Periodontal Status and the Risk of All‐Cause Dementia, Alzheimer's Disease and Vascular Dementia in a Community: The Hisayama Study

ABSTRACT

Aim

This study examined the associations of periodontal status with incident dementia and its subtypes in a Japanese community.

Materials and Methods

The 1396 participants aged ≥ 60 years without dementia were followed for 10 years. Periodontal status at baseline and 5 years was assessed using mean probing pocket depth (PPD), clinical attachment level (CAL) and percentage of bleeding on probing (%BOP), categorized into quartiles. Cox proportional hazards and time‐varying Cox models were used.

Results

During follow‐up, 267 participants developed dementia; 194 and 51 had Alzheimer's disease (AD) and vascular dementia (VaD), respectively. Compared with the lowest quartile, the highest quartiles of PPD, CAL and %BOP at baseline were associated with increased dementia risk (hazard ratio 1.72 [95% confidence interval 1.18–2.46], 1.59 [1.07–2.38] and 1.56 [1.09–2.22], respectively, all p for trend < 0.05). Elevated %BOP was associated with AD risk (p for trend = 0.023). Higher PPD tended to be associated with VaD (p for trend = 0.076). Similar associations for all‐cause dementia were observed in time‐varying Cox analyses.

Conclusions

Poor periodontal status was associated with dementia in older Japanese adults. These findings highlight the potential public health importance of preventing periodontitis to reduce the risk of dementia.

1. Introduction

Dementia reduces quality of life and is associated with serious public health challenges. It also imposes a heavy social and economic burden on caregivers and healthcare systems (GBD 2019 Dementia Forecasting Collaborators ²⁰²²). The global number of dementia cases increased from 20.2 million in 1990 to 43.8 million in 2016 largely due to population aging (GBD 2016 Dementia Collaborators ²⁰¹⁹) and is expected to increase to 152.8 million by 2050 (GBD 2019 Dementia Forecasting Collaborators ²⁰²²). Delaying the onset and reducing the incidence of dementia has received considerable attention because curative treatments are unavailable (Norton et al. 2014), and identifying modifiable risk factors may provide effective targets for preventive interventions (Livingston et al. 2024). Several determinants, including lifestyle behaviors and chronic health conditions, have been linked to the development of dementia and cognitive decline (Livingston et al. 2024).

Oral health indicators have increasingly been recognized as potential risk factors for dementia (Takeuchi et al. 2017; Demmer et al. ²⁰²⁰; Kiuchi et al. ²⁰²²; Wu et al. ²⁰²³). Several cohort studies have reported the association of tooth loss with increased dementia risk. However, evidence on periodontitis is limited (Asher et al. 2022). Findings on the longitudinal association between periodontitis and dementia are also inconsistent. Some cohort studies have reported this association (Lee, Lee, et al. ²⁰¹⁷; Yoo et al. ²⁰²³; Demmer et al. ²⁰²⁰; Choi et al. ²⁰¹⁹; Kim et al. ²⁰²²; Lee, Hu, et al. ²⁰¹⁷), while others have not (Arrivé et al. 2012; Holmer et al. ²⁰²²; Asher et al. ²⁰²⁵). The inconsistency might be due to methodological limitations, such as the accurate diagnosis of dementia and standardized periodontal assessment. The association between periodontitis and dementia tends to weaken in studies with longer follow‐up durations (Asher et al. 2022). Neuropathological and cognitive changes develop over many years before the clinical onset of dementia (Sonnen et al. 2009). Therefore, cohort studies with short follow‐up durations cannot rule out the possibility of pre‐existing cognitive decline at or before baseline leading to poor oral hygiene and periodontal deterioration, indicating a possible reverse causal relationship. Therefore, long‐term prospective studies are needed to clarify the temporal relationship between periodontal and cognitive health (Asher et al. 2022).

Dementia comprises multiple subtypes, of which Alzheimer's disease (AD) and vascular dementia (VaD) are the most common and differ etiologically and pathologically (Sokolovič et al. 2023). Several cohort studies have reported associations between periodontitis and the risk of developing dementia (Holmer et al. 2022; Lee, Lee, et al. ²⁰¹⁷; Lee, Hu, et al. ²⁰¹⁷; Choi et al. ²⁰¹⁹; Kim et al. ²⁰²²; Yoo et al. ²⁰²³; Arrivé et al. ²⁰¹²; Asher et al. ²⁰²⁵). However, only a few have examined associations with dementia subtypes (Yoo et al. 2023; Choi et al. ²⁰¹⁹; Demmer et al. ²⁰²⁰). The findings have been inconsistent, making it difficult to draw definitive conclusions. Accurate and standardized assessment of dementia subtypes is required to better understand the potential links between periodontitis and dementia.

