BACKGROUND: Previously proposed criteria for preoperatively identifying endometrial cancer patients at low risk for lymph node metastasis remain to be verified. For this purpose, a prospective, multicenter observational study was performed. METHODS: Eligible patients with histologically confirmed endometrial cancer underwent magnetic resonance imaging (MRI) and serum cancer antigen 125 (CA 125) testing before surgery. The following criteria were used to identify low-risk patients: 1) endometrioid-type cancer, 2) no evidence of deep myometrial invasion on MRI, 3) no enlarged lymph nodes on MRI, 4) no suspicious metastasis out of the uterine corpus, and 5) serum CA 125 levels less than 35 U/mL. Systematic pelvic and/or para-aortic lymphadenectomy was performed for all patients. The primary endpoint was estimation of the negative predictive value (NPV). RESULTS: From January 2012 to December 2014, 529 patients from 20 hospitals in 3 Asian countries were consecutively enrolled. According to our criteria, 272 patients (51.4%) were categorized into the low-risk group. Fifty-three of the 529 patients (10.0%) had lymph node metastases; these patients included 8 (2.9%) falsely categorized as low-risk. The sensitivity and specificity of the criteria were 84.9% and 55.5%, respectively. The NPV of 97.1% was higher than the predefined target endpoint of 96%. CONCLUSIONS: The low-risk criteria based on preoperative tests were confirmed to be reliable and accurate for identifying patients at low risk for lymph node metastasis. These criteria may facilitate patient counseling and surgical decision making. Cancer 2017;123:263–272.
Bibliographical noteFunding Information:
The National Cancer Center of South Korea sponsored this study (grant 1410060-2). Sokbom Kang and the Korean Gynecologic Oncology Group were solely responsible for the decision to submit the manuscript for publication. This study met its predetermined endpoint for the validation of our preoperative low-risk criteria for lymph node metastasis. To our knowledge, this is the first international, multicenter, prospective study showing that preoperative tests can predict low-risk patient groups with clinically acceptable accuracy. The main strengths of our study are the prospective recruitment and the multicenter design. The accuracy of our risk criteria has been tested in multiple independent cohorts (Fig.). The first external validation was performed simultaneously during the development of our risk model. Our criteria classified 43% of 180 patients as low-risk with a 98.6% NPV. The second and third external validations were performed in 2 independent Japanese cohorts. In one cohort, our criteria identified 42% of 137 patients as low-risk with a 96.5% NPV. The other cohort included 181 patients for whom the NPV according to our criteria was 99.0%. The hierarchical summary receiver operator characteristics curve indicated that 91% sensitivity and 54% specificity were obtained when these criteria were used for a pooled cohort of 1387 patients; this yielded an NPV of 98.2% at an assumed prevalence of 10% (Fig.). HSROC curve for the diagnostic performance of the KGOG preoperative criteria identifying patients at low risk for lymph node metastasis from the data pooled from the current study as well as previous studies (green line). Each circle represents a study; the size of each circle is proportional to the sample size of each study. The summary point (filled box) represents the summarized specificity and sensitivity (91% and 54%, respectively). The 90% confidence intervals for those operating points are shown as a dotted line. HSROC indicates hierarchical summary receiver operating characteristic; KGOG, Korean Gynecologic Oncology Group; ROC, receiver operating characteristic. Identifying endometrial cancer patients at low risk for lymph node metastasis is not a new concept. In 1987, researchers from the Gynecologic Oncology Group addressed the prognostic importance of lymph node status in endometrial cancer by analyzing more than 500 patient records; this prompted lymphadenectomy to be included during routine surgical staging. They characterized a low-risk patient as having less than a 5% probability of lymph node metastasis according to postoperative pathologic findings. Although the authors suggested that lymphadenectomy is not necessary for low-risk patients, the recommendation was largely ignored. However, since 2 randomized trials indicated no survival benefit from routine lymphadenectomy, the frequency of lymphadenectomy has been decreasing annually, and some practice guidelines now allow the omission of lymphadenectomy with caution, particularly for low-risk patients. Our study criteria produced a superior diagnostic odds ratio, area under the receiver operator characteristic curve, and NPV with a specificity similar to that for the ESMO criteria and a sensitivity similar to that for the SGO criteria. Therefore, we can recommend using the KGOG low-risk criteria for selective lymphadenectomy when MRI and serum CA 125 level data are readily available. Otherwise, the ESMO low-risk criteria may be a good alternative because of their better reliability in comparison with the SGO criteria. However, because we used MRI data to evaluate the