Approaches and limitations of phosphatidylinositol-3-kinase pathway activation status as a predictive biomarker in the clinical development of targeted therapy.

The central role played by the class I(A) phosphatidylinositol-3-kinase (PI3K) signaling node in human cancer is highlighted in the multiple mechanisms by which these signals become dysregulated. Many studies suggest that constitutive PI3K activation in human cancer contributes to drug resistance, including targeted agents and standard cytotoxic therapy. The combination of activation mechanisms and the multiple downstream cascades that emanate from the PI3K node contributes to the difficulty in measuring PI3K activation as a biomarker. Although many agents suppress the pathway in models, the challenge remains to translate this biology into a patient selection strategy (i.e., identify patients with "PI3K activated" tumors) and subsequently link this biomarker definition to drug responses in patients. The various genetic and epigenetic lesions resulting in pathway activation necessitate combined approaches using genetic, genomic, and protein biomarkers to accurately characterize "PI3K activated" tumors. Such a combined approach to pathway status can be assessed using a statistical stratification of patients in a randomized trial into "pathway on" and "pathway off" subsets to compare the treatment effect in each arm. Instead of considering individual biomarkers for their predictive ability, this strategy proposes the use of a collection of biomarkers to identify a specific "pathway on" patient population predicted to have clinical benefit from a pathway inhibitor. Here, we review the current understanding of the mechanisms of PI3K activation in breast cancer and discuss a pathway-based approach using PI3K as a predictive biomarker in clinical development, which is currently in use in a global phase 3 setting.