Welcome to the Etzioni Lab: Modeling and Data Science for Population Studies
The Etzioni Lab focuses on innovative statistical and computer modeling to study cancer control outcomes. Our team includes experts in simulation modeling, survival analysis, Bayesian methods, statistical programming, and data visualization. We aim to develop a deeper, mechanistic understanding of cancer progression to project intervention benefits and harms.
Multi-cancer early detection:
- Examining methods to estimate diagnostic performance across studies.
- Building models to support clinical recommendations for multi-cancer screening.
- Addressing evidence gaps in evaluating novel cancer diagnostics.
Major methodologic interests:
- Estimation of lead time and overdiagnosis.
- Quantifying overdiagnosis in prostate cancer screening in the US.
- Breast cancer screening overdiagnosis analyses.
- Development of Bayesian models for studying longitudinal biomarker trajectories and disease transitions.
- Studying prostate cancer progression in men on active surveillance and those with biochemical recurrence.
Applied research:
- Leads the Biostatistics Core for the Pacific Northwest Prostate Cancer SPORE.
- Central consulting resource for prostate cancer investigators at Fred Hutch and the University of Washington.
Recent projects have used modeling to:
- Develop a model to update breast cancer screening trials so that they reflect screening efficacy under contemporary treatments
- Establish best practices for overdiagnosis estimation in clinical trials and population studies
- Develop risk-stratified cancer screening policies that are tailored to the way in which risk of disease varies across population strata
- Develop a model for the natural history of PSA-recurrent prostate cancer and estimate the frequency of overdiagnosis of recurrence
- Study the harm-benefit tradeoffs of many competing prostate cancer screening policies to support national recommendations
- Project the impacts of novel prostate screening biomarkers on mortality and overdiagnosis given marker performance characteristics