Management of high-risk HPV-positive women for detection of cervical (pre)cancer

Expert Review of Molecular Diagnostics ISSN: 1473-7159 (Print) 1744-8352 (Online) Journal homepage: http://www.tandfonline.com/loi/iero20 Management of high-risk HPV-positive women for detection of cervical (pre)cancer Roosmarijn Luttmer, Lise M. A. De Strooper, Renske D. M. Steenbergen, Johannes Berkhof, Peter J. F. Snijders, Daniëlle A. M. Heideman & Chris J. L. M. Meijer To cite this article: Roosmarijn Luttmer, Lise M. A. De Strooper, Renske D. M. Steenbergen, Johannes Berkhof, Peter J. F. Snijders, Daniëlle A. M. Heideman & Chris J. L. M. Meijer (2016) Management of high-risk HPV-positive women for detection of cervical (pre)cancer, Expert Review of Molecular Diagnostics, 16:9, 961-974, DOI: 10.1080/14737159.2016.1217157 To link to this article: https://doi.org/10.1080/14737159.2016.1217157 2016 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group Accepted author version posted online: 26 Jul 2016. Published online: 05 Aug 2016. Submit your article to this journal Article views: 679 View related articles View Crossmark data Citing articles: 5 View citing articles Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalinformation?journalcode=iero20

EXPERT REVIEW OF MOLECULAR DIAGNOSTICS, 2016 VOL. 16, NO. 9, 961 974 http://dx.doi.org/10.1080/14737159.2016.1217157 PERSPECTIVE Management of high-risk HPV-positive women for detection of cervical (pre)cancer Roosmarijn Luttmer a,b, Lise M. A. De Strooper a, Renske D. M. Steenbergen a, Johannes Berkhof c, Peter J. F. Snijders a, Daniëlle A. M. Heideman a and Chris J. L. M. Meijer a a Department of Pathology, VU University Medical Center, Amsterdam, the Netherlands; b Department of Obstetrics & Gynecology, Diakonessenhuis, Utrecht, the Netherlands; c Department of Epidemiology & Biostatistics, VU University Medical Center, Amsterdam, the Netherlands ABSTRACT Introduction: Primary HPV-testing has been shown to provide a superior detection of women at risk of cervical (pre)cancer compared to cytology-based screening. However, as most high-risk HPV infections are harmless, additional triage testing of HPV-positive women is necessary to identify those with cervical (pre)cancer. In this paper, we compare the performance, advantages and limitations of clinically relevant available triage strategies for HPV-positive women. Areas covered: Many different colposcopy triage strategies, comprising both microscopy-based and molecular (virus/host-related) markers, have been suggested: Pap cytology, p16/ki-67 dual-stained cytology, HPV16/18 genotyping, viral DNA methylation and host cell DNA methylation. Literature search was limited to triage strategies that have achieved at least phase 2 of the five-phase framework for biomarker development and studies including large cohorts ( 100 hrhpv-positive women). Triage markers were stratified by sample type (cervical scrape, self-collected sample) and by study population (screening, non-attendee, referral). Expert commentary: At present, repeat Pap cytology and Pap cytology combined with HPV16/18 genotyping are the only triage strategies that have been robustly shown to be ready for implementation. Other strategies such as p16/ki-67 dual-stained cytology and host cell DNA methylation analysis, with or without additional HPV16/18 genotyping, are attractive options for the near future. ARTICLE HISTORY Received 19 April 2016 Accepted 22 July 2016 Published online 5 August 2016 KEYWORDS Human papillomavirus; cervical cancer; cervical intraepithelial neoplasia (CIN); screening; triage; cytology; genotyping; DNA methylation; p16/ki-67 dual-stained cytology; biomarker 1. Introduction With an estimated 528,000 new cases and 266,000 deaths per year [1,2], cervical cancer is an important public health problem. Globally, cervical cancer ranks fourth among the most common causes of cancer-related death in women each year [1,2]. The majority of the global cervical cancer burden (85%) occurs in low- and middle-income countries reflecting a disproportionate availability of cervical cancer prevention programs, that is, organized screening (secondary prevention) and prophylactic human papillomavirus (HPV) vaccination (primary prevention) [1]. Both prevention strategies have evolved greatly in the last decades. It has been established that cervical cancer results from a persistent infection with an oncogenic, so-called high-risk (hr) type of HPV, a common sexually transmitted virus [3]. This insight has led to the development of prophylactic HPV vaccines, which effectively protect against vaccine type-associated infections and cervical (pre)cancer [4 6]. A broader use of HPV vaccination, especially in the low- and middleincome countries, is required to reduce cervical cancer mortality drastically [7]. As long as organized implementation and uptake of HPV vaccination is limited, screening remains the cornerstone of cervical cancer prevention. Cervical carcinogenesis is a very slow process that lasts approximately 15 30 years from initial hrhpv infection to cancer development [8]. This process is characterized by cytomorphologically recognizable premalignant stages, which can be adequately treated to prevent progression to cancer [9]. For these reasons, cervical cancer is an ideal screening target. Indeed, the introduction of screening by cytological evaluation of cervical scrapes has substantially reduced the incidence of and mortality from cervical cancer in developed countries [10,11]. However, frequent repeats of Pap cytology (which refers to the evaluation of both conventional and liquid-based cervical scrapes) are necessary to compensate for its moderate sensitivity [12,13]. The recognition that hrhpv is the causative agent of cervical cancer has changed the ideas about cervical cancer screening. Large longitudinal studies have shown that screening can be improved by the introduction of a hrhpv test as a primary screening tool [14 26]. Compared to primary screening with Pap cytology, a primary hrhpv test has a superior sensitivity for cervical cancer [24] and its high-grade precursor lesions (cervical intraepithelial neoplasia grade 3, CIN3) [14 21,26]. Moreover, a negative hrhpv test has been shown to provide a significantly higher long-term reassurance against the presence or development of CIN3 or cervical cancer (together referred to as CIN3) compared to a normal Pap cytology result [22,23]. In addition, in contrast to cytology, hrhpv testing is a molecular and thus highly reproducible and objective strategy [27]. Based on these data, hrhpv-testing will replace CONTACT Chris J. L. M. Meijer cjlm.meijer@vumc.nl Department of Pathology, VU University Medical Center, De Boelelaan 1117, 1081 HV, Amsterdam, the Netherlands 2016 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.

