Modelling the ecological niche of emerging infectious diseases: The case of Mycobacterium ulcerans causing Buruli ulcer in humans

Background/Question/Methods

Ecological epidemiology often undertakes ecological niche modelling (ENM) of pathogenic agents, or their vectors, when seeking methods to control environmental diseases. In cases where the vector of a disease is unknown, we can use ENM to explore the likelihood that a particular species is a vector. However, ENM is approached differently by epidemiologists and ecologists; it requires careful consideration of the ecology of the species, and is often hampered by inappropriate selection of sample sites, ecologically unrealistic delineation of study areas, or inappropriate model evaluation techniques. An example of this is the pathogen Mycobacterium ulcerans, the causative agent of Buruli ulcer. Buruli ulcer is an emerging neglected tropical disease occurring in West Africa, French Guiana and Australia, transmitted to humans from an unknown environmental reservoir via an unknown vector. Recent work has suggested, controversially, that M. ulcerans could be transmitted by biting aquatic insects (Naucoridae). We investigate the ecological niches of M. ulcerans and Naucoridae, identifying the macro-scale environmental determinants of these species. We present findings on the ecological suitability of Naucoridae to act as vectors of M. ulcerans, and use this example to highlight the general differences in ecological and epidemiological approaches to niche modelling.

Results/Conclusions

We find ecological differentiation among nine families of aquatic insects, with further differentiation within the family according to developmental stage. The distribution of suitable habitat for these insects changes throughout the year, according to changes in rainfall between wet and dry seasons. We compare the niches of Naucoridae and M. ulcerans in West Africa, French Guiana and Australia, exploring the potential of Naucoridae to act as a host or vector of the pathogen. As a broader application to ecological epidemiology, we discuss our methodology and evaluate the impact of sample site selection to demonstrate that studies which delineate study area according to health district or political boundary are likely to be ecologically inappropriate. We explore the ability of the area under the receiver operating characteristic (AUC) to judge model performance, and argue that much confusion arises from inappropriate use of the AUC as an index of model performance in presence-only ecological niche modelling.