Background: Gastric cancer is typically diagnosed at a late stage, leading to poor prognoses. Helicobacter pylori is responsible for 70% of gastric cancers globally, and patients with this bacterial infection often present with early stages of the carcinogenic pathway such as inflammation or gastritis. Although many patients continue to progress to advanced-stage disease after antibacterial treatment, there are no follow-up screening protocols for patients with a history of H. pylori.
Methods: Several biomarkers (Lgr5, CD133, CD44) become upregulated during gastric carcinogenesis. A logistic regression model is developed using clinical data from 59 patients at different stages of the carcinogenic pathway to identify the likelihood of being at an advanced stage of disease for all combinations of age, sex, and marker positivity. Using these likelihood distributions and the observed rate of marker positivity increase, time to high likelihood (probability >0.8) of advanced disease for individual patients is predicted.
Results: A strong correlation between marker positivity and disease stage was found for all three markers. Disease stage was accurately classified by the respective regression models for more than 86% of retrospective patients. Highly patient-specific predictions of time to onset of dysplasia were made, allowing the classification of 17 patients initially diagnosed with intestinal metaplasia into high-, intermediate-, or low-risk categories.
Conclusions: We present an approach designed to integrate pathology, mathematics, and statistics for detection of the earliest precancerous, treatable lesion. Given the simplicity and robustness of the framework, such technique has the potential to guide personalized screening schedules to minimize the risk of undetected malignant transformation.
Keywords: Early detection of cancer; Helicobacter pylori; Stomach neoplasms.