The present staging system for esophageal carcinoma

Transcription

1 Esophageal Carcinoma: Depth of Tumor Invasion Is Predictive of Regional Lymph Node Status Thomas W. Rice, MD, Gregory Zuccaro, Jr, MD, David J. Adelstein, MD, Lisa A. Rybicki, MS, Eugene H. Blackstone, MD, and John R. Goldblum, MD Departments of Thoracic and Cardiovascular Surgery, Gastroenterology, Hematology and Medical Oncology, Biostatistics and Epidemiology, and Anatomic Pathology, The Cleveland Clinic Foundation, Cleveland, Ohio Background. The depth of tumor invasion (T) and regional lymph node status (N) are two factors that define the stage of an esophageal carcinoma. However, the arrangement of staging groups assumes that these factors are independent variables. A retrospective review of 359 consecutive patients undergoing esophageal resection was conducted to define the relationship between T and N and to determine whether T is a significant predictor of regional lymph node metastasis (N1). Methods. Primary treatment was operation without preoperative therapy. There were 295 (82%) adenocarcinomas, 55 (15%) squamous cell carcinomas, and 9 (3%) adenosquamous carcinomas. T status was Tis in 29 (8%) patients, T1 in 65 (18%), T2 in 37 (10%), T3 in 219 (61%), and T4 in 9 (3%). N status was N0 in 161 (45%) patients and N1 in 198 (55%). M status was M0 in 327 (91%) patients and M1 in 32 (9%). Stage was 0 in 29 (8%) patients, I in 58 (16%), IIA in 70 (20%), IIB in 22 (6%), III in 148 (41%), and IV in 32 (9%). Results. The likelihood of N1 disease occurring with increasing T was tested using the trend test. The percentage of patients with N1 disease is 0% for Tis, 11% for T1, 43% for T2, 77% for T3, and 67% for T4 (p < 0.001). This relationship existed for both adenocarcinoma and squamous cell carcinoma. Multivariable analysis identified increasing T, adenocarcinoma, and lack of welldifferentiated histologic features as significant predictors of N1 disease. Compared with a T1 patient, a T2 patient is 6 times more likely to have N1 disease, a T3 patient 23 times, and a T4 patient 35 times. Conclusions. We conclude that for patients with esophageal carcinoma, T is an important predictor of N and this association should be included with other established factors used in clinical staging and treatment decisions. (Ann Thorac Surg 1998;65:787 92) 1998 by The Society of Thoracic Surgeons The present staging system for esophageal carcinoma recognizes the importance of depth of tumor invasion (T) and regional lymph node status (N) in defining stage [1]. These factors are assumed to be independent, and the arrangement of staging groups is based on this hypothesis. Because both increasing T and the presence of regional lymph node metastasis (N1) greatly reduce survival, it is important to explore the relationship between these factors and determine their importance in clinical staging and treatment decisions. The purposes of this study were to define the relationship between depth of tumor invasion and regional lymph node status and to determine whether depth of tumor invasion is a predictor of regional lymph node metastasis. Patients and Methods Patients All records of patients with esophageal carcinoma who underwent esophagectomy at The Cleveland Clinic Presented at the Forty-fourth Annual Meeting of the Southern Thoracic Surgical Association, Naples, FL, Nov 6 8, Address reprint requests to Dr Rice, The Cleveland Clinic Foundation, 9500 Euclid Ave, Cleveland, OH ( ricet@cesmtp.ccf.org). Foundation between January 1, 1983, and August 1, 1997, were reviewed. Patients excluded from the study were those who received any treatment of the primary carcinoma before resection and pathologic staging. Patient age, sex, histologic cell type, histologic differentiation, presence of Barrett s mucosa, presence of signet ring cells, depth of tumor invasion (T), number of regional lymph nodes examined, number of regional lymph nodes with metastasis (N1), regional lymph node status (N), distant metastatic status (M), pathologic stage, and status of the resection margins were recorded. Surgical Therapy Esophageal resection, lymphadenectomy, and reconstruction using the stomach were performed by three different surgical approaches. Patients diagnosed preoperatively to have high-grade dysplasia (Tis N0 M0), stage I carcinomas (T1 N0 M0), or a performance status that would not allow thoracotomy underwent a transhiatal esophagectomy. Lymph node sampling was performed This article has been selected for the open discussion forum on the STS Web site: by The Society of Thoracic Surgeons /98/$19.00 Published by Elsevier Science Inc PII S (97)

