Relative change rate of neutrophil-to-albumin ratio (ΔNAR) during neoadjuvant chemotherapy predicts axillary pathological complete response in locally advanced breast cancer Authors: Yujie Zhao, Yuxiang Deng, Weicai Chen, Meiyu Liu, Zhaoming Peng, Yajie Liu, Lei Zhang Published: 2026 ABSTRACT AbstractBackgroundRoutine blood inflammatory-nutritional markers can predict neoadjuvant chemotherapy (NAC) response in locally advanced breast cancer. Neutrophil-to-albumin ratio (NAR) reflects inflammation and nutrition, but prior studies only use static baseline NAR without tracking its dynamic changes during NAC. Axillary pathological complete response (pCR) strongly guides surgery and prognosis. This study proposed ΔNAR, the relative change rate of NAR pre- vs. post-NAC, and evaluated its predictive value for axillary pCR.MethodsWe retrospectively enrolled 169 node-positive locally advanced breast cancer patients who finished standard NAC and radical surgery. Blood neutrophil and albumin were tested 7 days pre-NAC and preoperatively post-NAC to calculate ΔNAR. Receiver operating characteristic (ROC) curves determined the optimal ΔNAR cutoff. Chi-squared tests and univariate/multivariate logistic regression identified predictors of axillary pCR, defined as ypN0 (no invasive axillary tumor cells).ResultsROC analysis set the optimal ΔNAR cutoff at −19.29% (area under curve =0.663). The low ΔNAR subgroup (ΔNAR ≤−19.29%) had a markedly higher axillary pCR rate than the high ΔNAR group (P=0.001). Univariate analysis linked hormone receptor (HR) status, human epidermal growth factor receptor 2 (HER2) status, anti-HER2 chemotherapy and ΔNAR to axillary pCR. Multivariate regression confirmed that ΔNAR >−19.29% was independently associated with lower odds of axillary pCR (P=0.049).ConclusionsΔNAR acts as an independent biomarker for axillary pCR in locally advanced breast cancer. Patients with ΔNAR above −19.29% have lower odds of axillary pathological remission. Derived from routine blood tests at no extra cost, ΔNAR assists clinicians in preoperative prediction of axillary tumor regression and aids therapeuticplan optimization. Its utility for tailoring surgical plans remains to be further validated.Keywords: Breast cancer, neoadjuvant chemotherapy (NAC), axillary pathological complete response (pCR), neutrophil-to-albumin ratio (NAR), dynamic inflammatory-nutritional biomarker Highlight box Key findings • A dynamic inflammatory-nutritional biomarker ΔNAR [relative change rate of neutrophil-to-albumin ratio (NAR) pre- vs. post-neoadjuvant chemotherapy (NAC)] is developed, with an optimal cutoff value of −19.29%, for predicting axillary pathological complete response (pCR) in node-positive locally advanced breast cancer. What is known and what is new? • Static baseline inflammatory-nutritional markers (such as baseline NAR) have been reported to predict NAC response in locally advanced breast cancer. Axillary pCR is a critical endpoint that directly guides surgical strategy and prognosis assessment. • This study innovatively proposes ΔNAR as a dynamic index tracking NAR changes during NAC, and verifies its independent predictive value for axillary pCR, breaking the limitation of prior studies that only focused on static baseline indicators. What is the implication, and what should change now? • ΔNAR can serve as a noninvasive, cost-free risk-stratification tool to aid preoperative axillary tumor response assessment and support individualized clinical decisionmaking. It should be used only as an auxiliary marker rather than a standalone criterion for surgical de-escalation. For patients with ΔNAR above −19.29% (lower probability of axillary pCR), clinicians can consider more intensive systemic therapy or more prudent axillary surgical planning in advance. Introduction Among women, breast cancer is the most common cancer for both incidence and mortality worldwide (1). Neoadjuvant chemotherapy (NAC) is the standard therapeutic regimen for locally advanced breast cancer characterized by cT3–4 primary tumor or clinically positive axillary lymph nodes (cN+) (2). The core clinical benefits of NAC include tumor downstaging, increased probability of breast-conserving surgery, and early evaluation of tumor chemosensitivity (3). Attaining pathological complete response (pCR), especially axillary pCR (ypN0), is a well-established favorable prognostic marker associated with improved disease-free survival (DFS) and overall survival (OS), particularly for human epidermal growth factor receptor 2 (HER2)-positive and triple-negative breast cancer subtypes (4,5). Achieving axillary pCR also enables the omission of complete axillary lymph node dissection in selected clinical scenarios, which can substantially reduce adverse complications such as axillary lymphedema and other postoperative morbidities related to this invasive surgical procedure (6). Nevertheless, substantial heterogeneity exists in axillary tumor regression after standardized NAC, creating an urgent clinical demand for simple, accessible predictive biomarkers to stratify responders before treatment (7). Cumulative evidence reveals that systemic inflammation and host nutritional status jointly regulate tumor chemotherapeutic sensitivity (8). Composite peripheral blood biomarkers integrating inflammatory and nutritional indicators have attracted widespread research attention (9). Neutrophil-to-albumin ratio (NAR), a novel dual-index biomarker combining absolute neutrophil count (systemic inflammatory burden) and serum albumin (nutritional reserve), has exhibited predictive capacity for tumor progression and treatment response in lung cancer, oral squamous cell carcinoma, rectal cancer and gastrointestinal stromal tumors (10-13). Our previous single-center retrospective study confirmed that static pretreatment NAR independently predicted overall breast plus axillary pCR after NAC, with patients carrying higher baseline NAR demonstrating superior pathological remission rates (14). Notably, all existing relevant studies solely rely on single time-point pretreatment NAR, ignoring dynamic fluctuations of systemic inflammation and nutritional status throughout multi-cycle NAC (15). Chemotherapy exerts bidirectional biological effects: it induces anti-tumor inflammatory responses via immunogenic cell death, while simultaneously causing bone marrow suppression and nutritional consumption, both of which alter circulating neutrophil and albumin levels (16). Changes in inflammatory biomarkers may predict chemosensitivity more accurately compared with static inflammatory indicators (17). Axillary lymph node pathological status determines postoperative tumor staging, adjuvant treatment schemes and long-term prognosis (2). To date, no published research has explored the predictive value of dynamic NAR variation for isolated axillary pCR. In the present study, we adopted the patient cohort and electronic medical database as our prior static NAR research, defined ΔNAR as the percentage relative change of NAR before and after NAC, identified its optimal cutoff value for axillary pCR, and verified its independent predictive efficacy (14). This work aims to provide a convenient peripheral blood dynamic biomarker for preoperative evaluation of axillary tumor remission in breast cancer patients receiving NAC. We present this article in accordance with the STROBE reporting checklist (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-0411/rc). Methods Study population and data source This single-center retrospective observational study retrieved clinical data from the inpatient electronic medical database of Peking University Shenzhen Hospital covering November 2019 to June 2023. All consecutive female patients diagnosed with invasive breast adenocarcinoma who received full-cycle NAC followed by definitive breast radical surgery were screened. Notably, the study cohort enrolled herein differs from patients included in prior related studies, and we exclusively focused on patients presenting with clinically positive axillary lymph nodes at initial diagnosis. Details of the inclusion/exclusion criteria and patient screening process are provided in Figure 1. