Clinical Prediction Model for Anal High-Grade Squamous Intraepithelial Lesions Risk.

Abstract

Objective: Timely treatment of anal high-grade intraepithelial lesions (HSIL) prevents progression to anal cancer. Available screening tools (anal Pap test and high-risk human papillomavirus testing) have inconsistent and suboptimal performance, often leading to overreferral to high-resolution anoscopy, the gold standard test for HSIL diagnosis. We aimed to develop and externally validate a clinical prediction model for histologic HSIL to improve triage to high-resolution anoscopy among individuals at increased risk for anal cancer.

Materials and methods: Medical records from 2 institutions were reviewed to identify candidate predictors of histologic HSIL. A penalized logistic regression model with elastic net regularization was developed and internally validated with five-fold cross-validation. External validation was performed in a third institution cohort. Candidate predictors were age, sex, HIV status, history of anogenital HPV-related disease, immunosuppressant use, anal cytology, anal high-risk HPV (hrHPV) status, and interaction terms (HIV status*hrHPV infection) and (HIV status*history of anogenital HPV-related disease).

Results: The derivation dataset included 536 patients, 382 (71.3%) were people living with HIV, 168 (31.3%) were women, and HSIL prevalence was 21.1%. The area under the ROC on the derivation dataset was 0.80 (95% CI = 0.69; 0.90). The external validation dataset included 242 patients, 159 (65.7%) people living with HIV, 18 (7.4%) women, with HSIL prevalence of 37.2%. The final model included age, sex, anal cytology, anal hrHPV, immunosuppressant use, history of anogenital HPV-associated disease, and the 2 interaction terms. The area under the receiver operating characteristic (ROC) on the external validation dataset was 0.73 (95% CI = 0.67; 0.80).

Conclusions: This clinical prediction model demonstrated a promising performance and included objective factors that are easily obtained.