The study was approved by the Ethical Review Committee of Guanghua School of Stomatology, Sun Yat-Sen University (ERC-01). A written informed consent was obtained from each student’s guardian before the study.
The oral examination and saliva tests were performed in March 2018 in Foshan. Foshan is a medium-sized city in Guangdong Province in southern China with a population of 7.6 million. The GDP was CNY 124,324 (USD 18,018) per capita in 2017 (10). The water fluoride concentration is 0.16 mg/L (11).
According to Peduzzi, for logistic regression analysis, the number of events (death or illness) should be 5-10 times the number of independent variables (12). In total, 43 independent variables were investigated via logistic regression in this study, and the caries prevalence in 12-year-old adolescents was 38.5%; thus, the total sample size should be between 559 and 1117 (12)(3).
A random cluster sampling technique was employed in 6 middle schools according to the number of children that we needed to recruit. All Grade 7 students living in Foshan for the previous 2 years who reported no systematic illness and no antibiotic use for at least the preceding 2 weeks were invited to participate in the study.
The questionnaire was mainly designed with reference to the Fisher-Owens conceptual model of influence on children’s oral health (2). The structured questionnaire recorded demographic information, socioeconomic status, diet, self-reported oral health behaviors and past dental experiences. The questionnaire was self-administered and completed in schools by students with the guidance of guardians. The reliability of the questionnaire was assessed by an internal consistency test, and the Cronbach’s alpha was 0.75.
2.3 Oral examination
The oral examination was conducted in the classrooms. The examinee was lying in a portable dental chair. The teeth were examined visually with the help of a plain mouth mirror, a CPI probe and compressed air. The International Caries Detection and Assessment System (ICDAS) criteria were used (13). All tooth surfaces were first examined with a wet surface and then reexamined after the teeth were dried with compressed air. Filled surfaces and missing teeth due to caries (the reasons for missing teeth were obtained from the questionnaire) were also recorded. The DMFT was calculated both at the early clinical stage of decay, with enamel and dentine caries (at the D1 level, as in the ICDAS method), and at the later stages of decay, with dentine-only caries (at the D3 level, as in the ICDAS method), and these values were recorded as D1MFT and D3MFT, respectively. Individuals with lesions scoring from 3-6 were classified into the caries group, and the others were classified into the caries-free group in this study(14). The Plaque Index (PlI) was recorded according to the Silness and Löe scale (15).
The three examiners had taken a 12-hour e-learning course on ICDAS before the study. Ten percent of the subjects were randomly selected to obtain a duplicate examination in order to monitor the reliability of the diagnoses and calculate the interexaminer consistency. The Cohen’s kappa values were 0.85, 0.83 and 0.91 for caries examination.
2.4 Microbiological examination
The Microbiological assessment was performed by the Cariostat method, which is a colorimetric test for measuring caries activity based on the presence of acidogenic microorganisms (16)(Ramesh et al. 2013). Plaque was collected using the sterile swab included in the Cariostat kit (GangDa Medical Technology Co. LTD., Beijing, China) package. According to the Cariostat kit instructions, the examiners scrubbed the buccal surfaces of the maxillary molars and mandibular incisors 3-5 times and immersed the swab into the culture medium ampule. The samples were incubated at 37°C for 48 hours. The color of the culture medium was compared with the reference color on the color chart supplied with the Cariostat kit. The reference color was scored from 0 to 3 with every 0.5 grade as an interval. The color turns from blue to green and, ultimately, to yellow, indicating an increased acid production ability of the plaque in the sample.
2.5 Saliva tests
All saliva was collected after the students had rinsed their mouths with tap water. The unstimulated saliva was collected for 15 minutes. Students were asked to spit the saliva through a funnel into a scaled tube every 3 minutes. Then, the unstimulated saliva ﬂow rate (ml/min) was calculated. The saliva buffering capability was measured according to the Ericsson method (17). One milliliter of saliva was added to 3 ml of 3.3 mmol HCl within 5 minutes after collection and then allowed to stand for 20 minutes to remove CO2. The final pH of the saliva was evaluated by an electrical pH meter. The buffering capability of unstimulated saliva was recorded as ‘low’, ‘normal’ and ‘high’ when the pH value fell into the ranges of <4.25, 4.25-4.75 and >4.75, respectively.
2.6 Candidate Genes Selection
The major candidate gene categories identified to date include enamel formation genes, immune response genes, genes related to saliva, and genes related to taste and dietary habits (6). The candidate genes selected for this study are listed in Table 1. The Haploview 4.2 software package (http://www.broad.mit.edu/mpg/haploview/) was used to select tag SNPs in each candidate gene. SNPs reported having associations with dental caries in previous studies or located in a gene fragment that could have functional effects were selected.
2.7 DNA analysis
Genomic DNA was extracted from 1-ml saliva samples according to the manufacturer's instructions (Saliva DNA Sample Collection Kit, ZEESAN, Xiamen, China). The DNA concentration and purity of each sample were determined by spectrophotometry. Twenty-three SNPs in 14 genes were identified by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS).
2.8 Statistical analysis
The data were analyzed by SPSS version 22 for Windows (IBM Inc., Chicago, IL, USA). The differences in the characteristics between the caries and caries-free groups were compared by binary regression analysis and an independent samples T test. A binary regression model was also applied to analyze the environmental risk factors. For both the characteristics and environmental risk factors, each risk indicator with a P value of less than 0.2 in the univariate analysis was applied to stepwise multivariate analysis. Risk factors with a final P value in the multivariate analysis of less than 0.05 were used as covariates in further analysis.
The Hardy‐Weinberg equilibrium was estimated by the chi-square test. Genotypes of the candidate genes were analyzed under the hypothesis of an additive model and using the major allele as the reference. Each SNP was treated as a continuous exposure under additive model, coded 0,1, or 2(minor alleles). For direct associations, a binary regression model was applied. Each SNP with a P value of less than 0.2 in the univariate analysis would be applied to stepwise multivariate analysis. All possible pairwise SNP-SNP interactions between the selected SNPs were assessed by a bivariate logistic regression model. Significant characteristics and environmental risk indicators were used as covariables. The interaction between two SNPs was defined as the additional effect of their concomitant carriage on caries beyond the addition of their independent effects (departure from additivity): βinteraction = βobserved – (βSNP1 + βSNP2), where βobserved is the observed effect of the concomitant carriage of SNP1 and SNP2 (versus the carriage of neither) in bivariate binary regression and βSNP1 and βSNP2 are the independent effect of SNP1 and SNP2, respectively, in the same bivariate regression framework (bivariate refers only to the number of pairwise SNPs, as nongenetic covariables were systematically included for all analyses)(18).The highly conservative Bonferroni correction considering the number of tested SNPs was applied. There were 253 possible SNP-SNP pairs in total, so the cutoff value for multiple testing significance was set at P<0.0001 (0.05/253).