Periodontal Status, C-Reactive Protein, NT-proBNP, and Incident Heart Failure: The Atherosclerosis Risk in Communities Study

Abstract

Background:

Periodontal disease (PD), resulting from inflammatory host-response to dysbiotic subgingival microbiota, has been linked to cardiovascular disease, however its relationship to HF and its subtypes (HF with reduced ejection fraction (HFrEF) and HF with preserved ejection fraction (HFpEF)) is unexplored. We hypothesize the presence of PD is associated with increased risk of incident HF, HFpEF and HFrEF.

Methods:

6,707 participants (mean age=63±6) of the Atherosclerosis Risk in Communities Study with full-mouth periodontal examination at visit 4 (v4) (1996–1998) and longitudinal follow-up for any incident HF (v4-2018), or incident HFpEF and HFrEF (2005–2018) were included. Periodontal status was classified: healthy, PD (as per Periodontal Profile Classification (PPC)), or edentulous. Multivariable-adjusted Cox proportional hazards models were used to calculate hazard ratios (HRs), 95% confidence intervals (CIs) for the association between PPC levels and incident HF, HFpEF or HFrEF. Additionally, biomarkers of inflammation (C-reactive protein, (CRP)) and congestion (N-terminal brain natriuretic peptide, (NT-proBNP)) were assessed.

Results:

1,178 incident HF cases occurred (350 HFpEF, 319 HFrEF and 509 HF of unknown type) over a median of 13 years. 59% had PD, while 18% were edentulous. PD was associated with an increased risk for HFpEF [HR (95% CI: 1.35(0.98–1.86)] and significantly increased risk for HFrEF [1.69(1.18–2.43)], as was edentulism: HFpEF [2.00(1.37–2.93)], HFrEF [2.19(1.43–3.36)]. Edentulism was associated with unfavorable change in CRP and NT-proBNP, while PD was associated only with CRP.

Conclusion:

Periodontal status was associated with incident HF, HFpEF and HFrEF, as well as unfavorable changes in CRP and NT-proBNP.

INTRODUCTION

Heart failure (HF) is a growing public health burden, affecting >37.7 million people worldwide(1). In the US, it is estimated that >8 million people will be living with HF by the year 2030, and projected direct medical costs of HF will be doubling in the next several decades to $53 billion(1). Despite development of novel therapeutics, the mortality burden of HF remains high: up to half of HF patients die within 5 years of initial diagnosis(2, 3).

Among patients with clinical HF syndrome, 2 main phenotypes exist, HF with a reduced ejection fraction (HFrEF) and HF with a preserved ejection fraction (HFpEF). The epidemiology of HF has evolved such that overall incidence remains stable, yet the incidence of HFrEF has decreased and that of HFpEF has increased. These trends are largely attributable to the aging population and the rise in risk factors such as diabetes mellitus, obesity, and metabolic syndrome. Although the pharmacologic armamentarium to treat HFrEF has expanded, these therapeutics have neutral effect against HFpEF.

The pathophysiological mechanisms that lead to HF have been the subject of substantial research in the recent years. Chronic elevated levels of systemic inflammation and venous congestion are commonly observed in both HFrEF and HFpEF and are believed to be directly related to disease pathogenesis(4). Perturbations in the gut microbiota and impairment of gut mucosal barriers, which facilitate entry of endotoxins and gut metabolites into the circulation, have been observed in HFrEF(5). Such endotoxins and metabolites increase systemic inflammation, potentially contributing to progression of HFrEF. To date, microbiome studies in HF have focused solely on gut microbiota(5) among the HFrEF population, while the oral bacterial milieu and HFpEF populations have not been investigated in this context.

