Imatinib plasma levels: correlation with clinical benefit in GIST patients

Sir, 
 
The issue of suitability of therapeutic drug monitoring (TDM) for imatinib continues to fuel controversies. Unlike two previous studies in gastrointestinal stromal tumours (GIST) patients (Judson et al, 2005; Delbaldo et al, 2006), a recent clinical pharmacokinetic (PK) substudy carried out on the basis of the B2222 trial, evaluating imatinib in patients with unresectable or metastatic CD117-positive GIST, found a correlation between imatinib total exposure and clinical response. Trough levels over 1100 ng ml−1 predicted a better overall benefit rate (composite clinical outcome; Demetri et al, 2009). These results are thus in line with observations made in CML patients, showing trough levels over 1000 ng ml−1 to predict a better molecular response rate (Picard et al, 2007; Larson et al, 2008). 
 
These concentration–effect relationships confirm and strengthen our results obtained in a population of GIST patients, albeit smaller-sized (38 patients), which we had previously reported in the Journal (Widmer et al, 2008). We have indeed observed that higher free imatinib exposure predicts a higher probability of therapeutic response, when taking into account tumour KIT genotype. The strongest association was observed in patients harbouring exon 9 mutation or wild-type (wt) KIT, which is known to decrease tumour sensitivity towards imatinib (Heinrich et al, 2003). In fact, we found that, in our population of patients, free plasma concentration (the pharmacologically active species in plasma; i.e., imatinib fraction not bound to α1-acid glycoprotein (AGP); Widmer et al, 2006) was a better predictor of the clinical response rather than total concentration. This free exposure was derived from the total exposure using a mathematical model taking into account the AGP plasma level (Widmer et al, 2006). Moreover, we found a significant relationship between this free exposure and clinical response only in patients with exon 9 mutation and wt KIT. Of importance, we also observed significant correlations between total, as well as free, imatinib exposure and the occurrence of side effects (Widmer et al, 2008). 
 
To better compare our results with those of the B2222 PK substudy, we recomputed maximum a posteriori extrapolations for both total Cmin and free Cmin in our patient samples, rather than considering the global imatinib exposure (area under the plasma concentration–time curve) that was previously analysed (Widmer et al, 2008). Among the 38 GIST patients of the previous analysis, AGP plasma levels – required to calculate free Cmin – were available for 36 patients. All these patients were included in an observational study approved by the Ethics Committee of the Lausanne Faculty of Medicine. Informed written consent was obtained from all the participants. A specific population PK model (Widmer et al, 2006) was used for this extrapolation (using NONMEM, version VI 2.0, NONMEM Project Group, University of California at San Francisco, San Francisco, CA, USA). We investigated their correlation with clinical benefit, defined as response evaluation criteria in solid tumours (RECIST) complete response, partial response or stable disease, by logistic regression analysis (using Stata version 10.1, Stata Co., College Station, TX, USA). 
 
We found no significant overall association between total Cmin and response in our GIST population. Conversely, imatinib free Cmin was correlated with a clinical benefit, with responders having higher free levels than non-responders (RECIST progressive disease). This relationship did not reach significance over the whole patient sample series (i.e., irrespective of the KIT genetic profile; area under the ROC curve=0.594 and P=0.26 using logistic regression analysis on log2 values of free Cmin). However, focusing on exon 9 mutated and wt KIT cases allowed the identification of a clear relationship (area under the ROC curve=0.932 and P=0.013). The cutoff value of 20 ng ml−1 free imatinib plasma trough level corresponded to the best sensitivity (86%) and specificity (100%). The geometric average estimate of imatinib free fraction across our study samples was 1.0% (CV 45%). The mean daily doses of imatinib only tended to be slightly higher in exon 9 or wt KIT patients compared with exon 11 KIT patients (649 mg vs 590 mg daily, respectively; P=0.07 using t-test). Table 1 describes our GIST patient samples and Figure 1 shows the striking difference between the free Cmin values of responder and non-responder exon 9 or wt KIT patients (per-sample analysis). 
 
