Infection burden, periodontal pathogens, and their interactive association with incident all‐cause and Alzheimer's disease dementia in a large national survey

Abstract

INTRODUCTION

Relationships and interplay of an infection burden (IB) and periodontal pathogens or periodontal disease (Pd) markers with Alzheimer's disease (AD) and all‐cause dementia among US adults were examined.

METHODS

Less than or equal to 2997 participants from the National Health and Nutrition Survey III were linked to CMS‐Medicare [≥45 years (1988‐1994); ≤30 years follow‐up].

RESULTS

Hepatitis C (hazard ratio = 3.33, p = 0.004) and herpes simplex virus 2 were strongly associated with greater all‐cause dementia risk. Porphyromonas gingivalis and Streptococcus oralis were associated with greater AD risk at higher IB. The red‐green periodontal pathogen cluster coupled with higher IB count increased the risk of all‐cause dementia among minority racial groups. Pocket probing depth associated with dementia risk at lower IB in the overall sample.

DISCUSSION

Select viruses and bacteria were associated with all‐cause and AD dementia, while the IB interacted with Pd markers in relation to these outcomes.

Highlights

• Interplay of infection burden (IB) and periodontal disease with dementia was tested.

• ≤2997 participants from NHANES III were linked to Medicare.

• Hepatitis C and herpes simplex virus 2 strongly associated with dementia risk.

• Tetanus sero‐positivity increased Alzheimer's disease (AD) risk.

• Porphyromonas gingivalis and Streptococcus oralis associated with AD at higher IB.

• Red‐green periodontal cluster at high IB, increased dementia in racial minorities.

• Pocket probing depth associated with dementia risk at lower IB.

1. BACKGROUND

Alzheimer's disease (AD) is the most common form of dementia, affecting 10% of people over age 65 years in the US population. In 2019, the medical expenses related to AD were close to $290 billion. ¹ AD is associated with hyperphosphorylated tau proteins resulting in neurofibrillary tangles and amyloid‐beta accumulation. ² Although approximately 54% of AD risk is accounted for by multiple modifiable risk variables, a significant amount of unexplained variation remains. ³ Therefore, as long as treatment for AD and other related dementia remains ineffective, investigating alternate risk‐reduction pathways is crucial to pinpointing intervention targets. Further, nearly 40% of the population attributable risk for dementia is ascribed to several modifiable factors including early‐life lower education, mid‐life hearing loss, traumatic brain injury, hypertension, alcohol use, obesity and to later‐life smoking, depression, social isolation, physical inactivity, and diabetes. ⁴ Socioeconomic factors such as low education are thought to induce poorer lifestyles or exposure to air pollution, which can result in poorer health profiles that can trigger poor cognition and dementia risk. ⁵ , ⁶ , ⁷ , ⁸ Furthermore, dietary and nutritional factors including adherence to the Mediterranean or the MIND dietary patterns have also been associated with reduced risk of age‐related cognitive decline as well as reduced post mortem AD‐related pathologies including less beta‐amyloid load. ⁹ , ¹⁰

Related to AD, research has investigated the pathogens associated with periodontal disease (Pd) and the production of bacterial products that may pass through the blood‐brain barrier. ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ In Pd, gums infected with one or more oral bacteria lead to gingival inflammation, which causes the slow breakdown of connective tissue, the development of a periodontal pocket, the loss of bone within the alveoli, and ultimately edentulation. ¹⁶ According to the World Health Organization, severe Pd is estimated to affect around 19% of the global adult population, representing more than 1 billion cases worldwide, with main risk factors being poor oral hygiene and tobacco use. ¹⁷ Pd's possible contribution to AD is significant. A study using Medicare data linked to the third National Health and Nutrition Examination Surveys (NHANES III) found that periodontal pathogens (Pd markers) are associated with cognitive outcomes, with AD incidence linked to a composite factor of Campylobacter rectus and Porphyromonas gingivalis titers. ¹⁸ However, it is unclear whether these pathogens interact with persistent infections overall and across racial/ethnic groups.

An increasing body of research suggests that AD may be linked to exposure to bacterial and viral infections, including spirochetes, oral bacterial species, gastrointestinal microbes, and herpesviruses. ¹⁹ Host immune response to infectious agents may initiate or accelerate AD processes, such as the accumulation of amyloid beta (Aβ), pathogenic tau protein, and neuroinflammation. ¹⁹ Particularly in brain regions with AD pathology, infectious agents and their related molecules may breach blood‐brain barriers, accelerate Aβ deposition, and trigger neurotoxic microglial activation, resulting in a cycle of neurodegeneration brought on by pathogen‐associated molecular patterns (PAMPs). ²⁰ Further, dementia has been linked to genes associated with inflammatory processes, systemic inflammation has been linked to cognitive loss, and infection has been linked to dementia with a recent onset. ²¹ Similarly, animal models show that some viruses might cause AD pathology, although the evidence in human trials is still scarce and inconsistent. ²¹ Another theory postulates that systemic inflammation aids in the development of AD and similar dementias. Moreover, the antimicrobial protection model of AD contends that Aβ builds up in reaction to pathogenic agents. ²¹ A few studies have examined the burden of infectious disease in connection to AD and related dementias. Research using UK Biobank data linked to hospital records suggested a strong association of hospital‐treated infections with all‐cause dementia and neuroimaging markers of age‐related cognitive decline, including reduced white‐matter integrity. ²² , ²³ , ²⁴ Based on these findings, we hypothesized that synergistic interactions of periodontitis and/or periodontal pathogens with an infection burden (IB) based on key persistent viral, bacterial, and parasitic infections increase the risk of dementia. Accordingly, the main goal of this study was to determine whether, among adults, aged 45 years at baseline, the longitudinal correlations between Pd indicators and periodontal pathogens with incident AD and all‐cause dementia varied by the IB level. Examining whether a potential synergistic impact would be exclusive to a racial/ethnic group was a secondary objective. Finally, we explored whether IB level and individual infections were associated with AD and all‐cause dementia incidence.

2. METHODS

2.1. Database: NHANES‐CMS

The US National Center for Health Statistics (NCHS) sponsors the National Health and Nutrition Examination Study (NHANES), a set of cross‐sectional surveys that yield nationally representative statistics on the health and nutritional status of the United States population. A multistage, stratified probability cluster design is used in the sampling process. In‐home interviews gather basic health and demographic data, and in‐depth evaluations in medical research vehicles closely follow. ²⁵ Appendix I provides the Centers for Medicare & Medicaid Services (CMS‐Medicare) and National Death Index (NDI) linking methods. The National Institute on Aging's Intramural Research Program's Institutional Review Board approved the current research project.

RESEARCH IN CONTEXT

1. Systematic review: The authors of this work reviewed the literature relying mainly on PubMed searches. Periodontal disease (Pd) clinical and bacterial markers have been explored in a handful of studies in relation to incidence of all‐cause and Alzheimer's disease (AD) dementia. Despite evidence of an association between Pd and increased risk for dementia over time, the interplay with other infections, particularly viral infections, remains unclear. We therefore examined potential moderators of these relationships by considering the total infection burden (IB) in nationally representative sample of ≤2975 participants of the third National Health and Nutrition Examination Surveys (NHANES III, 1988‐1994) linked to Center for Medicare and Medicaid‐Medicare data up to end of 2018.

2. Interpretation: Despite a lack of an IB‐dementia relationship, hepatitis C (hazard ratio [HR] = 3.33, 95% confidence interval [CI]: 1.52‐7.26, p = 0.004) and herpes simplex virus 2 (HR = 1.30, 95% CI: 1.04‐1.62, p = 0.028) were strongly linked to dementia risk. In addition, tetanus (Clostridium tetani) seropositivity was also associated with increased AD risk (HR = 1.43, 95% CI: 1.05‐1.94, p = 0.029). Porphyromonas gingivalis (Pg) and streptococcus oralis (So) IgG seropositivity was associated with higher AD risk at higher IB levels (Pg: HR = 1.19, 95% CI:1.01‐1.40, p = 0.045; So: HR = 1.26, 95% CI: 1.06‐1.50, p = 0.011, per standard deviation [SD]). Among ethnic minorities, the periodontal pathogens from the red‐green pathogen cluster coupled with higher IB count [RedGreen×IB: γ = +0.52 ± 0.14, p = 0.001] increased risk of all‐cause dementia. In contrast, factor 1 (the largest factor) was associated with all‐cause dementia mainly at lower IB in the overall sample (HR = 1.21, 95% CI: 1.04‐1.42, p = 0.022, per SD), and an antagonistic interaction was noted for pocket probing depth and all‐cause dementia (IB⁻: HR = 1.59, 95% CI: 1.14‐2.21, p = 0.009, per SD).

