Axillary lymph node dissection (ALND) is notorious for its complications, including surgical site infections, neuropathic pain, lymphedema, and seroma formation.1 The introduction of sentinel lymph node biopsies (SLNB) resulted in fewer lymph node dissections. The AMAROS trial showed that axillary radiotherapy after a positive SLNB was equal to ALND for patients with breast cancer regarding survival, and the MSLT-II trial showed that immediate complete lymph node dissections did not improve melanoma-specific survival.2,3,4,5 However, ALND remains an important treatment modality for some patients. Therefore, reducing surgical morbidity after these procedures remains pivotal.

Seroma formation is one of the most frequent complications after ALND, with an incidence of up to 90%.1,6 As addressed by van Bemmel et al. in 2011, many studies address different techniques to prevent seroma formation, but no single intervention proved to be completely successful.6 In the past decade, more research has been performed regarding this topic.

One major disadvantage in the current literature is that the majority of studies do not differentiate between ALND as a standalone procedure or ALND in combination with mastectomy.7,8 The extent of the dissection is a proven risk factor for developing seroma, and therefore, patients undergoing ALND with mastectomy are more prone to develop seroma than after ALND as a standalone procedure or in combination with breast-conserving surgery.9 Results of these trials suggest that flap fixation after mastectomy alone seems to be the most promising solution for seroma in these procedures; this could possibly be effective for ALND as a standalone procedure as well.7,9,10 However, for ALND, flap fixation of the axillary dead space could cause other difficulties, and as such may compromise mobility of the ipsilateral arm.11,12

The aim of this systematic review is to identify the most effective measures for the prevention of seroma formation after a standalone ALND, using a network meta-analysis (NMA) to indirectly compare different techniques. The ultimate goal is to create a surgical guideline for surgeons performing ALND.

Patients and Methods

Design

This systematic review was performed in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement,13 including the network meta-analysis extension.14 Prior to initiation, a local protocol for study selection and the NMA was drafted.

Eligibility Criteria

Articles were eligible for inclusion if: (1) it was a comparative study, (2) it compared any measure meant to prevent seroma formation after ALND, and (3) it had overall seroma or CSS incidence as outcome measure. Articles were excluded if ALND was combined with a mastectomy (modified radical mastectomy, MRM) only, or if there was no distinction in ALND procedures with or without mastectomy. If there was a clear distinction between ALND with and without mastectomy, the article was included and only the standalone ALND group was included. In addition, articles were excluded if older than 25 years (published before April 1998) or not written in English.

Search and Selection

The search was performed by two authors (MS and MB). The initial search for this systematic review was developed for PubMed and adjusted for EMBASE and Cochrane Library databases. The last search was performed on 24 April 2023. A full, detailed description of all searches is presented in Supplementary Material A–C. Search outcomes were imported in Rayyan (Rayyan Systems Inc., Cambridge, USA, http://rayyan.qcri.org), in which duplicates were removed. Titles and abstracts were first assessed, after which full texts were screened for eligibility. Study selection was performed in a blinded, standardized manner by two authors (MS and MB).

Data Collection and Outcome Measures

Data collection was performed by one author (MS) and cross-checked for validity by a second author (LA). General study information was collected, including year of publication, number of participants, and study design. Patient characteristics gathered included age, BMI, neoadjuvant treatment, and concomitant surgery. The primary outcome was CSS, which is defined as seroma requiring intervention due to infection, patient’s discomfort, or delayed wound healing. Secondary outcomes collected were overall seroma, surgical site infections (SSI), and hematoma.

Bias Assessment for Individual Studies

Bias assessment was performed using the Cochrane tool for Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I) and the revised Cochrane Risk of Bias tool for randomized trials (RoB2).15,16 ROBINS-I assesses observational trials and categorizes in low, intermediate, serious, or critical risk of bias. RoB2 assesses randomized trials and categorizes in low risk, some concerns, or high risk of bias. The risk of bias was visualized for all trials using the Robvis tool.17

Statistical Analysis

The NMA was performed using the MetaXL software (MetaXL, Version 5.3, EpiGear International Pty Ltd). Odds ratios (OR) and 95% confidence intervals (CI) were used to estimate effects for categorical outcomes, which are presented in a league table. ORs are presented in a way that a value > 1 indicates a preference for the treatment described in the column of the league table. Transitivity was assessed by evaluating inconsistency across the network using the weighted pooled H-statistic. A value < 3 indicated minimal presence of inconsistency.

