Development and multicenter validation of an immunoinflammatory marker-MRI model for pathological complete response after neoadjuvant chemotherapy in invasive breast cancer

Introduction

Breast cancer (BC) is one of the most diagnosed malignant tumors and a leading cause of cancer-related death in women (1). Due to the heterogeneity of BC, patients with different subtypes exhibit varying rates of recurrence and metastasis, resulting in significant differences in survival outcomes (2). Neoadjuvant chemotherapy (NAC) is now widely adopted as a routine therapeutic strategy for patients with high-risk BC (3). This approach provides multiple clinical benefits, including reduction of tumor burden, facilitation of disease downstaging, improvement in surgical feasibility and breast-conserving rates, as well as the opportunity to assess tumor responsiveness to systemic treatment at an early stage (4,5). Pathological complete response (pCR) is characterized by the absence of residual invasive carcinoma in both the excised breast tissue and all evaluated regional lymph nodes following neoadjuvant systemic therapy (ypT0/Tis ypN0) (6,7). Studies have shown that pCR is closely related to disease-free survival (DFS) and overall survival (OS) (4,5) and is widely used as a surrogate endpoint in patients with BC who received NAC, with the aim of accelerating the evaluation of new treatment strategies. Therefore, it is necessary to identify a method to accurately predict which patients are likely to achieve a pCR and consequently derive the greatest benefit from NAC.

Several studies have incorporated inflammatory indicators into predictive models for pCR. For instance, Pu et al. constructed a nomogram including lymphovascular invasion, anemia, estrogen receptor (ER) status, Ki-67 expression, and NAC regimen, which achieved an area under the curve (AUC) of 0.758 for predicting pCR after NAC (8). Other studies have also evaluated inflammatory markers such as neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), lymphocyte-to-monocyte ratio (LMR), and systemic immune-inflammation index (SII) as predictors of treatment response in BC (6,9,10). Although these markers are easily obtained from routine blood tests and reflect the host immune-inflammatory status, their predictive performance varies considerably across studies, and models relying primarily on systemic markers often show only moderate discrimination.

Imaging-based approaches have also been explored for predicting pCR after NAC. Among available imaging techniques, breast magnetic resonance imaging (MRI) is considered the most informative modality for treatment response assessment because of its high soft-tissue resolution and functional imaging capability. Multiparametric MRI techniques, particularly dynamic contrast-enhanced MRI (DCE-MRI) and diffusion-weighted imaging (DWI), can provide parameters reflecting tumor vascularity and cellularity. Li et al. (11) reported that a pretreatment DCE-MRI-based nomogram incorporating tumor volume, time-to-peak enhancement, and androgen receptor status achieved AUCs of 0.84 and 0.79 in the training and validation cohorts, respectively. In addition, a systematic review evaluating diffusion-weighted MRI for predicting pCR showed that reported AUC values ranged from 0.50 to 0.92, indicating substantial variability among studies (12). These findings suggest that imaging features can provide valuable information on tumor morphology and microstructure, but their predictive accuracy remains heterogeneous.

One possible explanation is that each modality reflects different aspects of tumor biology. Systemic inflammatory markers mainly represent host immune and inflammatory responses (6,13), whereas MRI features primarily describe local tumor morphology, perfusion, and cellular density. Because response to chemotherapy is influenced by both tumor-intrinsic characteristics and host biological factors, reliance on a single modality may not sufficiently capture the complexity of treatment response (14,15). Integrating systemic immunoinflammatory markers with multiparametric MRI features may therefore provide complementary information and improve predictive performance. However, studies evaluating such combined approaches remain limited.

The objective of the current study was to construct a combined model based on clinicopathological variables, immunoinflammatory markers, and multiparameter MRI features at baseline to predict the probability of pCR in BC patients receiving NAC. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2841/rc).

