|Year : 2015 | Volume
| Issue : 3 | Page : 461-465
Population pharmacokinetic analysis to identify the possibility of interaction between anti-cancer agents
J Subramanian1, A Damre2, S Rohatagi3
1 Translational Research, Piramal Healthcare Limited, Mumbai, Maharashtra, India
2 Drug Metabolism and Pharmacokinetics, Piramal Healthcare Limited, Mumbai, Maharashtra, India
3 Translational Research; Drug Metabolism and Pharmacokinetics, Piramal Healthcare Limited, Mumbai, Maharashtra, India
|Date of Web Publication||18-Feb-2016|
Translational Research; Drug Metabolism and Pharmacokinetics, Piramal Healthcare Limited, Mumbai, Maharashtra
Source of Support: None, Conflict of Interest: None
Background: A number of molecularly targeted agents in oncology are tested in clinical studies in combination with conventional chemotherapy and/or radiotherapy. There is the possibility that the pharmacokinetics and dynamics of these targeted agents may be different when administered alone as against when administered in combination with other agents. AIM: The aim of this study is to understand the effects of addition of combination agents on the pharmacokinetics of a new, investigational, cyclin dependent kinase inhibitor anti-cancer drug (Compound A) using population pharmacokinetic (pop-PK) analysis. Materials And Methods: Integrated pop-PK analysis of data obtained from multiple phase I/II studies of Compound A, given alone or in combination with other agents. Results: A two compartmental model was found suitable to explain the pharmacokinetics of Compound A. No statistically significant influence of patient covariates or combination agents on the pharmacokinetic parameters of the central compartment was detected up to a significance level of 0.01. Model evaluation showed that the parameter estimates are stable and that the variability in the data was well reproduced by the model. Conclusions: This study represents the first time that a pop-PK analysis was performed in India for a targeted anti-cancer agent being developed in India. Such an analysis is useful to not only understand the influence of patient covariates and combination agents on the pharmacokinetics of a new investigational agent, but would also be valuable in the simulation of later phase clinical trials for the agent under development.
Keywords: Combination therapy, modeling and simulation, oncology, population pharmacokinetic modeling, targeted agents
|How to cite this article:|
Subramanian J, Damre A, Rohatagi S. Population pharmacokinetic analysis to identify the possibility of interaction between anti-cancer agents. Indian J Cancer 2015;52:461-5
|How to cite this URL:|
Subramanian J, Damre A, Rohatagi S. Population pharmacokinetic analysis to identify the possibility of interaction between anti-cancer agents. Indian J Cancer [serial online] 2015 [cited 2019 Jun 26];52:461-5. Available from: http://www.indianjcancer.com/text.asp?2015/52/3/461/176692
| » Introduction|| |
A number of molecularly targeted agents have gained approval from the Food and Drug Administration for the treatment of cancer and others are in various stages of clinical development., Increased pre-clinical and clinical experience with these agents has also led to an increased skepticism on whether single agent targeted therapies are capable of providing broad, long-term tumor control. Single agent targeted therapy is probably most effective only in those cancers where the growth and survival of the tumor is fully dependent on the signal inhibited by the agent. This scenario seems to be rare in most forms of solid tumors, though there are a few notable exceptions. Combining molecularly targeted agents with other targeted agents/conventional chemotherapy/radiotherapy is seen as a promising approach that can improve the clinical response and more importantly, provide sustained tumor control.
One of the considerations in the clinical development of combination therapies is the potential for pharmacokinetic interactions between agents. For example, increased exposure to gefitinib was noted in the carboplatin/paclitaxel regimen. Hence, assessment of pharmacokinetic interactions is seen as an important objective in clinical trials of drug combinations.
In the course of its initial clinical development, pharmacokinetic data on a new investigational agent is normally available from multiple trials, both where the agent is administered singly or administered in a combination regimen. An integrated analysis of such data using population pharmacokinetic (pop-PK) analysis could give valuable insights on the influence of addition of combination agents on the pharmacokinetic parameters of the new investigational agent. To illustrate this point, we study the case of Compound A, a cyclin dependent kinase (CDK) inhibitor, being evaluated in multiple proof-of-concept cancer clinical trials as monotherapy, as well as in combination with chemotherapy and radiotherapy.
| » Materials and Methods|| |
Time versus plasma concentration data from two phase I trials where Compound A was evaluated as a single agent in two different populations (studies A and B), a phase I/II trial where Compound A was evaluated in combination with chemotherapy (Gemcitabine, study C) and a phase I/II trial where Compound A was evaluated in combination with radiotherapy (study D) were used in the development of the pop-PK model. In all these trials, the route of administration for Compound A was intra-venous infusion. Data from all patients for whom both plasma concentration and demographic data were available was included in the pop-PK analysis.
