Development and internal validation of a nomogram based on HER2 status for predicting pathological complete response to neoadjuvant chemotherapy in triple-negative breast cancer

Introduction

Breast cancer (BC) stands as the most common malignant tumor and a leading cause of cancer-related mortality among women globally, imposing an enormous socioeconomic and healthcare burden on global populations (1). Triple-negative breast cancer (TNBC) is a subtype of BC that lacks estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) expression, accounting for approximately 15–20% of all BC (2). It is characterized by strong invasiveness, high risk of early recurrence and metastasis, and a lack of specific targeted therapeutic drugs, making neoadjuvant chemotherapy (NAC) the cornerstone of its clinical treatment (2,3). Pathological complete response (pCR) after NAC is a well-recognized strong prognostic marker for TNBC, with patients achieving pCR exhibiting significantly improved disease-free survival (DFS) and overall survival (OS) when compared with those with residual disease (3,4). Multiple meta-analyses have shown that NAC can achieve a pCR rate of 30–50% in TNBC (4,5), but the therapeutic response varies greatly among individual patients, and those with unfavorable NAC responses face heightened risks of treatment failure and poor prognosis (7). Therefore, developing accurate and personalized predictive model for NAC efficacy in TNBC is of great clinical significance for optimizing treatment regimens, avoiding ineffective chemotherapy, and improving patient outcomes.

In recent years, extensive efforts have been made to establish predictive models for TNBC NAC response, and various models based on clinicopathologic factors, radiomics features, and multi-biomarker signatures have been developed and validated. Clinicopathologic factor-based models are the most widely used in clinical practice, with core predictors including Ki-67 proliferation index, tumor-infiltrating lymphocytes (TILs), tumor stage, and histological grade (8,9). For example, Liao et al. constructed a nomogram for TNBC NAC efficacy prediction using Ki-67, T stage, and lymph node status, with a reported area under the receiver operating characteristic curve (AUC) of 0.72 (10). Radiomics-based models, leveraging non-invasive imaging features from magnetic resonance imaging (MRI) and ultrasound, have also shown promising predictive value, with AUC values ranging from 0.70 to 0.85 in single-center and small-sample studies (11,12); these models exhibit the advantage of real-time non-invasive assessment but are limited by poor standardization of imaging acquisition and post-processing across centers. Multi-biomarker signature models, integrating molecular biomarkers (e.g., BRCA1/2, PD-L1) and genomic signatures [e.g., homologous recombination deficiency (HRD)], have further improved predictive accuracy, with some studies reporting an AUC of up to 0.80 (13,14); however, these models often require complex and expensive detection techniques, limiting their clinical translation in resource-limited settings. Despite the progress of these existing models, none of them have incorporated the stratification of HER2-zero and HER2-low status, and the interaction between HER2 subtypes and other biomarkers in predicting TNBC NAC response remains largely unexplored, which constitutes a critical knowledge gap in the field of TNBC personalized treatment.

With the proposal and standardized definition of HER2-low expression in BC, the traditional “HER2-negative” classification of TNBC has been reshaped (8,9). According to the latest American Society of Clinical Oncology (ASCO)/College of American Pathologists (CAP) guidelines, HER2-zero is defined as no HER2 protein expression by immunohistochemistry (IHC 0) with negative in situ hybridization (ISH), while HER2-low is defined as low-level HER2 protein expression (IHC 1+ or IHC 2+/ISH−) (8). Accumulating evidence has identified that approximately 30–50% of TNBC patients exhibit HER2-low expression (9,15), and emerging studies have suggested that HER2 status (zero vs. low) may be associated with distinct chemotherapy sensitivity and prognosis in TNBC. However, current studies on the correlation between HER2 subtypes and TNBC NAC response remain controversial (16-18), and clinical treatment of TNBC still does not distinguish between HER2-zero and HER2-low statuses. More importantly, the underlying biological mechanisms by which HER2 status influences chemotherapy sensitivity in TNBC have not been fully elucidated, which restricts the clinical application of HER2 subtype stratification in TNBC treatment decision-making.

The differential chemotherapy sensitivity between HER2-zero and HER2-low TNBC is closely linked to their distinct molecular subtype distribution and genomic landscapes, which form the core biological rationale for this study. In terms of PAM50 intrinsic subtypes, HER2-zero TNBC is predominantly enriched in the basal-like subtype (accounting for >70% of cases), which is characterized by high proliferation activity, high expression of basal epithelial markers (e.g., CK5/6, EGFR), and a lack of Luminal differentiation (15,19). Basal-like TNBC typically exhibits high sensitivity to anthracycline and taxane-based chemotherapy—the standard NAC regimen for TNBC—due to its high genomic instability and rapid cell cycle progression (15,20). In contrast, HER2-low TNBC has a relatively higher proportion of luminal androgen receptor (AR) subtype and mesenchymal subtype (20,21), with the former exhibiting low proliferation activity and the latter characterized by strong invasive and metastatic potential, both of which are associated with relatively lower chemotherapy sensitivity. At the genomic level, HER2-low TNBC is more frequently enriched in PIK3CA mutations (mutation rate –20% vs. –10% in HER2-zero TNBC) (21,22), which leads to abnormal activation of the PI3K/AKT/mTOR signaling pathway, resulting in chemotherapy resistance and reduced pCR rates after NAC. Additionally, recent metabolomic studies have shown that HER2-low TNBC has a distinct lipid metabolism profile, with enhanced de novo fatty acid synthesis and lipid droplet accumulation, which not only promotes tumor cell survival under chemotherapy stress but also inhibits the cytotoxic effect of chemotherapeutic drugs (23). In contrast, HER2-zero TNBC exhibits a more glycolysis-dependent metabolic phenotype, which is more susceptible to chemotherapy-induced energy metabolism disruption (23). These biological differences between HER2-zero and HER2-low TNBC provide a solid molecular basis for their differential response to NAC, and highlight the necessity of incorporating HER2 subtype stratification into TNBC NAC efficacy prediction models.