Results

Search Results

A detailed overview of the search results is depicted in Fig. 1. In total, the search resulted in 2071 articles. After removal of duplicates, 1762 articles were screened, 85 were analyzed on the basis of full text, and 19 articles were eligible for inclusion. These results are summarized in Table 1.

Fig. 1
figure 1

PRISMA flow diagram of the article selection procedure

Full size image
Table 1 Evidence table of all included studies, with study and patients’ characteristics and outcome measures
Full size table

Methods

Nine of the included studies describe a randomized controlled trial (RCT).18,19,20,21,22,23,24,25,26 In addition, six prospective cohorts,27,28,29,30,31,32 one retrospective cohort,33 and three combined retrospective and prospective cohorts were included.34,35,36 Inclusion periods were all between 1995 and 2021.

Participants

In total, the 19 articles included 1962 patients. Follow-up ranged from 3 weeks to 1 year postoperatively. The mean age ranged from 48 years to 65 years and BMI from 23 kg/m2 to 30 kg/m2. Only six articles reported the number of patients undergoing neoadjuvant chemotherapy (NAC). This was between 14% and 43% of the patients included.

Described Techniques

A variation of techniques is described by the different studies. Nine studies report on different wound closure techniques, including axillary padding, application of tissue glue, or use of a sealant patch.19,21,23,24,25,26,28,30,35 Six studies show results of different drain policies (no drain, 24-h drainage, drainage until < 50 ml per 24 h, or drainage until < 50 ml per 24 h for two consecutive days and progressive drain removal). In the drainage until < 50 ml per 24 h group, thresholds varied between 30 ml and 50 ml per day. In the drainage until < 50 ml per 24 h for two consecutive days group, only 50 ml thresholds were described. Progressive drain removal is described as removing the drain 1–2 cm a day, as opposed to immediate drain removal.18,20,29,31,32,34 The remaining four studies report on the use of different surgical energy devices in the prevention of seroma formation.22,27,33, 36 The network structure is presented in Fig. 2. The weighted pooled H-statistic of 1.021 indicated minimal presence of inconsistency.

Fig. 2
figure 2

Network structure of the direct comparisons. PAD Padding, ED Energy devices, D24 24 hour drainage, ND No drain, PTG Padding + tissue glue, D50 Drainage until <50ml/24h, PRO Progressive drain removal, D50-2 Drainage until <50ml/24h for two consecutive days, TG Tissue glue, PAT Patches

Full size image

Risk of Bias Assessments

All included articles were assessed for their individual bias. A summary of these results is shown in Fig. 3 for randomized trials and in Fig. 4 for non-randomized trials. Only one trial was rated as low risk of bias. All other articles were considered to have some concerns/a moderate risk of bias (14 articles) or even high/serious risk of bias (4 articles). In most cases, this elevated risk was caused by lack of blinding in the randomized trials. Blinding was one of the main concerns in the non-randomized trials as well. In addition, confounders could also induce bias in the non-randomized trials. For bias in selection of reported results, only one of the articles referred to a preregistered protocol or statistical analysis plan. Therefore, these are indicated as moderate risks as well.

Fig. 3
figure 3

Bias assessment of the individual randomized trials using the RoB2-tool

Full size image
Fig. 4
figure 4

Bias assessment of the individual non-randomized trials using the ROBINS-I-tool

Full size image

Clinically Significant Seroma

The results of the NMA for clinically significant seroma are presented in Table 2. As previously described, the three main domains of interventions are drain policies, surgical energy devices, and wound closure techniques. In general, the only two techniques with statistically significant differences in OR for CSS are progressive drain removal and drain removal if drain output is below 50 ml per 24 h for two consecutive days. Regarding different drain policies, no drains and drainage for 24 h is significantly worse than both long-term drainage options. When compared with drainage until output is < 50 ml per 24 h, progressive drain removal (OR 0.13, 95% CI 0.03–0.55) and drainage until output is < 50 ml per 24 h for two consecutive days (OR 0.15, 95% CI 0.06–0.37) are significantly better. Comparing progressive drain removal and drain removal until output is < 50 per 24 h for two consecutive days did not show a statistically significant difference (OR 1.11, 95% CI 0.36–3.36).