Methods

This was a multicenter study. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committees of Henan Cancer Hospital (No. 2024-170-001) and Longyan First Affiliated Hospital of Fujian Medical University (No. LYREC2025-k105-01). Informed consent was waived in this retrospective study. A total of 376 patients with suspected invasive BC who underwent NAC followed by surgery at center 1 (Henan Cancer Hospital) between January 2018 and June 2020 were retrospectively collected as the training cohort. Inclusion criteria: (I) all patients underwent MRI scanning before treatment, including DWI and DCE sequences; (II) invasive BC with clinical stages T1–4, NX, and M0; and (III) complete clinical data and immunoinflammatory markers before treatment were available. Exclusion criteria: (I) a history of other malignancies (n=13); (II) poor image quality (n=16); and (III) missing data were excluded (n=2).

Patients who met the same inclusion and exclusion criteria were prospectively enrolled at center 1 (Henan Cancer Hospital) between June 2023 and June 2024 to form the internal validation cohort. In addition, patients treated at center 2 (Longyan First Affiliated Hospital of Fujian Medical University) between November 2022 and March 2024 were retrospectively collected as the external validation cohort to provide an independent dataset for model validation. The study flowchart is shown in Figure 1. Regarding sample size considerations, the development of the prediction model followed the commonly recommended events-per-variable (EPV ≥10) principle to ensure model stability. The number of outcome events in the training cohort satisfied this requirement. For the validation cohorts, no formal sample size calculation was performed; instead, all eligible patients available during the study period were included.

Figure 1 Flowchart in this study. Center 1, Henan Cancer Hospital; center 2, Longyan First Affiliated Hospital of Fujian Medical University. BC, breast cancer; DCE, dynamic contrast-enhanced; DWI, diffusion-weighted imaging; M, metastasis; MRI, magnetic resonance imaging; N, node; NAC, neoadjuvant chemotherapy; pCR, pathological complete response; T, tumor.

Clinicopathological data

Complete clinicopathological information was obtained from the integrated medical record system of our hospital, including age, menopausal status, clinical T stage, clinical N stage, molecular subtype, ER, progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2), and Ki-67 expression. Clinical staging was determined according to the 8^th edition of the American Joint Committee on Cancer (AJCC) staging system. NAC regimens were selected in accordance with the prevailing clinical guidelines at the time, with anthracycline- and taxane-based therapies constituting the principal treatment strategies. pCR is characterized by the absence of residual invasive carcinoma in both the excised breast tissue and all evaluated regional lymph nodes following neoadjuvant systemic therapy (ypT0/Tis ypN0), pCR was defined as Miller-Payne grade 5, while grades 1–4 were classified as non-pCR (7,16). Pathological evaluation criteria: Immunohistochemical positivity for ER and PR was defined as nuclear staining rates of ≥1%, respectively. HER2 was considered negative with scores of 0, 1+, and 2+ [without fluorescence in situ hybridization (FISH) amplification] and positive with scores of 3+ and 2+ (with FISH amplification). Ki-67 was categorized as negative (<20%) and positive (≥20%). BC was classified into four molecular subtypes based on the expression of ER, PR, HER2, and Ki-67: luminal A, luminal B, HER2-enriched, and triple-negative.

Standard hematology assessments and complete blood count analyses were conducted within 2 weeks before therapy. Fasting blood samples were obtained, including 2 mL of ethylenediaminetetraacetic acid (EDTA)-anticoagulated peripheral blood, 2 mL of plasma collected using sodium citrate as an anticoagulant, along with serum specimens. The inflammatory markers were calculated based on the following formulas: systemic inflammation response index (SIRI) was calculated as ratio of neutrophil count × monocyte count to lymphocyte count, pan-immune-inflammation value (PIV) was calculated as the ratio of platelet count × neutrophil count × monocyte count to lymphocyte count, NLR, PLR, and LMR were computed as the ratios of neutrophils to lymphocytes, platelets to lymphocytes, and lymphocytes to monocytes, respectively. The SII was derived according to the following formula: SII = platelet count × neutrophil count/lymphocyte count.

