Breast cancer-related arm lymphedema (BCRaL) is a prevalent and serious complication following breast cancer treatment. Axillary lymph node dissection (ALND), a key surgical procedure for breast cancer, is the primary cause of BCRaL, affecting 19.9% of patients, compared to 5.6% following sentinel lymph node biopsy (SLNB) (1). BCRaL induces upper limb heaviness, pain, numbness, and dysfunction, along with psychological issues like anxiety, depression, and suicidal tendencies, significantly impacting patients’ quality of life (2–6). Although BCRaL cannot be cured, existing interventions can alleviate the symptoms of early-stage BCRaL patients but are ineffective in treating patients with late-stage BCRaL (7, 8). Effective prediction and early identification of high-risk BCRaL patients are crucial for prevention and early intervention.

Several risk prediction models have been proposed to assess BCRaL risk, and the risk factors identified vary slightly between models, depending on the study population, sample size, and variables included in the analysis (9–13). Risk factors for BCRaL that are supported by the evidence include high body mass index (BMI), removal of increased lymph nodes, and receiving modified radical mastectomy (MRM), ALND, radiotherapy, or chemotherapy (14–19). Breast cancer treatment is becoming increasingly individualized, which may influence the occurrence of BCRaL. Neoadjuvant chemotherapy (NAC), a standard treatment for locally advanced breast cancer, has been recognized as a risk factor for BCRaL development (17, 20, 21). Existing prediction models have typically treated chemotherapy as a unified variable without adequately distinguishing between neoadjuvant and adjuvant approaches in risk quantification.

To address this gap, we conducted a retrospective analysis of breast cancer patients receiving both ALND and chemotherapy, with particular focus on differentiating chemotherapy types (neoadjuvant versus adjuvant) as distinct variables. Our study builds upon previous findings by not only validating the independent risk of NAC in a rigorously defined ALND patient cohort but also precisely quantifying its risk relative to adjuvant chemotherapy (AC). By incorporating this distinction into a comprehensive nomogram, we provide an enhanced risk assessment tool that enables more individualized surveillance and prevention strategies for postoperative breast cancer patients, ultimately aiming to reduce BCRaL incidence and improve quality of life.

This retrospective cohort study, with a median follow-up of 39 months, established a 20.6% incidence of BCRaL following ALND and chemotherapy. We developed and internally validated a BCRaL risk prediction model that specifically differentiates between NAC and AC as distinct variables. In addition to NAC, the model identifies elevated BMI, an increased number of positive lymph nodes, HER2-targeted therapy, and SCRT as independent predictors. By integrating these factors into a comprehensive nomogram, this study provides a validated tool that incorporates NAC to enable individualized risk assessment and supports targeted prevention strategies for patients undergoing ALND and chemotherapy.

Our investigation demonstrated that NAC significantly elevated BCRaL risk, showing a two-fold increase compared to AC. This finding aligns with existing literature, including a prospective study by Montagna et al. demonstrating NAC to be independently associated with BCRaL risk versus upfront surgery (OR = 2.10, 95% CI: 1.16-3.95) (22). Another investigation reported an even more pronounced 3.76-fold risk elevation (95% CI: 2.29-6.20) among ALND patients receiving NAC (23). Given the established role of NAC in the preoperative management of locally advanced breast cancer and selected triple-negative or HER2-positive subtypes (24, 25), a comprehensive understanding of its associated risks is clinically imperative. The elevated risk may be attributable to multimodal treatment sequencing. As proposed by Kim et al. (26), patients undergoing NAC typically receive radiotherapy shortly following ALND, potentially inducing secondary lymphatic injury before adequate collateral circulation develops, thereby compromising recovery. In contrast, Specht et al. observed a non-significant trend toward risk reduction with NAC (HR = 0.74, 95% CI: 0.37-1.48), hypothesizing that tumor downstaging might mitigate lymphatic damage (27).

