In the seventh edition of the American Joint Committee on Cancer (AJCC) staging system for esophageal cancer, tumor grade was introduced as an independent determinant of stage grouping in early stage tumors. With the significantly lower prognosis for poorly differentiated early stage adenocarcinomas, patients with these tumors may become candidates for neoadjuvant therapy given an accurate identification of these tumors with preoperative staging. The objective of the current study was to investigate the accuracy of preoperative histopathologic grading and the effect of preoperative grade on tumor stage/prognostic grouping.
Preoperative tumor grade was compared with postoperative tumor grade in 427 patients who underwent surgery without receiving neoadjuvant therapy for adenocarcinoma of the esophagus. The impact of preoperative tumor grade on stage/prognostic grouping was investigated.
The overall accuracy of preoperative tumor grade assessment was 76% when unknown differentiation was regarded as well/moderately differentiated as recommended by the AJCC, whereas accuracy was 73% after the exclusion of tumors with unknown grade. In patients who have tumors classified as T1 or T2 and lymph node-negative (N0) (T1-T2N0) disease, 16% were assigned to a lower stage group based on preoperative pathology, whereas 5% were assigned to a higher stage group. In the T1-T2N0 group, sensitivity for detecting a poorly differentiated tumor was 0.43 (95% confidence interval [CI], 0.30-0.56), whereas specificity was 0.94 (95% CI, 0.90-0.98).
The seventh edition of the American Joint Committee on Cancer (AJCC) Cancer Staging Manual1 has introduced several major modifications in the staging of gastric and esophageal cancer2 compared with the sixth edition.3 Most important, all tumors involving the gastroesophageal junction (GEJ) are now classified as esophageal cancers. The only exception is for GEJ tumors with an epicenter >5 cm distal to the GEJ, which are coded as gastric cancer. A second important change is the incorporation of histologic grade into the stage/prognostic grouping for both adenocarcinoma and squamous cell carcinoma esophageal cancer. For patients who have adenocarcinoma tumors classified as T1 or T2 and negative lymph nodes (N0) without metastasis (M0) (T1-T2N0M0), the degree of differentiation is now an independent determinant of stage/prognostic group. Well and moderately differentiated T1N0 adenocarcinomas are staged IA, whereas poorly differentiated T1N0 adenocarcinomas are grouped together with well to moderately differentiated T2N0 tumors in stage IB. Poorly differentiated T2N0 adenocarcinomas are stage IIA. For T3 or greater tumors and for tumors with positive lymph nodes, the degree of differentiation does not influence stage/prognostic grouping. A third change is the definition of lymph node (N) status, which is now based on the absolute number of positive lymph nodes and is synchronized with lymph node stage for gastric carcinoma. Additional changes include the definition of tumor stage for in situ carcinoma (Tis) and T4 and metastasis (M) classification.
The proposal of the 2010 AJCC staging system for esophageal cancer is based on a combined large international database: the Worldwide Esophageal Cancer Collaboration (WECC).4 This database contains information on >7000 patients and represents the practice of 13 institutions on 3 continents. However, for the staging system, only data from the 4627 patients who underwent surgery without receiving chemotherapy or radiotherapy were used. Therefore, the compliance of the staging system with patients who received preoperative or postoperative treatment is debatable.5 Another issue with this data set is the lack of information from preoperative biopsies. The introduction of tumor grade in the staging system is based entirely on postoperative pathologic evaluation. However, this information is unavailable when a patient is staged before surgery to determine the use of neoadjuvant therapy.
Because the use of preoperative chemotherapy and radiation has become increasingly established for the treatment of resectable esophageal and GEJ adenocarcinoma,6, 7 accurate preoperative staging has become an issue of increasing clinical relevance that is not limited to locally advanced tumors. In the AJCC seventh edition, poor differentiation/tumor grade is used as an independent predictor of poor survival in early stage tumors. The stage-specific 10-year survival rates are 66%, 51% and 38% for stage IA (T1N0 grade 1 or 2 [T1N0G1,G2]), stage IB (T1N0G3/T2N0G1,G2), and stage IIA (T2N0G3), respectively.8 Provided with these data, the group of patients with T2N0G3 tumors may become candidates for preoperative therapy, given that these tumors can be correctly identified preoperatively.