The present study aimed to investigate the association between periodontal health and the onset of dementia and its subtypes in a general population of older adults in Japan over 10 years.

2. Methods

2.1. Study Population

This study used data from the Hisayama Study, an ongoing population‐based prospective cohort study established in 1961 in the town of Hisayama, a suburb of the Fukuoka metropolitan area in southern Japan (Ninomiya 2018). The demographic characteristics and occupational distribution of Hisayama residents are similar to those of the general Japanese population according to national census statistics (Hata et al. 2013). Full community surveys of the health status, including assessments of neurological conditions, have been conducted for all residents aged ≥ 40 years at intervals of 1–2 years. Comprehensive surveys on cognitive impairment and dementia in older people have been conducted periodically since 1985 (Ohara et al. 2017). Dental examinations were initiated in 1998 and have been performed annually since 2003 for selected age groups to promote early detection and prevention of oral diseases. They were extended to all residents aged ≥ 40 years in 2007. Dental examinations of all residents were conducted in 2012, whereas in all other years, examinations were conducted on only a subset of residents. Therefore, this study did not use the data from those years.

A baseline screening survey was conducted between 2007 and 2008. Residents aged ≥ 60 years (n = 2002; participation rate, 86.3%) were excluded if they did not provide consent for participation (n = 6), had dementia at baseline (n = 163), had missing oral examination data (n = 240), were edentulous (n = 129) and had missing covariate data (n = 68). Finally, 1396 participants (622 males and 774 females) were included in this study (Figure S1). Among the participants who completed the baseline screening survey, repeated periodontal measurements in 2012 were available for 948 participants.

The study protocol was approved by the Kyushu University Institutional Review Board for Clinical Research (Approval Nos. 23061 and 23092). Informed written consent was obtained from all participants.

2.2. Outcome, Exposure and Covariates

Dementia was the primary outcome, periodontal status was the exposure and other related factors were covariates. The participants were followed up from the baseline examination in 2007–2008 until November 30, 2017. The procedures for the follow‐up survey for dementia have been described in detail elsewhere (Ohara et al. 2017). Dementia was diagnosed based on the criteria of the Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised (DSM‐III‐R) (American Psychiatric Association ¹⁹⁸⁷). AD was diagnosed based on the criteria of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association (NINCDS‐ADRDA) (McKhann et al. ¹⁹⁸⁴). VaD was diagnosed based on the guidelines of the National Institute of Neurological Disorders and Stroke and the Association Internationale pour la Recherche et l'Enseignement en Neurosciences (NINDS‐AIREN) (Román et al. 1993). Each case of suspected dementia was carefully reviewed by expert psychiatrists and stroke physicians. They examined all available medical information, including neuroimaging, to determine the presence or absence of dementia and the subtype. The neuropathological findings of the participants who underwent autopsy and were diagnosed with dementia were evaluated alongside their clinical records to confirm the definitive dementia subtype. The detailed neuropathological diagnostic procedures for the autopsied cases have been described previously (Fujimi et al. 2008). Of the 384 participants who died during the follow‐up, 241 underwent autopsy. Loss to follow‐up did not occur because information on survival and dementia status was complete for all participants.

The number of present teeth and periodontal status were assessed as indicators of oral health. The periodontal status of all teeth except the third molars was determined based on the probing at the mesio‐buccal and mid‐buccal sites using the National Health and Nutrition Examination Survey (NHANES) III method. The probing pocket depth (PPD), clinical attachment level (CAL) and bleeding on probing (BOP) were assessed. The calibration of the examiners for the periodontal examination has been reported previously (Furuta et al. 2021). The mean PPD, mean CAL and percentage BOP (%BOP) were determined. Periodontitis was diagnosed based on the 1999 and 2018 classification (Eke et al. 2012; Papapanou et al. ²⁰¹⁸). The number of teeth with PPD of ≥ 4 mm was also used as an additional criterion.

The covariates were selected based on prior literature (Supporting Information). Information on toothbrushing frequency, regular dental visits, education level, smoking habits, alcohol consumption, physical activity and use of antidiabetic and antihypertensive medications was obtained using a self‐administered questionnaire.

