Periodontitis is consistently associated with atherosclerotic cardiovascular disease (ASCVD) across large prospective cohorts, with pooled relative risks converging on 1.14 to 1.26 for coronary heart disease, myocardial infarction, and stroke across independent meta-analyses. The association survives multivariable adjustment, shows dose-response with disease severity and tooth loss, and is supported by coherent mechanisms: transient bacteremia, systemic inflammatory signalling, endothelial dysfunction, and immune cross-reactivity. Randomised trials demonstrate that periodontal therapy reduces C-reactive protein and, in two independent trials at 12 and 24 months, slows carotid intima-media thickening by effects of the same order, approximately 0.02\u20130.03 mm.<\/p>\n
None of this establishes causation. Mendelian randomisation using genetic instruments for periodontitis finds no effect on coronary artery disease, stroke, or subclinical atherosclerosis. The observational literature carries high heterogeneity, predominantly critical risk of bias, and demonstrable publication bias. The detection of periodontal pathogen DNA in atheroma is contradicted by well-conducted negative studies. One randomised trial with cardiovascular endpoints \u2014 PREMIERS, in 280 post-stroke patients \u2014 produced a hazard ratio of 0.65 (95% CI 0.30\u20131.38): directionally favourable, statistically inconclusive, and prespecified as non-superior.<\/p>\n
This review separates what is known from what is assumed, places periodontal inflammation within the apoB-centred model of atherogenesis, and assigns explicit confidence scores to the two propositions that matter.<\/p>\n
Precision here is not pedantry. Most of the confusion in the popular literature comes from treating \u201cgum disease\u201d as a single entity.<\/p>\n
Dental plaque<\/strong> is a structured microbial biofilm on the tooth surface. It is not food debris and it is not removed by rinsing. Left undisturbed, it matures and shifts toward a dysbiotic, anaerobe-rich community.<\/p>\n Calculus (tartar)<\/strong> is plaque mineralised by salivary calcium and phosphate. It cannot be removed by brushing and provides a retentive surface for further biofilm accumulation. Calculus is a plaque-retentive factor, not itself the microbial cause of gingival inflammation \u2014 a distinction worth preserving.<\/p>\n Gingivitis<\/strong> is reversible inflammation confined to the marginal gingiva, initiated by dental biofilm. Clinical signs are erythema, oedema, and bleeding on probing. No connective tissue attachment has been lost and no bone has been resorbed. In the current classification framework, bleeding on probing at \u226510% of sites defines gingivitis on an intact periodontium, with 10\u201330% localised and >30% generalised [60]. Gingivitis resolves with adequate biofilm control and is highly prevalent. It does not create new periodontitis-related attachment or bone loss, and it can occur either on an intact periodontium or on one previously reduced but stable after treatment.<\/p>\n The causal relationship between biofilm and gingivitis was established experimentally rather than inferred. L\u00f6e, Theilade and Jensen withdrew oral hygiene in healthy volunteers, observed plaque accumulation and gingival inflammation develop, then reinstituted hygiene and observed resolution [33]. This within-subject withdrawal-and-reinstitution design is as close to an experimental demonstration of causation as oral disease research offers, and it is the reason biofilm is described as a cause rather than an association. Note precisely what it establishes: plaque causes gingivitis<\/em>. It does not by itself establish that plaque causes periodontitis, which requires host susceptibility and time in addition to biofilm.