Abstract
Background
Obesity is linked to poor breast cancer outcomes, but evidence within molecular intrinsic subtypes is limited. We measured visceral and subcutaneous adipose tissue (VAT and SAT) on routine clinical imaging and examined survival and tumor microenvironment (TME) differences across PAM50 subtypes.
Methods
We sampled 1,377 individuals diagnosed with stage II-III breast cancer at Kaiser Permanente Northern California between 2005 and 2015, enriched for HER2 + and ER- tumors. VAT and SAT areas (cm2) were quantified at L3-level from clinically acquired CT scans. Tumor RNA profiling (NanoString BC360™) provided PAM50 and TME-related signatures. Cox models estimated associations of VAT and SAT with overall and disease-specific outcomes. Differential expression and signature analyses were stratified by PAM50 subtype.
Results
Participants had a mean (± SD) age of 56 ± 13 years. Intrinsic subtype distribution was: Basal-like (31%), HER2-E (25%), Luminal B (22%), and Luminal A (22%). Over a median follow-up of 11 years, patients with HER2-E tumors and high-VAT had higher breast cancer-specific mortality (HR = 2.14, 95%CI 1.25–3.65) and distant relapse (HR = 2.05, 95%CI 1.23–3.42). No VAT associations were observed in other subtypes. Transcriptomic analyses revealed enhanced abundance of cancer-associated fibroblasts (CAF) with corresponding stromal signatures exclusively in high-VAT HER2-E tumors. Higher CAF and stromal activation signatures were associated with worse outcomes in the full cohort. SAT showed no associations in any subtype.
Conclusions
Visceral adiposity identified a high-risk subset of HER2-E breast cancer characterized by high CAF and stromal activation signatures and inferior survival, highlighting an opportunity for VAT-targeted interventions and mechanistic studies to clarify adiposity-TME interactions.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1186/s13058-026-02337-1.
Keywords: Breast cancer, Tumor microenvironment, Body composition, Survival
Introduction
More than one-third of U.S. women diagnosed with breast cancer have a body mass index (BMI) over 30 kg/m2 at diagnosis, [1, 2] and obesity has been associated with worse breast cancer outcomes [3, 4]. Among other effects, obesity alters systemic metabolism, leading to dysregulation of immune and metabolic pathways in the breast tumor microenvironment (TME) [5–9]. These alterations, in turn, have been linked to breast cancer recurrence, progression, and mortality [10–12]. While the mechanisms through which obesity impacts breast cancer outcomes are still being explored, elucidating them is essential for developing targeted interventions to reduce risk of recurrence and breast cancer death.
Recent meta-analyses have shown that obesity is associated with increased all-cause and breast cancer-specific mortality in patients with estrogen receptor-positive (ER+) and human epidermal growth factor receptor 2-positive (HER2+) tumors, but not in those with triple-negative breast cancer (TNBC) [3]. This divergence suggests that the relationship between obesity and breast cancer outcomes may be fundamentally shaped by tumor biology. Given that intrinsic molecular subtype, defined by gene expression profiles, determine both prognosis and the composition of the TME, [13, 14] it is critical to examine obesity-related mechanisms within these subtypes rather than across broad immunohistochemistry (IHC) categories (e.g., ER+). However, most studies of obesity and the breast TME have either not stratified by subtype [9] or lacked sufficient sample sizes to investigate less common subtypes such as HER2-enriched (HER2-E) [6–9].
Further, although seminal work in animal models [15] and translational studies [16] has demonstrated that adiposity can influence breast cancer metastasis through modulating immune features of TME, no study has directly examined clinical body composition measures in relation to breast TME and survival outcomes. Unlike BMI, which does not distinguish between adiposity and muscle, imaging-based approaches such as computed tomography (CT) allow precise quantification of visceral (VAT) and subcutaneous (SAT) adipose tissue, as well as skeletal muscle.[17] While prior research has examined white adipose tissue within the breast, [18] CT segmentation provides an accurate characterization of patient-level body composition that may capture the impact of systemic factors on the TME and breast cancer outcomes. For example, VAT, a hallmark of obesity that is not well-characterized by BMI, is the most metabolically active adipose depot and contributes to systemic inflammation, which may influence the breast TME by promoting immune suppression and stromal remodeling [10–12]. To address these gaps, we conducted transcriptomic analyses within PAM50-defined molecular subtypes to examine associations between body composition, the breast TME, and breast cancer outcomes in a large, diverse cohort of stage II–III breast cancer patients.
Methods
Study population
Female patients diagnosed with first, primary stage II or III invasive breast cancer between 2005 and 2015 were identified through the Kaiser Permanente Northern California’s (KPNC) cancer registry. Patients were eligible if they were diagnosed between 18 and 89 years of age, received initial treatment at KPNC, had height and weight measurements within 6 months of diagnosis but before systemic therapy, and had self-reported race and ethnicity available in the medical record. Individuals were excluded if they were underweight (BMI < 18.5 kg/m2) or had implausible BMI (> 68 kg/m2), or pathology data was not available to classify their hormone receptor status. The final eligible population size is 3,216.
From this population, we sampled participants with available pre-treatment formalin-fixed, paraffin-embedded (FFPE) tumor specimens from the surgical resection or biopsy (for those received neoadjuvant therapy), stratified by their St. Gallen consensus surrogate subtype [19, 20]. Surrogate subtypes were IHC-based, classified by estrogen receptor status, with tumor grade used as a proxy for Ki67. The final sample was enriched for HER2-overexpressing subtypes and Black/African-American individuals (N = 1,690). Participants and their samples were subsequently excluded if the tumor specimen could not be retrieved, or the tissue was unable to be macrodissected due to insufficient tumor purity or amount. Participants were also required to have a computed tomography (CT) scan within 6 months of diagnosis and before systemic therapy and were excluded if the diagnostic CT did not include the third lumbar (L3) vertebrae (N = 1,426), as L3 is the common landmark used for body composition assessments. A total of 49 samples were further excluded due to failing NanoString quality control standards, resulting in a final cohort size of 1,377 (eFigure 1). Sampling proportions by IHC subtype and race for the current analytical cohort are shown in eTable 1.
