Abstract
Purpose:
Several data sets have demonstrated a correlation between lymphovascular invasion (LVI) and locoregional recurrence (LRR). Whether the observation of “extensive LVI” is a further and incremental determinant of LRR risk is unknown. We describe clinical outcomes in women with invasive breast cancer stratified by (1) absence of LVI (neg), (2) LVI focal or suspicious (FS-LVI), (3) usual (nonextensive) LVI (LVI), and (4) extensive LVI (E-LVI).
Methods and Materials:
Between December 2009 and August 2021, 8837 patients with early-stage breast cancer were treated with curative intent and were evaluable. Clinical-pathologic details were abstracted by retrospective review. The description of LVI was abstracted from pathology reports. Recurrence and survival outcomes were compared based on the extent of LVI. A matched propensity score analysis compared outcomes between patients with LVI versus E-LVI.
Results:
Of the 8837 patients studied, 5584 were negative, 461 had FS-LVI, 2315 had LVI, and 477 had E-LVI. Patients with E-LVI had an adverse risk profile compared with the other groups.
The 5- and 10-year LRR cumulative incidence estimates in patients with E-LVI were 9.6% (95% CI, 7.1-13) and 13% (95% CI, 10-17), respectively, which were significantly higher than those observed in the usual LVI group (6.8% [5.7-7.9] and 10% [8.8-12], respectively). A statistically significant difference in LRR was demonstrated in univariable (HR, 1.4; 95% CI, 1.03-1.89; P = .029) and multivariable regression analysis (HR, 1.62; 95% CI, 1.15-2.27; P = .005) compared with nonextensive LVI. In an alternative approach, we performed a 2:1 propensity score matching analysis comparing patients with LVI to those with E-LVI. The hazard ratio for LRR (HR, 1.47; CI 1.02-2.14; P = .041) was suggestive of a higher risk associated with E-LVI.
Conclusions:
Our work suggests that patients with E-LVI are at a higher risk for LRR than those with usual LVI. For patients who are borderline candidates for regional nodal irradiation or post-mastectomy radiation therapy, the finding of E-LVI might be decisive in favor of intensified treatment.
Introduction
Breast cancer is the most common female malignancy in the United States, affecting up to 1 in 8 women by age 70. Surgical treatment is typically preceded or followed by various (neo) adjuvant systemic therapies based on tumor-specific factors such as intrinsic subtype and individualized clinical risk as assessed by traditional relevant clinical and pathologic features. Adjuvant radiation therapy decisions are similarly made based on estimated risk of locoregional recurrence (LRR) based on clinical-pathologic risk determinants. Lymphovascular invasion (LVI) is a well-described pathologic finding seen among various malignancies, including those of the breast.1,2 Defined as the presence of malignant cells within vascular or lymphatic spaces, LVI has been associated with cancer progression and distant metastasis.3 The presence of LVI has also been linked to adverse outcomes in several malignancies, including in breast cancer.4–7
Multiple data sets have implicated LVI as a risk factor for LRR. Clinical studies have found LVI to be an independent predictor of LRR and have concluded that the risk for LRR is potentiated when LVI is combined with other known risk factors for LRR.8–10 LVI has consistently been identified as a significant independent risk factor for LRR in patients with limited nodal involvement and even in node-negative patients.8–13
What remains unknown is whether the extent of LVI, particularly the finding of “extensive LVI,” influences LRR risk incrementally. We used our large institutional database of breast cancer patients to retrospectively determine whether the finding of “extensive LVI” is associated with breast cancer recurrence, potentially identifying women who need intensified local-regional treatment.
Methods and Materials
Study population
We identified 8837 evaluable breast cancer patients presenting to our institution from 2009 to 2021 for whom comprehensive pathologic and treatment data were available. Relevant clinical data and pathologic features were collected, including patient age, tumor size, number of involved lymph nodes, LVI status, estrogen receptor (ER) status, progesterone receptor (PR) status, human epidermal growth factor receptor 2 (HER2) status, and histologic tumor grade. LVI is separately reported for each case at our institution as 1 of the required features. LVI was coded as either 1) absent (neg), (2) focal or suspicious LVI (FS-LVI), (3) usual (nonextensive) LVI (LVI), or (4) extensive LVI (E-LVI), based on descriptions available in the pathology reports. Cases in which LVI was noted without any further descriptors were coded as usual (nonextensive) LVI.
Extensive LVI diagnostic criteria
The morphologic criteria for the diagnosis of LVI, first outlined in 1983,14 are adopted by all major pathology organizations, including the College of American Pathologists. For duration of this study period, pathologist did not have a uniform definition for extensive LVI. To illustrate, guidelines from St. Gallen International Expert Consensus Conference15 included recommendations based on the presence of “extensive” LVI but without guidance on how to define “extensive.” For the first time, the College of American Pathologists protocol for specimens with invasive carcinoma (version 4.9.0.0; protocol posting date, June 2023)16 specifies guidelines on reporting “extensive LVI” as a subset of LVI reporting, with categories as follows: 1) not identified; 2) present, a) focal (LVI in 1 block only) or b) extensive (LVI in 2 or more blocks); and (3) cannot be determined.
