Julie Dudasova (1), Pavel Fiser (1), Ferdous Gheyas (2), Frank Liu (2), Jeffrey R. Sachs (2)
(1) MSD, Czech Republic (2) Merck & Co., Inc., Kenilworth, NJ, USA
Introduction: Protective efficacy of a vaccine is defined as the proportional reduction in risk of disease among vaccinated subjects compared to control (placebo vaccinated) subjects and is often assessed in randomized double-blinded controlled clinical trials [1]. Compared to other drugs and biologics, vaccines trials are particularly costly and lengthy [2,3]. This is due in part to the number of subjects in the trials, the costs of assays, and enrollment and observation periods of months to years to accrue the number of disease cases necessary to obtain a sufficiently precise estimate of efficacy.
In some trials, immune response post vaccination (immunogenicity) is evaluated in addition to the primary clinical endpoint, which enables identification of a potential immunological correlate of protection (CoP). CoP is a biomarker that can be used to reliably predict efficacy [4].
Establishing and using a CoP in vaccine clinical development can be valuable for:
- understanding the mechanism of protection
- identifying promising vaccine candidates
- evaluating immunogenicity in small-scale clinical trials and assisting go/no-go decisions
- reducing size or duration of large-scale clinical trials
- providing endpoints in bridging clinical trials
- providing additional support in vaccines evaluation.
Having a CoP for a particular pathogen (and vaccine mechanism) leads to faster development of more effective vaccines [5].
Objectives:
- To propose a new quantitative framework, Probability of Disease Bayesian Analysis (PoDBA), to estimate vaccine efficacy (and its confidence interval) using immune response biomarker data collected shortly after vaccination, when there is sufficient evidence that the biomarker is correlated with protection
- To compare the results from using PoDBA with the standard, case-counting based estimation using both real clinical trial data and simulated data
Methods: In the PoDBA method, we use a three-parameter sigmoid function to describe a relationship, the probability of disease (PoD) curve, between biomarker values and probability of disease. Vaccine efficacy is estimated by combining the PoD curve with biomarker distributions of vaccinated and control groups. The 95% confidence intervals (CIs) associated with estimated vaccine efficacy use non-parametric bootstrapping to estimate the uncertainty of the PoD curve and of the biomarker distributions.
Results: When applied to simulated data representing a wide range of scenarios, PoDBA produced accurate point estimates. The method outperformed case-count efficacy estimation in terms of precision; estimated efficacy values were closer to the true value in a larger percentage of simulations compared to the case-count method. The simulations demonstrated that a benefit of PoDBA is the ability to estimate efficacy with substantially narrower CIs than those of case-count efficacy. This enables better informed go/no-go decisions in vaccine development, as frequently continuing development (and approval) require the lower CI be above 25%. When applied to subject-level immunogenicity data from a clinical study of influenza vaccines, estimates of vaccine efficacy were accurate when compared to the standard estimates based on counting cases. Even though the case-count efficacy estimation is based on six times as many subjects as were in the immunogenicity sub-study used for PoDBA, the CIs of vaccine efficacy are similar for the two approaches. The PoDBA framework was also used on subject-level clinical trial data to estimate efficacy of vaccines against dengue and zoster. For both examples, the PoDBA-estimated efficacy and CI are consistent with the standard estimate, which implies that the biomarkers used for the estimations might be useful for predicting protection.
Conclusion: When there is sufficient evidence that an immune response biomarker is a correlate of protection, PoDBA provides an accurate and precise estimate of vaccine efficacy. The proposed method outperforms the case-count efficacy estimation on simulated data, as well as on data for influenza, zoster, and dengue vaccines. Such a method, that uses the whole distribution of antibody values and provides not only a point estimate of vaccine efficacy but also its CI, can be valuable in defining development strategy as well as in making informed data-driven decisions in vaccines research and development.
References:
[1] Halloran ME et al. Design and Analysis of Vaccine Studies. Springer, 2010.
[2] Gouglas D et al. Estimating the Cost of Vaccine Development against Epidemic Infectious Diseases: A Cost Minimisation Study. The Lancet, 6: 1386-1396, 2018.
[3] Morgan S et al. The Cost of Drug Development: A Systematic Review. Health Policy,100: 4-17, 2011.
[4] Plotkin SA et al. Plotkin’s Vaccines. Elsevier, 2017.
[5] Plotkin SA. Complex Correlates of Protection After Vaccination. Clinical Infectious Diseases, 56: 1458-1465, 2013.
Reference: PAGE () Abstr 9425 [www.page-meeting.org/?abstract=9425]
Poster: Methodology - New Modelling Approaches