Breast cancer remains the most commonly diagnosed malignancy and the leading cause of cancer-related mortality among women worldwide (Bray et al., 2024). Triple-negative breast cancer (TNBC), characterized by the absence of estrogen receptor, progesterone receptor, and HER2 amplification, represents a particularly aggressive subtype with limited therapeutic options and a heightened propensity for early recurrence (Bianchini et al., 2016; Foulkes et al., 2010). Within this heterogeneous disease, premenopausal patients often exhibit distinct clinicopathological and molecular features, including enriched immunogenic signatures and homologous recombination deficiency, which may influence both tumour behaviour and therapeutic response (Azim et al., 2012; Bareche et al., 2018).

Anthracycline and taxane-based chemotherapy remains the cornerstone of adjuvant treatment for early-stage TNBC (Liedtke et al., 2008). However, intrinsic and acquired chemoresistance significantly undermines its efficacy, leading to disparate clinical outcomes (Gonzalez-Angulo et al., 2011; Wahba and El-Hadaad, 2015). The mechanisms underlying chemoresistance are multifactorial, encompassing tumour-intrinsic factors such as apoptotic evasion, enhanced DNA damage repair, and cancer stem cell persistence, as well as microenvironmental influences including immune contexture and stromal composition (Holohan et al., 2013; Shibue and Weinberg, 2017; Gottesman, 2002).

Emerging evidence underscores the prognostic significance of stromal tumour-infiltrating lymphocytes (sTILs) in TNBC, where higher densities correlate with improved pathological complete response and survival following chemotherapy (M Geurts et al., 2024; Salgado et al., 2015). Concurrently, The biological and clinical significance of HER2-low breast cancer, defined as immunohistochemistry (IHC) 1+ or IHC 2+ with negative in situ hybridization (FISH), has recently come into focus. Supported by evidence from clinical trials of novel anti-HER2 antibody–drug conjugates, this subset has emerged as a therapeutically actionable entity (Saura et al., 2025; Modi et al., 2025). Growing preliminary evidence further suggests that HER2-low tumours may represent a distinct biological subtype with unique therapeutic vulnerabilities (Tarantino et al., 2020; Agostinetto et al., 2021). Recent studies, including those by Tarantino P and Schettini F et al., have begun to delineate the underlying biology of HER2-low breast cancer across subtypes, including triple-negative breast cancer (Tarantino et al., 2023; Schettini et al., 2025).

Despite these advances, validated biomarkers for stratifying chemosensitivity in premenopausal TNBC remain scarce. Most existing prognostic tools neither account for the unique biology of this demographic nor integrate readily accessible biomarkers reflective of tumour-immune interplay (Sparano et al., 2018; Parker et al., 2009). Therefore, we aimed to develop and validate a clinically applicable nomogram that incorporates key clinicopathological and microenvironmental variables-specifically sTIL density and HER2-low status-to predict disease-free survival (DFS) in premenopausal women with TNBC treated with anthracycline-taxane chemotherapy.

In this large, multicentre retrospective study, we developed and validated a prognostic nomogram specifically tailored for premenopausal women with TNBC receiving anthracycline–taxane chemotherapy. The model-incorporating T stage, N stage, sTIL density, and HER2 immunohistochemistry (IHC) status-demonstrated high predictive accuracy and clinical utility, addressing a critical unmet need in this high-risk population.

Our findings confirm the prognostic relevance of the tumour immune microenvironment in this cohort. Low sTIL density was independently associated with worse disease-free survival, corroborating previous studies that have consistently linked lower sTIL levels with inferior outcomes in patients treated with DNA-damaging chemotherapies (M Geurts et al., 2024; Denkert et al., 2018; Adams et al., 2019). Clinically, these results align with the concept that tumours with low sTILs frequently exhibit an immunologically “cold” phenotype, which may inform future therapeutic strategies aimed at enhancing antitumour immunity in such patients (Sharma et al., 2017; Emens and Middleton, 2015).

We also observed that HER2 IHC 2+/fluorescence in situ hybridization (FISH)-negative status—but not IHC 1+—independently predicted poorer disease-free survival, challenging several earlier reports that associated HER2-zero expression with inferior outcomes in TNBC (Miglietta et al., 2025; Won and Spruck, 2020). This discrepancy may reflect variations in patient selection, treatment setting, or HER2 scoring methodology. Multivariable analysis under the standard HER2-low definition (IHC 1+ or IHC 2+/FISH-negative) indicated that the prognostic risk was predominantly driven by the IHC 2+/FISH-negative subgroup. This suggests that HER2-low TNBC may not represent a homogeneous entity and that IHC 2+/FISH-negative tumours could harbour distinct biological properties—a hypothesis that warrants further validation.