In studies A and B (5-day treatment in a 21-day cycle), blood samples for pharmacokinetic assessments were collected in cycle 1 on day 1 and 5 at pre-dose and at 0.25, 0.5, 1, 2, 4, 6, 8, 12 and 24 h post-dose. In study C, Gemcitabine was administered on day 1 and Compound A was administered from days 1 to 5 per 21 day cycle. For assessment of pharmacokinetics of Compound A, blood samples were collected on days 1 and 5 at pre-dose and at 0.17, 0.5, 2, 4 and 24 h post start of a constant rate infusion administered over 0.5 h. In study D (5-day treatment in a 21-day cycle coupled with radiation), blood samples were collected on days 2 and 5 at pre-dose and at 0.17, 0.5, 2, 8 and 24 h post start of a constant rate infusion administered over 0.5 h.
Whole blood samples (approximately 5-6 mL) were collected in polypropylene containers using dipotassium ethylenediaminetetraacetic acid as an anticoagulant and immediately centrifuged at 4,000 g for 15 min at 4°C to separate plasma. Plasma was transferred to cryovials (2 aliquots of ~1.5 mL each) and stored at −70°C to −80°C until analysis by a validated liquid chromatography-mass spectrometry/mass spectrometry (LC-MS/MS) method in human plasma.
The concentrations of Compound A for pharmacokinetic evaluation were determined using a precise and accurate LC-MS/MS method in human plasma using carbamazepine as internal standard. Compound A was extracted from human plasma using liquid-liquid extraction procedure and injected into the liquid chromatography (LC) coupled with tandem mass spectrometry/mass spectrometry (MS/MS) detector. The reconstituted samples were chromatographed on ACE C18, 15 mm × 4.6 mm, 3 µm column using a mobile phase consisting of high-performance liquid chromatography grade methanol and formic acid buffer. The method was validated over a concentration ranges from 10.01 ng/mL to 5003.71 ng/mL. Compound A and internal standard were monitored by LC-MS/MS in the multiple reaction monitoring mode using the mass transition 402.20 → 359.00 amu for Compound A and 237.20 → 194.10 for carbamazepine. A weighted linear regression with weighting 1/concentration 2 was used to determine the concentration of Compound A in plasma samples.
Pop-PK models were developed using non-linear mixed effects modeling. Since an analysis of steady state did not reveal a statistically significant difference between day 1/day 2 and day 5 exposures for any of the studies, only day 1/day 2 time-plasma concentration data was utilized for the model development. Initial exploratory analysis (data not shown) had suggested that Compound A follows two-compartmental pharmacokinetics [Figure 1]. Inter-subject variability was described using an exponential function. A combination of additive and proportional error model was used to describe the residual variability. The first order conditional estimation with interaction was used as the estimation method.
The effects of age, body weight, gender, combination therapy (none/chemo/radiation) and tumor type (gastro-intestinal/genito-urinary/others) were studied on all the PK parameters in the two-compartmental model, namely, clearance (CL), volume of distribution of the central compartment (V1), inter-compartmental clearance (Q) and volume of distribution of the peripheral compartment (V2). Since all clinical studies included only subjects with adequate renal function (serum creatinine within 1.5 times the upper normal value), creatinine clearance was not included in the covariate analysis.
Covariates were entered in a multiplicative fashion. Continuous covariates were included in the pop-PK model as:
Ti = Ttypical* (Xi)θeff
and categorical (0/1) covariates were included as:
Ti = Ttypical* θeffXi
where Ti is the value of a PK parameter for the ith individual, Ttypical is the typical value of that parameter in the population, Xi is the value of the covariate for the ith individual and θeff is the effect of the covariate on the population value of the parameter.
For each pharmacokinetic parameter, covariates were included in the model one at a time. Full and reduced models (one variable less) were compared using the change in their respective objective function values (OFVs). The OFV is equal to minus twice the log-likelihood of the data and is an indicator of the goodness of fit of the model to the data. The decision to include the covariate was made if the change in the OFV was statistically significant at the 0.01 level using a Chi-squared test at 1 degree of freedom. During the initial individual testing of covariates, four covariates (combination therapy: Chemo, combination therapy: Radiation, weight and tumor type: Gastro-intestinal) were found to be statistically significantly associated with Q and V2 (P < 0.01). The final pop-PK model was selected using forward selection on the base two-compartmental model and including the above four covariates one at a time. Backward elimination from the full model including all the four covariates also resulted in the same final model. The full series of models tested is listed as supplementary information.
Plots of observed concentrations versus population predicted values were used as a visual check of the goodness of fit and sufficiency of the model. For a relatively complete model, the predicted versus observed plots should show a close scatter around the line of unity.
