Sampling and clinical examination method
The present study was a part of the 4th National Oral Health Survey, which is the most comprehensive oral health survey to date in China, and was carried out during 2015–2016. A multistage stratified sampling method was used to recruit adults from all 31 provinces, municipalities and autonomous regions in mainland China. The probability proportional to size (PPS) method was used to randomly select subjects [18]. In total, 4410 35- to 44-year-olds and 4431 65- to 74-year-olds completed the survey. The exclusion criteria for this analysis were participants who were edentulous and had a periodontal status that could not be examined for any reason, such as a nonstandard fixed prosthesis that covered the gingiva or the presence of calculus to such an extent that a periodontal examination was impossible.
The latest WHO recommendations (2013) with appropriate adjustments according to the actual situation were used for the clinical oral examinations. The examinations were conducted by three trained licenced dentists, while three other trained individuals with clinical experience acted as recorders in each province. The examinations were conducted with a mobile dental chair using artificial light, a disposable dental mirror, and a standard WHO Community Periodontal Index (CPI) probe [18, 19].
Variables
Independent variables
Caries were recorded for all tooth surfaces, but the observations were recorded for each tooth. We divided the caries into the following three types: type A, caries or a filling on the crown, representing crown caries; type B, caries or a filling involving both the crown and root, representing mixed caries; and type C, caries or a filling on the root, representing root caries. These three types together represent the overall caries situation, which we call type ABC (Fig. 1). Because residual roots involve both the crown and root, we classified residual roots as type B. Missing teeth of any cause in adults were recorded as code 5 in our national survey, which differs from the WHO recommendations. The number of decayed and filled teeth (DFT) was calculated to analyse its relationship with periodontitis. The kappa value of the inter-examiners’ reliabilities in the examination of dental caries was 0.97 in both the middle-aged and elderly groups [18, 20].
Dependent variables
Periodontal disease severity was evaluated by clinical attachment loss (CAL) in each tooth. A full mouth examination including the third molars was performed. The tooth with the worst CAL score was recorded for the person-level score. Probing was performed by “walking around” along gingival crevices with a standardized force of no more than 20 g. Each tooth was recorded according to its severity. CAL was scored as 0 (0–3 mm), 1 (4–5 mm), 2 (6–8 mm), 3 (9–11 mm), 4 (12 mm or more), 9 (tooth excluded) or X (tooth not present). For the analysis, periodontitis was divided into the following three groups according to CAL: ≤ 3 mm, 4–5 mm, and ≥ 6 mm. Before the field investigation, the examiners underwent training, and each examiner and reference examiner performed an examination of 10 to 15 subjects per group to assess their consistency. The kappa value was calculated to be greater than 0.6, suggesting good reliability [18].
Covariates
The covariates included in the statistical analysis were social economic status (SES), such as sex (female or male), area (urban or rural), education level (subjects were classified according to whether they had received nine-year compulsory education and were divided into groups with ≤ 9 years or > 9 years), and household income per capita (less than RMB 5000/person, RMB 5000 to RMB 15,000/person, more than RMB 15,000/person, or prefer not to answer); oral health-related behaviours, such as the frequency of dessert consumption (< twice a day or ≥ twice a day), frequency of tooth brushing (< once per day or ≥ once per day), use of dental floss (no or yes), use of a toothpick (no or yes), smoking status (never or current), and alcohol consumption (ceased, rarely/never, daily, or weekly); and diabetes history (no or yes) [21, 22].
Statistical analysis
The data analysis was carried out using SPSS 20.0. Chi-square tests were performed to compare periodontitis according to the participant characteristics. To determine the association between periodontitis and dental caries, ordinal logistic regression models were used. First, a bivariate analysis was performed. Then, the independent variables with P ≤ 0.25 based on the bivariate logistic analysis were further tested in the multivariate analysis. Three ordinal logistic regression models were constructed to measure the crude and adjusted effects of the DFT scores on periodontitis. In Model 1, the DFT score was introduced as the only independent variable. Then, in Model 2, the DFT score and SES were included. Finally, the DFT score, SES, oral health-related behaviours and diabetes were included in Model 3. A P-value < 0.05 in all two-sided statistical tests was considered significant. The statistical analyses did not include missing values or individuals who preferred not to answer.