Although nomograms have been recognized as user-friendly and intuitive predictive tools for clinical outcomes, with superior individualized risk assessment compared with single biomarkers or staging systems (10), existing TNBC NAC predictive nomograms have not considered the biological characteristics of HER2 subtypes. Moreover, the interaction between HER2 status and other clinically relevant biomarkers (e.g., AR, P53, Ki-67) in TNBC NAC response has not been investigated. Both the residual cancer burden (RCB) and Miller-Payne (MP) grading system are the preferred modalities for pathological evaluation of post-NAC specimens and are widely used for survival prediction (11), but they are post-treatment assessment tools and cannot guide pre-treatment NAC regimen optimization. Therefore, it is urgent to establish a pre-treatment predictive model that integrates HER2 subtype stratification and key biological biomarkers for TNBC NAC efficacy.

The current study aims to appraise the influence of HER2 status (zero vs. low) on pCR and DFS in TNBC patients receiving NAC, and to develop and internally validate a nomogram based on HER2 status and key biomarkers (AR, P53, Ki-67) for predicting NAC efficacy in TNBC. We hypothesize that HER2-zero TNBC has a higher pCR rate and better prognosis than HER2-low TNBC, and that the integration of HER2 subtype stratification can improve the predictive accuracy of the nomogram. This study is expected to fill the existing knowledge gap in TNBC NAC prediction, provide a practical and personalized predictive tool for clinical practice, and lay a biological foundation for HER2 subtype-based individualized treatment of TNBC. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0152/rc).

Methods

Data collection and patient selection

It was a retrospective analysis of clinical data from patients with early TNBC who had received standard NAC and underwent standard surgery in the Department of Breast, Gansu Provincial Maternity and Child Health Hospital (Gansu Provincial Central Hospital), from January 2015 to December 2023. The inclusion criteria were as follows: (I) female; (II) TNBC confirmed by core needle biopsy with pathological and IHC verification; (III) completion of the planned anthracycline-taxane based NAC regimen followed by curative-intent surgery; (IV) availability of complete clinical, pathological, and follow-up data with a minimum follow-up of 6 months. Exclusion criteria included: (I) inflammatory, bilateral, lactating, or gestational BC; (II) presence of other severe comorbidities or additional malignant tumors; (III) intolerance to chemotherapy leading to early termination or regimen modification; (IV) referral from distant regions with incomplete follow-up data; (V) recurrent or metastatic TNBC at initial diagnosis. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Research and Ethics Committee of the Institutional Review Board of the Gansu Provincial Maternity and Child Health Hospital (Gansu Provincial Central Hospital) (No. [2025]GSFY Ethics[45]). Written informed consent was obtained from all participants.

Within 2 weeks after the end of standard NAC (12), the surgical procedure was performed according to the tumor location, clinical stage and the patient’s informed wish for breast conservation or mastectomy. The MP grading system was used to evaluate the pathological response of NAC by two senior pathologists who were blinded to the patients’ clinical characteristics and follow-up outcomes. pCR was strictly defined as the absence of invasive cancer components in both the primary breast tumor bed and regional lymph nodes, with only scattered ductal carcinoma in situ allowed (MP grade 5). DFS was defined as the duration from the first day of postoperative pathological diagnosis (confirmation of pCR or non-pCR status) to the first occurrence of any disease-related event, including local recurrence, distant metastasis, or cancer-related death; patients without disease-related events were censored at the last follow-up date (May 11, 2024). The standard NAC regimen was anthracycline (epirubicin/doxorubicin) combined with taxane (paclitaxel/docetaxel) drugs, with specific doses adjusted according to the patient’s body surface area and renal/hepatic function, and no other chemotherapy regimens were used in this study.

Clinical data collection

Clinical and pathological data were collected from the hospital’s electronic medical record system and pathological archive database, including age, menstrual status, tumor T stage (T), lymph node status (N), histological grade (Grade), lymphovascular invasion (LVI), and the expression status of Ki-67, HER2, AR, and P53. The expression status of Ki-67, HER2, AR and P53 was assessed through standardized IHC and ISH testing conducted by the hospital’s central pathology laboratory, with all tests performed in accordance with the 2023 ASCO/CAP HER2 testing guidelines for BC. The mutant type of P53 (nuclear staining in ≥10% of tumor cells) was defined as P53 positive, while the wild type (nuclear staining in <10% of tumor cells) was defined as P53 negative. HER2-zero was defined as IHC 0 with negative ISH, and HER2-low was defined as IHC 1+ or IHC 2+ with negative ISH.

