Division of Gynecologic Oncology, Department of Obstetrics, Gynecology, and Reproductive Sciences, University of California at San Francisco School of Medicine, University of California, San Francisco Comprehensive Cancer Center, San Francisco, California
Division of Gynecologic Oncology, Department of Obstetrics, Gynecology, and Reproductive Sciences, University of California at San Francisco School of Medicine, University of California, San Francisco Comprehensive Cancer Center, 1600 Divisadero St., Box 1702, San Francisco, CA 94143-1702
The purpose of the current study was investigate the association between the number of lymph nodes examined and the probability of detecting at least a single lymph node involved by metastatic disease in patients with endometrioid corpus cancer.
Demographic, clinicopathologic, and surgical information were obtained from the National Cancer Institute between 1990 and 2001. A logistic regression model was used to investigate the relation between the number of lymph nodes identified and the probability of detecting at least a single positive lymph node.
Of 11,443 patients, the median age was 64 years (range, 22–74 years). In all, 78.7% had stage I disease, 10.3% had stage II disease, and 11.0% had stage III disease; 31.5% had grade 1 histology, 40.6% had grade 2 histology, and 24.3% had grade 3 histology. The median number of lymph nodes reported was 9 (range, 1–90 lymph nodes). The median number of lymph nodes and the percent of patients with positive lymph nodes have increased from 1988 to 2001. An increasing number of lymph nodes removed was associated with a higher likelihood of identifying those with lymph node metastases. Based on the logistic regression model, the largest increase in probability of detecting at least a single positive lymph node was observed when 21 to 25 lymph nodes were resected (odds ratio [OR] of 1.45; 95% confidence interval [95% CI], 1.08–1.94 [P < .01]). Removing greater than 25 lymph nodes did not improve the statistical probability (OR of 1.23; 95% CI, 0.94–1.61 [P = .13]).
Endometrial cancer is the most common gynecologic malignancy, with an estimated 41,200 new cases diagnosed in the U.S. in 2006.1 Although the majority of patients present with early-stage disease, the overall survival has not improved over the past 25 years. In fact, the number of deaths related to uterine cancer has increased from 5900 in 1995 to 7350 in 2006.1 One of the controversies related to the treatment of uterine cancer involves defining the role of lymph node assessment in the surgical management of this disease.
Surgical evaluation became part of the staging criteria for uterine carcinoma by International Federation of Gynecology and Obstetrics (FIGO) criteria in 1988. Prior studies by the Gynecologic Oncology Group have demonstrated the impact of lymphadenectomy on the prognosis of endometrioid cancer.2–4 Lymph node assessment is important for determining the extent of disease, the need for adjuvant therapy, and may have independent therapeutic value.5–7 As such, the National Comprehensive Cancer Network guidelines recommend a lymphadenectomy for all uterine cancer patients.8 However, to our knowledge, there is no consensus among gynecologic oncologists and pathologists with regard to the definition of an “adequate” lymphadenectomy. Given that lymph node status is 1 of the criteria on which the decision for adjuvant therapy is made, a recommendation on the minimum appropriate number of lymph nodes recovered that confirms lymph node negative disease is important.
In colorectal carcinoma, studies have been designed to determine the number of lymph nodes that should be removed to define an adequate surgical dissection.9–11 Based on results from institutional-based studies and large patient databases, the College of American Pathologists established guidelines for colorectal cancer surgery, stating that a minimum of 12 to 15 lymph nodes are required to confirm that regional lymph nodes are negative for disease.12 In uterine cancer, single institution-based studies have suggested that the removal of 10 pelvic and 5 para-aortic lymph nodes is required to constitute an adequate lymphadenectomy.6 However, the selection of the number of lymph nodes was an arbitrary decision. Furthermore, most of these prior studies have been based on small numbers of patients from single institutions, which lack the statistical power to determine the number of lymph nodes that should be removed to define an adequate surgical dissection.6, 13
In this large population-based study of 11,443 patients, 2 mathematical approaches were designed to: 1) examine the correlation between the percent of specimens with lymph node metastases and the number of recovered lymph nodes, and 2) estimate the probability of detecting at least 1 positive lymph node associated with the number of lymph nodes examined in endometrioid uterine cancer. This analysis can potentially estimate the minimum number of lymph nodes that are needed to define an “adequate” lymph node dissection.
