Utility of PET, CT, and EUS to Identify Pathologic Responders in Esophageal Cancer

Transcription

1 ORIGINAL ARTICLES: GENERAL THORACIC Utility of PET, CT, and EUS to Identify Pathologic Responders in Esophageal Cancer Stephen G. Swisher, MD, Mary Maish, MD, Jeremy J. Erasmus, MD, Arlene M. Correa, PhD, Jaffer A. Ajani, MD, Robert Bresalier, MD, Ritsuko Komaki, MD, Homer Macapinlac, MD, Reginald F. Munden, MD, Joe B. Putnam, MD, David Rice, MD, W. Roy Smythe, MD, Ara A. Vaporciyan, MD, Garrett L. Walsh, MD, Tsung T. Wu, MD, and Jack A. Roth, MD Departments of Thoracic and Cardiovascular Surgery, Radiology, Gastrointestinal Oncology, GI Medicine and Nutrition, Radiation Oncology, Nuclear Medicine, and Pathology, University of Texas M. D. Anderson Cancer Center, Houston, Texas Background. This study evaluates the utility of positron emission tomography (PET), endoscopic ultrasonography (EUS), and computed tomographic (CT) scans to predict pathologic response and survival following preoperative chemoradiation (CRT) in esophageal cancer. Methods. One hundred three sequential patients with locoregionally advanced esophageal cancer, who were treated with CRT and esophageal resection between May 2001 and November 2003 at the University of Texas M.D. Anderson Cancer Center, were retrospectively reviewed. PET, EUS, and CT were performed before (pre) or after (post) CRT and before surgical resection. PET standardized uptake value (SUV) was defined as maximal uptake in primary tumor. Results. Most patients were male (91 [88%]) with adenocarcinoma (90 [87%]). Pretreatment clinical stages were: IIA (42 [41%]), IIB (5 [5%]), III (50 [49%]), and IVA (6 [6%]). At the time of surgery, 58 patients (56%) had a pathologic response to CRT (<10% viable cells). Post- CRT measurements that correlated with pathologic response were: CT esophageal wall thickness (13.3 vs 15.3 mm, p 0.04), EUS mass size (0.7 vs 1.7 cm, p 0.01) and PET SUV (3.1 vs 5.8, p 0.01). Post-CRT PET SUV equal to or greater than 4 had the highest accuracy for pathologic response (76%). Univariate and multivariate Cox regression analysis demonstrated that a post-crt PET SUV equal to or greater than 4 was an independent predictor of survival (HR, 3.5, p 0.04). Conclusions. The FDG-PET SUV is the most accurate noninvasive test to predict long-term survival after preoperative CRT and before surgical resection. Post-CRT FDG-PET cannot, however, rule out residual microscopic disease so esophagectomy should remain a therapeutic option even if the post-crt imaging modalities are normal. (Ann Thorac Surg 2004;78: ) 2004 by The Society of Thoracic Surgeons The long-term survival of patients with locoregionally advanced esophageal cancer (stage II to IVA) treated with surgery alone ranges from 6% to 40% with median survivals of 9 to 24 months [1 6]. Because of these poor outcomes, multimodality approaches with chemotherapy and concurrent chemotherapy and radiotherapy (CRT) have been evaluated [7 10]. These strategies have not indicated clear survival benefits, except in patients who respond to induction therapy [4, 8, 10]. Preliminary studies suggest that positron emission tomography (PET) using a d-glucose analog, 18F-2-deoxy-D-glucose (FDG- PET), may be able to identify patients with a pathologic response to preoperative chemoradiotherapy before surgical resection. Additionally, FDG-PET imaging may be Accepted for publication April 12, Presented at the Fortieth Annual Meeting of The Society of Thoracic Surgeons, San Antonio, TX, Jan 26 28, Address reprint requests to Dr Swisher, Department of Thoracic and Cardiovascular Surgery, University of Texas M. D. Anderson Cancer Center, 1515 Holcombe Blvd, Unit 445, Houston, TX 77030; sswisher@mdanderson.org. able to predict long-term survival [11 13]. Traditional noninvasive staging modalities such as computed tomographic (CT) scans and endoscopic ultrasonography (EUS) have been reported to be less effective than FDG- PET imaging at predicting long-term survival [14, 15]. The ability to identify this subset of patients has implications for the clinical management of esophageal cancer patients undergoing multimodality therapy. Therefore, we evaluated patients with esophageal cancer undergoing preoperative CRT who had FDG-PET imaging, CT scans of the chest and abdomen, and endoscopic ultrasonography (EUS) before (pre) or after (post) CRT to determine the ability of these modalities to identify pathologic responders and predict long-term survivors before surgical resection. Material and Methods Patient Population We retrospectively evaluated 103 consecutive patients with primary esophageal cancer who were treated with preoperative chemoradiation and had EUS, FDG-PET, 2004 by The Society of Thoracic Surgeons /04/$30.00 Published by Elsevier Inc doi: /j.athoracsur

