Lymphedema and Post-Operative Complications after Sentinel Lymph Node Biopsy versus Lymphadenectomy in Endometrial Carcinomas—A Systematic Review and Meta-Analysis

Background: Lymph node dissection (LND) is recommended as staging procedure in presumed low stage endometrial cancer. LND is associated with risk of lower-extremity lymphedema and post-operative complications. The sentinel lymph node (SLN) procedure has been shown to have high diagnostic accuracy, but its effects on complication risk has been little studied. This systematic review compares the risk of lower-extremity lymphedema and post-operative complications in SLN versus LND in patients with endometrial carcinoma. Methods: A systematic search was conducted in PubMed and Cochrane Library. Results: Seven retrospective and prospective studies (total n = 3046 patients) were included. Only three studies reported the odds ratio of lower-extremity lymphedema after SLN compared to LND, which was 0.05 (95% CI 0.01–0.37; p = 0.067), 0.07 (95% CI 0.00–1.21; p = 0.007) and 0.54 (95% CI 0.37–0.80; p = 0.002) in these studies. The pooled odds ratio of any post-operative complications after SLN versus LND was 0.52 (95% CI 0.36-0.73; I2 = 48%; p < 0.001). For severe post-operative complications the pooled odds ratio was 0.52 (95% CI 0.28–0.96; I2 = 0%; p = 0.04). Conclusions: There are strong indications that SLN results in a lower incidence of lower-extremity lymphedema and less often severe post-operative complications compared to LND. In spite of the paucity and heterogeneity of studies, direction of results was similar in all studies, supporting the aforementioned conclusion. These results support the increasing uptake of SLN procedures in endometrial cancer.


Introduction
Endometrial cancer (EC) is the second most common gynecologic malignancy and the fourth leading cause of cancer death in women worldwide, with 382,069 new cases and 89,929 deaths in 2018 [1]. EC is mostly diagnosed in postmenopausal women, with up to 20% of women diagnosed before menopause [2,3]. The majority of cases are diagnosed at early-stage due to early symptomatology, mostly abnormal uterine bleeding, in 90% of cases [4]. The primary treatment of endometrial cancer is surgical (total hysterectomy and bilateral salpingo-oophorectomy) to remove and classify the tumor [4,5]. Lymph node dissection (LND) is recommended in clinically presumed low stage disease as a staging procedure according to the International Federation of Gynecology and Obstetrics (FIGO) system [6][7][8]. Lymph node involvement is one of the most important prognostic factors in EC [9][10][11]. However, LND does not improve survival according to two performed randomized trials, and results in increased peri-and post-operative morbidity compared

Methodological Quality of Risk of Bias
Risk of bias was assessed by R.J.A.H., B.F.M.S., and H.M.J.W. using the Newcastle-Ottawa scale (NOS) and overall judgment using the Agency for Healthcare Research and Quality standards [42]. In case of disagreement between the authors, discrepancies were discussed after which agreement was reached. Risk of bias was scored low if the selection domain was scored low risk in minimal three subdomains, if statistical corrections were performed for the effect of confounders, including age, BMI, comorbidities, and adjuvant therapy (comparability domain), and if a minimum of two subdomains were scored low risk. Risk of bias was scored high if the selection domain was scored low risk in maximum one subdomain or if the study did not perform statistical corrections for confounders or if the outcome domain was scored low risk in a maximum of one subdomain.

Statistical Analyses
Review Manager Version 5.3 was used [34]. Heterogeneity was assessed using the I 2 statistic and visual interpretation of the forest plots. An I 2 of >75% was considered as considerable heterogeneity, 50-90% represented substantial heterogeneity, and 30-60% moderate heterogeneity [43]. Odds ratios (ORs) and p values were calculated to compare the effect of SLN versus LND on lower-extremity lymphedema. A p value of ≤0.05 was considered statistically significant. A meta-analysis was conducted on any and severe post-operative complications using forest plots and pooled ORs based on Mantel-Haenszel fixed-effects method, because of the small number of studies and low incidence rates [43].

Search Results
Through PubMed and Cochrane library searches, 125 potentially relevant articles were identified as detailed in Figure 1. After screening, 106 were excluded because they either did not investigate SLN, LND or complications, were reviews or case reports, or assessed different cancer types. Full-text assessment was performed for the 19 remaining articles; 12 articles were further excluded because of above-mentioned reasons ( Figure 1). Seven studies were ultimately included in the review, which are numbered from I to VII ( Table 2). The numbering is used to reference.

Statistical Analyses
Review Manager Version 5.3 was used [34]. Heterogeneity was assessed using the I 2 statistic and visual interpretation of the forest plots. An I 2 of > 75% was considered as considerable heterogeneity, 50-90% represented substantial heterogeneity, and 30-60% moderate heterogeneity [43]. Odds ratios (ORs) and p values were calculated to compare the effect of SLN versus LND on lower-extremity lymphedema. A p value of ≤ 0.05 was considered statistically significant. A meta-analysis was conducted on any and severe post-operative complications using forest plots and pooled ORs based on Mantel-Haenszel fixed-effects method, because of the small number of studies and low incidence rates [43].