The analysis further identified HER2-targeted therapy as a significant predictor, with patients receiving this treatment exhibiting a 1.689-fold increased BCRaL risk. The biological mechanism linking HER2-targeted therapy to BCRaL risk remains unclear. One plausible hypothesis involves vascular endothelial growth factor-C (VEGF-C), a critical mediator of lymphangiogenesis (28, 29). HER2-overexpressing breast cancers demonstrate upregulated VEGF-C expression (30), and anti-HER2 agents have been shown to suppress VEGF-C levels (29, 31). Thus, targeted therapy might inadvertently inhibit lymphatic repair processes. However, it is also critical to consider that HER2-positive disease is characterized by more aggressive features and higher nodal burden. Therefore, while our model adjusted for nodal status, the observed association may partially reflect the underlying tumor biology rather than a direct treatment effect.

A strong association emerged between elevated BMI and BCRaL development, with each unit increase in BMI corresponding to a hazard ratio of 1.121. This study confirmed a significant association between elevated BMI and increased BCRaL risk (HR = 1.121), consistent with established literature (14, 15, 19, 32, 33). The BMI definitions vary across studies, with Jiang et al. establishing BMI>25 kg/m² as a risk threshold (OR = 1.73, 95%CI:1.30-2.31) (9), and Armer et al. identified continuous BMI increase as an independent risk factor (HR = 1.03, 95%CI:1.01-1.06) (20). Our findings demonstrate progressive risk elevation with rising BMI. Additional studies validate the prognostic utility of continuous BMI for predicting lymphedema severity (34). The underlying mechanisms likely involve dual pathways: obesity-induced inflammatory factors impair lymphatic structure and function, increasing leakage while decreasing pumping capacity (35–38); simultaneously, excess adipose tissue may mechanically compress the lymphatic system, impeding lymphatic fluid return (36, 39, 40). These pathophysiological changes collectively contribute to elevated lymphedema risk in high-BMI patients.

The study also revealed a dose-response relationship between nodal burden and lymphedema risk, demonstrating a 4% risk increase per additional positive lymph node. Similar findings have been reported in several previous studies (16, 41–44). Penn et al. discovered that an increase of one positive lymph node increased the BCRaL risk by 8% (43). Additionally, a meta-analysis showed a moderate level of evidence supporting the presence of metastatic lymph nodes as a risk factor for BCRaL occurrence (14). Positive lymph nodes may experience an increase in the BCRaL risk due to lymphatic blockage caused by tumoral infiltration (45).

Notably, SCRT was specifically associated with heightened BCRaL risk, whereas RT overall showed no significant association. The extensive axillary dissection in our cohort (median 20 nodes removed) may have obscured any additional risk from general RT, as suggested by Naoum et al. (46). However, we found that SCRT was associated with a nearly 1.7-fold increased BCRaL risk compared to the non-SCRT patient population (HR = 1.666, 95% CI: 1.033-2.687). This is consistent with Jung’s (HR = 1.93, 95% CI: 1.20-3.10) and Kim’s (HR = 2.74, 95% CI: 1.80-4.17) findings that breast RT with SCRT resulted in a greater risk of lymphedema development (47, 48). Anatomically, the supraclavicular region contains dense lymphatic networks, and SCRT likely impairs drainage capacity through tissue fibrosis and direct vessel damage (49), explaining its distinct risk profile compared to RT alone.

We successfully developed and validated a comprehensive nomogram incorporating five independent predictors: elevated body mass index (BMI), increased number of positive axillary lymph nodes, NAC, HER2-targeted therapy, and SCRT. The model demonstrated good discrimination with C-index values of 0.692 (training) and 0.719 (validation), along with consistent time-dependent AUC values (0.659-0.760) across 1-, 3-, and 5-year predictions. Calibration curves confirmed accuracy, while decision curve analysis established clinical utility superior to universal intervention strategies. With an optimal cutoff score of 82.69, the nomogram effectively stratifies patients into distinct risk categories, providing clinicians with a practical tool for individualized risk assessment and targeted preventive care in breast cancer management. A comparison of our model’s performance with recently published nomograms provides important context. The discriminative ability of our model (AUC 0.760 at 5 years) is lower than the exceptional performance (AUC 0.944) reported by Luo et al. in a large Chinese cohort (13). This discrepancy is likely attributable to fundamental differences in predictor selection and modeling approach. The model by Luo et al. incorporated a wider array of strong predictors, including comorbidities and postoperative complications, and aimed to predict risk within a fixed 12-month period using logistic regression. In contrast, our Cox model focused on treatment-specific factors and provides dynamic risk stratification over a 5-year horizon, a more complex predictive task. Our performance is, however, comparable to the validation AUC of 0.724 reported by Jiang et al. (9). Therefore, the clinical value of our nomogram is its specific focus on quantifying the impact of modern treatment modalities (e.g., NAC, targeted therapy) on long-term lymphedema risk, offering a complementary tool for personalized surveillance planning in high-risk patients undergoing contemporary multi-modal therapy.