Body composition
Body composition was assessed from CT scans taken at a median (interquartile range) of 33 days (15–52 days) from diagnosis. Centrally trained researchers quantified the cross-sectional area (cm2) of muscle and adipose tissue in centimeters squared at the third lumbar vertebra (L3), discriminating components by tissue-specific Hounsfield units (HU) ranges using SliceOmatic Software version 5.0 (TomoVision). We then scaled these measurements to patient height squared to obtain indices of subcutaneous (SAT index, cm2/m2) and visceral adipose tissue (VAT index, cm2/m2), and skeletal muscle (SMI, cm2/m2). High SAT index was defined as greater than or equal to 97.7 cm2/m2, and high VAT index was defined as greater than or equal to 42.5 cm2/m2, chosen from the average in the larger eligible cohort.
Gene expression
Slides containing at least 4mm [2] of invasive tumor with greater than 10% cellularity were prepared and marked for macrodissection. Marked regions were macrodissected from matched unstained 10-micron slides, and RNA was extracted. Expression levels were measured using the NanoString nCounter bead assay and scanned using the nCounter digital analyzer. NanoString BC360 panel was used to measure gene expression and evaluate proprietary expression signatures including PAM50 molecular subtype [21]. Expression levels are normalized by using a ratio of expression value to the geometric mean of housekeeping genes on the BC360 panel and log2 transformed.
Gene sets from multiple sources were used to calculate gene signature scores. NanoString BC360 gene signature scores were calculated from the BC360 panel by NanoString using their in-house normalization methods and gene weights. Hallmark gene sets [22] and single-cell programs from the Curated Cancer Cell Atlas (3CA) [23] were obtained from the Molecular Signatures Database (MSigDB) [24]. 3CA includes multiple cancer-associated fibroblast (CAF) programs, including the CAF-1 program. Signature scores from the Hallmark and 3CA gene sets were calculated from single-sample gene set enrichment analysis (ssGSEA) using the fgsea package [25].
As an additional validation, we used CAF abundance from tumor deconvolution methods (using the EPIC [26] and MCP-counter [27] abundance previously calculated by TIMER2 [28]) and signature scores from the 3CA CAF programs (all CAF program genes and restricted to the NanoString panel genes) in an independent cohort from the TCGA-BRCA PanCancer Atlas from The Cancer Genome Atlas (TCGA).
Covariates
Age at diagnosis, self-reported race and ethnicity, and measured BMI were extracted from the electronic health record (EHR). American Joint Committee on Cancer (AJCC) stage, tumor grade, and hormone receptor status were extracted from the cancer registry. IHC subtype was confirmed via medical chart review.
Outcomes
Overall survival and breast cancer-specific survival (breast cancer listed as the immediate or underlying cause of death) were ascertained from KPNC EHR data and mortality information from the state of California, the Social Security Administration, and the National Death Index through December 31, 2023.
Invasive disease-free survival was ascertained via death records and medical record review and defined as the time from diagnosis to the first occurrence of death by any cause; local, regional, or distant breast cancer recurrence; second primary cancer; or the end of study. Similarly, distant relapse-free survival was the time from diagnosis to the first distant recurrence, or breast cancer-specific death. For this outcome, patients were censored at death from any other cause prior to distant recurrence. For both outcomes, patients were censored upon disenrollment from the health system or end of study follow-up (December 31, 2023).
Statistical analysis
Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CI) for the association between SAT index and VAT index with overall survival and invasive disease-free survival. In cause-specific models for breast cancer-specific survival and distant relapse-free survival, participants were censored upon death from other causes, and death from any cause was treated as a competing event for distant relapse. Models were adjusted for age at diagnosis, stage, race, and SMI based on clinical relevance. All analyses were first conducted in the overall sample and then stratified by PAM50 subtype.
The limma R package was used to identify genes that were differentially expressed between tumors from women with high vs. low VAT index and SAT index. We set the false discovery rate (FDR) significance threshold to 0.1 to control for multiple testing.
Hierarchical clustering using Ward’s minimum variance method and k-means were used to create 3 clusters of anti-tumor immune microenvironment activity (TIME) referred to as low, medium, or high using 15/18 genes in the tumor inflammation signatures (TIS) gene set, [29] which includes gene markers from multiple immune processes and cell types (T cells, NK cells, antigen presentation, interferon gamma, T cell exhaustion) and has been correlated with response to PD-1 blockade in multiple tumor types. The 3 genes excluded (HLA-DQA1, HLA-DRB1, CD276) did not correlate as well as the rest of the 15/18 in the three TIME clusters.
We further evaluated associations between the identified differentially expressed gene signatures and breast cancer outcomes, using the same modeling approaches as in the body composition analyses described above. High, medium, and low level of a signature were defined based on the highest, middle, and lowest tertile with respect to the full cohort.
We performed all analyses using R statistical software version 4.3.1 “Beagle Scouts”. All hypothesis tests were two-sided at 0.05 significance level.
Results
Table 1 displays participant characteristics by PAM50 subtype. The mean age at diagnosis for the cohort was 56 ± 13 years old. Patients with Luminal tumors were more likely to be overweight or obese (Luminal A: 73%, Luminal B: 71%) compared to those with HER2-E (62%) or Basal-like (68%) tumors. A higher proportion of Asian/Pacific Islander patients had HER2-E tumors (21%), and a higher proportion of Black/African American patients had Luminal B (21%) or Basal-like (20%) tumors. Compared to the eligible population, the final cohort included a higher proportion of Black individuals and HER2-E tumors due to the intentional over-sampling (eTable 2).
Table 1.