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Research Ethics Committee of Peking University Shenzhen Hospital (No. 2026-111). As this study employed a retrospective analysis design, the requirement for individual informed consent was waived. Flowchart of patient selection. A total of 261 patients were initially screened, and 169 patients with clinically node-positive locally advanced breast cancer were finally enrolled after applying inclusion and exclusion criteria. NAC, neoadjuvant chemotherapy; NAR, neutrophil-to-albumin ratio; pCR, pathological complete response. Laboratory measurement and ΔNAR calculation Peripheral venous blood samples were collected within 7 days prior to the first cycle of NAC (pre-NAC baseline) and before surgical resection after completion of all NAC cycles (post-NAC). The interval between the last NAC cycle and post-NAC blood sampling was more than two weeks, so the effect of chemotherapy-induced myelosuppression on blood cell counts was minimal. All routine blood tests and serum albumin detection were performed by the central laboratory of Peking University Shenzhen Hospital following standardized operating procedures. NAR at each timepoint was calculated as: NAR = absolute neutrophil count/serum albumin concentration (g/L). The ΔNAR was defined as the primary research biomarker, calculated as percentage value with the formula: ΔNAR = [(post-NAC NAR − pre-NAC NAR)/pre-NAC NAR] × 100%. NAC regimens All patients received standardized NAC regimens formulated in accordance with National Comprehensive Cancer Network (NCCN) breast cancer clinical guidelines (2). HER2-negative patients received anthracycline-taxane combined regimens (such as the AC-T regimen). HER2-positive patients received chemotherapy combined with dual anti-HER2 targeted therapy (TCbHP, THP, AC-THP). No dose reduction or treatment delay was implemented during chemotherapy for all enrolled patients. Specific dosing instructions for each regimen can be found in Table S1. Surgical resection was performed 3–4 weeks after the final NAC cycle, and all patients underwent complete axillary lymph node dissection. Clinicopathological evaluation Immunohistochemistry (IHC) detection of estrogen receptor (ER), progesterone receptor (PR), HER2 and Ki-67 was performed on pretreatment core biopsy specimens. ER/PR positivity was defined as nuclear staining ≥1% tumor cells, tumors with either ER or PR positivity were considered hormone receptor (HR)-positive. A Ki-67 index cutoff of 20% was used to distinguish luminal A and luminal B subtypes. HER2 positivity was judged following American Society of Clinical Oncology (ASCO)/College of American Pathologists (CAP) 2018 guidelines (18). Molecular subtypes were categorized into luminal A, luminal B, HER2-enriched HR-positive, HER2-enriched HR-negative, and triple-negative breast cancer (2). Primary endpoint definition: axillary pCR All enrolled patients presented with clinically positive axillary lymph nodes at initial diagnosis. Axillary pCR (ypN0) was defined as complete absence of invasive tumor cells in all harvested axillary lymph nodes on postoperative surgical pathology, regardless of residual tumor status in the primary breast lesion (19). Patients with any invasive tumor deposits in axillary lymph nodes were classified as non-axillary pCR. Statistical analysis All statistical analyses were performed using SPSS 20.0 (IBM, Chicago, IL, USA), R statistical software (version 4.3.0; R Foundation for Statistical Computing, Vienna, Austria) and GraphPad Prism 7 software (GraphPad Software, Inc., San Diego, CA, USA). Enumeration data were presented as case number (percentage). Receiver operating characteristic (ROC) curve analysis was constructed to calculate the area under the curve (AUC), sensitivity, specificity and Youden index. The pre-NAC NAR, post-NAC NAR, and ΔNAR values corresponding to the maximum Youden index were defined as the optimal cutoff values for predicting axillary pCR. Intergroup comparisons of clinicopathological features among different subgroups were performed using Pearson’s χ2 test or continuity-corrected χ2 test. Logistic regression models were applied to screen predictive factors of axillary pCR. Variables with P<0.05 in univariate analysis were incorporated into multivariate logistic regression with the enter method to identify independent predictive biomarkers. Before multivariable modeling, multicollinearity among candidate predictors was assessed using variance inflation factors (VIFs); a VIF value <5 indicated no significant multicollinearity. Odds