Abstract
OBJECTIVE: To compare the performance of cytology, HPV16/18 genotyping and PAX1/SOX1 methylation for the triage of high-risk HPV-positive cervical samples. DESIGN: Retrospective analyses of archival samples collected from a large-scale prospective randomised controlled trial. SETTING/SAMPLE: HPV-positive women recruited from the general cervical screening population. METHODS: 403 HPV-positive samples including 113 normal, 173 low-grade cervical intraepithelial neoplasia (LG-CIN), 114 HG-CIN and three cervical cancers. All samples were assessed by liquid-based cytology, HPV genotyping and PAX1/SOX1 methylation. MAIN OUTCOME MEASURES: AUC (area under the curve), sensitivity and specificity for cytology, HPV16/18 genotyping and PAX1/SOX1 methylation for high-grade (HG) premalignant cervical lesions. RESULTS: PAX1 was more sensitive than cytology and HPV16/18 genotyping in detecting a HG lesion (CIN2+). The sensitivity for PAX1, SOX1, cytology and HPV16/18 were 73.5% (95% CI: 65.5-81.5), 41.9% (95% CI: 32.9-50.8), 48.7% (95% CI: 39.7-57.8) and 36.8% (95% CI: 28.0-45.5), respectively, and their respective specificities were 70.3% (95% CI: 65.0-75.6), 83.6% (95% CI: 79.3-87.9), 77.6% (95% CI: 72.8-82.5) and 67.1% (95% CI: 61.7-72.6), respectively. Overall, PAX1 gave the best AUC at 0.72. Adding SOX1 to PAX1 did not improve the AUC (0.68). Three hundred and twenty-two women who did not have a HG lesion at baseline were followed up for two rounds of screening. Fewer women developed a HG lesion with a normal baseline PAX1 compared to women with a normal baseline cytology or negative HPV16/18 (8.4% vs. 14.5% and 17.5%, respectively). CONCLUSION: PAX1 triage for referral to colposcopy in HPV-positive women may be superior to cytology and HPV16/18 genotyping.