The current tumor-node-metastasis (TNM) staging

Transcription

1 CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2006;4: Influence of the Number of Malignant Regional Lymph Nodes Detected by Endoscopic Ultrasonography on Survival Stratification in Esophageal Adenocarcinoma JAIME CHEN,* RONGHUI XU, GORDON C. HUNT,* MARY LEE KRINSKY,*, and THOMAS J. SAVIDES* *Division of Gastroenterology, Department of Internal Medicine, University of California, San Diego; Department of Family and Preventive Medicine, Division of Biostatistics and Bioinformatics, and Department of Mathematics, University of California, San Diego; and Section of Gastroenterology, Department of Internal Medicine, San Diego Veterans Affairs Hospital, San Diego, California Background & Aims: The nodal staging of esophageal cancer accounts for the absence or presence of metastatic lymph nodes (N0 or N1, respectively). Surgical data suggest that patients have worse survival when esophagectomy specimens contain higher numbers of regional malignant lymph nodes. It has been proposed that the staging system for esophageal cancer be modified to include the number of malignant lymph nodes. The aim of this study was to determine the influence of the number of malignant-appearing regional lymph nodes detected on endoscopic ultrasonography (EUS) on survival in patients with esophageal adenocarcinoma. Methods: Historical case series involved patients with esophageal adenocarcinoma who underwent EUS staging at a single center between 1994 and Endoscopy reports were reviewed to determine the number of malignant-appearing periesophageal lymph nodes seen on EUS examination. Subjects were categorized as having 0, 1 2, or >2 periesophageal lymph nodes. A regional cancer registry prospectively obtained survival data. Results: Among 85 patients with esophageal adenocarcinoma, the Kaplan-Meier curves showed distinct survival advantages in those with fewer malignant-appearing regional lymph nodes (P.0008). The median survivals were 66 months, 14.5 months, and 6.5 months for 0, 1 2, and >2 malignant-appearing lymph nodes, respectively. Survival was also influenced by celiac lymph nodes and tumor length, both of which were associated with increased number of malignant nodes. Conclusions: The number of malignant-appearing periesophageal lymph nodes detected by EUS is associated with improved survival stratification in patients with esophageal adenocarcinoma and should be considered in the presurgical staging of esophageal cancer. The current tumor-node-metastasis (TNM) staging for esophageal cancer accounts for only absence (N0) or presence (N1) of malignant regional lymph nodes in the nodal (N) staging. This was last revised by the American Joint Committee on Cancer in The staging systems for other mucosal gastrointestinal cancers, such as gastric and colorectal cancers, have been modified to incorporate the number of malignant lymph nodes for better staging accuracy. 1 During the past decade, several large surgical series have shown that esophageal cancer survival stratification within N1 could be improved by taking into account the number of malignant regional lymph nodes. Three retrospective surgical case series found that patients with esophageal cancer survived less if there were higher numbers of malignant regional lymph nodes at the time of resection. 2 4 Two of those 3 studies had mixed cohorts of squamous cell carcinoma and adenocarcinoma. 3,4 Rice et al 5 in 2003 published the results of a prospective study that confirmed the observations of those earlier publications. Their study, evaluating 480 esophageal cancer patients who had undergone esophagectomy (85% adenocarcinoma), demonstrated statistically significant survival advantage in those with 2 or less malignant regional lymph nodes compared with those with higher numbers of malignant nodes. It is unknown whether the same survival correlation holds true with preoperative endoscopic ultrasonography (EUS) determined numbers of periesophageal lymph nodes. EUS is a highly accurate method for nodal (N) staging of esophageal cancer, with 75% accuracy by using endosonographic criteria alone and nearly 90% accuracy when combined with fine-needle aspiration (FNA). 6 EUS offers a less invasive preoperative alternative to esophagectomy and lymphadenectomy for quantitating malignant lymph node numbers. Only 2 EUS Abbreviations used in this paper: EUS, endoscopic ultrasonography; FNA, fine-needle aspiration; GEJ, gastroesophageal junction; TNM, tumor-node-metastasis; UCSD, University of California, San Diego; VAMC, Veterans Affairs Medical Center by the American Gastroenterological Association Institute /06/$32.00 doi: /j.cgh