Search Results
Through PubMed and Cochrane library searches, 125 potentially relevant articles were identified as detailed in Figure 1. After screening, 106 were excluded because they either did not investigate SLN, LND or complications, were reviews or case reports, or assessed different cancer types. Full-text assessment was performed for the 19 remaining articles; 12 articles were further excluded because of above-mentioned reasons ( Figure 1). Seven studies were ultimately included in the review, which are numbered from I to VII   The studies are numbered I to VII. Abbreviations: EC endometrial carcinoma, HYST hysterectomy, SALP salpingo-oophorectomy, SLN sentinel lymph node biopsy, LND lymph node dissection, IR infrarenal paraaortic nodal staging, PN pelvic nodal staging, IM intramesenteric paraaortic nodal staging, NR not reported * laparoscopy includes both robotic and non-robotic approach ** radiation therapy also includes chemoradiation therapy and brachytherapy *** confirmed after personal communication with the corresponding author.

Study Design
Five retrospective studies, one prospective and one combined retrospective/prospective cohort study were included and coded study I-VII as visualized in Table 1 [24,[35][36][37]39,41,44]. Studies I-VI were conducted in tertiary care hospitals, study VII was compiled from register data including multiple hospitals (American College of Surgeons National Surgical Quality Improvement Program). Study and patient characteristics are provided in Table 2. Where needed, authors were contacted for additional information.

Participants
From a total of 5600 patients included in the studies, 2554 patients were excluded as only hysterectomy was performed. 3046 patients were finally included in this systematic review, consisting of all patients who had undergone either SLN or LND in the indicated time period, as further specified in the inclusion criteria. Although patient recruitment intervals varied, all patients were included between 2006 and 2018. Only study IV did not specify the end date of patient recruitment for the SLN cohort [39]. Patients' median age varied from 61.0 to 79.5 years and differed significantly between the two arms in study II [37]. Median BMI varied from 26.8 kg/m 2 to 38.1 kg/m 2 . In studies II, IV and VII, BMI was significantly different between the SLN and LND cohort [37,39,41].

Intervention
Short-and long-term risks, defined as post-operative complications and lower-extremity lymphedema respectively, were separately assessed in the current review. Studies I-IV, VI and VII investigated the effect of SLN and LND on post-operative complication risk, whereas studies I, III and V specifically investigated the risk on lower-extremity lymphedema ( Table 2). Studies I, II, IV, V and VII also studied the effect of hysterectomy alone, but only the groups treated with SLN or LND were considered for this review [24,36,37,39,41]. The retrospective cohort studies (I, II, IV-VII) did not specify for the extent of lymph node dissection, whereas the prospective cohort study (III) did. For comparability reasons, data of the subgroups of the prospective study were pooled. Information concerning surgery was obtained from the medical records in all seven studies. Surgical approach varied within and between studies and included either robotic-assisted laparoscopy, laparotomy, or regular laparoscopy. In studies II, IV, VI and VII, only laparoscopy was performed (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was the main surgical approach, but laparotomy was performed in a small percentage (2-20%) [24,35,36]. Conversion rates only were reported by studies I-III and were all below 2.7%. Adjuvant therapy was reported by studies III-VI and included brachy-, chemo-, and/or external beam radiotherapy. The number of patients receiving adjuvant therapy varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. tinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. tinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. tinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. tinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. tinel nodes removed after SLN was 4 per patient. In the LND cohorts, the media of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. tinel nodes removed after SLN was 4 per patient. In the LND co of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Tab   Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each i Newcastle-Ottawa Scale. tinel nodes removed after SLN was 4 per patient. In the LND cohorts, th of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. tinel nodes removed after SLN w of nodes removed per patient v

Methodological Quality of Ris
The risk of bias has been ev tinel nodes rem of nodes remov

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Tab   Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each i Newcastle-Ottawa Scale. blue (n = 1) [24,35,37,39]. Use of patent blue res (80%) compared to indocyanine green (96%-97.9 tinel nodes removed after SLN was 4 per patien of nodes removed per patient varied from 15-3

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is i blue (n = 1) [24,35,37,39]. Use o (80%) compared to indocyanine tinel nodes removed after SLN w of nodes removed per patient v

Methodological Quality of Ris
The risk of bias has been ev Table 3. Risk of bias summary: review authors' judgements ab Newcastle-Ottawa Scale.

Methodolog
The risk o Table 3. Risk of bias summary: review autho Newcastle-Ottawa Scale.