2 Ann Thorac Surg SWISHER ET AL 2004;78: PET, CT, AND EUS IN ESOPHAGEAL CANCER and CT studies before surgical resection at our institution between May 2001 and November All patients included in this study had histologically diagnosed adenocarcinoma or squamous cell cancer of the esophagus and underwent EUS, CT of the chest and abdomen, and FDG-PET imaging before surgical resection. The University of Texas M.D. Anderson s institutional review board approved this study. Treatment Plan All patients were treated either on institutional protocols (n 46) with induction chemotherapy followed by concurrent CRT or off-protocol (n 57) with concurrent CRT alone. Patients treated on institutional protocols received up to two cycles of CPT-11 (45 mg/m 2 ), taxotere (33 mg/m 2 ), and 5-fluorouracil (2 g/m 2 as a 24-hour infusion). This was followed by concurrent radiotherapy (50.4 Gy in 28 fractions) and CPT-11 (30 mg/m 2 per week for 5 weeks), taxotere (20 mg/m 2 per week for 5 weeks) and 5-fluorouracil (300 mg/m 2 per day). Patients treated offprotocol received concurrent CRT consisting of radiotherapy (50.4 Gy in 28 fractions) and chemotherapy consisting of either cisplatin and 5-fluorouracil, or taxol and carboplatin. Patients were evaluated before (pre-crt) or after (post-crt) chemoradiation with EUS, CT of the chest and abdomen, and FDG-PET imaging. These noninvasive studies were obtained 3 to 5 weeks after the completion of CRT. Surgical resection was performed 5 to 8 weeks after completion of CRT. The type of esophagectomy performed was dependent on the tumor s location and individual surgeon s preference. Either a transthoracic approach (Ivor-Lewis, or total) or a transhiatal approach was used to resect the tumor. Mediastinal and celiac lymph nodes were resected in all patients who underwent surgery. Only lower mediastinal lymph nodes were removed by transhiatal resection. Pathologic responses were determined at the primary tumor site at the time of resection: complete response (no viable tumor), microscopic residual ( 0% to 10% viable), or macroscopic residual ( 10% viable). These criteria for pathologic response were utilized because of previous experience demonstrating that they identified a subset of patients with increased long-term survival [4, 16]. Pathologic responders were defined as those having equal to or less than 10% viable cancer cells at the primary malignancy (ie, complete response and microscopic residual), whereas nonresponders were defined as those having more than 10% viable cancer cells at the primary malignancy (macroscopic residual). Short-term outcome after surgery including pathologic response was collected prospectively and long-term outcome information regarding overall survival was obtained from hospital records, patient follow-up, and The University of Texas M.D. Anderson tumor registry. Treatment Evaluation: FDG-PET, CT, EUS The FDG-PET images were obtained on a Siemens HR tomograph (Siemens Medical Systems, Hoffman Estates, IL; resolution, mm; full width half maximum [FWHM]). All patients fasted for at least 4 hours before PET. Imaging was performed 45 to 60 minutes after intravenous injection of FDG (15 to 20 mci). The area imaged extended from the middle of the neck to the upper thighs. Transmission scanning was done over the chest and abdomen from 5 or 6 overlapping bed positions and using a rotating germanium-68 pin source, corresponding emission images were also obtained. The PET images were reconstructed by software running standard vendor-provided reconstruction algorithms, and by ordered subset expectation maximization (OSEM). Emission data were corrected for scatter, random events, and dead-time losses using Siemens s software, and images were reconstructed both with and without attenuation correction. Images were analyzed visually