In the current study, information about LVI was extracted retrospectively from existing pathology reports, issued by many different pathologists over the course of 20 years, and no slides were reviewed. In the earlier years of the study, all pathologists on staff in our department were reviewing all types of cases with no subspecialty assignments. In July 2004, pathology subspecialty sign-out was introduced, and initially, 2 dedicated breast pathologists reviewed all breast cases, whereas between 10 and 14 breast pathologists have been on staff in the past decade. Given the lack of specific criteria for the diagnosis of extensive LVI, no specific criteria were uniformly applied in this study. Nonetheless, based on the experience of the senior breast pathologist in this study (E.B.), most pathologists likely reported “extensive LVI” as 1 or more of the following scenarios:
presence of tumor emboli in lymphovascular spaces away from the tumor (ie, in one or more tissue sections not showing invasive carcinoma)
presence of large tumor emboli that distend and nearly occlude the lymphovascular spaces (ie, large tumor emboli that mimic solid ductal carcinoma in situ)
presence of multiple tumor emboli in lymphovascular spaces around the main tumor mass in 2 or more tumor sections
A more precise quantification of the number of tumor emboli required for the diagnosis of “extensive LVI” is difficult, as the slides were not reviewed and in most cases were not available for review. Representative images of extensive LVI are shown in Figure 1.
Fig. 1.
Representative images of extensive lymphovascular invasion (LVI). (A) Multiple tumor emboli surrounded by a clear halo are present in lymphovascular spaces near a focus of stromal invasion (upper left corner). (B) Higher magnification view of the emboli of carcinoma in (A). (C) Large tumor emboli distend and occlude the lymphovascular spaces near a focus of stromal invasion (left and upper edges of this image), in a pattern that closely mimics ductal carcinoma in situ. (D) Carcinoma is present only in lymphovascular spaces in breast tissue, with no other evidence of carcinoma.
We also collected treatment parameters, including surgery type (mastectomy or partial mastectomy), margin status, and whether patients received chemotherapy, hormonal therapy, or radiation. Radiation therapy details, such as number of fields, doses, and use of bolus/boosts, were not encoded in the database and were not analyzed.
Statistical analysis
The primary outcome of interest was LRR, the time from surgery to first recurrence in the ipsilateral breast or lymph nodes, with death as a competing risk. Secondary outcomes of interest included overall survival and distant metastasis. If the patient had multiple re-excisions, the time from the last surgery was used. Patient and treatment characteristics were summarized using the median and interquartile range (IQR) for continuous variables and counts and percentages for categorical variables. Histopathologic characteristics were compared using Pearson’s χ2 test for categorical variables and the Kruskal-Wallis rank sum test for continuous variables. Univariable and multivariable regression models were built for overall survival (Cox regression) and LRR and distant metastasis (DM) (Fine-Gray regression) using clinically relevant covariates. In the multivariable models, interactions between LVI and clinically relevant factors were included, and backward stepwise selection was used to select final models. A propensity score matching analysis was conducted comparing LVI (controls) to E-LVI in a 2:1 ratio, using the MatchIt package with R version 4.3.2. Nearest neighbor matching was conducted without replacement using a caliper size of 0.06 times the standard deviation of the propensity scores.
Analyses were conducted on both the overall cohort and on the propensity score–matched cohort, which was built using clinical characteristics to ensure the similarity of those with LVI to those with E-LVI (ie, age, tumor size, nodal involvement, ER status, PR status, HER2 status, grade, surgical approach, axillary surgery, surgical margins, chemotherapy, and radiation). All statistical tests were 2-tailed with a type I error rate (α) of 0.05 and were performed using R version 4.2.1 (R Core Development Team).
Results
Patient characteristics
We identified 8837 patients with invasive breast cancer treated curatively at our center from 2009 to 2021 for whom complete relevant clinical and pathologic data were available (Table 1). Of the 8837-patient cohort, 5584 patients were explicitly reported to have no LVI, 2315 had LVI, 461 had FS-LVI, and 477 had E-LVI. Among all of the cohorts, the E-LVI patients had the youngest median age, at 57 years (range, 49-67); the largest median tumor size, at 2.50 cm; the highest median nodal positivity, at 2 (range, 1.00-6.00); the highest proportion of grade 3 tumors (94%); the highest proportion of mastectomies (67%); and the highest percentage of chemotherapy and radiation use, at 87% and 76%, respectively. On aggregate, these baseline characteristics taken together suggest that E-LVI confers a higher risk profile, even compared with LVI, and resulted in more aggressive courses of treatment.
