Next Article in Journal
Renal Function Deterioration in Postoperative (Adjuvant) Chemotherapy for Colon Cancer—Real-Life Data
Previous Article in Journal
Beyond Barriers: Achieving True Equity in Cancer Care
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Current Oncology
Volume 32
Issue 6
10.3390/curroncol32060350
curroncol-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Dynamic Evolution of Vascular Features Based on Magnetic Resonance Imaging to Predict Pathological Response, Patterns of Recurrence and Survival Outcomes in Breast Cancer Neoadjuvant Chemotherapy

by
Qiong Wu
1,†,
Mingxi Zhu
1,†,
Huaying Xie
2,
Xiaochuan Geng
3,
Yan Wang
1,
Ziping Wu
1,
Yanping Lin
1,
Shuguang Xu
1,
Yumei Ye
1,
Wenjin Yin
1,
Zhiguo Zhuang
3,
Jingsong Lu
1,* and
Liheng Zhou
1,4,*
1
Department of Breast Surgery, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai 200127, China
2
Department of Radiation Oncology, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai 200127, China
3
Department of Radiology, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai 200127, China
4
Punan Branch of Renji Hospital, School of Medicine, Shanghai Jiao Tong University (Punan Hospital in Pudong New District, Shanghai), Shanghai 200127, China
*
Authors to whom correspondence should be addressed.
†
These authors contributed equally to this work.
Curr. Oncol. 2025, 32(6), 350; https://doi.org/10.3390/curroncol32060350
Submission received: 29 March 2025 / Revised: 2 June 2025 / Accepted: 10 June 2025 / Published: 13 June 2025
(This article belongs to the Section Breast Cancer)
Browse Figures
Versions Notes

Simple Summary

Neoadjuvant chemotherapy (NAC) is commonly used to reduce tumor burden in breast cancer prior to surgery. However, accurate prediction of therapeutic response remains a clinical challenge. In this study, we used breast MRI scans to analyze the number of blood vessels passing through the tumor at different stages of treatment (baseline, after two cycles of NAC and before surgery). Our results showed that patients with a greater reduction in tumor-associated blood vessels were more likely to achieve breast pathological complete response and longer relapse-free survival. Moreover, distinct vascular change patterns were correlated with different recurrence types. These findings suggest that MRI-derived vascular features may serve as non-invasive imaging biomarkers for predicting treatment efficacy and recurrence patterns, potentially guiding personalized therapeutic strategies for intensified post-NAC adjuvant therapy.

Abstract

Magnetic resonance imaging (MRI) is essential for assessing responses to neoadjuvant chemotherapy (NAC) in breast cancer. This exploratory analysis included 182 patients with locally advanced breast cancer who underwent MRI at baseline after two NAC cycles and post-treatment. Vessel-through-lesion (VTL) was quantified as the number of vessels passing through the lesion, with ΔVTL1-2 and ΔVTL1-3 representing relative VTL changes after the second and third MRI, respectively, compared to baseline. Patients were randomly divided into training and validation sets (7:3). A bpCR-predictive nomogram incorporating ΔVTL1-2 and clinicopathological factors out-performed a clinicopathology-only model (validation AUC of 0.892 vs. 0.857). Recurrence patterns differed by ΔVTL1-3 subgroup, where the low ΔVTL1-3 group showed an early recurrence peak within two years, while the high ΔVTL1-3 group maintained a stable, low recurrence risk over five years, particularly among non-pCR patients. The RFS-predictive nomogram using ΔVTL1-3 showed strong concordance with RFS (validation AUCs of 0.918 at 1 year, 0.767 at 3 years, and 0.717 at 5 years). Risk stratification showed superior RFS in the low-risk group. This study is the first to report that dynamic vascular evolution predicts bpCR and serves as an independent prognostic factor, associated with distinct recurrence patterns, particularly in non-pCR patients. These findings support early identification of patients for intensified post-NAC adjuvant therapy, enabling more precise treatment.

1. Introduction

Neoadjuvant chemotherapy (NAC), followed by surgery and adjuvant therapy, constitutes the prevailing therapeutic paradigm for patients with locally advanced breast cancer (LABC). NAC facilitates tumor downstaging, thereby augmenting the prospects for breast-conserving surgery and diminishing the exigency for superfluous axillary lymph node dissections [1]. It has been elucidated that pathological complete response (pCR) post-NAC is indicative of enhanced survival outcomes, particularly in cases of human epidermal growth factor receptor-2 (HER2) positive/hormone receptor (HR) negative and triple negative pathologies [2]. A substantial cohort of patients continues to confront the peril of suboptimal tumor response. Consequently, investigative endeavors are imperative to segregate individuals with an elevated likelihood of pCR and to estimate the potential prognosis, thereby identifying potential predictive determinants for personalized tumor responses and survival.
Magnetic resonance imaging (MRI) of the breast is a widely recognized modality for the diagnosis and assessment of breast cancer. In patients undergoing NAC, initial breast MRI serves as a viable technique for delineating the extent of the lesion and assessing nodal status. While previous MRI-based paradigms have shown commendable efficacy in forecasting pCR and survival outcomes [3,4,5], they are encumbered by substantial constraints, such as a paucity of patient data, reliance on singular subtype-specific characteristic, and esoteric feature extraction methodologies. Thus, there is an imperative for further investigations to devise more universally accessible and precise models to anticipate tumor response to NAC and subsequent survival metrics. Contrast-enhanced MRI offers a means to assess tumor vascularity using contrast agent injection [6]. The identification of the vascular map obtained by MIP could offer a novel perspective on tumor characteristics. Studies have shown that unilateral elevation in breast vascularity is linked to ipsilateral invasive breast cancers [7,8,9]. In the neoadjuvant setting, vascular features might serve as predictors of tumor response to NAC [10]. Yet, the predictive value of the dynamic changes in vascular features for tumor response and survival remain unexplored.
In this study, we postulate that vascular features, in conjunction with clinicopathological characteristics, could facilitate precocious prediction of tumor responses and survival, as well as discriminating temporal recurrence patterns of breast cancer patients. We endeavor to quantify the dynamic evolution of vascular features discerned from MRI throughout NAC.

2. Materials and Methods

2.1. Patients and Treatment

We performed an exploratory analysis on breast cancer patients retrospectively enrolled from the prospective neoadjuvant database of Renji Hospital, School of Medicine, Shanghai Jiao Tong University. Eligible criteria included female patients with histologically confirmed primary unilateral breast cancer (T1 N1–3 M0 or T2–4 N0–3 M0) who received NAC and surgical treatment. Patients were excluded if they were pregnant, had metastatic breast cancer, bilateral breast cancer, had a previous history of malignancy other than breast cancer or did not complete the full process of NAC. Briefly, the therapeutic regimen comprised weekly 80 mg/m2 paclitaxel for 16 weeks, coupled with cisplatin on days 1, 8, and 15, out of every 28 days for 4 cycles. For HER2 positive disease, concurrent trastuzumab was administered as anti-HER2 target therapy on a weekly basis (first dose of trastuzumab was 4 mg/kg of body weight and the subsequent doses were 2 mg/kg). To evaluate the response to NAC, MRI assessments were conducted before NAC (baseline), subsequent to 2 NAC cycles (early treatment), and prior to surgical intervention post-NAC (preoperative). Following NAC, patients proceeded to receive further surgical treatment. After surgery, patients continued to undergo subsequent therapy at the discretion of investigators or according to the guidelines at that time. Specifically, HER2-positive patients continued to receive adjuvant trastuzumab after surgery for a total duration of one year. Between August 2014 and December 2018, a cohort of 182 patients, each having undergone both pre-NAC (baseline) and post-NAC MRI scans, was assembled for analysis. Among these patients, data of the second MRI scan, performed after two cycles of NAC, were available for 157 individuals.