Anti-infective periodontal therapy among patients with periodontitis (a condition characterized by destruction of tooth-supporting tissues and subgingival microbial dysbiosis(6)) targeting oral microbial dysbiosis has been shown to reduce systemic inflammation(7) and favorably modulates gene expression in circulating monocytes(8). Moreover, a previous meta-analysis of 20 randomized controlled trials found reductions in C-Reactive Protein (CRP) after anti-infective periodontal therapy(9), a conclusion also supported by an AHA Scientific Statement(10). A robust literature exists linking periodontitis to coronary artery disease, ischemic stroke and adverse cardiovascular outcomes in numerous cohorts, including in the Atherosclerosis Risk in Communities (ARIC) Study(10–14). However, despite the potential intersection between periodontitis, inflammation and HF, limited data exists examining the relationship between periodontitis and HF risk in large population-based studies. Additionally, there are no data investigating whether periodontitis is differentially associated with HFrEF vs. HFpEF.

Presently, we investigated the relationship between periodontitis and incident HF, including HFrEF and HFpEF. We additionally studied the association between periodontitis and levels of systemic inflammation (CRP) and congestion (N-terminal brain natriuretic peptide (NT-proBNP)) in a multi-center, community-based cohort of participants enrolled in the ARIC study.

METHODS

In 1987–1989 the ARIC study recruited 15,792 predominantly Black or White participants between the ages of 45 and 64 from four different US communities: Forsyth County in North Carolina, the city of Jackson in Mississippi, the northwestern suburbs of Minneapolis in Minnesota and Washington County in Maryland(15). All participants provided written informed consent and the study was approved by the IRB at each site. All participants who completed the fourth clinic visit (v4) (1996–1998) were eligible for the study (N = 11,656). Of these 3,790 were excluded, due to missing information on periodontal status, HF status, important covariables, or self-reported race other than Black or White as well as Black participants from Minnesota. In addition, 344 participants were excluded for prevalent HF at v4 and 8 participants were excluded for no follow-up after v4 yielding a baseline sample size of n=7,514 (Figure 1) for analyses of 3 periodontal status predicting incident HF. Additional analyses were conducted on three smaller subsets of participants (Figure 1) as follows: i) participants with CRP<10 mg/L measured at both at v4 and visit 5 (v5:2011–2013; n=3,621); ii) participants with NT-proBNP measured at both v4 and v5 (n=3,979); and iii) participants who were living and free of HF in 2005 when ARIC first began adjudication of HF subtype (HFrEF vs HFpEF) (n=6,707).

Figure 1. Flow chart.

Participant flowchart for the three main analyses performed: incidence HFpEF and HFrEF; HF; and longitudinal NT-proBNP and CRP.

Periodontal Assessments

Full-mouth periodontal examinations were conducted by trained examiners calibrated against a standard examiner. Examiners measured probing depth, gingival recession and bleeding on probing at six sites per tooth. Attachment loss was calculated from probing depth and gingival recession. Through these measurements, (attachment loss, probing depth and bleeding on probing) the Periodontal Profile Class (PPC) was created in ARIC as previously validated(16). The PPC categories were then collapsed into three categories for analyses: 1) healthy, 2) any periodontal disease and 3) edentulism.

Prevalent Heart Failure

ARIC defined HF at v4 using the following criteria: 1) reported use of HF medication, 2) the presence of HF according the Gothenburg criteria at ARIC v1, 3) or having developed incident HF prior to v4 based on the presence of ICD-9-CM code 428 in any hospitalization during follow-up(17). Participants with prevalent HF at v4 were removed from the analysis.

Incident Heart Failure

Incident HF up to 2005 was based on the presence of a HF-related ICD-9 discharge code ascertained through ARIC surveillance of cohort hospitalizations(17). Starting in 2005, ARIC implemented committee adjudication of HF hospitalizations based on chart abstraction previously described(18). Abstraction included results of imaging studies and LVEF when available and reviewers determined if evidence of an LVEF<50% at the time of hospitalization was present and recorded a numerical LVEF if available. Participants with quantifiable left ventricle ejection fraction (LVEF) were categorized as HFpEF (LVEF ≥50%) or HFrEF (LVEF <50%). If adjudication of the HF type was not possible, participants were defined as ‘unknown’ HF.