ratio (OR) and corresponding 95% confidence interval (CI) were recorded. A two-tailed P value <0.05 was defined as statistically significant. To further assess the incremental predictive value of ΔNAR, we constructed a baseline model incorporating molecular subtype and chemotherapy regimen. We then compared the AUC of the baseline model with that of the baseline model plus ΔNAR using the DeLong test. Net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were also calculated to quantify the added predictive benefit of ΔNAR. To evaluate the robustness of the predictive model, Bootstrap sensitivity analysis with 1,000 resamples was performed to estimate bias-corrected CIs. Results Baseline clinicopathological characteristics of the cohort Table 1 summarizes the clinicopathologic characteristics of the 169 patients enrolled in this study. The median patient age was 47 years (range, 26–68 years), and the median tumor size was 35 mm (range, 9–91 mm). At initial diagnosis, most patients (72.2%) were classified as clinical T2 (cT2) stage. In the overall cohort, 75.1% had HR-positive disease, 58.6% were HER2-negative, and 88.8% had a Ki-67 index greater than 20%. Additionally, 49.1% (83 out of 169) of patients achieved axillary pCR. The median values of the calculated pre-NAC NAR, post-NAC NAR, and ΔNAR were 0.095 (range, 0.034–0.230), 0.085 (range, 0.011–0.502), and −12.88% (range, −91.27% to 426.21%), respectively. Table 1Clinicopathologic features of patients involved in this study (n=169)CharacteristicsValues, n (%)Age (years) ≤50108 (63.9) >5061 (36.1)cT stage 112 (7.1) 2122 (72.2) 324 (14.2) 411 (6.5)cN stage 1122 (72.2) 26 (3.6) 341 (24.3)HR Negative42 (24.9) Positive127 (75.1)HER2 Negative99 (58.6) Positive70 (41.4)Ki-67 <20%19 (11.2) ≥20%150 (88.8)Molecular subtype Luminal A13 (7.7) Luminal B66 (39.1) HER2 enriched (HR+)48 (28.4) HER2 enriched (HR−)22 (13.0) Triple negative20 (11.8)Chemotherapy regimen Anthracycline and taxane100 (59.2) Chemo + trastuzumab + pertuzumab69 (40.8)Axillary pCR Yes83 (49.1) No86 (50.9)HER2, human epidermal growth factor receptor 2; HR, hormone receptor; N, node; pCR, pathological complete response; T, tumor. Clinicopathologic features of patients involved in this study (n=169) HER2, human epidermal growth factor receptor 2; HR, hormone receptor; N, node; pCR, pathological complete response; T, tumor. ROC curve‑derived optimal cutoff values of pre‑NAC NAR, post‑NAC NAR, and ΔNAR ROC curve analysis was performed to determine the optimal cut‑off values of pre‑NAC NAR (0.089), post‑NAC NAR (0.057) and ΔNAR (−19.29%) for predicting axillary pCR. The ROC curve analysis showed that pre‑NAC NAR had a poor discriminatory ability for predicting axillary pCR, with an AUC of 0.586 (95% CI: 0.500–0.672, P=0.054). Post‑NAC NAR yielded an AUC of 0.594 (95% CI: 0.508–0.679, P=0.04), also showing limited predictive performance. ΔNAR demonstrated a moderate discriminatory ability for predicting axillary pCR, with an AUC of 0.663 (95% CI: 0.580–0.743, P<0.001). Patients were dichotomized according to each optimal cut‑off value: pre‑NAC NAR ≤0.089 (n=71) vs. >0.089 (n=98); post‑NAC NAR ≤0.057 (n=44) vs. >0.057 (n=125); ΔNAR ≤−19.29% (n=75) vs. >−19.29% (n=94). The corresponding ROC curves are presented in Figure 2. ROC curve of ΔNAR for predicting axillary pCR. The optimal ΔNAR cutoff value was −19.29%. AUC, area under curve; NAC, neoadjuvant chemotherapy; NAR, neutrophil-to-albumin ratio; pCR, pathological complete response; ROC, receiver operating characteristic. Correlations of pre‑NAC NAR, post‑NAC NAR, and ΔNAR with clinicopathological features We further compared the distribution of clinicopathological variables between the low and high ΔNAR subgroups (Table 2). No statistically significant intergroup differences were detected with regard to age (P=0.11), cT stage (P=0.86), cN stage (P=0.77), HR status (P=0.40), HER2 status (P=0.12), Ki-67 index (P=0.23), and NAC regimen (P=0.09). None of the above clinical and pathological parameters exhibited a significant correlation with ΔNAR stratification. In addition, we also compared the proportions of clinicopathological variables between low‑ and high‑NAR subgroups stratified by pre‑NAC NAR and post‑NAC NAR. HER2 status, Ki‑67 