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study using the Newcastle-Ottawa Scale. varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study using the Newcastle-Ottawa Scale. varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study using the Newcastle-Ottawa Scale. varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study using the Newcastle-Ottawa Scale. (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study using the Newcastle-Ottawa Scale. varied widely among studies (for example 1-31% for chemora to perform SLN included indocyanine green (n = 3), patent b blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a low (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. T tinel nodes removed after SLN was 4 per patient. In the LND co of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Tab   Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each i Newcastle-Ottawa Scale. varied widely among studies (for example 1-3 to perform SLN included indocyanine green (n blue (n = 1) [24,35,37,39]. Use of patent blue res (80%) compared to indocyanine green (96%-97.9 tinel nodes removed after SLN was 4 per patien of nodes removed per patient varied from 15-3

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is i varied widely among studies (f to perform SLN included indo blue (n = 1) [24,35,37,39]. Use o (80%) compared to indocyanine tinel nodes removed after SLN w of nodes removed per patient v

Methodological Quality of Ris
The risk of bias has been ev (80%) compared to indocyanine green (96%-97.9%) tinel nodes removed after SLN was 4 per patient. In of nodes removed per patient varied from 15-36 [3

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is ind

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Adjuvant therapy was reported by studies III-VI and included brachy-, chemo-, and/or external beam radiotherapy. The number of patients receiving adjuvant therapy varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Adjuvant therapy was reported by studies III-VI and included brachy-, chemo-, and/or external beam radiotherapy. The number of patients receiving adjuvant therapy varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Adjuvant therapy was reported by studies III-VI and included brachy and/or external beam radiotherapy. The number of patients receiving adjuvan varied widely among studies (for example 1-31% for chemoradiotherapy). Tra to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and m blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio o (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average num tinel nodes removed after SLN was 4 per patient. In the LND cohorts, the media of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Adjuvant therapy was reported by studies III-VI and i and/or external beam radiotherapy. The number of patients r varied widely among studies (for example 1-31% for chemora to perform SLN included indocyanine green (n = 3), patent b blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a low (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. T tinel nodes removed after SLN was 4 per patient. In the LND co of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Tab   Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each i Newcastle-Ottawa Scale. Adjuvant therapy was reported by studi and/or external beam radiotherapy. The numb varied widely among studies (for example 1-3 to perform SLN included indocyanine green (n blue (n = 1) [24,35,37,39]. Use of patent blue res (80%) compared to indocyanine green (96%-97.9 tinel nodes removed after SLN was 4 per patien of nodes removed per patient varied from 15-3

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is i Adjuvant therapy was reported by studies III-V and/or external beam radiotherapy. The number of pa varied widely among studies (for example 1-31% for c to perform SLN included indocyanine green (n = 3), p blue (n = 1) [24,35,37,39]. Use of patent blue resulted in (80%) compared to indocyanine green (96%-97.9%) [24,3 tinel nodes removed after SLN was 4 per patient. In the of nodes removed per patient varied from 15-36 [36,39,

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicate  was performed (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was the main surgical approach, but laparotomy was performed in a small percentage (2-20%) [24,35,36]. Conversion rates only were reported by studies I-III and were all below 2.7%. Adjuvant therapy was reported by studies III-VI and included brachy-, chemo-, and/or external beam radiotherapy. The number of patients receiving adjuvant therapy varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. of bias summary: review authors' judgements about each risk of bias item for each included study using the ttawa Scale.  (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was the main surgical approach, but laparotomy was performed in a small percentage (2-20%) [24,35,36]. Conversion rates only were reported by studies I-III and were all below 2.7%. Adjuvant therapy was reported by studies III-VI and included brachy-, chemo-, and/or external beam radiotherapy. The number of patients receiving adjuvant therapy varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. was performed (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was the main surgical approach, but laparotomy was performed in a small percentage (2-20%) [24,35,36]. Conversion rates only were reported by studies I-III and were all below 2.7%. Adjuvant therapy was reported by studies III-VI and included brachy-, chemo-, and/or external beam radiotherapy. The number of patients receiving adjuvant therapy varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. was performed (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was the main surgical approach, but laparotomy was performed in a small percentage (2-20%) [24,35,36]. Conversion rates only were reported by studies I-III and were all below 2.7%. Adjuvant therapy was reported by studies III-VI and included brachy-, chemo-, and/or external beam radiotherapy. The number of patients receiving adjuvant therapy varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study using the Newcastle-Ottawa Scale.

Adequacy of the Follow-up
Overall Judgement was performed (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was surgical approach, but laparotomy was performed in a small percentage [24,35,36]. Conversion rates only were reported by studies I-III and were all bel Adjuvant therapy was reported by studies III-VI and included brachy and/or external beam radiotherapy. The number of patients receiving adjuvan varied widely among studies (for example 1-31% for chemoradiotherapy). Tra to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and m blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio o (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average num tinel nodes removed after SLN was 4 per patient. In the LND cohorts, the media of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study usin Newcastle-Ottawa Scale.

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Tab   Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each i Newcastle-Ottawa Scale.

