Predictability of pretreatment contrast-enhanced MRI for pathological complete response after neoadjuvant chemotherapy in breast cancer patients: validation of the literature results.
[BACKGROUND] Pretreatment MRI-based features predictive of pathological complete response (pCR) after neoadjuvant chemotherapy (NAC) in breast cancer (BC) patients have been reported, with inconsisten
APA
Oliveira C, Oliveira F, et al. (2026). Predictability of pretreatment contrast-enhanced MRI for pathological complete response after neoadjuvant chemotherapy in breast cancer patients: validation of the literature results.. European journal of radiology, 194, 112485. https://doi.org/10.1016/j.ejrad.2025.112485
MLA
Oliveira C, et al.. "Predictability of pretreatment contrast-enhanced MRI for pathological complete response after neoadjuvant chemotherapy in breast cancer patients: validation of the literature results.." European journal of radiology, vol. 194, 2026, pp. 112485.
PMID
41138318
Abstract
[BACKGROUND] Pretreatment MRI-based features predictive of pathological complete response (pCR) after neoadjuvant chemotherapy (NAC) in breast cancer (BC) patients have been reported, with inconsistent results. We aimed to evaluate which of these features could predict pCR in our dataset.
[MATERIALS AND METHODS] Retrospective study including women with BC of no special type submitted to pretreatment MRI, NAC, and surgery at Champalimaud Foundation, and approved by the institutional Ethics Committee. A literature review was performed on the PubMed database, including only English-written papers published in the last 5 years. Only MRI-based features from the one-minute postgadolinium injection on T1-weighted images compliant with the Image Biomarker Standardisation Initiative (IBSI) and reported to be predictive of pCR were selected to be tested. The selected features were extracted from the pretreatment MRI of each primary BC and statistically compared between responders (pCR) and non-responders (non-pCR).
[RESULTS] Our dataset included 203 tumors. pCR was observed in 86 tumors (43/130 luminal B-like, 30/57 triple-negative and 13/16 HER2-enriched surrogate molecular subtypes). After the literature review, 43 MRI-based features were selected to be tested. Only one MRI-based feature was significantly associated with NAC response in the entire dataset and 11 features were associated with NAC response in at least one tumor subtype. However, none of these associations survived correction for multiple comparisons.
[CONCLUSIONS] Approximately ¾ of the radiomic features previously reported as predictive of pCR were not confirmed in our dataset (even without correction for multiple comparisons). This raises concerns about the validity of previously published findings. Large multicentre studies are needed to evaluate the real potential of radiomic features on pCR prediction.
[MATERIALS AND METHODS] Retrospective study including women with BC of no special type submitted to pretreatment MRI, NAC, and surgery at Champalimaud Foundation, and approved by the institutional Ethics Committee. A literature review was performed on the PubMed database, including only English-written papers published in the last 5 years. Only MRI-based features from the one-minute postgadolinium injection on T1-weighted images compliant with the Image Biomarker Standardisation Initiative (IBSI) and reported to be predictive of pCR were selected to be tested. The selected features were extracted from the pretreatment MRI of each primary BC and statistically compared between responders (pCR) and non-responders (non-pCR).
[RESULTS] Our dataset included 203 tumors. pCR was observed in 86 tumors (43/130 luminal B-like, 30/57 triple-negative and 13/16 HER2-enriched surrogate molecular subtypes). After the literature review, 43 MRI-based features were selected to be tested. Only one MRI-based feature was significantly associated with NAC response in the entire dataset and 11 features were associated with NAC response in at least one tumor subtype. However, none of these associations survived correction for multiple comparisons.
[CONCLUSIONS] Approximately ¾ of the radiomic features previously reported as predictive of pCR were not confirmed in our dataset (even without correction for multiple comparisons). This raises concerns about the validity of previously published findings. Large multicentre studies are needed to evaluate the real potential of radiomic features on pCR prediction.
MeSH Terms
Humans; Breast Neoplasms; Female; Neoadjuvant Therapy; Magnetic Resonance Imaging; Contrast Media; Middle Aged; Treatment Outcome; Retrospective Studies; Reproducibility of Results; Adult; Aged; Sensitivity and Specificity; Chemotherapy, Adjuvant