<\/p>\n Periodontitis<\/strong> is a destructive inflammatory disease that can develop in susceptible individuals. The inflammatory infiltrate extends apically, the junctional epithelium migrates, and connective tissue attachment and alveolar bone are lost. The 2018 classification defines a case by interdental clinical attachment loss at two or more non-adjacent teeth, or alternatively by buccal\/oral attachment loss of \u22653 mm with pocketing >3 mm at two or more teeth, in each case after excluding non-periodontal causes such as recession, caries, and root fracture. It stages periodontitis by severity and complexity (I\u2013IV) and grades it by rate of progression and risk factors (A\u2013C), with smoking and glycaemic control explicitly incorporated as grade modifiers [59].<\/p>\n Periodontal pockets<\/strong> are the pathologically deepened sulci created by attachment loss. They are the anatomical feature of greatest systemic interest: inflamed, frequently ulcerated pocket epithelium overlies a richly vascularised connective tissue bed, an interface across which bacteria and bacterial products can enter the circulation.<\/p>\n Clinical attachment loss (CAL)<\/strong> is the reference measure of cumulative destruction. Alveolar bone loss<\/strong> is its radiographic correlate.<\/p>\n Bleeding on probing (BOP)<\/strong> indicates active inflammation. It is the standard clinical sign of gingival inflammation, and repeated absence of bleeding at a site is generally reassuring. An important exception: smoking suppresses gingival bleeding through vasoconstriction and altered immune response, so absence of visible bleeding in a smoker does not indicate periodontal health.<\/p>\n Tooth loss<\/strong> is a downstream marker of advanced disease, but a contaminated one. Teeth are lost to caries, trauma, and access-driven extraction decisions as well as to periodontitis. Cohorts that use tooth count as a periodontitis proxy are measuring something broader than periodontal destruction, and this weakens a substantial fraction of the epidemiological literature.<\/p>\n Why periodontitis dominates the cardiovascular literature.<\/strong> Three reasons. Gingivitis is highly prevalent, which limits its discriminative value in a cohort study. Gingivitis produces a small ulcerated surface area, so the plausible systemic inflammatory load is low. And gingivitis is transient and fluctuating, so it does not represent the sustained decades-long exposure that atherogenesis would require. Not everyone with gingivitis progresses; progression requires the interaction of dysbiotic biofilm, host susceptibility, dysregulated immune response, environmental exposure, and time.<\/p>\n Distinguishing causes from associations is essential, because the confounding structure of the cardiovascular question depends entirely on it.<\/p>\n Smoking, diabetes, obesity, age, socioeconomic position, and healthcare access each cause or strongly predict both<\/em> periodontitis and ASCVD. This is not incidental overlap; it is the dominant feature of the data. Any observed periodontitis\u2013ASCVD association of the magnitude actually reported must be interpreted against a confounding structure fully capable of generating it.<\/p>\n Three independent meta-analyses using different study sets and methods converge on a narrow band of effect:<\/p>\n The convergence across independent analyses is genuine evidence against a purely random association. It is not evidence against shared systematic bias: these syntheses draw on overlapping primary studies, and a bias common to that literature would propagate into all of them alike.<\/p>\n Three qualifications that are routinely omitted and should not be.<\/strong> First, Larvin\u2019s heterogeneity was I\u00b2 = 97.3% \u2014 the pooled point estimate summarises studies that do not agree with one another. Second, of 32 included studies, 21 were rated at critical<\/strong> risk of bias by ROBINS-I and the remaining 11 at serious risk; not one was rated low or moderate. Third, formal testing found significant publication bias<\/strong> (Egger\u2019s \u03b2 = 2.91, P = .004). A literature with critical risk of bias throughout and demonstrable publication bias, producing a pooled effect of 1.20, is not a literature from which causation can be read off.