Biomarkers that predict the effects of therapeutic or preventative treatments hold great promise for im- proving health outcomes and decreasing medical costs. For example, if a treatment is thought likely to bene?t only a subset of subjects, a biomarker that identi?es these subjects can be used to recommend the treatment to them, and allow others to pursue alternatives. Potential treatment selection biomarkers, also called predictive or prescriptive biomarkers, are being produced in abundance due to technological advancements and a heightened interest in ?personalized medicine?, and yet a comprehensive statistical framework for study design and analysis of biomarker performance is lacking. We propose to build on a successful research program to develop novel statistical methods for marker identi?cation and evaluation, with the ultimate goal of advancing such a framework.
Aim 1 develops novel statistical analysis methods for discovering and evaluating markers for guiding treatment. The ideal setting for marker evaluation is a randomized and controlled trial. For this setting we will develop methods that address challenges not accommodated by existing methodology. For early-phase marker studies, which are typically not random- ized trials, methods for evaluating a biomarker's potential performance do not yet exist and our research will ?ll this gap. Methods will also be developed that incorporate medical cost data into the evaluation of a biomarker.
Aim 2 will develop novel study designs for discovering and evaluating markers for guiding treatment. A basic ?rst step in study design is identifying what is the desired performance of the biomarker, and we will develop techniques for this. Study designs will then be developed to assess whether a marker achieves this standard. Early-phase studies? either cohort studies or single arm trials? will be sized to evaluate a marker's potential performance. Novel randomized trial designs will be developed to de?nitively assess marker performance and will require smaller sample sizes than existing designs. A sequence of studies will be put forth for developing a biomarker, from early-phase studies of potential performance to late-phase studies of actual performance. The research will be conducted by an inter-disciplinary team of investigators with extensive expertise in biomarker evaluation, clinical trial design and analysis, health economics, and clinical research. Collaborations with cooperative groups such as the Early Detection Re- search Network, focused on biomarker discovery and evaluation, and SWOG, focused on clinical trials of cancer prevention and therapeutic strategies, will ensure that the research has immediate application and impact. Studies to which the methods will be applied include one that aims to identify women at high risk for epithelial ovarian cancer who can be recommended prophylactic removal of the fallopian tubes at the time of hysterectomy; and another that seeks to identify women with estrogen-receptor-positive breast cancer who are unlikely to bene?t from adjuvant chemotherapy.

Public Health Relevance

Markers that predict the likelihood an individual will bene?t from a medical intervention for disease pre- vention or treatment hold great potential public health value. This proposal will develop standards for the statistical design and analysis of studies discovering and evaluating candidate markers. These standards will help distinguish the good markers from the bad, optimize how the markers are used to make treatment recommendations, and ensure that research studies are designed so that the markers can be properly evaluated.

National Institute of Health (NIH)
National Cancer Institute (NCI)
Research Project (R01)
Project #
Application #
Study Section
Cancer Biomarkers Study Section (CBSS)
Program Officer
Ossandon, Miguel
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Fred Hutchinson Cancer Research Center
United States
Zip Code
Dai, James Y; Liang, C Jason; LeBlanc, Michael et al. (2018) Case-only approach to identifying markers predicting treatment effects on the relative risk scale. Biometrics 74:753-763
Buas, Matthew F; Li, Christopher I; Anderson, Garnet L et al. (2018) Recommendation to use exact P-values in biomarker discovery research in place of approximate P-values. Cancer Epidemiol 56:83-89
Kang, Chaeryon; Janes, Holly; Tajik, Parvin et al. (2018) Evaluation of biomarkers for treatment selection using individual participant data from multiple clinical trials. Stat Med 37:1439-1453
Kerr, Kathleen F; LeBlanc, Michael; Janes, Holly (2017) Comparisons of cancer staging systems should be based on overall performance in the population. Clin Trials 14:659-660
Kerr, Kathleen F; Janes, Holly (2017) First things first: risk model performance metrics should reflect the clinical application. Stat Med 36:4503-4508
Kerr, Kathleen F; Brown, Marshall; Janes, Holly (2017) Reply to A.J. Vickers et al. J Clin Oncol 35:473-475
Janes, Holly; Brown, Marshall D; Crager, Michael R et al. (2017) Adjusting for covariates in evaluating markers for selecting treatment, with application to guiding chemotherapy for treating estrogen-receptor-positive, node-positive breast cancer. Contemp Clin Trials 63:30-39
Kerr, Kathleen F; Brown, Marshall D; Zhu, Kehao et al. (2016) Assessing the Clinical Impact of Risk Prediction Models With Decision Curves: Guidance for Correct Interpretation and Appropriate Use. J Clin Oncol 34:2534-40
Pepe, Margaret S; Janes, Holly; Li, Christopher I et al. (2016) Early-Phase Studies of Biomarkers: What Target Sensitivity and Specificity Values Might Confer Clinical Utility? Clin Chem 62:737-42
Kang, Chaeryon; Huang, Ying; Miller, Christopher J (2015) A discrete-time survival model with random effects for designing and analyzing repeated low-dose challenge experiments. Biostatistics 16:295-310

Showing the most recent 10 out of 32 publications