Evidence Library

Ridker and colleagues (New England Journal of Medicine, 2008) enrolled 17,802 apparently healthy adults with LDL cholesterol below 130 mg/dL and hs-CRP above 2.0 mg/L in the JUPITER trial. Rosuvastatin reduced the primary composite endpoint — myocardial infarction, stroke, arterial revascularization, hospitalization for unstable angina, or confirmed cardiovascular death — by 44 percent (hazard ratio 0.56, 95% CI 0.46 to 0.69), and the greatest benefit was observed in participants who achieved the lowest on-treatment CRP levels. Because rosuvastatin lowered both LDL cholesterol and CRP, JUPITER alone could not isolate inflammation as the responsible mechanism (Yousuf and colleagues, JACC, 2013). In CANTOS (Ridker, NEJM, 2017), canakinumab — an IL-1β-neutralizing monoclonal antibody — reduced recurrent cardiovascular events by approximately 15 percent in post-MI patients with no change in LDL cholesterol. COLCOT (Tardif, NEJM, 2019; n = 4,745 post-MI) reported a 23 percent reduction in ischemic events on low-dose colchicine. LoDoCo2 (Nidorf, NEJM, 2020; n = 5,522 chronic coronary disease) reported a 31 percent reduction. Pooled across four colchicine trials, the reduction in MI, stroke, or cardiovascular death was approximately 25 percent. The CIRT trial (Ridker, NEJM, 2019) of low-dose methotrexate was negative, and methotrexate did not lower IL-6, hs-CRP, or IL-1β. In CLEAR-SYNERGY/OASIS-9 (Jolly and colleagues, New England Journal of Medicine, 2025), 7,062 patients were randomized after acute myocardial infarction to low-dose colchicine or placebo. The primary cardiovascular outcome was essentially identical between groups (hazard ratio 0.99, 95% CI 0.85 to 1.16). This is the colchicine CLEAR-SYNERGY trial, not the unrelated CLEAR Outcomes trial of bempedoic acid. The ACC Scientific Statement on Inflammation and Cardiovascular Disease (Mensah and colleagues, Journal of the American College of Cardiology, 2025) recommended universal hs-CRP screening in both primary and secondary prevention. The thirty-year comparison behind that recommendation comes from the Women’s Health Study analysis (Ridker and colleagues, New England Journal of Medicine, 2024), in which a single baseline hs-CRP measurement spread thirty-year cardiovascular risk more widely than LDL cholesterol or lipoprotein(a) — adjusted hazard ratios for the highest versus lowest group of roughly 1.70 for hsCRP, 1.36 for LDL cholesterol, and 1.33 for Lp(a). A 2025 UK Biobank analysis of 448,653 participants reported that hs-CRP above 3 mg/L was associated with a 34 percent higher risk of MACE, 61 percent higher cardiovascular mortality, and 54 percent higher all-cause mortality. Figtree and colleagues (Journal of the American College of Cardiology, 2023) described a growing population of “SMuRF-less” patients — people who develop coronary disease despite none of the four standard modifiable risk factors — and noted these patients carry higher early mortality after a heart attack, not lower. Ridker and colleagues (European Heart Journal, 2026) then measured hs-CRP in 12,530 initially healthy women from the Women’s Health Study who had none of those standard risk factors, and followed them for three decades. Compared with women whose hs-CRP was below 1 mg/L, those above 3 mg/L had a 77 percent higher risk of coronary events, a 39 percent higher risk of ischemic stroke, and a 52 percent higher risk of total cardiovascular disease. The association was linear across the full range of hs-CRP and persisted after adjustment for body-mass index and kidney function.

What this means. Randomized trials establish that inflammation is a real cardiovascular risk pathway, but they do not support the simpler claim that every anti-inflammatory drug, in every setting, prevents events. Canakinumab in CANTOS and low-dose colchicine in COLCOT and LoDoCo2 reduced cardiovascular events; methotrexate in CIRT and colchicine begun during acute MI in CLEAR-SYNERGY did not. The 2025 American College of Cardiology Scientific Statement incorporates that mixed record directly: it calls inflammation clinically actionable, recommends universal hs-CRP screening in primary and secondary prevention, and cautions that not every anti-inflammatory intervention has succeeded.

The Periodontal Inflamed Surface Area (PISA) framework was introduced by Nesse and colleagues at the University of Groningen (Journal of Clinical Periodontology, 2008) to express the sum of bleeding pocket epithelium as a single numerical value in square millimeters, calculated from probing depth and bleeding on probing at one examination. Leira and colleagues at the University of Santiago de Compostela (Acta Odontologica Scandinavica, 2018) applied PISA to 80 participants in four groups of 20, classified by the CDC–AAP case definitions as periodontally healthy or as having mild, moderate, or severe periodontitis. Mean PISA was 2,309.42 ± 587.69 mm² in the severe group and 34.30 ± 16.48 mm² in the healthy group, a ratio of about 67 to 1; in that sample, a PISA of 130.33 mm² or more identified periodontitis with 98 percent sensitivity and 100 percent specificity. These are group means from a small clinical sample, not population averages, and the study measured no blood markers. The SONIC study (Miki and colleagues, Scientific Reports, 2021) randomly recruited 250 community-dwelling adults aged 69 to 71 and analyzed the 158 with complete data. In a generalized linear model adjusted for sex, education, smoking, diabetes, dyslipidemia, history of cancer, and BMI, a PISA of 500 mm² or more was associated with higher hs-CRP (risk ratio 1.77, 95% CI 1.05 to 2.98, P = 0.033); the unadjusted rank correlation was modest (rs = 0.181). Absolute hs-CRP values were low — medians of 0.71 and 0.37 mg/L above and below the cutoff — and only 16 participants had a PISA of 500 mm² or more. The Global Burden of Disease 2021 analysis (GBD 2021 Oral Disorders Collaborators, The Lancet, 2025) reported that severe periodontitis affected 1.07 billion people globally in 2021, with age-standardized prevalence largely unchanged since 1990. In the United States, national survey data (NHANES 2009–2014) found periodontitis in 42 percent of dentate adults aged 30 and older, and severe periodontitis in 7.8 percent (Eke and colleagues, Journal of the American Dental Association, 2018). Untreated caries affected roughly 2.5 billion people, underscoring the overall scale of oral disease worldwide.

What this means. In this sample, severe periodontitis carried roughly 67 times the bleeding, inflamed pocket surface of a healthy mouth, in a disease affecting more than a billion people with little change since 1990, and in community-dwelling older adults a larger inflamed surface tracked with higher hs-CRP after adjustment. Both findings come from single studies — one small clinical sample and one cross-sectional cohort — and PISA quantifies inflamed surface at one examination, not what crosses into circulation or how often the tissue bleeds day to day.

Teles and colleagues (Journal of Clinical Periodontology, 2024) performed a longitudinal study comparing salivary and serum inflammatory biomarkers against periodontitis progression. Salivary IL-1β, IL-6, MMP-8, and interferon-gamma were significantly associated with progression. Serum analytes were not. Bostanci and colleagues (Scientific Reports, 2021) reported that combinations of salivary MMP-8, MMP-9, and TIMP-1 achieved greater than 90 percent diagnostic accuracy for gingivitis and 100 percent accuracy for periodontitis compared to clinically healthy controls. The saliva chapter examines salivary biomarkers in clinical detail.

What this means. In the longitudinal Teles comparison, salivary markers tracked progression and serum markers did not.

Adolph and colleagues (Scientific Reports, 2017) analyzed the IPC cohort — 76,188 adults examined at the Centre d'Investigation Préventive et Clinique in Paris and followed for a mean of 3.4 years — to evaluate the relationship between cumulative oral disease burden and long-term mortality. Participants with a prior history of cancer or cardiovascular disease were excluded at the outset. Carrying three or more oral diseases simultaneously was associated with a hazard ratio of 3.39 for all-cause mortality (95% CI 2.51 to 5.42), estimated with Cox models using propensity scores calibrated for each oral exposure. The signal for non-cardiovascular, non-cancer mortality was larger — hazard ratio 4.71 (95% CI 1.74 to 12.7). The methodological strengths are the prospective design and the large sample. The limits are the relatively short mean follow-up, the residual confounding that no observational analysis can fully eliminate, the composite nature of the exposure, the measurement of oral disease through clinical assessment rather than continuous biomarker tracking, and the cohort's geographic specificity. None of those limits erases a hazard ratio of this magnitude after adjustment: in a large prospective cohort, mortality risk rose with the number of oral diseases, with the authors reporting a positive linear association. The design is observational, so the result establishes a strong, graded association rather than proof of cause.

GLP-1-based medications consistently move several inflammatory biomarkers in a favorable direction, but the size and meaning of that shift depend on the marker and the drug. Earlier meta-analyses reported reductions in hs-CRP across GLP-1 receptor agonists. A 2026 meta-analysis of 41 randomized trials found a significant pooled reduction in CRP/hs-CRP with GLP-1 receptor agonists overall, while pooled effects on IL-6, TNF-α, and MCP-1 were not consistently significant. In the tirzepatide trials included in that analysis, IL-6 fell significantly and CRP showed a favorable but non-significant pooled trend. A 2026 SURMOUNT-1 post hoc analysis separately found that tirzepatide lowered hs-CRP by roughly one-third and IL-6 by about 30 percent over long-term treatment, with the biomarker changes tracking, but not simply duplicating, weight loss.

Evidence boundary. Hs-CRP is a biomarker, not a cardiovascular outcome. These findings support an anti-inflammatory signal; they do not prove that the inflammatory change itself is what produces cardiovascular benefit. No large trial has tested periodontal treatment specifically in people already taking these medications, so whether periodontal therapy adds independently to the systemic inflammatory effect remains unknown.

What this means. The drug can meaningfully lower systemic inflammatory markers while active periodontal disease still requires its own treatment. Those are compatible facts, not competing explanations.

Witt and colleagues (Journal of the American Dental Association, 2026) used CT imaging to evaluate 37 Egyptian mummies — 23 males and 14 females, ages nineteen to sixty, spanning roughly 1650 BCE to 364 CE — selected from a larger Horus database of 137 mummies on the basis of previously documented vascular calcification. The cohort is therefore not a random sample; it is a vascular-disease cohort whose dental findings were examined retrospectively. Oral disease was extensive. Periapical lesions at the root tips averaged 5.58 per male and 4.21 per female. Periodontal bone loss, measured from the cementoenamel junction to the alveolar crest, averaged 6.48 mm in males and 5.98 mm in females. Advanced cavities affected 6.67 teeth per male and 7.64 per female on average. A composite oral inflammation score (combining periapical lucencies, caries, and bone loss) was correlated with the number of calcified vascular beds. The unadjusted association was significant (p < 0.010). After adjustment for age and sex it attenuated to a trend (p = 0.080) — age codrives both outcomes, and the sample is small. One mummy carried high oral inflammation but no detectable vascular calcification. In a post-hoc sensitivity analysis excluding that single outlier, the correlation between oral inflammation and vascular disease reached statistical significance (p = 0.027) — a secondary result, not the primary adjusted finding.

Periodontitis is associated with higher circulating C-reactive protein, but the more clinically useful question is whether treating the periodontal inflammation changes that systemic signal. A 2023 systematic review and meta-analysis of randomized clinical trials found that periodontal treatment lowered CRP, with the size of the reduction varying according to baseline inflammatory burden and disease severity. A 2026 meta-analysis of randomized trials likewise found a mean CRP reduction of about 0.58 mg/L, while rating the certainty of the evidence moderate to low. Other reviews have reached the same directional conclusion while emphasizing substantial heterogeneity across studies and time points.

What this means. Treating periodontitis can lower a systemic inflammatory marker, but there is no single CRP reduction that should be expected in every patient. The magnitude and clinical significance vary across populations and study designs.

Evidence boundary. CRP is a biomarker, not a cardiovascular outcome. These studies do not establish that periodontal treatment prevents heart attacks, strokes, or other systemic events.

The grouping itself comes from Socransky and colleagues (Journal of Clinical Periodontology, 1998), who analyzed more than 13,000 subgingival plaque samples from 185 subjects and identified five co-occurring complexes — red, orange, yellow, green, and purple. The red complex (P. gingivalis, T. forsythia, T. denticola) was most strongly tied to clinical disease; the orange complex (F. nucleatum, Prevotella intermedia, and others) was the bridge cluster.

Santi-Rocca and colleagues (Journal of Clinical Periodontology, 2025) re-examined subgingival plaque from 135 people (41 periodontally healthy, 47 with gingivitis and 47 with periodontitis) using modern 16S sequencing and proposed ten bacterial complexes rather than Socransky's five. One complex tracked with health and one with disease; the disease-associated complex contained all three red-complex species, additional Treponema types and members of the orange complex. The refinement adds detail without overturning the original picture: gingivitis sat between health and periodontitis, and several distinct community patterns marked the progression toward disease. Like the original, this is a description of which organisms travel together, from a single cohort; it does not show which organism causes disease.

The short answer: the same organisms that drive gum disease have been found expressing their virulence in the brain, the joints, and colorectal tumors — and in animal models, doing measurable damage there.

In the brain, Dominy and colleagues (Science Advances, 2019) detected P. gingivalis DNA and gingipain enzymes in the brains of Alzheimer's patients at autopsy, with gingipain levels correlating with tau and ubiquitin pathology; in mice, oral infection led to brain colonization and increased amyloid-beta, and gingipain inhibitors reduced bacterial load, blocked amyloid-beta, and rescued neurons. Xie and colleagues (Journal of Dental Research, 2025) added that P. gingivalis uses gingipains to degrade CD14, a receptor microglia need to clear amyloid-beta. For T. denticola, Su and colleagues (Journal of Molecular Neuroscience, 2021) showed oral infection induced hippocampal amyloid-beta in mice via beta- and gamma-secretase, and Ciccotosto and colleagues (Journal of Infectious Diseases, 2024) confirmed increased neuronal damage and neuroinflammation in an Alzheimer's model. None of this proves causation in humans; what it establishes is that recognized periodontal pathogens reach the brain, express their virulence there, and can be pharmacologically targeted.

In the joints, Wegner and colleagues (Arthritis & Rheumatism, 2010) characterized the P. gingivalis-specific enzyme that modifies host proteins in the way linked to rheumatoid arthritis antibodies — the molecular basis for the long-known epidemiological association.

These findings were tested in people. In the GAIN trial, 643 people with mild to moderate Alzheimer's disease were randomized to placebo or one of two doses of atuzaginstat, an oral drug that blocks P. gingivalis gingipains. The drug did not meet its co-primary cognitive and functional endpoints. A dose-related liver-toxicity signal led the U.S. Food and Drug Administration to place the program on full clinical hold in January 2022, and development was suspended. An apparent benefit in participants with detectable oral P. gingivalis came from a prespecified subgroup analysis and is exploratory. No peer-reviewed primary report of the full trial results was located; the outcome is documented in published reviews. The autopsy and animal evidence above has not, so far, translated into a treatment that works.

Evidence boundary. Most authors of the Dominy 2019 study were affiliated with Cortexyme, the developer of atuzaginstat.

In the colon, Zepeda-Rivera and colleagues (Nature, 2024) sequenced 135 F. nucleatum strains and found that one clade — Fna C2 — dominates the colorectal cancer niche, enriched in tumor tissue and in stool from 627 cancer patients versus 619 controls, carrying 195 genetic factors suited to the tumor environment. Galeano Niño and colleagues (Cancer Cell, 2026) showed the mechanism at single-cell resolution: tumor F. nucleatum disrupts epithelial cell contacts, induces cell-cycle arrest, and confers chemotherapy resistance.

Komiya and colleagues (Gut, 2019) cultured Fusobacterium nucleatum from paired tumor and saliva samples of 14 people with colorectal cancer and typed the strains. The organism was found in 8 of 14 tumors and in every saliva sample, and in 6 patients the tumor and saliva strains were identical. This is direct, if small-sample, evidence that tumor-associated F. nucleatum can share an origin with the same person's oral strain; it does not show that oral bacteria start tumors.

What this means. The mouth's disease-associated community is not confined to the mouth.

Tannerella forsythia. Settem and colleagues (Molecular Oral Microbiology, 2018) showed that T. forsythia produces methylglyoxal, a reactive compound that forms advanced glycation end products, the same class of molecules that accumulates in diabetes, and that its methylglyoxal drove inflammatory signaling in human immune cells; in periodontitis, methylglyoxal levels in gum-pocket fluid track with the organism's load. A later mouse study from the same group linked T. forsythia methylglyoxal to glycation products and bone loss in gum tissue (Settem and Sharma, Infection and Immunity, 2025). This is laboratory and animal evidence about local tissue chemistry; it does not show that T. forsythia changes blood sugar or vascular outcomes in people.

Prevotella intermedia. Kornman and Loesche (Infection and Immunity, 1982) found that the organism now called P. intermedia, which increases in the gum flora during pregnancy, could use estradiol or progesterone in place of vitamin K, an essential growth factor. In a longitudinal study of Japanese pregnant women, P. intermedia was among six periodontal species detected in placental tissue, although F. nucleatum, not P. intermedia, was the organism associated with threatened preterm labor (Ye and colleagues, Clinical Oral Investigations, 2020). More broadly, Prevotella species are strong drivers of Th17-type mucosal inflammation, the immune pathway implicated in rheumatoid arthritis and other chronic inflammatory disease (Larsen, Immunology, 2017); that link is described for the genus, not for P. intermedia in the placenta.

Campylobacter rectus. In a case-control study of 181 postpartum women, C. rectus in subgingival plaque was associated with a higher risk of preterm low birth weight in multivariable analysis (Buduneli and colleagues, Journal of Clinical Periodontology, 2005). In pregnant mice, C. rectus reached the placenta, inflamed it and restricted fetal growth, an effect that depended on the TLR4 immune receptor (Yeo and colleagues, 2005; Arce and colleagues, 2010 and 2012). The human evidence is associational.

Aggregatibacter actinomycetemcomitans (Aa). Its leukotoxin A is a pore-forming toxin that kills white blood cells (Krueger and Brown, Molecular Oral Microbiology, 2020). Konig and colleagues (Science Translational Medicine, 2016) showed that leukotoxin A drives hypercitrullination in neutrophils, generating the citrullinated proteins targeted by rheumatoid-arthritis autoantibodies, and that exposure to leukotoxic Aa was associated with those autoantibodies in people with rheumatoid arthritis. Aa is also one of the HACEK organisms that cause infective endocarditis (Talapko and colleagues, Microorganisms, 2024).

The short answer: dietary nitrate only becomes nitric oxide if oral bacteria convert it first, so an intact oral microbiome is a prerequisite for the cardiovascular benefit of greens.

The enterosalivary nitrate-nitrite-nitric oxide pathway was characterized by Lundberg, Weitzberg, and colleagues across a series of foundational papers (Nature Reviews Drug Discovery, 2008; Cell Metabolism, 2018). The salivary glands concentrate nitrate at roughly 10 to 20 times plasma concentration, and health-associated commensals — Rothia, Neisseria, Veillonella, Actinomyces — carry out the reduction to nitrite.

Two short-term human studies demonstrated the consequence of disrupting the pathway directly: Kapil and colleagues (Free Radical Biology and Medicine, 2013) found that one week of twice-daily chlorhexidine mouthwash reduced oral nitrate-reducing bacteria, lowered plasma nitrite, and raised resting systolic blood pressure by roughly 2 to 3 mmHg in healthy adults. Bondonno and colleagues (American Journal of Hypertension, 2015) extended the finding to people already treated for hypertension: three days of antibacterial mouthwash blunted oral nitrate reduction and raised systolic blood pressure by about 2.3 mmHg. Rosier and colleagues (Journal of Dental Research, 2018) reviewed how dysbiosis erodes the oral microbiome's health-maintaining functions, including the denitrification capacity that supports nitrate metabolism.

When the controlled trials are pooled, the picture is more modest. Farook and colleagues (International Journal of Dental Hygiene, 2024; published online 2023) combined five crossover trials and found an average systolic rise of about 1.6 mmHg (95% CI −0.15 to 3.33) that was not statistically significant, with high inconsistency between trials. Toonen and colleagues (International Journal of Dental Hygiene, 2026), restricting the analysis to chlorhexidine, found a similar non-significant difference (1.56 mmHg; 95% CI −0.67 to 3.80) and judged any blood-pressure change for patients prescribed chlorhexidine to be very small. The disruption of the nitrate pathway is consistent; the blood-pressure consequence, on current pooled evidence, is uncertain.

What this means. The nitric-oxide chapter develops the clinical detail, including antiseptic mouthwash use and dietary strategy.

Commensal streptococci dominate the first plaque that forms on a cleaned tooth surface (Diaz and colleagues, Applied and Environmental Microbiology, 2006) and can make up as much as 80 percent of plaque in periodontal health (Tang and colleagues, Scientific Reports, 2022). Tang and colleagues showed, in macrophages, that streptococcal hydrogen peroxide activates Nrf2, which in turn suppresses NF-κB, the master switch of inflammatory gene activation. Helliwell and colleagues (Infection and Immunity, 2025) found that gingival epithelial cells that had first taken up Streptococcus sanguinis membrane vesicles were protected from later damage by P. gingivalis vesicles. Herrero and colleagues (Scientific Reports, 2016) showed that peroxidase and blood components, the kind released into inflamed tissue, neutralize commensal hydrogen peroxide; in multispecies biofilms that neutralization produced 2- to 3-log (100- to 1,000-fold) increases in disease-associated species. Commensal streptococci also produce bacteriocins and other antimicrobials that restrain competitors (Baty and colleagues, Journal of Bacteriology, 2022). Streptococcus salivarius K12 was first developed as an oral probiotic for its strong activity against Streptococcus pyogenes, the cause of strep throat, and M18 for its activity against the mutans streptococci associated with tooth decay (Tagg and colleagues, Frontiers in Microbiology, 2023; a review by authors affiliated with the probiotics' developer). In the national NHANES survey of 8,237 U.S. adults, six genera, Veillonella, Streptococcus, Prevotella, Rothia, Actinomyces and Gemella, were present in more than 99 percent of people (Chaturvedi and colleagues, JAMA Network Open, 2025).

Evidence boundary. The immune and vesicle findings are laboratory studies, and the evidence for oral probiotics is strain-specific and still developing.

The short answer: the disease-associated organisms survive as a cooperative metabolic network that feeds itself, so the network has to be physically broken — not just chemically attacked.

The dependencies that hold a periodontal biofilm together are specific at the molecular level. Ding and colleagues (Molecular Oral Microbiology, 2023) described a network in which each organism produces what another needs: P. gingivalis makes glycine and isobutyric acid that T. denticola requires, T. denticola returns succinate and lactate that P. gingivalis uses, and T. forsythia — unable to build a critical piece of its own cell wall — scavenges that piece from F. nucleatum and P. gingivalis, supplying glucose to F. nucleatum in return.

F. nucleatum sits at the center for a further reason. Sakanaka and colleagues (mSystems, 2022) showed that it metabolizes amino acids supplied by commensal streptococci into putrescine, a molecule that accelerates P. gingivalis biofilm formation and helps the keystone pathogen spread to new sites.

