57 DNA methylation markers for endometrial cancer Sensitivity and specificity could be extracted from all studies reporting on marker panels, of which two (26, 30) also computed a 95% CI of the reported sensitivity and specificity values. Table 1: Characteristics of the included studies. Study (year) Cases (n) / controls (n) Tumor type (n) Cytological sample type(s) Methylation analysis Ref. BakkumGamez (2015) 38† / 37 E (31), CAH (1), CC (3), PS (3) Endometrial brushes, vaginal tampons Pyrosequencing (25) Chang (2018) 30 / 30 E (30) Cervical scrapes qMSP (16) Doufekas (2013) 41† / 38† E (38), CAH (1), U (2) Vaginal swabs MethyLight PCR (28) Huang (2017) 50 / 56 E (50) Cervical scrapes qMSP (24) Jones (2013) 18/13†‡ / 17† E (18/13) Vaginal swabs MethyLight PCR (29) Liew (2019) 46† / 38† E (33), PS (6), O (7) Cervical scrapes qMSP (13) Sangtani (2020) 38 / 27 E (31), CAH (1) CC (3), PS (3) Vaginal tampons Pyrosequencing (26) De Strooper (2014) 21 / 120 U Cervical scrapes qMSP (30) Wentzensen (2014) 37 / 37 E (30), CC (2), PS (2), M (3) Endometrial brushes Pyrosequencing (27) † Women presented with postmenopausal (28, 29) or abnormal bleeding (13, 25). ‡ Stage Ia / Stage Ib/II/III. CAH = complex atypical hyperplasia; CC = clear-cell carcinoma; E = endometrioid adenocarcinoma; O = other; M = mixed; PCR = polymerase chain reaction; PS = papillary serous carcinoma; qMSP = quantitative methylation-specific PCR; U = adenocarcinoma, histotype unknown. Risk of bias of all included studies Selected studies underwent quality assessments according to the QUADAS-2 tool, of which the results are presented in Figure 2. In over half of the studies, a high risk of bias was introduced during patient selection, mainly due to their case-control design in which patients were specifically selected based on their confirmed histological status. Additionally, handling DNA methylation levels of the markers without pre-specified thresholds (e.g. based on the most optimal marker performance) introduced a high risk of bias in six of the included studies. In two studies, the risk of bias score for the reference test was unclear. Even though these studies did perform pathological examination of the included samples, they did not specify which cancer histotypes were found or what specimen type was used (e.g. use of a biopsy or a larger surgical specimen) during this examination. Likewise, the risk of bias introduced during the flow and timing were scored unclear when the reference test was not accurately described or 3
RkJQdWJsaXNoZXIy MjY0ODMw