Dosimetric parameters of the axillary-lateral thoracic junction in predicting risk of breast cancer-related lymphedema.

Hong, Soon Woo; Kim, Eun-Kyu; Shin, Hee-Chul; Beom, Jaewon; Lim, Jae-Young; Kim, In Ah; Kim, Kyubo · Radiother Oncol · 2026

retrospective_cohort · Level III

Where this comes from

Abstract

Regional nodal irradiation (RNI) is associated with breast cancer-related lymphedema (BCRL), yet varying definition of extensive RNI among radiation oncologists leads to the inconsistencies in risk assessment. Recently, the axillary-lateral thoracic junction (ALTJ) has been proposed as a potential organ-of-interest, but its role remains controversial. This study investigates the predictive role of ALTJ dose-volumetric parameters in BCRL development. We identified patients with breast cancer who underwent surgery and adjuvant radiotherapy (2018 - 2020). BCRL was defined as newly developed upper arm lymphedema after radiotherapy, with International Society of Lymphology stage ≥ 2. Following retrospective ALTJ contouring, multivariable Cox regression and Random Survival Forest models were developed using clinical factors and ALTJ dosimetric parameters. Model performance was evaluated via Harrell's C-index and 24-month landmark analysis. We investigated 722 patients with a median follow-up of 30 months. 283 (39.2%) underwent axillary lymph node dissection (ALND). In multivariate Cox regression, along with ALND, ALTJ V<sub>35Gy</sub> > 50% was associated with an increased risk BCRL (p = 0.003, HR 3.07), whereas binary RNI status wasn't (p = 0.363). When comparing multivariable prognostic frameworks, Model B (incorporating ALTJ V<sub>35Gy</sub>) demonstrated a more favorable statistical fit than Model A (incorporating RNI). While Model B yielded a numerically higher C-index (0.753 vs. 0.749), this gain was not statistically significant. This study highlights the significance of the ALTJ as an organ-of-interest in breast cancer patients who underwent radiotherapy. Incorporating ALTJ dosimetric parameters into predictive models enhances prognostic accuracy, addressing the limitations of variable RNI field definitions.