index, and chemotherapy regimen were found to be associated with pre‑NAC NAR, whereas the remaining clinicopathological parameters showed no significant correlations with grouping based on either pre‑NAC NAR or post‑NAC NAR (Table S2). Table 2Relationships between ΔNAR and patient characteristicsCharacteristicsΔNAR (n=169)≤−19.29% (n=75), n (%)>−19.29% (n=94), n (%)P valueAge (years)0.11 ≤5043 (39.8)65 (60.2) >5032 (52.5)29 (47.5)cT stage0.86 1–259 (44.0)75 (56.0) 3–416 (45.7)19 (54.3)cN stage0.77 155 (45.1)67 (54.9) 2–320 (42.6)27 (57.4)HR0.40 Negative21 (50.0)21 (50.0) Positive54 (42.5)73 (57.5)HER20.12 Negative39 (39.4)60 (60.6) Positive36 (51.4)34 (48.6)Ki-670.23 <20%6 (31.6)13 (68.4) ≥20%69 (46.0)81 (54.0)Chemotherapy regimen0.09 Anthracycline and taxane39 (39.0)61 (61.0) Chemo + anti HER236 (52.2)33 (47.8)HER2, human epidermal growth factor receptor 2; HR, hormone receptor; N, node; NAR, neutrophil-to-albumin ratio; pCR, pathological complete response; T, tumor. Relationships between ΔNAR and patient characteristics HER2, human epidermal growth factor receptor 2; HR, hormone receptor; N, node; NAR, neutrophil-to-albumin ratio; pCR, pathological complete response; T, tumor. Associations of clinicopathological variables, pre‑NAC NAR, post‑NAC NAR, and ΔNAR with axillary pCR Chi-squared tests were performed to explore factors associated with axillary pCR (Table 3). Age (P=0.53), cT stage (P=0.29), cN stage (P=0.71), and Ki-67 index (P=0.26) showed no significant correlation with axillary pCR. Compared with the nonaxillary-pCR group, the axillary pCR group comprised more HRnegative (P=0.009) and HER2-positive (P<0.001) cases, and presented lower ΔNAR (P=0.001), higher pre-NAC NAR (P=0.03), and lower post-NAC NAR (P=0.03). In addition, a larger proportion of patients in the axillary pCR group received chemotherapy plus anti-HER2 agents (P<0.001). Table 3Relationships between axillary pCR and patient characteristicsCharacteristicsNon pCR (n=86), n (%)pCR (n=83), n (%)P valueAge (years)0.53 ≤5053 (49.1)55 (50.9) >5033 (54.1)28 (45.9)cT stage0.29 1–271 (53.0)63 (47.0) 3–415 (42.9)20 (57.1)cN stage0.71 161 (50.0)61 (50.0) 2–325 (53.2)22 (46.8)HR0.009 Negative14 (33.3)28 (66.7) Positive72 (56.7)55 (43.3)HER2<0.001 Negative69 (69.7)30 (30.3) Positive17 (24.3)53 (75.7)Ki-670.26 <20%12 (63.2)7 (36.8) ≥20%74 (49.3)76 (50.7)Chemotherapy regimen<0.001 Anthracycline and taxane70 (70.0)30 (30.0) Chemo + anti HER216 (23.2)53 (76.8)ΔNAR0.001 ≤−19.29%27 (36.0)48 (64.0) >−19.29%59 (62.8)35 (37.2)Pre-NAC NAR0.03 ≤0.08943 (60.6)28 (39.4) >0.08943 (43.9)55 (56.1)Post-NAC NAR0.03 ≤0.05716 (36.4)28 (63.6) >0.05770 (56.0)55 (44.0)HER2, human epidermal growth factor receptor 2; HR, hormone receptor; N, node; NAC, neoadjuvant chemotherapy; NAR, neutrophil-to-albumin ratio; pCR, pathological complete response; T, tumor. Relationships between axillary pCR and patient characteristics HER2, human epidermal growth factor receptor 2; HR, hormone receptor; N, node; NAC, neoadjuvant chemotherapy; NAR, neutrophil-to-albumin ratio; pCR, pathological complete response; T, tumor. Univariate and multivariate logistic regression analyses for independent predictors of axillary pCR Univariate analysis indicated that HR status (OR =0.382, 95% CI: 0.184–0.794, P=0.01), HER2 status (OR =7.171, 95% CI: 3.581–14.359, P<0.001), chemotherapy regimen (OR =7.729, 95% CI: 3.823–15.626, P<0.001), ΔNAR levels (OR =0.334, 95% CI: 0.178–0.627, P=0.001), pre-NAC NAR levels (OR =1.964, 95% CI: 1.056–3.655, P=0.03) and post-NAC NAR levels (OR =0.449, 95% CI: 0.221–0.912, P=0.03) were significantly correlated with axillary pCR. Age, cT stage, cN stage and Ki-67 index exhibited no univariate predictive value, as visualized in the forest plot (Figure 3, Table 4). Prior to multivariable modeling, VIFs were used to assess multicollinearity across all candidate predictors, with a VIF <5 defined as the cutoff for absence of substantial multicollinearity. No notable multicollinearity was identified for HR status, ΔTHR, and other candidate predictors (all VIFs <5). Nevertheless, severe multicollinearity (VIF >10) was detected between HER2 status and anti-HER2 targeted-therapy administration. Given that the administration of anti-HER2 targeted therapy was