This systematic narrative review was reported using the PRISMA 2020 framework for transparent identification, screening, eligibility assessment, and reporting of the literature. The review was designed to synthesize landmark molecular studies, contemporary clinical trials, validated genomic assays, and major clinical-practice guidelines rather than to perform a quantitative meta-analysis [9]. Searches were conducted through June 2025 (final search month: June 2025).

The information sources were PubMed/MEDLINE, Scopus, Web of Science, Cochrane Library, and Google Scholar. Reference lists of key molecular studies, clinical trials, and guideline documents were also screened. The search covered database inception through June 2025. The core concept combined breast cancer with molecular classification/intrinsic subtypes and clinically relevant treatment, biomarker, prognosis, or genomic-assay terms. Database-specific syntax was adapted to each platform; the reproducible search strings used for this revision are provided in Supplementary Table S1.

Eligible literature comprised peer-reviewed human studies addressing breast-cancer molecular classification, intrinsic subtypes, clinically validated biomarkers, genomic assays, targeted treatment, immunotherapy, antibody-drug conjugates, or precision-oncology approaches. Landmark molecular studies, randomized clinical trials, prospective validation studies, systematic reviews, and major international guidelines were prioritized. Non-human studies, duplicate publications, conference abstracts without adequate full text, and publications without direct relevance to the review question were excluded.

Titles and abstracts were screened against the predefined eligibility criteria, followed by full-text assessment of potentially relevant reports. Two reviewers independently screened records and extracted data. Disagreements were resolved by discussion and, when required, by a third reviewer. The extraction fields included study design, population, molecular classification or biomarker, clinical setting, treatment decision addressed, principal outcomes, and whether the evidence was prognostic, predictive, or treatment-eligibility defining.

At full-text review, reports were excluded when they did not address the review question or eligible population, used a non-clinical/research-only classification without a relevant management implication, represented a duplicate or overlapping cohort, lacked sufficient clinical information, or did not provide usable full text. The documented aggregate flow was 176 full-text reports assessed and 82 studies retained; therefore, 94 reports were excluded at full-text stage. The exclusion categories are listed in Supplementary Table S2.

Evidence was synthesized narratively. Particular attention was given to distinguishing intrinsic molecular subtype from IHC-defined clinical subtype, prognostic risk estimation from prediction of treatment benefit, and exploratory molecular associations from validated treatment-selection biomarkers. No quantitative meta-analysis was performed.

The original intrinsic taxonomy included Luminal A, Luminal B, HER2-enriched, basal-like, and normal-like patterns. Gene-expression profiling demonstrated reproducible biological patterns, while subsequent work established the clinical relevance and limitations of these intrinsic categories [1][2]. Importantly, intrinsic subtypes are transcriptomic constructs and are not interchangeable with routine IHC-defined clinical categories. The relationship between intrinsic molecular patterns and routine clinical testing is summarized in Table 1 and Figure 2. Intrinsic subtypes should not be reconstructed solely from ER, PR, HER2, and Ki-67 values; IHC and ISH remain the standard methods for defining clinical receptor status.

Table 1. Intrinsic molecular patterns and their relationship to routine clinical testing.

Figure 2. Relationship between intrinsic molecular classification, routine biomarkers, and treatment-specific biomarkers.

Hormone-receptor-positive, HER2-negative breast cancer constitutes the largest clinical group. Endocrine therapy is the foundation of treatment, but recurrence risk varies widely. The labels Luminal A and Luminal B are useful biological concepts rather than stand-alone treatment algorithms. Tumour size, nodal status, histological grade, receptor expression, proliferation, patient age/menopausal status, and genomic assay results may all contribute to adjuvant decisions. Table 1 summarizes the relationship between intrinsic molecular patterns and routine clinical testing.

CDK4/6 inhibitors have expanded adjuvant treatment options for selected higher-risk HR-positive/HER2-negative early breast cancer. In monarchE, abemaciclib plus endocrine therapy improved invasive disease-free survival in a high-risk, node-positive population, with sustained benefit at longer follow-up [10][11]. NATALEE subsequently demonstrated improved invasive disease-free survival with ribociclib plus a nonsteroidal aromatase inhibitor in a broader stage II/III HR-positive/HER2-negative population [12]. These results should be interpreted within their trial-defined populations rather than assigned generically to all Luminal B tumours.

HER2-enriched is a gene-expression subtype and is not defined by ERBB2 amplification. HER2-enriched tumours may show high expression of HER2-related genes and pathway activation, but transcriptomic HER2-enriched status is not equivalent to HER2 amplification or IHC 3+ status [13]. Therefore, HER2-enriched should not be used as a synonym for clinically HER2-positive disease. For treatment purposes, clinical HER2 status remains based on validated IHC and ISH testing.

For treatment purposes, clinical HER2 status remains based on validated IHC and ISH testing. HER2-positive early breast cancer may receive neoadjuvant and/or adjuvant HER2-directed therapy according to stage and risk, while advanced disease can be treated with HER2-directed agents according to treatment line and clinical context [14][15].

