Urine Collection in Cervical Cancer Screening – Comparison of Two HPV DNA Assays


 Background: To reach non-participants, reluctant to undergo clinician-based cervical cancer screening and cervico-vaginal self-sampling, urine collection for high-risk human papillomavirus detection (hrHPV) may be valuable. Using two hrHPV DNA assays, we evaluated the concordance of hrHPV positivity in urine samples in comparison with cervico-vaginal self-samples and cervical cytology samples taken by the general practitioner (GP). We also studied women's accept of urine collection and preferences towards the different sampling procedures.Methods: One hundred fifty paired self-collected urine and cervico-vaginal samples and GP-collected cervical cytology samples were obtained from 30 to 59-year-old women diagnosed with ASC-US within the Danish cervical cancer screening program. After undergoing cervical cytology at the GP, the women collected first-void urine and cervico-vaginal samples at home and completed a questionnaire. Each sample was hrHPV DNA tested by the GENOMICA® CLART and COBAS® 4800 assays. Concordance in hrHPV detection between sample types was determined using Kappa (k) statistics. Sensitivity and specificity of hrHPV detection in urine was calculated using cervical sampling as reference.Results: With the COBAS assay, urine showed good concordance to the cervico-vaginal (k=0.66) self-samples and cervical samples (k =0.66) for hrHPV detection. The corresponding concordance was moderate (k=0.59 and k=0.47) using CLART. Compared to cervical sampling, urinary hrHPV detection had a sensitivity of 63.9% and a specificity of 96.5% using COBAS; compared with 51.6% and 92.4% for CLART. Invalid hrHPV test rates were 1.8% for COBAS and 26.9% for CLART. Urine collection was well-accepted and 42.3% of the women ranked it as the most preferred future screening procedure.Conclusions: Urine collection provides a well-accepted screening option. With COBAS, higher concordance between urine and cervico-vaginal self-sampling and cervical sampling for hrHPV detection was found compared to CLART. Urinary hrHPV detection is feasible, but its accuracy may need to be improved before urine collection at home can be offered to non-participants reluctant to both cervical sampling and cervico-vaginal self-sampling.


Background
The introduction of high-risk human papillomavirus (hrHPV) DNA testing in cervical cancer screening programs allows nonparticipating women to self-sample a cervico-vaginal sample in their own home using a device and return the sample to the laboratory for hrHPV testing by mail [1][2][3]. Home-based cervico-vaginal self-sampling is a well-accepted screening method, proven to increase screening participation signi cantly compared to mailed reminders to attend clinician-based screening [1,4]. Nevertheless, even with the cervico-vaginal self-sampling offer, some women are still not engaged in screening, possibly explained by the women's uncertainty about proper self-sampling including fear of injuries and discomfort touching oneself [5]. Urine samples may be considered as an alternative self-sampling screening option as it is cheap, non-invasive and straightforward to collect [6][7][8].
Several studies have assessed the hrHPV concordance and agreement between urine samples versus cervico-vaginal selfsamples and clinician-collected cervical samples using a great variety of hrHPV DNA assays [9][10][11][12][13][14][15]. However, little attention has been given to the direct comparison of different hrHPV DNA assays within the same study [12,15]; thus no strong conclusions about assay effects can be made. Moreover, the performance of hrHPV testing on urine samples has mainly been evaluated in studies, where women obtained the urine samples in the clinic and the samples were transported to the laboratory immediately for analyses [9][10][11][12][13]15]. Yet, if urine collection is to be implemented in an organized screening program, a home-based setting with mailing of the samples to the laboratory would be desirable. Urine collection at clinics has been reported to be highly acceptable [10,13] and preferable over cervico-vaginal self-sampling and clinician-based screening [16,17]; yet data regarding women's acceptability and preferences towards home-based urine collection are lacking.

Aims
We evaluated the concordance of hrHPV positivity in urine samples collected at home in comparison with cervico-vaginal self-samples collected at home and cervical cytology samples collected by a general practitioner (GP) using two hrHPV DNA assays (COBAS® 4800 and GENOMICA CLART® HPV4S). We also assessed the women's acceptance of urine collection and preferences towards the different sampling procedures.

