THE USE OF PERIODONTAL MEASUREMENTS FOR IDENTIFYING DIABETES AMONG SAUDI ADULTS

Abstract

Objective: To develop and validate a prediction model, based on clinical and perceived periodontal measures, to identify adults with diabetes. Methods: Two case-control studies were conducted to recruit consecutive adult patients attending primary care clinics at King Khalid University Hospital, Riyadh, Saudi Arabia. Study 1 (n=250) was the sample for model development and Study 2 (n=150) was the sample for external validation. Data were collected using self-administered questionnaires, anthropometric measurements and full-mouth periodontal examinations. The periodontal prediction model (PPM) performance (calibration and discrimination) and its added value when combined with conventional risk assessment tools, the Finnish Diabetes Risk Score (FINDRISC) and the Canadian Diabetes Risk Questionnaire (CANRISK), were assessed adhering to current international recommendations. Results: In the development sample (Study 1), a PPM including 3 periodontal indicators (missing teeth, proportion of sites with periodontal pocket depth [PPD]>6mm and mean PPD) had an area under the curve (AUC) of 0.694 (95% CI: 0.612, 0.776), which fell to 0.618 (95% CI: 0.539, 0.739) during internal validation with 100 bootstrap replications. The AUC of FINDRISC but not of CANRISK improved significantly after adding the PPM to each tool (p=0.048 and 0.114, respectively), correctly reclassifying 26.8% and 9.8% of participants. In decision curve analysis, the addition of the PPM to both tools yielded greater net benefits than using either tool alone, although not through the full range of probability thresholds. In the external validation sample (Study 2), the AUC of the PPM was 0.514 (95% CI: 0.385, 0.642). Neither improvements in the AUC nor in clinical benefit were found when the PPM was added to FINDRISC or CANRISK. The performance of the PPM was improved after update via recalibration of existing periodontal predictors and selection of additional periodontal predictors. Conclusions: This study showed that a PPM consisting of three clinical periodontal measurements could play a role in identifying diabetes among Saudi adults and improving the performance of conventional diabetes risk assessment tools. These encouraging findings were not replicated when the PPM was assessed in an external sample. Although the PPM performance improved after update, further external validation is needed. The present findings should be interpreted with caution as the study recruited consecutive patients from a single site in the city of Riyadh, thus limiting their generalisability.