The objective of the current study was to investigate the accuracy of preoperative assessment of tumor grade of esophageal and GEJ adenocarcinoma by comparing preoperative grading on biopsies with postoperative surgical pathology specimens in individuals who did not receive neoadjuvant therapy. The second objective was to investigate the impact of preoperative grade on tumor stage/prognostic grouping as detailed in the seventh edition of the AJCC staging manual.
MATERIALS AND METHODS
Patients were identified from 2 prospectively maintained databases of gastric and esophageal cancer. Between January 1996 and November 2009, 1440 patients with adenocarcinoma of the distal esophagus or GEJ without metastatic disease underwent potentially curative surgery at Memorial Sloan-Kettering Cancer Center (MSKCC). January 1996 is the time point at which an institutional electronic medical record system was introduced and, thus, is the date from which additional information to the prospective database can be obtained. Patients who received preoperative chemotherapy or radiation were excluded, leaving 475 patients who did not receive neoadjuvant treatment. Because tumor grade was not assessed in patients with T0 or Tis disease (high-grade glandular dysplasia), these patients (N = 48) also were excluded. Overall, 427 patients with both preoperative biopsies and postoperative resection material were available for this analysis. Patient and pathologic tumor characteristics, treatments, and follow-up data were recorded prospectively. This study was approved by the Institutional Review Board of MSKCC.
Preoperative Staging and Histology
Preoperative staging was performed with various combinations of chest radiographs, computed tomography (CT) scans, positron emission tomography (PET) scans, endoscopic ultrasound (EUS) studies with biopsies, and diagnostic laparoscopy with biopsies. To avoid influence by imaging modalities on the accuracy of preoperative tumor grade analysis, each patient was assigned a preoperative stage based on the tumor grade assessed from preoperative biopsies combined with postoperative TNM classification. All patients underwent preoperative histopathologic evaluation by an in-house pathologist, either by an evaluation of the submitted slides from referring hospitals or by a review of endoscopic biopsy specimens obtained at MSKCC. Whenever possible, all material from patients who had multiple biopsies was reviewed. Tumor grade was defined as well, moderately, or poorly differentiated and reflected a recording of the poorest grade within the biopsy. In the final analysis, well and moderately differentiated tumors were grouped as 1 entity. When tumor grade was not mentioned in the pathology report, it was recorded as “unknown.” During the period from 1996 to 2003, pathologists of any subspecialty participated in the assessment of these tumors; but, since 2004, only specialized gastrointestinal (GI) pathologists have evaluated esophageal and GEJ tumors.
All patients underwent potentially curative resection of the esophagus, the GEJ, the stomach, or a combination with different types of approaches, depending on the tumor location and the preference of the surgeon. Surgical techniques included 3-phase esophagectomy (cervicothoracoabdominal), Ivor-Lewis esophagectomy (right thoracoabdominal), (left) thoracoabdominal esophagectomy, transhiatal (cervicoabdominal) esophagectomy, proximal gastrectomy, and total gastrectomy.
Postoperative Histology and Staging
Staging was performed according to the new AJCC staging guidelines (seventh edition, 2010).1 The depth of tumor invasion, the number of positive lymph nodes, margin status, and the grade of differentiation were recorded prospectively and used to calculate postoperative AJCC seventh edition T classification, N status, and stage/prognostic group. According to the AJCC stage-grouping recommendation, tumors with unknown grade were regarded as well/moderately differentiated tumors.
Tumor grade was recorded as well, well-to-moderately, moderately, moderately-to-poorly, or poorly differentiated. These were translated into a trichotomous system of well, moderately, and poorly differentiated tumors, and the poorest grade mentioned was recorded. In the final analysis, well and moderately differentiated tumors were grouped into 1 entity. Adenocarcinomas of the GEJ were classified according to a modification of criteria proposed by Siewert and Stein,9 with type I tumors defined as adenocarcinomas of the distal esophagus that may extend below the esophagogastric junction by <25% of the tumor mass, type II tumors defined as carcinomas that straddle the esophagogastric junction, and type III tumors defined as subcardial gastric carcinomas that involve the GEJ and may extend above the GEJ by <25% of the tumor mass.