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is i  [24,35,36]. Conversion rates onl Adjuvant therapy was rep and/or external beam radiother varied widely among studies (f to perform SLN included indo blue (n = 1) [24,35,37,39]. Use o (80%) compared to indocyanine tinel nodes removed after SLN w of nodes removed per patient v

Methodological Quality of Ris
The risk of bias has been ev Table 3. Risk of bias summary: review authors' judgements ab Newcastle-Ottawa Scale.

Methodolog
The risk o Table 3. Risk of bias summary: review autho Newcastle-Ottawa Scale.

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study using the Newcastle-Ottawa Scale.

Adequacy of the Follow-up
Overall Judgement approach varied within and between studies and included either robotic-assisted laparoscopy, laparotomy, or regular laparoscopy. In studies II, IV, VI and VII, only laparoscopy was performed (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was the main surgical approach, but laparotomy was performed in a small percentage (2-20%) [24,35,36]. Conversion rates only were reported by studies I-III and were all below 2.7%. Adjuvant therapy was reported by studies III-VI and included brachy-, chemo-, and/or external beam radiotherapy. The number of patients receiving adjuvant therapy varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study using the Newcastle-Ottawa Scale.

Adequacy of the Follow-up
Overall Judgement approach varied within and between studies and included either robotic-assisted laparoscopy, laparotomy, or regular laparoscopy. In studies II, IV, VI and VII, only laparoscopy was performed (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was the main surgical approach, but laparotomy was performed in a small percentage (2-20%) [24,35,36]. Conversion rates only were reported by studies I-III and were all below 2.7%. Adjuvant therapy was reported by studies III-VI and included brachy-, chemo-, and/or external beam radiotherapy. The number of patients receiving adjuvant therapy varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study using the Newcastle-Ottawa Scale.

Adequacy of the Follow-up
Overall Judgement approach varied within and between studies and included either robotic-assiste copy, laparotomy, or regular laparoscopy. In studies II, IV, VI and VII, only lap was performed (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was surgical approach, but laparotomy was performed in a small percentage [24,35,36]. Conversion rates only were reported by studies I-III and were all bel Adjuvant therapy was reported by studies III-VI and included brachy and/or external beam radiotherapy. The number of patients receiving adjuvan varied widely among studies (for example 1-31% for chemoradiotherapy). Tra to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and m blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio o (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average num tinel nodes removed after SLN was 4 per patient. In the LND cohorts, the media of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. approach varied within and between studies and included either robotic-assisted lapar copy, laparotomy, or regular laparoscopy. In studies II, IV, VI and VII, only laparosco was performed (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was the m surgical approach, but laparotomy was performed in a small percentage (2-20 [24,35,36]. Conversion rates only were reported by studies I-III and were all below 2.7% Adjuvant therapy was reported by studies III-VI and included brachy-, chem and/or external beam radiotherapy. The number of patients receiving adjuvant thera varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers us to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methyle blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the S (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of s tinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median num of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study using the Newcastle-Ottawa Scale.

Adequacy of the Follow-up
Overall Judgement approach varied within and between studies an copy, laparotomy, or regular laparoscopy. In st was performed (Table 2) [37,39,41,44]. In studi surgical approach, but laparotomy was per [24,35,36]. Conversion rates only were reported Adjuvant therapy was reported by studi and/or external beam radiotherapy. The numb varied widely among studies (for example 1-3 to perform SLN included indocyanine green (n blue (n = 1) [24,35,37,39]. Use of patent blue res (80%) compared to indocyanine green (96%-97.9 tinel nodes removed after SLN was 4 per patien of nodes removed per patient varied from 15-3

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is i approach varied within and bet copy, laparotomy, or regular la was performed (Table 2) [37,39 surgical approach, but laparo [24,35,36]. Conversion rates onl Adjuvant therapy was rep and/or external beam radiother varied widely among studies (f to perform SLN included indo blue (n = 1) [24,35,37,39]. Use o (80%) compared to indocyanine tinel nodes removed after SLN w of nodes removed per patient v

Methodological Quality of Ris
The risk of bias has been ev Table 3. Risk of bias summary: review authors' judgements ab Newcastle-Ottawa Scale.

