Objective To identify factors that affect successful adaptation of sentinel lymph node mapping and those that lead to unintended adipose-only sentinel lymph node identification.
Methods Surgical and pathological data were prospectively collected on patients with endometrial cancer who underwent sentinel lymph node mapping with indocyanine green with or without pelvic and/or para-aortic lymph node dissection between November 2013 and April 2017. All mapping cases were performed with the robotic system. Adipose-only specimens were defined as a sentinel lymph node without a pathologically identified lymph node after ultrastaging.
Results A total of 202 patients were included: 83% had endometrioid pathology, 12% serous, 3% carcinosarcoma, and 2% clear cell, with mixed pathology noted in 2%. The bilateral sentinel lymph node detection rate was 66%, and the rate of mapping at least a unilateral sentinel lymph node was 86%. Neither the bilateral nor the unilateral sentinel lymph node mapping rate changed with increased surgeon experience. The rate of adipose-only sentinel lymph node identification was more frequent when comparing the first 10 cases (37%), cases 11 – 30 (28%), and > 30 cases (9%) (P = 0.006). Body mass index > 30 kg/m2, uterine fibroids, The International Federation of Gynecology and Obstetrics (FIGO) grade, and histology were not found to have a statistically significant impact on either sentinel lymph node identification or adipose-only sentinel lymph node identification. Adipose-only sentinel lymph nodes were more likely with increased time from cervical injection to identification of the sentinel lymph node in the right hemipelvis. The median range was 28 min (14–73) for true sentinel lymph node identification vs 33 min (23–74) for adipose-only sentinel lymph node identification (P = 0.02).
Conclusion Patient and surgeon factors did not impact the identification of sentinel lymph nodes over time. Adipose-only sentinel lymph nodes were more frequently identified in the initial cases and represent a potential complication to adapting sentinel lymph node biopsy without lymphadenectomy. The increase in adipose-only sentinel lymph node identification that was associated with time from cervical injection may represent delayed or disrupted uptake of indocyanine green.
Statistics from Altmetric.com
Adipose-only sentinel lymph node dissection is a possible limitation to sentinel lymph node mapping with indocyanine green.
There is a learning curve associated with identification of adipose-only sentinel lymph nodes, and the number harvested decreases over time.
Increased time from cervical injection to opening of the retroperitoneal spaces and harvesting an sentinel lymph node increases the chance of AO sentinel lymph node dissection.
As the most common gynecologic malignancy in the United States and the sixth-most common malignancy affecting women worldwide, endometrial cancer places a significant burden on healthcare.1 Traditionally, endometrial cancer is treated with hysterectomy, bilateral salpingo-oophorectomy, and lymphadenectomy. However, two randomized controlled trials have questioned the value of lymphadenectomy owing to complications including lymphedema higher in the arm randomized to lymphadenectomy.2–4
Obesity is a frequent comorbidity in patients with endometrial cancer and compared with open surgery and 'straight-stick' laparoscopy, the robotic platform affords decreased morbidity and a decreased rate of conversion to laparotomy.5 The near-infrared imaging technology of the da Vinci Si (Intuitive Surgical Inc., Sunnyvale, CA) is an imaging modality used with indocyanine green (Akorn Inc., Lake Forest, IL) to identify node mapping, and indocyanine green has proven to be an effective tracer for sentinel lymph node mapping when compared with blue dye or technetium.6–13 Recent data from a randomized clinical trial suggests the superiority of indocyanine green over blue dye in sentinel lymph node identification.14
During attempted sentinel lymph node identification, occasions have been noted when a specimen thought to be the first node with tracer uptake has no lymph node present on final pathological report. Functionally, this is a failed mapping; however, the surgeon would not be aware of the failure during the operation. Data are lacking on this facet of sentinel lymph node identification, and it is not known whether there is a decrease in the rate of failed mapping (adipose-only sentinel lymph nodes) as the surgical case volume increases, or if there are other surgeon or patient factors that increase this rate of mis-identification. Adipose-only sentinel lymph nodes are concerning because a surgeon could be falsely reassured that a sentinel lymph node has been obtained when actual nodal tissue will not be evaluated, leading to inadequate staging. In patients with high-risk disease (histology other than endometrioid or grade 3 cancer), this is particularly concerning owing to the higher rate of nodal metastasis.15
The purpose of this study is to investigate the impact of patient and surgeon factors on the adaptation of an sentinel lymph node mapping algorithm. A secondary objective was to describe causes of mis-identifying adipose-only sentinel lymph nodes and factors related to their identification. We hypothesize that a sentinel lymph node mapping algorithm can be successfully adapted regardless of patient and surgeon factors.
