Prognostic Factors for the Survival of Patients with Esophageal Carcinoma in the U.S.

1434 Prognostic Factors for the Survival of Patients with Esophageal Carcinoma in the U.S. The Importance of Tumor Length and Lymph Node Status Mohamad A. Eloubeidi, M.D., M.H.S. 1,2 Renee Desmond, Ph.D. 2 Miguel R. Arguedas, M.D., M.P.H. 1 Carolyn E. Reed, M.D. 3 C. Mel Wilcox, M.D. 1 1 Department of Medicine, Division of Gastroenterology and Hepatology, The University of Alabama at Birmingham, Birmingham, Alabama. 2 Biostatistics Unit, Comprehensive Cancer Center, The University of Alabama at Birmingham, Birmingham, Alabama. 3 Division of Cardiothoracic Surgery and Comprehensive Cancer Center, The Medical University of South Carolina, Charleston, South Carolina. BACKGROUND. The current TNM classification system does not consider tumor length or the number of lymph nodes in the staging and classification scheme for patients with esophageal carcinoma. Using data from the National Cancer Institute SEER Program, the authors explored the effect of tumor length and number of positive lymph nodes on survival in patients with esophageal carcinoma. METHODS. Patients with esophageal adenocarcinoma or squamous cell carcinoma were subgrouped according to historic stage with localized, regional, or distant disease. Demographic factors (age at diagnosis, race, and gender) and tumor characteristics (morphology, histologic grade, tumor length, primary site, depth of invasion, number of positive lymph nodes, proportion of positive lymph nodes dissected, and distant metastatic sites) were examined. RESULTS. Overall factors that were associated with an increased mortality risk included increasing age at diagnosis, black race versus white race, histologic grade, primary tumor site in the lower esophagus and abdomen versus upper regions, and increasing depth of invasion. Among patients with regional disease, the number of positive lymph nodes ( 5 vs. 5) was related to an increasing risk (hazard ratio [HR], 1.29; 95% confidence interval [95%CI], 1.06 1.56). The proportion of positive lymph nodes compared with the number of lymph nodes dissected conferred an increased risk (HR, 1.63; 95%CI, 1.26 2.11). Among patients with distant disease, sites other than distant lymph nodes implied an increased mortality risk (HR, 1.37; 95%CI, 1.37 1.65). Tumor length was an independent predictor of mortality when controlling for depth of invasion in patients with localized disease (HR, 1.15; 95%CI, 1.08 1.21). CONCLUSIONS. Tumor length, the number of involved lymph nodes, and the ratio of positive lymph nodes are important prognostic factors for survival in patients with esophageal carcinoma. A revised TNM classification system for patients with esophageal carcinoma might consider adding tumor length and number of positive lymph nodes as two important prognostic factors. Cancer 2002;95:1434 43. 2002 American Cancer Society. DOI 10.1002/cncr.10808 Presented in part in poster form at Digestive Disease Week, Atlanta, Georgia, May 20 23, 2001. Address for reprints: Mohamad A. Eloubeidi, M.D., M.H.S., Division of Gastroenterology and Hepatology, 1530 3rd Avenue South, ZRB 636, Birmingham, AL 35294-0007; Fax: 205-975-6381; E-mail: meloubeidi@uabmc.edu Received September 28, 2001; revision received March 15, 2002; accepted April 30, 2002. KEYWORDS: esophageal carcinoma, survival, tumor length, lymph node status. Esophageal carcinoma is a relatively uncommon tumor and carries a poor prognosis in the United States. It is estimated by the American Cancer Society that 13,200 patients will develop the disease in 2001 and that 12,500 will die from it. 1 The majority of patients with esophageal carcinoma die within 1 year of diagnosis, and only 8 20% of patients are alive at 5 years. 2 Accurate staging is important in these patients, because it directs and influences the choice of therapy. 3 The current American Joint Committee on Cancer (AJCC) staging system 2002 American Cancer Society