<\/p>\n Fourth, and separately: Larvin found no difference in CVD risk between clinically diagnosed and self-reported periodontal disease (RR 0.97, 0.87\u20131.07). The finding is that diagnosis method did not detectably modify the association. One reading is reassuring for the self-report cohorts. Another is less so: if a crude self-reported exposure yields the same estimate as a calibrated clinical examination, the association may be tracking something broader than periodontal destruction \u2014 general health literacy, healthcare engagement, or socioeconomic position. The meta-regression cannot distinguish these readings, and neither should be presented as established.<\/p>\n A 2024 umbrella review of systematic reviews concluded that the association is consistently observed but that methodological heterogeneity across primary studies precludes any causal determination [6]. A 2026 meta-analysis of 30 cohort studies reported HR 1.31 (95% CI 1.13\u20131.48) [58]; its main pooled exposure deliberately spans a broader periodontal and oral-health category rather than exclusively clinically diagnosed periodontitis, and heterogeneity was substantial, so it should be read alongside Guo and Larvin rather than pooled with them.<\/p>\n PAROKRANK<\/strong> is a carefully phenotyped case-control study with prospective follow-up [7]. It recruited 805 patients under 75 with a first myocardial infarction and 805 controls matched for age, sex, and geographic area, all examined with standardised dental assessment including panoramic radiography, with approximately 100 candidate confounders collected. Periodontitis was independently associated with first MI. The pre-planned long-term follow-up of 1,587 participants [8] recorded the composite of all-cause mortality, non-fatal MI, non-fatal stroke, and heart failure hospitalisation in 187 of 985 periodontally healthy participants (19%) versus 174 of 602 with periodontitis (29%; P<0.0001). The adjusted hazard ratio was 1.26 (95% CI 1.01\u20131.57; P=0.038)<\/strong>. Two features temper this: the confidence interval\u2019s lower bound touches unity, and adjustment covered only age, smoking, and diabetes \u2014 a narrower set than the roughly 100 variables collected at baseline, which leaves more room for residual confounding than the study\u2019s design would have permitted.<\/p>\n SCAPIS<\/strong> (2026) is the most recent large-scale imaging cohort [9]. Among 29,056 Swedish participants, severe periodontitis was present in 6% (n=1,809) and was associated with severe coronary artery calcium score (CACS \u2265301, OR 1.69, 95% CI 1.39\u20132.06) and severe segment involvement score (>4 segments, OR 1.40, 1.17\u20131.67), with the strongest associations among individuals without<\/em> concomitant dental caries. Periodontitis was also associated with lower epicardial adipose tissue attenuation on cardiac CT (\u03b2 = \u22120.27 HU, \u22120.48 to \u22120.05), a radiographic signature of perivascular inflammation, and with a 42% higher risk of incident coronary heart disease (HR 1.42, 1.03\u20131.97), partially mediated by coronary atherosclerosis.<\/p>\n This is a particularly informative observational design because it integrates dental phenotype, coronary imaging, an inflammatory imaging marker, and incident CHD in one cohort, and the caries-stratified finding argues modestly against pure \u201cpoor oral health equals poor general health\u201d confounding. The investigators themselves frame severe periodontitis as a marker<\/em> of increased coronary risk; the design does not establish that periodontitis caused the imaging abnormalities or the events. The accompanying editorial reaches the same position, treating the mechanistic and causal questions as unresolved [53].<\/p>\n ARIC<\/strong> examined ischaemic stroke by subtype in a clinically examined cohort [35]. Periodontal disease classes were associated with cardioembolic stroke (HR \u2248 2.6, 1.2\u20135.6) and thrombotic stroke (HR \u2248 2.2, 1.3\u20133.8) after adjustment for age, sex, race and study centre, BMI, hypertension, diabetes, LDL, smoking status and pack-years, and education. Regular dental care use was associated with lower ischaemic stroke risk (HR \u2248 0.77, 0.63\u20130.94) \u2014 an observational association particularly vulnerable to healthy-user and healthcare-access confounding.<\/p>\n Established cardiovascular disease.