962 R. LUTTMER ET AL. Pap cytology as a primary test in many screening programs in the near future [28 30]. An important limitation of primary hrhpv-testing is its 2.5 4% lower specificity for cervical (pre)cancer compared to primary Pap cytology testing [25]. Most hrhpv-positive women have transient (so-called productive) infections; they do not have any clinical symptoms and clear the virus spontaneously. Although these women can develop CIN lesions representing productive lesions, they do not develop cervical cancer and the regression of the lesions coincides with viral clearance. Only a minority of hrhpv-positive women (20%) harbors a transforming infection, in which the normal viral life cycle is aborted and the viral early genes E6 and E7 are overexpressed in proliferating cells, leading to the development of clinically relevant precancer (a subset of CIN2 and the majority of CIN3) [31 33]. Still, a substantial part of CIN2 (43%) and CIN3 (32%) [34] regresses spontaneously following viral clearance. The median time between the onset of CIN2/3 and the genesis of cervical cancer has been estimated at 23.5 years [8]. Over a period of 30 years, approximately 30% of CIN3 progresses to cancer [35]. Hence, CIN2/3 represent a heterogeneous disease with a highly variable risk of progression to cancer, in part dependent on the time of existence [32,33]. Therefore, referral of all hrhpv-positive women to a gynecologist for colposcopy would overload the diagnostic chain and induce significant overtreatment, leading to unnecessary costs and higher risks of obstetric complications [36,37]. Thus, it is of great importance to implement a feasible triage strategy, which effectively identifies only those hrhpv-positive women with cervical precancers that have a high short-term risk of progression to cancer, as these women are in need of treatment. Currently, there is neither an optimal management strategy for hrhpv-positive women nor consensus on the criteria that should be met to justify implementation of a triage strategy. Ideally, the negative predictive value (NPV) for CIN3 should be determined longitudinally to assess the screening interval that can be maintained after a negative triage test [38]. In previous studies evaluating the clinical applicability of different triage strategies, a cross-sectional NPV threshold of 98% for CIN3 has been suggested, indicating that the CIN3 risk after a negative triage test result should be less than 2% [38 40]. This threshold is similar to the CIN3 risk of women with borderline or mild dyskaryosis cytology followed by a negative cytology result at 6 and 18 months follow-up (1.2%) [41], which is presently accepted in the Netherlands. In addition, the positive predictive value (PPV) has been suggested as a criterion for implementation, as it indicates the ability of a test to prevent unnecessary colposcopy referrals. Of note, this is a more subjective criterion dependent of the available resources and the screening interval used in different countries: in US settings, using a 3-year interval, a PPV threshold of 10% is considered acceptable [38], whereas in the Netherlands, where 5-year intervals are used, this threshold is 20% [39,40]. For biomarker development for early detection of cancer, a five-phase framework has been proposed [42,43]. The designated phases are (1) preclinical exploratory studies, (2) clinical assay development for clinical disease and assessment in noninvasive samples, (3) retrospective longitudinal repository studies, (4) prospective screening studies, and (5) prospective intervention studies. Phase 5 preferentially concerns a population-based randomized controlled trial where a new biomarker test is applied and evaluated against the reference [42,43]. This review will discuss the performance, advantages, and limitations of the clinically relevant available strategies for detection of cervical (pre)cancer in hrhpv-positive women. We chose to describe only triage tests that have achieved at least phase 2 of the described validation process, and preferably phase 3 or further [42,43]. Second, we only considered data obtained in cohorts of hrhpv-positive women with adequate sample size (roughly taking a total number of 100 hrhpv-positive women as a minimum). We furthermore focus on the accuracy of triage tests for CIN3. Most triage studies have used both CIN2 and CIN3 as histological end points. But as CIN2 reflects a heterogeneous group of lesions comprising both productive and transforming HPV infections, the end point CIN3 is more reproducible and comparable between studies [32,44]. In addition, a subdivision of triage markers into microscopy-based and molecular assays is made; the latter group is split into virus-related and host cell-related markers. For triage tests that allow application on self-collected samples as well as cervical scrapes, both sample types are discussed separately. Furthermore, we describe the available data stratified by the type of study population, as this might influence the reported triage test performance. Prospective evaluations of new triage tests among hrhpv-positive women recruited from population-based screening are relatively scarce. Cohorts of hrhpv-positive women selected among women who did not attend regular screening (further referred to as non-attendee populations), women visiting gynecological outpatient departments (gynecologic outpatient populations), or women referred to the colposcopist because of abnormal Pap cytology (colposcopy referral populations) have been regularly used as alternatives. 2. Triage strategies 2.1. Microscopy-based strategies 2.1.1. Pap cytology Currently, Pap cytology is the most advocated triage strategy for hrhpv-positive women [39,40,45,46]. Due to experience with Pap cytology as a primary cervical screening test for decades, Pap cytology is a well-studied strategy embedded in the healthcare infrastructure of many developed countries. The main limitation of Pap cytology is its subjective nature, which results in a suboptimal sensitivity for CIN3. In large population-based screening studies (ATHENA n = 4275; VUSA-screen n = 1303, POBASCAM n = 1100), triage of hrhpv-positive women by a single Pap cytology test had a sensitivity for CIN3 varying between 52.8% and 75.4%, with a corresponding specificity of 78.0 85.6% [39,40,47]. Of note, these estimates were obtained in study settings where cytologists were blinded to hrhpv-test results. Recently, it has been shown that providing cytologists with information on the hrhpv-positive status of participants increases the overall sensitivity of Pap cytology [48,49], probably due to the beneficial preselection of a population with a higher a priori cervical cancer risk. This is, however, at the cost