The accuracy of preoperative staging was calculated by determining the concordance between preoperative and postoperative pathologic grade assessed from biopsies and surgical specimens, respectively. The postoperative pathologic grade was used as the gold-standard reference point. Accuracy was expressed as the percentage of patients with the correct grade assigned. Although this yields a number that is easy to understand, it fails to reflect on the distribution of patients over different categories. Therefore, the Cohen weighted kappa test for agreement was used as an additional measurement of accuracy of preoperative tumor grade. In general, values of kappa from 0.20 to 0.39 are considered fair agreement, values from 0.40 to 0.59 are considered moderate, values from 0.60 to 0.79 are considered substantial, and a value of 0.80 or above is considered outstanding.10 Differences between groups were calculated by using the Pearson chi-square test. Survival estimates were calculated using the Kaplan-Meier method, and differences between survival estimates were analyzed with the log-rank test. All statistical analyses were performed with the SPSS statistical software package (version 17.0; SPSS, Inc., Chicago, Ill).
Demographic, pathologic, and surgical data are summarized in Table 1. Seventy-eight percent of the patients were men, and the mean age was 66.2 years. The average number of patients per year who underwent surgery without receiving preoperative therapy decreased over time and was 37 patients per year from 1996 to 1999, 32 patients per year from 2000 to 2004, and 24 patients per year from 2005 to 2009. Adenocarcinomas were classified as Siewert type I in 30% of tumors, type II in 46% of tumors, and type III in 25% of tumors. The survival of patients who had postoperative T1-T2N0 well/moderately differentiated/unknown adenocarcinomas was significantly longer compared the survival of patients who had poorly differentiated tumors of the same T1-T2N0 stage (80% vs 56%; P = .005) (Fig. 1). Patients with preoperative stage IA disease who were upstaged to stage IB on postoperative pathology had a significantly worse prognosis compared with patients who remained in stage IA on postoperative pathologic staging (P = .014) (Fig. 2), although there was no significant difference in overall survival for patients who were assigned to stage IB preoperatively and postoperatively (P = .454). This analysis could not be performed for patients who were staged incorrectly between stages IB to IIA, because the number of events in this group was too few.
Table 1. Patient Characteristics, Surgical Treatment, and Pathology Data
No. of Patients (%)
Abbreviations: SD, standard deviation.
Age: Mean±SD, y
Year of surgery
Postoperative pathology indicated that 37 of 427 tumors (9%) were well differentiated, 183 of 427 tumors (43%) were moderately differentiated, and 199 of 427 tumors (47%) were poorly differentiated. Postoperative grade was reported in 419 of 427 tumors (98%), and preoperative grade was reported in 302 of 427 of tumors (71%) (Table 2). Based on AJCC staging guidelines, unknown tumor grade was recorded as well/moderately differentiated.
Table 2. All T1-T4N0-N3 Tumors (N = 427)
Abbreviations: Well, well differentiated; Mod, moderately differentiated; Poor, poorly differentiated.
The accuracy of preoperative tumor grade assessment in the entire group (T1-T4N0-N3) was 76% (324 of 427 tumors; P < .001) (Table 3). The Cohen kappa test for agreement demonstrated a kappa value of 0.50 (P < .001), reflecting moderate agreement. Most discordance in the preoperative grading of biopsies was a result of under grading, and 88 of 301 preoperative tumors (29%) that were well and moderately differentiated were poorly differentiated tumors on postoperative surgical pathology (Table 3). After the exclusion of patients with unknown preoperative or postoperative grade, the accuracy of preoperative grading was 73% (238 of 301 tumors; P < .001), with a kappa value of 0.58 (P < .001), indicating moderate agreement.
Unknown grade was coded as well-moderately differentiated.
The concordance of preoperative grade assessment was comparable when analyzed by specialized GI pathologists (80%) or non-GI-pathologists (78%; P value nonsignificant). Differences in accuracy between the different Siewert types (type I, 72%; type II, 79%; type III, 86%) had borderline significance (P = .06).
Accuracy slightly increased during the consecutive time periods: Accuracy was 73%, 81%, and 83% for the periods 1996 to 1999, 2000 to 2004, and 2005 to 2009, respectively. This observation did not reach statistical significance.
