Methylation estimates the risk of precancer in HPV-infected women with discrepant results between cytology and HPV16/18 genotyping

Abstract

Background: Vigilant management of women with high-risk human papillomavirus (hrHPV) is necessary in cancer screening programs. To this end, we evaluated the performance of S5 (targeting DNA methylation in HPV16, HPV18, HPV31, HPV33, and human gene EPB41L3) to predict cervical intraepithelial neoplasia grade 2 or higher (CIN2+) in a sample of hrHPV-infected women referred to colposcopy in the FRIDA Study, a large screening trial in Mexico. A nested case-control sample with women referred to colposcopy either by atypical squamous cells of undetermined significance or higher (ASCUS+) in cytology and/or positive for HPV types 16 or 18 was tested by S5. Seventy-nine cases of CIN2+ were age-matched to 237 controls without a diagnosis of CIN2+ (<CIN2). DNA from exfoliated cervical cells was bisulfite converted and PCR amplified for S5 targets, and methylation was quantified at specific cytosines by pyrosequencing.

Results: The S5 classifier separated women with CIN2+ from <CIN2 with a highly significant area under the curve (AUC) of 0.75 (95% CI 0.69-0.82), while AUC for CIN3+ was 0.81 (95% CI 0.74-0.89). To optimize sensitivity and specificity for Mexico, an alternative S5 cutoff of 3.7 was implemented to account for overall higher methylation seen in our already triaged women. All three invasive cancers were detected by methylation or HPV16/18 but none by cytology. Sensitivity of S5 for CIN2+ was 62% (95% CI 50.4-72.7%), specificity was 73% (95% CI 66.9-78.5%), and adjusted PPV was 15.1% (95% CI 12.0-18.3%). In contrast, the crude sensitivity of HPV16/18 detection and cytology were 63.3% (95% CI 51.7-73.9%) and 57.0% (95% CI 45.3-68.1%) respectively; specificity was 29.1% (95% CI 23.4-35.3%) and 62.4% (95% CI 55.9-68.6%) respectively, while adjusted PPV was 6.4% (95% CI 4.9-8.1%) and 10.5% (95% CI 8.0-13.1%), respectively. Methylation testing could reduce colposcopy referrals by 30 to 50% with virtually no loss of sensitivity for CIN2+ and CIN3+.

Conclusions: S5 testing on hrHPV-positive women significantly increased diagnostic information compared to triage by HPV16/18 plus cytology and appears to have clinical utility as an additional test to substantially lessen burdens on colposcopy.

Trial registration: The FRIDA Study is registered in ClinicalTrials.gov , number NCT02510027.

Keywords: Cervical cancer; Cervical intraepithelial neoplasia; DNA methylation; EPB41L3; Human papillomavirus; S5 classifier; Triage.

Conflict of interest statement

JS and ELP have received funding for research projects from Qiagen, Roche, Beckton Dickinson, Gen-Probe, DICIPA, and Arbor Vita. The other authors declare no competing interests.

Figures

Fig. 1

Consort diagram of FRIDA nested case-control triage study showing the numbers of women in each step. Triage positive women included HPV16/18 positive (n = 508), ASCUS+ (HPV16/18 negative) (n = 277), and HPV16/18 positive and ASCUS+ (n = 78). *Prior to July 2015, 672 out of 863 triage-positive women attended the colposcopy evaluation. Then, 561 women out of 672 who underwent colposcopy had histology results recorded from April 2013 to the time of the study cutoff date on July 15. These 561 women are represented in our sampling frame, from which we selected all the CIN2+ cases (79) and a random selection of three controls per case matched by age (1 CIN1 and 2 NEG). In total, 316 participants were included in our methylation analysis: 79 CIN2+ cases, 79 CIN1, and 158 NEG. The cases included all the CIN2+ detected until July 2015. The three controls per case were randomly selected and matched by age from women with histological diagnoses of CIN1 or less. The remaining 245 triage-positive women with histology results of CIN1 and negative would have been controls but were not selected by the sampling method as we already had adequate power for the study. Abbreviations: hrHPV, high-risk human papillomavirus; HPV16/18, human papillomavirus type 16 or type 18; ASCUS+, atypical squamous cells of undetermined significance or worse; NEG, histologically negative; CIN, cervical intraepithelial neoplasia (of grades 1, 2, and 3)

Fig. 2

Comparison of S5 methylation classifier in histologically negative (NEG), CIN1, CIN2, CIN3, and cervical cancer cases (CC). The S5 classifier was significantly different between the following group comparisons: NEG vs CIN2 (p = 0.01), NEG vs CIN3 (p < 0.001), NEG vs CC (p < 0.001), CIN1 vs CIN2 (p = 0.03), CIN1 vs CIN3 (p < 0.001), CIN1 vs CC (p = 0.004), CIN2 vs CC (p = 0.007), and CIN3 vs CC (p = 0.02). Other comparisons were not significant (NEG vs CIN1 and CIN2 vs CIN3). Abbreviations: NEG, histologically negative; CIN, cervical intraepithelial neoplasia (of grades 1, 2, and 3); CC, cervical cancer. The top of box represents the upper quartile (p75), bottom the lower quartile (p25), and the line the median (p50). The upper whisker extends to the largest point of the inter-quartile range from the upper quartile. The lower whisker extends to the smallest point of the inter-quartile range from the lower quartile. The outliers are plotted as individual points for each lesion grade. The Cuzick test for trend was highly significant (p < 0.001)

Fig. 3

Receiver operator characteristic (ROC) and area under the curve (AUC) of S5 methylation for detecting CIN2+ or CIN3+. The blue diamond denotes the sensitivity and specificity of S5 at a cutoff of 3.7 for CIN2+. The red circle denotes the sensitivity and specificity at the S5 cutoff of 0.8 predefined for use in the UK for CIN2+. The cutoff for cytology alone was ASCUS+. Abbreviations: ASCUS+, atypical squamous cells of undetermined significance or worse

Fig. 4

Benefit of using the S5 classifier as a second triage test for colposcopy referral for a) CIN2+ and b) CIN3+ endpoints. S5 helped reduce unnecessary colposcopy referrals in both the “HPV16/18 neg, ASCUS+” and “HPV16/18 pos, normal cytology” groups. Under the current Mexican algorithm, all HPV16/18 positive and/or ASCUS positive are referred to colposcopy, but we propose to use S5 as second triage test for the discrepant triage groups (“HPV16/18 neg, ASCUS+” and “HPV16/18 pos, normal cytology”) which then decreases the false-positive rate. We therefore defined the following hypothesis to test S5 benefits. In group 1, women were HPV16/18 positive and ASCUS positive and must be referred to colposcopy without any additional triage procedure. In group 2, women were HPV16/18 negative and ASCUS positive and were called for colposcopy, but if we had used S5 methylation as a second triage test, it would have reduced by 50% (CIN2 endpoint) or 43% (CIN3 endpoint) the number of false positives referred to colposcopy. Using S5 methylation as a second triage for women in group 3, who were HPV16/18 positive with normal cytology, would have reduced unnecessary colposcopy referrals by 30% in CIN2+ and 28% in CIN3+. Abbreviations: ASCUS+, atypical squamous cells of undetermined significance or worse; CIN, cervical intraepithelial neoplasia (of grades 1, 2, and 3). The frequency shows the absolute number of women in each group