Screening of hyperglycaemia in dental settings

Background + Aims

• Diabetes mellitus (DM) is a significant global health issue, often remaining undiagnosed during its asymptomatic phase. Timely detection can prevent complications. Given the bidirectional link between diabetes and periodontitis, dental professionals are uniquely positioned to identify undiagnosed hyperglycaemia using periodontal measures.
• This systematic review investigates whether periodontal indicators (clinical, radiographic, or self-reported) can predict undiagnosed prediabetes or diabetes in adults, aiding early detection and referral for medical evaluation.

Materials + Methods

• This systematic review included studies (cross-sectional, prospective design) developing or validating prediction models using periodontal measures to identify undiagnosed hyperglycaemia.
  • Periodontal measured may have been determined by clinical examinations, radiographs, or self-reports. Whether they were used solely or alongside each other, they were referred to as periodontal prediction model (PPM).
• Studies were excluded if they were case reports, reviews, or studies without periodontal predictors.
• The outcome measure analysed was undiagnosed hyperglycaemia (pre-diabetes and diabetes) determined from laboratory or point-of care (POC) tests

Results

• A total of 10 studies were included and categorised into 2 groups:
  • 8 model development studies
  • 2 external validation studies
• Most models incorporated clinical periodontal parameters (e.g., pocket depth, bleeding on probing, missing teeth) alongside diabetes risk factors.
• Model development studies:
  • Of the included studies, only 2 reported on PPMs:
    • Lalla et al. (2011) – Developed a periodontal prediction model using the number of missing teeth and the percentage of teeth with probing pocket depth (PPD) ≥5 mm. This model achieved an area under the curve (AUC)of 0.65, indicating moderate diagnostic accuracy.
    • Montero et al. (2021) – Utilized the Basic Periodontal Examination (BPE) code as a predictor. The AUC was 0.62, suggesting moderate ability to identify undiagnosed hyperglycaemia.
•  External validation of model
  • Only one of the PPMs were validated (Lalla et al. (2011)). The model was validated in two additional studies:
  • Lalla et al. (2013): Temporal validation within the same dental clinic using the same predictors.
    • The AUC dropped to 0.58, showing reduced diagnostic performance compared to the development sample.
  • Acharya et al. (2018): External validation using electronic health records from primary care clinics in Wisconsin.
    • The AUC was 0.59, again demonstrating a decline in predictive accuracy in a new population.
• These results indicate that while the model worked moderately in its development sample, its performance weakened in other settings, highlighting the need for improvement and robust external validation.

Limitations