dictated by intrinsic tumor HER2 status in this real-world cohort, baseline HER2 status was retained in the multivariate logistic regression model, whereas targeted-therapy status was excluded. Both variables were evaluated during univariate screening. Forest plot of multivariate logistic regression analysis for independent predictors of axillary pCR, displaying ORs and 95% CIs. The vertical dotted line represents OR =1. CI, confidence interval; HER2, human epidermal growth factor receptor 2; HR, hormone receptor; N, node; NAC, neoadjuvant chemotherapy; NAR, neutrophil-to-albumin ratio; OR, odds ratio; pCR, pathological complete response; T, tumor. Table 4Association of patient/tumor characteristics to axillary pCR in univariate and multivariate analysisVariableUnivariate analysisMultivariate analysisOR (95% CI)P valueOR (95% CI)P valueAge (years) ≤501 (reference)––– >500.818 (0.436–1.534)0.53––cT stage 1–21 (reference)––– 3–41.503 (0.709–3.183)0.29––cN stage 11 (reference)––– 2–30.880 (0.449–1.727)0.71––HR Negative1 (reference)–1 (reference)– Positive0.382 (0.184–0.794)0.010.430 (0.188–0.984)0.046HER2 Negative1 (reference)––– Positive7.171 (3.581–14.359)<0.0016.896 (3.244–14.659)<0.001Ki-67 <20%1 (reference)––– ≥20%1.761 (0.657–4.717)0.26––Chemotherapy regimen Chemo only1 (reference)––– Chemo + anti HER27.729 (3.823–15.626)<0.001––ΔNAR ≤−19.29%1 (reference)–1 (reference)– >−19.29%0.334 (0.178–0.627)0.0010.380 (0.144–0.999)0.049Pre-NAC NAR ≤0.0891 (reference)–1 (reference)– >0.0891.964 (1.056–3.655)0.030.986 (0.428–2.270)0.97Post-NAC NAR ≤0.0571 (reference)–1 (reference)– >0.0570.449 (0.221–0.912)0.030.794 (0.279–2.262)0.67P value, in logistic regression analysis, variables found to be statistically significant (P<0.05) in univariate analysis were entered into a logistic regression multivariate model using enter method. CI, confidence interval; HER2, human epidermal growth factor receptor 2; HR, hormone receptor; N, node; NAC, neoadjuvant chemotherapy; NAR, neutrophil-to-albumin ratio; OR, odds ratio; pCR, pathological complete response; T, tumor. Association of patient/tumor characteristics to axillary pCR in univariate and multivariate analysis P value, in logistic regression analysis, variables found to be statistically significant (P<0.05) in univariate analysis were entered into a logistic regression multivariate model using enter method. CI, confidence interval; HER2, human epidermal growth factor receptor 2; HR, hormone receptor; N, node; NAC, neoadjuvant chemotherapy; NAR, neutrophil-to-albumin ratio; OR, odds ratio; pCR, pathological complete response; T, tumor. After adjustment for confounding factors in multivariate regression, three variables remained as statistically significant independent predictors of axillary pCR. HER2positive status was an independent positive predictor of axillary pCR (OR =6.896, 95% CI: 3.244–14.659, P<0.001). Higher ΔNAR level and HR-positive status independently reduced the likelihood of axillary pCR (OR =0.380, 95% CI: 0.144–0.999, P=0.049; OR =0.430, 95% CI: 0.188–0.984, P=0.046). Incremental predictive value of ΔNAR beyond clinical variables To further evaluate the incremental predictive value of ΔNAR, we constructed a baseline model incorporating molecular subtype and chemotherapy regimen. The clinical-only model including tumor molecular subtype and chemotherapy regimen achieved modest discrimination, with an AUC of 0.765 (95% CI: 0.696–0.835). After adding ΔNAR to this baseline model, the discriminatory performance improved slightly, with an AUC of 0.798 (95% CI: 0.731–0.865). The DeLong test demonstrated that the difference in AUC between the two models did not reach statistical significance (Z =−1.81, P=0.07). Nevertheless, continuous NRI was 0.529 (95% CI: 0.235–0.812, P<0.001) and IDI was 0.041 (95% CI: 0.003–0.113, P=0.006) after ΔNAR was incorporated into the clinical model. The CIs for both indices did not include zero, indicating statistically-significant reclassification and discrimination improvement provided by ΔNAR over existing clinical factors (Table S3). Discussion Accurate pretreatment prediction of axillary lymph node pathological remission is critical for individualized NAC regimen adjustment and axillary surgical strategy