The present narrative review was informed by targeted searches of PubMed/MEDLINE (https://pubmed.ncbi.nlm.nih.gov/) and ClinicalTrials.gov (https://clinicaltrials.gov/) from database inception until June 15, 2026. Searches were supplemented by contemporary oncology guidelines, including the ASCO neoadjuvant breast cancer guideline, the ASCO pembrolizumab rapid recommendation update, the CSCO Breast Cancer guidelines and the St. Gallen consensus recommendations, as well as pivotal-trial updates, regulatory documents when needed to reconcile data cut-offs and backward reference-list screening (5,9–11). English-language human study publications were prioritized.

Eligible evidence included randomized trials, prospective interventional studies, mature survival updates, clinically annotated prospective translational studies and major guidelines relevant to stage II–III TNBC. In this context, clinically annotated prospective translational studies were defined as prospective biomarker studies linked to clinical variables, treatment exposure and response or survival outcomes. Retrospective studies were used if they addressed clinically important evidence gaps, including treatment delivery, residual-risk assessment or biomarker interpretation. Case reports, non-clinical studies without a direct clinical decision link, duplicate publications and conference abstracts superseded by full peer-reviewed reports were excluded as primary evidence for the present review.

Evidence was prioritized in the following order: i) Practice-defining randomized evidence and mature survival updates, defined as randomized data with clinically meaningful pCR, EFS, DFS or OS endpoints; ii) boundary-defining randomized negative trials, defined as randomized studies that failed to support a proposed escalation, de-escalation or scheduling strategy and therefore helped define the limits of routine practice; iii) guidelines; iv) prospective translational studies; and v) hypothesis-generating observational evidence. Because the present review was narrative rather than systematic, duplicate screening, PRISMA flow reporting, formal risk-of-bias grading and meta-analytic pooling were not performed.

Full search concepts and database search strings are provided in Tables SI and SII.

In contemporary practice, the baseline clinical treatment method for the majority of fit patients with stage II–III TNBC is neoadjuvant systemic therapy followed by surgery, radiotherapy according to baseline and post-treatment locoregional risk, and response-guided adjuvant therapy (4–11,24,25). For patients eligible for immune checkpoint inhibition, the most evidence-supported pathway is a KEYNOTE-522-like regimen; namely, pembrolizumab combined with carboplatin-paclitaxel, followed by anthracycline-cyclophosphamide chemotherapy, surgery and completion of adjuvant pembrolizumab (6–9). Patients with residual invasive disease would then require individualized evaluation for evidence-based post-neoadjuvant options, including capecitabine in appropriate residual-disease populations and olaparib for germline BRCA1/2-mutated high-risk HER2-negative disease (24–26). This baseline is essential, because escalation and de-escalation should be judged against an established treatment pathway, not against an undefined chemotherapy reference.

Baseline evaluation must include pathological confirmation, clinical stage according to AJCC TNM criteria, organ function assessment and biomarker sampling, all of which can affect treatment planning (4–5,9–11,27). In TNBC, regional nodal evaluation warrants more attention than it currently receives in routine practice. This is because the additional ultrasound evaluation of regional nodal basins may upstage the disease, alter surgical planning and change the indication or extent of regional nodal irradiation (27–31). This matters because locoregional treatment should be integrated at baseline rather than only reconsidered after surgery.

Response assessment during treatment should be pragmatic and repetitive. Clinical examination combined with breast ultrasound or MRI remains the central approach, and early imaging response has been associated with pathological response in TNBC (32). However, imaging should not be used to justify automatic completion of a preplanned regimen in the face of unequivocal progression. When progression or persistent inoperability is documented, the priority should then shift to multidisciplinary reassessment, restoration of resectability if possible and preservation of local control. The practical value of interim assessment is therefore not merely prognostic. It is to decide whether the current pathway still makes clinical sense. This point has become more relevant in the current immunotherapy era. Certain patients experience notable shrinkage before the full number of cycles have been completed, whereas others develop toxicity that renders nominal treatment completion unrealistic. The key clinical question in this case is therefore not whether the patient remains ‘on study’, but whether the balance among tumor control, deliverability and postoperative options remains favorable. In practice, treatment escalation, schedule modification or earlier surgery may all be more rational compared with strict adherence to a preplanned timeline when the biology or tolerance profile has been altered.

