Establishment of obligate cause of carcinogenic human papillomavirus (HPV) to cervical cancer contributed to an increasing shift from cytology-based cervical cancer screening to HPV-based molecular technique worldwide [123]. A growing body of evidence showed superior sensitivity of HPV testing to cytology for detection of moderate or severe cervical intraepithelial neoplasia (CIN) [456]. Well-organized single round HPV testing could substantially decrease cervical cancer mortality in low-income countries [78].

Currently, much evidence on the clinical value of HPV-based cervical cancer screening is based on studies using hybrid capture 2 (HC2) test (HC2 test; QIAGEN Inc., Hilden, Germany; formerly Digene Corporation), which was the first commercial HPV testing approved by Food and Drug Administration (FDA) and collectively targeted 13 carcinogenic HPV types without discerning individual genotype [910]. SPF₁₀-line probe assay (LiPA) system (DDL Diagnostic Laboratory, Rijswijk, the Netherlands), based on broad spectrum amplification of DNA with SPF₁₀ as primers and reverse linear probe hybridization, has been widely used in clinical trial of HPV vaccination for research of identification of specific sequences in the L1 region of the HPV genome of 28 HPV types [11]. Previous studies have reported good agreements between HC2 and SPF₁₀-LiPA system in detecting carcinogenic HPV genotypes among young women aged 18–25 or under a colposcopy referral setting [1213]. However, few studies have prospectively evaluated the effect of discordant results of HC2 and HPV genotyping by use of SPF₁₀-LiPA on the detection of cervical cancer and precancer, i.e., HC2 positive but non-carcinogenic HPV positive identified by SPF₁₀-LiPA, indicating cross-reactivity in HC2 testing, which would provide more detailed information for clinical performance of HC2 assay in cervical cancer screening.

In the present study, based on a ten-year follow-up cohort among Chinese women, we sought to analyze discordant HPV results between HC2 and SPF₁₀-LiPA testing and particularly focused on the accumulative risk of cervical cancer for those infected with the cross-reacted HPV types of HC2 assay.

In this study, based on a subpopulation from a prospective cervical cancer screening cohort with well-organized follow-up in China, we confirmed a reasonable agreement on detection of carcinogenic HPV between HC2 and SPF₁₀-LiPA testing and revealed the accumulative risks to develop cervical cancer and precancers for women with different testing results between two assays. We took a particular interest in subsequent risks of CIN2+ for women with discrepant assay results. Our data showed that non-carcinogenic HPV types determined by SPF₁₀-LiPA and cross-reacted by HC2, accounted for 8.9% specimens. These cross-reacted HC2 positive women would be confronted with great long-term risks of CIN2+ that should not be ignored, indicating potential benefit from the cross-reactivity. Meanwhile those negative by HC2 albeit carcinogenic HPV positive by LiPA remained low risk of CIN2+, providing great assurance against cervical cancer and precancers by negative HC2 results in cervical cancer screening.

HPV results tested by HC2 assay and SPF₁₀-LiPA system have been previously compared but the agreement rates on carcinogenic HPV were slightly discrepant among studies [1213]. Safaeian and colleagues [13], using the data of women who participated in a randomized, double-blind, placebo-controlled phase III HPV16/18 Vaccine Trial at baseline in Guanacaste, Costa Rica, found a crude agreement of 88% between two assays whereas Chris Perrons, based on women referred to the colposcopy clinic, found an agreement of 70%. By contrast, our study, with the subpopulation composed of all HC2 positive specimens and some randomly selected HC2 negative specimens (the ratio was approximately 2:1) from 2 follow-up visits, revealed an agreement of 73% for carcinogenic HPV, which was extremely close to the findings of Chris Perrons' study [12]. Discrepancies on agreement among studies are probably due to variation of population with different HPV infection rate, sample size, and testing settings [1213].

Age and viral loads, previously reported to be associated with discordance of HC2 and SPF₁₀-LiPA assays among young women [13], were further analyzed in our study among middle-aged women. Although Safaeian and colleagues [13] found that carcinogenic HPV positive identified by SPF₁₀-LiPA but HC2 negative women were older than carcinogenic HPV negative identified by SPF₁₀-LiPA but HC2-positive women, our study did not show such remarkable differences. Instead, we found similar detection rate of carcinogenic HPV at all age levels. Thus, we excluded the potential impact of host characteristics such as age on the analytic performance and confirmed the stability of HPV DNA test. HPV viral load, indicated by signal strength in HC2 test, reflects the productivity of DNA replication in HPV life cycle and defines the course of HPV infection [17]. In our study, HPV viral loads dramatically differed by SPF₁₀-LiPA results in HC2 positive specimens while no similar patterns in HC2 negative specimens. We found that most HC2 positive but LiPA-carcinogenic HPV negative women were infected with low viral loads of HPV, which was consistent with Safaeian and colleagues' findings [13]. Considering that SPF₁₀-LiPA specifically targets the L1 region of HPV genome while HC2 targets the whole length, further investigation into the change of molecular integration status of HPV DNA for specimens with low viral loads might be needed for the interpretation of these findings.