performance of the SGO and ESMO guidelines, it should be noted that using intraoperative examination data might result in changes to such a comparison. In addition, because no cost-effectiveness analysis is provided among those algorithms, the choice of the optimal tool for evaluating the risk may remain with gynecologic oncologists. Our risk criteria are useful not only for advocates of selective lymphadenectomy but also for those who support routine lymphadenectomy for low-risk patients. There have been increasing demands for surgeons to help patients make informed decisions and to provide necessary information regarding treatment options. Although there is no level 1 evidence showing a survival benefit of lymphadenectomy, a population-based study indicated that the morbidity of lymphadenectomy is as serious as that of radiotherapy. Therefore, such information should clearly be made available to patients before surgery, and patients should be encouraged to actively participate in decision making. The major limitation of our study is our inability to determine whether the survival outcome could be compromised if routine lymphadenectomy is not offered to the population classified as low-risk by our criteria. However, such a trial would require a much larger number of patients and extensive resources. In a mathematical model, the difference in the predicted 3-year survival between hysterectomy without lymphadenectomy followed by radiotherapy for high-risk patients and staging followed by selective radiotherapy was estimated to be only 1% if the nodal metastasis rate was 4%. Therefore, investing resources in a trial to investigate the survival benefit of lymphadenectomy may be economically justifiable only for high-risk patients. Another issue is the significant discrepancy in surgical outcomes between low- and high-risk groups. In patients with poorer prognostic features, PALND was performed less frequently, and minimally invasive surgery was performed more frequently. Thus, there is a concern that these discrepancies could have interfered with the results of the current study. However, when we compared the frequencies of false-negatives according to such discrepancies among patients identified as low-risk, no significant difference was detected (Supporting Table [see online supporting information]). The frequencies of false-negatives were 2.8% and 3.2% for the PALND and no-PALND groups, respectively, and 3.0% and 2.4% for the laparoscopy and open-surgery groups, respectively. If we assume that the diagnostic performance shown in the PALND subgroup is true (90.9% sensitivity and 47.6% specificity), we can estimate the NPV in the entire population with Bayes' theorem. Because the prevalence of isolated para-aortic nodal metastasis is approximately 1%, regardless of patients' risk, the prevalence of nodal metastasis can be estimated to be 10.6%. At this rate, the calculated NPV was 97.8% (95% CI, 93.7%-99.2%), which was not inferior to the NPV (97.1%) in this study. The final issue is whether our risk criteria can be improved by the incorporation of several well-known risk factors such as tumor grade and size. However, our data did not support this notion (Table). By incorporating the tumor grade, we improved the sensitivity by 3.8% at the expense of an 11.2% loss of specificity; this meant that 10% of the patients would require unnecessary lymphadenectomy to detect 0.4% of patients with occult lymph node metastases. When the tumor size was incorporated as a predictive factor, the specificity was decreased by 31.2% with no NPV improvement. The data show that overfitting due to excessive predictors reduced the specificity of the risk model, and this eventually would impair the cost-effectiveness of selective lymphadenectomy. The diagnostic performances of our criteria among the subgroups according to these risk factors are summarized in Supporting Table (see online supporting information). In conclusion, endometrial cancer patients with a low risk of lymph node metastasis can be identified preoperatively with the KGOG criteria. We found the accuracy of the KGOG low-risk criteria to be close to the accuracy of postoperative pathologic data. The KGOG risk criteria will facilitate presurgical counseling and treatment decision making. We enrolled 537 consecutive patients with endometrial cancer. After the exclusion of 8 patients who did not meet the inclusion/exclusion criteria, 529 were included in the analysis (Fig. B). The overall prevalence of lymph node metastasis in the study population was 10.0% (53 of 529); this was similar to other studies. The median number of harvested lymph nodes per patient was 23. According to the KGOG low-risk criteria, 272 patients (51.4%) were classified as low-risk. Among those 272 patients, no gross abdominal metastasis was detected during surgery. The clinical characteristics, preoperative assessment outcomes, and surgical outcomes are summarized for each risk group in Table. Compared with the non–low-risk group, the KGOG low-risk group was slightly but significantly younger (P =.013) and had fewer grade I or II tumors (P <.001). Furthermore, low-risk patients were more frequently diagnosed by dilatation and curettage than other methods (P =.012). The rate of