Participant characteristics by PAM 50 subtypes
| Characteristic | PAM50 Subtype | Overall N = 1377 |
|||
|---|---|---|---|---|---|
| Luminal A N = 301 |
Luminal B N = 312 |
HER2-E N = 342 |
Basal N = 422 |
||
| Age at diagnosis, Mean (SD) | 59 (13) | 58 (13) | 55 (13) | 54 (14) | 56 (13) |
| BMI, n (%) | |||||
| <25 kg/m2 | 84 (28) | 92 (29) | 133 (39) | 133 (32) | 442 (32) |
| 25-<30 kg/m2 | 110 (37) | 89 (29) | 101 (30) | 119 (28) | 419 (30) |
| 30 + kg/m2 | 107 (36) | 131 (42) | 108 (32) | 170 (40) | 516 (37) |
| Race and ethnicity, n (%)a | |||||
| Asian / Pacific Islander | 46 (15) | 47 (15) | 71 (21) | 63 (15) | 227 (16) |
| Black / African-American | 40 (13) | 65 (21) | 27 (7.9) | 86 (20) | 218 (16) |
| Hispanic / Latinx | 24 (8.0) | 20 (6.4) | 28 (8.2) | 49 (12) | 121 (8.8) |
| White | 188 (62) | 176 (56) | 212 (62) | 219 (52) | 795 (58) |
| Other | 3 (1.0) | 4 (1.3) | 4 (1.2) | 5 (1.2) | 16 (1.2) |
| Diabetes status at diagnosis, n (%) | 45 (15) | 50 (16) | 37 (11) | 57 (14) | 189 (14) |
| SAT Index, n (%) | |||||
| Low SAT (< 97.7 cm2/m2) | 175 (58) | 160 (51) | 211 (62) | 223 (53) | 769 (56) |
| High SAT ( > = 97.7 cm2/m2) | 126 (42) | 152 (49) | 131 (38) | 199 (47) | 608 (44) |
| VAT Index, n (%) | |||||
| Low VAT (< 42.5 cm2/m2) | 176 (58) | 190 (61) | 216 (63) | 263 (62) | 845 (61) |
| High VAT ( > = 42.5 cm2/m2) | 125 (42) | 122 (39) | 126 (37) | 159 (38) | 532 (39) |
| SKM Index, n (%) | |||||
| High SMI ( > = 40 cm2/m2) | 203 (67) | 232 (74) | 232 (68) | 286 (68) | 953 (69) |
| Low SMI (< 40 cm2/m2) | 98 (33) | 80 (26) | 110 (32) | 136 (32) | 424 (31) |
| Tumor stage, n (%) | |||||
| II | 173 (57) | 160 (51) | 182 (53) | 291 (69) | 806 (59) |
| III | 128 (43) | 152 (49) | 160 (47) | 131 (31) | 571 (41) |
| ER Positive, n (%) | 283 (94) | 312 (100) | 147 (43) | 61 (14) | 803 (58) |
| PR Positive, n (%) | 188 (62) | 195 (63) | 75 (22) | 15 (3.6) | 473 (34) |
| HER2 Positive, n (%) | 41 (14) | 39 (13) | 264 (77) | 29 (6.9) | 373 (27) |
| Chemotherapy, n (%) | 241 (80) | 241 (77) | 301 (88) | 356 (84) | 1,139 (83) |
| Radiation, n (%) | 61 (20) | 73 (23) | 60 (18) | 101 (24) | 295 (21) |
Abbreviations: BMI Body mass index, HER2 Human epidermal growth factor receptor 2, IQR Interquartile range, ER Estrogen receptor, PR Progesterone receptor, SAT Subcutaneous adipose tissue, SD Standard deviation, SKM Skeletal muscle, VAT Visceral adipose tissue
aWe intentionally oversampled Black / African-American individuals for this study
Overall, 39% of participants had a high VAT index (vs. low) and 44% had a high SAT index (vs. low) (Table 1). Compared with other subtypes, HER2-E patients were less likely to have a high SAT index (38%), while Luminal A patients were more likely to have a high VAT index (42%).
Body composition and survival
With a median follow-up of 11 years (IQR: 6.9–14.1), 522 (38%) of participants died, with 323 deaths due to breast cancer.
Overall, high SAT index was significantly associated with all-cause mortality compared to low SAT index (HR = 1.23, 95% CI: 1.02–1.47), but not breast-cancer specific death (HR = 1.18, 95% CI: 0.93–1.48) (Table 2). There was no association between high VAT index and breast cancer-specific survival in the overall sample pooling all PAM50 subtypes (HR = 1.06, 95% CI: 0.83–1.35).
Table 2.
Association between VAT index and SAT index and breast cancer outcomes
| Exposure | All-cause mortality | Breast cancer-specific mortality | Invasive disease | Distant relapse |
|---|---|---|---|---|
| HR (95% CI)1 | HR (95% CI)2 | HR (95% CI)1 | HR (95% CI)2 | |
| Overall3 | ||||
| n = 522 | n = 323 | n = 577 | n = 249 | |
| VAT Index | ||||
| Low VAT | 1 (referent) | 1 (referent) | 1 (referent) | 1 (referent) |
| High VAT | 1.17 (0.97 to 1.41) | 1.06 (0.83 to 1.35) | 1.12 (0.94–1.34) | 1.10 (0.88–1.38) |
| SAT Index | ||||
| Low SAT | 1 (referent) | 1 (referent) | 1 (referent) | 1 (referent) |
| High SAT | 1.23 (1.02–1.47) | 1.18 (0.93–1.48) | 1.23 (1.03–1.47) | 1.24 (1.00–1.55) |
| Luminal A4 | ||||
| n = 113 | n = 52 | n = 130 | n = 47 | |
| VAT Index | ||||
| Low VAT | 1 (referent) | 1 (referent) | 1 (referent) | 1 (referent) |
| High VAT | 1.00 (0.67 to 1.48) | 0.74 (0.41 to 1.36) | 0.93 (0.64–1.35) | 0.90 (0.53–1.53) |
| SAT Index | ||||
| Low SAT | 1 (referent) | 1 (referent) | 1 (referent) | 1 (referent) |
| High SAT | 1.22 (0.82 to 1.81) | 1.03 (0.56 to 1.88) | 1.24 (0.86–1.80) | 1.20 (0.71–2.02) |
| Luminal B4 | ||||
| n = 129 | n = 81 | n = 150 | n = 77 | |
| VAT Index | ||||
| Low VAT | 1 (referent) | 1 (referent) | 1 (referent) | 1 (referent) |
| High VAT | 1.20 (0.82 to 1.77) | 0.98 (0.60 to 1.59) | 1.15 (0.81–1.64) | 1.03 (0.67–1.59) |
| SAT Index | ||||
| Low SAT | 1 (referent) | 1 (referent) | 1 (referent) | 1 (referent) |
| High SAT | 1.22 (0.83 to 1.78) | 1.14 (0.72 to 1.83) | 1.33 (0.94–1.89) | 1.27 (0.84–1.93) |
| HER2-E4 | ||||
| n = 101 | n = 65 | n = 112 | n = 49 | |
| VAT Index | ||||
| Low VAT | 1 (referent) | 1 (referent) | 1 (referent) | 1 (referent) |
| High VAT | 1.65 (1.07 to 2.54) | 2.14 (1.25 to 3.65) | 1.57 (1.04–2.37) | 2.05 (1.23–3.42) |
| SAT Index | ||||