Table 2 League table presenting the combined direct and indirect comparisons between different surgical measures for preventing CSS after ALND
Full size table

Energy devices are significantly worse than removing drains when output is < 50 ml per 24 h for 2 days (OR 4.09, 95% CI 1.65–10.14). All wound closure techniques are significantly worse than removing drains when output is below < 50 per 24 h for 2 days (patches: OR 5.05, 95% CI 1.10–23.10; tissue glue: OR 14.04, 95% CI 1.73–114.26; axillary padding: OR 4.55, 95% CI 1.05–20.49; and padding + tissue glue: OR 7.60, 95% CI 2.36–24.48). Progressive drain removal is also significantly better than tissue glue (OR 0.08, 95% CI 0.01–0.85) and axillary padding + tissue glue (OR 0.15, 95% CI 0.03–0.73).

Surgical Site Infections

Secondarily, all techniques were compared regarding SSI incidence, as described in Table 3. None of the techniques show superiority in this outcome measure. Specifically, shorter drainage periods (24-h drainage) or no drainage periods did not show reduced OR for SSI when compared with more conservative drainage policies.

Table 3 League table presenting the combined direct and indirect comparisons between different surgical measures for preventing SSI after ALND
Full size table

Discussion

In this systematic review, current literature is used to identify the most effective measures for preventing seroma formation after standalone ALND. The necessity for developing a surgical guideline is highlighted by the incidence of CSS in this article, which varies between 3% and 67%. Postoperative drainage remains the most effective way to reduce seroma-related complications after ALND. However, it is striking that drain policies vary widely across different studies. In this NMA, no drains or drains for 24 h are inferior to volume-controlled drainage. This is in accordance with a previously conducted systematic review by Droeser et al.37

One of the main concerns of long-term drainage was the incidence of surgical site infections.34 The results show no significant increase in infections in patients undergoing drainage for a longer duration. In addition, none of the other techniques seem to reduce the incidence of SSI after ALND. Therefore, with proper drain care, infections should not be one of the limiting factors when implementing longer drainage periods.

Remarkably, within the volume-controlled drainage group, multiple policies are described as well. Some surgeons remove drains when drain output is less than 50 ml in one day, whereas others will wait for drain output < 50 ml for two consecutive days. The latter is prone to result in longer drainage periods after surgery. However, results of this study show significantly less seroma formation after ALND when adhering to a more conservative drainage policy. These findings are supported by Shima et al., who concluded that early drain removal in ALND combined with mastectomy or breast-conserving surgery did not result in fewer cases with seroma.38

One of the studies described two different techniques to remove drains: progressively with 1–2 cm a day or immediately when drain output was low for two consecutive days. These techniques did not show a significant difference in a direct comparison by Frich et al.,20 which indicates complete drain removal is a safe technique.

For mastectomy and MRM, flap fixation reduces the incidence of CSS significantly.9,10,39 This technique does not seem to be the solution for standalone ALND patients. Axillary padding was introduced by Classe et al. to omit drain use after ALND.40 However, as indicated in the results of the NMA, axillary padding as well as other techniques to reduce dead space after surgery did not reduce CSS incidence to the same extent as it does after mastectomy. One thing worth mentioning is that none of the axillary padding groups included drain placement postoperatively. The combination of drain placement and axillary padding might be an option to reduce CSS after standalone ALND.

Limitations need to be discussed when interpreting the results of the current study. As shown in the bias assessment, there is a considerable risk of bias in almost all included studies. Among others, this is caused by the lack of blinding, which in most cases is impossible due to the nature of the intervention. In addition, there are few articles describing the same direct comparison. This is probably the main reason for the wide confidence intervals in the league table. Several results are based on indirect comparisons only. Lastly, the definition of CSS varies among studies. It is essential to describe a widely accepted definition, including the criteria used for performing seroma aspiration. It is becoming standard practice to only perform seroma aspirations when seromas are accompanied by signs of infection, delayed wound healing, or pain.

Conclusions

Current literature describes many different techniques to reduce seroma formation after axillary lymph node dissection in patients with breast cancer or melanoma. The only technique that consistently reduces the incidence of seroma formation, without increasing surgical site infection, is axillary drainage. The results of this systematic review support drain removal when drain output is < 50 ml per 24 h for two consecutive days irrespective of duration. Since drainage policies vary widely, an evidence-based guideline is needed.