All pathological evaluations were independently performed by two senior pathologists with more than 10 years of experience in breast pathological diagnosis, who were blinded to the patients’ clinical information, NAC response, and follow-up results. Consensus was reached through a joint review by the two pathologists and a third senior pathologist in case of diagnostic disagreements. Given that Ki-67 expression in TNBC is frequently above 20%, this study utilized the ROC curve derived from all 122 cases to determine an optimal cutoff value of 60% for Ki-67 expression. Accordingly, TNBC patients were categorized into a high Ki-67 expression group (>60%) and a low Ki-67 expression group (≤60%). Follow-up was carried out via regular outpatient visits (every 3 months for the first 2 years, every 6 months for 3–5 years, and annually thereafter) or telephone interviews, and all follow-up procedures were completed by two independent researchers who were blinded to the pathological evaluation results.

Statistical analysis

Sample size estimation was performed based on the principles of multivariate logistic regression modeling for predictive nomograms, with the primary outcome being pCR after NAC. The minimum events-per-variable (EPV) ratio required for stable multivariate regression analysis is 10:1 (24), and the ideal EPV ratio recommended for modern predictive modeling in heterogeneous tumor populations (e.g., TNBC) is 20:1 (25). In this study, the pCR was observed in 50 patients, and four independent predictive factors (HER2 status, AR, P53, Ki-67) were identified for nomogram construction through univariate Logistic regression screening (P<0.1), resulting in an EPV ratio of 12.5:1. This ratio meets the minimum statistical stability requirement for multivariate Logistic regression but does not reach the ideal 20:1 standard; sample size expansion was limited by the single-center retrospective study design and the incidence of TNBC. All clinical, pathological, and follow-up data of the 122 patients included in this study are complete with no missing values; therefore, no missing data imputation method was used.

Statistical analyses were conducted using SPSS 27.0. Categorical variables were analyzed using Chi-squared test or Fisher’s exact test. Survival curves of DFS were plotted using the Kaplan-Meier method, and statistical analysis was performed with the Log-rank test. Univariate Logistic regression analysis was first used to screen candidate predictive factors for pCR, with factors with a two-sided P value <0.1 included in the subsequent multivariate logistic regression analysis. Based on the results of multivariate analysis, the “rms” package of R (version 3.4.4) along with Zstats v0.90 (www.medsta.cn/software) was used to construct nomogram. Internal validation of the nomogram was performed using the non-parametric bootstrap resampling method (1,000 repetitions), which is a standard method for assessing internal optimism and reducing overfitting bias in predictive models (26). The bootstrap-corrected AUC was calculated to evaluate the adjusted discriminative ability of the nomogram after accounting for overfitting, and the calibration curve was corrected using bootstrap resampling to further verify the consistency between predicted and observed probabilities. The predictive performance of the model was evaluated using ROC curves, calibration curves, and decision curve analysis (DCA). All statistical tests were two-sided, and a P value <0.05 was considered statistically significant for all analyses. The present model was constructed based on available clinical and pathological parameters; some important biomarkers and treatment-related factors were not included due to data availability constraints.

Results

A total of 122 consecutive TNBC patients who underwent NAC and surgery were retrospectively enrolled (Table 1). Univariate analysis demonstrated no significant associations (≥0.1) between pCR and age, menstruation, Grade, T, N, and LVI (Table 2). Significant univariate predictors (P<0.10) included HER2, Ki-67, P53, and AR (Table 2). The pCR rate in the HER2‑zero group was significantly higher than that in the HER2‑low group (48.57% vs. 30.77%, P=0.048, marginal significance). In HER2-low group of TNBC patients, the pCR rates were 26.5% and 38.9% in AR-negative and AR-positive patients, respectively, with no statistically significant difference (χ²=0.852, P=0.35; Table 3). The pCR rates were 20.7% in P53-negative patients and 43.5% in P53-positive patients, and the difference was not statistically significant (χ²=3.127, P=0.07; Table 3). Regarding Ki-67 expression, the pCR rates were 25.0% in patients with low Ki-67 expression and 34.4% in those with high Ki-67 expression, with no statistically significant difference (χ²=0.508, P=0.47; Table 3). In the HER2-zero TNBC subgroup, high Ki-67 expression (>60%, the optimal cutoff value determined by ROC curve analysis in this cohort) was significantly associated with a higher pCR rate (P=0.009), suggesting that Ki-67 proliferation status has a subtype-specific predictive value for NAC efficacy in TNBC. In HER2-zero expression TNBC patients, the pCR rates were 40.4% for AR-negative patients and 72.2% for AR-positive patients, with a statistically significant difference (χ²=5.426, P=0.02; Table 3). The pCR rates were 41.3% in P53-negative patients and 62.5% in P53-positive patients, showing no statistically significant difference (χ²=2.836, P=0.09; Table 3). Regarding Ki-67 expression levels, the pCR rates were 22.2% in patients with low Ki-67 expression and 57.7% in those with high Ki-67 expression, with a statistically significant difference (χ²=6.735, P=0.009; Table 3).

