2022-RA-801-ESGO Challenge to individualize surgical treatment for uterine cancer by intraoperative-prediction of lymph node metastasis using mRNA biomarker and clinical variables

Introduction/Background The significance of lymphadenectomy in uterine cancer has not yet been completely established. However, excessive lymphadenectomy significantly reduces the postoperative quality of life. Hence, noninvasive and highly accurate lymph node metastasis (LNM) diagnostic methods are strongly needed as alternatives to dissection. Therefore, we attempted a novel approach for the LNM diagnosis, in which the probability of metastasis is calculated using preoperative clinical variables and biomarker measurements. Methodology Preoperative clinical variables included serum tumor marker values and magnetic resonance imaging findings etc. Each variable’s discrimination power was evaluated by univariate analysis and validated combination of variables that contributed most to the LNM discrimination.The most promising mRNA biomarkers that correlate with expression difference between -positive and -negative groups were identified by CAGE (Cap Analysis Gene Expression), a genome-wide analysis. Results Ten clinical variables that contributed most to the LNM discrimination were extracted. Two promising biomarkers, SEMA3D and Novel isoform of TACC2, and two companion markers were identified. For all uterine cancers, the calculated predictive probability values were significantly different between the LNM-positive and -negative groups (P = 1.39 × 10-10), and high diagnostic accuracy of 83.6% area under the curve (AUC) was obtained. The LNM diagnosis requires essentially minimize the time difference between the diagnosis and hysterectomy. Therefore, reverse transcription-polymerase chain reaction enabled quantification from RNA in one step within 30 min, for intraoperative diagnosis. Conclusion This diagnostic method uses rapid nucleic acid amplification for intraoperative quantification of biomarkers in the primary tissue. Furthermore, the predictive model combined with various clinical variables can be used to discriminate LNM with high accuracy and facilitate individualization of the surgical treatment.