Survival effect of different lymph node staging methods on ovarian cancer: An analysis of 10 878 patients

Abstract Background To compare the survival impact of several lymph node staging methods and therapeutic role of lymphadenectomy in patients with epithelial ovarian cancer who had undergone lymphadenectomy. Methods Data were retrospectively collected from the Surveillance, Epidemiology, and End Results program between 1988 and 2013. Results An increasing number of resected lymph nodes (RLNs) was associated with a significant improvement in survival of FIGO stage II and III disease. However, for FIGO stage IV patients, better survival was not significantly associated with a more extensive lymphadenectomy. A higher lymph node ratio (LNR) and log odds of positive lymph nodes (LODDS) were associated with poorer survival regardless of stage. Nevertheless, four‐category classification of LODDS was more suitable for stage IV patients when three‐category classification was compatible with stage I‐III disease. Multivariate analysis demonstrated that LODDS and LNR were significant independent prognostic factors, but not RLN classification. Conclusion Sixteen to thirty RLNs are recommended for stage I disease. For stages II and III patients, the more lymph node excision, the better the prognosis. However, lymphadenectomy was nonessential for stage IV patients. Considering staging methods, for stages II and III patients, three‐category classification of LODDS was recommended to evaluate the prognosis. For stage I and IV, three‐category classification of positive LNR was idoneous.


| METHODS
Demographic, clinicopathologic, and survival information were generated from the Surveillance, Epidemiology, and End Results (SEER) database (http://seer.cancer.gov/). All pathologically confirmed and surgically treated EOC patients from 1 January 1988 to 31 December 2013 were identified. The following were the inclusion criteria: (a) Lymph nodes status were available, (b) pathology limited to epithelial ovarian cancer, (c) only one primary cancer in the patient's lifetime, (d) the primary site and morphology were ovarian, (e) survival time was available, and (f) grade was clear. A total of 10 878 patients were finally obtained. Patients were excluded if they had incomplete medical records, pathology, or outcome data.
Resected lymph nodes were used to evaluate lymphadenectomy extent, which were divided into six groups: one lymph node removed; two or three lymph nodes removed; four or five lymph nodes removed; six to 15 lymph nodes removed; 16-30 lymph nodes removed; and 31 or more lymph nodes removed. Lymph node status was characterized by LNR and LODDS. LNR is the ratio of positive lymph nodes (PLNs) to RLNs. LODDS is defined as the log of odds between PLNs and the number of negative nodes. LNR and LODDS were first divided into a lot of groups. Through analysis, we found that there was little difference among some classifications. We amalgamated these classifications. Finally, LNR was subdivided into three groups to meet proportionality assumptions: 0 to <10%, 10% to <40%, and 40%-100%. LODDS were grouped as follows: LODDS< −1.0, −1.0 ≤ LODDS<−0.5, −0.5 ≤ LODDS<0, LODDS≥0.
Survival was estimated by the Kaplan-Meier method and assessed by the log-rank test. OS was defined as the time between diagnosis and death due to any cause or censored at the last follow-up time if no events had occurred. Cancer-specific survival (CSS) was defined as the time between diagnosis and cancer-specific death. To investigate the significance of RLNs, LNR, and LODDS, multivariate analysis was performed using the Cox proportional hazards model after adjusting for other patient features, including age, race, tumor size, and grade in different FIGO stages. Analyses were performed using the SPSS statistical software package, version 19.0 (IBM Corporation, Armonk, NY, USA) and the R version 3.3.0 software (Institute for Statistics and Mathematics, Vienna, Austria; www.r-project.org). All tests were twotailed, and statistical significance remained conventionally defined as P < 0.05 in all other cases.