Prognostic Factors in Esophageal Carcinoma/Eloubeidi et al. 1435 for esophageal carcinoma is based mostly on retrospective data from the Japanese Committee for Registration of Esophageal Carcinoma Cases, which focused on patients with squamous cell carcinoma of the cervical and thoracic esophagus. 4 Few studies with small numbers of patients from the western literature focused on staging issues in patients with esophageal carcinoma, in particular, the location of lymph nodes and the number of lymph nodes involved. 5 7 With the continued rise in adenocarcinoma of the distal esophagus, 8,9 these issues become even more important. The current TNM (1997 AJCC) classification system 10 does not consider the tumor length or the number of lymph nodes in the staging and classification scheme for patients with esophageal carcinoma. The objective of the current analysis was to explore the effect of tumor length and the number of lymph nodes on survival in patients with esophageal carcinoma in the United States. MATERIALS AND METHODS Patient Population Data concerning patients with primary incident esophageal carcinoma (International Classification of Diseases for Oncology [ICD-O], 2nd edition; C15.0 C15.9) 11 were available from the National Cancer Institute Surveillance, Epidemiology, and End Results (SEER) Program. 12 Only patients with disease diagnosed from 1988 or later were included, because information on length of primary tumor and extent of disease was coded for patients who were diagnosed from 1988 or later. White patients and black patients with morphologic subtype ICD-O adenocarcinoma (8140 8573) and squamous cell carcinoma (8050 8082) were included; 11 patients with other disease subtypes were excluded. Statistical Analysis Demographic factors considered in the analysis included age at diagnosis (decades), race (black vs. white), gender (male vs. female), and stage of disease (localized, regional, or distant). Tumor factors included morphology (adenocarcinoma vs. squamous cell carcinoma), histologic grade (1 4), primary site, invasiveness or extent of disease, tumor length, number of positive lymph nodes, number of lymph nodes dissected, and site of distant metastasis. Data on tumor length were abstracted from the 12-digit extent of disease field in the SEER database. The extent of disease field description is SSSEELPN- EXPE, where the first three digits (SSS) correspond to the size of the tumor. For the esophagus, the first three digits (SSS) are coded in millimeters; for instance, 003 3 mm, and 030 30 mm or 3.0 cm. For the field 001, the coding is foci only, which we have included as 1 cm. There are some codes that were not included in our analysis (998 entire circumference, and 999 not stated), because it was not possible to determine the size. The size of the primary tumor was determined from pathology reports, operative reports, endoscopic examinations, and radiographic reports in priority order. Continuous covariates were categorized for analysis. Tumor length (length of involved esophagus) was grouped into quintiles. The exact groupings of 20%, 40%, 60%, and 80% were not possible due to measurement rounding in the data. The number of involved lymph nodes was categorized as 0, 5, or 5, and a proportion was computed to compare the number of positive lymph nodes with the number of lymph nodes dissected as 0.10 and 0.10. Primary site of tumor was collapsed into two categories: 1) cervical, upper one-third and middle one-third of the esophagus, and thoracic esophagus; and 2) lower one-third of the esophagus and abdominal esophagus. Invasiveness (T classification) was categorized by tumor extension: 1) mucosa, lamina propria, muscularis mucosa, submucosa, and localized (not otherwise specified [NOS]); 2) muscularis propria; 3) adventitia and/or soft tissue invaded; and 4) contiguous extension to adjacent structures, including major blood vessels, thyroid gland, trachea, lung, bronchus, diaphragm, pleurae, stomach, cardia, hypopharynx, larynx, cervical vertebrae, ribs, and mediastinal structures. Survival was calculated from date of diagnosis to the date of last follow-up. Only patients who died of esophageal carcinoma (ICD-9 150.0 150.9) were considered to have died of disease. Cox proportional hazards analysis was used to determine significant predictors of survival by stage of disease (localized, regional, or distant, as defined by SEER historic stage). Because of a large percentage of missing data fields for tumor length and T classification (invasiveness), survival models were run both including and excluding these fields. For all analyses, P values 0.05 were considered statistically significant. RESULTS Descriptive Statistics Of 10,441 patients who were diagnosed after 1988, the majority of patients were white (72.2%), and their median age at the time of diagnosis was 68 years. Over 70% of patients were male (n 7615 patients) (Table 1). The percentages of patients who staged with local, regional, and distant disease were 25.1%, 24.1%, and 23.4%, respectively. The median survival was 1.2 years for patients with localized disease, 0.8 years for pa-