Different sensitivity to detect HC2 targeted genotypes and cross-reactivity with non-carcinogenic HPV types has been reported but reaching no unanimous conclusions [181920]. Our study found that among women positive by LiPA-carcinogenic genotypes, HPV35 and 58 were most likely to test positive by HC2 and among women positive by LiPA-non-carcinogenic genotypes, HPV82 and 73 were most likely to test positive by HC2, which coincided with the findings by Castle et al. [1920] using the PGMY09/11 L1 primer-based genotyping assays to detect the genotypes and with those by Preisler et al. [18] using CLART HPV assay to detect the genotypes. All HPV45 positive women were identified by HC2 in our study, however, only 72.7% were HC2 positive in Castle's study [19]. A possibility of 84.2% for HPV56, the second mostly likely HPV types detected by HC2, were identified in Castle's study [19], which was 74.5% in our study. Besides, other common non-carcinogenic types cross-reacted by HC2 were found to be HPV26, 69, 71, and 40 in our study but HPV11, 61, 67, 70, and 81 in Castle's study [20]. As a probable carcinogenic but not HC2-targeted HPV [21], HPV66, frequently reported to be cross-reacted by HC2 [202223], had the likelihood of 58.3% to be tested positive by HC2 in our study. The probability to cross-react with non-carcinogenic HPV genotypes greatly depends on analytic sensitivity of the intended tests used for HPV genotyping, which partially interpreted the discrepant findings among studies [181920]. Meanwhile cross-reactivity of HC2 with any non-carcinogenic type also disclosed a predilection of HC2 probe set to cross react with the HPV phylogenetic related clades that led to lack of perfect fidelity of the HC2 probe set [24].

The clinical implication of cross-reactivity by HC2 has been previously evaluated by calculating the sensitivity and specificity for the detection of high-grade cervical neoplasia from a cross-sectional perspective and little effect of cross reactivity with non-carcinogenic HPV types was found on the overall clinical performance of HC2 as a general screening test [20]. Few studies, however, gauged the potential clinical significance of this cross-reactivity of HC2 with non-carcinogenic HPV types for the prospective risk prediction of HPV testing for general screening. In our study those HC2 positive and LiPA-non-carcinogenic HPV positive women presented a noticeable subsequent CIN2+ risk within 10 years (9.97%; 95% CI=8.57%–11.50%), which indicated that the cross-reactivity of HC2 might be useful for the detection of the cervical cancer and precancers to some extent. This is perhaps not surprising as many cross-reactive genotypes were probably or possibly carcinogenic, i.e., HPV82, 69, or 53 [15].

In addition to cross-reactivity by HC2, we noticed an extremely low risk of developing CIN2+ for those HC2 negative but LiPA-carcinogenic HPV positive women, which indirectly coincided with the overall good assurance of HC2 against cervical cancer screening [30]. Considering the positive cutoff point for HC2 at 1 pg/mL (signifying at least 5,000 viral copies), this finding corroborated other previous studies that HPV viral copies (viral loads) and carcinogenicity of genotypes were 2 concurrent essential risk factors for the progression to high-grade cervical lesions [2526]. Interestingly, 9.2% women negative by any HPV types by LiPA but HC2 positive were also at a low risk of CIN2+ within subsequent 10 years in our study. Although it remained uncertain whether other types not detected by LiPA caused cross-reactivity and led to false positive rate to some extent, this additional finding indicated these small fractions of HC2 positive women might cause unnecessary referral to colposcopy and reduce the specificity of detecting high-grade cervical lesions. Given more preference attached to sensitivity than specificity on HPV testing, specific biomarkers were required to preclude them from referral or further follow-up. Besides, we confirmed that a HC2 negative test could provide enough assurance against cervical cancer and precancer at least ten years, which were consistent with other publications [272829].

There are several limitations in our study. Primarily, we randomly selected a small fraction of HC2 negative specimens for HPV genotyping using the SPF₁₀-LiPA system, which might lead to selection bias and reduce clinical specificity to CIN2+. However, our focus was to evaluate the long-term risk of women with discrepant assay results, our conclusions would not be influenced to some extent. Secondly, a validation study on specimens with discordant results should be conducted, i.e., using such accurate technology as HPV genome sequencing, which may provide more definite genotyping information. Finally, few CIN3 and cancer cases were identified over the course of 10-year follow-up for analysis of long-term CIN3+ risk even though this population-based cohort was established in a high prevalent area of cervical cancer because women were recommended for treatment immediately after CIN2+ were diagnosed.

In conclusion, based on a prospective cervical cancer cohort, we demonstrated that a HC2 positive result predicted high risk of developing cervical cancer and precancers, even though cross-reacted with untargeted HPV types. Meanwhile a HC2 negative test provided enough assurance against subsequent cervical cancer and precancers. These findings verified the good clinical performance of HC2 in cervical cancer screening.