Methodolog
The risk o Table 3. Risk of bias summary: review autho Newcastle-Ottawa Scale. rability reasons, data of the subgroups of the prospective study were pooled. Information concerning surgery was obtained from the medical records in all seven studies. Surgical approach varied within and between studies and included either robotic-assisted laparoscopy, laparotomy, or regular laparoscopy. In studies II, IV, VI and VII, only laparoscopy was performed (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was the main surgical approach, but laparotomy was performed in a small percentage (2-20%) [24,35,36]. Conversion rates only were reported by studies I-III and were all below 2.7%. Adjuvant therapy was reported by studies III-VI and included brachy-, chemo-, and/or external beam radiotherapy. The number of patients receiving adjuvant therapy varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study using the Newcastle-Ottawa Scale. rability reasons, data of the subgroups of the prospective study were pooled. Information concerning surgery was obtained from the medical records in all seven studies. Surgical approach varied within and between studies and included either robotic-assisted laparoscopy, laparotomy, or regular laparoscopy. In studies II, IV, VI and VII, only laparoscopy was performed (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was the main surgical approach, but laparotomy was performed in a small percentage (2-20%) [24,35,36]. Conversion rates only were reported by studies I-III and were all below 2.7%. Adjuvant therapy was reported by studies III-VI and included brachy-, chemo-, and/or external beam radiotherapy. The number of patients receiving adjuvant therapy varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study using the Newcastle-Ottawa Scale. rability reasons, data of the subgroups of the prospective study were pooled. Information concerning surgery was obtained from the medical records in all seven studies. Surgical approach varied within and between studies and included either robotic-assisted laparoscopy, laparotomy, or regular laparoscopy. In studies II, IV, VI and VII, only laparoscopy was performed (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was the main surgical approach, but laparotomy was performed in a small percentage (2-20%) [24,35,36]. Conversion rates only were reported by studies I-III and were all below 2.7%. Adjuvant therapy was reported by studies III-VI and included brachy-, chemo-, and/or external beam radiotherapy. The number of patients receiving adjuvant therapy varied widely among studies (for example 1-31% for chemoradiotherapy). Tracers used to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and methylene blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio of the SLN (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average number of sentinel nodes removed after SLN was 4 per patient. In the LND cohorts, the median number of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study using the Newcastle-Ottawa Scale. rability reasons, data of the subgroups of the prospective study were pooled. In concerning surgery was obtained from the medical records in all seven studie approach varied within and between studies and included either robotic-assiste copy, laparotomy, or regular laparoscopy. In studies II, IV, VI and VII, only lap was performed (Table 2) [37,39,41,44]. In studies I, III and V, laparoscopy was surgical approach, but laparotomy was performed in a small percentage [24,35,36]. Conversion rates only were reported by studies I-III and were all bel Adjuvant therapy was reported by studies III-VI and included brachy and/or external beam radiotherapy. The number of patients receiving adjuvan varied widely among studies (for example 1-31% for chemoradiotherapy). Tra to perform SLN included indocyanine green (n = 3), patent blue (n = 1) and m blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a lower success ratio o (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. The average num tinel nodes removed after SLN was 4 per patient. In the LND cohorts, the media of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Table 3. rability reasons, data of the subgroups of the prospective study concerning surgery was obtained from the medical records in approach varied within and between studies and included eith copy, laparotomy, or regular laparoscopy. In studies II, IV, VI was performed (Table 2) [37,39,41,44]. In studies I, III and V, surgical approach, but laparotomy was performed in a s [24,35,36]. Conversion rates only were reported by studies I-III Adjuvant therapy was reported by studies III-VI and i and/or external beam radiotherapy. The number of patients r varied widely among studies (for example 1-31% for chemora to perform SLN included indocyanine green (n = 3), patent b blue (n = 1) [24,35,37,39]. Use of patent blue resulted in a low (80%) compared to indocyanine green (96%-97.9%) [24,35,39]. T tinel nodes removed after SLN was 4 per patient. In the LND co of nodes removed per patient varied from 15-36 [36,39,44,45].

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is indicated in Tab   Table 3. Risk of bias summary: review authors' judgements about each risk of bias item for each i Newcastle-Ottawa Scale. rability reasons, data of the subgroups of the pr concerning surgery was obtained from the med approach varied within and between studies an copy, laparotomy, or regular laparoscopy. In st was performed (Table 2) [37,39,41,44]. In studi surgical approach, but laparotomy was per [24,35,36]. Conversion rates only were reported Adjuvant therapy was reported by studi and/or external beam radiotherapy. The numb varied widely among studies (for example 1-3 to perform SLN included indocyanine green (n blue (n = 1) [24,35,37,39]. Use of patent blue res (80%) compared to indocyanine green (96%-97.9 tinel nodes removed after SLN was 4 per patien of nodes removed per patient varied from 15-3

Methodological Quality of Risk of Bias
The risk of bias has been evaluated and is i rability reasons, data of the subg concerning surgery was obtaine approach varied within and bet copy, laparotomy, or regular la was performed (Table 2) [37,39 surgical approach, but laparo [24,35,36]. Conversion rates onl Adjuvant therapy was rep and/or external beam radiother varied widely among studies (f to perform SLN included indo blue (n = 1) [24,35,37,39]. Use o (80%) compared to indocyanine tinel nodes removed after SLN w of nodes removed per patient v

Methodological Quality of Ris
The risk of bias has been ev rability reasons concerning sur approach varie copy, laparotom was performed surgical appro [24,35,36]. Con Adjuvant and/or externa varied widely to perform SLN blue (n = 1) [24 (80%) compare tinel nodes rem of nodes remov

Methodolog
The risk o  The studies are numbered I to VII. Abbreviations: SLN sentinel lymph node procedure, LND lymph node dissection.