This is a retrospective cohort study of patients with a diagnosis of endometrial carcinoma who underwent a hysterectomy with planned pelvic lymphadenectomy (with or without para-aortic lymph node dissection) after performing sentinel lymph node mapping with indocyanine green and near-infrared imaging with the da Vinci Si surgical system between 2013 and 2017. Institutional Review Board approval was obtained from the Medical College of Wisconsin and all policies regarding private health information were followed. A standardized data collection form was implemented, and patient data and operative times of the following were prospectively collected: cervical injection with indocyanine green; first abdominal incision; docking of the robotic surgical system; opening of the right and left retroperitoneal spaces; identification of the first sentinel lymph node; and completion of lymph node dissection. Difficulties with the procedure were recorded as free text if documented by the primary surgeon and typically classified as either adhesive disease of the abdomen or inadequate visualization of the surgical field. Final specimen pathology, FIGO grade, lymph node pathology, presence of uterine/cervical fibroids, uterine/cervical dimensions, and uterine weight (in grams) were also documented.
Patients who were diagnosed with endometrial cancer were included irrespective of histology, stage of cancer, or medical comorbidities. Patients were excluded from final analysis if they were without endometrial cancer on pre-operative biopsy or final pathology, or if surgical factors required abandonment of the attempt at sentinel lymph node mapping (Figure 1).
Indocyanine green and infrared light were used for sentinel lymph node identification, and all cases were planned robotic-assisted total laparoscopic hysterectomies. The cervix was injected with 1 mL indocyanine green to a depth of 1–2 cm (deep injection) and 2–3 mm (superficial injection) at the 3 and 9 o’clock positions for a total of 4 mL. Indocyanine green was diluted in 20 mL sterile water to a concentration of 1.25 mg/mL. The first node on either side of the pelvis that was visually green with tracer leading to it was removed. All sentinel lymph node specimens removed were considered to contain lymphatic tissue at the time of surgery unless proven otherwise at the time of pathological examination. During this adaptation phase, pelvic lymphadenectomy was planned for each patient after indocyanine green mapping, though ultimately this was performed at surgeon discretion. However, if a node was not identified with indocyanine green, a side-specific pelvic lymphadenectomy was performed.16 Para-aortic lymph node sampling was performed at the discretion of the surgeon.
All lymph nodes were stained with hematoxylin and eosin. For hematoxylin and eosin negative sentinel lymph nodes, pathological ultrastaging was performed by cutting two adjacent 5 µm sections at two levels 50 µm apart. At each level one side was stained with hematoxylin and eosin and the other with immunohistochemistry using pan cytokeratin (DAKO, Agilent, CA, USA) AE1 and AE3 for four total slides per block. Lymph nodes were classified by The Cancer Staging Manual of the American Joint Committee on Cancer criteria.17 The presence of isolated tumor cells did not change patient final stage, while the presence of micrometastases was considered positive for malignancy for staging purposes. Lymph node specimens were defined as benign if no malignant cells were identified within the lymph node. Sentinel lymph node specimens were defined as adipose-only if there was only fibroadipose tissue without lymphatic tissue present in the specimen; the pathology reports indicated that there was 'fibroadipose tissue only' in these instances.
The primary outcome was dependent on the learning curve and any patient characteristics that would significantly affect both the sentinel lymph node mapping rate and the rate of adipose-only sentinel lymph nodes. The secondary outcome of this study was the adipose-only sentinel lymph node identification rate during sentinel lymph node biopsy. To compare categorical variables, Fisher’s exact test was used. For continuous variables, the Mann–Whitney test (Kruskal–Wallis test for more than two groups) was performed. An unadjusted P-value of <0.05 was considered significant. Tree analysis was used to investigate possible predictors of adipose-only sentinel lymph nodes. The classification tree was optimized by Gini and 10-fold cross validation, and the split criteria were 10 for parent nodes and five for terminal nodes (Gini is a criterion for splitting at each node of the tree which uses the probability of the outcome, for example sentinel node positivity). The statistical programs used were SPSS version 24 (IBM Software, Chicago, WI, USA) and Salford Systems CART.