3. Future directions: The study results suggest that aging immune cells may exacerbate the effects of periodontal infection and persistent viral, bacterial, and parasite infections on cognitive impairment and neurodegeneration. It highlights associations between hepatitis C, HSV‐2, and tetanus and the combined effects of infection burden and specific periodontal bacteria with dementia. Randomized controlled studies are needed to investigate the impact of eradicating chronic infections and specific periodontal bacteria on neurodegenerative diseases.

2.2. Study sample

Figure 1 shows a participant flowchart with the number of occurrences and the sample at risk. Initially, NHANES III participants who were 45 years of age or older were included. Complete seropositivity data that allowed us to compute the IB (both phases of NHANES III, 1988‐1994), mortality status (1988‐1994), CMS‐linkage data (1988‐1994), and complete data on all 19 periodontal pathogens (surplus serum for participants ages 40 years or older, 1988‐1994, antibodies to periodontal pathogens (SPSDEPPX) were required. Of 33,199 persons aged 1‐90 years (sample 1), 9787 were over age of 45 years (sample 2), while having complete data on both periodontal pathogen IgG and the other pathogens used to compute IB was restricted to a sample of 2997 individuals who were age 45 years or older (sample 3). This latter sample was reduced to a maximum of 2975 after exclusion for Health Maintenance Organization (HMO) utilization and taking incomplete data into consideration (sample 4). Additionally, 2398 participants aged 45 years or more at the Medical Examination Center (MEC) exam had data on Pd markers, namely clinical attachment loss (CAL) and pocket probing depth (PPD) simultaneously along with IB (sample 5). Individuals without CMS‐Medicare data were either censored at the time of death or considered to have no event of interest until the end of 2018. Most analyses were carried out on sample 4, while part of the analysis was conducted on a subsample of sample 5, which also included participants in sample 4 (sample 6, N = 2038). Figure 1 also shows the number of incident cases of AD and dementia from all causes at each level of selection between samples 2 and 5.

FIGURE 1.

Participant flowchart. Note: Both phases: 1988‐1994. AD, Alzheimer's disease; CMS, Centers for Medicare and Medicaid Services; NHANES III, Third National Health and Nutrition Examination Surveys.

2.3. Dementia and AD onset

All‐cause dementia and AD were defined using comprehensive data from the CMS Chronic Condition Data Warehouse Categories, which included 21 chronic diseases with different reference periods, amounts and types of qualifying claims, exclusions, and a set of ICD‐9/CPT4/HCPCS codes. ICD‐9 code 331.0 (any DX on the claim) from inpatient, Skilled Nurse Practitioner (SNP), Home Health Agency (HHA), Health Options Program [HOP], or carrier claims over a 3‐year period was used to diagnose AD. Any of the following diagnostic codes (331.19, 331.2, 331.7, 290.0, 290.10, 290.11, 290.12, 290.13, 290.20, 290.21, 290.3, 290.40, 290.41, 290.42, 290.43, 294.0, 294.1, 294.11, 294.8 and 797) could be used to diagnose dementia from all causes. The MEC examination date was used as the starting point for the calculation of the earliest occurrence when using time on study as the underlying time scale. Regarding the pre‐estimated earliest occurrence date, the follow‐up period was 1999‐2018. The follow‐up period was stated in months and was up to December 31, 2018. We estimated the earliest diagnostic date of AD/dementia for the years 1991‐1998 using the same technique. ²⁶

2.4. Dental examination and definitions of Pd markers

CAL and PPD were obtained from dental examinations conducted during both MEC visit stages in NHANES III to identify Pd. ²⁷ Each tooth in the maxilla and mandible, two randomly chosen quadrants, had its CAL and probing depth measured at the mid‐ and mesiobuccal locations (range: 14‐28 teeth). ²⁸ The distance from the base of the periodontal pocket to the free gingival margin was designated as PPD, and the distance between the cementoenamel junction and the periodontal pocket was designated as CAL. ²⁹ For 28 sites, mean CAL and PPD were determined. For the CAL/PPD measure, two non‐missing sites per tooth were needed, and in order to estimate the mean, at least one tooth measurement was needed. ²⁹ Pd has been defined and classified using a combination of PPD and CAL in prior studies, with higher cutoffs measured in mm, indicative of more severe disease. ³⁰ , ³¹ , ³² , ³³

2.5. Periodontal pathogens

Using checkerboard immunoblotting and a series of 1:1000 serum dilutions, serum immunoglobulin G (IgG) titers for 19 periodontal bacteria were determined, as previously reported. ³⁴ Those pathogens are:(1) Aggregatibacter actinomycetemcomitans (Aa) (American Type Culture Collection [ATCC] strains 43,718, 29,523, and 33,384); (2) Pg (ATCC strains 3327 and 53,978); (3) Tannerella forsythia (ATCC strain 43,037); (4) Treponoma denticola (Oral Microbiology, Gothenburg, Sweden [OMGS] strain 3271); (5) Cr (ATCC strain 33,238); (6) Eubacterium nodatum (En), ATCC strain 33,099); (7) Pi (ATCC strain 25,611); (8) Pn (ATCC strain 33,563); (9) Pm (ATCC strain 25,845); (10) Fn (ATCC strain 33,563); (11) Parvimonas micra aka Micromonas micros (mm) (ATCC strain 10,953); (12) Selenomonas noxia (Sn) (ATCC strain 43,541); (13) Eikenella corrodens (ATCC strain 23,834); (14) Co (ATCC strain 33,624); (15) Si (ATCC strain 35,037); (16) Streptococcus oralis (So, ATCC strain 35,037); (17) Streptococcus mutans (Sm) (ATCC strain 25,175); (18) Vellonella parvula (Vp) (ATCC strain 10,790); and (19) Actinomyces naeslundii (An) (ATCC strain 49,340). ³⁵ The human IgG standard curve was utilized to measure and compare IgG titers using a chemiluminescent signal‐measuring apparatus. ³⁴ At Columbia University College of Dental Medicine in New York, NY, 8153 preserved serums (Phase 2 then Phase 1 samples) were examined between 2003 and 2006. Using varimax rotation, eigenvalue and scree plot criteria, and factor analysis, we extracted independent common factors from people 45 years of age or older (Appendix II). Then, we concurrently included these variables in models to investigate relationships between periodontal infections and the incidence, mortality, and all‐cause dementia of AD. Last, using the methodology from a prior study, we adjusted mutually exclusive color‐coded clusters. ³⁵ To create the clusters, we added up all the Log_e‐transformed pathogen titers, and then we z‐scored them (Figure S1). ³⁵

2.6. Persistent infections and infection‐burden measure

Serology data on hepatitis A, B, and C viruses, herpes simplex types 1 (HSV‐1) and 2 (HSV‐2) viruses, varicella‐zoster virus, rubella, tetanus, and Toxoplasma gondii are made available among part of the NHANES III participants. Detailed descriptions of the lab procedures and criteria for seropositivity are found at https://wwwn.cdc.gov/nchs/data/nhanes3/manuals/labman.pdf and https://wwwn.cdc.gov/nchs/data/nhanes3/1a/lab‐acc.pdf.

In brief, a solid‐phase competitive enzyme immunoassay was measured using a solid‐phase competitive enzyme immunoassay for hepatitis A (HAV) and hepatitis C (HCV), while a sandwich enzyme immunoassay was used to measure hepatitis B (HBV) (Centers for Disease Control, 1996, Laboratory Procedures Used for the Third National Health and Nutrition Examination Survey (NHANES III), 1988‐1994, VII‐V‐1 to VII‐Z‐1). Although seropositivity for HAV can be due to a history of either vaccination or infection, the hepatitis A vaccine was not available in the United States until 1995, which occurred after the period of data collection for the NHANES III. ³⁶ Thus, all participants in the current study with antibodies to HAV had the antibodies due to a past infection and not vaccination. A positive reaction to glycoprotein gG‐1 and gG‐2 antigen by immunodot assay was used to measure seropositivity to HSV‐1 and HSV‐2 antibodies, respectively. ³⁷ Seropositivity determinations for hepatitis A, B, and C, and herpes 1, and 2 were made by NHANES. Varicella zoster and rubella seropositivities were identified using an indirect enzyme immunoassay. Antibodies were reported on an optical density index and respondents were seropositive for varicella zoster if the index was 1 or greater ³⁸ (Kilgore et al.) and seropositive for rubella if the index was 10 or greater. ³⁹ Tetanus seropositivity was identified using a solid‐phase enzyme‐linked immunosorbent assay (ELISA). Consistent with Gergen et al. (1995), ⁴⁰ we considered respondents positive for tetanus if antibodies were above 0.15 IU/mL. Toxoplasma gondii was identified using indirect enzyme immunoassay. The optical density of human gamma chain immunoglobulin was compared to a World Health Organization standard curve. The determination point for Toxoplasma gondii seropositivity was 33 IU/mL. An index was constructed to determine an overall count reflecting the IB. The final selected sample was then subdivided into two groups: 0 = below the median of IB, and 1 = Above the median of IB.