2 788 RICE ET AL Ann Thorac Surg ESOPHAGEAL CARCINOMA: T AND N 1998;65: in these patients. Patients judged to have T2 N0 M0 or greater stage carcinoma underwent esophageal resection through a left thoracoabdominal approach, or a right thoracotomy and upper abdominal midline laparotomy. An extensive lymphadenectomy was performed in these patients. Patients with cervical esophageal carcinomas that involved the larynx and pharynx underwent a pharyngolaryngoesophagectomy. The esophagogastric anastomosis was constructed either in the neck or chest by a single-layer sutured or stapled technique. Pathologic Analysis Esophageal resection specimens were evaluated pathologically using a standardized protocol in which the resection margins, the esophageal body, the gastroesophageal junction, and regional lymph nodes were extensively sampled. When gross lesions were identified, up to five sections of each lesion, which included the area of deepest penetration of the esophageal wall, were evaluated. When a gross lesion was not identified, at least ten sections of the esophagus were evaluated. All separately resected lymph nodes and all lymph nodes that were grossly identified in the resection specimen were evaluated pathologically. When small enough, the entire lymph node was submitted, and two levels of that node were examined histologically. Larger lymph nodes were bisected and two levels of each of the two portions of the bisected lymph node were examined. Statistical Analysis Descriptive information is summarized as frequencies and percentages for categorical variables, and as the median and range for continuous variables. Data for the variables studied were available for all patients. A onesided Cochran-Armitage test was used to determine whether increasing T status is associated with increasing risk of N1 disease, overall and by histologic cell type. Logistic regression analysis was used to identify univariable predictors of N1 disease. Variables that were significant in the univariable analysis at p 0.05 and that made clinical sense to include in a model to predict N1 disease were considered in a stepwise logistic regression model. Results for the final multivariable model are summarized as the p value, odds ratio, and 95% confidence interval for the odds ratio, along with probability of N1 disease for each combination of variables in the final multivariable model. Patients with Tis carcinomas were excluded from univariable analysis and multivariable analysis because there was a 0% prevalence of N1 disease in this group. All statistical tests were performed using a 5% level of significance. Results Three hundred fifty-nine patients were studied. Descriptive information for all patients and for histologic cell types is summarized in Table 1. The frequency of regional lymph node metastasis (N1) for each level of tumor invasion (T) is listed in Table 2. A sufficient number of patients with T1 adenocarcinomas allowed subdivision of T1 adenocarcinomas into two groups: (1) adenocarcinomas that breached the basement membrane but with invasion limited to the lamina propria and muscularis mucosa (T1 intramucosal), and (2) adenocarcinomas invading into the submucosa but not deeper (T1 submucosal). The difference in the prevalance of regional lymph node metastasis between patients with T1 intramucosal adenocarcinomas (2.8%) and patients with T1 submucosal adenocarcinomas (20.8%) was unlikely to be caused by chance (p 0.033). The prevalance of regional lymph node metastasis significantly increased with increasing depth of tumor invasion for all patients (p 0.001), for patients with adenocarcinoma (p 0.001), and for patients with squamous cell carcinoma (p 0.034). The number and percentage of patients with regional lymph