Determination of risk factors for the occurrence of recurrences, metastases, and somatic complications after treatment of locally advanced cervical cancer using modern information technologies

Background. According to current data, the average incidence rate of cervical cancer is 9.0 cases per 100,000 people in developed countries and 17.8 cases per 100,000 people in developing countries. At the same time, the percentage of advanced stages (III–IV) of the disease remains high. Continuous improvement of radiation therapy methods has made it possible to find optimal radiation doses, in particular, for areas of regional metastasis, to reduce the risks of damage to surrounding organs and healthy tissues. A parallel search for optimal combinations of cytostatics and modes of injection leads to increase in survival rates for cervical cancer patients. Unfortunately, in Ukraine, the opposite trend is currently observed: amid active warfare, delayed medical consultation, and the forced intensification of treatment processes, the incidence of reproductive organ tumors in women has sharply increased, particularly among young women under 30, which makes the need to improve methods to combat this disease even more pressing. One possible strategy for acquiring new knowledge could be analyzing already catamnestic data on the course and outcomes of the disease using more advanced mathematical methods, including hidden knowledge discovery technology (Data Mining) and mathematical statistics. Рurpose – identify risk factors for long-term oncological outcomes, immediate radiation complications, and distant somatic complications, as well as methods for their prevention and mitigation, in order to improve individualized patient management programs for locally advanced cervical cancer. Materials and methods. The analyzed sample was based on follow-up data from medical histories of patients with locally advanced cervical cancer treated at the State Organization «Grigoriev Institute for Medical Radiology and Oncology of the National Academy of Medical Sciences of Ukraine» from 1998 to 2021. Selection criteria included medical histories with a follow-up duration of more than one year or documentation of disease manifestation within the initial (up to one year) period from the start of treatment. Complete observations included fatal outcomes, recurrences, and metastasis development. The selected medical histories were entered into an electronic database created according to an infological model developed at the start of the study. For mathematical analysis of the database content, the WizWhy package in the Data Mining category and nonparametric statistical methods (STATISTICA package) were used, including the Mann–Whitney test (MWT) and Pearsonʼs chi-square test (PCT). Descriptive statistics were provided as either the median or Me (LQ; UQ), where Me is the median, LQ is the lower quartile, and UQ is the upper quartile.In terms of the level of evidence, the study is a retrospective case-control study. Results. It was established that the risk factors for the appearance of cancer-specific long-term consequences of treatment in the future are the localization of the infiltrate, the characteristics of the cervical volume, the level of tumor specificity of the SCC marker, the level of platelets at the beginning of treatment and the dynamics of their decrease. Early radiation complications of treatment were analyzed and prognostic factors for the appearance of these complications were found, somatic disorders that appeared in the long term after the treatment were analyzed. Conclusions. The application of Data Mining technology and methods of mathematical statistics on the digitized array of catamnesis data of patients with cervical cancer made it possible to identify a number of non-trivial dependencies that are very valuable for preventing the development of remote cancer effects and improving treatment algorithms.