What this means. This is why mechanical disruption, not antiseptics or single-organism therapies alone, remains the foundation of periodontal treatment.

The short answer: yes — a five-minute, no-laboratory flow measurement defines a clinical threshold, which makes saliva a modifiable ecosystem variable rather than a vague complaint.

A 2025 review in Advances in Experimental Medicine and Biology described saliva as a "guardian of oral microbial balance," noting that its antimicrobial proteins and antibodies are critical for preventing infection and maintaining the equilibrium of the resident microbiota. The Agostini and colleagues meta-analysis (Brazilian Dental Journal, 2018) synthesized 29 population-based studies and reported a pooled xerostomia prevalence of 23.0 percent (95% CI 18.0 to 28.0), higher, at 27.2 percent, in studies of older people only. More than 500 medications can cause dry mouth, with anticholinergics, antidepressants, antihistamines, antihypertensives, and diuretics most often implicated (Wolff and colleagues, Drugs R&D, 2017).

What this means. A whole-mouth unstimulated salivary flow at or below 0.1 mL/min — the standard threshold for salivary gland hypofunction — warrants active management.

The short answer: enough to start the shift. Intraoral pH falls stepwise — about 7.3 awake, 7.0 in nasal sleep, 6.6 during mouth-breathing sleep — and the mouth-breathing drop runs deeper and longer than waking pH ever sees.

Choi and colleagues (Journal of Oral Rehabilitation, 2016) recorded intraoral pH continuously across 48-hour periods in ten healthy adults, using a nose clip to force mouth breathing: mean pH measured 7.3 awake, 7.0 in nasal sleep, and 6.6 under forced mouth breathing (P < 0.01), the mouth-breathing fall both deeper and longer. This is a continuous intraoral pH recording across sleep, and the sample was small — directional evidence, not the final word. Salivary flow, the mouth's main acid buffer, falls to its lowest at rest, including overnight (Pedersen and colleagues, Journal of Oral Rehabilitation, 2018). Two pediatric studies show the consequence and its limit: Davidovich and colleagues (Clinical Oral Investigations, 2022) found children with obstructive sleep apnea carried higher cariogenic bacterial counts and worse dental status, scaling with apnea severity; Zhang and colleagues (International Dental Journal, 2026) found the same microbial shift — enrichment of Dialister, Anaeroglobus, and Scardovia — but no daytime difference in salivary pH.

What this means. Forced mouth breathing during sleep measurably lowered intraoral pH, and children with sleep-disordered breathing carried a more cariogenic community in a severity-graded pattern; the daytime pH null supports the interpretation that the acid exposure is concentrated during sleep rather than waking hours. The gum side of mouth breathing runs through a different route — plaque and inflammation — developed in the periodontal chapter.

The short answer: the association is real and rises with severity, and the mechanism is plausible — but the causal arrows are not yet nailed down.

Sleep-disordered breathing and periodontitis travel together. In a meta-analysis of ten studies and 43,296 participants, Liu and colleagues (Medicina Oral Patología Oral y Cirugía Bucal, 2023) found polysomnography-confirmed sleep-disordered breathing associated with periodontitis at an odds ratio of 1.83 (95% CI 1.52–2.20), rising with severity (OR 1.66 mild, 2.23 moderate, 2.66 severe). For the return arrow, Chen and colleagues (Frontiers in Immunology, 2026) describe a melatonin–cortisol axis in which fragmented sleep suppresses melatonin's overnight anti-inflammatory protection of periodontal tissue while dysregulated cortisol promotes immune suppression and dysbiosis. Carra and colleagues (Periodontology 2000, 2024) reviewed the same terrain and concluded that sleep disorders, obstructive sleep apnea in particular, are an emerging risk factor for periodontal health.

What this means. What's missing is trial evidence that treating one arm improves the other. The airway and sleep chapters develop both sides.

The short answer: it consistently places the bacteria most associated with tooth decay at a competitive disadvantage without disturbing the wider community — an effect some researchers have described as an oral prebiotic — but that microbial shift has translated into only a modest and inconsistent reduction in actual cavities.

The mechanism is metabolic. Streptococcus mutans, the principal decay organism, takes xylitol up but cannot ferment it into acid. Jurakova and colleagues (Molecular Oral Microbiology, 2023) found that growth on xylitol reprograms the S. mutans transcriptome in a pattern opposite to sucrose and glucose — reshaping its polysaccharide metabolism, cell wall, and early-biofilm virulence genes — though it did not abolish biofilm formation, so the effect is a handicap, not a kill. The population data are consistent: Söderling and Pienihäkkinen (Acta Odontologica Scandinavica, 2020) reviewed 21 trials and found all nine of fair-to-high quality reported lower mutans streptococci with habitual use, while the three using multispecies probes found no change in the overall microbiota — pressure on the pathogen, the broader community left intact.

The clinical payoff demands restraint. The Cochrane review (Riley and colleagues, 2015; 10 trials, 5,903 participants) could pool only one comparison — fluoride-plus-xylitol versus fluoride-only toothpaste, roughly a 13 percent reduction (PF −0.13) — and rated even that low-quality, drawn from two trials by one group in one population. The best adult trial, X-ACT (Bader and colleagues, 2013; n=691), found a 10 percent reduction that was not statistically significant, with no dose-response. Later meta-analyses capture the inconsistency: Marghalani and colleagues (2017) found a non-significant effect on very-low-quality evidence; Luo and colleagues (2024) found a significant one (SMD −0.50) but with most trials at moderate-to-high risk of bias.

What this means. The ecological effect is real and reproducible; the cavity-prevention effect is genuine but small, uneven, and clearest only when xylitol rides alongside fluoride — a reasonable adjunct that nudges the ecosystem in the right direction, not a substitute for fluoride and mechanical plaque removal.

The short answer: the bloodstream route is the best established; the oral-gut route is strong in animal models and accumulating in humans; microaspiration is well-documented in vulnerable populations; the neural route is emerging and rests on animal evidence.

The bloodstream is the most thoroughly evidenced. The 2025 American Heart Association Scientific Statement (Tran and colleagues, Circulation, 2025) described how periodontal bacteria enter circulation through ulcerated pocket tissue and reach the vasculature, with the keystone-pathogen framework (Hajishengallis, Darveau, and Curtis, Nature Reviews Microbiology, 2012) accounting for the outsized role of low-abundance P. gingivalis. The systemic signal is now quantified: Kindstedt and colleagues (Aging and Disease, 2026), in a 206-study meta-analysis, found periodontitis-associated cytokine elevations that held even in otherwise healthy people. The oral-gut route is the next best supported: Kitamoto and colleagues (Cell, 2020) showed in mouse models that swallowed oral bacteria both inflame the gut directly and train gut-homing immune cells — a two-hit mechanism, with human data still catching up. Microaspiration is well-documented in elderly and intensive-care populations, though less studied in healthy adults. The neural route is the most preliminary: oral organisms have been recovered from the brain tissue of Alzheimer's patients (Dominy and colleagues, Science Advances, 2019) and shown in animal studies to migrate along the trigeminal nerve, but the human evidence remains associative.

What this means. The book leans hardest on the bloodstream route and treats the neural route as promising but unproven.

The short answer: Streptococcus mutans, decay's lead organism, routinely enters the bloodstream during ordinary daily activity. One downstream consequence — infective endocarditis — is long established. A second set of links, to cerebral small-vessel bleeding and to the kidney disease IgA nephropathy, is real but emerging, confined to a specific strain, and not yet shown to cause disease in humans.

Streptococcus mutans is the principal organism of dental caries, and like other oral bacteria it produces transient bacteremia — brief passages into the bloodstream — during chewing, brushing, and dental work. The long-settled consequence is infective endocarditis: S. mutans is one of the viridans-group oral streptococci that can settle on heart valves, a risk that matters most in people with pre-existing valve abnormalities.

The newer line concerns a subset of strains carrying the cnm gene, which produces a collagen-binding protein (Cnm) that lets the bacterium adhere where a vessel wall is already damaged. In mice, Nakano and colleagues (Nature Communications, 2011) found that cnm-positive S. mutans increased intracerebral hemorrhage; in people, several cohorts have since tied oral carriage of these strains to cerebral microbleeds — tiny bleeds that predict stroke and cognitive decline. Watanabe and colleagues (Scientific Reports, 2016) reported an odds ratio near 14 among carriers of collagen-binding-active strains, and Hosoki and colleagues (Stroke, 2020) linked carriage to more microbleeds over time. A parallel strand ties the same strains to IgA nephropathy through human association studies (Misaki and colleagues, 2016; Ito and colleagues, 2019) and a rat model (Naka and colleagues, Communications Biology, 2024) in which Cnm binds IgA1 and deposits in the kidney's filtering units.

What this means. Here it is a specific bacterial trait, not decay itself, that does the systemic work. The brain and kidney links are biologically plausible and increasingly documented, but the human evidence is associational, the mechanisms are largely from animals, and most cohorts come from a single research network; a prospective multicenter study (RAMESSES) is underway to test whether the association is causal.

The short answer: the same periodontal bacteria colonize the titanium surface, and the surface itself appears to make matters worse rather than better.

de Waal and colleagues (Journal of Periodontology, 2017) compared 85 peri-implantitis patients to 69 healthy-implant controls and reported the highest single-study odds ratios for P. intermedia (OR 15.1) and T. forsythia (OR 13.3), with P. gingivalis, F. nucleatum, and T. denticola also elevated — though the wide confidence intervals reflect single-study power, not pooled certainty. Carvalho and colleagues (Clinical Oral Implants Research, 2023) pooled 12 studies and 1,233 participants and found more modest but more robust estimates. Their most informative addition was S. epidermidis — a skin commensal classically associated with biofilm infections on indwelling medical devices, suggesting the implant behaves more like a colonized device than like a natural tooth.

The mechanism is the implant surface itself. Sekar and colleagues (PLoS One, 2025) placed titanium and tooth-mimicking discs in healthy mouths and documented significantly more biofilm on the titanium, with electron microscopy showing measurable surface corrosion. Two further lines connect surface to disease: titanium dissolution products act as selective pressure that enriches pathogenic species while displacing beneficial early colonizers (Daubert and colleagues, 2018), and oral-pathogen acids and lipopolysaccharides drive corrosion of the protective oxide layer, releasing metal ions that further inflame the tissue (Gaur and colleagues, 2022). Kandaswamy and colleagues (2024) frame three possibilities: dysbiosis alone, corrosion alone, or a synergistic loop in which each amplifies the other.

What this means. Clinical validation of the synergistic model is still maturing, but the individual mechanistic legs are established. The materials chapter develops the material-composition question, including titanium sensitivity and the zirconia alternative.

The short answer: PISA adds up the bleeding pocket wall across every tooth into one square-millimeter figure — robust at population scale, with caveats at the individual chairside.

Leira and colleagues (Acta Odontologica Scandinavica, 2018) measured 80 participants in four groups of 20, classified by the CDC–AAP case definitions rather than the 2018 World Workshop stages. The periodontally healthy group averaged 34.30 ± 16.48 mm² and the severe periodontitis group 2,309.42 ± 587.69 mm², a ratio of about 67 to 1. In that sample, a PISA of 130.33 mm² or more identified periodontitis with 98 percent sensitivity and 100 percent specificity. The study measured no blood markers, and its groups do not translate directly into Stages I–IV.

Partial-mouth protocols built for large samples preserve the population gradient while systematically underestimating the individual value, which is why full-mouth charting remains the reference standard for one patient. Examiner technique meaningfully affects per-tooth values, and the 98 percent sensitivity derives from a single validation cohort rather than broad replication.

What this means. PISA is best read as a scale — nearly two orders of magnitude between the healthy and severe group means in one small study — rather than a precise per-person diagnostic or a population norm. It counts pocket sites that bleed on gentle probing at one examination; it does not measure how often the tissue bleeds day to day.

The short answer: when the major inflammatory comorbidities were excluded, the periodontal signal still held — periodontitis remained associated with elevated systemic inflammatory markers independently of them.

Kindstedt and colleagues (Aging and Disease, 2026) supplied the answer at scale: a meta-analysis of 206 studies, 17,953 participants, and 157 inflammatory proteins. The elevated markers were not a generic inflammation list — they fell into recognizable functional groups. Three are the core drivers of inflammaging itself (IL-1β, IL-6, TNF-α). One drives the mucosal Th17 response behind the oral-gut axis (IL-17). One is the enzyme by which periodontal tissue is physically destroyed and which the saliva chapter develops as a salivary progression marker (MMP-8). One is a neutrophil oxidant tied to coronary risk (myeloperoxidase). One is the bone-resorption signal behind the alveolar bone loss on radiographs (RANKL). And one pulls immune cells into the vascular plaques where the microbiome chapter found the same periodontal bacteria (CCL2).

What this means. The decisive result was the sensitivity analysis: restricted to participants with no diabetes, cardiovascular disease, or autoimmune diagnosis, the markers stayed elevated. Combined with treatment trials in which the same markers fall when periodontitis is treated, the evidence supports periodontitis as a source of systemic inflammatory load, not a bystander to it.

The short answer: continuously, through ordinary daily activity — the per-event doses are small, but they recur every day, and the cumulative exposure dwarfs anything a single dental procedure produces.

A 2024 systematic review measured bacteremia rates per event: dental scaling (deep cleaning) produced measurable bacteremia in nearly half of patients, toothbrushing in roughly one in six, and even chewing and flossing in a meaningful fraction of patients with active disease (Martins and colleagues, Oral Diseases, 2024). The American Heart Association's endocarditis-prevention guideline built its cumulative-exposure comparison on exactly this arithmetic — small per-event doses multiplied across a year of ordinary oral activity.

What this means. A mouth with active periodontal disease seeds the bloodstream every day (Forner and colleagues, Journal of Clinical Periodontology, 2006).

The short answer: it reliably improves the intermediate measures — vessel function, blood pressure, artery-wall structure — but the trials proving it prevents heart attacks have not yet been done.

Three trials moved the evidence forward in sequence, each on a different endpoint. Tonetti and colleagues (New England Journal of Medicine, 2007) randomized 120 patients with severe periodontitis to intensive treatment or community care and tracked flow-mediated dilation: a biphasic pattern in which endothelial function transiently worsened at 24 hours from the treatment's own inflammatory burst, then improved to a 2.0 percent absolute between-group difference at six months. Czesnikiewicz-Guzik and colleagues (European Heart Journal, 2019) randomized 101 patients with hypertension and periodontitis and reported a mean 11.1 mmHg systolic reduction at two months (95% CI 6.5 to 15.8). That is a large effect from a single small trial, and its weight rests as much on its convergence with a paired Mendelian randomization analysis across roughly 750,000 participants, which found genetic predisposition to periodontitis associated with elevated blood pressure. Orlandi and colleagues (European Heart Journal, 2025) then randomized 135 patients with 24-month follow-up and found the FMD improvement sustained plus measurably slower carotid intima-media thickness progression — moving the evidence from function and pressure into structural change. Meta-analyses replicate the effects: Lyu and colleagues (PLoS One, 2024) pooled 14 randomized trials (491 patients) and found improved flow-mediated dilation at short-term follow-up, and Luthra and colleagues (Journal of Clinical Periodontology, 2023), pooling 26 randomized trials and 2,579 participants, found that treatment lowered CRP by 0.69 mg/L (95% CI 0.40 to 0.97) at six months. One marker has resisted treatment: ADMA, which competes with nitric oxide at the vessel wall, showed no significant treatment effect in both Rapone and colleagues (2022) and Okada and colleagues (2021) — a null result the 2025 American Heart Association statement explicitly notes.

What this means. Randomized trials show periodontal treatment improves endothelial function, lowers blood pressure, and slows carotid wall thickening, while meta-analyses replicate the vascular and inflammatory effects. Trials large enough to test whether periodontal treatment prevents heart attacks, strokes, or deaths have not yet been done, and the 2022 Cochrane review rated that hard-endpoint evidence very low certainty.

The short answer: yes — the more completely the bacterial source is cleared, the larger the drop in systemic inflammation, though the field is deliberately limiting how it gets there.

The clearest evidence on antibiotic adjuncts comes from Luthra and colleagues (Journal of Clinical Periodontology, 2023): hs-CRP falls by roughly 1.10 mg/L when systemic antibiotics — usually the amoxicillin-metronidazole combination — are added to mechanical debridement, versus about 0.58 mg/L for debridement alone. The CRP doubling is robust; the clinical benefit is concentrated in Stage III and IV disease. Sgolastra and colleagues (2021) ranked the same combination highest for probing-depth and attachment gains in a network meta-analysis.

Cosgarea and colleagues (2020) found a shortened 3-day course comparable to the standard 7-day regimen at 12 months, and the 2020 Cochrane review rated long-term certainty for all antibiotic comparisons as very low. Among non-antibiotic alternatives, antimicrobial photodynamic therapy showed effects judged "very small and not clinically important" across 50 trials in the 2024 Cochrane synthesis, and de Molon and colleagues (2026) describe the field's broader shift from eradicating the pathogen to resolving the inflammation, with the gap between preclinical promise and clinical evidence still substantial.

What this means. In severe periodontitis, adding systemic antibiotics to mechanical debridement nearly doubled the fall in hs-CRP compared with debridement alone — in that comparison, more complete bacterial suppression produced a larger systemic inflammatory effect. Non-antibiotic adjuncts have not yet matched that evidence.

The short answer: a consistent association with COPD and sleep apnea, and a striking pandemic-era signal for COVID-19 severity — with the mechanism running through aspiration rather than the bloodstream.

A 2023 systematic review and meta-analysis (Molina and colleagues, Journal of Clinical Periodontology, 2023) reported periodontitis associated with chronic obstructive pulmonary disease at an odds ratio of 1.28 and with obstructive sleep apnea at 1.65. The pandemic produced two larger signals: the need for assisted ventilation in COVID-19 at an odds ratio of 6.24, and COVID-19 mortality at 2.26.

What this means. The respiratory associations are consistent and, for COVID-19 severity, large — but they are observational, and the aspiration mechanism means they weigh most in patients with compromised airway clearance.

The short answer: three major non-dental fields have now moved — cardiology cautiously, pulmonology decisively, and diabetes care furthest of all — though each moved in a different way, and most other specialties have not moved at all.

The 2025 American Heart Association Scientific Statement (Tran and colleagues, Circulation) takes a structurally careful position: it rates the evidence for the inflammatory association and for treatment's effect on intermediate outcomes — endothelial function, blood pressure, biomarkers — as strong, while calling the causal arrow on cardiovascular events themselves elusive at the current level of trial evidence. But its clinical-relevance language moves past hedging — it explicitly endorses regular dental screening and targeted periodontal care, and calls referral to a periodontist especially important for patients with cardiovascular disease.

The 2026 GOLD Report on chronic obstructive pulmonary disease moved on a lower evidentiary bar but in a stronger direction, integrating periodontal disease into COPD management. Rating the evidence low-to-moderate, GOLD judged it sufficient to recommend periodontal care for slower lung-function decline, fewer exacerbations, and reduced healthcare use.

Diabetes care has gone furthest. Its evidence base is the deepest in oral-systemic medicine, and the 2026 American Diabetes Association Standards of Care now write a periodontal step into routine diabetes evaluation — a different kind of move from GOLD's, folding dental screening into standard care rather than periodontal treatment into disease management. The diabetes chapter develops it in full.

What this means. Three major non-dental specialties — cardiology, pulmonology, and diabetes care — have now moved periodontal care into their clinical frameworks, in different forms and at different strengths. Across the wider guideline landscape, Espada-Salgado and colleagues (2026) found that operational recommendations remain concentrated in diabetes-periodontitis statements.

The short answer: the cellular mechanism is well mapped, but Mendelian randomization does not support long-term cortisol as a direct cause.

The bench mechanism is clear. Cortisol binds the glucocorticoid receptor inside the cell, and the activated complex suppresses NF-κB-driven inflammatory and antimicrobial gene expression. The downstream effects on periodontal immunity are specific: neutrophil chemotaxis, phagocytosis, and oxidative-burst capacity are all attenuated — weakening exactly the arm of innate immunity most responsible for controlling subgingival biofilm. On the bacterial side, in-vitro evidence suggests cortisol and the catecholamines released alongside it can act as signaling molecules that upregulate virulence in P. gingivalis and related organisms. Skrypnyk and colleagues (Clinical Science, 2025) reviewed this literature and integrated it with the broader psychoneuroimmunology of the stress axes.

Baumeister and colleagues (Frontiers in Endocrinology, 2023) triangulated prospective data from 3,388 participants over a median 6.9 years with two-sample Mendelian randomization across more than 45,000 cases and controls, and reported "no univocal evidence" that cortisol drives periodontitis — the observational signal was inconsistent across endpoints, and genetically proxied long-term cortisol was unrelated to disease.

What this means. The bench evidence maps how cortisol impairs periodontal immune defenses and, in vitro, how stress hormones can heighten bacterial virulence. The Mendelian randomization result answers a narrower question — whether genetically proxied long-term cortisol is itself the direct causal mediator — and does not show that stress has no periodontal effect.

The short answer: the visible findings each carry quantified predictive weight, and together with one chairside breathing test they amount to a real airway evaluation delivered in the chair the patient is already in. The soft-tissue findings each rest on measured evidence. Visible tongue scalloping carried a positive predictive value above 80 percent for moderate-to-severe OSA (Weiss 2005, replicated by Tomooka in a community cohort that ruled out sleep-clinic selection bias). The JAMA Rational Clinical Examination synthesis reported a Mallampati odds ratio of about 2.5 per grade, independent of body weight, and Friedman tongue position outperformed modified Mallampati head-to-head. Brodsky tonsillar grades 3 and 4 mark clinically meaningful narrowing (Kumar 2014), and the 2024 Camañes-Gonzalvo frenulum review consolidated the tongue-tie evidence into the chapter's cautious framing — plausible mechanism, evidence not yet mature for definitive claims. The hard-tissue and co-occurrence findings extend it. A mandibular torus predicted MAD response with an odds ratio of 2.85 (Diaz de Teran 2022; Palm 2014 reported 2.5 for thick tori). Dal Fabbro's 2025 synthesis framed nocturnal masseter activity as a response to airway compromise rather than isolated parafunction, and the TMD-OSA bidirectional link rests on the OPPERA cohort (Sanders 2013, adjusted hazard ratio 1.73 for OSA preceding first-onset TMD) plus a confirming 2024 meta-analysis. CBCT imaging advanced through Momany 2016 and Nayak 2026 into a tool that complements rather than replaces the visible exam.

What this means. Together, these chairside findings carry measured predictive weight for OSA — a meaningful airway screen, delivered in a chair most adults already visit twice a year, that identifies who needs a sleep study rather than replacing one.