2.2. MRI Protocols

MR imaging was conducted using a 3-T system (Ingenia, Philips Medical Systems, Best, The Netherlands) with patients in the prone position and both breasts suspended in a dedicated bilateral, four-channel, phased-array breast coil. Dynamic contrast-enhanced (DCE) imaging was performed with a T1-weighted high resolution isotropic volume examination (THRIVE) sequence after transverse diffusion-weighted imaging (DWI). Postprocessing involved image subtraction (contrast-enhanced minus unenhanced) and maximum intensity projection (MIP). Axial MIPs were obtained to visualize the whole vascular maps of both breasts, and vessel measurements were manually performed by an experienced breast surgeon blinded to the histopathologic data.

2.3. MRI Vascular Feature Extraction

Patients received three MRI scans, respectively, labeled as the first, second, and third MRI. We define Vessel-through-lesion (VTL) as the number of vessels crossing the lesion on the MIP image. VTLs of the three MRIs of a patient were measured at the same level under unified parameters, denoted as VTL1, VTL2, and VTL3. We subsequently define ΔVTL1-2 as the relative change in VTL after 2 cycles of NAC, calculated by
Δ V T L 1 - 2 = V T L 1 − V T L 2 V T L 1 × 100%
Similarly, ΔVTL1-3 is defined as the relative change in VTL completion of NAC, i.e.,
Δ V T L 1 - 3 = V T L 1 − V T L 3 V T L 1 × 100%
For both ΔVTL1-2 and ΔVTL1-3, a positive value indicates decreased number of vessels through lesion after NAC. The optimal cutoff of ΔVTL1-2 and ΔVTL1-3 were calculated (surv_cutpoint function; survminer package in R) to classify patients into high-ΔVTL1-2, low-ΔVTL1-2, high-ΔVTL1-3, and low-ΔVTL1-3 groups.

2.4. Data Collection

At baseline, clinical data including age, menopausal status, body mass index (BMI), clinical T stage, and clinical N stage were collected. Pathologic data from biopsy tissues, including histologic grade, estrogen receptor (ER) status, progesterone receptor (PR) status, HER2 status, and Ki-67 status, were obtained from Department of Pathology, Renji Hospital, Shanghai Jiaotong University School of Medicine. Hormone receptor (HR) positivity was defined as ER and/or PR ≥ 1% of nuclei stained using standard immunohistochemistry (IHC) methods. The cutoff value of Ki-67 was 50% [11,12]. The molecular types were defined according to the St. Gallen International Expert Consensus [13].

2.5. Outcomes

The outcomes were breast pathological complete response (bpCR), total pathological complete response (tpCR), relapse-free survival (RFS), disease-free survival (DFS), and overall survival (OS). The definition of bpCR is having no invasive cancer in the breast (ypT0/Tis). tpCR indicates the absence of any invasive cancer in the breast and lymph nodes (ypT0/Tis ypN0). RFS is the time from surgery to the first occurrence of locoregional, ipsilateral, contralateral, distant recurrence, or death from any cause. DFS is the time from surgery to the first occurrence of locoregional, ipsilateral, contralateral, distant recurrence, second primary cancer, or death from any cause. OS is the time from surgery to death from any cause.

2.6. Model Development and Validation

All the eligible patients were randomly assigned to the training set and validation set in a 7:3 ratio. The least absolute shrinkage and selection operator (LASSO) algorithm and 10-fold cross validation were used for selection of the parameters, according to which nomograms were constructed to individualize the predictive models for bpCR and RFS. Receiver operating characteristic (ROC) analysis was performed and the area under the curve (AUC) was calculated to assess the accuracy. The calibration curve was plotted to compare the consistency between actual outcomes and the predicted probabilities. The clinical utility of the models was evaluated using decision curve analysis (DCA).

2.7. Statistics

The chi-square test was used to compare categorical variables, and the Student’s t test was used to compare continuous variables. RFS, DFS, and OS were calculated using the Kaplan–Meier analysis and log-rank test. Logistic regression analyses were performed to calculate the odds ratios (ORs) and 95% confidence intervals (CIs). Cox proportional hazard regression was established to derive hazard ratios with 95% CIs. ΔVTL1-2-related calculations were performed only in the 157 patients with available MRI data after two NAC cycles. No imputation was used for missing imaging data. The individual risk scores calculated by the RFS prediction nomogram in the training set were used to obtain the optimal cutoffs (100 points) to provide the largest discrepancy in RFS, DFS, and OS (surv_cutpoint function; survminer package in R). The results were considered significant with p < 0.05. All analyses were performed by R software version 4.3.1 (http://www.R-project.org).

3. Results

3.1. Patient Clinicopathologic Characteristics

A flowchart of the study design is shown in Figure 1. Clinicopathological characteristics, vascular features, and survival data are well balanced between training set and validation set. Among 182 patients, 72 (39.56%) achieved bpCR (Table 1).
As categorical variables, 66 (42.04%) patients had high ΔVTL1-2 and 150 (82.42%) patients had high ΔVTL1-3 (Table 1). By investigating the relationship between vascular features and clinicopathological characteristics, we found that HR negativity (adjusted p = 0.045) and histologic grade-3 (adjusted p = 0.045) were associated with high ΔVTL1-2, and histologic grade-3 (adjusted p = 0.010) were associated with high ΔVTL1-3 (Table 2, Supplementary Tables S1 and S2).

3.2. Tumor Response to NAC

ΔVTL1-2 was higher in pCR group than in non-pCR group (p < 0.0001; Figure 2a). Meanwhile the pCR rate was higher in the high-ΔVTL1-2 group than in the low-ΔVTL1-2 group (p < 0.0001; Figure 2b). The multivariate logistic regression analysis suggested that ΔVTL1-2 served as an independent predictive factor for bpCR (OR 3.99; 95% CI 1.77–9.01; p < 0.001) (Supplementary Table S3). Subgroup analysis indicated that advantages in favor of high ΔVTL1-2 for bpCR across almost all subgroups. No interaction was detected between clinicopathological variables and ΔVTL1-2 for bpCR (Figure 3). Advantages in favor of high ΔVTL1-3 for bpCR were also detected across almost all subgroups (Supplementary Figure S1).

3.3. Model Development and Validation for bpCR

Eligible factors in predicting bpCR including ΔVTL1-2, age, HR status, HER2 status, clinical T stage, Ki-67 index, and BMI were selected with LASSO logistic regression in the training set (Figure 2c,d). ΔVTL1-2 was an independent predictor of bpCR (OR 3.56; 95% CI 1.37–9.28; p = 0.01), along with other clinicopathological factors (Figure 2e). A nomogram was created based on the multivariate model including the selected factors (Figure 2f). The accuracy of the model was evaluated using a calibration curve, respectively, in the training and validation set (Figure 2g,h).
ROC curves were created to compare the effectiveness of the different models with and without ΔVTL1-2. AUC of the model combining ΔVTL1-2 with clinicopathological factors for the training set was 0.819, while the model based on only clinicopathological factors reached an AUC of 0.768 (Figure 2i). For the validation set, AUCs of the two models were 0.892 and 0.857, respectively (Figure 2j). The DCA curve of the model with ΔVTL1-2 confirmed the superiority of the model with ΔVTL1-2 (Figure 2k,l).
High ΔVTL1-3 was also significantly associated with bpCR (Supplementary Figure S2A,B). A bpCR-predictive nomogram was also developed based on ΔVTL1-3 and seven other clinicopathological characteristics extracted by LASSO logistic regression in the training set (Supplementary Figure S2C–F). The model’s accuracy was evaluated using calibration curves (Supplementary Figure S2G,H). ROC curves compared the effectiveness of models with and without ΔVTL1-3, both indicating the superior predictive ability of the model with ΔVTL1-3 (Supplementary Figure S2I,J). DCA curves showed higher net benefit for the model with ΔVTL1-3 (Supplementary Figure S2K,L).
Additionally, we conducted exploratory analyses using ΔVTL1-2 and ΔVTL1-3 alone to build a predictive model for bpCR, compared with the model with only clinicopathological factors (including age, hormone receptor status, HER2 status, clinical T stage, Ki-67 index, BMI, and lymph node status). The MRI-only model achieved an AUC of 0.789 in the validation set, which was slightly lower than the clinicopathological model but still demonstrated favorable predictive capability (Supplementary Figure S3).

