Project 4: Statistical Innovations

点击翻译按钮获取中文摘要

ABSTRACT The underlying theme of Project 4 is to develop and implement statistical methods to improve cancer risk prediction and to understand cancer etiology in several potentially novel ways. The availability of metabolomic data for Projects 1, 2 and 3 has the potential to improve risk prediction for breast, colorectal and ovarian cancer, and to identify new pathways underlying cancer development, yet there is also a high likelihood of discovering false positives. The analytic approaches in Aim 1 of the project address multiple comparison issues as well as look at associations between cancer incidence and (a) individual metabolites, (b) metabolic pathways, and (c) metabolic signatures, while carefully controlling for risk factors using well-established risk scores for each cancer site. Aim 2 proposes a strategy to account for screening in colorectal cancer (CRC) risk models. Since screening is one of the most important protective factors for CRC and may be associated with other behavioral risk factors, this is of high priority. Aim 3 proposes an innovative strategy for combining effects of risk factors for both cancer incidence and cancer mortality to identify disease-free women who are at high risk of lethal cancer; this aim is key to modeling strategies across Project 1 to 3, all of which propose to consider cancer survival and lethality. Aim 4 seeks to optimize the use of our unique 30 years of longitudinal risk factor data so as to weight reported exposures over time as predictors of risk. It expands on the standard concept of latency which only considers risk factors years in the past using a binary cutoff, by allowing a smoother weighting of exposures over multiple time periods. Aim 4 will also have broad utility across all the Projects. Finally, Aim 5 is concerned with short-term (or promotion) effects of changes in risk factors on cancer incidence, while simultaneously controlling for long-term effects. It addresses the clinical question of whether changes in risk factors over the short-term will result in corresponding changes in risk. Given the aging of the NHS cohort and the US population, this is of high clinical priority. Overall, this Project will promote the use of novel and improved statistical applications across the Projects of this P01 application, and in the scientific community. The close synergy between statisticians, clinicians, and epidemiologists which results from this Program Project application provides substantial added value in efficiently realizing the Aims of each Project and simultaneously expanding both scientific methods and achievements.