Introduction Sentinel lymph node (SLN) dissection has been recognized as a valid tool for staging in patients with endometrial cancer. Several factors are predictors of recurrence and survival in endometrial cancer, including positive lymphovascular space invasion. The aim of this study is to formulate a pre-operative score that, in the event of no-SLN identification, may give an estimate of the true probability of lymphovascular space invasion and guide management.
Methodology This was a multi-institutional retrospective study conducted from January 2007 to December 2017. We included all patients with any grade endometrial tumor with a complete pathological description of the surgical specimen and with a minimum follow-up of 12 months. All patients underwent a class A hysterectomy according to Querleu and Morrow and bilateral salpingo-oophorectomy. Lymphadenectomy was performed based on patient risk of node metastases. In order to verify the predictive capacity of the parameters associated with lymphovascular space invasion status, grading, abnormal CA125 (>35 units/ml), myometrial invasion, and tumor size, a synthetic score was calculated. The score was introduced in the receiver operating characteristic curve model in which the binary classifier was represented by the lymphovascular space invasion status. The ideal cut-off was calculated with the determination of the Youden index. Sensitivity and negative predictive value of lymphovascular space invasion score was calculated in patients with lymph node metastasis.
Results Six hundred and fourteen patients were included in the study. The average age and BMI of patients were 64.8 (range 33–88) years and 30.1 (range 17–64) respectively. Of the 284 patients who underwent lymphadenectomy, 231 (81.3%) patients had no lymph node metastases, 33 (11.6%) patients had metastatic pelvic lymph nodes, 12 (4.2%) patients had metastatic aortic lymph nodes, and eight (2.8%) patients had both pelvic and aortic metastatic lymph nodes. Lymphovascular space invasion was associated with deep myometrial infiltration (P<0.001), G3 grading (P<0.001), tumor size ≥25 mm (P=0.012), abnormal CA125 (P<0.001), recurrence (P<0.001), overall survival (P<0.001), and disease-free survival (P<0.01). Of all patients with lymphovascular space invasion, 79% had an lymphovascular space invasion score ≥5. The score ranged from a minimum score of 1 to a maximum of 7. The score shows 78.9% sensitivity (95% CI 0.6971 to 0.8594), 65.3% specificity (95% CI 0.611 to 0.693), 29.4% positive predictive value (95% CI 0.241 to 0.353), and 94.4% negative predictive value (95% CI 0.916 to 0.964).
Conclusion We found that when lymphovascular space invasion score ≤4, there is a very low possibility of finding lymph nodal involvement. The preoperative lymphovascular space invasion score could complement the SLN algorithm to avoid unnecessary lymphadenectomies.
- SLN and lympadenectomy
- lymph nodes
- sentinel lymph node
- endometrial neoplasms
- uterine neoplasms
Statistics from Altmetric.com
Lymphovascular space invasion score may be a valid pre-operative tool to identify patients at high risk of lymph node metastasis.
Patients with lymphovascular space invasion score ≤4 have low risk of lymphovascular space invasion to the final pathology.
Low lymphovascular space invasion score may avoid unnecessary lymphadenectomies in patients where the sentinel lymph node is not detected.
Endometrial cancer is the most common gynecological cancer in industrialized countries, with 61 000 new cases and an increasing incidence of 2.5% every year.1 Several factors are implicated in predicting recurrence and survival such as age, grade, myometrial infiltration, tumor size, cervical invasion, and lymphovascular spaces involvement.2–4 These negative prognostic factors are commonly associated with each other and are often found together.5 Recent research has shown that age, lymphovascular space invasion, tumor size, and different molecular aspects intrinsic to the tumor, such as TP53 or L1CAM, should be considered determinants of high risk for recurrence.6–9 Among these, lymphovascular space invasion has been recognized as a potential key indicator for adjuvant therapy, even in the absence of other high-risk features, since it is independently and strongly associated with lymph node metastases, local and distant recurrence, and overall survival.10–12
Recently, sentinel lymph node (SLN) dissection has been recognized as a standard in the management of patients with both low- and high-risk endometrial cancer, in order to avoid the added morbidity of complete lymphadenectomy.13 14 Nevertheless, in some cases, the SLN technique may lead to overtreatment. According to the sentinel lymph node algorithm, in case of no SLN identification, a systematic lymphadenectomy is recommended for the risk of 30% of neoplastic lymphangitis.15 In case of failed mapping, even if the probability of lymph node involvement is low, a systematic lymphadenectomy is recommended in the hemipelvis where the SLN was not identified. Furthermore, the bilateral SLN detection ranges between 50% and 75% even in referral centers.16 17 In this way, about 25%–50% of patients undergoing the SLN technique will experience a full lymphadenectomy.
