Survival curves of patients with early gastric cancer without (N−V−) and with (N+V+) lymph node and lymph vessel involvement.
Borie F, Millat B, Fingerhut A, Hay J, Fagniez P, De Saxce B, for the French Associations for Surgical Research. Lymphatic Involvement in Early Gastric CancerPrevalence and Prognosis in France. Arch Surg. 2000;135(10):1218-1223. doi:10.1001/archsurg.135.10.1218
The prognosis of early gastric cancer (EGC) is considered better than that of invasive gastric carcinoma, with a 5-year survival rate of more than 90% after surgery. The prevalence of lymph node metastasis in EGC ranges from 8% to 20% and is associated with a poor prognosis.
The main prognostic factor of EGC in patients in France is lymphatic involvement.
Design, Setting, and Patients
From January 1979 to December 1988, 332 patients with EGC were operated on in 23 centers of 2 of the French Associations for Surgical Research. Clinical, pathological, and therapeutic data were reviewed, and the reckoning point was in June 1996.
Main Outcome Measures
The cumulative 5- and 7-year specific survival rates of EGC with or without lymphatic involvement.
The cumulative 5- and 7-year specific survival rates of 332 patients with EGC (mean follow-up time, 80 months), excluding both operative and unrelated mortality, were 92% and 87.5%, respectively. Thirty-four patients (10.2%) had metastatic lymphatic spread: 13 exclusively in the lymphatic vessels close to the tumor, 17 in at least 1 lymph node, and 4 in both the lymphatic vessels and nodes. The rate of lymph node involvement (regardless of lymphatic vessel involvement) correlated significantly with submucosal invasion (P = .05) and histologic undifferentiation (P = .03). Lymphatic vessel involvement correlated positively with lymph node involvement (P = .003). Since 5- and 7-year survival rates of the 13 patients with EGC who had lymphatic vessel involvement without lymph node involvement did not differ significantly from those of patients who had EGC with lymph node involvement (85% and 84% vs 72% and 63%, respectively [P = .42]), all patients with lymph node and/or lymphatic vessel involvement were considered unique. Prognosis was poorest in these patients according to both univariate analysis (94% for 298 without node or vessel involvement vs 78% for 34 with node and/or vessel involvement; P = .006) and multivariate analysis (P = .01). Submucosal invasion was a prognostic factor independent of lymphatic involvement (P = .05). Five- and 7-year survival rates did not differ when the group of 211 patients for whom less than 15 lymph nodes were retrieved were compared with those (n = 51) for whom 15 or more lymph nodes were retrieved (95.5% vs 92% and 95.5% vs 88%, respectively), whether according to univariate (P = .21) or multivariate (P = .31) analysis.
Our results suggest that both lymph node and lymphatic vessel involvement are important prognostic factors in patients with EGC. Lymphadenectomy in EGC is important to identify the high-risk population for whom prognosis is worse. The extent of lymphadenectomy (at least 15 nodes) in these patients, however, does not alter prognosis.
EARLY GASTRIC cancer (EGC) is defined as adenocarcinoma of the stomach in which the depth of invasion is limited to the mucosa and/or submucosa, regardless of lymph node (LN) metastatic status.1,2 Prognosis of EGC is good, with a 5-year survival rate of more than 90% after surgery.3 The lymphatic system plays an important role in the spread of gastric cancer.4 The prevalence of LN metastasis in EGC ranges from 10% to 20%,5- 9 as LN involvement has been reported to be an important prognostic factor.5- 8 To our knowledge, the prognostic role of lymphatic vessel (LV) involvement and/or LN involvement has never been evaluated in previous studies of EGC. This retrospective study was undertaken to elucidate the prevalence, characteristics, prognosis, and potential therapeutic implications of lymphatic spread in EGC.
During a 120-month period (January 1979 to December 1988), 208 men and 124 women with EGC were operated on in 23 centers of the French Associations for Surgical Research. These patients represented 11% of all cases of gastric cancer operated on, and 13.5% of all resected gastric cancer patients (n = 2466) in these centers. Mean age was 64 years (62 years ± 12 years for men, 66 years ± 13 for women [mean ± SD]).
