Relevance. The relevance of the topic is determined by the fact that digital solutions that have demonstrated high accuracy on curated research datasets are increasingly being proposed for use in healthcare organizations, where data are shaped by the clinical practices of a particular institution.
Objective. To analyze factors related to real-world data (RWD) that may lead to a decline in the performance of medical artificial intelligence (AI) models when they are transferred to new settings, and to substantiate approaches to their preliminary assessment before implementation in a healthcare organization.
Materials and methods. A scoping review was conducted to systematize publications on the applicability of medical AI models beyond the original dataset. The search was performed in Scopus, Web of Science Core Collection, PubMed, Google Scholar and eLIBRARY. RU for the period from 2016 to 2026; earlier sources were inc luded when necessary.
Results. The review identified the main types of medical data, causes of model performance degradation and approaches to assessing model applicability in a new clinical setting. Particular attention was paid to data shift as a cause of reduced accuracy, impaired calibration and an increased number of clinically significant errors. Approaches to external and local validation, assessment of model robustness in subgroups and post-implementation monitoring were considered. Practical requirements were also proposed for describing model applicability in a scientific publication or validation report.
Conclusion. The ability of a medical AI system to maintain its effectiveness when used outside the original environment should be assessed as a separate criterion of its reliability. Validation reports should provide a detailed and reproducible description of data sources, outcome definitions, testing conditions and results obtained in a specific healthcare organization, as well as a justified strategy for subsequent monitoring and maintenance of system performance.

Relevance. Discriminant analysis (from the Latin discriminatio — “distinction”) allows objects to be classified according to a threshold value (the discrimination constant). The article presents a clinical example of predicting the accuracy of a system of predictor variables using a discriminant function analysis model.
Objective. To substantiate the feasibility of using discriminant analysis as a statistical tool for monitoring health and assessing work capacity among workers in radiation-hazardous industries.
Materials and methods. In a sample of 232 individuals with verified toxic dust bronchitis and berylliosis, predictors were studied using linear discriminant analysis: fibrinogen level, forced vital capacity (FVC), lactate dehydrogenase (LDH) level, and the time required for an electrical impulse to travel from the sinoatrial node to the ventricles on the electrocardiogram (PQ interval). In 124 patients, the predictor values were outside the reference ranges (the “patients” group); in 108 individuals, the results of laboratory and instrumental tests were within the normal range (the “non-patients” group). Additionally, to test the working hypothesis, 23 clinical cases without a specified category (N=23) were considered: “pathology”/“ normal”. To determine which group a new patient belongs to, the values of the discriminant functions for new objects (patients) were calculated using the formula Vnew = New × V, applying the input data (test results) with the help of the discrimination score vector. The obtained result was compared with the constant.
Results. The discrimination constant obtained through calculations enabled the classification of the tested individuals (N=23) into the “patients” or “normal” categories with an accuracy of 91.3 %.
Conclusions. It is advisable to use discriminant function analysis in healthcare and clinical practice to address problems in clearly defined groups with pre‑selected parameters. Its application is also justified in combination with other statistical methods, as it allows for the effective classification of patients’ diagnoses. The limitations of the method include the requirement for a normal distribution of the analyzed sample and the complexity of interpreting the results in the selected model when working with new data.

Background. Mathematical modeling and computer simulation techniques, including artificial intelligence (AI), are increasingly being used throughout the entire life cycle of medicinal products — from preclinical research to post‑marketing surveillance. However, the legislation of the Eurasian Economic Union (EAEU) lacks unified definitions and systematic regulatory approaches for these tools, which hinders their full integration into regulatory practice.
Objective. To perform a comparative analysis of approaches to defining and formalizing the terms “mathematical modeling”, “modeling and simulation”, and “artificial intelligence” in the law of the EAEU and Russian Federation; to develop proposals for improving the EAEU regulatory framework in this field.
Methods. A content analysis was conducted on the current normative legal acts of the EAEU and the Russian Federation, as well as ICH guidelines (M15, E11A, M12, E14/S7B, etc.), EMA documents (PBPK guidance, population pharmacokinetics, AI discussion papers) and FDA guidance (MIDD, PBPK, AI in manufacturing). Based on the identified definitions and concepts, new definitions were proposed for implementation into EAEU law.
Results. It was found that the terms “mathematical model” and “mathematical modeling” are used fragmentarily in EAEU acts, without clear definitions, mainly in the context of specific methodological recommendations (bioequivalence, stability, impurities). The definition of “artificial intelligence” in the EAEU and the Russian Federation is unified. In international guidelines (ICH M15), the concept of Model‑Informed Drug Development (MIDD) has been introduced, where modeling and simulation (M&S) are considered as a system for generating evidence, and AI is considered as one of the M&S approaches. EMA and FDA also do not provide a single definition of mathematical modeling, but they develop specialized guidelines and programs (MIDD paired meetings, qualification procedures). It is proposed to introduce into Decision No. 78 of the EEC Council the definitions of “mathematical modeling and computer simulation methods” and “evidence obtained using these methods”, with an emphasis on verification, validation, and reproducibility of results.
Conclusion. The introduction of the proposed definitions will create a legal basis for the systematic use of mathematical modeling and AI in pharmaceutical development and clinical practice in the EAEU, harmonize regulatory approaches with international standards, and accelerate the introduction of new drugs while maintaining high safety and efficacy requirements.