Setting
In Denmark, cervical cancer screening is a nationwide program inviting women aged 23-64 years for liquid-based cervical cytology sampling at their GP (cervical sample) [18]. At present, women aged 23-59 years are screened with cytology, whereas women aged 60-64 years undergo hrHPV-based screening [18]. Women aged 30-59 years who are diagnosed with atypical squamous cells of undetermined signi cance (ASC-US) undergo routine re ex hrHPV triage testing, and ASC-US/hrHPV positive women are referred for colposcopy, whereas ASC-US/hrHPV negative women are referred back to the routine screening program [18]. This study was conducted in the Central Denmark Region (CDR), where all cervical cytologies are routinely handled and analyzed by the Department of Pathology, Randers Regional Hospital. In the CDR, the COBAS® 4800 (Roche Diagnostics, Switzerland) test is the routine test platform for hrHPV DNA analysis.

Study participants
Women eligible for this study were aged 30-59 years and diagnosed with ASC-US based on a cervical sample between June 2015 and March 2017. Exclusion criteria were pregnancy and having given birth within the last 3 months. The recruitment procedure has been described in detail in a previous publication [19]. In brief, eligible women were consecutively mailed a consent form and an information letter about the study explaining that they had to contact the investigator for oral information regarding the study and return a signed informed consent if they wanted to participate. Sample collection, processing and storage As per routine, the women had a cervical sample collected with a cervical cytobrush at their GP. After collecting cervical epithelial cells, the cervical brush was rinsed in 10 mL SurePath medium (BD Diagnostics, Burlington, NC) and mailed to the Department of Pathology, Randers Regional Hospital, for routine processing and testing as previously described [19]. The COBAS hrHPV testing was performed as per routine using the sample cell pellet from 1 ml SurePath medium. For this study, 100 µl of the residual puri ed DNA material was subsequently stored at -80 C prior to CLART testing.
After written informed consent, the women were mailed a self-sampling package to their home. The package included a dry brush device (Evalyn® Brush, Rovers Medical Devices, B.V, Oss, Netherlands) for cervico-vaginal self-sampling [20], a transportation tube with preservative media (Genelock, ASSAY ASSURE, Sierra Molecular, CA) for urine sampling, written and picture-based instructions showing the order of the self-sampling, pre-addressed return envelope, and a questionnaire. All women were instructed to collect the samples before undergoing colposcopy. The women were further asked to rstly collect the cervico-vaginal sample and secondly collect a rst-void urine sample (the rst of 10-12 ml of urine voided) in a plastic cup during their rst urination in the morning. The urine was transferred to the provided transportation tube by the participant. The women were urged to collect both samples and return the samples and the accompanying questionnaire by ordinary mail on the same day as the samples were taken and before an eventual colposcopy examination [19].
Upon arrival in the laboratory, the urine sample was stored overnight at 4˚C and then vortexed for 5 min. A 10-12 mL volume of urine was centrifuged at 3000 x RPM for 20 min at room temperature. After centrifugation, the cell pellet was resuspended in 1 mL 25% ethanol-buffered (TRIS) and stored at -80 C until further hrHPV testing. The dry brush head was transferred into 10 mL SurePath medium (BD Diagnostics, Burlington, NC) also stored overnight at 4˚C and then vortexed for 5 min. A 6.4 mL volume of the self-sample material was centrifuged at 3000 x RPM for 20 min at room temperature [19]. With the supernatant removed, the cell pellet was placed in 1 mL 25% ethanol-buffered (TRIS) and stored at -80 C until further hrHPV testing [19]. A volume of 6.4 mL was chosen to adjust for material volume used for cytology examination performed on the cervical sample [19].
HrHPV DNA testing Before the day of COBAS hrHPV DNA testing, the cell pellet material from the urine and cervico-vaginal self-samples were thawed overnight at 4˚C. Subsequently, the self-samples (1 mL volume) were vortexed for 15 s before being transferred to empty test tubes for DNA puri cation and hrHPV testing [19]. DNA was puri ed using the COBAS x480, and ampli cation and detection of hrHPV DNA were undertaken using the COBAS z480 analyzer [21]. From each self-sample, 100 µl of the residual puri ed DNA material were stored at -80 C, until further CLART hrHPV testing. The COBAS® assay is a fully automated realtime PCR-based method, separately detecting HPV16, HPV18, and 12 other hrHPV types (HPV31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66 and 68) including the beta-globin gene as an extraction and ampli cation control [21].
Before the day of the CLART hrHPV DNA testing (HPV4S, GENOMICA, Madrid, Spain), the residual puri ed DNA material from the self-samples and the cervical samples gained using COBAS x480, were thawed overnight at 4˚C and ve µl of the puri ed DNA material from the samples were used for the PCR ampli cation. The PCR ampli cation was performed using the CLART HPV4S ampli cation kit (GENOMICA) [22]. Detection was performed on the CLART microarray. The genotyping results were analyzed and automatically performed on the Clinical Array Reader (GENOMICA) [22]. CLART HPV4S detects 14 hrHPV genotypes individually (HPV16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68) and two low-risk HPV genotypes (6 and 11). Ampli cation of a spiked CFTR plasmid served as an internal control of the PCR process, while the internal control for human CFTR gene validates material su ciency in the sample [22].
For both assays, samples with an invalid test result (i.e. COBAS: no betaglobin gene detected, CLART: no human CFTR ampli cation detected, or no spiked CFTR plasmid ampli cation detected) were retested once on diluted samples, and the second result was considered de nitive. Every run included four water samples to measure contamination [19]. The laboratory personnel performed the hrHPV testing without knowledge of the hrHPV status of the cervical samples [19].