Because tumor grade only affects stage grouping in patients with T1-T2N0 disease, subgroup analyses were performed excluding T3 and T4 tumors and tumors with positive lymph nodes. Accuracy of preoperative grade assessment in T1-T2N0 tumors was 79% (156 of 197 tumors; P < .001) with a kappa value of 0.42 (Table 4). Most grading discordance was caused by preoperative under grading. After the conversion of T1-T2N0 tumors into their corresponding preoperative and postoperative stage/prognostic groups, 79% (156 of 197 tumors; P < .001) of patients were staged properly (Table 5) (kappa, 0.57; P < .001). Preoperative under staging occurred in 32 of 197 of these patients (16%), and over staging in occurred in 9 of 197 patients (5%). Sensitivity in this group to detect a poorly differentiated tumor was 0.43 (95% confidence interval [CI], 0.30-0.56), whereas specificity was 0.94 (95% CI, 0.90-0.98). This indicates that, of all poorly differentiated tumors in the T1-T2N0 group, 57% were not identified as such.
Table 5. Stage Grouping for All Patients With T1-T2N0 Tumors (N = 197)a
Postoperative Stage Group
Preoperative Stage Group
Accuracy, 156/197 = 0.79; Cohen kappa = 0.57.
Although T classification, N status, and M status are strong independent predictors of survival in patients with esophageal cancer,11-13 the sixth edition of the AJCC staging system for esophageal cancer has been challenged for its heterogeneity of outcome on survival within the different stage groups.14 During the past years, several pathologic prognostic factors have been proposed for incorporation into the TNM staging system. These include degree of differentiation,15 vascular and perineural invasion,16 extracapsular lymph node invasion,17 tumor length,18, 19 clearance of the proximal and distal resection margin,20, 21 and status of the circumferential margin.22 These proposals however, are based primarily on analyses from relatively small series of patients and, in most instances, from single-institution databases. Incorporation of a new factor into the AJCC staging system not only requires a structured mechanism of the proposed change,23 but the factor also has to be available in the collaborative WECC database.
In 1991, Robey-Cafferty et al demonstrated in their series of 69 patients with squamous cell carcinoma of the esophagus that the degree of tumor differentiation was an independent prognostic factor.24 In 2001, Dickson et al proposed the incorporation of tumor grade into the staging system based on a series of 139 consecutive patients who underwent surgery for GEJ carcinoma (mostly adenocarcinoma). Those authors demonstrated differences in 3-year overall survival for well differentiated tumors (33.3%) and moderately differentiated tumors (28.9%) versus poorly differentiated tumors (15.9%).15 Khan et al confirmed those results in a series of 219 patients with N0 squamous cell carcinoma and adenocarcinoma of the esophagus, demonstrating that tumor grade was an independent prognostic factor in univariate and multivariate analyses.25 Other studies also confirmed a correlation between tumor grade and prognosis in univariate analysis, but not in multivariate analysis.26-28 Recently, Thompson et al reported a study of 240 patients with mainly adenocarcinoma in which tumor grade was an independent prognostic variable in both univariate and multivariate analyses.14 In that study, patients were divided into 2 groups: well/moderately differentiated and poorly differentiated/undifferentiated. Furthermore, combined data in the WECC database also supports the incorporation of tumor grade into the 2010 AJCC staging system for esophageal cancer. The inclusion of postoperatively determined tumor grade into the staging system may provide outcome information and guidance for adjuvant therapeutic strategies. However, the question raised with this addition is: How reliable is the assessment of tumor grade in small preoperative biopsies? This is particularly relevant when neoadjuvant options are considered for poorly differentiated T2N0M0 adenocarcinomas.
In the patient group that we evaluated, the average number of patients without preoperative treatment decreased over the years. This is consistent with increased use of neoadjuvant therapy in locally advanced GEJ carcinoma. Poorly differentiated (G3) and early stage tumors were associated with a significantly lower survival rate compared with tumors that were graded as well differentiated (G1), moderately differentiated (G2), or unknown on postoperative pathology (Fig. 1). Preoperatively under staged IB tumors (into preoperative stage IA) were associated with lower survival compared with correctly staged IA tumors (Fig. 2), indicating the significance of accurate tumor grade assessment on preoperative staging.
In the current study, the entire cohort, including patients with “unknown” tumor grade (which was regarded as well/moderately differentiated according to the AJCC recommendation), preoperative tumor grade assessed in biopsies was concordant with that assessed in surgical specimens in 76% of patients. After excluding individuals with unknown tumor grade, overall concordance was 73% (kappa value, 0.58), consistent with moderate agreement.
