Ovarian cancer is one the highest causes of cancer deaths in women in Western countries.1 Treatment of most ovarian cancers consists of surgery and chemotherapy. Primary cytoreductive surgery is the cornerstone treatment of ovarian cancer, and is a major prognostic factor.2 Ovarian cancer principally disseminates by two routes: transcoelomic spread and through lymphatics. Since 1986, FIGO classification takes into account these two routes: stage-IIIC disease includes patients with peritoneal implants larger than 2 cm outside the pelvis, and/or metastatic retroperitoneal lymph nodes.3 If maximal cytoreduction of peritoneal disease is widely accepted as a major determinant of survival, the therapeutic value of systematic lymphadenectomy in women with ovarian cancer remains controversial, even if it has been shown to be feasible and safe. Retrospective studies have suggested a significant survival advantage following systematic lymphadenectomy;4,5 however, a prospective randomised clinical trial demonstrated a progression-free survival benefit, but no overall survival benefit, when systematic lymphadenectomy was compared with a selected lymphadenectomy limited to palpable bulky nodes.6 Additional data are thus needed to explore the value of lymphadenectomy in patients with all-stage ovarian cancer. Recently, Chan et al. have performed two large population-based studies to investigate the role of lymphadenectomy in patients with advanced-stage ovarian cancers. Their studies analysed the disease-specific survival outcomes of 6686 patients diagnosed with International Federation of Gynecology and Obstetrics (FIGO) stage-I ovarian cancer, and 13 918 patients diagnosed with advanced-stage epithelial ovarian cancer, to demonstrate the potential role of lymph node dissection.7,8 Because the two analyses were separated in two different publications, the staging effect of lymphadenectomy can be hardly distinguished from the therapeutic effect. Moreover, residual disease is an important prognostic factor,2 and the effect of lymphadenectomy should be studied with respect to this variable or a marker of maximal surgical effort. Therefore, we conducted a population-based study to analyse the therapeutic effect of lymphadenectomy on survival after adjusting for disease stage and extent of surgery.
Please cite this paper as: Rouzier R, Bergzoll C, Brun J, Dubernard G, Selle F, Uzan S, Pomel C, Daraï E. The role of lymph node resection in ovarian cancer: analysis of the surveillance, epidemiology, and end results (SEER) database. BJOG 2010;117:1451–1458.
Objective The therapeutic role of lymphadenectomy on the survival in patients with ovarian cancer is controversial. The aim of this study was to evaluate the survival impact of lymphadenectomy, depending on the disease stage and extent of the surgery.
Design The surveillance, epidemiology, and end results (SEER) registry provided ovarian cancer data from 17 registries.
Setting Surveillance, Epidemiology, and End Results database.
Population The study population comprised 49 783 patients.
Methods Survival was studied according to the number of lymph nodes removed, with stratifications on disease stage and extent of surgery.
Main outcome measure The 5-year cause-specific survival rate.
Results The median follow up for patients alive at the last follow-up visit was 39 months. The five-year cause-specific survival rates were 37, 62, and 71% for the groups in which no lymph nodes were examined, in which between one and nine nodes were examined, and in which ten or more nodes were examined, respectively (P < 0.001). Avoiding lymphadenectomy was deleterious in all stages of the disease. It was maximal for International Federation of Gynecology and Obstetrics (FIGO) stage-II patients, but there was no significant interaction between stage and extent of lymphadenectomy. The cause-specific survival was found to significantly increase when more nodes were resected, even if the surgical procedure consisted of debulking surgery or a pelvic exenteration.
Conclusion Our study suggests a beneficial effect of lymphadenectomy in epithelial ovarian tumours, regardless of the stage of disease and extent of surgery. However, potential biases inherent to this retrospective methodology, such as staging migration, defining the extent of residual disease, and the possibility that thorough lymphadenectomy may reflect the quality of cytoreductive surgery, preclude any formal conclusions on the therapeutic role of lymphadenectomy.