<\/strong> A cohort of 1,002 patients with angiographically documented CVD followed 10 years found severe periodontitis associated with recurrent MI, stroke\/TIA, or cardiovascular death, adjusted HR 1.26 (1.00\u20131.58) \u2014 a confidence interval whose lower bound touches unity. Brushing more than once daily (HR 0.74, 0.57\u20130.97) and interdental cleaning (HR 0.71, 0.52\u20130.99) were associated with fewer recurrences, findings vulnerable to the same healthy-user confounding [36].<\/p>\n Peripheral artery disease<\/strong> has a thinner literature. A nationwide matched Korean cohort reported PAD incidence of 2.40 versus 2.08 per 1,000 person-years in periodontitis versus controls, HR \u2248 1.15 (1.07\u20131.23) \u2014 a small absolute difference.<\/p>\n Mortality.<\/strong> Romandini and colleagues pooled 57 studies covering 48 cohorts and 5.71 million participants, reporting all-cause mortality RR 1.46 (95% CI 1.15\u20131.85) and cardiovascular mortality RR 1.47 (1.14\u20131.90) for periodontitis [37]. It is cited in the 2026 AHA statement as a principal mortality reference. Guo\u2019s pooled estimates give cardiac death RR 1.42 (1.10\u20131.84) and all-cause mortality RR 1.31 (1.07\u20131.61) [5].<\/p>\n The largest relevant dataset is a Korean national cohort of 3,779,490 individuals free of cardiovascular disease at baseline, followed a median of 10.4 years, in which 17,942 incident cardiovascular events occurred [18]:<\/p>\n An aHR of 1.05 restricted to one outcome in one age stratum, from claims-based exposure definition, is at the threshold of detectability and well within residual confounding. Using \u201cgingivitis\u201d and \u201cperiodontitis\u201d interchangeably in cardiovascular discussion is not a simplification; it is an error.<\/p>\n The pooled effect is roughly a 20% relative increase. In a population with 10% baseline ten-year ASCVD risk, that is about 2 absolute percentage points \u2014 approximately 20 excess events per 1,000 people over ten years, if the association were entirely causal<\/em>.<\/p>\n The exposed group is enriched for current smoking, poor glycaemic control, obesity, lower income, lower educational attainment, and reduced healthcare contact. Multivariable adjustment addresses the measured portion of each. It does not address:<\/p>\n Janket\u2019s meta-regression estimated that residual confounding inflated the pooled estimate by approximately 12.9%, while use of periodontal proxies deflated it by 29.7% [27]. These are model-dependent estimates from a 2003 analysis, illustrative of direction rather than definitive.<\/p>\n A more revealing observation comes from FINRISK. In a prospective analysis, elevated antibody response to periodontal pathogens predicted incident CVD (reported HR 1.87, 1.13\u20133.08), but most associations attenuated to non-significance when serum lipids entered the model; only the endotoxin\/HDL ratio remained independently predictive [38]. Periodontal inflammatory exposure and lipid biology are statistically entangled to a degree that should caution anyone treating them as separable risk channels.<\/p>\n Honest summary:<\/strong> an effect of this size is within the range that the known confounding structure could plausibly produce. That does not mean it is confounding. It means the observational studies conducted to date cannot distinguish the two possibilities. Better observational work \u2014 negative-control outcomes, repeated exposure measurement, quantitative bias analysis, triangulation across designs \u2014 could meaningfully strengthen or weaken the inference, but no observational design can settle it on its own.<\/p>\n Dental biofilm Human studies support several components of this pathway, though the individual arrows differ considerably in evidentiary strength. The arrow into<\/em> the plaque box \u2014 the quantitative contribution of periodontal inflammation to human atherogenesis relative to apoB retention \u2014 is hypothesised, not measured [1]. That is the whole of the disagreement in this field.<\/p>\n Periodontitis is associated with modest elevations in CRP, IL-6, and TNF-\u03b1. The 2026 AHA statement identifies chronic low-grade systemic inflammation, with elevated CRP, IL-6, and TNF-\u03b1 and reduced anti-inflammatory mediators such as adiponectin, as the principal indirect mechanism [1].