Patients with early gastric cancer were grossly classified according to the Japanese Society for Gastroenterology Endoscopic criteria10 as type I (protruded), type IIa (elevated) (both grouped together in our study as "elevated"), type IIb (flat), type IIc (depressed), and type III (excavated). Early gastric cancer was classified as intramucosal or submucosal according to the depth of tumor invasion,1,2 and histologically classified according to the International Union Against Cancer TNM classification.11 In our series, "well" or "moderately differentiated" EGC were said to be differentiated, and "undifferentiated" or "independent" EGC were considered to be undifferentiated.12 Although neither the method of LN sampling nor the exact group of LN dissection was known for all patients, the location of LN metastasis was recorded in each case. Demographic data of patients, follow-up, and survival information (death and cause of death) were obtained from each center. The reckoning point was June 1996.
Data were compared in patients with and without lymphatic involvement according to 8 clinicomorphological features, using the χ2 test for discrete parameters and the t test and Mann-Whitney test for continuous parameters. Sex, age, tumor site, size of tumor (largest diameter <15 mm, ≥15 mm of the lesion when unique, or the largest lesion when multiple6), gross aspect, histologic type, depth of infiltration, and lymphatic involvement were recorded for each patient. The number of LNs retrieved and analyzed was noted and the prognostic value of LN retrieval on survival was calculated according to whether less than 15 or 15 or more LNs13 were retrieved (International Union Against Cancer standards for invasive cancer). Survival curves were calculated according to the Kaplan-Meier method. For the purpose of survival analysis, both overall mortality (including postoperative deaths) and specific mortality (excluding postoperative deaths), respectively, were calculated. Stepwise regression analysis on all variables was performed. Significance levels for entry and retention in the Cox proportional hazard model were 0.1 (10%). The SAS statistical package version 6 (SAS Institute, Raleigh, NC) was used for the statistical evaluation on a Digital MicroVAX computer (Compac, Houston, Tex). The accepted level of significance was P<.05.
Overall, 34 patients with EGC (10.2%) had metastatic spread: 13 to LV only, usually close to the tumor; 17 to at least 1 LN; and 4 both to LN and LV. Of the 21 patients (6.3%)(10 men and 11 women; mean ± SD age, 63 ± 14 years) with LN involvement, 5 (24%) had undergone total gastrectomy, while 16 (76%) had a partial or subtotal gastric resection. One involved LN (5%) was located at the cardia, 7 (33%) in the fundus of the stomach or the antrofundic junction, and 13 (62%) in the antrum or pylorus. Lymph node involvement was demonstrated in 1 (3%) of 34 patients with EGC located at the cardia, in 7 (8.8%) of 80 patients with EGC located in the fundus or the antrofundic junction, and in 13 (6.1%) of 213 patients with EGC located in the antrum or pylorus.
Lymph node involvement was clinically suspected in 40 patients at operation, of whom 16 (34%) had intraoperative frozen sections. Overall, 47 patients had intraoperative frozen sections. The rates of true EGC with LN involvement were 17.5% and 2.5% in those patients, respectively, and 5.4% in the 262 patients with neither clinical suspicion nor positive intraoperative frozen section.
According to gross classification, 3 cases of EGC with LN involvement (14%) were elevated, 2 (10%) were flat, and 16 (76%) were excavated. The mean diameter of EGC with LN involvement was 28 mm vs 26 mm for EGC without LN involvement (difference not statistically significant). Seven cases of EGC with LN involvement (33%) were well-differentiated adenocarcinomas, and 13 (62%) were undifferentiated adenocarcinomas (1 case of EGC with LN involvement was not recorded). Six cases of EGC with LN involvement (28.5%) were completely intramucosal, and 15 (71.5%) infiltrated the submucosa.
Thirteen patients, 10 men and 3 women (mean ± SD age, 66 ± 14 years), had EGC with LV spread but without LN involvement. Five (38%) had undergone total gastrectomy, while 8 (62%) had a partial or subtotal gastric resection. One EGC with LV involvement was located at the cardia, 4 (38%) were in the fundus of the stomach and at the antrofundic junction, and 8 (61%) were in the antrum or pylorus. Lymphatic vessel involvement was found in 1 of 34 patients with EGC located at the cardia, in 4 (5%) of 80 located in the fundus or the antrofundic junction, and in 8 (4%) of 213 in the antrum or pylorus. According to gross classification, 3 cases of EGC with LV involvement (23%) were elevated, 1 was flat, and 9 (69%) were excavated. The mean diameter was 34 mm in EGC with LV but no LN involvement vs 26 mm for EGC with no LN or LV involvement. Eight cases of EGC with LV but no LN involvement (66%) were well-differentiated adenocarcinomas, while 4 (33%) were undifferentiated adenocarcinomas (1 EGC with LV involvement was not recorded). All of the 13 cases of EGC with LV but no LN involvement completely infiltrated the submucosa.
