Population
and study sample
The study population was the schoolchildren aged 6 years in 2016 who benefited from the dental sealant program; the program concerned 4329 children, was effectively offered to 3774 (87%) children and 2532 children participated (89%). The number of children to be selected in the 2017 study sample from that population was calculated to ensure enough precision of the the retention rates (estimated rate = 50%, precision 5%, n = 385). This number was increased to 550 considering a participation rate of 75%. Children were randomly selected using a computerised sampling method with a probability proportional to the region size.
Ethics approval and Consent to participate
Ethical approvals were obtained from the NC educational and health institutions. Schools were approached through local educational authorities. Data were recorded in a file registered within the ethical comity for data management (CNIL N° kpP1390145R). Explanatory letters and consent forms were sent to parents prior to the dental examinations and children whose parents returned written consent were examined. In order not to impact negatively the participation rate in the OHP program, it was decided not to collect individual social information nor behavioural ones at this stage.
Sealant program
The age of 6 years has been chosen because school is compulsory since age 6 and in order to allow sealant to be applied on first permanent molars as soon as their erupted. The recommendation for the dentists is to seal all the non-carious or with very early enamel lesions (ICDAS1), erupted first permanent molars whatever caries risk [13]. Caries risk is very high in NC and it was considered not feasible nor pertinent to differentiate children according to caries risk. Photopolymerised resin sealants were applied with cotton rolls for saliva isolation [14, 15]. Sealants were applied either at school in a mobile dentist ‘surgery or in public dental offices.
Study variables
The children’s dental status was evaluated in 2017 at school. Radiographs were not used. Six general dental practitioners performed the children’s examinations, after having been calibrated through a training course described in a previous study [4].
The presence of dental sealants was recorded. A sealant was considered as totally or partially present using standardised criteria [15]. The number of sealants applied in 2016 was compared with the number of sealants present during the 2017 examination. This allowed the calculation of the retention rates; total (totally or partially present), partial (partially present) and complete retention rate (totally present).
Dental caries were diagnosed at the dentinal threshold level (ICDAS4–6) for deciduous and permanent teeth [16]. The dental status recorded in 2016 during the sealant intervention was compared with that of the 2017 dental examination. This allowed the calculation of the carious increment for first permanent molars: ∆DT1rst molars = DT first molars in 2017 - DT first molars in 2016.
Socio-demographic variables (gender, region, public or private school) were recorded from the school register. The setting where dental sealants were applied in 2016 (school vs dental office) and the presence of a dental assistant during the sealing procedure were also retrieved.
Data analysis
Data entry was duplicated and errors were corrected before analysis. The statistical analyses were performed using the Statistical Package Stata software version 13.
Relationships between caries increment, retention rates, carious status, sealing process and socioeconomic variable (measured by an ecological variable, the region) were explored first using bivariate analysis. Chi-squared (with Rao-Scott correction), F Fisher-Snedecor, ANOVA or Student t-tests were used depending of the variables being considered.
Carious increment was explained using a mixed multiple random-effects regression. Mixed models were used including the examiner parameter as random effect [17–20]. Factors significantly associated with carious increment in bivariate analyses but also non-significant parameters considered as being relevant were integrated in the analyses. The Carious increment was not normally distributed (even with a normality transformation), thus 0-inflated Poisson model a used. The interactions between factors were tested.
We conducted a mediation analysis to assess the respective contributions of the retention rate and the region on caries increment. A mediation proportion was estimated, indicating how much of the whole increment value provided by an independent variable (the region) can be explained by the indirect path in which changes in this independent variable drives a change in the mediator (retention rate), and changes in the mediator then affect outcome [21, 22]. We performed multilevel mediation analysis with other explanatory variables and examiner effect being integrated. Three Models were conducted depending on the region being considered as the reference (A - South, B - Islands C—North). The complete and total retention rates were considered separately. Results were summarized using a graph giving the proportion of mediation proportion and significance of the mediation analysis associations.