Table 1.
Patient and treatment characteristics of the study population
| Characteristic | No N = 5584 | Focal/Suspicious N = 461 | Yes N = 2315 | Extensive N = 477 | P value |
|---|---|---|---|---|---|
| Age | 65 (56, 74) | 63 (55, 72) | 60 (52, 70) | 57 (49, 67) | <.001 |
| Tumor size (cm) | 1.20 (0.70, 1.80) | 1.60 (1.10, 2.20) | 1.90 (1.40, 2.60) | 2.50 (1.80, 3.40) | <.001 |
| Lymph nodes involved | <.001 | ||||
| 0 | 4377 (81%) | 283 (62%) | 945 (41%) | 80 (17%) | |
| 1 | 619 (11%) | 94 (21%) | 572 (25%) | 95 (20%) | |
| 2 | 182 (3.4%) | 25 (5.5%) | 238 (10%) | 62 (13%) | |
| 3 | 78 (1.4%) | 13 (2.9%) | 129 (5.6%) | 33 (7.0%) | |
| >3 | 167 (3.1%) | 39 (8.6%) | 403 (18%) | 203 (43%) | |
| Unknown | 161 | 7 | 28 | 4 | |
| ER | <.001 | ||||
| Negative | 772 (14%) | 54 (12%) | 309 (13%) | 100 (21%) | |
| Positive | 4812 (78%) | 407 (92%) | 2006 (87%) | 377 (79%) | |
| PR | 0.005 | ||||
| Negative | 1228 (22%) | 86 (19%) | 467 (20%) | 127 (27%) | |
| Positive | 4356 (64%) | 375 (81%) | 1848 (80%) | 350 (73%) | |
| HER2 | <.001 | ||||
| Negative | 4964 (83%) | 394 (85%) | 1930 (83%) | 366 (77%) | |
| Positive | 620 (17%) | 67 (15%) | 385 (17%) | 111 (23%) | |
| Intrinsic subtype | <.001 | ||||
| HR+/Her2− | 4398 (79%) | 360 (78%) | 1736 (75%) | 304 (64%) | |
| Her2+ | 620 (11%) | 67 (15%) | 385 (17%) | 111 (23%) | |
| Triple negative | 566 (10%) | 34 (7.4%) | 194 (8.4%) | 62 (13%) | |
| Histologic grade | <.001 | ||||
| 1 | 492 (9.7%) | 12 (2.8%) | 34 (1.5%) | 0 (0%) | |
| 2 | 1049 (28%) | 81 (19%) | 325 (15%) | 26 (5.6%) | |
| 3 | 3508 (69%) | 339 (78%) | 1845 (84%) | 437 (94%) | |
| Surgery | <.001 | ||||
| Breast conservation | 3758 (67%) | 278 (60%) | 1202 (52%) | 155 (33%) | |
| Mastectomy | 1825 (33%) | 183 (40%) | 1112 (48%) | 321 (67%) | |
| Margins | |||||
| <2 mm | 327 (5.9%) | 24 (5.2%) | 146 (6.3%) | 33 (6.9%) | |
| <1 mm | 493 (8.8%) | 43 (9.3%) | 230 (9.9%) | 55 (12%) | |
| Negative | 4623 (83%) | 375 (81%) | 1856 (80%) | 359 (75%) | |
| Positive | 122 (2.2%) | 16 (3.5%) | 74 (3.2%) | 27 (5.7%) | |
| Unknown | 19 (0.3%) | 3 (0.7%) | 9 (0.4%) | 3 (0.6%) | |
| Chemotherapy administered | 2460 (44%) | 298 (65%) | 1656 (72%) | 417 (87%) | <.001 |
| Radiation administered | 3508 63%) | 293 (64%) | 1530 (66%) | 362 (76%) | <.001 |
Statistics presented: median (IQR); n(%)
Cumulative incidence rates of LRR and DM
We used cumulative incidence rates to examine the association of E-LVI with outcomes. The median follow-up in all patients was 88 months. At 5- and 10-year follow-ups, E-LVI had LRR rates of 9.6% (95% CI, 7.1%-13%) and 13% (95% CI, 10%-17%), respectively, which were higher than those of nonextensive LVI, at 6.8% (5.7%-7.9%) and 10% (8.8%-12%), respectively (Fig. 2, P < .001). The corresponding rates in FS-LVI were 6.6% (4.5%-9.2%) and 9.9% (7.1%-13%), which were very similar to those of LVI. The corresponding rates were lowest in the LVI-negative group (4.0% and 6.8, respectively). Interestingly, the cumulative incidence of DM at 5- and 10-year follow-ups for the E-LVI group were higher (21% [18%-25%] and 28% [23%-32%]), compared with those for LVI (12% [11%-14%] and 18% [16%-20%], respectively) (Fig. 3, P < .001). The DM rate was lowest for LVI-negative patients (Fig. 3).
