Bridging Statistical Strategies for Censored Covariates (24w5160)


(University of North Carolina at Chapel Hill)

(Simon Fraser University)

Yanyuan Ma (Pennsylvania State University)


The Banff International Research Station will host the "Bridging Statistical Strategies for Censored Covariates" January 28 - February 2, 2024.

Diseases of aging, like Alzheimer, Parkinson, and Huntington disease, are expected to affect 153 million individuals worldwide by 2050. Treatments to prevent or slow these diseases will significantly decrease the projected impact, and modeling how disease symptoms worsen over time---the symptom trajectory---before and after a diagnosis can help evaluate if a treatment can prevent or slow a disease.

Yet modeling the symptom trajectory is not easy because these diseases of aging progress slowly over decades, so studies that track symptoms often end before a diagnosis can be made. This makes time to diagnosis \emph{right-censored} (i.e., a patient will reach the criteria for a diagnosis sometime after the last study visit, but exactly when is unknown), leaving researchers with the challenge of trying to model the symptom trajectory without full information about when diagnosis occurs.

The challenge creates a unique statistical problem of modeling the symptom trajectory as a function of a \emph{right-censored covariate}, time to diagnosis. Tackling this problem by modeling time to diagnosis has long been thought to be the best strategy, but new discoveries now show that when those models are even slightly wrong, that strategy produces biased results and incorrectly powered clinical trials. Ways to avoid this bias are so-called ``model-free strategies'' that do not require accurately modeling time to diagnosis. These model-free strategies recently emerged for related statistical problems, but have not yet been expanded for a right-censored covariate. This workshop will bring together clinician scientists, statisticians, and biostatisticians to bridge ideas and create working groups toward developing practical, model-free strategies for censored covariate problems. Results from the workshop can ultimately help to produce robust estimates of the disease symptom trajectory which can assist in designing clinical trials that are correctly powered to detect whether an experimental treatment is working.

The Banff International Research Station for Mathematical Innovation and Discovery (BIRS) is a collaborative Canada-US-Mexico venture that provides an environment for creative interaction as well as the exchange of ideas, knowledge, and methods within the Mathematical Sciences, with related disciplines and with industry. The research station is located at The Banff Centre in Alberta and is supported by Canada’s Natural Science and Engineering Research Council (NSERC), the U.S. National Science Foundation (NSF), Alberta’s Advanced Education and Technology, and Mexico’s Consejo Nacional de Ciencia y Tecnología (CONACYT).