Can axillary lymph node dissection be omitted in breast cancer patients with 1-2 positive sentinel nodes? A systematic review and meta-analysis

Abstract

Purpose

Axillary lymph node dissection (ALND) has traditionally been recommended for breast cancer patients with positive sentinel lymph nodes (SLNs). Although omitting ALND is now widely accepted for patients undergoing breast-conserving surgery (BCS) with limited sentinel lymph node biopsy (SLNB) involvement, based on trials such as Z0011, evidence for patients undergoing total mastectomy (TM) remains limited and conflicting. This meta-analysis aimed to evaluate whether ALND can be safely omitted in TM patients with 1-2 positive SLNs by comparing survival outcomes between ALND and SLNB alone groups.

Methods

A systematic search of PubMed, Embase, Cochrane Library, and Web of Science (up to December 2024) identified 29 studies (6 randomized controlled trials and 23 observational) involving a total of 146 407 patients. Survival outcomes, including overall survival (OS), disease-free survival (DFS), and recurrence-free survival (RFS), were pooled using random- or fixed-effects models. Heterogeneity and publication bias were assessed using I² statistic, Begg’s test, and Egger’s test. Subgroup analyses were performed based on study design, type of surgical (TM vs. BCS), and pathological T-stage.

Results

Overall, no significant differences in OS (OR = 0.93, 95% confidence interval [CI]: 0.83-1.04), DFS (OR = 1.02, 95% CI: 0.87-1.20), or RFS (OR = 1.08, 95% CI: 0.89-1.30) were observed between ALND and SLNB alone groups. However, in the TM subgroup, ALND was associate with improved OS (OR = 0.75, 95% CI: 0.62-0.90). Similarly, patients with T3-4 tumors demonstrate better OS outcomes with ALND. No significant differences in DFS or RFS were observed across subgroups.

Conclusion

SLNB alone provides comparable survival outcomes to ALND in early breast cancer (EBC) patients with 1-2 positive SLNs, supporting its safety in those undergoing BCS. However, while our analysis suggests a potential survival advantage with ALND in TM patients and those with advanced T-stage (T3-4), this observation requires cautious interpretation due to potential selection bias, residual confounding from unmeasured variables and confounding by indication where ALND may reflect more intensive multimodal therapy rather than causally improving OS. Therefore, these subgroup findings should not be considered practice-changing at this stage. Further high-quality prospective studies are warranted to validate these associations and optimize patient selection criteria.

Implications for practice.

This meta-analysis supports omitting ALND in early breast cancer patients with 1-2 positive SLNs undergoing BCS, reducing morbidity without compromising survival. However, ALND may improve survival in TM patients and those with advanced T3-4 tumors, highlighting the need for tailored surgical decisions. Clinicians should prioritize individualized risk assessment, balancing de-escalation benefits against potential risks in higher-risk subgroups. These findings refine current guidelines and emphasize context-specific management, potentially sparing many patients from ALND-related complications while guiding targeted use in select populations. Further research is warranted to validate the criteria for ALND omission in TM cohorts.

Introduction

For more than a century, breast cancer (BC) surgery has adhered to the Halstead model, which posits that BC is primarily a localized disease that spreads through the lymphatic system and can be cured by extensive surgical excision.¹ Numerous studies have demonstrated that axillary lymph node status is a key prognostic factor in BC. Since the 19th century, axillary surgery has been an integral component of the localized treatment for BC.² In the 20th century, Axillary lymph node dissection (ALND) became a standard treatment for early-stage breast cancer (EBC) in the twentieth century.³ However, ALND is associated with significant complications, including reduced range of motion, lymphedema, and neuralgia.⁴ Sentinel lymph node biopsy (SLNB), a minimally invasive procedure, has gradually replaced ALND as the standard approach for axillary staging in patients with clinically lymph node-negative EBC.^3,5,6 For patients having negative sentinel lymph nodes (SLNs), omitting ALND has become the consensus recommendation in most treatment guidelines.^5–7