2.7. Covariates

Covariates were chosen based on historical evidence for their correlations with periodontal and other persistent infections and/or cognitive impairment. The three demographic factors were age, sex, and race/ethnicity (non‐Hispanic [NH] White, NH black, Mexican‐American, and Other ethnic groups). In addition, we included numerous indicators of socioeconomic position, social support, lifestyle and health traits, and dietary quality. We also took into consideration blood biomarkers for dietary quality and dentate status markers, which we thought could be confounding factors (See Appendix III for details). We stratified all models by race/ethnicity and IB status.

2.8. Statistical analysis

We used Stata 18.0 (StataCorp, College Station, TX) for all analyses. ⁴¹ In order to account for the complexity of the survey design, we included primary sample units and strata that span 6 years (NHANES III, Phase 1, 1988‐1994). Using Taylor series linearization (i.e., svy: commands), standard errors were computed. ⁴¹ After accounting for the complexity of the sample design, multivariate imputed data (m = 5 imputations, 10 iterations) with chained equations ⁴² computed means and proportions across IB strata and measures of association.

The primary exposures of interest were the 19 periodontal pathogens added both independently (Model 2) and simultaneously (Model 1). Pre‐established color‐coded clusters and data‐driven factors taken out of those 19 periodontal infections were also thought to be significant predictors. After controlling for all confounders and stratifying by IB status, these periodontal pathogens were inserted as predictors into linear regression models with periodontitis measured with CAL and PPD as outcomes. Last, in the primary causal models, namely incident AD and all‐cause dementia, the means of CAL/PPD were used as independent predictors. We ran Cox proportional hazards models for two incidence outcomes, stratifying separately by IB status, in those models, defining time‐to‐event from any age greater than or equal to age 45 years since baseline visit (i.e., delayed entry) until death, censoring, or outcome of interest (AD incidence and all‐cause dementia incidence). We show completely adjusted models that take into consideration lifestyle and social support factors, dietary biomarkers, health‐related factors, socioeconomic, and demographic variables detailed in Appendix III. Weighted person‐months of follow‐up and the weighted sample in each model ([Person‐months_(weighted)]/[Persons_(weighted)]) were used to estimate the weighted mean times of follow‐up. For statistical significance, a type I error of 0.05 was used, and 0.10 was considered to be borderline (or marginal) significance. Using a familywise Bonferroni approach, multiple testing adjustment was carried out, assuming that each exposure, model, and stratum was a different hypothesis and accounting for result multiplicity (e.g., CAL/Probing depth: 2 outcomes; incident AD/all‐cause dementia: two outcomes). ⁴³ Consequently, type I error for main effects was lowered to 0.05 for findings that were marginally significant and to 0.025 for findings that were statistically significant. To evaluate potential synergistic effects, two‐way interactions were added between each periodontal exposure and IB status. As these interaction terms had less power than main effects, we deemed a type I error of 0.10 to be statistically significant. ⁴⁴ The study population, NH White and racial minority group combined (NH black, Mexican‐American and Others), individuals all had these interaction terms examined. To accommodate for multiple testing relevant to the outcome, the type I error for interaction terms was lowered to 0.05.

3. RESULTS

Based on Table 1 depicting study characteristics by IB status, among the final selected sample (unweighted N = 2975), the weighted proportion of IB ⁺ was 46.2% (unweighted N = 1690) with an IB ranging between five and nine seropositivities, while the remaining sample (unweighted N = 1285, weighted proportion: 53.8%) was IB ⁻ with seropositivities ranging between one and four. Individual weighted seroprevalence varied widely between types of pathogens ranging between 1‐2% for hepatitis C to over 99% for varicella. Remaining weighted seroprevalence proportions based on the larger sample aged 45+ were: 55% of hepatitis A, 7.4% for hepatitis B, 81.1% for HSV1, 24.9% for HSV2, 35.4% for toxoplasmosis gondii, 46.8% for tetanus, and 94.7% for rubella. The cumulative incidence proportions of AD and all‐cause dementia were estimated at 11.4% and 22.3%, respectively, for the IB ⁻ group; and at 14.3 and 26.2% for the IB ⁺ group (Table 1) and a difference was only detected for all‐cause dementia (OR = 1.23, 95% CI: 1.03‐1.48, p = 0.026). Moreover, compared with the IB ⁻ group, the IB ⁺ group was older on average by 3.1 years, with a greater proportion of males (50.0% vs. 41.0%), of racial minority groups (NH White was 75.1% in the IB⁺ group as opposed to 91.1% in the IB⁻ group), and of “widowed”/“other” categories of marital status, a lower PIR, fewer years of education, lower diet quality based on both HEI‐1995 and MAR, and poorer lifestyle based on several markers of physical activity and smoking, higher proportions of fair/poor self‐rated health (28% vs. 16%), as well as higher means on the co‐morbidity and allostatic load indices and a greater proportion obese (30% vs. 22%). Furthermore, the IB ⁺ group, compared with the IB⁻ group, was also characterized by lower vitamin D, folate, and selenium status (p < 0.05) coupled with greater ferritin level in blood (p < 0.001), and higher proportion partial or complete edentulation coupled with greater CAL and PPD. Reduced IgG levels against periodontal pathogens were generally observed in the IB⁺ group compared with the IB⁻ groups, with the exception of Pg IgG, with similar patterns observed for the factors and clusters.

TABLE 1.

Baseline characteristics of selected participants by infection burden (IB) status, NHANES III, 1988‐1994 (N = 2975). ^a