Kaplan-Meier survival analysis showed that DFS probability of the pCR group was significantly higher than that of the no-pCR group [hazard ratio (HR): 0.000, 95% confidence interval (CI): 0.000–Inf, log-rank test P<0.001; Figure 1]. The DFS probability of the HER2-zero group was higher than that of the HER2-low group (HR: 0.407, 95% CI: 0.164–1.012, log-rank test P=0.043; Figure 1). Notably, the CI of the HR includes values close to 1, suggesting the prognostic difference may be unstable and limited by the small sample size. In both HER2-zero expression and HER2-low expression TNBC patients, there were no statistically significant differences in DFS between AR-positive (χ²=0.589, P=0.44; Figure 2) and AR-negative subgroups (χ²=2.379; P=0.12; Figure 2). In HER2-zero expression TNBC patients, the difference in DFS between P53-positive and P53-negative patients was statistically significant (χ²=5.351, P=0.02; Figure 2). In contrast, no statistically significant difference in DFS was observed between P53-positive and P53-negative patients with HER2-low expression TNBC (χ²=0.001, P=0.97; Figure 2). Additionally, neither in HER2-zero expression nor HER2-low expression TNBC patients was there a statistically significant difference in DFS between patients with low Ki-67 expression (χ²=0.040, P=0.84; Figure 2) and those with high Ki-67 expression (χ²=0.244; P=0.62; Figure 2).

Figure 1 Kaplan-Meier curve of DFS according to efficacy and HER2: (A) The relationship between efficacy and DFS; (B) the relationship between HER2 and DFS. DFS, disease-free survival; HER2, human epidermal growth factor receptor 2; pCR, pathological complete response.

Figure 2 The disease-free survival curves of triple-negative breast cancer patients with different HER2 expression statuses regarding the expression of AR, P53, and Ki-67. (A,B) The survival curves of patients with zero and low HER2 expression based on the expression of Ki-67; (C,D) the survival curves of patients with zero and low HER2 expression based on the expression of P53; (E,F) the survival curves of patients with zero and low HER-2 expression based on the AR. AR, androgen receptor; HER2, human epidermal growth factor receptor 2.

Based on the result of multivariate logistic regression analysis, the nomogram was constructed to predict pCR (Figure 3). Internal validation via bootstrap resampling (1,000 repetitions) showed that the corrected AUC of the nomogram for predicting pCR was 0.76 (95% CI: 0.68–0.85), which was slightly lower than the original AUC (0.78) but still indicated moderate discriminative ability, suggesting a low risk of overfitting in the model within the current study population. The AUC of ROC curve for predicting pCR was 0.78 (95% CI: 0.70–0.87; Figure 4). The calibration curve for pCR showed close alignment with the ideal 45-degree line (Figure 5). Quantitative calibration analysis showed that the calibration curve had an intercept of 0.08 (95% CI: −0.12 to 0.28) and a slope of 0.94 (95% CI: 0.71–1.17), values that are close to the ideal calibration parameters (intercept =0, slope =1), indicating high consistency between the predicted pCR probabilities and actual observed probabilities. The mean absolute error (MAE) between predicted and observed probabilities was 0.06, reflecting a small average deviation of the model’s predictions. Formal Hosmer-Lemeshow test yielded a non-significant P value (P=0.84; Figure), further confirming the good calibration performance of the new model. The DCA curve demonstrated that the nomogram provided a favorable net benefit in this cohort compared to both “treat all” and “treat none” strategies for threshold probabilities between 20% and 100% (Figure 6). The relatively wide CIs in the multivariable model should be noted, which may be attributed to the limited sample size and subgroup stratification, indicating a degree of estimation uncertainty (Table 2). The nomogram was internally validated using bootstrap resampling (1,000 repetitions). No independent external validation cohort was included, which may limit the generalizability of the model.

Figure 3 Nomogram for predicting pCR in TNBC based on HER2 status. Clinical interpretation: (I) locate the value of each predictive factor (menstrual status, T stage, P53, HER2, AR, Ki-67) for an individual TNBC patient on the corresponding axis; (II) draw a vertical line upward to the points axis to obtain the score for each factor; (III) sum the scores of all factors to get the total points; (IV) draw a vertical line downward from the total points axis to the risk axis to obtain the predicted pCR probability for the patient. Note: This nomogram is for exploratory research reference only and not for direct clinical decision-making. AR, androgen receptor; HER2, human epidermal growth factor receptor 2; pCR, pathological complete response; T, tumor; TNBC, triple-negative breast cancer.

