The Feasibility and Oncological Safety of Axillary Reverse Mapping in Patients with Breast Cancer: A Systematic Review and Meta-Analysis of Prospective Studies

Objective The axillary reverse mapping (ARM) technique has recently been developed to prevent lymphedema by preserving the arm lymphatic drainage during sentinel lymph node biopsy (SLNB) or axillary lymph node dissection (ALND) procedures. The objective of this systematic review and meta-analysis was to evaluate the feasibility and oncological safety of ARM. Methods We searched Medline, Embase, Web of science, Scopus, and the Cochrane Library for relevant prospective studies. The identification rate of ARM nodes, the crossover rate of SLN-ARM nodes, the proportion of metastatic ARM nodes, and the incidence of complications were pooled into meta-analyses by the random-effects model. Results A total of 24 prospective studies were included into meta-analyses, of which 11 studies reported ARM during SLNB, and 18 studies reported ARM during SLNB. The overall identification rate of ARM nodes was 38.2% (95% CI 32.9%-43.8%) during SLNB and 82.8% (78.0%-86.6%) during ALND, respectively. The crossover rate of SLN-ARM nodes was 19.6% (95% CI 14.4%-26.1%). The metastatic rate of ARM nodes was 16.9% (95% CI 14.2%-20.1%). The pooled incidence of lymphedema was 4.1% (95% CI 2.9–5.9%) for patients undergoing ARM procedure. Conclusions The ARM procedure was feasible during ALND. Nevertheless, it was restricted by low identification rate of ARM nodes during SLNB. ARM was beneficial for preventing lymphedema. However, this technique should be performed with caution given the possibility of crossover SLN-ARM nodes and metastatic ARM nodes. ARM appeared to be unsuitable for patients with clinically positive breast cancer due to oncological safety concern.


Introduction
Breast cancer is the most common malignancy among women in the United States and is the second leading cause of cancer-related death [1,2]. The status of axillary lymph nodes is one of the most important prognostic factors for patients with breast cancer, and can directly guide adjuvant therapy choices [1]. Currently, axillary lymph node dissection (ALND) represents the standard treatment for patients with metastatic axillary lymph nodes [1]. However, ALND always carries an unacceptable high incidence of lymphedema, ranging from 6% to 57% [3]. For patients with clinically negative axilla, Sentinel lymph nodes biopsy (SLNB) is recommended for the surgical staging, with significantly decreased surgical complications compared with ALND [1,4]. Nevertheless, the incidence of lymphedema remains significant, ranging from 0 to 13% [5].
Since 2007, axillary reverse mapping (ARM) has been developed as a novel surgical approach to distinguish the lymphatic drainage pattern of the upper limb from that of the breast [6,7]. It could be performed accompanying with ALND or SLNB procedures. The successful identification and preservation of ARM nodes/lymphatics are prerequisites for ARM feasibility. However, the identification rates of lymphatics or nodes during ARM varied between previous studies [8,9]. As the converged ARM-SLN nodes were unlikely to be preserved during sentinel node biopsy, their proportion was also closely related to ARM feasibility [10,11]. In addition, the preserved ARM nodes should not contain metastasis. The metastatic rate of ARM nodes during ALND could reflect the oncological safety of ARM. So far, no published guideline has appraised the role of ARM in breast cancer [1,12]. Therefore, we carried out this systematic review and meta-analysis, aiming to assess the feasibility and oncological safety of ARM during SLNB or ALND procedures.