Compared with SLNB alone, the role of ALND in overall survival (OS) and disease-free survival (DFS) for patients with positive SLNs has been increasingly questioned. Several randomized controlled trials (RCTs),^8–12 including the landmark Z0011 trial,⁸ have demonstrated no significant improvement in OS or DFS when ALND was performed in patients with EBC who have limited SLN metastases and who undergo breast-conserving surgery (BCS). Those findings suggest that ALND may be safely omitted in those patients, even in the presence of positive SLNs. Although BCS is the preferred surgical option for patients with EBC, total mastectomy (TM) might be necessary in certain cases because of factors such as tumor size and location. Nevertheless, only patients receiving BCS were eligible to participate in the Z0011 trial, possibly because of the higher baseline risk in patients undergoing TM, thus limiting the generalizability of its conclusions. Furthermore, most other RCTs investigating this topic did not differentiate between BCS and TM, leaving it unclear whether ALND can also be omitted in patients with SLN-positive disease treated with TM. The present updated meta-analysis directly addresses that gap by examining the largest patient cohort to date (N = 146 407) and, most importantly, by incorporating findings from newly published, large-scale RCTs^13,14 specifically designed to evaluate axillary de-escalation strategies in patients with limited SLN disease undergoing TM. Nevertheless, as new evidence emerges and clinical practice guidelines continue to evolve, the practice of downgrading ALND is anticipated to also eventually extend to patients undergoing TM.¹⁵

In recent years, several retrospective observational studies and RCTs have investigated the feasibility of omitting ALND in patients with limited axillary lymph nodes after TM. However, conflicting results have been reported.^14,16–18 While some studies have shown that omission of ALND does not significantly affect OS,^16,17 but others have shown that OS is significantly improved in patients treated with TM who undergo ALND after SLNB reveals limited metastases.¹⁴ Similar inconsistencies have been observed with respect to DFS.^16–18 Those discrepancies might be attributable to variations in sample sizes and clinicopathologic characteristics across studies.

Given these conflicting findings, a comprehensive meta-analysis was warranted to evaluate the effect of ALND on survival outcomes, specifically OS and DFS, in patients with EBC and limited SLN metastases. Our objective was to determine whether ALND could be safely omitted for patients with SLN-positive disease undergoing TM, as is current practice in patients undergoing BCS. By integrating recent prospective data with existing retrospective studies, we aimed to develop a more precise and evidence-based assessment to guide clinical decision-making.

Methods

Search strategy and study selection

Eligible articles published up to December 31, 2024, were identified by a comprehensive search of the Web of Science, Embase, PubMed, and Cochrane Library databases. The following search strategy was used: (segmental mastectomy OR breast conserving surgery OR breast sparing surgery) AND (lymph node excision OR ALND OR ALND) AND (SLNB OR SLNB) AND (positive SLNs) AND (radiation therapy OR postoperative radiotherapy) AND (breast neoplasms OR breast tumors OR BCs OR breast carcinomas OR mammary cancers OR mammary neoplasms OR mammary carcinomas OR mammary tumors). No restrictions were applied regarding title, abstract, language, and MeSH terms. Two authors (Jinyi Xie and Ning Zhang) independently conducted the literature screening. Initial screening was based on titles and abstracts, followed by full-text review and examination of reference lists to identify potentially eligible studies. Any disagreements during the process of literature selection, quality assessment, data extraction, or analysis were resolved through discussion between at least 2 authors until consensus was reached. This systematic review has been registered with the International Prospective Register of Systematic Reviews (PROSPERO) maintained by the National Institute for Health Research (Registration number: CRD42024585933).

Inclusion and exclusion criteria

The following inclusion criteria were utilized for the eligible studies: (1) included patients with BC; (2) all patients underwent TM or BCS; (3) all patients underwent SLNB and had positive SLNs; (4) compared the survival differences between ALND and SLNB patient groups; (5) provided OS, DFS or recurrence-free survival (RFS); (6) full text was available and published in English. Comments, letters, conference abstracts, animal studies, case reports, reviews, and duplicate or overlapping publications were excluded from the meta-analysis.

Outcomes assessment

The primary survival outcomes assessed in the meta-analysis were OS, which was defined as the time from surgery to death from any cause. The secondary outcomes included DFS and RFS. DFS was defined as the time from surgery to the occurrence of any type of disease recurrence. RFS was defined as the time from surgery to the first recurrence of the disease, excluding death from other causes.

