Study design
This was a cross-sectional study nested within a cohort study of preschool children aged from 18 to 42-months, in which was evidenced caries polarization [19]. This study was conducted from March to September 2008 in the city of São Luís, Maranhão, Brazil. A convenience sample of children enrolled from community nurseries servicing low-income families, with incomes up to twice the Brazilian minimum wage (corresponding to U.S. $8,208.00/year) was selected. Four nurseries located in the poorest city district were sufficient to complete the sample size. All children attending these nurseries were invited to participate in the study.
A total of 260 parents signed informed consent forms for participation in the study. Exclusion criteria included the presence of debilitating systemic diseases, malnutrition, and antibiotic use that could interfere with colonization of Streptococcus mutans (SM) 30 days prior to the onset of the study. These factors were chosen because of their potential to influence the expression of caries.
Considering that this study used previously collected data, [19] ad hoc power analysis indicated that a sample of 226 children is able to detect a 2.5 prevalence ratio (PR), with an 80% test power and a 5% probability of type I error, assuming that 10% of children presented caries.
Data collection
The mothers or caregivers were interviewed using a structured questionnaire that was tested in a pilot study involving 25 mothers or caregivers. The questionnaire included demographic and social variables (mother’s age, mother’s years of education, family income), mother-child data (if the mother has untreated cavities of caries or not), who stays at home with the child, previous information about oral health (if she received information about oral health care by a dentist or by health professional), previous child hospitalization, antibiotic use in the last year by the child, child visits to the dentist, nighttime breastfeeding after 1 year and child oral hygiene practices (when teeth brushing started, who brushes the child’s teeth and daily toothbrush frequency).
A 24-hour food recall was used to acquire information about participants’ dietary habits (use of a night bottle and frequency of sucrose consumption between main meals in the form of juice, baby formula, snacks, and sweets). On a separate day, a dental surgeon assessed the visible plaque index (VPI) of each participant under natural light in the knee-to-knee position. To calculate VPI, a score corresponding to the sum of tooth sides with clinically visible plaque in relation to the total number of tooth sides present was determined [20].
The same examiner determined the number of teeth with carious decay, with extraction completed or indicated, and with fillings apparent by surface examination (dmfs index) for each child using a dental mirror (n° 5) and an exploratory probe with a rhombus point, as recommended by the World Health Organization [21]. The clinical exam was repeated one week later in 10% of the study population to evaluate intra-examiner agreement (kappa = 0.91). To identify the group of children in which caries disease was polarized, the Significant Caries (SiC) Index, adapted for deciduous teeth corresponding to the third of the sample with highest dmfs scores, was used [22].
To detect SM contamination, 1.7 mL of non-stimulated saliva was collected by pressing a sterile swab on the sublingual zone and dorsal tongue of each child until the swab was saturated (~5 seconds in each area). Each saliva sample was diluted in saline solution (0.9% NaCl) to produce decimal dilutions. Aliquots (0.1 mL) of each dilution were spread onto Mitis salivarius-bacitracin agar plates and incubated in a 5% CO2 chamber for 48 hours. After incubation, the number of colony-forming units (CFUs) was counted on each plate with >10 colonies. A single examiner, who did not have knowledge of patients’ caries status, performed the microbial counts in duplicate using a digital colony counter (model CP 600 Plus, Phoenix).
Theoretical model
A hierarchical theoretical model with the following eight blocks was employed: 1) age of child in months; 2) socioeconomic and demographic factors; 3) mother’s past history of caries (if the mother has untreated cavities of caries or not); 4) child’s health status; 5) dietary child habits; 6) child oral hygiene data; 7) VPI; and 8) level of SM contamination in the saliva (Figure 1).
Age was considered a potentially confounding factor, so the model was adjusted for this variable, which represented the block furthest from the outcome. This variable was added as a continuous variable.
The second block included socioeconomic and demographic variables as distal factors in the theoretical model since they can influence all variables in subsequent blocks. Were included in this block mother’s age in years, mother’s years of education (<8, 8–11, or >11), family income (<1 minimum wage, 1 minimum wage (U.S. $4,104.00/year), or >1 minimum wage), who stayed at home with the child when she/he was not in day care (the mother, others), and whether the mother or caregiver received information about oral health (yes or no) [19].
The health variable included for the mother in the third block, which may mediate the association between socioeconomic variables and childhood caries, was previous history of caries, self-reported by the mother. She reported if she had untreated cavities of caries or not at the time of interview (yes, no).
Health variables for the child were included in the fourth block since they may mediate the association between the variables included in the previous blocks and childhood caries. The following variables were used: previous child hospitalization reported by the mother or caregiver (0, 1, or ≥2 times), antibiotic use in the last year (yes, no), and whether the child had already visited the dentist (yes, no).
In the fifth block, dietary practices were included, assuming that these variables may be influenced by socioeconomic and demographic factors, and by mother and child health variables. The variables in this block were: nighttime breastfeeding after 1 year (yes, no), use of a night bottle (yes, no), and frequency of sucrose consumption between main meals in the form of juice, baby formula, snacks, and sweets (≤3 or >3 times/day).
Oral hygiene practices may be moderator variables of the association between diet and ECC, so they were included in the sixth block as follows: when teeth brushing started (0–6 months, 7–12 months, after 12 months), who does the brushing (the child alone, the caregiver), and the number of times a day the child’s teeth are brushed (up to 1 time, ≥2 times).
In the seventh block, whether a visible biofilm was present (<15% presence of visible plaque or ≥15%) was recorded assuming that this variable may be influenced by both hygiene and dietary practices.
In the final block the level of SM contamination in the saliva (<105 or ≥105 CFU – colony-forming units) was included as a biological variable nearest to the outcome that might be influenced by all variables in the previous blocks.
Statistical analyses
Multinomial logistic regression using forward selection and guided by a hierarchical approach was used [19, 23]. We evaluated possible associations between risk factors and ECC using a series of models, which allowed for the modeling of the dependent variable, structured as polytomous data divided according to three SiC categories (caries-free, few caries, high caries). The estimated coefficients were expressed as prevalence ratios (PRs) and their 95% confidence intervals were also calculated.
The hierarchical modeling started with the first block. The variables of the first block were adjusted simultaneously for each other and only those variables whose p value was <0.10 entered in subsequent models. Then variables of the second block were adjusted simultaneously for each other and for the variables whose p value was <0.10 in the previous step. The significance of each variable was considered at the time of entry into the model (p value <0.05). All other blocks were then added in succession following the same procedure.
All statistical analyses were performed using the STATA 10.0 program (Stata Corp., College Station, TX). Collinear variables were evaluated using the variance inflation factor.
Ethical aspects
The ethics committee of the University Hospital of the Federal University of Maranhão (Universidade Federal do Maranhão, UFMA) (case 33104-1251/2007) approved the study. The parents or legal representatives of each child that participated in the study read and signed informed consent forms. All children needing dental treatment were referred to the pediatric dental clinical of the UFMA.