Fig. 2.
Cumulative incidence curves of local-regional recurrence.
Fig. 3.
Cumulative incidence curves of distant metastasis.
Univariable analysis of the prognostic factors associated with overall survival, LRR, and DM among the overall cohort
Univariable analysis of relevant clinical and pathologic features as well as treatment characteristics revealed several well-established associations with LRR (Table 2). ER and PR positivity was associated with reduced LRR (HR, 0.47; 95% CI, 0.39-0.57; and HR, 0.56; 95% CI, 0.47-0.66; respectively; P < .001 for each), as was the case in patients who underwent radiation therapy (HR, 0.56; 95% CI, 0.48-0.66; P < .001). Higher histologic grades (2 and 3) and positive margins or close margins (≤1 mm) were associated with increased LRR (HR, 2.59 and 1.91, respectively; P < .001 for each). Using the presence of nonextensive LVI as the reference level condition, the absence of LVI was associated with decreased LRR (HR, 0.62; 95% CI, 0.52-0.74; P < .001), while E-LVI was associated with a higher risk of LRR (HR, 1.40; 95% CI, 1.03-1.89; P = .029).
Table 2.
Risk of locoegional recurrence in patients with no lymphovascular invasion (LVI) (Neg), focal/suspicious LVI (FS-LVI), LVI, and extensive LVI (E-LVI) using univariable and multivariable regression with usual LVI as the referent
| Characteristic | Univariable Fine-Gray Model | Multivariable Fine-Gray Model | ||||||
|---|---|---|---|---|---|---|---|---|
| N | HR | 95% CI | P value | N | HR | 95% CI | P value | |
| Age | 9041 | 0.99 | 0.98, 1.00 | .003 | 7950 | 0.98 | 0.970.99 | <0.001 |
| Lymphovascular invasion | 8837 | 7,950 | ||||||
| Yes (LVI) | — | — | — | — | ||||
| No (Neg) | 0.62 | 0.52, 0.74 | <.001 | 0.69 | 0.550.85 | <.001 | ||
| Focal/Suspicious (FS-LVI) | 0.99 | 0.70, 1.40 | >.9 | 1.13 | 0.791.61 | .5 | ||
| Extensive (E-LVI) | 1.40 | 1.03, 1.89 | .029 | 1.62 | 1.152.27 | .005 | ||
| Tumor size | 9048 | 1.16 | 1.11, 1.20 | <.001 | 7950 | 1.19 | 1.121.26 | <.001 |
| Positive nodes removed | 8852 | 1.01 | 0.99, 1.03 | .3 | 7950 | 0.99 | 0.961.01 | .4 |
| ER | 9059 | 7950 | ||||||
| Negative | — | — | — | — | ||||
| Positive | 0.47 | 0.39, 0.57 | <.001 | 0.63 | 0.440.90 | .012 | ||
| PR | 9059 | 7950 | ||||||
| Negative | — | — | — | — | ||||
| Positive | 0.56 | 0.47, 0.66 | <.001 | 0.76 | 0.551.05 | .10 | ||
| HER2 | 9059 | 7950 | ||||||
| Negative | — | — | — | — | ||||
| Positive | 1.00 | 0.79, 1.27 | >.9 | 0.76 | 0.58, 1.01 | .056 | ||
| Histologic Grade | 8315 | 7950 | ||||||
| 1 | — | — | — | — | ||||
| 2 | 2.02 | 1.12, 3.64 | .020 | 2.03 | 1.04, 3.95 | .038 | ||
| 3 | 3.39 | 1.95, 5.87 | <.001 | 2.70 | 1.42, 5.10 | .002 | ||
| Surgery Type | 9056 | 7950 | ||||||
| Breast Conservation | — | — | — | — | ||||
| Mastectomy | 0.93 | 0.79, 1.10 | .4 | 0.34 | 0.260.44 | <.001 | ||
| Axillary Surgery | 9057 | 7950 | ||||||
| ALN | — | — | — | — | ||||
| SLN | 0.88 | 0.73, 1.06 | .2 | 0.96 | 0.751.22 | .7 | ||
| None | 1.74 | 1.11, 2.73 | .016 | |||||
| Margins | 9059 | 7950 | ||||||
| Negative | — | — | — | — | ||||
| ≤1 mm | 1.91 | 1.51, 2.42 | <.001 | 1.88 | 1.452.43 | <.001 | ||
| ≤2 mm | 1.78 | 1.34, 2.37 | <.001 | 1.61 | 1.182.21 | .003 | ||
| Positive | 2.59 | 1.81, 3.69 | <.001 | 2.23 | 1.423.51 | <.001 | ||
| Unknown | 4.70 | 2.52, 8.75 | <.001 | 3.16 | 1.168.58 | .024 | ||
| Chemotherapy | 9054 | 7950 | ||||||
| No | — | — | — | — | ||||
| Yes | 1.08 | 0.92, 1.27 | .4 | 0.86 | 0.691.07 | .2 | ||
| Radiation therapy | 9056 | 7950 | ||||||
| No | — | — | — | — | ||||
| Yes | 0.56 | 0.48, 0.66 | <.001 | 0.29 | 0.230.38 | <.001 | ||
Similarly, ER/PR status, grade, margins, and use of radiation therapy were all significantly associated with DM and overall survival in the univariable analysis. E-LVI was associated with poorer overall survival (HR, 1.66; 95% CI, 1.33-2.07; P < .001) and an increased risk of DM (HR, 1.71; 95% CI, 1.38-2.11; P < .001).
