Tumor Marker Expression Is Predictive of Survival in Patients With Esophageal Cancer

Transcription

1 Tumor Marker Expression Is Predictive of Survival in Patients With Esophageal Cancer Thomas A. Aloia, MD, David H. Harpole, Jr, MD, Carolyn E. Reed, MD, Carmen Allegra, PhD, Mary-Beth H. Moore, BA, James E. Herndon II, PhD, and Thomas A. D Amico, MD Division of Cardiothoracic Surgery, Duke University Medical Center, Durham, North Carolina, Division of Cardiothoracic Surgery, Medical University of South Carolina, Charleston, South Carolina, and the National Cancer Institute, Bethesda, Maryland Background. This study was designed to determine the prognostic value of immunohistochemical tumor marker expression in a population of patients with nodenegative esophageal cancer treated with complete resection alone. Methods. Resection specimens were collected from 61 patients with node-negative T1 (n 31), T2 (n 14), and T3 (n 16) esophageal cancer. A panel of 10 tumor markers was chosen for immunohistochemical analysis, based on associations with differing oncologic mechanisms: apoptosis (p53), growth regulation (transforming growth factor-, epidermal growth factor receptor, and Her2-neu), angiogenesis (factor VIII), metastatic potential (CD44), platinum resistance (p-glycoprotein and metallothionein), 5-fluorouracil resistance (thymidylate synthetase), and carcinogenic detoxification (glutathione S-transferase- ). Results. Complete resection was performed in all patients (44 adenocarcinoma, 17 squamous cell carcinoma), with no operative deaths. Multivariable analysis demonstrated a significant relationship between cancer-specific death and the following variables: low-level P-gp expression (p 0.004), high-level expression of p53 (p 0.04), and low-level expression of transforming growth factor- (p 0.03). In addition, the number of involved tumor markers present was strongly predictive of negative outcome (p ). Conclusions. This study supports the prognostic value of immunohistochemical tumor markers, specifically the expression pattern of P-gp, p53, and transforming growth factor-, in patients with esophageal carcinoma treated with complete resection alone. (Ann Thorac Surg 2001;72:859 66) 2001 by The Society of Thoracic Surgeons Currently, 60% of patients with esophageal cancer present with advanced disease, and the overall mortality is approximately 90% [1]. The dismal prognosis for most patients with esophageal cancer has prompted the use of novel therapeutic strategies including the use of multimodality therapy. Despite apparent advances in therapeutic strategy, the widespread use of induction chemotherapy [2] or chemotherapy and radiation therapy [3] has not succeeded in improving overall survival. This lack of efficacy may be related to either a lack of potency in the therapeutic regimen or the application of these therapies to an inappropriate population. Whereas the development of superior chemotherapeutic agents may improve the efficacy of induction therapy, survival may also be improved by using tumor marker analysis to refine the current staging system and guide therapeutic decisions [4]. Immunohistochemical (IHC) marker analysis in patients with esophageal cancer can be used for screening and early diagnosis, prognosis, the detection of occult Presented at the Forty-seventh Annual Meeting of the Southern Thoracic Surgical Association, Marco Island, FL, Nov 9 11, Address reprint requests to Dr D Amico, Duke University Medical Center, Box 3496, Durham, NC 27710; damic001@mc.duke.edu. metastases, and the determination of chemoresistance. In this study, tumors from a population of patients found to be node-negative (N0) after complete resection for esophageal cancer underwent IHC marker analysis in the absence of chemotherapy or radiation therapy. Nodenegative patients were chosen to eliminate the significant confounding influence of nodal metastasis on survival, which may mask the effect of some of these markers. A panel of 10 tumor markers was chosen for IHC analysis. Six of these markers were chosen on the basis of their associations with differing oncologic mechanisms of importance in thoracic malignancy, including apoptosis (p53), growth regulation (transforming growth factor- [TGF- ], epidermal growth factor receptor [EGFR], Her2- neu), angiogenesis (factor VIII), and