<\/p>\n The interventional evidence quantifies the magnitude, and three independent syntheses agree closely:<\/p>\n Convergence on roughly \u22120.6 mg\/L at six months is one of the more reliable quantitative findings in this field. Two caveats matter more than the point estimate. Reductions were greatest in participants with baseline CRP above 3 mg\/L. And in the four trials assessing CRP at twelve months or beyond, no pooled treatment effect could be detected<\/strong> [14] \u2014 an absence of demonstrated durability in a very small subset, which is not the same as a demonstrated return to baseline.<\/p>\n Is this enough to matter?<\/strong> Two observations argue for caution. First, a reduction of this size is modest in absolute terms, and the relationship between CRP lowering and event reduction is not straightforward: statin outcome benefit tracks closely with apoB and LDL-C reduction, and the independent contribution of the accompanying anti-inflammatory effect remains contested rather than settled. Second, the effect does not persist beyond six months. Atherogenesis operates over decades. This is an inference rather than a demonstrated result<\/em>, but a biomarker change undetectable at twelve months is a poor candidate for a mechanism that must accumulate over forty years.<\/p>\n The Tonetti trial makes the same point from another angle: between-group CRP differences failed to reach significance at both two months (1.4 mg\/L; \u22121.0 to 1.8) and six months (1.4 mg\/L; \u22121.0 to 2.0), even while vascular function clearly improved [12]. Whatever produced the vascular benefit, it was not detectably CRP.<\/p>\n IL-6 findings generally run in the same direction; TNF-\u03b1 is more variable. The BMC 2024 meta-analysis found moderate-certainty evidence for IL-6 reduction; no statistically significant pooled reduction was detected for IL-1\u03b2 or TNF-\u03b1, with low certainty for those outcomes [39].<\/p>\n Transient bacteremia from the oral cavity is real and well documented. It occurs during dental procedures but far more frequently during ordinary activities \u2014 chewing, toothbrushing, flossing \u2014 because these happen multiple times daily. In periodontitis, the ulcerated pocket epithelium provides a large, chronically inflamed portal of entry.<\/p>\n Forner and colleagues quantified this directly in humans: bacteremia following scaling was significantly more frequent and of greater magnitude in participants with periodontitis than in those with gingivitis or periodontal health, and the magnitude correlated with gingival inflammation, plaque scores, and bleeding on probing [41].<\/p>\n The 2026 AHA statement describes this as the principal direct mechanism: systemic dissemination of oral pathogens including Porphyromonas gingivalis<\/em> and Aggregatibacter actinomycetemcomitans<\/em>, together with virulence factors such as lipopolysaccharide and gingipains, capable of triggering endothelial dysfunction and vascular inflammation [1].<\/p>\n The correct inference is not that brushing is dangerous.<\/strong> It is the reverse: a chronically inflamed, bleeding periodontal surface presents more opportunity for bacterial entry, and maintaining periodontal health reduces that interface. This is also precisely why endocarditis guidance emphasises daily oral health over procedural antibiotics [19].<\/p>\n The counter-argument remains important: transient oral bacteremia occurs in people without periodontitis too \u2014 its frequency and magnitude vary by activity and periodontal status \u2014 and the great majority of episodes are cleared without consequence.<\/p>\n This is where the field is least honest with itself, and where the two source drafts for this document differed most usefully.<\/p>\n Positive findings.<\/strong> Kozarov and colleagues demonstrated that P. gingivalis<\/em> and A. actinomycetemcomitans<\/em> recovered from a carotid plaque homogenate could invade host cells in vitro, indicating viability, although direct culture failed [22]. Haraszthy and colleagues reported periodontal pathogen identification in atheromatous plaques [23].