and quantitatively to determine the standardized uptake value (SUV) for local malignancy and metastases. To determine the sensitivity, specificity, and accuracy of FDG-PET in identifying residual disease after preoperative CRT and before esophagectomy, PET images were interpreted retrospectively by an experienced nuclear medicine specialist (H.M.) and radiologist (J.J.E.), both were blinded to all clinical data and their findings were recorded by consensus. Images were reviewed in all standard planes along with maximum intensity projection images. PET scans were reviewed in combination with CTs whenever possible. On visual analysis, an abnormality was defined as any instance when FDG uptake was substantially greater than mediastinal blood pool activity on the attenuationcorrected images. Maximum intensity of FDG uptake in the primary tumor, locoregional nodal involvement, and distant sites was assessed as present or absent. Attempts were made to assess local and total tumor burden by measuring the maximal value in each area. A pixelated region of interest (ROI) was outlined within any region of increased FDG uptake and, after correction for radioactive decay, analyzed semiquantitatively according to the following formula: SUV ratio 1153 mean ROI activity (mci/ml)/injected dose (mci)/ body wt (g) Endoscopic ultrasonography was used to assess the primary malignancy and nodes in the celiac axis, perigastric, and parahiatal regions. These studies were all performed at MDACC by dedicated EUS specialists. The primary esophageal tumor was assessed with serial images at 1-cm intervals. Tumor depth was assessed and recorded as well as the length of mucosal involvement. Ulcerations were not considered tumor following preoperative chemoradiation unless biopsies demonstrate residual viable tumor. Nodal metastases were considered present if EUS detected lymphadenopathy ( 10 mm in diameter) or if nodes were round and hypoechoic with sharp margins. EUS guided fine needle aspiration of the enlarged lymph nodes was not performed in the study patients. All CT scans of the chest and abdomen and pelvis were performed on a HiSpeed or Lightspeed Advantage scan- GENERAL THORACIC

3 1154 SWISHER ET AL Ann Thorac Surg PET, CT, AND EUS IN ESOPHAGEAL CANCER 2004;78: ner (General Electrical Medical Systems, Milwaukee, WI). Oral (2% barium sulfate) and intravenous (150 ml Optiray 320, 3 ml/s) contrast material was administered. CT scans of the chest and abdomen were performed 25 seconds and 55 to 65 seconds after contrast administration, respectively. Collimation of 3.75 to 7 mm (chest) and 7 to 7.5 mm (abdomen), and a pitch ratio of 1.5:1 was used. CT images of the chest and abdomen were reviewed on a picture archiving and communication system (PACS) workstation and measurements of maximal dimension of the primary tumor/esophageal wall thickness were performed retrospectively and independently by two observers (M.M., J.J.E.) using electronic calipers. The cross -sectional diameter of the thickest part of the lesion was identified and divided by two. Locoregional nodal involvement ( 10 mm in diameter) was noted. The clinical stage was defined according to the EUS assessment for the primary tumor and regional lymph nodes. Biopsies were not obtained of enlarged nodes. No distant metastases were documented in this study since patients were selected for the review only if they underwent surgery and demonstrated resectable disease following preoperative CRT. Observers were blinded to clinical data and findings were recorded by consensus. Statistical Analysis Survival probability analyses were performed using the Kaplan-Meier method. Survival was calculated from the date of surgery to the date of death or most recent follow-up contact. Analysis of survival excluded patients who died of noncancer-related causes within 30 days after surgery. Statistical significance was assessed by log-rank test. Independent predictive factors for survival were determined by Cox regression analysis. Univariate analyses were performed by 2 analysis. Two-tailed p values of equal to or less than 0.05 were considered significant. Data analysis was performed by our departmental biostatistician (A.M.C.) with SPSS (SPSS, Chicago, IL) and Graph Pad Prism (Graph Pad Prism, San Diego, CA) software. Statistical