From November 2013 to April 2017, there were 208 patients who underwent sentinel lymph node mapping with robotic-assisted total laparoscopic hysterectomy for endometrial cancer; 202 were included in the final analysis (Figure 1). The median age was 63 years (35–91), and the median body mass index was 34 kg/m2 (17–62) (Table 1). Most patients (85%) had endometrioid adenocarcinoma on final pathology. Serous (12%), carcinosarcoma (3%), and clear cell (2%) were included, with mixed pathology in 2% (Table 1). FIGO IA disease was the most common stage (62%), and FIGO I the most common grade (48%).
The rate of mapping at least a unilateral sentinel lymph node was 86%, and the bilateral sentinel lymph node mapping rate was 66%; this is similar to previously published studies, though there is a trend toward increased mapping rates with indocyanine green in recent studies.4 8–10 Twenty-eight patients had no sentinel lymph node identified on either side. Bilateral pelvic lymphadenectomy was completed in 82% of patients, and pelvic and para-aortic lymphadenectomy in 55%. When comparing the sentinel lymph node mapping rate in the first 10 cases to later cases, there was not a statistically significant difference in the bilateral or unilateral mapping rate when all surgeons were compared (P>0.31) or when surgeons were compared with themselves (P>0.6). A total of 288 sentinel lymph node specimens were removed altogether from 404 candidate hemipelvises. Metastatic disease was present in 21 of these specimens: nine were macrometastases (41%); three were micrometastases (14%); and nine were isolated tumor cells (45%).
Adipose-only sentinel lymph nodes were identified in 30 patients; 8 (26.7%) of these patients had bilateral adipose-only sentinel lymph nodes (Table 2). Unilateral adipose-only sentinel lymph nodes occurred in 22 (73.3%) patients, 12 (54.5%) of which had a right-sided adipose-only sentinel lymph node, and 10 (45.5%) a left-sided adipose-only sentinel lymph node. Three of the patients with bilateral adipose-only sentinel lymph nodes had no pelvic or para-aortic lymphadenectomy. Increased age of 66 (36-91) vs 61 years (36-91) was associated with an increased rate of an adipose-only sentinel lymph node (P=0.03). Prolonged time to identification of a sentinel lymph node was also associated with increased adipose-only sentinel lymph node identification. Increased time from cervical injection to opening of the RP space (P=0.005) and identification of a sentinel lymph node was significantly associated with an increased rate of adipose-only sentinel lymph nodes (P=0.02). The median time from injection to identification did not significantly change when cases were compared in order from 1 to 10, 11–30, and >30. The median time from cervical injection to identification of the right sentinel lymph node for all cases was 35, 25, 44, and 27 min for surgeons 1–4 respectively (P=1, P=0.07, P=0.28, and P=0.56, respectively). Typically, sentinel lymph node dissections began in the right hemipelvis (n=188, 93%). High-grade cancer was not associated with an increased adipose-only sentinel lymph node rate (P=0.31). Other uterine factors that were not found to affect the adipose-only sentinel lymph node rate were uterine weight (P=0.23) and uterine fibroids (P=0.22).
The number of cases performed by each surgeon was 19, 79, 40, and 64 for surgeons 1–4, respectively. The median number of cases was one per month for surgeons 1, 3, and 4, and two per month for surgeon 2. Inter- and intra-surgeon adipose-only sentinel lymph node rates were compared, which revealed a lower adipose-only sentinel lymph node rate over time and with the number of cases performed. To analyze chronologically, the first full year of data collection (November 2013–December 2014) was compared with the last full year (January 2016–December 2016). The rates of adipose-only sentinel lymph nodes identification were 30.6% and 3.9%, respectively, a statistically significant decline (P<0.001). The caseloads of each surgeon were compared in the following case increments: 0–10, 11–30, and >30. Surgeon 1 was included up to the second case increment (Fig 2, Supplemental Digital Content). There was a significant decrease in the rate of adipose-only sentinel lymph nodes when comparing the first 10 cases to cases 11–30 and >30 (P=0.006).