Lymph node involvement rates correlated significantly with submucosal invasion (P = .05), LV involvement (P = .003) (Table 1), and histological undifferentiation (P = .03). Lymphatic vessel involvement rates correlated significantly with submucosal invasion (P<.001) (Table 1). Other clinicopathological factors that were not correlated to lymphatic involvement are reported in Table 2. Nineteen (90%) of 21 positive nodes were found in the N1 LN chains, but in 2 patients (9.5%), positive nodes were detected around the N2 LN chains. Both of these patients had submucosal invasion and undifferentiated adenocarcinomas. The number of LNs harvested was significantly higher after total gastrectomy than after subtotal gastrectomy (median, 10 [range, 0-47] vs 7 [range, 0-51], respectively; P = .02). The number of LNs removed in EGC with LN involvement was higher than that in EGC without LN metastasis, but the difference was not statistically significant (median, 12 [range, 1-47] vs 9 [range, 0-51]; P = .10).
The prevalence of LN involvement did not differ statistically according to whether 15 or more LNs (n = 51 patients) vs less than 15 LNs (n = 281 patients)(10% vs 8%; P = .62) were retrieved and analyzed.
There were 19 patients for whom no LNs were found in the specimen, 51 patients for whom the number of LNs retrieved was not given (1 patient classified as having LNs and LV involvement), 211 patients for whom 1 to 14 LN were analyzed (27 had LN and LV involvement), and 51 patients for whom 15 or more LNs were retrieved with the specimen (6 of whom had LN and LV involvement).
Excluding the 70 patients with incomplete or missing data, no statistically significant difference was found when 5- and 7-year survival rates were compared between the group of 211 patients for whom fewer than 15 LNs were retrieved and the group of 51 patients for whom 15 or more LNs were retrieved (95.5% vs 92% and 95.5% vs 88%, respectively), whether calculated according to univariate (P = .21) or multivariate (P = .31) analysis.
The mean follow-up for the 332 EGC patients was 80 months, ranging from 0 to 197 months. At the reckoning date, 104 patients (30.6%) had died: 14 (4%) due to postoperative complications, 31 (9.3%) due to EGC, 55 (17%) due to unrelated diseases, and 4 due to unknown causes. Seven (20.5%) of 34 patients with EGC with or without LN or LV involvement died of EGC, 7 (20.5%) of unrelated diseases, 1 (3%) of postoperative complications, and 2 (6%) of unknown causes. Seventeen patients with EGC with LN involvement (50%) were still alive at the end of the study.
The cumulative 5-year and 7-year survival rates for the 332 patients with EGC were 82% and 72%, respectively, for the whole group, and 92% and 87.5%, respectively, when both operative and unrelated mortality were excluded.
Since cumulative 5- and 7-year survival rates of the 13 patients with EGC with LV involvement were not statistically different from those of the 21 patients with EGC with LN involvement (85% and 84% vs 72% and 63%, respectively [P = .42 for all]), they were considered as 1 group in the univariate and multivariate analyses.
Results of univariate analysis of the potential prognostic factors are reported in Table 3. Prognosis was significantly worse for patients with EGC with LN and LV involvement (Figure 1) when compared with those with EGC without LN or LV involvement (7-year survival was 94% for the 298 patients without involvement vs 78% for the 34 patients with involvement [P = .006]). Tumors located in the lower third of the stomach had a better outcome than in all other locations (P = .02). Five-year disease-free survival rates of 159 patients with intramucosal and 167 patients with submucosal EGC were 95% and 82%, respectively (P = .02).
Results of multivariate analysis are reported in Table 4. Lymphatic involvement was the most important independent prognostic factor (P = .001), followed by the tumor site in the upper two thirds of the stomach (P = .02), and submucosal invasion (P = .05) (Table 4). Submucosal invasion remained a prognostic factor independent of lymphatic involvement.