Acceptability and preferences
Together with the self-sampling package, the women received a questionnaire (see supplementary material) addressing among others the acceptance of urine collection, the clarity of the user instructions, and the con dence in correct execution of the urine collection. For analysis, the ve response categories were grouped into three: "Agree" (totally agree and agree), "Disagree" (disagree or totally disagree), and "Do not know". Women were also requested to rank if they preferred urine collection, cervico-vaginal self-sampling, or cervical sampling at GP as their future screening examination. For analysis, women who refrained to rank their preferences or ranked two or more sampling methods as their preferred method were coded as missing. Additionally, the women were asked to report the date of collecting their self-sample and whether they had engaged in sexual intercourse between the cervical sampling at the GP and self-sampling [19].

Statistical analyses
For analysis, only women with valid hrHPV results from paired urine, cervico-vaginal, and cervical samples for both hrHPV assays were included. Cohen's kappa (κ) was used to measure concordance in hrHPV positivity (any hrHPV and speci c genotypes) for urine vs. cervico-vaginal self-samples, urine vs. cervical samples by the COBAS and CLART assays. Concordance was de ned as "poor" (κ ≤ 0.20), "fair" (0.21 ≤ κ ≤ 0.40), "moderate" (0.41 ≤ κ ≤ 0.60), "good" (0.61 ≤ κ ≤ 0.80), or "very good" (κ ≥ 0.81) [24]. Comparing the presence of HPV16/18 and hrHPV other types between the samples; concordance was determined by the presence of at least one identical genotype in both samples; discordance was determined as no genotype similarities [19]. McNemars test was performed to compare proportions of hrHPV positive results between the paired sample types. We also calculated the overall percentage of agreement between the paired samples [19]. Strati ed by hrHPV assay, the sensitivity and speci city of hrHPV positivity (any hrHPV and speci c genotypes) in urine samples was calculated with 95% con dence intervals (CIs) using the cervico-vaginal self-sample or the cervical sample as technical gold references.
The acceptability of urine collection and preferences regarding the different sampling procedures were evaluated by descriptive statistics (proportions and 95% CIs). We tested whether women's preferences differed between age groups (30-39, 40-49 and 50-59 years). The χ2-test was used to test for differences in categorical data. For continuous data, medians and interquartile ranges (IQR) were calculated; the Mann Whitney rank sum test was used to test for differences. P-values < 0.05 were considered statistically signi cant. The statistical analyses were performed using STATA, version 16 (STATA College).
A sample size calculation has been reported elsewhere [19].     Of the 1,110 eligible women, 216 participated in the study by returning urine and cervico-vaginal self-samples. However, a total of 66 women (30.6%) were excluded from the analyses: Three women (1.4%) had colposcopy performed prior to selfsampling; four women (1.8%) had invalid hrHPV urine results on both assays; 58 women (26.9%) had invalid hrHPV urine results using CLART only; and one woman (0.5%) had invalid cervical hrHPV result using CLART only (Fig. 1) HrHPV positivity in urine, cervico-vaginal and cervical samples For both assays, the hrHPV positivity was statistically signi cant lower in urine samples compared to cervico-vaginal selfsamples except for HPV16/18 positivity using CLART ( Table 1). The hrHPV positivity in urine samples was also lower compared to cervical samples for both assays, but in this comparison only HPV16/18 positivity using COBAS was statistically signi cant (4.0% vs. 10.0%, p = 0.02) ( Table 1). No statistically signi cant differences were observed in hrHPV positivity (any and speci c genotypes) between the COBAS and CLART assay for the urine samples (p > 0.05 for all comparisons, data not tabulated).