	IB status
Selected participant characteristics	IB −	IB +	p‐value b (design‐based F‐test)2
Unweighted N (both phases)	(N = 1285)	(N = 1690)	IB+ vs. IB−
Weighted %	53.8%	46.2%
Socio‐demographic characteristics
Age (years)	58.06 ± 0.76	61.16 ± 0.67	<0.001
Sex, % male	41.09 ± 0.59	50.07 ± 0.02	0.001
Race/ethnicity, %
Non‐Hispanic White	90.0 ± 1.0	76.0 ± 3.0
Non‐Hispanic Black	4.0 ± 1.0	13.0 ± 2.0	<0.001
Mexican‐American	2.0 ± 1.0	4.0 ± 1.0	<0.001
Other	2.0 ± 1.0	7.0 ± 1.0	<0.001
Urban/rural area of residence, %			0.14
Urban	5.0 ± 8.0	44.0 ± 7.0
Rural	50.0 ± 8.0	56.0 ± 7.0
Household size	2.5 ± 0.06	2.54 ± 0.06	0.61
Marital status,%
Never married	5.0 ± 1.0	4.0 ± 1.0	0.985
Married	72.0 ± 2.0	63.0 ± 2.0	Ref
Divorced	9.0 ± 2.0	9.0 ± 1.0	0.613
Widowed	12.0 ± 1.0	18.0 ± 2.0	0.001
Other	2.0 ± 1.0	6.0 ± 1.0	0.001
Socioeconomic status
Poverty income ratio	3.55 ± 0.12	2.76 ± 0.1	<0.001
Education, years	12.54 ± 0.23	10.86 ± 0.19	<0.001
Nutritional factors
1995‐HEI total score	67.21 ± 0.55	64.2 ± 0.52	<0.001
MAR total score	74.86 ± 0.66	71.79 ± 0.62	0.001
Physical activity 0 = Less, 1 = Same, 2 = more
Compare activity for past month to past year, %
Less	27.0 ± 2.0	23.0 ± 2.0	0.013
Same	57.0 ± 2.0	66.0 ± 2.0	Ref
More	16.0 ± 2.0	11.0 ± 1.0	0.008
Active compared with men/women your age, %
Less	18.0 ± 2.0	18.0 ± 2.0	0.77
Same	42.0 ± 2.0	43.0 ± 2.0	Ref
More	40.0 ± 0.02	38.0 ± 2.0	0.42
Active now compared with self 10 years ago, %
Less	52.0 ± 2.0	62.0 ± 2.0	<0.001
Same	35.0 ± 2.0	28.0 ± 2.0	Ref
More	13.0 ± 2.0	11.0 ± 1.0	0.030
Smoking
# Cigarettes/day	6.89 ± 0.51	8.05 ± 0.52	0.094
Years smoked cigarettes	6.3 ± 0.47	8.46 ± 0.45	0.004
Alcohol use	7.14 ± 0.98	7.2 ± 0.8	0.961
Drug use, % yes	13.0 ± 2.0	13.0 ± 2.0	0.980
Social support
(1) In a typical week, how many times do you talk on the telephone with family, friends, or neighbors?	10.73 ± 0.75	9.39 ± 0.52	0.106
(2) How often do you get together with friends or relatives; I mean things like going out together or visiting in each other's homes? (per year)	104.58 ± 9.05	103.03 ± 6.07	0.862
(3) About how often do you visit with any of your other neighbors, either in their homes or in your own? (per year)	63.89 ± 10.85	66.02 ± 6.56	0.809
(4) How often do you attend church or religious services? (per year)	38.1 ± 2.46	36.83 ± 2.12	0.669
(5) Altogether, how often do you attend meetings of the clubs or organizations (per year)	13.96 ± 1.37	12.04 ± 0.97	0.258
Health‐related factors
Self‐rated health
Excellent/Very Good	52.0 ± 3.0	35.0 ± 3.0	Ref
Good	32.0 ± 2.0	37.0 ± 2.0	<0.001
Fair/Poor	16.0 ± 2.0	28.0 ± 2.0	<0.001
Co‐morbidity index	1.04 ± 0.05	1.27 ± 0.05	<0.001
AL components, %	10.73 ± 0.75	9.39 ± 0.52	<0.001
Low albumin	8.0 ± 1.0	9.0 ± 1.0	0.58
High CRP	29.0 ± 2.0	37.0 ± 3.0	0.005
High waist‐hip ratio	74.1 ± 2.0	88.0 ± 1.0	<0.001
High total cholesterol	29.0 ± 2.3	35.2 ± 1.8	0.084
Low HDL‐C	22.9 ± 1.3	25.6 ± 1.6	0.14
High glycated hemoglobin	7.0 ± 1.0	11.2 ± 1.2	0.020
High resting heart rate	6.9 ± 0.01	6.3 ± 0.8	0.59
High systolic blood pressure	25.6 ± 1.4	33.4 ± 2.0	0.001
High diastolic blood pressure	8.7 ± 1.2	9.5 ± 1.2	0.59
Allostatic load, mean ± SE	2.13 ± 0.08	2.55 ± 0.06	<0.001
Weight status
Underweight	1.5 ± 0.4	2.2 ± 0.6	0.12
Normal	40.9 ± 1.9	31.0 ± 1.6	Ref
Overweight	35.2 ± 2.0	36.8 ± 1.9	0.017
Obese	22.3 ± 1.3	30.0 ± 1.6	<0.001
Nutritional biomarkers, mean ± SE
25(OH)D	28.54 ± 0.49	26.85 ± 0.48	0.006
Folate	7.82 ± 0.31	7 ± 0.19	0.015
Vitamin C	163.78 ± 41.24	113.04 ± 34.37	0.195
Vitamin A	62.12 ± 0.6	63.27 ± 0.68	0.179
Total carotenoids	85.28 ± 1.74	82.81 ± 2.14	0.297
Vitamin E	1343.33 ± 24.47	1305.16 ± 22.06	0.205
Ferritin	136.79 ± 4.77	165.04 ± 6.28	<0.001
Normalized calcium	1.23 ± 0	1.23 ± 0	0.112
Selenium	125.14 ± 1.55	123.06 ± 1.23	0.029
Dental measures
Periodontitis, mean ± SE
CAL	1.61 ± 0.05	2.07 ± 0.09	<0.001
Pocket probing depth	1.51 ± 0.04	1.71 ± 0.04	<0.001
Dentate status
Completely edentulous	14.6 ± 1.5	22.5 ± 2.1	<0.001
Edentulous in one arch	11.7 ± 1.4	16.4 ± 1.3	0.003
Teeth complete	73.7 ± 2.4	61.2 ± 2.2	Ref
Periodontal pathogens c
P. Gingivalis (Pg) mix	5.30 ± 0.07	5.52 ± 0.08	0.008
P. Intermedia (Pi)	5.55 ± 0.06	5.56 ± 0.06	0.894
P. Nigrescens (Pn)	5.29 ± 0.05	5.11 ± 0.05	0.006
T. Forsythia (Tf)	4.33 ± 0.07	4.1 ± 0.06	0.001
A. Actinomycetemcomitans (Aa) mix	6.21 ± 0.06	6.13 ± 0.07	0.184
F. Nucleatum (Fn)	4.38 ± 0.05	4.24 ± 0.05	0.042
S. Oralis (So)	3.92 ± 0.07	3.77 ± 0.05	0.023
M. Micros (Mm)	5.09 ± 0.07	4.85 ± 0.06	0.003
C. Rectus (Cr)	4.05 ± 0.06	4.05 ± 0.05	0.994
E. Corrodens (Ec)	4.82 ± 0.08	4.75 ± 0.08	0.291
E. Nodatum (En)	6.74 ± 0.1	6.36 ± 0.08	0.001
S. Intermedius (Si)	4.74 ± 0.07	4.65 ± 0.06	0.160
C. Ochracea (Co)	5.01 ± 0.03	4.87 ± 0.05	0.005
V. Parvula (Vp)	3.56 ± 0.06	3.39 ± 0.06	0.007
A. Naeslundii (An)	6.26 ± 0.08	5.9 ± 0.09	<0.001
P. Melaninogenica (Pm)	5.42 ± 0.05	5.3 ± 0.05	0.026
S. Noxia (Sn)	3.29 ± 0.08	3.09 ± 0.06	0.009
T. Denticola (Td)	4.43 ± 0.1	4.33 ± 0.07	0.259
S. Mutans (Sm)	4.29 ± 0.06	4.14 ± 0.05	0.018
Factors (z‐scores), mean ± SE
Factor 1	0.02 ± 0.06	−0.03 ± 0.04	0.368
Factor 2	−0.09 ± 0.04	−0.12 ± 0.04	0.576
Factor 3	0.21 ± 0.04	0.02 ± 0.04	<0.001
Clusters (z‐scores), mean ± SE
Orange‐Red	−0.12 ± 0.04	−0.14 ± 0.04	0.800
Red‐Green	0.01 ± 0.06	−0.08 ± 0.05	0.042
Yellow‐Orange	0.05 ± 0.04	−0.08 ± 0.03	0.002
Orange‐Blue	0.16 ± 0.04	−0.04 ± 0.04	<0.001
Cumulative incidence of AD and all‐cause dementia and of AD mortality, %
AD dementia	11.0 ± 1.0	14.0 ± 1.0	0.054
All‐cause dementia	22.0 ± 2.0	26.0 ± 2.0	0.026

Abbreviations: 25(OH)D, 25‐hydroxyvitamin D; AD, Alzheimer's disease; CAL, clinical attachment loss; CRP, C‐reactive protein; HDL‐C, high density lipoprotein cholesterol; IB, infection burden; NHANES III, Third National Health and Nutrition Examination Survey; Pd, periodontal disease; SE, standard error.

^a Values are weighted means ± SEM or percent ± SEP, considering sampling design complexity (PSU and strata), across five imputations with 10 iterations.

^b Design‐based F‐test accounting for design complexity in terms of sampling weights, PSU and stratum. for categorical variables, this was the equivalent of a χ² test of independence. For continuous variables, it was the equivalent of a t‐test in a linear regression model with the variable being the outcome predicted IB status.

^c SD of Log_e transformed periodontal pathogens across IB status groups ranged between 1.3 and 2.0.

