Trial designs and biostatistics for molecular-targeted agents

Abstract

The heterogeneity of molecular alterations identified in cancer has transformed drug discovery and treatment paradigms for the disease. Today, many new compounds have been identified which target specific defects or cellular pathways that are dysregulated in multiple tumor types. Clinical research and practice under the paradigm of personalized medicine requires integration of biomarkers that measure the molecular phenotype of the individual’s tumor. Here, prospective biomarker-driven studies are characterized as marker-stratified, marker-enriched, or marker-directed designs with discussion of the strengths and limitations. Umbrella and basket trials allow for drug development to span traditional cancer types. Adaptive designs allow for biomarker-driven trials to gain efficiency based on iterative evaluations of the performance of the assay. Completed and ongoing trials in breast cancer are given throughout as examples of biomarker-driven studies. Statistical inferences and estimation must account for the complexity of designs, including procedures to account for multiple comparisons, and can use Bayesian statistics to model the flexibility of adaptive designs.