Dynamic contrast-enhanced magnetic resonance imaging-derived three-dimensional tumor volume improves prognostic stratification in breast cancer: a retrospective cohort study

Introduction

Background

Breast cancer is the most common malignancy affecting women globally (1,2). With the progressive implementation of diverse therapeutic modalities in clinical practice, optimizing treatment selection and sequencing to maximize efficacy while minimizing treatment-related toxicity has become a major focus in breast oncology research (3-5). Currently, clinical decision-making primarily relies on tumor-node-metastasis (TNM) staging and molecular subtyping, as different stages and subtypes require distinct multimodal treatment strategies (6,7). Despite ongoing advances in staging systems and treatment paradigms, maximum tumor diameter (TD) remains the principal criterion for T staging in the current American Joint Committee on Cancer (AJCC) system and in major international guidelines, including those of the National Comprehensive Cancer Network (NCCN) and European Society for Medical Oncology (ESMO) (8,9). Its continued use as a global standard is largely attributable to its simplicity, relatively good reproducibility, long-standing standardization, and feasibility across diverse clinical settings. Therefore, any novel anatomical assessment metric proposed as a complement to the existing staging framework must demonstrate not only theoretical advantages but also clinically meaningful value beyond that provided by maximum TD.

The TNM staging system itself has undergone several major revisions in response to advances in medical technology and evolving diagnostic and therapeutic concepts. The TNM staging system, introduced by French pathologist Pierre Denoix in 1943, classifies cancers according to tumor size (T), lymph node involvement (N), and distant metastasis (M) (10). Subsequent editions have progressively refined staging: the fourth edition clarified T4 lesions with local invasion (11); the fifth incorporated biomarkers such as estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2), emphasizing their prognostic relevance (12); the sixth integrated molecular characteristics to enhance clinical evaluation (13); and the eighth highlighted advanced imaging and histopathology, defining microinvasive carcinoma as ≤1.0 mm and pT1a invasive tumors as 1.0–1.5 mm (6,7). These continuous refinements have played a pivotal role in guiding clinical decision-making and addressing practical challenges in oncology.

Rationale and knowledge gap

The TNM staging system has continuously evolved with advances in medical technology. Tumor burden, which serves as the foundation for T-staging, has traditionally been assessed using maximum diameter due to historical technical limitations (6,14). However, marked morphological heterogeneity can lead to over- or under-estimation of disease severity when relying solely on diameter. Parker et al. analyzed 955 clear cell renal cell carcinoma patients who underwent surgery from 1980 to 2004 and found that tumor volume (TV) estimated from maximum diameter often differed from true TV, particularly in larger tumors (15). This highlights the need for more accurate measures of tumor burden to guide treatment in precision medicine.

Although previous studies have suggested that volumetric parameters may be superior to simple linear measurements, the controversy surrounding their clinical value remains unresolved. One major reason is the inconsistency of imaging modalities across studies. Ultrasound, mammography, and magnetic resonance imaging (MRI) differ in spatial resolution, soft-tissue contrast, delineation of tumor margins, and detection of multifocal or multicentric disease, thereby directly affecting the accuracy and comparability of tumor measurements. Compared with two-dimensional modalities such as ultrasound, MRI—particularly dynamic contrast-enhanced MRI (DCE-MRI)—offers advantages in visualizing lesion enhancement boundaries and three-dimensional spatial relationships, making it better suited for three-dimensional reconstruction and volumetric assessment (16-18).

With the increasing accessibility of DCE-MRI, its application in breast cancer has broadened, providing superior contrast resolution and enabling three-dimensional tumor reconstruction compared with mammography and ultrasound (19-22). Combined with advanced post-processing techniques, MRI allows precise volumetric measurement regardless of tumor shape, and multiple studies have confirmed its high accuracy (23-26).

However, volumetric assessment is not without challenges. In cases of multicentric or multifocal disease, poorly defined lesion margins, marked background parenchymal enhancement, or non-uniform segmentation criteria, volumetric evaluation may still be influenced by image quality, observer experience, and differences in post-processing methods. This is also one of the major reasons why, despite their theoretical advantages, volumetric parameters have not yet been incorporated into routine staging systems.

Moreover, evidence remains limited regarding whether TV provides additional prognostic information beyond conventional T staging based on maximum TD in non-metastatic, particularly early-stage, breast cancer.

Objective

Therefore, this retrospective study compared the prognostic performance of MRI-derived TV with that of conventional diameter-based T staging, aiming to determine whether TV may provide additional value for refining TNM staging. We present this article in accordance with the STROBE reporting checklist (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0067/rc).

Methods

Patients

This single-center retrospective cohort study was approved by the Ethics Committee of The First Affiliated Hospital of Nanjing Medical University (No. 2025-SR-1223) and was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Owing to the retrospective nature of the study, the requirement for informed consent was waived, and all patient data were anonymized prior to analysis.