1436 CANCER October 1, 2002 / Volume 95 / Number 7 TABLE 1 Characteristics of 10,441 SEER Patients with Esophageal Carcinoma 1988 1996 Characteristic No. (%) Age at diagnosis (yrs) 10 19 3 (0.0) 20 29 13 (0.1) 30 39 108 (1.0) 40 49 660 (6.3) 50 59 1767 (16.9) 60 69 3223 (30.9) 70 79 3161 (30.3) 80 89 1303 (12.5) 90 203 (1.9) Race White 8061 (77.2) Black 1796 (17.2) Other 584 (5.6) Gender Male 7615 (72.9) Female 2826 (27.1) State of disease In situ 215 (2.1) Localized 2623 (25.1) Regional 2513 (24.1) Distant 2439 (23.4) Unstaged 2651 (25.4) Tumor characteristics Morphology Squamous cell carcinoma 5595 (53.6) Adenocarcinoma 3849 (36.9) Other 997 (9.5) Histologic grade 1) Well differentiated 500 (4.9) 2) Moderately differentiated 3268 (31.3) 3) Poorly differentiated 3964 (38.0) 4) Undifferentiated 264 (2.4) Missing 2445 (23.4) Tumor length (cm) 3.0 814 (7.8) 3.0 4.1 1260 (12.1) 4.2 5.7 995 (9.5) 5.8 7.7 1047 (10.0) 7.8 1070 (10.2) Missing 5255 (50.4) Primary site Cervical esophagus 350 (3.4) Thoracic esophagus 357 (3.4) Abdominal esophagus 108 (1.0) Upper one third of esophagus 769 (7.4) Middle one third of esophagus 2472 (23.7) Lower one third of esophagus 4656 (44.6) Overlapping lesion 536 (5.1) Unknown 1193 (11.4) Invasivesss (T classification) T0 215 (2.1) T1 2625 (25.1) T2 556 (5.3) T3 901 (8.6) T4 1423 (13.7) Missing 4721 (45.2) tients with regional disease, and 0.3 years for patients presenting with distant disease (Fig. 1). Tumor Characteristics Greater than 50% of the invasive tumors were Grade 2 3 squamous cell carcinomas (53.6%). The majority of tumors originated in the middle one-third and the lower one-third of the esophagus (Table 1). For patients who had tumor length recorded in the data base, the median length was 5 cm. Survival For the survival analysis, patients were subgrouped into those localized, regional, or distant stage disease, as specified by the SEER criteria. Significant differences in patients and tumor characteristics were observed across stage, with the exception of gender (Table 2). Many patients with distant tumors did not have data recorded for tumor length (n 1292 patients) or T staging (n 2070 patients). Similarly, tumor length was not recorded for 43.8% of patients with localized tumors and for 31.3% of patients with regional tumors, whereas T staging was complete for all patients with localized tumors and regional tumors. The differences in clinical and tumor characteristics across the stage groupings was the basis for the development of separate predictive models of survival by stage. Localized disease Increasing age was associated with a significantly increased risk of mortality among patients with localized esophageal carcinoma (hazard ratio [HR], 1.27) (Table 3). Black race also conferred a significantly greater risk of dying compared with white race (HR, 1.38). Increasing length of tumor was associated with an increased hazard (HR, 1.15). Patients who had tumors arising in the lower one-third and abdominal esophagus had a significantly increased mortality risk compared with patients who had tumors in the cervical and upper esophagus. In a model that excluded tumor length, factors associated with an increased risk of dying included age at diagnosis, black race, male gender, histologic grade, and primary site in the lower esophageal region compared with upper regions. T classification (T2 vs. T1) was associated with a lower risk of mortality (HR, 0.82) controlling for all other covariates (data not shown). Regional disease Increasing age at diagnosis was a significant predictor of mortality (HR, 1.15) (Table 4). Although black patients tended to have an increased risk compared with