Our hypothesis is that the lymphovascular space invasion score will help in deciding whether or not to perform nodal dissection in patients diagnosed with endometrial cancer in whom no SLN is detected in one or both hemipelvis. The objective of this study is to formulate a pre-operative score that, in the event of failed SLN detection, may provide an estimate of the true probability of lymphovascular space invasion to the final pathology and guide the surgeon to correct surgical management, thereby avoiding overtreatment. Patients with failed mapping and low lymphovascular invasion score therefore could avoid full lymphadenectomy.
We have proposed this score only for patients with endometrial cancer because the therapeutic role of lymphadenectomy in this cancer is of doubtful utility.2 This score cannot be applied to other gynecological cancers, such as cervical or ovarian cancer.
This is a retrospective study of patients with a diagnosis of endometrial cancer treated at the Department of Gynecology and Obstetrics of Parma, at the Division of Gynecologic Oncology, Giovanni Paolo II Foundation of Campobasso, and at the Department of Gynecologic Oncology of Palermo from January 2007 to December 2017. We included all patients with any grade endometrial cancer with a complete pathological description of the surgical specimen and with a minimum follow-up of 12 months. All patients underwent a class A hysterectomy according to Querleu and Morrow and bilateral salpingo-oophorectomy.18 Lymphadenectomy was performed based on patient risk of lymph node metastases according to the Mayo criteria,19 (in case of grading G3, tumor diameter >2 cm, myometrial infiltration >50%). Patients who underwent exclusive SLN technique were removed from the study.
Disease stage and grade were defined according to the International Federation of Gynecology and Obstetrics (FIGO) guidelines,.20 Histological classification was based on the WHO classification.21 Lymphovascular space invasion was defined as the presence of neoplastic cells within the lymphatic spaces or close to blood vessels in the primary tumor. Overall survival was defined as the time elapsed between the date of surgery and death, and disease-free survival was defined as the time between the date of surgery and diagnosis of the first recurrence. A micro- and macroscopic description were provided by dedicated pathologists.
A dedicated sonographer performed transvaginal ultrasound preoperatively on all patients. Patients who were not able to undergo pelvic ultrasound underwent MRI.
The study was approved by the institutional review board or ethics committee at each participating institution, and all patients gave written informed consent.
Statistical analysis and score processing
Continuous variables were described as mean±SD. The categorical variables were described with absolute numbers and percentages. The univariate analysis was obtained with the t-Student and/or analysis of variance, when appropriate (post-hoc Tuckey test) for continuous variables. The association of categorical variables was initially analyzed by the chi-square test.
In order to verify the predictive capacity of the parameters associated with lymphovascular space invasion status, grading, abnormal CA125, myometrial invasion, and tumor size, a synthetic score was calculated: lymphovascular space invasion Score=Grade + CA125 +Myometrial invasion +Tumor size. The score was introduced in the receiver operating characteristic curve model in which the binary classifier was represented by the lymphovascular space invasion status. The ideal cut-off was calculated with the determination of the Youden index (J) according to the formula: J=sensitivity + specificity - 1. The association of all parameters with lymphovascular space invasion was verified by dichotomizing the following categorical variables: laparoscopic surgery vs laparotomy, age ≥65 years vs <65 years, BMI≥35 kg /m2 vs <30, FIGO stage IA vs IB - IV, grading G1/G2 vs G3, abnormal CA125 vs normal, previous radiotherapy vs no radiotherapy, previous chemotherapy vs no chemotherapy, tumor size ≥25 mm vs <25 mm, and myometrial invasion ≥50% vs<50%.