Risk Factor Measurements

Confounding variables were measured at the time of the periodontal exam (v4) via questionnaires, clinical exam and laboratory analysis of blood samples except for birth date, sex, race (Black or White) and education level (1:< high school degree, 2:high school, GED, vocational school and 3:college, graduate, or professional school) which were collected at v1. In addition, physical activity was assessed at v3 (1993–95) via a modified Baecke questionnaire. A physical activity index score (1=lowest and 5=highest) was calculated based on intensity and time. At v4, smoking status (never, former, current), income (<$25,000/year, $25,000–<$50,000/year, $50,000–<$75,000/year, ≥$75,000/year), oral hygiene and access to healthcare variables were collected including insurance status (private, Medicare/Medicaid only, none).

Participants fasted for 8 hours before the v4 clinical examination and blood was collected for the assessment of lipids (including HDL and LDL cholesterol [estimated via the Friedewald equation]). NT-proBNP was measured via stored plasma samples using an electrochemiluminescent immunoassay and CRP was measured via immunophelometric assay. Troponin was measured using EDTA plasma samples from ARIC visit 4 (1996–1998) stored at – 80°C, using a chemiluminescent immunoassay. Diabetes mellitus was defined as self-reported physician diagnosis, fasting glucose ≥126 mg/dL, ≥200 mg/dL if non-fasting, or reported pharmacological treatment. In addition, body mass index (BMI) was collected and defined as measured weight in kilograms divided by height in meters squared, and blood pressure was measured twice after a five-minute rest and the average was used in analyses. Use of anti-hypertensive medication was self-reported.

Prevalent coronary heart disease (CHD) at v4 was assessed based on prior cardiovascular revascularization, self-reported physician-diagnosed myocardial infarction prior to v1, presence of a previous myocardial infarction by ECG at v1 or incident CHD following v1. Incident CHD occurring after baseline but prior to HF was also included. CHD is a high-quality annual surveillance variable for incident CHD, including myocardial infarction, which has been validated and incorporates hospitalizations and deaths. All CHD outcomes were adjudicated by the ARIC Morbidity and Mortality Classification Committee(17).

Statistical Analysis

All analyses were conducted using SAS version 9.4. Percents or means (±SE) were presented across participant characteristics at v4 on three categories of baseline periodontal status. P-values for difference among periodontal categories were obtained via chi-square for categorical variables or F-tests from ANOVA for continuous variables.

Analyses for the association of periodontal status and the outcome of any type of incident HF considered follow-up time from v4 to the first incident HF diagnosis, lost to follow-up, death, or administrated administratively as of December 31, 2018 (or December 31, 2017 among participants from the Jackson site), whichever occurred first. We calculated incidence rates across the three periodontal disease categories. Hazard ratios and 95% confidence intervals (CI) were calculated using multivariable Cox proportional hazards regression for six models constructed as follows: Model 1: demographics, Model 2: Model 1 + health behaviors, heart failure risk factors and comorbidities, Model 3: Model 2 + NT-proBNP at v4, Model 4: Model 2 + CRP at v4, Model 5: Model 2 + troponin at v4, Model 6: Model 2 + incident CHD.

For our subgroup analyses of the association between periodontal status and HF subtypes, follow-up time began in 2005, when HF adjudication allowed for the distinction between HFpEF and HFrEF, and continued until a HF event developed, lost to follow-up, death, or December 31, 2018 (December 31, 2017 for Jackson site), whichever occurred first. Incidence rates across the three periodontal disease categories and hazard ratios along with 95%CI were calculated. The same covariates were adjusted for in the Cox models listed above. The proportional hazards assumption was tested for all analyses via a time*periodontal status interaction term, and the assumption was met for each.

We additionally performed subgroup analyses to explore the relationship between periodontal status and longitudinal change in CRP and NT-proBNP levels. NT-proBNP was log transformed due to non-normality for regression analyses. Analyses for both outcomes were performed using generalized multivariable linear regression models regressing visit 5 CRP (or change between visit 4 and 5) or LN transformed NT-proBNP (in separate analyses) on v4 periodontal status. Adjusted means (±SE) were presented across periodontal categories and p-values from 2 d.f. type 3 F-tests are presented. Differential survival and/or participation at visit 5 might introduced selection bias, therefore, inverse probability of attrition weighting (IPAW) was utilized to adjust for attrition.

