The new histological system for the diagnosis of adrenocortical cancer

Abstract

Introduction: Adrenocortical cancer (ACC) is a rare malignant tumor that originates in the adrenal cortex. Despite extensive molecular-genetic, pathomorphological, and clinical research, assessing the malignant potential of adrenal neoplasms in clinical practice remains a daunting task in histological diagnosis. Although the Weiss score is the most prevalent method for diagnosing ACC, its limitations necessitate additional algorithms for specific histological variants. Unequal diagnostic value, subjectivity in evaluation, and interpretation challenges contribute to a gray zone where the reliable assessment of a tumor’s malignant potential is unattainable. In this study, we introduce a universal mathematical model for the differential diagnosis of all morphological types of ACC in adults. Methods: This model was developed by analyzing a retrospective sample of data from 143 patients who underwent histological and immunohistochemical examinations of surgically removed adrenal neoplasms. Statistical analysis was carried out on Python 3.1 in the Google Colab environment. The cutting point was chosen according to Youden’s index. Scikit-learn 1.0.2 was used for building the multidimensional model for Python. Logistical regression analysis was executed with L1-regularization, which is an effective method for extracting the most significant features of the model. Results: The new system we have developed is a diagnostically meaningful set of indicators that takes into account a smaller number of criteria from the currently used Weiss scale. To validate the obtained model, we divided the initial sample set into training and test sets in a 9:1 ratio, respectively. The diagnostic algorithm is highly accurate [overall accuracy 100% (95% CI: 96%-100%)]. Discussion: Our method involves determining eight diagnostically significant indicators that enable the calculation of ACC development probability using specified formulas. This approach may potentially enhance diagnostic precision and facilitate improved clinical outcomes in ACC management.