From a treatment-design perspective, three baseline questions remain especially important: Whether the patient is considered sufficiently high-risk for recurrence, based on stage II–III disease, nodal positivity, larger tumor burden or other adverse clinical features, to justify platinum and/or perioperative immunotherapy; whether germline BRCA1/2 mutations or other molecular features that will influence post-neoadjuvant therapy are present; and whether the anticipated treatment intensity is realistic in light of marrow reserve, comorbidity and expected treatment adherence. The answers to these questions will typically determine outcome as much as any nominal difference between regimens.

Anthracycline-taxane chemotherapy remains the most frequently applied neoadjuvant platform for the treatment of early-stage TNBC, particularly non-metastatic stage II–III disease. This is because it most readily delivers pCR rates and provides the scaffold on which the majority of modern perioperative strategies have been tested (5–8,33–38). Accordingly, the interpretation of platinum, immunotherapy and post-neoadjuvant escalation results cannot be analyzed separately from the backbone treatment on which these data were generated.

The continuing role of dose-dense anthracycline-taxane therapy should also be stated clearly, especially for patients who are not appropriate candidates for immune checkpoint inhibition. KEYNOTE-522 established pembrolizumab-based perioperative therapy as the best-supported regimen for patients with high-recurrence-risk early-stage TNBC, but it did not directly compare its efficacy with dose-dense anthracycline-taxane therapy (6–9,45). Therefore, although a KEYNOTE-522-like perioperative pathway is currently the preferred evidence-supported approach for immune checkpoint inhibitor-eligible high-risk patients, dose-dense anthracycline-taxane chemotherapy remains a reasonable standard option for selected patients with active autoimmune diseases, at a high risk of severe immune-related adverse events, prior organ transplantation or with other contraindications to pembrolizumab.

Platinum is a selective intensification strategy reserved for especially high-recurrence-risk populations, rather than a mandatory addition to all TNBC neoadjuvant regimens. Current randomized and pooled evidence supports higher pCR rates, but the magnitude and consistency of survival benefits remain much less stable compared with the pathological response datapoint itself. In addition, platinum comes at the cost of substantially higher rates of grade ≥3 hematological toxicity in intensified regimens (33–38). For that reason, anthracycline-taxane chemotherapy remains the standard backbone treatment option, whereas platinum is best positioned as a conditional intensifier when baseline recurrence risk, expected benefit, marrow reserve and treatment deliverability are considered together. This positioning refers specifically to the use of platinum in chemotherapy-alone neoadjuvant platforms. When chemotherapy is used without immunotherapy, platinum is best considered as a selective intensifier, the benefits of which must be weighed against hematological toxicity and treatment completion. By contrast, for patients who are candidates for the KEYNOTE-522 regimen, carboplatin is not an empirical intensification step, but is instead part of the validated chemoimmunotherapy backbone on which the pCR, event-free survival (EFS) and overall survival (OS) data are generated (6–9). Substituting or omitting carboplatin within that pathway should therefore be considered a regimen modification rather than simple de-escalation. Longer-term follow-up from BrighTNess, a randomized phase III trial, further validated this point by EFS: carboplatin, not veliparib, mainly accounted for the observed long-term benefit, whereas adding the PARP inhibitor veliparib to a carboplatin-containing regimen did not improve long-term efficacy (38). This is clinically informative, because it corrects a common overinterpretation that mechanistic appeal does not automatically justify regimen proliferation. In patients with early-stage TNBC, ‘more targeted’ should not be interpreted as synonymous with ‘superior’, especially when the incremental component has not improved survival endpoints.

