This package addresses this exact challenge via the multinomial-Poisson transformation. You can call glmer.mp() to build your model and Anova.mp() to get an ANOVA-like table of effects. https://cran.r-project.org/web/packages/multpois/index.html