Female patients with breast cancer who underwent surgical treatment at The First Affiliated Hospital of Nanjing Medical University between January 2016 and June 2019 were retrospectively screened. To minimize selection bias, all consecutive patients who met the eligibility criteria during the study period were included. Given the retrospective and exploratory nature of the study, no formal prospective sample size calculation was performed. The study sample size was determined by the total number of eligible patients with complete imaging, clinicopathological, and follow-up data available during the predefined recruitment period. Because the number of overall survival (OS) events was limited, OS analyses were considered supportive, whereas disease-free survival (DFS) was predefined as the primary endpoint for prognostic model evaluation.

The inclusion criteria were as follows: (I) histologically confirmed invasive ductal carcinoma with no evidence of distant metastasis at diagnosis; (II) availability of preoperative 3.0-T breast MRI suitable for three-dimensional tumor volumetric analysis; (III) normal major organ function and no history of other malignancies; (IV) complete clinicopathological and follow-up data; (V) no receipt of neoadjuvant therapy prior to surgery; and (VI) receipt of standardized treatment in accordance with contemporaneous clinical practice guidelines. Patients with distant metastasis at diagnosis, inadequate MRI quality for volumetric assessment, incomplete clinicopathological or follow-up data, prior neoadjuvant therapy, or a history of other malignancies were excluded.

Imaging acquisition and tumor segmentation

All patients underwent preoperative breast MRI using a 3.0-T scanner (Siemens MAGNETOM Trio Tim system, Siemens Healthcare). The imaging protocol included conventional T1-weighted imaging (T1WI) and DCE-MRI. DCE-MRI was performed using a gradient-echo sequence, with five post-contrast phases acquired. Images demonstrating the highest tumor-to-normal tissue contrast, the clearest enhancement margins, and the absence of motion artifacts were selected for tumor segmentation.

Tumor segmentation and volumetric measurements were performed using 3D Slicer software (version 5.5.0; www.slicer.org) by one radiologist with more than 10 years of experience and three breast surgeons, each with over 5 years of clinical experience, as illustrated in Figure 1. To reduce measurement bias and ensure comparability across patients, all observers underwent standardized training prior to segmentation, including calibration exercises using reference cases to harmonize tumor boundary delineation criteria. Inter-observer agreement was assessed and demonstrated good reproducibility, with a mean Dice similarity coefficient of 0.88±0.08.

Figure 1 Three-dimensional tumor reconstruction based on 3.0-T dynamic contrast-enhanced breast magnetic resonance imaging using 3D Slicer software. (A-C) Representative axial (A), sagittal (B), and coronal (C) MRI slices showing the tumor (highlighted in green) for segmentation. (D) Three-dimensional rendering of the segmented tumor. Tumor volume was calculated as 4.87242 cm³ with a software measurement precision of 0.00001 cm³. 3D, three-dimensional; MRI, magnetic resonance imaging.

Pathological evaluation and molecular marker assessment

Clinicopathological data were retrospectively retrieved from the institutional electronic medical record system, including patient age, pathological T stage, regional lymph node status, histological grade, immunohistochemical (IHC) findings, and lymphovascular invasion (LVI) status. Pathological T stage and regional lymph node status (LNS) were determined according to the AJCC 8th edition TNM staging system. Maximum TD was measured postoperatively on pathological specimens as the largest cross-sectional diameter (cm, 0.1 cm accuracy). Histological grade was classified as grade I–III based on the Nottingham grading system.

IHC markers included ER, PR, HER2, and Ki-67 proliferation index. Hormone receptor positivity (HR+) was defined as ≥1% nuclear staining for ER and/or PR. HER2 status was determined according to ASCO/CAP guidelines: tumors with immunohistochemistry (IHC) scores of 0, 1+, or 2+ without HER2 gene amplification on in situ hybridization, were classified as HER2-negative (HER2−), whereas tumors with IHC 3+ or IHC 2+ with HER2 gene amplification on in situ hybridization were classified as HER2-positive (HER2+). Ki-67 expression was evaluated in accordance with the St. Gallen International Expert Consensus, with ≥20% defined as high proliferation (Ki-67-high) and <20% as low proliferation (Ki-67-low). LVI was defined as the presence of tumor cells within lymphatic or blood vessels.

Follow-up and endpoint definitions

To minimize information bias, unified data sources and standardized endpoint definitions were applied across the entire cohort. Follow-up commenced at the time of surgery and was conducted through scheduled outpatient visits, supplemented by telephone interviews and review of electronic medical records. In routine clinical practice, patients were generally followed every 3–6 months during the first 2 years after surgery and every 6–12 months thereafter, depending on clinical status and treatment requirements. Follow-up was censored on September 30, 2024.

