Background

Globally, Breast cancer (BC) accounted for about 2.26 million cases in 2020, surpassing the number of lung cancers [1,2,3]. Moreover, BC is the 5th most common cause of cancer-related deaths [1].

The proportion of early-stage breast cancers continues to rise due to universal screening [4]. Advances in chemotherapy, targeted therapy, endocrine therapy, and immunotherapy have greatly improved breast cancer survival [5,6,7]. Due to early diagnosis and better prognosis, breast-conserving surgery is often recommended. Although there were some concerns about the surgery at the beginning [8], the exploration and improvement of the surgical style has not stopped since 1970s [9, 10]. In 1996, sentinel lymph node biopsy (SLNB) was adopted in BC staging and treatment and has promoted the development of breast-conserving surgery (BCS, or known as breast-conserving therapy) for BC [11, 12]. For early-stage BC, breast-conserving therapy extends disease-specific survival relative to mastectomy [13] but a high local recurrence risk (15-year relapse rate of 15.9-21.4%) has been observed [14].

Numerous breast cancer prognostic factors have been identified, including disease pathological stage, molecular subtype, lymph node invasion, lympho-vascular invasion, and histological grade [15,16,17,18,19]. Clinical studies show that lympho-vascular invasion correlates with breast cancer lymph node metastases and poor prognosis [15, 16]. A positive margin is associated with increased local recurrence (LR) in early-stage BC patients receiving BCS [20, 21]. Moreover, lympho-vascular invasion and extranodal tumor extension are risk indicators of breast cancer related lymphoedema [22, 23], which may affect the further treatment plan [24]. However, correlation between lympho-vascular invasion and LR or survival in breast cancer after breast-conserving therapy is controversial [25,26,27,28]. LVI is not systematically taken into account in decisions on breast cancer surgery (not mentioned in the National Comprehensive Cancer Network (NCCN) [29], Saint Gallen guidelines [30], or the European Society for Medical Oncology (ESMO) recommendations [31]). This meta-analysis of published data aimed to establish the prognostic significance of LVI in breast-conserving surgery.

Methods

Literature search and study selection

A systematic literature screening was done on Cochrane library, PubMed, Web of Science, and Embase, from inception to December 1st, 2021, including all prospective and retrospective investigations. The search terms used on PubMed were: ((breast conservative therapy) OR (breast-conserving surgery) OR (reserved mastectomy)) AND ((lymphovascular invasion) OR (lympho-vascular invasion) OR (lympho vascular invasion) OR (tumor thrombus) OR (carcinoma embolus)). Other search strategies are shown in Supplementary Material “Search strategy”. The searches confirmed with the Preferred Reporting Items For Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement [32, 33]. PRISMA 2020 checklist is shown in Supplementary Material “PRISMA Checklist”. Inclusion criteria were: 1) The study contains lympho-vascular invasion data after breast-conserving surgery; 2) Study has sufficient information for 95% confidence interval (95% CI) and Hazard Ratio (HR) analyses of the outcomes; 3) The study performed multivariate analyses. Exclusion criteria were: 1) Use of non-standard treatment; 2) Lack of distinction between LVI-unknown and LVI-positive patients in the analysis; 3) The survival data weren’t compared within BCS patients, for example: BCS patients with mastectomy patients. Case reports, letters, commentary articles, and conference abstracts were excluded.

Retrieval and quality assessment of data

Two independent researchers (YZ and FT) performed retrieval and quality assessment of data. The information extracted included number of patients included in studies, year of publication, first author’s name, study type, median follow-up months, breast cancer subtype, treatment type, and outcomes. The HR and 95% CIs were extracted from each study and classified by different outcomes. Prospective studies were assessed using Cochrane RoB 2.0 tool [34]. The Newcastle–Ottawa scale (NOS) was applied to analyze retrospective studies [35].

