Diagnostic accuracy of extended HPV DNA genotyping and its application for risk-based cervical cancer screening strategy

Objectives: To evaluate the consistency of 14 high-risk HPVs (hr-HPVs) detection between extended HPV DNA genotyping and a well-validated partial HPV genotyping kit, and to explore the diagnostic accuracy of risk strati ﬁ cation strategy based on extended HPV genotyping for cervical cancer (CC) screening. Methods: Baseline data from a clinical trial of recombinant HPV 9-valent vaccine in China was analyzed. All enrolled women aged 20 – 45 years received cervical cytology, HPV detection by extended and partial HPV genotyping kits. Those who met the indications would further receive colposcopy. The primary endpoints were cervical intra-epithelial neoplasia 2/3 or worse (CIN2 + /CIN3 + ). Results: A total of 8,000 women were enrolled between April 2020 and July 2020 and 83/33 cases were diagnosed as CIN2 + /CIN3 + . The overall agreement between the extended and partial HPV genotyping was 92.66 %. And the agreement further increased with the progression of lesions, which lead to similarly high sensitivity and negative predictive value of these kits. A strati ﬁ ed triage strategy of CC screening was constructed based on the immediate CIN2 + /CIN3 + risk of speci ﬁ c HPV. Compared with the conventional HPV primary CC screening strategy, the risk-based strategy had higher speci ﬁ city for CIN (CIN2 + : 94.84 vs. 92.46 %, CIN3 + : 96.05 vs. 91.92 %), and needed fewer colposcopies for detecting one cervical disease. Conclusions: Extended HPV genotyping had good agreement with a well-validated partial HPV genotyping CC primary screening kit in hr-HPV detection. Extended HPV genotyping could facilitate risk-based strati ﬁ ed management strategy and improve the diagnostic accuracy of primary CC screening


Introduction
Cervical cancer (CC) is the fourth most common female papillomavirus (hr-HPV) genotypes infection is considered to be the main cause of high-grade squamous intraepithelial lesion and CC [2].Although HPV vaccination is an effective primary prevention strategy for CC, CC screening remains an important approach for early detection of cervical precancerous lesions and cancers, given the short supply of vaccines in China.As a conventional CC screening strategy, HPV primary screening triaged by cervical cytology is recommended by multiple international guidelines [3][4][5].
Extended HPV genotyping was thought as a highly promising method for risk assessment of specific HPV, since it could distinguish each specific HPV.But, there is no FDA-approved extended HPV genotyping kit for CC screening up to date.A number of clinical trials for varied extended HPV genotyping kits have been completed or are undergoing in China.Yaneng extended HPV DNA genotyping kit (BioScience, Shenzhen, China) could simultaneously detect 14 hr-HPVs (same as Cobas 4,800), 4 medium-risk HPVs (26, 53, 73, and 82), and 5 low-risk HPVs (6,11,42,43, and 81) separately, which covered almost all oncogenic HPV genotypes [8].Previous studies have suggested that Yaneng extended HPV DNA genotyping kit displayed good agreement with Cobas 4,800 and Cervista (Hologic, Madison, WI, USA) in the detection of 14 hr-HPVs, but the oncogenic risk of varied hr-HPV genotypes has not been further explored [9,10].Thus, present study tried to evaluate the consistency of 14 hr-HPVs detection between extended and partial HPV DNA genotyping and explore the CC screening performance of extended HPV DNA genotyping using risk stratification strategy of specific hr-HPV infection among women aged 20-45 through a large-scale population-based cross-sectional study.