The variables that were significantly associated with lymphovascular space invasion on univariate analysis were introduced in a binary logistic regression model with multivariate OR and 95% CI. In order to normalize the effect of age at the time of diagnosis, a third model of multivariable analysis was performed using Cox regression, in which the variables resulting significantly associated with lymphovascular space invasion status at univariate analysis were also introduced, with the consequent calculation of the corresponding HR values and 95% CI. Disease-free survival and overall survival were analyzed using the Kaplan–Meier method of cumulative survival and were compared with each other using the Mantel–Cox logarithmic rank test. Finally, the sensitivity and negative predictive value of the lymphovascular space invasion score was calculated in patients with lymph node metastasis.
A total of 726 patients underwent surgery for endometrial cancer and, of these, 614 patients met the inclusion criteria and were included in the analysis. Patient characteristics are shown in Table 1. The average age and BMI of patients were 64.8 (range 33–88) years and 30.1 (range 17–64), respectively. Of the 614 patients, 232 (37.8%) patients underwent laparoscopy, 296 (48.2%) patients underwent laparotomy, and 86 (14%) patients had vaginal surgery. All patients underwent a class A hysterectomy according to Querleu and Morrow,18 bilateral salpingo-oophorectomy and 228 patients underwent concomitant pelvic lymphadenectomy, and 56 pelvic and para-aortic lymphadenectomy. Of the 284 patients who underwent lymphadenectomy, 231 (81.3%) patients were found to have no lymph nodal disease, 33 (11.6%) patients had metastatic pelvic lymph nodes, 12 (4.2%) patients had metastatic aortic lymph nodes, and eight (2.8%) patients had both pelvic and aortic lymph node metastases. In 95 (15.5%) patients evidence of lymphovascular space invasion was found, while in 519 (84.5%) patients none was noted. Eighty four (13.7%) recurrences were recorded, including 15 lymph nodal, 33 vaginal (pelvic), and 49 distant relapses. As shown in Table 2, lymphovascular space invasion was statistically associated with deep myometrial infiltration (P<0.001), type II tumor (P=0.003), G3 grading (P<0.001), tumor size ≥25 mm (P=0.012), abnormal CA125 (P<0.001), recurrence (P<0.001), radiotherapy (P<0.001), chemotherapy (P<0.001), worse overall survival (P<0.001), and worse disease-free survival (P<0.01) at univariate analysis. Furthermore, by assessing the site of recurrence, lymphovascular spaces were associated with a greater frequency of lymph nodal and distant metastases P=0.008 and P<0.001, respectively.
Patients with lymphovascular space invasion had a median survival of 50.6 (range 7–131) months compared with 79.4 (range 9–144) months of those with negative lymphovascular space invasion (P<0.001). Mean disease-free survival was 42.5 months and 74.5 months in patients with positive and negative lymphovascular space invasion, respectively (P<0.001). The Kaplan–Meier curves in Figure 1A&B show the overall survival and disease-free survival in patients with lymphovascular space invasion compared with those without. At multivariable analysis, grading G3 (adjOR 1.831, 95% CI 0.638; 5.252), abnormal CA125 (adjOR 3.506, 95% CI 1.270; 9.680), relapse (adjOR 4.090, 95% CI 1.488; 11.246), type II tumor (adjOR 1.399, 95% CI 0.310; 6.307), and deep myometrial infiltration (adjOR 17.004, 95% CI 1.776; 162.8) are more associated with lymphovascular space invasion (Table 2). In particular, myometrial infiltration ≥50% increases the probability of positive lymphovascular space invasion by more than 17 times.
