Background
Uterine cervical cancer is one of the most leading causes of cancer death among women worldwide, especially among women in developing countries1,2. According to the updated data from International Agency for Research on Cancer (IARC)3, it is estimated that of all the 604,127 new cases and 341,831 deaths4,5 of cervical cancer worldwide in 2020, and more than 1/6 of those cases occurred in China6. It was well acknowledged that persistent infection with high-risk human papillomavirus (hrHPV) is the cause of almost all cervical cancer7. However, only a small proportion of hrHPV infections persist8 and develop into cervical squamous intraepithelial lesion (SIL) which may, if left untreated, progress to cancer.
In 2019, the World Health Organization issued a call for action to eliminate cervical cancer9, and then advocated a series of approaches includes increasing HPV vaccine coverage10, increasing screening of women aged more than 30 years with hrHPV testing and treatment for hrHPV-positive (hrHPV+) women that are suspicious of cervical cancer precursor lesions. However, great discrepancies exist in the popularity and quality of screening methods. A combination of hrHPV+ DNA testing and cytology detection has been implemented as a routine classifier for cervical cancer screening in many countries, there are still several limitations including that cytology is subjective and requirement for pathologists, and hrHPV test with low specificity and might result in high colposcopy referral 8. Although colposcopy is helpful to detect the cervical lesion, most HPV infections will not give rise to (pre)malignant disease, increased unnecessary colposcopy referrals would lead to unnecessary treatment and further would negatively affect childbearing11,12. Thus, an adequate screening classifier is urgently needed for women with abnormal HPV and cytology results13.
An objective triage strategy which could be automated and incorporated molecular test combining with HPV detection might be able to solve these issues. As a primary form of epigenetic inheritance14, DNA methylation15,16 has been extensively studied and widely used for tumor classification17,18, early detection19, therapy target, and predictive biomarker20,21 of metastasis and recurrence22. What’s more, methylation assays can be automated, have accurate quantitation, are robust to operator variations and can be performed in the same specimen as the HPV testing. Of the more than 100 human methylation biomarker23,24 genes detected so far in cervical tissue13, several biomarkers have been repeated shown to have elevated methylation in cervical cancers and high-grade squamous intraepithelial lesion, which including ZNF67117, TERT20, SOX120, CADM125, MAL25, FAM19A426, miR-124-227, PAX128, JAM329, and EPB41L3 30.
The S5 methylation classifier is a test based on DNA methylation of the late regions L1 and L2 of HPV16, HPV18, HPV31 and HPV33 combined with the promoter region of human tumor suppressor gene EPB41L330 that identifies women with HSIL or more worse lesions. Here, S5 was developed in a colposcopy study, and well tested in the UK, Canada, and Mexico.
Although earlier studies in developed countries have shown that the performance of the S5 classifier to detect precancer lesion, there are few studies validating S5 in developing countries and such low and middle incomes settings. In this study, we figured out the efficiency and potential of S5 methylation in a screening population in China, including the 2246 women with abnormal results between HPV and cytology (2246 women who had any abnormal results for HPV and cytology). We compared the ability of the S5 methylation, repeated conventional cytology and HPV16/18 genotyping in detecting HSIL+ cases at the timepoint of 6 months, 12 months, 18 months and 24 months within the 2-year endpoint, among women with abnormal results between HPV and cytology selected from a multi-center study (the study recruited women from routine opportunistic screening services of China, a middle-income country) and attempted to find the evidence to support S5 as an optimal triage classifier.