Sensitivity and specificity analyses were performed in patients who underwent pelvic lymphadenectomy after identification of an sentinel lymph node with indocyanine green (Table 3). Adipose-only sentinel lymph nodes were included in this analysis because, at the time of surgery, they were treated as sentinel lymph node specimens. There were two mapped sentinel lymph nodes that were pathologically benign with malignancy found outside of the sentinel lymph node after lymphadenectomy. These were an adipose-only sentinel lymph node (false identification of sentinel lymph node) and a non-sentinel lymph node pelvic node positive for isolated tumor cells. A decision tree was used to examine important predictors with the primary outcome being adipose-only sentinel lymph nodes (Figure 2). There were 174 patients available for evaluation, as 28 failed sentinel lymph node mapping. Time from cervical injection to identification of a sentinel lymph node>25 min was statistically significant (P<0.001). For those adipose-only sentinel lymph nodes that were identified more than 35 min from injection with indocyanine green, age >64 years was a significant factor (P=0.004).
An adipose-only sentinel lymph node can be a 'false negative' identification while adapting a sentinel lymph node mapping technique as it represents a failure to evaluate sentinel lymph nodes for malignancy if that is the surgical intent. Risk factors for an increased rate of adipose-only sentinel lymph nodes include limited sentinel lymph node mapping experience with indocyanine green (≤30 cases of sentinel lymph node identification), patient age >64 years, and time to identification of sentinel lymph node >25 min after injection of indocyanine green dye. Surgeon factors related to adaptation of an sentinel lymph node mapping algorithm include operative time, surgical complications, sentinel lymph node mapping rate, and number of cases required for competence (ie, learning curve). Prior studies in minimally invasive surgery and gynecologic oncology define the learning curve as the number of cases needed to yield increased sentinel lymph node mapping rates and decreased operative time, citing 14–30 cases as the number needed for proficiency.18–20 During adaptation of sentinel lymph node mapping with indocyanine green, our recommendation is to continue systematic pelvic lymph node dissection, and for cases of delayed identification or patient age >64 years to either perform lymph node dissection or ensure a sentinel lymph node is present in the specimen prior to completion of surgery by surgeon ex vivo sentinel lymph node specimen palpation or intraoperative frozen section.
Patient factors that have been studied in relation to sentinel lymph node mapping include body mass index, The International Federation of Gynecology and Obstetrics grade, and histology of cancer.8 9 15 21 22 Higher body mass index has been associated with decreased mapping rates with indocyanine green,8 and it was expected to be associated with a higher adipose-only sentinel lymph node rate, as higher body mass index is associated with increased retroperitoneal adipose tissue.23 However, these factors were not associated with a decreased rate of mapping, nor did they influence the chance of harvesting an adipose-only sentinel lymph node. Body mass index was equal in the adipose-only sentinel lymph node and true sentinel lymph node groups (Table 2). Additionally, neither the presence of uterine fibroids nor uterine weight was associated with a decrease in the mapping rate or an increase in the number of adipose-only sentinel lymph nodes: these factors have not been assessed in prior studies. There was no difference in adipose-only sentinel lymph node rates between high- and low-grade disease or between the different histologic subtypes of endometrial cancer. This is reassuring for the adaptation of a sentinel lymph node mapping algorithm because patients with high-risk disease were at no higher risk of adipose-only sentinel lymph nodes.
