A prospective pilot evaluation of vaginal and urine self-sampling for the Roche cobas 4800 HPV test for cervical cancer screening

This pilot study sought to evaluate the feasibility of utilizing vaginal self-swabs and urine samples for HPV-based cervical cancer screening in 700 women who had undergone conventional Pap smear screening via the national cervical cancer program in Korea. The cobas 4800 HPV test was utilized to detect HPV in the self-samples. Pap smear results revealed three cases of atypical squamous cells of undetermined significance, 649 cases of negative for an intraepithelial lesion or malignancy, and 48 non-specific inflammatory findings. High-risk HPV was detected in 6.7% of urine samples and 9.6% of vaginal self-swab samples. The overall agreement for HPV 16/18 between urine and vaginal self-swab samples was 99.1% (95%CI 98.1% to 99.6%). Colposcopic biopsy revealed one cervical intraepithelial neoplasia (CIN) 3 lesion, 12 CIN1 lesions, and 23 normal or chronic cervicitis lesions. In conclusion, urine and vaginal self-swab sampling was feasible and deemed a potential alternative for HPV detection in women who hesitate to participate in cervical cancer screening programs. Meanwhile, due to overall lower rates of abnormal cytology and sexual risk behaviors in Korea, a larger sample size than expected is needed to assess the sensitivity of CIN2+ detection via self-samples.


Results
At enrollment, 12 women had missing or invalid test results, leaving 700 women available for analysis. The median age of the participants was 52 years (range 21-71 years). Among the study participants, cytology revealed three cases of ASC-US, 649 cases of negative for an intraepithelial lesion or malignancy (NILM), and 48 non-specific inflammatory findings. The self-sample results according to cytology are listed in Table 1.
High-risk HPV (hrHPV) was detected in 6.7% of the urine samples and in 9.6% of the vaginal self-swab samples (Table 1) The overall agreement between urine and vaginal self-swab results for any hrHPV was 94.9% (95% CI 93.0 to 96.4%), with a kappa of 0.66, and the overall agreement for HPV 16/18 was 99.1% (95% CI 98.1% to 99.6%), with a kappa of 0.75. A statistically significant difference in detection of hrHPV was observed between urine and vaginal self-swab samples (p < 0.001, Table 2). Of 26 urine_negative/vaginal_positive samples, most were of other hrHPV type (5 cases were HPV16/18), and were all classified as NILM based on cytology.
Among women with NILM or non-specific inflammation cytology, 33 underwent colposcopy (Table 3). Colposcopic findings revealed CIN1 in 11, chronic cervicitis in 16, and normal findings in six women. Among the women with chronic cervicitis (n = 16), 12 were positive for other hrHPV upon testing of their urine and vaginal self-swab samples; the four remaining women all had HPV-negative urine samples, with two being negative and the other two having HPV 16/18 positive vaginal self-swab samples. Cases with CIN1 (n = 11) and normal colposcopic results (N = 6) showed the same results between urine and vaginal self-swab samples with regards to HPV DNA positivity.
In this study, only three cases of ASC-US were included. In one woman with both negative urine and vaginal self-swab samples, colposcopic biopsy revealed chronic cervicitis. Meanwhile, in the other two women positive for hr-HPV, colposcopic biopsy revealed CIN3 and CIN1, respectively (Table 4).