Background. Anovulatory infertility remains one of the leading causes of impaired fertility in women of reproductive age. The efficacy of in vitro fertilization (IVF) programs demonstrates interindividual variability, which necessitates the identification of reliable predictive markers to optimize personalized stimulation protocols. The role of pharmacogenetic variants of cytochrome P450 genes in predicting IVF outcomes in this condition remains insufficiently studied, which determines the relevance of this pilot study.
Objective. To evaluate the association of polymorphic variants of cytochrome P450 genes (CYP2A6, CYP2C9, CYP2C19, CYP2D6, CYP3A4, CYP3A5) with the achievement of clinical pregnancy in IVF programs in patients with anovulatory infertility.
Materials and methods. This prospective pilot cohort study included 96 patients divided into two groups: those who achieved pregnancy after their first IVF attempt (group 1, n = 48) and those with three or more failed cycles in their history (group 2, n = 48). Genotyping of 19 single‑nucleotide polymorphisms was performed using whole‑genome typing on Infinium Global Screening Array‑24 v3.0 biochips. Statistical analysis included the Pearson χ² test, Fisher's exact test, and the Monte Carlo permutation test (10,000 permutations).
Results. Statistically significant associations were identified between polymorphisms in CYP2A6 (rs8192733, rs56113850, rs57897628) and CYP2C19 (rs4244285) and IVF outcomes. For rs8192733, associations were found for the G/G genotype of CYP2A6 with an unsuccessful IVF cycle (OR = 0.105; 95 % CI 0.105–0.316; p < 0.001) and for the C/C and C/G genotypes with a successful procedure (OR = 4.795; 95 % CI 1.212–18.963; p = 0.031 and OR = 3.176; 95 % CI 1.202–8.395; p = 0.031, respectively). For the rs56113850 polymorphism in CYP2A6, the C/T genotype was associated with a favorable outcome (OR = 3.095; 95 % CI 1.284–7.458; p = 0.010), while the T/T genotype was associated with a negative outcome (OR = 0.286; 95 % CI 0.100–0.812; p = 0.015). The homozygous A/A genotype of the rs57897628 polymorphism in CYP2A6 was associated with a lower probability of pregnancy (OR = 0.256; 95 % CI 0.091–0.725; p = 0.005). For the CYP2C19 gene (rs4244285), the G/A genotype was associated with effective IVF (OR = 7.500; 95 % CI 2.016–27.901; p = 0.001).
Conclusion. According to the obtained results, CYP gene polymorphisms can serve as potential pharmacogenetic markers for predicting IVF effectiveness in patients with anovulatory infertility.