node metastasis is summarized in Table 3 for clinical and postresection factors other than depth of tumor invasion. Logistic regression analysis identified depth of tumor invasion (p 0.001), adenocarcinoma (p 0.032), absence of well-differentiated carcinoma (p 0.001), no Barrett s esophagus (p 0.001), and signet ring cell (p 0.014) as univariable predictors of regional lymph node metastasis (Table 4). Univariable logistic regression analysis also identified positive resection margin (p 0.001) and total number of nodes resected (p 0.001) as predictors of regional lymph node metastasis (Table 5). Because these factors are not assessed until the time of pathologic staging (after resection), they were not included in the multivariable analysis of clinical predictors of regional lymph node metastasis. Logistic regression analysis identified increasing depth of tumor invasion (p 0.001), adenocarcinoma (p 0.001), and lack of well-differentiated carcinoma (p 0.014) as multivariable predictors of regional lymph node metastasis (Table 4). Compared with a patient with a T1 carcinoma, a patient with a T2 carcinoma is 5.9 times more likely to have regional lymph node metastasis, a patient with a T3 carcinoma is 23 times more likely, and a patient with a T4 carcinoma is 34.8 times more likely to have regional lymph node metastasis than a patient with a T1 carcinoma (Table 4). The probability of regional lymph node metastasis for increasing depth of tumor invasion and combinations of histologic cell type and differentiation are listed in Table 6. Comment T and N The presence of lymphatics in the mucosa makes the esophagus unique among gastrointestinal hollow viscous organs [2 5]. Mucosal and submucosal lymphatics form a complex interconnecting network that extends the length of the esophagus and exceeds capillaries in number [6]. Intermittently, lymphatics pierce the muscularis propria and drain into regional lymph nodes in the periesophageal tissue or directly into the thoracic duct [7]. This unusual lymphatic anatomy allows early and widespread dissemination of esophageal carcinoma (Fig 1). As soon as an esophageal carcinoma breaches the basement

3 Ann Thorac Surg RICE ET AL 1998;65: ESOPHAGEAL CARCINOMA: T AND N 789 Table 1. Descriptive Statistics a Variable All Patients (n 359) AD (n 295) SQ (n 55) AS (n 9) Male 301 (83.8%) 262 (88.8%) 32 (58.2%) 7 (77.9%) Barrett s mucosa 127 (35.4%) 124 (42.0%) 1 (1.8%) 2 (22.2%) T status is 29 (8.1%) 27 (9.2%) 2 (3.6%) 0 (0.0%) 1 65 (18.1%) 60 (20.3%) 5 (9.1%) 0 (0.0%) 1-intramucosal 38 (10.6%) 36 (12.2%) 2 (3.6%) 0 (0.0%) 1-submucosal 27 (7.5%) 24 (8.1%) 3 (5.5%) 0 (0.0%) 2 37 (10.3%) 24 (8.1%) 12 (21.8%) 1 (11.1%) (61.0%) 182 (61.7%) 29 (52.7%) 8 (88.9%) 4 9 (2.5%) 2 (0.7%) 7 (12.7%) 0 (0.0%) N1 198 (55.2%) 170 (57.6%) 23 (41.8%) 5 (55.6%) M1 32 (8.9%) 29 (9.8%) 2 (3.6%) 1 (11.1%) Pathologic stage 0 29 (8.1%) 27 (9.2%) 2 (3.6%) 0 (0.0%) I 58 (16.2%) 54 (18.3%) 4 (7.3%) 0 (0.0%) IIA 70 (19.5%) 43 (14.6%) 23 (41.8%) 4 (44.4%) IIB 22 (6.1%) 16 (5.4%) 5 (9.1%) 1 (11.1%) III 148 (41.2%) 126 (42.7%) 19 (34.6%) 3 (33.3%) IV 32 (8.9%) 29 (9.8%) 2 (3.6%) 1 (11.1%) Poor 141 (39.3%) 123 (41.7%) 13 (23.6%) 5 (55.6%) Moderately poor 34 (9.5%) 32 (10.9%) 2 (3.6%) 0 (0.0%) Moderate 105 (29.2%) 71 (24.1%) 31 (56.4%) 3 (33.3%) Moderately well 10 (2.8%) 6 (2.0%) 4 (7.3%) 0 (0.0%) Well 69 (19.2%) 63 (21.4%) 5 (9.1%) 1 (11.1%) Signet ring cells 44 (12.3%) 43 (14.6%) 0 (0.0%) 1 (11.1%) Positive margins 50 (13.9%) 45 (15.3%) 3 (5.5%) 2 (22.2%) Age at operation (y) 64 (26 85) 64 (26 85) 64 (36 84) 64 (58 75) Lymph nodes resected 11 (0 53) 11 (0 53) 11 (0 36) 7 (3 29) Positive lymph nodes 4 (1 24) 5 (1 24) 2 (1 9) 2 (1 12) Negative lymph nodes 8 (0 52) 8 (0 52) 11 (0 31) 7 (1 28) a Data are shown as n (%) for categoric variables; median (range) for continuous variables. AD adenocarcinoma; AS adenosquamous carcinoma; SQ squamous cell carcinoma. membrane and invades the lamina propria it encounters lymphatics. Deeper invasion into the muscularis mucosa and submucosa increases exposure to lymphatics and the potential for longitudinal lymphatic extension and lateral metastasis to regional lymph nodes (N1) and the thoracic duct [8]. Increasing