Selection
In study V, only 49% of the patients that were contacted via email participated in the study [36]. Although power was still sufficient (88%), selection bias could not be ruled out, and was judged high risk of bias for study cohort representativeness. In study VII, patients were included from a large database comprising secondary and tertiary care hospitals [41]. The number of contributing hospitals was not stated. Although the SLN cohort was small (n = 144) compared to the LND cohort (n = 1089), numbers were still comparable to the other studies. Therefore, the study cohort representativeness was judged low risk of bias.
In all studies, participants for both arms were obtained from the same cohort (Table 3) [24,[35][36][37]39,41,44]. Ascertainment of the treatment received was obtained from the hospital records in all studies. Only in study V, absence of study outcome (lower-extremity lymphedema) at study start was specified [36]. Absence of post-operative complications prior to surgery was not reported in any study. Although e.g., preoperative existence of urine tract infections or low hemoglobin levels necessitating post-operative treatment cannot completely be ruled out, their likelihood was considered very low and 'absence of study outcome (post-operative complications) at study start' was left unscored in the Risk of Bias table (Table 3).

Comparability
Although all seven studies controlled for age and BMI for both arms, only studies II, V and VII controlled for comorbidities, such as diabetes and hypertension [36,37,41] ( Table 3). In study III, patients were matched for BMI and uterine size, but not for significant differences in baseline characteristics [35]. The other six studies applied various methods to correct for imbalanced baseline characteristics.

Outcome
In all but studies IV and VI, a specific tool was applied to assess post-operative complications (Table 3) [24,[35][36][37]41]. The follow-up periods for lower-extremity lymphedema and post-operative complications were deemed sufficient in six studies. In study I, lowerextremity lymphedema was assessed within 90 days, which was considered too short for all lower-extremity lymphedema to occur [24]. Due to lack of standardized in-hospital follow-up, adequacy of follow-up was judged high risk of bias in study IV [39].
Based on the above-mentioned observations, an overall judgement has been made regarding risk of bias in the included studies. Overall, study III was judged moderate risk of bias in view of significant differences in baseline characteristics and no evaluation of presence of lymphedema at the start of the study [35]. All other studies were judged low risk of bias.

Outcome Evaluation: Lower-Extremity Lymphedema
Studies I, III and V investigated the risk on lower-extremity lymphedema after SLN and LND [24,35,36]. Lower-extremity lymphedema was measured using a different grading system in all three studies (Table 1). In the SLN cohort, lower-extremity lymphedema incidences in studies I, III and V were 0%, 1.3% and 27.2%. In the LND cohort, the incidences were 10.1%, 18.1% and 40.9%. Incidences in study V were remarkable higher compared to the other two studies. The OR of lower-extremity lymphedema after SLN compared to LND ranged between studies: 0.05 (95% CI 0.006-0.365; p = 0.007) in study III, 0.07 (95% CI 0.00-0.21; p = 0.067) in study I and 0.54 (95% CI 0.365-0.799; p = 0.002) in study V. These results suggest a markedly decreased lower-extremity lymphedema incidence after SLN. In studies III and V, this was statistically significant (p = 0.003; p = 0.002) [35,36]. Results of all three studies are illustrated in Table 4. A meta-analysis was not performed due to the small number of studies available and substantial heterogeneity (I 2 = 73%) resulting from the different lower-extremity lymphedema assessments and subsequently reported incidences. Only study V ruled out lower-extremity lymphedema prior to the intervention [36]. Adjuvant radiotherapy was reported in studies III and V. In the LND cohorts, more radiotherapy was applied compared to the SLN cohort [35,36]. The incidence of lower-extremity lymphedema does not seem proportional to the frequency of applied radiation therapy. The studies are numbered I, III and V. Abbreviations: SLN sentinel lymph node biopsy, LND lymph node dissection, MSKCCSSEGS memorial sloan kettering cancer center surgical secondary events grading system, CTC common toxicity criteria, NR not reported, PRO patient-reported outcome. * radiation therapy also includes chemoradiation therapy and brachytherapy ** p < 0.05 *** not provided per surgery group † LND included pelvic and/or para-aortic LND ‡ LND included infrarenal and inframesenteric paraaortic and pelvic LND.
Although we were unable to compare edema in other localizations, study III did specify incidence of truncal edema and showed a difference in incidence in line with lower-extremity lymphedema (6/105 LND cohort vs. 1/73 SLN cohort).