The current anthracycline-free literature is informative but should not be overstated. NeoPACT, a single-arm phase II clinical trial, previously showed that pembrolizumab combined with carboplatin-docetaxel can produce a relatively high pCR rate and encouraging 3-year EFS rate in patients with high-risk TNBC, supporting the feasibility of a non-anthracycline chemoimmunotherapy platform (39). However, this single-arm design lacked a randomized comparator and cannot replace randomized evidence. The Neo-N randomized phase II trial, which evaluated nivolumab timing on a carboplatin-paclitaxel backbone, did not show that a lead-in immunotherapy window was superior to concurrent initiation in terms of pCR (40). WSG-ADAPT-TN, a randomized phase II neoadjuvant de-escalation trial, has contributed invaluable de-escalation framework data. Specifically, short-course anthracycline-free therapy may be reasonable in tumors that show early treatment sensitivity, assessed by early response and pathological response rather than by a pre-treatment test alone, but the key is biological selection and early response, not empirical subtraction, for all patients; the final analysis supported the prognostic relevance of pCR and molecular markers after de-escalated therapy (41). NeoSTOP reached a similar conclusion from another angle by directly comparing anthracycline-free vs. anthracycline-containing carboplatin-based neoadjuvant chemotherapy in a randomized phase II trial. Anthracycline omission reduced toxicity, but equivalence claims remain limited as the study enrolled 100 patients and follow-up was not powered to establish definitive long-term survival equivalence (46).

In addition, two recent analyses support a cautious interpretation of anthracycline-free strategies. A 2025 meta-analysis of platinum-based anthracycline-free regimens suggested that such approaches are feasible and can generate substantial pCR rates, but heterogeneity is pronounced and survival evidence remains less stable compared with standard anthracycline-taxane regimens (47). Another comparative study of anthracycline-free vs. anthracycline-containing chemoimmunotherapy also reported that efficacy may be similar in certain selected settings, particularly in node-negative disease, but such retrospective or non-standardized comparisons are not sufficient to justify redefining current standard treatment procedures (48). Likewise, differences in dose intensity matter. Dose-dense anthracycline may retain value in patients with higher-burden disease (45), whereas cisplatin-taxane alternatives, such as the HELEN-001 phase II trial, can improve pCR in selected subgroups, including germline BRCA1/2 mutation carriers, without yet establishing a new universal standard (49). A 2024 network meta-analysis also suggested that although intensified regimens may rank higher for pCR in indirect cross-trial comparisons, standard chemotherapy remains one of the most reliable treatments for long-term survival endpoints (13). However, real-world pembrolizumab-based data add a practical warning: Toxicity-related chemotherapy dose reduction or discontinuation can adversely affect the likelihood of pCR, meaning that deliverability is not a peripheral issue in regimen selection (23).

Biologically, the appeal of platinum is high in the patient subtype enriched for DNA repair defects. Clinically, however, the experience of moving PARP inhibitors, such as olaparib, into the neoadjuvant phase is a warning against broad inference. In PARTNER, a randomized clinical trial, adding olaparib to platinum-based neoadjuvant chemotherapy in germline BRCA wild-type TNBC did not improve pCR, EFS or OS (50), meaning that knowledge of a plausible mechanism is not sufficient. Standard chemotherapy remains the core therapeutic scaffold of perioperative TNBC treatment regimen expansion, where platinum is a reasonable intensifier in genuinely high-recurrence-risk disease. However, the use of platinum should be judged together with toxicity, deliverability and the intended postoperative plan. Positioned in this manner, platinum directly addresses the central conundrum between legitimate escalation and avoidable toxicity, while acknowledging that TNBC heterogeneity does not support a universal one-size-fits-all intensified regimen.

Among the recent developments in early-stage TNBC treatment, perioperative immunotherapy has had a notable impact on clinical practice. However, positive studies are not interchangeable: A trial may be positive for pCR without having mature EFS or OS evidence, and studies may differ in checkpoint inhibitor, chemotherapy backbone, timing and risk population. KEYNOTE-522 remains the only perioperative immunotherapy strategy with convincing level I evidence, defined here as randomized phase III evidence, reporting pCR, EFS and OS endpoints (6–8). This hierarchy is also reflected in contemporary guidelines and consensus statements, including ASCO, CSCO and St. Gallen recommendations, which treat pembrolizumab-based chemoimmunotherapy as the most optimally-supported therapeutic option for high-risk stage II–III TNBC (9–11). Representative perioperative immunotherapy trials and pivotal or boundary-defining studies are summarized in Table I, listing study design, population, treatment arms or regimens, primary endpoints, key efficacy findings, survival updates and grade ≥3 adverse-event or treatment-related adverse-event rates.