| Low SAT | 1 (referent) | 1 (referent) | 1 (referent) | 1 (referent) |
| High SAT | 1.21 (0.78 to 1.87) | 1.45 (0.84 to 2.48) | 1.28 (0.84–1.95) | 1.49 (0.89–2.51) |
| Basal4 | ||||
| n = 179 | n = 125 | n = 185 | n = 76 | |
| VAT Index | ||||
| Low VAT | 1 (referent) | 1 (referent) | 1 (referent) | 1 (referent) |
| High VAT | 1.07 (0.78 to 1.46) | 0.93 (0.63 to 1.37) | 1.00 (0.73–1.36) | 0.96 (0.66–1.4) |
| SAT Index | ||||
| Low SAT | 1 (referent) | 1 (referent) | 1 (referent) | 1 (referent) |
| High SAT | 1.31 (0.96 to 1.78) | 1.25 (0.87 to 1.80) | 1.23 (0.91–1.67) | 1.29 (0.90–1.85) |
Abbreviations: CI Confidence interval, HER2-E Human epidermal growth factor receptor 2 enriched, HR Hazard ratio, SAT Subcutaneous adipose tissue, VAT Visceral adipose tissue
1Cox Proportional Hazards model
2Cause-specific Cox Proportional Hazards model
3Models adjusted for age at diagnosis, stage, race, skeletal muscle index (low vs. high), and PAM50 subtype
4Stratified models adjusted for age at diagnosis, stage, race, and skeletal muscle index (low vs. high)
When stratified by PAM50 subtype, however, patients with HER2-E tumors and a high VAT index had twice the risk of breast cancer-specific death (HR = 2.14, 95% CI: 1.25–3.65) and distant relapse (HR = 2.05, 95% CI: 1.23–3.42) compared to those with a low VAT index. No association was observed in other subtypes (Table 2). SAT was not significantly associated with breast cancer-specific outcomes within any PAM50 subtype.
BC360 gene expression
Principal component analysis (PCA) of overall gene expression revealed that the first principal component strongly corresponded to PAM50 subtype classification, as expected (eFigure 2a). The second principal component primarily captured variation by TIME (eFigure 2b), suggesting that TIME contributes substantially to transcriptional heterogeneity beyond intrinsic tumor subtype in our sample.
Differential gene expression
At an FDR-adjusted significance threshold of 0.1, we identified 30 genes that were differentially expressed between high vs. low VAT index in HER2-E, while no VAT index-related differentially expressed genes were identified in other PAM50 subtypes (Fig. 1). Similarly, based on multiple gene signature sets (BC360, Hallmark, 3CA), differential signatures by VAT index were only identified in HER2-E but not in other subtypes (Fig. 2). Across these three gene sets, CAF-related signatures (e.g., BC360 stroma and CAF-related single-cell programs in 3CA) were enriched in patients with high VAT compared to those with low VAT. Signatures such as BC360 stemness and epithelial-to-mesenchymal transition were also more enriched in high-VAT patients, and the concurrent enrichment of CAF-related signatures in this group suggests that these signatures likely arise from the TME as well.
Fig. 1.
Differential gene expression by VAT index stratified by PAM50 subtype
Fig. 2.
Differential expressed gene signatures by VAT index stratified by PAM50 subtype
(a) BC360; (b) Hallmark; (c) 3CA
The 25 genes overexpressed among women with high VAT index and HER-2 tumors were found in CAF-related gene sets, and included: 9/50 of the genes of the 3CA CAF-1 program (COMP, FBN1, ASPN, DCN, SFRP2, COL6A3, THBS2, SFRP4, COL11A1); FAP, COL6A3, ADAM12 which make up the BC360 stromal signature; and VIM, DDR2, HED1, FHL1 which are part of the mammary stemness signature (eTable 3).
We compared the associations between VAT and CAF-1 and stroma signatures observed in the HER2-E subtype with HER2 + cases across other subgroups (eFigure 3). Across the IHC subtypes, HER2+/HR- is most similar to HER2-E, whereas the association for HER2+/HR + has a smaller effect size in the same direction that is not significant. For HER2 + PAM50 subtypes, sample sizes are extremely small, and the patterns are not consistent with those observed in HER2-E.
Between high and low SAT index, the only differences detected at the pathway level were higher hypoxia-related pathway activity among patients with high SAT index in basal-like subtype and increased angiogenesis pathway in Luminal B tumors, compared to low SAT index (eFigure 4).
Identified pathways and survival
To confirm their prognostic importance in multivariable models, we then examined the associations between signatures including CAF-1, stroma, and mammary stemness with breast cancer outcomes in the overall sample (n = 1,377) and within each PAM50 subtype.
After adjusting for age, stage, race, and PAM50 subtype, high CAF-1 expression was associated with increased risk of all-cause (HR = 1.53, 95% CI 1.20–1.93) and breast cancer-specific mortality (HR = 1.86, 95% CI 1.37–2.52) as well as invasive disease (HR = 1.51, 95% CI 1.21–1.89), and distant relapse (HR = 1.81, 95% CI 1.36–2.41) (Table 3). Each 1-unit increase in stroma and mammary stemness signature score was also associated with higher risks of breast cancer-specific mortality (stroma: HR = 1.32, 95% CI: 1.14–1.52; mammary stemness: HR = 1.16 95% CI 1.07–1.26). Similar associations were observed for all-cause mortality, invasive disease and distant relapse, with higher signature scores associated with worse outcomes (Table 3).
Table 3.