Figure 4 ROC curve of the nomogram for predicting pCR in TNBC based on HER2 status. Clinical interpretation: the area under the ROC curve (AUC =0.78, 95% CI: 0.70–0.87) indicates a moderate to good discriminative ability of the model to distinguish TNBC patients who will achieve pCR from those who will not after NAC. CI, confidence interval. AUC, area under the ROC curve; CI, confidence interval; HER2, human epidermal growth factor receptor 2; NAC, neoadjuvant chemotherapy; pCR, pathological complete response; ROC, receiver operating characteristic; TNBC, triple-negative breast cancer.

Figure 5 Calibration curve of the nomogram for predicting pCR in TNBC based on HER2 status. Clinical interpretation: the close alignment of the actual observation curve (solid line) with the ideal 45° prediction curve (dashed line) indicates good calibration of the model, meaning the predicted pCR probabilities generated by the nomogram are highly consistent with the actual observed pCR rates in clinical practice (Hosmer-Lemeshow test, P=0.84). HER2, human epidermal growth factor receptor 2; pCR, pathological complete response; TNBC, triple-negative breast cancer.

Figure 6 Decision curve analysis of the nomogram for predicting pCR in TNBC based on HER2 status. Clinical interpretation: the blue curve (nomogram) is above the “treat all” (red line) and “treat none” (blue line) curves at a threshold probability of 20% to 100%, indicating that the model provides a higher clinical net benefit when used to predict pCR in this range; that is, the model can help clinicians identify TNBC patients who are more likely to benefit from NAC with fewer misclassifications in this threshold range. HER2, human epidermal growth factor receptor 2; NAC, neoadjuvant chemotherapy; pCR, pathological complete response; TNBC, triple-negative breast cancer.

Discussion

Studies have shown that TNBC patients have a higher pCR rate compared to other molecular subtypes (13). In early-stage TNBC, achieving pCR is associated with significantly improved DFS (14). The findings of this study also confirm that DFS is longer in patients who achieve pCR than in those who do not achieve pCR. HER2 is a crucial driver gene and prognostic biomarker for BC, providing an important reference for selecting therapeutic regimens and evaluating prognosis in BC patients (27). With the publication of results from the DESTINY-Breast04 trial, HER2-low expression has emerged as a research hotspot in the field of BC (28). Yi et al. demonstrated that HER2-low expression does not affect the pCR in BC, and there is no statistically significant difference in prognosis between patients with HER2-low expression and those with HER2-zero expression (16). This study demonstrates that HER2 status significantly influences the pCR of TNBC patients undergoing NAC. The HER2-zero group exhibits a higher pCR rate compared to HER2-low group, indicating that TNBC patients with HER2-zero status may respond more favorably to NAC. This observation may be attributed to the following aspects. Firstly, the molecular mutation profile may represent an underlying mechanism (19). Second, the immune micro-environment may also play a crucial role (15). Third, the methods and criteria for detecting HER2 status vary across studies (20). Among TNBC patients with HER2-zero expression, there were statistically significant differences in the pCR rates between AR-negative and AR-positive patients, as well as between Ki-67-negative and Ki-67-positive patients. In contrast, the expressions of AR, P53 and Ki-67 were not correlated with the pCR rates in TNBC patients with HER2-low expression.

Studies have found no statistically significant difference in DFS between the HER2-low expression group and the HER2-zero expression group (21). However, multiple other studies suggest that HER2-low patients may have a better DFS than HER2-zero patients after receiving NAC. For example, a single-center study showed that the 3-year DFS of HER2-low patients was significantly higher than that of HER2-zero patients (17). Meta-analyses also support the association between HER2-low expression and longer DFS (18). In TNBC, the DFS difference between HER2-low and HER2-zero is inconsistent. Some study has found no significant difference (22), while other suggests that HER2-low patients have a slightly longer DFS, though not reaching statistical significance (23). Additionally, achieving pCR is closely associated with improved DFS, regardless of HER2 status (29). This study revealed a higher DFS probability in the HER2-zero TNBC group than in the HER2-low group, contradicting prior meta-analyses. Key contributors include marked heterogeneity in study populations, methods, and insufficient statistical power, alongside unadjusted confounders, TNBC molecular subtype diversity, and variable follow-up durations. Current robust evidence from large-sample meta-analyses confirms no significant DFS difference associated with HER2-low status overall. Future research should prioritize standardized definitions, larger samples, and extended follow-up to clarify its prognostic value. Among patients with HER2-zero expression, there was a statistically significant difference in DFS between P53-positive and P53-negative cases (χ²=5.351, P=0.02). In all other subgroups regardless of HER2 expression status, the expressions of AR, P53 and Ki-67 were not correlated with DFS.

Multivariate analysis of this study revealed that HER2 was an independent influencing factor for pCR. Based on the analysis of the multivariate Logistic regression model, we constructed a nomogram model to predict the pCR in TNBC patients, aiming to provide a guiding basis for clinical diagnosis and treatment. The ROC curve and calibration curve for pCR showed the model’s promising preliminary predictive discriminative ability and reliability. DCA integrates the preferences of patients or decision-makers into the analysis, aligning with the practical needs of clinical decision-making. The result of DCA indicates that using the nomogram for clinical decision-making could lead to better patient outcomes by identifying those who would benefit most from additional interventions, while avoiding unnecessary treatments for low-risk patients.