Background. Sixth-generation (6G) networks, currently being standardized by the International Telecommunication Union under the designation IMT-2030, are being designed not merely as data transmission media but as distributed sensing platforms. Their key capability — Integrated Sensing and Communication (ISAC) — enables the detection of individuals, the tracking of their movement, and the registration of chest‑wall micro‑movements induced by respiration and heartbeat, including through dielectric barriers and without any device carried by the observed subject. The feasibility of measuring heart rate (HR) through a wall, in both single‑subject and two‑subject scenarios simultaneously, has been experimentally confirmed.
Objective. To systematize the evidence base for ISAC‑based monitoring of vital signs, to formulate a catalog of ethical issues, and to propose a practical toolkit for ethics committees evaluating research proposals involving ISAC technologies, as well as to analyze cultural‑anthropological aspects of the perception of invisible monitoring in Russian society.
Methods. A narrative review of peer‑reviewed publications (PubMed/PMC, Frontiers, MDPI, arXiv), regulatory documents of the ITU‑R, the General Data Protection Regulation (GDPR), the EU Artificial Intelligence Act (EU AI Act), as well as the Russian Federal Law No. 152‑FZ of 27 July 2006 "On Personal Data" and official clarifications of Roskomnadzor; materials from research institutes (Barkhausen Institut, 6G Flagship, VTT) covering the period 2016–2026. Additionally, sociological and anthropological studies on the perception of privacy and surveillance in the post‑Soviet context were consulted. The ethical analysis was conducted within the framework of the principles of respect for autonomy, beneficence, non‑maleficence, and justice.
Results. Radar methods (CW, FMCW, UWB) and OFDM communication signals enable the registration of respiratory rate and HR at distances ranging from tens of centimeters to several meters, including behind physical barriers. The principal ethical barrier is the impossibility of obtaining informed consent from individuals who passively enter the sensing zone. Eleven ethical issues were identified (with an additional issue on cultural perception of invisible surveillance). A legal analysis was conducted: ISAC‑derived data qualify as both health data and biometric data within the meaning of Article 9 of the GDPR, while identification scenarios fall under the restrictions of Article 5 of the EU AI Act. Under Russian law, such data qualify as biometric personal data (Article 11 of Federal Law No. 152‑FZ) and special categories of personal data (Article 10 of Federal Law No. 152‑FZ), the processing of which requires written consent (Article 9 of Federal Law No. 152‑FZ), as confirmed by official clarifications of Roskomnadzor regarding remotely obtained physiological data. A 14‑item checklist for ethics committees was developed. It is shown that the cultural characteristics of Russians (a high level of trust in state institutions combined with a historical experience of surveillance) create a paradoxical situation: the technology may be accepted passively, but at the same time generate latent discontent and a decreas e in autonomy.
Conclusions. Prior to the commercial deployment of 6G (approximately 2030), a regulatory window exists for embedding privacy safeguards into the architecture of the standard. ISAC research involving human subjects requires adaptation of the requirements of the Declaration of Helsinki to a situation in which the object of observation is not only the research participant but also any individual within the coverage zone. Cultural features of privacy perception must be taken into account when developing transparency and consent mechanisms.

Background. Irrational use and improper disposal of medicines contribute to pharmaceutical pollution of the environment, highlighting the need for objective quantitative methods to assess the volume of unused drugs in clinically meaningful units.
Objective. To evaluate the applicability of the ATC/DDD pharmacoepidemiological methodology for the analysis of expired and unused medicines collected from the public for safe disposal.
Methods. From November 2022 to April 2024, a separate collection programme for unused and expired medicines was conducted in Kazan, Russian Federation. All collected medicines were classified according to the WHO ATC system, and DDD (Defined Daily Dose) values were assigned based on the WHO ATC/DDD Index 2024. Comparative analysis was performed using DDDs and mass of active substance (mg).
Results. A total of 4,737 medicine packs were collected, corresponding to 1,213 brand names and 707 INNs. The ATC/DDD analysis included 2,793 packs (299 INNs). The highest number of defined daily doses was found in group C (Cardiovascular system) — 11,377.34 DDDs, and group A (Alimentary tract and metabolism) — 11,263.31 DDDs. The top individual substances included cholecalciferol, cyanocobalamin, prednisolone, folic acid, acetylsalicylic acid, and atorvastatin. Mass-based assessment (mg) showed a different distribution and did not reflect the clinical relevance of unused medicines.
Conclusion. This is the first study to apply ATC/DDD analysis for quantitative assessment of unused and expired medicines collected from the public. The approach enables standardized evaluation of pharmacotherapy withdrawn from use, identifies drug classes most commonly wasted, and may serve as an additional tool for monitoring the rationality of drug prescribing and pharmaceutical environmental risk.