Prognostic Factors in Esophageal Carcinoma/Eloubeidi et al. 1437 FIGURE 1. Kaplan Meier survival curves for 7575 patients with esophageal carcinoma stratified by tumor stage. white patients, the P value was not significant (P 0.0803). Similar to localized disease, primary site was a predictor with tumors arising in the lower onethird and the abdominal esophagus conferring an increased risk compared with tumors arising in upper sites. Tumor length was a borderline significant predictor controlling for other factors (P 0.0555). The depth of invasion was correlated with an increased mortality risk as well as the number of positive lymph nodes ( 5 vs. 5) (Fig. 2). The mortality risk increased significantly as the number of positive lymph nodes relative to the number of lymph nodes dissected increased. In a model that excluded tumor length, there were no differences in the variables that were significant compared with the model, including tumor length (not shown). Distant disease Factors that were predictive of mortality risk among patients with distant disease that also were significant for patients with localized and regional disease included age at diagnosis, black race, and histologic grade (Table 5). Patients with a morphologic diagnosis of adenocarcinoma had a higher risk of mortality (HR, 1.21) compared with patients with squamous cell carcinoma. Tumor length was associated with an increased risk, but the association was not significant (P 0.0830). The site of distant metastasis (other vs. lymph node) was associated with a significantly greater mortality risk (HR, 1.37). In a multivariate model that excluded tumor length, there were no differences in the factors that remained significant compared with the model, including tumor length (not shown). For patients with distant disease, T staging was excluded from all models due to the large amount of missing data for this field. Tumor length Further explorations of the correlation between 2-year survival and tumor length are shown in Figure 3 and Table 6. Figure 3 includes patients who were staged with localized disease. There is a significant difference between survival curves 1 and 2 (P 0.001) and survival curves 2 and 3 (P 0.04). Table 6 shows the 1-year and 2-year survival rates for patients with clinically localized disease (n 1471 patients) and patients with known lymph node negative status (n 580 patients). The correlation between tumor length also was significant for patients with lymph node negative disease. A nonlinear model was fit to describe the correlation between tumor length and mortality. Although the model demonstrated no naturally occurring break points, there is a suggestion of a leveling off of risk at about 3 4 cm, as illustrated in Figure 4. The correlation was modeled as a negative exponential function: f(t) 1 0.9424e ( 0.3285t 0.02100t2), where t is the measured tumor length, and f(t) is the 2-year disease specific mortality rate. Treatment Factors The majority of patients in this cohort received radiation therapy (61.5%) compared with surgery (34.3%). A greater proportion of patients who had localized and regional disease underwent surgery compared with

1438 CANCER October 1, 2002 / Volume 95 / Number 7 TABLE 2 Characteristics of SEER Patients with Esophageal Carcinoma 1988 1996 by Stage of Disease Characteristic a Localized no. (%) Regional no. (%) Distant no. (%) Age at diagnosis (yrs) 10 19 0 (0.0) 2 (0.08) 1 (0.04) 20 29 2 (0.08) 1 (0.04) 3 (0.12) 30 39 19 (0.7) 23 (0.92) 40 (1.6) 40 49 137 (5.2) 193 (7.7) 196 (8.0) 50 59 418 (15.9) 507 (20.2) 472 (19.4) 60 69 845 (32.2) 840 (33.4) 802 (32.9) 70 79 830 (31.6) 710 (28.3) 704 (28.9) 80 89 332 (12.7) 221 (8.8) 207 (8.5) 90 40 (1.5) 16 (0.6) 14 (0.6) Race White 1980 (80.1) 1924 (81.9) 1864 (80.8) Black 493 (19.9) 426 (18.1) 443 (19.2) Gender Male 1857 (70.8) 1885 (75.0) 1998 (77.8) Female 766 (29.2) 628 (25.0) 541 (22.2) Tumor characteristics Morphology Squamous cell carcinoma 1594 (63.8) 1406 (58.6) 1132 (51.1) Adenocarcinoma 903 (36.2) 994 (41.4) 1083 (48.9) Histologic grade 1) Well differentiated 177 (8.5) 111 (5.1) 81 (4.1) 2) Moderately differentiated 971 (46.4) 896 (41.1) 674 (34.4) 3) Poorly differentiated 897 (42.9) 1091 (50.0) 1114 (56.9) 4) Undifferentiated 46 (2.2) 83 (3.8) 88 (4.5) Tumor length (cm) 0.1 2.9 352 (23.9) 214 (12.4) 104 (9.1) 3.0 4.1 392 (26.6) 454 (26.3) 224 (19.5) 4.2 5.7 269 (18.2) 374 (21.7) 196 (17.1) 5.8 7.7 260 (17.6) 352 (20.4) 279 (24.3) 7.8 202 (13.7) 331 (19.2) 344 (30.0) Primary site Cervical esophagus 91 (3.5) 125 (5.0) 57 (2.3) Thoracic esophagus 114 (4.4) 86 (3.4) 59 (2.4) Abdominal esophagus 24 (0.91) 26 (1.0) 30 (1.2) Upper one third of esophagus 210 (8.1) 182 (7.2) 141 (5.8) Middle one third of esophagus 763 (29.1) 603 (24.0) 464 (19.0) Lower one third of esophagus 1139 (43.4) 1247 (49.6) 1258 (51.6) Overlapping lesion 120 (4.6) 141 (5.6) 158 (6.5) Unknown 162 (6.2) 103 (4.1) 272 (11.2) Invasivesss (T classification) T1 2259 (86.1) 282 (11.2) 84 (22.8) T2 364 (13.9) 172 (6.8) 20 (5.4) T3 0 (0.0) 858 (34.1) 43 (11.7) T4 0 (0.0) 1201 (47.8) 222 (60.2) a The numbers for each characteristic do not reflect the total numbers due to missing data. patients who had distant disease. Information on chemotherapy was not coded as a treatment variable in the SEER data base. Treatment factors, obviously, are related to lower mortality. In this analysis, due to incomplete information on all treatment modalities and the possibility of selection bias in treatment regimens, treatment factors were not considered in the multivariable models. DISCUSSION Staging patients with esophageal carcinoma precedes therapy, because treatment modalities are tailored to the stage of disease. 