EXPERT REVIEW OF MOLECULAR DIAGNOSTICS 963 of a significant decrease in specificity, which can be seen as the negative consequence of a more meticulous cytological evaluation [49,50]. In a hrhpv-positive gynecologic outpatient population, a sensitivity of 80.8% and a specificity of 63.3% of Pap cytology were found [51]. The reported PPVs of a single Pap cytology test for triage of hrhpv-positive women range between 14.1% and 42.2% in screening populations [39,40,47]. Longitudinal studies in screening populations have shown that a single negative Pap cytology result in a hrhpv-positive woman yields insufficient reassurance against CIN3 (NPV 94.3 95.1%) [39,40] to justify direct dismissal from further follow-up, as the required NPV threshold of 98% [38] is not met. To solve this problem, repeat Pap cytology (at baseline and 6 or 12 months follow-up) has been suggested, which yields an acceptable NPV for CIN3 of 98.5 99.3% [39,40]. Based on these data, and on the broad experience with Pap cytology, hrhpv-based screening with repeat cytology triage at 0 and 6 months will be implemented in the Netherlands in 2017. However, repeat testing is inevitably associated with loss to follow-up, which has been shown to be as high as 28 33% [39,40]. Furthermore, it has been shown that Pap cytology cannot be reliably performed on self-sampled material [52 55]. Another general limitation is the suboptimal sensitivity of Pap cytology for cancers: even after the preselection of hrhpvpositive women, not all cervical carcinomas are detected [24,56]. In summary, Pap cytology has been robustly validated (phase 4 [43]) as a triage marker for hrhpv-positive women, however with the important limitations of a suboptimal reproducibility, a significant susceptibility to pre-investigational knowledge of the HPV status, and the necessity of repeat testing. 2.1.2. p16/ki-67 dual-stained cytology The search for a way to improve the reproducibility of cytology has led to the development of the combined immunostaining of the proteins p16 INK4A and Ki-67 in cervical cytology specimens (referred to as p16/ki-67 dual-stained cytology) [57 60]. Simultaneous expression of the antiproliferative p16 protein and the proliferation marker Ki-67 should exclude each other in cells under normal physiological conditions [61 64]. The simultaneous expression of p16 and Ki-67 in one cell is indicative for cell cycle dysregulation and appears indicative for a transforming hrhpv infection [61 64]. Three recent studies have addressed the performance of p16/ki-67 dualstained cytology for triage of hrhpv-positive women, in population-based screening cohorts (n = 1509 [65] and n = 396 [66]) and a gynecologic outpatient population (n = 446 [67]). In these studies, p16/ki-67 dual-stained cytology had a consistently high sensitivity for CIN3 (86.9 93.8% which was comparable to that of Pap cytology [65 67]. The specificity for CIN3 of p16/ki-67 dual-stained cytology was estimated at 51.2% [67] to 56.9% [65], which is high compared to Pap cytology (44.9% [67] to 48.7% [65]). Because p16/ki-67 dualstained cytology can be adequately interpreted by relatively untrained staff, it has a decreased interobserver variability compared to Pap cytology [68,69]. The threshold for positivity of the assay is the presence of p16 and Ki-67 in one single cell [57 60,65,67,69], which emphasizes the need for sufficient cellularity of specimens with sufficient intact abnormal cells to guarantee the clinical applicability of this triage strategy. Concordantly, p16/ki-67 dual-stained cytology is not reliably applicable to self-sampled material [70]. Although data on the long-term CIN3 risk in p16/ki-67 dual-stain negative women are currently lacking, results from the studies with short-term follow-up (up to 2 years) [65] are promising, reporting NPVs of a single p16/ki-67 dual-stained cytology test between 97.4% and 99.4% [65,67]. In a population-based screening cohort, Wentzensen et al. found a PPV of 12.4% [65], whereas in an outpatient population Luttmer et al. found a PPV of 29.9% [67]. In summary, p16/ki-67 dual-stained cytology could serve as a more reproducible alternative to Pap cytology with similar sensitivity but higher specificity for CIN3. 2.1.3. Other microscopy-based strategies Several other microscopy-based strategies have been suggested for application in cervical cancer screening, such as the detection of a gain of chromosome 3q or the human telomerase gene (TERC) in cytological specimens by fluorescence in situ hybridization [71 74]. To the best of our knowledge, these strategies have not yet been evaluated in sufficiently large clinical cohorts of hrhpv-positive women. 2.2. Molecular triage strategies 2.2.1. Virus-related markers 2.2.1.1. HPV genotyping. The detection of specific hrhpv types (referred to as HPV genotyping) is based on the understanding that carcinogenic potential differs significantly between hrhpv types. HPV16 is the genotype associated with the highest (pre)cancer risk, followed by HPV18 [75,76]. Concordantly, the proportion of HPV16 and HPV18 positivity increases with histological severity. Together, HPV16 and HPV18 are responsible for the development of 52 64% of CIN3 lesions and approximately 70% of cervical carcinomas [75,76]. Women with a single HPV16 and/or HPV18 positive test have a fivefold increased risk of developing CIN3 during the following 10 years compared to women infected with other hrhpv types [77,78]. Therefore, the detection of HPV16 and/or HPV18 might serve to identify women with an elevated risk of cervical (pre)cancer. Inherent to the genotype restriction, the use of HPV16/18 genotyping for triage of hrhpv-positive women has a limited CIN3 sensitivity, which has been estimated at 59.5 65.4% in population-based screening studies [39,47]. In relatively highrisk populations, comprising screening non-attendees or gynecology patients, HPV16/18 genotyping had a CIN3 sensitivity of 72.2 74.8% [51,79]. HPV16/18 genotyping has been shown to yield a very reasonable CIN3 specificity (72.5 75.6% in a screening and non-attendee population [40,79] and 57.4% in a gynecologic outpatient population [51]. However, the NPV of HPV16/18 genotyping varies between 90.6% and 93.4% (regardless of population type [39,51,79], indicating that the residual CIN3 risk in nonhpv16/18 hrhpv-positive women does not allow safe dismissal of these women from further follow-up [38]. The reported PPVs of HPV16/18 genotyping in screening populations vary between 15.5% and 26.1% [39,47].

964 R. LUTTMER ET AL. An advantage of HPV16/18 genotyping is its applicability to self-sampled material. The performance of HPV16/18 genotyping has been evaluated on both lavage- and brush-based selfsamples of hrhpv-positive non-attendees, revealing an accuracy which appears to be comparable to cervical scrapes ( CIN3 sensitivity 65.4 69.4%, specificity 65.0 70.9%, PPV 25.0 48.5%, NPV 85.4 92.4% [80,81]). It is obvious that regardless of sample type HPV16/18 genotyping does not suffice as a direct stand-alone triage strategy for hrhpv-positive women, but the combination of HPV16/18 genotyping with other tests has been frequently suggested. In screening populations, the combined use of HPV16/18 genotyping and Pap cytology has been shown to yield a high sensitivity (78.2 96.6% [39,40,47]). Moreover, these studies found a high NPV (97.1 98.8% [39,40]) that approaches the required 98% threshold [38], indicating that the remaining CIN3 risk in women who test both cytology-negative and HPV16/18-negative justifies their dismissal to the next screening round. For this reason, combined HPV16/18 genotyping and Pap cytology is an appealing strategy that could eliminate the need for repeat testing which is necessary for Pap cytology alone and thereby circumvent the problem of loss to follow-up. The reported specificity of combined cytology and HPV16/18 genotyping varies between 53.6% and 63.2% in screening populations [39,40], 47.3% in non-attendees [79], and 37.5% in a gynecologic outpatient population [51]. Relatively low PPVs (12.6 28.5%) in screening populations [39,40,47] indicate that this strategy might induce some over-referral. Following these data, Pap cytology combined with HPV16/ 18 genotyping has been approved for triage of hrhpv-positive women in the USA [30] and might also serve as a feasible strategy in other countries. As several clinically validated hrhpv-assays provide an automatic readout of HPV16 and HPV18 results [82,83], genotyping is an easy additive option. Especially countries with low-quality cytological assessment or limited possibilities for call and recall in screening might benefit from such a direct triage strategy. In summary, the combination of HPV16/18 genotyping with Pap cytology is the only currently available strategy that enables safe triage of hrhpv-positive women without repeat testing. 