If adipose-only sentinel lymph nodes are concerning, bilateral adipose-only sentinel lymph nodes are of particular concern. Of the eight patients that had this pathology noted, only three had no other sentinel lymph node evaluable as the others underwent pelvic and/or para-aortic lymph node dissection. However, if a sentinel lymph node mapping algorithm is adapted to replace lymphadenectomy, bilateral adipose-only sentinel lymph nodes are possible and would result in no nodal tissue whatsoever for pathologic evaluation. Though not specific to reducing the risk of an adipose-only sentinel lymph node, techniques proposed for increasing the rate of sentinel lymph node mapping with indocyanine green and decreasing 'failed' mapping include color-segmented fluorescence to differentiate higher from lower levels of indocyanine green uptake, ex vivo palpation of the specimen, and use of the Pinpoint Mapping System (Novadaq Technologies, Bonita Springs, FL) if using conventional laparoscopy.22 24 25 In the current study, there was no protocol for intraoperative confirmation of actual nodal tissue in a sentinel lymph node specimen. However, the presence of a learning curve is reassuring, as the number of adipose-only sentinel lymph nodes decreased with increasing case experience. Considering that time from cervical injection with indocyanine green to identification of the first sentinel lymph node did not significantly change with increasing surgical experience, prolonged time from injection to identification should lead the surgeon to consider performing a frozen section or lymphadenectomy regardless of surgeon experience. Lastly, patients older than 64 years were at increased risk of an adipose-only sentinel lymph node, which may be due to an age-related decline in lymphatic function.26
The main strengths of the current study are its description of a novel limitation to sentinel lymph node mapping in endometrial cancer (adipose-only sentinel lymph nodes) and the detailed prospective assessment of surgeon factors as they relate to sentinel lymph node mapping, specifically intraoperative timing. The surgeon factors included that were, to our knowledge, not examined in prior studies were timing from cervical injection with indocyanine green to opening of the RP space and identification of the first sentinel lymph node. Close examination and description of the surgical learning curve as it relates to both the identification of sentinel lymph nodes as well as adipose-only sentinel lymph nodes is another strength of the study. Dual analysis, chronologically and by caseload, showed a significant difference that was advantageous as the learning curve was confirmed through two separate means. Though surgeon 1 did not show improvement in the AO sentinel lymph node rate, this is most likely owing to the much smaller number of cases performed compared with the other three surgeons, especially considering this analysis showed a further decrease in AO sentinel lymph nodes after 30 cases. Our sample size is similar to recent studies9 and larger than the median (N=48) and mean (N=86) sample sizes of studies included in a recent meta-analysis.27 Lastly, the injection protocol for indocyanine green was standardized based on a prior study, and all surgeons used the same technique for injection.8
Though they were aware that certain timepoints were being documented, surgeons were not aware that timing information would be used to examine the learning curve or the chances of removing an actual lymph node. Whether or not the surgeon thought that an actual lymph node was being obtained was not recorded. Because the surgeons were planning to perform pelvic lymphadenectomy, there could have been inherent bias and they could have chosen to simply collect what appeared visually green, regardless of whether they were fully convinced that this was a lymph node. Lymphadenectomy was not performed on every patient, which would limit the validity of the sensitivity and specificity analysis. However, cases where pelvic lymphadenectomy was not completed were excluded from the sensitivity and specificity calculations, controlling for this limitation. Additionally, the median number of cases completed per month was one per surgeon except for surgeon 2. The speed of this learning curve could have had an unmeasured impact on the adipose-only sentinel lymph node rate. Also, the sensitivity of our analysis was lower than previous studies. Though there were two sentinel lymph nodes that mapped negative and subsequently had metastatic pelvic lymph node disease, one was an adipose-only sentinel lymph node and the other had a pelvic node that was positive for isolated tumor cells. Lastly, the single institution design of this study is a limiting factor.
Though previous studies reported that certain patient factors limited adaptation of a sentinel lymph node mapping algorithm, none were found to affect the rate of mapping in the current analysis.15 20–22 Surgeon factors initially limited adaptation but a learning curve in relation to identification of adipose-only sentinel lymph nodes was present, which decreased with experience beyond 30 cases. Preliminary evidence is presented that suggests a need to be cognizant of intraoperative timing after indocyanine green injection. In conclusion, an algorithm whereby intraoperative frozen section or lymphadenectomy is performed through the first 30 cases, especially in patients>64 years old or those cases with delayed identification >25 min, should be considered when adopting a sentinel lymph node mapping algorithm with indocyanine green and the robotic surgical platform.
Contributors JH: Primary author, contributed to both data collection and manuscript construction, and approved the final article. EB, JR, DU, WB: one of the surgeons included in the final analysis, contributed to manuscript construction, had editing comments, and approved the final article. MN, PS: responsible for statistical analysis, contributed to manuscript construction, had editing comments, and approved the final article
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial, or not-for-profit sectors.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.