Using the surveillance, epidemiology, and end results (SEER) registry (http://seer.cancer.gov) public use data tapes provided from 17 registries (Alaska Natives, Atlanta, Connecticut, Detroit, Hawaii, Iowa, Los Angeles, New Mexico, San Francisco–Oakland, San Jose–Monterey, Seattle–Puget Sound, Rural Georgia, Utah, Greater California, Kentucky, Louisiana, and New Jersey), we identified patients diagnosed with ovarian carcinoma from 1 January 1988 to 31 December 2004. The SEER registry includes patients from 1973, but information concerning the disease extent (nodes) and type of surgery is unavailable for patients managed before 1988.
Case listings were generated using codes specific for primary organ site, extent of disease, age at diagnosis, year of diagnosis, survival in months, and cause of death. Lymphadenectomy extent was characterised by the number of lymph nodes removed, and lymph node status was characterised by the number of metastatic nodes reported. Analysis was restricted to patients with information available concerning the number of lymph nodes removed (code < 95).
Patients were grouped based on the extent of surgery: hysterectomy versus no hysterectomy, omentectomy versus no omentectomy, and debulking or exenteration versus no debulking or exenteration. Four groups were generated for the multivariate analysis (no hysterectomy, hysterectomy, omentectomy, and debulking or exenteration). In the SEER database, debulking or exenteration were defined as the partial or total removal of the tumour mass, and could involve removal at multiple organ sites. We considered that debulking surgery or exenteration correspond to an extensive ovarian and peritoneal cytoreduction. To study the effect of the extent of lymphadenectomy, we used the same cut-offs as Chang et al.7,8, and divided the cohort into three groups: no lymph nodes removed; between one and nine (limited lymphadenectomy) lymph nodes removed; and ten or more lymph nodes removed (extensive lymphadenectomy).
Overall survival (OS), cause-specific survival (CSS), and other cause survival (OCS) were measured by SEER as the time from diagnosis to date of death, and date of last follow up, respectively. We used the Kaplan–Meier product limit method to describe OS, CSS, and OCS, and the log-rank test to assess differences between patient groups. Survival was studied according to the number of lymph nodes resected, with stratification on disease stage and the extent of surgery. The multivariate survival analysis was performed using a Cox proportional hazard model. The multivariate model included age (<50 versus >50 years), FIGO stage, histological grade and cell type, extent of surgery, number of resected lymph nodes, and lymph node status. All tests were two tailed, and P < 0.05 was considered to denote a significant difference. All analyses were performed using the R package with the Design, Hmisc, and Survival libraries (http://lib.stat.cmu.edu/R/CRAN).
We identified 54 838 patients with ovarian cancer during the study period. Information on the number of lymph nodes harvested was unavailable for 5055 patients. Therefore, the study population comprised 49 783 patients. The mean number of lymph nodes removed was 3.7 (standard deviation, 8.2). No lymph nodes were harvested in 33 980 patients (62%). The median patient age was 64 years (mean, 63 years; standard deviation, 16 years). Table 1 summarises the patient characteristics according to the number of nodes removed: most patients were white. Patients less than 50 years of age were significantly more likely to have a lymphadenectomy: 25 and 23% had between one and nine nodes, and ten or more nodes, resected, respectively. In patients older than 50 years, 16 and 12% had between one and nine nodes, and ten or more nodes, resected, respectively. Other differences between the populations of women who underwent lymphadenectomy and those who did not are reported in Table 1.
|No lymphadenectomy||Lymphadenectomy 1–9 nodes||Lymphadenectomy >10 nodes||P|
|>50 years||28 353||72||6195||16||4566||12|
|Site of treatment|
|sex cord stromal||511||63||168||21||130||16|
The median follow up was 39 months (range, 0–203 months). In the whole population, the 5- and 10-year overall survival rates were 38 and 28%, respectively. CSS rates were 46 and 38% at 5 and 10 years, respectively. The 5-year CSS rates were 37, 62, and 71% for the groups in which no lymph nodes were examined, in which between one and nine nodes were examined, and in which ten or more nodes were examined, respectively (P < 0.001). When restricted to epithelial tumours (including clear cell; n = 46 278; Figure 1), the 5-year CSS rates were 34, 61, and 70% for the groups in which no lymph nodes were examined, in which between one and nine nodes were examine, and in which ten or more nodes were examined, respectively (P < 0.001).