3.4. Model Development and Validation for RFS

ΔVTL1-3 was shown to be an independent predictor of RFS according to the univariate (hazard ratio 0.28; 95% CI 0.14–0.57; p < 0.001) and multivariate cox regression analysis in the whole set (hazard ratio 0.23; 95% CI 0.11–0.50; p < 0.001) (Supplementary Table S4). The annual recurrence hazard curves of the different subgroups exhibited significant temporal heterogeneity. The low ΔVTL1-3 group exhibited a pronounced peak in recurrence and metastasis hazard within the first 2 years, with the peak occurring around the 20th month. In contrast, the high ΔVTL1-3 group showed a relatively steady and lower recurrence risk, with no noticeable peak observed within the 5-year period. Within the first two years post-operation, the incidence of RFS and DFS events was significantly higher in the low ΔVTL1-3 group compared to the high ΔVTL1-3 group (Figure 4a,b), which was particularly evident in the non-pCR population (Figure 4c,d).
With LASSO regression, ΔVTL1-3, age, clinical T stage, BMI were selected for valuable factors of RFS prediction (Figure 5a,b). We established a nomogram based on these factors (Figure 5c,d), and compared the predictive value of this model with that based on tpCR for RFS. As shown in Figure 6, the predictive model with ΔVTL1-3 revealed a stronger correlation with RFS in the training set (AUCs for 1-year, 3-year, and 5-year RFS of 0.745, 0.772, and 0.703, respectively, Figure 6a) and the validation set (AUCs for 1-year, 3-year, and 5-year RFS of 0.918, 0.767, and 0.717, respectively, Figure 6b), outperforming the model using tpCR in the training set (AUCs for 1-year, 3-year, and 5-year RFS of 0.522, 0.655, and 0.607, respectively, Figure 6c) and the validation set (AUCs for 1-year, 3-year, and 5-year RFS of 0.702, 0.613, and 0.582, respectively, Figure 6d).
The DCA curves indicated that using the predictive model with ΔVTL1-3 added more benefit than using tpCR in both sets (Supplementary Figure S4). The corresponding calibration curve showed promising agreement between the predicted RFS and the observed results in the training set (Supplementary Figure S5).
Compared with the clinicopathological model for RFS, the MRI-only model showed better performance, with AUCs of 0.944, 0.808, and 0.816 at 1, 3, and 5 years, respectively, in the validation set, which were higher than the corresponding AUCs of the clinicopathological model (Supplementary Figure S6).

3.5. Risk Stratification of the RFS Predictive Model

Patients were stratified into two risk groups (high-risk vs. low-risk) according to the optimal cutoff of risk scores for RFS. The high-risk group showed poorer RFS than the low-risk group presented by Kaplan–Meier curves (Figure 7a–c; p = 0.0022 for the training set, p = 0.0038 for the validation set; p < 0.0001 for both of them combined). In the whole set, the 1-year, 3-year, and 5-year RFS rates were 89.3%, 63.9%, and 61.0%, respectively, for high-risk patients, versus 99.3%, 92.8%, and 89.5%, respectively, for low-risk patients. The Kaplan–Meier curves in various subgroups according to HR, HER2, and pCR status also showed good distinguishability of RFS prediction when stratified by risk groups (Supplementary Figure S7A–C). In order to evaluate the ability of the nomogram, the optimal cutoffs of risk score were applied for risk stratification for DFS and OS. The Kaplan–Meier curves for DFS and OS revealed favorable predictive value of the nomogram (Supplementary Figures S8 and S9).

4. Discussion

To the best of our knowledge, this study is the first to observe that dynamic evolution of vascular features can effectively predict bpCR. A predictive model based on changes in vascular features after two cycles of NAC could early forecast bpCR, exhibiting superior predictive performance compared to models relying on clinical-pathological factors. Additionally, our study is also the first to identify that the different dynamic evolution of vascular features post-NAC serves as an independent prognostic factor, which is associated with distinctly different temporal patterns of post-surgical recurrence and metastasis, and further clues a fundamental heterogeneity in tumor characteristics, especially apparent in non-pCR patients. Furthermore, a predictive model for prognosis was developed based on changes in vascular features during NAC, with predictive efficacy surpassing that based on tpCR.
Breast MRI angiography has become an integral part of standard breast examinations with the use of contrast-enhanced imaging. The typical MIP images generated from postprocessing subtracted images not only identify enhancing lesions but also present a vascular map of vessels within the breast [8,14,15]. This study aims to provide a more profound understanding of how vascular dynamics during NAC correlates with tumor characteristics, angiogenesis, and tumor microenvironment, offering potentially valuable insights for clinical practice. Angiogenesis, the formation of new blood vessels, is crucial for the growth and metastasis of solid tumors, which require a sustained blood supply for rapid proliferation [16]. This process is regulated by a complex interplay of proangiogenic factors, such as vascular endothelial growth factor (VEGF), which can also act directly on tumor cells, promoting tumor progression and creating a malignant positive feedback loop [17]. Bahhnassy et al. found that triple-negative breast cancer (TNBC) patients exhibit higher VEGF expression compared to non-TNBC patients, with elevated levels of VEGF-A in the blood associated with poor prognosis [18]. Furthermore, various malignancies, including esophageal cancer, colon cancer, and osteosarcoma, demonstrate different VEGF expression levels, which are correlated with tumor stage, molecular grade, and prognosis [19,20,21,22]. High VEGF expression is often linked to metastasis and poor outcomes [23,24]. The varying expression of angiogenesis-related factors suggests that tumor tissues differ in their sensitivity to treatment [25]. These changes can also be monitored throughout the course of treatment to assess therapeutic response. Kim et al. found that a decrease in VEGF transcripts is significantly associated with favorable treatment outcomes [26]. The aforementioned studies partially support our findings. The different dynamic changes in ΔVTL observed in this study might be related to corresponding fluctuation of angiogenic factors such as VEGF and further reflect the temporal heterogeneity of the tumor, at least partially representing a comprehensive manifestation of the tumor’s intrinsic characteristics.
Our study focused on predicting pathological response to NAC by calculating the relative change in the number of vessels compared to baseline (ΔVTL) and found that integrating ΔVTL1-2 with clinicopathological factors effectively predicts bpCR. Additionally, models constructed using ΔVTL1-3 also demonstrated good predictive performance. Bufi et al. found that patients with an asymmetric increase in breast vascularity in the affected breast had a higher likelihood of achieving pCR [10]. Similarly, Martincich et al. explored the relationship between changes in the number of large vessels in the affected breast and pathological response to NAC, observing a significant decrease in average vessel counts between responders and nonresponders (p = 0.015) [27]. Preclinical studies have also highlighted the anti-angiogenic activity of paclitaxel, contributing to its anti-tumor efficacy in vivo [28,29,30,31]. Lau et al. demonstrated that paclitaxel exhibits antiangiogenic activity by downregulating VEGF in a highly vascularized transgenic murine Met-1 breast cancer model, administered at non-cytotoxic doses, paclitaxel significantly reduced intratumoral angiogenesis and also suppressed VEGF expression in the tumors [32]. Studies have shown that paclitaxel can inhibit endothelial cell proliferation, chemotaxis, and morphogenesis in vitro—functions closely related to angiogenesis—and suppress angiogenesis itself in vivo using chick and mouse models [28,33]. Additionally, paclitaxel also downregulated VEGFR2 expression and upregulated thrombospondin-1 (TSP-1) expression in a metastatic mouse model [34]. All these studies were consistent with us, at least partially, and have provided evidence supporting that different dynamic change patterns of vessels during NAC, which can be detected early and conveniently through changes in ΔVTL, might be considered as a surrogate for tumor heterogeneity, and then further reflect sensitivities to treatment.
Through the study of the time distribution of recurrence risk in subgroups with different patterns of vascular changes, we observed for the first time that the low ΔVTL1-3 group experienced a rapid hazard peak in recurrence and metastasis, while the high ΔVTL1-3 group maintained consistently low recurrence levels without a noticeable peak. This disparity was particularly pronounced in the non-pCR subgroup. This might suggest that the differences in microenvironment and tumor heterogeneity between the two subgroups result in different reactivation times and growth patterns of dormant tumor cells in metastasis, displaying distinct recurrence patterns for the two subgroups. Gasparini et al.’s study showed that patients with a high angiogenesis index in their primary tumors have a higher risk of recurrence compared to patients with poorly vascularized tumors [35]. Moreover, patients with high intratumoral microvessel density (IMD) or elevated VEGF protein levels have significantly worse prognoses compared to those with low IMD or VEGF levels [36,37]. The aforementioned studies, either directly or indirectly, support our findings. Tumor dormancy has long been a clinical challenge, as dormant tumor cells can, in some cases, re-enter the proliferative phase and cause tumor recurrence [38,39]. The “angiogenic switch” has been confirmed as a critical step in tumor progression, marking the transition of the tumor from a dormant state to one of angiogenesis, which promotes tumor growth [40]. During this process, the balance between pro-angiogenic and anti-angiogenic factors is disrupted, favoring angiogenesis. Research suggests that this switch is regulated by various signals, such as VEGFA and TSP-1 [41]. In this study, the low ΔVTL1-3 subgroup showed no significant reduction in blood vessels, as tumor cells may have a strong ability to secrete angiogenic factors and be less affected by treatment. The micro-metastatic foci may maintain the attributes of the primary lesion and keep some abilities to actively secrete angiogenesis-related factors like VEGF, which can induce angiogenesis and stimulate early recurrence and metastasis post-operation. It seems reasonable to hypothesize that the heterogeneity in dynamic changes in vascular features during NAC reflects the tumor’s intrinsic traits and the nature of its micro-metastases. Low ΔVTL1-3 tumors are more prone to early recurrence, which is also observed in non-pCR subgroup, underscoring the value of tailoring follow-up strategies and treatment according to subgroups with different vascular change patterns. Especially in non-pCR patients, intensified treatment may be particularly significant for those with limited vascular reduction, although further research is needed to confirm this.
There are several limitations in this study. Firstly, it is a retrospective analysis. However, our patient selection was derived from a prospectively maintained database, it is exploratory and has been validated in both the training and validation sets. Therefore, this could indicate the existence of underlying intrinsic patterns to some degree, providing clues for future prospective research. Secondly, with a relatively small sample size, further validation in larger cohorts is expected. Thirdly there is an imbalance in the distribution of patients across different subtypes in the dataset, with a predominance of Luminal-type patients. Future studies should aim to expand the sample size, with a particular focus on triple-negative subtypes.