The short answer: yes — even "just snoring" with no measurable apnea responds to airway-focused treatment, and current guidance now treats it as worth treating. A 2024 randomized trial in JAMA Otolaryngology compared mandibular advancement against a four-component airway-and-positional bundle — an external nasal dilator, nasal saline lavage with mometasone, mouth taping, and lateral positional therapy — in patients with primary snoring. Both arms improved sleeping-partner-reported Clinical Global Impression meaningfully: mandibular advancement produced a 91 percent responder rate, the multimodal bundle 58 percent (Ioerger 2024). The 2024 JAMA Internal Medicine Network review on obstructive sleep apnea went further, stating explicitly that mandibular advancement devices are reasonable first-line therapy for primary snoring even in the absence of formal sleep-disordered breathing on polysomnography.

What this means. The tools that treat primary snoring are largely the ones a dental practice already reaches for.

The short answer: the vibration of habitual snoring is linked to progressive nerve damage in the soft palate — shown in biopsy, longitudinal and dose-response studies and confirmed by a 31-study systematic review, though the individual studies are small. The anchor is a 2018 soft-palate biopsy study of 22 surgical patients and 12 healthy controls (Shah et al.), which found axon density in long-term snorers and obstructive sleep apnea patients reduced to roughly one-third that of controls. Longitudinal data point the same way: across six to seven years of follow-up, untreated snorers showed worsening palatal sensory loss, with cold-detection thresholds deteriorating from 4.2 to 11.0 degrees Celsius (Svanborg 2020). A dose-response relationship reinforces it — snoring duration correlated with sensory neuropathy at r = 0.47, and apnea severity at r = 0.41 (Sunnergren et al. 2011) — and Patel’s 2018 systematic review of 31 studies confirmed the broader trend.

Evidence boundary. The direction is consistent across biopsy, longitudinal, dose-response and review evidence; what the evidence base still lacks is large controlled trials.

The short answer: the jaw grows in response to the load the diet and early-life environment place on it, and the modern environment removed most of that load — so the bone that houses the airway came in smaller. The 2020 Kahn et al. BioScience synthesis drew on two converging anthropological records. The first is the skull record: comparisons of preindustrial with modern crania show progressive narrowing of dental arches, more crowding, and reduced palatal volume — the Carlson and van Gerven 1977 study of Nubian crania across a 1,500-year window is the most-cited example, with parallel patterns in European, Native American, and East Asian series. The second is the contemporary record: roughly 30 to 40 percent of children in industrialized populations now need orthodontic intervention, against far lower historical baselines. The mechanism has direct experimental support. When Corruccini and Beecher raised squirrel monkeys on artificially softened food, the soft-diet animals developed narrower dental arches, crowded premolars, and rotated teeth — within a single generation (Corruccini and Beecher 1982). The same pattern shows up in humans wherever a population's diet softens: Corruccini documented it in a rural Kentucky community and among the Pima, where the cohort raised on refined commercial food had measurably narrower palates and more malocclusion than elders raised on traditional fare (Corruccini and Whitley 1981; Corruccini, Potter, and Dahlberg 1983). Surveying these transitions across world populations, he found the speed of the malocclusion rise tracked the speed of urbanization — a signature that points to environment, not genes (Corruccini 1984). The Moss Functional Matrix Hypothesis supplies the principle beneath all of it: bone form follows the muscles, soft tissues, and mechanical forces acting on it during development, so when the load drops, the structure that bears it grows smaller. The breastfeeding literature works through the same principle by different physiology — the Boronat-Catalá 2017 meta-analysis found that longer breastfeeding is inversely associated with posterior crossbite and class II malocclusion.

What this means. Because the jaw is shaped by the load it receives while it is still growing, early, load-restoring intervention can influence it.

The short answer: they were validated on obese, older men, so their thresholds are tuned to that body — and a lean woman can clear every cutoff while her airway does not, until the test is rescored for her or replaced by an objective one. STOP-BANG is a risk questionnaire, not a diagnosis. At the standard cutoff it catches better than nine in ten moderate-to-severe cases, but only about a third of the people it flags turn out to have OSA (Pivetta 2021), and it was calibrated on a presurgical sample that was roughly two-thirds men with an average BMI near obese. In the lean or female patient it was never built for, it loses the sensitivity that is its whole justification. The Epworth Sleepiness Scale is the one most often misread as an OSA test. It measures the propensity to doze, and it tracks only weakly with how much someone actually stops breathing at night — an apnea-hypopnea index and an Epworth score can move almost independently (Kiciński 2016), and in heavier patients the scale misses most OSA outright (Singla 2018). It reads a symptom, not the disease. A newer generation of questionnaires was built squarely for the patient STOP-BANG misses. NoSAS drops the symptom questions and scores only neck, weight, snoring, age, and sex; across cohorts it matches STOP-BANG's overall accuracy while trading some sensitivity for higher specificity — fewer false alarms, but more true cases slipping through (Chen 2021). GOAL, a four-item tool, performs comparably (Duarte 2023), and No-Apnea is built as a near-pure rule-out: it catches almost everyone with disease, at the cost of flagging many who don't have it. None is clearly better than STOP-BANG across the board — and the gap they were meant to close does not close on its own. In a general-population study, NoSAS caught fewer than six in ten women with moderate-to-severe disease (about 55 percent, against 87 percent in men); only when the threshold was lowered specifically for women, with a female neck-circumference cutoff, did detection in women rise to match men's (Bauters 2020). The lesson for the missed patient is not that a smarter questionnaire rescues her — it is that the same questionnaire has to be scored against her, not against the heavy older man it was tuned on. The one instrument built from the ground up around female symptom patterns, OSAF, is promising but rests on a single cohort and awaits outside validation.

Low testosterone in a middle-aged man is usually multifactorial, with age and body composition doing most of the work; obstructive sleep apnea is one contributor among several, and its independent effect is genuinely contested. The practical question is less what caused the low number than what to do about it, and the practical concern runs in the other direction — exogenous testosterone can worsen sleep apnea, which is why the Endocrine Society's clinical practice guideline recommends against starting testosterone in men with untreated severe obstructive sleep apnea (Bhasin and colleagues, 2018). It is a strong recommendation the guideline itself grades on low-quality evidence, and its scope is narrow: untreated and severe — treated apnea, or milder disease, is not what it addresses.

Evidence boundary. None of this makes testosterone therapy wrong, and none of it makes the airway the cause of the low number. The point is narrower: in a man who has never been screened, evaluating the breathing before or alongside testosterone is prudent — it can surface a modifiable contributor and avoids pushing on an untreated severe airway with a hormone known to aggravate it. Where testosterone is genuinely indicated, it stays indicated.

The short answer: the cohort, biomarker, and per-protocol trial evidence converge on a real cardiovascular load, and four other loads — metabolic, sexual, oncologic, immune — ride the same underlying biology.

Cardiovascular. Adequate CPAP adherence lowers systolic blood pressure by 2 to 3 mmHg in mixed populations (Bratton 2015; 51 randomized trials, with each additional hour of nightly use associated with a further systolic reduction) and 5 to 7 mmHg in resistant hypertension (Iftikhar 2014; Sun 2024), alongside reductions in CRP, IL-6, and TNF-α (Zhu 2025; Xie 2013) and improved arterial stiffness (Chalegre 2021); the 2023 European Society of Hypertension guideline treats the resistant-hypertension evidence as actionable. The mortality data are the case at its strongest: the Wisconsin Sleep Cohort found an adjusted all-cause-mortality hazard ratio near 3.8 in severe untreated OSA, and the Sleep Heart Health Study found a 1.7-fold rise in coronary-heart-disease mortality in men under 70. The hard-endpoint trials are where the case is most misread: SAVE (2016), RICCADSA (2016), and ISAACC (2020) each reported negative primary intent-to-treat results; two of the three — SAVE and RICCADSA — showed significant benefit in their adherent per-protocol subgroups, while ISAACC did not. The design issue is shared — intent-to-treat measures the effect of assigning treatment, not receiving it, and average adherence fell below the threshold the mechanism requires (the treatment chapter develops this). The real case is the one the cohorts, biomarker meta-analyses, AF data, and per-protocol arms converge on — not the negative-trial headlines. The supporting cohort signals are specific: severe REM-predominant OSA raised incident systolic non-dipping roughly 2.5-fold and diastolic non-dipping about 4-fold (Mokhlesi 2015), with nocturnal blood-pressure elevation confirmed independently of BMI in more than two thousand patients (Lin 2025); atrial fibrillation runs two- to four-fold higher in severe OSA, AF recurrence after ablation rose 31 percent in untreated patients, and CPAP cut that recurrence by 42 to 44 percent across three meta-analyses (Mehra 2022); the Penn State cohort reported an all-cause-mortality hazard ratio of 1.59 and a cerebrovascular odds ratio of 1.74 for mild-to-moderate OSA in adults under sixty, rising three- to four-fold alongside cardiometabolic comorbidity. The age-modification is quantitative: an incident-hypertension hazard ratio of 3.62 in young and middle-aged adults versus 1.36 over sixty (Vgontzas 2019). Metabolic. Sleep-disordered breathing is associated with insulin resistance even after accounting for body weight. In the Sleep Heart Health Study, moderate-to-severe sleep-disordered breathing was associated with higher odds of fasting glucose intolerance after adjustment for age, sex, BMI, and waist circumference; other referred cohorts likewise found apnea severity and oxygen nadir independently related to insulin resistance. Treatment evidence is less uniform. A 2021 meta-analysis of seven randomized trials in people with type 2 diabetes and OSA found improvements in HbA1c, fasting glucose, and insulin resistance. A larger 2025 meta-analysis of 74 studies found fasting glucose declined, while pooled effects on HbA1c, insulin, and HOMA-IR were favorable but not statistically significant. The strongest synthesis is therefore modest and mixed rather than uniformly reproducible glycemic benefit.

Sexual. The periodontal chapter framed erectile dysfunction as the cardiovascular canary in a smaller vascular bed; OSA carries the same vascular load. Meta-analyses put erectile dysfunction roughly 1.8-fold higher in OSA and female sexual dysfunction at a similar 2.0-fold (Liu 2015; Kellesarian 2018). Severe OSA is also associated with lower serum testosterone, though the effect concentrates in the most severe disease and is partly entangled with obesity — one community cohort found obesity accounts for it entirely (Su 2022; Clarke 2020). CPAP can improve erectile function, but the controlled evidence is limited and mixed (Li 2019). The nitric-oxide chapter carries the full nitric-oxide biology.

Oncologic — with the most careful framing. Intermittent hypoxia is biologically distinct from steady hypoxia, and its reoxygenation cycles generate reactive oxygen species. A 2021 Cheng and Li meta-analysis found a dose-response with all-cause cancer incidence (14 / 36 / 59 percent elevation across mild / moderate / severe), and the Wisconsin cohort found a 4.8-fold cancer-mortality hazard over 22 years. But a large 2023 cohort found OSA severity not independently associated with cancer after adjustment — while nocturnal hypoxemia was — and community cohorts have been inconsistent. The signal is real; the population-level evidence is mixed.

Immune. Dong and Yue (Frontiers in Immunology, 2025) propose OSA be understood as an immune-modulating disorder; the inflammasome evidence beneath that proposal is the next entry.

What this means. The five loads share one biology — intermittent hypoxia and autonomic stress — expressed through five organ systems.

The short answer: it is the repeated swing between low and restored oxygen — not the low oxygen itself — that drives an inflammatory cascade, thousands of times a night. Intermittent hypoxia is not a milder version of sustained hypoxia; the cellular pathways only partly overlap, and the difference is the substrate of the OSA-specific inflammatory cascade. A 2020 Prabhakar synthesis (J Clin Invest) documented the architecture: each reoxygenation phase generates a burst of reactive oxygen species through mitochondrial and NADPH-oxidase pathways, recurring with every apneic cycle. Across thousands of cycles per night, the cumulative oxidative load switches on transcriptional machinery the body normally reserves for acute injury — HIF-1α, NF-κB, and the NLRP3 inflammasome. The inflammasome is the convergence point. Once assembled, it activates caspase-1, which matures IL-1β and drives the downstream cascade of IL-6, TNF-α, and the systemic inflammatory state the biomarker literature measures. This explains why CPAP's anti-inflammatory effect is dose-dependent and time-delayed: removing the nightly hypoxic cycles closes the upstream input to a cascade that takes weeks of consistent suppression to recede. Díaz-García (2022) showed it directly — monocytes from severe-OSA patients carried elevated NLRP3 activity scaling with both AHI and hypoxemic burden, measurable in circulating immune cells. Dong and Yue's 2025 reframing of OSA as an immune-modulating disorder follows from this evidence rather than asserting it independently. The pandemic made the consequence concrete. A 2023 Thorax cohort of more than 300,000 OSA patients found elevated risks of testing positive for COVID-19, hospitalization, and ICU admission after adjustment (Kendzerska), and a 21-study meta-analysis reported a composite poor-outcome odds ratio of 1.72 (Hariyanto and Kurniawan).

What this means. Intermittent hypoxia drives a specific inflammatory cascade that has been measured in patients’ own immune cells, scaling with apnea severity — which is why treating the airway also lowers systemic inflammatory markers.

Three measured changes sit underneath the story above. The first is nitric oxide. The paranasal sinuses produce it continuously and release it into the inspired airstream — the 1995 Lundberg discovery in Nature Medicine that established the sinus pathway. Mouth breathing bypasses that delivery. The second change is acidity, and it deepens with severity. Across the OSA spectrum salivary pH steps down — 6.40 in mild disease, 6.15 in moderate, 5.87 in severe — toward the 5.5 enamel-demineralization threshold, while salivary flow roughly halves over the same range (Makeeva 2021). The third change is the community that acidity selects for. Mouth-breathing children carry a higher relative abundance of opportunistic organisms — Acinetobacter, Neisseria, Streptococcus pneumoniae — with upregulated oxidative-stress salivary proteins (Fan 2020); a 2026 cohort of 8-to-12-year-olds found increased Dialister, Streptobacillus, and Scardovia, genera tied to caries and periodontal disease, alongside elevated secretory IgA (Zhang 2026). The periodontal signal follows the same gradient. A 2023 meta-analysis of 43,296 participants reported an odds ratio of 1.83 for the association between sleep-disordered breathing and periodontitis, rising with severity — 1.66 mild, 2.23 moderate, 2.66 severe (Liu et al.). A 2026 bidirectional Mendelian randomization analysis found genetic evidence for a causal direction from OSA to periodontitis, but not the reverse (Yu et al.). It is a single analysis, carrying the phenotype and assumption limits Bashir and colleagues (2025) catalogued for oral-health MR.

What this means. The move from nose to mouth is a measurable change in three chemistries — nitric oxide, pH, and microbial ecology — and the periodontal evidence runs in the direction the chemistry predicts: the more severe the airway dysfunction, the heavier the load that follows.

Some patients with upper airway resistance do not fit the familiar picture of high blood pressure and obvious apnea. In a sleep-clinic study of 4,409 referred patients, low resting blood pressure was reported in 23 percent of those classified with upper airway resistance syndrome, compared with 0.06 percent of those with obstructive sleep apnea. UARS patients who also reported orthostatic intolerance had larger drops in systolic blood pressure during tilt-table testing than comparison groups. Related early work in children with sleep-disordered breathing described low blood pressure, dizziness on standing, and chronically cold hands and feet in a subgroup with mild orthostatic symptoms.

Evidence boundary. This is an older and relatively narrow evidence base. Much of it comes from the same research lineage, and the cold-extremity observation in children has not been robustly independently replicated. UARS itself is also variably treated as a distinct syndrome or as part of the broader sleep-disordered-breathing spectrum. None of these symptoms is diagnostic of UARS or any airway disorder.

What this means. In selected patients, low blood pressure and orthostatic symptoms may coexist with subtle sleep-disordered breathing. They are clues that can justify a broader evaluation, not findings that identify the diagnosis by themselves.

The short answer: measurable traits — including lower BMI, positional dependence, mandibular retrusion and female sex — help predict response, and substantial AHI reductions occur across the severity spectrum; the well-documented dental cost is small, predictable, and worth naming up front.

The most consistent metabolic and positional predictors of MAD response — lower body mass index, smaller neck circumference, positional dependence, lower baseline AHI — were consolidated in a 2022 Camañes-Gonzalvo review building on Chen 2020 and Mostafiz 2019, and Shen 2012 added craniofacial predictors (retroglossal airway dimension, mandibular position, anterior face height). The 2024 Liao meta-analysis of 42 studies quantified the severity gradient — 48 percent AHI reduction in mild OSA, 67 percent in moderate, 62 percent in severe — and the 2026 Hamza retrospective Mayo Clinic cohort of 595 patients identified mandibular retrusion, female sex, and higher baseline AHI as the strongest response predictors. Newer mandibular-movement monitoring now measures MAD adherence as objectively as CPAP records its own, closing a gap that confounded the comparison for two decades.

The occlusal cost is real and quantifiable. Hamoda and colleagues (2019) followed MAD users for an average of 12.6 years and reported roughly 1.5 mm of overjet decrease, 1.2 mm of overbite decrease, and 8 degrees of incisor proclination, with negligible skeletal change. It is predictable enough that the clinician can tell the patient the magnitude to expect.

Combination therapy is the active frontier. Lai 2019 found MAD plus expiratory positive-airway-pressure valves outperformed MAD alone in incomplete responders; Aishah 2023 demonstrated endotype-informed combinations selected by each patient's underlying mechanism; and the 2026 Harris interdisciplinary paper (with Lavigne, of the 2020 screen/refer/treat model, as co-author) extends this into a precision-medicine pathway.

What this means. Response can be anticipated from measurable traits, and the dental cost is quantified well enough to name before treatment begins.

The short answer: no — the cardiometabolic benefit is not explained by weight loss alone: improvement in OSA metrics independently mediated part of the reductions in inflammation, insulin resistance, and triglycerides.

SURMOUNT-OSA was two parallel trials: one in patients with moderate-to-severe OSA not using CPAP, one in patients already on CPAP. Malhotra and colleagues randomized 469 patients with obesity to weekly tirzepatide (titrated to 10 or 15 mg) or placebo for 52 weeks, with absolute AHI reductions of roughly 25 events per hour in the active arms. The benefit also appeared in patients who were already being treated with CPAP.

The 2026 Malhotra Nature Medicine mediation analysis recast how to read the trial. Using formal causal-mediation methods, it decomposed the cardiometabolic benefits — lower hs-CRP, HOMA-IR, triglycerides, and systolic blood pressure — into the proportions attributable to weight loss, AHI reduction, and hypoxic-burden reduction. Improvement in the OSA metrics independently mediated part of the reductions in hs-CRP, HOMA-IR, and triglycerides; systolic blood pressure was mediated significantly through weight alone and through the combined weight-plus-OSA pathway, and no significant mediation was shown for diastolic blood pressure. The companion 2025 time-course analysis showed AHI reduction beginning by Week 4 and diverging from placebo by Week 20.

The broader class adds a related signal on the airway itself. The 2016 SCALE Sleep Apnea trial of liraglutide reported AHI improvement not fully proportional to weight loss (Blackman), and semaglutide reports are directionally consistent. Candidate mechanisms — reduced parapharyngeal and tongue fat (Wang 2020 showed tongue-fat reduction to be the primary mediator of AHI improvement with weight loss), altered ventilatory control through brainstem GLP-1 receptors, anti-inflammatory effects on airway tissue — each have support, but none is yet isolated as the driver.

What this means. In the mediation analysis, improvement in OSA metrics accounted for part of the cardiometabolic benefit independently of weight loss — reinforcing why residual airway burden still matters even when metabolic treatment is working.

The short answer: exercise and myofunctional therapy have the most credible, replicable evidence; positional therapy works in a defined subset; Buteyko, mouth taping, and breathing trainers are weaker bets that turn on patient selection.

Exercise has the most consistent independent evidence of anything in this section. In the 2020 Gottlieb and Punjabi JAMA review, the odds of moderate-to-severe OSA fell in a clear dose-response — odds ratios of 0.62, 0.39, and 0.31 across one-to-two, three-to-six, and seven-plus hours of weekly vigorous activity, roughly 69 percent lower odds at the top of the range, independent of body weight, in cross-sectional data. A 2024 twelve-RCT meta-analysis put the intervention effect near a 7-event-per-hour AHI reduction (Lin 2024). The catch is durability: the benefit reverses within a year of stopping (Berger 2021).

Myofunctional therapy illustrates the gap between observational and randomized synthesis. The 2015 Camacho meta-analysis that produced the widely-cited 50 percent AHI reduction pooled randomized and pre-post observational designs together — and pre-post designs inflate effect sizes by losing the regression-to-the-mean correction that randomization supplies. The RCT-only Saba 2024 synthesis (seven randomized trials, 310 adults; 10.2 events per hour) and the Pisoni 2025 overview of nine systematic reviews (21 primary studies, 716 patients; roughly 9.5 events per hour) both converged near 10 events per hour: lower than the headline, but more credible. The mechanism has an older and more memorable anchor: four months of daily didgeridoo practice lowered the AHI by 6.2 events per hour in moderate OSA (Puhan 2006, BMJ) — the same upper-airway-muscle conditioning that modern myofunctional therapy now operationalizes.

Positional therapy helps a defined subset, patients whose events are supine-driven: a 2023 Thorax meta-analysis of vibrotactile devices found they cut the AHI by about nine events per hour — a 43 percent reduction — and reduced time spent on the back by roughly 70 percent, though daytime sleepiness and quality of life barely moved and follow-up was short (ALQarni 2023).

Buteyko and mouth taping share a structural limit: the evidence closest to sleep-disordered breathing is indirect. Buteyko's best trials are in asthma, where breathing-exercise reviews have found benefit for quality of life and symptoms (Bruton 2005; Santino 2020), and a 2026 meta-analysis of seven Buteyko RCTs found improved asthma control as an adjunct to standard care — though the quality-of-life benefit fell mostly to children and the authors still judged the overall evidence limited (Wu 2026). No comparable trial tests it as a treatment for sleep apnea. An active-comparator RCT of Buteyko against diaphragmatic breathing, in acute coronary syndrome patients rather than an apnea cohort, found no between-group advantage in sleep quality, fatigue, or oxygenation (Safarabadi 2026). The Yang 2024 mouth-taping study visualized obstruction directly via drug-induced sleep endoscopy and found the airway worsened with taping in 22 percent of patients — capturing the mechanism rather than inferring it. The Rhee 2025 and Fangmeyer 2025 reviews each stressed how small and selectively sampled the taping evidence remains.

One last intervention deserves brief mention: inspiratory muscle training — about 30 minutes of daily resistive breathing — is sometimes raised as a way to condition the respiratory muscles. The evidence in sleep apnea is early and the trials small and variable, so it stays a curiosity to be aware of rather than something to recommend.

What this means. Exercise and myofunctional therapy carry randomized evidence of roughly 7 to 10 fewer events per hour, and positional therapy works in supine-predominant disease; Buteyko, mouth taping, and inspiratory training remain selection-dependent and need better trials.

The short answer: nerve stimulation, jaw-advancement surgery, and adult palate expansion each have a maturing evidence base — and the candidate population for each is widening, not narrowing.

Hypoglossal nerve stimulation has matured across three generations of trials reporting roughly 60 percent AHI reduction with adherence above CPAP, and newer proximal and bilateral approaches are widening the candidate population to patients earlier devices excluded.