Data Availability

ARIC data are available through NIH NHLBI-sponsored Biologic Specimen and Data Repository Information Coordinating Center at https://biolincc.nhlbi.nih.gov/

RESULTS

Among 7,514 participants, 55% were female, 20% were Black and the mean age±SD at baseline (v4) was 63±6 years (range=53–75). The distribution of periodontal disease based on PPC was as follows: 23% healthy, 59% had periodontal disease, 18% were edentulous. Table 1 shows general characteristics across levels of PPC and demonstrates that participants with worse periodontal status and the edentulous generally have a more adverse cardiovascular risk profile.

Table 1.

Baseline Characteristics of Participants by Baseline PPC Classification, ARIC, 1996–1998 (N = 7514).

	Healthy	Periodontal Disease	Edentulism	P-value
N	1746	4420	1348
Demographics
Age (years)	62 (5)	63 (6)	64 (6)	<.0001
Male	34%	51%	43%	<.0001
Race-field center				<.0001
White-Minneapolis (n=2290)	46%	28%	12%
White-Washington (n=2236)	15%	29%	37%
White Forsyth (n=1863)	36%	20%	20%
Black-Forsyth (n=175)	1%	2%	3%
Black-Jackson (n=1591)	2%	21%	28%
Education 2				<.0001
Basic	5%	16%	43%
Intermediate	42%	44%	41%
Advanced	53%	41%	16%
Behaviors
Cigarette smoking status				<.0001
Current	7%	14%	25%
Former	41%	45%	42%
Never	52%	41%	33%
Physical activity 1	2.7 (0.8)	2.6 (0.8)	2.4 (0.7)	<.0001
Dental
Tooth count	26 (3)	20 (8)	0 (0)	<.0001
Insurance				<.0001
Other	32%	40%	54%
Medicare/Medicaid	66%	54%	37%
None	2%	6%	9%
Physiologic
Body mass index (kg/m2)	27 (5)	29 (5)	30 (6)	<.0001
Systolic blood pressure (mmHg)	123 (17)	127 (18)	130 (20)	<.0001
Diastolic blood pressure (mmHg)	70 (9)	71 (10)	70 (11)	<.0001
Anti-Hypertensive medication	29%	38%	51%	<.0001
Diabetes mellitus	8%	14%	25%	<.0001
HDL cholesterol (mg/dL)	54 (17)	49 (16)	48 (15)	<.0001
LDL cholesterol (mg/dL)	121 (32)	123 (33)	125 (34)	0.02
Prevalent coronary heart disease at visit 4	4%	6%	10%	<.0001
Prevalent stroke at visit 4	1%	2%	3%	<.0001
C-Reactive Protein Visit 4 median (25%, 75%)	1.6 (0.8, 3.5)	1.9 (0.9, 4.0)	2.3 (1.1, 4.5)	0.0002
C-Reactive Protein Visit 5 median (25%, 75%)	1.5 (0.8, 2.8)	1.8 (0.9, 3.3)	2.2 (1.1, 4.1)	<.0001
NT-proBNP (pg/mL) at Visit 4 median (25%, 75%)	59.9 (31.6, 107.3)	51.5 (24.8, 99.6)	61.8 (31.5, 106.9)	0.002
NT-proBNP at Visit 5 (pg/mL) median (25%, 75%)	127.2 (67.0, 241.0)	130.3 (66.0, 255.7)	163.0 (86.6, 320.5)	<.0001
Troponin Visit 4 (ng/g)	1.53 (0.52)	1.68 (0.60)	1.77 (0.70)	<.0001

Data are presented as % or mean (SD)

P-values for difference amongst categories were obtained from chi-square or analysis of variance tests.

¹Physical activity data was collected at visit 3

²Education variable: Basic = less than a high school degree, Intermediate = high school, GED, or vocational school and Advanced = college, graduate, or professional school

Periodontal Status and Incident Heart Failure

The cumulative incidence of HF between v4 and 2018 was 21% (1,558 incident cases) and the incidence density was 112.3 events per 10,000-person years (95%CI: 106.8–118.0). Baseline periodontal status was associated with any incident HF after multivariable adjustment (Supplement Table 1). HRs(95%CIs) for incident events among participants who had periodontal disease or who were edentulous (vs. healthy) in model 2 were 1.14(0.98–1.33) and 1.58(1.32–1.89), respectively (p for any difference <0.0001).

