An example of current project: Modeling for efficiency HIV Therapeutic vaccine during analytical treatment interruption
Although antiretroviral treatment (ART) have drastically improved the quality of life and life expectancy of people living with HIV, they cannot cure HIV. Stopping ART is accompanied by a rapid rebound of the viral load, with a harmful side-effect on the immune system (decline in CD4+ T lymphocytes). Consequently, HIV therapeutic vaccine development is warranted in the perspective of HIV cure or long-term viral control. Their efficacy is typically assessed in trials with Analytic treatment interruption (ATI), in which ART are interrupted over a period of time. Thanks to the data of 3 HIV therapeutic clinical trials, LIGHT, DALIA, ILIADE, that included a total of 265 patients who had HIV RNA load < 50 copies/mL for at least 6 months previous to the trials, we have evaluated what would be the major virological indicators to be used as primary endpoint in HIV therapeutic vaccine trials to characterise viral load dynamics and assess viral control. We compared (1) time to rebound (2) set point, (3) peak value of HIV RNA load, (4) slope of the viral rebound reflecting its speed and (5) area under the curve during ATI normalized by the calculation time (nAUC). Normalized AUC appears as a good primary endpoint for ATI protocols studies. Further work, will consist in combining this with a better use of imputation methods based on mechanistic models in a way that is acceptable in clinical protocols. This work has allowed the VRI Data Science division to better define the primary and secondary virological endpoints of the coming EHVA and VRI HIV therapeutic clinical trials.
|Edouard Lhomme, VRI Data Science division, SISTM team, Inserm U1219 – Inria, who has developed a new statistical approach to analyze the T cell response in vaccine clinical trial. This innovative approach may soon replace the conventional statistical approach used until now to analyze this type of data
Evaluation of immunogenicity is a key step in the clinical development of a novel vaccine strategy. T cell responses to vaccine candidates are often assessed using intracellular cytokine staining (ICS), which allows the characterization of subsets of specific cytokine-producing T cells after ex vivo antigenic stimulation. The conventional statistical approach currently used to analyze ICS data is to compare, between vaccine regimens, the percentages of cells producing a cytokine of interest after ex vivo stimulation by vaccine antigens, after subtracting the non-stimulated response (measured on unstimulated cells) of each sample. Subtraction of the non-stimulated response aims at capturing the specific response to the antigen used for stimulation, but raises methodological issues related to measurement error, decreased statistical power, and potential biased estimates. We needed a more efficient statistical approach to analyze ICS responses in vaccine trials for accurate estimation of vaccine effects, and we therefore developed a new approach using a bivariate model.
The bivariate linear regression model that we proposed jointly estimates the non-stimulated and antigen-specific ICS responses. These responses are both modelled according to the vaccine effect as the main explanatory variable, and the stimulated response is additionally adjusted on the non-stimulated response. The model captures the vaccine effect on the stimulated response, usually a key outcome in phase 2 vaccine clinical trials. The model also provides an estimation of the vaccine effect on the non-stimulated response and the effect of the non-stimulated response on the stimulated response (see Figure 1).
We benchmarked the statistical performances of the model in comparison with conventional approaches. Our results showed that this novel method is more flexible than conventional analyses methods, leading to a systematically accurate interpretation of the vaccine effect and more detailed results. We detected that the use of the conventional approach with subtraction of the non-stimulated response can lead to erroneous results and should no longer be recommended for the analysis of cellular immunogenicity in vaccine clinical trials.
Figure 1. Bivariate linear model for jointly estimating the non-stimulated and antigen-specific responses measured by intra cellular cytokine assay with a in vitro stimulation by peptide A and B.
The results will be published in the coming months. In the article, the statistical code will be available to use the bivariate model in both SAS and R statistical softwares.
For those who are non-comfortable with statistical software, we also built a user-friendly interface called VICI where immunologist can easily analyze their ICS data with the bivariate model and visualizing the results. A first version is available here (link: https://shiny-vici.apps.math.cnrs.fr/), any feedback are welcome.
With no available licensed vaccine or therapy against Ebola, the outbreak of Ebola virus disease in 2014-2016 caused more than 11 000 deaths in West Africa. At the moment an outbreak is still ongoing in Democratic Republic of Congo. The scientific community is working on vaccine solutions that could protect populations from Ebola virus disease.
The Data Science Division has successfully modelled the immune response to an heterologous Ebola vaccine in the context of the EBOVAC research project (Ebola development vaccine study, www.ebovac.org) and these results are published in the Journal of Virology. EBOVAC1 and EBOVAC2 are two Ebola vaccine development projects coordinated by Inserm and London School of Hygiene and Tropical Medicine (LSHTM), promoted by Janssen, a Johnson & Johnson pharmaceutical company and funded by the Innovative Medicines Initiative IMI2, where the MIC (Immunomonitoring core) and Clinical Core of the VRI are also deeply involved. These two projects are assessing the safety, tolerability and immunogenicity of new vaccines regimens by conducting several phase 1 and 2 clinical trials in Europe and Africa. These new vaccines regimens, intended to enhance the immune response and increase the duration of the response, are based on a prime with Ad26.ZEBOV vaccine followed by a boost with the MVA-BN-Filo,.
Based on data from EBOVAC 1 phase 1 trials in East Africa and Europe (UK, Kenya, Tanzania and Uganda), we modelled the dynamics of the humoral immune response from 7 days after the boost immunization onwards to estimate the durability of the response and understand its variability. The results show that antibody production is maintained by a population of long-lived cells. Estimation suggests that half of these cells can persist at least five years in humans.
These predictions are very encouraging for the use of this vaccine. New data from EBOVAC 1&2 phase 2 trials performed in Europe and Africa will be soon available and will help to check model predictions.
Link to scientific publication
Dynamics of the humoral immune response to a prime-boost Ebola vaccine: quantification and sources of variation
Chloé Pasin, Irene Balelli, Thierry Van Effelterre, Viki Bockstal, Laura Solforosi, Mélanie Prague, Macaya Douoguih, Rodolphe Thiébaut