Maxillomandibular advancement remains the most effective single intervention for severe OSA with craniofacial compromise, advancing both jaws to enlarge the airway at multiple levels; the Holty and Guilleminault meta-analysis reported about 86 percent surgical success across the pooled studies. The procedural cost is substantial, so the patient choosing MMA is trading one high-acuity surgery with durable efficacy for years of nightly device use.

The adult maxillary expansion protocols run on a precision-medicine principle keyed to how far the palatal suture has closed: EASE (endoscopically-assisted surgical expansion) for a fully closed suture, DOME (distraction osteogenesis) for gradual separation over weeks, and MARPE (temporary-anchorage expansion) in younger adults whose suture retains some patency, often without surgery.

What this means. Protocol choice is driven by skeletal maturation, deficiency magnitude, and procedural commitment.

The short answer: in the elderly it demonstrably does — daily oral care cuts pneumonia by roughly 40 percent — and institutional medicine has now begun extending the principle.

The intervention evidence comes from aspiration pneumonia in the elderly. A 2002 Yoneyama randomized trial in nursing-home residents found that daily toothbrushing plus periodic professional care reduced pneumonia incidence by about 40 percent over two years, and the Sjögren 2008 systematic review confirmed similar effect sizes for fever, lower-respiratory infection, and mortality across institutional cohorts.

Across the wider respiratory literature, the Molina 2023 synthesis pooled more than thirty studies across four respiratory-disease populations, reporting periodontitis associated with COPD at an odds ratio of 1.28, with obstructive sleep apnea at 1.65, and with COVID-19 at 6.24 for assisted ventilation and 2.26 for mortality; heterogeneity was moderate for COPD and higher for OSA, and the COVID-19 estimates came from a smaller subset with consistent results across hospital systems.

What this means. Oral care is a demonstrated respiratory intervention — in institutional populations it cut pneumonia by about 40 percent — and the 2026 GOLD Report has begun to extend the principle to COPD.

The short answer: sleep apnea is linked to a measurably higher risk of later cognitive decline, with oxygen loss the most consistent signal — but no study yet shows that treating it prevents dementia.

A meta-analysis of prospective studies found sleep-disordered breathing associated with a 26 percent higher risk of developing cognitive impairment (risk ratio 1.26, 95% CI 1.05 to 1.50), along with a small worsening in executive function (Leng and colleagues, JAMA Neurology, 2017). In a prospective study of 298 older women, those with sleep-disordered breathing were more likely to develop mild cognitive impairment or dementia (adjusted odds ratio 1.85, 95% CI 1.11 to 3.08); measures of oxygen desaturation tracked with that risk, while measures of sleep fragmentation and sleep duration did not (Yaffe and colleagues, JAMA, 2011).

The proposed mechanism runs through the brain's waste-clearance system, often called the glymphatic system. That pathway was first described in mice in 2012, and the roughly 60 percent expansion of the brain's interstitial space during sleep was measured in sleeping mice, not people. In humans, a meta-analysis of imaging studies found an indirect marker of glymphatic function lower in obstructive sleep apnea, most markedly in severe disease (Ghaderi and colleagues, Sleep Medicine, 2025). In a small study, untreated sleep apnea was associated with lower spinal-fluid amyloid-beta 42 and a higher tau-to-amyloid ratio, the pattern linked to Alzheimer's disease, while CPAP-treated patients did not show those changes (Liguori and colleagues, Sleep, 2017). In another small study, a subgroup of 12 patients treated effectively with CPAP improved on cognitive testing and in brain glucose uptake over twelve months (Fernandes and colleagues, Journal of Neurology, 2022).

What this means. Sleep apnea belongs in the conversation about cognitive aging: the association is consistent, oxygen loss is the most specific signal, and small studies suggest some brain markers and some measures of thinking may improve with consistent CPAP use.

Evidence boundary. These are associations and small studies. They do not show that sleep apnea causes dementia, or that treating it prevents dementia. The 2024 Lancet Commission on dementia did not include sleep apnea among its 14 modifiable risk factors (Livingston and colleagues, 2024).

The 2001 Kato study (Journal of Dental Research) used overnight polysomnography to compare ten sleep bruxism patients against ten matched controls, measuring brain wave activity (electroencephalogram, EEG) and heart rate in the seconds preceding each episode of rhythmic masticatory muscle activity — the technical term for the jaw-muscle bursts that produce grinding. EEG activity increased four seconds before the jaw muscles contracted in 79% of episodes. Heart rate accelerated one cardiac cycle before muscle activity. The authors’ own framing placed the grinding downstream of the arousal: bruxism, they concluded, is "a powerful oromotor manifestation secondary to micro-arousal."

The 2006 Huynh follow-up (Journal of Sleep Research) used a three-group comparison of moderate-to-severe bruxers, light bruxers, and non-bruxer controls. In the moderate-to-severe group, the autonomic balance shifted toward the sympathetic — the fight-or-flight branch — approximately eight minutes before bruxism onset, with the cross-correlation reaching statistical significance at P ≤ 0.03. The shift was specific to that group, time-locked to the episodes, and concentrated in the second and third sleep cycles of the night, with both arousal and grinding indexes rising before each REM transition.

Evidence boundary. Two small independent polysomnographic studies — ten patients per group in Kato, twenty per group in Huynh — found the same sequence: autonomic shift, cortical arousal, heart-rate change, then jaw-muscle activity; whether that upstream shift causes the grinding or only predicts it has not been tested experimentally.

The 2020 Aarab trial (Journal of Clinical Sleep Medicine) studied 18 patients with OSA (mean AHI 22.0 ± 16.0 events/hour) in a randomized crossover design with ambulatory polysomnography. With the mandibular advancement device in place, the respiratory arousal index fell significantly (Z = −2.896; P =.004), and jaw-closing muscle activity time-related to respiratory arousals fell significantly (Z = −3.434; P =.001). Jaw-closing muscle activity time-related to nonrespiratory arousals showed no significant change. Because this was measured in OSA patients rather than in patients with separately diagnosed sleep bruxism, the authors noted that confirmation in comorbid sleep bruxism is still needed.

The 2009 Landry-Schönbeck crossover study (International Journal of Prosthodontics) compared a mandibular advancement device against a mandibular occlusal splint in 12 patients with sleep bruxism. The advancement device reduced sleep bruxism episodes per hour by 39% at mild protrusion and 47% at pronounced protrusion from baseline, both reaching statistical significance (P <.04). The occlusal splint reduced episodes by 34%, which did not reach significance (P =.07).

Evidence boundary. A 2022 re-analysis of crossover data (Abe et al., Clinical Oral Investigations) showed that both mandibular advancement devices and occlusal splints reduced rhythmic masticatory muscle activity independently of measurable changes in heart rate variability — the autonomic signature presumed to mediate the earlier sequence. The clean dichotomy that the advancement device treats the airway cause while the night guard manages the downstream consequence is therefore a useful simplification, not a settled mechanism.

The 2018 Lobbezoo consensus (Journal of Oral Rehabilitation) updated the 2013 working definition of bruxism. Beyond separating sleep from awake bruxism and treating bruxism in otherwise healthy individuals as a behavior that can be a risk factor rather than a disorder, it concluded that standardized cutoff points for diagnosing the presence or absence of bruxism in healthy adults should not be used — masticatory muscle activity should be assessed across a continuum.

A 2024 narrative review of bruxism, temporomandibular disorders, and headache (Voß et al., Pain) confirmed that awake and sleep bruxism have different impacts on the pathophysiology of primary headache disorders. Awake bruxism correlated more strongly with tension-type headache; sleep bruxism’s contribution to headache was harder to disentangle from concurrent temporomandibular disorder pain. The authors emphasized that treatment of one condition may produce changes in the others without establishing causality, and called for greater consistency in diagnostic criteria.

Evidence boundary. The consensus changes how bruxism is assessed — on a continuum, separately for sleep and wake — rather than testing how any single treatment performs.

A long-standing hypothesis holds that sleep bruxism may have a protective function in the context of airway obstruction — that the jaw activity helps restore airway patency in patients with sleep-disordered breathing.

A 2024 case-control polysomnographic study (Cid-Verdejo et al., Sleep Medicine) examined 37 subjects with and without OSA. Sleep bruxism was associated with OSA at an odds ratio of 0.15 (95% CI 0.036-0.68), and OSA patients had fewer average sleep bruxism episodes than non-OSA patients. Sleep bruxism episodes correlated negatively with the apnea-hypopnea index. The authors interpreted these findings as consistent with a protective role for sleep bruxism in subclinical and mild OSA.

A 2025 polysomnographic study of 241 patients with mild and moderate OSA (Przegrałek et al., Journal of Sleep Research) reached a different conclusion. Bruxers had longer duration of hypopnea and apnea+hypopnea events than non-bruxers (P =.001 and P =.004 at the BEI ≥ 2 threshold) — the opposite of what the protective hypothesis predicts. The authors concluded that any protective effect of sleep bruxism may be limited to milder airway obstruction or may diminish over time.

Evidence boundary. The relationship between sleep bruxism and respiratory events is not unidirectional and is not consistent across studies; it varies at the individual level, and whether bruxism is protective, neutral, or additive depends on the patient in front of you. Clinicians who treat bruxism as airway-protective for every patient are extrapolating beyond what the data support.

The 2018 Melo systematic review (Journal of Oral Rehabilitation) followed PRISMA methodology. Five cross-sectional studies met inclusion criteria. Sleep bruxism was diagnosed by questionnaires or clinical examination — not by polysomnography, which is the reference standard for sleep bruxism diagnosis. The risk of bias across the included studies was mixed: one low, three moderate, one high. The GRADE quality of evidence for the overall body of evidence was rated "very low."

The agents associated with increased odds of sleep bruxism (adjusted ORs vs. non-users): paroxetine 3.63 (95% CI 2.15-6.13), venlafaxine 2.28 (95% CI 1.34-3.86), duloxetine 2.16 (95% CI 1.12-4.17), and methylphenidate in adolescents 1.67 (95% CI 1.03-2.68). The agents not associated with increased odds: citalopram, escitalopram, fluoxetine, mirtazapine, and sertraline.

A 2021 narrative review (de Baat et al., Journal of Oral Rehabilitation) extended the landscape to medications and addictive substances potentially inducing or attenuating sleep and awake bruxism. The authors concluded that the evidence base on both sides of the ledger is insufficient for definitive conclusions, including the literature on medications that may reduce bruxism, which is smaller than the inducing-side literature.

Evidence boundary. Melo and colleagues explicitly recommended caution given the very low overall quality of evidence. The signal is consistent across the agents that showed an effect — and informatively absent for the agents that did not — but the evidence base is observational, cross-sectional, and assembled from non-polysomnographic bruxism diagnoses. Causal direction has not been established with interventional data. A 2023 analysis of the World Health Organization's pharmacovigilance database (Montastruc, British Journal of Clinical Pharmacology) complicates the agent-by-agent picture in two ways. First, the bruxism signal tracked more closely with a drug's norepinephrine-transporter affinity than its serotonin-transporter affinity — consistent with why paroxetine, which carries unusual norepinephrine activity for an SSRI, and the SNRIs show the signal while purer serotonergic agents do not. Second, in that database sertraline carried one of the highest bruxism reporting rates — the opposite of its null result in the cross-sectional review above. Pharmacovigilance and cross-sectional designs answer different questions; the clean "these agents yes, those no" list is less settled than it looks. The clinical translation: the medication-bruxism timeline belongs in the differential diagnostic conversation, not that any specific medication should be discontinued on the basis of this review alone.

Two 2024 meta-analyses on BoNT-A for temporomandibular disorders reached opposite conclusions.

A 2024 meta-analysis of 15 randomized controlled trials of BoNT-A for muscular TMD specifically (Li et al., Journal of Oral Rehabilitation) found BoNT-A significantly more effective than placebo for muscular TMD pain at one month (mean difference −1.92, 95% CI −2.87 to −0.98) and at six months (MD −2.08, 95% CI −3.19 to −0.98). Optimal dose appeared to be 60-100 units bilaterally.

A 2024 meta-analysis of 14 RCTs of BoNT-A for TMD more broadly (Saini, Ali Abdullah Almoyad, et al., PLoS One) concluded that BoNT-A was not significantly better than placebo for pain reduction, describing their findings as "statistically insignificant for the BTX group." The abstract's own interval estimates are not fully consistent with that conclusion: those reported at one month (MD −1.71; 95% CI −2.87 to −0.5) and three months (MD −1.53; 95% CI −2.80 to −0.27) do not cross zero, while the six-month interval does (MD −1.33; 95% CI −2.74 to 0.77).

Evidence boundary. In muscular TMD specifically, pooled randomized trials show significant pain reduction at one and six months; across mixed TMD, Saini and colleagues judged the overall effect non-significant. The two analyses differ in three documented ways: (a) the underlying RCT population (muscular TMD specifically versus mixed TMD), (b) dose and injection-site variation across trials, and (c) follow-up duration. A longer-term concern compounds the short-term uncertainty: repeated BoNT-A injection of the masseter over years has been associated with disuse osteopenia of the mandibular condyle and alveolar bone — the 2021 de Baat review names this as an emerging concern. Measurable short-term pain relief has to be weighed against the bone biology that supports the same dentition decades later.

Ahmad et al. (Journal of Dental Research 2018; 97[9]:1023-1030) showed the molecular pathway in primary TMJ fibrochondrocytes. 17β-estradiol exposure upregulated matrix metalloproteinase-9 (MMP-9) and MMP-13 — the protein-degrading enzymes most directly implicated in cartilage extracellular matrix breakdown. The effect was mediated specifically through estrogen receptor alpha (ER-α). Selective ER-α antagonism abolished the MMP induction; selective ER-β blockade did not. The downstream signaling extended through ERK activation, NF-κB, and ELK-1 — the canonical inflammation-coupled pathways that connect hormone receptor activation to enzymatic matrix degradation.

Evidence boundary. The Ahmad 2018 study was conducted in primary mouse TMJ fibrochondrocytes in cell culture. The mechanism is well-established at the molecular and cellular level. Extrapolation to human TMJ degeneration in vivo is supported by the parallel sex and reproductive-stage patterns in the population evidence, but is not the same as a clinical trial of estrogen modulation in TMD. It is a candidate biological explanation for a robust clinical pattern, consistent with but not yet proven by the human evidence.

A 2020 systematic review and meta-analysis of observational studies (Cuenca-Martínez et al., Journal of Clinical Medicine, La Salle University Madrid) examined the relationship between craniocervical features and TMD across imaging, range-of-motion, and pain-threshold designs. The pooled analysis found a clinically relevant association between cervical and mandibular disability (SMD 0.72), moderate evidence for reduced cervical range of motion in TMD patients, and lower pressure pain thresholds in cervical musculature. The same meta-analysis found no significant difference in cervical alignment (SMD 0.02) or craniocervical position (SMD −0.09) between TMD patients and controls. Functional findings replicate; postural findings do not.

A 2021 controlled rehabilitation trial (Oleksy et al., BioMed Research International) studied patients with idiopathic neck pain who did not report TMJ complaints. A three-week cervical-only rehabilitation programme reduced forward-head-posture prevalence from 76% to 28% and produced simultaneous improvement in cervical range of motion and TMJ clinical condition. Cervical intervention modified jaw function in patients who had not presented for jaw care.

What this means. The evidence supports a functional jaw-neck relationship, not a simple postural-cause model. Treatment aimed at head posture as the cause of TMD is not supported; cervical rehabilitation, however, has improved jaw function, in patients with neck pain rather than TMJ complaints.

Observational data show that systemic autoimmune disease and TMD co-occur, but not which drives which. A 2024 Mendelian randomization study (Chen et al., Frontiers in Immunology, Jiangyin People’s Hospital) used genetic instruments from genome-wide association studies covering fourteen autoimmune diseases to test causal direction. Genetically predicted rheumatoid arthritis raised the risk of TMD (OR 1.12, 95% CI 1.05–1.19, P < 0.001). Multiple sclerosis raised TMD risk (OR 1.06, 95% CI 1.03–1.10, P = 0.001). Two immune-cell phenotypes were identified as mediating the RA→TMD pathway, and one mediated the MS→TMD pathway.

A 2025 Mendelian randomization study (Gu et al., Medicine (Baltimore), Anhui Medical University) extended the finding set. Rheumatoid arthritis (OR 1.087), multiple sclerosis (OR 1.081), and ankylosing spondylitis (OR 1.355) each genetically increased TMD risk. A reverse causal connection was also detected — TMD genetically increased rheumatoid arthritis risk (OR 1.099) — suggesting the relationship may be bidirectional at the population level.

What this means. Two independent Mendelian-randomization analyses found genetically predicted rheumatoid arthritis and multiple sclerosis associated with higher TMD odds, with ankylosing spondylitis added in the second analysis. The effects are modest — odds ratios near 1.1, and 1.36 for ankylosing spondylitis — and the reverse TMD-to-RA signal rests on one study.

Häggman-Henrikson et al. (Journal of Oral Rehabilitation 2024; 51[1]:202-217) performed a pairwise random-effects meta-analysis pooling 27 studies that reported on implant failure in patients with "probable" bruxism (diagnosed by combined self-report and clinical examination per the 2018 Lobbezoo consensus). The pooled analysis covered 2,105 implants in probable bruxers (138 failures) against 10,264 implants in non-bruxers (352 failures). The odds ratio for implant failure in bruxers was 2.189 (95% CI 1.337-3.583; P =.002). Meta-regression showed that follow-up time did not modify the effect — the risk persisted across the duration of observation. Marginal bone loss data were available from 18 studies but were not separable by bruxism status; this is a limitation the authors named explicitly.

Chrcanovic et al. (Scientific Reports 2025; 15[1]:22752) followed 30 probable bruxers and 21 non-bruxers, all rehabilitated with implant-supported restorations, in a long-term clinical examination protocol. Bruxers had significantly higher maximum bite force (P =.023) and a significantly higher proportion of tooth and veneer porcelain fractures relative to total tooth or prosthetic unit count (P =.045). The relationship between fracture frequency and bite force did not differ significantly between bruxers and non-bruxers (P =.054) — the higher bite force, rather than a separate vulnerability, appears to drive most of the additional fracture burden.

What this means. Bruxism roughly doubled implant-failure odds across 27 observational studies (OR 2.19), with bite force the apparent driver of prosthetic fracture; whether treating bruxism first lowers failure has not been tested prospectively.

The short answer: the mechanistic disruption is real and consistent, but the acute blood pressure signal in healthy young adults is small and not yet statistically settled.

The main counter-evidence comes from a 2024 systematic review and meta-analysis with trial sequential analysis by Farook and colleagues at King Saud Bin Abdul Aziz University for Health Sciences, in the International Journal of Dental Hygiene. Pooling five controlled crossover trials, the weighted change in systolic blood pressure was 1.59 mmHg (95% CI –0.15 to 3.33), and in diastolic 0.46 mmHg (95% CI –0.72 to 1.64). Neither reached statistical significance, and the trial sequential analysis was explicit that the data do not yet conclusively support blood pressure elevation from mouth rinse use. Toonen and colleagues (International Journal of Dental Hygiene, 2026), restricting the analysis to chlorhexidine, found a similar non-significant difference (1.56 mmHg; 95% CI −0.67 to 3.80) and judged any blood-pressure change for patients prescribed chlorhexidine to be very small.

The case against routine daily chlorhexidine rests instead on findings Farook does not contest: the acute disruption of oral nitrite, plasma nitrite, and microbial composition in every pooled trial, the longer-horizon Joshipura 2017 diabetes signal, and the Clarke 2019 exercise finding. A 2024 randomized trial by Bescos and colleagues in the Journal of Oral Microbiology added a useful nuance: chlorhexidine raised the relative abundance of nitrite-producing species while impairing their activity. Abundance and activity are not the same thing.

What this means. The longer-term and exercise-related outcomes are arguably more clinically relevant than resting blood pressure in a young, healthy sample.

The short answer: no. The food matrix matters, and an isolated nitrate salt does not reliably reproduce what whole beetroot does.

The cardiovascular benefit attributed to dietary nitrate shows up most consistently in trials of nitrate-rich whole foods — beetroot juice, leafy greens, beetroot extract. When nitrate is delivered as an isolated salt, the picture changes. A 2025 Nutrition Research meta-analysis by Forster, Waclawovsky, and Stefani pooled six randomized trials of chronic sodium nitrate supplementation in adults over 50 (181 participants; mean PEDro risk-of-bias score 8.16, indicating low risk of bias). The pooled effect on systolic blood pressure was –3.81 mmHg, with a confidence interval that crossed zero (–10.05 to 2.43); diastolic blood pressure and resting heart rate were similarly non-significant.

Vegetable-sourced nitrate arrives inside a food matrix — polyphenols, vitamin C, potassium, betalains — that guards against the formation of harmful N-nitroso compounds and may support endothelial nitric oxide through routes independent of the enterosalivary pathway. The pathway this chapter describes is necessary for nitrate's nitrite-derived contribution to nitric oxide. It is not sufficient to explain everything whole beetroot does for the cardiovascular system.

The historical safety concern — methemoglobinemia — is negligible at dietary intakes in healthy adults. A 2025 case report by Kono and colleagues in Medicine documented a 60-year-old woman who reached a methemoglobin level of 25 percent from excessive leafy-green consumption on an unbalanced diet — a documented but rare pattern. The infant susceptibility that drove the original regulatory caution reflects developmental differences in an enzyme (cytochrome b5 reductase) and does not apply to the adult audience of this book (Iolascon 2021).

What this means. The old safety concern is largely a concern about the wrong sources, a thread the cancer chapter picks up in the cancer context.

The short answer: estradiol drives the classical nitric oxide pathway, so when estradiol falls, the bacterial pathway carries relatively more of the load — just as its own capacity is contracting.

Estradiol increases transcription of endothelial nitric oxide synthase, increases its activating phosphorylation, and raises the availability of BH4 — the cofactor whose depletion uncouples the enzyme into a reactive-oxygen-species producer. The withdrawal of estradiol at menopause lowers NOS-dependent nitric oxide through all three mechanisms simultaneously.

A 2025 randomized controlled trial by Alvares and colleagues in the American Journal of Physiology — Heart and Circulatory Physiology: twelve weeks of nitrate-rich beetroot extract in postmenopausal women aged 60 to 85 improved femoral artery flow-mediated dilation, raised microvascular reactivity in the tibialis anterior at both eight and twelve weeks, and produced serum that drove 1.8-fold greater endothelial tube formation in an angiogenesis assay. Three independent vascular readouts moved in the same direction, in a sample of 16. Two crossover trials add the dynamics: Benjamim and colleagues (2023) showed that in inactive postmenopausal women with hypertension, beetroot juice lowered post-exercise systolic blood pressure by 9.28 mmHg versus placebo and raised flow-mediated dilation at rest and after exercise; Delgado Spicuzza and colleagues at Penn State (2024) found the protection in early-postmenopausal women was lost within 24 hours of the last dose, and that acute beetroot improved macrovascular function in late- but not early-postmenopausal women. A 2025 meta-analysis (Raimundo et al.) of ten beetroot trials in postmenopausal women confirmed an autonomic-recovery signal, with one heart-rate-variability index (RMSSD) improved by 6.68 ms and the others null.