2.3. Follow‐Up Survey, Oral Examination and Measurement of Other Risk Factors

Detailed information on the follow‐up survey, oral examination and measurement of other risk factors is provided in Supporting Information.

2.4. Statistical Analysis

The participants were grouped into quartiles based on their mean PPD, CAL and %BOP. The quartiles of the mean PPD were ≤ 1.90, 1.91–2.28, 2.29–2.83 and ≥ 2.84 mm. Those of the mean CAL were ≤ 2.25, 2.26–2.79, 2.80–3.46 and ≥ 3.47 mm. Those of the %BOP were ≤ 3.9, 4.0–15.9, 16.0–33.9 and ≥ 34.0%. The mean values or frequencies of the risk factors across the four groups based on the quartiles of the mean PPD were determined and compared using linear or logistic regression analysis. The incidence of dementia was calculated using the person‐years method. The age‐ and sex‐adjusted incidence of total dementia was determined using a Poisson regression model that included age and sex. The log‐transformed person‐years were used as the offset term.

A directed acyclic graph was constructed to clarify the assumed relationships among periodontal status, vascular risk factors, stroke and dementia and to distinguish potential confounders from mediators (Figure S2). Since a history of stroke and antihypertensive treatment may lie on the causal pathway between periodontal status and dementia, these factors were not included in the multivariable models to reduce potential overadjustment.

The Cox proportional hazards model was used to determine the age‐ and sex‐adjusted and multivariable‐adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for all‐cause dementia and its subtypes based on the mean PPD, CAL and %BOP. The covariates were age, sex, educational level, systolic blood pressure, diabetes, serum total cholesterol, body mass index (BMI), current smoking and drinking, regular exercise and number of teeth present at baseline. The Cox proportional hazards models were also tested using the mean PPD, CAL and %BOP as continuous variables. The HRs for all‐cause dementia and its subtypes based on oral health behaviours such as toothbrushing frequency and regular dental visits were determined. Other models, including those for periodontitis based on the 1999 (no, moderate, severe) or 2018 (no, stage I–II, III–IV) classifications and the number of teeth with PPD of ≥ 4 mm (0, 1–2, 3–4, ≥ 5 teeth), were also tested.

To account for temporal changes in periodontal status over time, we performed a time‐varying Cox proportional hazards regression analysis using the Andersen‐Gill counting‐process style of data input, which allows the incorporation of time‐dependent exposures (Andersen and Gill 1982). The baseline periodontal status was applied to the initial interval (from baseline to 5 years). For the subsequent interval (beyond 5 years), periodontal status was updated with the measurements obtained at the second examination for participants who underwent dental examination in 2012. Participants without repeated dental examinations contributed to follow‐up information only at the initial interval. The covariates were fixed at baseline.

Several sensitivity analyses were conducted to assess the robustness of the findings. First, lag analyses were performed by excluding participants who developed dementia within 2 or 3 years of follow‐up to reduce the potential influence of reverse causality. Second, to account for death as a competing risk, subdistribution HRs were estimated using the Fine–Gray model (Austin and Fine 2017). Third, inverse probability of treatment weighting (IPTW) based on propensity scores was used to further address potential confounders. Covariate balance after weighting was evaluated using standardized mean differences, and acceptable balance was defined as a maximum standardized mean difference < 0.10 (Austin et al. 2021). Fourth, E‐values were calculated to assess the potential impact of unmeasured confounders on the observed associations (VanderWeele and Ding ²⁰¹⁷). Fifth, an additional analysis, including edentulous individuals who were excluded from the primary analysis, was conducted by classifying them as a separate category in addition to quartiles of periodontal measures among dentate participants. Finally, potential non‐linear associations between periodontal status and dementia were explored using restricted cubic spline analyses.

Statistical analyses were performed using SAS (version 9.4; SAS Institute, Cary, NC). Statistical significance was set at p < 0.05 (two‐sided).

3. Results

3.1. Baseline Characteristics

The mean PPD of the 1396 participants was 2.44 ± 0.86. The baseline characteristics of the study population based on the mean PPD, CAL and %BOP are provided in Tables 1, S1, and S2.

TABLE 1.

Baseline characteristics according to quartiles of mean PPD.