depth of tumor invasion (T) is predictive of regional lymph node metastasis (N1). However, increasing depth of tumor invasion reflects longitudinal tumor growth. It is longitudinal growth in the lymphatic-rich submucosa that may be the most important factor that results in regional lymph node metastases. Analysis of data available for T1 tumors indicates a small but important prevalance of regional lymph node metastasis in T1-intramucosal tumors. However, once the submucosa is invaded (T1-submucosa), nearly a 10-fold increase occurs in regional lymph node metastasis. Consequently, 98% of T1-intramucosal esophageal carcinomas are potentially resectable for cure, but only 80% of T1- submucosal carcinomas are potentially resectable for cure. Clinical Staging Depth of tumor invasion and regional lymph node status are correlated variables in the determination of the stage of an esophageal carcinoma. Because increasing depth of tumor invasion and regional lymph node metastasis greatly reduce survival, clinical (pretreatment) staging should be used for prognostication and to guide therapy. Unfortunately, a perfect clinical staging tool does not exist. Computed tomographic scanners do not allow an accurate assessment of depth of tumor invasion. In this evaluation, computed tomography is most useful in excluding T4 carcinomas by demonstrating the maintenance of fat planes between the tumor and adjacent structures. In the assessment of regional lymph node status, size of the node is the only factor measured, thus limiting the usefulness of computed tomography in this clinical assessment. Endoscopic esophageal ultrasonography is an excellent clinical staging tool and provides the best clinical estimate of depth of tumor invasion. Multiple factors, including size, shape, border, internal

4 790 RICE ET AL Ann Thorac Surg ESOPHAGEAL CARCINOMA: T AND N 1998;65: Table 2. N1 by T Status Cell Type/T Status Total n No. N1 Percent N1 p Value All patients Tis T T1-intramucosal T1-submucosal T T T Adenocarcinoma Tis T T1-intramucosal T1-submucosal T T T Squamous Tis T T T T a a a a Significant (p 0.05) by one-sided Cochran-Armitage test for linear trend. echo characteristic, and proximity to the primary carcinoma, are evaluated to determine the status of regional lymph nodes. Most studies, however, lack histologic correlation with esophageal ultrasonographic findings, except those with esophageal ultrasonographic-directed fine-needle aspiration of suspicious regional lymph nodes. Sheathed fine-needle aspiration needles are not available. Therefore, regional lymph nodes located immediately adjacent to the primary tumor, where the needle must traverse the tumor, cannot be accurately assessed. A negative fine-needle aspiration result raises the question of sampling error. Thoracoscopy allows direct biopsy of regional nodes, but unlike mediastinoscopy in the staging of bronchogenic carcinoma, it is not a simple outpatient procedure. General anesthesia and one-lung ventilation are required. Unilateral sampling is generally the only procedure performed. In most instances the primary tumor must be dissected to gain access to regional lymph nodes. Because this dissection is not confined by definite tissue boundaries, a potential exists for tumor dissemination throughout the pleural space. If laparoscopy is added, this becomes a formidable procedure with operating times approaching that of resection. Despite the shortcomings of present staging technologies, precise clinical staging of esophageal carcinomas can be obtained. The assessment of depth of tumor invasion (T) by esophageal ultrasonography and the best clinical evaluation of regional lymph node status (N), individualized for each patient, may be augmented by the consideration of the relationship between T and N. Clinical Predictors of Regional Lymph Node Metastasis Clinical predictors of regional lymph node metastasis may also be used to complement clinical staging. Of all factors in this study, depth of tumor invasion was the most useful clinical predictor. The likelihood of regional lymph node metastasis increased dramatically with deeper invasion, a clear reflection of an increasing exposure to lymphatics