Study design
Totally 8,000 women were recruited for evaluating the efficacy of recombinant human papillomavirus 9-valent vaccine, in a multicenter phase III randomized clinical trial approved by NMPA and sponsored by Bovax Biotechnology (Shanghai, China, NCT04422366).The women received 3 doses of HPV vaccine at baseline, 2nd, and 6th month, and would be followed up with cytology and HPV detection every half year for 5 years.This trial was approved by the Ethics Committee of Guangxi Center for Disease Prevention and Control (approval number: GXIRB2019-0044-1).Eligible criteria included healthy women aged 20-45 years with sexual history, who did not report previous abnormal results on CC screening and were willing to provide written informed consent.Women who had a history of HPV infection, abnormal cervical cytology, cervical intraepithelial neoplasia (CIN) or worse, vulvar intraepithelial neoplasia or worse, vaginal intraepithelial neoplasia or worse, HPV vaccination, pregnant or lactating, or with other vaccine contraindications were excluded.Only baseline data were collected and analyzed in present study.
At baseline visit, two cervical samples were collected including one for cervical cytology (Hologic, USA) and Cobas 4,800 partial HPV genotyping detection, and another for Yaneng extended HPV genotyping test according to the manufacturer's instructions.Cytology was reported using the classification of the Bethesda System [11].Women who met the following criteria would receive colposcopy at baseline visit: (1) HPV 16/18 positive; (2) abnormal cytology worse than atypical squamous cells of undetermined significance (>ASC-US) cytology; (3) Other 12 hr-HPV genotypes positive and ASC-US cytology.Due to the ethical consideration, women with normal cervical cytology and negative HPV/other 12 hr-HPV genotypes positive test were considered lesion-free at baseline visit and would be followed up at the scheduled time.The primary endpoints were histological cervical intraepithelial neoplasia 2/3 or worse (CIN2+/CIN3+).

Statistical analysis
The consistency of both HPV detection methods was evaluated by Cohen's kappa statistics [12].Diagnostic accuracy was assessed by sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), and 95 % confidence intervals (95 % CI) were computed using Wilson method.Chi-square test was used for pairwise comparison of two detective methods and p<0.05 was recognized as statistically significant.
The prevalence of hr-HPV genotypes by extended HPV genotyping and their immediate risk of CIN2+/CIN3+ were calculated by three previously published approaches [13][14][15], including minimum estimate (Min.) that was only for single HPV infection, any type estimate (Any.) was used for all HPV infection and repeated calculation for multiple infections, and hierarchical attribution estimate (Hier.) was to attribute the multiple HPV infections to a specific HPV genotype according to the HPV ranking of Any.Hr-HPV genotypes were further grouped into different risk groups by their immediate risk of CIN2+/CIN3+ based on extended HPV genotyping combined with cytology.All statistical analyses were conducted using SAS 9.4.

Demographics and clinical information of study population
As shown in Supplemental Figure S1, 7,999 women aged 20-45 were successfully enrolled between April 2020 and July 2020.After excluding 3 unsatisfied cytology results and 9 loss of follow-up, a total of 7,987 were included for evaluating the efficacy of CC screening.The demographics and clinical information of study population were listed in Supplemental Table S1.In brief, partial HPV genotyping revealed 761 (9.51 %) women with hr-HPV positive, including 184 (2.3 %) HPV16/18 and 627 (7.84 %) other 12 hr-HPV genotypes.Extended HPV genotyping revealed 1,114 (13.93 %) women with hr-HPV positive including 293 (3.66 %) HPV16/18, 893 (11.16 %) other 12 hr-HPV genotypes, and 937 (84.11 %) women with single hr-HPV infection.And 644 women were referred for colposcopy and 83/33 cases were diagnosed as CIN2+/CIN3+ at baseline.

Consistency of extended and partial HPV genotyping in hr-HPV detection and diagnostic accuracy
We compared the consistency of 14 hr-HPV genotypes results between extended and partial HPV genotyping.The overall agreement was 92.66 % for 14 hr-HPVs, 98.57 % for HPV16, 99.41 % for HPV18, and 93.85 % for other 12 hr-HPV genotypes (Supplemental Table S2).Interestingly, the agreement further increased with the progression of lesions with high kappa coefficients for HPV16, HPV18, and other 12 hr-HPV genotypes (Table 1).These results indicated extended and partial HPV genotyping methods had excellent consistency for 14 hr-HPV detection, and it was reasonable to use extended HPV genotyping to perform HPV test.In addition, both methods had similarly high sensitivity, high NPV, and good PPV for the detection of CIN2+/CIN3+ when using conventional HPV primary screening strategy (triaged by cytology).The specificity of both methods were also similar, although the differences were significant (Table 2).
The distribution of hr-HPV genotypes in women with CIN2+/CIN3+ was shown in Figure 1D-F and Supplemental Table S4.HPV ranking among CIN2+ and CIN3+ by Any. was further adopted for Hier.1 and Hier.2 respectively.HPV16, 52, 33, 58, and 18 were the top 5 HPV genotypes that both related to CIN2+ and CIN3+ among all three methods.The different top ranking HPV in women with HPV infection and CIN2+/ CIN3+ indicated that different HPV carried different cervical disease risks.