MATERIALS AND METHODS
Demographic, clinicopathologic treatment, and survival information on 11,443 women diagnosed with endometrioid uterine carcinoma from 1990 to 2001 were extracted from the Surveillance, Epidemiology, and End Results (SEER) database from the National Cancer Institute (NCI). Data are reported from 9 population-based registries that represent approximately 14% of the U.S. population.14
Patients who did not undergo lymph node dissection were excluded from the analysis. Moreover, women with uterine papillary serous carcinoma, clear cell carcinoma, and sarcomas were not included in our study. Those with stage IV cancer were excluded due to the low rate of lymph node dissection and limited information regarding lymph node status.
The dependent variable of interest is the probability of detecting 1 or more positive lymph nodes among all dissected lymph nodes. Multivariate logistic regression models were used to determine the probability of detecting at least a single positive lymph node after controlling for the independent variables stage, grade of disease, and year of diagnosis. Independent variables included in the model were selected based on a significant univariate association with the dependent variable and goodness of fit as tested by likelihood ratio tests. To allow for the possibility of nonlinear increase in the log odds ratio (OR) with increasing number of lymph nodes dissected, several dichotomous indicator variables were created to represent the independent variable number of dissected lymph nodes in addition to testing number of lymph nodes as a continuous variable. The best model with number of lymph nodes represented by 5 design variables was selected based on tests for goodness of fit. Statistical Analysis was performed using the STAT 8.0 software program (StataCorp, College Station, Tex).
The demographic and clinicopathologic characteristics of the study population are provided in Table 1. Of the 11,443 women the median age was 64 years (range, 22–74 years). The majority (86.2%) were aged ≥50 years. White patients comprised 87% of the study group, African Americans 4.9%, Asians 5.7%, and other ethnic groups comprised 2.4%. Of the study cohort, 9009 (78.7%) had stage I disease, 1211 (10.3%) had stage II disease, and 1223 (11.0%) had stage III disease. In all, 31.5%, 40.6%, 24.3%, and 3.6% of patients had grade 1, 2, 3, and unknown grade of disease, respectively.
Table 1. Demographic and Clinicopathologic Characteristics (n = 11,443)
No. of Patients
No. of lymph nodes recovered
No. of positive lymph nodes
There was a wide range in the number of lymph nodes reported from the surgical staging. The median lymph node count was 9 (range, 1–90 lymph nodes). Greater than half of the patients (53.3%) had ≤10 lymph nodes identified, approximately one-fourth (26.3%) had 11 to 20 lymph nodes identified, and one-fifth (20.4%) had >20 lymph nodes identified. The median number of lymph nodes was similar by stage of disease, with medians of 9, 11, and 10 lymph nodes identified in patients with stage I, II, and III disease, respectively. Over the time periods from 1988–1990, 1991–1993, 1994–1996, 1997–1999, and 2000–2001 the median number of lymph nodes identified increased from 6, 7, 9, and 11, to 12, respectively. The percent of patients with lymph node metastases also increased from 4.4%, 4.4%, 6.9%, 5.4%, to 6.0% over these same time periods (Fig. 1). Of the 11,443 patients in our study, 638 (5.6%) had lymph node metastases. The median number of lymph nodes identified for lymph node-positive patients was 12. The correlation between increasing number of lymph nodes identified and number of lymph node metastasis is shown in Figure 2. It is important to evaluate the cumulative percentage of specimens with lymph node metastases based on the number of recovered lymph nodes to identify the point after which an increase in the percent of specimens with lymph node metastases is no longer significant. In this large database analysis, >85% of women with lymph node metastasis were identified after the removal of up to 25 lymph nodes. However, the yield of detecting lymph node disease decreases after resecting 30 lymph nodes.
