Byung Joo Chae

Objective: Accurate prediction of breast cancer (BC) recurrence after curative treatment is essential for optimizing surveillance strategies and personalized patient management. We aimed to prospectively validate a previously developed artificial intelligence (AI)-based breast cancer recurrence prediction model in a real-world clinical setting.Methods: This prospective real-world validation study included BC patients who underwent curative surgery followed by adjuvant therapy at a tertiary cancer center. The AI model integrated clinicopathologic variables and longitudinal follow-up data, including laboratory and imaging findings, to estimate individualized 2-year recurrence risk. Patients were stratified into low- and high-risk groups based on a predefined risk threshold. Model performance was evaluated using recurrence outcomes during follow-up.Results: Of 829 patients initially identified, 400 were included in the final analysis after applying the eligibility criteria. The model classified 391 patients (97.8%) as low risk and 9 (2.2%) as high risk. During the predefined two-year follow-up period, 7 patients (1.8%) developed recurrence, including 6 in the low-risk group (1.5%) and 1 in the high-risk group (11.1%). The model demonstrated excellent specificity (98.0%) and negative predictive value (98.5%) for identifying patients at low risk of recurrence.Conclusion: Prospective real-world validation demonstrated that the AI-based prediction model effectively identified patients at low risk of breast cancer recurrence. These findings support the clinical feasibility of AI-assisted recurrence risk stratification for personalized post-treatment surveillance, although further external validation in larger multi-center cohorts is warranted.

Background and purposeRecent trials have demonstrated the safety of sentinel lymph node biopsy (SLNB) omission in selected patients with early breast cancer. However, criteria for identifying patients suitable for partial breast irradiation (PBI) without surgical axillary evaluation remain undefined. This study aimed to identify the candidates for PBI in pT1 breast cancer patients.Materials and methodsA retrospective analysis was conducted on 5,097 patients with pT1 breast cancer who underwent breast surgery with axillary evaluation between 2016 and 2020. All patients had clinically node-negative disease (cN0) confirmed by preoperative axillary ultrasound. Univariable and multivariable logistic regression analyses were used to identify independent predictors of axillary lymph node metastasis (ALNM).ResultsThe overall ALNM rate was 13.9% (708/5,097). The mean age was 51.3 years. Multivariable analysis identified lymphovascular invasion (LVI) as the strongest independent predictor (odds ratio [OR] 5.57, 95% confidence interval [CI] 4.61–6.74, p < 0.001), and LVI-positive patients (14.3%) had a 42.1% ALNM rate. pT1c stage was also a significant independent predictor (OR 2.13, 95% CI 1.71–2.65, p < 0.001). A simple classification using LVI and tumor size stratified patients into three groups: low-risk (LVI-negative, tumor ≤1.0 cm; 40.1% of patients), intermediate-risk (LVI-negative, tumor >1.0 cm; 45.7%), and high-risk (LVI-positive; 14.3%). The corresponding ALNM rate was 4.9%, 13.0%, and 42.1%, respectively.ConclusionsPatients with pT1a-b tumors without LVI represent a low-risk population with an axillary metastasis rate below 5%. These findings suggest the feasibility of PBI in this low-risk subgroup without SLNB, warranting validation through prospective randomized trials.

BackgroundEndocrine therapy (ETx) for hormone receptor–positive breast cancer frequently induces menopausal symptoms that may compromise treatment adherence. However, longitudinal symptom trajectories compared with natural menopausal progression remain unclear.MethodsThis longitudinal cohort study included 5,734 women from two prospective cohorts. Menopausal symptoms were assessed using the Menopause Rating Scale (MRS). Women with breast cancer (n=2,583) were recruited from the Breast Cancer Information Grand Round for Survivorship cohort, and women without breast cancer (n=3,151) from the Kangbuk Samsung Health Study. Linear mixed-effects models estimated longitudinal changes in MRS scores according to breast cancer status, menopausal status, and ETx use. Difference-in-differences (DiD) analyses used premenopausal women without breast cancer as the reference group.ResultsAmong premenopausal women, ETx was associated with greater increases in total MRS scores compared with women without breast cancer (DiD 3.30; 95% CI, 2.04–4.58 at 1 year). Premenopausal women receiving ETx experienced sustained symptom increases over 3 years, particularly those receiving ovarian function suppression. Regimens including gonadotropin-releasing hormone agonists were associated with the greatest increases in symptom burden. Among postmenopausal women, ETx was associated with modest changes in total MRS scores but a significant increase in urogenital symptoms (DiD 0.71; 95% CI, 0.26–1.16).ConclusionsETx for breast cancer is associated with sustained increases in menopausal symptoms beyond those observed during natural menopause. The excess symptom burden is most pronounced among premenopausal women receiving ovarian function suppression. These findings highlight the importance of proactive symptom monitoring and supportive care strategies to improve long-term adherence to endocrine therapy.

Objective: Accurate prediction of breast cancer (BC) recurrence after curative treatment is essential for optimizing surveillance strategies and personalized patient management. We aimed to prospectively validate a previously developed artificial intelligence (AI)–based BC recurrence prediction model in a real-world clinical setting.Methods: This prospective validation study included BC patients who underwent curative surgery followed by adjuvant therapy at a tertiary cancer center. The AI model integrated clinicopathologic variables and longitudinal follow-up data, including laboratory and imaging findings, to estimate individualized 2-year recurrence risk. Patients were stratified into low- and high-risk groups based on a predefined risk threshold. Model performance was evaluated using recurrence outcomes during follow-up.Results: Of 829 patients initially identified, 529 met eligibility criteria, and 400 were randomly selected. Among them, 391 were classified as low-risk and 9 as high-risk. Significant differences in clinicopathologic factors (T/N stage, ER, PR, HER2, Ki-67) were observed between groups, excluding age, adjuvant radiotherapy, and targeted therapy. During follow-up, 7 patients (1.8%) experienced recurrence, 6 in the low-risk group (1.5%) and 1 in the high-risk group (11.1%).Conclusion: This prospective validation demonstrates that the AI-based model can effectively identify BC patients at low risk of recurrence in a real-world clinical setting. The model shows potential as a clinical decision support tool to guide individualized surveillance strategies, although further multi-center studies with larger high-risk populations are warranted to enhance sensitivity and generalizability.

Lee JE, et al. J Breast Cancer. 2026 ;29:e46. https://doi.org/10.4048/jbc.2026.0248

Byun D, et al. J Breast Cancer. 2026 Aug;29(4):354-363. https://doi.org/10.4048/jbc.2026.0018

Lee JE, et al. J Breast Cancer. 2026 ;29:e24. https://doi.org/10.4048/jbc.2026.0038

Woen D, et al. J Breast Cancer. 2026 ;29:e12. https://doi.org/10.4048/jbc.2025.0272

Park WK, et al. J Breast Cancer. 2026 Jan;29:e5. https://doi.org/10.4048/jbc.2025.0290