In this relatively well-educated East Bay Area sample, the mean of the highest household education level approached ‘some graduate or professional school’, with a range of ‘some grade school’ to ‘graduate or professional degree’ (mean = 6.7, SD = 1.4, range 1-8). Four percent of parents had only a high school education or less, 17% had some college or a two-year degree, 22% were college graduates, and 57% had at least some professional or graduate education beyond college. Despite relatively high education levels on average, nearly 16% of study households had incomes less the federal poverty level for a family of four. Family Financial Stressors were rated an average of 2.4 on the 1-5 point scale (5 corresponding to the highest level of stress), with mean scores ranging from 1 to 5. The mean raw, pre- and post-school salivary cortisol levels were 7.9 nmol/L (SD = 4.4; range = 2.4 - 25) and 4.7 nmol/L (SD = 4.8; range = 1.0 - 40.1 ), respectively. The mean pre- and post-stress reactivity protocol cortisol levels were 5.0 nmol/L (SD = 4.6; range = 1.2 - 36.0) and 4.3 nmol/L (SD = 3.3; range = 1.1 - 26.8). Approximately one third of children (34%) showed an increase in salivary cortisol over the course of the reactivity protocol. Oral MS counts averaged 50,964 CFU/mL and ranged from none to 2,500,000 CFU/mL. Oral LB averaged 7,372 CFU/mL, with a range of none to 70,000 CFU/mL. No oral Cariogenic Bacteria of either species were found in a subset of 42 children (45%). As shown in , the 94 study children had a positively skewed distribution of ∑DMFS, with a mean of 4.3 surfaces with lesions (SD = 7.7; range = 0-38). Forty-four children (47%) had at least one caries lesion, while 50 (53%) had none. All reported analyses were also run using a count of decayed, missing and filled teeth (i.e., ΣDMFT), rather than dental surfaces. Resulting associations were comparable in direction and magnitude, but the stronger predictions were derived from the ΣDMFS analyses. Only the latter results are thus presented here.
displays Pearson correlation coefficients among independent and dependent variables. SES was significantly and inversely associated with Family Financial Stressors, Basal Salivary Cortisol Secretion, Cariogenic Bacteria, and Dental Caries. Basal Salivary Cortisol Secretion was positively associated with Family Financial Stressors and Dental Caries, and Cariogenic Bacteria was strongly linked to Dental Caries. Because Salivary Cortisol Reactivity was unrelated to Dental Caries in the Study 1 sub-sample, this variable was omitted from further analysis.
Bivariate Associations Among Independent Variables (1-5) and Dental Caries (6)—Study 1 (Pearson Correlation Coefficients, N = 94 children)
SES, Family Financial Stressors, Basal Salivary Cortisol Secretion, and Cariogenic Bacteria were sequentially entered as predictors of ∑DMFS counts in the Poisson component of the computed ZIP regression models (). Basal Salivary Cortisol and Cariogenic Bacteria were the strongest bivariate predictors of Dental Caries, and from a theoretical perspective, salivary cortisol could plausibly suppress mucosal immunity against cariogenic bacteria. In light of these considerations, an interaction term for these two predictor variables was entered as a final step in the computed Poisson model. In addition, because those children with no oral pathogens accounted substantially for the large numbers of zeroes among ∑DMFS scores, Cariogenic Bacteria counts were used to predict the presence or absence of ∑DMFS in the models’ logistic component. In the step at which each variable entered the model, SES, Basal Salivary Cortisol Secretion and Cariogenic Bacteria bore strong, independent associations with counts of Dental Caries. The coefficient for SES was -0.16, corresponding to an exponentiated odds ratio of .85 or a 15% decrease in the number of caries for every one unit increase in SES. Similarly, the coefficients for Basal Salivary Cortisol Secretion and Cariogenic Bacteria were 0.19 and 0.12, corresponding to odds ratios of 1.21 and 1.13 or 21% and 13% increases in caries for each one log10 unit increase in cortisol secretion and counts of oral bacteria. Further, the coefficient for SES diminished by half when Cariogenic Bacteria was added to Model 4, indicating that the SES - Dental Caries association was partially mediated by the number of Cariogenic Bacteria.