Table 2 shows findings for individual titers as potential predictor for incidence of AD and all‐cause dementia, adjusting in Model 2 for all potential confounders and in Model 1 additionally adjusted for all other titers. While most of these associations were null and did not differ across IB groups, two notable associations were detected. Specifically, Pg and So IgG were associated with higher AD risk at higher IB levels in Model 1 (Pg: HR = 1.19, 95% CI:1.01‐1.40, p = 0.045; So: HR = 1.26, 95% CI: 1.06‐1.50, p = 0.011, per standard deviation, SD). In contrast, titers against Aa and Fn were positively associated with incident AD within the IB ⁻ group, and several findings indicated inverse associations of specific titers against AD and all‐cause dementia incidence, including T. Denticola in the IB ⁺ group for all‐cause dementia (Model 1).

TABLE 2.

Periodontal pathogens’ serum IgG association with incident Alzheimer's Disease and all‐cause dementia in multiple Cox proportional hazards model, overall and restricted by Infection burden (IB) status, interaction by IB status (overall and by race/ethnicity): NHANES III, 1988‐1994. ^a , ^b

	IB −, c			IB +, d
Parameter	Loge(HR)	(SE)	p‐value	Loge(HR)	(SE)	p‐value
AD incidence
P. Gingivalis (Pg) mix
Model 1	+0.04	(0.11)	0.71	+0.18	(0.11)	0.12
Model 2	+0.11	(0.12)	0.39	+0.17	(0.08)	0.047 *
P. Intermedia (Pi)
Model 1	−0.43	(0.20)	0.037 *	+0.20	(0.26)	0.45
Model 2	+0.02	(0.16)	0.90	+0.14	(0.09)	0.11
P. Nigrescens (Pn)
Model 1	+0.21	(0.30)	0.49	−0.29	(0.19)	0.12
Model 2	+0.10	(0.17)	0.58	+0.01	(0.09)	0.92
T. Forsythia (Tf)
Model 1	+0.33	(0.18	0.068	−0.02	(0.16)	0.88
Model 2	+0.30	(0.18)	0.11	+0.12	(0.09)	0.19
A. Actinomycetemcomitans (Aa) mix
Model 1	+0.39	(0.19)	0.048 *	−0.01	(0.16)	0.96
Model 2	+0.26	(0.19)	0.17	+0.18	(0.09)	0.057
F. Nucleatum (Fn)
Model 1	+0.35	(0.19)	0.046 *	+0.15	(0.20)	0.45
Model 2	+0.24	(0.16)	0.13	+0.10	(0.08)	0.20
S. Oralis (So)
Model 1	−0.05	(0.25)	0.84	+0.32	(0.21)	0.14
Model 2	+0.24	(0.16)	0.13	+0.23	(0.09)	0.011 **
M. Micros (Mm)
Model 1	−0.12	(0.18)	0.54	+0.12	(0.15)	0.44
Model 2	−0.09	(0.16)	0.54	+0.17	(0.09)	0.073
C. Rectus (Cr)
Model 1	−0.08	(0.22)	0.70	−0.09	(0.14)	0.52
Model 2	+0.19	(0.19)	0.31	+0.13	(0.08)	0.11
E. Corrodens (Ec)
Model 1	+0.01	(0.18)	0.94	−0.11	(0.17)	0.50
Model 2	+0.21	(0.14)	0.15	+0.08	(0.08)	0.33
E. Nodatum (En)
Model 1	+0.08	(0.17)	0.64	−0.04	(0.22)	0.86
Model 2	+0.10	(0.15)	0.51	+0.07	(0.10)	0.47
S. Intermedius (Si)
Model 1	−0.55	(0.29)	0.065	−0.14	(0.11)	0.20
Model 2	+0.00	(0.23)	0.98	+0.07	(0.10)	0.47
C. Ochracea (Co)
Model 1	+0.01	(0.17)	0.95	−0.25	(0.14)	0.08
Model 2	+0.24	(0.15)	0.94	+0.05	(0.09)	0.61
V. Parvula (Vp)
Model 1	+0.41	(0.21)	0.052	+0.03	(0.15)	0.87
Model 2	+0.31	(0.19)	0.11	+0.13	(0.09)	0.17
A. Naeslundii (An)
Model 1	−0.13	(0.15)	0.38	−0.03	(0.19)	0.89
Model 2	+0.02	(0.14)	0.90	+0.04	(0.09)	0.65
P. Melaninogenica (Pm)
Model 1	+0.04	(0.25)	0.87	+0.14	(0.23)	0.53
Model 2	+0.13	(0.18)	0.49	+0.07	(0.11)	0.48
S. Noxia (Sn)
Model 1	−0.47	(0.28)	0.10	+0.02	(0.22)	0.92
Model 2	+0.18	(0.17)	0.31	+0.13	(0.09)	0.16
T. Denticola (Td)
Model 1	−0.20	(0.16)	0.24	−0.11	(0.16)	0.49
Model 2	+0.11	(0.12)	0.39	+0.03	(0.08)	0.74
S. Mutans (Sm)
Model 1	+0.33	(0.26)	0.20	+0.12	(0.20)	0.49
Model 2	+0.27	(0.18)	0.14	+0.19	(0.10)	0.052
All‐cause dementia incidence
P. Gingivalis (Pg) mix
Model 1	+0.08	(0.07)	0.28	+0.04	(0.09)	0.63
Model 2	+0.13	(0.08)	0.092	−0.003	(0.07)	0.97
P. Intermedia (Pi)
Model 1	−0.20	(0.17)	0.24	−0.12	(0.15)	0.43
Model 2	+0.07	(0.11)	0.51	−0.04	(0.05)	0.42
P. Nigrescens (Pn)
Model 1	−0.03	(0.21)	0.87	−0.02	(0.11)	0.86
Model 2	+0.07	(0.11)	0.56	−0.08	(0.06)	0.20
T. Forsythia (Tf)
Model 1	+0.05	(0.10)	0.62	+0.06	(0.13)	0.64
Model 2	+0.15	(0.11)	0.19	+0.01	(0.07)	0.93
A. Actinomycetemcomitans (Aa) mix
Model 1	+0.06	(0.16)	0.72	+0.09	(0.12)	0.47
Model 2	+0.13	(0.12)	0.28	+0.05	(0.08)	0.50
F. Nucleatum (Fn)
Model 1	+0.15	(0.15)	0.73	+0.10	(0.15)	0.49
Model 2	+0.13	(0.10)	0.20	−0.01	(0.07)	0.90
S. Oralis (So)
Model 1	+0.02	(0.25)	0.93	+0.13	(0.16)	0.45
Model 2	+0.18	(0.09)	0.051	+0.06	(0.07)	0.41
M. Micros (Mm)
Model 1	−0.06	(0.12)	0.62	+0.12	(0.07)	0.11
Model 2	+0.07	(0.10)	0.49	+0.06	(0.07)	0.41
C. Rectus (Cr)
Model 1	+0.00	(0.14)	0.99	−0.07	(0.10)	0.48
Model 2	+0.14	(0.11)	0.21	−0.02	(0.06)	0.81
E. Corrodens (Ec)
Model 1	+0.13	(0.14)	0.38	+0.09	(0.11)	0.41
Model 2	+0.17	(0.08)	0.027 *	−0.05	(0.08)	0.50
E. Nodatum (En)
Model 1	−0.12	(0.12)	0.29	−0.08	(0.13)	0.41
Model 2	+0.02	(0.09)	0.83	−0.05	(0.08)	0.50
S. Intermedius (Si)
Model 1	−0.33	(0.18)	0.065	−0.11	(0.10)	0.28
Model 2	+0.01	(0.13)	0.94	−0.00	(0.06)	1.00
C. Ochracea (Co)
Model 1	−0.12	(0.11)	0.29	−0.02	(0.09)	0.83
Model 2	+0.12	(0.10)	0.21	−0.05	(0.07)	0.52
V. Parvula (Vp)
Model 1	+0.20	(0.14)	0.17	+0.10	(0.11)	0.34
Model 2	+0.18	(0.11)	0.096	−0.02	(0.07)	0.74
A. Naeslundii (An)
Model 1	+0.05	(0.09)	0.55	−0.02	(0.13)	0.89
Model 2	+0.03	(0.08)	0.67	−0.05	(0.06)	0.41
P. Melaninogenica (Pm)
Model 1	+0.20	(0.18)	0.27 e	−0.07	(0.14)	0.61
Model 2	+0.16	(0.12)	0.19 e	−0.05	(0.06)	0.41
S. Noxia (Sn)
Model 1	−0.09	(0.19)	0.61	−0.19	(0.14)	0.17
Model 2	+0.16	(0.10)	0.12 e	+0.05	(0.07)	0.45
T. Denticola (Td)
Model 1	−0.13	(0.13)	0.29	−0.20	(0.09)	0.032 *
Model 2	+0.09	(0.07)	0.22	−0.04	(0.06)	0.49
S. Mutans (Sm)
Model 1	+0.27	(0.20)	0.19	+0.08	(0.12)	0.51
Model 2	+0.22	(0.11)	0.061	+0.04	(0.06)	0.55

Abbreviations: 25(OH)D, 25‐hydroxyvitamin D; AD, Alzheimer's disease; HR, hazard ratio; IB, infection burden; IgG, immunoglobulin G; NHANES III, Third National Health and Nutrition Survey. See full list of abbreviations for periodontal pathogen titers.