metastatic potential (CD44). Four additional markers of resistance to commonly used chemotherapy agents in esophageal cancer were included: platinum (p-glycoprotein [P-gp], glutathione S-transferase- [GST- ], metallothionein [MT]) and 5-fluorouracil (thymidylate synthetase [TS]). This set of markers has been studied in patients undergoing trimodality therapy, but their potential role as natural history prognostic markers has not been examined [5]. The primary goal of this study was to determine the relative prognostic power of IHC markers in esophageal cancer patients treated with surgical resection alone. A 2001 by The Society of Thoracic Surgeons /01/$20.00 Published by Elsevier Science Inc PII S (01)

2 860 ALOIA ET AL Ann Thorac Surg IMMUNOHISTOCHEMICAL STAGING OF ESOPHAGEAL CANCER 2001;72: second goal was to compare the prognostic power of these variables to the pathologic variables used in the current staging system. The addition of tumor markers with greater prognostic value than the currently used pathologic variables may allow refinement of the current staging system and more accurate allocation of therapy [6]. Finally, IHC profile analysis of neoadjuvant therapy naive esophageal tumors may serve as reference data when analyzing the changes in marker expression that occur during neoadjuvant therapy. Material and Methods Population Pathologic specimens were collected from patients with completely resected node-negative (T1 T3, N0) esophageal cancer, treated from 1988 to 1999 at two universityaffiliated hospitals. None of these patients received induction or postoperative chemoradiotherapy. All study patients had an en bloc esophageal resection with negative margins, including celiac axis and periesophageal lymph nodes. All operative samples were submitted to institutional pathology for routine pathologic analysis. Gross examination included determination of tumor size and location in the esophagus. Routine microscopic analysis, including determination of histologic subtype and lymph node status, was also performed. For entry into statistical analysis each patient had to live 60 days. These consecutive node-negative patients were selected from more than 300 patients who underwent resection alone, the majority being node-positive. Immunohistochemical Methods After institutional pathology department evaluation, tumor specimens were collected for IHC analysis at a single institution. A single pathologist verified histologic staging, and then the study material was formalin-fixed, paraffin-embedded, and serially sectioned at 4 to 5 m. Sections were plated on glass slides, deparaffinized, and underwent IHC analysis using a standard horseradish immunoperoxidase protocol [4] and automated immunostainers (BioGenex Laboratories, Inc, San Ramon, CA). The mouse monoclonal IgG antibodies used in IHC staining, including the source and antibody dilution, were as follows: from BioGenex Laboratories, Inc (San Ramon, CA), anti-p-gp (1:40), anti-p53 #1801 (1:200), anti-gst- (1:150), anti-her2-neu #CB11 (1:20), anti- EGFR (1:20), and antiangiogenesis factor VIII (1:100); from DAKO Corporation (Carpinteria, CA), anti-mt #E9 (1:150); from Boehringer Ingelheim Bioproducts Partnership (Heidelberg, Germany), anti-cd44 #v6 (1:50); and from Oncogene Research Products (Cambridge, MA), anti-tgf- (1:100). The anti-ts antibody is a monoclonal antibody from the laboratory of Carmen Allegra, PhD. Thymidylate synthetase immunostaining and slide evaluation were performed by Dr Allegra at the National Cancer Institute using a previously validated methodology [7]. Representative samples of IHC-stained tissues with selected biologic markers are shown in Figure 1. Immunohistochemical Marker Scoring Known positive blocks were simultaneously prepared with each tissue assay, as well as IgG-negative control slides. Individual slides were read by three independent observers blinded to clinical information and classified as either positive or negative for the respective antibody on a semiquantitative scale: 0 (none), 1 (1% to 20%), 2 (21% to 50%), and 3 ( 50%). This reproducible scale measures the number of tumor cells stained, not the intensity of stain present, which may vary with the age of the paraffin blocks. The markers tested were dichotomized by the median score, as determined by statistical analysis. High-level