Impact of periodontitis on diabetic complications

Background + Aims

• Chronic kidney disease (CKD) involves impaired kidney function, increasing the risk of cardiovascular disease, infection, and mortality. Despite medical advances, CKD prevalence is rising globally.
• Periodontitis, a common inflammatory oral disease, has been proposed as a potentially modifiable risk factor for CKD.
• Evidence suggests periodontal pathogens and inflammatory markers may impair kidney function by affecting endothelial cells in nephrons.
• Both diabetes and periodontitis are linked to CKD, with diabetes potentially mediating this relationship.
• This study used mediation analysis to investigate possible pathways linking periodontitis, diabetes and subsequent CKD.

Materials + Methods

• The study was an observational cohort design, utilising data from the Electric Generation Authority of Thailand (EGAT) project. The EGAT project is a prospective cohort study investigating non-communicable disease risk factors, with health surveys conducted every five years.
• Data from the 2003 survey (EGAT2/2) served as the baseline, with follow-ups in 2008 (EGAT2/3) and 2013 (EGAT2/4).
• Participants were eligible if they attended at least one of the surveys in 2003, 2008 or 2013; completion of all three surveys was not required. Exclusions applied to those with baseline kidney dysfunction (eGFR <60 ml/min/1.73 m²), refusal of periodontal examination, systemic conditions requiring antibiotic prophylaxis, or complete tooth loss.
• Baseline characteristics, health behaviours, and medical history were collected via self-administered questionnaires.
• Physical examinations and fasting blood tests were conducted by trained personnel from Ramathibodi Hospital.
• Periodontal health was assessed by calibrated periodontists from Chulalongkorn University using periodontal probing depth (PPD) and gingival recession (RE) measurements at six sites per tooth. Clinical attachment level (CAL) was calculated, and inter-examiner reliability was assessed using weighted kappa scores. Periodontitis severity was quantified as the percentage of proximal sites with CAL ≥5 mm, and the CDC/AAP case definition was also applied.
• Diabetes was diagnosed based on fasting blood glucose levels ≥126 mg/dl or the use of anti-diabetic medications.
• The outcome of interest was chronic kidney disease (CKD), defined as eGFR <60 ml/min/1.73 m², calculated using the CKD-EPI equation.
• Multiple imputation using chain equations (MICE) addressed missing data, assuming it was missing at random.
• Mediation analysis was performed using structural causal diagrams to explore two pathways:
  • Periodontitis as the independent variable, diabetes as the mediator, and CKD as the outcome
  • Diabetes as the independent variable, periodontitis as the mediator, and CKD as the outcome.
• The generalised structural equation model (GSEM) was used to account for within- and between-participant variability over time.
• The causal mediation effects were estimated using the product of coefficients method, with bootstrap analysis (1,000 replications) to generate bias-corrected confidence intervals.
• Statistical analyses were performed, with significance set at p < 0.05.

Results

• 2,635 participants were included in the final analysis
  • Mean age at baseline: 47.7 years
  • Gender at baseline: 73% male
  • Diabetes at baseline: 7.7%
  • Hypertension at baseline: 27.3%
  • Dyslipidaemia at baseline: 68.6%
  • Moderate periodontitis at baseline: Approximately 50%
  • Severe periodontitis at baseline: Approximately 30%
• Over the 10-year follow-up:
  • 272 new CKD cases were identified, with a cumulative incidence of 10.3 cases per 100 persons.
  • Ten-year cumulative CKD incidence increased with periodontitis severity:
    • 7.2 cases per 100 persons in those with no/mild periodontitis
    • 9.6 cases per 100 persons in moderate periodontitis
    • 13.9 cases per 100 persons in severe periodontitis.
• Missing data ranged from 0.38% to 18.3%, and multiple imputation using the MICE method was applied to address this.
• Causal Diagram A (Periodontitis → Diabetes → CKD):
  • Greater periodontitis extent was significantly associated with diabetes in the adjusted model (p < .001).
  • Diabetes and periodontitis were independently associated with subsequent CKD in the adjusted model.
  • The model estimated that 42.4% (95% CI: 25.4%–65.4%) of the total modelled effect of periodontitis on CKD was mediated through diabetes.
  • For each one-percentage-point increase in proximal sites with CAL ≥5 mm, the estimated CKD odds ratios were 1.010 (95% CI: 1.005–1.015) for the direct pathway and 1.007 (95% CI: 1.004–1.013) for the indirect pathway through diabetes. These describe relative odds, not absolute increases in CKD probability.
• Causal Diagram B (Diabetes → Periodontitis → CKD):
  • Diabetes was associated with a 4.8-percentage-point higher proportion of proximal sites with CAL ≥5 mm after adjustment.
  • The model estimated that 6.5% (95% CI: 2.5%–13.8%) of the total modelled effect of diabetes on CKD was mediated through periodontitis.
  • The total modelled odds ratio for CKD in participants with diabetes was 2.091 (95% CI: 1.519–2.828), compared with those without diabetes.
• Sensitivity analyses showed that results depended partly on the periodontal definition. In diagram A, the direct association was not significant when modified cumulative probing depth was used. In diagram B, mediation through periodontitis was significant for only two of the periodontal measures examined.

Limitations

• CKD was identified using eGFR, without measuring proteinuria, a key marker of kidney damage. The absence of proteinuria data could have impacted the accuracy of CKD classification and affected the effect sizes of associated risk factors.
• Data collection occurred at three time points over a decade, with 5-year gaps between surveys. This interval may have introduced uncertainty regarding changes in health status or risk factors during follow-ups, although time-varying analyses were used to address this limitation.
• The study population, comprising EGAT employees, may limit the generalisability of findings. The cohort had higher education and income levels compared to the general Thai population, which could influence health behaviours and outcomes.
• Other diseases or behaviours may act as additional mediators or moderators. Residual confounding and the assumptions required for mediation analysis limit causal interpretation.
• The mediation percentages are estimates within the specified statistical models, not percentages of CKD cases caused by diabetes or periodontitis. This observational study did not test whether periodontal treatment prevents CKD.

Conclusion

• Periodontitis and diabetes were associated with subsequent CKD, and mediation models suggested possible direct and indirect pathways linking these conditions.
• The findings support further investigation of oral and kidney health, but do not establish that periodontal treatment prevents CKD or slows its progression.