Article Text
PDF
Abstract
Objective To determine whether the rate of sentinel lymph node (SLN) dissections that do not yield a lymph node on pathological analysis ('empty packet dissection') changes with increasing surgeon experience in the setting of patients undergoing minimally invasive hysterectomy and SLN mapping using indocyanine green dye for endometrial cancer.
Methods All patients undergoing SLN mapping using indocyanine green dye at the time of minimally invasive hysterectomy for endometrial cancer or complex atypical hyperplasia were identified between January 2013 and September 2017 at our institution. All surgeons had prior experience performing SLN mapping for endometrial cancer using other methods. The rate of empty packet dissections and SLN counts were evaluated using a logistic regression model analysis.
Results In total, 236 patients undergoing SLN mapping for either endometrial cancer (85%) or complex atypical hyperplasia (15%) were identified from a prospectively maintained database. When examining all six surgeons together, the percentage of empty packet dissections decreased with increasing number of procedures performed. Each additional procedure was associated with a 3.6% reduction in the odds of an empty packet SLN dissection. After adjusting for individual surgeons, each additional procedure was associated with a 4.9% reduction in the odds of an empty packet. The expected odds of an empty packet after 10 additional procedures decreased by 40.1% (95% CI 12.4% to 58.6%). The addition of two covariates (age and body mass index) did not contribute significantly to the model (likelihood ratio test: X2=2.75, p=0.25). The rate of empty packets appeared to stabilize after approximately 30 procedures. The number of SLNs removed did not change with increasing surgeon experience.
Conclusion The rate of empty packet SLN dissections using indocyanine green dye decreases with increasing number of procedures. This stabilizes after 30 procedures, suggesting completion of a learning curve.
- sentinel lymph node dissection
- endometrial cancer
- learning curve
Statistics from Altmetric.com
HIGHLIGHTS
Within the first 25 procedures, the empty packet rate was 20% compared to 7% in the following 25 procedures.
A total of 25–30 sentinel lymph node (SLN) mapping procedures defines learning curve in avoiding empty packet dissection.
We suggest criteria to mitigate the risk of performing an unsuccessful SLN dissection.
Introduction
Sentinel lymph node (SLN) mapping has increasingly been recognized as an effective technique for lymphatic assessment in patients undergoing surgical staging of endometrial cancer.1 2 While SLN mapping is standard practice for other solid tumors such as breast cancer and melanoma, lymphatic mapping in endometrial cancer is not without controversy. This controversy stems from the wide variability of bilateral SLN mapping rates observed in the literature, as well as continued uncertainty about optimal injection strategy.3 In a recently completed prospective trial (FIRES) evaluating the false-negative rate of SLN mapping using indocyanine green dye for endometrial cancer, SLNs were identified bilaterally in only 52% of patients. The authors cited the learning curve for SLN dissection as a potential important contributing factor to the low rate of bilateral mapping.4
The relationship between surgeon experience and learning curve for SLN mapping is well described in the breast cancer literature. Surgeon experience clearly impacts the rate of successful SLN mapping in this setting.5 6 Numerous studies have sought to define an optimal experience threshold where the rates of false-negative and failed mappings are minimized. A reduction in the risk of SLN mapping failure appears to occur in breast cancer surgery after approximately 20 cases.6
A similar learning curve has not yet been established for SLN mapping of endometrial cancer. As our surgeons had previous experience with SLN mapping using cervical injection of isosulfan blue dye, we were unable to determine learning curve characteristics with SLN mapping overall. However, we did anecdotally observe an increase in the rate of SLN dissections without a lymph node present on histopathological analysis during our initial experience switching from isosulfan blue to indocyanine green dye. Therefore, our objective was to determine whether the rate of SLN packets that do not yield a lymph node on pathological analysis ('empty packet dissection') changes with increasing surgeon experience in the setting of patients undergoing SLN mapping using indocyanine green dye during minimally invasive hysterectomy for endometrial cancer. Additionally, we sought to determine whether the number of lymph nodes obtained during SLN mapping changes with surgeon experience.