2.2.1.2. HPV DNA methylation analysis. Methylation analysis of the HPV genome has been suggested as a method to identify women with cervical (pre)cancer [84]. Viral methylation, which generally leads to the silencing of viral genes, is one of many protective mechanisms exerted by human host cells to inhibit viral invasion [84 91]. The assessment of viral methylation has been described for various hrhpv types [90]. In general, findings on the association of the methylation of specific viral CpG sites and cervical disease are inconsistent. Inconsistencies might be explained by small sample sizes, differences in DNA methylation detection method, different sample types, and/or the specific CpGs analyzed. To date, only a specific combination of CpGs within the L1 and L2 genes of HPV16 and HPV18 has shown consistent hypermethylation with increasing disease severity between different studies. For other HPV types, the available information is scarce (reviewed by Clarke et al. [84]). Altogether these data show that methylation analysis of viral DNA has not yet matured to an efficient triage assay. On the other hand, combined assessment of methylation of both viralspecific CpGs (HPV16/18/31/33-L1) and human host cell DNA (EPB41L3) has been clinically validated as a triage marker in a hrhpv-positive screening population (n = 341), yielding a CIN3 sensitivity of 84% (95% confidence interval [CI] 62 94%) and a corresponding specificity of 63% (95% CI 58 68%) [86]. A general limitation of viral methylation analysis is that for each hrhpv-type with unique DNA sequence, a unique assay has to be developed and clinically validated, making short-term implementation challenging. 2.2.2 Host cell-related markers 2.2.2.1. Host cell DNA methylation analysis. Methylation of the host cell genome is a frequently observed epigenetic phenomenon in cervical carcinogenesis [92]. HPV-E6- and -E7-modulated activation of DNA-methyltransferase activity and chromatin remodeling lead to hypermethylation of CpG-islands in the host genome [93]. In the context of promoter regions, this hypermethylated state generally results in transcriptional repression [94]. Several candidate tumor-suppressor genes have been identified as consistently methylated in cervical cancer [32,92]. Methylation levels of certain host cell genes (such as CADM1, MAL, mir124-2, JAM3, TERT, C13ORF, EPB41L3, ANKRD18CP, GFRA1, CDH6, LHX8, GATA4, PRDM14, and FAM19A4) in cervical scrapes have been shown to increase with the severity of underlying cervical (pre)cancer [51,95 99]. Furthermore, host cell methylation analysis of CADM1, MAL, mir124-2, and FAM19A4 in cervical scrapes allows the discrimination of high-grade CIN with a long-term preceding hrhpv-infection from those with a short-term hrhpv infection [95,96]. Concordantly, in both cervical scrapes and selfsampled cervicovaginal lavages, methylation levels of FAM19A4 have been shown to correspond with the estimated CIN lesion volume [100]. Methylation levels of CADM1, MAL, mir124-2, and FAM19A4 are extraordinarily high in cervical scrapes of women with carcinomas [96,101], resulting in a 100% sensitivity for carcinomas of methylation assays targeting these genes, even when very high thresholds are used to define methylation positivity. As previously described, all these findings underline that in contrast to cytology host cell promoter methylation analysis specifically detects so-called advanced CIN lesions, which harbor a cancer-like methylation profile and have a high short-term risk of progression to cancer [32,33]. Therefore, host cell DNA methylation analysis might be of value as a marker for the detection of cervical (pre)cancer among hrhpv-positive women [32,92]. A variety of methylation markers has been studied in clinical cohorts of hrhpvpositive women [51,80,81,96,97,99,102 106]. In general, these studies were cross-sectional or had a short-term follow-up and the majority of studies were performed in nonscreening cohorts. A selection of the most relevant studies, based on the criteria mentioned in the introduction section, has been listed in Tables 1 4.

EXPERT REVIEW OF MOLECULAR DIAGNOSTICS 965 Table 1. Triage of hrhpv-positive women by host DNA methylation analysis on physician-taken cervical scrapes. Target gene(s) Reference n Age Follow-up Sensitivity CIN3 Specificity CIN3 NPV CIN3 PPV CIN3 Screening populations FAM19A4 [96] 218 33 63 18 months 75.8 (61.1 90.4) 67.0 (60.3 73.8) 93.9 (89.9 98.0) 29.1 (19.5 38.7) CADM1/MAL [102] 234 33 63 18 months 68.4 (53.6 83.2) 75.5 (69.5 81.5) 92.5 (88.4 96.6) 35.1 (24.3 46.0) Non-attendee populations JAM3 [97] 128 33 63 Median 15 months 82.0 (65.5 93.2) 88.0 (80.0 94.0) na na C13ORF18 [97] 128 33 63 Median 15 months 65.0 (46.5 80.2) 91.0 (83.9 96.2) na na TERT [97] 128 33 63 Median 15 months 76.0 (58.8 89.2) 60.0 (49.0 69.6) na na EPB41L3 [97] 128 33 63 Median 15 months 88.0 (72.5 96.6) 61.0 (50.0 70.6) na na CADM1/MAL [107] 364 33 63 12 months 69.4 (57.9 80.8) 71.2 (66.1 76.3) 91.9 (88.4 95.4) 33.1 (25.0 41.2) Gynecologic outpatient populations FAM19A4 [51] 287 30 66 Cross-sectional 88.3 (80.2 96.5) 62.1 (55.8 68.4) 95.3 (91.9 98.7) 38.1 (30.1 46.2) FAM19A4 [51] 221 18 29 Cross-sectional 50.0 (32.1 67.9) 81.7 (76.2 87.2) 91.2 (87.0 95.5) 30.0 (17.3 42.7) Colposcopy referral populations JAM3 [104] 143 32 83 Cross-sectional 70 83 na na JAM3 [105] 267 Mean 44 ± SD10 Cross-sectional 86.0 73.0 93.8 43.8 JAM3 [105] 224 Mean 42 ± SD10 Cross-sectional 80.6 80.3 94.8 48.1 JAM3 [99] 152 IQR 32 43 Cross-sectional 80 76 na na C13ORF18 [104] 143 32 83 Cross-sectional 62 97 na na C13ORF18 [99] 152 IQR 32 43 Cross-sectional 54 88 na na TERT [104] 143 32 83 Cross-sectional 54 88 na na TERT [99] 152 IQR 32 43 Cross-sectional 90 43 na na EPB41L3 [104] 143 32 83 Cross-sectional 70 83 na na EPB41L3 [99] 152 IQR 32 43 Cross-sectional 85 68 na na ANKRD18CP [99] 152 IQR 32 43 Cross-sectional 74 71 na na CDH6 [99] 152 IQR 32 43 Cross-sectional 83 57 na na GFRA1 [99] 152 IQR 32 43 Cross-sectional 64 85 na na GATA4 [99] 152 IQR 32 43 Cross-sectional 88 33 na na LHX8 [99] 152 IQR 32 43 Cross-sectional 91 47 na na C13ORF18/JAM3/ANKRD18CP [99] 152 IQR 32 43 Cross-sectional 85 61 na na JAM3/GFRA1/ANKRD18CP [99] 152 IQR 32 43 Cross-sectional 85 63 na na JAM3/ANKRD18CP [99] 152 IQR 32 43 Cross-sectional 85 64 na na DLX1/ITGA4/RXFP3/SOX17/ZNF671 ( 2/5 positive) [106] 103 30 81 Cross-sectional 96.2 (80.4 100) 76.6 (65.6 85.5) na na DLX1/ITGA4/RXFP3/SOX17/ZNF671 ( 2/5 positive) [106] 114 18 29 Cross-sectional 43.8 (19.8 70.1) 87.8 (79.6 93.5) na na CIN3: Cervical intraepithelial neoplasia grade 3 or worse; NPV: negative predictive value; CI: confidence interval; PPV: positive predictive value; IQR: interquartile range; SD: standard deviation; na: not applicable.