expression was defined as the higher values for each marker. Discrepant scores were resolved by consensus [5, 8]. For TS evaluation, each slide was assigned a score for intensity and staining pattern. Intensity scores range from0to3, and the staining pattern was either focal or diffuse. For intensity, 0 is no staining, 1 is light or trace staining, 2 is definite staining of light to moderate intensity, and 3 is bright or dark intensity. Slides with 50% or less of malignant cells stained at the assigned intensity level were considered focal, whereas those with greater than 50% stained were scored as diffuse, with the final score being adjusted for the amount of tumor stained [7]. Angiogenesis factor VIII was scored by counting the stained microvessels in 10 consecutive fields ( 20) within the area of highest microvessel density. Using this method there was little intraobserver variability. Reported scores are the average of the observer s individual scores [9]. Statistical Analysis The Kaplan-Meier product limit estimator and the log rank test were used to describe the relationship between individual study variables and outcome. Variables that achieved statistical significance at p less than or equal to 0.20 in univariable analysis were then placed into multivariable models. Using Cox proportional hazards model the interrelationship between variables and their relative impact on outcome was analyzed. The Kaplan-Meier product limit estimator and the log rank test were also used to analyze the impact of the number of negative prognostic markers present on cancer-specific death. The 2 test was used to analyze the correlation between histologic type and IHC variables. Results The mean age of patients in the cohort was 62 years, with a range of 35 to 83 years. There were 50 men and 11 women enrolled in the study. Routine pathologic examination demonstrated 44 patients with adenocarcinoma and 17 patients with squamous cell carcinoma. Applying the American Joint Committee on Cancer staging defini-

3 Ann Thorac Surg ALOIA ET AL 2001;72: IMMUNOHISTOCHEMICAL STAGING OF ESOPHAGEAL CANCER 861 Fig 1. Representative samples of immunohistochemically stained esophageal tumor specimens. Areas of brown coloration indicate immunohistochemical marker staining. Antibodies included are (clockwise from upper left) p53, Her2-neu, glutathione S-transferase-, and P-glycoprotein. All photomicrographs were imaged at 40. (GST- glutathione S-transferase- ; P-gp p-glycoprotein.) tions, 31 patients were T1N0, 14 patients were T2N0, and 16 patients were T3N0 [10]. Mean clinical follow-up was 32 months. At the time of statistical analysis, 27 patients were alive without disease, 3 patients were alive with recurrent disease, and 27 patients had succumbed to disease. There were 4 patients who died of noncancer causes, with no evidence of disease recurrence. Overall median cancer-specific survival time was 37 months (Fig 2). Pathologic T stage was a significant predictor of disease-specific death. Median survival for patients with T1 and T2 tumors was 92 months, whereas those with T3 tumors had a median survival of only 8 months (p 0.02; Fig 3). In contrast to T stage, histologic subtype had no impact on outcome. Patients with adenocarcinoma had a median survival of 38.3 months whereas those with squamous cell carcinoma of the esophagus had a median survival of 39.8 months (p 0.92). Subset analysis was unable to detect any association between marker expression and histology except for P-gp. Only 3 of 17 (18%) patients with squamous cell carcinoma showed highlevel expression of P-gp, whereas 27 of 44 (61%) patients with adenocarcinoma showed high-level expression of this marker ( 2 test, p 0.04). Univariable Analysis of Immunohistochemical Tumor Markers Univariable analysis of the tumor markers demonstrates that low-level expression of P-gp (p 0.192) and TGF- (0.016), as well as high-level expression of GST- (p 0.015), p-53 (p 0.080), angiogenesis factor VIII (p 0.040), and EGFR (p 0.11), are correlated with diseasespecific death (Table 1). Survival curves for each of these variables are presented in Figure 4. Expression of Her2- neu, CD44, MT, and TS showed no association with outcome in our patient population. In addition, no association with survival was identified for the combination of TGF- with EFGR.