3 It is now clear that patients who have lymph node involvement have worse survival compared with patients who are without lymph node involvement. 13 New staging modalities, such as positron emission tomography scans, endoscopic ul-

Prognostic Factors in Esophageal Carcinoma/Eloubeidi et al. 1439 TABLE 3 Prognostic Factors for Survival from Localized Esophageal Carcinoma: Cox Multivariate Model TABLE 4 Prognostic Factors for Survival from Regional Esophageal Carcinoma: Cox Multivariate Model Factor Hazard ratio (95% CI) P value Factor Hazard ratio (95% CI) P value Age at diagnosis (10 yrs) 1.27 (1.18 1.37) 0.0001 a Race White 1.0 0.0014 a Black 1.38 (1.13 1.68) Gender Female 1.0 0.1942 Male 1.12 (0.94 1.33) Morphology Squamous cell carcinoma 1.0 0.1994 Adenocarcinoma 0.87 (0.70 1.08) Histologic grade (1 4) 1.08 (0.96 1.21) 0.2121 Tumor length (1 5) 1.15 (1.08 1.21) 0.0001 a Primary site Cervical, upper one third, middle one third, thoracic 1.0 0.0313 a Lower one third, abdominal 1.22 (1.02 1.47) Invasiveness (T1 T2) 0.86 (0.71 1.04) 0.1115 95% CI: 95% confidence interval. a P 0.05. Age at diagnosis (10 yrs) 1.15 (1.06 1.25) 0.0013 a Race White 1.0 0.0803 Black 1.28 (0.97 1.68) Gender Female 1.0 0.8714 Male 0.98 (0.79 1.23) Morphology Squamous cell carcinoma 1.0 0.3181 Adenocarcinoma 1.10 (0.91 1.34) Histologic grade (1 4) 1.06 (0.93 1.22) 0.3102 Tumor length (1 5) 1.07 (1.00 1.15) 0.0555 Primary site Cervical, upper one third, middle one third, thoracic 1.0 0.0001 Lower one third, abdominal 1.55 (1.24 1.93) Invasiveness (T1 T4) 1.17 (1.07 1.30) 0.0013 a No. of positive lymph nodes 5 1.0 0.0124 a 5 1.29 (1.06 1.56) Proportion of positive lymph nodes 0.10 1.0 0.0013 a 0.10 1.63 (1.26 2.11) 95% CI: 95% confidence interval. a P 0.05. trasound (EUS), and minimally invasive surgery, allow accurate preoperative staging and provide prognostic information for overall survival. More emerging data suggest that both the location and the number of involved lymph nodes are important prognostic factors. We explored these factors using a well-defined population of patients with esophageal carcinoma in the United States using the National Cancer Institute SEER data base. Across all stages, patient age and tumor site were important predictors of survival with esophageal carcinoma. The major factors that predicted survival for patients with localized disease were ranked in the order of their relative contribution to the multivariable analysis (chi-square test); these included age at diagnosis, tumor length, race, and disease site. For patients with regional disease, the significant predictors in ranked order included disease site, proportion of positive lymph nodes, depth of invasion, age at diagnosis, and number of positive lymph nodes. The factors of importance for patients with distant disease were age at diagnosis, site of metastasis, tumor grade, race, and morphology. It is interesting to note that tumor length was an independent risk factor controlling for depth of invasion in patients with localized disease. The survival of these patients worsened as the tumor became longer. For example, the 2-year survival for patients with tumors measuring 1 cm in greatest dimension was 78% compared with 18% in patients with tumors measuring 9 10 cm in greatest dimension. The fact that patients with tumors measuring 10 cm have a slightly better survival compared with patients in the preceding group may be attributed to the small number of patients in this group. More importantly, in patients with lymph node negative disease, those with shorter tumors had a better survival compared with patients who had longer tumors. For example, the 2-year survival rate for patients with 1-cm tumors was 84% compared with a rate of 14% for patients with tumors measuring 10 cm. It is important to note, however, that the correlation