Associations between differentially expressed gene signatures and breast cancer outcomes (n = 1,377)
| Signature | All-cause mortality (Event N = 522) |
Breast cancer-specific mortality (Event N = 323) |
Invasive disease (Event N = 577) |
Distant relapse (Event N = 249) |
|---|---|---|---|---|
| HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | |
| Cancer-associated fibroblasts | ||||
| CAF1 Low | 1 (referent) | 1 (referent) | 1 (referent) | 1 (referent) |
| CAF1 Medium | 1.36 (1.09–1.71) | 1.67 (1.26–2.22) | 1.31 (1.06–1.62) | 1.64 (1.26–2.13) |
| CAF1 High | 1.53 (1.20–1.93) | 1.86 (1.37–2.52) | 1.51 (1.21–1.89) | 1.81 (1.36–2.41) |
| CAF2 Low | 1 (referent) | 1 (referent) | 1 (referent) | 1 (referent) |
| CAF2 Medium | 1.29 (1.03–1.61) | 1.62 (1.22–2.14) | 1.33 (1.07–1.64) | 1.64 (1.26–2.14) |
| CAF2 High | 1.56 (1.22–1.99) | 1.89 (1.39–2.57) | 1.66 (1.32–2.09) | 1.95 (1.46–2.60) |
| Stroma (per 1 unit increase) | 1.19 (1.06–1.33) | 1.32 (1.14–1.52) | 1.23 (1.11–1.37) | 1.34 (1.18–1.54) |
| Mammary stemness | 1.11 (1.04–1.19) | 1.16 (1.07–1.26) | 1.14 (1.07–1.22) | 1.17 (1.08–1.27) |
Abbreviations: CAF Cancer-associated fibroblast, CI Confidence interval, HR Hazard ratio
When stratified by PAM50 subtypes, the observed associations between signatures and breast cancer outcomes were similar to those in the overall sample, with large but nonsignificant effect sizes in HER2-E subtype due to small number of events (eTable 4). When further adjusting for TIME categories, the effect estimates stayed similar (results not shown).
CAF and TIME by PAM50 subtypes
Given that CAF-related gene signatures were associated with the differences in gene expression and survival by VAT and that these differences were only seen in HER2-E tumors but not in other subtypes, we hypothesized that this overall observation could be related to the balance of anti-tumor TME and immunomodulatory TME cell types and processes between PAM50 subtypes. We focus on TIS as a measure of anti-tumor TIME activity and CAF-1 as the immunomodulatory cell type of interest.
As expected from the degree of immunogenicity by PAM50 subtype, TIS scores showed a trend from lowest in Luminal A to highest in Basal tumors, with Luminal B and HER2-E tumors ranking second and third, respectively (Fig. 3). CAF-1 scores displayed the opposite behavior as TIS score for Luminal A and Basal tumors but had a different order for Luminal B and HER2-E, with HER2-E tumors being the second highest in CAF-1 score, and high VAT tumors HER2-E having a similar median score as Luminal A tumors (Fig. 3).
Fig. 3.
Distribution of CAF signature and TIS score across PAM50 subtypes in study cohort. Abbreviations: CAF: cancer-associated fibroblast; TIS, tumor inflammation signature
Furthermore, when stratified by TIME activity (low/medium/high), survival differences by CAF-1 score were most pronounced in the high TIME group (eFigure 5). This suggests that in HER2‑E tumors, VAT‑associated increases in CAF‑1 can counteract a robust anti-tumor TIME activity, which may help explain the distinct association between VAT and breast cancer outcomes in this group.
To validate potential consequences of CAF enrichment, we verified that the CAF-1 signatures (both the full 50-gene 3CA CAF-1 program and the 9 CAF-1 related genes identified in our cohort using BC360) showed the same distribution pattern across PAM50 subtypes in TCGA (eFigure 6). Similarly, multiple deconvolution-based estimates of CAF abundance across PAM50 subtypes mirrored the pattern seen with the CAF-1 signature (eFigure 6). Importantly, both CAF abundance and CAF-1 signature levels were associated with overall survival in TCGA (eFigure 7), reinforcing their potential prognostic relevance and validated findings in our cohort.
Discussion
In this study of 1,377 breast cancer survivors, we found that excess VAT was associated with higher all-cause and breast cancer-specific mortality, as well as increased risks of recurrence in HER2-E subtype. Among HER2-E patients with high VAT, genes and signatures related to cancer-associated fibroblasts and stroma were enriched. Elevated activity of these signatures was further associated with worse breast cancer outcomes.
Though several studies have examined the association between imaging-based adiposity and survival outcomes in nonmetastatic breast cancer, [30, 31] none have stratified by molecular subtype. However, evidence from metastatic breast cancer aligns with our observations: a high total abdominal fat index was linked to worse progression-free survival in HER2 + patients defined by IHC, [32] mirroring our findings with VAT in HER2-E cases defined by PAM50.
We did not observe significant associations between high VAT and breast cancer outcomes in other intrinsic molecular subtypes. In luminal B tumors, the null finding could be due to the smaller subgroup sample size, limiting the statistical power to detect significant differences despite HRs suggesting poorer outcomes numerically (all-cause mortality HR = 1.20, 95% CI: 0.82–1.77; invasive disease HR = 1.15, 95% CI: 0.81–1.64). In contrast, the HR estimates for luminal A and basal-like tumors were close to 1.0 across outcomes, indicating that any VAT association in these subtypes is likely small in our cohort rather than simply undetected because of limited statistical power. There have been mixed findings in the prior literature regarding obesity and breast cancer outcomes, especially among TNBC [3, 33, 34]. However, most previous studies have relied on BMI and IHC subtypes, whereas our study evaluated CT-derived adiposity and PAM50 intrinsic subtypes. Although IHC-defined TNBC is not identical to PAM50-defined basal-like tumor, these groups overlap substantially, and the aggressive biology and early relapse pattern characteristic of TNBC may reduce the relative contribution of VAT to prognosis in basal-like tumors as such modifiable risk factors may exert their effects over a longer time period [35]. These differences in exposure measurement and subtype classification may contribute to discrepancies between our findings and the existing literature. Future studies using imaging-based adiposity measures and intrinsic molecular subtyping are needed to validate our findings.
