A combined PAX1/JAM3 methylation and HPV viral load model for exploratory risk stratification of ≥CIN2 among women with positive cervical screening results

BACKGROUND: High-risk human papillomavirus (Hr-HPV) testing is highly sensitive for cervical cancer screening but lacks specificity, leading to unnecessary colposcopy referrals. Host-gene methylation and HPV viral load capture different aspects of disease risk and may support triage. METHODS: In this prospective real-world triage cohort, 126 women with Hr-HPV positivity and/or abnormal cervical cytology identified through opportunistic cervical screening and clinical referral were enrolled. All participants underwent cytology, HPV genotyping with viral load quantification, and PAX1/JAM3 methylation testing. Histopathology served as the reference standard. A logistic regression model integrating methylation status and viral load was constructed to evaluate its performance for detecting ≥CIN2, with subgroup analysis in ASCUS. RESULTS: Methylation levels, expressed as ΔCt values, decreased with increasing lesion severity and were associated with HPV viral load. In the overall screening-positive cohort, the logistic regression model discriminated ≥CIN2 better than either marker alone. In the ASCUS subgroup, the biomarker-only logistic regression model showed the highest observed AUC among the three evaluated approaches, although this subgroup finding remained exploratory. Scenario-specific strategies showed different trade-offs: the OR rule provided higher sensitivity, whereas the AND rule improved specificity but had lower sensitivity and should be interpreted as a confirmatory triage approach rather than a stand-alone screening rule. CONCLUSION: Combining PAX1/JAM3 methylation with HPV viral load may refine ≥CIN2 risk stratification among women with positive cervical screening results, particularly among women with ASCUS. However, given the referral-enriched nature, limited sample size, and single-center design, the clinical applicability of this approach requires validation in larger, more representative cohorts.