Aims To assess the performance of a clinically validated human papillomavirus (HPV) test (the Cobas 4800 HPV test) in urine and self-taken vaginal specimens within a colposcopy population and to assess HPV prevalence before and after treatment across the different biospecimens.
Methods A total of 100 women attending a colposcopy clinic provided three biospecimens (a clinician-taken liquid-based cytology sample (LBC), a self-taken vaginal sample and a urine sample) for HPV testing. HPV prevalence and concordance was compared across the biospecimens and clinical performance relative to the detection of cervical intraepithelial neoplasia (CIN)2+ and CIN3+ was assessed. A total of 39 women retuned at 6 months for a post-treatment follow-up appointment, and HPV concordance in all biospecimens was measured relative to their original HPV status.
Results 65 cases of CIN2+ were detected in the baseline population; sensitivity for CIN2+ was 92% (82 to 97) for the vaginal and the LBC sample and 80.0 (68% to 88%) for the urine sample. In the follow-up (post treatment) population, women were twice as likely to be HPV positive in their urine or vaginal sample compared with the equivalent LBC sample.
Conclusions Vaginal and LBC samples showed very similar performance for the detection of CIN2+ in this population using the Cobas HPV test; further validation of these findings in screening contexts will be of value. Self-taken samples may have less utility in a ‘test of cure’ setting—given the higher prevalence of HPV relative to LBC.
Statistics from Altmetric.com
The direction of travel for cervical screening is for a human papillomavirus (HPV) test to be performed ‘first’ with reflex of positives to a more specific triage.1 The Netherlands is due to commence HPV primary screening in 2016,2 the American Society for Clinical Pathology recommend this approach for women >303 in the USA and several European countries including the UK have implementation projects underway.
Given that HPV tests rely on molecular amplification and do not require intact cells, there is scope for using biospecimens that are less or non-invasive including self-taken vaginal and urine samples.4 ,5 Such approaches have been considered as particularly relevant for those women who are often referred to as the 'hard to reach' population. However, as HPV testing becomes used increasingly for screening and disease management, it is pertinent to consider these options for those beyond the hard to reach group.
With respect to self-sampling, there is more evidence on the acceptability and performance of self-taken vaginal samples in the literature.6 However, urine constitutes a cheap non-invasive biospecimen on which HPV testing can be performed and reviews on performance are available.7 Several studies have focused on analytical rather than clinical performance often using in-house assays and/or have focused on the assessment of urine as a biospecimen for immunisation surveillance.8 ,9 While these studies are valuable, it is important that clinical performance of self-sampling (including urine testing) relative to disease is assessed using established assay platforms common to the routine service of laboratory medicine. However, there is little information on the performance of clinically validated HPV tests (approved for use in cervical disease management) in urine and in self-collected vaginal samples.
In the present evaluation, we assessed the performance of a high-throughput clinically validated HPV test—the Cobas 4800 HPV Test (Roche Molecular Systems, California, USA)—in liquid-based cytology (LBC) and self-collected urine and vaginal samples. In addition, as women were returning for follow-up post treatment—HPV status in the different biospecimens was assessed both before and after treatment.
Materials and methods
Women with abnormal cytology referred to a National Health Service (NHS) Colposcopy Clinic (Dumfries and Galloway Royal Infirmary), provided a sample of urine and a self-collected vaginal sample in the clinic. Prior to undertaking colposcopy, the clinician collected an LBC sample. Cytology classification was according to the British Association for Cervical Pathology. Any biopsy and/or treatment was as routinely indicated.
Out of the initial group, a total of 39 women received follow-up at 6 months from initial visit and provided an additional urine and vaginal sample prior to a further LBC and colposcopy.
Sample capture and processing
Urine was collected into a universal container before collection of the vaginal sample. The vaginal samples were self-collected using Rovers Cervex-Brush (Oss, The Netherlands). Women were advised to insert the brush into the vagina and slowly rotate it a few times. The brush was subsequently suspended in 5 mL of ThinPrep, PreservCyt Solution (Hologic, UK). A 20 mL volume of urine was centrifuged within 24 h for 10 min at 3000 rpm and the cell pellet suspended in 5 mL of ThinPrep. All samples were processed by Cobas 4800 HPV test according to manufacturer's instructions; the test is fully automated and detects 14 high-risk HPV types reporting HPV 16, 18 and ‘others’: 31, 33, 35, 39, 45, 51, 51, 56, 58, 59, 66 and 68.