Variable	Mean PPD a				p
	Q1 (n = 348)	Q2 (n = 348)	Q3 (n = 350)	Q4 (n = 350)
Age, years b	69.3 (0.4)	69.9 (0.4)	71.0 (0.4)	71.3 (0.4)	< 0.001
Men c	34.2	43.4	46.0	54.6	< 0.001
Education, ≤ 9 years	29.8	35.1	35.8	35.8	0.110
Hypertension	52.4	60.0	59.0	66.5	0.001
Systolic blood pressure, mm Hg	132.8 (1.0)	133.8 (0.9)	134.1 (0.9)	136.6 (0.9)	0.007
Diastolic blood pressure, mm Hg	79.0 (0.5)	79.7 (0.5)	79.4 (0.5)	80.6 (0.5)	0.055
Antihypertensive medication	35.7	40.3	45.0	48.4	< 0.001
Diabetes	16.1	19.1	25.1	23.9	0.003
Serum total cholesterol, mmol/L	210.6 (1.7)	207.1 (1.7)	206.6 (1.7)	208.6 (1.7)	0.405
Obesity	19.5	26.9	28.2	32.5	< 0.001
Body mass index	22.6 (0.2)	23.2 (0.2)	23.3 (0.2)	23.5 (0.2)	0.001
History of stroke	3.3	3.9	5.5	4.2	0.383
Current smoking	5.6	6.8	8.5	15.1	< 0.001
Current drinking	40.5	42.0	45.9	43.4	0.344
No regular exercise	59.3	64.6	60.8	66.7	0.116
Number of present teeth	21.3 (0.4)	21.5 (0.4)	20.5 (0.4)	16.4 (0.4)	< 0.001
Toothbrushing ≤ 1 time per day	26.3	30.8	35.5	39.2	< 0.001
No regular dental visit	60.8	68.1	69.4	76.4	< 0.001

Note: Age‐ and sex‐adjusted mean values (SEs) or proportions. Abbreviation: PPD, periodontal pocket depth.

Abbreviation: PPD, periodontal pocket depth.

^a Range of mean PPD: Q1, ≤ 1.90 mm; Q2, 1.91–2.28 mm; Q3, 2.29–2.83 mm; Q4, ≥ 2.84 mm.

^b Adjusted for sex.

^c Adjusted for age.

3.2. Incident Dementia During Follow‐Up

All 267 participants (102 men and 165 women) who developed all‐cause dementia were followed up for 10 years. Median follow‐up time was 3753 days (approximately 10.3 years). Brain imaging was performed in 206 participants, neuropathological examination at autopsy was conducted in three participants, and both procedures were performed in 43 participants. Overall, morphological examinations were performed in 252 participants with dementia (94.4%). Fifteen patients had mixed AD and VaD, but each subtype was counted independently in the analysis. AD and VaD were diagnosed in 194 and 51 participants, respectively.

3.3. Association Between Periodontal Status and Dementia Risk

The age‐ and sex‐adjusted incidence of total dementia increased with higher mean PPD and %BOP (p for trend = 0.026 and 0.015, respectively) (Figure 1). Table 2 shows the estimated HRs and 95% CIs for all‐cause dementia and the subtypes based on the quartiles of mean PPD, CAL and %BOP. The multivariable model revealed positive associations of the highest quartiles of mean PPD (HR, 1.72; 95% CI, 1.18–2.49), mean CAL (HR, 1.59; 95% CI, 1.07–2.38) and %BOP (HR, 1.56; 95% CI, 1.09–2.22) with the development of all‐cause dementia. The highest quartile of %BOP was associated with a high risk of AD (HR, 1.61; 95% CI, 1.06–2.44) (Table 2). The higher quartile of mean PPD was marginally associated with a higher risk of VaD (HR, 2.20; 95% CI, 0.89–5.44; p for trend = 0.076) (Table 2). Using the continuous value of periodontal status, %BOP was associated with the development of AD (HR, 1.80; 95% CI, 1.04–3.09; p = 0.035) (Table 3). The significant association between the continuous values of mean PPD and VaD was evident (HR, 1.40; 95% CI, 1.10–1.78; p = 0.006). This contrasted with the analysis based on the categorization of mean PPD into quartiles.

FIGURE 1.

Age‐ and sex‐adjusted incidence of dementia between 2007 and 2017 based on the quartiles of the baseline mean PPD (A), mean CAL (B) and %BOP (C) of older Japanese adults (aged ≥ 60 years). The incidence of dementia was determined using the person‐years method. The age‐ and sex‐adjusted incidence of total dementia was determined using a Poisson regression model that included age and sex. The log‐transformed person‐years were used as the offset term.

TABLE 2.