The stability of the final pop-PK model was assessed using a non-parametric (unstratified) bootstrap analysis. A total of 500 bootstrap samples were generated from the original dataset. The parameters for the final model were estimated for each of these replicate samples and the median and 90% confidence intervals (CI) for these parameter estimates were calculated and compared with the parameters estimated from the original dataset.
The ability of the final pop-PK model to reproduce the variability in the observed data was assessed through a visual predictive check (VPC). For the VPC, the final model and the corresponding parameter estimates were used to simulate the concentration data 1,000 times for each patient. At each nominal time point, the median and the 90% CI were calculated for the simulated data. The actual observed concentrations was overlaid with the simulated data and visually compared. The percentage of actual observations falling outside 90% CI was noted.
PDx-Pop version 4.0 running NONMEM (version 7, ICON Development Solutions, Ellicott City, MD, USA) running on a PC (Pentium 4, MS Windows XP) was used for the development of the pop-PK model. All graphs were created using R (2.12.1) software.
| » Results|| |
Subject demographic information for the data used for developing the pop-PK model is summarized in [Table 1].
The parameter and variability estimates for the two-compartmental model before the inclusion of any subject covariates is given in [Table 2]. The final model obtained after covariate analysis using forward selection or backward elimination was the same and given by:
TVCL = θ1; TVV1= θ2;
TVQ = θ3 * (θ5) Chemo * (θ6)Radiation; TVV2= θ4 * (body weight)θ7
The parameter and variability estimates for the final model are given in [Table 3]. A significant influence of any of the available covariates on the clearance or volume of distribution of the central compartment was not detected. Covariate analysis showed an influence of combination agents on the inter-compartmental clearance [Table 3]. This was however not reflected as a statistically significant difference in the peak exposures of Compound A. A one-way ANOVA to compare the effect of combination agents (none, chemotherapy, radiation) on dose normalized Cmax (Cmax_D) was not statistically significant at the 0.05 significance level (P = 0.75) [Figure 2].
|Table 3: Parameter estimates for the final model after inclusion of covariates|
Click here to view
Plots of observed versus population and individual predicted observations for the final pop-PK model are shown in [Figure 3] and weighted residual plots are shown in [Figure 4]. These plots show that the model predictions are in reasonable agreement with the observed data.
|Figure 3: Observed versus predicted plot for the best population pharmacokinetic model after covariate analysis|
Click here to view
|Figure 4: Weighted residual (WRES) plots (left) WRES against PRED (right) WRES against Time. Dashed horizontal lines at WRES = ±4 are provided as a reference|
Click here to view
In the bootstrap analysis, 83% of the bootstrap runs converged successfully. The median values for the pharmacokinetic parameters estimated using bootstrap are in close agreement with the parameters estimated from the original dataset, implying that the parameter estimates are on an average stable [Table 3]. The results of the bootstrap analysis also showed narrow CIs for all parameter and variability estimates, except for the estimates of the volume of distribution of the peripheral compartment.
The results of the VPC demonstrate that the variability in the observed data is well reproduced in the simulation using the final model [Figure 5]. Only 7% of the actual exposure values were seen to fall outside the 90% confidence bounds obtained through simulations.
|Figure 5: Visual predictive check for the final population pharmacokinetic model. Around 7% of the actual observed exposure values are seen to fall outside of the 90% confidence interval bounds obtained through simulations. (Inset) Actual and simulated exposures after removing outliers|
Click here to view
| » Discussion|| |
Targeted agents in oncology are reportedly associated with highly variable pharmacokinetics. Population pharmacokinetic analyses are conducted in an attempt to explain this variability in terms of subject characteristics., However, most of these studies have been conducted for pharmacokinetic data from trials where the drug has been either administered only as monotherapy,, or when it has been administered only in a combination regimen. Since it is likely that the already high variability in the pharmacokinetics of targeted agents may be compounded in a combination regimen, an integrated analysis is important. Such analysis would be valuable in delineating the effects of the combination agent on the pharmacokinetic parameters while accounting for differences in other subject covariates.
The objective of the present study was an integrated pop-PK analysis of pharmacokinetic data obtained from multiple trials of a CDK inhibitor, being developed in India, in order to understand the influence of subject characteristics and combination agents on the pharmacokinetics of this targeted agent. Such an integrated pop-PK analysis has not been reported for any CDK inhibitor. The value of such analysis has been reported time and again, as for example through the recent study that showed the synergistic anticancer effect of the capecitabine and docetaxel combination in metastatic breast cancer through a population based approach.