between tumor length and survival is not linear and appears to have cut-off points, breaking into three groups: patients with short tumors measuring 3 cm in greatest dimension, patients with tumors measuring 3 10 cm in greatest dimension, and patients with tumors measuring 10 cm in greatest dimension. For patients with regional disease, tumor length was of borderline significance (P 0.0555), which also was true for patients with distant disease (P 0.0830). Increasing depth of invasion was a predictor of regional lymph node metastasis and poorer survival. However, increasing depth of invasion reflected longitudinal tumor growth. It is suggested that longitudinal growth in the lymphatic-rich submu-

1440 CANCER October 1, 2002 / Volume 95 / Number 7 FIGURE 2. Kaplan Meier survival curves for 1340 patients with lymph node status stratified by the actual number of metastatic lymph nodes. TABLE 5 Prognostic Factors for Survival from Distant Esophageal Carcinoma: Cox Multivariate Model Factor Hazard ratio (95% CI) P value Age at diagnosis (10 yrs) 1.17 (1.09 1.26) 0.0001 a Race White 1.0 0.0328 a Black 1.28 (1.02 1.60) Gender Female 1.0 0.5544 Male 1.06 (0.87 1.29) Morphology Squamous cell carcinoma 1.0 0.0067 a Adenocarcinoma 1.21 (1.05 1.38) Histologic grade (1 4) 1.17 (1.03 1.32) 0.0152 a Tumor length (1 5) 1.05 (0.99 1.12) 0.0830 Primary site Cervical, upper one third, middle one third, thoracic 1.0 0.3332 Lower one third, abdominal 1.10 (0.91 1.34) Distant site Lymph node 1.0 0.0008 a Other 1.37 (1.14 1.65) 95% CI: 95% confidence interval. a P 0.05. cosa that may be the most important factor resulting in regional lymph node metastasis. 14 It is likely that, beyond a certain length, micrometastases already have occurred. Therefore, a 3-cm tumor length cut-off point was used in the proposed modified TNM classification system (Table 7). We also showed that, the higher the number of lymph nodes involved, as well as the ratio of positive lymph nodes, the poorer the survival of patients with esophageal carcinoma. One study 6 found that the number of involved lymph nodes, irrespective of the depth of tumor invasion, was correlated with the subsequent development of systemic metastasis: 0% for patients with no involved lymph nodes, 45% for patients with 1 4 involved lymph nodes, 80% for patients with 5 9 involved lymph nodes, and 92% for patients with 10 involved lymph nodes. In addition, the presence and the number of lymph node metastases in these patients was the most significant predictor of survival. 6 Those investigators found that the 5-year survival rate was 23% in patients with 4 involved lymph nodes, but the rate decreased to 12% for patients with 4 lymph node metastases. Correspondingly, the likelihood of developing recurrent disease increased from 53% to 90%, respectively. 6 Patients with 10 involved lymph nodes had a particularly poor prognosis, and all developed recurrent disease or died of their disease. Patients with a ratio of the number of lymph nodes removed to the number involved after en bloc resection of 0.1 had a significantly better survival rate. 6 Korst and colleagues 5 evaluated the influence of the number and location of involved lymph nodes on survival in 195 patients with esophageal carcinoma. In their study, 5 patients with no lymph node involvement had the best survival compared with other groups. Although there was no significant difference in survival between patients who had one, two, or three positive lymph nodes, there was a significant difference in survival between this overall group and patients who had four or more positive lymph nodes (HR, 1.77). The importance of this finding lies in the fact that these lymph

Prognostic Factors in Esophageal Carcinoma/Eloubeidi et al. 1441 FIGURE 3. Kaplan Meier survival curves for 1471 patients with clinically localized esophageal carcinoma subgrouped by tumor length. TABLE 6 One-Year and 2-Year Survival Rates from Esophageal Carcinoma for Patients with Localized Stage Disease and Patients with Known Lymph Node Negative Disease Stratified by Tumor Length Tumor size (cm) Clinical localized disease (n 1471 patients) Lymph node negative disease (n 580 patients) One-year survival rate SE Two-year survival rate SE One-year survival rate SE Two-year survival rate SE 1 81.6 3.8 71.8 4.7 89.4 3.8 83.7 4.8 1.1 2.0 63.6 4.1 50.8 4.3 78.4 5.0 75.0 5.3 2.1 3.0 57.9 3.2 35.6 3.3 73.0 4.6 57.1 5.3 3.1 4.0 62.3 3.4 39.0 3.5 67.5 4.6 46.1 5.0 4.1 5.0 52.5 3.3 34.4 3.3 72.2 5.9 44.4 6.9 5.1 6.0 48.7 4.2 26.8 3.9 59.6 6.7 42.7 7.0 6.1 7.0 50.0 4.8 33.8 4.6 58.8 7.5 54.1 7.6 7.1 8.0 39.9 6.2 18.2 5.2 53.2 10.0 40.2 10.0 8.1 9.0 46.8 8.2 35.8 8.0 56.3 13.6 40.2 13.9 9.1 10.0 32.9 6.0 17.8 5.2 55.6 16.8 55.6 16.8 10.0 54.1 8.8 36.2 8.9 47.6 15.2 14.3 13.4 SE: standard error. nodes now can be identified accurately and preoperatively by EUS. One recent study found that the number of lymph nodes detected by EUS was correlated with survival. Patients with greater than three peritumoral lymph nodes detected by EUS had a worse prognosis compared with patients who had less than three lymph nodes detected. 