Quality assessment

Our meta-analysis included both RCTs and observational studies. Accordingly, we employed the Newcastle–Ottawa Scale (NOS) to assess the quality of observational studies,¹⁹ while the Cochrane Collaboration’s Risk of Bias Tool (RoB2) was used to evaluate RCTs.²⁰ Two authors (Jinyi Xie and Ning Zhang) conducted the quality assessments respectively. The NOS is a widely accepted tool for evaluating the quality of observational studies. It comprises 3 dimensions encompassing 8 items and utilizes a semiquantitative star rating system. Each item (except for comparability, which can receive up to 2 stars) can be awarded a maximum of 1 star, yielding a total score ranging from 0 to 9. Higher scores indicate higher methodological quality. The NOS scores for the included observational studies are presented in Table 1. In contrast, the Cochrane Collaboration’s RoB2 for randomized trials categorizes potential sources of bias into 7 domains: (1) random sequence generation (selection bias), (2) allocation concealment (selection bias), (3) blinding of participants and personnel (performance bias), (4) blinding of outcome assessment (detection bias), (5) incomplete outcome data (attrition bias), (6) selective reporting (reporting bias), and (7) other sources of bias. Each domain was rated as having a low, high, or unclear risk of bias. For those rated as “unclear,” justification was provided. A summary of the risk of bias assessments for the relevant RCTs is shown in Figure 1.

Table 1.

Characteristics of the included studies.

Author	Year	Country	Study design	Pathologic T stage	Breast surgery	No. of pts with sentinel lymph node biopsy (SLNB)	No. of pts with axillary lymph node dissection (ALND)	Follow-up time	Survival outcomes	Risk of bias (NOS/RoB2 score)
Arisio et al.	2019	Italy	Observational	T1-3	Breast-conserving surgery (BCS)/Total mastectomy (TM)	211	406	84.4 months (SLNB: 74 months ALND: 90 months)	Overall survival (OS), recurrence-free survival (RFS)	9
Bilimoria et al. (1)	2009	USA	Observational	T1-3 (macro-metastasis)	BCS/TM	1458	18 617	63 months (SLNB: 64 months ALND: 62 months)	OS	9
Bilimoria et al. (2)	2009	USA	Observational	T1-3 (micro-metastasis)	BCS/TM	530	1673	63 months (SLNB: 64 months ALND: 62 months)	OS	9
Chun et al.	2024	Korea	Observational	T1-3	TM	237	406	65 months	DFS	9
Cocco et al.	2022	USA	Observational	T1-2	BCS/TM	3030	5446	57.9 months	OS	8
Wild et al.	2024	Netherlands	Randomized controlled trial (RCT)	T1-2	TM	219	437	6.0 years	OS	Low risk
FitzSullivan et al.	2017	USA	Observational	T1-4	TM	96	546	66 months	OS, RFS	7
Gao et al.	2022	China	Observational	T1-2	BCS/TM	268	893	36 months	OS, RFS	8
Giuliano et al.	2017	USA	RCT	T1-2	BCS	446	445	9.3 years	OS, DFS	Low risk
Jo et al.	2020	Korea	Observational	T1-2	BCS	90	271	94.38 months (SLNB: 69.57 months ALND: 101.38 months)	DFS	8
Joo et al.	2019	Korea	Observational	T1-4	TM	158	1539	93 months	OS, DFS	9
Kim et al.	2020	Korea	Observational	T1-2	TM	179	704	54 months	OS	8
Lee et al.	2018	Korea	Observational	T1-2	BCS	1268	3174	SLNB: 30.86 months ALND: 47.24 months	OS	9
Park et al.	2014	Korea	Observational	T1-2	BCS/TM	197	2384	42.0 months (SLNB: 41.1 months ALND: 42.1 months)	OS	9
Reyna et al.	2021	USA	Observational	T3-4	TM	769	1148	SLNB: 40.3 months ALND: 52.3 months	OS	9
Schwieger et al.	2024	USA	Observational	T1-2	TM	8427	11574	74.8 months (SLNB: 68 months ALND: 79.7 months)	OS	9
Sola, et al.	2013	Spain	RCT	T1-3	BCS/TM	121	112	62 months	DFS	Low risk
Sun et al.	2021	USA	Observational	T1-3	TM	128	201	51 months	OS	7
Wang et al.	2014	USA	Observational	T1-2	BCS	393	876	73 months	OS	9
Yamamoto et al.	2012	Japan	Observational	T1-2	BCS/TM	152	308	SLNB: 60 months ALND: 79 months	OS	9
Yi et al.	2013	USA	Observational	T1-2	BCS/TM	188	673	5.2 years (SLNB: 5.5 years ALND: 4.9 years)	DFS	9
Zaveri et al.	2023	USA	Observational	T1-2	TM	126	422	84 months (SLNB: 65.3 months ALND: 87.3 months)	OS, DFS	9
Boniface et al.	2024	Sweden, Denmark, Germany, Greece, Italy	RCT	T1-3	BCS/TM	1335	1205	46.8 months	OS, RFS	Low risk
Fu et al.	2014	USA	Observational	T1-4	TM	106	108	43.6 months	OS, RFS	9
Galimberti et al.	2018	Australia, Belgium, Denmark, France, Italy, New Zealand, Peru, Slovenia, Switzerland	RCT	T1-3	BCS/TM	469	465	9.7 years	OS, DFS	Low risk
Gou et al.	2022	USA	Observational	T1-2	BCS	17143	17797	99.0 months	OS	8
Houvenaeghel et al.	2023	France	Observational	T1-2	BCS	185	1266	70.0 months (SLNB: 65.4 months ALND: 71.2 months)	OS, DFS	7
Lim et al.	2021	Singapore	Observational	T1-2	TM	92	168	61 months	OS, DFS	7
Picado et al.	2018	USA	Observational	T1-2	TM	13821	20422	53 months	OS	9
Tinterri et al.	2022	Italy	RCT	T1-2	BCS/TM	440	439	34.0 months	OS, RFS	Low risk

