Clinical utility of postchemoradiation endoscopic brush cytology and biopsy in predicting residual esophageal adenocarcinoma

Authors


Abstract

BACKGROUND:

Esophageal adenocarcinoma generally carries a poor prognosis. Treatment with combination chemoradiation (CRT) followed by esophagectomy is becoming common. A pathologic complete response is uncommon but predicts improved survival. Identifying the subset of patients with residual carcinoma has potential management implications. Post-CRT endoscopic brush cytology and biopsy may detect residual tumor; however, the accuracy and clinical value of these methods remain unclear.

METHODS:

Sixty-seven patients with esophageal adenocarcinoma who underwent preoperative CRT and post-CRT endoscopic brush cytology and biopsy followed by esophagectomy were identified. By using esophagectomy histology as the gold standard, the performance of cytology and biopsy was evaluated in diagnosing residual carcinoma. Two pathologists independently reviewed all false-negative and false-positive cases and resolved disagreements by consensus.

RESULTS:

The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of cytology for diagnosing residual carcinoma were 26%, 95%, 92%, 35%, and 45%, respectively. For biopsy, these rates were 13%, 90%, 75%, 31%, and 36%, respectively. Sampling error accounted for false-negative diagnoses in approximately 66% of cytology analyses and 98% of biopsy analyses. Approximately 33% of false-negative cytology analyses and 1 false-negative biopsy analysis were caused by the under–recognition of tumor cells. Major diagnostic pitfalls included obscuring acute inflammation, necrosis, tumor cells that mimicked benign cells with radiation/reactive atypia, and the under recognition of mucin-containing adenocarcinoma cells.

CONCLUSIONS:

Brush cytology and biopsy were specific but not sensitive methods for predicting residual cancer after CRT. However, cytology was superior. The current results indicated that brush cytology can be used alone to diagnose residual esophageal carcinoma, and awareness of specific diagnostic pitfalls will help pathologists improve its accuracy. Cancer (Cancer Cytopathol) 2009. © 2009 American Cancer Society.

Adenocarcinoma of the esophagus and the esophagogastric junction (EGJ) is aggressive and has a poor prognosis. The majority of these patients present with locally advanced tumors or regional lymph node metastasis. The 5-year survival rate for patients with locoregionally advanced esophageal carcinoma (stages IIA to IVA) remains poor and ranges from 6% to 40% when patients undergo surgery alone.1-6 Multimodality therapy with preoperative combination chemotherapy and external beam radiation therapy (CRT) followed by esophagectomy has been used to treat patients with locoregionally advanced esophageal and EGJ carcinomas because of their poor long-term outcome if they undergo surgery alone.7-10 This strategy has produced mixed results in several randomized controlled trials.11, 12 However, the outcome of patients after preoperative CRT is much better if no residual carcinoma is found in the resected specimen, a result that represents a pathologic complete response. A pathologic complete response is achieved in <50% of patients, and these patients have a much better long-term survival than patients with residual carcinoma in multiple nonrandomized studies.13-21 Identifying this subset of patients has potential implications for management, because they may benefit from avoiding esophagectomy-related morbidity and mortality.22 Post-CRT endoscopic brush cytology and biopsy are used to diagnose residual tumor before a definitive resection is performed. However, the clinical value of these 2 diagnostic methods in predicting the presence of residual carcinoma is not clear, and data on the accuracy of these techniques are lacking.23 The objectives of this study were 1) to evaluate the usefulness of endoscopic brush cytology and biopsy in predicting residual carcinoma after CRT and 2) to determine diagnostic pitfalls and ways to improve diagnostic accuracy in evaluating these specimens.

MATERIALS AND METHODS

Patient Population and Specimens

Sixty-seven patients with locally advanced esophageal or EGJ adenocarcinoma who underwent preoperative CRT and had post-CRT endoscopic brush cytology and biopsy followed by esophagectomy were identified from the database at the University of Maryland Medical Center between January 1995 and February 2008. All patients received pretreatment staging workup, including endoscopic biopsy for diagnosis, computed tomography (CT) scans of the chest and abdomen, and endoscopic ultrasound. The regimens for preoperative CRT varied but, in general, consisted of a 30-day treatment with combined 5-fluorouracil, cisplatin, and paclitaxel and a total radiation dose of 50.4 Gray. All patients underwent repeat endoscopic examination 4 to 6 weeks after CRT. Brush cytology was performed in the area previously noted to contain tumor. Endoscopic mucosal biopsy using large-capacity forceps was taken from suspicious areas. Biopsies were also taken in 4 quadrants every 1 cm to 2 cm in the esophagus from the area previously noted to contain tumor regardless of the endoscopic appearance. The time interval between cytology/biopsy and resection varied from 4 days to 84 days (average, 23 days). Fifty-six of 67 patients (84%) underwent an esophagectomy within 30 days after cytology/biopsy.