To evaluate the effect of the extent of lymph node dissection, our study cohort was divided into six groups: Patients who had 1, 2-3, 4-5, 6-15, 16-30, and ≥31 nodes were reported. We determined whether the influence of RLNs on CSS and OS was modified by FIGO stages. When FIGO stages were studied separately, the beneficial effect of lymphadenectomy with different extent was not observed for all FIGO stages (Table 2a; Figure 2). The results indicated that in FIGO stages II and III, an increasing numbers of lymph nodes were associated with a significant improvement in 5-y and 10-y CSS or OS. The findings of 1, 2-3, 4-5, 6-15, 16-30, and ≥31 lymph nodes were associated with 10-y OS of 37.1%, 41.3%, 63.5%, 60.9%, 67.7%, and 68.4% in FIGO II stage, and 17.6%, 22.4%, 23.0%, 28.0%, 29.9%, and 30.7% in FIGO III stage, respectively (P < 0.0001). However, for 3703 patients with stage I disease, the optimal survival rate was achieved in the 16-30 RLNs group, instead of the ≥31 group (OS: P = 0.0371; CSS: P = 0.0154). In 1658 EOC cases with stage IV, better OS was not found to be significantly associated with a more extensive lymphadenectomy (P = 0.0538).
The effects of LNR and LODDS are shown in Table 2b and Figure 3. The 5-and 10-y CSS and OS rates were found to be significantly decreased when more positive lymph nodes appeared within all FIGO stages. In an analysis of 5504 patients who had FIGO stages II and III disease, the 10-y OS was 61.2% for patients with an LNR of 0 to <10%, 53.1% for those with an LNR of 10% to <40%, and 18.5% for those with an LNR of 40%-100% (P < 0.0001). In FIGO II, the 10-y OS rates were 31.4%, 26.3%, and 16.4% for groups in which LNR was between 0 and 10%, between 10 and 40%, and more than 40%, respectively (P <0.0001). Notably, the 10-y OS rates in FIGO I were 81.0%, 68.1%, and 46.9% of three-category LNR (P <0.0001), which indicated positive lymph nodes, especially the 0%-10% and 10%-40% groups, were more important in FIGO I. Additionally, the effect of LODDS was investigated in subgroups of patients with FIGO stages I-IV disease ( Figure 4). Generally, a higher LODDS was significantly associated with poorer CSS (P < 0.0001) and OS (P < 0.0001). Nonetheless, differences exist in various FIGO stages. The four-category system is suitable for FIGO IV patients, with 10-y OS of 26.1%, 23.1%, 12.6%, and 9.9%, respectively. Three categories (LODDS<−1, −1 ≤ LODDS<0, and LODDS≥0) are compatible with FIGO stages I-III.     On multivariate analysis, LNR persisted as a significant and independent prognostic factor of OS (Table 3a) and CSS (Table 3b), regardless of the stage of the disease. The hazard ratios (HRs) of LNR on OS in FIGO stages I-IV were 2.569, 1.670, 1.234, and 1.259, respectively. LODDS was also a significant prognostic factor, which was associated with a poorer CSS and OS.
In contrast, the number of lymph nodes resected was a significant prognostic factor only for FIGO I-III on OS (HR: 0.895, 95% confidence interval and OS (HR: 0.966, 95% CI: 0.920-1.014, P = 0.165), suggesting that not all stages of patients could benefit from lymphadenectomy. In addition, multivariate analysis also indicated that age and grade were significant prognostic factors for OS (Table 3) and CSS (Table 4).

| DISCUSSION
Ovarian cancer is the most common cancer and has the highest death rate in female malignancy. Although surgery is definitely necessary in the treatment of EOC, the role of lymphadenectomy is controversial. In the present study, we investigated the prognostic value of RLNs, LNR, and LODDS in EOC patients with different disease stages. Our results showed that in different stages, the prognostic value of lymph node parameters was inconformity. In FIGO stage II and III, an increasing number of RLNs was associated with a significant improvement in survival. However, for stage I disease, the optimal survival rate was achieved in the 16-30 RLNS group, and in stage IV patients, a more extensive lymphadenectomy did not yield survival benefits. In all disease stages, a higher LNR and LODDS were significantly associated with poorer survival. Furthermore, the two-category system of LNR was suitable for FIGO stage II and III patients, and the three-category system was suitable for FIGO I and IV patients. Three LODDS categories (LODDS<−1, −1 ≤ LODDS<0, and LODDS≥0) were compatible with FIGO I-III disease. However, for FIGO IV patients, the four-category system (LODDS<−1, −1 ≤ LODDS<−0.5, −0.5 ≤ LODDS<0, and LODDS≥0) is more suitable. Multivariate analysis demonstrated that LODDS and LNR were significant independent prognostic factors regardless of the disease stage, but not RLN classification.
