Abstract Plant disease often increases with N, decreases with CO 2 , and increases as biodiversity is lost (i.e., the dilution effect). Additionally, all these factors can indirectly alter disease by changing host biomass and hence density-dependent disease transmission. Yet over long periods of time as communities undergo compositional changes, these biomass-mediated pathways might fade, intensify, or even reverse in direction. Using a field experiment that has manipulated N, CO 2 , and species richness for over 20 years, we compared severity of a specialist rust fungus ( Puccinia andropogonis ) on its grass host ( Andropogon gerardii ) shortly after the experiment began (1999) and twenty years later (2019). Between these two sampling periods, two decades apart, we found that disease severity consistently increased with N and decreased with CO 2 . However, the relationship between diversity and disease reversed from a dilution effect in 1999 (more severe disease in monocultures) to an amplification effect in 2019 (more severe disease in mixtures). The best explanation for this reversal centered on host density (i.e., aboveground biomass), which was initially highest in monoculture, but became highest in mixtures two decades later. Thus, the diversity-disease pattern reversed, but disease consistently increased with host biomass. These results highlight the consistency of N and CO 2 as drivers of plant disease in the Anthropocene and emphasize the critical role of host biomass—despite potentially variable effects of diversity—for relationships between biodiversity and disease.