KEYNOTE-522 remains the reference framework to date, because it delivered the most complete evidence chain. Pembrolizumab added to neoadjuvant chemotherapy and continued after surgery was found to improve pCR and EFS, and the 2024 update reported an OS benefit (6–8). The importance of this sequence of results cannot be overstated. Early-stage TNBC has seen various regimens improve pCR, including platinum intensification and non-pembrolizumab immunotherapy trials, but far fewer have shown conclusive event-driven benefit and fewer still have extended that benefit to OS (6–8,33–38,51–55). For these reasons, KEYNOTE-522 serves as the benchmark against which other prospective perioperative immunotherapy strategies should be compared and then interpreted. This distinction is now reflected in formal guidance. The ASCO rapid recommendation update now specifically endorses pembrolizumab with neoadjuvant chemotherapy followed by adjuvant pembrolizumab for stage II–III or otherwise high-risk early-stage TNBC, rather than extrapolating benefit across all immune checkpoint-inhibitor regimens (9). However, the KEYNOTE-522 data should be interpreted by the publication version and data cut-off. The pCR result of 64.8 vs. 51.2% is derived from the first interim analysis among the first 602 randomized patients, the EFS result reported in 2022 used a later data cut-off and showed 36-month EFS of 84.5 vs. 76.8% [hazard ratio (HR), 0.63; 95% CI, 0.48–0.82], and the 2024 OS update with a March 22, 2024, data cut-off and median follow-up of 75.1 months reported 60-month OS of 86.6 vs. 81.7% (HR, 0.66; 95% CI, 0.50–0.87) (6–8). These figures should therefore serve as complementary updates rather than a single simultaneous analysis, and Table I specifies which update is being cited.

IMpassion031, a randomized phase III clinical trial, reported similar trends but did not provide identical evidence. The original report documented pCR benefits from atezolizumab treatment (51,52). The final 2025 analysis suggested favorable trends in EFS, disease-free survival and OS, while adding informative ctDNA analyses, but wider confidence intervals and postoperative treatment heterogeneity limit interpretation (51).

Negative or neutral trials are equally important for defining the clinical boundaries of perioperative immunotherapy, rather than simply refuting its activity. NeoTRIP, a randomized study of atezolizumab combined with carboplatin-paclitaxel, failed to demonstrate a convincing efficacy advantage on its primary framework when paired with a non-anthracycline carboplatin-paclitaxel backbone (53). This result does not invalidate the survival benefit observed with pembrolizumab in KEYNOTE-522. Instead, it shows that the effect of immune checkpoint blockade in early-stage TNBC is inseparable from the chemotherapy platform, schedule and risk population in which it is tested. A common clinical mistake is to extrapolate the OS finding of pembrolizumab administered with an anthracycline-taxane platform to unvalidated non-anthracycline regimens within the same drug class; current evidence does not support that shortcut. CamRelief, a randomized clinical trial of camrelizumab plus chemotherapy, likewise supports programmed cell death protein 1-based chemoimmunotherapy, but the survival-level evidence reported remains less complete compared with KEYNOTE-522 (55). Taken together, these studies are important for biological consistency, but they do not erase the hierarchy between exploratory or preliminary benefit parameter findings and a regimen already supported by concrete survival endpoint data.

The question of timing has also not been settled in favor of more complex schedules. Neo-N, a randomized phase II trial, did not support lead-in nivolumab treatment over concurrent administration (40). More aggressive chemotherapy-free or window-only immunotherapy strategies remain hypothesis-generating. Adaptive nivolumab/ipilimumab approaches and short atezolizumab windows suggest that a biologically immune-enriched subset, defined by favorable immune microenvironment features or early immune activation, may respond without the full conventional chemotherapy (56,57). However, toxicity, strict selection criteria and lack of confirmatory randomized outcome data keep these approaches in the research domain at present. In routine care, they should be interpreted as information leveraged for future biomarker development rather than as substitutes for established chemoimmunotherapy.

In addition, endpoint interpretation warrants similar degrees of discipline. A shift in pCR may yield beneficial information in TNBC, but a clinically useful perioperative regimen should also show that it can change the distribution of residual disease, not merely the binary pCR threshold. The residual cancer burden (RCB) analysis of KEYNOTE-522 is therefore particularly informative, because pembrolizumab improved not only RCB-0 but the overall migration toward lower residual disease classes (16). That type of effect is more persuasive compared with isolated pCR gains, especially when later postoperative decisions are increasingly being based on residual burden rather than on a single yes/no pathological definition.