We also studied OCS to evaluate whether the lymphadenectomised and non-lymphadenectomised women were at similar risk of death from other causes than ovarian cancer. Among the deaths, 23, 20, and 18% were caused by disease other than ovarian cancer for the groups in which no lymph nodes were examined, in which between one and nine nodes were examined, and in which ten or more nodes were examined, respectively (P < 0.001). This demonstrates the patient selection for lymphadenectomy in the SEER database. Nonetheless, the difference of survival between groups observed for OCS (13%) was statistically different from the difference for CSS (36%) (P < 0.001). Therefore, even if groups with and without lymphadenectomy are different, the value of lymphadenectomy can be considered as independent of patient health status.
As lymphadenectomy is aimed to upstage I–IIIB disease to IIIC disease, the extent of lymphadenectomy in these patients may reflect the quality of staging. When FIGO stages were studied separately, the beneficial effect of lymphadenectomy was observed for all FIGO stages (Figure 2). The deleterious effect of the omission of lymphadenectomy was more pronounced in patients with FIGO stage II compared with stage I (Figure 2). The survival difference between the groups in which between one and nine nodes were examined and in which ten or more nodes were examined was higher in patients with FIGO stage-IIIA and -IIIB disease. These results suggest that the detrimental effect related to the omission of lymphadenectomy was higher in patients with a higher risk of nodal involvement, namely patients with a higher disease stage.
We also studied CSS with a classification based on the extent of surgery. This is of particular interest in patients with locally advanced disease stages, where the cytoreduction and the extent of surgery may influence the outcome. The 5- and 10-year CSS rates were found to increase when more nodes were resected, regardless of the extent of surgery (Figure 3). The 5-year CSS rate was 80.4% in patients with more than ten nodes resected, compared with 24.5% for patients with no lymphadenectomy, for the group of patients with no omentectomy performed (P < .001). In the omentectomy group, the CSS rate was 64.4% in patients with more than ten nodes resected versus 32.3% in the no nodes resected group (P < 0.001). The 5-year CSS rate was 78.4% in patients with more than ten nodes resected compared with 30.9% in patients with no lymphadenectomy for patients with no debulking/exenteration needed, or performed (P < 0.001). In the debulking/exenteration group, the CSS was 44.6% higher in patients with more than ten nodes resected, compared with 22.5% in patients with no lymphadenectomy (P < 0.001). Patients with between one and nine nodes resected always had intermediate survival. However, the prognostic difference was less pronounced in patients that underwent a debulking or an exenteration.
We analysed CSS in each FIGO stage according to histologic type and lymphadenectomy (Table 2). Results favoured lymphadenectomy in epithelial tumour whatever the stage of the disease. For clear cell and sarcoma, there was a significant survival benefit in patients with stage-III and -IV disease. It fell short of reaching statistical significance in stage-II clear cell carcinoma. Concerning sex cord stromal and germ cells tumours, no clear significant survival benefit was observed in patients who underwent lymphadenectomy.
|no lymph.||Lymph.||P||no lymph.||Lymph.||P||no lymph.||Lymph.||P||no lymph.||Lymph.||P|
|Sex cord stromal|
When analysed in a multivariate model, the number of lymph nodes resected was a significant prognostic factor (Table 3). An interaction term between the extent of lymphadenectomy and FIGO stage did not significantly improve the survival model (P = 0.3), suggesting that patients benefit from lymphadenectomy regardless of the stage of the disease. In contrast, the interaction between the extent of lymphadenectomy and the extent of surgery significantly improved the survival model. The extent of lymphadenectomy improved survival to a lesser degree when an extensive ovarian and peritoneal cytoreduction was performed (P < 0.001) (Figure 3; Table 3).
|age < 50 years||1|
|age > 50 years||1.32||1.25–1.39||<0.0001|
|sex cord stromal||0.84||0.62–1.14||0.26|
|Extent of surgery|
|Extent of lymphadenectomy|
|>10 nodes resected||1|
|1–9 nodes resected||1.20||1.11–1.29||<0.0001|
|no nodes resected||2.13||1.97–2.31||<0.0001|
|no positive nodes|
In this study, we demonstrate that there is an association between lymphadenectomy and survival in patients with ovarian cancer, whatever the stage of the disease and the extent of surgery. This finding theoretically suggests a therapeutic effect of lymphadenectomy, but bias inherent to this retrospective methodology must be acknowledged.