Statistical analysis

Pooled HR with 95% CIs (95% CI) were determined for all extracted outcomes (OS, DFS, EFS, LR, LRR, DM, BR, BCSS and DSS). I2 test was used to assess statistical heterogeneity. I2 >56% indicated significant heterogeneity. I2 < 31% indicated homogeneity. I2 between 31% and 56% indicated mild heterogeneity [36]. The meta-analysis was done by applying the random effect model. Egger, Funnel plots, and Begg tests were utilized for the assessment of publication bias. P ≤0.05 (2-sided) indicated as statistical significance. All analyses were carried out with the Stata, version 16.0 (Stata Corporation, College Station. TX).

Results

Study selection and quality assessment

Our search strategy identified 716 records and 5 additional records were identified from references in these studies (Fig. 1). After 258 duplicate records removal, 458 records remained. Of these, 372 were excluded after title and abstract review. Of the remaining 86 records, 5 reports couldn’t be retrieved, 19 were excluded because of unrelated topic, 52 weren’t meeting inclusion criteria or meeting exclusion criteria, as shown in the flow diagram (Fig. 1). Finally, 15 full-text studies involving 21,704 patients were included this meta-analysis (Table 1). Among these, 3 were prospective and 12 were retrospective. The characteristics and quality of the studies are shown on Tables 1, 2, 3.

Fig. 1
figure 1

PRISMA 2020 flow diagram for meta-analysis

Full size image
Table 1 The characteristic of each involved study
Full size table
Table 2 Cochrane RoB 2.0 tool evaluate for the RCTs
Full size table
Table 3 NOS scale for the cohort studies
Full size table

Prognosis of lympho-vascular invasion after breast-conserving surgery

Of the prospective studies, Freedman GM et al.[42] showed the outcomes of LRR and OS (HR = 1.254, 95% CI 0.944–1.667, p = 0.12), Mittendoff EA et al.[43] analyzed LRR (HR = 1.49, 95% CI 1.02–2.17, p = 0.039), and Magee B et al.[37] identified LR (HR = 1.78, 95% CI 1.035–3.063, p = 0.037) as outcomes. Additionally, the study of Freedman GM et al.[42] only contained HR for LRR, except for the 95% CI and p-value; Magee B et al. showed eβ (equal to Hazard Ratio), and p-value, in order to have the unification, we transferred these data into HR and 95% CI according to the functions conducted by Altman DG et al.[48] Because the three prospective studies involved different outcomes, the meta-analysis could not be applied on them. Thus, the meta-analysis was conducted on the retrospective studies assessed by random effects model. The HR and 95% CI data were separately pooled from each study. Outcomes included: overall survival (OS), disease-free survival (DFS), event-free survival (EFS), local recurrence (LR), loco-regional recurrence (LRR), distant metastases (DM), breast recurrence (BR), breast cancer specific survival (BCSS), and disease specific survival (DSS). According to the definitions in these studies, BR is equal to LR and BCSS is equal to DSS. Major outcomes in each study are shown on Table 1. The meta-analysis on retrospective studies showed that lympho-vascular invasion (LVI) after breast-conserving surgery significantly worsened OS, DFS, LRR, BCSS, DM and LR (Fig. 2). Results of heterogeneity tests are shown on Fig. 2. There were only two studies included in the BCSS, and the heterogeneity was relatively high (I2 = 73.8%, p = 0.051). On the other hand, the conclusion of each study both showed the significant difference in the BCSS outcomes, so we accepted this heterogeneity. Mild heterogeneity was also observed in DFS (I2= 50.5%, p = 0.133). Thus, we evaluated its similarity to that of BCSS and accepted the heterogeneity. Taken together, this meta-analysis found that LVI is a significant prognostic factor in early-stage breast cancer after BCS.

Fig. 2
figure 2

Forest plot of meta-analysis and cumulative meta-analysis in primary outcomes

Full size image

Publication bias and sensitivity analyses

Funnel plots indicated a symmetric distribution of included studies. Begg and Egger tests (Begg = 0.112 > 0.05, Egger = 0.279 > 0.05) revealed no publication bias in these studies (Fig. 3), which was confirmed by sensitivity analyses.