966 R. LUTTMER ET AL. lavage FAM19A4 [100] 198 18 29 Self-collected lavage 37.5 (18.1 56.9) 90.8 (86.5 95.1) 91.3 (87.1 95.5) 36.0 (17.2 54.8) CIN3: Cervical intraepithelial neoplasia grade 3 or worse; CI: confidence interval; NPV: negative predictive value; PPV: positive predictive value. Table 3. Triage of hrhpv-positive women by host DNA methylation analysis combined with Pap cytology (on physician-taken cervical scrapes). scrape FAM19A4 and/or ASC-US [51] 221 18 29 Physician-taken scrape Sensitivity CIN3 Specificity CIN3 NPV CIN3 PPV CIN3 18 months 86.8 (76.1 97.6) 64.8 (58.1 71.5) 96.2 (93.0 99.5) 32.4 (23.3 41.4) 12 months 88.7 (80.8 96.6) 53.6 (48.0 59.3) 95.9 (92.9 98.9) 28.2 (21.9 34.5) Table 2. Triage of hrhpv-positive women by host DNA methylation analysis on self-collected samples. Target gene(s) Reference n Age Sample type Follow-up Sensitivity CIN3 Specificity CIN3 NPV CIN3 PPV CIN3 Non-attendee populations MAL/mir124-2 [80] 1019 33 63 Self-collected lavage 12 months 73.5 (66.3 80.6) 47.2 (43.9 50.6) 91.4 (88.8 93.9) 19.0 (15.8 22.2) FAM19A4/mir124-2 [81] 389 33 63 Self-collected 12 months 70.5 (60.4 80.6) 67.8 (62.7 73.0) 90.2 (86.4 94.0) 35.5 (28.0 43.0) lavage FAM19A4/mir124-2 [81] 254 33 63 Self-collected 12 months 69.4 (58.8 80.1) 76.4 (70.2 82.6) 86.3 (81.0 91.6) 53.8 (43.6 63.9) brush Gynecologic outpatient populations FAM19A4 [100] 252 30 66 Self-collected Cross- 78.4 (67.1 89.7) 73.1 (67.0 79.3) 93.0 (89.1 97.0) 42.6 (32.6 52.5) sectional Crosssectional Target gene(s) and additional test Reference n Age Sample type Follow-up Screening population CADM1/MAL and/or ASC-US [102] 234 30 60 Physician-taken scrape Non-attendee population CADM1/MAL and/or ASC-US [107] 364 33 63 Physician-taken scrape Gynecologic outpatient populations FAM19A4 and/or ASC-US [51] 287 30 66 Physician-taken Crosssectional Crosssectional 96.7 (92.1 100) 32.2 (26.1 38.2) 97.3 (93.7 100) 27.4 (21.4 33.4) 90.0 (79.3 100) 44.5 (37.5 51.6) 96.6 (92.8 100) 20.3 (13.5 27.1) CIN3: Cervical intraepithelial neoplasia grade 3 or worse; CI: confidence interval; NPV: negative predictive value; PPV: positive predictive value. 2.2.2.2. Host cell DNA methylation analysis on cervical scrapes.. FAM19A4 methylation analysis [98] has been evaluated on cervical scrapes from a screening cohort (n = 234 [96]) and a gynecologic outpatient population (n = 508 [51]). FAM19A4 methylation analysis has been shown to be dependent on age, with a substantially higher sensitivity (75.8 88.3%) and lower specificity (62.1 67.0%) for CIN3 in women aged 30 years [51,96] compared to younger women ( CIN3 sensitivity 50%, specificity 81.7%; Table 1 [51]). Regardless of age, the NPV of FAM19A4 methylation analysis for CIN3 was relatively consistent, ranging between 91.2% and 95.3% [51]. Combined methylation analysis of both CADM1/MAL on hrhpv-positive cervical scrapes has yielded a comparable accuracy in hrhpv-positive scrapes of screening attendees (n = 234 [102]) and non-attendees (n = 364 [107], Table 1), with a CIN3 sensitivity of 68.4 69.4% and a corresponding specificity of 71.2 75.5%. Methylation analysis of JAM3 has been clinically evaluated in hrhpv-positive cohorts of non-attendees [97] and women referred to colposcopy [99,104,105], with a sensitivity for CIN3 ranging between 70.0% and 86.0% and a specificity of 73.0 88.0% (Table 1). In a cross-sectional study design, the NPV was estimated at 93.8 94.8% (Table 1) [105]. Methylation analysis of the single host genes TERT, C13ORF18, or EPB41L3 for triage of hrhpv-positive women had a more variable accuracy (Table 1) [97,104]. Recently, a novel set of single methylation markers (ANKRD18CP, CDH6, GFRA1, GATA4, LHX8) was described in a hrhpv-positive colposcopy referral population (n = 152) [99], yielding promising results for especially ANKRD18CP, with a CIN3 sensitivity of 74% and specificity of 71%, and GFRA1, with a CIN3 sensitivity of 64% and a specificity of 85% (Table 1) [99]. In a colposcopy referral population (n = 152 [99], Table 1), combined methylation analysis of C13ORF18/ JAM3/ANKRD18CP, JAM3/GFRA1/ANKRD18CP, and JAM3/ ANKRD18CP has been evaluated in hrhpv-positive cervical scrapes, with a CIN3 sensitivity of 85% and a specificity of 61 64%. In another colposcopy referral population (n = 117 [106]), methylation analysis of DLX1/ITGA4/RXFP3/SOX17/ ZNF671 yielded a clinically relevant accuracy especially among women aged 30 years ( CIN 3 sensitivity 96.2%, specificity 76.6%). In this study, methylation analysis was considered positive when at least two out of five genes were methylated [106]. 2.2.2.3. Host cell DNA methylation analysis on self-collected samples. In contrast to most candidate triage markers in hrhpv-positive women, host cell DNA methylation analysis has been shown feasible in self-sampled material. In a large randomized controlled trial among hrhpv-positive screening non-attendees (n = 1019) [103], combined methylation analysis of MAL/mir124-2 on self-sampled cervicovaginal lavage material yielded a CIN3 sensitivity of 73.5% and a specificity