Our study is the first to identify VAT-associated enhanced CAF activation in HER2-E breast TME, with consistent findings from multiple gene sets and verified CAFs activity in an external dataset. Previous studies have largely focused on either IHC subtype-based ER + 10 or HER2- tumors, [7] or did not stratify by HER2 status [6]. Two studies of obesity specifically assessed the PAM50-identified HER2-E tumors, [8, 9] but did not detect differential expression of stromal or CAF genes, potentially due to smaller sample size or because of reliance on BMI, which has only a modest correlation with VAT as a measure of adiposity [36]. Recent population studies have shown that metabolic syndrome, which is tightly associated with excess VAT, [37] increases all-cause mortality among every IHC subtype but breast cancer-specific mortality only among HER2 + disease [38]. This suggests that visceral adiposity may uniquely exacerbate tumor-promoting stromal dynamics in HER2 + disease. In our exploratory analyses, the association between VAT and CAF was consistent in all IHC HER2 + tumors, while no association was observed in HER2 + Luminal A, Luminal B, and Basal groups, though the sample sizes are extremely small. Taken together, these findings suggest that the observed association is driven by the HER2-E subtype and not solely by HER2 + status. Larger studies are needed to determine whether the positive VAT and CAF relationship is generalizable across HER2 + tumors in intrinsic subtypes other than HER2-E subtype, and whether ER status modifies this association.
The enrichment of stromal/CAF genes and gene signatures in VAT-high HER2-E tumors suggests that upregulation of CAF activity may mediate the link between excess VAT and worse breast cancer outcomes. The biological plausibility of this finding is further supported by observations that obesity is associated with increased CAF activity [39–41] as well as stromal desmoplasia, [42] a fibrotic stromal response characterized by activated CAF programs. Certain CAFs subtypes and stromal desmoplasia have been associated with poorer survival in breast cancer and other types of cancer [43–49]. However, we acknowledge that our analyses only captured CAF gene sets that do not resolve specific CAF subtypes (e.g. myofibroblast or inflammatory CAFs).
A potential explanation for why differential gene expression by VAT is observed only in the HER2-E subtype relates to the coexistence of high anti-tumor TIME activity and stromal remodeling. HER2-E tumors exhibit relatively high immune infiltration [50] and CAF abundance compared to other intrinsic molecular subtypes, and our data suggested that high VAT in this context could further enhance CAF-related programs. Notably, we observed that the adverse association between CAF activity with breast-cancer survival was stronger in TIME-high compared to TIME-low tumors. This pattern indicates that in HER2‑E tumors, despite high immune infiltration, VAT‑linked increases in CAF programs may counterbalance otherwise favorable anti‑tumor immune activity through stromal remodeling, thereby restricting immune efficacy and potentially leading to reduced treatment response [51, 52]. Future studies integrating single-cell and spatial resolution with functional studies in experimental models are needed to experimentally validate this hypothesis.
As the first study to conduct transcriptomic analyses of the breast TME by body composition phenotypes, these data contribute to the literature on obesity and breast cancer outcomes by more precise adiposity measurements than previous studies that relied on BMI alone. Further, we stratified analyses by PAM50 subtypes and over-sampled less common but aggressive subtypes such as HER2-E, addressing a significant gap in the literature. Additionally, we examined multiple clinically relevant outcomes, including breast cancer-specific mortality, invasive disease-free survival, and distant relapse, offering insights into specific survival mechanisms.
Although one of the largest studies of its kind, sample sizes within individual subtypes in our cohort remain relatively small, limiting statistical power to conduct further stratified analyses by race, ethnicity, or genetic signatures. The data availability in the retrospective EHR also limited our ability to adjust for or stratify by other clinically important variables, such as menopausal status. Further, we used bulk RNA expression and a gene expression panel, which limits the ability to resolve cell type-specific contributions to the observed gene expression changes. Future studies should investigate differential single-cell RNA expression by body composition phenotypes to more precisely map which cell populations drive these associations.
In sum, higher VAT is associated with worse breast cancer outcomes in patients with HER2-E breast tumors, likely mediated by upregulation of genes related to stroma and CAFs. These alterations lead to higher risks of invasive disease and distant relapse, as well as worse breast cancer-specific and all-cause mortality. Findings from our study suggest that VAT-targeted interventions may be particularly beneficial for patients with HER2-E breast cancer. Future single-cell and mechanistic studies are needed to validate the hypothesized interactions between VAT, TIME and CAFs.
Supplementary Information
Acknowledgements
This work was supported by grants from the National Cancer Institute (R01CA251589) and Victoria’s Secret Global Fund for Women’s Cancers in Partnership with Pelotonia and AACR (SSM). We thank Emma Armstrong for her assistance with adjudicating breast cancer recurrence outcomes, and Aline Alencar Giongo, Tabata Alves Domingos, Ashka Patel, Rodrigo Abreu, Oliver Pigeon, Roberto Peixoto and Claire Capaldi for their contributions in analyzing the pathology specimens.
Authors’ contributions
E.C.F.: Conceptualization.A.C., S.F., J.G.T.Z., and E.C.F.: Data analysis.A.C., S.F., J.G.T.Z., W.Y.C., D.D., S.M., E.A.M., R.J., B.J.C., C.Q., E.M.C.F.: Investigation.E.C.F. and W.C.: Funding acquisition.A.C., S.F., and E.C.F.: Drafted the initial manuscript.J.G.T.Z., W.Y.C., D.D., S.M., E.A.M., R.J., B.J.C., C.Q.: Reviewed and revised the manuscript.
Funding
This work was supported by grants from the National Cancer Institute (R01CA251589) and Victoria’s Secret Global Fund for Women’s Cancers in Partnership with Pelotonia and AACR (SSM).
Data availability
Datasets were generated by the authors and are not publicly available because they contain protected health information (PHI) of the participants but are available from the corresponding author on reasonable requests with appropriate ethics approvals and data sharing agreements.
Declarations
Ethics approval and consent to participate
This data-only study is approved by the KPNC Institutional Review Board with waiver of informed consent due to the retrospective nature. The research was conducted in accordance with the ethical principles of the Declaration of Helsinki.
Consent for publication
Not applicable.