Logistic regression models were developed to estimate the probability of identifying at least 1 positive lymph node as a function of the independent variable total number of lymph nodes identified in the lymphadenectomy specimens. These probabilities are expressed in ORs (Tables 2–4). In the overall study group, the removal of 21 to 25 lymph nodes resulted in a 45% increase (OR of 1.45; 95% confidence interval [95% CI], 1.08–1.94 [P < .01]) in the probability of detecting at least 1 positive lymph node relative to the likelihood of obtaining a positive lymph node when only 1 to 5 lymph nodes are removed (OR of 1). However, removing more than 25 lymph nodes did not appear to significantly increase the statistical probability of detecting at least 1 positive lymph node (see Table 2). The OR for >25 lymph nodes was lower than for 21 to 25 lymph nodes. However, the 95% CIs for these 2 groups overlap. Therefore, the probabilities for finding at least 1 positive lymph node are not significantly different from each other. Given that the risk of metastatic lymph node spread is relatively uncommon in grade 1 tumors, a separate analysis was performed excluding patients with grade 1 disease. Of patients with stage I to stage III, grade 2 to 3 cancers, the removal of 21 to 25 lymph nodes resulted in a 51% increase in the probability of detecting at least 1 positive lymph node relative to the likelihood of obtaining a positive lymph node when only 1 to 5 lymph nodes were removed (OR of 1.51; 95% CI, 1.11–2.06 [P < .01]) (see Table 3). In another subset analysis of patients with stage III disease only, the largest increase in the probability occurred as the number of lymph nodes examined increased from 6 to 10 to 11 to 15 lymph nodes (OR of 1.71; 95% CI,1.18–2.48 [P < .01]) (see Table 4). Nevertheless, the removal of 21 to 25 lymph nodes significantly enhanced the probability of identifying at least 1 positive lymph node by nearly 2-fold (OR of 1.99; 95% CI, 1.29–3.10 [P < .01]).
Table 2. Odds Ratio of Detecting at Least 1 Positive Lymph Node, Stages I-III (n = 10,998)
No. of patients
95% Confidence interval
Year of diagnosis is coded as 0 indicates 1988–1992, 1 indicates 1992–1997, and 2 indicates 1998–2001.
Grade is coded as 0 indicates grade 1, 1 indicates grade 2, and 2 indicates grade 3.
Lymphadenectomy is an important component of the surgical staging of uterine corpus cancer. The ability to identify lymph nodes with metastatic involvement or other poor prognostic factors is not only prognostic, but can direct postoperative care and potentially affect survival.2 However, to our knowledge, there is no consensus regarding the definition of an “adequate” lymph node dissection or the number of lymph nodes required to be removed to appropriately stage patients as having lymph node-negative disease.
Prior single-institutional studies have shown a survival advantage associated with the extent of lymphadenectomy.5, 7, 13 Conversely, others have questioned the therapeutic value of lymphadenectomy in uterine cancers even in randomized clinical trials.15, 16 Recently, Chan et al17 performed a review of 12,333 endometrioid uterine cancer patients to determine the potential therapeutic role of lymphadenectomy. In the intermediate/high-risk patients (stage IB, grade 3; stage IC and II-IV, all grades), a more extensive lymph node resection (1, 2–5, 6–10, 11–20, and >20 lymph nodes) was associated with improved 5-year disease-specific survivals across all 5 groups at 75.3%, 81.5%, 84.1%, 85.3%, and 86.8%, respectively (P < .001). However, their data did not show a survival advantage associated with the extent of lymphadenectomy in those with low-risk cancers defined as stage IA, all grades, and stage IB grades 1–2. Furthermore, to our knowledge, there are no guidelines to define an “adequate” lymph node dissection to determine lymph node status in uterine cancer. Although investigators claimed that removing 15 lymph nodes (10 pelvic and 5 para-aortic) is required for an adequate lymphadenectomy, this arbitrary number was based on a small study group without statistical modeling.18
Our analysis showed that a higher number of recovered lymph nodes was associated with a greater chance of detecting at least 1 positive lymph node metastasis. One of the obvious explanations is that recovering a higher number of lymph nodes decreases the chances of missing a metastatic lymph node.10 In addition, a more thorough sampling of lymph nodes that are located anatomically farther away from the uterus can improve the detection of skip metastases.18 Recovering a higher number of lymph nodes reduces the chance of missing a lymph node metastasis and allows for the accurate staging of uterine cancer patients. Accordingly, a lymph node sampling of fewer than 5 lymph nodes is likely to be inadequate to assure the surgeon or patient that there are no lymph node metastases. In fact, our data showed that only 27.6% of all patients with lymph node metastases were diagnosed when ≤5 lymph nodes were recovered. Conversely, it is unlikely that the removal of an additional 5 lymph nodes after resecting more than 40 lymph nodes is going to significantly improve the yield of finding a positive lymph node. Thus, there must exist a certain cutoff limit after which a further increase in the number of recovered lymph nodes will no longer affect the accuracy of staging but may contribute to higher surgical morbidity. In other words, there must be an estimated range of recovered lymph nodes that will assure the clinician to a reasonable degree of certainty that the patient does not have lymph node metastases without resecting additional lymph nodes.
