We fit a class of non-linear, stochastic, susceptible-infectious-susceptible (SIS) models which allow positive or negative feedback, to data from a recent community-randomized trial in Ethiopia, and make predictions using model averaging.
Trachoma programs use mass distributions of oral azithromycin to treat the ocular strains of Chlamydia trachomatis that cause the disease. There is debate whether infection can be eradicated or only controlled. Mass antibiotic administrations clearly reduce the prevalence of chlamydia in endemic communities. However, perfect coverage is unattainable, and the World Health Organization's goal is to control infection to a level where resulting blindness is not a public health concern. Here, we use mathematical models to assess whether more ambitious goals such as local elimination or even global eradication are possible.
The models predict that reintroduced infection may not repopulate the community, or may do so sufficiently slowly that surveillance might be effective. The preferred model exhibits positive feedback, allowing a form of stochastic hysteresis in which infection returns slowly after mass treatment, if it returns at all. Results for regions of different endemicity suggest that elimination may be more feasible than earlier models had predicted.
If trachoma can be eradicated with repeated mass antibiotic distributions, it would encourage similar strategies against other bacterial diseases whose only host is humans and for which effective vaccines are not available.