Comparison of conventional and risk-based primary screening strategies by extended HPV genotyping
Risk-based management strategy was further constructed, and women who met the following criteria should receive colposcopy: (1) HPV 16/18 positive; (2) Group An HPV positive with cytology ASC-US or worse; (3) Group B HPV positive with cytology LSIL or worse; (4) Group C HPV positive and cytology High-grade (Table 3).The risk-based management strategy demonstrated significantly higher specificity in detecting CIN2+/CIN3+ (Table 4) and CIN3+ (Table 2) compared with the conventional primary screening strategy based on Yaneng extended and Cobas partial HPV genotyping methods, respectively.And there were no significant differences in sensitivity, PPV, and NPV in predicting CIN2+/CIN3+ among these strategies.The number of colposcopies for detecting one cervical disease using risk-based strategy was significantly lower than that of conventional primary screening strategy in the detection of CIN2+/CIN3+ (CIN2+: 6.30 vs. 8.54, p=0.0203;CIN3+: 10.81 vs. 21.09,p<0.001) and Cobas-based primary screening strategy in the detection of CIN3+ (CIN 3+: 10.81 vs. 16.91,p=0.0009).These results indicated that risk-based CC screening strategy was more accurate and cost-effective in predicting cervical lesions and cancer.

Discussion
Current study demonstrated that extended HPV genotyping exhibited good agreement with the partial HPV genotyping for detecting 14 hr-HPV genotypes, and the agreement could be further improved with the progression of cervical lesions.Moreover, there was an excellent consistency between extended and partial HPV genotyping for diagnostic accuracy using conventional HPV primary screening strategy.A risk-based precision CC screening strategy was further constructed based on the immediate CIN2+/CIN3+ risks of specific HPV, which improved the efficacy of CC screening significantly.And the number of colposcopies performed to detect one CIN2+/CIN3+ lesion was also lower compared to the conventional primary screening strategy.
The overall prevalence of 14 hr-HPVs infection in present study was slightly lower than previously reported (ranging from 9.9 to 23.84 %) in China [16][17][18][19], which might attribute to the fact that women with HPV infection history were excluded at baseline.Due to the excellent consistency of extended and partial HPV genotyping for HPV detection and diagnostic accuracy using conventional HPV primary screening strategy, extended HPV genotyping was further used for assessing the distribution and the risk of carcinogenesis for specific HPV genotype.Yu et al. reported that the prevalence of hr-HPV with descending order was HPV 16, 52,  [20].Our data revealed generally similar results.Moreover, the prevalence of HPV31 and HPV33 was ranked tied for 10th in the population, but the cases of CIN2+/CIN3+ for HPV 31 and HPV 33 were ranked the 6th and the 3rd by Min.estimate.This discordance of the HPV ranking between the population and the patients with CIN2+/CIN3+ suggested that the infectivity and carcinogenicity was discordant for specific HPV genotype.
HPV16 and 18 were universally known as the most and second common types in cervical cancer.However, it was debated for non-16/18 hr-HPV that whether the genotypes should be further distinguished for primary CC screening [3,21,22].Although Monsonego et al. [22] supported that pooled non-16/18 hr-HPV could provide sufficient information for screening, their results showed HPV31, 33, 52 and 45 carried high CIN3+ risk (7.9, 5.4, 4.4 and 4.3 % successively) in aged ≥30.These results suggested that a detailed genotype analysis should be essential for clinical purpose, since 2019 ASCCP guideline had suggested 4 % risks for CIN3+ should be    [23].Our result also observed that HPV33, 16, 58, and 31 carried high risk of CIN2+/CIN3+ (>4 %).Surprisingly, we even found that HPV33 carried a higher CIN3+ risk than HPV16, which was consistent with Monsonego' report, they found that HPV33 carried the highest CIN3+ risk in women aged 25-29 [22].In addition, Schiffman et al. reported that HPV45 ranking 7th for the risk of CIN3+ and 9th for CIN2+ in women aged ≥30 with negative cytology and positive hr-HPVs after a 3-year follow-up [24].But in current study, HPV45 showed low oncogenic risk for both CIN2+ and CIN3+, which might attribute to the small sample size and lack of follow-up.So a larger population-based study focused on HPV33 and HPV45 should be conducted to investigate their pathogenicity in the future.
To construct the stratified management strategy based on CIN2+/CIN3+ risk, we further divided other 12 hr-HPV genotypes into high-risk, medium-risk, and low-risk groups based on their oncogenic risk.There were significant differences in other 12 hr-HPV risk classification between CIN2+ and CIN3+ groups, CIN2+ group had more high-and medium-risk HPV genotypes than CIN3+ group, which might be that some CIN2 lesions were not the true precursor stage of CC [25].While CIN3+, the immediate precursor lesion of CC, was a better predictor of CC compared with CIN2+ [26,27].Our results suggested that the risk-based HPV genotyping screening strategy could effectively improve the specificity in the detection of CIN2+ and/or CIN3+ compared to conventional HPV primary screening strategy with Yaneng extended or Cobas partial HPV genotyping and maintain high sensitivity, which would decrease unnecessary referrals for colposcopy in a screening setting for women aged 20-45 with hr-HPV infection.Previous studies have revealed that the number of colposcopies required to detect one case of CIN2+ by conventional primary screening strategy ranged from 6 to 8, and CIN3+ were 11-13 [28,29], and the number of current study was slightly higher than previous study for detecting one CIN3+.The possible reason was that the participants in current analysis were aged 20-45 and the younger women would be more likely to clear the HPV virus due to their immunocompetence.
Our study adopted the international recognized method of HPV detection combined with cervical cytology for CC screening, which could fully assess the cervical diseases.In addition, we focused to evaluate the women aged 20-45 who had no previous abnormal results for CC screening, which could reduce the impact of past history and provide more accurate guidance for clinical application.There were also several limitations.Firstly, present study was lack of followup data on the natural history of different HPV genotypes infection as a cross-sectional study.Secondly, the sample size of current study was not large enough to accurately assess the oncogenic risk of rare HPV genotype.