EXPERT REVIEW OF MOLECULAR DIAGNOSTICS 967 Table 4. Triage of hrhpv-positive women by host DNA methylation analysis combined with HPV16/18 genotyping (on both physician-taken cervical scrapes and self-collected samples). Target gene(s) and additional test Reference n Age Sample type Follow-up Sensitivity CIN3 Specificity CIN3 NPV CIN3 PPV CIN3 Non-attendee populations MAL/mir124-2 and/or HPV16/18 [80] 1019 33 63 Self-collected lavage 12 months 77.6 (70.8 84.3) 54.8 (51.5 58.1) 93.5 (91.4 95.7) 22.4 (18.8 26.1) FAM19A4/mir124-2 and/or HPV16/18 [81] 389 33 63 Self-collected lavage 12 months 88.5 (81.4 95.6) 46.0 (40.4 51.5) 94.1 (90.3 97.8) 29.1 (23.3 34.9) FAM19A4/mir124-2 and/or HPV16/18 [81] 254 33 63 Self-collected brush 12 months 84.7 (76.4 93.0) 54.9 (47.7 62.2) 90.1 (84.5 95.7) 42.7 (34.6 50.8) Gynecologic outpatient populations FAM19A4 and/or HPV16/18 [51] 287 30 66 Physician-taken scrape Cross-sectional 93.3 (87.0 99.6) 37.0 (30.7 43.3) 95.5 (91.1 99.8) 28.1 (21.9 34.4) FAM19A4 and/or HPV16/18 [51] 221 18 29 Physician-taken scrape Cross-sectional 90.0 (79.3 100) 48.7 (41.6 55.8) 96.9 (93.4 100) 21.6 (14.1 28.8) FAM19A4 and/or HPV16/18 [100] 252 30 66 Self-collected lavage Cross-sectional 92.2 (84.8 99.5) 43.8 (36.9 50.6) 95.7 (91.5 99.8) 29.4 (22.3 36.4) FAM19A4 and/or HPV16/18 [100] 198 18 29 Self-collected lavage Cross-sectional 83.3 (68.4 98.2) 53.4 (46.0 60.9) 95.9 (91.9 99.8) 19.8 (12.0 27.6) CIN3: Cervical intraepithelial neoplasia grade 3 or worse; CI: confidence interval; NPV: negative predictive value; PPV: positive predictive value.

968 R. LUTTMER ET AL. of 47.2% (Table 2) [80,103]. Combined methylation analysis of FAM19A4 and mir124-2 was evaluated in a subset of these hrhpv-positive non-attendees using self-collected lavages (n = 389), as well as hrhpv-positive non-attendees from another cohort who submitted self-collected brushes (n = 254) [81]. This marker panel yielded a similar clinical performance for both devices ( CIN3 sensitivity 69.4 70.5%, specificity 67.8 76.4%), resulting in a similar sensitivity and higher specificity compared to the previously described MAL/ mir124-2 marker panel (Table 2) [81]. When comparing the different host cell DNA methylation markers (in both cervical scrapes and self-collected samples), several technical aspects should be considered. The combined methylation analysis of multiple genes in a multiplex qmsp assay, as was described for CADM1/MAL/mir124-2 [101] and FAM19A4/mir124-2 [81], is an efficient method with the important advantage of an internal control. In contrast to other suggested triage strategies, multiplex qmsp results in a high reproducibility and is suitable for a high-throughput setting, using a minimal amount of material [108]. The reported NPVs of different methylation marker combinations for CIN3, in both cervical scrapes and lavages, range from 86.3% to 95.3% (Tables 1 and 2). Searching for a way to increase the sensitivity and NPV that are currently reached by sole methylation testing, the combination of methylation analysis with other tests, such as Pap cytology (Table 3) and HPV16/18 genotyping (Table 4) has been suggested. 2.2.2.4. Combining host cell DNA methylation analysis with other markers. A few studies have evaluated the accuracy of methylation analysis (CADM1/MAL, FAM19A4) combined with Pap cytology (Table 3) [102,107]. In both a screening [102] and a non-attendee population [107], CADM1/MAL methylation analysis combined with Pap cytology yielded a high sensitivity for CIN3 of 86.8 88.7% with a corresponding specificity of 53.6 64.8% (Table 3). In a gynecologic outpatient population, combined FAM19A4 methylation analysis with Pap cytology had a sensitivity of 96.7%, however with a relatively low specificity of 32.2% among women 30 years of age [51]. In women <30 years, combined FAM19A4 methylation analysis and Pap cytology had a CIN3 sensitivity of 90.0% with a corresponding specificity of 44.5% (Table 3) [51]. These data are in agreement with the notion that most CIN3 lesions in young women (<30 years) are regressive lesions. A fully molecular alternative triage strategy is the combination of methylation analysis with HPV16/18 genotyping, which is applicable to both cervical scrapes and self-sampled material (Table 4). In self-collected samples of a large population of non-attendees, methylation analysis of MAL/mir124-2 combined with HPV16/18 genotyping had a CIN3 sensitivity of 77.6% at a specificity of 54.8% [80]. Also in self-samples of non-attendees, methylation analysis of FAM19A4/mir124-2 was evaluated in combination with HPV16/18 genotyping, yielding a higher sensitivity (84.7 88.5%) at a comparable specificity (46.0 54.9%) [102]. In a gynecologic outpatient population, FAM19A4 methylation analysis combined with HPV16/18 genotyping in cervical scrapes had a sensitivity of 90.0 93.3% and a specificity of 37.0 48.7% [51]. Regardless of population type, marker panel and age, the NPV of combined methylation analysis and HPV genotyping was relatively high (90.1 96.9%; Table 4). In summary, host cell DNA methylation analysis, possibly in combination with HPV16/18 genotyping, could serve as an attractive triage marker for hrhpv-positive women, with the advantage of applicability on self-collected samples. 2.2.2.5. Other molecular strategies. Besides the discussed strategies, several other molecular strategies, both virus- and host cell-related, have been suggested for detection of cervical (pre)cancer. For example, the detection of specific micrornas has shown promising results in one clinical study [109]. Furthermore, detection of the E6 protein [110,111], and E6/ E7 mrna [66,112 114] was evaluated in some clinical cohorts of hrhpv-positive women, however with disappointing results. These assays require a relatively complex genotype-specific approach and have not been shown to reach the desired NPV in hrhpv-positive women; others did not reach the required phase of marker validation and are therefore beyond the scope of this review [42,43]. 