# Optimizing Oral Cancer Screening: Latent Class Analysis of Chairside Adjuncts in a High-Risk Dental Cohort

## TL;DR

This study evaluates diagnostic tests for oral cancer and finds that combining exams improves detection accuracy in high-risk patients.

## Contribution

A novel two-tier screening strategy using latent class analysis improves oral cancer detection in dental practice.

## Key findings

- Brush cytology showed best standalone accuracy with 83.7% sensitivity and 80.7% specificity
- Combining conventional exam with autofluorescence followed by brush cytology optimizes detection
- Latent class analysis provides unbiased performance estimates without relying on single reference test

## Abstract

Introduction: Oral cancer remains a major public health challenge in high-burden regions, and late diagnosis contributes to poor survival. Conventional oral examination (COE) may be limited by subjectivity and low specificity, prompting the evaluation of adjuncts, such as toluidine blue (TB) staining, autofluorescence imaging, and oral brush cytology. This study applied latent class analysis (LCA) to estimate the unbiased diagnostic performance of multiple chairside tests in a single cohort.

Materials and methods: A cross-sectional diagnostic study was conducted on 100 adults (aged 18-70 years) with visible oral lesions. Sequential testing included COE, VELscope autofluorescence (LED Dental Inc., White Rock, British Columbia, Canada), OralCDx brush cytology (CDx Diagnostics, Suffern, NY, USA), and 1% TB staining, which were performed under standardized conditions by calibrated examiners (kappa > 0.80). Incisional biopsy was performed for lesions that were positive in ≥2 tests. LCA modeled the true disease status without universal verification; traditional metrics used histopathology as a reference.

Results: Histopathology confirmed dysplasia or malignancy in 43 patients (43%). COE showed high sensitivity (83.7%), but low specificity (54.4%). TB staining yielded a sensitivity of 69.8% and specificity of 86.0%, and autofluorescence showed a sensitivity of 60.5% and specificity of 89.5%. Brush cytology achieved a balanced accuracy (sensitivity 83.7%; specificity 80.7%), with the strongest correlation with histopathology (r = 0.64). LCA identified two latent classes, brush cytology and TB, demonstrating superior class discrimination. The inter-test agreement was highest between COE and TB (r = 0.61).

Conclusions: No single test was found to be optimal. Brush cytology offered the best standalone accuracy, whereas COE and autofluorescence served as sensitive initial screens. A two-tier sequential strategy comprising COE/autofluorescence followed by brush cytology maximizes case detection and reduces false positives. The LCA provides robust and unbiased estimates for real-world screening. Integration into routine dental practice with structured training can improve early detection of oral dysplasia in high-risk populations.