MRI scanning

Center 1: MRI scans were performed using a Skyra 3.0-T and Prisma 3.0-T MRI scanner (Siemens Healthineers, Erlangen, Germany) with a 16-channel breast phased-array coil, in a prone position with feet first. The sequences included T2-weighted imaging (T2WI) [repetition time (TR), 3,500 ms; echo time (TE), 81 ms; slice thickness, 4 mm; field of view (FOV), 384 mm × 360 mm], T1-weighted imaging (T1WI) (TR, 5.36 ms; TE, 2.41 ms; slice thickness, 1.5 mm; FOV, 416 mm × 416 mm), DWI (TR, 6,460 ms; TE, 57 ms; slice thickness, 4 mm; FOV, 112 mm × 224 mm; b-value, 50, 1,000 s/mm²), and DCE- T1WI (TR, 4.18 ms; TE, 1.31 ms; slice thickness, 2 mm; FOV, 640 mm × 560 mm). DCE-T1WI was performed using the TWIST-VIBE sequence, with dynamic acquisition initiated 21.1 seconds after intravenous injection of gadolinium-diethylenetriaminepentaacetic acid (Gd-DTPA) at a flow rate of 2.5 mL/s via a high-pressure injector, for a total dose of 0.2 mmol/kg, capturing 40 phases of enhanced images.

Center 2: MRI scans were performed using an Ingenia 3.0-T and an Achive 3.0-T MRI scanner (Philips Healthcare, Andover, MA, USA) with a 7-channel breast phased-array coil, in a prone position with feet first. The sequences included T2WI (TR, 3,772 ms; TE, 70 ms; slice thickness, 4 mm; FOV, 280 mm × 340 mm), T1WI (TR, 542 ms; TE, 8 ms; slice thickness, 4 mm; FOV, 280 mm × 339 mm), DWI (TR, 6,391 ms; TE, 77 ms; slice thickness, 4 mm; FOV, 280 mm × 341 mm; b-value 0, 800 s/mm²), and DCE-T1WI (TR, 4.4 ms; TE, 2.1 ms; slice thickness, 1.2 mm; FOV, 240 mm × 379 mm). DCE-T1WI was performed using the dyn-eTHRIVE sequence, with dynamic acquisition initiated 48.8 seconds after intravenous injection of Gd-DTPA at a flow rate of 2.5 mL/s via a high-pressure injector, for a total dose of 0.2 mmol/kg, capturing 8 phases of enhanced images.

MRI image analysis

MRI characteristics: two radiologists (F.M. and J.Y.) with over 10 years of experience in breast MRI diagnosis analyzed the images based on the 2013 5^th edition of the Breast Imaging Reporting and Data System (BI-RADS). Both radiologists were blinded to the pathological results and clinical outcomes during image evaluation to minimize potential bias. In case of discrepancies, consensus was reached through discussion. For multiple lesions, the largest lesion was evaluated. The MRI characteristics recorded included the amount of fibro glandular tissue (FGT), background parenchymal enhancement (BPE), lesion location, number, maximum diameter, shape, margin, enhancement pattern, time-intensity curve (TIC) type, and peritumoral edema. Mean apparent diffusion coefficient (ADC_mean) values were measured using Syngo.Via software (Siemens Healthineers) and IntelliSpace Portal (Philips Healthcare). The same two radiologists independently performed the measurements in a blinded manner. Three circular regions of interest (ROI) were selected from the solid part of the tumor, avoiding large vessels and visibly cystic or necrotic areas with an area of 10–40 mm² at the largest tumor section and the adjacent upper and lower layers, and then the average value was calculated.