2 574 CHEN ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 4, No. 5 studies exist in the literature to date that look at quantitating malignant lymph nodes in the survival stratification of esophageal cancer. 7,8 Neither specifically focused on esophageal adenocarcinoma, which is the predominant type in North America. This purpose of this study was to determine whether the number of malignant-appearing regional lymph nodes on EUS correlates with survival in patients with esophageal adenocarcinoma. Because celiac lymph node metastases and tumor length also have prognostic implications for survival, the secondary aim of the study was to determine whether a relationship exists between these 2 factors and the number of malignant periesophageal lymph nodes. Methods Study Design This is a historical case series using existing medical record data at the University of California, San Diego (UCSD) Healthcare System and the San Diego Veterans Affairs Medical Center (VAMC) Hospital. This study was approved by the Human Research Protections Program at both institutions. All EUS examinations were performed by 1 of 3 experienced endosonographers. Case Identification Potential cases were identified via an electronic search of existing gastrointestinal endoscopy electronic medical records at UCSD and the San Diego VAMC from January 1, 1994 January 1, The search parameters used were for upper EUS procedures with the indications of tumor of the esophagus or gastroesophageal junction (GEJ). The patient population consisted of those from within the UCSD or VA healthcare systems as well as outside referrals from the community practices in the San Diego area. Endoscopic Ultrasonography Technique Protocol At our institution, EUS examinations for esophageal cancer begin with a standard forward viewing video endoscope to assess the location of the tumor and degree of stenosis. The tumor length is recorded as part of the endoscopy report. EUS examination is then performed with the radial endoechoscope (Olympus GF-UM20 with a UM20 ultrasound processor; Olympus Corporation, Lake Success, NY) or an Olympus GF-UM130 echoendoscope with an Olympus EUS-M30 ultrasound processor (Olympus Corporation). The endosonographers in this study adhered to the standard criteria for determination of malignant-appearing periesophageal lymph nodes. These endosonographic parameters included size greater than 1 cm in the short axis, hypoechoic echotexture, well-defined borders, and rounded shape. 9 The endosonographers reported whether they thought the lymph nodes were malignant by taking into account all 4 EUS criteria, but they generally did not report specifically which criteria were used. EUS-guided transesophageal FNA was not routinely performed into the lymph nodes. Esophageal dilation was performed at the discretion of the gastroenterologist if stenosis prevented the passage of the endoechoscope. An over-the-wire, non-video 8-mm echoscope was available for use (Olympus MH908; Olympus Corporation) at the discretion of the endosonographer in instances in which the echoendoscope could not traverse the tumor. Inclusion and Exclusion Criteria To be included in the study, patients had undergone a prior computed tomography scan that did not show any obvious distant metastatic disease. In cases in which multiple EUS procedures were performed (ie, before and after chemoradiation), only the initial EUS procedure was recorded. Only patients who had not received any therapy (surgery, chemotherapy, or radiation therapy) were included in this study. The endoscopy report and/or pathology report had to state a confirmed diagnosis of adenocarcinoma of the esophagus or GEJ. For tumors labeled as GEJ on the endoscopy database, the tumor distribution was reviewed on the endoscopy report. We included only GEJ tumors that had more than half of the tumor mass on the esophageal side. This is based on the literature that showed that GEJ junctional tumors defined in this manner behave like cancers of the esophageal body. 10 There had to be survival data from the UCSD or San Diego VAMC institutional tumor registry for a subject to be included in our study. Data Collection Data were abstracted by a single investigator ( J.C.) by using standardized collection forms designed for this study. At both UCSD Healthcare and the San Diego VAMC, there was a graded transition of medical records from paper charts to computer-based electronic medical records during the study period. Electronic endoscopy databases were