^a Models were adjusted for age, sex, race/ethnicity, poverty income ratio, education (years), urban‐rural area of residence, household size, marital status, nutritional factors (HEI, MAR), nutritional biomarkers (25(OH)D, folate, vitamin C, vitamin A, total carotenoids, vitamin E, ferritin, selenium, and normalized calcium), lifestyle (smoking, drug use, alcohol, physical activity), health‐related factors (self‐rated health, co‐morbidity index, allostatic load, weight status), dentate status and social support variable, as well as Phase of NHANES III. Covariates (other than exposures) were imputed and analysis is across five imputations with 10 iterations. Model 1: adjusted for all other periodontal pathogens; Model 2: one periodontal pathogen at a time.

^b Periodontal pathogen exposures were Log_e transformed and then standardized into z‐scores. See abbreviation list for full name of each pathogen.

^c Unweighted N = 1186 for Model 1 and 1194‐1267 for Model 2, unweighted mean follow‐up time for smallest sample: 252 months (AD incidence), 246 months (all‐cause dementia).

^d Unweighted N = 1569 for Model 1 and N = 1580‐1672 for Model 2, unweighted mean follow‐up time for smallest sample: 224 months (AD incidence), 216 months (all‐cause dementia).

^e p < 0.10 for IB (above median = 1, below median = 0) by Periodontal pathogen exposures (Log_e transformed and then standardized into z‐scores) 2‐way interaction in unstratified model.

^* p < 0.05, marginally significant after correction for multiple testing.

^** p < 0.025, significant after correction for multiple testing.

^┼ p < 0.05, significant interaction.

According to Table 3, among racial minorities, several clusters including the Red Orange, Red‐Green, and Yellow‐Orange clusters and factors (mainly factors 1 and 2) interacted synergistically with IB⁺ status [e.g. Red‐Green × IB γ = +0.52 ± 0.14, p = 0.001]. Specifically, higher titers against periodontal pathogens from the Red‐Green cluster coupled with higher IB count increased the risk of all‐cause dementia in these racial groups. In contrast, based on Table 4, PPD was associated with all‐cause dementia mainly at lower IB in the overall sample (IB⁻: HR = 1.59, 95% CI: 1.14‐2.21, p = 0.009, per SD). In fact, according to the findings presented in supplementary Table IV.1 (Appendix IV), periodontal pathogen titers were generally directly associated with clinical Pd markers, namely CAL and PPD, with a significantly stronger association of Pg with PPD at higher IB level. Other titers such as Aa showed an antagonistic interaction with IB with respect to both CAL and PPD. In addition, and based on Table 5, despite a lack of IB‐dementia relationship, hepatitis C (hazard ratio [HR] = 3.33, 95% CI: 1.52‐7.26, p = 0.004) and herpes simplex virus 2 (HSV‐2) (HR = 1.30, 95% CI: 1.04‐1.62, p = 0.028) were strongly linked to all‐cause dementia risk, while tetanus seropositivity was associated with increased AD risk (HR = 1.43, 95% CI: 1.05‐1.94, p = 0.029).

TABLE 3.

Periodontal pathogens’ serum IgG (factor scores and pre‐defined clusters) association with AD mortality, AD incidence and all‐cause dementia incidence in multiple Cox proportional hazards model, overall and stratified by IB status and race/ethnicity: NHANES III, 1988‐1994 ^a , ^b

	IB −, c			IB +, d			Interaction with IB status, overall			Interaction with IB status, NH whites			Interaction with IB status, Other racial/ethnic groups
Parameter	Loge(HR)	(SE)	p‐Value	Loge(HR)	(SE)	p‐value	Loge(HR)	(SE)	p‐value	Loge(HR)	(SE)	p‐value	Loge(HR)	(SE)	p‐value
Factor scores
AD incidence
Factor 1: So/Sn/Td	+0.28	(0.15)	0.064	+0.17	(0.09)	0.06	−0.11	(0.17)	0.53	−0.25	(0.19)	0.19	+0.71	(0.22)	0.001 ** , ┼
Factor 2:Pi/Pn/Pm	+0.03	(0.17)	0.89	+0.09	(0.1)	0.37	+0.09	(0.19)	0.63	−0.01	(0.21)	0.97	+0.71	(0.28)	0.02 * , **
Factor 3:An/En	+0.06	(0.14)	0.66	−0.06	(0.11)	0.60	−0.07	(0.19)	0.71	−0.10	(0.21)	0.65	+0.51	(0.35)	0.15
All‐cause dementia incidence
Factor 1: So/Sn/Td	+0.19	(0.08)	0.022 **	+0.06	(0.08)	0.47	−0.15	(0.12)	0.23	−0.24	(0.13)	0.074 **	+0.60	(0.16)	<0.001 ** , ┼
Factor 2:Pi/Pn/Pm	+0.03	(0.12)	0.81	−0.06	(0.07)	0.42	−0.09	(0.12)	0.46	−0.17	(0.14)	0.23	+0.37	(0.17)	0.029 ** , ┼
Factor 3:An/En	−0.02	(0.1)	0.85	−0.09	(0.08)	0.25	−0.12	(0.11)	0.28	−0.14	(0.13)	0.29	+0.23	(0.19)	0.23
Pre‐defined clusters
AD incidence
Cluster 1: Orange‐Red	−0.11	(0.15)	0.45	+0.13	(0.13)	0.32	+0.09	(0.18)	0.62	−0.06	(0.19)	0.73	+1.00	(0.28)	0.001 * , **
Cluster 2: Red‐Green	+0.41	(0.23)	0.09	−0.05	(0.19)	0.81	−0.10	(0.2)	0.61	−0.24	(0.22)	0.28	+0.75	(0.2)	0.001 ┼
Cluster 3: Yellow‐Orange	−0.05	(0.27)	0.84	+0.15	(0.2)	0.47	−0.05	(0.19)	0.80	−0.19	(0.2)	0.33	+0.83	(0.19)	0.001 ** , ┼
Cluster 4: Orange‐Blue	−0.02	(0.13)	0.91	−0.07	(0.11)	0.52	−0.03	(0.18)	0.86	−0.09	(0.2)	0.66	+0.64	(0.25)	0.01 * , **
All‐cause dementia incidence
Cluster 1: Orange‐Red	+0.02	(0.12)	0.88	−0.09	(0.09)	0.35	−0.14	(0.12)	0.23	−0.28	(0.14)	0.047 ┼	+0.57	(0.17)	0.001 ** , ┼
Cluster 2: Red‐Green	+0.27	(0.16)	0.10	−0.03	(0.15)	0.82	−0.18	(0.12)	0.15	−0.29	(0.14)	0.046 ┼	+0.52	(0.14)	0.001 ┼
Cluster 3: Yellow‐Orange	−0.09	(0.2)	0.66	+0.13	(0.12)	0.28	−0.16	(0.12)	0.20	−0.27	(0.13)	0.046 ┼	+0.65	(0.14)	0.001 ** , ┼
Cluster 4: Orange‐Blue	−0.05	(0.08)	0.55	−0.08	(0.08)	0.28	−0.13	(0.11)	0.23	−0.18	(0.12)	0.17	+0.38	(0.15)	0.01 ** , ┼

Abbreviations: 25(OH)D, 25‐hydroxyvitamin D; AD, Alzheimer's disease; HR, hazard ratio; IB, infection burden; IgG, immunoglobulin G; NHANES III, Third National Health and Nutrition Examination Survey. See full list of abbreviations for periodontal pathogen titers.