4 862 ALOIA ET AL Ann Thorac Surg IMMUNOHISTOCHEMICAL STAGING OF ESOPHAGEAL CANCER 2001;72: Table 1. Immunohistochemical Marker Variables Used in Univariable Analysis of Cancer-Specific Death a Marker Expression Subgroup n No. of Deaths Median Survival (mo) p Value Fig 2. Cancer-specific survival analysis for the entire cohort plotted as percent survival versus time in months (mo.) using the Kaplan- Meier product limit estimator. Number of patients at risk at 1-, 3-, and 5-year intervals is given along the x-axis. Multivariable Analysis Immunohistochemical markers with statistical significance in univariable analysis at a p value less than 0.2 were entered into multivariable analysis. A summary of the multivariable regression analysis model is provided in Table 2. This analysis supports the independent prognostic significance of low-level expression of P-gp (p 0.006) and TGF- (0.005), and high-level expression of GST- (p 0.016) and p53 (p 0.041). Pathologic T stage data were then added to the IHC marker data, creating a second multivariable model (Table 3). This analysis demonstrates that low-level P-gp expression, low-level TGF- expression, and high-level expression of p53 are predictive of poor outcome in this patient population. The impact of T stage on cancer-specific death did not achieve statistical significance in this analysis. Cancer-Specific Survival in Relation to the Number of Prognostic Markers Present Using the statistically significant factors from the first multivariable model (low-level expression of P-gp and p53 Low-level High-level TGF- Low-level High-level EGFR Low-level High-level Her2-neu Low-level High-level 14 3 NR Factor VIII Low-level NR High-level CD 44 Low-level High-level P-gp Low-level High-level GST- Low-level 24 7 NR High-level MT Low-level High-level TS Low-level High-level a Each variable is listed along with the subgroup analyzed, the number of patients in each subgroup, the number of cancer-specific deaths in each subgroup, the median survival of each subgroup, and the statistical p value associated with comparison of each variable with cancer-specific death using the log rank test. EGFR epidermal growth factor receptor; GST- glutathione S-transferase- ; MT metallothionein; NR not reached; P- gp P-glycoprotein; TGF- transforming growth factor- ; TS thymidylate synthetase. TGF-, and high-level expression of GST- and p53), an additional statistical model was created to assess the additive impact of these prognostic markers on cancerspecific survival. For each patient, the number of negative prognostic markers present was summed. Six patients lacked negative marker values completely. Forty patients expressed either one or two negative markers, 12 patients expressed three negative markers, and 3 patients expressed all four negative markers. Patients with no negative markers had a median survival exceeding the length of the study period. Patients with one to two negative markers had a median survival of 49 months, patients with three negative markers had a median survival of 7 months, and patients with four negative markers had a median survival of only 2 months. Survival analysis demonstrated that the number of negative markers present strongly correlates with outcome (p ; Fig 5). Fig 3. Cancer-specific survival analysis comparing patients with T1, T2, and T3 esophageal tumors plotted as percent survival versus time in months (mo.) using the Kaplan-Meier product limit estimator. Number of patients at risk at 1-, 3-, and 5-year intervals is given along the x-axis. Comment The use of IHC tumor markers in the staging of esophageal cancer is being investigated as a method to select patients with comparatively more favorable tumor biology to receive induction or adjuvant therapy. In contrast

5 Ann Thorac Surg ALOIA ET AL 2001;72: IMMUNOHISTOCHEMICAL STAGING OF ESOPHAGEAL CANCER 863 Fig 4. (A) Cancer-specific survival analysis comparing patients with low-level expression of P-glycoprotein (P-gp) versus those with P-glycoprotein high-level expression, plotted as percent survival versus time in months (mo.) using the Kaplan-Meier product limit estimator. Number of patients at risk at 1-, 3-, and 5-year intervals is given along the x-axis. (B) Cancer-specific survival analysis comparing patients with lowlevel expression of transforming growth factor- (TGF- ) versus those with transforming growth factor- high-level expression, plotted as percent survival versus time in months (mo.) using the Kaplan-Meier product limit estimator. Number of patients at risk at 1-, 3-, and 5-year intervals is given along the x-axis. (C) Cancer-specific survival analysis comparing patients with low-level expression of glutathione S-transferase- (GST- ) versus those with glutathione S-transferase- high-level expression, plotted as percent survival versus time in months (mo.) using the Kaplan-Meier product limit estimator. Number of patients at risk at 1-, 3-, and 5-year intervals is given along the x-axis. (D) Cancer-specific survival analysis comparing patients with low-level expression of p53 versus those with p53 high-level expression, plotted as percent survival versus time in months (mo.) using the Kaplan-Meier product limit estimator. Number of patients at risk at 1-, 3-, and 5-year intervals is given along the x-axis. (E) Cancer-specific survival analysis comparing patients with low expression of factor VIII versus those with factor VIII high-level expression, plotted as percent survival versus time in months (mo.) using the Kaplan-Meier product limit estimator. Number of patients at risk at 1-, 3-, and 5-year intervals is given along the x-axis.