The therapeutic effect of lymphadenectomy is controversial for two reasons. First, patients in whom lymphadenectomy has not been performed may have been understaged. The therapeutic effect of lymphadenectomy may in fact reflect the negative impact of understaging. The pathological examination of a lymphadenectomy specimen is de facto the best evaluation of nodal status, and is more accurate than preoperative imaging or preoperative palpation.9,10 Second, the omission of lymphadenectomy may reflect an insufficiently extensive surgery. Lymphadenectomy is thus a surrogate marker for radicality of surgery, and patients for whom lymphadenectomy has not been performed may have a poorer prognosis because of incomplete cytoreduction.
Regarding understaging, it has recently been reported that the omission of lymphadenectomy in stage-I ovarian cancer may decrease survival.7 Patients with apparent stage-I disease, but for whom lymphadenectomy has been omitted, may in fact have stage-IIIc disease in the case of occult nodal metastasis. Thus, the survival difference may be explained by an underestimation of the stage of the disease. In our study, we show that the detrimental effect of understaging is logically more pronounced in patients with a higher risk of nodal metastases, namely patients with higher stage disease. Patients with apparent stage-I disease had a 10–24% risk of nodal metastasis compared with 20–30% for patients with stage-II disease.11 Few data are available for patients with stage-IIIA–B disease. Our results highlight the risks of undertreatment associated with the omission of lymphadenectomy in patients with early stage disease. Patients should be clearly informed of this risk.
The deleterious effect of lymphadenectomy omission may be related to the presence of metastatic nodes. For women with stage-II ovarian cancer, the risk of incomplete cytoreductive surgery can be estimated to be relatively low, but the incidence of lymph node involvement is around 20%. In this subgroup of women with a relatively low risk of intraperitoneal relapses, the impact of lymphadenectomy was particularly dramatic, underlying its specific role in survival. In patients with stage-IIIC and -IV disease, the understaging effect of lymphadenectomy omission is theoretically non-existent. Even in women with advanced stages of ovarian cancer (FIGO stages III/IV), for whom the risk of incomplete cytoreductive surgery and intraperitoneal recurrence increases, our data have demonstrated the positive impact of lymphadenectomy on survival.
This reinforces the concept that lymphadenectomy should be systematically recommended in women with ovarian epithelial cancer—even with advanced stages of the disease—if preoperative conditions of the patients remain favourable regarding blood loss and anaesthetic status. Our study confirmed the data of Chan et al.8 that included patients from 1988 to 2001. However, our data included patients from the SEER database from 1988 to 2004, a study period where taxans were more widely used. Moreover, our study provides additional evidence of the therapeutic effect of lymphadenectomy because we evaluated and quantified the interaction between clinicopathologic characteristics and the significance of the prognostic impact of lymphadenectomy. We also report in our study the effect of lymphadenectomy in patients with non-epithelial tumours: our results suggest that lymphadenectomy provides few or no benefits in patients with sex cord stromal and germ cell tumours. However, the relatively low number of patients (n = 2973) limits the robustness of our findings for these histologic types.