When compared with Western countries, the annual incidence of gastric carcinoma is approximately 8 times higher in Japan: 50 per 100,000 for men and 25 per 100,000 for women.14 In Japan, as many as 30% to 40% of all diagnosed gastric cancers are EGC.3 In our experience, the percentage of EGC represented 13.5% of all gastric cancers resected, comparing favorably with those reported in other Western countries (6%-16%).15- 17 The high incidence and good prognosis (5-year survival rate of >95%) of EGC in Japan, however, might be explained by Japanese pathologists focusing more on nuclear factors and glandular structures than Western pathologists and, therefore, adenoma and dysplasia being diagnosed as EGC.18
Studies of more than 300 patients have rarely been encountered in the literature because of the low incidence of gastric carcinoma and EGC in Western countries. Only 1 investigation,5 reporting on 414 patients, studied more patients than ours (n = 332). The large number of cases in our series and the long follow-up period (median, 84 months) allowed us to evaluate the prognostic factors with a multivariate analysis.
Four studies have previously attempted to identify prognostic factors in EGC according to the Cox regression model.5- 8 In all studies, ours included, the main prognostic variable for survival was LN metastasis.5- 8 In contrast to our series, however, other prognostic factors, as determined by multivariate analysis, such as histologic type, age, or tumor size were also reported.5- 7 In these studies, however, either the number of patients included was small (159 patients)7 or the mean follow-up period was less than 80 months.5 In our study, univariate and multivariate analyses highlighted 3 prognostic factors. The first and most important was lymphatic involvement. The second was the influence of EGC located in the upper two thirds of the stomach (P = .02). Until now, this factor has only been reported in a population-based series.19 The third prognostic factor, also determined in the study reported by Ranaldi et al,5 was submucosal invasion, an independent prognostic factor of survival (P<.05). The level of significance reached by submucosal involvement for independent significance, however, was close to 5%. Due to the retrospective character of our study and the number of centers involved, lymphadenectomy was not standardized. It is therefore possible that some of the cases of submucosal EGC considered as not having LN involvement might have been classified otherwise if a larger lymphadenectomy had been performed. Thus, the correlation between submucosal invasion and LN involvement could have been stronger, thereby cancelling out the independent prognostic value of submucosal involvement.
These apparently paradoxical results may be explained by the size of the respective studies and the different lengths of follow-up. However, no explanation for the negative influence of cancer located in the upper two thirds of the stomach could be found.
Our study was the first published study to find that EGC with LV involvement had the same prognosis as EGC with LN involvement and that LV without LN involvement was a poor prognostic factor. This may be assimilated to the findings of Siewert et al,20 who spoke of "micrometastasis" occurring in this setting. Other clinicopathological characteristics that have been found to correlate with LN involvement are submucosal invasion and tumor size.7,9,12,21,22 Depth of invasion and LN involvement seem to be strongly correlated. This can be explained by the distribution of lymph capillaries within the gastric mucosa, which are found only in the deep lamina propria adjacent to and within the muscularis mucosae. In contrast, the mucosa has a rich supply of blood capillaries, many of which are adjacent to the basal lamina of gastric glands and to the surface epithelium.23
The present study compared clinicopathological features of patients with EGC with or without lymphatic involvement. Lymph node involvement was associated with submucosal invasion (P = .05), undifferentiated adenocarcinoma (P = .03), and LV invasion (P = .003). Submucosal invasion was a predictive factor for lymphatic involvement only. Few authors have reported other clinicopathological characteristics that enhance the risk of lymph node involvement, such as LV invasion and gross types IIa+IIc and I+IIa.9,21 Boku et al22 reported that LN metastasis was more frequent in tumors located in the lower third of the stomach, suggesting that the location in the lower third of the stomach had a poor influence on prognosis. In contrast, we found that EGC located in the lower third was associated with a better prognosis than EGC in the upper two thirds of the stomach.
Several authors have stated that for invasive carcinoma of the stomach, the cutoff value of 15 or more LNs retrieved was an independent prognostic factor for survival.13 This cutoff value might be too high for EGC according to our data. When we compared survival in the 27 patients with LN involvement (n = 211) who had 1 to 14 LNs retrieved with the 6 patients (n = 51) who had 15 or more LNs retrieved, there was no significant difference in survival (5- and 7-year survival rates were 95.5% vs 92%, and 95.5% vs 88%, respectively). Also, the fact that the prevalence of LN/LV involvement is lower in EGC (45.5% for invasive cancer24 compared with 10%-20% in EGC5- 9), several hundred specimens would be required to show any significant differences or survival benefits in routine lymphatic resection.