Multivariable analyses of the prognostic factors associated with LRR, overall survival, and DM among the overall cohort
Given the baseline differences between patient cohorts, a multivariable model was constructed to identify independent risk factors associated with LRR, DM, and overall survival by controlling for the effects of relevant clinicopathologic and treatment features (Table 3).
Table 3.
Multivariable analysis of risk of locoregional recurrence after backward stepwise selection
| Characteristic | N | HR | 95% CI | P value |
|---|---|---|---|---|
| Age | 7950 | 0.98 | 0.97, 0.99 | <.001 |
| Lymphovascular invasion | 7950 | |||
| Yes (LVI) | — | — | ||
| No (Neg) | 1.07 | 0.63, 1.81 | .8 | |
| Focal/Suspicious (FS-LVI) | 1.83 | 0.82, 4.10 | .14 | |
| Extensive (E-LVI) | 2.19 | 1.01, 4.75 | .047 | |
| Tumor size | 7950 | 1.11 | 1.01, 1.21 | .037 |
| ER | 7950 | |||
| Negative | — | — | ||
| Positive | 0.50 | 0.40, 0.62 | <.001 | |
| Histologic grade | 7950 | |||
| 1 | — | — | ||
| 2 | 1.95 | 1.0, 3.81 | .052 | |
| 3 | 2.56 | 1.35, 4.87 | .004 | |
| Surgery type | 7950 | |||
| Breast conservation | — | — | ||
| Mastectomy | 0.52 | 0.34, 0.78 | .002 | |
| Margins | 7950 | |||
| Negative | — | — | ||
| ≤1 mm | 1.81 | 1.39, 2.35 | <.001 | |
| ≤2 mm | 1.61 | 1.17, 2.20 | .003 | |
| Positive | 2.43 | 1.60, 3.69 | <.001 | |
| Unknown | 3.29 | 1.18, 9.18 | .023 | |
| Chemotherapy | 7950 | |||
| No | — | — | ||
| Yes | 0.79 | 0.63, 0.98 | .034 | |
| Radiation therapy | 7950 | |||
| No | — | — | ||
| Yes | 0.49 | 0.32, 0.74 | <.001 | |
| Lymphovascular invasion × tumor size | 7950 | |||
| No | 1.21 | 1.06, 1.38 | .005 | |
| Focal/Suspicious | 1.15 | 0.88, 1.52 | .3 | |
| Extensive | 1.01 | 0.86, 1.18 | >.9 | |
| Lymphovascular invasion × mastectomy | 7950 | |||
| No | 0.46 | 0.27, 0.78 | .004 | |
| Focal/Suspicious | 0.40 | 0.15, 1.08 | .070 | |
| Extensive | 0.69 | 0.34, 1.39 | .3 | |
| Lymphovascular invasion × radiation therapy | 7950 | |||
| No | 0.40 | 0.24, 0.68 | <.001 | |
| Focal/Suspicious | 0.46 | 0.19, 1.12 | .086 | |
| Extensive | 0.75 | 0.37, 1.54 | .4 |
Reassuringly, risk determinants known to predict LRR maintained significance (except for PR positivity [HR, 0.76; 95% CI, 0.55-1.05; P = .10]), suggesting that the analysis was sound. Notably, the presence of E-LVI retained significance as a predictor of LRR (HR, 1.62; 95% CI, 1.15-2.27; P = .005) relative to LVI as a reference. The group without LVI had a lower risk of LRR (HR, 0.69; 95% CI, 0.55-0.85; P < .001), while the group with FS-LVI did not behave differently compared with the reference (LVI), suggesting that FS-LVI carried the same implication as LVI. The use of radiation therapy and mastectomy was associated with a lower hazard of LRR in the multivariable analysis (Table 3; HR, 0.29; 95% CI, 0.23-0.38; and HR, 0.34; 95% CI, 0.26-0.44; P < .001). E-LVI was not an independent predictor of DM or overall survival in the multivariable analysis relative to LVI as a reference.