Another issue is whether node sampling is as efficient as extensive lymphadenectomy for survival. In the present study, for women with early stages of ovarian cancer, we demonstrated that lymphadenectomy with more than ten nodes removed increased the survival compared with those with less than ten nodes removed. Our results are in contrast with those of Maggioni et al.,12 who reported in a randomised trial including 268 women with FIGO stages-I/II ovarian cancer that no difference in recurrence or death by disease was observed between women undergoing a node sampling and those who underwent a lymphadenectomy. For patients with advanced ovarian cancer, our data are in agreement with those of Beneditti Panici et al.,6 who conducted a randomised trial to compare complete lymphadenectomy with resection of bulky nodes only. They showed that the adjusted risk for the first event was significantly lower in the systematic lymphadenectomy arm than in the no-lymphadenectomy arm: the 5-year progression-free survival rates were 31.2 and 21.6%, respectively (hazard ratio, HR = 0.75; P = 0.01). The risk of death was similar in both arms (HR = 0.97; 95% CI = 0.74–1.29; P = 0.85). In the control group the median number of lymph nodes resected was four, and this is comparable with the group in which between one and nine nodes were resected in the current study. In the univariate analysis, we showed that there was a significant association between the extent of lymphadenectomy and survival. However, the surgical effort to resect ovarian and peritoneal disease may have biased our results. After stratification for this factor, we obtained similar results: the positive effect of lymphadenectomy was observed regardless of the extent of surgery. In the study by Beneditti Panici et al.,6 when residual disease was taken into account, the hazard ratios for occurrence of the first event and mortality from any cause were almost unchanged. In the Scotroc study including 1077 patients from the UK (n = 689) and other countries (Europe, United States, and Australia; n = 388), UK patients with no visible residual disease had a less favourable progression-free survival rate compared with patients recruited from non-UK centres who were similarly debulked (HR = 1.85; 95% CI = 1.16–2.97; P = 0.010).13 This observation seemed to be related to surgical practice, primarily lymphadenectomy. Combined with our results, these data suggest that lymphadenectomy may have a positive impact on prognosis, independent of cytoreduction. One should keep in mind that the majority of ovarian cancer patients died from peritoneal disease. Therefore, there is no doubt that the maximum surgical effort must be concentrated on the peritoneal cavity in order to achieve a ‘complete debulking’ with no peritoneal residual tumour.
There are several weaknesses in our study that must be acknowledged. First, the majority of patients did not have lymphadenectomy. When 51% of stage-I and 56% of stage-II cases had no lymph nodes removed, the distinction between understaging and detrimental therapeutic effect is challenging. It would have been interesting to determine the survival of patients with advanced stage disease based on positive nodes only. Unfortunately, this subgroup cannot be individualised from the SEER database. Second, in the absence of a detailed description of disease extent and cytoreduction, we used a surrogate marker: the surgical procedures that were performed. It can be reasonably assumed that surgical procedures were adapted to the extent of disease and surgical effort. However, they also correlate with the quality of surgery, and the omentectomy, for example, was associated with the best outcome, whereas patients who had no hysterectomy had the worse outcome (Table 2). Unfortunately, a description of carcinomatosis and surgical procedures performed to obtain complete cytoreduction are optimally reported only through relatively complex scores, as reported by Sugarbaker, Eisenkop, or Aletti.14–16,17 Nevertheless, even with these scores, the quality of exploration (mobilisation of the liver, exploration of the omental bursa, etc.) may be associated with discrepancies between surgeons. With the surrogate marker that we used we have shown that a lack of lymphadenectomy was detrimental, independent of the surgical procedures performed.
Our results suggest that no factor, apart from sex cord stromal and germ cell histology, was associated with a lack of benefit from lymphadenectomy in ovarian cancer patients. Our results contrast with findings of a randomised controlled trial.6 Further studies are warranted to determine if our finding is related to a ‘better staging’ effect, a ‘marker of more adequate surgery’ effect, or a therapeutic effect of lymphadenectomy. Results of this study cannot be regarded as conclusive because of the bias inherent to the retrospective methodology, including stage migration and the possibility that the extent of lymphadenectomy is a surrogate marker for the radicality of surgery.
Disclosure of interests
No conflicts of interest.
Contribution to authorship
Study concept and design: RR, CP, and ED. Analysis and interpretation of data: RR, CB, FS, and GD. Drafting of the manuscript: RR, CB, JLB, and CP. Statistical expertise: RR. Study supervision: RR, SU, CP, and ED.
Details of ethics approval
This study was conducted according to the institutional and ethical rules concerning research, and was exempted from patient informed consent.
No funding source.