Discussion
This pilot study is the first prospective study to use non-invasive, urine and vaginal self-swab sampling for cervical cancer screening at a tertiary hospital in Korea. However, our results could not confirm whether HPV screening based on urine or vaginal swab sampling shows increased sensitivity compared to cytology for detection of  CIN2+ or CIN3+ cancer due to the limited sample size: only one participant (1/700) was found to have CIN2+ or CIN3+ disease in this pilot study. HPV detection rates for the general population of Korea have been found to vary according to age groups, region, socioeconomic status, and methods for detecting HPV DNA, ranging from 10% to 15% 19 . In this study, hrHPV was detected in 6.7% of urine samples and in 9.6% of vaginal self-swab samples. HPV16/18 was detected in 1.5% of urine samples and in 2.0% of vaginal self-swab samples. Other hrHPV was detected in 5.2, and 7.6% of the urine and vaginal self-swab samples, respectively. Although a statistically significant difference in the detection rate of hrHPV was noted between urine and vaginal self-swab samples (p < 0.001), the agreement for HPV 16/18 was relatively high (99.1%, 95%CI 98.1~99.6%), with a kappa of 0.75. Discrepancies arose mainly in the positive samples for hrHPVs other than HPV 16 and 18.
A previous study indicated that the sensitivities for detecting HPV were 96.4% in vaginal self-samples and 83.9% in urine samples relative to cervical samples acquired by a clinician 20 . A recent meta-analysis showed that HPV 16/18 detection in urine has a pooled sensitivity of 73% (56% to 86%) and a specificity of 98% (91% to 100%) 21 . Similar to these studies, for hrHPV detection using Roche_HPV in urine samples we found relative sensitivities of 70.4~79.2% and relative specificities of 93.5~100.0% 18 . Increasing the sensitivity of hrHPV detection in urine samples is needed prior to the use thereof in cervical cancer screening programs 18,22 . One of the reasons for the relatively inferior sensitivity of urine samples in the detection of hrHPV is due to the use of random-void urine samples with no preservatives, instead of an initial stream of urine. Stanczuk et al. suggested that the sensitivity of the cobas 4800 HPV test for the detection of CIN2+ could be increased if the cut-off value was increased to 45 PCR cycles 17 . We also showed that RQ-PCR with an adjusted cut-off C T value increased the relative sensitivity of HPV detection in urine samples in our previous study 18 . Thus, if the adjusted cut-off values of RQ-PCR-based tests would be applied for urine samples, detection of HPV DNA using urine samples could be promising for full prospective study for cervical cancer screening.
The reported prevalences of HPV infection in several studies have been found to vary. In several studies, subjects comprised highly-selected, high-risk individuals referred for further evaluation with abnormal cytology in primary cervical screening 23 . Compared to the general population, the prevalence of abnormal cytology (12-83%) or HPV infection (4-91%) in referred populations has been shown to be much higher 24 . Due to overall lower rates of abnormal cytology and lower sexual risk behavior in Korea, a larger sample size than expected may   Table 3. HrHPV results and cervical biopsies in women with NILM cytology (n = 33). *Urine not obtained (n = 1).  be needed to comprehensively evaluate the effectiveness of self-sampling in detecting CIN2 disease or greater. For a sufficient number of women with CIN2 disease or greater in the ATHENA trial, an estimated 45,000 women were required for analysis: calculations were based on reported rates of ASC-US cytology 25 . The prevalence of ASCUS in the ATHENA study was 4.1% 26 . In this pilot study, however, an unexpectedly low ASCUS rate was discovered, although this finding is similar to rates of abnormal cytological results (0.52%) reported for the national cervical cancer screening program in Korea in 2014 3 . In addition, the number of CIN2 or CIN3 cases included in this pilot study was also unexpectedly low, compared to that in previous studies 27,28 . Thus, a much larger sample size than that in the ATHENA study is needed for definitive study. In Korea, women aged 30-39 years have had the second-lowest (27.7-44.9%) participation rate in cervical cancer screening, but the highest rate of abnormal results, compared to other age groups, since 2012 4 . This participation rate is much lower than that in Europe 29 and the US 30 . In addition, cervical cancer incidence rates have been increasing in younger, sexually active groups 4 . For early detection in screening programs, the use of non-invasive self-sampling HPV tests may help with promoting participation among younger women and in preventing cervical cancer in Korea. Indeed. among the participants in the present study, overall satisfaction was 91.4% with the urine sampling and 92.7% with the vaginal self-swab sampling, which was higher than that with clinician-guided cervical swabs (88.1%) (data not shown). Additionally, embarrassment with vaginal self-swabs and with urine sampling was significantly lower than that with clinician-guided Pap smear.
This study had several limitations, including the relatively small sample size. The number of women with ASCUS was very small (n = 3), and only one of the women who underwent colposcopy had a CIN2+ disease. In addition, not all women with NILM underwent colposcopy. Thus, we could not assess the sensitivity of CIN2+ detection by hrHPV DNA testing, although population-based cervical cancer screening samples were collected. In addition, the use of random-void urine samples with no preservatives could compromise the performance of RQ-PCR assays using urine samples.
In summary, Roche_HPV showed relatively good concordance between urine and vaginal self-samples. However, hrHPV DNA detection using urine samples seemed to be inferior to HPV DNA detection using vaginal self-swab samples. A large prospective study should be performed to determine if urine and vaginal self-swab samples may be an effective alternative for HPV detection in women who hesitate to participate in cervical cancer screening programs that use conventional methods.