Cytologic and Histologic Preparations

A smear slide was made using the cytology brush and immediately was fixed in 95% alcohol for Papanicolaou staining. The brush was then cut off and put into a CytoRich preservative fluid (TriPath Imaging, Inc, Burlington, NC) to prepare 1 thin-layer slide.

Tissue biopsy was fixed directly in 10% formalin and embedded in a paraffin block. Three levels of routine hematoxylin and eosin-stained slides were made for review. For esophagectomy specimens, areas with ulceration or scarring, which indicated the therapy field, were submitted completely for histologic evaluation if no macroscopically identifiable cancer was noted. Patients were assigned to 1 of 3 groups based on the histology of their resected specimen. Group A was comprised of patients who had a complete pathologic response (defined as the absence of tumor cells in both the esophagus and the lymph nodes), Group B consisted of patients who had a partial response (defined as from 1-50% residual carcinoma in the resected specimen), and Group C included patients who had no response (defined as >50% residual carcinoma in the resected specimen).24

We reviewed all pathologic reports and all available slides from cytology, biopsy, and resection. Diagnostic performance of post-CRT brush cytology and biopsy were evaluated by using histologic diagnosis of the esophagectomy specimen as the gold standard. Sensitivity, specificity, positive predictive value, negative predictive value, and accuracy for cytology and biopsy were determined. We identified all false-negative and false-positive cases for cytology and biopsy. Two pathologists independently reviewed all false-negative biopsy cases, 27 false-negative cytologic cases that had slides available for review, and all false-positive cytology and biopsy cases to analyze the diagnostic pitfalls. Disagreements between the 2 reviewers were resolved by consensus review. The University of Maryland Baltimore Institutional Review Board approved this study.

RESULTS

Demographic features and histopathologic findings are summarized in Table 1. There was a predominance of men (85%; 57 of 67 patients), and the mean age was 63 years. For positive brush cytology cases, either abundant residual tumor cells are present (Fig. 1A) with morphology similar to pretreatment (Fig. 1B), or only a few small groups/clusters of residual tumor cells are seen (Figs. 1C and 1D). In the background, numerous benign and reactive squamous and columnar cells and acute inflammation are always seen. Most postchemoradiation endoscopic biopsies show mainly inflamed granulation tissue, fibrinoinflammatory exudates and no viable tumor cell (Fig. 2A), or rarely they are positive for residual tumor with scant less well-preserved tumor cells (Fig. 2B). The esophagectomy specimen in all cases demonstrated varied degrees of treatment effect with mucosal ulceration, granulation tissue with acute inflammation, fibrinoinflammatory exudates, tumor necrosis with varied degrees of viable residual tumor in the mucosa, and reactive stromal and endothelial cells (Figs. 2 and 3). In some cases, residual tumor was present only in the muscularis propria and subadventitia (Figs. 2D, 3A, and 3B); whereas, in other cases, scant, rare residual tumor cells were difficult to observe without cytokeratin immunostaining (Figs. 3C and 3D). In this series, 17 of 67 patients (25%) achieved a pathologic complete response with no residual tumor identified in the esophagus or lymph nodes, 8 of 67 patients (12%) had a partial response with from 1% to 50% residual carcinoma, and 42 of 67 patients (63%) were nonresponders who were found to have >50% residual carcinoma in the resected specimen.

Figure 1.

(A) A large volume of residual tumor was observed on brush cytology (Papanicolaou stain, original magnification ×40). (B) The same tumor demonstrated similar morphology on pretreatment brush cytology (Papanicolaou stain, original magnification ×40). (C and D) Scant residual tumor in a small cluster is shown (Papanicolaou stain, original magnification ×60 in Panel C and ×40 in Panel D).

Figure 2.