Based on the established nomogram cut-off score of 82.69, we propose distinct clinical management pathways: Low-risk patients (score <82.69) receive standardized lymphedema education. High-risk patients (score ≥82.69) are recommended a comprehensive prevention strategy including intensified 3–4 month surveillance for 2–3 years, preventive compression sleeve use during strenuous activities, and weight management support for those with elevated BMI.

The developed prediction model offers several distinct advantages for clinical practice. First, it precisely targets high-risk patients undergoing ALND combined with chemotherapy, ensuring service to those most needing individualized management. Second, it innovatively integrates modern comprehensive treatment elements, including NAC and HER2-targeted therapy. Third, it provides multi-timepoint dynamic risk assessment at 1, 3, and 5 years, overcoming the limitations of traditional static prediction and supporting long-term management strategy development. Fourth, clinical decision curve analysis confirms the model’s significant net benefit and superior clinical utility compared to extreme intervention strategies. While demonstrating clinical utility, this study has several limitations that should be considered when interpreting the results. First, the assessment of BCRaL was based on circumferential measurements, which may introduce measurement bias. Although all measurements were performed by trained therapists using a standardized protocol, potential inter-observer variability remains a concern. Second, due to the retrospective nature of this study, several potentially important predictors were not available for analysis, including the precise level of ALND, operative time, surgical site infection, post-operative complications, and specific comorbidities such as hypertension; other potential risk factors, such as genetic background, lifestyle psychological factors, and financial factors (14, 50, 51), were not included in the model. The omission of these variables may have influenced the observed associations. Third, while we adjusted for key clinicopathological factors, including the number of positive lymph nodes, residual confounding may persist. Specifically, HER2-targeted therapy and SCRT are often indicators of more advanced disease, and despite our multivariate adjustments, it remains challenging to fully disentangle their independent effects from the underlying tumor biology and overall treatment intensity. Fourth, although the model showed strong performance during 1, 3, and 5-year follow-up periods, breast cancer patients may need extended monitoring to evaluate late complications or chronic lymphedema development. Thus, future studies should extend the follow-up period to verify the model’s predictive ability over a longer time frame. Fifth, as this was a single-center study conducted in a Chinese population, the generalizability of our nomogram to other ethnic groups and healthcare settings requires validation. Future external validation in multi-center and international cohorts is essential to ensure the model’s broad applicability and stability. Sixth, with the ongoing evolution of axillary management toward less invasive approaches, the applicability of our model may be most relevant for patients who still require comprehensive ALND due to advanced nodal disease. Finally, the study lacked a pre-specified sample size calculation, a common limitation in retrospective modeling, highlighting the need for confirmation in future, explicitly powered prospective studies.

This study successfully developed and internally validated a predictive nomogram for BCRaL risk specifically targeting patients undergoing ALND and chemotherapy. The model innovatively integrates modern treatment modalities (NAC, HER2-targeted therapy, SCRT) and traditional risk factors (BMI, number of positive lymph nodes), enabling precise identification of high-risk patients. Validation results demonstrated robust predictive accuracy and clinical utility, with consistent performance maintained across 1-, 3-, and 5-year follow-up periods. It is important to note that this model was not externally validated, making its implementation a critical future research step. The proposed tool should be used only as a reference for identifying high-risk patients. Its clinical utility in guiding precision prevention strategies remains exploratory and requires validation through future multicenter studies.