Fig. 3
figure 3

Publication Bias analysis for the meta-analysis

Full size image

Discussion

Lympho-vascular invasion correlates with lymph node metastases and poor breast cancer prognosis [15, 16]. Breast-conserving surgery is a standard treatment option for early-stage breast cancer. However, few studies have focused on the prognostic role of LVI after breast-conserving surgery. This meta-analysis involved studies on breast-conserving surgery that contain LVI data. The prospective studies included showed the significant longer LR and LRR in LVI positive patients than LVI negative patients who underwent breast-conserving therapy, whereas with the similar OS [37, 42, 43]. However, we could not meta-analyze the prospective studies due to insufficient data. The conclusions of the retrospective studies included in our meta-analysis were controversial. DFS, LRR and BCSS were significantly different between LVI-positive and negative patients, studies showed a poor prognosis in LVI-positive patients; however, controversial conclusions were conducted in OS, DM and LR. In this meta-analysis, we conclude that early-stage breast cancer patients after breast-conserving surgery with LVI showed poorer OS, DFS, LRR, BCSS, DM and LR than those without LVI.

We noticed that some studies chose OS as outcome, showed different trends in HR or p-value: In the study of Nichol AM et al. [26] the HR is less than 1 (HR = 0.77) with a p-value greater than 0.05 (p = 0.43), while Dinshaw KA et al. [25, 49] showed the significance in the OS (p = 0.001) and the HR is greater than 1 (HR = 2.01) (Fig. 2). Sensitivity analyses after exclusion of both studies and redoing the meta-analysis with the remaining studies also revealed significant difference (Fig. 4), and low heterogeneity (I2 = 0%, p = 0.954). Indicating that all these studies should be included in the meta-analysis.

Fig. 4
figure 4

Forest blot of survival data for sensitivity analyses

Full size image

A variety of patient, treatment, and pathologic factors have been reported to be associated with increased risk of local recurrence after breast conservation therapy. For breast cancer, positive microscopic margins are associated with a ≥ twofold higher risk of local recurrence relative to negative margins [20]. Thus, re-excision to achieve negative margins should be done for most patients with positive margins [50]. Our findings show that LVI positive breast cancer patients undergoing breast conservation therapy have a twofold higher risk of local recurrence relative to LVI-negative patients. Local recurrence after breast-conserving surgery for invasive cancer can influence patient survival. The Early Breast Cancer Trialists Collaborative Group (EBCTCG) found that 1 life is saved at 15-year follow-up for every 4 local recurrences prevented at 10 years after lumpectomy [51]. Our results show poor survival in LVI-positive breast cancer patients with breast conservation therapy. Thus, patients undergoing breast conservation therapy need to know about the predictive value of LVI on local recurrence and survival and mastectomy, with or without breast reconstruction should be considered, especially for patients who need positive margin re-excision. Recent studies show that breast reconstruction after a mastectomy has similar results to breast-conserving surgery in terms of quality of life [52].

Our meta-analysis based on retrospective studies may carry bias due to retrospective data analysis. The advent of molecular subtyping of breast cancer has changed the paradigm for breast cancer treatment. Neoadjuvant therapy has been standard care for human epidermal growth factor receptor 2 overexpressing, and triple negative breast cancers [53]. LVI in patients treated with neoadjuvant chemotherapy was an independent predictor of local recurrence, distant metastasis, and overall survival in all breast carcinomas [54]. Clinical studies on if LVI is an independent prognostic factor in stage T1N0M0 breast cancer, and if further systemic treatment or mastectomy can improve the prognosis of LVI-positive patients LVI are needed.

Conclusions

We find that early-stage breast cancer patients after breast-conserving surgery with LVI showed poorer OS, DFS, LRR, BCSS, DM and LR than those without LVI. Mastectomy or its combination with radical systemic therapies could be considered, especially for patients who need a second surgery. How to change the impact of LVI on the local recurrence rate and long-term survival in patients who undergo breast-conserving surgery may be a valuable research direction in the future.