15 In patients with distant metastasis, we detected a survival difference according to the site of metastasis. A metastasis to a solid organ indicated a worse prognosis compared with metastasis to a distant lymph node. This observation has been corroborated by other smaller studies. 5,7 The current TNM classification system clearly makes that distinction, because M1a disease designates celiac lymph node disease for distal carcinoma, and M1b designates metastasis to solid organs, like the liver. 10 In this analysis, the exact site of distant lymph nodes could not be determined. However, according to the revised 1997 AJCC classification system, patients with M1a disease are considered to have metastatic disease. A recent study 16 evaluated the survival of patients with esophageal carcinoma who had M1a/M1b disease. Christie and colleagues 16 showed that, even though the survival of patients with M1a disease was 6 months longer compared with the survival of patients with M1b disease,

1442 CANCER October 1, 2002 / Volume 95 / Number 7 FIGURE 4. Observed and predicted 2-year mortality rates for 1471 patients with clinically localized esophageal carcinoma based on a mathematic model ( 0.3250t 0.021001t2) f(t) 1 0.942e derived from the Surveillance, Epidemiology, and End Results program data base. In the model, t is the measured tumor length, and f(t) is the 2-year disease specific mortality rate. TABLE 7 Proposed new TNM Classification System for Esophageal Carcinoma Primary tumor (T) Tx: Tumor cannot be assessed T0: No evidence of primary tumor Tis: Carcinoma in situ T1: Tumor invades lamina propria or submucosa T2: Tumor invades muscularis propria T3: Tumor invades adventitia T4: Tumor invades adjacent structures T a : Tumor length 3cm T b : Tumor length 3cm Regional lymph nodes (N) Nx: Regional lymph nodes cannot be assessed N0: No regional lymph node metastasis N1: Regional lymph node metastasis N1 a : 5 lymph nodes N1 b : 5 lymph nodes Distant metastasis (M) Mx: Distant metastasis can not be assessed M0: No distant metastasis M1: Distant metastasis Tumors of the lower thoracic esophagus M1a: Metastasis in celiac lymph nodes M1b: Other distant metastasis Tumors of the midthoracic esophagus M1a: Not applicable M1b: Nonregional lymph nodes and/or other distant metastasis Tumors of the upper thoracic esophagus M1a: Metastasis in cervical lymph nodes M1b: Other distant metastasis the authors did not consider it clinically important. We also showed that preoperative evaluation with EUS can predict long-term survival in patients who have esophageal carcinoma with or without celiac axis lymph node involvement. Patients with N1 disease or with celiac adenopathy, as detected by EUS, had a worse survival compared with patients who had N0 disease or patients without celiac adenopathy. 17 The strength of this analysis is the large number of patients from a population-based study and is reflective of the current practices and population in the United States. It included patients with both esophageal adenocarcinoma and squamous cell carcinoma. However, we note the limitation of this study. Of patients with localized disease, paradoxically, T-stage 2 versus T-stage 1 was associated with lower mortality controlling for other factors, although the difference was not significant. This may have been due in part to misclassification. The extent of disease variable for T staging included classifying 65.7% of the patients with localized disease as localized, NOS, and these patients were classified as T1 by the investigators, although they may have included patients with disease to the muscularis propria, which would be classified T2. With the current data base, there is no mechanism for detecting these patients, which may comprise 10 20% of the patients in this category. In addition, accurate staging technology, such as EUS, was not used widely. Therefore, the accuracy for the classification of T classificationin the nonsurgical group probably was compromised. Finally and more importantly, the extent of surgical dissection of lymph nodes is not well docu-