Methods
Study Design
Following institutional review board approval, we identified all patients who underwent robotic-assisted total laparoscopic hysterectomy with SLN mapping using indocyanine green dye for endometrial cancer or complex atypical hyperplasia at our institution between January 2013 and September 2017. Three surgeons were fellowship-trained in SLN dissection and robotic surgical techniques while four surgeons were not. All surgeons were skilled minimally invasive surgeons, each performing a minimum of 25 robotic surgical procedures annually for endometrial cancer or complex atypical hyperplasia. One surgeon was excluded from the statistical analysis as she had previous experience with SLN mapping using indocyanine green dye at another institution prior to joining our faculty.
Indocyanine green dye was injected into the cervix at the time of uterine manipulator placement. One 25 mg phial of indocyanine green dye was reconstituted in 10 mL aqueous solvent and further diluted in 10 mL normal saline (1.25 mg/mL). Four milliliters of solution were then injected into the cervical stroma. For both the 3 and 9 o'clock positions, one milliliter of solution was injected superficially (submucosa) and one milliliter injected deep (~8 mm). After intra-cervical injection was performed, the robot was docked and our previously published SLN algorithm performed by all surgeons.7
Outcomes
The rate of empty packet dissections was defined as the proportion of procedures in which the SLN dissections did not yield a lymph node on final pathological analysis after the procedure. If a unilateral dissection did not yield a lymph node, then it was counted as a procedure with an empty packet dissection. The total number of SLNs identified on pathological analysis after each procedure was also determined.
Statistical Analysis
To assess outcomes with increasing numbers of procedures, procedures for each individual surgeon were ordered and numbered according to date. Since only one surgeon performed more than 50 procedures during the study period, the analysis was limited to the first 50 procedures for each surgeon.
Outcomes (empty packet rate and total number of SLNs identified) were calculated across all surgeons within groups of five procedures for the first 50 procedures: within the first five procedures performed, within the second five procedures performed, and so forth. The relationship between the probability of an empty packet and the number of procedures (within the first 50 procedures) was assessed using logistic regression. Logistic regression models were adjusted for individual surgeon effects as well as patient age and body mass index (BMI). Likelihood ratio tests were used to compare between logistic regression models and Wald confidence intervals were constructed for odds ratios.
For the outcome of empty packets, 'learning curves' were constructed for each individual surgeon and combined across all surgeons. The learning curve, or cumulative empty packet rate for a fixed number of procedures, was defined as the percentage of empty packet dissections within that number of procedures. Comparison of empty packet rates for different numbers of procedures was performed using Fisher’s exact test. The statistical package R version 3.5.0 was used for the analysis.
Results
During the study period, 236 patients underwent SLN mapping with indocyanine green dye intra-cervical injection during robot-assisted total laparoscopic hysterectomy for complex atypical hyperplasia or endometrial cancer. The clinical and pathological features of the 236 patients can be found in Table 1. SLN mapping was bilateral in 82.2% of patients, unilateral in 16.1% of patients, and neither side in 1.7% of patients. Of the 232 patients in which at least one SLN was identified, at least one specimen labeled as a SLN was actually an empty packet dissection in 33 patients (14%). Therefore, 82.2% of patients had bilateral SLNs mapped and dissected but only 73% (173 patients) of patients had bilateral SLNs mapped, dissected, and confirmed as 'lymph node tissue' on final pathological analysis. Nodal metastases were identified in 8.5% (n=20 of 236 patients) of patients overall and in 6.4% of patients by SLN dissection alone. A total of 86 patients underwent complete lymphadenectomy after SLN dissection was performed. There were five false-negative SLN dissections. In these cases, a SLN was negative for tumor and a lymph node metastasis was identified in an ipsilateral non-SLN. There was one patient in whom an empty packet dissection was identified on the same side as a non-SLN lymph node metastasis. The percentages of at least one empty packet dissection per SLN dissection ranged from 0% to 40% according to the individual surgeon.