lymph node metastasis was 2.9% and 17.5% in the low-risk and non–low-risk groups, respectively (P <.001). As expected, all pathologic factors indicating a poor prognosis were less frequent in the low-risk group. Minimally invasive surgeries, such as laparoscopic surgery and robotic surgery, were more frequently performed in the low-risk group (P =.012). The decision to undergo minimally invasive surgery was significantly associated with various MRI features showing the anatomical extent of disease, such as myometrial invasion (P =.004), extrauterine tumor extension (P <.001), and nodal enlargement (P <.001), but not age or histologic features. Moreover, surgeons harvested fewer lymph nodes (P =.001) and performed PALND less often (P <.001) in patients identified as low-risk. Abbreviation: CA 125, cancer antigen 125; FIGO, International Federation of Gynecology and Obstetrics; KGOG, Korean Gynecologic Oncology Group; MRI, magnetic resonance imaging. Data from preoperative tests. Data from postoperative pathology reports. The sensitivity and specificity of the KGOG risk criteria were 84.9% (95% confidence interval [CI], 72.4%-93.3%) and 55.5% (95% CI, 50.9%-60.0%), respectively; they produced an NPV of 97.1% (95% CI, 94.3%-98.7%). Other performance indicators, such as the area under the receiver operator characteristic curve and the diagnostic odds ratio, are summarized in Table. In addition, the diagnostic performance of the KGOG risk criteria was compared with the performance of 2 alternative sets of criteria from the SGO and ESMO low-risk guidelines (Table). The KGOG criteria showed a higher diagnostic odds ratio (7.0) than the SGO and ESMO criteria (3.7 and 4.1, respectively). The sensitivity of the SGO criteria was superior (90.6%); it was significantly higher than the sensitivity of the ESMO criteria (75.5%, P =.009) but was not significantly higher than the sensitivity of the KGOG criteria (84.9%, P =.38). The ESMO criteria showed the highest specificity (57.4%), but this was not significantly different from the specificity of the KGOG criteria (55.5%, P =.40). However, the SGO criteria showed significantly lower specificity (27.9%) than the KGOG and ESMO criteria (P <.0001 for both). The NPV from the KGOG criteria was higher (97.1%) than the NPVs of the SGO and ESMO criteria (96.4% and 95.5%, respectively). Finally, we compared the preoperative KGOG risk criteria with the postoperative criteria. The KGOG criteria were compared with the conventional low-risk criteria using postoperative pathology data; the low-risk group comprised patients with an endometrioid-type histology, grade I/II disease, and a myometrial invasion depth less than 50%. There was no difference in sensitivity between the 2 criteria (P = 1.0); however, the postoperative pathology-based criteria showed better specificity (67.2%) than the KGOG criteria (55.5%, P <.001). The NPV was thus 97.1% for our criteria, which was very close to that of the postoperative pathology-based criteria (97.6%). Abbreviations: CI, confidence interval; ESMO, European Society of Medical Oncology; KGOG, Korean Gynecologic Oncology Group; ROC, receiver operator characteristic; SGO, Society of Gynecologic Oncology. The 2 sets of criteria were generated from SGO guidelines (endometrioid type, grade I/II, myometrial invasion < 1/2, and tumor size < 2 cm) and ESMO guidelines (endometrioid type, grade I/II, and myometrial invasion < 1/2), respectively. Predictions were based on preoperative data such as magnetic resonance imaging and preoperative biopsy, except for tumor size. Predictions were based on postoperative pathology. Pathologic data from surgical specimens were used. We tested whether including other risk factors could improve the diagnostic performance of the KGOG criteria (Table). First, we integrated an unfavorable tumor grade (III or not determined) as an additional risk factor; the sensitivity increased from 84.9% to 88.7% without statistical significance (exact McNemar test, P =.50), whereas the specificity was significantly reduced from 55.5% to 44.3% (P <.0001). Second, we integrated the tumor size (≥2 cm). This increased the sensitivity slightly from 84.9% to 90.6% (P =.25) but decreased the specificity significantly from 55.5% to 28.8% (P <.0001). Abbreviations: CI, confidence interval; KGOG, Korean Gynecologic Oncology Group; ROC, receiver operator characteristic. The exact McNemar test was applied. The characteristics of the 8 patients who were falsely classified as being in the low-risk group are summarized in Table. One had grade III disease, and another had deep myometrial invasion in the final pathologic specimen; none had both features simultaneously. In the low-risk group, various clinical variables that were not included in the criteria were tested for a possible association with lymph node metastasis (Supporting Table [see online supporting information]); none were significantly associated in the univariate analysis. Abbreviation: LVSI, lymphovascular space invasion. Without lymphadenectomy, 24 of 272 low-risk patients (8.8%) would have been found to be high-risk patients after surgery on the basis of their postoperative pathologic outcomes: 5 would have been reallocated to a high-risk group because of grade III