DFS was defined as the interval from surgery to the first occurrence of local recurrence, regional recurrence, distant metastasis, disease progression, or death from any cause. Patients without an event were censored at the date of last follow-up. OS was defined as the interval from surgery to death from any cause, with patients alive at last follow-up censored accordingly.

Suspected recurrence or progression was assessed using imaging examinations performed during follow-up, including breast ultrasound, mammography, breast MRI, chest or abdominal computed tomography, bone scintigraphy, and other clinically indicated imaging modalities. Whenever feasible, recurrent or metastatic lesions were further verified by histopathological or cytological examination obtained through biopsy or surgery. In the absence of pathological confirmation, recurrence or progression was determined on the basis of imaging findings together with the corresponding clinical assessment documented in the medical records.

Statistical analysis

The distribution of continuous variables was assessed using the Shapiro-Wilk test. Normally distributed variables are presented as mean ± standard deviation, whereas non-normally distributed variables are expressed as median with interquartile range (IQR). Categorical variables are reported as counts and percentages.

Spearman rank correlation analysis was used to evaluate the association between TV and maximum TD. Comparisons of non-normally distributed continuous variables were performed using the Mann-Whitney U test. Survival outcomes were analyzed using the Kaplan-Meier method, with differences between groups assessed by the log-rank test. Prognostic factors were evaluated using univariable and multivariable Cox proportional hazards regression models. Clinically relevant covariates, including lymph node-positive count and LVI, were incorporated into multivariable models to control for potential confounding. Model discrimination was quantified using the concordance index (C-index). Model robustness was further assessed through internal validation with 1,000 bootstrap resampling iterations. All statistical analyses were performed using R software (version 4.5.1; R Foundation for Statistical Computing). All tests were two-sided, and a P value <0.05 was considered statistically significant.

Results

Baseline clinicopathological characteristics of patients

The study cohort consisted of 574 patients with invasive breast cancer, with a mean age of 49.67±9.80 years, of whom 53.8% were aged ≤50 years. The median TV and TD were 2.07 cm³ [interquartile range (IQR), 0.95–3.84 cm³] and 2.00 cm (IQR, 1.50–2.50 cm), respectively. Postoperative pathological T-stage distribution was as follows: T1, 57.1%; T2, 41.5%; and T3, 1.4%. Regional lymph node metastasis was present in 38.7% of patients; among those with nodal involvement, the median number of positive lymph nodes was 0 (IQR, 0–1). According to TNM staging, stage I, II, and III disease accounted for 41.1%, 46.5%, and 12.4% of cases, respectively. Molecular subtypes were predominantly HR+/HER2− (62.9%), and a high Ki-67 index was observed in 72.6% of patients. Histological grade III tumors accounted for 51.4% of cases, and LVI was present in 24.4%. The median follow-up time was 81.0 months (IQR, 72.0–92.0 months) for DFS and 82.0 months (IQR, 73.0–93.0 months) for OS (Table 1).

Correlation between TV and TD

In the overall cohort, TV was strongly positively correlated with TD, with Spearman’s correlation coefficient r=0.781 (95% CI: 0.734–0.820, P<0.001). Further analysis stratified by TD indicated that in the TD ≤2.5 cm subgroup (n=447, 77.9%), the correlation between TV and TD was r=0.723 (95% CI: 0.653–0.773, P<0.001), whereas in the TD >2.5 cm subgroup (n=127, 22.1%), the correlation markedly decreased to r=0.341 (95% CI: 0.141–0.494, P<0.001) (Table 2). The scatter plot visually demonstrated that for larger tumors, TD may not accurately reflect TV (Figure 2). An illustrative case further shows a substantial discrepancy between TD and TV, where TD is relatively large but the tumor has a flat morphology, resulting in a relatively small TV (Figure 3).

Figure 2 Relationship between maximum tumor diameter and tumor volume (scatter plot).

Figure 3 Three-dimensional dynamic contrast-enhanced magnetic resonance imaging–based reconstruction of the tumor. (A-C) Representative MRI slices in axial (A), sagittal (B), and coronal (C) planes, showing a tumor (highlighted in green). (D) 3D reconstruction of the tumor. The tumor has a maximum diameter of 6.5 cm but a relatively small tumor volume of 3.10555 cm³, displaying an elongated and flat morphology, indicating that tumor diameter alone may overestimate the actual tumor burden. 3D, three-dimensional; MRI, magnetic resonance imaging.

TV staging correlates with prognostic risk

The 574 patients were stratified according to overall OS and DFS into OS status groups (survival vs. death) and DFS status groups (disease-free vs. recurrence/metastasis) to compare TV. The median TV was 4.82 cm³ in the death group and 2.01 cm³ in the survival group (P<0.001), and 5.13 cm³ in the recurrence/metastasis group versus 1.98 cm³ in the disease-free group (P<0.001), indicating that TV may serve as a potential prognostic indicator in breast cancer (Figure 4). Based on optimal cutoff values, TV was further categorized into V1 (TV ≤2 cm³), V2 (2 cm³ < TV ≤5 cm³), and V3 (TV >5 cm³). Kaplan-Meier analysis revealed significant differences in 5-year DFS among the V1, V2, and V3 groups (94.2%, 87.3%, and 75.0%, respectively) and in 5-year OS (97.5%, 93.2%, and 82.6%, respectively), demonstrating significant prognostic stratification (log-rank test, P<0.001) (Figure 5).

