Abstract
BACKGROUND: While benign mammary hyperplasia frequently undergoes spontaneous regression, clinicians currently lack validated serological biomarkers for personalized surveillance strategies. OBJECTIVE: To evaluate whether the serum prolactin-to-estradiol ratio [PER; prolactin (ng/mL) ÷ estradiol (pg/mL)] can predict radiological regression in benign hyperplasia. METHODS: This retrospective cohort study (January 2020-December 2024) enrolled women (18-55 years) with biopsy-confirmed ductal or lobular hyperplasia. Baseline fasting prolactin and estradiol were measured using duplicate electrochemiluminescence immunoassays (WHO-traceable) from routine clinical samples. Follow-up biopsies were performed only when imaging triggers were met. Multivariable logistic regression and interval-censored Cox models assessed associations with demographic, reproductive, and lesion covariates. Performance was evaluated using C-statistics, calibration, ROC analysis, and decision curves. RESULTS: Among 1,645 participants completing follow-up (94.7%), 790 (45.5%) demonstrated radiological regression. Patients with regression had significantly lower mean PER compared to non-regressors (0.161 ± 0.086 vs. 0.232 ± 0.136; p < 0.001), reflecting a composite hormonal environment of reduced prolactin and relatively elevated estradiol. PER demonstrated a strong inverse dose-response relationship with regression probability (adjusted OR per 0.1-unit increase = 0.15; 95% CI: 0.10-0.22) and time-to-regression (adjusted HR per 0.1-unit decrease = 1.76; 95% CI: 1.54-2.01), with regression rates declining from 60.3% in the lowest PER tertile to 28.0% in the highest. At the optimal cutoff of PER ≤0.185, discriminative performance reached an AUC of 0.664, significantly outperforming clinical variables alone (AUC 0.529), with net clinical benefit confirmed across a broad range of decision thresholds. Predictive effects remained consistent across age, menstrual phase, and histological subtype (all interaction p > 0.08), with a modest but significant BMI interaction (p = 0.042). CONCLUSION: Low PER independently and reliably predicts spontaneous regression of benign breast hyperplasia. External validation studies and point-of-care assay development are needed before clinical implementation.