Search:

Upcoming Events

2018-11-02

Tatsuki Koyama, PhD, Department of Biostatistics, Vanderbilt University

"What we don’t talk about when we talk about biostatistics: Challenges in data merging and treatment definition in a large-scale observational study."

As biostatisticians in a medical research project team, we are generally responsible for data analysis. This includes creating the analysis plan, executing statistical analyses, and drafting the results. Occasionally, in a truly collaborative setting, our contributions go far beyond data analysis. We may additionally be involved in generating hypotheses, defining the endpoints, or cleaning and preparing data sets. In this talk, I will discuss the CEASAR project (Comparative Effectiveness Analysis of Surgery And Radiation), focusing on the topics and issues that are not generally discussed in a research presentation by a biostatistician. CEASAR is an ongoing multi-site research study conducting long-term follow-up on more than 3,600 men diagnosed with localized prostate cancer between 2010 and 2011, and has resulted in 18 publications so far. The primary findings based on the three-year follow-up data were published in January, 2017, and this talk is mainly based on this paper (Barocas et al). In this project, a team of biostatisticians are responsible for everything associated with the data and data analysis. Data for this project came from multiple sources (patient questionnaires, medical chart abstraction, and cancer registry); the information was sometimes contradictory across data sources for some of the basic variables such as PSA (prostate specific antigen), Gleason score, treatment received, and date of treatment. Determining the treatment for each patient was the most difficult task in this project. In making this and other determinations, some of our related decisions about data cleaning and data analysis were made without informing the PI’s. I will present some of the issues we faced in data merging, treatment definition, missing data, and handling journal review processes.

Stay connected TwitterFacebookLinkedInYouTubeInstagram