t-threads
private beta
search about

Assistant Professor of Biostatistics at @UF specializing in emerging infectious diseases and vaccine study design. @HarvardBiostats PhD. Tweets my own. She/her.

Jun 26, 2020 9:08 PM
copied to clipboard

What is key to a successful vaccine efficacy trial? Picking the right trial sites! If we place the trial in outbreak hotspots, we can quickly determine if the vaccine is effective for preventing disease.

A thread on our proposed strategy. 1/11 https://www.mobs-lab.org/uploads/6/7/8/7/6787877/ensembleforecastvaccines_12jun20.pdf

tweet media


Since our work on an Ebola ring vaccination trial in West Africa, our group’s research has focused on innovative strategies for implementing vaccine trials during outbreaks. How can we design these trials so they are more likely to succeed? 2/11 https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(15)61117-5/fulltext


A central challenge is that outbreaks are unpredictable across space and time. They start and stop, and hotspots change. Start a trial in one place (China, New York City), and the outbreak could end before you have accrued enough data. 3/11

tweet media


To avoid having to stop trials early that yield only inconclusive findings, we recommend using “core protocols.” These are large trials where new sites can be added over time, even pausing between outbreaks for diseases that pop up intermittently. 4/11 https://www.nejm.org/doi/full/10.1056/NEJMsb1905390


Imagine a trial that started in New York City, but now is picked up in Florida. This allows us to keep gathering data until we are sure that the vaccine works. 5/11


So a key question then is where do we start our trial. Mathematical and statistical models can be used to project future disease incidence. While their accuracy can be limited beyond short-term horizons, they are valuable tools for planning. 6/11


A simple approach is to look at which locations have the most new cases each day, but does this account for how much testing is being done? Models can integrate all available data (hospitalizations, deaths) to put areas on a level footing so we can make better comparisons. 7/11


Individual models have different strengths and weaknesses, so we suggest combining different models. @reichlab in particular has led great work on the value of “ensemble modeling.” This approach has previously been explored to design Zika trials. 8/11 https://www.npr.org/sections/health-shots/2020/05/13/855038708/combining-different-models-new-coronavirus-projection-shows-110-000-deaths-by-ju


For planning COVID vaccine trials, we suggest the formation of a modeling consortium, where independent groups submit short-term projections for potential vaccine trial sites, and these are combined in a stacking procedure. This gives a distribution for future incidence. 9/11

tweet media


Sites with highest projected incidence are prioritized (plus other considerations). This process can be repeated throughout the trial to help investigators decide whether they should continue at an existing site or whether they should shift the trial to a new location. 10/11


We believe that this data-driven approach will improve COVID vaccine efficacy trial planning. Reach out if you are interested in discussing further. With @apastorepiontti Zach Madewell @datcummings Matt Hitchings @KeyaJoshi3 @rebeccajk13 @alexvespi @betzhallo @ilongini 11/END