Which 'imperfect vaccines' encourage the evolution of higher virulence?

Background and objectives: Theory suggests that some types of vaccines against infectious pathogens may lead to the evolution of variants that cause increased harm, particularly when they infect unvaccinated individuals. This theory was supported by the observation that the use of an imperfect vaccine to control Marek's disease virus in chickens resulted in the virus evolving to be more lethal to unvaccinated birds. This raises the concern that the use of some other vaccines may lead to similar pernicious outcomes. We examine that theory with a focus on considering the regimes in which such outcomes are expected. Methodology: We evaluate the plausibility of assumptions in the original theory. The previous theory rested heavily on a particular form of transmission-mortality-recovery trade-off and invoked other assumptions about the pathways of evolution. We review alternatives to mortality in limiting transmission and consider evolutionary pathways that were omitted in the original theory. Results: The regime where the pernicious evolutionary outcome occurs is narrowed by our analysis but remains possible in various scenarios. We propose a more nuanced consideration of alternative models for the within-host dynamics of infections and for factors that limit virulence. Our analysis suggests imperfect vaccines against many pathogens will not lead to the evolution of pathogens with increased virulence in unvaccinated individuals. Conclusions and implications: Evolution of greater pathogen mortality driven by vaccination remains difficult to predict, but the scope for such outcomes appears limited. Incorporation of mechanistic details into the framework, especially regarding immunity, may be requisite for prediction accuracy. Lay Summary: A virus of chickens appears to have evolved high mortality in response to a vaccine that merely prevented disease symptoms. Theory has predicted this type of evolution in response to a variety of vaccines and other interventions such as drug treatment. Under what circumstances is this pernicious result likely to occur? Analysis of the theory in light of recent changes in our understanding of viral biology raises doubts that medicine-driven, pernicious evolution is likely to be common. But we are far from a mechanistic understanding of the interaction between pathogen and host that can predict when vaccines and other medical interventions will lead to the unwanted evolution of more virulent pathogens. So, while the regime where a pernicious result obtains may be limited, caution remains warranted in designing many types of interventions.