The most common drug-drug interactions (DDIs) of an investigational agent are associated with alterations in drug clearance, primarily due to inhibition or induction of drug-metabolizing enzymes, particularly cytochromes P450 (P450s) leading to changes in drug concentrations. Hence, in the pre-clinical setting, in vitro drug metabolism data is increasingly being used to predict the DDI potential of compounds in the clinic. Accordingly, for Compound A too, the in vitro drug metabolism data was initially used to assess its clinical DDI potential.In vitro CYP inhibition studies in human liver microsomes indicated that Compound A inhibited CYP2C19, CYP2D6 and CYP3A4. However, Compound A being administered intravenously in the clinic and due to its moderate-high clearance, the concentrations in plasma were expected to decline rapidly following its administration. Moreover, Compound A was metabolized by multiple CYPs and also directly glucuronidated. Thus, a metabolism-based pharmacokinetic DDI was also not anticipated in the clinic for Compound A. This was further substantiated by the pop-PK analysis of the clinical data.
The integrated pop-PK analysis failed to detect any statistically significant influence of patient covariates or combination agents on the clearance or volume of distribution of the central compartment (CL and V1) up to a significance level of 0.01. Though some of these covariates seemed to influence the parameters related to the peripheral compartment, incorporation of combination agents did not result in a statistically significant difference in the (dose normalized) peak exposure levels of Compound A. The influence seen on the peripheral compartmental parameters may therefore only signify statistical significance but not clinical importance. Interestingly, the pharmacokinetics of Compound A with and without combination with Gemcitabine was earlier studied in mice, the pre-clinical animal model. Analysis of data from this pre-clinical study had also shown no significant differences in disposition (Cmax, t1/2 and AUClast) of Compound A in the presence of Gemcitabine (data not shown).
Evaluation of the final pop-PK model using bootstrap showed the parameter estimates to be stable. The VPC following simulations from the final pop-PK model further showed that the variability in the original data was well reproduced by the model. This makes the final pop-PK model useful for simulations of later phase trials of Compound A either as monotherapy or in combination with chemo or radiotherapy.
Though the inclusion of covariates in the pop-PK model reduced some of the variability in the estimates of the pharmacokinetic parameters, there is still considerable variability in some of the parameter estimates. Incorporation of pharmacogenetic covariates is likely to explain some of the residual inter-individual variability in the parameters. However, these covariates were not measured in the trials and hence could not be analyzed.
| » Conclusions|| |
To conclude, a single integrated analysis of data pooled from multiple proof-of-concept clinical studies helped understand the effects of subject covariates and combination agents on the pharmacokinetics of a new CDK inhibitor anti-cancer agent. This analysis also reiterates the importance of collecting as much information as possible on subjects even for early stage clinical studies. Such information would be valuable in explaining the variability in the pharmacokinetics early-on, appropriately designing further late phase trials and ultimately in the personalization of treatment.
| » References|| |
Kwak EL, Clark JW, Chabner B. Targeted agents: The rules of combination. Clin Cancer Res 2007;13:5232-7.
Fojo T, Parkinson DR. Biologically targeted cancer therapy and marginal benefits: Are we making too much of too little or are we achieving too little by giving too much? Clin Cancer Res 2010;16:5972-80.
Humphrey RW, Brockway-Lunardi LM, Bonk DT, Dohoney KM, Doroshow JH, Meech SJ, et al
. Opportunities and challenges in the development of experimental drug combinations for cancer. J Natl Cancer Inst 2011;103:1222-6.
Hammond LA. Pharmacokinetic evaluation of gefitinib when administered with chemotherapy. Clin Lung Cancer 2003;5 Suppl 1:S18-21.
Beal S, Sheiner LB, Boeckmann A, Bauer RJ. NONMEM User's Guides (1989-2009). Ellicott City, MD, USA: Icon Development Solutions; 2009.
R Development Core Team. R:A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. Vienna, Austria: ISBN 3-900051-07-0; 2010. Available from: http://www.R-project.org/
Baker SD, Hu S. Pharmacokinetic considerations for new targeted therapies. Clin Pharmacol Ther 2009;85:208-11.
Jain L, Woo S, Gardner ER, Dahut WL, Kohn EC, Kummar S, et al
. Population pharmacokinetic analysis of sorafenib in patients with solid tumours. Br J Clin Pharmacol 2011;72:294-305.
Petain A, Kattygnarath D, Azard J, Chatelut E, Delbaldo C, Geoerger B, et al
. Population pharmacokinetics and pharmacogenetics of imatinib in children and adults. Clin Cancer Res 2008;14:7102-9.
Zandvliet AS, Schellens JH, Dittrich C, Wanders J, Beijnen JH, Huitema AD. Population pharmacokinetic and pharmacodynamic analysis to support treatment optimization of combination chemotherapy with indisulam and carboplatin. Br J Clin Pharmacol 2008;66:485-97.
Frances N, Claret L, Bruno R, Iliadis A. Tumor growth modeling from clinical trials reveals synergistic anticancer effect of the capecitabine and docetaxel combination in metastatic breast cancer. Cancer Chemother Pharmacol 2011;68:1413-9.
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5]
[Table 1], [Table 2], [Table 3]