 
 
Figure 1 
 
Box plots and individual values of extrapolated free trough concentrations (free Cmin) in exon 9 mutated or wt KIT patients, contrasting responders' samples (complete response, partial response, or stable disease; n=14; median=25.7 ng ml ... 
 
 
 
 
 
Table 1 
 
Left part: description of free trough plasma concentrations (free Cmin) deduced from imatinib and AGP levels available in samples from 36 GIST patients; right part: description of the samples from seven patients with exon 9 or wt KIT identified among ... 
 
 
 
This per-sample analysis was performed because imatinib doses administered to each patient during the course of this 3-year-long observational study could be increased or decreased. The concentration, and possibly the response or adverse events related to treatment, may therefore vary at some point for a given patient. Interestingly, a similar analysis carried out on a per-patient basis (i.e., expressing only a single mean free Cmin and one median response for each individual patient) provided a similar relationship and cut-off, without, however, reaching statistical significance because of the limited number of patients with exon 9 or wt KIT. The results from our observational study should therefore still be considered cautiously and will have to be confirmed in a larger patient population. In fact, extrapolation of free concentrations still remains to be formally confirmed by direct measurement of free plasma levels of imatinib in patients' blood, an aspect that is currently being addressed in ongoing studies initiated at our Institution with GIST and CML patients. 
 
Altogether, the results from the B2222 substudy and our results call for attention to imatinib concentration–effect relationships in the management of GIST patients, which deserve further in-depth exploration. In our opinion, future investigations should take into account not only total plasma concentrations but also free plasma concentrations (either measured or computed with regard to AGP and albumin levels), as well as the tumour genotype. We are currently expanding our set of observations in a larger GIST population in that endeavour. We also agree with the authors of the B2222 PK substudy regarding the urgent need for prospective controlled trials to assess whether a TDM programme would optimise the treatment outcomes in GIST patients, as it has been called for in CML patients (Blasdel et al, 2007). Similar considerations might apply as well to other new tyrosine kinase inhibitors currently fleshing out our armament of targeted anticancer agents. Convenient analytical methods already exist for their measurement in patients' plasma (Haouala et al, 2009), but, as for imatinib, further clinical evaluation of their concentration–response relationships and of the benefit of optimising their concentration exposure is warranted in cancer patients.


