Development and internal validation of a prognostic model for predicting periodontitis progression using clinical and behavioural factors

To develop and internally validate the PERI-Score, a pragmatic prognostic index for estimating 36-month periodontitis progression from routinely available clinical, systemic, and behavioural information. In this prospective single-centre cohort study, 350 consecutive adults with periodontitis received non-surgical periodontal therapy and were followed for 36 months during supportive periodontal therapy (SPT). Progression was operationally defined as clinical attachment loss, represented by an increase (worsening) in clinical attachment level (CAL) of ≥ 2 mm at ≥ 20% of assessable matched sites. The PERI-Score was developed using Cox proportional hazards modelling. Performance was evaluated using a stratified 70:30 development/internal-validation split and 1,000 bootstrap resamples; both subsets originated from the same cohort and therefore represent internal validation. A total of 118 progression events were observed (33.7%). The strongest predictors were Stage III/IV periodontitis (HR 2.64), active smoking (≥ 10 cigarettes/day; HR 2.31), PPD ≥ 5 mm at ≥ 4 sites (HR 2.18), SPT non-compliance (< 75% attendance; HR 2.08), and Grade C periodontitis (HR 1.92), with additional contributions from HbA1c ≥ 7.0%, CAL ≥ 5 mm at ≥ 4 sites, BOP ≥ 30%, Class II/III furcation involvement, FMPS ≥ 30%, and age ≥ 50 years. The internal-validation AUC was 0.843 (95% CI 0.786-0.900), and observed progression increased across the three PERI-Score strata. The PERI-Score is an internally validated candidate prognostic model for 36-month progression during treated periodontal care. It combines baseline disease severity and susceptibility with modifiable systemic and behavioural factors and a longitudinal marker of SPT adherence. It may support risk communication and structured monitoring, but independent multicentre external validation is required before its thresholds are used to determine treatment or recall protocols.