3. Expert commentary With the upcoming introduction of primary hrhpv-testing in cervical cancer screening in several countries, the question how to manage hrhpv-positive women is becoming increasingly urgent. At present, repeat Pap cytology and Pap cytology combined with HPV16/18 genotyping are the only robustly validated strategies for colposcopy triage of hrhpvpositive women. These strategies have been approved for implementation in the Netherlands [28] and the USA [30]. Other molecular strategies such as p16/ki-67 dual-stained cytology and host cell DNA methylation analysis, with or without additional HPV16/18 genotyping, are attractive options for the near future. The combination of Pap cytology and methylation analysis, which have been suggested to serve as complementary markers in terms of detection of early and advanced CIN [32], could serve as a sensitive strategy that, in contrast to Pap cytology alone, detects all carcinomas and virtually all CIN3. Table 5 provides an overview of test characteristics (that is, sensitivity, specificity, NPV, PPV) of Pap cytology, HPV16/18 genotyping, p16/ki-67 dual-stained cytology, and methylation analysis in a single (outpatient) study population [51,67,100,114]. A general limitation of microscopy-based strategies is the fact that these can only be adequately performed on cervical scrapes collected by health-care professionals. Fear or embarrassment related to the cervical scraping procedure and other practical reasons related to visiting a physician [115] restrain about 30% of eligible women from participation in cytology-based screening in the Netherlands [116]. This problem, which also exists in other industrialized countries, could be circumvented by offering these women a hrhpv-test on self-sampled material [103,117], which has been shown to be a feasible alternative to hrhpv-testing on cervical scrapes [118,119]. However, the use of Pap cytology or p16/ki-67 dual-stained cytology for triage would still require a visit to a physician. Ideally, both

EXPERT REVIEW OF MOLECULAR DIAGNOSTICS 969 Table 5. Overview of triage markers for hrhpv-positive women on cervical scrapes of a single study population. Triage marker n1/n1 Sensitivity (95%CI) (%) n2/n2 Total group (n = 421) CIN3 Pap cytology 65 / 75 86.7 (79.0 94.4) 161 / 346 46.5 (41.3 51.8) 94.2 (90.6 97.7) 26.0 (20.6 31.4) HPV16/18 genotyping 58 / 75 77.3 (67.9 86.8) 196 / 346 56.6 (51.4 61.9) 92.0 (88.4 95.7) 27.9 (21.8 34.0) p16/ki-67 dual-stained cytology 71 / 75 94.7 (89.6 99.8) 177 / 346 51.2 (45.9 56.4) 97.8 (95.6 99.9) 29.6 (23.8 35.4) FAM19A4 methylation analysis 57 / 75 76.0 (66.3 85.7) 246 / 346 71.1 (66.3 75.9) 93.2 (90.1 96.2) 36.3 (28.8 43.8) Pap cytology and/or HPV16/18 genotyping 71 / 75 94.7 (89.6 99.8) 91 / 346 26.3 (21.7 30.9) 95.8 (91.8 99.8) 21.8 (17.3 26.3) FAM19A4 methylation analysis and/or HPV16/18 genotyping 70 / 75 93.3 (87.7 99.0) 148 / 346 42.8 (37.6 48.0) 96.7 (93.9 99.5) 26.1 (20.9 31.4) CIN2 Pap cytology 130 / 152 85.5 (79.9 91.1) 149 / 269 55.4 (49.4 61.3) 87.1 (82.1 92.2) 52.0 (45.8 58.2) HPV16/18 genotyping 93 / 152 61.2 (53.4 68.9) 154 / 269 57.2 (51.3 63.2) 72.3 (66.3 78.3) 44.7 (38.0 51.5) p16/ki-67 dual-stained cytology 131 / 152 86.2 (80.7 91.7) 160 / 269 59.5 (53.6 65.3) 88.4 (83.7 93.1) 54.6 (48.3 60.9) FAM19A4 methylation analysis 90 / 152 59.2 (51.4 67.0) 202 / 269 75.1 (69.9 80.3) 76.5 (71.4 81.6) 57.3 (49.6 65.1) Pap cytology and/or HPV16/18 genotyping 139 / 152 91.4 (87.0 95.9) 82 / 269 30.5 (25.0 36.0) 86.3 (79.4 93.2) 42.6 (37.3 48.0) FAM19A4 methylation analysis and/or HPV16/18 genotyping 121 / 152 79.6 (73.2 86.0) 122 / 269 45.4 (39.4 51.3) 79.7 (73.4 86.1) 45.1 (39.2 51.1) Subgroup: women 30 years (n = 236) CIN3 Pap cytology 45 / 52 86.5 (77.3 95.8) 83 / 184 45.1 (37.9 52.3) 92.2 (86.7 97.8) 30.8 (23.3 38.3) HPV16/18 genotyping 39 / 52 75.0 (63.2 86.8) 102 / 184 55.4 (48.3 62.6) 88.7 (82.9 94.5) 32.2 (23.9 40.6) p16/ki-67 dual-stained cytology 50 / 52 96.2 (90.9 100) 91 / 184 49.5 (42.2 56.7) 97.8 (94.9 100) 35.0 (27.1 42.8) FAM19A4 methylation analysis 47 / 52 90.4 (82.4 98.4) 114 / 184 62.0 (54.9 69.0) 95.8 (92.2 99.4) 40.2 (31.3 49.1) Pap cytology and/or HPV16/18 genotyping 48 / 52 92.3 (85.1 99.6) 43 / 184 23.4 (17.3 29.5) 91.5 (83.5 99.5) 25.4 (19.2 31.6) FAM19A4 methylation analysis and/or HPV16/18 genotyping 49 / 52 94.2 (87.9 100) 67 / 184 36.4 (29.5 43.4) 95.7 (91.0 100) 29.5 (22.6 36.5) CIN2 Pap cytology 86 / 99 86.9 (80.2 93.5) 77 / 137 56.2 (47.9 64.5) 85.6 (78.3 92.8) 58.9 (50.9 66.9) HPV16/18 genotyping 59 / 99 59.6 (49.9 69.3) 75 / 137 54.7 (46.4 63.1) 65.2 (56.5 73.9) 48.8 (39.9 57.7) p16/ki-67 dual-stained cytology 89 / 99 89.9 (84.0 95.8) 83 / 137 60.6 (52.4 68.8) 89.2 (83.0 95.5) 62.2 (54.3 70.2) FAM19A4 methylation analysis 72 / 99 72.7 (64.0 81.5) 92 / 137 67.2 (59.3 75.0) 77.3 (69.8 84.8) 61.5 (52.7 70.4) Pap cytology and/or HPV16/18 genotyping 91 / 99 91.9 (86.6 97.3) 39 / 137 28.5 (20.9 36.0) 83.0 (72.2 93.7) 48.1 (41.0 55.3) FAM19A4 methylation analysis and/or HPV16/18 genotyping 83 / 99 83.8 (76.6 91.1) 54 / 137 39.4 (31.2 47.6) 77.1 (67.3 87.0) 50.0 (42.4 57.6) Subgroup: women < 30 years (n = 185) CIN3 Pap cytology 20 / 23 87.0 (73.2 100) 78 / 162 48.1 (40.5 55.8) 96.3 (92.2 100) 19.2 (11.7 26.8) HPV16/18 genotyping 19 / 23 82.6 (67.1 98.1) 94 / 162 58.0 (50.4 65.6) 95.9 (92.0 99.8) 21.8 (13.2 30.5) p16/ki-67 dual-stained cytology 21 / 23 91.3 (79.8 100) 86 / 162 53.1 (45.4 60.8) 97.7 (94.6 100) 21.6 (13.5 29.8) FAM19A4 methylation analysis 10 / 23 43.5 (23.2 63.7) 132 / 162 81.5 (75.5 87.5) 91.0 (86.4 95.7) 25.0 (11.6 38.4) Pap cytology and/or HPV16/18 genotyping 23 / 23 100.0 (100 100) 48 / 162 29.6 (22.6 36.7) 100 (100 100) 16.8 (10.5 23.0) FAM19A4 methylation analysis and/or HPV16/18 genotyping 21 / 23 91.3 (79.8 100) 81 / 162 50.0 (42.3 57.7) 97.6 (94.3 100) 20.6 (12.7 28.4) CIN2 Pap cytology 44 / 53 83.0 (72.9 93.1) 72 / 132 54.5 (46.1 63.0) 88.9 (82.0 95.7) 42.3 (32.8 51.8) HPV16/18 genotyping 34 / 53 64.2 (51.2 77.1) 79 / 132 59.8 (51.5 68.2) 80.6 (72.8 88.4) 39.1 (28.8 49.3) p16/ki-67 dual-stained cytology 42 / 53 79.2 (68.3 90.2) 77 / 132 58.3 (49.9 66.7) 87.5 (80.6 94.4) 43.3 (33.4% 53.2) FAM19A4 methylation analysis 18 / 53 34.0 (21.2 46.7) 110 / 132 83.3 (77.0 89.7) 75.9 (68.9 82.8) 45.0 (29.6 60.4) Pap cytology and/or HPV16/18 genotyping 48 / 53 90.6 (82.7 98.4) 43 / 132 32.6 (24.6 40.6) 89.6 (80.9 98.2) 35.0 (27.0 43.0) FAM19A4 methylation analysis and/or HPV16/18 genotyping 38 / 53 71.7 (59.6 83.8) 68 / 132 51.5 (43.0 60.0) 81.9 (73.6 90.2) 37.3 (27.9 46.6) CIN3: Cervical intraepithelial neoplasia grade 3 or worse; CIN2: cervical intraepithelial neoplasia grade 2 or worse; CI: confidence interval; PPV: positive predictive value; NPV: negative predictive value; n1: number of test positive disease cases; N1: total number of disease cases; n2: number of test negative non-disease cases; N2: total number of non-disease cases. Specificity (95%CI) (%) NPV (95%CI) (%) PPV (95%CI) (%)

