ABSTRACT
Cervical cancer screening is based on cytologic analysis and high-risk human papillomavirus (HR-HPV) testing, each having their drawbacks. Implementation of new biomarker-based methods may improve screening accuracy. Here, the levels of 25 microRNAs (miRNAs, miRs) and 12 mRNAs involved in cervical carcinogenesis in 327 air-dried Papanicolaou-stained cervical smears from patients with cervical precancerous lesions, cancer, or without the disease were estimated by real-time PCR. Using logistic regression analysis, small-scale miRNA-based, mRNA-based, and combined molecular classifiers were built based on paired ratios of miRNA or mRNA concentrations; their ability to detect high-grade cervical lesions and cancer was then compared. Combined mRNA-miRNA classifiers manifested a better combination of sensitivity and specificity than miRNA- and mRNA-based classifiers. The best classifier, combining miR-375, miR-20, miR-96, CDKN2A, TSP4, and ECM1, predicted high-grade lesions with diagnostic sensitivity of 89.0%, specificity of 84.2%, and a receiver-operating characteristic area under the curve of 0.913. Additionally, in a subsample of the same specimens, the levels of MIR124-2 and MAL promoter methylation, HR-HPV genotypes, and viral loads were analyzed. The relative high-grade lesion risk estimated by the classifier correlated with the frequency of MAL and MIR124-2 methylation but not with the HR-HPV genotype or viral load. The results support the feasibility of cellular biomarker-based methods for cervical screening and patient management.