TNBC is defined clinically by the absence of ER and PR expression and lack of HER2 overexpression/amplification according to accepted testing criteria. It is biologically heterogeneous, with basal-like, mesenchymal, luminal-androgen-receptor, immune-related, and other proposed transcriptomic groups. These research classifications have not been established as routine treatment-selection biomarkers [16].

In early TNBC, pembrolizumab combined with neoadjuvant chemotherapy and continued after surgery improved event-free outcomes in the KEYNOTE-522 population [17][18]. PD-L1 expression did not define eligibility for the early-stage KEYNOTE-522 regimen. In metastatic TNBC, however, PD-L1 testing can guide pembrolizumab-containing therapy in appropriate patients, including the KEYNOTE-355 population with PD-L1 combined positive score thresholds specified by the trial and applicable guidance [19]. Thus, PD-L1 is a treatment-specific biomarker in defined metastatic settings and is not a universal molecular definition of TNBC or a requirement for early-disease pembrolizumab.

Germline BRCA1/2 status is another clinically validated biomarker. In OlympiA, adjuvant olaparib improved invasive and distant disease-free outcomes in patients with high-risk, HER2-negative early breast cancer carrying germline pathogenic or likely pathogenic BRCA1/2 variants after local treatment and chemotherapy [20]. This illustrates the distinction between a molecular alteration that identifies a treatment opportunity and an intrinsic subtype that describes tumour biology.

Sacituzumab govitecan has demonstrated activity in previously treated metastatic TNBC, including in the phase 3 ASCENT trial [21]. Its use should be described according to the disease setting and prior-treatment context rather than as therapy selected solely by an intrinsic TNBC molecular subclass.

HER2-low is not an intrinsic molecular subtype. It is a treatment-relevant IHC expression category generally referring to HER2 IHC 1+ or IHC 2+ with negative ISH in the context of clinical-trial eligibility. The ASCO/CAP guideline update retained the established HER2 testing framework while acknowledging the therapeutic relevance of trastuzumab deruxtecan in tumours with IHC 1+ or 2+ without amplification [14][22]. The biological heterogeneity of HER2-low tumours means that HER2-low should not be presented as a discrete intrinsic molecular class.

DESTINY-Breast04 demonstrated improved progression-free and overall survival with trastuzumab deruxtecan compared with physician’s-choice chemotherapy in previously treated HER2-low advanced breast cancer [22]. The clinical implication is treatment-specific: low HER2 expression can provide a target for an antibody-drug conjugate even when the tumour is not HER2-positive by conventional amplification/overexpression criteria.

Emerging evidence has also examined very low HER2 expression, including HER2-ultralow categories. These categories are treatment- and jurisdiction-specific and should not be interpreted as new intrinsic molecular subtypes. Their clinical use should follow the applicable regulatory indication, assay, and guideline.

Multigene assays have an important but specific role in selected HR-positive/HER2-negative early breast cancers. Their clinical roles differ by assay and population. Some assays are primarily prognostic, whereas prospective evidence supports chemotherapy-benefit assessment for specific assays in defined populations [4]-[8]. Table 2 summarizes the validated clinical context and the distinction between prognostic and predictive utility.

Table 2. Genomic assays: validated populations and prognostic versus predictive roles.

TAILORx showed that many women with HR-positive/HER2-negative, node-negative disease and intermediate 21-gene recurrence scores can avoid adjuvant chemotherapy without compromising invasive disease-free outcomes, although age and recurrence-score range modify the magnitude of chemotherapy benefit [4][5]. RxPONDER further demonstrated that menopausal status matters in node-positive disease: postmenopausal women with one to three positive nodes and recurrence scores of 25 or lower did not derive a clinically meaningful chemotherapy benefit, whereas premenopausal women did derive benefit [6]. These findings illustrate why genomic scores should not be interpreted independently of clinical and patient characteristics.

MINDACT provided prospective evidence for use of a 70-gene signature in selected patients with discordant clinical and genomic risk, particularly those with high clinical risk and low genomic risk [7]. Its principal contribution is risk stratification and chemotherapy decision support in its validated population, not a universal prediction of chemotherapy sensitivity.

Next-generation sequencing can identify alterations with therapeutic implications, particularly in advanced disease. Examples include PIK3CA alterations in HR-positive/HER2-negative breast cancer, ESR1 alterations associated with endocrine resistance, and AKT-pathway alterations. Germline BRCA1/2 status can identify candidates for PARP inhibition in appropriate settings. Molecular testing should therefore be linked to a specific therapeutic decision; broad genomic profiling does not automatically imply that every detected alteration is clinically actionable.

Precision oncology also increasingly includes circulating tumour DNA, minimal residual disease approaches, transcriptomic signatures, digital pathology, artificial intelligence, and multi-omics integration. These technologies are promising but vary substantially in analytical and clinical validation. Research findings should therefore be distinguished from biomarkers already incorporated into treatment guidance. The clinical questions and corresponding validated uses are summarized in Table 3.

Table 3. Clinical integration of molecular information in contemporary breast-cancer management.