Competing interests
J.G.T.Z. reports ownership of stocks in the biotechnology exchange-traded funds IDNA, IBB, and XBI; owns stocks in Novo Nordisk and GRAIL; and previously owned stocks in the exchange-traded fund CNCR, Adaptive Biotechnologies, 2seventy bio, and bluebird bio. R.J. received research funding from Lilly, Novartis and Recursion and served on advisory boards for Lilly, Pfizer, Astra Zeneca, Novartis and Carrick Therapeutics. D.D. received research funding from Canon, Inc and serves as a consultant for Asuragen (BioTechne), Inc.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Bellini A, et al. The effect of body mass index on breast cancer stage and breast cancer specific survival. Breast Cancer Res Treat. 2025;211:649–56. 10.1007/s10549-025-07678-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Bandera EV, et al. Association of Body Mass Index, Central Obesity, and Body Composition With Mortality Among Black Breast Cancer Survivors. JAMA Oncol. 2021;7:1186–95. 10.1001/jamaoncol.2021.1499. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Pang Y, Wei Y, Kartsonaki C. Associations of adiposity and weight change with recurrence and survival in breast cancer patients: a systematic review and meta-analysis. Breast Cancer. 2022;29:575–88. 10.1007/s12282-022-01355-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Lohmann AE, et al. Association of Obesity With Breast Cancer Outcome in Relation to Cancer Subtypes: A Meta-Analysis. JNCI: J Natl Cancer Inst. 2021;113:1465–75. 10.1093/jnci/djab023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Devericks EN, Carson MS, McCullough LE, Coleman MF, Hursting SD. The obesity-breast cancer link: a multidisciplinary perspective. Cancer Metastasis Rev. 2022;41:607–25. 10.1007/s10555-022-10043-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Wang J, et al. Early-Life Body Adiposity and the Breast Tumor Transcriptome. JNCI: J Natl Cancer Inst. 2020;113:778–84. 10.1093/jnci/djaa169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Nguyen HL, et al. Obesity-associated changes in molecular biology of primary breast cancer. Nat Commun. 2023;14:4418. 10.1038/s41467-023-39996-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Kwan ML, et al. Association of high obesity with PAM50 breast cancer intrinsic subtypes and gene expression. BMC Cancer. 2015;15:278. 10.1186/s12885-015-1263-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Toro AL, Costantino NS, Shriver CD, Ellsworth DL, Ellsworth RE. Effect of obesity on molecular characteristics of invasive breast tumors: gene expression analysis in a large cohort of female patients. BMC Obes. 2016;3:22. 10.1186/s40608-016-0103-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Fuentes-Mattei E, et al. Effects of Obesity on Transcriptomic Changes and Cancer Hallmarks in Estrogen Receptor–Positive Breast Cancer. JNCI: J Natl Cancer Inst. 2014;106. 10.1093/jnci/dju158. [DOI] [PMC free article] [PubMed]
- 11.Smith LA, et al. Weight Loss Reverses the Effects of Aging and Obesity on Mammary Tumor Immunosuppression and Progression. Cancer Prev Res (Phila). 2025;18:453–63. 10.1158/1940-6207.Capr-24-0514. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Creighton CJ, et al. A gene transcription signature of obesity in breast cancer. Breast Cancer Res Treat. 2012;132:993–1000. 10.1007/s10549-011-1595-y. [DOI] [PubMed] [Google Scholar]
- 13.Parker JS, et al. Supervised risk predictor of breast cancer based on intrinsic subtypes. J Clin Oncol. 2009;27:1160–7. 10.1200/jco.2008.18.1370. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Li J, Wu J, Han J. Analysis of Tumor Microenvironment Heterogeneity among Breast Cancer Subtypes to Identify Subtype-Specific Signatures. Genes (Basel). 2022;14. 10.3390/genes14010044. [DOI] [PMC free article] [PubMed]
- 15.Quail DF, et al. Obesity alters the lung myeloid cell landscape to enhance breast cancer metastasis through IL5 and GM-CSF. Nat Cell Biol. 2017;19:974–87. 10.1038/ncb3578. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.McDowell SAC, et al. Obesity alters monocyte developmental trajectories to enhance metastasis. J Exp Med. 2023;220. 10.1084/jem.20220509. [DOI] [PMC free article] [PubMed]
- 17.Rubino F et al. Definition and diagnostic criteria of clinical obesity. Lancet Diabetes Endocrinol 10.1016/S2213-8587(24)00316-4. [DOI] [PMC free article] [PubMed]
- 18.D’Esposito V, et al. Mammary Adipose Tissue Control of Breast Cancer Progression: Impact of Obesity and Diabetes. Front Oncol. 2020;10–2020. 10.3389/fonc.2020.01554. [DOI] [PMC free article] [PubMed]
- 19.Goldhirsch A, et al. Strategies for subtypes–dealing with the diversity of breast cancer: highlights of the St. Gallen International Expert Consensus on the Primary Therapy of Early Breast Cancer 2011. Ann Oncol. 2011;22:1736–47. 10.1093/annonc/mdr304. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Sweeney C, et al. Intrinsic subtypes from PAM50 gene expression assay in a population-based breast cancer cohort: differences by age, race, and tumor characteristics. Cancer Epidemiol Biomarkers Prev. 2014;23:714–24. 10.1158/1055-9965.Epi-13-1023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Bastien RR, et al. PAM50 breast cancer subtyping by RT-qPCR and concordance with standard clinical molecular markers. BMC Med Genomics. 2012;5:44. 10.1186/1755-8794-5-44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Liberzon A, et al. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst. 2015;1:417–25. 10.1016/j.cels.2015.12.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Gavish A, et al. Hallmarks of transcriptional intratumour heterogeneity across a thousand tumours. Nature. 2023;618:598–606. 10.1038/s41586-023-06130-4. [DOI] [PubMed] [Google Scholar]
- 24.Liberzon A, et al. Molecular signatures database (MSigDB) 3.0. Bioinformatics. 2011;27:1739–40. 10.1093/bioinformatics/btr260. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Korotkevich G, et al. Fast gene set enrichment analysis. bioRxiv. 2021;060012. 10.1101/060012. [DOI]
- 26.Racle J, de Jonge K, Baumgaertner P, Speiser DE, Gfeller D. Simultaneous enumeration of cancer and immune cell types from bulk tumor gene expression data. Elife. 2017;6. 10.7554/eLife.26476. [DOI] [PMC free article] [PubMed]