Clinical and pathological features (N=236)
When considering all surgeons together, the percentage of empty packet dissections decreased with increasing numbers of procedures performed (Figure 1). Each additional procedure was associated with a 3.6% reduction in the odds of an empty packet dissection (Table 2). After adjusting for individual surgeons, each additional procedure was associated with a 4.9% reduction in the odds of an empty packet dissection. Thus, after 10 additional procedures, the expected odds of an empty packet dissection would be decreased by 40.1% (95% CI 12.4% to 58.6%). Similar results were seen after additional adjustments for age and BMI. The addition of these two additional covariates did not contribute significantly to the model (likelihood ratio test: X2=2.75, p=0.25; see Table 2). The cumulative empty packet rate for all surgeons as a group and individual surgeons according to number of procedures performed is shown in Figure 2. The cumulative empty packet dissection rate for all surgeons appears to stabilize after 25–30 procedures. Within the first 25 procedures across all surgeons the empty packet rate was 19/95 (20%), compared with an empty packet rate of 5/68 (7%) in the 26–50 procedures (Fisher’s exact test: p=0.03). When considering the single surgeon who performed more than 51 procedures, the empty packet rate was 8/25 (32%) in the first 25 procedures, compared with 9/97 (9%) in subsequent procedures (Fisher’s exact test: p=0.009). With respect to the number of lymph nodes obtained during SLN dissection, we found no association between the average number of SLNs removed by surgeons and the number of procedures performed. The relationship between the probability of detecting a SLN bilaterally or unilaterally and the number of procedures performed was also assessed. With respect to the bilateral or unilateral detection rate, there was no statistically significant relationship between the probability of detection and the number of procedures performed, even after adjusting for individual surgeon effect.
Percentages of empty packet dissection by number of procedures.
Cumulative empty packet rate by number of procedures for individual surgeons (dashed lines) and all surgeons (solid line).
Relationship between the probability of an empty packet dissection and the number of procedures performed: logistic regression results
Discussion
An increasing body of literature has established the ability of SLN mapping to accurately predict lymph node status for patients with endometrial cancer. As initial concerns about diagnostic accuracy have dissipated, many surgeons are now comfortable with omitting completion lymphadenectomy in patients who have undergone apparently successful bilateral SLN identification. This has led to new concerns about the frequency and clinical implications of 'empty packet' dissections (ie, no lymph nodes identified on pathological analysis of an apparent SLN). The risk of empty packet dissections is small but problematic. In our high-volume center, we identified an empty packet dissection in 14% of patients. Fortunately, we found that the odds of an empty packet SLN dissection decreases with increasing surgeon experience. The risk of an empty packet dissection appears to stabilize after 25–30 cases but does not completely disappear.
In the current study, we also evaluated the average number of SLNs removed with increasing surgeon experience. We hypothesized that as surgeon experience with SLN dissection using indocyanine green dye mapping increased, the number of lymph nodes removed would decrease. The rationale for this hypothesis was that indocyanine green dye mapping results in improved visualization of lymphatic channels and could initially lead to over-sampling. We found no such association between surgeon experience and the number of lymph nodes removed over time.
Nodal metastasis is an important prognostic factor for patients with endometrial cancer.8 Though the risk of nodal metastasis is small in patients with low-grade tumors, the implications of missing a nodal metastasis due to misidentification of a lymph node are significant. If no lymph nodes are identified in a SLN specimen on final pathology, clinicians are left with the dilemma of whether to consider a patient adequately staged. Furthermore, the advent of pathological ultrastaging has allowed for the detection of small-volume metastases in lymph nodes. Whether information obtained through ultrastaging provides an opportunity to improve clinical outcomes due to the use of adjuvant therapy remains unproven.9 Regardless, an empty packet dissection leaves the clinician without information that they intended to gather intra-operatively.
Our results are consistent with the robust literature evaluating the learning curve for SLN identification in breast cancer and melanoma.10–13 Cox et al sought to examine the learning curve for SLN mapping in breast cancer using a prospective analysis of 16 surgeons who performed 2255 cases. They found that the failed mapping rate was 10% after 22 cases but fell to 5% after 63 cases. Importantly, their data also suggest that surgeons who infrequently perform SLN dissections may have higher failure rates even in the context of high lifetime case numbers. The failure rate for breast surgeons performing fewer than three SLN biopsies per month was 13.7% compared with a failure rate of 2.2% for surgeons performing more than six cases per month.5 These results led to the publication of a consensus statement recommending that breast surgeons document their own experience with SLN mapping and to continue to perform complete axillary dissection until able to demonstrate a consistent, acceptable detection rate (>90%) and false-negative rate (<5%).13 The learning curve for SLN dissection is also well described in the melanoma literature and is approximately 30 cases.12 While we are unaware of any breast cancer or melanoma literature explicitly addressing the risk of empty packet dissections, our results do corroborate a similar learning curve when performing SLN mapping for endometrial cancer using indocyanine green dye.