Hazard ratios for the development of dementia according to periodontal status.

Outcome	No. of event	No. at risk	Age‐ and sex‐adjusted	Multivariable‐adjusted a
			HR (95% CI)	HR (95% CI)
All‐cause dementia
Mean PPD
Q1	48	348	1	1
Q2	67	348	1.50 (1.04, 2.18)	1.47 (1.01, 2.14)
Q3	73	350	1.45 (1.01, 2.10)	1.49 (1.03, 2.16)
Q4	79	350	1.65 (1.15, 2.37)	1.72 (1.18, 2.49)
p for trend			0.014	0.007
Mean CAL b
Q1	45	350	1	1
Q2	67	342	1.30 (0.89–1.90)	1.31 (0.89–1.91)
Q3	80	355	1.33 (0.92–1.93)	1.37 (0.94–2.00)
Q4	75	349	1.46 (0.99–2.13)	1.59 (1.07–2.38)
p for trend			0.065	0.027
%BOP c
Q1	51	345	1	1
Q2	56	352	1.24 (0.85–1.82)	1.27 (0.86–1.88)
Q3	79	363	1.37 (0.96–1.94)	1.38 (0.96–1.98)
Q4	81	336	1.59 (1.12–2.26)	1.56 (1.09–2.22)
p for trend			0.008	0.014
Alzheimer disease
Mean PPD
Q1	41	348	1	1
Q2	50	348	1.34 (0.88, 2.02)	1.34 (0.88, 2.03)
Q3	49	350	1.17 (0.77, 1.77)	1.22 (0.80, 1.85)
Q4	54	350	1.35 (0.90, 2.04)	1.40 (0.92, 2.13)
p for trend			0.249	0.183
Mean CAL
Q1	34	350	1	1
Q2	51	342	1.31 (0.85–2.03)	1.32 (0.85–2.05)
Q3	55	355	1.21 (0.78–1.87)	1.24 (0.79–1.93)
Q4	54	349	1.43 (0.92–2.22)	1.52 (0.95–2.42)
p for trend			0.185	0.125
%BOP
Q1	36	345	1	1
Q2	40	352	1.28 (0.82–2.02)	1.36 (0.86–2.17)
Q3	60	363	1.49 (0.98–2.25)	1.54 (1.01–2.36)
Q4	58	336	1.62 (1.07–2.46)	1.61 (1.06–2.44)
p for trend			0.018	0.023
Vascular dementia
Mean PPD
Q1	7	348	1	1
Q2	11	348	1.57 (0.61, 4.07)	1.47 (0.56, 3.83)
Q3	15	350	1.87 (0.76, 4.62)	1.77 (0.71, 4.40)
Q4	18	350	2.35 (0.98, 5.69)	2.20 (0.89, 5.44)
p for trend			0.049	0.076
Mean CAL
Q1	8	350	1	1
Q2	9	342	0.94 (0.36–2.45)	0.93 (0.36–2.42)
Q3	16	355	1.41 (0.60–3.36)	1.36 (0.56–3.26)
Q4	18	349	1.77 (0.75–4.17)	1.76 (0.71–4.35)
p for trend			0.097	0.134
%BOP
Q1	13	345	1	1
Q2	13	352	1.11 (0.51–2.40)	1.09 (0.49–2.41)
Q3	7	363	0.46 (0.18–1.15)	0.43 (0.17–1.10)
Q4	18	336	1.34 (0.65–2.73)	1.23 (0.60–2.53)
p for trend			0.788	0.959

Abbreviations: BOP, bleeding on probing; CAL, clinical attachment level; CI, confidence interval; HR, hazard ratio; PPD, periodontal pocket depth.

^a The model included age, sex, educational level, systolic blood pressure, diabetes, serum total cholesterol, body mass index, current smoking, current drinking, no regular exercise and number of present teeth as covariates.

^b Range of mean CAL: Q1, ≤ 2.25 mm; Q2, 2.26–2.79 mm; Q3, 2.80–3.46 mm; Q4, ≥ 3.47 mm.

^c Range of %BOP: Q1, ≤ 3.9%; Q2, 4.0%–15.9%; Q3, 16.0%–33.9%; Q4, ≥ 34.0%.

TABLE 3.

Hazard ratio for the development of dementia and its subtypes according to the continuous value of periodontal status.