- 27.Becht E, et al. Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression. Genome Biol. 2016;17:218. 10.1186/s13059-016-1070-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Li T, et al. TIMER2.0 for analysis of tumor-infiltrating immune cells. Nucleic Acids Res. 2020;48:W509–14. 10.1093/nar/gkaa407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Ayers M, et al. IFN-γ-related mRNA profile predicts clinical response to PD-1 blockade. J Clin Invest. 2017;127:2930–40. 10.1172/jci91190. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Cheng E, Kirley J, Cespedes Feliciano EM, Caan BJ. Adiposity and cancer survival: a systematic review and meta-analysis. Cancer Causes Control. 2022;33:1219–46. 10.1007/s10552-022-01613-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Poltronieri TS, Pérsico RS, Viana LV. Body adipose tissue depots and treatment outcomes for women with breast cancer: A systematic review. Clin Nutr. 2024;43:1033–42. 10.1016/j.clnu.2024.03.010. [DOI] [PubMed] [Google Scholar]
- 32.Palleschi M, et al. Computed tomography based analyses of body mass composition in HER2 positive metastatic breast cancer patients undergoing first line treatment with pertuzumab and trastuzumab. Sci Rep. 2022;12:3385. 10.1038/s41598-022-07143-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Harborg S, et al. Overweight and prognosis in triple-negative breast cancer patients: a systematic review and meta-analysis. npj Breast Cancer. 2021;7:119. 10.1038/s41523-021-00325-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Vaz-Goncalves L, Protani MM, Saunus JM, Colditz GA, Reeves MM. Continental differences in the association between excess body weight and prognosis in triple-negative breast cancer: a meta-analysis. Breast Cancer Res Treat. 2025;209:215–27. 10.1007/s10549-024-07538-w. [DOI] [PubMed] [Google Scholar]
- 35.Dent R, et al. Triple-negative breast cancer: clinical features and patterns of recurrence. Clin Cancer Res. 2007;13:4429–34. 10.1158/1078-0432.Ccr-06-3045. [DOI] [PubMed] [Google Scholar]
- 36.Camhi SM, et al. The relationship of waist circumference and BMI to visceral, subcutaneous, and total body fat: sex and race differences. Obes (Silver Spring). 2011;19:402–8. 10.1038/oby.2010.248. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Shah RV, et al. Visceral adiposity and the risk of metabolic syndrome across body mass index: the MESA Study. JACC Cardiovasc Imaging. 2014;7:1221–35. 10.1016/j.jcmg.2014.07.017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Loroña NC, et al. Metabolic Syndrome and Risks of Breast Cancer Outcomes for Luminal, Triple-Negative, and HER2-Overexpressing Subtypes. Cancer Epidemiol Biomarkers Prev. 2025;34:117–24. 10.1158/1055-9965.Epi-24-1167. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Strong AL et al. Obesity Enhances the Conversion of Adipose-Derived Stromal/Stem Cells into Carcinoma-Associated Fibroblast Leading to Cancer Cell Proliferation and Progression to an Invasive Phenotype. Stem Cells Int 2017, 9216502 (2017). 10.1155/2017/9216502. [DOI] [PMC free article] [PubMed]
- 40.Strong AL, Burow ME, Gimble JM, Bunnell BA. Concise review: The obesity cancer paradigm: exploration of the interactions and crosstalk with adipose stem cells. Stem Cells. 2015;33:318–26. 10.1002/stem.1857. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Freese KE, et al. Adipose-derived stems cells and their role in human cancer development, growth, progression, and metastasis: a systematic review. Cancer Res. 2015;75:1161–8. 10.1158/0008-5472.Can-14-2744. [DOI] [PubMed] [Google Scholar]
- 42.Seo BR, et al. Obesity-dependent changes in interstitial ECM mechanics promote breast tumorigenesis. Sci Transl Med. 2015;7:301ra130. 10.1126/scitranslmed.3010467. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Talia M, et al. Cancer-associated fibroblasts (CAFs) gene signatures predict outcomes in breast and prostate tumor patients. J Translational Med. 2024;22:597. 10.1186/s12967-024-05413-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Zhao Y, et al. Stromal cells in the tumor microenvironment: accomplices of tumor progression? Cell Death Dis. 2023;14:587. 10.1038/s41419-023-06110-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Chen Y, McAndrews KM, Kalluri R. Clinical and therapeutic relevance of cancer-associated fibroblasts. Nat Rev Clin Oncol. 2021;18:792–804. 10.1038/s41571-021-00546-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Beck AH, et al. Systematic analysis of breast cancer morphology uncovers stromal features associated with survival. Sci Transl Med. 2011;3:108ra113. 10.1126/scitranslmed.3002564. [DOI] [PubMed] [Google Scholar]
- 47.de Kruijf EM, et al. Tumor-stroma ratio in the primary tumor is a prognostic factor in early breast cancer patients, especially in triple-negative carcinoma patients. Breast Cancer Res Treat. 2011;125:687–96. 10.1007/s10549-010-0855-6. [DOI] [PubMed] [Google Scholar]
- 48.Dekker TJ, et al. Prognostic significance of the tumor-stroma ratio: validation study in node-negative premenopausal breast cancer patients from the EORTC perioperative chemotherapy (POP) trial (10854). Breast Cancer Res Treat. 2013;139:371–9. 10.1007/s10549-013-2571-5. [DOI] [PubMed] [Google Scholar]
- 49.Moorman AM, Vink R, Heijmans HJ, van der Palen J, Kouwenhoven EA. The prognostic value of tumour-stroma ratio in triple-negative breast cancer. Eur J Surg Oncol. 2012;38:307–13. 10.1016/j.ejso.2012.01.002. [DOI] [PubMed] [Google Scholar]
- 50.Griguolo G, et al. Immune microenvironment characterisation and dynamics during anti-HER2-based neoadjuvant treatment in HER2-positive breast cancer. NPJ Precis Oncol. 2021;5:23. 10.1038/s41698-021-00163-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Costa A, et al. Fibroblast Heterogeneity and Immunosuppressive Environment in Human Breast Cancer. Cancer Cell. 2018;33:463–e479410. 10.1016/j.ccell.2018.01.011. [DOI] [PubMed] [Google Scholar]
- 52.Perrone M, Talarico G, Chiodoni C, Sangaletti S. Impact of Immune Cell Heterogeneity on HER2 + Breast Cancer Prognosis and Response to Therapy. Cancers. 2021;13:6352. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Datasets were generated by the authors and are not publicly available because they contain protected health information (PHI) of the participants but are available from the corresponding author on reasonable requests with appropriate ethics approvals and data sharing agreements.