6 864 ALOIA ET AL Ann Thorac Surg IMMUNOHISTOCHEMICAL STAGING OF ESOPHAGEAL CANCER 2001;72: Table 2. Multivariable Model of Outcome Based on Cox s Proportional Hazards Analysis of Immunohistochemical Variables on Cancer-Specific Death a Marker Relative Risk Z Value p Value P-gp present TGF- present GST- present p53 present Factor VIII overexpression EGFR present a Variables that reached a significance level of p 0.2 in univariable survival analysis including low-level expression of P-glycoprotein and TGF- and high-level expression of glutathione S-transferase-, p53, and factor VIII were used to create this model. Each marker is listed with the impact of the presence of each marker on survival, the statistical Z value, and the statistical p value. EGFR epidermal growth factor receptor; GST- glutathione S-transferase- ; P-gp P-glycoprotein; TGF- transforming growth factor-. to several previous studies that have analyzed the prognostic effect of individual tumor markers on outcome in patients with esophageal carcinoma [11 13], this study analyzes a comprehensive panel of markers, related to a spectrum of oncologic mechanisms. One set of markers was chosen for their associations with differing oncologic mechanisms with significance in thoracic malignancy (p53, TGF-, EGFR, Her2-neu, factor VIII, CD44), whereas a second group was chosen on the basis of their ability to predict treatment response (P-gp, GST-, MT, and TS). Of the markers in the first group, only high-level expression of p53 and low-level expression of TGF- correlated with poor outcome. Previous investigations of this group of prognostic markers in esophageal cancer have not been definitive. Our data support two previous studies in which mutation of p53 was associated with decreased survival in patients with esophageal carcinoma [14, 15]. Another study, however, in which 204 patients with esophageal cancer were analyzed with Table 3. Multivariable Model of Outcome Based on Cox s Proportional Hazards Analysis of Immunohistochemical Variables With the Addition of Pathologic Tumor Stage on Cancer-Specific Death a Marker Relative Risk Z Value p Value T stage P-gp present TGF- present GST- present P53 present a Each variable is listed with the impact of the presence of each marker on survival, the statistical Z value, and the statistical p value. GST- glutathione S-transferase- ; P-gp P-glycoprotein; T stage pathologic tumor stage; TGF- transforming growth factor-. Fig 5. Cancer-specific survival analysis comparing patient groups determined by the number of negative prognostic markers present (0, 1 to 2, 3, and 4 markers), plotted as percent survival versus time in months (mo.) using the Kaplan-Meier product limit estimator. respect to p53 expression found no correlation between marker expression and outcome [16]. Although the current study supports a prognostic role for p53 high-level expression in esophageal carcinoma, it should be noted that the IHC assay detects elevated p53 protein levels, and does not necessarily reflect p53 mutation status. The inability to prove conclusively that p53 alterations have prognostic significance may relate to the biology of esophageal cancer, suggesting superseding mutations. Previous esophageal cancer studies have shown a correlation between high-level TGF- expression and poor prognosis [17, 18]. The proposed mechanism for this relationship involves an interaction between the EGFR and its agonist, TGF-, leading to the development of autocrine or paracrine tumor growth cascades and a dedifferentiated phenotype with enhanced metastatic potential. Our data, however, suggest that in patients with node-negative esophageal cancer, low-level expression of TGF- is associated with a worse prognosis. This association and its apparent opposition to previous reports may be related to a lack of interaction between TGF- and EGFR in early stage patients. In our study population, EGFR expression was a significant predictor of survival in univariable analysis; however, this association lost significance in multivariable analysis. In addition, the combination of TGF- expression and EGFR expression did not correlate with prognosis. The discordance of our results with previous reports may be related to a lack of interaction between TGF- and EGFR, reflecting early pathologic stage and relative tumor cell differentiation. In the oncology literature many novel tumor markers are being identified as potential markers of resistance or sensitivity to chemotherapy when they may actually be natural history prognostic markers, with no relationship to the therapy. The second set of IHC variables used in this study shed light on this important distinction. Our previous work in patients undergoing trimodality therapy for esophageal cancer indicates that high-level expression of P-gp, GST-, and TS are associated with poor outcome [5]. The expression pattern of MT in this popu-