integrated into the main systems within this time period as well. As a result, sources of medical record data included the paper chart, the electronic chart, and electronic endoscopy databases. The study subjects were assigned unique study numbers such that the data were completely de-identified during analysis. The EUS procedure report was used to establish the date of index examination and the EUS-assessed tumor stage (T), nodal stage (N), and possible metastatic stage (M). Information on the number of malignant-appearing regional lymph nodes was obtained from each EUS report. The exact number of malignant-appearing lymph nodes by EUS was recorded, if reported. This was often the case if only 1 or 2 malignant nodes were seen on EUS examination. If descriptors such as several or numerous or many were used to describe the number of malignant lymph nodes, then it was designated as 2. Malignant celiac lymph nodes were included in the total lymph node count. The presence of enlarged celiac lymph nodes on

3 May 2006 ESOPHAGEAL CANCER SURVIVAL BASED ON EUS LYMPH NODE NUMBERS 575 EUS and the endoscopically determined length of the tumor were recorded. For each subject TNM criteria were applied to the EUS findings to assign the patients to EUS-based stage I, II, III, or IV cancer, on the basis of the th Edition American Joint Committee on Cancer, Cancer Staging Manual. 1 Survival Data Survival data were obtained from the UCSD and San Diego VAMC Cancer Registry offices. These offices are a part of the California Cancer Registry network. In the state of California all cancers are a required reportable disease, and all cancer patients are followed indefinitely by their local tumor registry office until the time of patient death or loss of contact. These cancer registry offices are staffed by trained personnel who prospectively track cancer patients by checking routinely with primary physicians, examining medical charts and records, contacting patients by letter, and examining public records such as Social Security claims status and information from the Department of Motor Vehicles. Survival data are provided as the date of death or the date of last living contact. For this study, death was defined as all-cause mortality. The date of patient death was recorded for those who had died. Those patients who were still living had their date of last living contact recorded. The date of index EUS examination was assigned as time zero in calculating survival time. All data from the study forms were entered into a computer database. Data Analysis The following patient characteristics were considered for possible association with overall survival, where overall survival was defined as the time between the initial EUS examination and the time of death or last living contact. Lost to follow-up cases were handled as censored in the survival analysis. The malignant-appearing periesophageal lymph nodes were categorized as 0, 1 2, or 2; the basis for these cutoffs are adapted from the surgical literature. 5 Tumor length was determined endoscopically and categorized into 2 groups ( 2.9 cm vs 3 cm). This cutoff at 3 cm was based on results from a previous study indicating that a survival difference was most significant about this threshold. 11 Finally, celiac lymph node involvement and tumor location were also considered for possible association with survival. Univariate survival analysis was performed by plotting the Kaplan-Meier curves of survival probabilities for each category of the predictor variables, accompanied by a log-rank test for difference in survival among the categories. A multivariate proportional hazards model was used to assess the independent predictability of each variable in the presence of the others, as well as their joint ability to predict overall survival. The proportional hazard assumption was checked by plotting the log-log of the survival curves of each subgroup. 