Prognostic Factors in Esophageal Carcinoma/Eloubeidi et al. 1443 mented. It is not clear to us how many patients underwent radical, two-field or three-field lymphadenectomy. A more extensive lymphadenectomy results in the removal a greater number of uninvolved lymph nodes, thereby driving down the lymph node ratio. This emphasizes the fact that the critical number is not the ratio itself but, instead, the total number of involved and removed lymph nodes. Consequently, a systematic lymphadenectomy is fundamental to allow an accurate assessment of the patient s lymphatic tumor burden and to determine the likelihood of longterm survival. In conclusion, in addition to the well-described factors that have an impact on the survival of patients with esophageal carcinoma, we have shown that tumor length, the number of lymph nodes, and the ratio of positive lymph nodes are important prognostic factors for the survival of these patients. A revised TNM classification system for esophageal carcinoma may consider tumor length and the number of positive lymph nodes as two important prognostic factors. Our results need to be validated by prospectively collecting information in a more standardized fashion on tumor length, depth of tumor invasion (by EUS), number of lymph nodes, and status in patients with esophageal carcinoma who undergo surgery to further delineate the correlation of these important variables with mortality. REFERENCES 1. Greenlee RT, Hill-Harmon MB, Murray T, Thun M. Cancer statistics, 2001. Ca Cancer J Clin. 2001;51:15 36. 2. Lightdale CJ. Esophageal cancer. Am J Gastroenterol. 1999; 94:20 29. 3. Rice TW, Adelstein DJ. Precise clinical staging allows treatment modification of patients with esophageal carcinoma. Oncology (Huntingt). 1997;11:58 62. 4. Japanese Committee for Registration of Esophageal Carcinoma. A proposal for a new TNM classification of esophageal carcinoma. Jpn J Clin Oncol. 1985;15:625 636. 5. Korst RJ, Rusch VW, Venkatraman E, et al. Proposed revision of the staging classification for esophageal cancer. J Thorac Cardiovasc Surg. 1998;115:660 669. 6. Nigro JJ, DeMeester SR, Hagen JA, et al. Node status in transmural esophageal adenocarcinoma and outcome after en bloc esophagectomy. J Thorac Cardiovasc Surg. 1999;117: 960 968. 7. Steup WH, De LP, Deneffe G, et al. Tumors of the esophagogastric junction. Long-term survival in relation to the pattern of lymph node metastasis and a critical analysis of the accuracy or inaccuracy of ptnm classification. J Thorac Cardiovasc Surg. 1996;111:85 94. 8. Blot WJ, Devesa SS, Kneller RW, Fraumeni JF Jr. Rising incidence of adenocarcinoma of the esophagus and gastric cardia [see comments]. JAMA. 1991;265:1287 1289. 9. Blot WJ, Devesa SS, Fraumeni JF Jr. Continuing climb in rates of esophageal adenocarcinoma: an update. JAMA. 1993;270:1320. 10. American Joint Committee on Cancer. AJCC cancer staging manual, 5th ed. Philadelphia: Lippincott-Raven, 2000:65 68. 11. Percy C, Van Holten V, Muire C. International classification of diseases for oncology, 2nd ed. Geneva: World Health Organization, 1990. 12. National Cancer Statistics Branch. Surveillance, Epidemiology and End Results (SEER) Program: public-use CD-ROM (1973 1997). Bethesda: National Cancer Statistics Branch, 2000. 13. Rice TW, Blackstone EH, Adelstein DJ, et al. N1 esophageal carcinoma: the importance of staging and downstaging. J Thorac Cardiovasc Surg. 2001;121:454 464. 14. Rice TW, Zuccaro GJ, Adelstein DJ, et al. Esophageal carcinoma: depth of tumor invasion is predictive of regional lymph node status. Ann Thorac Surg. 1998;65:787 792. 15. Burtin P, Kaassis M, Aube K, et al. ASA classification and lymph node extent at endosonography are independent predictive factors of survival of patients with esophageal cancer [abstract]. Gastrointest Endosc. 2000;51:AB119. 16. Christie NA, Rice TW, DeCamp MM, et al. M1a/M1b esophageal carcinoma: clinical relevance. J Thorac Cardiovasc Surg. 1999;118:900 907. 17. Eloubeidi MA, Wallace MB, Hoffman BJ, et al. Predictors of survival for esophageal cancer patients with and without celiac axis lymphadenopathy: impact of staging endosonography. Ann Thorac Surg. 2001;72:212 219.