E-LVI and LVI propensity score matching analysis
In an alternative statistical approach to handling the imbalance in salient features between patients with E-LVI and LVI, we attempted to demonstrate the finding in a 2:1 propensity score matching analysis. A total of 378 E-LVI patients were matched with 756 LVI patients based on age, size, receptors, grade, surgery type, margins, and chemotherapy/radiation therapy application, although number of lymph nodes involved was unable to be balanced (Table S1). In the univariable analysis between these matched cohorts, a statistical difference was noted in overall survival (HR, 1.34; 95% CI, 1.00-1.80; P = .05) and in LRR (HR, 1.47; 95% CI, 1.02-2.14; P = .041), but not in DM (HR, 1.21; 95% CI, 0.92-1.59; P = .2) (Table 4). Notably, over 70% of patients in both cohorts received radiation therapy.
Table 4.
Risk of death, locoregional recurrence (LRR), distant metastasis (DM), and any invasive recurrence in a 2:1 propensity score matching analysis
| Propensity score matching analysis | |||||
|---|---|---|---|---|---|
| N | HR | CI | P value | ||
| Overall survival | 1134 | ||||
| LVI | — | — | |||
| E-LVI | 1.34 | (1.00, 1.80) | .050 | ||
| LRR | 1134 | ||||
| LVI | — | — | |||
| E-LVI | 1.47 | (1.02, 2.14) | .041 | ||
| DM | 1134 | ||||
| LVI | — | — | |||
| E-LVI | 1.21 | (0.92, 1.59) | .2 | ||
| Invasive rec | 1134 | ||||
| LVI | — | — | |||
| E-LVI | 1.18 | (0.92, 1.51) | .2 | ||
Discussion
We have shown a graduated increase in LRR risk going from no LVI to nonextensive LVI to extensive LVI. The association of E-LVI with LRR was shown by 1) cumulative incidence rates 2), univariable and multivariable Fine-Gray analyses, and (3) propensity score matching cohorts. Another novel and previously undescribed finding from our work is that “focal” or “suspicious” LVI carries a similar LRR risk as “usual” LVI (Fig. 2 and Tables 2 and 3). This report represents the largest, and to our knowledge, first analysis of E-LVI in invasive breast cancer, characterizing it as an adverse risk factor for LRR above and beyond the known effect of nonextensive LVI. The finding from the propensity score matching analysis further confirmed this effect. Taken in aggregate, these findings are highly suggestive of an incremental risk associated with the finding of E-LVI in terms of LRR and other disease endpoints relative to usual LVI.
Several prior studies have compared the risk of LRR with respect to LVI. These studies did so with breast conservation surgery (BCS), mastectomy, or both. For example, Dinshaw et al evaluated a cohort of 1022 women with pathologic stage I and II breast cancer treated with BCS with or without systemic adjuvant therapy. They concluded that LVI was the most important independent risk factor for LR (HR, 2.85; 95% CI, 1.68-4.83) and LRR (HR, 2.36; 95% CI, 1.52-3.66). They also found significant increases in the actuarial 5- and 10-year LRR in LVI patients versus their non-LVI counterparts.4 Furthermore, Nichol et al studied a cohort of 722 patients consisting of all women aged 70 to 79 years referred between August 1999, and May 2009, with pT1 pN0 M0 or pT1 cN0 M0 ER-positive breast cancer who were treated with lumpectomy achieving negative margins. They demonstrated that patients with LVI had lower event-free survival and had higher rates of LRR. Multivariate analysis demonstrated a significantly higher risk of events with grade 3 versus grade 1 histology (HR, 2.7; 95% CI, 1.4-5.3; P = .005), LVI-positive versus LVI-negative status (HR, 2.1; 95% CI, 1.02-4.5; P = .046), and treatment with hormonal therapy (HT) alone (HR, 2.2; 95% CI, 1.1-4.4; P = .01). Of note, the 10-year event-free-survival rate was 85% (95% CI, 73%-91%) for HT alone versus 90% (95% CI, 87%-92%) for HT–radiation therapy (log-rank P = .01).5 Similarly, Freedman et al evaluated 1478 women with T1mic, T1a, T1b, T1c, or T2 breast cancer with known LVI status treated with breast-conserving surgery and radiation therapy from February 1980 to August 2007. This study showed that LVI correlated with an increased risk of 5- and 10-year LRR (3.4% and 8.6%, respectively; P = .05) compared with no LVI (1.6% and 5.6%, respectively; P = .05). However, in that study, LVI was not found to be an independent predictor of LRR. The authors note that their results may have been confounded by the “all or nothing” indication of the presence of LVI recorded in their data, which is true of almost all studies. They state that this indication may lead to a relative dilution of the observed effect of LVI between cases with focal or minimal LVI and extensive LVI.17