Statistical analysis

Data were analyzed using MedCalc 20.1.0 and R 4.0.2 software. In center 1, the minimum sample size of the internal validation was calculated according to the method of test diagnosis in the reference (17). Assuming a null hypothesis AUC of 0.50 and an alternative AUC of 0.70, with an anticipated distribution of 70% non-pCR and 30% pCR cases and a statistical power of 90%, the minimum sample size required for the internal validation cohort was estimated to be 100 patients. Receiver operating characteristic (ROC) curve analysis was applied in the training cohort to identify optimal cutoff values for inflammatory biomarkers, with the threshold corresponding to the maximum Youden index selected as the optimal cutoff. Interobserver agreement between the two radiologists was assessed using the intraclass correlation coefficient (ICC), and an ICC value greater than 0.75 was considered indicative of good agreement. Independent sample t-test was performed the measurement data with normal distribution between groups. Mann-Whitney U test was used to compare the non-normal distribution measurement data and categorical data between groups. Chi-squared test or Fisher’s exact test was used for comparison of non-categorical data. Feature selection was conducted using the least absolute shrinkage and selection operator (LASSO) regression algorithm, with 10-fold cross-validation employed to select the most informative variables. Multivariate logistic regression analysis was used to screen the independent factors for predicting pCR of invasive BC, and then models were constructed and visualized as a nomogram. The ROC curve and the AUC were used to evaluate the performance of the three prediction models in predicting pCR, and the DeLong test was used for comparison. An AUC value between 0.70 and 0.80 was considered acceptable discrimination, 0.80–0.90 was considered good discrimination, and values above 0.90 were considered excellent performance in the context of clinical prediction models (18). All statistical tests were two-sided, and a P value <0.05 was considered statistically significant. LASSO regression analysis was operated with the “glmnet” package.

Results

Patient features

Overall, a total of 345 female patients, with 116 (33.62%) pCR and 229 (66.38%) non-pCR; including 187 patients (pCR group: 65; non-pCR group: 122; a mean age of 46.82±9.65 years) in the training set, 104 patients (pCR group: 33; non-pCR group: 71; a mean age of 46.77±9.30 years) in the internal validation set, and 54 patients (pCR group: 18; non-pCR group: 36; a mean age of 49.02±8.87 years) in the external validation set.

Using the training set data, the Youden index was used to calculate the best cut-off values of SIRI, PIV, PLR, LMR, NLR, and SII, which were 0.461, 303.774, 156.039, 4.656, 1.304, and 946.470, respectively.

Table 1 presents the comparison of clinicopathological variables between patients with and without pCR across the different cohorts. In the training set, clinical T stage, clinical N stage, molecular subtype, ER status, PR status, Ki-67 index, PLR, LMR, and SII differed significantly between the pCR and non-pCR groups (P<0.05). In the internal validation set, significant between-group differences were identified for clinical T stage, molecular subtype, ER status, PR status, and PLR (P<0.05). In the external validation set, only clinical T stage, Ki-67 expression, and PLR demonstrated statistically significant differences between the two groups (P<0.05). No significant differences were observed for other features in the three sets.

MRI features

Comparison of MRI features between the pCR and non-pCR groups is shown in Table 2. Enhancement pattern, maximum diameter, shape, TIC type, and ADC_mean were significantly associated with pCR in the training set (P<0.05). In the internal validation set, enhancement pattern, TIC type, and ADC_mean have significant differences between the pCR and non-pCR groups (P<0.05). In the external validation set, enhancement pattern, maximum diameter, and ADC_mean have significant differences between the pCR and non-pCR groups (P<0.05) (Figures 2,3).

Figure 2 A 60-year-old female patient with invasive BC in the left breast achieved a pCR after NAC. (A) Axial fat-suppressed T1WI shows an equal signal with relatively clear boundaries. (B) Axial fat-suppressed T2WI shows a slightly high signal with surrounding peritumoral edema. (C) Axial apparent diffusion coefficient map shows an uneven low signal, ADCmean =988×10⁻⁶ mm²/s. (D,E) Axial and sagittal enhanced T1WI show marked enhancement of the lesion. (F) Photomicrograph (H&E, ×200) of histologic specimen shows no clear residual cancer in the tumor bed area (MP, level 5). ADCmean, mean apparent diffusion coefficient; BC, breast cancer; H&E, hematoxylin and eosin; MP, Miller-Payne grading system; NAC, neoadjuvant chemotherapy; pCR, pathological complete response; T1WI, T1-weighted imaging; T2WI, T2-weighted imaging.

Figure 3 A 33-year-old female patient with invasive BC in the left breast did not achieve a pCR after NAC. (A) Axial fat-suppressed T1WI shows an equal and low signal. (B) Axial fat-suppressed T2WI shows a slightly high signal with surrounding peritumoral edema, central necrosis. (C) Axial apparent diffusion coefficient map shows an uneven low signal, ADCmean =820×10⁻⁶ mm²/s. (D,E) Axial and sagittal enhanced T1WI show rim enhancement of the lesion with relatively clear boundaries. (F) Photomicrograph (H&E, ×200) of histologic specimen shows residual cancer (MP, level 3). ADCmean, mean apparent diffusion coefficient; BC, breast cancer; H&E, hematoxylin and eosin; MP, Miller-Payne grading system; NAC, neoadjuvant chemotherapy; pCR, pathological complete response; T1WI, T1-weighted imaging; T2WI, T2-weighted imaging.

Feature selection and prediction model establishment

In the training set, the univariate analysis showed that clinical T stage, clinical N stage, molecular subtypes, PLR, LMR, SII, ER, PR, Ki-67, enhancement pattern, maximum diameter, shape, TIC type, and ADC_mean were significantly associated with pCR (P<0.05). Then, LASSO regression analysis showed that clinical T stage, clinical N stage, PLR, LMR, SII, ER, PR, Ki-67, enhancement pattern, maximum diameter, TIC type, and ADC_mean were important in predicting pCR after NAC in invasive BC patients (Figure 4). Subsequently, the results of multivariable logistic regression analysis showed that clinical T stage, LMR, SII, enhancement pattern, and ADC_mean were the independent factors (Table 3).

Figure 4 Feature selection with the LASSO method in the training set. (A) The tuning parameter (λ) selection in the LASSO model used 10-fold cross-validation with minimum criteria. The optimal λ value was 0.0009932. (B) LASSO coefficient profiles of texture features could predict pCR in invasive BC, yielding 12 best coefficients at the selected logarithm (λ). BC, breast cancer; LASSO, least absolute shrinkage and selection operator; pCR, pathological complete response.

Based on the features identified above, three prediction models were developed: one of which was a clinical model (clinical T stage + LMR + SII), the MRI model (enhancement pattern + ADC_mean), and the combined model (clinical + MRI model).

To facilitate independent validation and future clinical application, the final logistic regression model was defined as follows:

Logit(p)=−10.814+3.771×(Tstage1)+2.449×(Tstage2)+1.744×(Tstage3)−1.750×(LMR)+1.896×(SII)+2.721×(enhancementpattern1)+1.973×(enhancementpattern2)+0.006×(ADCmean)[1]

where p represents the predicted probability of achieving pCR after NAC.

The predicted probability was calculated using the logistic transformation:

p=11+e−logit(p)[2]

Evaluation of the performance of different models

In the training set, AUC [95% confidence interval (CI)] for the combined model in predicting pCR was 0.820 (0.758–0.873), which was higher than that for clinical model [0.731 (0.661–0.793)] and MRI model 0.708 (0.637–0.772) (Z=3.326 and 3.491, P<0.05) (Table 4, Figure 5). In the internal validation set, AUC (95% CI) for the combined model [0.810 (0.722–0.881)] was also higher than that for the clinical model [0.756 (0.662–0.835)] and the MRI model [0.691 (0.593–0.778)] (Z=2.052 and 2.379, P<0.05). In the external validation set, AUC (95% CI) for the combined model [0.799 (0.668–0.896)] was also higher than that for the clinical model [0.754 (0.608–0.854)] and the MRI model [0.645 (0.503–0.771)] (Z=1.194 and 1.957, P=0.23 and 0.050) (Table 4, Figure 5).

Figure 5 ROC curves of the prediction models in the training (A), internal validation (B), and external validation (C) cohorts. MRI, magnetic resonance imaging; ROC, receiver operating characteristic.