disease, 18 would have been reallocated because of stage IB disease, and 1 would have been reallocated because of stage II disease. The rate of lymph node metastasis for the 24 patients was 8.3%. As for the 248 patients who would have remained in the low-risk group after surgery, the rate of lymph node metastasis was 2.4%. In comparison, 10.9% (15 of 138) and 13.6% (39 of 286) of the patients classified as low-risk by the SGO and ESMO guidelines, respectively, would have been found to be high-risk patients after surgery. Surgical staging was required no later than 8 weeks after MRI. All patients were classified according to our previously published KGOG preoperative risk criteria (Fig. A). Patients with no positive predictors (endometrioid-type histology, no deep myometrial invasion, no enlarged lymph nodes, no suspicious extrauterine spread, and serum CA 125 levels less than 35 U/mL) were considered low-risk. All patients underwent surgical staging, including systemic pelvic and/or para-aortic lymphadenectomy (PALND), regardless of the assessment outcome. There were no restrictions on surgical routes, such as transabdominal, laparoscopic, and robotic approaches. Peritoneal washing cytology was not mandatory. Systemic lymphadenectomy in both the pelvic and para-aortic regions up to the level of the renal vessels was recommended. However, PALND was left to the discretion of the surgeons. If a gross abdominal metastasis was discovered during surgery, the patient was classified as high-risk, regardless of the preoperative assessment outcome. All the harvested lymph nodes were stained with hematoxylin and eosin. Tissue specimens were obtained from all patients and were locally reviewed by gynecologic pathologists; no central review was performed to reflect the real-world variation of expertise across hospitals. (A) Criteria for determining a low risk of lymph node metastasis in endometrial cancer as proposed by KGOG and (B) Statement for Reporting Studies of Diagnostic Accuracy diagram reporting the flow of participants through the study. CA 125 indicates cancer antigen 125; KGOG, Korean Gynecologic Oncology Group; MR, magnetic resonance. Between January 2012, and December 2014, 529 consecutive patients with biopsy-confirmed endometrial cancer were enrolled into a prospective, multicenter cohort study from 20 hospitals in 3 countries: Korea, Japan, and China. Approval from local institutional review boards was obtained for all centers. All participating centers were tertiary hospitals that regularly performed surgical care for endometrial cancer and had multidisciplinary teams including specialized gynecologic oncologists, radiologists, and pathologists. Participating institutions registered all eligible patients for surgical staging and performed MRI and serum CA 125 testing before surgery. Enrolled patients, all of whom provided written informed consent, underwent systemic lymphadenectomy. Patients with histologic features suggesting squamous cell carcinoma or carcinosarcoma on preoperative biopsy were excluded. This study is registered at ClinicalTrials.gov (NCT01527396). Before surgery, all patients underwent endometrial biopsy. Tissue specimens were obtained via endometrial curettage, hysteroscopic biopsy, or pipelle biopsy; suction or brush biopsy was not allowed. No central review was performed to allow interobserver variability. Patients underwent MRI of the pelvis with a body coil; images were obtained with a 1.5- or 3.0-T magnet after gadolinium enhancement. At all the centers, MRI was routinely used for preoperative workup in endometrial cancer patients. To reflect real-world variation in clinical settings, no specific protocol for MRI imaging was required. Lymph nodes with short axes longer than 1 cm were considered enlarged. Serum CA 125 levels were determined by radioimmunoassay with the preferred assaying methods of each institution. Biopsy and MRI results were assessed by evaluators blinded to nodal involvement data. However, surgeons were not blinded to preoperative test results. All of the tests were performed within 8 weeks of surgical staging. Surgical staging was required no later than 8 weeks after MRI. All patients were classified according to our previously published KGOG preoperative risk criteria (Fig. A). Patients with no positive predictors (endometrioid-type histology, no deep myometrial invasion, no enlarged lymph nodes, no suspicious extrauterine spread, and serum CA 125 levels less than 35 U/mL) were considered low-risk. All patients underwent surgical staging, including systemic pelvic and/or para-aortic lymphadenectomy (PALND), regardless of the assessment outcome. There were no restrictions on surgical routes, such as transabdominal, laparoscopic, and robotic approaches. Peritoneal washing cytology was not mandatory. Systemic lymphadenectomy in both the pelvic and para-aortic regions up to the level of the renal vessels was recommended. However, PALND was left to the discretion of the surgeons. If a gross abdominal metastasis was discovered during surgery, the patient was classified as high-risk, regardless of the preoperative assessment outcome. All the harvested lymph nodes were stained with hematoxylin and eosin. Tissue specimens were obtained from all