample frequency per patient was 3 (range 1–11).

Patient and tumour characteristics are described in Table 1. More than half of the cohort consisted of men (n = 60, 56.5 %), and the median age was 60 years (range 28–87) (Table 1).

Table 1 Characteristics of the patients
Full size table

An overview of the distribution of the calculated imatinib C min values in the patients studied in this cohort is given in Table 2. The median steady-state C min was 1082 µg/L. Sixty patients (55.6 %) had adequate C min values at steady state (Fig. 1). Overall, 32.4 % of patients showed low imatinib C min values in >75 % of their samples, and 33.3 % of patients showed adequate imatinib C min values in all measured samples. Exposure to imatinib showed larger inter- and intrapatient variabilities, with relative standard deviations of 54 and 23 %, respectively. No significant change over time was found. The slope was estimated at a negligible 0.00004 day−1, with a relative standard error of 25 % (P > 0.05).

Table 2 Characteristics of the 421 available plasma imatinib samples
Full size table
Fig. 1
figure 1

Distribution of the 108 patients’ first representative trough plasma imatinib concentration (C min) values. The dotted red line indicates a C min of 1000 µg/L

Full size image

The median time on imatinib was 27 months (range 1–161). Within the recorded follow-up period, 12 patients treated with palliative intent stopped imatinib because of progressive disease. No statistically significant difference in the TTP was found between patients with low steady-state C min values (n = 27) and those with adequate C min values (n = 35) in univariate Cox regression (hazard ratio 1.64, 95 % confidence interval 0.611–5.61; P = 0.43) (Fig. 2). In multivariate analysis correcting for the imatinib dose, sex and KIT mutational status, the association between the C min and TTP remained nonsignificant (hazard ratio 0.60, 95 % confidence interval 0.53–6.35; P = 0.34).

Fig. 2
figure 2

Time to progression of gastrointestinal stromal tumour patients on imatinib treated with palliative intent as a function of the trough plasma imatinib concentration (C min) at steady state. The dashed blue line indicates patients with an imatinib C min ≥1000 µg/L (n = 35), and the solid red line indicates patients with an imatinib C min <1000 µg/L (n = 27)

Full size image

Of the 62 evaluable patients treated with palliative intent, 5 (8.1 %) were non-responders. The median C min values were 1270 µg/L in patients showing a radiological response and 920 µg/L in non-responders (P = 0.23) (Fig. 3). In the neo-adjuvant setting, no difference in imatinib C min values was found between responders and non-responders, as all but two patients had a response.

Fig. 3
figure 3

Box plot of trough plasma imatinib concentration (C min) values measured at steady state in non-responders (n = 5) and responders (n = 57) to palliative imatinib treatment. The median C min values were 920 µg/L in non-responders and 1271 µg/L in responders

Full size image

No clinical characteristic (age, type of surgery, sex, extent of resection) was predictive of low imatinib C min values. Also, no association with tumour characteristics, such as the location (P = 0.54), tumour status at registry entry (P = 0.23) and mutation status (P = 0.48), was found. Four patients (3.7 %) discontinued imatinib treatment because of adverse events. No association with the imatinib C min was found (P = 0.40).

4 Discussion

Several studies have linked higher imatinib C min values to better treatment outcomes [4, 10–13, 15]. In CML, a threshold of ≥1000 µg/L has been recommended on the basis of several studies [20–22]. In GIST patients, a threshold of ≥1100 µg/L has been suggested [20–22]. This is based on a study by Demetri et al. [4], in which patients in the lowest C min quartile (<1100 µg/L) had a shorter TTP and decreased clinical benefit.

In our cohort, we found that a large proportion of patients were underexposed to imatinib even when a relatively low threshold of ≥1000 µg/L was used (Table 2; Fig. 1). Although 92.6 % of patients received imatinib doses of 400 mg or higher, >40 % of our patients had imatinib C min values <1000 µg/L in the first steady-state sample, and only one third of patients had adequate C min values in every sample (Table 2). This suggests that GIST patients in routine clinical care have a higher risk of underexposure, which may even result in less clinical benefit [4].

The higher average C min found by Demetri et al. [4] may have been due to a higher imatinib dose, as patients were randomized to receive either 400 or 600 mg once daily. But other studies in both CML and GIST patients have also described higher concentrations than those observed in our cohort [11, 15]. This could be explained by the fact that those previous studies were performed in a selected and regulated trial setting. In our cohort, no patient selection was made other than the diagnosis of GIST and treatment with imatinib. Although concomitant medication was strictly monitored to prevent possible interactions, no strict exclusion criteria for this study were set considering any concomitant medication causing an interaction for which no replacement was possible. Also, no exclusion criteria were set for comorbidities and laboratory results. Moreover, in routine clinical care, lack of patient compliance could be a factor.

Besides the large percentage of underexposure in the first steady-state sample (relative standard deviation 54 %), we also found a large intrapatient variability of 23 %. Only one third of patients had adequate C min values in every sample. This is in accordance with the findings reported by Yoo et al. [8], who also found large inter- and intrapatient variabilities of 44.7 and 26.5 %, respectively.

An earlier prospective pharmacokinetic study found a significant decrease in systemic exposure to imatinib of almost 30 % within 90 days [9]. The authors hypothesized that this was a consequence of lower oral bioavailability with time, possibly due to upregulation of drug transporters or CYP3A4. Another explanation could be that the decrease in exposure to imatinib resulted from a decrease in α1-acid glycoprotein (AGP) as a consequence of the impressive activity of imatinib treatment [23]. In our cohort, the large variability could not be explained by a change in C min values over time. A later study also did not find a time-dependent decrease in exposure in a cohort of 65 patients, supporting our finding [24].

No clinical characteristic was found to be predictive of low imatinib C min values. Although previous studies have reported lower imatinib C min values after major gastrectomy, no correlation between C min values and the extent of surgery was found in our study [8].

While previous studies have found a correlation between higher imatinib C min values and better clinical outcomes [4, 10–12], our results show that in routine clinical care, underexposure seems to be a substantial issue. Although no statistically significant relationships between exposure to imatinib and treatment response were found, we did find a trend towards responders having higher C min values than non-responders in the palliative setting (Fig. 3), and the same trend was found in neo-adjuvant patients. However, no correlation between the C min and TTP was found in the palliative subgroup of patients (Fig. 2). This lack of statistically significant differences could have been caused by the small number of non-responders and the limited number of progression events.

Our study gives a new and representative insight into underexposure to imatinib in GIST patients in routine clinical care. We have shown that underexposure is a substantial problem in routine clinical care and that there are large inter- and intrapatient variabilities. Given the fact that several studies have described a correlation between C min values and response, pharmacokinetically guided dose individualization—also known as therapeutic drug monitoring (TDM)—should be considered. One study attempted to demonstrate the benefits of TDM of imatinib but failed to do so because of small patient numbers and limited physician adherence to TDM recommendations [25].

A prospective clinical trial to assess the benefit of C min-guided imatinib dose adjustments in GIST patients is needed. Ideally, such a trial should use a relevant clinical endpoint, such as progression-free survival, because previous studies have found clear correlations between exposure to imatinib and efficacy, and we have now shown that underexposure is a frequent problem in routine clinical care of imatinib-treated GIST patients.