However, not all stage/prognostic groups are affected by tumor grade. In the AJCC seventh edition, only T1-T2N0 tumors are assigned to 3 separate stage groups when the tumor is well and moderately differentiated or poorly differentiated. In T1-T2N0 tumors, the concordance for tumor grade was slightly higher compared with concordance for the entire cohort (79% vs 76%); therefore, 21% of the patients in this group had an “inappropriate” stage group assigned based on preoperative biopsies: under staging occurred in 16%, and over staging occurred in 5%.
The differences in concordance of tumor grade assessed by GI-pathologists and non-GI-pathologists were not statistically significant. To the best of our knowledge, this has not been described previously. Siewert type had borderline significant differences favoring higher accuracy of tumor grade assessment in Siewert type III tumors.
Several factors may account for the discordance in tumor grade assessment in biopsy specimens and surgical resection specimens. These include sampling issues, technical quality of the specimen, and, to a lesser extent, experience of the pathologist. It is noteworthy that a significant number of GEJ adenocarcinomas revealed intratumoral heterogeneity (Fig. 3), exhibiting mixed populations of well-to-moderately differentiated and poorly differentiated histopathology within the same tumor. Thus, a biopsy specimen may not always represent the dominant component of the tumor grade in the entire lesion. Therefore, sampling bias may be responsible for discordance in both upgrading and downgrading biopsy specimens and resection specimens, respectively. A second significant factor responsible for tumor grading discordance is the suboptimal preparation of biopsy specimens, which may include an excessive air-dry effect before formalin fixation, tumor tissue adjacent to ulcer and necrosis, or thermal/mechanically generated crush artifacts in diminutive specimens. In these situations, upgrading from a well/moderately differentiated tumor to a poorly differentiated tumor is a more likely consequence than downgrading from a poorly differentiated tumor to a well/moderately differentiated tumor (Fig. 4).
The very recent introduction of the 2010 staging system into clinical practice has precluded the development of treatment algorithms for the different stage/prognostic groups, and these remain to be established. To our knowledge, no prospective studies have been performed in the subset of patients with early stage, poorly differentiated tumors, and most current clinical trials of neoadjuvant therapy for esophageal adenocarcinoma apply different inclusion criteria and usually include patients with clinically T2-T3 (cT2-T3)N0 disease29 and cT1-T3N1 disease.7 However, results from the French Federation for Digestive Cancer (FFCD) 9901 trial indicated that preoperative chemoradiation followed by surgery has a negative impact on postoperative mortality in patients with early stage esophageal cancer compared with surgery alone without a significant difference in overall survival.30 However, the majority of patients in that trial had squamous cell carcinoma, and no subgroup analyses were performed on patients with poorly differentiated, early stage tumors. Furthermore, a WECC database analysis revealed that, when the subgroup of patients with T2N0G3 tumors underwent surgery only, they had a significantly worse 10-year survival rate (38%) compared with other patients with early stage but lower grade tumors (66% and 51%, respectively).8
Because the new AJCC staging system has revealed a prognostic difference for patients with stage IA, IB, and IIA esophageal adenocarcinoma that is stratified with the combination of tumor classification (T1-T2) and tumor differentiation, it is likely that patients who have early esophageal and GEJ tumors that are poorly differentiated may become candidates for neoadjuvant therapy. However, in our current study, we demonstrated that the sensitivity for grading a poorly differentiated, early stage tumors is only 0.43. Given this low sensitivity, there exists the potential risk that >50% of these patients would not receive therapy that, otherwise, may have been recommended.
With the increasing use of neoadjuvant chemotherapy and radiotherapy in esophageal cancers, it is evident that therapeutic management strategy should be evaluated based on a combination of clinical, radiographic, and pathologic assessments. In future modifications of the AJCC staging system, this might be addressed by capturing clinical staging information in the WECC database. Precise pathologic identification is particularly pertinent when assessing tumor differentiation in individuals with early stage and lymph node-negative esophageal cancer.
This research was funded in part by Mushett Family Foundation, the “Prof. Michael van Vloten” Foundation, the Dutch Cancer Society (KWF Kankerbestrijding), the Dutch Digestive Foundation, and the Jo Keur Foundation.