Figure 1.

Risk of bias assessment for included randomized controlled trials (RCTs). This figure presents the methodological quality evaluation of 6 RCTs across 7 domains of bias risk, using symbols to denote judgments: + (low risk), − (unclear risk), and x (high risk). Domains assessed include: D1: Random sequence generation (selection bias), D2: Allocation concealment (selection bias), D3: Blinding of participants and personnel (performance bias), D4: Blinding of outcome assessment (detection bias), D5: Incomplete outcome data (attrition bias), D6: Selective reporting (reporting bias), Other bias: Additional sources of bias. Studies are labeled with first authors and publication years. Assessments were performed following the Cochrane collaboration’s risk of bias tool.

Data extraction

Two authors (Jinyi Xie and Ning Zhang) independently extracted the following data from each eligible studies, including (1) author name, (2) publication year, (3) country, (4) study design, (5) pathologic T stage, (6) type of breast surgery, (7) number of patients undergoing SLNB or ALND, (8) duration of follow-up time, and (9) survival outcomes (OS, DFS, and RFS). Data on OS, DFS, and RFS were extracted directly from text, tables or survival curves. The number of patients in each group was also extracted to enable comparative analysis between the SLNB and ALND groups.

Statistical analysis

Statistical analyses were conducted using Review Manager version 5.4, StataMP 17.0 and R studio version 4.4.1. Odds ratios (ORs) with 95% confidence intervals (CIs) were pooled to evaluate the association between survival outcomes and whether ALND was performed following SLNB. The pooled results were presented using forest plots. Heterogeneity among studies was assessed using the I² statistic, which quantifies the proportion of variation across studies that is due to heterogeneity rather than chance. A random-effects model was applied if significant heterogeneity was detected (P < .10 or I² > 50%). In these instances, subgroup analyses were performed based on breast surgery type, pathologic T stage, or study design to explore the potential sources of heterogeneity. Conversely, a fixed effect model was adopted if heterogeneity was not significant (P > 0.10 or I² < 50%).^21,22 Moreover, publication bias was evaluated using both Begg’s and Egger’s tests. Egger’s test yielded a P-value of 0.814, while Begg’s test yielded a P-value of 1.00; both values exceeded the 0.05 threshold, indicating no significant publication bias. A funnel plot was also ­generated to visually assess publication bias and is presented in Figure S1 (see online supplementary material for a color version of this figure).