<\/p>\n Negative findings, from studies designed to find it.<\/strong> Aimetti and colleagues examined subgingival plaque and carotid atheromas from patients with advanced chronic periodontitis. All subgingival samples were positive for at least one target organism, with T. forsythia<\/em> 69.7%, P. gingivalis<\/em> 63.6%, T. denticola<\/em> 54.5%, P. intermedia<\/em> 45.4%, A. actinomycetemcomitans<\/em> 33.3%. Bacterial DNA was detected in 31 of 33<\/strong> endarterectomy specimens. None tested positive for periodontal pathogen DNA<\/strong> [42]. That result is not a failure of sensitivity \u2014 the assay found abundant bacterial DNA in the same specimens. It found the wrong bacteria.<\/p>\n Cairo and colleagues ran a controlled design with 26 dentate and 26 edentulous patients undergoing carotid endarterectomy. Subgingival samples from dentate test patients showed T. forsythensis<\/em> 79%, F. nucleatum<\/em> 63%, P. intermedia<\/em> 53%, P. gingivalis<\/em> 37%. No periodontal bacterial DNA was detected in any carotid sample in either group<\/strong> [43].<\/p>\n Fiehn and colleagues set out specifically to determine whether viable<\/em> oral bacteria could be recovered from atherosclerotic lesions. Culture for the target periodontal organisms was negative, while bacterial DNA was detectable by PCR in the specimens tested [44]. Detection and viability are not the same finding, and this study separates them directly.<\/p>\n Why this does not establish causation.<\/strong> Five reasons, each independently sufficient:<\/p>\n Finding P. gingivalis<\/em> DNA in an atheroma proves exposure at most.<\/strong> It does not establish colonisation, viability, biological activity, direction of travel, or causal acceleration of atherosclerosis.<\/p>\n Tonetti et al., NEJM 2007<\/strong> randomised 120 patients with severe periodontitis to intensive periodontal treatment or community-based care [12]. The findings were biphasic and instructive:<\/p>\n Czesnikiewicz-Guzik et al., EHJ 2019<\/strong> found FMD improved by approximately 1.7 percentage points at two months in hypertensive patients [11].<\/p>\n Kapellas et al., Hypertension 2014<\/strong> randomised 273 Aboriginal Australians with periodontitis to full-mouth scaling in a single visit versus no treatment, with 12-month carotid intima-media thickness and 3- and 12-month pulse wave velocity as prespecified primary endpoints [45]. At 12 months, cIMT fell by \u22120.023 mm (\u22120.038 to \u22120.008) in the treatment arm and rose by +0.002 mm (\u22120.017 to +0.022) in controls; the reported between-group difference was \u22120.026 mm (\u22120.048 to \u22120.003; P=0.03). Pulse wave velocity did not differ. Attrition was substantial \u2014 endpoints could be calculated for 169 participants at 3 months and 168 at 12 months out of 273 randomised, i.e.\u00a0roughly 38% loss.<\/p>\n Orlandi et al., EHJ 2025<\/strong> randomised 135 participants with periodontitis and no major systemic comorbidity (68 intensive, 67 control) to intensive versus control periodontal treatment, with cIMT at 24 months as the primary outcome [13]. cIMT was lower in the intensive group by 0.023 mm at 24 months (P<0.0001). The between-group difference in change from baseline at 24 months was \u22120.022 mm (95% CI \u22120.027 to \u22120.018), and a post hoc analysis stratifying by median baseline cIMT found a consistent \u22120.02 mm (\u22120.03 to \u22120.02; P<0.0001) in both strata. The between-group FMD difference at 24 months was 2.66 percentage points (2.24\u20133.09). FMD improved within two months and remained higher throughout. Glycoprotein acetyls, an integrated inflammation marker, were reduced. No between-group differences in blood pressure, pulse wave velocity, anthropometrics, or metabolomic markers.<\/p>\n The convergence is the interesting part.<\/strong> Two independent randomised trials, eleven years apart, in radically different populations (remote Indigenous Australians with high comorbidity burden; healthy urban Londoners), using different treatment protocols and follow-up durations, produced effects of the same order: \u22120.026 mm at 12 months and approximately \u22120.02 mm at 24 months. That is either a real and reproducible effect or a coincidence of two small trials. It deserves to be taken seriously.