12 Crosstabulations were given to describe the association between the numbers of malignant periesophageal lymph nodes and both celiac lymph node metastasis and length of esophageal tumor, accompanied by Fisher exact test for the association. Results Study Population Between the years of 1994 and 2004, 85 subjects were identified who had undergone EUS examination for a confirmed diagnosis of esophageal adenocarcinoma and were accounted for by the cancer registry. Fifty-three of the 85 subjects had documented mortality at the time of last subject follow-up by the cancer registry. Subject Demographics The median age was 65 years (range, years). There were 57 white subjects, 8 non-white subjects, and 20 whose race was not disclosed in the hospital information system electronic medical record. Seventyeight subjects were men. Among 85 patients with esophageal adenocarcinoma, the EUS-TNM stage was 19% T1, 26% T2, 42% T3, and 13% T4, 69% had N1 disease, and 12% had celiac lymph nodes. Treatment data were available for 31 of the 85 patients. Chemoradiation was the sole therapy in 15 of those 31 patients. An additional 4 patients received chemoradiation along with esophagectomy. Six other patients had esophagectomy only. Surgical resection was performed only in those with stage I III cancers. Other treatments that were rendered included photodynamic therapy, tumor ablation with argon plasma coagulation, endoscopic mucosal resection, and palliative stenting. Nontraversable strictures were encountered in 13 cases, 8 of which had attempted dilations. Three dilations were successful in eventually allowing the echoendoscope to traverse the stricture. The over-the-wire, nonvideo 8-mm echoscope (Olympus MH908) was used in 1 patient. A total of 10 of 85 (12%) patients did not have their strictures traversed by the echoendoscope. Outcomes by Regional Lymph Node Status Having more EUS malignant-appearing regional lymph nodes was associated with decreased survival (Figure 1). The median survivals were 66 months, 14.5 months, and 6.5 months for 0, 1 2, and 2 malignant lymph nodes, respectively. Outcomes by Endoscopic Ultrasonography Based Tumor-Node-Metastasis Stage and Tumor Location Advanced EUS TNM stage was associated with decreased survival (Figure 2). Median survival by EUS stage was as follows: 66, 17.5, 10.5, and 11.5 months for stages I, II, III, and IV, respectively. Survival was not statistically significantly different depending on whether the tumor was located in the esophageal body (n 65)

4 576 CHEN ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 4, No. 5 Figure 1. Kaplan-Meier curve showing survival for esophageal adenocarcinoma based on the number of malignant-appearing periesophageal lymph nodes (LN) detected by EUS. Figure 3. Kaplan-Meier curve showing survival for esophageal adenocarcinoma based on EUS evaluation of celiac lymph nodes (CLN). or at the GEJ (n 20), with median survival time of 18.5 and 12.5 months, respectively (P.11). Outcomes by Celiac Lymph Node Status Of the 85 patients enrolled in this study, the celiac lymph node status was available for 75 patients. Ten patients had strictures that prevented evaluation of the celiac lymph node status. The presence of malignantappearing celiac lymph nodes was associated with worse survival (Figure 3). The median survival was 21 months for patients without celiac lymph node metastases compared with a median of 11.5 months for those with celiac lymph node metastases (P.04). Cross-tabulations and multivariate analysis showed that there is a statistically significant association between having malignant celiac lymph nodes and a higher likelihood of regional lymph node involvement (P.01) (Table 1). Outcomes by Tumor Length There were tumor length data for 65 subjects. Patients with tumor lengths 2.9 cm had a better prognosis than those with tumor lengths 2.9 cm (Figure 4). Subanalysis of tumor lengths 2.9 cm showed that there was no statistical significance in patient survival for those who had tumors 3 6 cm long compared with those with tumors greater than 6 cm. With crosstabulation and multivariate analysis, the association between having longer tumor length and regional lymph node metastases (Table 2) was statistically significant (P.01). Multivariate Proportional Hazards Regression A multivariate proportional hazards model was fitted to the data with all 4 predictors: regional lymph node status, TNM stage, celiac lymph node status, and tumor length. Because of the very limited sample size and, in particular, number of deaths, the individual predictor did not remain significant for overall survival while adjusting for other predictors. However, the overall prediction model with all 4 predictors was significant in predicting overall survival, with a P value of.048 for Table 1. Cross-Tabulation of Determined Celiac Lymph Node Status to the Number of EUS-Detected Malignant-Appearing Regional Lymph Nodes Nonmalignant CLN (n 66) Malignant CLN (n 9) Figure 2. Kaplan-Meier curve showing survival for esophageal adenocarcinoma based on EUS-determined TNM cancer stage. 0 regional metastatic LN 25/66 (38%) 0/9 (0%) 1 2 regional metastatic LN 24/66 (36%) 3/9 (33%) 2 regional metastatic LN 17/66 (26%) 6/9 (67%) CLN, celiac lymph nodes; LN, lymph nodes.

