Diabetes is a chronic condition that profoundly impacts health and well-being, and its prevalence is increasing rapidly worldwide.1 In addition to diabetes, pre-diabetes, a condition in which glucose levels are elevated but do not reach the diagnostic criteria for diabetes, has been receiving more attention in recent years. Pre-diabetes not only is a predictor of future diabetes but also is already associated with several health outcomes.2 3
The number of teeth serves as a straightforward and widely used measure of oral health status and is particularly suitable for large-scale studies.4 Previous research indicates that a lower number of teeth is associated with increased mortality and morbidity,5–9 including diabetes onset.7 10–12 The number of teeth could indicate exposure to periodontal disease,5–7 a chronic inflammatory oral disease that can destroy supporting periodontal tissues,13 and is the primary cause of adult tooth loss.14 The low-grade systemic inflammation triggered by this disease may be one explanation for the association between the number of teeth and diabetes.15 On the other hand, changes in dietary habits related to a decline in masticatory function due to tooth loss might also explain the association.16 17 However, the potential association between number of teeth and pre-diabetes, rather than diabetes, remains unclear due to limited studies.18 19 Furthermore, while several studies have been conducted on the association between periodontal disease and pre-diabetes, their results are inconsistent.20–22
In Japan, nearly all dental claims for periodontal disease include information on the number of teeth. Using these data, examinations of the associations between the number of teeth and subsequent health outcomes at a nationwide level are feasible.19 23 In this study, we analysed data from Japanese oral health claims combined with annual check-up data to examine the association between the number of teeth and the onset of pre-diabetes using one of the largest sample sizes reported.
This retrospective cohort study used the National Database of Health Insurance Claims and Specific Health Checkups of Japan (NDB), which comprised the information on all of the country’s health insurance claims and Specific Health Checkup results.24 The Ministry of Health, Labour and Welfare of Japan (MHLW) owns and manages this database. The claims data in the database include clinical details such as age group, sex, diagnosis codes from the 10th edition of the International Classification of Disease (ICD-10), dental formula information and codes for procedures and prescriptions. The NDB has documented nearly all claims for medical and dental treatments received under Japan’s universal health insurance coverage, which provides coverage to most of Japan’s residents. As a result, the database provides near-complete coverage of the healthcare utilisation data in Japan.
In 2008, the MHLW introduced an annual health screening programme called the Specific Health Checkups. Health insurance providers have been mandated to offer this programme to all members aged 40–74 years. This check-up programme consists of anthropometry, laboratory tests and a self-administered questionnaire regarding health and lifestyle. Details of the programme, including the specific measurements used, are provided elsewhere.25 The results of Specific Health Checkups are stored in the NDB, which includes information including age, sex, current smoking status, frequency of alcohol consumption, amount of alcohol consumption, exercise habit, self-reported medical history (cardiovascular disease, cerebrovascular disease and renal disease), self-reported medication (hypertension, lipidaemia and diabetes), body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), triglyceride (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), haemoglobin A1c (HbA1c) and fasting blood glucose (FBG).
We obtained de-identified extracted NDB data consisting of dental claims data and Specific Health Checkup records from the MHLW. To link dental claims and Specific Health Checkup data, we used an anonymised identifier called ID1N, which was generated by combining the individual’s date of birth, sex, the insurer’s identification number and the identification number within the insurer’s system.
Figure 1 is the flow chart of the subjects enrolled in this study. We initially included subjects who met the following criteria: those who (1) participated in the Specific Health Checkup programme every year from fiscal year (FY) 2015 to FY 2018 and (2) had dental insurance claims data with a diagnosis of periodontal disease (ICD-10 code: K05.3) during the baseline year (FY 2016). We then excluded subjects who met any of the following exclusion criteria: those (1) with a self-reported medical history of cardiovascular disease, cerebrovascular disease or anaemia or missing data related to these conditions at the Specific Health Checkups at the baseline year (FY 2016); (2) with missing data regarding the number of teeth or were edentulous; (3) with no records for HbA1c, FBG or self-reported medication for diabetes in the records of Specific Health Checkups during the year before the baseline year (FY 2015), the baseline year (FY 2016) or the follow-up year (FY 2018); (4) who were classified as having pre-diabetes or diabetes in the Specific Health Checkup during the baseline year (FY 2016) or 1 year prior (FY 2015); or (5) with missing covariate data.