Results

Study characteristics

The initial database search yielded 2307 publications, of which 882 were duplicates. A review of titles and abstracts found 1364 publications to be ineligible, and they were excluded. A subsequent search of the full text of the remaining 61 studies excluded 40 studies for irrelevance, lack of survival outcomes data, or incomplete data. Eight studies^{11,13,16,23–27} located through other means (quoted or cited literature, etc.) were added, yielding 29 studies,^{8,10,11,13,14,16–18,23–43} of which 6 were RCTs and 23 were observational studies (Figure 2). The 8 additional studies had not been retrieved in the original database search because of terminology gaps, indexing limitations, and conceptual drift. The studies by Bilimoria et al. were categorized into Karl Y. Bilimoria (1) and Karl Y. Bilimoria (2), with 1 focusing on macrometastases and the other on micrometastases in lymph nodes. All included studies involved patients with BC and limited lymph node metastases who underwent either BCS or TM. Some patients received ALND; others received only SLNB. Table 1 presents detailed characteristics of the included studies.

Figure 2.

Flow chart of literature selection. The schematic diagram illustrated the systematic identification and screening of studies for inclusion. Initial database searches yielded 2315 records, with 8 additional studies identified through other sources. After removing 882 duplicates, 1433 titles/abstracts were screened for relevance. Of these, 69 full-text articles were assessed for eligibility, with 1364 studies excluded due to irrelevant topics, inappropriate article types and studies not published in English. Ultimately, 29 studies met the predefined inclusion criteria and were retained for analysis.

Survival outcomes

OS comparison

The difference in OS between patients receiving SLNB alone and those receiving ALND was evaluated in 25 studies. Our meta-analysis using a random-effects model (Figure 3) yielded a pooled OR of 0.93 (95% CI: 0.83-1.04), indicating no statistically significant difference in OS between the SLNB and ALND groups and suggesting that ALND does not confer a survival benefit. However, substantial heterogeneity was observed (I² = 80.5%), prompting further subgroup analyses based on study type, surgical approach, and pathological tumor stage.

Figure 3.

Forest plot comparing overall survival between sentinel lymph node biopsy (SLNB) alone and axillary lymph node dissection (ALND) groups. Pooled analysis of 26 studies demonstrates no significant difference in overall survival between SLNB alone and ALND (odds ratio [OR] = 0.93, 95% confidence interval [CI]: 0.83-1.04; P = .19). Results were calculated using a random-effects model (Mantel–Haenszel method) with significant heterogeneity across studies (I² = 80%, P < .001). Horizontal lines represent individual study effect sizes (95% CI); the diamond indicates the pooled estimate.

The subgroup analyses demonstrated consistency across RCTs and observational studies, with no significant OS difference between the SLNB and ALND groups (Figure 4). When stratified by surgical modality, patients who underwent TM had a pooled OR of 0.75 (95% CI: 0.62-0.90), indicating improved OS in the ALND group and suggesting that ALND might be a protective factor in this subgroup. In contrast, and consistent with the overall findings, patients who underwent BCS or whose surgical type was not clearly specified experienced no significant difference in OS whether they received SLNB or ALND (Figure 5). In terms of pathological tumor stage, only one study focused on patients with T3-4 disease, reporting higher OS in the ALND group. For patients staged at T1-2, T1-3, and T1-4, pooled analyses found no significant differences in OS between the SLNB and ALND groups (Figure 6).

Figure 4.

Comparison of overall survival (OS) in axillary lymph node dissection (ALND) versus sentinel lymph node biopsy (SLNB) in different study designs. A random-effects meta-analysis was performed, stratified by study design (observational vs. randomized controlled trials [RCTs]). Pooled ORs with 95% confidence intervals are shown. Heterogeneity was high overall (I² = 80%, τ²=0.04, P < .001) and within subgroups (observational: I² = 82%, RCT: I² = 77%). No significant difference in OS was observed between ALND and SLNB overall (OR = 0.93, 95% CI: 0.83-1.04; P = .19) or within subgroups (P = .81 for subgroup differences).

Figure 5.