Growing preclinical and clinical evidence indicates that HER2-low breast cancers exhibit altered receptor trafficking, activation of compensatory signalling pathways (e.g., PI3K/AKT, MAPK), and specific tumour–stroma interactions, which may collectively compromise chemotherapy efficacy (Tarantino et al., 2020; Schettini et al., 2021; Modi et al., 2022). Translationaly, the recent efficacy of novel antibody–drug conjugates such as trastuzumab deruxtecan in HER2-low metastatic breast cancer underscores the clinical relevance of this subset and highlights its potential susceptibility to targeted approaches (Banerji et al., 2019; Gennari et al., 2021). Our data suggest that premenopausal patients with HER2-low TNBC—particularly those with IHC 2+/FISH-negative disease—represent a candidate population for adjuvant antibody–drug conjugate trials. Such a strategy could eventually shift the therapeutic paradigm from empirical chemotherapy toward biomarker-guided targeted treatment.

Moreover, TNBC is a molecularly heterogeneous disease, encompassing distinct subtypes such as those described by Lehmann BD et al. (basal-like, immunomodulatory, mesenchymal, etc.) (Lehmann et al., 2011). The prognostic role of sTILs and HER2-low status may vary across these subtypes, a nuance not captured in our retrospective analysis. Future prospective studies integrating transcriptomic subtyping with tumour microenvironment features are warranted to refine personalised prognostic models.

By restricting our analysis to premenopausal women, we minimised confounding from age-related biological differences, such as the higher prevalence of PIK3CA mutations in older patients (Ciriello et al., 2015; Cancer Genome Atlas Network, 2012). This enhances the biological coherence and clinical applicability of our model for this specific demographic. The nomogram’s performance (C-index >0.86) surpasses that of many conventional prognostic tools and genomic assays (Paik et al., 2004; Cardoso et al., 2019), supporting its utility in routine practice for risk-adapted therapeutic decision-making.

From a translational perspective, our model provides a rationale for personalised adjuvant strategies. High-risk patients-those with low sTILs and HER2-low tumours-may benefit from treatment intensification, including capecitabine maintenance (Wang et al., 2021; Masuda et al., 2017) or PARP inhibition in BRCA-mutated cases (J Tutt et al., 2021; Robson et al., 2017). Conversely, low-risk patients could be spared unnecessary toxicity. Furthermore, this tool may facilitate refined patient selection for clinical trials evaluating immunomodulatory agents or novel antibody–drug conjugates in the early-stage setting.

Several limitations of our study warrant consideration. First, its retrospective design introduces potential selection bias, despite multicentre enrolment. Second, all participants were of Chinese ancestry; external validation in diverse ethnic and geographic populations is essential to confirm generalisability (Eniu et al., 2008). Third, although patients were randomly assigned to training and validation sets, some baseline variables (e.g., family history, postoperative pathological T staging, Ki-67 index, and EGFR status) showed uneven distributions. While our model maintained high predictive performance across cohorts, such imbalances could theoretically introduce selection bias or affect generalizability. Future validation in larger, multi-ethnic cohorts is warranted to confirm the model’s stability. Fourth, although HER2 and sTIL assessments followed international guidelines and underwent central review, HER2 scoring reproducibility remains a challenge in TNBC, and sTIL evaluation is semi-quantitative. Future iterations could benefit from digital pathology, automated image analysis, or standardized immune-based assays to improve objectivity and reproducibility (Klauschen et al., 2018; Vanguri et al., 2022). Fifth, we did not incorporate germline BRCA status, a known determinant of TNBC prognosis and treatment response (Tutt et al., 2010; Tung et al., 2020). Integrating genomic and transcriptomic data in future prospective studies may further refine predictive accuracy and biological insight (Curtis et al., 2012; Pereira et al., 2016).

In conclusion, we have identified and validated low stromal TILs and HER2-low status as key biomarkers of intrinsic resistance to anthracycline-taxane chemotherapy in premenopausal TNBC. Beyond prognostication, these factors illuminate distinct biological pathways underlying treatment failure-an immunosuppressive microenvironment and the unique biology of HER2-low tumors. This mechanistic insight provides a compelling rationale for future pharmacological strategies: leveraging immunotherapy to inflame sTIL-low tumors and deploying novel ADCs to target HER2-low subtypes. Our findings thus offer a practical framework for patient stratification and pave the way for more personalized, mechanism-driven adjuvant therapy trials in this challenging disease.