as a tumor grows along and through the esophageal wall. Patients with adenocarcinoma were significantly more likely to present with regional nodal metastases than patients with squamous cell carcinoma. A similar result could not be established for patients with adenosquamous carcinoma because of the small number of patients. However, the prevalence of regional lymph node metastasis was nearly identical between adenocarcinoma and adenosquamous carcinoma. A less well differentiated neoplasm had a greater potential for regional nodal metastasis demonstrated by multivariable analysis. The presence of Barrett s esophagus, one of the factors accounting for the increasing prevalence of esophageal adenocarcinoma, was not predictive of regional lymph Table 3. Pattern of N1 Disease Variable Total n No. N1 Percent N1 Clinical Sex Male Female Age at operation (y) Barrett s mucosa No Yes Cell type Adenocarcinoma Squamous Adenosquamous Well Not well Signet ring Yes No Postresection Margins Positive Negative No. of lymph nodes resected

5 Ann Thorac Surg RICE ET AL 1998;65: ESOPHAGEAL CARCINOMA: T AND N 791 Table 4. Univariable and Multivariable Clinical Factors Predicting N1 Disease Variable Univariable Results Multivariable Results OR 95% CI p Value OR 95% CI p Value Sex Male/female NC Age at operation Per 10-year increase NC Barrett s mucosa No/yes NS Signet ring cells Yes/no NS T status 2/ / / Cell type AD/SQ AS/SQ Not well/well AD adenocarcinoma; AS adenosquamous carcinoma; CI confidence interval; OR odds ratio; NC not considered in multivariable model; NS not significant in multivariable model; SQ squamous cell carcinoma. node metastasis. Although lack of Barrett s esophagus was a univariable predictor of regional lymph node metastasis, it appeared to be highly associated with the three multivariable predictors, but did not add additional information to the multivariable model. In this series, many patients with Barrett s esophagus were identified through screening programs and, in general, had early mucosal involvement. Many patients with early Barrett s associated tumors had more differentiated adenocarcinomas. Age and sex, so important in the recent epidemic of adenocarcinoma of the esophagus and esophagogastric junction, were not predictive of regional lymph node metastasis. Postresection Predictors of N1 Although factors identified at pathologic staging are not available for clinical staging, analysis of these data indicates the need for excellent surgical technique in the Table 5. Univariable and Multivariable Postresection Factors Predicting N1 Disease Variable Univariable Results OR 95% CI p Value OR Multivariable Results 95% CI p Value Margins Positive/negative NC No. of lymph nodes resected Per 5 node increase NC CI confidence interval; OR odds ratio; NC not considered in multivariable model; NS not significant in multivariable model. resection of esophageal carcinoma. Univariable analysis demonstrated that regional lymph node metastases are more common when the surgical margins are positive, a reflection of the intramural extent of the tumor. When a resection is undertaken, patients with regional lymph node metastasis are more likely to have positive resection margins, and vice versa. The total number of lymph nodes resected was a univariable predictor of regional lymph node metastasis. This suggests that surgical removal of all regional lymph nodes (lymphadenectomy) is important, both prognostically and therapeutically. The relationship between the number of lymph nodes resected and the likelihood of lymph node metastasis must be considered when attempting to assess the role of pathologic staging in this disease. Conclusions We conclude that for patients with esophageal carcinoma, the probability of regional lymph node metastasis Table 6. Probability of N1 Disease According to T Status, Cell Type, and Cell Type T Status T1 T2 T3 T4 Squamous Well (1.4) a Not well Adenocarcinoma Well 5.0 (23.6) a 54.6 (64.6) a Not well Adenosquamous Well (1.7) a 9.4 (28.7) a (37.9) a Not well (6.5) a (28.9) a 61.3 (70.6) a a This combination of cell type, differentiation, and T status does not exist in the data; thus, this should be interpreted cautiously.