Comparison of overall survival (OS) in axillary lymph node dissection (ALND) versus sentinel lymph node biopsy (SLNB) in different surgical types. Forest plot of pooled odds ratios (ORs) with 95% confidence intervals (CIs) from our meta-analysis comparing OS outcomes in patients undergoing ALND versus SLNB. Subgroup analyses were performed based on breast-conserving surgery (BCS) or total mastectomy (TM). Random-effects models were applied due to significant heterogeneity (I² = 80%, P < .001). Overall, no statistically significant survival difference was observed between ALND and SLNB (odds ratio [OR] = 0.93, 95% confidence interval [CI]: 0.83-1.04; P = .19), though subgroup analyses revealed differential effects between BCS and TM cohorts (test for subgroup differences: P = .006). Squares represent study-specific estimates; diamond indicates pooled effect size.

Figure 6.

Comparison of overall survival (OS) in axillary lymph node dissection (ALND) versus sentinel lymph node biopsy (SLNB) in different pathologic T stages. Forest plot meta-analysis of pooled odds ratios (ORs) with 95% confidence interval (CI) for OS in patients undergoing ALND versus SLNB, stratified by pathologic T stage subgroups (T1-3, T1-2, T1-4, T3-4). A random-effects model was applied due to significant heterogeneity (I² = 80%, P < .001). Subgroup differences were statistically significant (χ² = 15.53, P = .001). Overall, no significant OS difference was observed (OR = 0.93, 95% CI: 0.83-1.04, P = .19), though reduced mortality with SLNB was noted in the T3-4 subgroup (OR = 0.62, 95% CI: 0.50-0.77, P < .001). Study weights and individual trial data are shown.

DFS and RFS comparisons

Ten studies reported DFS outcomes for the SLNB and ALND groups. Our meta-analysis using a fixed-effects model resulted in a pooled OR of 1.02 (95% CI: 0.87-1.20), indicating no significant difference in DFS between the groups and suggesting that ALND does not confer a DFS benefit. The heterogeneity between studies was minimal (I² = 0.0%) (Figure S2, see online supplementary material for a color version of this figure).

Six studies reported RFS outcomes for the SLNB and ALND groups. A fixed-effects model revealed a pooled OR of 1.08 (95% CI: 0.89-1.30), indicating no significant difference in RFS between the groups. Receiving an ALND did not significantly improve the RFS rate. The I² was 30.3%, indicating little heterogeneity between the studies (Figure S3, see online supplementary material for a color version of this figure).

Discussion

Our meta-analysis included 6 RCTs and 23 observational studies assessing the efficacy of ALND compared with SLNB alone in patients with EBC and SLN metastases. Our findings indicate that, in RCTs and observational studies alike, SLNB alone yields OS outcomes comparable to those for ALND. ALND did not confer a significant survival advantage, except in patients receiving TM or those with T3-4 stage disease, suggesting that ALND might be able to be selectively omitted in certain patients with EBC and limited SLN involvement. Furthermore, compared with ALND, SLNB alone was not associated with a significant reduction in DFS or RFS.

Several previous meta-analyses^44–46 have similarly examined survival outcomes for ALND and SLNB alone in patients with EBC and metastatic SLNs, typically reporting no significant differences. For instance, a meta-analysis by Glechner et al.,⁴⁴ which included 1 RCT and 2 observational studies, provided preliminary evidence supporting the safety of omitting ALND. However, the limited number of studies—primarily because of the availability of eligible data at the time—might have constrained the statistical power and robustness of the conclusions. Moreover, the DFS data analyzed were derived solely from the Z0011 trial, limiting the generalizability of the findings. Another meta-analysis⁴⁵ that incorporated 5 RCTs and 7 observational studies, also concluded that ALND could be safely omitted in patients with EBC and SLN metastases. However, the lack of subgroup analyses based on surgical approach might limit the applicability of those results. Similarly, Huang et al.,⁴⁶ analyzed 1 RCT and 6 retrospective studies, finding that, compared with ALND in patients having 1 or 2 positive SLNs, SLNB alone was associated with comparable OS, DFS, and RFS outcomes, alongside a substantially lower incidence of lymphedema. Nevertheless, the use of inconsistent effect measures—some studies reporting ORs and others hazard ratios (HRs)—and their separate analyses might have introduced potential bias.

