Beschreibung:
<jats:title>Abstract</jats:title>
<jats:p>Gastric cancer (GC) is often diagnosed at an advanced stage and consequently remains the third most common cause of cancer-related death worldwide. Early detection and surgical intervention has been found to reduce GC-associated mortality, yet an efficient and cost-effective screening program still does not exist.</jats:p>
<jats:p>Several proteins including Lgr5, CD44 and CD133 have been found to be upregulated during the H. pylori-associated Correa pathway of carcinogenesis, and have the potential to act as biomarkers for early detection. The tools of mathematical oncology facilitate the development of models capable of both reproducing and predicting the increase in expression of these markers, and thus the patient-specific progression of disease. These predictions can allow the identification of optimal screening times to minimize the risk of undetected malignant transformation. Here, we present a combined mathematical-statistical approach to identifying such optimal screening times in a manner which is both biomarker-independent and H.pylori status-independent, thus allowing improved generalizability and the potential for broader application in the clinical setting. Logistic regression models are developed for each biomarker of interest (Lgr5, CD133, CD44) to determine the likelihood of a patient being either early in the Correa pathway (gastritis or metaplasia) or late in the pathway (dysplasia or carcinoma) based on current age, sex and biomarker-positive cell fraction obtained from immunohistochemical staining of biopsy tissue samples. Calibrated models for all three biomarkers were able to accurately classify disease stage as determined by pathology report in more than 85% of an initial cohort of 59 patients.</jats:p>
<jats:p>Mathematical models describing the temporal evolution of the expression of each marker of interest are then defined based on clinical data. Coupled with the statistical tool for identifying the likelihood of being at each stage of disease for all combination of input parameters, this framework can forecast time to a clinically-significant endpoint such as the development of dysplasia or adenocarcinoma for future patients, allowing the suggestion of screening times at which the likelihood of progression reaches a certain threshold. The results suggest that for 13 out of 16 patients in our initial test cohort (81%), models independently developed using three different markers all recommend similar follow-up screening times, showing promise for the general applicability and robustness of such a tool.</jats:p>
<jats:p>Citation Format: Heiko Enderling, Rachel Walker, Jose Pimiento, Jaime Mejia, Domenico Coppola. Computational modeling to suggest patient-specific screening schedules for early detection of gastric cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 4544. doi:10.1158/1538-7445.AM2017-4544</jats:p>