selection. In certain clinical scenarios, patients with negative sentinel lymph node status after NAC can avoid complete axillary lymph node dissection or receive adjuvant radiotherapy as an alternative to dissection, which markedly alleviates postoperative complications such as upper-extremity lymphedema induced by axillary surgery (6). Static single-time-point inflammatory biomarkers cannot reflect dynamic changes of host inflammation and nutritional status during multiple cycles of chemotherapy (20). In this study, we proposed a novel dynamic biomarker ΔNAR defined as the relative percentage change of NAR before and after NAC, and validated that ΔNAR ≤−19.29% independently predicts axillary pCR in locally advanced breast cancer patients based on the single-center cohort as our previous static NAR research. NAR simultaneously reflects two core host physiological dimensions: systemic inflammation represented by circulating neutrophils and nutritional reserve reflected by serum albumin (10). Tumor cells trigger systemic inflammatory cascades to induce neutrophil proliferation and activation; activated neutrophils secrete reactive oxygen species, tumor necrosis factor alpha (TNF-α) and nitric oxide to modulate anti-tumor immune responses during chemotherapy (21). Serum albumin participates in antioxidative stress, microcirculation regulation and immune homeostasis; tumor consumption and chemotherapy-induced malnutrition cause hypoalbuminemia, which impairs lymphocyte anti-tumor function and reduces tumor chemosensitivity (22). Changes in inflammatory biomarkers may predict chemosensitivity more accurately compared with static inflammatory indicators (17). Our results demonstrated that patients with greater relative NAR reduction (lower ΔNAR) were more likely to achieve complete elimination of axillary invasive tumor lesions, which may be explained by two potential mechanisms. First, in patients with lower ΔNAR, NAC alleviates tumor-driven chronic systemic inflammation, reducing the recruitment of immunosuppressive neutrophils [polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs)] to axillary lymph nodes (23). These neutrophils upregulate programmed death-ligand 1 (PD-L1), inhibiting cytotoxic T cell activity and promoting tumor escape (24). Notably, NAC can induce the polarization of anti-tumor neutrophils toward an immunosuppressive N2 phenotype, a shift driven by chemotherapy-induced tumor microenvironment changes [e.g., increased transforming growth factor beta (TGF-β), interleukin (IL)-4, and IL-13 secretion] and potentially linked to neutrophil extracellular trap (NET) formation—a key mediator of chemotherapy resistance (25,26). NETs, formed by neutrophils in response to chemotherapy, trap and activate latent TGF-β via integrin-αvβ1 and MMP9, triggering epithelial-to-mesenchymal transition (EMT) in cancer cells and exacerbating resistance (26). Conversely, high ΔNAR patients have persistent systemic inflammation, causing sustained immunosuppressive neutrophil proliferation, elevated NAR, and impaired axillary tumor clearance. This aligns with prior findings that circulating granulocytic MDSC (G-MDSC) levels are lower in NAC-responsive patients than non-responders, and clinical evidence showing elevated plasma NET levels correlate with progressive disease in chemotherapy-treated metastatic breast cancer patients (23,26). Together, these highlight the key roles of MDSC regulation, neutrophil polarization, and NET formation in NAC response. Second, patients with low ΔNAR usually maintain relatively stable serum albumin levels during NAC, indicating preserved nutritional status and intact anti-tumor immune function, avoiding the severe immunosuppression caused by hypoalbuminemia (27). In contrast, patients with high ΔNAR often suffer from chemotherapy-related malnutrition and bone marrow suppression, leading to blunted systemic anti-tumor inflammatory responses and residual axillary tumor infiltration. Consistent with our previous NAR-related study (14), univariate logistic regression analyses verified that patients receiving chemotherapy combined with anti-HER2 targeted therapy were more likely to achieve axillary pCR, which may be attributed to synergistic anti-tumor effects of anti-HER2 