Sir,
The issue of suitability of therapeutic drug monitoring (TDM) for imatinib continues to fuel controversies. Unlike two previous studies in gastrointestinal stromal tumours (GIST) patients (Judson et al, 2005;Delbaldo et al, 2006), a recent clinical pharmacokinetic (PK) substudy carried out on the basis of the B2222 trial, evaluating imatinib in patients with unresectable or metastatic CD117-positive GIST, found a correlation between imatinib total exposure and clinical response. Trough levels over 1100 ng ml À1 predicted a better overall benefit rate (composite clinical outcome; Demetri et al, 2009). These results are thus in line with observations made in CML patients, showing trough levels over 1000 ng ml À1 to predict a better molecular response rate (Picard et al, 2007;Larson et al, 2008).
These concentration -effect relationships confirm and strengthen our results obtained in a population of GIST patients, albeit smaller-sized (38 patients), which we had previously reported in the Journal (Widmer et al, 2008). We have indeed observed that higher free imatinib exposure predicts a higher probability of therapeutic response, when taking into account tumour KIT genotype. The strongest association was observed in patients harbouring exon 9 mutation or wild-type (wt) KIT, which is known to decrease tumour sensitivity towards imatinib (Heinrich et al, 2003). In fact, we found that, in our population of patients, free plasma concentration (the pharmacologically active species in plasma; i.e., imatinib fraction not bound to a 1 -acid glycoprotein (AGP); Widmer et al, 2006) was a better predictor of the clinical response rather than total concentration. This free exposure was derived from the total exposure using a mathematical model taking into account the AGP plasma level (Widmer et al, 2006). Moreover, we found a significant relationship between this free exposure and clinical response only in patients with exon 9 mutation and wt KIT. Of importance, we also observed significant correlations between total, as well as free, imatinib exposure and the occurrence of side effects (Widmer et al, 2008).
To better compare our results with those of the B2222 PK substudy, we recomputed maximum a posteriori extrapolations for both total C min and free C min in our patient samples, rather than considering the global imatinib exposure (area under the plasma concentration -time curve) that was previously analysed (Widmer et al, 2008). Among the 38 GIST patients of the previous analysis, AGP plasma levels -required to calculate free C min -were available for 36 patients. All these patients were included in an observational study approved by the Ethics Committee of the Lausanne Faculty of Medicine. Informed written consent was obtained from all the participants. A specific population PK model (Widmer et al, 2006) was used for this extrapolation (using NONMEM, version VI 2.0, NONMEM Project Group, University of California at San Francisco, San Francisco, CA, USA). We investigated their correlation with clinical benefit, defined as response evaluation criteria in solid tumours (RECIST) complete response, partial response or stable disease, by logistic regression analysis (using Stata version 10.1, Stata Co., College Station, TX, USA).
We found no significant overall association between total C min and response in our GIST population. Conversely, imatinib free C min was correlated with a clinical benefit, with responders having higher free levels than non-responders (RECIST progressive disease). This relationship did not reach significance over the whole patient sample series (i.e., irrespective of the KIT genetic profile; area under the ROC curve ¼ 0.594 and P ¼ 0.26 using logistic regression analysis on log 2 values of free C min ). However, focusing on exon 9 mutated and wt KIT cases allowed the identification of a clear relationship (area under the ROC curve ¼ 0.932 and P ¼ 0.013). The cutoff value of 20 ng ml À1 free imatinib plasma trough level corresponded to the best sensitivity (86%) and specificity (100%). The geometric average estimate of imatinib free fraction across our study samples was 1.0% (CV 45%). The mean daily doses of imatinib only tended to be slightly higher in exon 9 or wt KIT patients compared with exon 11 KIT patients (649 mg vs 590 mg daily, respectively; P ¼ 0.07 using t-test). Table 1 describes our GIST patient samples and Figure 1 shows the striking difference between the free C min values of responder and nonresponder exon 9 or wt KIT patients (per-sample analysis).
This per-sample analysis was performed because imatinib doses administered to each patient during the course of this 3-year-long observational study could be increased or decreased. The concentration, and possibly the response or adverse events related to treatment, may therefore vary at some point for a given patient. Interestingly, a similar analysis carried out on a per-patient basis (i.e., expressing only a single mean free C min and one median response for each individual patient) provided a similar relation-Published online 23 February 2010 ship and cut-off, without, however, reaching statistical significance because of the limited number of patients with exon 9 or wt KIT. The results from our observational study should therefore still be considered cautiously and will have to be confirmed in a larger patient population. In fact, extrapolation of free concentrations still remains to be formally confirmed by direct measurement of free plasma levels of imatinib in patients' blood, an aspect that is currently being addressed in ongoing studies initiated at our Institution with GIST and CML patients.
Altogether, the results from the B2222 substudy and our results call for attention to imatinib concentration -effect relationships in the management of GIST patients, which deserve further in-depth exploration. In our opinion, future investigations should take into account not only total plasma concentrations but also free plasma concentrations (either measured or computed with regard to AGP and albumin levels), as well as the tumour genotype. We are currently expanding our set of observations in a larger GIST population in that endeavour. We also agree with the authors of the B2222 PK substudy regarding the urgent need for prospective controlled trials to assess whether a TDM programme would optimise the treatment outcomes in GIST patients, as it has been called for in CML patients (Blasdel et al, 2007). Similar considerations might apply as well to other new tyrosine kinase inhibitors currently fleshing out our armament of targeted anticancer agents. Convenient analytical methods already exist for their measurement in patients' plasma (Haouala et al, 2009), but, as for imatinib, further clinical evaluation of their concentrationresponse relationships and of the benefit of optimising their concentration exposure is warranted in cancer patients.  Overall response Free C min (ng ml -1 ) Non-responders Responders Figure 1 Box plots and individual values of extrapolated free trough concentrations (free C min ) in exon 9 mutated or wt KIT patients, contrasting responders' samples (complete response, partial response, or stable disease; n ¼ 14; median ¼ 25.7 ng ml À1 ) vs non-responders' (progressive disease; n ¼ 19; median ¼ 10.1 ng ml À1 ).