significance was accepted as p less than Results Patients and Treatments Table 1 describes the demographics of the 103 patients evaluated. All patients selected for this study were required to have resectable and nonmetastatic disease to allow pathologic correlation with the noninvasive studies. Forty-four patients underwent CRT on-protocol with chemotherapy and concurrent chemoradiation, whereas 57 patients underwent CRT off-protocol with concurrent chemoradiation alone. Post-CRT evaluations were performed 4 to 6 weeks after the completion of radiation therapy while pre-crt evaluations were performed 1 to 3 weeks before initiation of CRT. Noninvasive studies performed outside of MDACC were not included because of differences in technique and equipment. Available MDACC noninvasive studies reviewed included: 64 Table 1. Demographics (n 103) Number Percent Gender Male Female Age (median, range) (34 79) Histology ACA Squamous cell Tumor location Upper 1/ Middle 1/ Lower 1/ GEJ Clinical stage a Stage IIA Stage IIB Stage III Stage IVA Pathologic outcome Response ( 10% viable cells) Non-Response ( 10% viable cells) a Clinical stage determined by CT scan, EUS, and PET prior to CRT therapy. CT computed tomography; EUS endoscopic ultrasound; PET positron emission tomography. patients with FDG-PET imaging both before and after CRT; 13 patients only before preoperative CRT; and 26 patients only after CRT. Sixty-seven patients had CT scans of the chest and abomen both before and after CRT; 1, only before CRT; and 31, only after CRT. Ninetytwo patients had EUS both before and after CRT; and 11, only after CRT. Fifty-eight patients (56%) demonstrated a pathologic response following CRT ( 10% viable cells in primary malignancy), although 45 patients (44%) were nonresponders ( 10% viable cells in primary malignancy). Of the 103 patients who had surgery, 73 had a right transthoracic esophagectomy with either a chest anastomosis (Ivor-Lewis) or a cervical anastomosis (total). Lymph node dissections were performed in the abdomen and chest only. A transhiatal approach with a cervical anastomosis and no thoracic incision was performed in 30 patients. The median hospital stay was 12 days, with 1 day in the intensive care unit. Anastomotic leaks were noted in 12 patients (12%). Three patients (2.9%) died perioperatively before discharge or within 30 days of surgery. Treatment Evaluation Multiple retrospective assessments were performed with pre-crt and post-crt evaluations including CT scan of esophageal wall thickness and assessment of TNM stage); EUS of the length of mucosal involvement and assessment of TNM stage; and PET assessment of SUV of primary malignancy, primary regional distant. Analyses were performed both of the post-crt measurement

4 Ann Thorac Surg SWISHER ET AL 2004;78: PET, CT, AND EUS IN ESOPHAGEAL CANCER Table 2. CT, EUS, and PET Assessment of Pathologic Response a Response b Non-Response b Total p Value c Post CRT Evaluation: d CT Esophageal Thickness (mm) N Mean Std. Deviation EUS Mass Size (cm) 5 N Mean Std. Deviation PET SUV Primary N Mean Std. Deviation SUV Primary Regional Distant N Mean Std. Deviation Pre Post CRT Evaluation: e (% Decrease) CT (% Decrease) Esophageal Thickness (mm) N Mean Std. Deviation 14.2% % % EUS (% Decrease) Mass Size (cm) N Mean Std. Deviation 86.1% % % PET (% Decrease) SUV Primary N Mean Std. Deviation 58% % % SUV Primary Regional Distant N Mean Std. Deviation 49.9% % % GENERAL THORACIC a Diagnostic studies performed before (Pre) and/or after (Post) CRT and prior to surgical resection; b Pathologic response defined as 0 to 10% viable cancer cells in specimen. Pathologic non-response defined as 10% viable cancer cells in specimen; c p value 0.05 defined as significant. Evaluation performed of pathologic response vs non-response; d Percent decrease following CRT determined by formula: [(pre post)/pre] * 100. and the relative percent change with treatment [(prepost)/pre] 100. The measurements that correlated most closely with pathologic response are listed in Table 2. Table 2 demonstrates the post-crt evaluation was able to distinguish responders from nonresponders more consistently than the percent decrease with pre-crt and post-crt studies although this study was not powered to prove this statistically. Assessments of TNM status were