Additional analyses considered the cumulative incidence of the HF subtypes (HFpEF and HFrEF) among n=6,707 participants for whom HF subtype adjudication was available. Among this sample, the cumulative incidence of HFrEF or HFpEF between 2005 and 2018 was 4.8% and 5.2% while the cumulative incidence of any HF was 18% (1,178 incident cases). The incidence densities of HFrEF or HFpEF were 43.3/10,000-person years (95%CI:38.7–48.4) and 47.5/10,000-person years (95%CI:42.7–52.8). After multivariable adjustment, periodontal disease and edentulism were both associated with HF outcomes (Table 2). Hazard ratios for the combined incident HFpEF or HFrEF among participants who had periodontal disease or who were edentulous (vs. healthy) were 1.50(1.18–1.91) and 2.08(1.57–2.77), respectively (p for any difference=0.001). Results were modestly stronger for HFrEF than for HFpEF (Table 2). Adjustment for levels of NT-proBNP, CRP, troponin or incident CHD did not meaningfully attenuate these findings (Table 2, Models 3, 4, 5, 6). Hazard ratios for incident HF of unknown type among participants who had periodontal disease or who were edentulous (vs. healthy) were 0.86(0.67–1.10) and 1.02(0.75–1.39), respectively (p for any difference=0.22).

Table 2.

Multivariable Adjusted Hazard Ratios (95% confidence interval) for Incident Heart Failure With Reduced or Preserved Ejection Fraction From 2005–2018 by Periodontal Category Measured in 1996–98, ARIC (N = 6707).

	Healthy	Periodontal Disease	Edentulism	P-value
	N = 1646	N = 3994	N = 1067
Incident Heart Failure with Preserved Ejection Fraction
# Events	52	204	94
IR (95% CI) *	2.6 (2.06–3.32)	4.67 (4.14–5.24)	9.31 (7.79–11.05)
Model 1	Ref	1.66 (1.21–2.28)	2.83 (1.95–4.11)	<.0001
Model 2	Ref	1.35 (0.98–1.86)	2.00 (1.37–2.93)	0.001
Model 3	Ref	1.33 (0.96–1.84)	1.87 (1.27–2.76)	0.004
Model 4	Ref	1.36 (0.98–1.88)	1.99 (1.36–2.93)	0.001
Model 5	Ref	1.34 (0.96–185)	2.01 (1.37–2.96)	0.001
Model 6	Ref	1.39(0.99,1.95)	2.11(1.41, 3.14)	0.001
Incident Heart Failure with Reduced Ejection Fraction
# Events	38	202	79
IR (95% CI) *	1.92 (1.44–2.52)	4.62 (4.10–5.19)	7.82 (6.43–9.43)
Model 1	Ref	1.96 (1.37–2.80)	2.97 (1.95–4.53)	<.0001
Model 2	Ref	1.69 (1.18–2.43)	2.19 (1.43–3.36)	0.001
Model 3	Ref	1.74 (1.20–2.54)	2.04 (1.31–3.17)	0.005
Model 4	Ref	1.82 (1.25–2.65)	2.31 (1.49–3.59)	0.001
Model 5	Ref	1.82 (1.25–2.65)	2.27 (1.46–3.53)	0.001
Model 6	Ref	1.74(1.17, 2.59)	2.36(1.48, 3.75)	0.001
Incident Heart Failure with Preserved or Reduced Ejection Fraction Combined
# Events	90	406	173
IR (95% CI) *	4.56, (3.80–5.43)	9.29 (8.54–10.08)	17.13 (15.05–19.43)
Model 1	Ref	1.79 (1.41–2.27)	2.89 (2.18–3.82)	<.0001
Model 2	Ref	1.50 (1.18–1.91)	2.08 (1.57–2.77)	<.0001
Model 3	Ref	1.51 (1.18–1.93)	1.93 (1.44–2.59)	<.0001
Model 4	Ref	1.55 (1.22–1.98)	2.12 (1.59–2.84)	<.0001
Model 5	Ref	1.54 (1.20–1.97)	2.11 (1.58–2.82)	<.0001
Model 6	Ref	1.53(1.19, 1.98)	2.19(1.62, 2.97)	<.0001

^*Unadjusted IR (incidence rate) is per 100,000 person years.

Model 1: adjusts for baseline age, gender, race/center, education, insurance.

Model 2: adjusts for model 1 + cigarette status, physical activity, BMI, LDL, hypertension medication, CHD, diabetes, SBP.

Model 3: adjusts for model 2 + baseline NT-proBNP.

Model 4: adjusts for model 2 + baseline CRP.

Model 5: adjusts for model 2 + baseline troponin.

Model 6: adjusts for model 2 + incident CHD.

After full multivariable adjustment, HRs(95%CIs) for periodontal disease and edentulism predicting incident CHD were: 1.16(0.95–1.42) and 1.42(1.11–1.82). Incident CHD was a strong predictor of incident HFrEF/HFpEF combined: HR(95%CI)=3.14(2.53–3.90) and findings were also statistically significant for both HFpEF [HR(95%CI)=2.14(1.52–3.02)] and HFrEF [HR(95%CI)=4.34(3.26–5.78)] modeled separately.

Periodontal Status, CRP and NT-proBNP

CRP (mean±SD) levels at v4 and v5 were 2.7±2.4 and 2.4±2.1 mg/L, respectively. At v4 and v5, the median levels of NT-proBNP were 81.0±123.8 pg/mL and 291.2±885.8 pg/mL, respectively. In longitudinal analyses, mean CRP and NT-proBNP levels were elevated among participants with periodontal disease and edentulous compared to periodontally health participants; these results remained after multivariable adjustment (Table 3, Figure 2).

Table 3:

Associations Between Baseline Periodontal Status (1996–1998) and Mean Values of Change in CRP (n=3,621) or NT-proBNP (n=3,979) Among Participants in the ARIC Study.

	Healthy	Periodontal disease1	Edentulism	P-Value*
	N = 1034	N = 2183	N = 404
Mean CRP Change (visit 5 – visit 4)
Model 1	−0.40±0.08	−0.38±0.05	−0.23±0.12	0.47
Model 2	−0.49±0.08	−0.35±0.05	−0.07±0.11**	0.02
Model 3	−0.51±0.07	−0.30±0.04**	−0.03±0.09**	0.0003
Mean LN NT-proBNP change (visit 5 – visit 4)
Model 1	1.02±0.03	1.06±0.02	1.11±0.04	0.26
Model 2	1.05±0.03	1.06±0.02	1.07±0.04	0.94
Model 3	1.04±0.03	1.06±0.02	1.15±0.04**	0.06

Results presented as adjusted mean values of change in CRP±SE mg/L or LN transformed NT-proBNP pg/mL.

Inverse probability weights were included for all models.

Model 1: adjusts for baseline age, gender, race/center, education, insurance.

Model 2: adjusts for model 1 + cigarette status, physical activity, BMI, LDL, hypertension medication, CHD, diabetes, SBP.

Model 3: adjusts for model 2 + baseline CRP in analyses of CRP change and baseline NT-proBNP in models of NT-proBNP change.

^*p-values are 2 d.f. type 3 F-tests

^**p-value <0.05 for comparison with healthy participants.

Figure 2. Association Between Baseline Periodontal Status (1996–1998) and Follow-up CRP (2011–2013) Among N=3,621 Participants in the ARIC Study.

Results presented as adjusted mean values of CRP±SE.

Inverse probability weights were included for all models.

Model 1: adjusts for baseline age, gender, race/center, education, insurance.

Model 2: adjusts for model 1 + cigarette status, physical activity, BMI, LDL, hypertension medication, CHD, diabetes, SBP.

Model 3: adjusts for model 2 + baseline CRP

p-values are 2 d.f. type 3 F-tests were significant for all models (p-value = <.0001, 0.0003, 0.0003 respectively.)

*p-value <0.05 for comparison with healthy participants.