What this means. In postmenopausal women the pathway becomes a larger relative contributor to vascular health, it remains responsive to dietary nitrate, and the protection depends on continued intake. No trial has yet been powered for cardiovascular event reduction in this population — but the mechanism, the surrogate endpoints, and the autonomic signal all point the same way.

The short answer: recovery, not performance. The cardiovascular response to exercise improves; race times largely do not.

The recovery half is robust. A 2024 meta-analysis by Benjamim and colleagues in Free Radical Biology and Medicine pooled 26 randomized trials and found that nitrate blunted the rise in systolic blood pressure during exercise by 2.81 mmHg (95% CI –5.20 to –0.41) and amplified the post-exercise drop by 3.53 mmHg (95% CI –5.65 to –1.41). This is the meta-analytic replication of Clarke 2019: dietary nitrate consistently improves the cardiovascular response to acute exercise.

The performance half does not hold up. A 2023 systematic review by Hogwood and colleagues in Sports Medicine — Open pooled eleven trials of chronic nitrate combined with at least three weeks of training. The effect on VO₂max was a standardized mean difference of 0.18 (95% CI –0.09 to 0.44), and on time to exhaustion 0.08 (95% CI –0.21 to 0.37) — both confidence intervals crossing zero. Chronic nitrate did not amplify training adaptations beyond the training itself.

What this means. The operative benefit of dietary nitrate in exercise is cardiovascular recovery — post-exercise hypotension, flow-mediated dilation, parasympathetic reactivation — not race time.

The short answer: the observational signal is consistent and source-specific, but randomized trials have not confirmed it. This is a replicated association, not a proven mechanism.

Four large cohorts point the same way. The Rotterdam Study followed 9,543 people for a median of 14.5 years and linked higher vegetable-derived nitrate to lower dementia risk. A 2025 analysis of the Danish Diet, Cancer and Health cohort — 54,804 participants over roughly 27 years — confirmed it, with a stronger signal for early-onset dementia, while animal- and tap-water-sourced nitrate ran the opposite way. The Australian Imaging, Biomarkers and Lifestyle Study found the association strongest in APOE ε4 carriers. And a 2024 analysis of the Australian Diabetes, Obesity and Lifestyle Study by Rajendra and colleagues (9,149 adults, 17 years) found the highest plant-source nitrate intake carried a 57 percent lower risk of dementia-related mortality — 66 percent for vegetable-sourced nitrate alone — while processed-meat-source nitrate doubled it. A 2024 source-stratification of the Danish cohort by Bondonno and colleagues reported the same split for all-cause mortality: plant-sourced nitrate protective, animal-, processed-meat-, and tap-water-sourced nitrate harmful.

A 2022 review by Shannon and colleagues noted that randomized trials of nitrate supplementation on cognitive function have produced predominantly null results, and that interventional confirmation has not been established.

What this means. Across large European and Australian cohorts, higher vegetable-sourced nitrate intake tracks with lower dementia risk, while source-stratified analyses point the opposite way for animal- and processed-meat-sourced nitrate. Randomized cognition trials have not yet confirmed a benefit.

A 2025 review in Microorganisms by Riyoko Tamai and Yusuke Kiyoura at Ohu University School of Dentistry frames saliva as the body's first-line constraint on Candida albicans, with the histatin family and β-defensins identified as the dominant antifungal effectors and salivary mucins as the primary adhesion barrier (Tamai 2025). A 2025 review in Advances in Experimental Medicine and Biology described saliva as a "guardian of oral microbial balance" whose antimicrobial proteins and antibodies are crucial for preventing oral infections and maintaining ecological equilibrium of the resident microbiota. The clinical reality these reviews map is quantitative. Normal unstimulated whole saliva flow is approximately 0.3 to 0.5 mL/min. Salivary gland hypofunction is diagnosed at 0.1 mL/min or below — a reduction of roughly two-thirds to four-fifths from the normal unstimulated range of 0.3–0.5 mL/min. A 2018 meta-analysis of 29 population-based studies reported a pooled xerostomia prevalence of 23.0 percent (95% CI 18.0 to 28.0), higher, at 27.2 percent, in studies of older people only (Agostini and colleagues, Brazilian Dental Journal, 2018).

Evidence boundary. The salivary protein composition data is robust at the molecular level; the population-prevalence estimates rest on self-reported xerostomia, which is a subjective measure that does not perfectly track sialometric hyposalivation. The relationship is directional but imprecise.

The magnitude and time course were mapped precisely by Dawes and Macpherson (Caries Research, 1992), who tracked salivary flow in 24 adults across nine gums and lozenges. Flow peaked in the first minute at roughly 10 to 12 times the unstimulated rate (baseline 0.47 mL/min), then fell along a power curve; within about ten minutes it was indistinguishable from chewing flavorless gum base, and by twenty minutes had settled to about 2.7 times resting flow. Critically, sugar-containing and sugar-free gums produced the same flow response. The primary driver was the act of chewing rather than the sweetener itself. Aguirre-Zero and colleagues (Caries Research, 1993) confirmed it directly: across a crossover of no gum, sucrose, sorbitol, and xylitol gum in ten adults, no regimen significantly changed salivary flow rate (xylitol's distinct effect was on plaque pH, not flow). The Cochrane review of topical dry-mouth therapies (Furness and colleagues, 2011; 36 trials, 1,597 participants, only one at low risk of bias) found that chewing gum increases saliva in people with residual secretory capacity, but that no gum type outperforms another or the saliva substitutes — and that no topical therapy reliably relieves the symptom. The ISOO/MASCC/ASCO guideline (Mercadante and colleagues, Journal of Clinical Oncology, 2021) lists sugar-free gum and lozenges among reflex stimulants for established hypofunction, reserving pilocarpine and cevimeline for radiation-damaged glands — stimulation presupposes surviving tissue.

Evidence boundary. The flow response to gum is real but modest, transient, and contingent on functioning glands; the sweetener is not the active ingredient.

For patients taking multiple medications, anticholinergic burden is a common, cumulative driver of salivary dysfunction that single-drug review can miss. The mechanism is muscarinic-receptor blockade on salivary acinar cells. Validated scales (the Anticholinergic Cognitive Burden Scale, the Anticholinergic Drug Scale, the Anticholinergic Risk Scale) score individual medications and sum the load across a regimen. The 2023 Michail study (Archives of Oral Biology) measured serum anticholinergic activity directly in 55 middle-aged adults and showed a negative linear relationship with unstimulated whole salivary flow, with each pmol/mL increase in serum anticholinergic activity reducing flow by 0.058 mL/min (Michail 2023). A 2023 study of 321 community-dwelling older adults (mean age 81.6 years) found high anticholinergic burden was associated with hyposalivation more strongly than with subjective xerostomia, with relative risks of 1.47 to 2.07 for unstimulated and stimulated flow reduction across multiple scoring scales (Stenbäck et al., Acta Odontologica Scandinavica, 2023). A 2025 preliminary study identified a CYP2D6 rs28371725 polymorphism associated with severe hyposalivation at a genotype-level relative risk of 12.75, though with very wide confidence intervals (1.45–112.12) from a single-study cohort — a hypothesis-level pharmacogenetic signal that may explain why some patients on modest anticholinergic loads experience disproportionate dryness (Korczeniewska et al., Frontiers in Pharmacology, 2025).

Evidence boundary. The Michail and Stenbäck findings are direct clinical observations; the Korczeniewska pharmacogenetic signal is preliminary and from a single study.

The Qu 2016 review in Journal of Dental Research identified sialin as the molecular basis for active nitrate transport into the salivary glands. At the molecular level, sialin sits on the plasma membrane of salivary acinar cells and functions as an electrogenic 2NO₃⁻/H⁺ transporter: it moves two nitrate ions into the acinar cell along with one proton, against the concentration gradient, using the proton gradient as the energy source. Up to 25% of circulating nitrate is taken up by the gland and concentrated 10- to 20-fold in saliva (Qu 2016). The species gap was measured empirically at Karolinska Institutet (Montenegro and colleagues, Redox Biology, 2016, with Sundqvist, Nihlén, Hezel, Carlström, Weitzberg, and Lundberg): after standardized nitrate administration (0.1 mmol/kg), plasma nitrate reached approximately 300 μM in humans, rats, and mice — but salivary nitrate rose to over 6,000 μM in humans (a 20-fold concentration step) while in rats and mice it never exceeded plasma levels. The species difference matters for translational research: most rodent studies of dietary nitrate and the nitric oxide pathway bypass the salivary concentration step that defines the system in humans.

Evidence boundary. This is one of the more striking translational-physiology gaps in the field. It does not invalidate rodent mechanistic work on downstream nitric oxide signaling, but it does mean that any rodent finding that depends on salivary concentration cannot be extrapolated to humans without empirical verification — and the converse is also true. The human data on salivary nitrate kinetics, by contrast, is direct and robust.

In a 2023 study of 117 adults — 32 healthy controls, 49 with gingivitis, and 36 with periodontitis (Lee et al., Frontiers in Endocrinology) — a multivariable logistic regression modeled the odds of above-average salivary cortisol. Compared with healthy controls, periodontitis carried an odds ratio of 256.8, but with a 95% confidence interval running from roughly 26 to 2,556. Female sex (OR 6.4) and self-reported psychological stress (OR 6.0) were also independent predictors.

Evidence boundary. The point estimate should not be read as a stable effect size. A confidence interval spanning two orders of magnitude, a small sample, an outcome dichotomized at the sample-average cortisol level (a statistical threshold, not a clinical cortisol cutoff), and group cortisol means that barely separate between periodontitis (278 ng/mL) and healthy controls (127 ng/mL) are consistent with sparse-data or near-separation instability — the setting in which logistic models return implausibly large, unstable odds ratios. Two further limits matter: the same model found no significant difference for periodontitis versus gingivitis (OR 0.55), so the extreme value is specific to the periodontitis-versus-healthy-control contrast; and because this is an observational, cross-sectional comparison, it cannot establish that periodontitis caused the elevated cortisol. What survives is direction, not magnitude — periodontitis, female sex, and psychological stress each tracked with higher salivary cortisol, consistent with an inflammation-to-HPA-axis-to-cortisol loop.

Salivary cortisol is one of the most extensively validated applications of saliva diagnostics. LC-MS/MS validation studies report salivary–serum correlations in the range of r = 0.82 for cortisol and r = 0.90 or higher for cortisone across multiple assays. The Endocrine Society and the American Association of Clinical Endocrinology both endorse late-night salivary cortisol as a first-line screening test for Cushing's syndrome. For research applications, the cortisol awakening response and the full diurnal curve are workhorses.

Evidence boundary. Most consumer-marketed "adrenal fatigue" or "cortisol panel" testing extends salivary cortisol well beyond its validated clinical use. The instrument is robust for specific endocrine questions (Cushing's screening, HPA reactivity research, sleep-disorder workup). It is not robust for diagnosing a free-floating fatigue syndrome, and the framework of "adrenal fatigue" itself has been explicitly rejected by the Endocrine Society as unsupported by physiologic evidence.

Periodontal inflammation status should be reviewed before interpreting any abnormal result, because periodontitis is associated with higher salivary cortisol and can confound interpretation.

The validation question for salivary cortisol is not whether the assay tracks free serum cortisol — it does — but whether the most consumer-marketed metric, the cortisol awakening response, reflects the HPA regulatory biology it is widely assumed to measure. A 2023 study by Abelson and colleagues in Psychoneuroendocrinology tested this question directly in 140 healthy adults: nine saliva samples per day across six days, paired against five standardized HPA regulatory tests including ACTH stimulation, dexamethasone/CRH, metyrapone challenge, dexamethasone suppression, and the Trier Social Stress Test (Abelson 2023). The diurnal cortisol decline was linked to feedback sensitivity, indexed by dexamethasone suppression. Morning cortisol levels were linked to adrenal sensitivity. The cortisol awakening response itself showed limited linkage to HPA regulatory biology — the paper "raises questions about the biological meaning of other curve components, including morning cortisol levels, and perhaps CAR."

Evidence boundary. This is one well-conducted study, not a definitive negative result. But it is the most rigorous attempt to map specific curve components to specific HPA regulatory mechanisms, and the CAR — the metric most heavily marketed to consumers — was the component with the least clear regulatory linkage.

The 2021 Bostanci study in Scientific Reports established that salivary MMP-8 in combination with MMP-9 and TIMP-1 exceeded 90% diagnostic accuracy for gingivitis and reached 100% accuracy for periodontitis versus health (Bostanci 2021). The 2025 Li meta-analysis in the Journal of Clinical Periodontology pooled 8 studies and 2,048 subjects and reported the full performance figures: sensitivity 0.59 (95% credible interval 0.42–0.75), specificity 0.82 (95% CrI 0.68–0.93), and hierarchical summary AUROC 0.77 (95% CrI 0.74–0.81), at moderate GRADE certainty (Li 2025). The 2025 Aji review (Dentistry Journal, with Sorsa and Räisänen at Helsinki and Karolinska) described the chairside aMMP-8 point-of-care test with a 20 ng/mL cut-off as optimized for active disease; with a pooled specificity near 0.82, false positives still occur — which matters because the test is increasingly proposed for patient-administered home use, not only for professional screening. The 2024 Xanthopoulou study (Clinical and Experimental Dental Research, with Räisänen, Sorsa, and Sakellari) extended the assay to peri-implant disease, with mean aMMP-8 of 13.65 ng/mL in healthy implants, 32.33 ng/mL in mucositis, and 73.07 ng/mL in peri-implantitis; AUC 0.814, sensitivity 71.7%, specificity 77.8%, and overall accuracy 77.5% above a 33.7 ng/mL cut-off (Xanthopoulou 2024).

Evidence boundary. AMMP-8 testing is positioned as adjunctive, not replacement, for clinical periodontal assessment. Its strength is detecting active tissue destruction that probing depth alone cannot identify in real time. Its limitations are that it does not localize the destruction to specific sites in the way that pocket-by-pocket probing does, and that it is not yet integrated into most medical record systems or insurance reimbursement codes in the United States. The European and Scandinavian uptake is further along than the American.

In a 2024 Gut randomized trial by Zhu and colleagues, forty-nine healthy volunteers were randomized between PPI (n=23) and H2RA (n=26) for seven days; saliva and stool were sequenced by metagenomic next-generation sequencing before and after intervention. PPI use produced significantly higher oral-to-gut microbial transmission than H2RA use, with measurable colonization of the gut by oral-origin species including disease-associated taxa. The machine learning classifier discriminating PPI users from non-users achieved AUROC 0.924; the equivalent H2RA classifier achieved 0.509, indistinguishable from chance (Zhu 2024). The taxonomic detail across the supporting evidence is consistent. The 2018 Mishiro study found increased Streptococcus in fecal samples and decreased Neisseria and Veillonella in the oral carriage of PPI users. The Macke 2020 systematic review found PPI use raised Streptococcaceae, Enterococcaceae, Lactobacillaceae, Veillonella, Pasteurellaceae, and Enterobacteriaceae while reducing Bifidobacteriaceae, Ruminococcaceae, and Lachnospiraceae — a pattern the 2024 Fossmark review confirmed across 13 experimental trials, centered on Streptococcaceae, Veillonellaceae, and related oral-derived families. Combined, the literature supports a coherent mechanism: pharmacologic hypochlorhydria removes the gastric-acid barrier that normally constrains oral microbes to the oral cavity.

Evidence boundary. The trials cited are short (7 days to 4 weeks) and conducted in healthy volunteers, not chronic PPI users in their tenth year of therapy. The cross-sectional observational data in long-term PPI users are consistent with the same pattern but cannot establish causation. There is no randomized trial of multi-year PPI exposure on long-term oral-to-gut transmission and downstream disease outcomes — that trial would not be ethical or feasible. The clinical translation has to integrate mechanistic trials, short-term RCTs, and observational long-term data into a single judgment, and that judgment is that PPIs as a class meaningfully reshape the upper-GI microbiome in a direction that favors disease-associated species.

The clinical and preclinical GLP-1 receptor agonist literatures point in opposite directions. The clinical reports — Mawardi 2023, the Khan 2025 FAERS analysis, the O'Driscoll 2026 review — describe xerostomia as a meaningful adverse-event signal. Mawardi's three-patient series documented severe dryness with a mean modified Schirmer value of 9 mL at three minutes; the Khan analysis drew on 9,746 adverse-event reports across five GLP-1 receptor agonists, with semaglutide signals spanning dry mouth, dysgeusia, anosmia, and Bell's palsy, plus a class-wide GERD signal. A 2025 study (Jung et al., European Journal of Pharmacology) reports the opposite directional finding in a preclinical aging-rat model: D-galactose-induced aging (300 mg/kg intraperitoneally daily for 7 weeks) produced age-related decline in salivary secretion, which was significantly reversed by Exendin-4 (a GLP-1 receptor agonist) at two doses over a 2-week treatment period — with preserved acinar cell morphology and reduced apoptotic cell death in salivary gland tissue (Jung 2025).

Evidence boundary. This is a rodent D-galactose-induced aging model, not a chronic-dosing study in healthy adult animals. The salivary gland protection observed in this specific model may not predict human xerostomia in chronic GLP-1 receptor agonist users at typical metabolic doses, where the dominant pharmacologic effect is the well-characterized delayed gastric emptying and centrally mediated appetite suppression that drives the metabolic effects.

What this means. At the cellular level, GLP-1 receptor signaling appears to be salivary-gland protective in an aging model; at the clinical level, dry mouth is a documented adverse-event signal for the drug class. The resolution — pending prospective human salivary gland imaging and functional data — favors the clinical signal, which is what shows up in practice. The preclinical signal warrants further investigation but does not yet justify reframing GLP-1 receptor agonists as salivary-protective in humans.

The overnight oral environment in numbers

The fall in salivary flow to negligible levels across sleep onset is not gradual and not modest, and the mechanism that produces it is specific. Salivary secretion is dominantly parasympathetic, driven by muscarinic acetylcholine signaling from cranial nerves VII and IX to the salivary acinar cells; the autonomic transition into sleep withdraws that drive almost completely. The negligible nighttime flow described in the Dawes 2008 review is therefore not a flow reduction in the usual sense — it is closer to a baseline secretory state with the active stimulus removed. The clinical implication is that the daytime measurement most clinicians take in the chair represents the waking state of parasympathetic drive, not the secretory ceiling under maximal drive and not the secretory floor the patient lives in for the eight hours after lights-out. Hydration, medication review, and the daytime sialometry from the saliva chapter each measure a state the convergence window does not occupy.

The magnitudes of the downstream consequences scale with the flow reduction but are not identical to it. Antimicrobial peptide concentrations at the salivary outlet drop in proportion to flow; their effective concentrations at the periodontal pocket and tongue dorsum depend additionally on the clearance, dilution, and degradation that occur between secretion and the tissue surfaces — most of which also slow with reduced flow. The buffering effect compounds the same way: bicarbonate from the parotid and submandibular glands falls in parallel with secretion, and the buffering capacity that holds the daytime mouth near pH 7.0 within minutes of any perturbation operates at a fraction of that capacity overnight. A mouth-breathing pattern that would acidify the resting mouth and resolve within minutes during the day can hold the oral environment at pH 6.5 or lower for hours during sleep, particularly in patients whose nasal airway has been compromised.

What the literature has not yet mapped with precision is the phase relationship between the circadian rhythms of immune cycling and the architecture of fragmented sleep. Neutrophil trafficking through the periodontal pocket and secretory IgA on the mucosal surfaces both follow circadian timing in healthy adults; whether the phase shifts in fragmented sleep are uniform across cell populations, whether the timing of immune arrival at the periodontal margin shifts in proportion to the sleep-architecture disruption, whether recovery of phase alignment lags or matches recovery of consolidated sleep — these remain incompletely mapped. The directional claim that fragmented sleep mistimes the oral immune response is well-supported. The precision mapping that would tell a clinician how much mistiming, which immune populations, and over what recovery interval is still developing. The mechanistic picture this chapter draws is robust at the level of direction and conservative at the level of quantification.

What this means. The overnight oral environment is a different physiological state than the daytime one, not simply a reduced version of it. Sialometry, medication review, and the chairside measurements clinicians take during the day reflect a state the convergence window does not occupy. The directional biology — lower flow, lower peptide delivery, lower buffering, mistimed immune arrival — is robust. The precision quantification of how much mistiming and over what recovery interval is still in development.

Carroll 2016 — what replicated and what did not

Carroll and colleagues at UCLA examined healthy older adults — a population with elevated baseline senescent-cell burden — restricting sleep for one night and drawing peripheral blood the following morning. The measurements included DNA damage response gene expression, the senescence-associated secretory phenotype gene expression profile, and p16INK4a — a cyclin-dependent kinase inhibitor whose accumulation in peripheral blood mononuclear cells indexes human cellular aging. The senescence marker p16INK4a rose on the initial microarray (p<0.01) but did not replicate on confirmatory RT-PCR in the same study, leaving the SASP gene expression signal (p<0.05) as the robust load-bearing finding. The DNA damage response signal reached trend level (p=0.08). The conclusion the authors drew — that a single night of partial sleep deprivation activates cellular senescence machinery measurably — is carried by the SASP gene expression result rather than by p16INK4a specifically. The distinction matters because p16INK4a is the more familiar senescence anchor in the longevity literature; what the Carroll study established with statistical robustness was the downstream secretory program, not the upstream cell-cycle-arrest marker. The molecular changes are likely reversible with recovery sleep; what the finding establishes is that the cellular machinery of aging is responsive at the scale of single nights rather than years. The cumulative burden over the years of an untreated pattern of disrupted sleep is what the inflammaging chapter develops as inflammaging.

What this means. The cellular machinery of aging responds to sleep at the scale of single nights, not only across decades. The senescence-associated secretory phenotype — the inflammatory output of aging cells that the inflammaging chapter develops as the engine of inflammaging — is the load-bearing signal here, not the more familiar p16INK4a marker. A single night’s changes are likely reversible; the cumulative cost of a chronic pattern of disrupted sleep is what shows up over time in the tissues the secretory program has been irritating.

Social jet lag — the timing-vs-duration mechanism

The biological circadian rhythm is governed by the suprachiasmatic nucleus of the hypothalamus, which entrains peripheral clocks in nearly every tissue of the body through hormonal signaling — most prominently cortisol, melatonin, and the autonomic nervous system. The peripheral clocks of immune cells, salivary glands, gingival fibroblasts, and the endothelium each carry cell-autonomous oscillators that depend on the central clock for phase information. When sleep timing shifts by two or more hours between weekdays and weekends, the peripheral clocks receive misaligned phase signals from sleep-wake behavior and from the cortisol-melatonin axis. Misalignment is associated with measurable elevations in inflammatory cytokines independent of total sleep duration — the Alqaderi 2023 finding in adolescents, the Faraut 2022 finding in shift workers, and a developing literature on workplace and lifestyle-driven circadian disruption converging on the same pattern. The mechanism appears to operate through both reduced melatonin amplitude — Carra and colleagues 2024 Periodontology 2000 documented melatonin's role as a free-radical scavenger and anti-inflammatory modulator in periodontal tissue specifically — and through phase-shifted cortisol release that elevates baseline inflammatory tone. The translational implication is that the sleep-related pattern is broader than the obstructive sleep apnea phenotype the sleep-medicine literature has historically emphasized. Sleep timing irregularity is an oral-inflammatory input even in the patient whose airway and sleep duration appear adequate.