^a Models were adjusted for age, sex, race/ethnicity, poverty income ratio, education (years), urban‐rural area of residence, household size, marital status, nutritional factors (HEI, MAR), nutritional biomarkers (25(OH)D, folate, vitamin C, vitamin A, total carotenoids, vitamin E, ferritin, selenium, and normalized calcium), lifestyle (smoking, drug use, alcohol, physical activity), health‐related factors (self‐rated health, co‐morbidity index, allostatic load, weight status), dentate status, and social support variables. Covariates (other than exposures) were imputed and analysis is across five imputations with 10 iterations.

^b 19 Periodontal pathogen exposures (both phases) were Log_e transformed and then standardized into z‐scores. Factor analysis was conducted from which 5 factors were extracted each explaining > 4% of total variance. Factor loading details, after varimax rotation, is presented in Table II.1. Factors were labeled based on up to 3 highest loadings. See Figure S1 and methods section for definition of each cluster.

^c Unweighted N = 1186.

^d Unweighted N = 1569.

^* p < 0.05, marginally significant main effect after correction for multiple testing.

^** p < 0.025, significant main effect after correction for multiple testing.

^┼ p < 0.05, significant interaction.

TABLE 4.

Mean CAL and PPD and their association with AD incidence and all‐cause dementia incidence in multiple Cox proportional hazards model, overall and stratified by IB level (below vs. above median): NHANES III, 1988‐1994, linked to CMS‐Medicare up to 2018. ^a , ^b

	IB− , c			IB +, d
Parameter	Loge(HR)	(SE)	p‐value	Loge(HR)	(SE)	p‐value
AD incidence
Model 1: CAL	+0.214	(0.087)	0.018**	+0.001	(0.094)	1.00
Model 2: PPD	+0.274	(0.229)	0.24	+0.046	(0.156)	0.77
All‐cause dementia incidence
Model 1: CAL	+0.112	(0.084)	0.19	−0.020	(0.060)	0.74
Model 2: PPD	+0.463	(0.163)	0.009 *** , e	−0.052	(0.111)	0.64

Abbreviations: 25(OH)D, 25‐hydroxyvitamin D; AD, Alzheimer's disease; CAL, clinical attachment loss; CMS, Centers for Medicare & Medicaid Services; HR, hazard ratio; IB, infection burden; NHANES III, Third National Health and Nutrition Examination Survey; PPD = pocket probing depth.

^a Models were adjusted for age, sex, race/ethnicity, poverty income ratio, education (years), urban‐rural area of residence, household size, marital status, nutritional factors (HEI, MAR), nutritional biomarkers (25(OH)D, folate, vitamin C, vitamin A, total carotenoids, vitamin E, ferritin, selenium, and normalized calcium), lifestyle (smoking, drug use, alcohol, physical activity), health‐related factors (self‐rated health, co‐morbidity index, allostatic load, weight status), dentate status and social support variables. Covariates (other than exposures) were imputed and analysis is across five imputations with 10 iterations. Mean follow‐up time for all‐cause dementia overall was 233 months and for AD dementia was 226 months.

^b CAL/PPD expressed as is, per 1 mm.

^c Unweighted N = 911.

^d Unweighted N = 1127.

^e p < 0.10 for null hypothesis that IB×CAL or IB×PPD 2‐way interaction = 0 in separate unstratified models, that included main effects of IB and CAL (or PPD).

^* p < 0.05, marginally significant main effect after correction for multiple testing.

^** p < 0.025, significant main effect after correction for multiple testing.

TABLE 5.

Infection burden (above vs. below median), individual infections and their associations with incidence of AD and all‐cause dementia: NHANES III, 1988‐1994; N = 2974. ^a

Parameter	Loge(HR)	(SE)	p‐value
AD incidence
IB: Above vs. below median	+0.04	(0.10)	0.68
Hepatitis A	+0.10	(0.15)	0.50
Hepatitis B	+0.16	(0.28)	0.58
Hepatitis C	+0.78	(0.80)	0.34
Herpes simplex virus 1	−0.15	(0.23)	0.50
Herpes simplex virus 2	+0.16	(0.19)	0.39
Varicella zoster	−0.09	(0.51)	0.86
Rubella	−0.50	(0.27)	0.074
Tetanus	+0.36	(0.06)	0.029
Toxoplasmosis	−0.16	(0.12)	0.19
All‐cause dementia incidence
IB: Above vs. below median	+0.04	(0.10)	0.68
Hepatitis A	+0.10	(0.15)	0.49
Hepatitis B	+0.13	(0.24)	0.59
Hepatitis C	+1.20	(0.40)	0.004
Herpes simplex virus 1	−0.18	(0.16)	0.26
Herpes simplex virus 2	+0.26	(0.11)	0.028
Varicella zoster	+0.13	(0.30)	0.68
Rubella	−0.42	(0.21)	0.054
Tetanus	+0.19	(0.11)	0.096
Toxoplasmosis	−0.08	(0.09)	0.38

Abbreviations: AD, Alzheimer's disease; IB, infection burden.

^a Models were adjusted for age, sex, race/ethnicity, poverty income ratio, education (years), urban‐rural area of residence, household size, marital status, nutritional factors (HEI, MAR), nutritional biomarkers (25(OH)D, folate, vitamin C, vitamin A, total carotenoids, vitamin E, ferritin, selenium, and normalized calcium), lifestyle (smoking, drug use, alcohol, physical activity), health‐related factors (self‐rated health, co‐morbidity index, allostatic load, weight status), dentate status and social support variables. Covariates (other than exposures) were imputed and analysis is across five imputations with 10 iterations.

4. DISCUSSION

This study examines incident all‐cause and AD dementia in relation to the interplay of an IB (and individual pathogens), periodontal pathogens, and periodontitis clinical markers, based on exposure data from the NHANES III (1988‐1994) and follow‐up through December 31, 2018. In adjusted models, hepatitis C and HSV‐2 were strongly associated with all‐cause dementia risk. Given their worldwide and our study's estimated seroprevalence, (https://www.who.int/news‐room/fact‐sheets/detail/herpes‐simplex‐virus; https://iris.who.int/bitstream/handle/10665/277005/WHO‐CDS‐HIV‐18.46‐eng.pdf?ua = 1), exposures to HSV‐2 and HCV at mid‐life could be important modifiable factors for preventing dementia onset later in life. Further, P. gingivalis and S. oralis seropositivities were associated with AD risk at higher IB. Furthermore, the red‐green periodontal pathogen cluster coupled with higher IB count increased all‐cause dementia risk among minority racial groups. PPD was associated with dementia risk at lower IB levels in the overall sample.

Prior research also indicates that herpes, flaviviruses, human immunodeficiency virus (HIV), hepatitis viruses, severe acute respiratory syndrome coronavirus 2 (SARS‐CoV2), Ljungan virus (LV), influenza A virus, and Borna disease virus, may increase AD risk. ⁴⁵ Further, HSV‐2 serologic titers have been inversely associated with whole‐brain cortical thickness, suggesting that HSV‐2 seropositivity may contribute to accelerated brain aging, consistent with our present findings. ⁴⁶ HCV is considered an opportunistic infection among HIV patients. ⁴⁷ Antiviral therapy and subsequent response have been linked to a reduced likelihood of developing dementia in individuals infected with the hepatitis C virus (HCV), as approximately 7.2% of vaccinated patients and 10.2% of unvaccinated patients developed AD during follow‐up for the Tdap/Td vaccine. ⁴⁸ , ⁴⁹

Animal studies of herpesvirus (HCV) have primarily focused on the liver, but liver infections can indirectly contribute to brain pathology. ⁴⁸ In mice with restricted Aβ synthesis, brains still showed uptake of Aβ, leading to AD‐like symptoms as they aged. ⁴⁸ Injection of HCV core protein into mice brains resulted in neurodegeneration and over‐activation of ERK, which causes cell death and loss of neuronal density. ⁵⁰ , ⁵¹ , ⁵² HCV's neurotoxic effects may be additive with other viruses, as evidence suggests the involvement of herpesvirus, particularly HSV‐1, in AD. ⁵³ Much evidence points toward involvement of herpesvirus in AD, mostly HSV‐1, possibly due to its higher prevalence in AD cases compared with HSV‐2. ⁵⁴ Apolipoprotein E4 (ApoE4) lipoprotein dosage in mouse brains affects the latency of cerebral HSV‐1 infections, suggesting an interaction between these risk factors. ⁵⁵ However, young mice infected with HSV‐1 did not accumulate Aβ plaques, and Aβ aggregates in aged mice were not associated with HSV‐1, questioning its interaction with AD. ⁵⁶ The relationship between HSV‐2 and AD remains unknown.

