Development of a Clinical Prediction Model for Systemic Therapy Escalation in Oral Lichen Planus: A Two-Center Retrospective Cohort Study

Background and Objectives: Oral lichen planus (OLP) is a chronic immune-mediated disorder in which the need for systemic therapy remains difficult to predict using standardized clinical criteria. This retrospective two-center cohort study aimed to develop a multivariable clinical prediction model for systemic therapy escalation in patients with histopathologically confirmed OLP. Material and Methods: Medical records of 141 adult patients treated between 2022 and 2025 were reviewed. Baseline demographic, clinical, and laboratory variables were analyzed using univariate analyses, followed by multivariable logistic regression with backward stepwise selection and Firth penalized likelihood estimation. Discrimination was assessed by the area under the ROC curve (AUC), and internal validation was performed by bootstrap resampling (1000 samples) with repetition of the full selection procedure. Results: Four independent predictors of systemic therapy were identified: concomitant cutaneous lichen planus (penalized odds ratio (OR) 13.96, 95% CI 3.94–55.84), autoimmune thyroiditis (OR 12.67, 3.26–54.88), genital lichen planus (OR 10.21, 2.36–51.57) and oral Candida superinfection (OR 4.48, 1.10–18.82). The apparent AUC was 0.889 (95% CI 0.801–0.976); after bootstrap internal validation, the optimism-corrected AUC was 0.84 and the calibration slope 0.73, indicating moderate overfitting. Candida superinfection was the least stable predictor (selected in 43% of bootstrap samples). A prognostic risk matrix was developed to estimate the probability of systemic therapy escalation according to different combinations of predictors. Conclusions: The identified clinical factors and derived risk matrix may provide a preliminary framework for risk stratification and individualized therapeutic decision-making in patients with OLP. However, given the retrospective design and limited number of outcome events, these findings should be interpreted cautiously. Prospective external validation in larger independent cohorts is required before the model can be considered for routine clinical use.