<\/p>\n It also deserves proportionate interpretation. The absolute difference is 23 micrometres. cIMT is an imperfect surrogate with a mixed record in drug development, and cIMT regression has not consistently predicted event reduction. Both trials were open-label with substantial or moderate attrition.<\/p>\n Pooled FMD evidence is weaker than individual trials suggest.<\/strong> The BMC Oral Health 2024 meta-analysis of 21 RCTs found the pooled FMD treatment effect not statistically significant<\/strong>: 1.70% (\u22121.63 to 5.03, P=0.32, I\u00b2=53%), graded low certainty [39]. Lyu\u2019s 14-trial meta-analysis (491 participants) found a strong short-term effect (\u22643 months, WMD \u22123.78, 95% CI \u22125.49 to \u22122.07, P<0.0001, in the authors\u2019 sign convention) but a non-significant<\/strong> six-month estimate (WMD \u22120.96, \u22122.06 to 0.14, P=0.09) [30]. Orlandi\u2019s individual trial sustained an FMD difference to 24 months [13].<\/p>\n These are reconcilable rather than contradictory, and the reconciliation is informative: the acute-to-subacute FMD response to periodontal therapy is well established; its persistence is not.<\/strong> Individual well-conducted trials with sustained periodontal maintenance show durable improvement; pooled estimates across a heterogeneous literature, in which maintenance and control-arm handling vary widely, do not.<\/p>\n Several pathways are biologically coherent and partly demonstrated:<\/p>\n2. Causes and risk factors for periodontal disease<\/h3>\n
\n\n
\n \nFactor<\/td>\n Role<\/td>\n Strength and interpretation<\/td>\n<\/tr>\n<\/thead>\n \n Dysbiotic dental biofilm<\/strong><\/td>\n Initiating cause of plaque-induced gingivitis; necessary component of periodontitis pathogenesis<\/td>\n Causal.<\/strong> Experimental withdrawal\/reinstitution evidence [33]; necessary but not sufficient<\/td>\n<\/tr>\n \n Inadequate plaque control<\/strong><\/td>\n Increases biofilm accumulation and maturation<\/td>\n Causal exposure.<\/strong> Directly modifiable; improvement reverses gingivitis<\/td>\n<\/tr>\n \n Smoking<\/strong><\/td>\n Strong risk and progression modifier; impairs neutrophil function, reduces gingival blood flow, worsens treatment response, masks bleeding<\/td>\n Causal.<\/strong> Dose-related; incorporated into periodontal grading. Also a strong independent cause of ASCVD \u2014 the central confounder<\/td>\n<\/tr>\n \n Diabetes \/ glycaemic control<\/strong><\/td>\n Strong susceptibility and progression modifier; bidirectional<\/td>\n Causal.<\/strong> Explicit grading modifier. Cochrane finds moderate-certainty evidence that periodontal therapy improves HbA1c<\/td>\n<\/tr>\n \n Host inflammatory response<\/strong><\/td>\n Tissue destruction is predominantly host-mediated (MMPs, RANKL-driven osteoclast activation)<\/td>\n Causal mechanism.<\/strong> Hyper-responsive phenotypes progress faster<\/td>\n<\/tr>\n \n Age<\/strong><\/td>\n Strong risk indicator<\/td>\n Mostly not an independent biological cause.<\/strong> Reflects cumulative exposure, immunosenescence, comorbidity, and lifetime opportunity for attachment loss<\/td>\n<\/tr>\n \n Genetics<\/strong><\/td>\n Modifies susceptibility<\/td>\n Partial.<\/strong> Periodontitis has a heritable component, but estimates vary widely by design and known variants explain only a minority of susceptibility. GWAS loci individually explain little variance \u2014 a limitation that becomes decisive in Section 4.6<\/td>\n<\/tr>\n \n Obesity \/ metabolic dysfunction<\/strong><\/td>\n Associated inflammatory-metabolic modifier<\/td>\n Probable contributor.<\/strong> Observational evidence substantial; some Mendelian randomisation supports causal effects of adiposity traits<\/td>\n<\/tr>\n \n Medications<\/strong><\/td>\n Gingival enlargement (phenytoin, ciclosporin, some calcium channel blockers); xerostomia from many drug classes<\/td>\n Indirect.