The time-dependent Cox regression analysis, correcting for age, showed a statistically significant correlation in patients with lymphovascular space invasion with deep myometrial infiltration (HR 2,883, 95% CI 0.675; 12.312), grade G3 (HR 1.234, 95% CI 0.642; 2.443), tumor size >25 mm (HR 2.220, 95% CI 1.002; 4.918), abnormal CA125 (HR 1.787, 95% CI 0.925; 3.453), and relapse (HR 1.621, 95% CI 0.823; 3.193). In the analyzed series, 18.6% of the patients (53 in total) had lymph node metastasis. The sensitivity and negative predictive value of the “lymphovascular space invasion score” in predicting nodal metastases is 100%, 95% CI 83.9% to 100%, 95% CI 43.9% to 100%, respectively.
Lymphovascular space invasion score
The lymphovascular space invasion score is based on a series of pre-operative parameters to which a value of 0,1, or 2 is assigned: the score, therefore, ranges from a minimum score of 1 to a maximum of 7. A score of 0 was given in the case of normal CA125. A score of 1 was given in the case of G1/G2, abnormal CA125, myometrial invasion <50%, and tumor size 0–24 mm. A score of 2 was given in the case of G3, myometrial infiltration >50%, and tumor diameter ≥25 mm. Table 3 shows the relative score assigned for each pre-operative factor analyzed. Table 4 shows the distribution of patients with lymphovascular space invasion in relation to their lymphovascular space invasion score. Of all patients with positive lymphovascular space invasion, 95 (79%) patients had lymphovascular space invasion score ≥5. A total of 29.5% of patients with lymphovascular space invasion had a score of 5, 27.4% patients had a score of 6, and 22.1% of patients had a score of 7. The lymphovascular space invasion score showed 78.9% sensitivity (95% CI 0.6971 to 0.8594), 65.3% specificity (95% CI 0.611 to 0.693), 29.4% positive predictive value (95% CI 0.241 to 0.353), and 94.4% negative predictive value (95% CI 0.916 to 0.964) with a likelihood ratio of 2.276.
In case of lymphovascular invasion score ≤4, there is a low probability of lymphovascular space invasion in the final pathology. The score has a negative predictive value of 94.4% and a sensitivity of 78.9%. Despite the progress achieved in the sentinel node technique with the advent of new dyes and different infiltration techniques, bilateral detection rates are still below optimal levels. Khoury–Collado et al, in a series involving 266 patients report at least one-sided uptake in 84% of cases and bilateral uptake in only 67% of cases, although 12% of their patients had received a second infiltration into the uterine fundus and 28% an additional preoperative cervical injection with Tc99.22 The same results are also reported during the FIRES clinical trial in which bilateral detection was only 52% while at least one-side uptake was 86%.14
Although the SLN technique is an alterantive to complete lymphadenectomy, failing to detect it may lead to unnecessary lymphadenectomy in certain patients. Data in the literature imply that in approximately one-third of patients, a lymphadenectomy is performed at least on one hemipelvis when in the majority of cases lymph nodes will prove negative.22 This is even more important in patients with low risk of lymph node metastases.12
To design this score, as a first step, we performed a thorough analysis of the literature by looking at the prognostic factors that most heavily influence survival in patients diagnosed with endometrial cancer. Among them, the lymphovascular space invasion plays a major role as an independent factor even in multivariate analysis, worsening the overall survival and increasing the risk of both local and distant recurrence.6 10 23 24 These results have also been shown in our series. We also registered the role of other poor prognostic indicators, such as elevated CA125, deep myometrial infiltration, grade G3recurrence, and tumor size ≥25 mm. In light of our findings, we decided to use lymphovascular space invasion as the focus of our score.