Figure 4 Relationship between tumor volume and survival outcomes in breast cancer patients. (A) Comparison of TV between patients alive and deceased. (B) Comparison of TV between patients who remained disease-free and those who experienced recurrence or progression. Tumor volume is shown as median with interquartile range; P values were calculated using the Mann-Whitney U test. DFS, disease-free survival; OS, overall survival; TV, tumor volume.

Figure 5 Kaplan-Meier survival curves according to tumor volume groups. (A) Overall survival and (B) disease-free survival stratified by tumor volume (V1–V3), V1 (TV ≤2 cm³), V2 (2 cm³ < TV ≤5 cm³), and V3 (TV >5 cm³). Kaplan-Meier curves show significantly decreased survival with increasing tumor volume (log-rank, P<0.001).

Cox regression analyses and multivariable model comparison for DFS

Univariable Cox regression analyses were first performed to examine the associations of TD, TV, and other clinicopathological variables with survival outcomes (Table 3). For DFS, both TD- and TV-based groupings were significantly associated with prognosis. Compared with the V1 group (TV ≤2 cm³), patients in the V2 group (2 cm³ < TV ≤5 cm³) and V3 group (TV >5 cm³) had significantly increased risks of DFS events, with hazard ratios (HRs) of 3.100 (95% CI: 1.356–7.087, P=0.007) and 8.148 (95% CI: 3.620–18.340, P<0.001), respectively. Likewise, compared with the T1 group (TD ≤2 cm), the HRs for the T2 group (2 cm < TD ≤5 cm) and T3 group (TD >5 cm) were 2.490 (95% CI: 1.352–4.586, P=0.003) and 9.159 (95% CI: 3.044–27.557, P<0.001), respectively. For OS, both TV- and TD-based classifications were associated with poorer prognosis; however, within the TD-based grouping, only the comparison between T2 and T1 reached statistical significance. In addition, a higher number of positive lymph nodes, lymph node involvement, and LVI were all associated with poorer DFS and OS (all P<0.05).

Because DFS was predefined as the primary endpoint, multivariable Cox regression models were subsequently constructed for DFS analysis (Table 4). Based on the univariable results and clinical relevance, the number of lymph node-positive (LN+) nodes and LVI were incorporated into the baseline model (Model 1). TD grouping and TV grouping were then added separately to generate Model 2 and Model 3, respectively. In Model 2, compared with the T1 group, the adjusted HRs were 2.227 (95% CI: 1.185–4.187, P=0.01) for T2 and 4.691 (95% CI: 1.424–15.457, P=0.01) for T3. In Model 3, compared with the V1 group, the adjusted HRs were 2.721 (95% CI: 1.181–6.271, P=0.02) for V2 and 6.069 (95% CI: 2.640–13.950, P<0.001) for V3. These findings indicate that both TD- and TV-based classifications remained independently associated with DFS after adjustment for LN+ count and LVI.

Model discrimination was assessed using the C-index. The baseline model (Model 1) yielded a C-index of 0.679, which increased to 0.702 after the addition of TD grouping (Model 2) and to 0.744 after the addition of TV grouping (Model 3). Following 1,000 bootstrap resampling iterations, the optimism-corrected C-index values were 0.681 (95% CI: 0.597–0.773) for Model 1, 0.704 (95% CI: 0.619–0.795) for Model 2, and 0.749 (95% CI: 0.674–0.818) for Model 3.

The proportional hazards assumption was further examined using Schoenfeld residual tests (Table 5). No significant violation was observed in any of the three models, supporting the validity and robustness of the Cox regression analyses.

The TV-based model demonstrates superior prognostic performance in subgroups characterized by HER2−, Ki-67-high, and LN+

Model performance was evaluated across tumor size, clinicopathological features, molecular phenotype, and proliferative activity. In the LN+ subgroup (n=222), the TV-based model achieved a C-index of 0.778, higher than the TD-based model at 0.752 (difference =0.027, P=0.02). In the lymphovascular invasion-negative (LVI−) subgroup (n=434), C-index was 0.708 versus 0.663 for the TD-based model (difference =0.045, P=0.006). In the HER2− subgroup (n=447), the TV-based model had a C-index of 0.717 compared with 0.683 for the TD-based model (difference =0.034, P=0.02). In the Ki-67-high subgroup (>20%, n=417), C-index was 0.715 versus 0.680 (difference =0.035, P=0.03). Across other subgroups, the TV-based model consistently showed higher C-indices than the TD-based model, though the differences did not reach statistical significance (all P>0.05) (Table 6).