5. Conclusions

In conclusion, we first reported that the dynamic evolution of vascular features can effectively predict bpCR and that the different dynamic evolution of vascular features post-NAC serves as an independent prognostic factor, which is associated with distinctly different temporal post-surgical recurrence and metastasis patterns, especially apparent in non-pCR subgroup patients. It is a novel and practical direction of integration of dynamic alterations in vascular features with clinical-pathological parameters to formulate predictive models for the early anticipation of tumor pathological response and prognosis. These predictive instruments serve as invaluable tools for the identification of patients likely to benefit from NAC and better individualizing adjuvant treatment plans for patients after NAC, particularly in designing adjuvant chemotherapy and intensified adjuvant therapy for non-pCR patients.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/curroncol32060350/s1, Figure S1: Subgroup analysis for bpCR according to ΔVTL1-3; Figure S2: Feature selection, model development and model validation for bpCR prediction; Figure S3: ROC analysis of MRI-only model and clinicopathological model for bpCR; Figure S4: Decision curve analysis of 1-year, 3-year, 5-year RFS for predictive model with ΔVTL1-3 vs tpCR alone; Figure S5: Calibration curve for based on nomogram and tpCR alone in the training set; Figure S6: ROC analysis of MRI-only model and clinicopathological model for RFS; Figure S7: Kaplan–Meier survival curves of RFS according to the risk score predicted by the nomogram in subgroups; Figure S8: Kaplan–Meier survival curves of DFS according to the risk score predicted by the nomogram; Figure S9: Kaplan–Meier survival curves of OS according to the risk score predicted by the nomogram; Table S1: Relationships of VTL1, VTL2 and VTL3 with clinicopathological characteristics; Table S2. Relationships of ΔVTL1-2 and ΔVTL1-3 with clinicopathological characteristics; Table S3: Univariate and multivariate analysis for predictive factors of bpCR in the whole set; Table S4: Univariate and multivariate analysis for predictive factors of RFS in the whole set.

Author Contributions

Conceptualization: Q.W., L.Z., J.L., Z.Z. and H.X.; formal analysis, investigation, methodology, software and visualization: Q.W., M.Z., H.X., J.L., Z.Z. and X.G.; data curation and validation: Q.W., M.Z. and L.Z.; investigation and resources: Y.W., Z.W., S.X., Y.L., Y.Y. and W.Y.; funding acquisition: L.Z. and J.L.; project administration: Z.Z., L.Z. and J.L.; writing—original draft preparation: Q.W.; writing—review and editing: Z.Z., L.Z. and J.L. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the Academic Leaders Training Program of Pudong Health Bureau of Shanghai (grant number PWRd2024-06), New Quality Clinical Specialty Program of High-end Medical Disciplinary Construction in Shanghai Pudong New Area (grant number 2025-PWXZ-12, Hypermicrosurgical treatment of lymphedema), the National Natural Science Foundation of China (grant number 82173115 and 82103695), the Science and Technology Commission of Shanghai Municipality (grant number 20DZ2201600), the Shanghai Municipal Key Clinical Specialty, Shanghai Rising-Star Program (grant number 22QC1400200) and the Shanghai Municipal Health Commission Health Industry Clinical Research Special Project (grant number 202340085).

Institutional Review Board Statement

The study was conducted in accordance with the Declaration of Helsinki and approved by the Independent Ethics Committee of Renji Hospital (Approval number: LY2022-028-B, date: 28 October 2022).

Informed Consent Statement

Written informed consent has been obtained from the patient(s) to publish this paper.

Data Availability Statement

Data available on request due to privacy and ethical restrictions.

Acknowledgments

We would like to thank the investigators and patients who participated in the present study.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
AUCarea under the curve
bpCRbreast pathological complete response
DCAdecision curve analysis
DFSdisease-free survival
ERestrogen receptor
HER2human epidermal growth factor receptor-2
HRhormone receptor
LABClocally advanced breast cancer
LASSOleast absolute shrinkage and selection operator
MIPmaximum intensity projection
MRImagnetic resonance imaging
NACneoadjuvant chemotherapy
OSoverall survival
PRprogesterone receptor
RFSrelapse-free survival
ROCreceiver operating characteristic
TNBCtriple-negative breast cancer
tpCRtotal pathological complete response
TSP-1thrombospondin-1
VEGFvascular endothelial growth factor
VTLvessel through lesion

