Developing an automated cytology-based DNA ploidy test for detecting oral squamous cell carcinoma in Taiwan: A...

Oral squamous cell carcinoma (OSCC) remains a global health issue, particularly in underserved settings such Taiwan where access to early diagnosis is limited. DNA ploidy-based oral cytology has shown strong diagnostic performance, but its applicability across populations require evaluation.

In this prospective feasibility study, brush and oral rinse samples were collected from patients with OSCC, low-grade dysplasia, reactive lesions, and normal mucosa. DNA cytometry was performed to quantify ploidy profiles. Diagnostic performance was evaluated using published algorithm and cohort-specific thresholds derived from stratified 10-fold cross-validation.

A total of 118 patients were included (50 OSCC, 41 normal, 22 reactive lesions, and 5 low-grade dysplasia). Applying the published algorithm to brushing samples yielded high sensitivity (0.97), but much poor specificity (0.27), highlighting cohort-specific limitations. New thresholds were therefore derived using 10-fold cross-validation based on the percentage of abnormal cells. Among 82 adequate brush samples, a threshold of 3.7% abnormal cells achieved 0.75 sensitivity 0.75 and 0.86 specificity. For 95 adequate rinse samples, a 1.4% threshold yielded 0.77 sensitivity and 0.80 specificity. Thresholds based on non-diploid cells performed worse. Covariate-adjusted logistic regression further improved sensitivity to 0.94 (brush) and 0.87 (rinse). In 105 paired samples, rinse testing demonstrated fewer inadequate samples (12.5% vs. 22.5%) and higher cell yield.

This is the first study assessing automated DNA cytology in an areca nut-prevalent population. Findings emphasize the need for population-specific calibration and support rinse-based cytology as a scalable self-administered screening approach. Large multicenter validation is warranted.