Discussion

This study showed that clinical T stage, LMR, SII, enhancement pattern, and ADC_mean were the independent factors for predicting pCR of invasive BC, and they can effectively distinguish pCR from non-pCR in invasive BC patients. A combined model based on clinicopathological variables, immunoinflammatory markers, and MRI characteristics can predict post-NAC pCR in invasive BC patients preoperatively, with the combined model (clinical + MRI model) achieving the highest diagnostic efficiency in the training, internal validation, and external validation sets.

Clinicopathological features and immunoinflammatory markers have been previously shown to predict pCR after NAC in invasive BC (6-10). Wang et al. (4) showed that lower clinical T stage, higher LMR, and lower SII were associated with a higher likelihood of achieving pCR. Mosalem et al. (19) reported that high NLR and LMR were associated with a higher likelihood of pCR. According to the research results of Dong et al. (20), clinical T and N stages, SIRI, and NLR are independent prognostic factors for pCR in invasive BC patients, and a low clinical stage is a favorable factor for pCR. In this study, low clinical T stage, high LMR, and low SII were favorable factors for pCR in invasive BC, which was consistent with the results of the appeal study. However, the results of some studies are inconsistent with the present study, such as Peng et al. (21), who suggested that lower LMR is a favorable predictor of the efficacy of NAC in invasive BC patients. The possible reasons are the large difference in patient sample size between the two studies, and related to the inconsistency of LMR cut-off values.

Previous studies have shown that MRI features and parameters can be used for noninvasive prediction of pCR after NAC in invasive BC, and all have good diagnostic efficacy (17,22,23). ADC value can be used to quantitatively evaluate the diffusion of water molecules in tumors, which is widely used. Partridge et al. (22) reported that DWI can reflect the cytotoxic effect of chemotherapy, and mid-treatment ADC is a predictive marker for pCR. Han et al. (17) showed that the maximum diameter and enhancement pattern of DCE-MRI were independently associated with pCR after NAC. Li et al. (24) reported that both DWI and DCE-MRI can well predict the pathological response of invasive BC to NAC, and there is no significant difference in diagnostic performance between them. Tang et al. (25) showed that early clinical staging was associated with pCR in invasive BC. Choi et al. (26) used the pre-treatment ADC value to predict the response of invasive BC to NAC, and the results showed that a high pre-treatment ADC value could better predict pCR. Consistent with the results of this study, ADC_mean, enhancement pattern, and clinical T stage were favorable factors for pCR in invasive BC in this study.

Based on the analyses described above, three predictive models were developed: a clinical model, an MRI-based model, and a combined model. In the training cohort, both the clinical and MRI models demonstrated moderate predictive performance, with AUC values of 0.731 and 0.708, respectively. By contrast, the combined model achieved superior discrimination, yielding an AUC of 0.820 in the training set, 0.810 in the internal validation set, and 0.799 in the external validation set, outperforming both the clinical and MRI models. Wang et al. (4) developed a nomogram to predict pCR in invasive BC based on systemic immune-inflammatory markers before and after treatment, with an AUC of 0.754. Partridge et al. (22) used ADC values to assess the response to NAC for invasive BC and showed an AUC of 0.63 for the entire cohort. Wu et al. (27) showed that the AUC of the combination of clinicopathological features and pretreatment systemic inflammatory response index for predicting pCR to NAC in HER2-positive invasive BC was 0.74. Chen et al. (28) showed that the AUC of the MRI model in predicting pCR to NAC for invasive BC was 0.769. The results of this study showed that the AUC of the combined model in the training and validation groups was higher than those of the above studies. This indicated that clinicopathological features combined with inflammatory markers and MRI characteristics can effectively predict pCR in invasive BC.

There are limitations in this study. Firstly, the relatively small size of the external validation cohort (n=54) represents a notable limitation and may affect the precision and generalizability of the model performance. Further studies with larger sample sizes are needed to confirm our findings. Second, differences in patient characteristics were noted between the training and external validation sets; however, these discrepancies may mirror real-world clinical conditions and contribute to a more conservative and realistic assessment of model performance. Thirdly, MRI scans were acquired using machines from two different vendors across two centers, which may have introduced measurement biases.