Outcome	Age‐ and sex‐adjusted	p	Multivariable‐adjusted a	p
	HR (95% CI)		HR (95% CI)
All‐cause dementia
Mean PPD	1.12 (0.99–1.27)	0.065	1.15 (1.01, 1.32)	0.034
Mean CAL	1.04 (0.93–1.16)	0.483	1.05 (0.94–1.19)	0.386
%BOP	1.92 (1.24, 2.98)	0.004	1.94 (1.22, 3.10)	0.005
Alzheimer disease
Mean PPD	1.01 (0.86, 1.19)	0.891	1.02 (0.87, 1.21)	0.799
Mean CAL	0.99 (0.87, 1.14)	0.923	0.99 (0.86, 1.14)	0.893
%BOP	1.84 (1.10, 3.09)	0.021	1.80 (1.04, 3.09)	0.035
Vascular dementia
Mean PPD	1.36 (1.10, 1.67)	0.004	1.40 (1.10, 1.78)	0.006
Mean CAL	1.19 (0.97–1.47)	0.098	1.21 (0.95–1.53)	0.127
%BOP	1.94 (0.71–5.29)	0.193	1.88 (0.65–5.41)	0.241

Abbreviations: BOP, bleeding on probing; CAL, clinical attachment level; CI, confidence interval; HR, hazard ratio; PPD, periodontal pocket depth.

^a The model included age, sex, educational level, systolic blood pressure, diabetes, serum total cholesterol, body mass index, current smoking, current drinking, no regular exercise and number of present teeth as covariates.

3.4. Time‐Varying Cox Regression Analyses

In the analyses incorporating repeated periodontal measures as time‐varying exposures, higher time‐varying mean PPD, mean CAL and %BOP were associated with an increased risk of all‐cause dementia (Table 4). For AD, the highest quartiles of time‐varying mean PPD and CAL were associated with an increased risk. Time‐varying %BOP in the continuous value was positively associated with AD (p = 0.032) (Table S3). No association was observed between time‐varying periodontal status and VaD (Table S3).

TABLE 4.

Association between repeated periodontal measures and incident dementia using time‐varying Cox proportional hazards models.

Exposure	Multivariable‐adjusted a
	HR (95% CI)
All‐cause dementia
Quartiles b
Time‐varying mean PPD
Q1	1
Q2	1.16 (0.78–1.73)
Q3	1.16 (0.78–1.72)
Q4	1.67 (1.13–2.48)
p for trend	0.017
Time‐varying mean CAL
Q1	1
Q2	1.24 (0.84–1.84)
Q3	1.48 (0.98–2.23)
Q4	1.81 (1.17–2.78)
p for trend	0.006
Time‐varying %BOP
Q1	1
Q2	1.05 (0.62–1.80)
Q3	0.99 (0.60–1.63)
Q4	1.47 (1.03–2.08)
p for trend	0.035
Continuous values
Time‐varying mean PPD	1.15 (0.99–1.34)
Time‐varying mean CAL	1.09 (0.95–1.25)
Time‐varying %BOP	1.01 (0.99–1.02)

Note: Periodontal measures were treated as time‐varying exposures.

Abbreviations: BOP, bleeding on probing; CAL, clinical attachment level; CI, confidence interval; HR, hazard ratio; PPD, periodontal pocket depth.

^a The model included age, sex, educational level, systolic blood pressure, diabetes, serum total cholesterol, body mass index, current smoking, current drinking, no regular exercise and number of present teeth as covariates.

^b Quartile cut points were defined using the baseline distribution of mean PPD, CAL and %BOP and were applied to both baseline and follow‐up periodontal measurements.

3.5. Sensitivity Analyses

The results of the sensitivity analyses are presented in Tables S4–S6 and Figure S3. In lag analyses excluding dementia cases occurring within the first 2 and 3 years after baseline, the overall pattern of associations remained generally consistent in the 2‐year lag analysis, although several associations were attenuated in the 3‐year lag analysis (Table S4). However, the association between %BOP, all‐cause dementia and AD persisted after the lag analyses. Similar associations were observed in Fine–Grey subdistribution hazard models accounting for the competing risk of death (Table S4). After IPTW, all covariates were balanced across mean PPD quartiles (maximum standardized mean difference < 0.10) (Table S5). In the IPTW analyses, the association of mean PPD and CAL with all‐cause dementia remained significant, whereas the association involving %BOP was attenuated (Table S4). The E‐values for the associations of the highest quartiles of mean PPD, CAL and %BOP with all‐cause dementia were 2.83, 2.56 and 2.49, respectively (Table S6). When edentulous participants were included, no association was observed between edentulousness and risk of dementia (Table S4). The restricted cubic spline analyses did not show non‐linearity in the association of mean PPD and %BOP with all‐cause dementia (p = 0.504 and 0.114, respectively) (Figure S3). A non‐linearity was observed for mean CAL (p = 0.044) (Figure S3).

