In previous studies we have shown that hypermethylation of CpG islands in proximity to the genes DLX1, ITGA4, RXFP3, SOX17, and ZNF671 correlated with the presence of cervical precancerous lesions and cervical cancer [12]. The molecular diagnostic test GynTect based on these results was designed and developed to allow detection of these marker regions by standard methods in cervical smears collected in the denaturing specimen transport medium (STM), which is originally used for QIAGEN’s Digene HPV test [13]. Utilization of this medium has, however, its limitations, the most important being that from STM only molecular tests but no cytology can be performed. Furthermore, DNA stability is limited in STM. In contrast, cervical smear material collected in liquid-based cytology media can be used more flexibly. As a main advantage, the cellular material preserved in this medium can be used for cytology as well as for molecular biology tests. This enables the performance of triage tests from the sample taken for the initial screening test, a feature which is increasingly demanded as a prerequisite for diagnostics [14].
In this study we evaluated whether GynTect is suitable for using residual material from liquid-based cytology samples and thus fulfils this prerequisite. Furthermore, we compared the results with cobas HPV testing including genotyping for HPV 16/18, and CINtec Plus data obtained from the same samples. The latter test is also discussed as triage test for patients with abnormal cytology or positive HPV tests.
GynTect might provide the possibility to test if a woman with an abnormal cytology finding in the Pap smear and/or a positive HPV test result, has a precancerous lesion that requires follow-up and treatment. For this purpose we used samples for which the cytology findings and, for all Pap-abnormal cases, the histopathology results were available.
GynTect showed a very good technical performance, since among all 280 PreservCyt samples only 4 (1.4%) and among the 352 SureThin samples, only 6 (1.7%) were tested invalid with GynTect.
The excellent performance of GynTect from LBC samples was directly visible in the Ct values obtained for the control DNA regions detected in the test system. In many samples less than 28 cycles were observed as Ct value for the control markers included in the test, due to very good DNA quality. Not too surprising, several of the samples with normal cytology showed Ct values also for the cancer marker regions, but in the range above 37 cycles. Therefore, in contrast to the GynTect analysis of STM samples [13], for using the samples out of LBC medium a threshold for the marker Ct values in relation to the controls was set. Using a delta Ct threshold to the ACHE control region of 9 for the five markers ASTN1, DLX1, ITGA4, RXFP3, and SOX17 and a delta Ct threshold of 10 for ZNF671, we achieved a very good specificity (< 98%) in the NILM group combined with an excellent sensitivity for cancer cases (100%) and a detection rate for the different precancerous stages CIN1, 2 and 3 increasing with grade. This confirmed the results we obtained with samples collected in STM medium [13].
All five cervical cancer samples included in the study were GynTect-positive, each of them for at least four of the six GynTect markers. The high sensitivity for cancer cases was already shown previously analysing 123 (123/123 detected) and 5 (5/5 detected) cases, respectively [12, 13]. A detection rate of > 60% among the CIN3 samples included in the study also confirms results obtained in previous studies [12, 13]. Furthermore, the data show that the GynTect score is related to the histopathologic severity of the lesion. In fact, the higher the CIN grade, the more GynTect markers are detected in the LBC samples. This may reflect the fact that methylation increases with the severity of the lesion.
It is well-known that not all CIN3 lesions proceed to cervical cancer [15, 16]. In several observational studies CIN2/3 short-term regression rates around 30% were reported [17,18,19,20]. Very recently, Tainio and colleagues in a meta-analysis summarized CIN2 progression and regression rates of more than 3000 women out of 36 studies [21]. Regression and progression rates ranging from 3 months to up to 60 months were analysed, whereas most data were available for 24 months observation. Overall, a 50% regression rate (11 studies including 1470 women) and an 18% progression rate (9 studies, 1445 women) was observed looking at a 24 months interval. In a subgroup analysis including 1069 women younger than 30 years, regression rates were 60% and only 11% showed progression to higher lesions [21]. One of these studies performed by Loopik et al. demonstrated that among the 211 women < 25 years included in the study, the long-term regression rate of CIN2 lesions was even 71%, whereas the overall progression rate in this study was similar to the other studies (16.6%) [22].
Taking these observations into account with respect to the GynTect-positive results obtained for the three CIN stages, one may hypothesize that only lesions from women tested GynTect-positive may progress to higher grade or cancer while lesions from women tested negative for GynTect are likely to regress to normal. Further studies will show whether regression of CIN lesions correlates with a negative GynTect result.