(A) This view of a postchemoradiation endoscopic biopsy demonstrates mainly inflamed granulation tissue and no viable tumor cell (H&E stain, original magnification ×4). (B) Scant, less well–preserved, positive residual tumor was observed on biopsy (H&E stain, original magnification ×4). (C) Extensive mucosal tumor necrosis and inflamed granulation tissue with scattered, rare, viable tumor cells in the deeper mucosa and muscularis propria were observed in a specimen of resected esophagus (H&E stain, original magnification ×2). Inset: Residual tumor was observed at higher magnification (H&E stain, original magnification ×10). (D) A large area of residual tumor was present only in the deep muscularis propria and subadventitia (H&E stain, original magnification ×4). Inset: Residual tumor was observed at higher magnification (H&E stain, original magnification ×10).

Figure 3.

(A) This view of a postchemoradiation esophagectomy specimen reveals mucosal ulceration with deep residual tumor (H&E stain, original magnification ×2). (B) Residual deep tumor was observed at higher magnification (H&E stain, original magnification ×10). (C) A single residual tumor cell was surrounded by reactive stromal cells (H&E stain, original magnification, ×40). (D) Immunostaining for cytokeratin was positive for this tumor cell (original magnification ×40).

Table 1. Clinical and Histopathologic Findings in 67 Patients With Esophageal and Esophagogastric Junction Adenocarcinoma
VariableNo. of Patients (%)
  1. SD indicates standard deviation.

Sex 
 Men57 (85)
 Women10 (15)
Age, y 
 Mean±SD63±11
 Range28-90
Pathologic response 
 Complete17 (25)
 Partial: 1-50% of residual tumor8 (12)
 Nonresponse: >50% of residual tumor42 (63)

By using the presence or absence of residual tumor in the esophagectomy specimen as the gold standard, the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy for cytology to diagnose residual carcinoma were 26%, 95%, 92%, 35%, and 45%, respectively; and the respective rates for mucosal biopsy were 13%, 90%, 75%, 31%, and 36% (Tables 2 and 3). Among the 47 cases with residual carcinoma in the resected esophagus, 3 were diagnosed by both brush cytology and biopsy, 9 were diagnosed by brush cytology only, and 3 were diagnosed by biopsy only. In 1 biopsy-positive case, the original cytology was called atypical but was positive by consensus review. Combining cytology and biopsy results, sensitivity increased only slightly to 32%. Compared with biopsy, brush cytology performed better in all diagnostic parameters.

Table 2. Endoscopic Brush Cytology and Mucosal Biopsy in Predicting Residual Esophageal Adenocarcinoma After Combination Chemotherapy and External Beam Radiotherapy
Diagnostic MethodNo. of Esophagectomies
Residual CarcinomaNo Residual CarcinomaTotal
Cytology   
 Positive12113
 Negative351954
Mucosal biopsy   
 Positive628
 Negative411859
Total no.472067
Table 3. Comparison of Diagnostic Performance of Endoscopic Brush Cytology and Mucosal Biopsy in Predicting Residual Esophageal Adenocarcinoma After Combination Chemotherapy and External Beam Radiotherapy
VariableBrush Cytology, %Mucosal Biopsy, %
Sensitivity2613
Specificity9590
Accuracy4536
Positive predictive value9275
Negative predictive value3531

Brush cytology did not predict the presence or absence of residual lymph node disease. Only 2 of 13 patients (15%) who had positive cytology results, compared with 13 of 54 patients (24%) who had negative cytology results, were found to have positive lymph node disease determined on resection (P = .7, Fisher exact test).

Specimens were available for review in 27 of 35 false-negative cytology cases and in all 41 false-negative biopsy cases. Table 4 lists causes of the false-negative readings in cytology cases. Sampling error accounted for the false-negative diagnosis in 19 cytology cases (71%) and in 40 of 41 biopsy cases (98%), which had no tumor cells present in the sampled specimen. In the cytology specimens, the majority of cellular components were benign squamous cells, columnar/glandular cells, and acute inflammatory cells. Most biopsies in these cases had granulation tissue with acute inflammation and necrosis (Fig. 2A). Eight cytology cases (30%) and 1 biopsy case were called negative because of under recognition of tumor cells. Carcinoma cells were scant and were obscured by acute inflammation and necrosis in all 8 cases. In half of those cases, tumor cells were misinterpreted as benign cells with radiation or reactive atypia and in the other half, mucin-containing adenocarcinoma cells were not recognized (Figs. 4A and 4B). Overinterpretation of repair/reactive atypia as cancer cells was the cause for 1 false-positive cytology case and 2 false-positive biopsy cases (Fig. 5).