Outcome Evaluation: Post-Operative Complications
Studies I-III, IV, VI and VII investigated the effect of SLN and LND on post-operative complications, as visualized in Figure 2. The pooled ORs of any, and of severe postoperative complications were 0.52 (95% CI 0.36-0.73; p < 0.001) and 0.52 (95% CI 0.28-0.96; p = 0.04) among the included studies respectively. Heterogeneity in the analysis on the presence of post-operative complications is moderate (I 2 = 48%) and mainly caused by study I [24]. Pooled analysis without this study showed reduced heterogeneity (I 2 = 0%) and pooled OR remained comparable (0.62; 95% CI 0.43-0.92). Follow-up length was adequate in all six studies and varied from 30 to 42 days. The extent of LND differed between studies: study II included only pelvic LND, studies I, III, IV and VI also added para-aortic LND [24,35,37,39,44]. Study VII did not specify the extent of lymphadenectomy [41].

Discussion
In this systematic review and meta-analysis, the short-and long-term risks, defined as post-operative complications and lower-extremity lymphedema respectively, after SLN and LND procedures in EC patients were assessed.
Both short-and long-term risks were significantly reduced after SLN compared to LND. The incidence of any as well as severe (grade III-V) post-operative complications was significantly lower after SLN compared to LND. This reduction is also clinically very relevant in view of the elderly, often comorbid, population affected by endometrial cancer. Although grading systems differed in all studies, they could, in a nearly 1:1 manner, be converted into the CDC system, thus ensuring comparability (Table 1,  Supplementary Tables S1-S3). Surgical mode of access was comparable across the study arms in the included studies and thus did not contribute to the lower complication rates after SLN [24,[35][36][37]39,41,44]. Decreased incidence of post-operative complications results in several benefits, as decreased bodily stress response and decreased surgical morbidity enable enhanced recovery after surgery (ERAS) and shorter hospital stay [46]. ERAS programs comprise a multimodal and multidisciplinary approach to optimize patient recovery after surgery following specific evidence-based recommendations including preoperative, intraoperative, and post-operative aspects [47]. As demonstrated in the included studies, the SLN results in reduced surgery time and blood loss [24,[35][36][37]39,41,44]. Also, procedures more often can be performed minimally invasive, ensuring less occurrence of post-operative ileus and earlier mobilization and discharge. As such, SLN further enhances ERAS adherence.
Besides short-term risks such as post-operative complications, importantly, lowerextremity lymphedema was clearly reduced after SLN compared to LND. The reduction in lower-extremity lymphedema may be even more important, because it can result in reduced QOL and increased morbidity in (former) cancer patients. However, with only three studies focusing on lower-extremity lymphedema and especially with the widely varying methodology used, the size of this effect, although studies are unanimous on the direction, needs to be interpreted with caution.