patients and were locally reviewed by gynecologic pathologists; no central review was performed to reflect the real-world variation of expertise across hospitals. (A) Criteria for determining a low risk of lymph node metastasis in endometrial cancer as proposed by KGOG and (B) Statement for Reporting Studies of Diagnostic Accuracy diagram reporting the flow of participants through the study. CA 125 indicates cancer antigen 125; KGOG, Korean Gynecologic Oncology Group; MR, magnetic resonance. Our criteria were developed on the basis of a logistic risk model and were designed to identify patients with a less than 4% probability of having lymph node metastasis. Therefore, the predefined endpoint of the study was a false negative predictive value (NPV), which was calculated as 1 − NPV, of less than 4%. The sample size was determined on the basis of an estimated lymph node metastasis prevalence of 10% in a population with suspected early-stage endometrial cancer; hence, a cohort of at least 513 patients was targeted to achieve at least 80% power. Patients not classified as low-risk by the KGOG criteria were deemed to be true-positives if they were pathologically confirmed to have a lymph node metastasis after surgery. Sensitivity was defined as the ratio of true-positive results to the total positive samples obtained. Specificity was defined as the number of true-negatives divided by the sum of the true-negatives and false-positives. The diagnostic odds ratio was calculated as follows: (sensitivity/[1 − specificity])/([1 − specificity]/sensitivity). In addition, low-risk criteria from 2 other practice guidelines, the Society of Gynecologic Oncology (SGO) guidelines (endometrioid type, grade I/II, myometrial invasion < 1/2, and tumor size < 2 cm) and the European Society of Medical Oncology (ESMO) guidelines (endometrioid type, grade I/II, and myometrial invasion < 1/2) were compared with the KGOG criteria with the McNemar test. We assumed that the tumor grade and the myometrial invasion depth were determined by MRI and biopsy before surgery because neither guideline specified the methods used to measure these parameters. The tumor size was extracted from postoperative pathologic reports. All statistical analyses were performed with Stata 11.0 (StataCorp, College Station, Texas). All P values less than.05 were considered significant. Endometrial cancer is the most common gynecological cancer and the fourth most common cancer overall in developed countries. Since systemic pelvic lymphadenectomy was endorsed as a staging procedure for the surgical treatment of endometrial cancer, its role has been greatly debated among gynecologic oncologists. The arguments raised against routine lymphadenectomy are based on 2 factors: the absence of evidence showing improved survival due to the procedure and the compromised quality of life for patients who undergo lymphadenectomy. However, many still advocate its routine use because it can provide prognostic information and allow tailoring of adjuvant therapy. A substantial proportion of endometrial cancer patients do not have high-risk features such as deep myometrial invasion or a higher tumor grade; they are called low-risk. Because this population has a <5% chance of developing lymph node metastasis overall and seldom requires adjuvant treatment, there is an approaching consensus that routine lymphadenectomy is not necessary in these patients. Indeed, many practice guidelines are beginning to allow omitting lymphadenectomy in low-risk populations. However, because many risk factors used for determining risk are obtained postoperatively, risk assessments using surrogate risk factors from preoperative or intraoperative examinations have been studied. Nevertheless, a preoperative assessment of risk factors such as myometrial invasion and tumor grade is believed to be not as reliable as a postoperative assessment. To demonstrate the reliability and accuracy of a preoperative assessment of the risk of lymph node metastasis, the Korean Gynecologic Oncology Group (KGOG) recently performed a multicenter study (KGOG-2014) in which we proposed a risk-prediction model based on data from preoperative procedures such as magnetic resonance imaging (MRI), biopsy, and serum cancer antigen 125 (CA 125) testing. With this model, criteria were developed for discriminating patients whose probability of lymph node metastasis is less than 4%. Moreover, we successfully validated the model's performance in several independent cohorts. In addition, a cost-effectiveness analysis indicated that selective lymphadenectomy using the KGOG criteria would be less costly ($6454 vs $7079 in Korea and $23,995 vs $26,318 in the United States) and more effective (6.91 vs 6.85 quality-adjusted life-years [QALYs] in Korea and 6.87 vs 6.81 QALYs in the United States) than routine lymphadenectomy in both Korea and the United States. To eliminate a verification bias and test the model's reliability when variability of the testing method or observer is allowed, the KGOG performed a prospective, international observation study: the Preoperative Assessment of Lymph Node Metastasis in Endometrial Cancer study.
© 2016 American Cancer Society
All Science Journal Classification (ASJC) codes
- Cancer Research