Biomarker enrichment is attractive, but the current data do not support a single universal immunotherapy selector. DetermaIO is an immune-related 27-gene expression signature, tumor-specific MHC-II expression is an immunohistochemical or transcriptomic tumor-cell antigen-presentation marker, and GeparNuevo-derived immune signatures are translational biomarker models derived from a clinical trial dataset; together, these studies suggest that composite immune marker profiles may outperform single markers, such as programmed death-ligand 1 (PD-L1), staining alone (17–18,58–60). However, they remain mainly translational analyses rather than decision-defining tests. Currently, the most defensible position is that perioperative immunotherapy has established clinical activity, meaning improvement in pathological response and survival endpoints in appropriately selected high-risk patients, but treatment selection should remain anchored to conclusive randomized evidence, not to isolated biomarkers (6–9). Contemporary reviews therefore frame the main unresolved questions less as whether checkpoint blockade is active and more as how to select patients appropriately, how to integrate pembrolizumab with other post-neoadjuvant options, such as capecitabine or olaparib, and how to limit the burden of immune-related toxicity and financial hardship (12,14,61). Preserving this evidence hierarchy helps avoid the clinical mistake of treating promising but exploratory biomarker signals as practice-defining evidence, clarifying the present boundary between prognostic enrichment and validated treatment selection.

Radiotherapy decisions after neoadjuvant therapy remain more nuanced than a simple yes/no discussion based on the yp stage. Current data support a relatively stable principle, whereby benefit from postmastectomy or regional nodal radiotherapy is the most consistent in patients with persistent nodal disease, especially ypN2-3 (28–30). By contrast, omission of regional nodal irradiation is more controversial in ypN0 populations and must be interpreted in combination with initial nodal status, subtype and residual burden (28–31). In TNBC, reducing this decision to a ‘ypN0 equals no radiotherapy’ rule is likely clinically inappropriate, even in patients who achieved pCR. Locoregional recurrence risk remains dependent on the baseline nodal burden, primary tumor extent and biological behavior of the original disease. Therefore, yp status alone should not be used as a universally simplified standard.

The NSABP B-51/RTOG 1304 randomized clinical trial provided an important reference point for patients who present with biopsy-proven cN1 disease, which then converted to ypN0 after neoadjuvant chemotherapy. In the randomized trial, regional nodal irradiation did not significantly improve the interval free from invasive breast cancer recurrence compared with omission of regional nodal irradiation (5-year rate, 92.7 vs. 91.8%; HR, 0.88; 95% CI, 0.60–1.28) (31). These data support a more evidence-based discussion of regional nodal irradiation de-escalation in ypN0 responders, while still requiring caution in patients with TNBC with high baseline burden, incomplete systemic response or other adverse risk features. These results and the eligibility framework are reported in the primary NSABP B-51/RTOG 1304 publication (31).

However, the more difficult question concerns the timing of radiotherapy relative to postoperative systemic therapy. In residual TNBC, a multi-center Korean retrospective comparative study suggested that concurrent radiotherapy and capecitabine may be associated with superior disease-free and OS compared with sequential treatment, without a major penalty in clinically significant toxicity (62). However, the non-randomized design means this should be treated mainly as a preliminary finding that warrants future validation.

A similarly cautious approach is warranted when radiotherapy is combined with olaparib or pembrolizumab. RADIOPARP, a phase I clinical trial, established feasibility and dose escalation boundaries for concurrent olaparib and radiotherapy by testing dose levels and monitoring acute and 1-year toxicity rather than demonstrating superiority (63,64). Real-world pembrolizumab-radiotherapy reports suggest that concurrent treatment is generally feasible, but these series are small and are more suitable as safety-oriented implementation data instead of evidence of incremental efficacy (65,66). Modern surgical series also suggest that neoadjuvant chemoimmunotherapy does not necessarily compromise surgical deliverability or cause prohibitive delay to radiotherapy, but these are implementation data rather than efficacy data (67). In a small prospective observational series of residual TNBC, persistent ctDNA positivity before radiotherapy was associated with poor distant control despite local treatment, underscoring that radiotherapy cannot substitute for systemic risk management when molecular residual disease is already present (68).