Several authors have attempted to formulate a score that could predict the risk of lymph node infiltration, but most scores are based on postoperative molecular analysis and often lead to more adjuvant therapy rather than offering additional information for intraoperative management. Yang et al25 report a lymph node metastases predicting score based on progestin receptors Ki-67, and CA125 with 84.6% sensitivity and 67.4% specificity.26 Before the integration of the sentinel lymph node technique, Mariani et al devised a prediction score for risk of lymph node metastases.19 The authors identified 328 patients at low risk of lymph node metastases and relapse not requiring lymphadenectomy or adjuvant radiotherapy. In this group of patients, the authors reported 5% of lymph node metastasis and an overall and disease-free survival at 5 years of 96% and 97%, respectively. In their series, tumor diameter ≤2 cm had a decisive role, in this subgroup of patients, in fact, even in cases of lymphovascular space invasion no lymph node metastasis was found among patients with smaller tumors. Conversely, tumors>2 cm with lymphovascular space invasion showed 27% node metastases. The authors, therefore, identified low-risk patients with tumor diameter ≤2 cm, grade G1/G2, myometrial infiltration <50%, and endometrioid histotype. The recognition of these characteristics, however, takes place during the operation. Finally, Mariani et al did not report a score, therefore a comparison with our lymphovascular space invasion score is not possible. We also consider that our score is not a substitute for the Mayo criteria but it does support it, especially in centers where the accuracy of frozen section may not be as high as in the Mayo clinic.
In line with Mitamura et al27 (using MRI) or Zhou et al28 (using CA125), all the parameters included in our score can be assessed pre-operatively.27 29 This may allow planning of a tailored surgical strategy before and not during surgery. Rather than focusing on a single parameter, our score is more comprehensive but still a rather simple system. Furthermore, not all hospitals have a frozen analysis during the operation, so after the validation, our score could be a useful tool. In our series, patients who underwent exclusive SLN technique were excluded, to avoid the potential bias of false negative cases.
A possible weakness of our study could be the fact that the tumor size and the degree of tumor infiltration are considered preoperatively with MRI and ultrasound. Preoperative evaluation of these factors is less accurate than the final pathology. However, the sensitivity of these tests is comparable to the definitive pathology and to frozen section.28 30 In our centers, in line with the literature, the diagnostic accuracy of transvaginal ultrasound in the detection of deep myometrial infiltration and stromal invasion is 80% and 91% respectively, compared with the final pathology.25 31 Furthermore, basing a score on preoperative parameters was our main objective, and this bias reproduces daily clinical practice. Another weakness of the lymphovascular space invasion score could be the fact that grading assessed at pre-operative biopsy does not always correspond to definitive histological examination. Of 494 patients with G1-G2 tumor, 88 (17.8%) were G3 at final pathology. Our data are in line with the literature.32 As shown by Helpman et al33 on a large series of 653 patients, as many as 18% of patients considered G1 were actually G2, and 2% were G3 on definitive histological examination. Huang et al34 reported on a series of 360 patients, 13.9% of patients with preoperative G1 tumors were actually G2 at final pathology, and 35.6% of patients with preoperative G2 tumors were G1. To avoid this issue in our scoring system, we considered G1 and G2 together. Lastly, in our study, a total of 86 patients underwent vaginal surgery, and therefore none had a lymphadenectomy. However, the choice of this approach has been reserved for patients with pelvic organs prolapse, with low performance status, and age over 75 years. Of these patients, only four had a relapse and only one had nodal recurrence. Furthermore, more than 80% of these patients had pathological stage IA and therefore at low risk of nodal involvement.
We would like to conclude by proposing the use of this score in the setting of the sentinel node algorithm to avoid unnecessary lymphadenectomy in cases where no SLN is detected. (Figure 2) We also consider that the negative predictive value of the score is high (94.4%), indicating that in the event of a lymphovascular space invasion score ≤4, there is a very low likelihood of lymph node involvement. Additional studies to validate this score are required to integrate it into daily clinical practice.
Contributors Conception of the work: CVA, BR. Analysis, interpretation, and collection of data: CVA, BR, SG, CV, CA, GG, AG, PM. Drafting the work: BR, US, SG, CV. Final approval: BR, SG, CV.
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.
Data availability statement Data are available upon reasonable request. None.
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.