Figure 4.

(A) This cluster of mucinous tumor cells, which mimics radiation atypia, was missed in the original cytologic diagnosis (Papanicolaou stain, original magnification ×60). (B) Another small cluster of residual tumor (red arrow) with 2 stromal cells that had radiation changes (black arrow) from the same case is shown (Papanicolaou stain, original magnification ×60). (C) A pretreatment mucosal biopsy and (D) an esophagectomy specimen demonstrated similar tumor morphology (H&E stain, original magnification ×40).

Figure 5.

(A and B) False-positive brush cytology revealed clusters of reactive glandular cells. (B) Inset: A large atypical cell is observed with radiation changes (Papanicolaou stain, original magnification ×40). (C and D) A false-positive mucosal biopsy is shown. (C) Inflamed granulation tissue with a few large cells that had radiation atypia was overinterpreted as residual tumor cells (H&E stain, original magnification ×10). (D) Higher magnification revealed false-positive cells (H&E stain, original magnification ×20).

Table 4. Causes of False-Negative Brush Cytology Cases
Cause of False-Negative ResultsNo. of Patients (%)
Sampling error 
 No tumor cells on cytology 
 No surface viable tumor/rare tumor cells on resection15 (56)
 Viable surface tumor present on resection4 (15)
Interpretation error 
 Obscured by acute inflammation/necrosis8 (30)
 Mimic radiation atypia4 (15)
 Mucinous tumor cells4 (15)

DISCUSSION

Accurate assessment of residual cancer status after combination CRT therapy may assist in the selection of optimal candidates for esophagectomy after this therapy. Endoscopic examination with esophageal biopsy and/or brush cytology is an easy, simple, and accurate procedure with which to provide a pathologic diagnosis for esophageal or EGJ carcinoma. It is widely accepted as the standard initial morphologic diagnostic procedure in this setting.25-29 However, its clinical importance and accuracy for predicting residual cancer status after CRT remains unclear. Previous studies that evaluated endoscopic biopsy for predicting residual cancer after CRT have reported results similar to ours, including high positive predictive value (range, 92-100%) and specificity (range, 85-100%), but low sensitivity (range, 22-54%) and negative predictive value (range, 11-58%).8, 14, 23, 30, 31 Additional studies have reported false-negative preoperative biopsies in 40% to 66% of patients who previously had what was determined to be a complete response to CRT.14, 32, 33 It also has been demonstrated that esophageal biopsies taken from patients who completed CRT underestimated the histologic response within diseased tissue in up to 29% of patients.34

To the best of our knowledge, the current report is the first in the English literature that evaluates endoscopic brush cytology and mucosal biopsy simultaneously in predicting residual esophageal cancer after CRT. It is interesting to note that the current study indicates that cytology has better diagnostic performance in all parameters (Table 3). The likely reasons for these findings are 2-fold: 1) brush cytology can sample a wider area than biopsy; and 2) the preservation and survival of scant, viable, residual tumor cells in post-CRT specimens is most likely better in cytologic preparation compared with the biopsy preparation process. Considering other advantages, such as shorter turn-around time and lower cost for cytology, we suggest that endoscopic brush cytology can be used alone in this setting. The high specificity and positive predictive value indicate that, when cytology or biopsy results are positive, the patient most likely has residual tumor. The significantly lower sensitivity for both endoscopic biopsy and brush cytology is mainly because of significant treatment effects, including mucosal tumor necrosis as well as fibrosis, inflammation, and repair/reactive changes. In 98% (40 of 41 cases) of false-negative biopsy cases and in 70% (19 of 27 cases) of false-negative cytology cases, no tumor cells were identified, even with repeat evaluation of the specimens. No mucosal viable tumor cells or only rare scattered tumor cells were identified in the resected esophagus in 79% (15 of 19 cases) of these cytologic sampling error cases. Only 4 of these cases (21%) were the result of true sampling error with viable tumor identified in the mucosa of the resected esophagus. However, in the 27 false-negative cytology cases, failure to recognize tumor cells was observed in 8 cases (30%), allowing some room for improvement. Awareness of diagnostic pitfalls, such as obscuring acute inflammation and necrosis as well as tumor cells that mimic radiation/reactive atypia, will help pathologists reach the correct diagnosis. We have found it very helpful to differentiate tumor cells from radiation/reactive atypical cells by comparing the tumor morphology in the pretreatment specimen with the post-treatment specimen, especially in the case of adenocarcinoma with mucinous features (Figs. 1A, 1B, and 4). In a few cases, scattered, large, atypical cells with cytoplasmic fine vacuoles, which were reported as radiation change in the original report, had the same morphology as the tumor identified both in the pretreatment biopsy and in the resected esophagus (Figs. 4C and 4D).

Other diagnostic modalities, such as CT, endoscopic ultrasonography (EUS) with or without fine-needle aspiration (FNA), and positron emission tomography (PET), have been used to predict pathologic response and survival after preoperative CRT in esophageal cancer.35-40 Swisher et al retrospectively evaluated the utility of these 3 modalities in 103 patients and reported a 56% pathologic response to CRT (≤10% viable cells). Esophageal wall thickness (13.3 mm vs 15.3 mm), EUS mass size (0.7 cm vs 1.7cm), and PET standardized uptake value (SUV) (3.1 vs 5.8) were correlated with pathologic response. Post-CRT PET with an SUV ≥4 was found to have the highest accuracy for pathologic response (76%) and was an independent predictor for survival. However, those authors suggested that no modality could exclude the presence of residual microscopic disease after CRT and recommended esophagectomy even if the post-CRT imaging modalities were normal.39 Cerfolio et al prospectively studied the accuracy of EUS/FNA, CT, and fluorodeoxyglucose (FDG)-PET/CT in restaging 48 patients. Those authors reported that none of the 3 tests reliably differentiated the tumor classification of esophageal cancer after CRT except for FDG-PET to identify T0 disease or to exclude T4 disease. FDG-PET/CT was found to be superior to both chest CT and EUS-FNA for predicting lymph node disease and for predicting complete responders, with sensitivity, specificity, accuracy, positive predictive value, and negative predictive value of 87%, 88%, 88%, 76%, and 94%, respectively, compared with CT (27%, 91%, 71%, 57%, and 73%, respectively) and EUS (20%, 94%, 70%, 60%, and 67%, respectively). The higher false-positive rate noted with FDG-PET is caused mainly by metabolic activity from leukocytes and macrophages associated with post-CRT inflammation leading to an elevated SUV.40

Agarwal et al reported that, after preoperative CRT and surgery, patients who had residual tumor in the regional lymph nodes had lower actuarial survival at 1 year, 2 years, and 3 years after surgery compared with patients who had a pathologic complete response or patients who had residual tumor only in the esophagus. Patients who had significant residual lymphadenopathy detected by EUS after CRT had significantly worse postoperative survival. In 8 patients, Agarwal et al observed that reliable cytologic identification of residual malignancy was technically feasible by EUS-FNA after CRT.41 Yang et al observed no difference in esophagectomy specimen tumor stage between biopsy-positive patients and biopsy-negative patients. More biopsy-positive patients demonstrated persistent lymph node disease in their study (69.2% vs 28.8%; P = .01).23 Our study did not reveal the same pattern.

Schneider et al prospectively studied 80 patients with localized esophageal cancer (T2-T4, N0-N1, M0) and assessed the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of endoscopy, biopsy, and EUS in determining histomorphologic tumor response to preoperative CRT. Sensitivities for the detection of residual disease were 60%, 36% and 7%, respectively, for the 3 modalities.31 Accuracy for the identification of histomorphologic regression also was lacking, with values of 47%, 47%, and 50%, respectively, reported. Because of poor sensitivity and low accuracy, those authors concluded that each of the 3 modalities should be removed from the clinical management of patients who complete preoperative CRT.

In summary, both endoscopic esophageal brush cytology and biopsy after CRT are specific, but not sensitive, diagnostic methods for predicting residual cancer in patients with esophageal adenocarcinoma; however, brush cytology is superior to biopsy. Brush cytology can be used without mucosal biopsy in patients after CRT to diagnose residual carcinoma. Unfortunately, to the best of our knowledge, noninvasive methods to accurately evaluate and predict the presence of residual cancer status in the esophagus after CRT are lacking. Efforts must continue to identify a better predictive modality.

Conflict of Interest Disclosures

The authors made no disclosures.

Ancillary