6 792 RICE ET AL Ann Thorac Surg ESOPHAGEAL CARCINOMA: T AND N 1998;65: Fig 1. The lymphatic anatomy of the esophagus. Lymphatics enter the mucosa to lie just below the basement membrane of the epithelium and drain the lamina propria and muscularis mucosa. The submucosa is richly supplied with an interconnecting network of lymphatic channels that run the length of the esophagus. Lymphatics intermittently pierce the muscularis propria to drain into regional lymph nodes or directly into the thoracic duct. (Reprinted with permission from the Cleveland Clinic Foundation.) (N1) can be predicted from an assessment of (1) pathologic diagnosis of adenocarcinoma, (2) a less than welldifferentiated histology, and (3) most importantly, increasing depth of tumor invasion (T). This assessment identifies those patients at particularly high risk for regional lymph node metastasis and should be included with other established factors used in clinical staging and treatment decision making. References 1. Fleming ID, Cooper JS, Henson DE, et al, ed. AJCC cancer staging manual. 5th ed. Philadelphia: Lippincott-Raven, 1997: Goseki N, Koike M, Yoshida M. Histopathologic characteristics of early stage esophageal carcinoma. A comparative study with gastric carcinoma. Cancer 1992;69: Bogomoletz WV, Molas G, Gayet B, Potet F. Superficial squamous cell carcinoma of the esophagus. A report of 76 cases and review of the literature. Am J Surg Pathol 1989;13: Sugimachi K, Ikebe M, Kitamura K, Toh Y, Matsuda H, Kuwano H. Long-term results of esophagectomy for early esophageal carcinoma. Hepatogastroenterology 1993;40: Sabik JF, Rice TW, Goldblum JR, et al. Superficial esophageal carcinoma. Ann Thorac Surg 1995;60: Liebermann-Meffert D, Duranceau A. Anatomy of the esophagus. In: Zuidema GD, Orringer MB, ed. Shackelford s surgery of the alimentary tract. 4th ed. Philadelphia: W.B. Saunders, 1996: Murakami G, Sato I, Shimada K, Dong C, Kato Y, Imazeki T. Direct lymphatic drainage from the esophagus into the thoracic duct. Surg Radiol Anat 1994;16: Riquet M, Saab M, Le Pimpec Barthes F, Hidden G. Lymphatic drainage of the esophagus in the adult. Surg Radiol Anat 1993;15: DISCUSSION DR HARVEY I. PASS (Detroit, MI): Doctor Rice, you are well known as one of the leaders in endoesophageal ultrasound, and I have a couple of questions related to that issue because it is mentioned in your abstract. You have now defined pathologically that depth is important with nodal involvement. What are we talking about as far as this nodal involvement? Are we talking about microscopic disease? Are we talking about gross pericapsular disease or enlarged nodes when we look at the pathologic specimens? Does this mean that patients who, on an endoesophageal ultrasound, have a greater depth of penetration should have random, if possible, transesophageal biopsy of any lymph nodes even if the nodes are not enlarged? DR RICE: Concerning lymph node metastasis, the nodal status (N) is determined at pathologic, microscopic review of the resection specimen and includes all regional nodal metastases both microscopic and macroscopic. Esophageal ultrasound is very accurate in the prediction of depth of tumor invasion (T) and good in the assessment of regional lymph node status (N). We use nodal size, shape, border, echo characteristics, and proximity of the node to the tumor in the assessment of regional lymph node status (N). Depth of tumor invasion (T) as determined by endoscopic ultrasound is another predictor of regional lymph node status (N). For example, a poorly differentiated T3 adenocarcinoma should have metastasized to regional lymph nodes (T3 N1 M0) 88% of the time and be free of regional lymph node metastases (T3 N0 M0) only 12% of the time. In our experience, regardless of the depth of tumor invasion (T), patients with regional lymph node metastasis (N1) have a 7% survival 5 years after surgical resection. Whenever possible, transesophageal needle aspiration of regional lymph nodes should be used to confirm regional lymph node status (N).