Han Liu: Advancing anticancer dose selection and individualization

Abstract

For many anticancer therapies, dose selection has historically been driven more by the need to demonstrate efficacy and expedite development than by a systematic characterization of the dose that optimally balances benefit and risk. Increasing recognition of the clinical consequences of suboptimal dosing has, however, underscored the need for a stronger quantitative foundation for both dose selection during development and dose individualization in clinical practice. The overall aim of this thesis was therefore to address methodological limitations in commonly applied pharmacometric analyses in oncology and thereby strengthen the evidence base for dose optimization.

To this end, pharmacometric modeling and simulation methods were developed and applied to clinical data for bintrafusp alfa and sunitinib. For bintrafusp alfa, a multistate framework integrating tumor response, progression, dropout, and death was used to evaluate dose–response in a dose-randomized study in non-small cell lung cancer. Although no individual transition showed a statistically significant dose effect, simulation of the full disease course supported a clinically meaningful survival advantage for 1200 mg over 500 mg, illustrating the value of integrative model-based evidence for dose selection.

For sunitinib in metastatic renal cell carcinoma, the thesis showed that commonly cited exposure–efficacy and toxicity–survival associations were substantially distorted by time-dependent bias and baseline confounding. After appropriate adjustment, the exposure–efficacy relationship was essentially flat within the observed range, and toxicity markers such as hypertension were not independently predictive of improved survival. In contrast, higher exposure was consistently associated with poorer tolerability, including earlier and more frequent dose reductions, a higher risk of treatment discontinuation due to toxicity, and greater patient-reported side-effect burden. Female patients had substantially higher exposure than males without improved efficacy, but with poorer tolerability, supporting exposure-driven sex differences in treatment burden. Model-based simulations further supported reduced starting doses in females and proactive therapeutic drug monitoring-guided dose de-escalation as strategies to improve tolerability without compromising expected efficacy.

Overall, this thesis demonstrates the value of integrative, bias-aware pharmacometric approaches for oncology dose optimization.