Axillary management raises a parallel issue. Surgical de-escalation after neoadjuvant therapy is attractive, but for TNBC it cannot be assumed that less axillary surgery necessarily translates into less regional treatment. Ongoing data from modern cohorts suggest that residual nodal disease burden after preoperative therapy continues to matter, where radiotherapy decisions should not simply mirror surgical de-escalation algorithms without conclusive clinicopathological and stage-based information (28–30). The practical conclusion is modest but important: Radiotherapy should be planned as part of the perioperative sequence, but concurrent systemic intensification remains selective and should be interpreted cautiously. In this manner, radiotherapy integration can contribute to local control without blurring the broader boundary between justified multimodality intensification and unsupported combination escalation.

When chemotherapy, immunotherapy, radiotherapy and postoperative escalation are considered across one treatment course, the key issue is not the number of components added, but whether the sequence remains deliverable and evidence-based. Response-adapted treatment is an attractive concept in TNBC, but it should not mean empirical escalation in response to slow early shrinkage or empirical de-escalation in response to an encouraging interim response. Adjustment should be reasonable only when the evolving response, toxicity profile and evidence of benefit all align. WSG-ADAPT-TN, a randomized phase II clinical trial, suggested that abbreviated anthracycline-free therapy may be appropriate in certain highly sensitive tumors rather than as a routine de-escalation strategy for all patients (41). By the same logic, treatments proven in the post-neoadjuvant residual-disease setting should not be moved upstream without prospective validation. In routine care, combined-modality intensification should remain selective.

The value of biomarkers for early-stage TNBC is not to replace clinical staging, but to resolve heterogeneity within patient groups with apparently similar clinical burden. PD-L1 expression and stromal tumor-infiltrating lymphocytes (TILs) have been found to be associated with higher pCR rates, with more favorable outcomes in some datasets (17–18). Even so, their role in routine practice should not be overstated. The ASCO neoadjuvant guidelines do not support using these markers alone to determine whether established neoadjuvant therapy should be given (5). Despite their prognostic value, PD-L1, TILs, homologous recombination deficiency (HRD) and related baseline markers remain to be fully validated as stand-alone selectors that can safely exclude patients from established perioperative therapies.

An equally cautious interpretation is warranted for HRD. Broad HRD labels do not always predict platinum or PARP inhibitor sensitivity consistently, where more functional readouts may be more informative. In GeparSixto, a randomized clinical trial, low RAD51 foci formation was observed to be a more viable marker compared with conventional HRD labeling for predicting platinum-associated benefit (19). This distinction matters, because TNBC is frequently extrapolated erroneously as if all DNA damage-related biomarkers are clinically interchangeable. In this context, ‘conventional HRD markers’ refer mainly to germline or somatic BRCA1/2 mutation status and genomic-scar assays based on measures such as loss of heterozygosity, telomeric allelic imbalance and large-scale state transitions (78). RAD51 foci formation is different, because it is a functional readout of homologous recombination activity rather than a static genomic scar (19,79). This distinction may explain why RAD51 performs differently from conventional HRD parameters in predicting platinum sensitivity (19,79).

Phase II clinical trial programs further illustrate why HRD-related signals require clinical context. In I-SPY2, an adaptively randomized phase II platform trial, durvalumab plus olaparib and paclitaxel increased estimated pCR in patients with high-risk HER2-negative disease compared with standard paclitaxel followed by doxorubicin-cyclophosphamide (80). GeparOLA, a randomized phase II trial, compared paclitaxel/olaparib with paclitaxel/carboplatin in HRD-defined HER2-negative disease and reported long-term survival on follow-up after the neoadjuvant comparison (81). These studies support the prospective refinement of functional HRD assessment, but they do not validate a broad range of HRD markers as standalone selectors for routine PARP-based neoadjuvant therapy.