To compare specificity, sensitivity, PPV and NPV, CINtec Plus was performed for all cytology-abnormal samples plus 59 of the NILM samples. Within this subgroup of randomly chosen NILM cases the CINtec Plus specificity was comparable to GynTect. The cytologically and histopathologically abnormal samples, however, were all positive for the CINtec Plus test. Thus, the test would not allow a differentiation between lesions prone to progression and lesions which may persist or regress. This is well reflected by the PPV (CIN3+) for CINtec Plus with 69.2% compared to GynTect with 74.5%. Other studies using CINtec Plus as triage option showed similar results, even though not all had a 100% sensitivity for CIN1+ samples.
False-positive rates for CINtec Plus reported in the group with normal cytology are in the range of 27–55% [23, 24] and specificity for CIN3+ between 51.3–82.1% [6, 7], being highest in the PALMS trial (94.8%, [7]). CINtec Plus has the ability to specify a triage population out of cytologically abnormal women more precisely than HPV testing [6, 7]. But the GynTect methylation marker panel seems to be more specific with no loss of sensitivity for cancer cases. However, this study only included 5 cancer cases.
HPV genotyping was performed for all samples, and stratification for HPV16/18 positivity leads to a better specificity than HPV in total or CINtec Plus, but still has a lower specificity compared to GynTect for both, CIN2+ and CIN3+. NPV is slightly better for CIN2+ and similar for CIN3+ compared to GynTect. It has to be noted, however, that HPV-positively tested patients who are negative for HPV16/18 – as is the case for one of the five cancer cases included in the study – nonetheless may be detected in the triage following the HPV test, e.g. another cytology.
Cytology results of the 74 CIN lesions from this cohort are difficult to compare to GynTect, HPV or CINtec Plus results since all histologically confirmed CIN lesions have originally been found due to an abnormal cytology result. Therefore, we did only compare GynTect, CINtec Plus and cobas HPV results, because all these were second-line tests after the initial Pap testing in this study.
Methylation markers are discussed as a tool for triage in cervical cancer screening programs, since they have the potential of being more specific than other biomarkers such as p16/Ki-67 (CINtec Plus) or HPV testing, with excellent sensitivity for cervical cancer cases. A recent work from Ochs and colleagues [25] came to the conclusion, that about 50% of all conisations registered at the Hospital in Lucerne, Switzerland, between 2000 and 2014, were performed on women without serious precancerous lesion. This underlines the medical need for more specific diagnostics before referring to conisation.
Meijer et al. established four methylation markers (CADM1, MAL, FAM19A4, mit124–2) with FAM19A4 and mir124–2 being the most promising ones [26,27,28]. Comparing their clinical performance, sensitivity for cancer and CIN3 seem to be similar to the GynTect markers with 100% cancer detection and around 2/3 CIN3 detection (42.1–100% [27, 28] and 68.8% CIN3 detection [29]). In another study POU4F3 is described as a promising methylation marker, with clinical sensitivity and specificity for CIN3+ of 74 and 89%, respectively [30]. A DNA methylation marker combination PAX1 and ZNF582 is also discussed as a candidate biomarker for triaging suspicious cervical samples [31].
Regarding the specificity, especially for mir124–2 and FAM19A4 a comparison is difficult because no distinction between CIN1 and “no CIN” was made [27, 28]. The GynTect marker panel shows higher specificity for the NILM group with 1.5% detection (GynTect) compared to > 20% for PAX1/ZNF582 or 17.3 and 12.4% for PAX1 and ZNF582 alone [31] or 13% for the panel CADM1/MAL [32]. As well POU4F3 seems to be less specific (specificity CIN3+ 61.4%, with 82.7% sensitivity [33]), even though no data only for the NILM group were reported.
Our results demonstrate that the molecular diagnostic test GynTect based on methylation of the marker regions ASTN1, DLX1, ITGA4, RXFP3, SOX17, and ZNF671 has very good performance using liquid based cytology samples. On one hand, GynTect displayed superior specificity in inconspicuous samples, on the other hand, the test showed excellent sensitivity in detecting the relevant cancer cases. Nevertheless, sample size in this study was small and thus, the power of the study is limited. Further studies have to confirm the results shown in this article.