7 Ann Thorac Surg ALOIA ET AL 2001;72: IMMUNOHISTOCHEMICAL STAGING OF ESOPHAGEAL CANCER 865 lation did not correlate with outcome variables. Our current data indicate that low-level expression of P-gp and high-level expression of GST- correlates with poor outcome in patients with node-negative tumors treated with resection alone, although the impact of GST- was diminished when pathologic T stage was included in the multivariable model. Expression of MT and TS did not correlate with outcome in this population. From these data we can measure the relative impact of each of these variables as natural history markers versus their impact as markers of chemotherapy response. In the case of TS, comparative analysis of the data indicates that this marker has little or no power as a natural history marker, but is a significant marker of 5-fluorouracil resistance. In the case of GST-, it is likely that high expression of this marker is a stronger predictor of platinum response and less important as a natural history marker. The expression of MT appears to have little impact on outcome in esophageal cancer patients regardless of nodal status or therapy. High-level expression of P-gp, which functions as an ATP-dependent efflux pump, is associated with resistance to chemotherapeutic agents, such as platinum and taxanes [4]. Although the mechanism of chemoresistance associated with high-level expression of P-gp is well described, the independently positive prognostic effect of P-gp high-level expression in this study of patients treated with resection alone is not obvious. We speculate that P-gp may function to efflux a variety of growth factors or carcinogens. Cells lacking in P-gp receptors may retain these factors and progress to develop malignant phenotypes. It is possible that high-level expression of the P-gp protein is associated with tumors with a higher degree of differentiation, and therefore less metastatic potential. Further studies must be designed to correlate P-gp with other potentially positive prognostic factors. In univariable analysis, T status was found to be an independent predictive variable. Esophageal cancer patients with stage T1 tumors had a median survival more than 4 years longer than esophageal cancer patients with T3 tumors. When combined with other IHC markers with significance in univariable analysis, however, the effect of T stage on cancer specific death lost statistical significance. In this analysis, low-level P-gp expression, lowlevel TGF- expression, and high-level expression of p53 were predictive of poor outcome with relative risk ratios of 0.2, 1.6, and 0.3, respectively. These data provide early evidence that IHC characterization of esophageal cancer tumors may be a valuable adjunct to the current staging system, which relies heavily on the prognostic impact of pathologic T stage to determine overall stage and prognosis. Finally, a prognostic model was created to examine the effect that the presence of multiple IHC markers had on survival in this patient population. The model included analysis of four IHC markers shown to have individual prognostic importance in multivariable analysis (P-gp, GST-, p53, and TGF- ). Detailed analysis of these data show that the presence of multiple IHC markers is, in a step-wise fashion, predictive of survival. Patients with no negative markers present had a median survival exceeding the length of the study period. In comparison, patients with greater than three negative markers present had a median survival less than 6 months. Although these results must be confirmed in a prospective study, these data indicate that it may be possible to select a subpopulation of patients with node-negative esophageal cancer for induction or adjuvant therapy using IHC staging to stratify cancer-specific risk. This study evaluates the prognostic value of a panel of 10 IHC markers in patients with node-negative esophageal cancer treated with surgical resection alone. Analysis of the data demonstrates that low-level P-gp expression, low-level TGF- expression, and high-level expression of p53 are significantly negative prognostic factors in patients with esophageal cancer. Furthermore, analyzing patients according to the number of negative prognostic markers present enhances risk stratification. Prospective validation of this model is required to determine whether IHC marker profiles can be used to select patients with node-negative esophageal cancer for adjuvant therapy. The authors are indebted to both Allyson Parr and Debbie Conlon for their expert assistance in the preparation and analysis of tissues from the cohort of patients presented in this paper. References 1. Lund O, Hasenkam JM, Aagaard MT, Kimose HH. Time related changes in characteristics of prognostic significance in carcinomas of the oesophagus and cardia. Br J Surg 1989;76: Kelsen DP, Ginsberg R, Pajak TF, et al. Chemotherapy followed by surgery compared with surgery alone for localized esophageal cancer. N Engl J Med 1998;339: Bosset JF, Gignoux M, Triboulet JP, et al. Chemotherapy followed by surgery compared with surgery alone in squamous-cell cancer of the esophagus. N Engl J Med 1997;337: D Amico TA, Massey M, Herndon JE, et al. A biologic risk model for stage I lung cancer: immunohistochemical analysis of 408 patients with the use of ten molecular markers. J Thorac Cardiovasc Surg 1999;117: Harpole DH, Moore MB, Herndon JE, et al. The prognostic value of molecular marker analysis in patients treated with trimodality therapy for esophageal cancer. Clin Cancer Res 2001;7: D Amico TA, Harpole DH. Molecular biology of esophageal cancer. Chest Surg Clin North Am 2000;10: Edler D, Blomgren H, Allegra CJ, et al. Immunohistochemical determination of thymidylate synthetase in colorectal cancer: methodological studies. Eur J Cancer 1997;33: Gerdes AM, Nielsen O, Mohr U, et al. Correlation between molecular genetic analyses and immunohistochemical evaluation of the epidermal growth factor receptor and p185her2. Anticancer Res 1998;18: Harpole DH, Richards WG, Herndon JE, et al. Angiogenesis and molecular biologic substaging in patients with stage I non-small cell lung cancer. Ann Thorac Surg 1996;61: Digestive system. Esophagus. In: Fleming ID, Cooper JS, Henson DE, eds. AJCC cancer staging manual, 5th ed. Philadelphia: Lippincott-Raven, 1997:65 9.