In this current report a mathematical model was designed to investigate the association between the number of lymph nodes examined and probability of detecting at least a single lymph node metastasis. The validity of this statistical model requires a large number of patients with an extensive range of lymph nodes resected and sufficient percentage of patients with lymph node metastases. Using 11,443 patients from the National Cancer Registry, a logistic regression model was employed showing that removing 21 to 25 lymph nodes had a 45% increase in the probability of detecting at least 1 positive lymph node compared with removing only 1 to 5 lymph nodes. The number of lymph nodes calculated from this mathematical model is higher than the median number of lymph nodes reported from single institutions, ranging from 4 to 15 lymph nodes.5, 6, 13, 19
An established guideline that recommends the removal of more than 20 lymph nodes in all patients with endometrioid uterine cancer may pose a challenge for the surgeon as well as the pathologist. Furthermore, lymphadenectomy has been associated with an adverse event rate of 18%, including prolonged ileus or bowel obstruction (2.6%), deep venous thrombus (2.6%), and lymphocysts (2.4%) as the more common complications.7 For the pathologist, dissecting lymph nodes is laborious, particularly if the lymph nodes are small. Thus, the challenge may lie in the judicious use of resources including time, cost, and energy required to recover the appropriate number of lymph nodes. Nevertheless, it is clinically feasible to establish such a standard in endometrial cancer. In fact, greater than 20% of all patients in the current report had >20 lymph nodes resected. Moreover, it is likely that many of these patients in this current study did not receive surgical staging procedures by gynecologic or surgical oncologists. The clinical feasibility of resecting more than 20 lymph nodes may not be possible in the medically compromised or morbidly obese patients. Ideally, it would be useful to have a lymph node count at the time of surgery. Because pathologists typically do not identify the number of lymph nodes resected for frozen sections, it is not possible to know how many lymph nodes are being removed at the time of surgery. Therefore, we recommend that a complete lymphadenectomy be performed. Furthermore, in a recent surgical trial completed by the Gynecologic Oncology Group comparing open versus laparoscopic surgery for uterine cancer, the median lymph node count was greater than 20 lymph nodes.20 Thus, with recommended surgical guidelines advocating comprehensive lymphadenectomy, the removal of more than 20 lymph nodes is clinically feasible.
In the situation in which a patient has already undergone surgery but has not had an adequate number of lymph nodes identified, it is important to weigh the risks and benefits of additional surgery. In addition, the pathologist should be encouraged to review the macroscopic specimen to identify additional lymph nodes. Using Figure 2, the surgeon is able to evaluate the additional gain in finding a positive lymph node based on the number of lymph nodes identified. With the primary surgical specimen available, one can examine other histological factors such as depth, grade, lymphovascular space invasion, cell type, and tumor size—all factors that can potentially guide the surgeon on the overall benefit of a more extensive lymph node dissection.