What this means. Sleep timing irregularity is an inflammatory input in its own right, distinct from inadequate sleep duration. The patient whose total hours are sufficient but whose weekday-weekend schedules differ substantially is still receiving the input. A sleep-related oral-inflammatory finding in such a patient is not anomalous; it reflects a biology that operates through circadian misalignment whether or not the airway is involved.

The cellular mechanism, the TwinsUK signature, and the by-system evidence

The cellular mechanisms of mercury toxicity are well characterized. Mercuric chloride at subcytotoxic concentrations dose-dependently increases the release of pro-inflammatory cytokines IL-1β and TNF-α from human peripheral blood mononuclear cells while simultaneously suppressing IL-1Ra and IL-10 — the cytokine-balance shift that maps directly onto the unopposed-inflammation phenotype of inflammaging. At the mitochondrial level, mercury depolarizes the inner mitochondrial membrane, increases hydrogen peroxide production, depletes intracellular glutathione by approximately 50 percent, and increases lipid peroxidation by approximately 68 percent. These are the same mitochondrial-stress signals that activate the NLRP3 inflammasome, release mitochondrial DNA into the cytosol where it triggers cGAS–STING and IL-1β maturation, and contribute to the inflammaging loop developed in the inflammaging chapter. The TwinsUK study by van de Streek and colleagues (2025) in 800-plus female twins identified 873 differentially expressed genes in skin tissue associated with subtoxic plasma mercury, with the strongest functional enrichment in mitochondrial oxidative phosphorylation (P = 3.02 × 10⁻¹⁰), including mitochondrial DNA-encoded oxidative-phosphorylation genes such as MT-ND4L and MT-ND5. The caveat: associations were tissue-specific. Whole blood and adipose tissue showed no signal, and lead and selenium were null in any tissue. The authors interpret skin as a tissue both directly exposed to the environment and metabolically active, showing the clearest signature of mercury's transcriptional effects.

By body system, the epidemiology the narrative compresses runs as follows. Kidney: a pediatric study found urinary NAG, a tubular-damage marker, rising dose-responsively with urinary mercury from a threshold near 1.45 µg/g creatinine; a Geier re-analysis of the Casa Pia amalgam trial reported a dose-dependent rise in GST-α that the original Woods analysis did not — real under that model, but a contested re-analysis rather than independent replication; the New England Children’s Amalgam Trial found roughly doubled microalbuminuria (adjusted OR 1.8); and Zuo and colleagues’ 2025 outpatient cohort (22,670 of 58,864 German outpatients) found blood mercury inversely correlated with eGFR with no identifiable safe threshold. Liver: in that same 58,864-outpatient cohort mercury was the strongest metal correlate of ALT elevation, with a 560-participant panel study echoing it across ALT, AST, and GGT. Cardiovascular: occupational meta-analyses run dose-related — Hu and colleagues’ 2021 pooled estimate (about 34,000 participants) found a 68 percent rise in cardiovascular mortality — while a 348,000-participant dietary-methylmercury meta-analysis found no distinctive signal, the discrepancy turning on mercury species and exposure magnitude. Thyroid: an eighteen-study meta-analysis found blood mercury raises TSH, strongest in children. Lung: Miao and colleagues (2023, roughly 1,800 students) found each doubling of blood mercury associated with measurable decrements in FVC, FEV1, and peak flow. Reproductive: a 2019 systematic review associated mercury with subfertility, sperm DNA damage, and adverse reproductive outcomes.

The evidence is mechanistically consistent at the cellular level and mixed at the epidemiologic level; the clinical question is what mercury contributes in a specific patient with a specific cumulative burden.

Exposure, dose, and oral methylation. The quantitative exposure figures the narrative compresses run as follows. It has been estimated that ten amalgam surfaces raise urinary mercury concentrations by approximately 1 μg/L, roughly doubling background levels. The daily inhaled dose from a typical restoration burden is roughly one to five micrograms, with higher estimates — up to seventeen micrograms a day — in patients with many surfaces and heavy occlusal wear from bruxism or frequent gum chewing. The World Health Organization's provisional tolerable weekly intake for inorganic mercury works out to about forty micrograms a day for a seventy-kilogram adult. A 2022 analysis by Geier and Geier of the NHANES dataset, representing 158 million American adults, found that roughly half exceed the California EPA mercury-vapor safety limit from amalgam and about one in ten exceed the less restrictive US EPA limit — a nationally representative federal survey, but an interpretation from the same investigators behind the contested Casa Pia re-analysis, controversial within mainstream toxicology. On the organic-mercury question: in 1983, Heintze and colleagues at the University of Lund demonstrated in vitro that three common oral streptococcal species — Streptococcus mitior, S. mutans, and S. sanguis — methylate mercury from dental amalgam, methylmercury appearing in the bacterial cells when mercuric chloride or pulverized amalgam was the source. The 2019 integrated review by Jirau-Colón and colleagues identifies organic-mercury formation from amalgam in the oral cavity as one of four core issues in the dental amalgam dilemma — alongside constant vapor release, gene-regulation effects, and the epidemiological associations with Alzheimer's and Parkinson's. The 1983 demonstration is in vitro; contemporary work has not extended it at population scale, so the in vivo magnitude remains the open question while the mechanism is established.

What this means. Mercury at amalgam-level exposures is not a question of catastrophic poisoning; it is a question of continuous low-grade molecular stress on the same systems — inflammasome activation, mitochondrial impairment, glutathione depletion — that the inflammaging chapter develops as the engines of inflammaging. The TwinsUK skin signature is real, the cellular mechanisms are characterized, and the unanswered question is the magnitude of the systemic contribution in any given patient, not whether the contribution exists.

The randomized trials, Cochrane synthesis, and direct brain measurement

The New England Children's Amalgam Trial (NECAT) and the Casa Pia Children's Dental Amalgam Trial, both published in the Journal of the American Medical Association (JAMA) in 2006, are the two randomized controlled trials designed to test amalgam safety in children. Together they enrolled more than 1,000 children randomized to amalgam versus composite restorations, followed for five to seven years, with neurobehavioral, cognitive, and biomarker assessment. Their primary conclusions were the same: no statistically significant differences in neurobehavioral outcomes between groups. The 2021 Cochrane systematic review by Worthington and colleagues synthesized this evidence and concluded that there is "low to very low certainty evidence" of any neurobehavioral difference between amalgam and resin composite, and that the evidence base is "insufficient to confirm or refute a link between dental amalgam and adverse health effects." The conclusion cuts two ways: low to very low certainty is not a declaration that amalgam is safe; it is a declaration that the evidence is insufficient to establish safety. Cochrane also noted that no studies assessed effects in adults and that follow-up was limited to five to seven years. For a book whose central claim is that cumulative multi-decade burden matters, the absence of long-term adult randomized data is not a counterargument. It is the argument. Direct pathological measurement tells a more conservative story than mechanistic literature would predict. A 2016 JAMA study by Morris and colleagues measured brain mercury directly in 203 deceased participants and found no significant correlation between brain mercury concentrations and Alzheimer's disease neuropathology, with no significant interaction by ApoE ε4 status; unexpectedly, brain mercury was significantly associated with lower odds of macroinfarcts and microinfarcts. A 2019 cross-reference review of mercury's effects against 70 factors associated with Alzheimer's reported that all 70 could be attributed to mercury toxicity, including amyloid pathology, neurofibrillary tangles, and phosphorylated tau. Cross-referencing a broad-spectrum toxin's effects against a complex disease's features will generate extensive overlap for many heavy metals, reflecting mercury's biological reach rather than specific causal evidence.

The American Dental Association's 2023 evidence-based guideline, while incorporating the FDA's 2020 advisory in four at-risk groups, concluded that dental amalgam is a safe and effective restorative material for populations outside those groups. The World Health Organization acknowledges amalgam's mercury content but notes that current evidence does not support the claim that amalgam causes systemic disease in the general population. The European Commission's 2024 policy update called for amalgam phase-down primarily on environmental grounds — mercury contamination of water systems — rather than patient-health grounds.

What this means. These positions are not wrong about what population-level studies have shown. They are answering a different question. Population studies average across individuals who differ enormously in their cumulative burden, and do not stratify by concurrent periodontitis, sleep-disordered breathing, ApoE ε4 status, kidney function, or gut barrier integrity. The patient with five amalgam surfaces, ApoE ε4/ε4, elevated urinary NAG, and severe periodontitis is not the average population subject. The evidence is most useful when read as evidence about what mercury can do under the right biological conditions, rather than what it does in a population-averaged analysis that cannot see those conditions.

Why a population-averaged amalgam study can miss a real effect in a subset of carriers.

ApoE comes in three common allelic variants — ε2, ε3, and ε4 — encoded by single-nucleotide differences in the APOE gene on chromosome 19. ApoE2 is broadly neuroprotective; ApoE3 is the most common and functionally average; ApoE4 is associated with reduced amyloid-β clearance, impaired lipid transport, less efficient repair of oxidative damage, and reduced binding affinity for divalent metals including mercury. One ε4 allele carries roughly threefold elevated lifetime risk for late-onset Alzheimer’s; two carry more than tenfold. Roughly 15 to 25 percent of the population carries at least one ε4 allele. The mercury relevance follows from the metal-binding deficit: an ApoE4 carrier buffers and clears central-nervous-system mercury less competently, so an exposure inconsequential on a population average can matter in the carrier who also carries amalgam — which is why a study that averages across ApoE genotypes can report a null while a real effect exists in a genetically defined subset.

What this means. A null result in the average patient does not rule out an effect in a genetically defined subset of carriers.

A 2026 systematic review and meta-analysis (Kothari et al., Journal of Oral Biology and Craniofacial Research) pooled eleven studies and 365 patients and found that amalgam replacement was associated with markedly higher odds of lesion resolution: pooled odds ratio 122.18 (95% CI 55.64–268.26, p < 0.001), with greater benefit when lesions directly contacted amalgam and patch testing was positive. The authors named the limits: between-study heterogeneity and indications of publication bias temper confidence in the pooled magnitude.

Evidence boundary. The directional finding — lesion resolution follows amalgam replacement with striking consistency across the available studies — survives the caveat; the point-estimate magnitude requires it. This applies to compatible oral lichenoid reactions with demonstrated amalgam contact or positive patch testing, not to routine amalgam removal in unaffected patients.

Cohort and meta-analytic evidence, the endocrine bridge, and the 1.5 vs 0.7 mg/L question

Bashash et al, published in Environmental Health Perspectives in 2017, followed 299 mother-infant pairs in the ELEMENT project. Maternal urinary fluoride was measured prospectively during pregnancy; offspring IQ was assessed at ages 4 and 6 to 12 using the McCarthy Scales' General Cognitive Index and the Wechsler Abbreviated Scale of Intelligence. A 0.5 mg/L increase in maternal urinary fluoride — roughly the interquartile range of cohort exposure — predicted a 3.15-point lower GCI at age 4 (95% CI: -5.42, -0.87) and a 2.50-point lower IQ at age 6 to 12 (95% CI: -4.12, -0.59). The effect was robust to adjustment for socioeconomic status, lead exposure, gestational age, and maternal IQ. Green 2019 in JAMA Pediatrics analyzed the Canadian MIREC cohort, measuring maternal urinary fluoride prospectively during pregnancy in 512 mother-child pairs. A 1 mg/L increase in maternal urinary fluoride was associated with a 4.49-point lower IQ in male offspring; the female-offspring relationship was attenuated and less clear. The 2024 NTP monograph synthesized 72 studies through systematic review methodology, applying the NTP framework for assessing study quality and grading evidence. The conclusion — moderate confidence for an association between fluoride exposures above 1.5 mg/L in drinking water and reduced IQ in children — represented the highest evidence-quality grade NTP applies short of high confidence. The monograph noted that data at lower exposures — closer to the 0.7 mg/L used in US community fluoridation — were less consistent, and that effects below 1.5 mg/L could not be ruled out with the available evidence. The Taylor 2025 meta-analysis in JAMA Pediatrics, authored by the National Institute of Environmental Health Sciences and the National Toxicology Program with ICF analytic contractors, extended the NTP synthesis. The analysis pooled 74 studies covering 20,932 children (22 low risk-of-bias of 74; mostly cross-sectional). The pooled standardized mean difference at the group level was -0.45 (95% CI: -0.57 to -0.33; p < 0.001). At the individual level, each 1 mg/L increase in urinary fluoride was associated with a 1.63-point IQ decrease across 13 studies, and 1.14 points in the low risk-of-bias subset. When restricted to drinking water concentrations below 1.5 mg/L, the overall pooled analysis was null. The inverse association persisted in two subsets — urinary-fluoride exposure measures, more accurate than drinking-water proxy, and the low risk-of-bias drinking-water subset. The authors' own conclusion captured the calibration: limited data and uncertainty in the dose-response association below 1.5 mg/L when fluoride exposure was estimated by drinking water alone.

The endocrine bridge. The thyroid evidence runs in structural parallel to the IQ evidence and increasingly connects to it. Hall and colleagues (Science of the Total Environment, 2023) analyzed thyroid function in 1,508 pregnant women from the same Canadian MIREC cohort that produced the Green IQ findings — the same population, the same exposure window, the same author team, examining a different organ. A 0.5 mg/L increase in drinking-water fluoride was associated with 1.65-fold increased odds of primary hypothyroidism (95% CI 1.04–2.60); among women with normal thyroid peroxidase antibody status — that is, without autoimmune thyroid disease — the adjusted odds ratio for primary hypothyroidism rose to 2.85 (95% CI 1.25–6.50), the effect larger where the hypothyroidism was not autoimmune in origin. Children born to women with fluoride-associated primary hypothyroidism had lower IQ at ages 3 to 4, concentrated in boys: an 8.42-point reduction in Full-Scale IQ (95% CI −15.33 to −1.50). The authors interpreted thyroid disruption as a candidate mechanism for the developmental neurotoxicity seen in the Green IQ data — a mechanistic bridge linking maternal exposure to a measurable endocrine intermediate. A 2026 comprehensive review by Patial and Thakur (Biological Trace Element Research) describes the pathways: fluoride interferes with iodine metabolism, disrupts T3 and T4 production, and induces apoptosis in thyroid follicular cells.

What this means. The September 2024 district-court ruling added a legal-regulatory finding of unreasonable risk under TSCA; the Ninth Circuit vacated that ruling on procedural grounds in May 2026 without deciding the underlying scientific question. What none of these lines of evidence established is causal certainty at 0.7 mg/L specifically — the exposure level used in most U.S. community water systems. The Taylor authors themselves are explicit about that uncertainty. The question the evidence does support taking seriously is whether the 0.7 mg/L threshold, set in an era before the maternal-exposure cohort data existed, still represents the right calibration for a delivery vehicle that exposes pregnant women without their explicit choice.

The antimicrobial mechanism, the community-shift literature, the endothelial-nitric-oxide evidence, and the human-exposure gap.

The antimicrobial mechanism. Fluoride is a broad-spectrum antimicrobial, and the mechanism is now established at the molecular level. Banerjee and colleagues (Proceedings of the National Academy of Sciences, 2025) demonstrated that fluoride disrupts the Clp protease complex in Streptococcus mutans, the keystone cariogenic species. Without Clp protease activity, the alternative sigma factor ComX accumulates uncontrolled, driving upregulation of bacteriocins and autolysins while suppressing immunity peptides; the cell wall degrades and the bacterium dies. The authors explicitly note the mechanism is not selective for cariogenic species — it is a broad pathway, which raises the ecological question the dental field has rarely asked.

The community-shift literature. In vitro, the effect is a shift rather than a collapse. López-López and Mira (Microbial Ecology, 2020) exposed oral biofilms to 500 ppm sodium fluoride — a topical-rinse concentration — and characterized the response with 16S rRNA sequencing and metatranscriptomics. Streptococcus oralis, a health-associated commensal, dropped ten-fold in both DNA and RNA, and sugar-metabolism gene expression shut down — but Rothia mucilaginosa, one of the genera responsible for oral nitrate reduction, rose eight-fold, and Streptococcus salivarius rose four-to-five-fold. The community restructured; it did not collapse. Jayanandan and colleagues (Odontology, 2025) reviewed antimicrobial-mouthwash dysbiosis and named the pathway explicitly — suppression of nitrate-reducing bacteria associated with reduced nitric-oxide bioavailability — across chlorhexidine (worst, 40–60% diversity reduction), cetylpyridinium chloride and polyhexamethylene biguanide (milder), herbal rinses (balance preserved), and fluoride-arginine formulations (promoting beneficial bacteria 30–50% while marginally upregulating antimicrobial-resistance genes); pure topical fluoride was not covered as a discrete intervention. Gondo and colleagues (Journal of Dentistry, 2026) found formulation-specific effects on saliva-derived biofilm viability and diversity, suggesting chemical form matters at least as much as elemental exposure.

The human-exposure gap. Moran and colleagues (Toxicology Letters, 2023) asked directly whether fluoride exposure affects the human microbiome and concluded that no studies have examined the effects of ingested fluoridated water on the human microbiome. Animal data (Yasuda 2017) show fluoride depleting acidogenic taxa in oral but not gut communities; human data are absent.

The endothelial nitric-oxide evidence — three layers that do not converge. At the cellular level, Huang and colleagues (Biological Trace Element Research, 2018) demonstrated in human umbilical vein endothelial cells that fluoride suppresses nitric-oxide synthesis through the PI3K/AKT/eNOS pathway, reducing phosphorylation of both AKT and endothelial nitric-oxide synthase — work the original investigators themselves described as preliminary. At the organ-system level, Yu and colleagues (Biological Trace Element Research, 2026) synthesized evidence that excess fluoride initiates and aggravates hypertension through endothelial dysfunction — an endothelin-1/nitric-oxide imbalance, oxidative stress, renin-angiotensin-aldosterone-system dysregulation, atherosclerosis, and direct myocardial damage. At the population level, Davoudi and colleagues (International Archives of Occupational and Environmental Health, 2021) meta-analyzed six studies: high-fluoride-exposed populations showed systolic blood pressure 6.49 mmHg higher (95% CI 3.73–9.25) and diastolic 4.33 mmHg higher (95% CI 1.39–7.26), with a pooled odds ratio for essential hypertension of 2.14 (95% CI 1.02–4.49).

The nitrate-reducer bridge question. Whether topical fluoride produces an antiseptic-pattern depletion of the nitrate-reducing community is unanswered in the systematic-review literature.

What this means. Endothelial nitric-oxide suppression is direct; the population hypertension signal is real — but the six meta-analyzed studies are predominantly from fluoride-endemic regions (Iran, China, India) where natural drinking-water fluoride runs 2–10 mg/L, well above the 0.7 mg/L used in U.S. community water fluoridation. Whether fluoride at 0.7 mg/L, sustained across decades, produces measurable cardiovascular effects in U.S. populations — and whether through the oral-microbiome pathway or direct endothelial mechanisms — remains unanswered at population scale.

The CBCT-versus-periapical detection comparison has been made repeatedly and the direction of the finding is consistent: cone-beam computed tomography detects more apical lesions per tooth than the two-dimensional periapical radiograph, with the detection ratio commonly running in the range of two- to four-fold on a per-tooth basis depending on the population and the reference standard. Estrela and colleagues (Journal of Endodontics, 2008) compared the two modalities directly in a sample of endodontically treated teeth and reported a detection gap that has anchored the subsequent literature. Kruse and colleagues (International Endodontic Journal, 2015) pooled the available studies in a systematic review and meta-analysis, with the synthesis confirming both the direction and the order of magnitude of the gap. The American Association of Endodontists and the American Academy of Oral and Maxillofacial Radiology joint position paper on CBCT in endodontics, published in 2015, reviewed the detection literature alongside the radiation-dose calculus and named specific clinical indications — including suspected endodontic pathology not visible on conventional imaging and assessment of healing in previously treated teeth with persistent or new symptoms — where CBCT is the appropriate modality.

What this means. First, the false-positive rate: not every CBCT-detected periapical radiolucency is active disease. The exact rate varies by reader experience and reference standard; what the literature consistently shows is that it is meaningful enough that the imaging finding is the beginning of the clinical question, not the end of it. Second, the reference-standard problem: the histological confirmation of a CBCT-detected lesion as active disease versus healing versus scar requires biopsy, which is rarely obtained for asymptomatic findings. The detection-rate ratios in the literature are therefore radiographic-against-radiographic comparisons, not radiographic-against-histology. The direction holds. The exact magnitude carries that caveat.

The biology of persistent post-treatment endodontic infection has been developed in the literature along three converging lines. Stuart and colleagues (Journal of Endodontics, 2006) reviewed the role of Enterococcus faecalis in failed endodontic treatment and consolidated the evidence on its disproportionate recovery from persistently symptomatic or asymptomatic-but-lesion-positive treated teeth. The mechanisms underlying that disproportionate recovery — biofilm organization, dentinal-tubule penetration, calcium-hydroxide tolerance, and capacity for genetic exchange of resistance traits — are reviewed alongside the recovery data. Siqueira and Rôças, in their 2014 review of the causes and management of post-treatment apical periodontitis with colleagues Ricucci and Hülsmann, broadened the picture beyond E. faecalis alone: the post-treatment microbiology is a narrower, more selected community than the primary-infection microbiology, but it is not monospecies. The review consolidates the three pathways the literature has identified — persistent intraradicular infection that survives initial treatment, secondary intraradicular infection introduced during or after treatment, and extraradicular infection in the periapical tissues themselves — and treats them as the dominant explanatory frame for treatments that fail by the imaging-and-symptoms criterion despite technically correct execution. Ricucci and Siqueira (Journal of Endodontics, 2009) added the histological dimension, demonstrating biofilm architecture directly on the canal walls of teeth extracted for treatment failure, with bacterial colonies organized into the structured communities the planktonic-versus-biofilm paradigm predicts.

What this means. The limit of this evidence is on the question the clinician most wants answered: what proportion of clinically successful root canal treatments harbor residual infection at this microbiological level? That question has no clean answer at the population scale, because the teeth examined microbiologically are by definition the teeth extracted — typically for clinical failure or for unrelated reasons. The biology of the residual-infection mechanism is established. The prevalence of clinically silent residual infection in apparently successful treatments is a quantity the evidence base infers from imaging studies, not measures directly from microbiology.