In the present study, HER2‑zero TNBC showed a higher pCR rate and longer DFS compared with HER2‑low TNBC, with marginal statistical significance. The HR for DFS had a wide CI crossing near 1, indicating the result may be unstable due to the relatively small sample size. These findings are inconsistent with several large cohort studies and meta‑analyses, which reported no significant prognostic difference between HER2‑low and HER2‑zero TNBC. The discrepancy may be attributed to single‑center design, limited sample size, heterogeneity in detection methods, and baseline characteristics. Therefore, our results should be considered as preliminary observational findings rather than definitive clinical evidence.

This nomogram can be directly applied in clinical research as a preliminary exploratory tool: clinicians can calculate the predicted probability of pCR for patients based on their HER2 status, AR, P53, Ki-67 expression, and menstrual status using the nomogram. This nomogram can provide a preliminary quantitative reference for clinicians to assess the individual pCR potential of TNBC patients after NAC, and the predicted pCR probability can be used as a tentative indicator for clinical decision-making discussion, rather than a direct basis for formulating chemotherapy intensification or targeted therapy strategies. Notably, this nomogram has not yet undergone prospective multicenter validation and has not been incorporated into clinical treatment guidelines for TNBC. Therefore, it is not ready for direct clinical decision-making at present, and any clinical treatment adjustment based on the model’s prediction needs to be combined with comprehensive clinical evaluation and multidisciplinary discussion. Additionally, for patients with HER2-zero and P53-positive status, follow-up strategies can be appropriately simplified for research reference only; whereas HER2-low patients require enhanced long-term follow-up, which also needs to be verified by more clinical evidence.

The intriguing differential pCR rates and DFS between HER2-zero and HER2-low TNBC subgroups observed in our study is a critical discovery that supplements the understanding of TNBC heterogeneity. Unlike previous studies with conflicting results, our retrospective analysis with a median 48-month follow-up clearly demonstrated that HER2 status (zero vs. low) is an independent factor associated with NAC response and prognosis in TNBC. This finding suggests that the traditional “HER2-negative” classification for TNBC is overly simplistic, and HER2 subtyping may serve as a potential biomarker for personalized NAC in TNBC, providing a new direction for subsequent clinical research on TNBC stratification and targeted therapy exploration for HER2-low subgroups.

The core innovation of this study is reflected in three interconnected dimensions. First, the research perspective is unique, which opens new avenues for understanding the heterogeneity of TNBC. Second, the research outcomes extend beyond theoretical exploration, having been successfully translated into a practical personalized prediction tool that can be directly applied in clinical settings. This tool enables quantitative assessment of individual patient risks and assists clinicians in making diagnostic and therapeutic decisions, thereby fully realizing the value transition from fundamental theoretical exploration to direct clinical application. Third, this study further analyzed the associations between the expressions of AR, P53 and Ki-67 and both pCR and prognosis among TNBC patients with different HER2 expression statuses, thereby enabling a more in-depth understanding of the relationship between HER2 expression status and TNBC prognosis.

Several important limitations of this study should be acknowledged, most of which are related to the methodological and statistical framework, and these limitations may affect the generalizability and clinical application of the constructed nomogram. First, this is a single-center retrospective study with a relatively small sample size, which leads to an EPV ratio of 12.5:1 for the multivariate logistic regression model. Although this ratio meets the minimum statistical stability requirement (10:1) for predictive modeling, it is lower than the ideal 20:1 EPV ratio recommended for modern predictive modeling in heterogeneous tumor populations such as TNBC (25). This relatively low EPV ratio may increase the risk of overfitting and limit the stability of the model, and the inclusion of multiple predictors and further subgroup stratification by HER2 status combined with AR, P53, and Ki-67 expression substantially increased the risk of model overfitting and unstable coefficient estimates. The relatively wide CIs in the multivariable logistic regression analysis also reflect the limited statistical power and precision of estimates in this study. The observed differences in pCR and DFS between HER2‑zero and HER2low groups were of marginal significance (P=0.048 and P=0.043), and the HR CI for DFS included values near 1. This suggests the findings may be unstable and strongly influenced by the limited sample size, which further constrains the reliability and generalizability of the results. Second, the follow-up duration of the study is relatively short, and the study only focused on DFS as the prognostic outcome, without OS, which is a more clinically meaningful long-term prognostic endpoint. In addition, although the study included core clinical and pathological factors, it did not incorporate other potential predictive factors such as TILs, BRCA1/2 mutation status, and radiomics features, which have been shown to be associated with NAC response in TNBC (8,12), and the exclusion of these factors may limit the predictive accuracy of the nomogram. Third, only standard anthracyclinetaxane NAC regimens were included, and the generalizability to other regimens remains unclear. Additionally, the most significant design flaw of this study is the absence of an independent external validation cohort. Although Bootstrap resampling (1,000 repetitions) was used for internal validation and showed a low risk of overfitting within the current study population, internal validation alone cannot substitute for external validation, which is the gold standard for verifying the generalizability and clinical utility of predictive models (26,30). Bootstrap resampling only assesses the internal optimism of the model and cannot reflect the model’s performance in different patient populations. In addition, we did not perform an internal-external validation via temporal or geographical splitting of the current dataset, which is a feasible alternative for single-center studies with limited sample size (30). Another important limitation is that the nomogram does not incorporate several clinically and biologically relevant variables known to affect pCR and survival, including specific chemotherapy regimens, platinum use, TILs, germline BRCA mutation status, molecular TNBC subtypes, comorbidities, and treatment delays. These factors are strongly associated with NAC response and prognosis, and their absence may lead to residual confounding in the predictive model. This may limit the comprehensiveness and generalizability of the current model. Finally, this nomogram was established and internally validated only by bootstrap resampling, without an independent external validation cohort. As a prediction model derived from a single‑center retrospective cohort, it is prone to optimistic performance estimates, and its generalizability to other populations, regions, or treatment regimens remains unclear. Therefore, this model is still exploratory and hypothesis‑generating rather than ready for routine clinical use. Future multi‑center, prospective studies with diverse populations and independent external validation are urgently needed to verify the predictive accuracy, stability, and generalizability of this nomogram before clinical implementation. Future model development should integrate platinum administration, TILs, BRCA status, TNBC molecular subtyping, comorbidities, and treatment timing to minimize residual confounding and further improve predictive accuracy and clinical applicability.

