MRI-guided de-escalation in HER2-positive early breast cancer: precision progress with necessary prudence

Introduction

Neoadjuvant systemic therapy (NST) incorporating dual human epidermal growth factor receptor 2 (HER2) blockade has transformed outcomes in stage II–III HER2-positive breast cancer. Contemporary regimens combining trastuzumab and pertuzumab achieve pathological complete response (pCR) rates exceeding 60% and 3-year event-free survival (EFS) approaching 90–95% (1-5). These improvements, however, are achieved at the cost of substantial toxicity, including hematologic suppression, neuropathy, cardiotoxicity, and cumulative treatment burden.

The TRAIN-3 study by Louis et al. introduces a response-adapted strategy that individualizes chemotherapy duration according to radiological complete response (rCR) assessed by breast magnetic resonance imaging (MRI) (6). Rather than administering a fixed six to nine cycles of chemotherapy, treatment was truncated upon rCR, with patients proceeding to early surgery. The recently reported 3-year EFS outcomes provide important evidence supporting the feasibility of MRI-guided de-escalation.

As de-escalation strategies mature, however, the critical issue shifts from whether treatment can be reduced to whether the tools guiding reduction are sufficiently reliable and biologically comprehensive.

Summary of TRAIN-3 findings

TRAIN-3 enrolled 467 patients with stage II–III HER2-positive breast cancer (6). Patients received paclitaxel, carboplatin, trastuzumab, and pertuzumab, with MRI reassessment every three cycles. Surgery was performed upon rCR or after a maximum of nine cycles.

At a median follow-up of 40.1 months, 3-year EFS was 92.2% in hormone receptor (HR)-negative tumors and 92.0% in HR-positive tumors (6). Patients undergoing surgery after only one to three cycles demonstrated particularly favorable outcomes, with 3-year EFS of 96.1% in HR-negative and 98.6% in HR-positive disease. Toxicity increased proportionally with chemotherapy duration.

These data suggest that early MRI response may guide chemotherapy truncation in selected patients. However, interpretation requires scrutiny of how rCR was defined and operationalized.

MRI as a decision trigger: reliability and generalizability

In TRAIN-3, rCR was defined as absence of pathological contrast enhancement in the original tumor bed and was assessed locally across 43 hospitals without central radiology review (6). Although pragmatic and reflective of real-world practice, this approach introduces variability at the very decision point that determines treatment cessation.

Inter-observer variability in MRI response assessment following NST has been demonstrated in prospective multi-reader imaging platforms, including analyses from the ACRIN 6657/I-SPY studies, where agreement in residual disease estimation and complete response classification was moderate rather than uniform (7). When imaging functions as a therapeutic gatekeeper, reproducibility becomes essential. Overcalling rCR risks premature cessation of systemic therapy; undercalling risks unnecessary continuation.

This concern is particularly relevant in HR-positive disease. In TRAIN-3, the negative predictive value of MRI for pCR in HR-positive tumors was 53%, indicating that nearly half of radiological complete responders harbored residual invasive disease (6). This finding should be interpreted in the context of the lower prevalence of pCR in HR-positive disease, which inherently reduces negative predictive value even when the intrinsic performance of the test is similar. Although subsequent adjuvant therapy mitigates risk, this level of diagnostic precision raises concerns regarding generalizability without imaging standardization.

Future implementation of MRI-guided de-escalation will likely require structured acquisition protocols, reader calibration frameworks, and formal quality assurance to ensure consistent performance across institutions. Standardization of imaging protocols will be particularly important to support reproducibility and external validity. Emerging evidence suggests that multiparametric MRI combined with radiomic or machine-learning approaches may further improve predictive performance for pathological response (8). In addition to enhancing discrimination, automated quantitative imaging biomarkers may help reduce inter-reader and intra-reader variability by providing more objective response metrics. While radiomics does not fully eliminate variability in image interpretation, algorithmic support may complement protocol standardization and reduce subjective interpretation. Overall, technical refinement, imaging standardization, and reproducibility safeguards will be central to the safe clinical translation of MRI-guided treatment de-escalation.

Biological enrichment and the need for randomized validation

The excellent outcomes observed among patients treated for only one to three cycles should be interpreted within a biological framework. These tumors exhibited rapid and profound sensitivity to HER2-directed therapy and likely represent a biologically enriched subset. Their favorable prognosis reflects tumor biology as much as therapeutic duration.

Although benchmarking against historical trials provides contextual reassurance (1-5), cross-trial comparisons remain inherently limited by differences in staging, eligibility, treatment era, and adjuvant protocols. A randomized comparison in patients achieving rCR after three cycles—contrasting early surgery with continuation of chemotherapy—would provide more definitive evidence than a single-arm phase 2 design can offer. However, the excellent short-term outcomes observed may challenge equipoise, underscoring the practical difficulty of conducting definitive trials in biologically favorable populations.