A total of 109 women were recruited—however, analysis was confined to 100 women from whom three samples (urine, vaginal sample, LBC) were captured. Sensitivity, specificity, positive predictive value and negative predictive value for the detection of cervical intraepithelial neoplasia (CIN)2+ and CIN3+ were performed for each specimen type (at baseline). A Fisher's exact test was performed to assess differences in sensitivity and specificity across the specimen types. Cross-tabulations were performed to assess the difference in the distribution of discordant results between biospecimen types. Agreement for ‘any’ HPV positivity and also positivity for HPV 16 and/or 18 was performed across the biospecimen types.
In the group of women who returned for follow-up post treatment, evidence of HR-HPV persistence in the separate biospecimens was assessed descriptively. According to the read-out channels of the Cobas test (HPV 16, HPV 18 or HPV other), women were described as having a fully concordant infection (exact match between baseline and follow-up), a partially concordant infection (at least one channel in common between baseline and follow-up) or having no concordance (no channel in common between their baseline and follow-up).
Characteristics of referral population
The median age of the population (n=100) was 27.5 years of age (range 21–60). Women from whom a histology result was not available and/or did not provide three evaluable biospecimens were excluded from performance analysis. In total, 51 and 49 women had low-grade and high-grade referral cytology, respectively.
A total of 29 women had biopsies and 71 had large loop excision of the transformation zone. Associated histology revealed that 65 women had CIN2+ and 32 had CIN3+; 9=no CIN, 26=CIN1, 33=CIN2, 28=CIN3, 3=high-grade glandular lesions and 1=cancer. All biopsies were routinely indicated (as part of standard of care) and no study-driven biopsies were taken. Histology was performed according to routine protocols in line with the requirements of the NHS Scottish Cervical Cancer Screening Programme. Review histology is not performed as part of routine practice; however, for study purposes, all biopsies graded CIN2+ associated with HPV-negative (liquid) samples were reviewed post hoc, blinded to the original result. No discordance between first and second CIN diagnosis was observed.
Clinical performance of the Cobas in varied biospecimens relative to histologically confirmed high-grade disease
Overall, 92, 91 and 79 HPV infections were detected in the LBC, vaginal and urine sample of the recruits, respectively. Of the 65 CIN2+ cases, 5 were HPV negative using the LBC biospecimen compared with 5 and 13 cases in the vaginal and urine sample(s), respectively. Cross-tabulations for the 65 cases associated with CIN2+ can be seen in online supplementary data 1, tables A–C. Of the five cases of CIN2+ not detected via the LBC sample, none were detected via the equivalent urine sample with only 1/5 being detected in the equivalent vaginal sample.
Sensitivity for CIN2+ was 92% (82 to 97) for both the vaginal sample and the LBC sample and 80.0 (68% to 88%) for urine (table 1). Differences in sensitivity of the vaginal and urine sample (vs LBC) did not reach statistical significance (p=1.000 and 0.073, respectively). Specificity for all biospecimen types was low and ranged from 8.5% (LBC sample) to 22.8% (urine sample), and the differences between LBC versus vaginal sample and LBC versus urine were not statistically significant (p=1.000 and 0.188, respectively). Of the 32 CIN3+ cases (table 2), all were detected by the LBC and vaginal sample (100% (86.6 to 100)), whereas the urine sample ‘missed’ three cases with a sensitivity of 90.6 (73.8 to 97.5), p=0.238.
HPV-type distribution and agreement
Of the 92 HPV-positive LBC samples, 40 harboured HPV 16 infection (43.5%) and 11 harboured HPV 18 infection (11.9%). Comparatively, of the 91 positive vaginal samples, 40 (43.9%) and 10 harboured (11.0%) HPV 16 and 18, respectively. Of the 79 positive urine samples, 35 (43.7%) and 6 (7.5%) were positive for HPV 16 and 18, respectively. Consequently, the proportions of HPV 16 and 18 are similar in each biospecimen type. Cross-tabulations of HPV 16 and 18 are presented in online supplementary data 2 (tables A–F); agreement for HPV 16 was 91% (84 to 96) in LBC versus urine, 96% (90 to 99) for LBC versus vaginal sample and 91% (84 to 96) for vaginal sample versus urine sample. For HPV 18, agreement for the aforesaid comparisons was 93% (86 to 97), 99% (95 to 100) and 94% (87 to 98).
Overall agreement for any Cobas positivity between LBC versus urine, LBC versus vaginal sample and vaginal sample versus swab was 84% (75 to 91), 94% (87 to 98) and 84%, respectively, as indicated in tables 3⇓–5.
