Get started
Work

Sequential trial emulation for observational vaccine effectiveness studies

Year: 2026
Type: dissertation
Abstract: Vaccines are a cornerstone of public health. Because vaccine effects can depend on complex biological, epidemiological, and social processes which unfold over time, statistical considerations for measuring vaccine effectiveness (VE) are often context dependent (Halloran et al., 2010, Ch. 3). Evidence from randomized controlled trials is used to establish safety and efficacy of vaccines and forms the basis for vaccine approval. Once efficacious vaccines are available, it may be infeasible or unethical to conduct additional randomized controlled trials, and in turn, vaccine studies often rely on surveillance data or other routinely collected data (Halloran and Hudgens, 2018). Recent decades have seen rapid development of principled approaches for causal inference in the observational setting. This dissertation considers methods for assessing COVID-19 and pertussis VE via target trial emulation (TTE; Hernán and Robins, 2016; Hernán et al., 2016). The TTE approach involves conceptualizing hypothetical randomized trials, then mimicking the design and analysis of these trials using observational data. The first project describes a nested trial emulation (NTE; Hernán et al., 2005, 2008) method for estimating COVID-19 VE among elders in the Abruzzo region of Italy in 2021. COVID-19 vaccine effects may vary over calendar time (as new viral variants emerge), time since vaccination, or both. A series of weekly trials are emulated and trial-specific effect estimates are used to assess VE trends across both time scales. A NTE inverse probability weighted estimator and a test for heterogeneity in VE across trials are proposed. The second project extends the two-time-scale NTE method described above to accommodate competing risks. In the context of COVID-19, competing risks commonly occur. For example, if an individual dies from causes unrelated to COVID-19, they cannot subsequently die from COVID-19, so non-COVID-19-related death is a competing event for death from COVID-19. Common approaches, such as treating the competing event as a censoring event, may lead to misleading conclusions. In the second project, results obtained using a NTE analysis that accommodates competing events are compared with those obtained from a NTE analysis which treats competing events as censoring events. Causal inference methods often rely on the assumption of no interference between individuals, i.e., an individual’s potential outcomes do not depend on other individuals’ treatment. In many VE studies, this assumption is unlikely to hold because widespread vaccination can have indirect effects that reduce the risk of disease for both the vaccinated and the unvaccinated. The third project proposes a sequential trial emulation method for settings with a time-varying interference structure. A parametric g-formula approach is considered for estimating population-level vaccine effects. The proposed methods are motivated by a large observational database containing diphtheria, tetanus toxoids, and acellular pertussis vaccination records and pertussis diagnoses for children in King County, Washington. BIBLIOGRAPHYHalloran, M.E., Longini, I.M., and Struchiner, C.J. (2010). Design and Analysis of Vaccine Studies. Springer, New York, NY.Halloran, M.E. and Hudgens, M.G. (2018). Estimating population effects of vaccination using large, routinely collected data. Statistics in Medicine, 37(2):294–301. Hernán, M.A., Robins, J., and Garcia Rodríguez, L. (2005). Discussion on ‘Statistical issues arising in the women’s health initiative’ by Prentice, Ross L. et al. Biometrics, 61(4):922–930. Hernán, M.A., Alonso, A., Logan, R., Grodstein, F., Michels, K.B., Stampfer, M.J., Willett, W.C., Manson, J.E., and Robins, J.M. (2008). Observational studies analyzed like randomized experiments: An application to postmenopausal hormone therapy and coronary heart disease. Epidemiology, 19(6):766–779. Hernán, M.A. and Robins, J. M. (2016). Using big data to emulate a target trial when a randomized trial Is not available. American Journal of Epidemiology, 183(8):758–764. Hernán, M.A., Sauer, B.C., Hernández-Díaz, S., Platt, R., and Shrier, I. (2016). Specifying a target trial prevents immortal time bias and other self-inflicted injuries in observational analyses. Journal of Clinical Epidemiology, 79:70–75.
DOI: 10.17615/p6yq-0g05
Language: English
Open Access status: green
OpenAlex ID: W7118708373
Cites:
Cited by:
Related to: