Is systemic inflammation a missing link between periodontitis and hypertension? Results from two large population‐based surveys

The primary objective was to investigate the relationship between periodontitis and hypertension in two independent large surveys. The secondary objective was to ascertain whether systemic inflammation had a mediation effect in the association.

Objective. The primary objective was to investigate the relationship between periodontitis and hypertension in two independent large surveys. The secondary objective was to ascertain whether systemic inflammation had a mediation effect in the association.

Introduction
Hypertension is a complex multifactorial disorder. Its prevalence exceeds 31% worldwide with more than 1.13 billion people affected [1]. Elevated blood pressure (BP) is strongly linked to cardiovascular complications, increasing morbidity and mortality [2]. Experimental and observational evidence supports a prominent role of systemic inflammation both in the initiation and in the progression of hypertension [3]. The management of this condition, however, is still a challenge and it represents an increasing burden for society.
Periodontitis is one of the most common inflammatory disorders worldwide with >46% of adults in the United States diagnosed with the disease [4]. Strong evidence supports the role of a dysbiotic dental biofilm in the development of periodontitis. Further, a cluster of modifiable risk factors is shared between periodontitis and leading noncommunicable diseases (NCDs) (cardiovascular diseases, cancer, chronic respiratory diseases and diabetes) [5].
Patients with periodontitis exhibit not only gingival inflammation but also endothelial dysfunction, increased bacterial burden (endotoxins and exotoxins dissemination), metabolic dysregulation and systemic inflammation [6][7][8][9]. A bidirectional link has been proposed between periodontitis and other metabolic disorders such as metabolic syndrome and diabetes [10,11].
Hypertension has been linked to periodontitis but evidence from intervention trials is limited [12]. A possible causal relationship between these two conditions has been proposed recently using Mendelian randomization. This analysis confirmed a link between genetic variants linked to periodontitis and elevated BP phenotypes in a large UK population study [13]. The exact mechanisms mediating this association remain unknown raising the question of whether inflammation or bacterial burden could play a prominent role. Given that systemic inflammatory biomarkers such as C-reactive protein (CRP) and leucocyte counts have been correlated with both periodontitis and hypertension, we hypothesized that systemic inflammation could be a mediator between the two diseases. Therefore, confirmation of this association in large independent studies with a focus on mediators needs to be unravelled prior to undertaking interventional trials investigating the treatment of periodontitis as a target nonpharmacological treatment for hypertension. Accordingly, the primary aim of this study was to investigate the association between periodontitis and hypertension using two representative surveys of the US and Korean populations. The secondary aim was to ascertain the role of systemic inflammation in mediating this association.

Material and methods
Two population-based surveys were analysed and hereby reported in compliance with the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) guidelines (Supplemental checklist).
was conducted in accordance with the 1975 Declaration of Helsinki and participants provided written consent. These survey waves were selected for this study as were the first ones containing a detailed periodontal examination, measurements of average BP, serum concentrations of high-sensitivity CRP (hs-CRP) and white blood cell counts (WBC).
Exclusion criteria used in the final sample analysis were (i) age < 30 years in the NHANES (N = 6451) and age < 19 years in the KNHANES (N = 1345) as no periodontal data were collected for younger individuals; We extracted data on socio-demographic, healthy lifestyle behavioural factors, anthropometric measurements, medical history, oral examination, mean BP and biochemical parameters (Table S1).

Blood pressure measurements
In both cross-sectional studies, sitting BP was measured using a standardized protocol [16]. Average measurements of systolic and diastolic arterial pressure (SBP and DBP) were obtained from three consecutive readings. Participants were then categorized as normal, prehypertensive and hypertensive according to the Joint National Committee 7 guidelines [17]. Further, hypertension was defined as values of SBP ≥ 140 mmHg or DBP ≥ 90 mmHg or the use of antihypertensive medication [18]. The number of participants taking antihypertensive medications was also calculated.

Periodontal examination and dental exposure variables
The analysis was conducted using both established case definitions and continuous measures (fullmouth indices) of periodontitis. However, different protocols were used. In the NHANES, a full-mouth periodontal assessment was carried out at six sites per tooth and periodontitis (exposure) was defined as mild, moderate or severe [19]. Continuous aggregate dental variables (number and percentage of sites) were then created to indicate (a) the extent of periodontal lesions with probing pocket depths (PPD) of ≥4 mm, ≥5 mm, ≥6 mm and (b) the extent of loss of periodontal tissue attachment (AL) of ≥3 mm, ≥4 mm, ≥5 mm, ≥6 mm as previously described [20].
In the KNHANES study, participants presenting with community periodontal index (CPI) scores of 3 and 4 (at least in one sextant) were defined as having worse periodontal status, whereas those presenting scores of 0, 1 and 2 represented controls with better periodontal status. Continuous measures of periodontal lesions were then created as follows: (a) CPI cumulative score (the sum of only CPI scores of 3 or 4 of all sextants) and (b) CPI continuous score (sum of all CPI scores of all sextants) as previously described [21].

Statistical analysis
Data analyses were performed with STATA version 15.0 (StataCorp, College Station, Tex, USA) and R Software (version 3.5.2). Continuous variables are reported as mean AE standard error (SE), whereas categorical variables are expressed as percentages. Simple differences between participants with or without periodontitis were assessed by independent t-test (for continuous variables) or chi-square test (for categorical variables). Normality assumptions were checked, and a logarithmic transformation of hs-CRP was used for parametric analyses. Different measures of oral disease exposure (categorical and continuous) were adopted to test the association between periodontal status and BP. Further, circulating levels of WBC and hs-CRP were used as biomarkers of systemic inflammation and possible mediators of the association between periodontal status and hypertension. Univariate analyses were performed for all continuous variables comparing the groups of participants with periodontitis/worse periodontal status and the rest of the study sample. All those variables with statistically significant associations were then used in the multivariate models.
Multivariate logistic regression models were created to test potential associations between periodontitis case definitions or continuous measures of periodontal lesions with hypertension, SBP ≥ 140 mmHg or hs-CRP > 2mg L À1 as outcome variables. Multivariate linear regression models were then constructed to investigate the association between periodontal (both categorical and continuous) and arterial BP (mean SBP/DBP) variables. Similar analyses were performed with hs-CRP and WBC values (exposures) and hypertension (as categorical or continuous outcomes). Odds ratios (ORs) and 95% confidence intervals (CI) were calculated as well as b coefficient with standard errors. A fully adjusted model (Model 1) included age, gender, ethnicity, smoking, education level and chronic medical conditions as covariates (as previously reported) in both surveys [18]. In addition to these, body mass index (BMI), alcohol consumption, creatinine and physical activity were included in the multivariable models of the KNHANES survey (as they all presented univariate association with the outcome variables).
Structural equation modelling (SEM) was then used to estimate whether the association between periodontitis and hypertension was mediated by WBC or CRP using R Software [22]. Four different and prespecified routes were used as follows: direct (route 1) and indirect (route 2, 3, 4) mediation effects with their 95% CI were estimated: Route 1: Periodontitis (exposure) ? Hypertension (outcome).
When hs-CRP or WBC was introduced as independent exposure variables, the odds of hypertension and SBP ≥ 140 mmHg ranged from 1.0 to 1.4 in the US and from 1.0 to 1.1 in the Korean survey.
Only some continuous measurements of periodontitis (mean PPD and mean CAL) were associated with greater odds of hypertension, SBP ≥ 140 mmHg and of hs-CRP ≥ 2 mg L À1 in the fully adjusted model and in those participants not taking antihypertensive medications (Figure 1).

Linear regression analyses
Linear regression analyses confirmed that periodontitis (assessed both as categorical and as continuous variables) was associated with mean SBP. These findings were confirmed in the subgroup of participants not taking antihypertensive medications in the US survey (Table 3). In the Korean survey, the cumulative CPI score was consistently associated with SBP and DBP and this was also confirmed in participants not taking BP medications. Higher WBC counts were associated with mean SBP in both surveys, whilst higher hs-CRP levels were associated with SBP and DBP only in the US study (Table 3). Further, we observed a negative association between DBP and the number or percentage of gingival sites with attachment loss of ≥3 mm and of sites with probing depth ≥ 6 mm in the NHANES (model 2) and with the cumulative CPI score in the KNANES (models 1 and 2) (Table 3). Lastly, both US and Korean participants with severe periodontitis or worse periodontal status (CPI 3-4) exhibited greater systemic inflammation as assessed by hs-CRP serum levels and by WBC when compared to those without periodontitis or better periodontal status (CPI 0-2) and this difference was independent of other common confounders (Table 4).

Mediation analyses
The association between periodontitis and hypertension (categorical) was mediated by CRP (b AE SE = 0.010 AE 0.003; P < 0.001) in the NHANES dataset, whilst WBC (b AE SE = 0.001 AE 0.001; P = 0.221) was only an indirect mediator of the association (indirect route linked to hs-CRP) (Model A, unadjusted) ( Fig. 2a and Table S2). When repeating the same analysis in the KNHANES database, both hs-CRP (b AE SE = 0.003 AE 0.001; P = 0.015) and WBC (b AE SE = 0.004 AE 0.001; P = 0.004) acted as mediators of the association between worse periodontal status and hypertension ( Fig. 2b and Table S2). Models B, D, C replicated these results when the analyses applied for continuous periodontal (PPD, CAL in NHANES and CPI continuous in KNANES) and BP (SBP) variables, in both adjusted and unadjusted models (Table S2).

Discussion
The analysis of two of the largest population surveys with available dental and general health data demonstrated that both categorical and continuous measures of periodontitis were consistently associated with hypertension and SBP independent of other common cardiovascular risk factors. Participants in the United States with severe periodontitis had higher odds for SBP ≥ 140 mmHg when compared to participants without periodontitis and all findings were confirmed in participants not taking antihypertensive medications. Systemic inflammation defined by two commonly measured biomarkers (hs-CRP and WBC) was not only associated independently with periodontitis, SBP and diagnosis of hypertension but acted as a modest mediator of these associations. This analysis confirmed that participants with periodontitis have a 20-60% greater chance of presenting also a concomitant diagnosis of hypertension and a 10% to 2.5 times greater chance of SBP ≥ 140 mmHg. This is consistent with previous studies reporting that patients with periodontitis have on average 4.5 higher mean SBP (95% CI: 2.88-6.11) than participants without periodontitis [12]. In the present study, a higher mean SBP of 6   compared to those without the disease. The magnitude of this association could have important public health implications if we consider that high sodium intake with the diet is linked to a 6.0mmHg higher average SBP [23]. Further, SBP is a strong independent risk predictor for coronary heart disease events, stroke, heart failure and end-stage renal disease [24,25].
Negligible associations of measures of periodontitis with DBP have been reported [20,26]. Interestingly, we observed a negative linear association between two measures of continuous disease and DBP in both datasets. This observation has not been reported previously, and it seems to be in contrast with the findings related to SBP. At this stage, it is speculative to suggest a biological explanation of these findings. Authors consider important the role of residual confounding from other traditional risk factors variables (i.e. age, gender, ethnicity) as the estimates of association between DBP and some of the continuous measures of periodontitis tended to be greater in multivariate fully adjusted models. DBP is not considered on its own as a strong predictor for CVD events [27,28], and the role of inflammation on affecting this measure of blood pressure is unclear. Further research should be conducted to ascertain the degree of association between diastolic pressure and periodontal inflammation as well as investigate potential biological mechanisms linking them.
When comparing the findings between the two surveys, a stronger association of categorical and continuous variables of periodontitis and gingival inflammation with hypertension and SBP was observed in the US dataset when compared with the Korean data survey. Similar findings were found for the association between biomarkers of inflammation (CRP and leucocytes counts) with measures of arterial blood pressure. Whilst the two sample populations present ethnic and socio-economic differences including a higher proportion of current smokers, adiposity and chronic medical  A recent intervention study assessing the impact of periodontitis treatment on arterial blood pressure confirmed a substantial reduction of SBP after 2 months (mean difference of 11.1 mmHg) [13]. This preliminary evidence suggests that periodontal treatment could represent a novel nonpharmacological intervention for hypertension of similar magnitude of other lifestyles adjustments (weight loss, increasing physical activity, salt or alcohol intake reduction or smoking cessation) with an average reduction of SBP ranging from 4.6 to 6.4 mmHg [29][30][31]. However, larger and longer RCTs are needed.