Exposures, outcomes and covariates
We used the number of teeth, excluding third molars, derived from the dental claims during the baseline year as the exposure. We determined the number of teeth by using the dental formula data for periodontal disease diagnoses, as in studies that used Japanese dental claims data.19 23 Given the skewed distribution of the number of teeth among the participants, we categorised the number of teeth into four groups (26–28, 20–25, 15–19 and 1–14) as previously described,6 with the modifications of excluding third molars and edentate status. We assigned the ordinal numbers 4–1 to the above categories, respectively, and calculated estimates per decrease in tooth category in regression models. Concurrently, we treated the number of teeth as a continuous variable without categorisation, presenting the estimates corresponding to every five fewer teeth.
The outcome was pre-diabetes or diabetes observed at the Specific Health Checkup during the follow-up year. Pre-diabetes was defined as an HbA1c value 5.7%–6.4% or FBG of 100–125 mg/dL, and diabetes was defined as HbA1c ≥6.5%, FBG ≥126 mg/dL or self-reported diabetes medication, in accord with the criteria described by the American Diabetes Association (ADA).26
We selected the following covariates based on previous studies11 20 27: age, sex, BMI, current smoking, excessive alcohol drinking, regular exercise, hypertension, dyslipidaemia and baseline HbA1c level. We used the results of the subjects’ Specific Health Checkups during the baseline year to define all these variables. Age was divided into the following categories: <45, 45–54, 55–64 and ≥65. BMI (kg/m2) was classified into distinct categories as follows: underweight (<18.5), normal (18.5–24.9), overweight (25.0–29.9) and obese (≥30.0). Excessive alcohol drinking and regular exercise were defined as described.28 Hypertension was defined as SBP ≥140 mm Hg, DBP ≥90 mm Hg or self-reported medication for hypertension. Dyslipidaemia was defined as TG ≥150 mg/dL, LDL-C ≥140 mg/dL, HDL-C <40 mg/dL or self-reported medication for dyslipidaemia. We also included changes in BMI (defined as a difference in the subject’s BMI between the follow-up and baseline) as a covariate to examine the mediational role of weight changes in the association.
We used a multivariate logistic regression to estimate the ORs and their 95% CIs for the onset of pre-diabetes or diabetes at the follow-up examination. We began with crude models, followed by adjustments for age and sex, and we further adjusted for BMI, current smoking, excessive alcohol drinking, regular exercise, hypertension, dyslipidaemia and baseline HbA1c level. We also used a model that additionally included changes in BMI to examine its mediational role. To explore potential heterogeneity, we performed a subgroup analysis by fitting models for each subgroup according to age group, sex, current smoking and BMI. The statistical significance of differences between estimates was obtained as described.29 For sensitivity analysis, we used the WHO/Information, Education & Communication (WHO/IEC) criteria, which define pre-diabetes as an HbA1c value of 6.0%–6.4% or an FBG of 110–125 mg/dL,30 31 and we repeated the entire analysis. We further incorporated the criteria used in the National Health and Nutrition Survey in Japan.32 According to this survey, ‘individuals strongly suspected of having diabetes’ (diabetes) were defined as having an HbA1c value of ≥6.5% or self-reported medication for diabetes, and ‘individuals who cannot deny the possibility of diabetes’ (pre-diabetes) were defined as having an HbA1c value of 6.0%–6.4%. Additionally, we treated age and BMI as continuous variables and applied natural spline functions with 5 df to account for potential non-linear relationships. We defined statistical significance using a two-tailed p value threshold of <0.05. Python 3 (ver. 3.8.5) and R (ver. 4.0.5) were used for the data handling and statistical analyses.
Patient and public involvement
Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
This nationwide retrospective cohort study examined the association between the number of teeth and the occurrence of pre-diabetes or diabetes 2 years later among middle-aged Japanese adults. Our findings indicate that a lower number of teeth were associated with an increased incidence of pre-diabetes or diabetes, and the estimated ORs increased monotonically as the number of teeth decreased.