5 May 2006 ESOPHAGEAL CANCER SURVIVAL BASED ON EUS LYMPH NODE NUMBERS 577 Figure 4. Kaplan-Meier curve showing survival based on endoscopic determination of tumor length of esophageal adenocarcinoma. the likelihood ratio test against the null model of no predictors. Discussion This is the first study to evaluate the relationship between the number of EUS-detected malignant-appearing periesophageal lymph nodes and survival in patients with esophageal adenocarcinoma. Only 2 other EUS studies have looked at the relationship between malignant lymph node count and patient survival in esophageal cancer. 7,8 Burtin et al 8 showed that malignant regional lymph node count was a predictor of survival in patients with esophageal cancer. The study was a retrospective series with a European patient population. The type of esophageal cancer was not specified, and the data were only presented in abstract form. In a more recent series involving 313 patients from Japan, Natsugoe et al 7 found that the combined number of both regional and distant metastatic lymph nodes correlated with patient survival. The data reported worse 5-year survival for those with higher numbers of malignant lymph nodes. Squamous cell cancers (94%) accounted for the majority of that study. Our study is unique because it strictly evaluates esophageal adenocarcinoma, which is the most common form of esophageal cancer in the United States. We confirmed that higher counts of malignant-appearing lymph nodes as determined by EUS correlated with shorter patient survival. The results of our EUSbased study were consistent with the surgical experience of Rice et al. 5 Their survival rate for 2 or less nodes was 20% at 5 years, whereas those with greater than 2 nodes had 5% survival rate at 5 years. Our results were very similar (Figure 1). In this study we used the same cut-off values as Rice et al. Previous retrospective surgical series had proposed various other cutoff values as well as using the ratio of malignant to total lymph nodes dissected. We chose to adapt the format of Rice et al because it was a prospective study on a North American population and had robust statistical analysis. Our TNM-stage survival curves confirmed that our study population was similar to those of other published cohorts. We included GEJ tumors if the majority of the tumor burden was on the esophageal side. The literature suggests that GEJ tumors defined in this manner behave similarly to esophageal body tumors. 10 Of the 85 cases of esophageal cancer in this study, 20 were located in the GEJ. Subanalysis disclosed no statistically significant difference in survival seen between those with GEJ tumors and those with esophageal body tumors. The poor prognosis associated with having malignant celiac lymph nodes has been well-described in the surgical literature and more recently with EUS. In our study, patients with malignant celiac node involvement tended to have higher numbers of periesophageal lymph node metastases. It raises the possibility that malignant celiac lymph nodes confer worse prognosis because they are generally associated with a greater number of lymph node metastases in patients with esophageal adenocarcinoma. Recently tumor length has been proposed as an independent predictor of survival. Of 1471 patients with clinically localized disease, Eloubeidi et al 11 found that longer tumors correlated with shorter survival. The results were not statistically significant, and the survival curve seemed to level off beyond 3-cm length. This association was not seen in another study in which the cohort all had malignant celiac lymph nodes. 13 Our study showed that tumor length correlated with survival in a statistically significant manner, and that tumor length is associated with a greater number of lymph node metastases in patients with esophageal adenocarcinoma. Further prospective study will be required to define better the influence of esophageal cancer length on survival. Table 2. Cross-Tabulation of Endoscopically Determined Esophageal Adenocarcinoma Tumor Length to the Number of Malignant-Appearing Regional Lymph Nodes Detected With EUS Length, cm (n 20) Length, 2.9 cm (n 45) 0 regional metastatic LN 12/20 (60%) 9/45 (20%) 1 2 regional metastatic LN 6/20 (30%) 20/45 (44%) 2 regional metastatic LN 2/20 (10%) 16/45 (36%) LN, lymph nodes.

