Understanding the factors determining the distribution of parasites and pathogens in natural systems is essential for making predictions about the spread of emerging infectious disease. Here, we report the distribution of the fungal anther‐smut disease, caused by Microbotryum spp., on populations of the European wildflower Silene vulgaris over a range of elevations. A survey of several geographically distinct mountains in the southern French alps found that anther‐smut disease was restricted to high elevations, rarely observed below 1300 m despite availability of hosts below this elevation. Anther smut causes host‐sterility, and is recognized as a model system for natural host–pathogen interactions, sharing common features with vector‐borne and sexually‐transmitted disease in animals. In such systems, many biotic and abiotic factors likely to change over ecological gradients can influence disease epidemiology, including host spatial structure, pathogen infectivity, host resistance, and vector behavior. Here, we tested whether host population size, density, or connectivity also declined across elevation, and whether these epidemiologically relevant factors explained the observed disease distribution. We found that while none of these factor means changed across elevation, disease was significantly more likely to occur at both higher elevations and in larger populations, the majority of which were found above 1300 m. The break in disease incidence was also associated with an apparent scarcity of these larger host populations between 1000 and 1300 m in elevation. Examining variation in climatic factors among host populations, we also showed that the probability of disease was higher in areas with historically colder, wetter, and more stable conditions. The restricted distribution of anther‐smut disease in high‐elevation S. vulgaris provides an opportunity for empirical study on range limits and disease distribution in natural alpine communities that are considered particularly sensitive to the effects of climate change.