Using the LBC sample as a technical gold standard for HPV positivity, the sensitivity (any HPV positivity) of the assay was significantly lower in the urine sample (84.8% vs 96.7%, p=0.009), whereas there was no significant difference in LBC versus vaginal sample. When stratifying sensitivity according to HPV type, sensitivities for HPV 16 and HPV 18 in vaginal samples were 95.0% and 90.1%, respectively (relative to LBC). For the urine samples, sensitivities for HPV 16 and 18 relative to the LBC sample were 82.5% and 45.4%, respectively—although lower, this did not quite reach significance (p=0.154 and 0.063 for 16 and 18, respectively).
Assessment of HPV positivity and concordance in the follow-up population
A total of 39 women attended for follow-up with 7, 15 and 17 having received a baseline diagnosis of CIN1, CIN2+ and CIN3+, respectively. A total of 35/39 women were HPV positive in at least one biospecimen and 18 women remained HPV positive (in at least one biospecimen). Women who remained positive were more likely to be positive on their vaginal sample as a total of 8, 16, and 13 LBC, vaginal and urine samples were positive, respectively (table 6), although two LBC follow-up samples were invalid for HPV detection. HPV-positive agreement between the biospecimen types at the follow-up appointment is presented in supplementary data 3 (tables A–C). To summarise, agreement between LBC and urine, LBC and vaginal sample and vaginal sample and urine was 84% (68 to 94), 76% (59 to 88) and 94% (73 to 96), respectively.
We have shown that in a colposcopy setting self-taken vaginal samples are equivalent to LBC samples for the detection of CIN2+. The sensitivity of the urine assay for detecting CIN2+ and CIN3+ was 80% and 90.6%, respectively. While this was lower than the more cellular samples, these results were generated through application of an assay optimised to detect CIN2+ in cervical samples. It is feasible that adjustment of the analytical sensitivity of this (and indeed other molecular assays), specifically for less cellular samples, may provide performance-related benefits for urine testing.
Although HPV testing of urine is topical and was the focus of a high-profile recent review,7 this is one of the few studies that compares the head-to-head performance of urine versus LBC (and vaginal sample) within a setting where disease outcomes are known. Frequently the sensitivity/specificity of urine testing is described relative to a technical (rather than a disease) gold standard. We do acknowledge that the sensitivity of the urine and vaginal samples may be inflated in this referral population (given the disease-enriched nature) and is not directly transferable to performance in a screening population. However, we are currently assessing the performance of urine testing (using the Cobas) within such a screening context in the same Scottish region as this study,10 and it will be of interest to compare the two data sets.
The number of women from who we received follow-up samples (n=39) was small, so extrapolation is limited and performed with caution. Notwithstanding this caveat, women were more likely to remain HPV positive in their vaginal or urine sample post-treatment compared with their LBC sample. The 23% figure of post-treatment positivity in the LBC sample is also consistent with national Scottish figures on test of cure where positivity is 22% at 6 months.11 The positivity in the vaginal and urine samples in the present analysis is effectively double that observed for cervical positivity, bringing into question the use of self-sampling in a test of cure context—further, larger studies to assess this would be welcome.
To summarise, this study confirms the utility of self-taken vaginal swabs for the detection of CIN2+ using a well-established commercial HPV assay. While sensitivity of the urine assay for CIN3+ was >90%, further development and optimisation of HPV assays for urine will be welcome, particularly for use in screening contexts.
Take home messages
Self-taken vaginal samples are equivalent to liquid-based cytology (LBC) samples for the detection of cervical intraepithelial neoplasia (CIN)2+ in a colposcopy setting.
Compared with the LBC or vaginal sample, only three fewer cases of CIN3+ were missed in the urine sample (29 vs 32).
Post treatment, women may be more likely to test human papillomavirus positive on a urine or vaginal sample (compared with LBC), suggesting that self-taken samples may have limited utility as biospecimens for test of cure.
We thank Dr K Nale for expert advice and review of histology.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Correction notice This article has been corrected since it was published Online First. The provenance and peer review statement has been amended.
Handling editor Slade Jensen
Contributors GAS: principal investigator of study and responsible for study design, oversight and delivery of study (including dissemination). HC: lead clinician involved in patient-facing activities and collation of clinical data. GB: provided input into study design, governance and manuscript preparation. AF: laboratory lead—oversight of specimen manipulation and HPV testing including quality aspects. LG: laboratory scientist—delivered HPV testing relevant to study. CG: statistician—provided statistical support during manuscript preparation. KC: provided support for raw data analysis and manuscript drafting at various stages.
Competing interests GAS and GB: associated institution has received research monies/associated consumables from Roche Molecular Systems.
Ethics approval The study was approved by the NHS West of Scotland Research Ethics Service, ref: 11/AL/0084 and sponsored by NHS Dumfries and Galloway.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.