Main ndings
With the COBAS assay, urine samples showed good concordance in hrHPV detection compared with cervico-vaginal and cervical samples, while moderate hrHPV concordance was found between samples using the CLART assay. Compared to cervical sampling, urinary hrHPV detection had a sensitivity of 63.9% and a speci city of 96.5% using COBAS; compared with 51.6% and 92.4% for CLART. The invalid hrHPV test rates of urine were 1.8% and 26.9% for the COBAS and CLART assay, respectively. Home-based urine collection was well-accepted and women ranked it as the most preferred future screening procedure.

Strengths and limitations
Main strengths of our study were the use of two clinically validated hrHPV DNA assays, approved for primary hrHPV screening [22,25,26] and the fact that the women served as their own controls, limiting potential biases. Additionally, the women collected the urine samples at home without supervision from healthcare professionals, which from an implementation point-of-view is the most appropriate setting to evaluate urine collection before becoming routine.
A limitation of the study is the time interval between samples. Part of the discrepancy in hrHPV concordance between urine and cervical samples may be explained by a hrHPV infection acquired or cleared between sampling [19]. Yet, the questionnaire data enabled us to interpret discordant results, and the time interval between collecting the samples was not an issue when comparing urine samples with cervico-vaginal self-samples.
Since our study was conducted in a referral population diagnosed with ASC-US, the concordance results cannot be generalized directly to a screening population. Furthermore, as we enrolled women who had already attended the screening program, the acceptability of urine collection may differ from women reluctant to participate in routine screening. A generalization of our results to routine screening programs should therefore be done with caution.