Conclusions

This study demonstrated that clinical T stage, LMR, SII, enhancement pattern, and ADC_mean were the independent factors for predicting pCR of invasive BC, a combined model based on clinicopathological features, combined with inflammatory markers and MRI characteristics, can effectively predict pCR in invasive BC.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2841/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2841/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2841/prf

Funding: This work was supported by Henan Medical Science and Technology Research Project (No. LHGJ20240123).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2841/coif). The authors have 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. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committees of Henan Cancer Hospital (No. 2024-170-001) and Longyan First Affiliated Hospital of Fujian Medical University (No. LYREC2025-k105-01). Informed consent was waived in this retrospective study.

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. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. Han X, Cao W, Wu L, et al. Radiomics Assessment of the Tumor Immune Microenvironment to Predict Outcomes in Breast Cancer. Front Immunol 2021;12:773581. [Crossref] [PubMed]
3. De La Cruz LM, Harhay MO, Zhang P, et al. Impact of Neoadjuvant Chemotherapy on Breast Cancer Subtype: Does Subtype Change and, if so, How? : IHC Profile and Neoadjuvant Chemotherapy. Ann Surg Oncol 2018;25:3535-40. [Crossref] [PubMed]
4. Wang H, Huang Z, Xu B, et al. The predictive value of systemic immune-inflammatory markers before and after treatment for pathological complete response in patients undergoing neoadjuvant therapy for breast cancer: a retrospective study of 1994 patients. Clin Transl Oncol 2024;26:1467-79. [Crossref] [PubMed]
5. Spring LM, Bar Y, Isakoff SJ. The Evolving Role of Neoadjuvant Therapy for Operable Breast Cancer. J Natl Compr Canc Netw 2022;20:723-34. [Crossref] [PubMed]
6. Dowling GP, Daly GR, Hegarty A, et al. Predictive value of pretreatment circulating inflammatory response markers in the neoadjuvant treatment of breast cancer: meta-analysis. Br J Surg 2024;111:znae132. [Crossref] [PubMed]
7. Cortazar P, Zhang L, Untch M, et al. Pathological complete response and long-term clinical benefit in breast cancer: the CTNeoBC pooled analysis. Lancet 2014;384:164-72. [Crossref] [PubMed]
8. Pu S, Wang K, Liu Y, et al. Nomogram-derived prediction of pathologic complete response (pCR) in breast cancer patients treated with neoadjuvant chemotherapy (NCT). BMC Cancer 2020;20:1120. [Crossref] [PubMed]
9. Costigliola R, Marino AV, Fioretto I. How the blooming effect can interfere on choroidal vascularity index assessments using optical coherence tomography. Quant Imaging Med Surg 2023;13:558-9. [Crossref] [PubMed]
10. Xue LB, Liu YH, Zhang B, et al. Prognostic role of high neutrophil-to-lymphocyte ratio in breast cancer patients receiving neoadjuvant chemotherapy: Meta-analysis. Medicine (Baltimore) 2019;98:e13842. [Crossref] [PubMed]
11. Li Y, Chen Y, Zhao R, et al. Development and validation of a nomogram based on pretreatment dynamic contrast-enhanced MRI for the prediction of pathologic response after neoadjuvant chemotherapy for triple-negative breast cancer. Eur Radiol 2022;32:1676-87. [Crossref] [PubMed]
12. Zhu Y, Zhang S, Wei W, et al. Intra- and peritumoral radiomics nomogram based on DCE-MRI for the early prediction of pathological complete response to neoadjuvant chemotherapy in breast cancer. Front Oncol 2025;15:1561599. [Crossref] [PubMed]
13. Fiste O, Mavrothalassitis E, Kokkalis A, et al. Inflammation-related biomarkers as predictors of pathological complete response in early-stage breast cancer. Clin Transl Oncol 2025;27:2453-60. [Crossref] [PubMed]