Currently, mechanisms connecting Pd or periodontal infections with cognitive impairment are uncertain. One possible route is vascular transmission of bacterial infections from areas affected by Pd to other organs, potentially impacting cognitive function. ⁵⁷ Brain inflammation caused by inflammatory substances released by Pd such as cytokines, chemokines, and prostaglandins is another potential mechanism by which Pd may contribute to AD. ⁵⁸ There is evidence of a direct association between Pg and the observed outcomes, and there seems to be a combined effect with IB, but this association was only observed in the group who received both Pg and IB. In murine models, lipopolysaccharides and gingipains produced by Pg promote the formation of Aβ, along with the migration and inflammation of microglia, which are two characteristic AD pathologies. Exposing mice to Pg many times led to buildup of Aβ in the hippocampus and neurodegeneration. ⁵⁹ , ⁶⁰ , ⁶¹ , ⁶² Pg is extremely susceptible to inhibitors of proton‐pumping ATPase, such as curcumin and citreoviridin, resulting in the inhibition of bacterial growth. ⁶³ Nevertheless, evidence from randomized controlled trials is required to infer causality. Another periodontal pathogen showing synergism with IB in relation to AD is S. oralis. This periodontal bacterium was recently found to be the most abundant in AD patient saliva based on a recent study. ⁶⁴ We also found that the first factor, which loaded strongly on S. oralis and 14 other periodontal bacteria, was associated with all‐cause dementia more strongly at lower IB. This suggests a different pathway by which this bacterium and group of periodontal bacteria may alter dementia risk including AD and vascular dementia. Synergism was found between Pg and IB in relation to AD and between red‐green cluster and IB in relation to all‐cause dementia among racial minorities. The component most strongly associated with all‐cause dementia, HCV, is more prevalent in lower socioeconomic groups and has been shown to interact synergistically with Pd in relation to poor cognitive outcomes, with evidence of common mechanistic pathways between Pd and HCV. ⁶⁵ , ⁶⁶

Previous studies have linked periodontal infections and clinical markers with dementia risk. Elevated levels of Pg IgG have been linked to difficulties in delayed verbal recall and impaired ubtraction. ¹¹ One study found a correlation between elevated levels of Pg IgG and higher robability of difficulties in delayed verbal recall and impaired subtraction. ¹¹ In another study, A. naeslundii, F. nucleatum, and P. intermedia titers were associated with AD. ¹² , ¹³ PPD has been associated with AD in older US adults, but no link was found with clinical attachment loss (CAL). ¹⁸ In our present study, PPD was associated with all‐cause dementia when IB was lower, suggesting that present Pd can alter dementia risk when the number of persistent viral, bacterial, and parasitic infections is relatively lower. Oral health issues (Pd) are linked to unfavorable cognitive outcomes, such as dementia development. ⁶⁷ A recent meta‐analysis of 39 studies revealed that PPD not only elevates the risk of cognitive impairments but also exhibits a dose‐response relationship with Pd severity. ⁶⁸ This finding was corroborated by a subsequent meta‐analysis. ⁶⁷ A comprehensive study using UK Biobank data revealed a strongly positive correlation between oral health issues and incident dementia. ⁶⁷ This investigation contributes to the increasing amount of data linking Pd and dementia, suggesting some dependence on other persistent infections.

The present study employed a large nationally representative sample with CMS‐linkage with up to 30 years of follow‐up, hence enhancing its robustness. We conducted an evaluation of the occurrence of AD and the development of all types of dementia, in relation to levels of antibodies in the blood against periodontal infections, other persistent infections, and their interactions, while also using data obtained from dental examinations. It is, to our knowledge, the first study to investigate associations of Pd markers with dementia traits across various levels of persistent IB. It is crucial to determine if the primary cause of AD is viral, bacterial, or parasitic or a combination of all four, considering that other persistent infections such as viruses, bacteria, parasites, and fungi have been linked to AD and all‐cause dementia.

It is also possible that in some cases, the infectious pathogen itself might not directly cause neurodegeneration but rather immune activation is responsible for at least some of the injury to the brain. ⁶⁹ There are three possible scenarios for an organism's response to a pathogen: (a) generating an antibody response, which is an indicator of the pathogen, and the pathogen itself increasing or decreasing the risk of dementia; (b) failing to generate an antibody response, which is not particularly virulent or pathogenic; (c) or failing to generate an antibody response, which increases the risk of dementia. These scenarios may reflect early and enduring antibody responses to infection or changes in immune function, such as immunosenescence and infection control. Interpreting findings of an association between seropositivity (IB) and a health outcome has been challenging, ⁷⁰ and potential confounding by socioeconomic status (SES) has also been discussed in studies examining the association of IB with myocardial infarction. ⁷¹

Our present study investigated the occurrence of all‐cause and AD dementia in relation to persistent infections and periodontal pathogens. However, it has limitations, such as the use of partial‐mouth examinations, potential residual confounding bias due to genetic risk factors, ⁷² and the inclusion of only nine infectious pathogens in the IB measure. Other pathogens, such as Chlamydia pneumoniae, human herpes virus 6, and Epstein‐Barr virus, have also been associated with dementia risk. The study also excluded some pathogens due to data availability on the entire NHANES III sample, including H. pylori. Future analyses may uncover other interactive effects with other prevalent pathogens, such as cytomegalovirus, other herpesviruses, and influenza. Additionally, the study only investigated remote exposure, not recent exposure to these pathogens. The study also suggests that tetanus seropositivity is likely due to vaccination, ³³ as older adults who have been vaccinated are more likely to engage in healthcare for other reasons, leading to more diagnoses, including AD.

In conclusion, the study suggests that aging immune cells may exacerbate the effects of periodontal infection and persistent viral, bacterial, and parasite infections on cognitive impairment and neurodegeneration. It highlights associations between hepatitis C and HSV‐2, and the combined effects of IB and specific periodontal bacteria with dementia. In general, our study finding implies that future studies should focus on understanding the common pathogenesis between Pd and dementia, including AD, while incorporating IB with other non‐periodontal pathogens, as a potential effect modifier. More specifically, randomized controlled studies are needed to investigate the impact of eradicating chronic infections and specific periodontal bacteria on neurodegenerative diseases. Early diagnosis and prevention of poor oral health at mid‐life, patient education, as well as screening for various pathogens, may help mitigate their potential effects on neurodegeneration in later life. Additional work is needed to identify age‐specific intervention targets in clinical settings. Effective strategies must be developed to mitigate periodontal pathogens over the life course, ensuring better oral health outcomes at each stage of life.

AUTHOR CONTRIBUTION

May A. Beydoun: Study concept; plan of analysis; data management; statistical analysis; literature search; write‐up of the manuscript; revision of the manuscript. Hind A. Beydoun: Plan of analysis; literature search and review; write‐up of parts of the manuscript; revision of the manuscript. Dawson Hedges: Plan of analysis; literature search and review; write‐up of parts of the manuscript; revision of the manuscript. Lance D. Erickson: Plan of analysis; literature search and review; write‐up of parts of the manuscript; revision of the manuscript. Shawn D. Gale: Plan of analysis; literature search and review; write‐up of parts of the manuscript; revision of the manuscript. Jordan Weiss: Assistance with statistical analysis; write‐up of parts of the manuscript; revision of the manuscript. Ziad El‐Hajj: Literature search and review; write‐up of parts of the manuscript; revision of the manuscript. Michele K. Evans: Study concept; study supervision; write‐up of parts of the manuscript; revision of the manuscript. Alan B. Zonderman: Study concept; plan of analysis; write‐up of parts of the manuscript; revision of the manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the supporting information.

CONSENT STATEMENT

All human subjects provided informed consent in the original NHANES study. The present study was approved by the institutional review board of the National Institutes of Health.

Supporting information

Supporting information

Supporting information

Supporting information

Beydoun MA, Beydoun HA, Hedges DW, et al. Infection burden, periodontal pathogens, and their interactive association with incident all‐cause and Alzheimer's disease dementia in a large national survey. Alzheimer's Dement. 2024;20:6468–6485. 10.1002/alz.14141