The systemic-inflammatory dimension of chronic apical periodontitis has been developed along three main lines. The cohort and cross-sectional epidemiology is the first. Caplan and colleagues (Journal of Dental Research, 2006) followed 708 men in the VA Dental Longitudinal Study for up to 32 years, with incident endodontic lesions assessed radiographically; incident endodontic lesions were associated with time to coronary heart disease after adjustment, but only among men 40 or younger. Costa and colleagues (Journal of Endodontics, 2014) reported a cross-sectional study of one hundred three patients undergoing coronary angiography, with chronic apical periodontitis associated with an approximately 2.8-fold higher prevalence of significant coronary artery disease. The mechanistic biomarker work is the second line. Cotti and colleagues (Journal of Endodontics, 2011) studied forty young men with asymptomatic apical periodontitis against twenty controls and reported elevated circulating interleukin-1, interleukin-2, interleukin-6, and tumor necrosis factor alpha, alongside elevated asymmetric dimethylarginine — the endogenous inhibitor of nitric oxide synthesis — and reduced endothelial flow-mediated dilation reserve. The design is cross-sectional and the sample size is modest; the directional findings connect endodontic infection to vascular function at the level of the molecular intermediary, not only the inflammatory marker. The pre/post-treatment biomarker work is the third line. Jakovljevic and colleagues (International Endodontic Journal, 2024) systematically reviewed sixteen studies of patients with chronic apical periodontitis before and after successful endodontic treatment; from a six-study subset of those sixteen, the meta-analytic estimate showed a significant reduction in serum high-sensitivity C-reactive protein at six months — from approximately 2.26 mg/L at baseline to approximately 1.28 mg/L post-treatment (Z = 2.03, p =.04). The pooled studies are observational before-and-after designs without an untreated control arm — patients measured at baseline and again at six months — so the pre-to-post decline cannot be cleanly separated from regression to the mean or secular change, a limitation the Newcastle–Ottawa appraisal the review applied is designed to surface. The borderline test statistic reflects a relatively small meta-analytic synthesis.

What this means. In pooled pre–post studies, successful endodontic treatment was followed by a roughly 40 percent fall in hs-CRP; without controlled trials at the scale the periodontal literature has reached, that is evidence of association and reversibility, not yet causal proof.

The comparative-outcomes literature has been synthesized in several systematic reviews and meta-analyses, with the contemporary picture more favorable to retention of the natural tooth than the older surgical-endodontics literature suggested. Torabinejad and colleagues (Journal of Prosthetic Dentistry, 2007, with a follow-on synthesis in the Journal of Endodontics in 2015) compared root canal treatment and restoration, implant-supported single crowns, fixed partial dentures, and extraction without replacement across the available outcomes literature, with five- and ten-year survival the principal endpoints. The headline finding was that the survival of endodontically retained teeth and the survival of implant-replaced teeth converge over the medium term, with the divergence at longer follow-up depending heavily on the population-level variables (periodontal history, smoking status, maintenance regularity) the implant literature has identified as predictors. Kang and colleagues (2015) compared retreatment and surgical endodontics directly in a meta-analysis, with the contemporary surgical-endodontics literature (microsurgery, biocompatible retrograde materials, operative microscope) outperforming the historical surgical-endodontics literature substantially. Setzer and colleagues (Journal of Endodontics, 2010) consolidated the contemporary surgical-endodontics outcomes specifically and reported the magnitude of the improvement over older techniques.

What this means. First, the success-criterion problem: the comparative literature uses heterogeneous outcome definitions — radiographic success (no detectable lesion), functional success (the tooth or implant in service), and symptom-free survival — and the numbers vary depending on which is applied. Second, the patient-selection problem: the comparative literature draws from clinical populations that are not random samples of the asymptomatic-CBCT-finding population this chapter has been describing. The five- and ten-year numbers reflect the populations treated in the published studies, which are weighted toward symptomatic cases. The applicability of those numbers to the asymptomatic-CBCT-detected case is an inference, not a direct extrapolation.

The American Association of Endodontists and the American Academy of Oral and Maxillofacial Radiology joint position paper (2015) reviews the indications, the radiation-dose considerations, and the comparative imaging context for CBCT in endodontic practice. The indications named include diagnosis in cases where conventional radiography has not resolved the clinical question, assessment of complex anatomy, assessment of healing in previously treated teeth where new or persistent symptoms have emerged, evaluation of resorption, and treatment planning for surgical endodontics. The screening of asymptomatic patients without indication is explicitly not endorsed. The effective dose of a small-field-of-view dental CBCT — the protocol relevant for endodontic indication — falls in a range below the dose of a comparable medical computed-tomography study and above the dose of a periapical radiograph; the exact values vary by machine, voxel size, and exposure parameters.

What this means. The threshold question this chapter named — at what asymptomatic-patient-marker-elevation does the indication threshold shift toward imaging — is not addressed in the position paper, because the position paper's framework is clinical indication for symptoms or known disease, not the longevity-medicine framework of unexplained inflammatory elevation in an otherwise optimized patient. Closing that gap requires evidence the literature has not yet generated.

Dominy and colleagues (Science Advances, 2019) combined three lines of evidence. In postmortem brain tissue, P. gingivalis gingipain antigens (RgpB, Kgp) were detected at significantly higher rates in patients who had died with clinical and pathological Alzheimer's disease than in age-matched non-demented controls, with bacterial DNA corroboration in a subset. In mice, chronic oral infection produced hippocampal colonization within six weeks, followed by amyloid-beta accumulation in the same regions. In vitro, gingipains cleaved tau at sites associated with neurofibrillary tangle formation. The gingipain inhibitor COR388 (atuzaginstat) reduced hippocampal P. gingivalis burden and amyloid accumulation in the mouse model and advanced into the GAIN trial (NCT03823404), in which 643 people with mild to moderate Alzheimer's disease were randomized to placebo or one of two doses of atuzaginstat. The drug did not meet its co-primary cognitive and functional endpoints. A dose-related liver-toxicity signal led the U.S. Food and Drug Administration to place the program on full clinical hold in January 2022, and development was suspended. An apparent benefit in participants with detectable oral P. gingivalis came from a prespecified subgroup analysis and is exploratory. No peer-reviewed primary report of the full trial results was located; the outcome is documented in published reviews.

What this means. The GAIN outcome refutes the simpler therapeutic claim — that inhibiting one bacterial enzyme prevents cognitive decline in established Alzheimer's disease — but does not refute the underlying mechanism. The mechanism survives the trial. The therapeutic translation does not yet exist.

Evidence boundary. Most authors of the Dominy 2019 study were affiliated with Cortexyme, the developer of atuzaginstat.

Chen and colleagues' 2026 meta-analysis of longitudinal studies (about 35.7 million participants) reported pooled associations of roughly odds ratio 1.26 for dementia and 1.40 for mild cognitive impairment in adults with greater tooth loss, adjusted for the standard demographic and medical confounders the component studies used. Among participants with significant tooth loss who had functional dentures — chewing function restored despite the missing teeth — the pooled estimate was close to neutral (odds ratio 1.03, 95% CI 0.82–1.28).

Evidence boundary. Because this was a subgroup analysis, it does not establish that dentures themselves prevent dementia: denture use is not randomly assigned, and people who obtain and wear functional dentures differ in access to care and health behaviors. The confidence interval also remains compatible with a modest effect. What the finding suggests is that chewing function may help explain part of the association — pointing intervention toward restoring function, not tooth count alone.

Ermini and colleagues (NPJ Parkinson's Disease, 2024) examined postmortem substantia nigra tissue from Parkinson's patients and matched controls. P. gingivalis gingipain antigens were abundant in dopaminergic neurons in both groups; the Parkinson's-specific finding was not higher abundance but ultrastructural colocalization of gingipains with aggregates of phosphorylated alpha-synuclein. (The higher-in-disease result belongs to Dominy 2019, in Alzheimer's temporal cortex.) In vitro, the lysine-gingipain (Kgp) cleaved alpha-synuclein at biologically relevant sites, generating fragments prone to the aggregation that drives Lewy body pathology — a parallel to the Dominy tau finding.

At population level, Jeong and colleagues analyzed a nationwide Korean retrospective cohort of about 6.8 million adults, identifying periodontitis from insurance diagnostic codes. The fully adjusted hazard ratio for incident Parkinson's disease was 1.09 (95% CI 1.06–1.12), an association the authors characterize as weak; it did not strengthen with periodontal severity, and the severe-disease estimate was not statistically significant. Periodontitis with concurrent metabolic syndrome showed a somewhat higher association (hazard ratio 1.17, 95% CI 1.12–1.22, compared with neither condition). Coded periodontal disease preceded the Parkinson's diagnosis, but the database was not designed prospectively, and prodromal disease before diagnosis cannot be excluded. Yilmaz and colleagues (Journal of Periodontology, 2023) detected elevated neurofilament light chain and S100B in the gingival crevicular fluid of Parkinson's patients. The remaining literature is cross-sectional — oral- and gut-microbiome composition studies (Li 2022; Mihaila 2019) — or review-level (Wang 2024; Zavala-Medina 2026). No published study has followed the oral microbiome forward in time against Parkinson's progression.

What this means. The mechanism rests on one substantia-nigra cohort and parallels the better-validated Alzheimer's mechanism; the weak population-level association is consistent with it. No interventional trial of an oral-pathogen-targeting agent has been conducted in Parkinson's disease. The therapeutic claim has not been tested.

Pi and colleagues (Journal of Affective Disorders, 2024) applied two-sample Mendelian randomization, which uses genetic variants fixed at conception as instruments for an exposure, because those variants precede, and so are largely independent of, environmental confounders. The method's principal weakness is horizontal pleiotropy — variants that affect the outcome through other pathways; the Pi study tested for it with MR-Egger regression and weighted-median estimators, with sensitivity analyses consistent across approaches. The estimate for major depressive disorder → periodontitis was OR 2.15 (95% CI 1.24–3.75). The reverse direction, using painful-gum instruments, returned OR 96.48, but its 95% CI (2.66–3495.28) spans a roughly 1,300-fold range, identifying direction while leaving magnitude profoundly uncertain. The authors flag both estimates as potentially confounded by smoking and alcohol. MR supports causal direction only for the specific exposure-outcome pairs tested, not for related pathologies the instruments never measured.

Around this causal core, Neupane and colleagues (Neurobiology of Disease, 2026) scoping-reviewed 165 longitudinal studies and catalogued 35 oral→mental associations, 32 mental→oral and 5 bidirectional, with periodontitis linked to Alzheimer's and dementia outcomes in 16 studies. A separate Mendelian randomization analysis (Nolde and colleagues, 2022) found no causal relationship between depression and periodontitis in either direction. On cognitive outcomes, Dibello and colleagues (2024) reported a meta-analytic relative risk for cognitive impairment of 3.01 (95% CI 1.52–5.95), with substantial between-study heterogeneity, graded moderate-quality and described by the authors as controversial, and a dementia relative risk of 1.22.

What this means. Dibello also found no significant periodontitis → depression association (RR 1.07, 95% CI 0.95–1.21, crossing the null). With Pi's MDD → periodontitis arrow supported and the reverse not, "bidirectional" here means documented in both directions, not equally weighted in both.

Toups, Tanio, Bredesen, and colleagues (2026) reported, in a non-peer-reviewed preprint, a randomized controlled trial in seventy-three patients with mild cognitive impairment or early dementia. The precision-medicine arm addressed each patient's identified contributors — metabolic, nutritional, sleep and circadian, physical and cognitive — together with the oral environment: periodontal management, oral-microbiome diagnostics, UARS dental-splint therapy where indicated, and evaluation of dental-material exposures including amalgam mercury. The control arm received standard of care. Over nine months the primary endpoint, a Neurocognitive Index, rose from 92.0 to 106.2 in the precision-medicine arm and fell from 96.9 to 92.4 in the control arm (p < 0.001; Cohen's d = 1.11 [95% CI 0.56–1.66]). An open-label pilot (Toups and colleagues, Journal of Alzheimer's Disease, 2022, n = 25) and a methodology review (Bredesen, Toups, and colleagues, Journal of Alzheimer's Disease, 2023) preceded it. The authors set the effect at roughly seven times that of lecanemab in CLARITY-AD (van Dyck and colleagues, NEJM, 2023) and four times that of donanemab in TRAILBLAZER-ALZ 2 (Sims and colleagues, JAMA, 2023). That comparison is neither a head-to-head trial nor a comparison of safety: lecanemab produced ARIA-E (brain edema or effusion) in 12.6% of treated patients and donanemab in 24.0% — adverse events the integrated protocol has no analogue to.

What this means. The trial is a preprint, not yet peer-reviewed, and open-label, so attention bias cannot be excluded. The protocol was multi-component by design, so no analysis could isolate the contribution of mercury evaluation, dental-splint therapy, or microbiome remediation. Patients scoring below 18 on the Montreal Cognitive Assessment were excluded, and seven dropped from the precision-medicine arm, leaving sixty-six analyzable. The cross-trial comparison spans different populations, endpoints, and durations. What the trial establishes is a hypothesis-generating signal: an integrated protocol including the oral environment produced cognitive improvement in a small randomized comparison at an effect size warranting larger-scale investigation. It does not establish that the oral environment is the dominant contributor within the protocol, that the integrated approach replaces monoclonal antibody therapy, or that the effect translates to the biomarker-confirmed early-Alzheimer's population the antibodies were tested in. The closest standalone interventional signal — periodontal therapy tested on its own — is Bajaj and colleagues' 2018 proof-of-concept study in cirrhosis-related cognitive impairment, where periodontal therapy improved cognitive measures and reduced endotoxemia over thirty days: a specialized population and a surrogate endpoint, not a dementia-prevention trial.

Two routes from mouth to brain carry distinct evidentiary signatures. The trigeminal pathway is supported only by animal evidence, with an anatomical rationale: Pisani and colleagues (Int J Environ Res Public Health, 2022) reviewed the corridor by which tissue-invasive organisms — spirochetes most reliably — could travel along continuous tissue from the oral cavity into brainstem nuclei, bypassing the blood-brain barrier, and Su and colleagues (J Mol Neurosci, 2021) supplied the animal-model evidence on T. denticola hippocampal colonization. The indirect oral-gut-brain pathway — synthesized in Liang and colleagues (CNS Neuroscience and Therapeutics, 2024), with broader gut-brain-axis context from Chandra and Vassar (Immunological Reviews, 2024) — begins with up to about 1.5 liters of swallowed saliva a day: surviving oral organisms contribute to gut microbiome composition, and the gut's own signaling completes the path through vagal afferents and gut-derived metabolites (short-chain fatty acids, lipopolysaccharide, neurotransmitter precursors), a subset of which reach the brain.

What this means. The trigeminal pathway rests on animal-model evidence only; the oral-gut-brain pathway by mechanistic evidence and converging observational data. Quantitative partitioning of their relative contributions to neurodegenerative outcomes in humans has not been established. The integrated clinical strategy addresses both because the evidence does not support closing one while leaving the other open.

Three lines of evidence converge on IL-1 as the molecular conductor of age-related myeloid skewing, and together they make the case that the skew is a sustained signal rather than a permanent rewrite.

Pietras et al. (Nat Cell Biol, 2016) demonstrated that chronic IL-1 exposure drives hematopoietic stem cells toward precocious myeloid differentiation through a PU.1-dependent program, eroding self-renewal capacity and priming the cells to fail under transplantation stress — but found these effects fully reversible on IL-1 withdrawal, identifying a regulatory circuit rather than a fixed lesion.

Kovtonyuk et al. (Blood, 2022) localized the signal in aging: older mice produce more IL-1α/β in steady-state marrow, carry higher circulating levels of microbe-associated molecular patterns (specifically TLR4 and TLR8 ligands), and — decisively — show restored, unbiased lymphomyeloid output when the microbiota is suppressed with antibiotics or IL-1 signaling is pharmacologically blocked. The paper names the microbiome/IL-1/IL-1R1 axis as a "self-sustaining and therapeutically partially reversible" driver of HSC inflammaging.

Mitchell et al. (Passegué lab; Nat Cell Biol, 2023) added the niche: the aging marrow microenvironment degrades — fewer endosteal osteoprogenitors, expansion of LepR⁺ stromal cells, sinusoidal deterioration — and the damaged endosteum produces IL-1β that drives the inflamed character of the central marrow. Niche deterioration, HSC dysfunction, and defective regeneration were all ameliorated by blocking IL-1.

Evidence boundary. All three are mouse studies. IL-1 blockade is available in people — canakinumab, the drug used in the CANTOS cardiology trial discussed in the inflammation chapter — but no approved therapy targets IL-1 to reverse the aging of the blood system, and the human translation may not look like the mouse experiments. What the animal work establishes is the wiring: IL-1 is the line into the marrow, chronic IL-1 retunes the factory toward inflammation, and the retuning is not a one-way ratchet.

Agarwal et al. (Starczynowski lab; Nature, 2025) provide a direct molecular link between the microbial world and age-related clonal hematopoiesis. The mechanism: ADP-D-glycero-β-D-manno-heptose (ADP-heptose), an intermediate in Gram-negative lipopolysaccharide biosynthesis, accumulates in the circulation of older individuals as age-related intestinal permeability increases. ADP-heptose binds its cytosolic receptor ALPK1, triggering transcriptional reprogramming and NF-κB activation that confers a competitive proliferative advantage on DNMT3A-mutant pre-leukemic clones. The study also reports ADP-heptose's association with increased inflammation and cardiovascular risk in CHIP, and frames the ADP-heptose–ALPK1 axis as a candidate therapeutic target to slow progression from CHIP toward overt leukemia.

Evidence boundary. The source organ in this study is the gut (the paper's MeSH indexing is "Gastrointestinal Microbiome"; the tracing is to age-related intestinal permeability). An oral Gram-negative contribution to the same circulating ADP-heptose pool is biochemically plausible — ADP-heptose is a conserved Gram-negative LPS-pathway intermediate, and the periodontal pocket is a breached Gram-negative reservoir — but the oral-pathogen-to-marrow demonstration has not been performed. The oral contribution is a motivated hypothesis, not an established fact.

The simple story — CHIP is bad, more inflammation, more disease — is mostly true and importantly incomplete.

The selection mechanism. Caiado et al. (Blood, 2023) established that aging-related IL-1 selectively expands TET2-mutant clones, and that pharmacologic IL-1 inhibition impairs that pre-leukemic expansion — placing CHIP growth downstream of the same IL-1 signal that drives myeloid skewing.

The protective counter-signal. In a meta-analysis by Bouzid et al. (Nat Med, 2023), people with CHIP had a lower risk of Alzheimer's dementia (OR 0.64, P = 3.8 × 10⁻⁵), with Mendelian-randomization analyses supporting a possible causal direction. The qualifier that matters is mutation specificity — Matatall et al. (Cell Stem Cell, 2025) found the protection is TET2-specific (a 47% lower risk of late-onset Alzheimer's), with no protection from DNMT3A. Two further studies set the bounds: the protective signal replicates but only for larger clones (Jakubek et al., Alzheimers Dement, 2025, HR 0.62 at variant allele fraction ≥ 8%), and at least one large cohort finds no Alzheimer's association at all while CHIP raises the risk of vascular neurodegeneration and ALS (Liu et al., J Intern Med, 2024).

What this means. CHIP is best read as context-, mutation-, and clone-size-specific rather than globally good or bad.

GDF-15 (growth differentiation factor 15) rises with age along a non-linear, J-shaped curve rather than at a fixed rate (Welsh et al., Clin Chem Lab Med, 2022), and in large proteomic studies it is among the circulating proteins most strongly associated with age, multimorbidity, and mortality (Tanaka et al., Aging Cell, 2018; eLife, 2020) — while also being elevated across cardiovascular, renal, metabolic, and mitochondrial disease. A 2025 study from Martin Picard's group at Columbia (Huang et al., Mol Genet Metab) established three things relevant here: GDF-15 is reliably quantifiable in saliva; salivary gland secretory cells are themselves high expressers of it (profiled across 48 tissues in the GTEx dataset); and saliva GDF-15 carries a steep awakening rhythm — highest at waking, declining by 61.2% within 45 minutes. Salivary levels tracked neurological symptoms, fatigue, and functional capacity in patients with mitochondrial disease.

Evidence boundary. This is early, and the population studied was rare mitochondrial disease, not periodontitis — the salivary form is a promising, validated-but-preliminary measurement, not a clinical test of oral-inflammatory aging yet. The collection caveat is not a footnote: a salivary GDF-15 result is close to meaningless without a fixed time-since-waking, given a decline that halves the value inside an hour. The connection to GrimAge and the other aging clocks is taken up in the biological-age chapter.

Charles Serhan's group defined the framework of specialized pro-resolving mediators (SPMs) — resolvins, protectins, maresins, and lipoxins — as an active, agonist-driven resolution program (Nature, 2014), and Thomas Van Dyke and Serhan (J Dent Res, 2003) advanced periodontitis as a paradigm of failed resolution — a chronic lesion sustained partly because the resolution program is impaired, not only because the bacterial challenge is high.

In the key preclinical study, Hasturk et al. (J Immunol, 2007) treated established periodontitis in rabbits with Resolvin E1 as a monotherapy. RvE1 produced complete restoration of the local lesion, regeneration of lost bone, and reduction in the systemic markers C-reactive protein and IL-1β — while structurally related lipids (PGE₂, leukotriene B₄) worsened disease.

Evidence boundary. The regeneration result is animal data, in a rabbit model, with a degree of effect that human therapy has not matched. SPM supplementation in people is heterogeneous and not standard care — the human translation is the unfinished part.

The strongest evidence that periodontal treatment lowers blood sugar is the 2022 Cochrane review led by Terry Simpson and colleagues. The review included 35 randomized trials and 3,249 participants overall. Thirty of those trials, including 2,443 participants, contributed to the pooled three-to-four-month estimate: periodontal treatment lowered HbA1c by 0.43 percentage point compared with no treatment or usual care, a modest effect supported by moderate-certainty evidence. At six months, the pooled reduction was 0.30 point across 12 studies and 1,457 participants. At twelve months, the estimate of 0.50 point came from a single study of roughly 264 participants, so the long-term durability remains much less certain. Nearly all of the evidence in this chapter — the Cochrane trials, the epidemiology, the salivary work — comes from type 2 diabetes. The type 1 literature points the same way but is far thinner; read the claims here as claims about type 2 disease unless stated otherwise. A 2025 non-randomized cohort by Chee and colleagues, controlling for changes in diabetes medication, found larger reductions sustained to twelve months, greatest in patients who started with the poorest control. A cohort cannot isolate the effect of treatment, so this is supportive, not definitive; the long-term size of the effect is still being pinned down. A later retrospective cohort did not find a significant HbA1c reduction after periodontal treatment, a useful real-world counter-signal that does not overturn the randomized evidence but reinforces that the effect is not universal.

What this means. This is not a claim that periodontal treatment manages diabetes, or replaces anything. It is a claim that one input into blood sugar control has been sitting in the mouth, unaddressed, and that removing it helps — a smaller statement than the headlines around this topic sometimes make. It is also one of the few statements in oral-systemic medicine backed by a stack of randomized trials.