Discussion

Key findings

Tumor staging is a critical determinant of prognosis and treatment decisions in breast cancer. The widely used AJCC TNM system enables standardized reporting of cancer data worldwide. With advances in diagnostic techniques and increased disease awareness, a growing proportion of patients are diagnosed with early-stage breast cancer (27-29), typically without lymph node involvement or distant metastasis, rendering accurate T staging particularly important for postoperative management. Early-stage breast cancer is classified as T1–T3, with most patients falling into the T1 or T2 categories, and T1 further subdivided into T1a–c to refine prognostic stratification (20). Treatment strategies are adapted according to T stage and molecular subtype (30-33), underscoring the clinical importance of tumor size in current management paradigms.

Despite its central role, conventional T staging relies on maximum TD as a surrogate for tumor burden, raising concerns about whether TD adequately reflects the true extent of disease. The present study directly addressed this issue by systematically comparing the prognostic value of three-dimensional TV with that of TD. Although TD and TV were strongly correlated overall, this association weakened substantially in tumors larger than 2.5 cm. Notably, some irregularly shaped tumors exhibited large diameters but relatively small volumes, such that lesions measuring up to 4 cm in diameter had volumes comparable to those of 2 cm tumors—a phenomenon also reported by Sevgül Fakı in papillary thyroid cancer (34). These findings indicate that TD approximates tumor burden only under certain conditions and may overestimate or underestimate disease severity in individual patients.

Using preoperative 3.0-T DCE-MRI combined with semi-automatic segmentation in 3D Slicer, TV was directly measured, enabling more precise quantification of tumor burden. To evaluate the prognostic value of TV relative to TD, univariate Cox regression was first performed to identify variables associated with outcomes, followed by multivariable Cox regression analyses. Given the limited number of events in this early-stage, non-metastatic cohort, DFS was selected as the primary endpoint. A baseline model incorporating the number of LN+ metastases and LVI was constructed (Model 1), to which TD (Model 2) or TV (Model 3) was subsequently added.

Model 3 (TV + LN+ + LVI) yielded a C-index of 0.744, which was higher than that of Model 2 (0.702) and Model 1 (0.679). In multivariable Cox regression analysis, compared with the V1 group (TV ≤2 cm³), both the V2 group (>2–5 cm³) and the V3 group (>5 cm³) were independently associated with poorer DFS, with HRs (95% CIs) of 2.721 (1.181–6.271) and 6.069 (2.640–13.950), respectively. In the TD-based model, compared with the T1 group, the HRs (95% CIs) for the T2 and T3 groups were 2.227 (1.185–4.187) and 4.691 (1.424–15.457), respectively. In addition, the bootstrap-corrected C-index of the TV-based model [0.749 (95% CI: 0.674–0.818)] was also higher than that of the TD-based model [0.704 (95% CI: 0.619–0.795)]. These findings indicate that, after adjustment for LN+ count and LVI, both TV- and TD-based classifications remained independently associated with DFS, while the model incorporating TV showed a numerically higher level of discrimination.

Subgroup analyses further identified patient populations deriving particular benefit from TV-based staging. In LN+ patients, the C-index was higher for the TV model than for the TD model (0.778 vs. 0.752; difference =0.027, P=0.02), potentially reflecting more aggressive tumor biology, vascular invasion, and higher proliferative activity in this subgroup. In HER2− patients, TV also improved prognostic discrimination over TD (0.717 vs. 0.683; difference =0.034, P=0.02), supporting the value of more precise anatomic staging in settings where targeted therapies are limited. Similarly, in the Ki-67-high subgroup (n=417), the TV-based model demonstrated higher predictive accuracy than the TD-based model (0.715 vs. 0.680; difference =0.035, P=0.03), likely because TV more accurately captures volumetric accumulation associated with active proliferation, whereas TD may underestimate tumor burden in morphologically heterogeneous tumors.

Strengths and limitations

This study has several notable strengths. First, it focused on a well-defined cohort of early-stage, non-metastatic breast cancer patients, a population in which T staging plays a pivotal role in postoperative risk stratification and therapeutic decision-making. Second, TV was derived from high-resolution preoperative 3.0-T DCE-MRI using semi-automatic three-dimensional segmentation, enabling direct quantification of tumor burden rather than reliance on surrogate linear measurements. This approach addresses an important methodological limitation of many previous breast cancer studies, in which TV was indirectly estimated using simplified geometric formulas. Third, prognostic performance was systematically evaluated using multivariable Cox regression models, C-index comparisons, and clinically meaningful subgroup analyses, thereby strengthening the robustness and interpretability of the findings.

Several limitations should also be acknowledged. First, this was a single-center retrospective analysis, and no formal prospective sample size calculation was performed because the cohort consisted of all consecutive eligible patients within the prespecified study period. Although this design reflects real-world clinical practice, it may still be subject to the inherent selection bias of retrospective studies. Second, the number of outcome events was relatively limited, particularly for OS, which may have reduced the statistical power to detect differences in model performance for low-event-rate survival endpoints. Accordingly, DFS was selected as the primary endpoint, whereas OS findings should be regarded as supportive and interpreted with caution. Third, although all patients received postoperative treatment in accordance with contemporary clinical guidelines, variability in adjuvant treatment strategies remained inevitable, and unmeasured biological heterogeneity may also have introduced residual confounding. In addition, although the inter-observer Dice coefficients indicated acceptable segmentation reproducibility, the semi-automatic segmentation approach remained partly operator-dependent. These factors may affect the stability and generalizability of the findings and underscore the need for further prospective multicenter validation.

Comparison with similar research

Our findings are consistent with an expanding body of evidence demonstrating that volumetric tumor parameters outperform diameter-based measurements across multiple cancer types. In papillary thyroid cancer, Sevgül Fakı et al. showed that tumors with relatively large diameters may exhibit small volumes due to irregular growth patterns, resulting in discordance between diameter-based measurements and actual tumor burden (34). Similar observations have been reported in colorectal cancer (35), esophageal cancer (36,37), head and neck squamous cell carcinoma (38), and oral cancer (39).

Notably, the CAIRO5 study demonstrated that total metastatic volume predicted disease progression more accurately than lesion count in metastatic colorectal cancer (40), reinforcing the concept that volumetric burden better reflects biological aggressiveness. In head and neck cancer radiotherapy, Zahid et al. reported that TV-based monitoring improved outcome prediction compared with conventional metrics (41). In breast cancer, although several studies have explored associations between TV and prognosis, most relied on indirect volume estimation using ellipsoid formulas, which may inadequately capture complex three-dimensional tumor morphology (42). By contrast, our study provides direct imaging-based volumetric assessment, extending the relevance of TV to early-stage breast cancer, a setting in which TD has traditionally dominated staging systems.

Explanations of findings

Several factors may explain the superior prognostic performance of TV. TD represents a one-dimensional measurement and implicitly assumes regular tumor geometry, an assumption that becomes increasingly invalid as tumors enlarge and develop heterogeneous or infiltrative growth patterns. As tumor size increases, asymmetric volumetric expansion may occur without proportional increases in maximum diameter, leading to misclassification of tumor burden when TD alone is used.

In contrast, TV integrates spatial growth across all dimensions and may better reflect cumulative malignant cell burden, proliferative activity, and the extent of invasive fronts. This may explain why the advantage of TV-based staging was particularly evident in LN+, HER2−, and Ki-67-high subgroups, which are often characterized by aggressive biological behavior and limited therapeutic targets. In such contexts, reliance on TD may obscure clinically meaningful risk differences that are more accurately captured by volumetric assessment.

Implications and actions needed

The findings of this study have important clinical and research implications. Clinically, incorporating TV into preoperative assessment may improve risk stratification beyond conventional T staging, particularly in early-stage breast cancer patients with aggressive biological features. TV-based staging could help identify patients who may benefit from intensified systemic therapy or closer surveillance, even when classified as lower risk by diameter-based criteria.

From a research perspective, these results support further refinement of anatomic staging systems by integrating volumetric metrics derived from modern imaging. Future studies should prioritize prospective, multicenter validation across diverse imaging platforms and populations to enhance generalizability. The implementation of automated deep learning-based segmentation techniques may further improve efficiency, reproducibility, and clinical feasibility. Ultimately, integrating TV into precision staging frameworks may represent a meaningful step toward more individualized prognostic assessment and treatment planning in breast cancer.

Conclusions

Three-dimensional reconstruction of breast tumors using 3D Slicer enabled direct measurement of TV from preoperative MRI. In this single-center retrospective cohort, TV-based models showed higher discrimination than diameter-based models for prognostic assessment, particularly for DFS. These findings suggest that TV may complement conventional diameter-based staging; however, prospective multicenter studies are required before volumetric assessment can be considered for integration into future clinical staging frameworks.

Acknowledgments

The authors gratefully acknowledge the support of the Department of Breast Surgery and the Department of Radiology at The First Affiliated Hospital of Nanjing Medical University for their assistance in data collection and imaging acquisition. We also thank all clinicians, radiologists, and research staff who contributed to patient management and follow-up. The authors appreciate the technical support provided by the 3D Slicer development community.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0067/rc

Data Sharing Statement: Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0067/dss

Peer Review File: Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0067/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0067/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. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The retrospective study was approved by the Ethics Committee of The First Affiliated Hospital of Nanjing Medical University (No. 2025-SR-1223). The requirement for informed consent was waived due to the retrospective nature of the study and the use of anonymized clinical and imaging data.

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. Siegel RL, Kratzer TB, Giaquinto AN, et al. Cancer statistics, 2025. CA Cancer J Clin 2025;75:10-45. [Crossref] [PubMed]
2. Loibl S, Poortmans P, Morrow M, et al. Breast cancer. Lancet 2021;397:1750-69. [Crossref] [PubMed]
3. Gradishar WJ, Moran MS, Abraham J, et al. Breast Cancer, Version 3.2024, NCCN Clinical Practice Guidelines in Oncology. J Natl Compr Canc Netw 2024;22:331-57. [Crossref] [PubMed]
4. Waks AG, Winer EP. Breast Cancer Treatment: A Review. JAMA 2019;321:288-300. [Crossref] [PubMed]
5. Xiong X, Zheng LW, Ding Y, et al. Breast cancer: pathogenesis and treatments. Signal Transduct Target Ther 2025;10:49. [Crossref] [PubMed]
6. Giuliano AE, Connolly JL, Edge SB, et al. Breast Cancer-Major changes in the American Joint Committee on Cancer eighth edition cancer staging manual. CA Cancer J Clin 2017;67:290-303.
7. Amin MB, Greene FL, Edge SB, et al. The Eighth Edition AJCC Cancer Staging Manual: Continuing to build a bridge from a population-based to a more “personalized” approach to cancer staging. CA Cancer J Clin 2017;67(2):93-9.
8. Loibl S, André F, Bachelot T, et al. Early breast cancer: ESMO Clinical Practice Guideline for diagnosis, treatment and follow-up. Ann Oncol 2024;35:159-82. [Crossref] [PubMed]
9. Gradishar WJ, Moran MS, Abraham J, et al. NCCN Guidelines® Insights: Breast Cancer, Version 5.2025. J Natl Compr Canc Netw 2025;23:426-36. [Crossref] [PubMed]
10. Rami-Porta R. The TNM classification of lung cancer-a historic perspective. J Thorac Dis 2024;16:8053-67. [Crossref] [PubMed]
11. American Joint Committee on Cancer. Manual for staging of cancer. 4th ed. Philadelphia: Lippincott; 1992.
12. American Joint Committee on Cancer. AJCC cancer staging manual. 5th ed. Philadelphia: Lippincott-Raven; 1997.
13. Singletary SE, Connolly JL. Breast cancer staging: working with the sixth edition of the AJCC Cancer Staging Manual. CA Cancer J Clin 2006;56:37-47; quiz 50-1.
14. Rezo A, Dahlstrom J, Shadbolt B, et al. Tumor size and survival in multicentric and multifocal breast cancer. Breast 2011;20:259-63. [Crossref] [PubMed]
15. Jorns J, Thiel DD, Lohse CM, et al. Three-dimensional tumour volume and cancer-specific survival for patients undergoing nephrectomy to treat pT1 clear-cell renal cell carcinoma. BJU Int 2012;110:956-60. [Crossref] [PubMed]
16. Tsarouchi M, Hoxhaj A, Portaluri A, et al. Breast cancer staging with contrast-enhanced imaging. The benefits and drawbacks of MRI, CEM, and dedicated breast CT. Eur J Radiol 2025;185:112013.
17. Azcona Sáenz J, Molero Calafell J, Román Expósito M, et al. Preoperative estimation of the pathological breast tumor size in architectural distortions: a comparison of DM, DBT, US, CEM, and MRI. Eur Radiol 2025;35:5635-45. [Crossref] [PubMed]
18. Daly AE, Anderman KJ, Holt LR, et al. Sizing It Up: Concordance between Breast Imaging and Pathologically Determined Tumor Measurement. Ann Surg Oncol 2025; [Crossref]
19. Gilbert FJ, Payne NR, Allajbeu I, et al. Comparison of supplemental breast cancer imaging techniques-interim results from the BRAID randomised controlled trial. Lancet 2025;405:1935-44. [Crossref] [PubMed]
20. Teichgraeber DC, Guirguis MS, Whitman GJ. Breast Cancer Staging: Updates in the AJCC Cancer Staging Manual, 8th Edition, and Current Challenges for Radiologists, From the AJR Special Series on Cancer Staging. AJR Am J Roentgenol 2021;217:278-90.
21. Cortadellas T, Argacha P, Acosta J, et al. Estimation of tumor size in breast cancer comparing clinical examination, mammography, ultrasound and MRI-correlation with the pathological analysis of the surgical specimen. Gland Surg 2017;6:330-5. [Crossref] [PubMed]
22. Shamir SB, Sasson AL, Margolies LR, et al. New Frontiers in Breast Cancer Imaging: The Rise of AI. Bioengineering (Basel) 2024;11:451. [Crossref] [PubMed]
23. Kovacs L, Eder M, Hollweck R, et al. Comparison between breast volume measurement using 3D surface imaging and classical techniques. Breast 2007;16:137-45. [Crossref] [PubMed]