DISCUSSION

We found baseline periodontal status (i.e., periodontal disease and edentulism) to be associated with incident HF during 13 years of follow-up. The observed associations were stronger among adjudicated subtypes of HFpEF and HFrEF, while analyses for unknown (unadjudicated) HF type were null. Results for overall HF and HFpEF/HFrEF specifically remained significant after extensive adjustment for socio-demographic, behavioral and HF risk biomarkers along with other pre-existing health conditions and incident CHD. We observed that periodontal status was associated with elevated levels of CRP and NT-proBNP longitudinally.

While numerous studies have reported associations between periodontal status and coronary artery disease(19)^,(20), stroke(14) or diabetes(11), few have examined HF. To our knowledge, this is the first large population-based study to examine the relationship between periodontal status and incident HF, and the first study to consider HFrEF and HFpEF separately.

One potential mechanistic explanation for this association relates to the chronic pro-inflammatory challenge present in the dysbiotic subgingival microbial communities that characterize and drive periodontal disease. Prior studies(21, 22) have reported periodontal status to be associated with systemic inflammation, and randomized trials have demonstrated a reduction in CRP(9) following anti-infective periodontal therapy. Our findings support prior publications and demonstrate longitudinal associations between baseline periodontal status and CRP levels assessed over a decade later.

Inflammatory biomarkers, including interleukin-6 (IL-6), tumor necrosis factor-α (TNF-α), and CRP, are known to be elevated in HFrEF and HFpEF patients and are predictors of poor clinical outcomes(23–25). Data from MESA show IL-6 and CRP to be strongly and independently associated with all incident HF(26, 27). Framingham data indicate that 5-year risk of developing HF increases by 60% and 68% per tertile increment of TNF-α and IL-6 levels, respectively(28).

Inflammation represents a physiologic response intended to provide protection and promote healing in the setting of injury. Paradoxically, this process can promote tissue damage if unchecked(29). Inflammatory cytokines exert direct effects on myocardial and vascular cells that predispose individuals to HF(30, 31). TNF-α promotes cardiac apoptosis, hypertrophy and fibrosis, and alters calcium handling in the myocytes, leading to a direct negative inotropic (systolic)(32) and lusinotropic (diastolic) effect(31). IL-6 promotes myocyte hypertrophy and increases myocardial stiffness by reducing the phosphorylation of titin(33). The identification of upstream triggers of chronic systemic inflammation is a promising avenue of research for reducing HF incidence generally, and HFpEF more specifically given its limited therapeutic options. Given the aforementioned knowledge linking inflammation to both periodontal disease and HF, coupled with the demonstrated ability to reduce inflammation by intervening on oral dysbiosis, the potential for anti-infective periodontal therapy for the reduction in risk for HF development merits further consideration.

Our findings linking periodontal status to levels of NT-proBNP are novel and bolster the hypothesis that chronic subgingival dysbiosis might contribute to cardiac abnormalities prior to overt HF development. A recent randomized control trial among patients with type 2 diabetes found that periodontal treatment significantly reduced the ratio of early mitral inflow to tissue velocity of the mitral annulus (E/e’ ratio), an echocardiographic indicator of pre-clinical diastolic dysfunction associated with development of HFpEF(34), and observed a non-significant decreased trend of NT-proBNP levels(35).

Additionally, our results found a relationship between periodontal status and incident HFpEF and HFrEF. While this may be surprising since HF is a phenotypically heterogeneous syndrome(36–38), it fits with the overarching hypothesis that most phenotypic HF subtypes of HFrEF or HFpEF – whether they be characterized predominantly by functional, morphological or biomarker abnormalities(39) – stem from a common, chronic and central inflammatory milieu ultimately leading to HF. We hypothesize that subgingival microbial dysbiosis contributes to this upstream chronic inflammatory milieu. In that sense, a good portion of HF subtypes (likely with few exceptions, e.g. infiltrative cardiomyopathies) are possible consequences. Moreover, there is evidence that the microbiome might contribute to relevant pathways for HFrEF and HFpEF risk such as hypertension(40, 41) and diabetes(11, 13). These results may inform HF prevention and treatment strategies, particularly for HFpEF, which currently lacks effective therapies(42).