The association runs both ways and is strongest in severe disease. A 2020 meta-analysis of 53 observational studies by Chen-Zhou Wu and colleagues found that type 2 diabetes raised the risk of developing periodontitis by 34% and severe periodontitis raised the incidence of type 2 diabetes by 53%, with a weaker and less robust effect for mild periodontitis. Meta-analyses limited to prospective cohorts report smaller associations, closer to 20 to 30 percent in either direction: the relationship is real and bidirectional, but its size varies with study design and with how periodontitis is defined. A large Taiwanese National Health Insurance cohort by Shih-Yi Lin and colleagues, in people without diabetes at baseline, likewise found a more modest association for the gum-disease-to-diabetes direction (adjusted hazard ratio 1.19). Dana Graves and colleagues' 2025 review sets out the biology in both directions: high blood sugar impairs periodontal immune defenses and tilts bone metabolism toward resorption, while inflammatory signals from a diseased periodontium contribute to systemic insulin resistance. The reviewers also note that oral dysbiosis can deplete the nitrate-reducing bacteria the body uses to maintain nitric oxide signaling, a thread the chapter on the mouth as a nitric oxide factory picks up.

What this means. Whether periodontitis can initiate diabetes, as opposed to worsening it, remains uncertain; the reviewers are explicit that causality in that direction is not established.

The idea that a saliva sample might flag diabetes rests on a small but real body of work. In a 2020 study in Scientific Reports, Xiaoyu Sun and colleagues sequenced the salivary microbiomes of healthy people, periodontitis patients, and periodontitis patients who also had type 2 diabetes; classification models separated the groups with accuracies between 96% and 100% depending on the comparison. A 2022 study by Michi Omori and colleagues took an independent approach — 42 elderly patients with type 2 diabetes matched to 42 controls — and a random-forest model distinguished them with an area under the curve of 0.94. Sun and colleagues also observed that after effective glycemic control — in their cohort, with metformin — the salivary community shifted back toward a healthier profile, with several periodontitis-associated species declining.

What this means. These are proof-of-concept results, not a validated diagnostic. The samples are small, the cohorts are specific, and no model has been prospectively tested as a screening tool in the general population. The metformin observation suggests the mouth tracks the metabolic state rather than registering it once — a reason to keep watching this work rather than to act on it yet.

A 2025 narrative review by Paras Ahmad and colleagues catalogued the preclinical work: in cell and animal models, agents such as liraglutide and exenatide promoted bone-forming activity in periodontal ligament stem cells, reduced inflammatory cytokines, and limited alveolar bone loss — apparently through direct receptor activation in periodontal tissue, independent of their effect on blood sugar. A 2020 study by Noritaka Sawada and colleagues is representative: in rats with experimentally induced periodontitis, liraglutide reduced inflammatory infiltration, lowered tumor necrosis factor expression, shifted macrophages away from the inflammatory M1 type, and reduced bone resorption.

What this means. The mechanistic case is interesting; the clinical case is barely begun. The direct-periodontal evidence is almost entirely from animals and cell culture, and the human data remain very limited and observational. Ahmad and colleagues conclude that the drugs' role looks more like preserving periodontal and peri-implant health than treating established disease — and even a real anti-inflammatory effect does not remove the bacterial biofilm, which is why mechanical treatment stays necessary. A 2026 scoping review by Angeliki Polymeri and colleagues reached the same bottom line: the preclinical signals are genuine, but no clinical trial has yet tested periodontal endpoints. Most important: no trial has tested the combination this chapter is really about — periodontal treatment plus a GLP-1 — so the claim that their benefits add is reasonable inference, not demonstrated fact.

A 2020 microsimulation model by Sung Eun Choi and colleagues at Harvard, published in Diabetes Care, projected that nonsurgical periodontal treatment for people in the United States with type 2 diabetes and periodontitis would be cost-saving — a net savings of $5,904 per person — while reducing microvascular complications (nephropathy, neuropathy, retinopathy) by roughly 18–20%. Cost-saving is a stronger claim than cost-effective: the treatment more than pays for itself. The essential caveat is that this is a model built on trial and survey inputs, not a prospective rollout in a real health system. A 2025 systematic review of economic evaluations concluded that periodontal treatment is cost-effective in this population, reported lower healthcare costs with low-to-moderate certainty, and noted that transferability across health systems still needs study. The integration runs the other way too. Evanthia Lalla and colleagues at Columbia (2011) found that, in dental patients with no prior diabetes diagnosis but at least one diabetes risk factor, two findings a dentist already records — the proportion of deep periodontal pockets and the number of missing teeth — combined with a point-of-care HbA1c test correctly identified about 92% of those who turned out to have diabetes or prediabetes. That performance belongs to this pre-selected, higher-risk group, not to dental patients at large. The 2026 ADA Standards of Care situate dental and periodontal assessment within the comprehensive evaluation of the patient with diabetes.

What this means. The economic case rests on a modeled cost-saving result, with a later systematic review supporting cost-effectiveness; whether savings hold across health systems is not yet established. The dental visit as a screening point for undiagnosed diabetes is supported by validation data in a higher-risk group and is now reflected in the 2026 ADA Standards.

Two surface proteins do two different jobs. Rubinstein and colleagues (Cell Host Microbe, 2013) showed that the FadA adhesin binds E-cadherin on colon epithelial cells and augments Wnt/β-catenin signaling — a growth pathway already central to colorectal carcinogenesis; fadA gene levels in patient tissue ran 10- to 100-fold higher in adenomas and cancers than in normal colon. Gur and colleagues (Immunity, 2015) identified the second job: the Fap2 lectin binds TIGIT, an inhibitory receptor on T cells and natural killer cells, switching off the immune attack on the tumor — and the binding works on human TIGIT, not mouse, which matters for how the lab work translates.

The strain that does this is specific: the colorectal niche is dominated not by the ordinary mouth-dwelling form but by one subspecies-level clade — the F. nucleatum animalis C2 group established earlier in this book — and single-cell mapping of these tumors shows the bacteria clustered tightly with the cancer cells whose behavior they disturb. The clinical correlates followed. Yu and colleagues (Cell, 2017) traced chemotherapy resistance to a specific route: the organism engages TLR4 and MYD88 innate-immune signaling and a set of microRNAs to switch on autophagy, the cellular recycling program tumor cells lean on to survive treatment. Borozan and colleagues (Cancer Epidemiol Biomarkers Prev, 2022), sequencing tumors from nearly 2,000 patients, found its presence marked worse colorectal-cancer-specific survival — a hazard ratio of 2.13 in nonhypermutated, microsatellite-stable tumors, partly through a shift to later stage. On the route question, Abed and colleagues (Front Cell Infect Microbiol, 2020) whole-genome-sequenced paired oral and tumor isolates from the same patients, found that in some patients they matched, and showed in mouse models that bloodborne bacteria colonized tumors more efficiently than swallowed ones.

Evidence boundary. The human tumor-tissue association is strong, the strain match in some patients is direct human evidence that the route exists, and the mechanisms are unusually well resolved for this field. But the growth-promoting and chemoresistance steps are shown in cell culture and animal models; F. nucleatum is one of several organisms found in these tumors; and none of this establishes that the mouth starts the cancer. What the evidence supports is a passenger that, once present, makes the disease worse.

Wang and colleagues (J Dent Res, 2024) pooled nineteen studies and roughly 16.6 million people. Periodontal disease tracked with higher gastrointestinal-cancer risk overall (hazard ratio 1.31, 95% CI 1.16–1.49) and at specific sites: colorectal 1.21 (1.05–1.39), esophageal 1.39 (1.15–1.68), gastric 1.13, and pancreatic 1.35 — but the pancreatic confidence interval runs from 1.00 to 1.82, meaning its lower bound sits exactly at no effect. Severe periodontal disease carried the largest point estimate, 1.79, with a wide interval (1.07–2.99).

Evidence boundary. These are observational associations. The dose-response — more severe disease, larger estimate — makes pure confounding a less complete explanation, but cannot rule it out: periodontal disease and gastrointestinal cancer share risk factors, from smoking to diet to chronic inflammation. The pancreatic estimate should be read as suggestive at best, given where its interval ends. The epidemiology establishes breadth and consistency. It does not establish that the mouth causes these cancers, and it says nothing about whether treating the mouth lowers the risk.

Acetaldehyde linked to alcohol is classified as a Group 1 carcinogen — the highest, "definite" tier — by the International Agency for Research on Cancer (Secretan and colleagues, Lancet Oncol, 2009). Salaspuro (Best Pract Res Clin Gastroenterol, 2017) estimated that after a drink, more than 70% of the acetaldehyde in the mouth is produced by oral microbes rather than by the body's own enzymes; specific oral streptococci and Neisseria carry out the conversion (Kurkivuori and colleagues, Oral Oncol, 2007; Tagaino and colleagues, Sci Rep, 2019).

Two features make it relevant to a longevity reader. First, it is modifiable: as far back as the late 1990s, Homann and colleagues (Carcinogenesis, 1997, and in heavy drinkers and smokers, 2000) showed a chlorhexidine rinse lowered salivary acetaldehyde, proof the bacterial source can be reduced. Second, the risk is gene-dependent: Tsai and colleagues (Int J Cancer, 2014) found that poor oral hygiene multiplied head-and-neck cancer risk most in people carrying the ALDH2 and ADH1B variants that already impair alcohol metabolism. Recent reviews place this within the broader alcohol-and-cancer picture (Morford and colleagues, JAMA, 2025; Gapstur and colleagues, N Engl J Med, 2023).

What this means. This is an amplifier of a known risk factor, not an independent oral pathway to cancer. The mouth does not generate meaningful acetaldehyde without alcohol to work on. What the evidence supports is that the oral microbiome modulates how much of a known carcinogen forms locally after drinking — a modifiable contribution, not a standalone cause.

The chemistry: peptidylarginine deiminase (PAD) enzymes convert arginine residues in proteins to citrulline, and antibodies against the altered proteins — anti-citrullinated protein antibodies, or ACPA — are the defining serologic marker of rheumatoid arthritis. Wegner and colleagues (Arthritis & Rheumatism, 2010; established earlier in this book) characterized P. gingivalis's own version, called PPAD, citrullinating host proteins such as fibrinogen and α-enolase inside the periodontal pocket.

P. gingivalis is the only known human oral pathogen that makes its own PPAD — but it is not the only organism that can citrullinate. Gabarrini and colleagues (J Dent Res, 2018) showed that animal Porphyromonas species (P. gulae, P. loveana) carry PPAD homologues. And Konig and colleagues (Sci Transl Med, 2016) showed that a second oral pathogen, Aggregatibacter actinomycetemcomitans, drives citrullination by a wholly different route: its leukotoxin makes host neutrophils hypercitrullinate. It does not make its own enzyme.

Evidence boundary. Silva and colleagues (Arthritis Care Res, 2022) meta-analyzed nine studies (388 patients) and found periodontal treatment moved rheumatoid-arthritis disease activity in the favorable direction (standardized mean difference −0.88), but graded the certainty low to very low — the effect, they noted, is likely to shift as better trials arrive. The mechanism is among the cleanest in oral-systemic medicine, and the treatment signal points the right way. But the clinical effect is modest and the evidence thin.

Kouris and colleagues (J Clin Periodontol, 2026) randomized 53 chronic-kidney-disease patients to nonsurgical periodontal treatment or no treatment and followed them six months. Within the treated group, estimated glomerular filtration rate (eGFR) — the standard measure of how well the kidneys filter — rose by about 10 to 11 mL/min/1.73m² depending on the calculation used; high-sensitivity C-reactive protein fell by 84% relative to the control group; and HbA1c dropped 0.54% in the patients who also had diabetes.

Evidence boundary. A 10-unit eGFR gain would be an unusually large and clinically striking change — which is exactly why the size of the claim demands that the size of the trial be kept in view. This is roughly fifty patients over six months, with the eGFR change measured within the treated group rather than as a between-group difference. He and colleagues (J Periodontal Res, 2023) reviewed the surrounding literature and found a consistent bidirectional relationship but heterogeneous, lower-certainty evidence on treatment. If a larger, longer trial reproduces the kidney-function effect, it would be among the most significant findings in the field. Until then it is a striking early signal — the kind that either replicates or evaporates.

The null is unusually well established. Iheozor-Ejiofor and colleagues (Cochrane Database of Systematic Reviews, 2017) pooled 15 randomized trials and 7,161 women: periodontal treatment during pregnancy produced no clear effect on preterm birth (relative risk 0.87, 95% CI 0.70–1.10) and only a low-certainty signal toward less low birth weight. Polyzos and colleagues (BMJ, 2010) found that once the analysis was restricted to the higher-quality trials, even that signal disappeared — the apparent benefit lived in the weaker studies.

The mechanism explains why the worry is real even though treatment-during-pregnancy fails. Parhi and colleagues (Cell Reports, 2022) showed F. nucleatum reaches the placenta using Fap2 — the same lectin from the colorectal story — to bind Gal-GalNAc, a sugar increasingly exposed on the placenta as gestation advances; Fap2-expressing bacteria reduced fetal survival in mice by about 70%, while a Fap2-deficient mutant did not. Han and colleagues (Infect Immun, 2004) first showed that intravenous F. nucleatum causes fetal loss in mice; Coppenhagen-Glazer and colleagues (Infect Immun, 2015) characterized the adhesin.

Evidence boundary. The older idea of a resident placental microbiome has largely not survived contamination scrutiny; reviews by Xu and Han and by Figuero and colleagues (Periodontology 2000, 2022 and 2020) now describe the placenta as colonized only pathologically, in adverse pregnancies, not as a normal microbial habitat. The mechanism is molecular and animal-model. It explains why timing might matter; it does not prove that earlier treatment changes human outcomes.

The signal comes from separating the trials by what they added. Merchant and colleagues (JAMA Network Open, 2022) found that scaling plus chlorhexidine was associated with lower preterm birth (relative risk 0.56) and low birth weight (0.47), while scaling alone was not (preterm RR 1.03). Salama and colleagues (BMC Pregnancy and Childbirth, 2024) found a similar direction (RR 0.44) but flagged severe heterogeneity (I² = 91%) and funnel-plot asymmetry; Wu and colleagues (Front Public Health, 2024) saw chlorhexidine-containing strategies lead in a network analysis. The mechanism fits the chapter: chlorhexidine specifically reduces Fusobacteria (Bescos and colleagues, Sci Rep, 2020). A dedicated trial is now underway (Yasuoka and colleagues, Trials, 2025; 480 women) — though it tests sustained chlorhexidine through pregnancy, the very pattern the next point cautions against.

That same Bescos study, with a systematic review by Senkus and Crowe-White (Crit Rev Food Sci Nutr, 2020), shows the cost: chlorhexidine also strips the nitrate-reducing oral bacteria behind nitric oxide production, lowering salivary and plasma nitrite and nudging blood pressure up — the argument the nitric-oxide chapter builds in detail. An umbrella review of daily mouthwashes (Koromantzos and colleagues, J Clin Periodontol, 2026) found that prolonged daily use yields only limited plaque-and-gingivitis benefit and that alternatives, including cetylpyridinium chloride and essential-oil rinses, are reasonable for sustained use; cetylpyridinium chloride alone reduced preterm birth in one high-risk trial (Jeffcoat and colleagues, Am J Obstet Gynecol, 2011). A gentler lever exists too: omega-3 fish oil suppressed F. nucleatum-driven placental inflammation in mice (Garcia-So and colleagues, JCI Insight, 2019), with no targeted human trial yet.

Evidence boundary. A consistent signal, weakened by trial bias, one outlier study, and populations that may not generalize — biologically coherent and genuinely promising, but unproven until a dedicated trial reports. Read with the nitric-oxide chapter: a short, targeted course around treatment is defensible; daily long-term chlorhexidine as a wellness habit is not.

Most epigenetic clocks estimate accumulated biological age — a single snapshot. DunedinPACE, introduced by Belsky and colleagues (eLife, 2022), measures something more useful for anyone trying to intervene: the current rate of aging, expressed as biological years accrued per calendar year. A value of 1.0 means a person is aging one biological year per year; 1.2 means twenty percent faster.

The clock was built from the Dunedin birth cohort, which tracked nineteen separate indicators of organ-system integrity across two decades in the same people, then distilled that two-decade trajectory of decline into a single blood test. In validation across multiple datasets, DunedinPACE was associated with morbidity, disability, and mortality, with effect sizes comparable to the GrimAge clock and adding predictive information beyond it. It also showed faster aging in adults who had experienced childhood adversity — evidence that the measure responds to lived biological burden, not just inherited genetics.

Evidence boundary. DunedinPACE is a research instrument, validated for population studies, not a diagnostic test for individuals. It has high test-retest reliability, but a single value carries meaningful measurement uncertainty, and the clinically actionable question — does lowering a specific exposure move the number in a given person — is exactly the kind of within-individual claim the tool was designed to study but cannot yet answer at the bedside.

Mendelian randomization (MR) is the design that can test the "isn't this just a poverty marker?" objection. Because the gene variants a person inherits are fixed at conception and shuffled randomly, they are not influenced by income, stress, or smoking, so an association built from them is largely free of social confounding.

Peng and colleagues (Clinical Epigenetics, 2025) found that genetically predicted GrimAge acceleration was associated with increased risk of periodontitis (odds ratio roughly 1.12 to 1.16 across the FinnGen and GLIDE datasets), with no association for caries, ulcers, or leukoplakia — a specificity that points to inflammation rather than "bad teeth" in general. Huang and colleagues (J Clin Periodontol, 2024) independently found the same direction: GrimAge acceleration causally associated with periodontitis risk (OR 1.16). But when Song and colleagues (J Periodontal Res, 2025) tested the reverse arrow — does periodontitis causally accelerate the clock? — the MR result was null (no significant effect on BioAgeAccel or PhenoAgeAccel), even though their cross-sectional data showed the expected association. Li and colleagues (Sci Rep, 2024) added a caution from the forward direction itself: using different clocks — BioAge and PhenoAge rather than GrimAge — their Mendelian randomization found no causal effect of accelerated aging on periodontitis, even as the cross-sectional NHANES association held. The forward signal, so far, is clock-dependent.

What this means. The genetic signal that replicates — in GrimAge-based analyses — runs from faster aging to greater periodontitis susceptibility. The observational associations between periodontitis severity and accelerated biological aging are robust and dose-dependent across large NHANES samples, but those are exactly the confounded correlations MR exists to scrutinize; on the reverse direction, the one test so far — using BioAge and PhenoAge, the same clocks that missed even the forward signal — found no effect, and a GrimAge-based test of that direction is not yet in hand. The mouth is demonstrably readable in the aging signal. That it drives the signal remains, for now, unproven.

Dental cone-beam computed tomography (CBCT), taken for implant planning and other routine dental purposes, frequently captures the cervical region — and with it, incidentally, calcifications in the carotid arteries. AlSakr and colleagues (J Periodontol, 2021) found an association between intracranial carotid-artery calcifications visible on CBCT and periodontitis. Amarin and colleagues (J Prosthodont, 2023) documented calcified head-and-neck atheromas detected incidentally during routine dental evaluation. The clinical question is what those incidental findings predict. Lim and colleagues (Clin Oral Investig, 2021), in a systematic review, examined whether carotid calcifications on dental radiographs predict adverse vascular events. On the evidence, these incidental findings are a signal worth acting on — a prompt for medical referral — rather than a validated standalone screening test for stroke or cardiovascular death. Newer longitudinal evidence strengthens the idea that these calcifications can function as risk markers. In the PAROKRANK program, long-term follow-up has linked bilateral carotid calcifications on panoramic dental radiographs with later cardiovascular events and mortality. That prognostic signal adds weight to referral, but it does not turn dental imaging into a population screening test.

Evidence boundary. The case here is not that dentists should diagnose vascular disease or that a calcification on a CBCT is a verdict. It is narrower: the finding is often captured and sometimes noted, but the referral pathway that would make it matter does not reliably exist.

On December 16, 2025, the American Heart Association released an updated scientific statement, “Periodontal Disease and Atherosclerotic Cardiovascular Disease,” later published in the 2026 print issue of Circulation. The statement synthesizes more than a decade of new evidence since the 2012 statement, including Mendelian-randomization studies, periodontal-treatment interventions, inflammatory markers, vascular measures, bacteremia, and systemic inflammation. It also highlights the disproportionate burden of periodontal disease in under-resourced populations. The statement is deliberately calibrated. It concludes that the accumulated evidence supports a significant association between periodontal disease and atherosclerotic cardiovascular disease, while definitive causation and effective intervention strategies remain undetermined. Periodontal treatment can improve intermediate measures such as inflammatory markers and some vascular risk factors, but trials have not established that treating periodontal disease prevents heart attacks, strokes, or other cardiovascular events. A Danish nationwide cohort of more than 100,000 first-time heart-attack patients provides a useful observational example. Patients who had received periodontal treatment in the year before their heart attack had a three-year recurrent-event risk similar to those presumed periodontally healthy, while patients with no prior dental visits had substantially higher risks of cardiovascular death and heart failure.

Evidence boundary. The Danish finding cannot separate the effect of periodontal treatment from the many ways people who seek regular dental care differ from those who do not, including health behavior and socioeconomic factors. It is observational evidence, not proof that treatment prevents recurrent cardiovascular events.

What this means. The institutional position has moved well beyond dismissal: the association is substantial enough to warrant attention across medicine and dentistry. What remains unproved is the stronger claim that periodontal treatment itself prevents cardiovascular events.

The economic case for integrating periodontal care into diabetes management is promising, but it needs the same calibration as the biological case. A 2020 model-based cost-effectiveness analysis by Choi, Sima, and Pandya estimated that providing nonsurgical periodontal treatment to U.S. adults with type 2 diabetes and periodontitis would be cost-saving from a healthcare perspective. In that modeled population, the projected net saving was $5,904 per person, with an estimated gain of 0.6 quality-adjusted life-years. Those numbers came from a microsimulation built from clinical-trial and survey inputs; they were not the result of a prospective health-system rollout. A 2025 systematic review of economic evaluations found periodontal treatment cost-effective across several health-system models and found cost savings in some settings, including the U.S. Choi model. The review also emphasized that transferability across countries, payers, and health systems remains uncertain.

What this means. “Cost-effective” means the health gained appears worth the cost. “Cost-saving” is the stronger claim that total spending actually falls. Current evidence supports cost-effectiveness overall and suggests cost savings in some settings and models; it does not justify promising that every payer or health system will spend less.

On the dental side, caries (tooth decay) and periodontal inflammation are distinct processes; a mouth can be cavity-free while carrying substantial inflamed periodontal surface area, the burden earlier chapters quantified as comparable to a sizable wound. On the medical side, the high-sensitivity CRP assay read against cardiovascular thresholds is frequently not ordered in routine care — the interpretive gap the inflammation chapter develops.

Evidence boundary. The strength of the evidence varies across the measures this chapter discusses: high-sensitivity CRP is well validated for cardiovascular risk stratification, while routine salivary-flow testing or bruxism assessment in someone with no symptoms rests on a thinner base for changing outcomes.

What this means. The point is not that every reader needs every measure. It is that the specific reassurances of a routine checkup answer narrower questions than patients usually assume, and that the gap between "passed the exam" and "evaluated for the processes that drive long-term health" is real, documented, and — for most of these domains — measurable with tests and questions that already exist.