At the MD Anderson Cancer Center, Dominici et al examined a total of 819 patients with invasive breast cancer who underwent mastectomy from January 2000 through December 2005. No patient received neoadjuvant chemotherapy (NAC). In a univariable analysis, the presence of LVI was seen to be associated with LRR (n, 10; HR, 5.08; 95% CI, 2.06-12.5; P < .01). In a multivariable analysis, triple-negative status was one of the strongest predictors of LRR. Triple-negative status and LVI increased the risk of LRR at 60 months to 30%.18 Jagsi et al conducted a retrospective analysis of a cohort of 870 node-negative patients (excluding T4 patients) treated with modified radical mastectomy without radiation therapy at Massachusetts General Hospital between 1980 and 2000. In a multivariable analysis, they found that LVI remained a significant independent predictor (n, 59; HR, 3.2; P = .0088).13
Finally, the presence of post-NAC LVI retains prognostic information. Hamy et al analyzed a cohort of 1033 T1-3Nx M0 patients with invasive breast cancer. Most patients were treated with NAC using an anthracycline- and taxane-based chemotherapy regimen (n = 713 [69%]) and either BCS (n = 703 [68%]) or mastectomy (n = 330 [32%]). Patients were designated as either post-NAC LVI positive or negative. The presence of post-NAC LVI differed significantly between breast cancer subtypes (luminal breast cancer, 42.2%; triple-negative breast cancer [TNBC], 19.4%; HER2-positive breast cancer, 19.2%; P < .0001) and tumor grade (grade 1-2L, 35%; grade 3, 24.3%; P = .0001). The presence of post-NAC LVI was also significantly associated with increased nodal involvement (N−, 14.5%; 1-3 N+, 44.5%; ≥N+, 60.5%; P < .0001). In a univariable analysis, LVI was associated with poor disease-free survival in these patients (HR, 2.54; 95% CI, 1.96-3.31; P < .001). However, this differed in magnitude between each subtype (luminal breast cancer [HR, 1.83; P = .003], TNBC [HR, 3.73; P < .001], and HER2-positive breast cancer [HR = 6.21, P < .001]).19
Limitations
Our findings must be interpreted in the context of our study design and its inherent limitations. Our results are subject to potential confounding by other risk factors for LRR. We performed a multivariable and propensity score matching analysis to reduce the effect of potential confounders on our primary outcome; nonetheless hidden confounders could exist. In particular, we were unable to fully balance number of involved lymph nodes in the propensity score matching despite trying different methods and matching algorithms or parameters. However, we note here that the number of involved lymph nodes was not shown to be a significant factor for LRR in any of our analyses, including multivariable regression and propensity score matching analysis. Another limitation of a retrospective study design is that there may have been variability in the reporting of LVI. In other words, the potential for misclassification bias arises. It is crucial to remember that not all institutions and pathologists will consistently identify LVI. This is further compounded by the lack of a uniform pathologic definition for “extensive LVI” in our study and was based on a subjective determination by the examining pathologist. To summarize, variability in reporting LVI owing to artifacts, technical reasons, interobserver variability, and lack of a consistent definition may undermine its prognostic value. Finally, biases can result from the data abstraction process, as abstractors are not blinded to the etiologic relation being studied or the hypothesis being tested. However, our institutional database is curated independently of any particular study question, and we further account for this potential by creating a standardized and systemic abstraction form to guide the abstraction process while using precise definitions to promote consistency between abstractors.
Conclusion
Our work suggests that patients with E-LVI are at a higher risk for LRR despite maximal standard of care treatment. Our finding may be especially informative and useful in those patients who have borderline indications for PMRT or regional nodal irradiation. The finding of E-LVI may be dispositive in these patients, that is they should receive PMRT/regional node irradiation. Hence, the finding of E-LVI could be clinically actionable, especially as omission-of-sentinel node biopsy trials are reported and fewer pathologic nodal evaluations are performed.
Supplementary Material
Disclosures:
A.J.K. has received research funding from Clovis Oncology, Merck, and Varian, unrelated to this work, and honoraria from Exact Sciences, as well as travel reimbursement from Accuray. M.M. has received honoraria from Roche. S.N.P. has received consulting fees and travel reimbursements from Rain Therapeutics Scientific and Repare Therapeutics and travel reimbursements from Varian Medical Systems. This research was funded in part through the National Institute of Health/National Cancer Institute Cancer Center Support Grant P30 CA008748.
Footnotes
Supplementary material associated with this article can be found in the online version at doi:10.1016/j.ijrobp.2024.04.073.
Data Sharing Statement:
Research data are stored in an institutional repository and will be shared upon request to the corresponding author.