Our updated meta-analysis, which included 29 studies with a total of 146 407 patients, significantly expands upon previous work and strengthens the statistical power and credibility of the results. The quality of the included observational studies was assessed using NOS,¹⁹ with 14 studies scoring 9, 5 scoring 8, and 4 scoring 7, indicating high overall quality. The RCTs were evaluated using RoB2,²⁰ with 1 trial identified as having a high risk of bias in 2 out of 7 domains. Importantly, our analysis incorporated 2 newly published RCTs from 2024.^13,14 The inclusion of those 2 trials, alongside 4 other RCTs and a substantial body of observational data, provides a uniquely comprehensive and contemporary evidence base. The registry study by Wild et al.¹⁴ which included 1090 patients demonstrated that among patients with clinical T1-2N0 BC who underwent TM and had low-volume SLN metastases, the 5-year regional recurrence rate was low and comparable for patients with and without ALND. Critically, omission of completion axillary treatment did not compromise oncologic safety, even without adjuvant chest wall radiotherapy, supporting de-escalation strategies in carefully selected patients (eg, older age, favorable tumor biology). The multicenter trial by Boniface et al.,¹³ which included 2766 patients, also provided strong evidence supporting the safety of omitting ALND in clinically node-negative patients with T1-3 tumors and 1-2 macrometastatic SLNs. It also demonstrated that omitting ALND is noninferior to using ALND for RFS (5-year rates: 89.7% vs. 88.7%), with most patients receiving nodal radiotherapy and systemic therapy. The trial also included underrepresented subgroups (eg, patients undergoing mastectomy, and those with T3 tumors or extracapsular extension), enhancing generalizability.

Furthermore, we conducted subgroup analyses based on study type, surgical approach, and pathological grade. The results consistently showed no significant difference in both RCTs and observational studies, thereby enhancing the robustness of our findings. However, the fact that the findings were derived predominantly from observational studies constitutes the primary limitation of this meta-analysis. Despite rigorous quality assessment, residual confounding by unmeasured variables and selection bias (eg, systemic preference for ALND in healthier patients) cannot be excluded. These biases might artificially inflate the perceived benefits of ALND. Notably, in the TM subgroup, OS was significantly higher in the ALND group than in the SLNB alone group, suggesting that ALND might be necessary in patients who receiving TM with limited SLN metastasis. Nonetheless, this finding warrants careful consideration. First, patients undergoing TM are also likely to be systematically different from those receiving BCS in ways that affect outcomes, including more aggressive disease and lower rates of adjuvant radiotherapy, which cannot be adjusted for using aggregated data. Furthermore, ALND might not be the causal agent of the observed OS benefit, as it could instead reflect a more intensive overall treatment strategy, including more extensive surgery and intensified systemic therapy, compared with patients receiving SLNB alone, who might be selected for less intensive approaches. Such factors constitute important sources of selection bias and residual confounding that our analysis cannot fully address. Similarly, among EBC patients with pathological grade T3-4 tumors, ALND was associated with better OS, indicating a protective effect. This finding is plausible, as patients with T3-4 tumors are at a more advanced disease stage and ALND may offer a more comprehensive approach to improving survival outcomes in such cases. Nonetheless, this finding should be interpreted with caution given the relatively small sample size of T3-4 cases (n = 7812) compared with the total cohort (N = 146 407). That imbalance limits the statistical power for this specific subgroup and highlights the need for future studies focusing on patients with advanced-stage disease.