Several studies have found an association between a decreased number of teeth and diabetes incidence,7 10–12 which is consistent with our findings. For instance, a longitudinal study from Finland reported that having ≥9 missing teeth, in comparison with 0–1 missing teeth, was associated with incident diabetes.7 Likewise, a study from the USA found an increased incidence of diabetes among subjects missing ≥25 teeth compared with those missing 0–8.10 Additionally, a recent study from Korea reported that an increased number of missing teeth was associated with a new onset of diabetes.11 Using periodontal disease as an exposure yielded similar results,10 11 33 34 though there are some exceptions.35 36
In contrast, studies that explore the association between the number of teeth and pre-diabetes are scarce, and their results are inconclusive. A cross-sectional study from Korea reported that a decreased number of teeth was associated with increased pre-diabetes prevalence; however, the results were insignificant except for a linear trend.18 Another cross-sectional study from Japan reported that pre-diabetes was associated with a decreased number of remaining teeth.19 Several other studies have examined periodontal disease and the new onset of pre-diabetes, with varied conclusions.20–22 For example, a study conducted in Taiwan revealed that periodontal disease was linked to incident pre-diabetes in participants who were normoglycaemic at baseline.20 However, a study of Hispanic/Latino adults in the USA found no such association.22 The precise explanations for these conflicting results are unclear, but differences in the study populations, the definitions of exposure and outcome, the study designs, and the employed statistical models may have contributed.
One potential explanation for the observed association between the number of teeth and pre-diabetes could be periodontal disease, the primary reason for tooth loss among adults.14 Numerous longitudinal studies have reported an association between periodontal disease exposure and the onset of pre-diabetes or diabetes.10 11 20 33 34 Periodontal disease may cause low-grade systemic inflammation through persistent bacteraemia and/or the dissemination of periodontal inflammation mediators into the bloodstream.15 This systemic inflammation may then impair insulin signalling and resistance,37 resulting in elevated glycaemic levels.
Another possible explanation could relate to altered masticatory performance due to tooth loss. Having fewer teeth impairs masticatory performance, which may lead to a shift in diet selection toward fewer vegetables and fibres,16 17 subsequently increasing the incidence of diabetes.38 39 Additionally, impaired masticatory performance can reduce satiety, leading to increased food intake.40 We considered BMI changes as a critical mediator of these pathways and tested models with BMI change between baseline and follow-up as an additional covariate. However, the estimates remained essentially consistent whether or not the model included BMI changes, suggesting that these hypothesised pathways may not explain the observed association. Nevertheless, the follow-up duration in the present study was limited to 2 years, and future research with a longer study period is required to obtain a clearer picture.
Our stratification analysis did not reveal any clear association modification by age, sex, BMI or current smoking habit. Although a few of these tested modifications showed statistical significance, the differences in the magnitude of the associations were negligible and not robust. Similar to our present findings, previous results on the association between periodontal disease and diabetes did not discern consistent patterns.10 11 33 35 Our results suggest that these factors may not modify the association between the number of teeth and the incidence of pre-diabetes or diabetes. However, further studies are necessary to thoroughly test this possibility.
This study has several important limitations. First, due to the lack of data, it was not possible to consider certain vital covariates, such as health behaviours, educational level, household income, family history of diabetes or any other unmeasured confounders. This limitation prevented any definitive inference of causality for the observed associations. Additionally, this may have introduced bias in either direction into our estimations. Second, we used the number of teeth exclusively as the exposure definition, but the specific reasons for tooth loss were not considered as they were not available in the employed database. While periodontal disease is the primary reason for tooth loss in the studied population,14 other causes, such as dental caries or injury, may have contributed to tooth loss. Moreover, detailed timings of tooth loss were also unavailable, limiting the exploration of temporal relationships. And finally, other oral health information, including periodontal disease status and masticatory function, was also not provided in the database. Third, the follow-up period was limited to 2 years, which may limit the generalisability of our results to longer-term associations. Fourth, our study lacked information regarding the specific subtypes of diabetes, although it is known that type 2 diabetes is dominant among Japanese patients.41 Fifth, our sample was restricted to individuals who had undergone a dental visit with a diagnosis of periodontal disease and who regularly participated in Specific Health Checkups. It has been suggested that the characteristics of individuals who undergo Specific Health Checkups may differ from those of the general population.42 Additionally, an investigation of Japanese subjects revealed that having a family dentist was associated with socioeconomic status.43 Our present findings may thus have been affected by selection bias, and their generalisability is therefore limited. Notably, however, the Specific Health Checkup introduced a new questionnaire item in 2018 about chewing ability.44 Using the responses to this questionnaire item might help to mitigate some of the selection bias.
This post was originally published on https://bmjopen.bmj.com