970 R. LUTTMER ET AL. young women (<35 years) and older women with a child wish this approach might be useful to prevent overtreatment and unnecessary harm of the cervix [51]. 4. Five-year view Figure 1. Concept of the complementary value of host DNA methylation analysis and HPV16/18 genotyping in the diagnosis of transforming CIN2/3 and cervical cancer. It is expected that within 5 years, several countries will have implemented hrhpv-based cervical cancer screening. Although Pap cytology (with or without additional HPV16/18 genotyping) is likely to remain the triage strategy of choice in the coming years, more longitudinal studies will probably confirm the safety of alternative promising (partly) molecular strategies such as p16/ki-67 dual-stained cytology or CADM1/MAL or FAM19A4 (/mir124-2) methylation analysis. CADM1/MAL or FAM19A4 (/mir124-2) methylation, possibly in combination with HPV16/ 18 genotyping, could reduce unnecessary referrals and treatment through the specific detection of advanced CIN3 and cervical carcinomas among hrhpv-positive women. In addition, the applicability of these markers on self-sampled material could greatly improve participation rates in cervical cancer screening. the screening hrhpv-test and the secondary triage test should be applicable to self-sampled material. This would enable direct referral to colposcopy in triage-test-positive women, without the necessity of an additional physiciantaken scrape. Therefore, host cell DNA methylation analysis could serve as a more appealing triage strategy. In both cervical scrapes and lavages the NPVs of CADM1/ MAL and FAM19A4(/mir124-2) methylation analysis (86.3 95.3%) are below the suggested 98% threshold [38]. However, given that methylation analysis of CADM1/MAL and FAM19A4(/mir124-2) has a very high sensitivity for carcinomas and CIN3 with a long-term previous hrhpv-infection [95,96], we argue that methylation-negative CIN3 lesions have a low short-term risk of progression to cancer. From this point of view, one might consider the use of (CADM1/MAL and FAM19A4(/mir124-2)) methylation analysis for triage of hrhpvpositive women accepting a somewhat lower NPV, thereby allowing CIN2/3 with low short-term progression risk to remain undetected. However, with the currently available data, detection of the majority of CIN3 and CIN2 lesions remains desirable in clinical practice. For this purpose, we suggest the combined use of (CADM1/MAL or FAM19A4(/mir124-2)) methylation analysis and HPV16/18 genotyping, as these markers are complementary in the detection of cancer and advanced CIN2/3 (by methylation analysis) and early CIN2/3 lesions (by HPV16/18 genotyping) (Figure 1). In addition, the use of this triage combination could provide clinical guidance to gynecologists in the management of CIN2/3 lesions. Given the high specificity of (CADM1/MAL or FAM19A4/mir124-2) methylation analysis for advanced CIN, we propose that a methylation-positive result in a woman with CIN2/3 justifies treatment by large loop excision of the transformation zone (LLETZ). Conversely, we propose that a methylation-negative result could be an extra reason to perform regular follow-up of a CIN2/3 lesion instead of a primary LLETZ, at the discretion of the gynecologist. Especially in Key issues HPV-based cervical cancer screening will replace cytologybased screening in several countries in the near future. Women who test HPV-positive require a secondary (triage) test to prevent over-referral and over-treatment Pap cytology has been validated as a safe triage strategy, however with the necessity of repeat testing, leading to loss to follow-up. Also, knowledge of the HPV status of women influences the Pap cytology result Combined Pap cytology and HPV16/18 genotyping has been validated as a safe and direct triage strategy The performance of Pap cytology varies considerably between laboratories, populations and countries. Molecular tests will considerably improve reliability of cervical cancer screening p16/ki-67 dual-stained cytology could serve as a more reproducible and specific alternative to Pap cytology, possibly enabling direct triage CADM1/MAL or FAM19A4(/mir124-2) methylation analysis, preferably with but also without additional HPV16/18 genotyping, could serve as fully molecular triage strategy In contrast to other triage markers, CADM1/MAL or FAM19A4 (/mir124-2) methylation is especially sensitive for advanced CIN3 lesions and cancer, and the latter marker combination has the advantage of applicability to self-collected samples Longitudinal evaluation of the negative predictive value of p16/ki-67 dual stained cytology and CADM1/MAL or FAM19A4(/mir124-2) methylation analysis is required to determine the optimal re-evaluation interval for hrhpvpositive women who test negative for these triage assays. Funding This work was supported by the SME Instrument in the Horizon 2020 Work Programme of the European Commission (Valid-screen 666800) and FP7 ERC-ADV grant Mass-care (322986).