Several lines of evidence now implicate inflammation in the development and progression of vascular diseases. For the last 3 decades, inflammation has been recognized as a common denominator of early vascular dysfunction, leading onto the development of atheroma and vascular complications [32]. Recent proof-of-concept evidence suggests that targeting upstream inflammation by selective drugs results in reduced morbidity and mortality [33]. This could also be applicable in hypertension. Experimental and human studies have documented several pathways by which elevated inflammatory markers such as CRP and circulating leucocytes are associated with an increased risk of incident hypertension including a derangement of the renin-angiotensin system, increased oxidative stress and downregulation of nitric oxide leading to increased endothelial stiffness and dysfunction [34]. A recent review identified a number of potential sources of extravascular inflammation including periodontitis as a potential factor influencing vascular risk [32]. It is now well documented that patients with periodontitis have elevated levels of CRP and WBC [7,35].
In the mediation analysis, our findings suggest that CRP and WBC mediate partly the association between periodontitis and hypertension, although the effect is rather modest in nature (only 2% of the association explained by the model for the Korean survey whilst up to 7% in the US survey). Similar findings were recently reported for CRP (5.4%), WBC (4.2%) and ferritin (10.2%) as mediators of the total association between a continuous measure of periodontitis and high/uncontrolled BP (≥130/80 mmHg) [36]. An alternative pathway implicated in hypertension and cardiovascular injury relates to the activation of innate and adaptive immune cells such as monocyte/macrophages, and B and T lymphocytes [37]. Damageactivated molecular patterns from the vasculature and pathogen-activated molecular patterns from opportunistic diseases such as periodontitis can exacerbate the inflammatory cascade by activation of Th1 and Th17 lymphocytes, with kidneys and vasculature injuries aggravating a pro-hypertensive status, which results in progressive raised BP [38,39].
Our analyses point perhaps towards a more prominent role of the gut and oral microbiome and their dysbiosis on hypertension [40]. Periodontal pathogens may well play a role influencing the gut microbiome as well as exerting a direct vascular effect. Swallowing Gram-negative oral bacteria or their end-products may trigger metabolic endotoxemia and systemic inflammation contributing to cardio-metabolic disorders [41]. Lastly, experimental studies confirmed that periodontal bacteria can cause lower nitric oxide bioavailability and vascular dysfunction [42] and patients with periodontitis exhibit less nitrate-reducing bacteria [43]. These novel mechanistic hypotheses warrant further investigation.
Cross-sectional designs preclude any inference on a possible temporal and/or causal association between periodontitis and hypertension. In the attempt of mitigating this limitation, we performed the analysis in two large surveys as to identify common patterns of association and minimizing spurious findings. Two different periodontal assessments and case definitions were adopted in each survey which could be considered a limitation but could also show that the association remains significant irrespectively. Whilst in the NHANES, a recognized case definition was used [19], in the Korean survey a simplified clinical index (CPI) was selected, which is known to have risks of overestimation of the extent but underestimation of the prevalence of periodontitis [44]. In the attempt to overcome some of these limitations, we included a panel of measures of periodontal lesions to detect whether simple categorical associations were replicated when using other exposure variables. Another limitation to consider is the effect of antihypertensive medications on gingival inflammation and increased probing depths as well as on the overall association between periodontitis and hypertension [45]. Sensitivity analyses were therefore performed by repeating the models in the group of participants not taking antihypertensive medications. We cannot, however, exclude that our analyses missed some common risk determinants for hypertension (abdominal obesity, salt intake, use of anti-inflammatory drugs, hormone treatments and stress) as well as unmeasured confounders associated with both periodontitis and hypertension (residual confounding). Future mechanistic and clinical studies should investigate further the role of periodontal-driven systemic inflammation and microbial burden as a risk factor for the development and management of hypertension and its complications.

Sources of funding
We would like to acknowledge that contribution of this work was undertaken at UCLH/UCL who

Patient and public involvement
This research was done without patient involvement. Patients were not invited to comment on the study design and were not consulted to develop patient-relevant outcomes or interpret the results. Patients were not invited to contribute to the writing or editing of this document for readability or accuracy. periodontal

Supporting Information
Additional Supporting Information may be found in the online version of this article: Table S1. Variables description for Korean and American databases. Table S2.