To better determine the influence of grade and stage of disease on the estimated number of lymph nodes required to detect at least 1 positive lymph node, we restricted the analysis to grade 2 and grade 3 disease and found that the predictive value was not significantly different. This suggests that the comprehensive staging procedures remain important for patients with all grades of cancer. Similarly, a prior study on patients presenting with grade 1 endometrial cancers showed that lymphadenectomy impacted postoperative treatment decisions in 29% of these women.21 In our current study, a subset analysis was also performed on stage III patients. In this high-risk group of women only 11 to 15 lymph nodes were required to show a significant increase in the probability of detecting at least 1 lymph node metastasis. This subset analysis supports this mathematical model in that the threshold for the number of lymph nodes needed to find a positive lymph node is significantly lower in advanced-stage patients with high risk for lymph node metastasis. Nevertheless, the trend of the OR continues to increase from 1.71 (P<.01) for 11 to 15 lymph nodes, and up to 1.99 (P<.01) for removing 21 to 25 lymph nodes in patients with stage III disease.
The limitations of this study include the lack of information regarding the location and size of the lymph nodes resected, subsets of stage IIIA disease (eg, serosal extension, adnexal spread, or positive peritoneal cytology), specialty of the surgeons, and lack of uniformity in pathological processing. Furthermore, the number of lymph nodes resected can also be confounded by anatomic variations between patients,22–24 differences in the pathological processing and analysis,11, 25, 26 as well as the extent of surgical dissection.5, 6, 13, 19, 20 Although the American College of Surgeons have established that a minimum of 12 lymph nodes are required to define comprehensive colorectal cancer surgery, it remains to be determined if similar guidelines should be instituted for endometrioid uterine cancer. One of the potential differences lies in the quality of the lymph node specimen from uterine cancer compared with colorectal cancer surgery. In uterine cancer the lymph node containing tissues are submitted as multiple specimens after surgical dissection of fatty lymph node tissue from blood vessels and nerves. In contrast, colorectal surgical specimens are removed en bloc, in which the architecture of the lymph node chains are preserved, facilitating lymph node identification within the specimen. One of the other limitations associated with this analysis is the assumption that all lymph nodes are equivalent and interchangeable. Nevertheless, in a study of greater than 11,000 patients, one may question whether subtle differences such as anatomic variations, pathologic processing, and surgical technique can affect the overall results. Moreover, the strength of the data lies in the large sample size, which allows for mathematical modeling even within subgroups of patients. Furthermore, this study population reflects real-life practices combining patients from both community and academic centers, whereas previous studies report on selected, high-risk patients cared for by academicians based in tertiary care centers. As such, our group has proposed additional analyses of other databases to validate the models developed in this current study.
One of the surgical limitations of resecting 20 to 25 lymph nodes is that surgeons are not capable of knowing the number of lymph nodes resected during surgery. Thus, some centers have adopted a system to perform a thorough analysis on the tumor specimen and lymph nodes to determine the adequacy of the surgery during the operation. However, it is unlikely that nontertiary care centers or community hospitals will have the dedicated pathology staff and resources to perform definitive intraoperative analyses.6 Another potential surgical innovation that may assist the surgeon in identifying the lymph node status during surgery is the use of sentinel lymph node evaluation. This may enhance the surgeons' intraoperative diagnostic ability to accomplish adequate staging procedures with less extensive lymph node resections. However, the majority of these approaches are still under clinical investigation.
With the increase in the number of deaths per year associated with uterine cancer, advances in treatment are required in the management of this disease. In particular, the surgical management of uterine cancer needs to be standardized. Given the low median number of lymph nodes resected in the overall group, both surgeons and pathologists should aim to increase the number of lymph nodes recovered to assure a patient that she does not have lymph node metastases that warrant adjuvant therapy. If the estimation that greater than 20 lymph nodes are required to define “adequate” staging by this mathematical model is validated in the clinical setting, all uterine cancer patients should undergo consultation with a surgical oncologist with the training and expertise in performing comprehensive staging procedures, including extensive lymphadenectomy, when clinically feasible. Clearly, the definition of “adequate” lymph node dissection in the staging of uterine cancer demands further investigation.
We thank Alex McMillan and Philip Lavori for statistical assistance and support.