To address the above limitations and further optimize the nomogram, we have developed a detailed and feasible follow-up research plan. First, we will initiate a multicenter prospective cohort study in collaboration with multiple tertiary hospitals specializing in breast disease in China, aiming to collect a large sample of TNBC patients (target EPV ratio ≥20:1) to validate the constructed nomogram. This multicenter study will include patients with different clinical characteristics, treatment regimens, and geographical regions, and will perform an independent external validation to verify the generalizability of the model. Second, we will perform an internal-external validation via temporal splitting of the current single-center dataset (2015–2020 as the modeling set, 2021–2023 as the internal validation set) in the subsequent analysis, which will provide additional evidence for the model’s performance. Third, we will extend the follow-up duration of the current study population to collect OS data and other long-term prognostic endpoints, and incorporate additional predictive factors (e.g., TILs, BRCA1/2 mutation, radiomics features) into the nomogram to further improve its predictive accuracy. Fourth, we will expand the study population to include TNBC patients receiving platinum-based and other NAC regimens, to investigate the applicability of the nomogram in different chemotherapy regimens and to explore the interaction between HER2 status and chemotherapy regimens on NAC response.

In addition, future studies will also focus on exploring the molecular mechanisms underlying the differential NAC response between HER2-zero and HER2-low TNBC, such as the immune microenvironment, genomic mutations, and metabolic reprogramming, which will provide a more solid biological basis for the clinical application of HER2 subtype stratification in TNBC.

Conclusions

HER2 status plays a critical role in the prognostic evaluation of TNBC, as patients with HER2-zero status are more likely to achieve a pCR after NAC. The nomogram constructed in this study showed favorable preliminary discriminative ability for predicting pCR in TNBC patients. Given the limited sample size and risk of overfitting, this tool may serve as a preliminary user-friendly reference for clinicians to individualize efficacy assessment, pending further validation in larger multi-center cohorts.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0152/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0152/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0152/prf

Funding: This study was supported by General Project of Gansu Joint Research Fund (No. 25JRRA1241).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0152/coif). The authors have no conflicts of interest to declare.