Several study limitations are present. Changes in periodontal status could not be considered, as periodontal examinations were conducted only at ARIC v4. In addition, the only time-varying confounder adjustment was done for CHD, thus time dependency may have influenced the evolution of HF. Differential follow-up time related to baseline participant health could cause selection bias (e.g., survivor bias), although our longitudinal analyses assessing biomarkers utilized IPAW weighting, with survival analyses performed to minimize these concerns. Further, asymptomatic HF may have been present at baseline and could have been more common among individuals with periodontal disease or the edentulous, giving the spurious appearance of increased incidence in these groups. However, our analyses for incident HFpEF/HFrEF starting in 2005 removed incident HFpEF/HFrEF events occurring in the ~5–7 years between the baseline periodontal evaluation and initial follow-up. It is important to note that several participants were excluded mainly due to missing periodontal status and having prevalent HF at baseline. These exclusions could have introduced selection bias that would have likely attenuated our results towards the null. In order for selection bias to push our HRs away from the null, very strong selection bias that is synergistic between periodontal and incident HF status would have to be present. This could only occur if people who do not develop HF are more likely lost to follow up among periodontally healthy than those with periodontitis.

This study also had many strengths including a large cohort of participants followed for up to 22 years (1996–2018) and the ability to specify HF subtypes in a subgroup of participants. Robust data collection enabled rigorous control for confounding, though residual confounding may remain.

We have observed baseline periodontal status to be related to levels of systemic CRP, NT-proBNP, troponin, and incident HF among a diverse, population-based sample of adults. Our findings support nascent literature linking oral infection to HF risk and support the need for more research exploring the potential for anti-infective periodontal therapies as a preventative strategy for minimizing HF burden. If future studies provide evidence of causal association, the population-level implications could be notable given the high prevalence of treatable periodontal disease and the burden of HF on our aging society.

Figure 3. Association Between Baseline Periodontal Status (1996–1998) and Follow-Up Mean NT-proBNP (2011–2013) Among N=3,979 Participants in the ARIC Study.

Values are presented as adjusted geometric mean transformed NT-proBNP and 95% confidence intervals. Inverse probability weights were included for all models.

Model 1: adjusts for age, gender, race/center, education, insurance.

Model 2: adjusts for model 1 + cigarette status, physical activity, BMI, LDL, hypertension medication, CHD, diabetes, SBP.

Model 3: adjusts for model 2 + baseline NT-proBNP.

p-values are 2 d.f. type 3 F-tests were significant for all models 1 and 2 (p-value = <0.0001, 0.001, 0.06 respectively.)

*p-value <0.05 for comparison with healthy participants.

Central Illustration: Periodontal status and incident heart failure among HFpEF and HFrEF participants in ARIC.

Cox regression of time to first HFpEF or HFrEF according to periodontal status. Follow-up time began in 2005, when HF adjudication allowed for the distinction between HFpEF and HFrEF.

Competency in Medical Knowledge:

Oral health may often be overlooked by physicians as a potential precursor to systemic disease. However, our research illustrates the intersection between inflammation, chronic periodontal infections/periodontitis, and heart failure. Since periodontal status was strongly associated with HF, HFpEF, and HFrEF, cardiologists should consider asking about oral health among HF patients.

Translational outlook:

This is the first longitudinal study to assess the relationship between baseline periodontal status and incident HFpEF or HFrEF. Critically, our results show a strong relationship between periodontal status and incident HF. These findings support the need for future intervention studies that directly test whether anti-infective periodontal prevention and treatment strategies reduce inflammation and risk of incident heart failure.

Abbreviations:

HF

Heart failure

PD

Periodontal disease

PPC

Periodontal Profile Classification

LVEF

Left ventricle ejection fraction

HFrEF

Heart failure with reserved ejecting fraction

HFpEF

Heart failure with preserved ejecting fraction

NT-proBNP

N-terminal pro–B-type natriuretic peptide

CRP

C-reactive protein

HRs

Hazard ratios

ARIC

The Atherosclerosis Risk in Communities Study