Local control of the treated lesion is a commonly used endpoint in radiation oncology trials comparing two radiation doses or methods of delivery or assessing the impact of adding an additional therapy. It is often assumed that a treatment which improves local control will likely also improve overall survival. However this assumption is controversial and often debated. In cancers such as pancreatic cancer, treatment consists of 3 or more components (e.g. chemo, RT and targeted agents or immunotherapy). In early trials seeking to identify dose levels for such tri-modality therapy tradeoffs must be made between dose levels of RT and the systemic treatments. To makes these trade-offs optimally, requires knowledge of the relative impact of local vs distant progression on overall survival.
In this project we will conduct a meta-analysis of the impact of both local and distant progression on OS using individual patient data from a large number of completed phase III trials. Local and distant progression will be modeled as time-dependent covariates in Cox models for OS also including baseline covariates such as age, gender and randomization arm. We will estimate the relative impact (hazard ratio) of local and distant progression on the hazard of death within distinct cancer types (e.g. Head and Neck, Lung, …). Using multi-state models we will also seek to estimate the expected improvement in OS at fixed timepoints resulting from varying degrees of improvement in LC. Finally, we will assess the extent to which local or distant progression can serve as a surrogate endpoint for overall survival in trials of radiation therapy.