Predicting Laryngeal Involvement in Pemphigus Vulgaris Using a Clinical Risk Score

Importance: Deep laryngeal involvement occurs in 10% to 40% of patients with pemphigus vulgaris and can cause life-threatening airway complications. However, fiber-optic laryngoscopy-the standard detection method-is invasive and causes patient discomfort. Evidence-based criteria for determining which patients require laryngoscopic examination are lacking. Objective: To develop and internally validate a clinical prediction model for deep laryngeal involvement in patients with pemphigus vulgaris. Design, Setting, and Participants: Retrospective cohort study of consecutive patients with pemphigus vulgaris evaluated at Rabin Medical Center, a tertiary dermatology referral center in Israel, between January 2015 and December 2022. All patients had a confirmed pemphigus vulgaris diagnosis based on clinical presentation, histopathology, and direct immunofluorescence as well as complete laryngoscopic examination data. Data analysis was performed from January to November 2024. Main Outcomes and Measures: The primary outcome was deep laryngeal involvement confirmed by flexible fiber-optic laryngoscopy. A prediction model was developed using Firth penalized logistic regression with multiple imputation for missing data and bootstrap validation for internal validation. Results: Among 247 patients with pemphigus vulgaris (mean [SD] age, 61.5 [15.7] years; 135 [54.7%] female), 49 (19.8%) had deep laryngeal involvement confirmed by laryngoscopy. Three clinical predictors were identified: hoarseness (odds ratio [OR], 2.50 [95% CI, 1.10-5.69]), dysphagia (OR, 6.22 [95% CI, 2.78-13.92]), and Pemphigus Disease Area Index score with a significant sex interaction (P = .03). A simplified scoring system stratified patients into low-risk (0-1 points; 0% observed involvement), intermediate-risk (2-3 points; 15.3%), and high-risk (≥4 points; 66.7%) categories. The model achieved an area under the receiver operating characteristic curve of 0.78 (95% CI, 0.71-0.84), with a negative predictive value of 92.6%. Conclusions and Relevance: In this study, the clinical prediction model using routinely assessed parameters achieved good discrimination and high negative predictive value. If externally validated, this tool may help identify low-risk patients who can safely defer laryngoscopy while prioritizing higher-risk patients for evaluation.