Dynamic biomarkers are more promising compared with static markers for the next stage of treatment personalization. WSG-ADAPT-TN, a randomized phase II neoadjuvant trial, and related analyses suggest that early changes in stromal TILs, tumor cell density and proliferative markers, such as Ki-67, can improve discrimination beyond baseline measures alone (18,41,82). Early ultrasound tumor-volume reduction after two cycles of anthracycline/taxane-based therapy has likewise shown useful discriminatory value for identifying optimal responders, illustrating that interim assessment can contribute more than simple anatomic restaging when interpreted within the treatment context (32). Similarly, the prognostic value of RCB appears to be more accurate when interpreted together with the baseline clinical stage, as shown in I-SPY2 analyses (83). These observations support a broader shift in how perioperative TNBC should be conceptualized: From a one-time pre-treatment categorization to a trajectory-based model, in which early responses actively influence later decisions. At present, these aforementioned dynamic markers may help interpret evolving sensitivity, but they remain to be fully validated as standalone tools for routine regimen omission, substitution or switching outside the current structured clinical contexts.

Among the post-treatment markers, RCB provides more discriminatory information compared with the simple pCR/non-pCR dichotomy. This is clinically important because postoperative decisions now increasingly depend on the amount of residual disease rather than on the binary absence or presence of residual invasive tumor alone (16,83). Analyses from KEYNOTE-522 and I-SPY2 showed that RCB, when interpreted together with initial staging, can further refine long-term risk assessment beyond pCR alone (16,83). In practical terms, RCB is currently one of the strongest biomarkers for guiding post-neoadjuvant management, although it functions primarily as a prognostic marker rather than a treatment-selective biomarker.

Composite immune and genomic tools remain of clear interest, but their current utility remains in enrichment and risk communication. TNBC-DX is a genomic test, DetermaIO is an immune-related gene-expression signature, tumor-specific MHC-II expression is a tumor antigen-presentation marker, and related integrative models may improve the biological resolution and help identify response-enriched subsets (58–59,84). Mature tertiary lymphoid structures may add another layer of stratification, since higher tertiary lymphoid structure maturity has recently been associated with superior responses to neoadjuvant therapy and more favorable prognosis in TNBC (85). However, they have not replaced established clinical decision frameworks. The most defensible hierarchy at present is therefore to let baseline markers refine the expected sensitivity and let dynamic markers show whether sensitivity is emerging, before allowing post-treatment markers, such as RCB and ctDNA, to redefine residual risk. At present, these integrated tools mainly support residual-risk refinement rather than the independent replacement of established post-neoadjuvant decision frameworks.

ctDNA is emerging as one of the most informative tools for refining residual risk across the neoadjuvant and post-neoadjuvant continuum in TNBC, but its most notable role is currently prognostic rather than interventional. Current data consistently show that ctDNA positivity after neoadjuvant therapy is associated with residual disease, earlier recurrence and inferior survival outcomes (69–77). This makes ctDNA clinically valuable for risk stratification and for the design of enriched post-neoadjuvant trials. However, it does not yet make ctDNA a validated routine trigger for treatment escalation in unselected practice.

Post-treatment interpretation tends to be most accurate when ctDNA is read together with residual burden rather than in isolation. In residual-disease populations, ctDNA and RCB have been reported to provide complementary prognostic information, including in patients with intermediate-burden disease, where ctDNA separates patients with materially different relapse risks (71,74). In this context, materially different relapse risks means clinically meaningful differences in recurrence probability despite similar pathological residual-burden categories. A post-treatment hotspot-panel study similarly suggested that ctDNA can optimize risk assessment when tissue response alone is insufficient to identify which non-pCR patients remain the most vulnerable to recurrence (75). The practical implication is not that ctDNA can replace pathology, but that it adds another molecular layer to the residual-disease phenotype. Assay sensitivity is therefore not a technical footnote but a clinical boundary. Comparative studies in early-stage TNBC have revealed meaningful differences between molecular residual disease platforms, such that epigenomic approaches may further increase surveillance sensitivity (76–77). These advances may explain why ctDNA-negative results are not automatically equivalent across studies. However, c-TRAK TN, a clinical trial using ctDNA mutation tracking to detect molecular residual disease and trigger intervention, remains the clearest reminder of the present limit. Namely, ctDNA positivity can identify a group at particularly high risk, but intervention triggered at molecular relapse frequently occurs when occult metastatic disease is already present (69). This is a core reason why ctDNA, despite its potent prognostic value, cannot yet be treated as a routine trigger for treatment escalation in standard clinical practice. The major unanswered question is no longer whether ctDNA is prognostic, but when it should trigger intervention, in which patient cohort and if so, with which treatment regimen.