8 866 ALOIA ET AL Ann Thorac Surg IMMUNOHISTOCHEMICAL STAGING OF ESOPHAGEAL CANCER 2001;72: Bottger TC, Youssef V, Dutkowski P, et al. Expression of CD44 variant proteins in adenocarcinoma of Barrett s esophagus, and its relation to prognosis. Cancer 1998;83: al-kasspooles MA, Moore JH, Orringer MB, et al. Amplification and over-expression of EGFr and Her2-neu genes in human esophageal adenocarcinomas. Int J Cancer 1993;54: Peters WH, Wobbes T, Roelofs HM. Glutathione S- transferase in oesophageal cancer. Carcinogenesis 1993;14: Casson AG, Tammemagi M, Eskandarian S, et al. p53 alterations in oesophageal cancer. Association with clinicopathological features, risk factors, and survival. Mol Pathol 1998; 51: Patel DD, Bhatavdekar JM, Chikhlikar PR, et al. Clinical significance of p53, nm23, and bcl-2 in T3 4N1M0 esophageal carcinoma: an immunohistochemical approach. J Surg Oncol 1997;65: Sarbia M, Porschen R, Borchard F, et al. p53 protein expression, and prognosis in squamous cell carcinoma of the esophagus. Cancer 1994;74: Yoshida K, Kuniyasu H, Yasui W, et al. Expression of growth factors and their receptors in human esophageal carcinomas: regulation of expression by epidermal growth factor and transforming growth factor alpha. J Cancer Res Clin Oncol 1993;119: Iihara K, Shiozaki H, Tahara H, et al. Prognostic significance of transforming growth factor-alpha in human esophageal carcinoma. Implication for the autocrine proliferation. Cancer 1993;71: DISCUSSION DR JOE PUTNAM (Houston, TX): Doctor Aloia, I appreciated your comments and certainly appreciated your moving thoracic surgeons from an anatomic staging basis to a molecular biologic staging basis. This is particularly timely given the number of prospective studies that are ongoing at single institutions and, as well, some of the national multicenter studies. Your use of molecular markers for identifying patients at risk certainly attests to the interests of your institution. One of the two questions that I have is how the use of immunohistochemistry is the most appropriate technique or whether other techniques, such as polymerase chain reaction, may be more helpful. Second, have you used this technique for biopsy specimens in attempting to identify these markers before definitive treatment? I certainly enjoyed your comments. DR ALOIA: Thank you very much. Those are both excellent questions. From your introductory comments, I think that single-institution studies definitely suffer from a low N, and this study may not avoid that problem. We did pool together two institutions to get the N we presented here. As you probably know, the number of patients that present with this level of early stage disease is few among the total number that we see. In terms of the question of immunohistochemistry versus techniques that have a higher sensitivity, such as polymerase chain reaction, I think that there are several comments to be made there. Number one is that our laboratory is very facile with immunohistochemistry: we are very comfortable reading the stains, we use automated stainers, and our algorithms are tried and true. We have started to perform polymerase chain reaction analysis, but that is fraught with questions about specificity as well. So the answer to that question is we are comfortable with our immunohistochemistry. We are trying other modalities as we speak. As I said in the introduction, we did have a cohort of 112 patients in whom we performed a nearly identical analysis. These patients did not have early disease. They were candidates for induction therapy, and they were enrolled in a trial for induction therapy. We enlisted a similar panel of molecular markers, using immunohistochemistry on the biopsy specimens, and there are several interesting points that we found. We did find some predictors of outcome from that, which made the basis for our current study, but we found, of course, that you have a small amount of tissue to analyze and there are only so many slides you can cut from these core biopsies that are not used by routine pathology. As well, you have a sample bias; you are only taking a part of the tumor. We have done that and that work is ongoing and I think it will be fruitful. Polymerase chain reaction technologies may help us out by being able to apply multiple markers or gene arrays to small pieces of tissue obtained at biopsy.