Conclusions
We showed that extended HPV genotyping had good agreement with a well-validated partial HPV genotyping kit for hr-HPV detection consistency and primary CC screening accuracy.In addition, our new risk stratification strategy based on extended HPV genotyping not only retains the sensitivity but also greatly improves the specificity of CC primary screening.Which could reduce the unnecessary colposcopies and guide the clinical stratification management for HPV positive population.Taken all together, current study provided a new triage strategy for CC screening, but a larger sample study is still needed for evaluating the diagnostic accuracy of our risk-based screening strategy.

Figure 1 :
Figure 1: The prevalence and the immediate CIN2+/CIN3+ risk of specific HPV genotypes in women aged 20-45.(A-C) The distribution of specific HPV genotype in 7,999 women estimated by Min.(A), by Any.(B), and by Hier.(C).(D-F) The prevalence of specific HPV genotype in CIN2+/CIN3+ estimated by Min.(D), by Any.(E), and by Hier.(F).(G-I) The immediate CIN2+/CIN3+ risks of specific HPV genotype estimated by Min.(G), by Any.(H), and by Hier.(I).

Table  :
The consistency of extended and partial HPV genotyping in the detection of HPV by the disease status.
a HPV: repeated calculation for women with HPV infection; b the results of HPV detection by extended and partial HPV genotyping were completely consistent; c other  hr-HPV types: repeated calculation for women with HPV or  infection.

Table  :
The efficacy of conventional HPV primary screening triaged by cytology for identifying CIN+ and CIN+.

Table  :
Risk-based management strategy according to different hr-HPV genotypes grouping., , and ; Group B: HPV, , and ; Group C: HPV, , , , and ; b CIN+: Group A: HPV; Group B: No; Group C: HPV, , , , , , , , , , and ; c we could not provide a definite follow-up interval in current management strategy because no follow-up data was available for cumulative risk at  or  years.