Updated: Feb 7, 2022
Within the STriTuVaD project, the consortium is developing a methodology to generate in silico patients to enhance the data from the Phase II clinical trial that started in September with the first patients’ enrollment.
But how exactly is the information from computer simulations combined with data from the clinical trial?
The approach is detailed in a paper authored by Dr Kiagias et al., recently published on Frontiers in Medical Technology, where the authors propose a Bayesian hierarchical method for combining in silico and in vivo data into an augmented clinical trial using the UISS-TB simulator developed in the project.
The simulation platform
The Universal Immune System Simulator (UISS) previously described here, is an agent-based model (ABM) able to simulate the evolution of tuberculosis in the lung after RUTI vaccination.
This ABM produces in silico data from a number of biological entities and chemical species (e.g., cytokines) for an individual virtual patient, identified and characterised through an initial vector of 22 features (e.g. lymphocytes subpopulation levels, bacterial load, BMI).
From the virtual patient to the virtual cohort
To create whole cohorts of virtual patients, the group of Prof Juarez and Dr Kiagias at the University of Sheffield tuned a novel approach especially for UISS-TB, that samples the features mentioned above either at once or sequentially and, based on the joint distribution of the population characteristics, simulates each virtual patient using UISS-TB and the recorded endpoint of the clinical trial.
The goal of this strategy is to reduce the number of real patients needed to test the efficacy of the adjuvant vaccine. It aims at doing so by combining the information from the in silico data coming from the UISS-TB simulator with the in vivo data from the clinical trial that is taking place in India, and this is done using a Bayesian approach (detailed in this paper by Juarez et al.)
The augmented clinical trials
Both sources of information, in vivo and in silico, are combined using a novel statistical coherent Bayesian approach capable of propagating the uncertainty from both sources of information onto the posterior distribution of the clinical endpoint. The contribution of the in silico experiment is controlled by a measure of compatibility with the in vivo data and weighted accordingly into the combined trial.
Indeed, it is important to balance the information between the two sources and to not overwhelm the information from the in vivo trial.
Using the information from in silico models to enhance clinical trials would allow to decrease their size and duration, that currently can reach between 5 and 7 years from Phase I to Phase III, and potentially speeding up the commercialisation of drugs or vaccines. These improvements translate into a lower final cost to the public, a crucial aspect for those countries with the highest rate of incidence.