Beyond pCR: micrometastatic disease and molecular residual risk

Several de-escalation strategies in early breast cancer have historically relied on pCR rather than imaging as the primary decision-making endpoint. Adaptive platform trials such as I-SPY have used predicted pCR to guide treatment escalation and de-escalation, underscoring the central role of pathological response in treatment individualization (7).

Although pCR is strongly associated with improved outcomes (1-3), it does not guarantee eradication of systemic micrometastatic disease. Lin et al. demonstrated that patients achieving pCR but remaining circulating tumor DNA (ctDNA)-positive after NST experienced significantly inferior recurrence-free survival compared with ctDNA-negative counterparts (9). These findings suggest that molecular minimal residual disease may supersede pathological response in prognostic relevance. Integration of molecular monitoring alongside imaging may therefore provide a more robust framework for safe de-escalation, particularly in HR-positive disease where late relapse biology predominates (10,11).

In parallel with dynamic molecular monitoring, baseline biological stratification may further refine patient selection. Genomic assays such as HER2DX integrate Erb-B2 receptor tyrosine kinase 2 (ERBB2) expression, immune signatures, proliferation markers, and clinical variables to generate prognostic and predictive risk scores in HER2-positive early breast cancer (12). Emerging validation studies suggest that genomic risk remains heterogeneous even among patients achieving pCR, underscoring that imaging-defined complete response does not uniformly equate to systemic cure. Incorporation of genomic profiling alongside imaging and ctDNA assessment may provide a more biologically coherent framework for safe treatment de-escalation (9-12).

Residual disease and surgical de-escalation

Management of residual invasive disease has evolved toward antibody-drug conjugate-based escalation. The DESTINY-Breast05 trial demonstrated improved invasive disease-free survival with trastuzumab deruxtecan compared with trastuzumab emtansine in patients with residual HER2-positive early breast cancer (13), reinforcing biologically targeted escalation strategies. In addition to escalation based on residual disease at surgery, serial MRI assessment during neoadjuvant therapy may allow earlier treatment adaptation. Early identification of non-responders could permit transition to alternative HER2-directed therapies, including antibody-drug conjugates, consistent with adaptive strategies explored in trials such as I-SPY, potentially overcoming treatment resistance earlier in the disease course (7,14). This strategy merits further investigation in prospective studies.

The logical extension of chemotherapy de-escalation is surgical de-escalation. Early prospective data suggest that vacuum-assisted biopsy-confirmed pCR may permit selective omission of breast surgery in carefully chosen exceptional responders (15). However, surgical omission presents distinct challenges. Unlike chemotherapy truncation—where systemic therapy continues—elimination of surgery removes a definitive locoregional control modality. Accurate identification of true complete responders therefore requires minimization of sampling error, imaging-pathology concordance, and assurance that residual microscopic disease will not compromise long-term local control. Rigorous validation and extended follow-up remain essential.

Conclusions

TRAIN-3 represents an important advance toward individualized chemotherapy duration in HER2-positive early breast cancer. The 3-year EFS outcomes are reassuring, and toxicity reduction is clinically meaningful.

However, safe implementation of MRI-guided de-escalation requires attention to imaging reproducibility, integration of molecular biomarkers, longer-term follow-up—particularly in HR-positive disease—and ideally randomized validation. As treatment intensification becomes increasingly biologically targeted, de-escalation must meet equivalent evidentiary standards.

Precision de-escalation is achievable. Precision confirmation remains the necessary next step.

Acknowledgments

Artificial intelligence tools were used to assist in drafting and language refinement. All contents were critically reviewed and approved by the authors. The manuscript adheres to the Transparency in the Reporting of Artificial Intelligence (TITAN) guidelines 2025 for transparent AI reporting.

Footnote

Provenance and Peer Review: This article was a standard submission to the journal. The article has undergone external peer review.

Peer Review File: Available at https://tbcr.amegroups.com/article/view/10.21037/tbcr-2026-1-0015/prf

Funding: None.

Conflicts of Interest: Both authors have completed the ICMJE uniform disclosure form (available at https://tbcr.amegroups.com/article/view/10.21037/tbcr-2026-1-0015/coif). K.M. serves as an unpaid editorial board member of Translational Breast Cancer Research from May 2025 to June 2027. K.M. received honoraria from Merit Medical and Q Medical corporations for academic advisory services. The other author has 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.