24. Killaars RC, Preuß MLG, de Vos NJP, et al. Clinical Assessment of Breast Volume: Can 3D Imaging Be the Gold Standard? Plast Reconstr Surg Glob Open 2020;8:e3236. [Crossref] [PubMed]
25. Choppin SB, Wheat JS, Gee M, et al. The accuracy of breast volume measurement methods: A systematic review. Breast 2016;28:121-9. [Crossref] [PubMed]
26. Bulstrode N, Bellamy E, Shrotria S. Breast volume assessment: comparing five different techniques. Breast 2001;10:117-23. [Crossref] [PubMed]
27. Benitez Fuentes JD, Morgan E, de Luna Aguilar A, et al. Global Stage Distribution of Breast Cancer at Diagnosis: A Systematic Review and Meta-Analysis. JAMA Oncol 2024;10:71-8. [Crossref] [PubMed]
28. Ryzhov A, Corbex M, Piñeros M, et al. Comparison of breast cancer and cervical cancer stage distributions in ten newly independent states of the former Soviet Union: a population-based study. Lancet Oncol 2021;22:361-9. [Crossref] [PubMed]
29. Kong X, Liu Z, Cheng R, et al. Variation in Breast Cancer Subtype Incidence and Distribution by Race/Ethnicity in the United States From 2010 to 2015. JAMA Netw Open 2020;3:e2020303. [Crossref] [PubMed]
30. Prat A, Pineda E, Adamo B, et al. Clinical implications of the intrinsic molecular subtypes of breast cancer. Breast 2015;24:S26-35. [Crossref] [PubMed]
31. Ye F, Dewanjee S, Li Y, et al. Advancements in clinical aspects of targeted therapy and immunotherapy in breast cancer. Mol Cancer 2023;22:105. [Crossref] [PubMed]
32. Burstein HJ, Curigliano G, Thürlimann B, et al. Customizing local and systemic therapies for women with early breast cancer: the St. Gallen International Consensus Guidelines for treatment of early breast cancer 2021. Ann Oncol 2021;32:1216-35.
33. Wang J, Wu SG. Breast Cancer: An Overview of Current Therapeutic Strategies, Challenge, and Perspectives. Breast Cancer (Dove Med Press) 2023;15:721-30. [Crossref] [PubMed]
34. Fakı S, Tam AA, İnce N, et al. Comparison of Tumor Diameter and Tumor Volume in Terms of Aggressive Tumor Behavior and Prognosis in Papillary Thyroid Cancer. Cancers (Basel) 2025;17:1367. [Crossref] [PubMed]
35. Bhooshan N, Sharma NK, Badiyan S, et al. Pretreatment tumor volume as a prognostic factor in metastatic colorectal cancer treated with selective internal radiation to the liver using yttrium-90 resin microspheres. J Gastrointest Oncol 2016;7:931-7. [Crossref] [PubMed]
36. Wang R, Zhou X, Liu T, et al. Gross Tumor Volume Predicts Survival and Pathological Complete Response of Locally Advanced Esophageal Cancer After Neoadjuvant Chemoradiotherapy. Front Oncol 2022;12:898383. [Crossref] [PubMed]
37. Chen Y, Zhang Z, Jiang G, et al. Gross tumor volume is the prognostic factor for squamous cell esophageal cancer patients treated with definitive radiotherapy. J Thorac Dis 2016;8:1155-61. [Crossref] [PubMed]
38. Moumoulidis PT, Pavlopoulou A, Ünal R, et al. Nodal tumor volume as a prognostic factor for head and neck squamous cell carcinoma: a systematic review. Front Biosci (Landmark Ed) 2021;26:235-45. [Crossref] [PubMed]
39. Tofanelli M, Boscolo Nata F, Giudici F, et al. Is there a role for tumor volume in prediction of prognosis for oral cancer? Am J Otolaryngol 2023;44:103740. [Crossref] [PubMed]
40. Michiel Zeeuw J, Wesdorp NJ, Ali M, et al. Prognostic value of total tumor volume in patients with colorectal liver metastases: A secondary analysis of the randomized CAIRO5 trial with external cohort validation. Eur J Cancer 2024;207:114185. [Crossref] [PubMed]
41. Mohsin N, Enderling H, Brady-Nicholls R, et al. Simulating tumor volume dynamics in response to radiotherapy: Implications of model selection. J Theor Biol 2024;576:111656. [Crossref] [PubMed]
42. Hwang KT, Han W, Lee SM, et al. Prognostic influence of 3-dimensional tumor volume on breast cancer compared to conventional 1-dimensional tumor size. Ann Surg Treat Res 2018;95:183-91. [Crossref] [PubMed]