References
- 1.Sampat MB, Sirsat MV, Gangadharan P. Prognostic significance of blood vessel invasion in carcinoma of the breast in women. J Surg Oncol 1977;9:623–632. [DOI] [PubMed] [Google Scholar]
- 2.Lauria R, Perrone F, Carlomagno C, et al. The prognostic value of lymphatic and blood vessel invasion in operable breast cancer. Cancer 1995;76:1772–1778. [DOI] [PubMed] [Google Scholar]
- 3.Song YJ, Shin SH, Cho JS, Park MH, Yoon JH, Jegal YJ. The role of lymphovascular invasion as a prognostic factor in patients with lymph node-positive operable invasive breast cancer. J Breast Cancer 2011;14:198–203. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Dinshaw KA, Budrukkar AN, Chinoy RF, et al. Profile of prognostic factors in 1022 Indian women with early-stage breast cancer treated with breast-conserving therapy. Int J Radiat Oncol Biol Phys 2005;63:1132–1141. [DOI] [PubMed] [Google Scholar]
- 5.Nichol AM, Chan EK, Lucas S, et al. The use of hormone therapy alone versus hormone therapy and radiation therapy for breast cancer in elderly women: A population-based study. Int J Radiat Oncol Biol Phys 2017;98:829–839. [DOI] [PubMed] [Google Scholar]
- 6.Lee BM, Chang JS, Cho YU, et al. External validation of IBTR! 2.0 nomogram for prediction of ipsilateral breast tumor recurrence. Radiat Oncol J 2018;36:139–146. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Chen SY, Tang Y, Song YW, et al. [Prognosis and risk factors of 1 791 patients with breast cancer treated with breast-conserving surgery based on real-world data]. Zhonghua Zhong Liu Za Zhi 2018;40:619–625. 10.3760/cma.j.issn.0253-3766.2018.08.011 [in Chinese]. [DOI] [PubMed] [Google Scholar]
- 8.Truong PT, Lesperance M, Culhaci A, Kader HA, Speers CH, Olivotto IA. Patient subsets with T1-T2, node-negative breast cancer at high locoregional recurrence risk after mastectomy. Int J Radiat Oncol Biol Phys 2005;62:175–182. [DOI] [PubMed] [Google Scholar]
- 9.Yildirim E, Berberoglu U. Can a subgroup of node-negative breast carcinoma patients with T1-2 tumor who may benefit from postmastectomy radiotherapy be identified? Int J Radiat Oncol Biol Phys 2007;68:1024–1029. [DOI] [PubMed] [Google Scholar]
- 10.Wallgren A, Bonetti M, Gelber RD, et al. Risk factors for locoregional recurrence among breast cancer patients: Results from International Breast Cancer Study Group trials I through VII. J Clin Oncol 2003;21:1205–1213. [DOI] [PubMed] [Google Scholar]
- 11.Floyd SR, Buchholz TA, Haffty BG, et al. Low local recurrence rate without postmastectomy radiation in node-negative breast cancer patients with tumors 5 cm and larger. Int J Radiat Oncol Biol Phys 2006;66:358–364. [DOI] [PubMed] [Google Scholar]
- 12.Taghian AG, Jeong JH, Mamounas EP, et al. Low locoregional recurrence rate among node-negative breast cancer patients with tumors 5 cm or larger treated by mastectomy, with or without adjuvant systemic therapy and without radiotherapy: Results from five national surgical adjuvant breast and bowel project randomized clinical trials. J Clin Oncol 2006;24:3927–3932. [DOI] [PubMed] [Google Scholar]
- 13.Jagsi R, Raad RA, Goldberg S, et al. Locoregional recurrence rates and prognostic factors for failure in node-negative patients treated with mastectomy: Implications for postmastectomy radiation. Int J Radiat Oncol Biol Phys 2005;62:1035–1039. [DOI] [PubMed] [Google Scholar]
- 14.Rosen PP. Tumor emboli in intramammary lymphatics in breast carcinoma: Pathologic criteria for diagnosis and clinical significance. Pathol Annu 1983;18:215–232. [PubMed] [Google Scholar]
- 15.Goldhirsch A, Wood WC, Coates AS, 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;22:1736–1747. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.College of American Pathologists. Protocol for the examination of resection specimens from patients with invasive carcinoma of the breast. Accessed Jan 07, 2024. https://documents.cap.org/protocols/Breast.Invasive_4.9.0.0.REL_CAPCP.pdf.
- 17.Freedman GM, Li T, Polli LV, et al. Lymphatic space invasion is not an independent predictor of outcomes in early stage breast cancer treated by breast-conserving surgery and radiation. Breast J 2012;18:415–419. [DOI] [PubMed] [Google Scholar]
- 18.Dominici LS, Mittendorf EA, Wang X, et al. Implications of constructed biologic subtype and its relationship to locoregional recurrence following mastectomy. Breast Cancer Res 2012;14:R82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Hamy AS, Lam GT, Laas E, et al. Lymphovascular invasion after neoadjuvant chemotherapy is strongly associated with poor prognosis in breast carcinoma. Breast Cancer Res Treat 2018;169:295–304. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.