References

  1. Mauri, D.; Pavlidis, N.; Ioannidis, J.P.A. Neoadjuvant Versus Adjuvant Systemic Treatment in Breast Cancer: A Meta-Analysis. JNCI J. Natl. Cancer Inst. 2005, 97, 188–194. [Google Scholar] [CrossRef] [PubMed]
  2. Cortazar, P.; Zhang, L.; Untch, M.; Mehta, K.; Costantino, J.P.; Wolmark, N.; Bonnefoi, H.; Cameron, D.; Gianni, L.; Valagussa, P.; et al. Pathological complete response and long-term clinical benefit in breast cancer: The CTNeoBC pooled analysis. Lancet 2014, 384, 164–172. [Google Scholar] [CrossRef]
  3. Tahmassebi, A.; Wengert, G.J.; Helbich, T.H.; Bago-Horvath, Z.; Alaei, S.; Bartsch, R.; Dubsky, P.; Baltzer, P.; Clauser, P.; Kapetas, P.; et al. Impact of Machine Learning With Multiparametric Magnetic Resonance Imaging of the Breast for Early Prediction of Response to Neoadjuvant Chemotherapy and Survival Outcomes in Breast Cancer Patients. Investig. Radiol. 2019, 54, 110–117. [Google Scholar] [CrossRef]
  4. Bitencourt, A.G.V.; Gibbs, P.; Rossi Saccarelli, C.; Daimiel, I.; Lo Gullo, R.; Fox, M.J.; Thakur, S.; Pinker, K.; Morris, E.A.; Morrow, M.; et al. MRI-based machine learning radiomics can predict HER2 expression level and pathologic response after neoadjuvant therapy in HER2 overexpressing breast cancer. EBioMedicine 2020, 61, 103042. [Google Scholar] [CrossRef]
  5. Ramtohul, T.; Djerroudi, L.; Lissavalid, E.; Nhy, C.; Redon, L.; Ikni, L.; Djelouah, M.; Journo, G.; Menet, E.; Cabel, L.; et al. Multiparametric MRI and Radiomics for the Prediction of HER2-Zero, -Low, and -Positive Breast Cancers. Radiology 2023, 308, e222646. [Google Scholar] [CrossRef]
  6. Zhao, S.; Li, Y.; Ning, N.; Liang, H.; Wu, Y.; Wu, Q.; Wang, Z.; Tian, J.; Yang, J.; Gao, X.; et al. Association of peritumoral region features assessed on breast MRI and prognosis of breast cancer: A systematic review and meta-analysis. Eur. Radiol. 2024, 34, 6108–6120. [Google Scholar] [CrossRef]
  7. Schmitz, A.C.; Peters, N.H.; Veldhuis, W.B.; Gallardo, A.M.; van Diest, P.J.; Stapper, G.; van Hillegersberg, R.; Mali, W.P.; van den Bosch, M.A. Contrast-enhanced 3.0-T breast MRI for characterization of breast lesions: Increased specificity by using vascular maps. Eur. Radiol. 2008, 18, 355–364. [Google Scholar] [CrossRef]
  8. Sardanelli, F.; Iozzelli, A.; Fausto, A.; Carriero, A.; Kirchin, M.A. Gadobenate dimeglumine-enhanced MR imaging breast vascular maps: Association between invasive cancer and ipsilateral increased vascularity. Radiology 2005, 235, 791–797. [Google Scholar] [CrossRef]
  9. Carriero, A.; Di Credico, A.; Mansour, M.; Bonomo, L. Maximum intensity projection analysis in magnetic resonance of the breast. J. Exp. Clin. Cancer Res. 2002, 21, 77–81. [Google Scholar]
  10. Bufi, E.; Belli, P.; Di Matteo, M.; Giuliani, M.; Tumino, M.; Rinaldi, P.; Nardone, L.; Franceschini, G.; Mulé, A.; Bonomo, L. Hypervascularity Predicts Complete Pathologic Response to Chemotherapy and Late Outcomes in Breast Cancer. Clin. Breast Cancer 2016, 16, e193–e201. [Google Scholar] [CrossRef]
  11. Humbert, O.; Riedinger, J.-M.; Charon-Barra, C.; Berriolo-Riedinger, A.; Desmoulins, I.; Lorgis, V.; Kanoun, S.; Coutant, C.; Fumoleau, P.; Cochet, A.; et al. Identification of Biomarkers Including 18FDG-PET/CT for Early Prediction of Response to Neoadjuvant Chemotherapy in Triple-Negative Breast Cancer. Clin. Cancer Res. 2015, 21, 5460–5468. [Google Scholar] [CrossRef] [PubMed]
  12. Masuda, H.; Masuda, N.; Kodama, Y.; Ogawa, M.; Karita, M.; Yamamura, J.; Tsukuda, K.; Doihara, H.; Miyoshi, S.; Mano, M.; et al. Predictive factors for the effectiveness of neoadjuvant chemotherapy and prognosis in triple-negative breast cancer patients. Cancer Chemother. Pharmacol. 2011, 67, 911–917. [Google Scholar] [CrossRef] [PubMed]
  13. Coates, A.S.; Winer, E.P.; Goldhirsch, A.; Gelber, R.D.; Gnant, M.; Piccart-Gebhart, M.; Thürlimann, B.; Senn, H.J. Tailoring therapies—Improving the management of early breast cancer: St Gallen International Expert Consensus on the Primary Therapy of Early Breast Cancer 2015. Ann. Oncol. 2015, 26, 1533–1546. [Google Scholar] [CrossRef] [PubMed]
  14. Sardanelli, F.; Fausto, A.; Menicagli, L.; Esseridou, A. Breast vascular mapping obtained with contrast-enhanced MR imaging: Implications for cancer diagnosis, treatment, and risk stratification. Eur. Radiol. Suppl. 2007, 17, 48–51. [Google Scholar] [CrossRef]
  15. Nissan, N.; Massasa, E.E.M.; Bauer, E.; Halshtok-Neiman, O.; Shalmon, A.; Gotlieb, M.; Faermann, R.; Samoocha, D.; Yagil, Y.; Ziv-Baran, T.; et al. MRI can accurately diagnose breast cancer during lactation. Eur. Radiol. 2023, 33, 2935–2944. [Google Scholar] [CrossRef]
  16. Liu, Z.L.; Chen, H.H.; Zheng, L.L.; Sun, L.P.; Shi, L. Angiogenic signaling pathways and anti-angiogenic therapy for cancer. Signal Transduct. Target. Ther. 2023, 8, 198. [Google Scholar] [CrossRef]
  17. Lorenc, P.; Sikorska, A.; Molenda, S.; Guzniczak, N.; Dams-Kozlowska, H.; Florczak, A. Physiological and tumor-associated angiogenesis: Key factors and therapy targeting VEGF/VEGFR pathway. Biomed. Pharmacother. 2024, 180, 117585. [Google Scholar] [CrossRef]
  18. Bahhnassy, A.; Mohanad, M.; Shaarawy, S.; Ismail, M.F.; El-Bastawisy, A.; Ashmawy, A.M.; Zekri, A.R. Transforming growth factor-β, insulin-like growth factor I/insulin-like growth factor I receptor and vascular endothelial growth factor-A: Prognostic and predictive markers in triple-negative and non-triple-negative breast cancer. Mol. Med. Rep. 2015, 12, 851–864. [Google Scholar] [CrossRef]
  19. Inoue, K.; Ozeki, Y.; Suganuma, T.; Sugiura, Y.; Tanaka, S. Vascular endothelial growth factor expression in primary esophageal squamous cell carcinoma. Association with angiogenesis and tumor progression. Cancer 1997, 79, 206–213. [Google Scholar] [CrossRef]
  20. Han, G.; Wang, Y.; Bi, W.; Jia, J.; Wang, W.; Xu, M. Effects of vascular endothelial growth factor expression on pathological characteristics and prognosis of osteosarcoma. Clin. Exp. Med. 2016, 16, 577–584. [Google Scholar] [CrossRef] [PubMed]
  21. DuBois, S.; Demetri, G. Markers of angiogenesis and clinical features in patients with sarcoma. Cancer 2007, 109, 813–819. [Google Scholar] [CrossRef] [PubMed]
  22. Ishigami, S.I.; Arii, S.; Furutani, M.; Niwano, M.; Harada, T.; Mizumoto, M.; Mori, A.; Onodera, H.; Imamura, M. Predictive value of vascular endothelial growth factor (VEGF) in metastasis and prognosis of human colorectal cancer. Br. J. Cancer 1998, 78, 1379–1384. [Google Scholar] [CrossRef] [PubMed]
  23. Yang, Y.; Cao, Y. The impact of VEGF on cancer metastasis and systemic disease. Semin. Cancer Biol. 2022, 86, 251–261. [Google Scholar] [CrossRef]
  24. Vimalraj, S. A concise review of VEGF, PDGF, FGF, Notch, angiopoietin, and HGF signalling in tumor angiogenesis with a focus on alternative approaches and future directions. Int. J. Biol. Macromol. 2022, 221, 1428–1438. [Google Scholar] [CrossRef]
  25. Yang, S.X.; Steinberg, S.M.; Nguyen, D.; Wu, T.D.; Modrusan, Z.; Swain, S.M. Gene expression profile and angiogenic marker correlates with response to neoadjuvant bevacizumab followed by bevacizumab plus chemotherapy in breast cancer. Clin. Cancer Res. 2008, 14, 5893–5899. [Google Scholar] [CrossRef]
  26. Kim, R.; Kawai, A.; Wakisaka, M.; Sawada, S.; Shimoyama, M.; Yasuda, N.; Hidaka, M.; Morita, Y.; Ohtani, S.; Arihiro, K. Immune correlates of the differing pathological and therapeutic effects of neoadjuvant chemotherapy in breast cancer. Eur. J. Surg. Oncol. 2020, 46, 77–84. [Google Scholar] [CrossRef]
  27. Martincich, L.; Bertotto, I.; Montemurro, F.; Ponzone, R.; Carbonaro, L.A.; Regge, D.; Sardanelli, F. Variation of breast vascular maps on dynamic contrast-enhanced MRI after primary chemotherapy of locally advanced breast cancer. AJR Am. J. Roentgenol. 2011, 196, 1214–1218. [Google Scholar] [CrossRef]
  28. Belotti, D.; Vergani, V.; Drudis, T.; Borsotti, P.; Pitelli, M.R.; Viale, G.; Giavazzi, R.; Taraboletti, G. The microtubule-affecting drug paclitaxel has antiangiogenic activity. Clin. Cancer Res. 1996, 2, 1843–1849. [Google Scholar]