Additional analyses of oral health behaviours and alternative definitions of periodontal status are presented in Tables S7 and S8, and Supporting Information.

4. Discussion

This study demonstrated the associations between all periodontal parameters, including mean PPD, mean CAL and %BOP and total dementia. Conventional and time‐varying Cox analyses consistently supported an association between periodontal status and dementia, although findings for dementia subtypes differed by analytical approach. These findings support a potential role of poor periodontal health in dementia development.

Several previous longitudinal studies have reported an association between periodontitis and dementia (Lee, Lee, et al. ²⁰¹⁷; Demmer et al. ²⁰²⁰; Choi et al. ²⁰¹⁹; Kim et al. ²⁰²²; Lee, Hu, et al. ²⁰¹⁷; Yoo et al. ²⁰²³). However, some studies have reported no association (Holmer et al. 2022; Arrivé et al. ²⁰¹²; Asher et al. ²⁰²⁵). Most previous studies were based on national health insurance databases (Lee, Hu, et al. ²⁰¹⁷; Lee, Lee, et al. ²⁰¹⁷; Choi et al. ²⁰¹⁹; Kim et al. ²⁰²²; Yoo et al. ²⁰²³), which provide large sample sizes but often lack detailed clinical periodontal measurements. The diagnoses of periodontitis and dementia are typically based on diagnostic codes (such as those of the International Classification of Diseases), which may lead to misclassification or underestimation of diseases and their severity. Additionally, information on confounding factors such as education level and lifestyle factors is often unavailable in the national health insurance database. In contrast, periodontal status and dementia were clinically assessed in the present study, and a wide range of potential confounders was considered, strengthening the validity of our findings.

The present study showed that %BOP and mean PPD were associated with the risks of developing AD and VaD, respectively; however, mean CAL was not associated with dementia subtypes in the baseline analyses. PPD and BOP are indicators of active and progressive periodontal inflammation (Chaves et al. 1993; Heitz‐Mayfield ²⁰²⁴). The possible mechanisms underlying the association between periodontal status and dementia may be explained by an increase in the inflammatory state. Periodontal inflammation induces the production of proinflammatory cytokines such as interleukin (IL)‐1, IL‐6 and tumour necrosis factor‐α. These cytokines can enter the systemic circulation and subsequently influence the central nervous system (Martínez‐García and Hernández‐Lemus ²⁰²¹). Persistent systemic inflammation may impair the blood–brain barrier, allowing inflammatory mediators to penetrate the brain tissue and induce neuroinflammation and neuronal damage (Galea 2021). Periodontal inflammation, may contribute to systemic inflammation that promotes neuroinflammation and neurodegeneration, which lead to AD. Similarly, the periodontal pathogen burden is higher and may induce vascular endothelial dysfunction (Li et al. 2022) and lead to cerebral hypoperfusion and vascular injury. These may induce neuropathological changes and result in vascular cognitive impairment (Rajeev et al. 2023). Impaired cerebral perfusion due to cerebral small‐vessel diseases may induce oxidative stress, and dysfunction of the neurovascular unit may increase neuroinflammation. These factors may accelerate brain aging via not only microvascular injury but also neurodegeneration (Tarasoff‐Conway et al. ²⁰¹⁵). Thus, periodontitis may increase the risk of dementia by causing neuroinflammatory and vascular changes.

The pattern of associations differed between the conventional and time‐varying Cox analyses. In the conventional models, %BOP was associated with AD risk. In the time‐varying analyses incorporating repeated periodontal measurements, mean PPD and CAL were associated with AD in quartile‐based analyses, while %BOP was also associated with AD when analysed as a continuous variable. These findings suggest that both active periodontal inflammation and the cumulative periodontal burden over time may be relevant to AD development.

The observed associations were generally consistent across several sensitivity analyses. The results of the Fine–Gray subdistribution hazard models and IPTW analyses supported the primary findings, although attenuation of some associations was observed after IPTW. In the 3‐year lag analysis, some associations were attenuated; however, the association between %BOP and dementia generally persisted. Taken together, these findings provide additional support for the observed associations between periodontal status and dementia risk.