Lymph node removal and analysis were not standardized in this study. Moreover, when the study began in 1979, endoscopic resection and preoperative endoscopic ultrasound were not performed for all patients and, therefore, preoperative analysis was not possible. Because of the relationships found between undifferentiated carcinoma and submucosal involvement, and between LN and LV involvement, further investigation is recommended. The generally accepted criteria for endoscopic resection of EGC6 are well-differentiated intramucosal adenocarcinoma of superficial, elevated lesions (IIa, diameter <2 cm), and well-differentiated intramucosal adenocarcinoma of a superficial, depressed lesion (IIc, diameter <1 cm) without ulcers or scars inside.25,26 Although the indications for endoscopic mucosal resection are narrow and strict, the precise preoperative diagnosis, while essential, is questionable when using abdominal computed tomography and endoscopic ultrasonography.27
Lymph node involvement might be considred suspect by preoperative investigations. Endoscopic ultrasonography of gastric cancer has been shown to be superior to computed tomography or ultrasonography in predicting the T and N stages.27 However, false-positive and false-negative results due to invisible microscopic invasion, technical failures, artifacts, and peritumoral inflammatory or fibrous changes may be responsible for diagnostic failures.20,28 Despite its limited use in our study, and notwithstanding the retrospective character of our study, endoscopic ultrasonography is not reliable enough to indicate the appropriate therapeutic procedures. Similarly, our results suggest that intraoperative procedures, clinical suspicion of LN involvement, and/or frozen section of suspected nodes are of no therapeutic value. In addition, because LV involvement shares the same prognostic value as LN involvement and cannot currently be diagnosed preoperatively, elaborate preoperative workup of LN involvement is currently of little value.
Most LN metastasis in EGC is confined to N1 and/or N2 lymph nodes. According to previous studies6,21,28 and our experience, the incidence of N1 involvement ranges from 80% to 90% in EGC with LN involvement. The incidence of N2 and N3 involvement is lower, ranging from 10% to 19% and from 0% to 1%, respectively.6,21,28 These observations seem to favor R2 gastrectomy.20 However, until now, no randomized, controlled study has proven the superiority of extended lymphadenectomy over standard lymphadenectomy in patients undergoing total or subtotal gastrectomy for EGC. A prospective, controlled, randomized study for invasive gastric cancer has shown no difference in survival rate between the 2 surgical techniques.29
Further evaluation of adjuvant therapy and follow-up might be required for patients with lymphatic involvement. At the present time, there is no proven effective treatment available for node-positive ECG. Moreover, our results suggest that evaluation of the benefits of adjuvant therapy might be difficult or even impossible owing to the small number of patients with node-positive EGC (10%-20%5- 9) and to the favorable long-term survival5 (85% at 5 years) observed in these patients.
In conclusion, our results suggest that lymphatic involvement in EGC is a poor prognostic factor and that lymphadenectomy is important to identify the high-risk population for which prognosis is worse. The extent of lymphadenectomy (at least 15 LNs removed) in these patients, however, does not alter prognosis.
Presented at the 6th Annual Meeting of the European Surgical Association, London, England, April 24, 1999.
The following centers in France participated in this study: Adloff Michel (Strasbourg), Baumel Hugues (Montpellier), Chipponi Jacques (Clermont-Ferrand), De Calan Loïc (Tours), Delaitre Bernard (Cochin), Dubois Francois (Paris), Escat Jean (Toulouse), Fagniez Pierre-Louis (Créteil), Fékété Francois (Clichy), Gayet Pierre (Clichy), Gignoux Marc (Caen), Gouzi Jean-Luc (Toulouse), Hay Jean-Marie (Colombes), Huguier Michel (Paris), Huten Noël (Tours), Julien Michel (Créteil), Laborde Guy (Pau), Langlois Odile (Maubeuge), Lenriot Jean-Pierre (Lonjumeau), Levard Hugues (Paris), Marescaux Jacques (Strasbourg), Millat Bertrand (Montpellier), Samama Guy (Caen), Tenières Paul (Rouen), Testard Jacques (Rouen), Triboulet Jean-Pierre (Lille), and Veyrières Michel (Pontoise).
Corresponding author: Abe Fingerhut, MD, FACS, FRCS, Service de Chirurgie Viscérale, Hôpital Léon Touhladjian, 10 rue du Champs Gaillard, 78303 Poissy, France (e-mail: AbeFinger@aol.com). Reprints: French Associations for Surgery Research, 8 avenue des Peupliers, 92270 Bois-Colombes, France.