  29. Ng, S.S.; Figg, W.D.; Sparreboom, A. Taxane-mediated antiangiogenesis in vitro: Influence of formulation vehicles and binding proteins. Cancer Res. 2004, 64, 821–824. [Google Scholar] [CrossRef]
  30. Grant, D.S.; Williams, T.L.; Zahaczewsky, M.; Dicker, A.P. Comparison of antiangiogenic activities using paclitaxel (taxol) and docetaxel (taxotere). Int. J. Cancer 2003, 104, 121–129. [Google Scholar] [CrossRef] [PubMed]
  31. Pasquier, E.; Carré, M.; Pourroy, B.; Camoin, L.; Rebaï, O.; Briand, C.; Braguer, D. Antiangiogenic activity of paclitaxel is associated with its cytostatic effect, mediated by the initiation but not completion of a mitochondrial apoptotic signaling pathway. Mol. Cancer Ther. 2004, 3, 1301–1310. [Google Scholar] [CrossRef]
  32. Lau, D.H.; Xue, L.; Young, L.J.; Burke, P.A.; Cheung, A.T. Paclitaxel (Taxol): An inhibitor of angiogenesis in a highly vascularized transgenic breast cancer. Cancer Biother. Radiopharm. 1999, 14, 31–36. [Google Scholar] [CrossRef]
  33. Vacca, A.; Ribatti, D.; Iurlaro, M.; Merchionne, F.; Nico, B.; Ria, R.; Dammacco, F. Docetaxel Versus Paclitaxel for Antiangiogenesis. J. Hematotherapy Stem Cell Res. 2002, 11, 103–118. [Google Scholar] [CrossRef]
  34. Jiang, H.; Tao, W.; Zhang, M.; Pan, S.; Kanwar, J.R.; Sun, X. Low-dose metronomic paclitaxel chemotherapy suppresses breast tumors and metastases in mice. Cancer Investig. 2010, 28, 74–84. [Google Scholar] [CrossRef]
  35. Gasparini, G.; Biganzoli, E.; Bonoldi, E.; Morabito, A.; Fanelli, M.; Boracchi, P. Angiogenesis sustains tumor dormancy in patients with breast cancer treated with adjuvant chemotherapy. Breast Cancer Res. Treat. 2001, 65, 71–75. [Google Scholar] [CrossRef]
  36. Gasparini, G.; Barbareschi, M.; Boracchi, P.; Verderio, P.; Caffo, O.; Meli, S.; Dalla Palma, P.; Marubini, E.; Bevilacqua, P. Tumor angiogenesis predicts clinical outcome of node-positive breast cancer patients treated with adjuvant hormone therapy or chemotherapy. Cancer J. Sci. Am. 1995, 1, 131–141. [Google Scholar]
  37. Gasparini, G.; Toi, M.; Miceli, R.; Vermeulen, P.B.; Dittadi, R.; Biganzoli, E.; Morabito, A.; Fanelli, M.; Gatti, C.; Suzuki, H.; et al. Clinical relevance of vascular endothelial growth factor and thymidine phosphorylase in patients with node-positive breast cancer treated with either adjuvant chemotherapy or hormone therapy. Cancer J. Sci. Am. 1999, 5, 101–111. [Google Scholar]
  38. Tufail, M.; Jiang, C.H.; Li, N. Tumor dormancy and relapse: Understanding the molecular mechanisms of cancer recurrence. Mil. Med. Res. 2025, 12, 7. [Google Scholar] [CrossRef]
  39. Tamamouna, V.; Pavlou, E.; Neophytou, C.M.; Papageorgis, P.; Costeas, P. Regulation of Metastatic Tumor Dormancy and Emerging Opportunities for Therapeutic Intervention. Int. J. Mol. Sci. 2022, 23, 13931. [Google Scholar] [CrossRef]
  40. Dudley, A.C.; Griffioen, A.W. Pathological angiogenesis: Mechanisms and therapeutic strategies. Angiogenesis 2023, 26, 313–347. [Google Scholar] [CrossRef] [PubMed]
  41. Kazerounian, S.; Lawler, J. Integration of pro- and anti-angiogenic signals by endothelial cells. J. Cell Commun. Signal. 2018, 12, 171–179. [Google Scholar] [CrossRef] [PubMed]
Curroncol 32 00350 g001
Figure 1. Flowchart of study design.
Figure 1. Flowchart of study design.
Curroncol 32 00350 g001
Curroncol 32 00350 g002
Figure 2. Feature selection, model development, and model validation for bpCR prediction. (a) ΔVTL1-2 between pCR and non-pCR groups (ΔVTL1-2 as a continuous variable). "****" indicates p < 0.0001, same with (b). (b) The pCR rates of patients with low-ΔVTL1-2 and high-ΔVTL1-2 (ΔVTL1-2 as a binary categorical variable). (c,d) Feature selection for predicting bpCR using LASSO-logistic regression. (c) LASSO algorithm using minimum penalty criteria from 10-fold cross-validation. ΔVTL1-2, age, hormone receptor status, HER2 status, clinical T stage, Ki-67 index, and BMI were extracted for model development. The red dots denote the mean deviances of the cross validation folds under different penalty λ. (d) LASSO coefficient profiles of candidate features. Each ldenotes the LASSO coefficient of a candidate feature. (e) Forest plot illustrating factors selected by LASSO regression in predicting bpCR using multivariate logistic regression. (f) Nomogram built for predicting bpCR based on multivariate logistic regression in the training set. (g) Calibration of the nomogram for the training set. (h) Calibration of the nomogram for the validation set. (i) Receiver operating characteristic curves (ROC) of different predictive models for the training set. (j) ROC of different predictive models for the validation set. (k) Decision curve analysis (DCA) of different predictive models for the training set. (l) DCA of different predictive models for the validation set.
Figure 2. Feature selection, model development, and model validation for bpCR prediction. (a) ΔVTL1-2 between pCR and non-pCR groups (ΔVTL1-2 as a continuous variable). "****" indicates p < 0.0001, same with (b). (b) The pCR rates of patients with low-ΔVTL1-2 and high-ΔVTL1-2 (ΔVTL1-2 as a binary categorical variable). (c,d) Feature selection for predicting bpCR using LASSO-logistic regression. (c) LASSO algorithm using minimum penalty criteria from 10-fold cross-validation. ΔVTL1-2, age, hormone receptor status, HER2 status, clinical T stage, Ki-67 index, and BMI were extracted for model development. The red dots denote the mean deviances of the cross validation folds under different penalty λ. (d) LASSO coefficient profiles of candidate features. Each ldenotes the LASSO coefficient of a candidate feature. (e) Forest plot illustrating factors selected by LASSO regression in predicting bpCR using multivariate logistic regression. (f) Nomogram built for predicting bpCR based on multivariate logistic regression in the training set. (g) Calibration of the nomogram for the training set. (h) Calibration of the nomogram for the validation set. (i) Receiver operating characteristic curves (ROC) of different predictive models for the training set. (j) ROC of different predictive models for the validation set. (k) Decision curve analysis (DCA) of different predictive models for the training set. (l) DCA of different predictive models for the validation set.
Curroncol 32 00350 g002
Curroncol 32 00350 g003
Figure 3. Subgroup analysis for bpCR according to ΔVTL1-2. Notes: ORs and 95% CIs were obtained from univariate logistic regression model. Interaction p-values were shown between ΔVTL1-2 and subgroups. Abbreviations: VTL, vessel through lesion; T, tumor; HR, hormone receptor; HER2, human epidermal growth factor receptor 2; BMI, body mass index; CI, confidential interval.
Figure 3. Subgroup analysis for bpCR according to ΔVTL1-2. Notes: ORs and 95% CIs were obtained from univariate logistic regression model. Interaction p-values were shown between ΔVTL1-2 and subgroups. Abbreviations: VTL, vessel through lesion; T, tumor; HR, hormone receptor; HER2, human epidermal growth factor receptor 2; BMI, body mass index; CI, confidential interval.
Curroncol 32 00350 g003
Curroncol 32 00350 g004
Figure 4. Time-dependent hazard rate by ΔVTL1-3 for RFS and DFS. (a) Smoothed hazard rate estimates for RFS. (b) Smoothed hazard rate estimates for DFS. (c) Smoothed hazard rate estimates for RFS in non-pCR patients. (d) Smoothed hazard rate estimates for DFS in non-pCR patients.
Figure 4. Time-dependent hazard rate by ΔVTL1-3 for RFS and DFS. (a) Smoothed hazard rate estimates for RFS. (b) Smoothed hazard rate estimates for DFS. (c) Smoothed hazard rate estimates for RFS in non-pCR patients. (d) Smoothed hazard rate estimates for DFS in non-pCR patients.
Curroncol 32 00350 g004
Curroncol 32 00350 g005