Search Strategy
This systematic review and meta-analysis was conducted according to guidelines from the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [13]. We searched Medline (Ovid format), Embase, Web of science, Scopus, and Cochrane Library were searched from their inception until September 2015. We used the following Mesh Terms or key words in the search: "axillary reverse mapping", "lymphatic arm drainage", and "breast cancer". The search strategy was shown in S1 Table. The language was restricted to English. The references included articles were manually searched for additional relevant records. resect ARM nodes, and a low rate of converged SLN-ARM nodes. Thus, the primary outcomes were defined as the overall identification rate of lymph nodes and/or lymphatics, the rate of positive resected ARM nodes, and the rate of converged SLN-ARM nodes. The second outcome was the incidence of lymphedema during follow-up. The occurrences of lymphedema measured within 3 months of ARM procedure were excluded because arm-related changes during this timeframe potentially represented acute surgery-related response [14]. In addition, we tried to assess the influences of preoperative neoadjuvant chemotherapy and axillary metastasis on the metastasis rate of ARM nodes. The staging of breast cancer was defined according to the 2015 NCCN guideline [1]. We compared the results between pN 0-1 and pN 2-3 stages of breast cancer.

Data Extraction and Quality Assessment
Two authors (CH and BY) independently extracted all data, with discrepancies resolved by consensus or discussion with a third investigator (WSZ). The following data were extracted: author, publication year, location, number of patients, mean/median age, mapping material for ARM, number of ARM procedures during SLNB and/or ALND, outcomes, and study period. Data on ARM during SLNB or ALND procedures were extracted separately. The quality of included studies was assessed by using the Agency for Healthcare Research and Quality (AHRQ) checklist [15].

Statistical Analysis
The event rates for outcomes of interest were combined to determine the pooled rates and accompanying 95% confidence intervals (CIs). The Comprehensive Meta-Analysis statistical package (CMA Version 2.2, Biostat, Englewood, NJ) was used to conduct all meta-analyses by employing random-effects models. The heterogeneity across the results of included studies was assessed by using I 2 statistics and the χ 2 -test. Low, moderate and high heterogeneity was set at I 2 values of 25%, 50% and 75%, respectively [16]. We did separate analyses for ARM-SLNB procedures and ARM-ALND procedures. Subgroup analyses were performed according to different ARM mapping materials (blue dye, isotope, or fluorescence) and different locations (Asia, Europe, or North America). Meta-regression analyses (unrestricted maximum likelihood) were performed to determine whether the pooled rates were modulated by sample sizes. The publication bias was inspected visually by the funnel plots and statistically by the Egger's test [17,18]. A P value of less than 0.1 was considered statistically significant when assessing heterogeneity or publication bias. In other ways, a P value of 0.05 was regarded as significance level.

Study Selection
Our initial searches identified 95 publications, including 43 records in Medline, 53 records in Embase, 2 records in the Cochrane Library, 51 records in the Web of Science, and 55 records in the Scopus. After removing 142 duplicates, we screened 62 publications by titles and abstracts. Thirty-two records were eligible for full-text assessment. Further, one trial protocol [19], one postmortem study [20], and one case report were excluded [21]. One study were discarded because the outcomes of interest were not reported [22]. Twenty-eight studies were included into qualitative synthesis. Subsequently, one retrospective study and three studies with overlapping population were discarded [3,10,23,24]. The remaining 24 publications were pooled into meta-analysis, involving 2709 patients [5-9, 11, 25-42]. (Fig 1)

