Competing Heterogeneities in Vaccine Effectiveness Estimation
Understanding waning of vaccine-induced protection is important for both immunology and public health. The epidemiological data indicate that protection from the flu and COVID-19 vaccines could wane within a year. Our recent studies showed that the previously proposed extension of the Cox proportional hazards model by utilizing scaled Schoenfeld residuals used to estimate waning of vaccine effectiveness (VE) fails to accurately capture fast intraseasonal waning, especially when vaccination is spread over months. However, a relatively simple method based on including time-vaccine interaction in the model, with further proposed optimization, performs significantly better (Nikas et al., Clinical Infectious Diseases 2022). Population heterogeneities, both in underlying (pre-vaccination) susceptibility and vaccine response, add additional challenge in VE estimation, as they can cause measured VE to change over time even in the absence of pathogen evolution and any actual waning of immune responses. We use a multi-scale agent-based model parameterized using epidemiological and immunological data to investigate the effect of these heterogeneities. Our study suggests heterogeneity is more likely to ‘bias’ VE estimates downwards towards faster waning of immunity but a subtle bias in the opposite direction is also plausible.