General discussion and future perspectives 175 8 analysed 15 methylation markers, consisting of ANKRD18CP, ASCL1, C13orf18, EPB41L3, GHSR, JAM3, LHX8, PAX1, POU4F3, SOX1, SST, ST6GALNAC5, ZIC1, ZNF582 and ZSCAN1, in a training-test-set approach, and reported ASCL1 and LHX8 among the markers with the highest AUC for CIN3+ (AUC of ≥ 0.7) 71. A methylation marker panel was constructed consisting of ANKRD18CP, LHX8 and EPB41L3 which was most discriminative for CIN3+ detection 71 with a CIN3+ test-set sensitivity of 84% and specificity of 71%. These results need further confirmation in independent validation series. To date, most other studies on methylation analysis using self-collected samples have primarily focused on under-/ never-screened women and referral populations. These studies reported favourable triage performance for identifying CIN3+ of DNA methylation analysis of CADM1, MAL, miR124-2, JAM3, EPB41L3, C13orf18, ZNF582, PAX1, SOX1, ASTN1, DLX1, ITGA4, RXFP3, SOX17, ZFN671, ASCL1, LHX8 and ST6GALNAC5 7, 8, 40, 70, 72-74. Our study (Chapter 4) showed relatively low clinical performance for detecting CIN3+ compared to these other studies. A possible explanation is that the performance of new methylation tests may be overly optimistic when the population used for the development or validation of the test is not representative of the target screening population. In particular, under-/never-screened women and referral populations are not representative of women who are invited to regular screening. The incidence of CIN3 in under/never-screened women is higher than in a population of screening attendees 75 and CIN3 lesions identified in a regular screening programme have a relatively short duration since onset, are relatively small, and are known to have lower methylation levels 8, 26, 46, 64. It is worth noting that both our study (Chapter 4) and the study of De Waard et al. 71 took place within a screening setting, which aligns more closely with the target screening population. De Waard’s study however included a selection of specific samples, which could have had an influence on the test’s performance. Chapter 4 addressed the practical application of a methylation test in real-life screening scenarios. An accepted criterion for acceptability of a triage strategy is an NPV for end-point CIN3+ of at least 98% (i.e., a residual CIN3+ risk of 2%) and a positive predictive value (PPV) of at least 20% (The Netherlands) 48, 76 or 5-10% (USA) 54. In Chapter 3 and Chapter 4, the PPV of the ASCL1/LHX8 test was > 20%, but the NPV was below 98% or, equivalently, the CIN3+ risk was above 2%. A study by Vink et al. reported PPV of 33% (95% CI 27.4-38.5%) and NPV of 94.8% (95% CI 93.1-96.4%) for another methylation marker panel, i.e., FAM19A4/miR124-2 methylation 77. Together these findings indicate that follow-up is required after a negative methylation test. An approach that was considered in this thesis for increasing the NPV is adding HPV16/18 genotyping. Although the combination of HPV16/18 genotyping with
RkJQdWJsaXNoZXIy MjY0ODMw