Genome-wide association analysis in host genetic characteristics of progression to high-grade cervical intraepithelial neoplasia or higher for women with human papillomavirus infection and normal cytology.

Abstract

6033 Background: Human papillomavirus (HPV) testing is widely used for cervical cancer screening. The hazard ratio of developing cervical intraepithelial neoplasia grade 2 or higher (CIN2+) in HPV-positive/ normal cytology women is 20–34 fold as compared to those with HPV-negative/normal cytology. HPV-positivity would cause substantial anxiety. Apart from viral factors such as high-risk (hr) types, it is important to identify host characteristics for predicting outcome. Methods: An initial genome-wide association study (GWAS) of single nucleotide polymorphisms (SNPs) by Affymetrix Axiom™ Genome-Wide Human Arrays was conducted on 505 cases with histological diagnosis of CIN2+ (group D1) versus 920 female controls. An additional set of 2315 female controls from the Taiwan Biobank genotype array were added in the discovery stage. The identified 29 CIN2+ -associated SNPs from GWAS (p < 5 x 10−6) were verified in an independent cohort (group D2 [n = 306] versus group N [n = 600]). Group N were HPV-negative/normal cytology women from a population-based cervical cytology and HPV co-test study. A cohort with HPV-positive/normal cytology (group P, n = 755) underwent follow-up and was served as the prediction set. The predictive validity was analyzed by logistic regression and receiver operating characteristic (ROC) curve analysis. Results: Thirty-three individuals of the group P progressed to CIN2+ (median follow-up: 23.7 months, range 4.0–122.1). A risk-predictive panel of 8 SNPs rs3097662, rs35979982, rs7763822, rs4282438, rs3128927, rs7759943, rs213194, rs17835649 which were significant in the replication (p < 0.05) was used to train models for disease risk prediction using the combination of GWAS and verification sets. Two prediction models were finalized and determined using 7 SNPs for hr- and low-risk (lr) HPV groups respectively (sensitivity 0.72 and 0.75, specificity 0.651 and 0.884, area under the ROC curve 0.703 and 0.701). Among group P with hr-HPV, those carried < 6 risk-alleles had significantly decreased hazard (log-rank p < 0·001) of progression to CIN2+ than those with ≧6 risk-alleles, while among group P with lr-HPV, those with predictive probability of ≥ 0·095 had a cumulative risk of progression of 10% at 3 years. Conclusions: Two risk-predictive SNP panels including 7 SNPs with hr- or lr-HPV groups can assist risk stratification among HPV-positive/ normal cytology women. These panels could be further tested in other ethnic populations.