Discussion
The results of our meta-analysis demonstrated that the pooled identification rate of ARM lymphatics or nodes was 82.8% during ALND, which was much higher than that during SLNB (38.2%). This discrepancy was supported by the suggestion that the majority of lymphatics draining the upper extremity may be located deeper than the SLNs [9]. Most studies used blue dye alone as mapping material for ARM identification. Compared with blue dye alone, the fluorescent imaging and combined use of blue dye with radioisotope seemed to be more sensitive for detecting ARM lymphatic systems during ALND procedures. In addition, the detection failure may be attributed to the existence of learning curve, the insufficient time interval elapsing from blue dye injection to initiation of surgery, the potential location of nodes outside the  axilla area, or the extensive metastasis of ARM nodes obstructing the lymphatic drainages [6,25,34].
The crossover rate of SLN-ARM nodes by using blue dye injection was 7.8%. It would be difficult to preserve the converged SLN-ARM nodes during SLNB. Thus, the ability of ARM to prevent lymphedema may be limited after removing these converged nodes. With respect to the pathologic status of resected ARM nodes, the overall metastatic rate was 16.9%. This may be explained by that the ARM nodes were located in the central nodal group for breast lymphatic drainage, which was also supported by the existence of converged SLN-ARM nodes [43]. Additionally, the numerous interconnections shared by lymphatic drainages of the arm and the breast may contribute to ARM metastasis [20]. Preoperative neoadjuvant chemotherapy did not significantly decrease the risk of ARM-nodes metastasis. However, patients with extensive axillary metastasis carried an increased risk of metastasis to the ARM nodes. Therefore, ARM may be contraindicated for patients with clinically positive breast cancer.
The incidence of lymphedema post ARM procedures was 4.1% during follow-up. Recently, a meta-analysis of 72 studies showed that the pooled incidence of arm lymphedema was 19.9% in ALND, and was 5.6% in SLNB [14]. As only 1 ARM study was included in this meta-analysis [30], the pooled results represented the overall incidence of lymphedema in non-ARM procedures. Therefore, it appeared that ARM was effective in preventing lymphedema. A higher incidence of lymphedema was revealed for ARMs during ALND procedures compared with ARMs during SLNB procedures. This discrepancy may be attributed to that the majority of lymphatics draining the upper extremity were located deep to the plane of SLNs, thus causing more disruptions of the lymphatic during ALND [44]. In one study comparing the ARM-nodes preservation group with the ARM-nodes resection group, patients with preserved ARM nodes experienced significantly decreased incidence of lymphedema [33]. In accordance, several studies demonstrated that lymphedema mostly occurred accompanying with the resection of ARM lymphatic nodes or lymphatics [5,7,24,30,31,34,42].
We were aware of the limitations regarding this meta-analysis. Except for 1 RCT comparing the incidence of lymphedema between ARM and non-ARM procedures [42], most publications were single-arm studies of ARM procedures, which precluded the availability of direct comparison effect estimates. Thus, the efficacy of ARM in preventing lymphedema could not be thoroughly evaluated by controlled groups. We could only try to compare it with previous metaanalysis results. The efficacy outcome did not serve as one of our main objectives. Although meta-analysis of RCTs provided the best evidence, our pooled results from non-randomized studies were of clinical significance to inform the design of subsequent trials that evaluate the long-term efficacy of ARM in preventing lymphedema [45]. Additionally, the clinical features, such as ages, breast cancer stages, and preoperative NAC, were heterogeneous among included studies. For example, several studies clearly excluded patients who had received NAC [5,9,27,31,42]. Besides, the definition, measurement and follow-up duration of lymphedema were inconsistent across included studies. Some clinical variables may be associated with the risk of lymphedema, such as body mass index and the receipt of radiation therapy or chemotherapy, which were not adjusted or balanced in most studies [14]. Further well-designed RCTs were warranted to provide more convincing evidence.
We noted that a review has described ARM in depth recently [43]. In comparison, the distinct features and strengths of our study lied in the following aspects. Our study represented the first meta-analysis regarding ARM procedures, which included a large number of prospective studies through comprehensive literature search. The rates relating to the feasibility and oncological safety of ARM procedures were statistically summarized, with separate exploration for SLNB and ALND. The impact of NAC and axillary status on the metastasis of ARM nodes were firstly systematically analyzed. Besides, the included studies were critically appraised by quality tool, displaying moderate to high methodological qualities. The heterogeneity was carefully explored by subgroup analyses and meta-regression analyses. No publication bias was detected for included studies.

Conclusion
The ARM technique was feasible for patients undergoing ALND, but was limited by unsatisfying identification rate of ARM nodes for patients undergoing SLNB. ARM appeared to be beneficial for decreasing the occurrence of arm lymphedema. However, clinicians should prudently perform this procedure in light of the possibility of crossover SLN-ARM nodes or metastatic ARM nodes. Patients with clinically positive breast cancer may be unsuitable for ARM due to potentially increased risk of ARM-nodes metastasis.