We studied the population of two villages, Barisey-la-Côte and Barisey-au-Plain, located in Lorraine region, Northeastern France. Since 1991, Barisey-la-Côte has been the site of a long-term survey on cat population biology and epidemiology [10, 11], including a specific interest in toxoplasmosis. That rural site was considered appropriate for the study of the local variability of prevalence because rural area people often live in the same houses for long periods whereas, in urban areas, part of the inhabitants comes from other areas or countries and have specific risks . A first information sheet was distributed via mailboxes. Then, from March to September 2004, two field investigators contacted directly each household. Consenting subjects aged at least 7 years old were asked to answer a questionnaire and give a blood sample. Each questionnaire was filled out by the investigator and given an anonymous number. Blood samples were collected by nurses after written consent was obtained from adult participants or minor's parents. The protocol has received the agreement of the Comité Consultatif de Protection des Personnes dans la Recherche Biomédicale (Centre Régional de Lutte Contre le Cancer Léon Bérard, Lyon (France) and the study was declared to the Commission Nationale de l'Informatique et des Libertés.
In the two villages, field investigators contacted directly 422 persons and failed to contact only 12 other identified individuals. Eighty-four percent of contacted persons accepted to fill out the questionnaire (353 answers) and 61% accepted to give a blood sample (256 samples). A previous positive serological status was more frequently a cause of refusal to provide a new blood sample than a negative status: 17/18 previously positive individuals versus 12/48 previously negative individuals. We included in the analysis the 17 previously positive subjects and thus analyzed the serological status of 273 subjects. The study involved 138 households. Twenty-five households (18%) had a single member, 75 (54.5%) two members, 21 (15%) three members, 11 (8%) four members, and 6 (4.5%) had 5 members.
The questionnaire sought information on: a) demographic characteristics: sex, age, localization of the accommodation; b) exposure to soil oocysts: number of years spent in the village, occupation related to agriculture, having a garden, time spent in the garden for gardening or other activities; c) exposure to contamination by food: use of bottled or tap water, frequency of meat consumption (beef, pork, poultry, lamb, mutton, goat, horse, rabbit, and game), frequency of consumption of undercooked beef, lamb, or pork, frequency of consumption of raw vegetables; d) contact with cats: presence of cats around the house, number of cats fed at home, cat litter in the house, total number of cats bred during life; e) history of toxoplasmosis: previous serological testing for toxoplasmosis and result, medical problem related to toxoplasmosis. The following classes were used to describe the frequency of meat or vegetable consumption: less than weekly, weekly, daily, more that once per day.
Sera were kept frozen at -20°C and sent to the Department of Parasitology (Croix-Rousse hospital, Lyon, France) to be blindly tested for Toxoplasma-specific IgG and IgM using ELISA (Enzygnost Toxoplasmosis; Dade-Behring, Marburg, Germany) with a lysate of the parasite as antigen.
According to the results of the IgM serology, there were no recent infections in the study population. Among the 256 subjects who had an IgM serology, the titer was zero in 248 (97%) and between 3 and 5 (under the threshold of positivity) in the remaining eight subjects. Therefore, the seroprevalence was determined only on IgG serology results. Among the 256 subjects, 116 (45.3%) had a clearly negative result (titer under 10 IU), 29 (11.3%) an uncertain result (titer between 10 and 25 IU) and 111 (43.4%) a clearly positive result (titer over 25 IU, this being the threshold for positivity in the present study).
The seroprevalence of T. gondii (with its 95% confidence interval) was estimated among the 273 subjects with known serological status. A cubic regression spline was used to represent the functional form of the link between seroprevalence and age.
A mixed logistic regression model with two levels (individual and household) was used to quantify the effects of individual and household factors on the probability of seropositivity. This model allowed us to take into account the fact that individuals from the same household would be exposed to a common environment and to quantify and explain the heterogeneity of seropositivity among households. The individual factors considered for this model were the demographic characteristics (location, age, and sex), the eating habits (consumption of raw vegetables, meat, or water from different origins), and the contact with cats and soil. To explain heterogeneity of seropositivity among households, these individual factors were aggregated to build household factors: the mean age of household members and various percentages of household members having the same eating habits or contact patterns with cats and soil. The parameters of the model were estimated with a Bayesian method. The analysis was carried out with WinBugs 1.4 (MRC Biostatistic Unit, Cambridge, UK) that uses Gibbs sampling.