Discussion
In this systematic review and meta-analysis, the short-and long-term risks, defined as post-operative complications and lower-extremity lymphedema respectively, after SLN and LND procedures in EC patients were assessed.
Both short-and long-term risks were significantly reduced after SLN compared to LND. The incidence of any as well as severe (grade III-V) post-operative complications was significantly lower after SLN compared to LND. This reduction is also clinically very relevant in view of the elderly, often comorbid, population affected by endometrial cancer. Although grading systems differed in all studies, they could, in a nearly 1:1 manner, be converted into the CDC system, thus ensuring comparability (Table 1, Supplementary Tables S1-S3). Surgical mode of access was comparable across the study arms in the included studies and thus did not contribute to the lower complication rates after SLN [24,[35][36][37]39,41,44]. Decreased incidence of post-operative complications results in several benefits, as decreased bodily stress response and decreased surgical morbidity enable enhanced recovery after surgery (ERAS) and shorter hospital stay [46]. ERAS programs comprise a multimodal and multidisciplinary approach to optimize patient recovery after surgery following specific evidence-based recommendations including preoperative, intraoperative, and post-operative aspects [47]. As demonstrated in the included studies, the SLN results in reduced surgery time and blood loss [24,[35][36][37]39,41,44]. Also, procedures more often can be performed minimally invasive, ensuring less occurrence of post-operative ileus and earlier mobilization and discharge. As such, SLN further enhances ERAS adherence.
Besides short-term risks such as post-operative complications, importantly, lowerextremity lymphedema was clearly reduced after SLN compared to LND. The reduction in lower-extremity lymphedema may be even more important, because it can result in reduced QOL and increased morbidity in (former) cancer patients. However, with only three studies focusing on lower-extremity lymphedema and especially with the widely varying methodology used, the size of this effect, although studies are unanimous on the direction, needs to be interpreted with caution.
Risk factors for lower-extremity lymphedema after EC include amongst others (extensive) lymph node dissection, post-operative radiotherapy, and obesity [48][49][50]. As visualized in Table 4, there were differences in how often adjuvant radiotherapy was applied, with a higher frequency in the LND cohort. In study III, the LND cohort received even up to three times more frequently radiation therapy, which may have added to the lower-extremity lymphedema risk on top of the effect of the lymphadenectomy procedure [17,48,49,51,52]. However, the increased incidence of lower-extremity lymphedema does not seem proportional to the differences in applied radiation therapy between SLN and LND cohorts. We therefore argue that the increased incidence in the included studies is mainly explained by the surgical differences as radiation therapy may have contributed to, but cannot explain, the effect observed. It was noted that the extent of LND was not identical between and within studies, neither was the number of nodes removed. It is known that radicality of LND may affect the risk of lower-extremity lymphedema and post-operative complication rates, since para-aortic nodal dissection and removal of large number of nodes (n > 30) results in extended surgery times and increased complexity [9][10][11]49]. As we were unable to correct for this imbalance, LND extent remains a possible confounder, although reflecting regular clinical variation. The low incidence of lower-extremity lymphedema in the SLN groups fits with the assumption that a lower number of nodes removed better preserves lymphatic flow and thus avoids lower-extremity lymphedema [4,6]. Since a causal relationship between number of nodes removed and lower-extremity lymphedema incidence has not been investigated in the included studies, the aforementioned assumption could not be confirmed.
Another possible confounding factor includes BMI. Obesity is known to increase risk of lower-extremity lymphedema independent of surgery due to preferential adipose deposition in the lower extremities and impaired lymphatic transport against gravity in the lower extremities leading to subclinical edema [50]. Also, abnormal lymphatic valves, decreased interstitial fluid transport capacity, lymphatic inflammation, and changes in lymph angiogenesis caused by obesity can affect lymphatic function [50,53]. Although patients' BMI differed substantially between studies (26.8-38.1 kg/m 2 ), it did vary much less between the studies focusing on lower-extremity lymphedema development and especially between study cohorts within studies. Therefore, we feel it is unlikely that (differences in) BMI has affected the results and confounded the effects on lower-extremity lymphedema incidence. Ideally though, lower-extremity lymphedema should be ruled out before the intervention, as was done in study V [36]. However, if this had affected the results, it would have resulted in reduced incidence in lower-extremity lymphedema in this study, rather than the higher incidence that was observed, again suggesting that the observed differences are truly due to the surgical procedures.
Challenges in the current review differ for both outcomes. As depicted earlier, lower-extremity lymphedema was assessed using different methodologies, including grading based on medical records, by a specialized physiotherapist and by means of self-report [24,35,36]. The different methods can result in different incidence reporting, with self-reporting likely resulting in high(er) incidences of lower-extremity lymphedema. These differences were noted in the current review and also reported by others [16,17]. This underscores the need for a standardized grading system for studies comparing lowerextremity lymphedema. Follow-up lengths differed widely across the three studies and was deemed suboptimal in one, as lower-extremity lymphedema may start as late as 18 months or more after surgery and thus some cases may have gone unnoticed [16,54]. As such, also the long follow-up in study V may have contributed to the higher percentages of lower-extremity lymphedema in both cohorts [36]. Regarding lower-extremity lymphedema, the above-mentioned heterogeneity in assessment and follow up lengths should be noted as they likely have contributed to the substantial heterogeneity (I 2 = 73%), prohibiting meta-analysis. In spite of their different methodologies, all studies do agree that lower-extremity lymphedema occurs significantly less often after SLN compared to LND. Although further studies should be encouraged to test this association on prolonged follow-up, it is unlikely that this effect will disappear.
Considering post-operative complications, moderate study heterogeneity (I 2 = 48%) was mainly caused by one study I [24]. Differences in baseline characteristics, including adjuvant therapy and number of lymph nodes removed, are unlikely to have significantly contributed as they would have contributed similarly to heterogeneity in the analysis of severe complications, which was not observed. There is however a more pronounced favorable effect of the SLN on 'all post-operative complications' compared to the other studies whereas this is much more in line with the other studies when considering 'severe complications' only. This may thus explain (part of) the observed heterogeneity. Length of follow-up in all six studies studying post-operative complications was 30-42 days and deemed sufficient to identify all complications [24,35,37,39,41,44]. Finally, the SLN procedure in EC was still in its infancy in the inclusion period of all studies. This implies that results for the SLN groups of all studies may further improve, now that more experience in performing SLN has been gained.
As mentioned previously, diagnostic accuracy of the SLN has been proven in literature both prospectively and retrospectively [26][27][28][29][30][31]. The current review is the first to provide a systematic overview of the available literature on lower-extremity lymphedema and post-operative complications after SLN and LND in EC, with thorough quality assessment. Further foundation can be given to the application of the SLN with reported reduced complications. It is considered unlikely that a large randomized controlled trial will be conducted comparing the complications after SLN and LND in view of the high uptake and positive oncological outcomes of the SLN procedure. Therefore, it is needed to consider the evidence from the available retrospective studies to be the best quality evidence we have.
In conclusion, this systematic review further supports the application of SLN, showing a significant decrease in the known complications associated with the LND procedure, including post-operative complications and lower-extremity lymphedema. However, as short-and long-term complications are not non-existent after SLN; the decision to perform any type of lymph node dissections remains to be taken on clear clinicalgrounds.