Pharmacometrics to improve clinical benefit assessment in oncology

Sammanfattning: The high attrition rate in oncology drug development calls for new approaches that would increase the understanding of drugs’ efficacy and safety profiles. This thesis focuses on the development of pharmacometric models to characterize and quantify the relationships between drug exposure, circulating and imaging biomarkers, adverse effects, overall survival (OS), and patient-reported outcomes (PROs).In axitinib-treated metastatic renal cell carcinoma patients, exposure-driven changes in soluble VEGF receptor 3 were linked to tumor size dynamics, which could in turn predict OS better than biomarker- or hypertension-related predictors. In sunitinib-treated gastro-intestinal stromal tumor (GIST) patients, the tumor metabolic response was sensitive to sunitinib dosing schedule and a substantial inter-lesion variability was quantified. A more pronounced decrease in tumor metabolism for the lesion that best responds to treatment after one week was predictive of longer OS. In imatinib-treated GIST patients, tumor volume better detected size changes of liver metastases and were slightly more predictive of OS than conventional tumor diameters, while tumor density had no predictive value.A new modeling approach, the minimal continuous-time Markov model (mCTMM), was developed to facilitate the analysis of ordered categorical scores with Markovian features, e.g. fatigue or hand-foot syndrome grades. The mCTMM is applicable when existing approaches are not appropriate (non-uniform assessment intervals) or not easily implemented (variables with large number of categories).An item response theory pharmacometric framework was established to describe longitudinal item-level data of a PRO questionnaire, the Functional Assessment of Cancer Therapy-Breast (FACT-B). Four correlated latent well-being variables characterized the multi-dimensional nature of FACT-B. When applied to data from breast cancer patients, the progression of physical well-being was typically better in patients treated with ado-trastuzumab emtansine (T-DM1) than with capecitabine-plus-lapatinib-treated patients. No relationship was identified between T-DM1 exposure and any of the latent variables.In summary, the developed models advance the use of pharmacometrics in assessing the clinical benefit of anti-cancer therapies. They provide a quantitative understanding of the desired and adverse responses to drugs, and their relationships to exposure and long-term clinical outcome. Such frameworks may help to early assess response to therapy and optimize dosing strategies for investigational or existing therapies.