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. Gianni L, Pienkowski T, Im YH, et al. 5-year analysis of neoadjuvant pertuzumab and trastuzumab in patients with locally advanced, inflammatory, or early-stage HER2-positive breast cancer (NeoSphere): a multicentre, open-label, phase 2 randomised trial. Lancet Oncol 2016;17:791-800. [Crossref] [PubMed]
2. Schneeweiss A, Chia S, Hickish T, et al. Long-term efficacy analysis of the randomised, phase II TRYPHAENA cardiac safety study: Evaluating pertuzumab and trastuzumab plus standard neoadjuvant anthracycline-containing and anthracycline-free chemotherapy regimens in patients with HER2-positive early breast cancer. Eur J Cancer 2018;89:27-35. [Crossref] [PubMed]
3. van der Voort A, van Ramshorst MS, van Werkhoven ED, et al. Three-Year Follow-up of Neoadjuvant Chemotherapy With or Without Anthracyclines in the Presence of Dual ERBB2 Blockade in Patients With ERBB2-Positive Breast Cancer: A Secondary Analysis of the TRAIN-2 Randomized, Phase 3 Trial. JAMA Oncol 2021;7:978-84. [Crossref] [PubMed]
4. Hurvitz SA, Martin M, Jung KH, et al. Neoadjuvant Trastuzumab Emtansine and Pertuzumab in Human Epidermal Growth Factor Receptor 2-Positive Breast Cancer: Three-Year Outcomes From the Phase III KRISTINE Study. J Clin Oncol 2019;37:2206-16.
5. Pérez-García JM, Cortés J, Ruiz-Borrego M, et al. 3-year invasive disease-free survival with chemotherapy de-escalation using an (18)F-FDG-PET-based, pathological complete response-adapted strategy in HER2-positive early breast cancer (PHERGain): a randomised, open-label, phase 2 trial. Lancet 2024;403:1649-59. [Crossref] [PubMed]
6. Louis FM, van der Voort A, van Ramshorst MS, et al. MRI-based personalisation of neoadjuvant chemotherapy duration in HER2-positive early breast cancer (TRAIN-3): primary results from a multicentre, single-arm, phase 2 study. Lancet Oncol 2026;27:349-60. [Crossref] [PubMed]
7. Hylton NM, Blume JD, Bernreuter WK, et al. Locally advanced breast cancer: MR imaging for prediction of response to neoadjuvant chemotherapy--results from ACRIN 6657/I-SPY TRIAL. Radiology 2012;263:663-72. [Crossref] [PubMed]
8. Ye T, Jin S, Xu H, et al. Can radiomics from dynamic contrast-enhanced MRI effectively predict response to neoadjuvant chemotherapy in breast cancer?: A meta-analysis. Clin Radiol 2026;92:107085. [Crossref] [PubMed]
9. Lin PH, Tsai LW, Lo C, et al. ctDNA Detected after Neoadjuvant Therapy for HER2-Positive Breast Cancer Is Associated with Inferior Outcomes and May Inform Adjuvant Therapy. Cancer Res Commun 2026;6:105-14. [Crossref] [PubMed]
10. Huober J, Holmes E, Baselga J, et al. Survival outcomes of the NeoALTTO study (BIG 1-06): updated results of a randomised multicenter phase III neoadjuvant clinical trial in patients with HER2-positive primary breast cancer. Eur J Cancer 2019;118:169-77. [Crossref] [PubMed]
11. Piccart-Gebhart M, Holmes E, Baselga J, et al. Adjuvant Lapatinib and Trastuzumab for Early Human Epidermal Growth Factor Receptor 2-Positive Breast Cancer: Results From the Randomized Phase III Adjuvant Lapatinib and/or Trastuzumab Treatment Optimization Trial. J Clin Oncol 2016;34:1034-42. Erratum in: J Clin Oncol 2019;37:356.
12. Llombart-Cussac A, Pérez-García J, Brasó-Maristany F, et al. HER2DX Genomic Assay in HER2-Positive Early Breast Cancer Treated with Trastuzumab and Pertuzumab: A Correlative Analysis from the PHERGain Phase II Trial. Clin Cancer Res 2024;30:4123-30. [Crossref] [PubMed]
13. Loibl S, Park YH, Shao Z, et al. Trastuzumab Deruxtecan in Residual HER2-Positive Early Breast Cancer. N Engl J Med 2026;394:845-57. [Crossref] [PubMed]
14. Barker AD, Sigman CC, Kelloff GJ, et al. I-SPY 2: an adaptive breast cancer trial design in the setting of neoadjuvant chemotherapy. Clin Pharmacol Ther 2009;86:97-100. [Crossref] [PubMed]
15. Kuerer HM, Valero V, Smith BD, et al. Selective Elimination of Breast Surgery for Invasive Breast Cancer: A Nonrandomized Clinical Trial. JAMA Oncol 2025;11:529-34. [Crossref] [PubMed]