The inclusion of translational medicine in the NIH Roadmap and FDA Critical Path Initiatives has motivated a reanalysis of how preclinical and clinical research may be bridged to hasten drug development and yield a clearer understanding of the relationships between preclinical and clinical studies. Academic drug discovery and development programs have arisen as a means to foster a “from the bench to the bedside” strategy. These programmatic efforts, although not always explicitly stated, renew the importance of PK/PD principles and investigations throughout the drug development process, in part, due to the quantitative pharmacological information that can be provided. In this regard, and as highlighted by the FDA (http://www.fda.gov/oc/initiatives/criticalpath/stanski/stanski.html
), model-based drug development is a key component of the translational science effort. However, much of the current focus has been in the clinical domain, and definitive model-based approaches in the preclinical setting have not been forthcoming. The current project was conducted based on the equivalent PK/PD dosing strategy (9
) that is representative of model-based drug development, and may prove to be a technique that enhances our ability to select the most efficacious drugs in an era of molecular and targeted chemotherapy.
A new component of the model development approach was to link the PD model to a tumor growth model, and thus, facilitate quantitative comparisons between drugs and tumor types in a preclinical setting. The efficacy model and link to the PD model were based on an approach recently described by Bueno L, et al (13
) that analyzed a pSMAD inhibitor. The approach of using two intermediate compartments (i.e
. INH1 and INH2) seemed applicable to our study since inhibition of pERK is amongst a cascade of events that result in apoptosis and register as a retardation in tumor size. The control and gefitinib-treated efficacy models were constructed to readily determine the drug effect as INH2 is nonzero only for gefitinib treatments, and Kgexp
and gamma were equal in the control and drug-treated models. This attribute of distinguishing the drug-dependent features enables the use of simulations to see how different dosing regimens alter tumor size relative to control, and may permit the elimination of costly size-based efficacy studies. It should be appreciated that the complex mechanisms of gefitinib activities involve other signaling pathways, such as PI3K-AKT, which may reveal other useful PD endpoints. Whether one or multiple PD endpoints can best be formulated into tumor size models will require further studies.
The equivalent PK/PD dosing strategy was designed to overcome the limitations of semi-empirical tumor-size based studies, and place PK/PD investigations at the center of preclinical drug development. The hypothesis that equal degrees of pERK inhibition afforded by gefitinib in the sensitive and resistant tumor variants will produce equal degrees of tumor growth retardation was supported, and suggests further examination of equivalent PK/PD dosing, for instance, to assess relationships between target inhibition and tumor growth for a series of tumors and drug candidates.
The development of a clinical PK/PD model for gefitinib based on patient PK properties and the preclinical pERK model was a novel feature of the investigation, and demonstrated through simulations the types of questions that can be addressed. Although the model is viewed as speculative it is believed that the model-based approach to address pharmacological questions may generate increased interest to obtain clinical data to further develop and validate tumor-based models. It was shown that patients possessing the vIII EGFR mutation in tumors require about a 50% lower dose than patients with wild-type EGFR, assuming normal PTEN and no other genetic differences in the two tumor types. The 24-hour dosing intervals, regardless of tumor type, maintained suppression of pERK over the 24-hour period once steady-state was achieved, although the extent of inhibition never exceed 80% at doses of 500 mg/d. The simple physiologically-based blood flow-limited brain tumor model is able to predict gefitinib concentrations due to changes in tumor volume, blood flow and drug distribution into tumor. The latter was used as a variable to represent possible changes in drug penetration and associated maximum brain tumor gefitinib concentrations within and between patients due to alterations in the BBB. The simulations showed the limited conditions in which 50% pERK inhibition at the nadir could be achieved, and baits the question of the degree and duration of inhibition that leads to antitumor activity. This type of uncertainty in the meaning of target inhibition (i.e
. pEGFR) and downstream effectors (i.e
. pERK and pAKT) has been expressed by Lassman and co-workers (18
) based on sparse data collected in brain tumor patients. It is believed that through the continued collection of sparse brain tumor samples from patients and the application of predictive models that these uncertainties can be mitigated. The application of the preclinical PD pERK model to patients was done without any variable scaling to account for possible differences between mouse xenograft gliomas and patient tumors. The pertinent PD variables in the target-response model are Kin
, and how these may be related between species will have to be addressed in a broader context with more patient data. It is also appreciated that the complexities of drug resistance, feedback signalling and crosstalk when multiple drugs are used will require a reassessment of the PD model. In summary, although the clinical PK/PD model is speculative it highlights a future direction of establishing links between preclinical and clinical models based on tumor measurements, sites that cannot be frequently sampled in patients.
In conclusion, this investigation developed a preclinical PK/PD/efficacy model predicated upon a molecular determinant of drug activity that highlighted the use of equivalent PK/PD dosing. Further, a clinical PK/PD model was also devised based on cancer patient data and naive scaling of a PD model that demonstrated target inhibition as a function of drug accumulation and tumor type. It is proposed that these preclinical and clinical modelling approaches can provide a foundation for drug development, and lead to the rational selection and use of anticancer drug candidates in different tumor types.