Another study suggested a similar trend in the learning curve for SLN identification in endometrial cancer, although with an older injection dye. Khoury-Collado et al performed a prospective study of 115 patients undergoing SLN mapping for endometrial cancer at Memorial Sloan-Kettering Cancer Center.14 The overall SLN detection rate utilizing cervical injection of isosulfan blue dye was 85%. Increasing surgical volume was associated with significantly increased detection rates as demonstrated by an increase in SLN identification from 78% in the first 27 months of the study period to 94% in the next 15 months (p=0.018). SLN detection rates per surgeon also increased significantly after 30 cases (77% vs 94%, p=0.033). In this study, only one surgeon performed enough cases (95/115 total cases) to emerge from the learning curve period encountered in the breast cancer literature.14
One of the strengths of our study is the number of surgeons who performed SLN dissection during the study period. This allowed for multiple individual analyses evaluating the odds of an empty packet dissection over time. Our study uniquely evaluates not only mapping rates but also the risk of empty packet dissections. We were also able to adjust for BMI, a patient characteristic known to increase the risk of failed SLN identification.7 One limitation of our study was that we evaluated surgeons with previous SLN dissection experience using isosulfan blue dye; therefore, a true learning curve assessment for the novice gynecologic oncology surgeon performing SLN dissection was not possible.
There are currently no consensus recommendations for the minimum number of procedures that should be performed in order to achieve competency with SLN mapping for endometrial cancer. However, our results suggest that the risk of empty packet dissections does decrease after 25–30 cases, consistent with other disease sites. As SLN mapping techniques improve, efforts should focus not only on maximizing mapping rates but also minimizing the risk of dissections that do not yield lymph nodes on pathological analysis. These results have led us to critically examine our SLN mapping practices. We have postulated the following criteria to mitigate the risk of performing an empty packet dissection at our institution:
Surgeons must have a thorough understanding of lymphatic anatomy in the pelvis. Most importantly, SLN are rarely found in parametria in patients with endometrial cancer. Additionally, dilated lymphatic channels can often be identified lateral to the internal iliac vessels before reaching the true lymph nodes along the pelvic sidewall. Surgeons should be skeptical if a 'lymph node' is identified in these areas. Lymph nodes will have a distinct capsule and will not be translucent under close inspection (see Figure 3).
On removal of any SLN specimen, the surgeon or experienced assistant must palpate all SLNs to ensure nodal tissue present.
If still unsure, we recommend sending the potential specimen to pathology for rapid gross assessment. The pathologist should be able to confirm whether nodal tissue is present without degrading their ability to perform histopathological analysis. If an empty packet is identified on gross assessment, the surgeon should either attempt to identify the correct nodal packet (usually adjacent to the empty packet) or proceed with a side-specific lymphadenectomy.
Image of an 'empty packet' dissection. A potential sentinel lymph node (SLN) is identified in natural light (left) and with fluorescence imaging (right). The tissue is a dilated lymphatic channel traversing between the parametria and left pelvic sidewall and was found to have no lymph node on histologic evaluation. SLNs are rarely found medial to the pelvic sidewall. Potential SLNs should always be palpated on removal to ensure nodal tissue is present.
With these guidelines in place, we hope to reduce the risk of empty packet dissections at our institution. As in the breast cancer literature, we recommend that all surgeons track their SLN mapping accuracy and rate of empty packet dissections to ensure that they are fulfilling the best interests of their patients using this technique.
Acknowledgments
The authors would like to acknowledge support for the statistical analysis from the National Center for Research Resources and the National Center for Advancing Translational Sciences (NCATS) of the National Institutes for Health through Grant No. 1UL1TR001079.
References
Footnotes
Contributors LT contributed to study design, data collection, manuscript writing, data analysis, and manuscript revision. LJ performed data analysis. EJT contributed to study design, data collection, manuscript writing, data analysis, and manuscript revision. RS, SW and AF contributed to manuscript writing and revision.
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.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.