Study design and population
The source population for this study comprised the participants of the Korean National Health and Nutrition Examination Survey from 2007 to 2015 (KNHANES IV–VI). Detailed information on this study has been given previously21,22. Briefly, the KNHANES is a nationwide surveillance system used to assess data from individuals aged more than 1 year by conducting health interviews, health examinations and nutrition surveys. From the database of this survey, baseline data on sociodemographic and lifestyle factors, body mass index (BMI), biomarkers, dental behaviours, and self-reported disease history were obtained. To obtain outcome information, this database was linked to two other administrative claims databases in South Korea, including the Health Insurance Review and Assessment Service (HIRA) 2007–2019 and the National Health Insurance Service-National Health Information Database (NHIS-NHID) 2007–2015. The HIRA is a database containing detailed medical records of Korean residents who visited pharmacies, hospitals or physicians23, while the NHIS-NHID is a database organised by the Korean National Health Insurance Service that contains information on the health care utilisation, health screening, sociodemographics and mortality of the Korean population24. By combining the three databases, information on participants’ medical records from the HIRA and death information from the NHID were obtained. The Korea Health Information Service organised and merged the three databases.
The baseline or index period of this study was set from 2009 to 2015, and the participants were followed up until 2019. A washout period from 2007 to 2008 was applied to minimise the risk of reverse causality of the outcome. For better understanding, the scheme of this study is detailed in Supplementary Fig. 1.
To evaluate the association between PD and DM as well as CVD, three schemes were used: (1) the analysis of PD as the outcome, (2) the analysis of DM as the outcome, and (3) the analysis of CVD as the outcome. Among all the participants of the KNHANES 2007–2015 (n = 28,711), 535 participants recruited during the washout period (2007–2008) and 6386 participants aged less than 20 years were excluded. Thus, 16,790 participants remained in the study. Three dataset groups were formed. For the first, participants with a history of PD before the index period (n = 6257) were excluded, leaving 10,533 eligible participants for the analysis. For the second, participants with a history of diabetes (n = 2267) were excluded, leaving 14,523 participants. For the third, participants with a history of CVD (n = 2475) were excluded; thus, 14,315 participants were included. The detailed flow of study participant selection is shown in Fig. 1.
Ascertainment of periodontal disease
PD in the present study included gingivitis and periodontitis. Information on PD was obtained from the HIRA and identified according to the International Classification of Diseases 10 (ICD-10). The incidence was ascertained if the participant had visited a physician at least twice within a month with the periodontitis diagnostic code K05 or one of the following health claim-of-treatment codes was used: U2240, U1010, U1020, U1051, U1052, U1071, U1072, U1081, U1082, U4412, U4413 or U4414. Details regarding disease identification are provided in Supplementary Table 1.
Ascertainment of diabetes mellitus
DM cases as exposure were ascertained using the ICD-10 codes from the medical record data in the HIRA database before the index period and self-reported disease history from the KNHANES. Participants were considered to have DM if there was a medical record of prescription of diabetes medication or the provision of ICD-10 codes E11, E12, E13 or E14 at least twice within 1 year in the medical record before the index period or if the participant reported a history of being diagnosed with diabetes by a physician in the KNHANES data. Furthermore, DM was ascertained based only on the medical record data in the HIRA during the follow-up period. The details of disease identification are provided in Supplementary Table 1 and Supplementary Fig. 1.
Ascertainment of cardiovascular disease
In this study, CVD included cases of coronary heart disease and heart failure. CVD was ascertained using the ICD-10 codes from the medical record data in the HIRA database and self-reported disease history in the KNHANES data. Coronary heart disease was confirmed based on medical records indicating hospitalisation for at least 2 days because of the disease codes I20–I25 before the index period, the self-reported diagnosis of angina, or the diagnosis of myocardial infarction by a physician. Participants were considered to have heart failure if there was a medical record indicating hospitalisation for at least 2 days because of the disease code I50 before the index period. Furthermore, CVD was ascertained using the medical record data from the HIRA database during the follow-up period. The details of disease identification are provided in Supplementary Table 1 and Supplementary Fig. 1.
To address potential bias from potential covariates, some other variables were included in this study, such as sociodemographic factors (sex, income level, age and residential area), lifestyle factors (smoking status, smoking pack-years, alcohol consumption and physical activity), BMI, biomarkers (aspartate transaminase (AST), alanine aminotransferase (ALT), fasting plasma glucose (FPG), cholesterol and blood pressure), dental behaviour (brushing frequency, use of dental floss, use of an interdental brush and use of rinsing solution) and oral health at baseline. A history of other diseases (tooth decay, cancer, cerebrovascular diseases, hypertension, hyperlipidaemia, arthritis, rheumatic diseases, gastrointestinal diseases and infectious diseases) was also included. The inclusion of the covariates was based on previous studies reporting associations between some of these variables and PD25,26,27, DM1,28,29, and CVD30,31.
Smoking status was classified as never smoker, former smoker or current smoker based on participant reports regarding current smoking habits. Smoking pack-years were calculated by multiplying the number of cigarette packs smoked per day by the total number of smoking years, and participants were categorised based on their total smoking pack-years as light smokers (0.1–20 pack-years), moderate smokers (20.1–40 pack-years) and heavy smokers (> 40 pack-years)32. Participants were divided into non-drinkers and drinkers according to their alcohol consumption habits. Furthermore, participants were grouped into underweight (< 18.5 kg/m2), normal (18.5–22.9 kg/m2), overweight (23–24.9 kg/m2) and obese (≥ 25 kg/m2) BMI categories following the recommended classification for Koreans33.
Oral health at baseline was represented by the score of the Decayed, Missing, and Filled Teeth (DMFT) index. The DMFT index is one of the indices used to assess oral health based on the number of decayed, missing and filled teeth34. Disease history was identified using ICD-10 codes and self-reports. The ascertainment of disease history is presented in Supplementary Table 1. A separate category was allocated for missing information on each covariate and included in the analysis.
Cox proportional hazard regression with 95% confidence intervals (CIs) was applied to obtain the hazard ratios (HRs) of PD, DM and CVD. The time variable was defined as the duration from the index date to the diagnosis of outcome, death or end of follow-up, whichever came first. Participants who did not have PD or died before the endpoint of follow-up were treated as censors. Similarly, in the other groups, those who did not have DM or CVD or died before the end of follow-up were treated as censors.
Five models were assessed in the main analysis: model 1 was adjusted for age and sex; model 2 was adjusted for age, sex, lifestyle factors and BMI; model 3 was adjusted for age, sex, lifestyle factors, BMI and dental behaviour; model 4 was adjusted for age, sex, lifestyle factors, BMI, DM and CVD; and model 5 was adjusted for age, sex, lifestyle factors, BMI, dental behaviour, DM and CVD. A stratified analysis by sex along with a heterogeneity test using Cochran’s Q test and the I2 test was also performed35. All analyses were performed using SAS statistical software (version 9.4; SAS Institute, Cary, NC, USA).
Ethics approval and consent to participate
This study was approved by the Institutional Review Board of the Seoul National University Hospital, Seoul, Korea (IRB No: E-1911-054-1078).