6 578 CHEN ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 4, No. 5 There are several limitations to this study. The endosonographers based their decision about whether a lymph node was involved on the basis of standard criteria for nodal staging of esophageal cancer including size greater than 1 cm in the short axis, hypoechoic echotexture, well-defined borders, and rounded shape. 9 If a lymph node has all of these features, there is an 80% 100% chance of malignancy; however, only 25% of lymph nodes have all these features. 9,14 It is important to note that sometimes proximity of the lymph node to the esophageal tumor might influence the endosonographer s decision about whether the lymph node appears malignant, although this has not been studied formally. In general, endosonographers use a combination of these criteria to make a subjective decision about whether periesophageal lymph nodes appear malignant. It is reported that the overall N-staging accuracy for esophageal cancer lymph nodes with diagnostic EUS is approximately 75%. 15 The use of EUS-guided FNA might increase this accuracy of regional lymph node staging slightly to approximately 85%. 6 The patients received a variety of therapies that could impact survival. The retrospective nature of this study prohibited knowledge of all treatments received, and the cancer registry did not record specific therapies. Despite this limitation, the survival based on EUS stage is similar to other published data, suggesting the patient population studied was similar to that in other published trials. 5,13,16 In the San Diego area, esophageal adenocarcinoma patients with EUS staging consistent with T3 and/or N1 disease generally undergo preoperative chemoradiation. Those with T1, N0 or T2, N0 usually undergo resection followed by possible postoperative chemoradiation, whereas those with T4 or M1b disease usually are not offered surgery. Individualized decisions are made for patients with M1a esophageal adenocarcinoma regarding whether they would undergo surgery after chemoradiation. Another limitation of this study was that the exact numbers of malignant nodes were not always reported. The endosonographers all performed systematic EUS evaluation to look for the presence of malignant-appearing lymph nodes. In abstracting data for this study, it was noted that 1 or 2 malignant nodes were almost always reported numerically. Beyond 2 nodes there was more variability in how the quantity was reported, sometimes numerically and other times descriptively. For instance, 3 or 4 malignant nodes were sometimes reported numerically, but often the designation a few or several was used instead. In this study an assumption was therefore made that descriptors such as few and several meant having more than 2 malignant lymph nodes. The 3 endosonographers in this study agreed that this was a reasonable assumption of how they wrote their reports. This underscores that practicing endosonographers often describe the general number of malignantappearing lymph nodes, but they are not consistently reporting an exact number of lymph nodes because the staging system has not required this. However, it seems reasonable that endosonographers can report the number of malignant-appearing lymph nodes, especially if using categories such as 0, 1 2, or 3, without adding significantly to procedure difficulty or time. This is consistent with what should currently be done for EUS staging of rectal and gastric cancers. Another limitation of this retrospective study regarding classifying the lymph nodes as malignant-appearing was that the morphologic features were not systematically recorded. The general practice of the endosonographers was not to describe in detail the morphologic appearance of each lymph node or to record whether it was peritumoral. Instead, the endosonographers simply recorded whether the nodes appeared to them likely to be malignant (on the basis of criteria such as short-axis diameter greater than 1 cm, round shape, hypoechoic echotexture, well-defined borders, and close association to the tumor). In addition, although the endosonographers might have recorded the location of the malignantappearing lymph nodes from the incisors, they did not necessarily report whether the node was immediately adjacent to the tumor mass. Given these limitations, these results reflect more of a real-world scenario in which decision making is subjective and made by the endosonographer on the basis of the overall appearance and not just on isolated morphologic features. Total lymph node count included both periesophageal and potentially celiac lymph nodes. The reason this was done was that this is similar to the surgical resection literature, and because it is sometimes very difficult to decide how to classify lymph nodes around a distal esophageal junction or gastroesophageal in terms of whether they are peritumoral lymph nodes versus celiac lymph nodes. Including all lymph nodes in the total count makes this EUS staging much simpler than trying to subclassify the nodes. FNA was not routinely performed for several reasons. Intervening tumor within the esophagus did not always make FNA possible without the risk of tumor tracking. During the earlier years of this study period, EUS/FNA equipment was not available at the institutions. In addition, with particularly high numbers of malignantappearing lymph nodes, it was usually not practical to

