P01: A COVID-19 Epidemiologic Model to Enhance Efficiency Through Evidence-Based Site Selection for Vaccine and Treatment Trials
Poster Presenter
Debra Schaumberg
Vice President and Global Head, Startegic Development Consulting Chair
Evidera | PPD United States
Objectives
Due to variability in the local occurrence of repeated waves of COVID-19 community spread globally, an epidemiological model capable of predicting intensity of new COVID-19 cases over time at the level of a vaccine or treatment trial site’s catchment population is needed to inform site selection.
Method
We developed a hybrid dynamic transmission model of the COVID-19 caseload trajectory at a local level (WAVE). The model is calibrated to granular weekly case data and used to forecast expected cases around each potential trial site for the period of recruitment and primary endpoint follow-up.
Results
The model was successfully calibrated based on granular weekly case data at localized levels to capture the initial waves of exponential growth in new COVID-19 cases in each locality up to imposition of non-pharmaceutical interventions (NPIs), followed by an exponential decay in transmission after NPIs. The parameters are then carried forward until inflection points for subsequent waves are identified, at which time corresponding parameters are re-estimated to capture repeated cycles of regrowth and decay of new COVID-19 cases over time. Forecasts are output weekly for expected COVID-19 cased loads based on the most recent parameter sets for each metropolitan statistical area in the US, or similar level of granularity in other geographies. These forecasts undergo expert interpretation and are then contextualized with additional feasibility considerations for each trial’s specific protocol requirements. WAVE projections have been instrumental in selection of sites for several phase 3 vaccine trials, a basket treatment trial, and additional treatment trials. Sites were recommended for activation based on the expected case intensity at the time of patient recruitment and primary endpoint follow-up. For example, for an early phase 3 vaccine trial with US enrollment initiated in July 2020, sites throughout New England were rejected by WAVE while those in most California cities were recommended, before the summertime surges had emerged. In areas that WAVE predicted would surge, experience validated the WAVE predictions, and the trial met the case numbers needed to support its highly accelerated timelines for Emergency Use Authorization.
Conclusion
By leveraging a combination of growth and decay functions, a COVID-19 epidemiological model can fit granular case data and project expected COVID-19 case numbers at each location over time with enough accuracy to inform trial site selection. The model is instrumental in aiding trial efficiency through evidence-based enhanced feasibility, and in enabling Sponsors to achieve highly accelerated development timelines for COVID-19 vaccines and treatments. The model can also be used to inform non-COVID-19 trials and to monitor ongoing trails providing early warnings of impending case onslaught such that measures can be implemented to maintain trial integrity.