Compared to the previous FIGO staging system, the revised FIGO staging system utilizes PLNs to characterize stage III. 33,34 However, because PLNs is not reflex the information of negative lymph nodes, it is not appropriate to make predictions. LNR, combining positive lymph nodes and resected number, should be utilized to reduce the potential bias. 18 In stage IIIC ovarian cancer patients with lymphadenectomy, increasing LNR leads to significantly decreased OS, 18,35 which was similar to the results of our study. In our study, we demonstrated that, in FIGO stage III patients, LNR within 0 to ≤40% group had survival advantage over LNR more than 40% group. Furthermore, this two-category system is also suitable for stage II disease. However, due to the significant difference of survival between LNR>0 to ≤10% and LNR>10 to ≤40%, dividing LNR into three parts (0 to ≤10%, 10% to ≤40%, and >40%) was more suitable in FIGO stage IV. Because the positive nodes were more common in advanced-stage disease, studies about LNRs have always been focused on those patients. Therefore, the role of LNR in early-stage disease was ambiguous. Our research demonstrated that classifying LNR into three parts (0 to ≤10%, 10% to ≤40%, and >40%) was associated with decreased OS in FIGO stage I. Based on these findings, our study recommends the adaptation of the two-category system of LNR in FIGO stage II and III, and the three-category system in FIGO stage I and IV. Through this method, the survival status of patients can be predicted more accurately.
Log odds of positive lymph nodes, as an emerging indicator of lymph node status, has been studied in various cancers. [19][20][21][22][23] Little is known about the role of LODDS in ovarian cancer. Xu et al 24 found LODDS as an independent prognostic factor for predicting survival in patients with EOC regardless of the tumor stage. Although LODDS exhibits discriminatory ability in predicting survival, there is no study concerning the diverse role of LODDS in different FIGO stages. Our study demonstrated that in FIGO stage I to III patients, the three classification method (LODDS<−1, −1 ≤ LODDS<0, and LODDS≥0) could stratified patients clearly. However, in FIGO IV, classification of LODDS into four categories is more suitable for prediction.
The therapeutic value of lymphadenectomy in ovarian cancer remains controversial. In early stages, survival of patients with lymph node dissection was significantly better than that of patients without lymph node dissection, 36 whereas, Maggioni et al 37 reported that, in FIGO stage I-II patients, there was no difference between women undergoing a node sampling and those who underwent a lymphadenectomy. In our study, we found that lymphadenectomy had diverse effects on stage I and II disease. In FIGO stage I, a large number of resected lymph nodes did not represent a better outcome. The patients with [16][17][18][19][20][21][22][23][24][25][26][27][28][29][30] RLNs have the best prognosis. When RLNs numbered more than 30, the survival rate decreased. In FIGO stage II, the minimum number of RLNS for a good prognosis is 4, preferably up to 16. According to our study, in early-stage ovarian cancer, the goal of the extent of lymphadenectomy is to remove 16-30 lymph nodes. In advanced-stage lymphadenectomy, there were also several opinions. Pereira 38 reported that removing

0.000
The bold values present p < 0.05. more than 40 lymph nodes may yield survival benefit. Iwase, 39 however, indicated that systematic lymphadenectomy did not confer therapeutic benefits, which is similar to our results in FIGO stage IV. In FIGO III, we suggest that the number of RLNs should be more than 6.
In conclusion, our data suggest that the prognostic information provided by LNRs or LODDS should be deciphered according to patients' FIGO stage. Flexible application of LNR or LODDS in different FIGO stages to characterize ovarian cancer patients might predict outcomes more precisely. Furthermore, taking FIGO stage into consideration when carrying out lymphadenectomy may avoid a lack of benefit in patients with ovarian cancer, which can further contribute to the individualized clinical decision. To prevent the bias inherent to the retrospective methodology, the results of this study need to be confirmed in future prospective studies.