Figure 5. Feature selection and model development for RFS prediction. (a,b) Feature selection for predicting RFS using LASSO-cox regression. (a) LASSO algorithm using minimum penalty criteria from 10-fold cross-validation. ΔVTL1-3, age, clinical T stage, and BMI were extracted for model development. The red dots denote the mean deviances of the cross validation folds under different penalty λ. (b) LASSO coefficient profiles of candidate features. Each line denotes the LASSO coefficient of a candidate feature. (c) Forest plot illustrating factors selected by LASSO regression in predicting RFS using multivariate cox regression. (d) Nomogram built for predicting RFS based on multivariate cox regression.
Figure 5. Feature selection and model development for RFS prediction. (a,b) Feature selection for predicting RFS using LASSO-cox regression. (a) LASSO algorithm using minimum penalty criteria from 10-fold cross-validation. ΔVTL1-3, age, clinical T stage, and BMI were extracted for model development. The red dots denote the mean deviances of the cross validation folds under different penalty λ. (b) LASSO coefficient profiles of candidate features. Each line denotes the LASSO coefficient of a candidate feature. (c) Forest plot illustrating factors selected by LASSO regression in predicting RFS using multivariate cox regression. (d) Nomogram built for predicting RFS based on multivariate cox regression.
Curroncol 32 00350 g005
Curroncol 32 00350 g006
Figure 6. ROC analysis of different predictive models for RFS. (a) ROC of predictive model with ΔVTL1-3 for the training set. (b) ROC of predictive model with ΔVTL1-3 for the validation set. (c) ROC of tpCR alone for RFS prediction for the training set. (d) ROC of tpCR alone for RFS prediction for the validation set. Abbreviations: VTL, vessel through lesion; tpCR, total pathological complete response; AUC, area under curve; RFS, relapse-free survival.
Figure 6. ROC analysis of different predictive models for RFS. (a) ROC of predictive model with ΔVTL1-3 for the training set. (b) ROC of predictive model with ΔVTL1-3 for the validation set. (c) ROC of tpCR alone for RFS prediction for the training set. (d) ROC of tpCR alone for RFS prediction for the validation set. Abbreviations: VTL, vessel through lesion; tpCR, total pathological complete response; AUC, area under curve; RFS, relapse-free survival.
Curroncol 32 00350 g006
Curroncol 32 00350 g007
Figure 7. Kaplan–Meier survival curves of RFS according to the risk score predicted by the nomogram in (a) training set, (b) validation set, and (c) whole set.
Figure 7. Kaplan–Meier survival curves of RFS according to the risk score predicted by the nomogram in (a) training set, (b) validation set, and (c) whole set.
Curroncol 32 00350 g007
Table 1. Clinicopathological characteristics, MRI vascular features and outcome data of patients in the training and validation set.
Table 1. Clinicopathological characteristics, MRI vascular features and outcome data of patients in the training and validation set.
VariableWhole Set
(n = 182)		Training Set
(n = 127)		Validation Set
(n = 55)		Statisticp-Value
Age, n (%)   χ2 = 0.1960.658
    <3514 (7.69)11 (8.66)3 (5.45)  
    ≥35168 (92.31)116 (91.34)52 (94.55)  
Menopausal status, n (%)   χ2 = 0.0520.819
    Premenopausal87 (47.8)60 (47.24)27 (49.09)  
    Postmenopausal95 (52.2)67 (52.76)28 (50.91)  
Clinical T stage, n (%)   χ2 = 0.5800.446
    T1–T236 (19.78)27 (21.26)9 (16.36)  
    T3–T4146 (80.22)100 (78.74)46 (83.64)  
Nodal status, n (%)   χ2 = 0.2780.598
    Negative26 (14.29)17 (13.39)9 (16.36)  
    Positive156 (85.71)110 (86.61)46 (83.64)  
HER2 status, n (%)   χ2 = 0.8290.363
    Negative110 (60.44)74 (58.27)36 (65.45)  
    Positive72 (39.56)53 (41.73)19 (34.55)  
Ki-67 index, n (%)   χ2 = 0.7510.386
    <50%108 (59.34)78 (61.42)30 (54.55)  
    ≥50%74 (40.66)49 (38.58)25 (45.45)  
Histologic grade, n (%)   χ2 = 2.4810.289
    G120 (10.99)13 (10.24)7 (12.73)  
    G260 (32.97)38 (29.92)22 (40.00)  
    G3102 (56.04)76 (59.84)26 (47.27)  
Molecular subtype, n (%)   χ2 = 1.7620.623
    Luminal A-like19 (10.44)14 (11.02)5 (9.09)  
    Luminal B-like122 (67.03)82 (64.57)40 (72.73)  
    HER2-enriched21 (11.54)17 (13.39)4 (7.27)  
    Basal-like20 (10.99)14 (11.02)6 (10.91)  
BMI, n (%)   χ2 = 0.0230.881
    <25137 (75.27)96 (75.59)41 (74.55)  
    ≥2545 (24.73)31 (24.41)14 (25.45)  
bpCR, n (%)   χ2 = 0.0060.936
    Non-pCR110 (60.44)77 (60.63)33 (60.00)  
    pCR72 (39.56)50 (39.37)22 (40.00)  
ΔVTL1-2, %, Mean ± SD34.50 ± 38.3632.59 ± 39.0838.83 ± 36.70t = −0.9380.349
ΔVTL1-3, %, Mean ± SD53.72 ± 39.5954.34 ± 40.8952.28 ± 36.73t = 0.3220.748
ΔVTL1-2, n (%)   χ2 = 0.0040.950
    Low91 (57.96)63 (57.80)28 (58.33)  
    High66 (42.04)46 (42.20)20 (41.67)  
ΔVTL1-3, n (%)   χ2 = 0.3180.573
    Low32 (17.58)21 (16.54)11 (20.00)  
    High150 (82.42)106 (83.46)44 (80.00)  
RFS rate, % (95% CI)   t = 0.1670.875
    1-y97.2 (94.8–99.6)97.6 (94.9–100.0)96.3 (91.5–100.0)  
    3-y86.8 (82.0–92.0)86.0 (80.0–92.4)88.7 (80.6–97.7)  
    5-y83.6 (78.3–89.4)83.2 (76.7–90.2)84.6 (75.3–95.0)  
OS rate, % (95% CI)   t = 0.4720.661
    1-y100.0 (100.0–100.0)100.0 (100.0–100.0)100.0 (100.0–100.0)  
    3-y94.0 (90.6–97.5)93.7 (89.6–98.0)94.5 (88.7–100.0)  
    5-y91.7 (87.8–95.8)90.6 (85.6–95.8)94.5 (88.7–100.0)  
Note: All p-values adjusted for multiple comparisons using the Benjamini–Hochberg method (FDR) were >0.05. Abbreviations: T, tumor; ER, estrogen receptor; PR, progesterone receptor; HER2, human epidermal growth factor receptor 2; BMI, body mass index; bpCR, breast pathological complete response; HR, hormone receptor; VTL, vessel through lesion; RFS, relapse-free survival; OS, overall survival; CI, confidential interval.
Table 2. Relationships of ΔVTL1-2 and ΔVTL1-3 with clinicopathological characteristics.
Table 2. Relationships of ΔVTL1-2 and ΔVTL1-3 with clinicopathological characteristics.
CharacteristicsΔVTL1-2 ΔVTL1-3
Low,
n = 91		High, n = 66p-ValueAdjusted
p-Value		Low,
n = 32		High, n = 150p-ValueAdjusted
p-Value
Age(yr)  11  0.2750.393
    <357 (7.69%)5 (7.58%)  4 (12.5%)10 (6.67%)  
    ≥3584 (92.3%)61 (92.4%)  28 (87.5%)140 (93.3%)  
Menopausal status  0.7390.924  0.1010.337
    Premenopausal45 (49.5%)30 (45.5%)  20 (62.5%)67 (44.7%)  
    Postmenopausal46 (50.5%)36 (54.5%)  12 (37.5%)83 (55.3%)  
BMI  0.8580.953  11
    <2568 (74.7%)51 (77.3%)  24 (75.0%)113 (75.3%)  
    ≥2523 (25.3%)15 (22.7%)  8 (25.0%)37 (24.7%)  
HR status  0.0090.045  0.1670.347
    Negative11 (12.1%)20 (30.3%)  3 (9.38%)33 (22.0%)  
    Positive80 (87.9%)46 (69.7%)  29 (90.6%)117 (78.0%)  
HER2 status  0.320.457  0.2080.347
    Negative58 (63.7%)36 (54.5%)  23 (71.9%)87 (58.0%)  
    Positive33 (36.3%)30 (45.5%)  9 (28.1%)63 (42.0%)  
Ki-67 status  0.0520.150  0.5490.686
    <50%62 (68.1%)34 (51.5%)  21 (65.6%)87 (58.0%)  
    ≥50%29 (31.9%)32 (48.5%)  11 (34.4%)63 (42.0%)  
Clinical T Stage  0.0750.150  11
    T1–211 (12.1%)16 (24.2%)  6 (18.8%)30 (20.0%)  
    T3–480 (87.9%)50 (75.8%)  26 (81.2%)120 (80.0%)  
Nodal status  0.1140.190  0.1750.347
    Negative16 (17.6%)5 (7.58%)  7 (21.9%)19 (12.7%)  
    Positive75 (82.4%)61 (92.4%)  25 (78.1%)131 (87.3%)  
Molecular type  0.0690.150  0.0560.280
    Luminal A-like10 (11.0%)4 (6.06%)  7 (21.9%)12 (8.00%)  
    Luminal B-like67 (73.6%)41 (62.1%)  22 (68.8%)100 (66.7%)  
    HER2-enriched6 (6.59%)12 (18.2%)  1 (3.12%)20 (13.3%)  
    Basel-like8 (8.79%)9 (13.6%)  2 (6.25%)18 (12.0%)  
Histologic grade  0.0070.045  0.0010.010
    G111 (12.1%)6 (9.09%)  6 (18.8%)14 (9.33%)  
    G239 (42.9%)14 (21.2%)  17 (53.1%)43 (28.7%)  
    G341 (45.1%)46 (69.7%)  9 (28.1%)93 (62.0%)  
Note: p-values adjusted by Benjamini–Hochberg method (FDR control). Abbreviations: VTL, vessel through lesion; BMI, body mass index; HR, hormone receptor; HER2, human epidermal growth factor receptor 2; T, tumor.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Wu, Q.; Zhu, M.; Xie, H.; Geng, X.; Wang, Y.; Wu, Z.; Lin, Y.; Xu, S.; Ye, Y.; Yin, W.; et al. Dynamic Evolution of Vascular Features Based on Magnetic Resonance Imaging to Predict Pathological Response, Patterns of Recurrence and Survival Outcomes in Breast Cancer Neoadjuvant Chemotherapy. Curr. Oncol. 2025, 32, 350. https://doi.org/10.3390/curroncol32060350