Participants.
To evaluate the performance of Roche_HPV for HPV DNA detection, paired urine and vaginal self-swab samples were collected prospectively from 712 women who visited the National Cancer Center in Korea between July 2016 and November 2016. This study was approved by the Institutional Review Board of the National Cancer Center, Seoul (IRB No. NCC2014-0183 and NCC2015-0066). All women provided informed consent to participate in the study. We confirmed that all experiments were performed in accordance with relevant guidelines and regulations.
Sample collection, processing, and testing. The research ethics committee of the Institutional Review Board of the National Cancer Center approved the guidelines and regulations for all study methods. The collecting and processing procedure was the same as that described in a previous study 12 . Vaginal swabs were collected via a dry cone-shaped flocked swab (52980 C, FLOQSwabs, Copan Italia, Brescia, Italy) stored in a dry tube.
Study participants also provided urine samples of approximately 30 mL. Before high-risk hrHPV DNA testing using Roche_HPV, each urine sample was washed with Roche Transport Medium. Briefly, 5 mL of each urine sample was mixed with 5 mL of Roche medium and then centrifuged at 3000 rpm for 15 min. After removing the supernatant, the pellet was resuspended in 2 mL of Roche medium and used for HPV testing.
The self-samples were compared to cervical cytology samples, which were collected by a clinician using ThinPrep solution (PreservCyt Solution, Hologic, UK) as per routine protocol in the cancer screening program. Colposcopy was performed and directed biopsies were obtained if clinically indicated. In addition, colposcopy with biopsy was performed for women with hrHPV DNA positive results and, if desired, for those with normal cytological results.
All samples were tested with Roche_HPV based on RQ-PCR as described in a previous study 18,31 and in accordance with the manufacturer's instructions. Roche_HPV is a fully automated system for DNA extraction and detection, and provides specific genotype information for HPV types 16 and 18, while concurrently detecting the other 12 hrHPV types in a pooled analysis 12 .
To investigate the participants' satisfaction with all three sampling methods, we administered a survey questionnaire. Satisfaction and embarrassment were assessed via two respective questions for each sampling method: for example, "Overall, how satisfied were you with the urine sampling?" and "How embarrassed were you with urine sampling?" The items were answered on a 4-point Likert scale, ranging from "very satisfied" to "very unsatisfied. " Overall satisfaction was defined as "Yes" with a Likert score of 1-2 and "No" with a Likert score of 3-4, while embarrassment was defined as "Yes" with a Likert score of 3-4 and "No" with a Likert score of 1-2. In total, 551 participants completed to this questionnaire, and the overall response rate was 78.7%.

Statistical analyses.
Kappa coefficients with 95% confidence intervals (CIs) were calculated to estimate agreement between the results of Roche_HPV tests. McNemar's test was applied to assess the significance of differences between two correlated proportions. Statistical significance was defined as p < 0.05. All statistical analyses were performed using NCSS software, version 10 (NCSS Statistical Software, Kaysville, UT, USA).