<\/strong> Complicates plaque control rather than causing periodontitis<\/td>\n<\/tr>\n \n Dry mouth \/ hyposalivation<\/strong><\/td>\n Removes a major protective mechanism<\/td>\n Indirect modifier.<\/strong> From medication, radiotherapy, Sj\u00f6gren disease, diabetes<\/td>\n<\/tr>\n \n Diet and nutrition<\/strong><\/td>\n Refined carbohydrate supports dysbiotic biofilm; severe vitamin C deficiency directly impairs gingival collagen<\/td>\n Weak relative to biofilm, smoking, and glycaemic control<\/strong><\/td>\n<\/tr>\n \n Socioeconomic status and healthcare access<\/strong><\/td>\n Powerful determinant of exposure and treatment<\/td>\n Mostly not a biological cause<\/strong> \u2014 a determinant of biofilm control, professional care, smoking, diet. Independently associated with ASCVD. Difficult to capture fully with conventional variables such as income and education<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n The confounding structure<\/h3>\n
3. Association with cardiovascular disease<\/h3>\n
Pooled estimates<\/h3>\n
\n\n
\n \nSource<\/td>\n Studies \/ participants<\/td>\n Outcome<\/td>\n Pooled estimate (95% CI)<\/td>\n<\/tr>\n<\/thead>\n \n Larvin 2021 [34]<\/td>\n 32 cohorts (30 pooled)<\/td>\n All CVD<\/td>\n RR 1.20 (1.14\u20131.26)<\/strong><\/td>\n<\/tr>\n \n <\/td>\n <\/td>\n CHD<\/td>\n RR 1.14 (1.08\u20131.21)<\/strong><\/td>\n<\/tr>\n \n <\/td>\n <\/td>\n Stroke<\/td>\n RR 1.24 (1.12\u20131.38)<\/strong><\/td>\n<\/tr>\n \n <\/td>\n <\/td>\n MI<\/strong><\/td>\n RR 1.12 (0.96\u20131.30)<\/strong> \u2014 not significant<\/em><\/td>\n<\/tr>\n \n <\/td>\n <\/td>\n Severe periodontal disease<\/td>\n RR 1.25 (1.15\u20131.35)<\/strong><\/td>\n<\/tr>\n \n Guo 2023 [5]<\/td>\n 39 cohorts \/ 4,389,263<\/td>\n MACE<\/td>\n RR 1.24 (1.15\u20131.34)<\/strong><\/td>\n<\/tr>\n \n <\/td>\n <\/td>\n CHD<\/td>\n RR 1.20 (1.12\u20131.29)<\/strong><\/td>\n<\/tr>\n \n <\/td>\n <\/td>\n MI<\/td>\n RR 1.14 (1.06\u20131.22)<\/strong><\/td>\n<\/tr>\n \n <\/td>\n <\/td>\n Stroke<\/td>\n RR 1.26 (1.15\u20131.37)<\/strong><\/td>\n<\/tr>\n \n <\/td>\n <\/td>\n Cardiac death<\/td>\n RR 1.42 (1.10\u20131.84)<\/strong><\/td>\n<\/tr>\n \n <\/td>\n <\/td>\n All-cause mortality<\/td>\n RR 1.31 (1.07\u20131.61)<\/strong><\/td>\n<\/tr>\n \n Janket 2003 [27]<\/td>\n 9 cohorts<\/td>\n Future CVD<\/td>\n RR 1.19 (1.08\u20131.32)<\/strong>; 1.44 (1.20\u20131.73)<\/strong> if aged \u226465<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n Individual cohorts of note<\/h3>\n
Gingivitis alone<\/h3>\n
\n
How much could be confounding?<\/h3>\n
\n
4. Could periodontal disease actually worsen atherosclerosis?<\/h3>\n
4.1 The hypothesised pathway<\/h4>\n
\n\u2193
\nGingivitis\u00a0 \u2192\u00a0 [susceptible host + persistent exposure]\u00a0 \u2192\u00a0 Periodontitis
\n\u2193
\nUlcerated periodontal pocket epithelium
\n\u2193\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 \u2193
\nTransient bacteremia\u00a0\u00a0\u00a0\u00a0\u00a0 IL-6, TNF-\u03b1, CRP
\n\u2193\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 \u2193
\nInnate immune and endothelial activation
\n\u2193
\nOxidative stress \u00b7 impaired NO \u00b7 leukocyte adhesion
\n\u2193
\nApoB lipoproteins retained in artery wall\u00a0 \u2192\u00a0 PLAQUE\u00a0 \u2192\u00a0 inflammation, progression
\n\u2193
\nPlaque disruption \/ thrombosis
\n\u2193
\nMI \u00b7 ischaemic stroke \u00b7 PAD<\/p>\n4.2 Systemic inflammation<\/h4>\n
\n\n
\n \nSynthesis<\/td>\n Design<\/td>\n CRP effect<\/td>\n<\/tr>\n<\/thead>\n \n Luthra 2023 [14]<\/td>\n Meta-analysis, 19 RCTs at 6 months (26 trials, n=2,579 total)<\/td>\n \u22120.69 mg\/L (\u22120.97 to \u22120.40)<\/strong>, I\u00b2=66%<\/td>\n<\/tr>\n \n BMC Oral Health 2024 [39]<\/td>\n Meta-analysis, 21 RCTs<\/td>\n \u22120.63 mg\/L (\u22121.02 to \u22120.24)<\/strong>, I\u00b2=66%; GRADE low<\/strong> certainty<\/td>\n<\/tr>\n \n Umbrella review 2025 [40]<\/td>\n Recalculated random-effects across reviews<\/td>\n \u22120.58 mg\/L (\u22120.91 to \u22120.25)<\/strong><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n 4.3 Bacteremia<\/h4>\n
4.4 Bacteria in atherosclerotic plaque \u2014 a genuinely contradictory literature<\/h4>\n
\n
4.5 Endothelial dysfunction and vascular structure<\/h4>\n
\n
4.6 Immune mechanisms<\/h4>\n
\n