7 May 2006 ESOPHAGEAL CANCER SURVIVAL BASED ON EUS LYMPH NODE NUMBERS 579 perform FNA on all those nodes. Despite infrequent use of EUS/FNA, the fact that our survival curves were comparable to those from the study by Rice et al 5 suggests that the EUS assessment of lymph node malignancy was sufficiently accurate by endosonographic features alone. It is likely that if EUS-guided FNA were performed of the malignant-appearing lymph nodes that were separate from the primary esophageal wall tumor, that a more accurate evaluation of the number of malignant lymph nodes could be achieved. 6 Strictures precluded complete examination in 12% of the cases, even after accounting for stricture dilation. This potentially limited the detection of the total number of malignant-appearing regional lymph nodes as well as determination of the celiac lymph node status. The results from this study were from a single center experience with historical data. Future studies should emphasize data collection in a prospective manner with a larger sample size so that multivariate analysis can be performed. Age at diagnosis and racial background are other factors that have been demonstrated to affect survival outcome and should also be included in future studies. 11 In conclusion, increasing numbers of EUS-detected malignant-appearing regional lymph nodes in esophageal adenocarcinoma correlated with worse patient survival. This study supports the recommendation for modifying the existing nodal staging system for esophageal cancer to include the number of lymph nodes. This would make the regional nodal staging system for esophageal cancer more similar to those already used for gastric and colorectal cancer. The descriptive quantitation of malignant regional nodes via EUS examination should be incorporated into future nodal staging studies. References 1. Greene FL, Page DL, Fleming ID, eds. AJCC cancer staging manual. 6th ed. New York: Springer, 2002: Roder JD, Bushc R, Stein HJ, et al. Ratio of invaded to removed lymph nodes as a predictor of survival in squamous cell carcinoma of the oesophagus. Br J Surg 1994;81: Holscher AH, Bollschwiler E, Bumm R, et al. Prognostic factors of resected adenocarcinoma of the esophagus. Surgery 1995;118: Korst RJ, Rusch VW, Venkatramen E, et al. Proposed revision of the staging classification for esophageal cancer. J Thorac Cardiovasc Surg 1998;115: Rice TW, Blackstone EH, Rybicki LA, et al. Refining esophageal cancer staging. J Thorac Cardiovasc Surg 2003;125: Vazquez-Sequeiros E, Wiersema MJ, Clain JE, et al. Impact of lymph node staging on therapy of esophageal carcinoma. Gastroenterology 2003;125: Natsugoe S, Yoshinaka H, Shimada M, et al. Number of lymph node metastases determined by presurgical ultrasound and endoscopic ultrasound is related to prognosis in patients with esophageal carcinoma. Ann Surg 2001;234: Burtin P, Kaassis M, Aube C, 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:AB Catalano MF, Sivak MV Jr, Rice T, et al. Endosonographic features predictive of lymph node metastasis. Gastrointest Endosc 1994;40: Steup WH, De Leyn P, 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: Eloubeidi MA, Desmond R, Arguedas MR, et al. Prognostic factors for the survival of patients with esophageal carcinoma in the US: the importance of tumor length and lymph node status. Cancer 2002;95: O Quigley J, Xu R. Encyclopedia of biostatistics, 1st ed. Chichester, UK: Wiley Publishing, 1998: Marsman WA, van Wissen M, Bergman JJ, et al. Outcome of patients with esophageal carcinoma and suspicious celiac lymph nodes as determined by endoscopic ultrasonography. Endoscopy 2004;36: Bhutani MS, Hawes RH, Hoffman BJ. A comparison of the accuracy of echo features during endoscopic ultrasound (EUS) and EUS-guided fine-needle aspiration for diagnosis of malignant lymph node invasion. Gastrointest Endosc 1997;45: Rosch T. Endosonographic staging of esophageal cancer: a review of literature results. Gastrointest Endosc Clin N Am 1995; 5: Reed CE. Surgical management of esophageal carcinoma. Oncologist 1999;4: Address requests for reprints to: Thomas J. Savides, MD, UCSD Thornton Hospital, Division of Gastroenterology, 9320 Campus Point Drive, Mail Code 0956, La Jolla, California Supported by an ASGE Career Development Award (to T.J.S.) and a NCI Grant P30 CA23100 (to R.X.). Presented in part as a poster abstract at the Digestive Disease Week conference in Chicago, Illinois, on May 19, The associated abstract was published in Gastrointest Endosc 2005;61:AB273.