Health technology assessment (HTA) worldwide has entered a period of significant change. This was the theme of the 2026 HTAi Global Policy Forum. The current challenge is considered through the health technology trilemma, which brings together the financial sustainability of the health system, value-creating innovation and equity of access. What is new is not the growth in expenditure itself, which historically became one of the drivers of HTA, but the simultaneous intensification of processes that alter the balance between all three elements of the trilemma. Opportunities to increase health budgets are shrinking amid competition from other public priorities. Multi-million technologies for small patient groups are entering the market at the same time as interventions for common diseases with large target populations. Decisions are being made within shorter timeframes on the basis of less mature evidence, disparities in access persist, the cost of evidence generation is rising and political pressure is increasing. Under these conditions the focus shifts from whether the system is willing to pay to whether it is able to fund a technology at the required scale and within the required timeframe. Ability to pay does not replace cost-effectiveness and budget impact analysis; rather, it requires more precise definition of target patient groups, uptake volumes and rates, the distribution of expenditure over time and the consequences of uncertainty. Retaining previous approaches increases the risk of politicized and non-transparent decisions, court-enforced access, displacement of other types of care and loss of the role of HTA in evidence-based resource allocation. Eight key signals and conclusions of the forum are presented, aimed at preserving the legitimacy of HTA, supporting value-creating innovation, making priorities and opportunity costs explicit, addressing equity in practice and reducing low-value care. Russian examples demonstrate the limitations of current regulation in funding technologies with multiple indications and an evidence base that is still being formed.

Background. Oncology databases have become a foundation for developing models of tumor sensitivity to drug and radiation therapy; however, their clinical value is determined not by data volume but by the quality of the response phenotype, external validation, and evidence of treatment-effect modification.
Objective. To perform a critical translational synthesis of oncology databases and computational resources used to predict tumor radiosensitivity and chemosensitivity.
Materials and methods. A narrative critical review was conducted across five groups of resources: clinico-molecular atlases, pharmacogenomic panels, radiosensitivity databases, functional dependency maps, and clinical and infrastructure platforms. The response phenotype, reproducibility, external validation, transferability across model systems, risk of bias, and clinical interpretability were assessed.
Results. The greatest translational maturity is currently demonstrated by homologous recombination deficiency in the context of DNA-damaging therapy and genomic-adjusted radiation dose as a research predictor of radiotherapy benefit. Contemporary DepMap resources, clonogenic radiosensitivity databases, organoid models, radiomics, and federated learning expand validation opportunities but do not eliminate the gap between association and clinical utility.
Conclusions. Models should undergo cross-database, external clinical and, where feasible, prospective validation. Control of batch effects, cohort representativeness, and phenotype standardization are particularly important. The immediate practical role of oncology databases is to prioritize biomarkers and support study design rather than autonomously replace clinical decision-making.

Introduction. Harmonization of pharmacovigilance systems in candidate countries with European Union (EU) legislation is a prerequisite for integration and for ensuring medicinal product safety. The Republic of Albania, a candidate country since 2014, shows mixed results in this area.
Objective. To conduct a comparative analysis of the Albanian pharmacovigilance system against European standards in terms of legal framework, institutional structure, monitoring procedures and risk management, in order to identify key barriers to harmonization.
Materials and methods. The study is based on an analysis of Albanian Law No. 105/2014 and its draft amendments (2022), scientific publications (Hoxha et al., 2024; Shkreli et al., 2023; Roshi et al., 2021), audit reports from the Albanian High State Control (KLSH) for 2019 and 2024, and EUROSAI data. A comparative descriptive analysis was applied across four key criteria at both the de jure and de facto levels.
Results. A significant gap was found between formal legislation and practical implementation. The “Shkreli paradox” was identified: high declared reporting rates of adverse drug reactions (ADRs) among physicians (57.3 %) and pharmacists (58.6 %) contrast with the negligible number of official signals recorded by the regulatory authority (AKBPM). Institutional weakness of AKBPM was documented (lack of laboratory accreditation, staff shortages) along with information isolation from the EudraVigilance database. Patient awareness of drug therapy risks remains low (only 51 % knew about allergic reactions to NSAIDs). Positive signals include the agreement with the Italian Medicines Agency (AIFA) and participation in the PHARM‑EU project.
Conclusion. The Albanian pharmacovigilance system lags significantly behind European standards in operational parameters. Key barriers are the broken data transmission chain, institutional weakness of the regulator, and lack of access to EudraVigilance. To achieve genuine harmonization, priorities should include establishing a national electronic ADR reporting system, accrediting the AKBPM laboratory, and initiating negotiations with the EMA for observer status.