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. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Research and Ethics Committee of the Institutional Review Board of the Gansu Provincial Maternity and Child Health Hospital (Gansu Provincial Central Hospital) (No. [2025]GSFY Ethics[45]). Written informed consent was obtained from all participants.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Filho AM, Laversanne M, Ferlay J, et al. The GLOBOCAN 2022 cancer estimates: Data sources, methods, and a snapshot of the cancer burden worldwide Int J Cancer 2025;156:1336-46.
2. Chaudhary LN, Wilkinson KH, Kong A. Triple-Negative Breast Cancer: Who Should Receive Neoadjuvant Chemotherapy? Surg Oncol Clin N Am 2018;27:141-53. [Crossref] [PubMed]
3. Xia LY, Hu QL, Zhang J, et al. Survival outcomes of neoadjuvant versus adjuvant chemotherapy in triple-negative breast cancer: a meta-analysis of 36,480 cases World J Surg Oncol 2020;18:129.
4. Xia LY, Hu QL, Zhang J, et al. Survival outcomes of neoadjuvant versus adjuvant chemotherapy in triple-negative breast cancer: a meta-analysis of 36,480 cases World J Surg Oncol 2020;18:129.
5. Wu J, Yan X, Fang J, et al. Platinum-based neoadjuvant chemotherapy in triple-negative breast cancer: An updated systematic review and meta-analysis Medicine (Baltimore) 2025;104:e44740.
6. Li ZY, Zhang Z, Cao XZ, et al. Platinum-based neoadjuvant chemotherapy for triple-negative breast cancer: a systematic review and meta-analysis J Int Med Res 2020;48:300060520964340.
7. Zhao Y, Schaafsma E, Cheng C. Gene signature-based prediction of triple-negative breast cancer patient response to Neoadjuvant chemotherapy Cancer Med 2020;9:6281-95.
8. Baez-Navarro X, van Bockstal MR, Jager A, et al. HER2-low breast cancer and response to neoadjuvant chemotherapy: a population-based cohort study Pathology 2024;56:334-42.
9. de Moura Leite L, Cesca MG, Tavares MC, et al. HER2-low status and response to neoadjuvant chemotherapy in HER2 negative early breast cancer Breast Cancer Res Treat 2021;190:155-63.
10. Liao XW, Gao JB, Sun H, et al. Prediction of neoadjuvant chemotherapy efficacy and prognostic biomarker analysis in patients with triple-negative breast cancer Front Pharmacol 2025;16:1553831.
11. Xu X, Zhao W, Liu C, et al. The residual cancer burden index as a valid prognostic indicator in breast cancer after neoadjuvant chemotherapy BMC Cancer 2024;24:13.
12. Telli ML, Hellyer J, Audeh W, et al. Homologous recombination deficiency (HRD) status predicts response to standard neoadjuvant chemotherapy in patients with triple-negative or BRCA1/2 mutation-associated breast cancer Breast Cancer Res Treat 2018;168:625-30.
13. Khamajeet A, Cairncross L, Molabe H, et al. Pathological response of breast cancer to neoadjuvant chemotherapy at a single tertiary centre S Afr J Surg 2025;63:62-7.
14. de Graaf M, Gielen RCAM, Balduzzi S, et al. Pathological complete response and survival after neoadjuvant chemotherapy in patients with stage I TNBC: a registry-based study ESMO Open 2025;10:105923.
15. Liu JJ, Zhang Y, Zhang SC, et al. Analysis of factors influencing the efficacy of NAC and prognosis between HER2-zero and HER2-low HR negative breast cancer Front Cell Dev Biol 2024;12:1417271.
16. Yi Xiaoli, Hu Shasha, Zhang Yan. The influence of low HER-2 expression on the efficacy and prognosis of neoadjuvant therapy for breast cancer Chinese Journal of Breast Disease (Electronic Edition) 2023;17:340-6.
17. Qiao W, Guo W, Liu Q, et al. Pathological complete response and prognosis after neoadjuvant chemotherapy in patients with HER2-low breast cancer Ann Diagn Pathol 2023;64:152125.
18. Xia LY, Cao XC, Yu Y. Survival outcomes in HER2-low versus HER2-zero breast cancer after neoadjuvant chemotherapy: a meta-analysis World J Surg Oncol 2024;22:106.
19. Yi X, Hu S, Ma M, et al. Effect of HER2-low expression on neoadjuvant efficacy in operable breast cancer Clin Transl Oncol 2024;26:880-90.
20. Ma Y, Jiao D, Zhang J, et al. HER2-Low Status Was Associated With Better Breast Cancer-Specific Survival in Early-Stage Triple-Negative Breast Cancer Oncologist 2024;29:e309-18.
21. Yi X, Hu S, Ma M, et al. Effect of HER2-low expression on neoadjuvant efficacy in operable breast cancer Clin Transl Oncol 2024;26:880-90.
22. Morante Z, Ferreyra Y, Otoya I, et al. Influence of HER2 Changes on Survival Outcomes After Neoadjuvant Chemotherapy in Peruvian Patients With Triple-Negative Breast Cancer Breast J 2025;2025:3770655.
23. Shi Z, Liu Y, Fang X, et al. Efficacy and prognosis of HER2-Low and HER2-Zero in triple-negative breast cancer after neoadjuvant chemotherapy Sci Rep 2024;14:16899.
24. Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors Stat Med 1996;15:361-87.
25. Vickers AJ, Elkin EB. Decision curve analysis: a novel method for evaluating prediction models Med Decis Making 2006;26:565-74.
26. Steyerberg EW, Harrell FE Jr, Borsboom GJ, et al. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis J Clin Epidemiol 2001;54:774-81.
27. Bogdanovska-Todorovska M, Petrushevska G, Janevska V, et al. Standardization and optimization of fluorescence in situ hybridization (FISH) for HER-2 assessment in breast cancer: A single center experience Bosn J Basic Med Sci 2018;18:132-40.
28. Modi S, Jacot W, Yamashita T, et al. Trastuzumab Deruxtecan in Previously Treated HER2-Low Advanced Breast Cancer N Engl J Med 2022;387:9-20.
29. Yıldırım S, Başoğlu T, Doğan A, et al. The impact of carboplatin on pathologic complete response and survival based on HER2 low and HER2 zero status in triple negative breast cancer patients receiving neoadjuvant chemotherapy: a multicenter real-world analysis BMC Cancer 2025;25:833.
30. Steyerberg EW, Vickers AJ, Cook NR, et al. Assessing the performance of prediction models: a framework for traditional and novel measures Epidemiology 2010;21:128-38.