Interpretation and comparison with previous studies
Even though rst-void urine was used, which has been proven to contain signi cantly more HPV DNA than the subsequent part [8,27], we found lower hrHPV positivity in urine samples as compared with corresponding cervical and cervico-vaginal samples for both assays. Thus, the lower hrHPV positivity in urine may be due to specimen type differences rather than the assay itself [15]. Our result corresponds to most comparative studies [9-12, 15, 28, 29], but not all [13,14,27].
COBAS showed higher concordance for hrHPV detection (any type) between rst-void urine and cervical samples than CLART  [10]; all using a combination of rst-void urine samples and COBAS in referral populations. As our study was the only one with an interval between the urine and cervical sample, some of the differences in concordance may be attributed to this factor. Other reasons for the differences are likely explained by variations in sampling procedure (at home vs. clinic), type of preservative media [27], (pre-) analytical processing protocols [13], storage conditions [27], the volume of urine collected [27], and study populations (differences in abnormal cytology prevalence). The concordance between urine and cervical samples for hrHPV detection using COBAS remained robust (κ = 0.61), when including the 58 women who had an invalid test result by CLART only in the analysis (data not shown).
A meta-analysis found a pooled sensitivity of 77% and speci city of 88% for hrHPV detection in urine samples compared with cervical samples [8]. In our study, lower sensitivity (63.9% and 51.6% for COBAS and CLART, respectively) but higher speci city (96.5% and 92.4% for COBAS and CLART, respectively) was found between the urine and cervical samples for both assays. Thus, more than 50% of the cervical hrHPV infections were not detected in the urine samples. However, it is possible that adjustment of the COBAS assay cut-off value for positive result could provide performance-related bene ts for urine collection, as reported elsewhere [28]. Despite small numbers, CLART showed higher concordance between urine and cervical samples for the detection of HPV16/18 than COBAS (κ = 0.65 vs. 0.34), whereas COBAS performed better than CLART for the detection of hrHPV other types (κ = 0.70 vs. 0.38). In comparison with our previous study which was conducted within the same study population (n = 213), good concordance (κ = 0.70) was found between cervico-vaginal self-sampling and cervical sampling for hrHPV detection using COBAS including acceptable sensitivity (80.9%) and speci city (91.6%) [19]. Additionally, cervico-vaginal self-sampling was well-accepted, but almost 10% of the women expressed concerns about proper sampling.
In the present study, overall agreement between urine and cervico-vaginal self-samples for detection of hrHPV (any type) using COBAS (87.3%) was in line with previous data [11].
Neither the COBAS nor the CLART assays are currently CE-marked for urinary testing; yet, urine samples analyzed on CLART resulted far more often in invalid hrHPV test results than on COBAS (26.9% vs. 1.8%). This indicates that the CLART assay may be more severely affected by the presence of PCR inhibitors in urine and the lower amount of cells in urine, both known to reduce assay sensitivity [30,31].
In comparison, invalid test rates of urine samples have in other studies been reported to range between 0 and 4% using PCRbased HPV DNA assays [9,10,[12][13][14][15]29]. Whether optimizing the (pre)-analytical processing protocols could lead to better results for the CLART assay warrants further exploration. Indeed, our results support that future research should focus on optimizing the urinary (pre)-analytical procedures to improve accuracy, but also compare accuracy of hrHPV testing in paired urine and cervical samples using different combinations of urine collection methods and hrHPV assays. The ongoing VALUDES study seeks to address this current lack in evidence [32].
For screening purposes, detection of hrHPV in urine would be useful only if it can identify women who have underlying CIN2+, which is a treatable screening endpoint [33]. Accuracy of hrHPV testing with COBAS for CIN2 + detection using urine samples has proven to be signi cantly lower than compared with cervical and cervico-vaginal sampling [28]. Although our study was not designed to evaluate the clinical accuracy of urine, we did nd that up to half of the CIN2 + cases were missed by urine collection.
For use in screening, a high acceptability of the method is of great importance if wanting to reach non-participants through urine collection. Our results were consistent with the literature, showing urine collection to be highly acceptable [10,13].
Similar to other studies, urine was the most preferred screening method [13,16,17]. We did not observe any differences in preferences between age groups. Con dence in correct execution of the urine collection procedure is essential, as insecurity could lead to mistrust of the test results and cause the woman to worry. Despite women in our study performed urine collection at home with no specially designed urine collection device [34], only 3.4% of the women expressed concerns about collecting the urine sample correctly. This is lower than the 20% reported by women performing clinic-based urine collection [10] and even lower than the 10% of women expressing concerns about performing home-based cervico-vaginal selfsampling as reported previously [19].

Conclusions
This study showed that home-based urine collection was well-accepted and ranked as the most preferred future screening procedure. The COBAS assay performed better than CLART with respect to higher hrHPV concordance between urine and both cervico-vaginal self-sampling and cervical sampling as well as fewer invalid hrHPV test results. This study thus con rms the utility of urinary hrHPV detection although its accuracy may need to be improved before urine can serve as an alternative screening option to reach non-participants reluctant to undergo GP-based cervical sampling and cervico-vaginal self-sampling.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.