Personalized treatment algorithm and prognostic model for the risk of recurrence of endometrial hyperplasia in women of late reproductive age

Abstract

DOI: 10.52705/2788-6190-2026-01-5
УДК 618.14-007.61-06:616-008.9-08-039.76

Recurrent endometrial hyperplasia (REH) in women of late reproductive age (36–45 years) is characterized by a high rate of repeated episodes, risk of progression, and limited effectiveness of standard progestin therapy. Hormonal and metabolic disturbances, particularly insulin resistance, obesity, and progesterone deficiency, are considered key factors contributing to a recurrent course of the disease.
The objective: to develop and clinically implement a personalized treatment algorithm for recurrent endometrial hyperplasia based on hormonal and metabolic profiling, and to create a validated prognostic model for recurrence risk.
Materials and methods. A prospective cohort study (2022–2025) included 120 women aged 36–45 years. The main group (n=60) consisted of patients with recurrent non-atypical hyperplasia (≥ 2 recurrences within 24 months), the comparison group (n = 30) included women with newly diagnosed hyperplasia, and the control group (n = 30) comprised practically healthy women. Within the main group, two subgroups were identified: standard therapy (n = 30) and a personalized algorithm with metabolic correction (n = 30). The follow-up period was 24 months. The primary endpoint was recurrence. Multivariate logistic regression, ROC analysis, internal validation using repeated resampling (1000 iterations), calibration assessment, and reclassification indices were applied. Results. The 24-month recurrence rate in the REH group was 38.3 % compared to 16.7 % in primary hyperplasia (p < 0.01). Independent predictors of recurrence were identified as HOMAIR > 3.5 (OR = 2.84), BMI ≥ 30 kg/m2 (OR = 2.31), progesterone < 6 nmol/L (OR = 2.58), and triglycerides ≥ 1.7 mmol/L (OR = 1.96). The prognostic model demonstrated high discriminative ability (AUC = 0.84; after internal validation – 0.82) and adequate calibration (p = 0.62). The addition of metabolic variables improved model performance (ΔAUC = +0.09). In the personalized treatment subgroup, the recurrence rate decreased to 16.7% compared to 38.3% with standard therapy (RR = 0.44; ARR = 21.6 %; NNT = 4.6).
Conclusions. Recurrent endometrial hyperplasia has a systemic hormonal–metabolic nature. Integration of insulin resistance assessment and other metabolic parameters into the treatment algorithm significantly reduces the risk of recurrence. The developed prognostic model is statistically robust and suitable for clinical risk stratification.