Low frequency of toothbrushing was associated with a high risk of all‐cause dementia, independent of periodontal status (Table S7). One possible explanation is that infrequent toothbrushing promotes the accumulation of subgingival plaque and bacterial burden (Waerhaug 1981), which may contribute to systemic inflammation (Arimatsu et al. 2014) and increase dementia risk.

This study had some limitations. First, a small number of participants developed VaD. Therefore, the statistical power to detect associations with VaD was limited, and the corresponding estimates should be interpreted with caution. Further studies with a larger number of patients with VaD are required to confirm these findings. Second, information regarding nutritional status and depression was not available. Both factors have been reported to be associated with periodontitis and cognitive decline (Aida et al. 2026) and may therefore act as potential confounders in the association between periodontal status and dementia. Although we adjusted for various relevant factors, residual confounding by such unmeasured factors cannot be excluded. Nevertheless, the E‐values for the highest quartiles of mean PPD, CAL and %BOP were 2.83, 2.56 and 2.49, respectively, suggesting that an unmeasured confounder would need to be relatively moderately associated with both periodontal status and dementia to fully explain the observed findings. Third, the partial‐mouth assessment was used to evaluate the periodontal status, and the lingual or palatal sites were excluded. Consequently, the severity of periodontitis may have been underestimated, which may have influenced the observed association between periodontitis and dementia. Finally, this study involved residents of a single municipality in Japan. Therefore, our findings have limited generalizability.

5. Conclusion

This study confirmed the association between periodontal status and dementia risk in a general older Japanese population followed up for 10 years. It specifically revealed the association of periodontal inflammation, as indicated by periodontal indicators such as PPD and BOP, with AD and VaD. Periodontal care to improve and maintain periodontal health may help reduce the risk of dementia.

Author Contributions

Michiko Furuta: contributed to conception, design, data interpretation, performed statistical analyses and drafted and critically revised the manuscript. Tomoyuki Ohara: contributed to data acquisition and interpretation and drafted and critically revised the manuscript. Satoko Sakata: contributed to data acquisition and interpretation and critically revised the manuscript. Mao Shibata: contributed to data acquisition and interpretation and critically revised the manuscript. Emi Oishi: contributed to data acquisition and interpretation and critically revised the manuscript. Yoshihiko Furuta: contributed to data acquisition and interpretation and critically revised the manuscript. Shinya Kageyama: contributed to data interpretation, drafted and critically revised the manuscript. Mikari Asakawa: contributed to data interpretation, drafted and critically revised the manuscript. Yoshihiro Shimazaki: contributed to data acquisition and critically revised the manuscript. Toru Takeshita: contributed to data acquisition and drafted and critically revised the manuscript. Toshiharu Ninomiya: contributed to conception, design, data acquisition and interpretation, study coordination and acquisition of funding and drafted and critically revised the manuscript. All authors gave their final approval and agree to be accountable for all aspects of the work.

Funding

This study was supported in part by the Ministry of Education, Culture, Sports, Science and Technology of Japan (JSPS KAKENHI Grant Number 24K13242) and Health and Labor Sciences Research Grants of the Ministry of Health, Labor and Welfare of Japan (23FA1022).

Ethics Statement

The present study was approved by the Kyushu University Institutional Review Board for Clinical Research (Approval Nos. 23061 and 23092).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Baseline characteristics according to quartiles of mean CAL.

Table S2: Baseline characteristics according to quartiles of %BOP.

Table S3: Association between repeated periodontal measures and incident dementia using time‐varying Cox proportional hazards models.

Table S4: Results of sensitivity analyses.

Table S5: Covariate balance before and after IPTW for mean PPD quartiles.

Table S6: E‐values for the associations between periodontal measures and dementia outcomes.

Table S7: Hazard ratios for all‐cause dementia according to oral health behaviour at baseline.

Table S8: Hazard ratios for dementia according to different definitions of periodontal condition at baseline.

Figure S1: Participant flow chart.

Figure S2: Directed acyclic graph (DAG) representing presumed the relationship of periodontal status, dementia and related factors.

Figure S3: Restricted cubic spline analyses of the association between periodontal status and the risk of all‐cause dementia.

Data Availability Statement

The datasets generated and analysed in the present study are not publicly available because they contain confidential clinical and demographic data of the study participants. However, further information about the datasets may be obtained from the principal investigator of the Hisayama Study (Toshiharu Ninomiya) upon reasonable request for purposes of replicating procedures and results.