An integrated prediction model of recurrence in endometrial endometrioid cancers

Marina D Miller,1 Erin A Salinas,1 Andreea M Newtson,1 Deepti Sharma,1 Matthew E Keeney,2 Akshaya Warrier,1 Brian J Smith,3 David P Bender,1,4 Michael J Goodheart,1,4 Kristina W Thiel,1 Eric J Devor,1,4 Kimberly K Leslie,1,4 Jesus Gonzalez-Bosquet1,41Department of Obstetrics and Gynecology, University of Iowa Carver College of Medicine, Iowa City, IA, USA; 2Department of Pathology, University of Iowa Carver College of Medicine, Iowa City, IA, USA; 3Department of Biostatistics, University of Iowa College of Public Health, Iowa City, IA, USA; 4Holden Comprehensive Cancer Center, University of Iowa Carver College of Medicine, Iowa City, IA, USAObjectives: Endometrial cancer incidence and mortality are rising in the US. Disease recurrence has been shown to have a significant impact on mortality. However, to date, there are no accurate and validated prediction models that would discriminate which individual patients are likely to recur. Reliably predicting recurrence would be of benefit for treatment decisions following surgery. We present an integrated model constructed with comprehensive clinical, pathological and molecular features designed to discriminate risk of recurrence for patients with endometrioid endometrial adenocarcinoma.Subjects and methods: A cohort of endometrioid endometrial cancer patients treated at our institution was assembled. Clinical characteristics were extracted from patient charts. Primary tumors from these patients were obtained and total tissue RNA extracted for RNA sequencing. A prediction model was designed containing both clinical characteristics and molecular profiling of the tumors. The same analysis was carried out with data derived from The Cancer Genome Atlas for replication and external validation.Results: Prediction models derived from our institutional data predicted recurrence with high accuracy as evidenced by areas under the curve approaching 1. Similar trends were observed in the analysis of TCGA data. Further, a scoring system for risk of recurrence was devised that showed specificities as high as 81% and negative predictive value as high as 90%. Lastly, we identify specific molecular characteristics of patient tumors that may contribute to the process of disease recurrence.Conclusion: By constructing a comprehensive model, we are able to reliably predict recurrence in endometrioid endometrial cancer. We devised a clinically useful scoring system and thresholds to discriminate risk of recurrence. Finally, the data presented here open a window to understanding the mechanisms of recurrence in endometrial cancer.Keywords: endometrial cancer, recurrence risk, prediction model