AMA Style

Wu Q, Zhu M, Xie H, Geng X, Wang Y, Wu Z, Lin Y, Xu S, Ye Y, Yin W, et al. Dynamic Evolution of Vascular Features Based on Magnetic Resonance Imaging to Predict Pathological Response, Patterns of Recurrence and Survival Outcomes in Breast Cancer Neoadjuvant Chemotherapy. Current Oncology. 2025; 32(6):350. https://doi.org/10.3390/curroncol32060350

Chicago/Turabian Style

Wu, Qiong, Mingxi Zhu, Huaying Xie, Xiaochuan Geng, Yan Wang, Ziping Wu, Yanping Lin, Shuguang Xu, Yumei Ye, Wenjin Yin, and et al. 2025. "Dynamic Evolution of Vascular Features Based on Magnetic Resonance Imaging to Predict Pathological Response, Patterns of Recurrence and Survival Outcomes in Breast Cancer Neoadjuvant Chemotherapy" Current Oncology 32, no. 6: 350. https://doi.org/10.3390/curroncol32060350

APA Style

Wu, Q., Zhu, M., Xie, H., Geng, X., Wang, Y., Wu, Z., Lin, Y., Xu, S., Ye, Y., Yin, W., Zhuang, Z., Lu, J., & Zhou, L. (2025). Dynamic Evolution of Vascular Features Based on Magnetic Resonance Imaging to Predict Pathological Response, Patterns of Recurrence and Survival Outcomes in Breast Cancer Neoadjuvant Chemotherapy. Current Oncology, 32(6), 350. https://doi.org/10.3390/curroncol32060350

Article Metrics

Back to TopTop