We have found that communities with higher percentages of younger children, nasal discharge, facial flies, and number of years of education of the head of the household are associated with higher community prevalence of chlamydia infection in univariate analysis. Community-level facial flies and years of education of head of household are significantly associated with the prevalence of chlamydia infection in multivariate analysis. Flies are thought to be important in trachoma transmission but studies have been done only on the individual level 
or have used clinical outcomes rather than more objective laboratory measurements 
. Chlamydia DNA has been found on 15% of flies in areas hyperendemic for trachoma 
and so flies are believed to be vectors for the spread of chlamydia in endemic areas 
. Our study is the first to show the association between community-level fly density flies and community-level ocular chlamydia infection.
Other studies performed on the individual-level have concluded that general education is associated with less trachoma 
, but our study in Niger found the opposite and chlamydia infection was correlated directly with the self- reported years of education of the head of the household. Trachoma is felt to be disease which aggregates among the poor and uneducated 
, and studies have shown health education can improve trachoma control 
. However, a study in Mali showed the odds of trachoma was higher in households where children attended a traditional school compared to households where children attended a modern school. Mali shares a border with Niger of greater than 860 km and is likely more similar to Niger than more distant areas. The structure of the schools and the way in which children congregate in these schools is just as important as the years of education that have been received.
We have used community-level predictors and community-level outcomes because trachoma is a communicable disease; poor hygiene and specific behavior of others in the same household and neighborhood may increase the risk of infection in the entire community 
. For these reasons the WHO strategy for reducing trachoma is implemented at the community level with community-wide interventions. Other trachoma studies have been cluster-randomized but then analyzed at the level of the individual 
. Like our study in Niger, the Tanzania arm of the PRET study found that facial flies are associated with ocular chlamydia infection 
. However, this individual-level study also showed that facial cleanliness is a risk for ocular chlamydia infection, an association which we were unable to identify. Reinfection is known to occur at the community level and an individual-level approach does not take this into account. In our study, we look for risk factors for higher community prevalence of ocular chlamydia infection.
The clinical exam for trachoma has been shown to be unreliable and poorly correlated with infection in some situations 
. We chose to use the more objective, masked, outcome of ocular chlamydia infection by PCR in our study. Note that we did include the clinical exam in the multivariate predictor model of infection by design, because TF and TI are consequences
of infection rather than causes
of it. However, in a different context, programs that have access to clinical surveys may be interested in the associated level of infection that we have measured here (). It is important to keep in mind the clinical exam for trachoma (TF) is close to 90% sensitive but only 30% specific in latent class analysis, leading to overtreatment in some situations 
. Furthermore, the poor correlation between the clinical signs of trachoma and the laboratory evidence of infection with C. trachomatis
, becomes more problematic as the community-level of infection decreases following mass treatment 
. Nevertheless, the clinical exam is inexpensive and easy to perform. It will remain an important tool for the WHO in their treatment guidelines and continued efforts to understand the relationship between clinical trachoma and infection with the causative bacteria is critical.
Multivariate analysis of clinical trachoma as a predictor of chlamydia infection in children 0–5 years.
There are several limitations to this study. First, we evaluated only a sample of sentinel children in the communities rather than all individuals and this may have produced bias in our estimates. Note that laboratory workers were masked to the identities of the sentinel children chosen and the communities in which they lived. Second, although all of the selected communities for inclusion in the study were evaluated successfully as planned, some individuals who were randomly selected for inclusion were missing; if individuals were not missing at random, this could also have created bias. All analyses were done on an intention-to-treat basis with no adjustments for missing individuals. Third, the evidence for risk factors of infectious ocular chlamydia that we have found in rural Niger may not be generalizable to other trachoma-endemic areas in other countries and continents because of differences in important, unmeasured cultural or environmental characteristics. The finding that more years of education of the head of household is associated with higher levels of chlamydia infection is counterintuitive. The study questionnaire asked specifically for the number of years of formal education that were completed. Religious education, a form of learning that is very common in this area, was not included in this estimate. If the education of adults is by methods other than formal education in households, our measurement could have been an underestimate of the total amount of education interfering with our ability to capture an association with infection.
In summary, we have found that facial flies and years of education of the head of the household are associated with community-level prevalence of ocular chlamydia infection. Further analyses will be performed as treatment begins in the study and continues over the next 3 years.