The impact of adjuvant radiotherapy on the observed outcomes in TM subgroups also warrants specific consideration. Guidelines from the US National Comprehensive Cancer Network recommended comprehensive regional nodal irradiation to include any undissected axilla at risk for patients with 1-3 positive nodes whether receiving BCS or TM⁷; however, compared with patients treated with BCS, those undergoing TM are less likely to receive radiotherapy; this differential use might influence survival independently of axillary surgery. A further uncertainty concerns radiotherapy practices in TM patients with 1-2 positive SLNs. In clinical practice, management in this setting varies widely. For example, some centers routinely deliver chest wall irradiation with regional nodal fields in lieu of ALND, whereas others restrict treatment to the chest wall alone without nodal coverage. In some cases, patients receive no radiotherapy. Those differences in radiotherapy approach can substantially affect locoregional control and survival, independent of axillary surgery. Consequently, variability in radiotherapy delivery across the included studies might have confounded the apparent OS advantage observed with ALND in patients staged T3-4 undergoing TM. Future prospective trials with standardized reporting and explicit adjustment for radiotherapy practices are required to disambiguate the various effects. Given the lack of information concerning radiotherapy, we were unable to conduct further subgroup analyses. Similarly, reporting of adjuvant systemic therapies (eg, CDK4/6 inhibitors) was often incomplete across the included studies. Such heterogeneity could represent a potential confounding factor, as these treatments might have influenced survival outcomes independently of the extent of axillary surgery, and could not be adjusted for in our analysis. Besides, subgroup analyses based on variables such as the number of SLNs could not be performed due to insufficient documentation of these data. An additional limitation is the considerable heterogeneity in how DFS and RFS were defined across the included studies, ranging from time to any recurrence (locoregional or distant) to composite endpoints including death from any cause (Table S1). This variability complicates precise interpretation and between-study comparability of outcomes.

Despite possible confounding effects of unmeasured factors on survival outcomes, our primary finding is that OS was equivalent between the SLNB alone group and the ALND group, which reinforces the feasibility of sparing ALND in appropriately selected EBC patients while advocating for a more tailored strategy in TM and advanced stage cases. Consideration of ALND omission is typically most appropriate for patients with characteristics suggesting favorable tumor biology and a lower risk of aggressive recurrence. These characteristics can include older age (particularly when life expectancy considerations are weighed against surgical morbidity), favorable tumor biology (eg, estrogen receptor and/or progesterone receptor positivity, low-grade histology, absence of lymphovascular invasion), and postmenopausal status, where endocrine therapy provides potent systemic control. In such low-risk TM cohorts, the potential morbidity benefits of avoiding ALND might outweigh the marginal survival benefit suggested in our overall TM subgroup analysis, especially when combined with comprehensive adjuvant systemic therapy. Ultimately, the decision to omit ALND in a patient undergoing TM must be individualized, incorporating tumor biology, staging details, comorbidities, patient preference, and multidisciplinary team discussion, further underscoring the need for personalized axillary management strategies in this population.

Conclusion

In conclusion, our meta-analysis indicates that omitting ALND in EBC patients with 1-2 positive SLNs yields survival outcomes comparable to those undergoing ALND, particularly among those treated with BCS. By contrast, subgroup analyses suggest a potential survival advantage from ALND in patients undergoing TM and in those with advanced tumor stage (T3-4), underscoring the need for individualized decision-making based on surgical approach and tumor biology. However, these findings should be interpreted cautiously, as residual confounding and selection bias may partly account for the observed benefit. While our results support the ongoing trend toward de-escalation of axillary surgery, they also emphasize the importance of caution when extending omission strategies to higher-risk subgroups. The inclusion of recent RCTs strengthens the robustness of our findings, but prospective validation is required before practice-changing recommendations can be made. Overall, our study highlights the feasibility of sparing ALND in appropriately selected patients while advocating a more tailored approach for TM and advanced-stage disease.

Author contributions

Jinyi Xie (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Validation, Visualization, Writing—original draft), Tong Wan (Resources, Software, Visualization), Jie Hao (Investigation, Resources, Validation), Siyue Zhang (Conceptualization, Resources), Haoyu Wang (Investigation, Project administration), Xiaochen Zhang (Investigation, Project administration), Pengfei Zhu (Formal analysis, Software, Visualization), Lijuan Wang (Conceptualization), Bing Chen (Conceptualization), Wenjing Zhao (Conceptualization), Qifeng Yang (Conceptualization, Methodology, Project administration, Supervision), and Ning Zhang (Conceptualization, Methodology, Project administration, Resources, Supervision, Visualization, Writing—review & editing)

Supplementary material

Supplementary material is available at The Oncologist online.

Funding

This work was supported by Mount Taishan Scholar Young Expert (No. tsqn202306345), Natural Science Foundation of Shandong Province (ZR2024MH002) and National Natural Science Foundation of China (No. 82373267).

Conflicts of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.