Annual symposium: Translational Biostatistics – Bridging the Gap Between Complex Statistical Analyses and Clinically Actionable Information

April 29, 2022 | 9:30am – 1:00pm

This year’s annual symposium will focus on efforts to overcome translational barriers between statistical analyses of clinical studies and results that can be readily comprehended by clinicians and patients. The speakers will describe a number of opportunities for closing the gap and enhancing translation.

 

Agenda:

9:30 – 10:15am     
Lost in Translation and How to Fix it?
Lee-Jen (LJ) Wei, PhD
Professor, Department of Biostatistics, Harvard T.H. Chan School of Public Health

10:15 – 11:00am
Real World Evidence with Multi-Institutional EHR Data: Assessing Comparative Effectiveness of COVID Vaccines
Tianxi Cai, ScD
John Rock Professor of Population and Translational Data Science, Harvard T.H. Chan School of Public Health
Professor of Biomedical Informatics, Harvard Medical School

11:00 – 11:10am
Break

11:10 – 11:55am
How Should We Compare Survival Outcomes in Clinical Trials?
Rick Chappell, PhD
Professor, Department of Statistics, University of Wisconsin-Madison
Professor, Department of Biostatistics & Medical Informatics, University of Wisconsin Medical School

11:55am – 12:40pm
Recurrent Discussions of Recurrent Events
Brian Claggett, PhD
Assistant Professor, Harvard Medical School
Chief Statistician, Cardiac Imaging Core Laboratory and Clinical Trials Endpoints Center, Brigham and Women’s Hospital

12:40 – 1:00pm
Discussant 
Marc A. Pfeffer, MD, PhD
Distinguished Dzau Professor of Medicine, Harvard Medical School
Senior Member, Cardiovascular Division, Brigham and Women’s Hospital

 

Abstracts:

Tianxi Cai: Real World Evidence with Multi-Institutional EHR Data: Assessing Comparative Effectiveness of COVID Vaccines

Tianxi Cai, ScD
John Rock Professor of Population and Translational Data Science, Harvard T.H. Chan School of Public Health
Professor of Biomedical Informatics, Harvard Medical School

While clinical trials and cohort studies remain critical sources for studying disease progression and treatment response, they have limitations including the generalizability of the study findings to the real world, the limited ability to examine subgroup effects or test broader hypotheses, and the cost in performing these studies. In recent years, due to the increasing adoption of electronic health records (EHR) and the linkage of EHR with specimen bio-repositories and other research registries, integrated large datasets now open opportunities to generate real world evidence (RWE). Generating reliable RWE with EHR studies, however, remain highly challenging due to heterogeneity across healthcare centers in their patient population and health dynamics. In addition, sharing detailed patient level data cross institutions remains infeasible due to privacy constraints. In this talk, I will discuss federated approaches to study COVID vaccine effects across different patient subgroups to highlight both the value and challenges in using multi-institutional EHR data for RWE.

Rick Chappell: How Should We Compare Survival Outcomes in Clinical Trials?

Rick Chappell, PhD
Professor, Department of Statistics, University of Wisconsin-Madison
Professor, Department of Biostatistics & Medical Informatics, University of Wisconsin Medical School

Textbooks describing how to analyze time-to-event outcomes in clinical trials tend to list a limited range of topics. Differences are often quantified using hazard ratios from the Cox model and its associated score, the log-rank test. Weighted rank tests may be presented, along with comparisons of landmarks and quantiles. All these have their disadvantages in terms of interpretation, convenience, and/or power. I will discuss alternatives to the above in the context of non-inferiority and superiority trials. In addition to restricted mean life, additive hazard, and mixed additive-multiplicative hypotheses I will also comment on weighted log-rank tests. In the context of treatments with presumed delayed responses, such as many immunotherapy studies in cancer, researchers have proposed that early events be down-weighted. However, this has the consequence of rewarding early deaths.

Brian Claggett: Recurrent Discussions of Recurrent Events

Brian Claggett, PhD
Assistant Professor, Harvard Medical School
Chief Statistician, Cardiac Imaging Core Laboratory and Clinical Trials Endpoints Center, Brigham and Women’s Hospital

In many disease settings, patients experience multiple clinical events which may mark the severity or progression of disease. Historically, the primary analyses of Phase III clinical trials have preferred to focus only on the timing and occurrence of the patients’ first events. Despite intuitive advantages to collecting and analyzing more patient-level data, obstacles to embracing analyses of recurrent events have included lack of familiarity with appropriate methodology. Here we describe the evolution in the role of recurrent-event analyses in heart failure clinical trials.

Lee-Jen (LJ) Wei: Lost in Translation and How to Fix It?

Lee-Jen (LJ) Wei, PhD
Professor, Department of Biostatistics, Harvard T.H. Chan School of Public Health

One of the main goals of conducting a clinical, comparative study is to obtain robust, clinically interpretable treatment effect estimates with respect to harm-benefit perspectives at the patient’s level via efficient and reliable quantitative procedures. To accomplish this goal, it is important to know how to effectively  translate new developments in basic data science research into clinical research and practice. Unfortunately, some commonly used statistical procedures are not translational. That is, results of the analysis may be misinterpreted or difficult to comprehend. A notorious example is use of the p-value for clinical decision making, which is not an appropriate quantifier for assessing the clinical utility of a new therapy or strategy. In this talk, we will discuss several translational problems and present possible remedies.