14. Zheng G, Peng J, Shu Z, et al. Predicting pathological complete response to neoadjuvant chemotherapy in breast cancer patients: use of MRI radiomics data from three regions with multiple machine learning algorithms. J Cancer Res Clin Oncol 2024;150:147. [Crossref] [PubMed]
15. Li W, Le NN, Onishi N, et al. Diffusion-Weighted MRI for Predicting Pathologic Complete Response in Neoadjuvant Immunotherapy. Cancers (Basel) 2022;14:4436. [Crossref] [PubMed]
16. Ogston KN, Miller ID, Payne S, et al. A new histological grading system to assess response of breast cancers to primary chemotherapy: prognostic significance and survival. Breast 2003;12:320-7. [Crossref] [PubMed]
17. Han X, Yang H, Jin S, et al. Prediction of pathological complete response to neoadjuvant chemotherapy in patients with breast cancer using a combination of contrast-enhanced ultrasound and dynamic contrast-enhanced magnetic resonance imaging. Cancer Med 2023;12:1389-98. [Crossref] [PubMed]
18. Biesheuvel CJ, Vergouwe Y, Steyerberg EW, et al. Polytomous logistic regression analysis could be applied more often in diagnostic research. J Clin Epidemiol 2008;61:125-34. [Crossref] [PubMed]
19. Mosalem O, Alsubait S, Kherallah S, et al. Impact of pre-treatment lymphocyte monocyte ratio and neutrophil lymphocyte ratio on pathologic response in breast cancer patients receiving neoadjuvant chemotherapy. J Clin Oncol 2021;39:e12623. [Crossref]
20. Dong J, Sun Q, Pan Y, et al. Pretreatment systemic inflammation response index is predictive of pathological complete response in patients with breast cancer receiving neoadjuvant chemotherapy. BMC Cancer 2021;21:700. [Crossref] [PubMed]
21. Peng Y, Chen R, Qu F, et al. Low pretreatment lymphocyte/monocyte ratio is associated with the better efficacy of neoadjuvant chemotherapy in breast cancer patients. Cancer Biol Ther 2020;21:189-96. [Crossref] [PubMed]
22. Partridge SC, Zhang Z, Newitt DC, et al. ACRIN 6698 trial: Quantitative diffusion-weighted MRI to predict pathologic response in neoadjuvant chemotherapy treatment of breast cancer. J Clin Oncol 2017;35:11520. [Crossref]
23. Gelardi F, Sollini M, Zanca R, et al. Utility of [18F]FDG PET/CT in predicting pathological complete response to neoadjuvant therapy in breast cancer: A prospective study. J Clin Oncol 2024;42:3070. [Crossref]
24. Li Z, Li J, Lu X, et al. The diagnostic performance of diffusion-weighted imaging and dynamic contrast-enhanced magnetic resonance imaging in evaluating the pathological response of breast cancer to neoadjuvant chemotherapy: A meta-analysis. Eur J Radiol 2021;143:109931. [Crossref] [PubMed]
25. Tang S, Xiang C, Yang Q. The diagnostic performance of CESM and CE-MRI in evaluating the pathological response to neoadjuvant therapy in breast cancer: a systematic review and meta-analysis. Br J Radiol 2020;93:20200301. [Crossref] [PubMed]
26. Choi BB. Effectiveness of ADC Difference Value on Pre-neoadjuvant Chemotherapy MRI for Response Evaluation of Breast Cancer. Technol Cancer Res Treat 2021;20:15330338211039129. [Crossref] [PubMed]
27. Wu HY, Yang JR, Hung CC, et al. A nomogram to predict pathological complete response of neoadjuvant systemic therapy in HER2-positive breast cancer based on clinicopathologic characteristics and pretreatment systemic inflammation response index: A dual-center study. J Clin Oncol 2022;40:e12590. [Crossref]
28. Chen P, Wang C, Lu R, et al. Multivariable Models Based on Baseline Imaging Features and Clinicopathological Characteristics to Predict Breast Pathologic Response after Neoadjuvant Chemotherapy in Patients with Breast Cancer. Breast Care (Basel) 2022;17:306-15. [Crossref] [PubMed]