Zero-Inflated Poisson Regression Models Predicting ΣDMFS—Study 1 (N = 94 children)
The interaction of Basal Salivary Cortisol x Cariogenic Bacteria was also a significant predictor of ∑DMFS scores. This interaction was inspected using a scatterplot of the two component variables predicting ∑DMFS (). The highest number of Dental Caries was identified among children with the combination of high Basal Salivary Cortisol Secretion and high Cariogenic Bacteria counts, and the coefficient for the interaction was 0.23, corresponding to an odds ratio of 1.26. The ZIP logistic regression component revealed, at each model level, significant effects of Cariogenic Bacteria on the presence or absence of Dental Caries. To examine the possibility that the prior exfoliation of primary teeth could have biased ∑DMFS scores and confounded the identified associations, regression analyses were repeated using counts of only decayed or filled (but not missing) dental surfaces (DFS) as the dependent variable; coefficients for both the Poisson and logistic portions of these models were virtually unchanged from those derived for ∑DMFS scores. In addition, the Vuong statistic for the final model was significant, indicating that the zero-inflated model fit the observed data better than an ordinary Poisson regression (z = 3.74 - 5.41, p < .001).
Interaction of Basal Salivary Cortisol Secretion and Cariogenic Bacteria Predicting ∑DMFS— Study 1 (N = 94 children)
Multivariate regression models thus indicated that: a) lower SES, higher Basal Salivary Cortisol Secretion, and higher counts of Cariogenic Bacteria predicted the number of caries lesions; b) children with the highest number of Dental Caries were those with the combination of high Salivary Cortisol Secretion, high Cariogenic Bacteria, low SES and high Family Financial Stressors; and c) counts of Cariogenic Bacteria were a significant predictor of the binary presence or absence of Caries.
Study 2: Salivary cortisol and the dental microanatomy of exfoliated teeth
Sample and Methods
To examine associations among salivary cortisol and the thickness and density of dental tissue compartments, a second sub-sample of children was recruited during their first grade year, from 2004 - 2006. This sub-sample comprised 38 children ranging in age from 5.9 to 6.8 years (17 girls and 21 boys), recruited from all three PAWS cohorts, and consisting of children who lost a tooth during the 9 months of the first grade academic year and from whom we were able to collect the tooth for analysis (see sub-study flow chart, ). Again the subsample did not differ from the larger, longitudinal study sample on socioeconomic status (F = .91, p = NS). During their kindergarten year, when basal and reactive cortisol levels were assessed, 18 children (47%) had attended a morning class and 20 (53%) an afternoon class. Children taking medications that could result in erroneous measures of salivary cortisol were excluded from the sample.
SES, Family Financial Stressors, Basal Salivary Cortisol Secretion, and Salivary Cortisol Reactivity were measured using the methods described for Study 1.
Dental Compartment Thickness and Density
Exfoliated primary mandibular incisors were collected from the children. Parents/guardians were instructed to call the project coordinator for tooth collection immediately after the tooth fell out. The teeth were then transferred to sterile water with 0.1% thymol, sterilized by overnight gamma radiation, and stored at 4°C.
A single investigator (YJ), blind to other study data, measured the thickness and density of the enamel and dentin compartments by scanning with a microtomography (μCT) scanner (Scanco Medical AG, Bassersdorf, Switzerland). A small x-ray tube with a micro focal spot was used as a source, and the detector consisted of a linear CCD-array. A scout view scan was obtained first, followed by automatic positioning, measurement, and offline reconstruction (see Jiang, Zhao, Mitlak, Wang, Genant, & Eriksen, 2003
). Images with isotropic resolution of 21 μm were obtained with 70 KVp and 85 μA. For subsequent image analyses, a subset of the original μCT section, containing entirely enamel and readily distinguished from dentin by obvious differences in mineral density, was selected. Enamel data were thresholded into binary data sets, and enamel tissue was segmented from non-enamel in gray-scale images with a fixed thresholding procedure. Boundaries between enamel and dentin and between dentin and pulp cavity were manually traced to measure the thickness and density of the two layers.
Preliminary analyses duplicated those of Study 1. OLS regression was used to estimate the direct and interactive associations of independent variables with dental compartment measures. Each predictor variable was centered at its mean, and where a significant interaction was found, the moderator effect was probed using scatter plots as in Study 1.