targeted therapy and chemotherapy that amplify systemic inflammatory anti-tumor responses (28). Multivariate regression further verified that ΔNAR retained the independent predictive value after adjusting for HR status, HER2 status, and treatment regimens, proving that dynamic NAR variation provides unique predictive information unavailable from traditional static clinical and pathological markers (29). Compared with gene sequencing, liquid biopsy and imaging predictive models, ΔNAR has prominent clinical translational advantages: it is calculated from routine blood tests without extra cost or additional invasive examinations, and is widely accessible in primary medical institutions. The optimal cutoff value of −19.29% derived from this cohort can be directly applied to preoperative risk stratification. For patients with ΔNAR >−19.29%, clinicians may consider intensified neoadjuvant regimens or auxiliary anti-inflammatory interventions to improve axillary tumor regression; patients with ΔNAR ≤−19.29% have high probability of axillary pCR. While this provides theoretical support for deescalated axillary surgery, the safety of clinical decisions relying solely on ΔNAR cannot be confirmed, as all patients in this cohort underwent complete axillary lymph node dissection without treatment deescalation. ΔNAR should be used only as an auxiliary marker; clinicians need to integrate tumor subtype, chemotherapy regimen, imaging response and other parameters when evaluating the possibility of axillary deescalation therapy. This study has several limitations. First, this is a retrospective single-center analysis with a moderate sample size of 169 patients, and external multi-center cohorts are required to validate the −19.29% cutoff value of ΔNAR. Second, we did not explore the underlying molecular mechanism linking ΔNAR and axillary tumor chemosensitivity; future basic experiments are needed to clarify neutrophil-albumin mediated immune regulation in lymph node metastatic microenvironment. Third, we did not conduct long-term follow-up to analyze the correlation between ΔNAR and postoperative survival endpoints including DFS and OS, which will be supplemented in our subsequent prospective cohort research. Fourth, all patients in our cohort received complete axillary lymph node dissection. Consequently, we could only demonstrate an association between ΔNAR and axillary pCR. Our study cannot verify the safety of de-escalated axillary surgery guided by ΔNAR. Prospective trials are essential before ΔNAR can be incorporated into real-world clinical decision-making. Conclusions Dynamic relative ΔNAR during NAC is an independent predictive biomarker for axillary pCR in locally advanced breast cancer. Patients with ΔNAR lower than −19.29% exhibit significantly higher axillary pCR rates. As a low-cost, easily accessible peripheral blood dynamic inflammatory-nutritional index, ΔNAR can assist clinicians in the preoperative prediction of axillary lymph node remission and help optimize NAC regimens for breast cancer patients. Its utility to guide individualized axillary surgical strategies remains to be validated in prospective cohorts. Supplementary The article’s supplementary files as Acknowledgments None. Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Research Ethics Committee of Peking University Shenzhen Hospital (No. 2026-111). As this study employed a retrospective analysis design, the requirement for individual informed consent was waived. Footnotes Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-0411/rc Funding: This work was supported by the Sanming Project of Medicine in Shenzhen (No. SZSM202211036), Shenzhen Science and Technology Program (No. JCYJ20240813115830041), and Open-ended Research Topic of Hubei Clinical Research Center for Reproductive Medicine (No. 2025RMOF028). Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-0411/coif). The authors have no conflicts of interest to declare. Data Sharing Statement Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-0411/dss References Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-0411/prf Supplementary Materials The article’s supplementary files as Data Availability Statement Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-0411/dss