not useful (data not shown). From the post-crt evaluations (Table 2) we developed thresholds for each modality to differentiate pathologic responders from nonresponders (CT [esophageal thickness, 14.5 mm]; EUS [mucosal mass length, 1cm]; PET [primary SUV, 4]; and PET [primary regional distant SUV, 6]). We then utilized these thresholds to assess the sensitivity, specificity and accuracy of each modality for pathologic nonresponse. Post-CRT PET SUV of the primary was the most accurate (76%) in assessment of pathologic nonresponse (Table 3). Long-Term Outcome We utilized the determined thresholds that differentiated pathologic responders from nonresponders and performed Kaplan-Meier analyses of the three modalities. Patients who died in the hospital or within 30 days of surgery were excluded from this analysis (n 3) to allow assessment of tumor related survival. Only post-crt PET SUV equal to or greater than 4 of the primary predicted long-term survival (Figs 1 3). The 18-month survival of patients with a post-crt FDG-PET SUV greater than or equal to 4 was 34% compared with 77% for patients with an SUV less than 4 (p 0.01). Univariate Cox regression analyses of multiple preoperative factors (including type of surgical resection) demonstrated that only histology and post-crt PET SUV predicted longterm survival (Table 4). Multivariate analysis of the three modalities and histology showed that only post-crt

5 1156 SWISHER ET AL Ann Thorac Surg PET, CT, AND EUS IN ESOPHAGEAL CANCER 2004;78: Table 3. Accuracy in Predicting Pathologic Non-Response a Sensitivity b Specificity c Accuracy d Post-CRT CT Esophageal Thickness 51% 69% 62% ( 14.5mm) Post-CRT EUS Mass Size ( 1cm) 56% 75% 68% Post-CRT PET SUV Primary ( 4) 62% 84% 76% Post-CRT PET SUV Prim. Reg. Dist. ( 6) 69% 78% 75% a Pathologic non-response 10% viable cancer in primary tumor; b Sensitivity TP / (TP FN); c Specificity TN / (TN FP); d Accuracy (TP TN) / (TP FP TN FN). FDG-PET and CT thickness were independent predictors of long-term survival (Table 5). Comment The long-term survival of patients with locoregionally advanced esophageal cancer treated with surgery alone is poor, with 3-year survival rates ranging from 6% to 40% and a median survival of 9 to 24 months for patients with adenocarcinoma [1 6]. In an attempt to improve survival, a multimodality approach to therapy using preoperative chemoradiation (CRT) has been advocated. However, of the three randomized studies evaluating preoperative CRT [7 9] strategies in adenocarcinoma, only one study has shown a statistical survival advantage. In all these studies, however, improved survival has been noted in the subset of patients who have had a complete or almost complete pathologic response to induction therapy [7 9]. The ability to identify these patients could potentially allow more appropriate selection of multimodality treatments for patients who are otherwise treated Fig 2. Kaplan-Meier survival analysis of esophageal cancer patients according to post-crt EUS size (mm) at primary tumor site (p 0.17). (CRT concurrent chemotherapy and radiotherapy; EUS endoscopic ultrasonography.) as a homogenous group. We have previously evaluated the ability of FDG-PET imaging to predict pathologic response and long-term survival [17] and have found that it is able to identify a subset of patients with macroscopic residual cancer and poor long-term prognosis. In this study we retrospectively compared FDG-PET imaging with other commonly used modalities (CT and EUS) to determine the modality that optimally predicts pathologic response and survival before surgical resection. Our study resulted in several important findings. First, we confirmed in a larger group of patients that FDG-PET imaging could identify patients who have a large amount of residual disease (macroscopic residual, 10% viable, pathologic nonresponse) after preoperative CRT. Although FDG-PET imaging is unable to distinguish microscopic residual ( 10% viable) disease from a complete pathologic response, it is able to separate out a group of patients with macroscopic residual disease (Table 2). Fig 1. Kaplan-Meier survival analysis of esophageal cancer patients according to post-crt PET SUV at primary tumor site (p 0.01). (CRT concurrent chemotherapy and radiotherapy; PET positron emission tomography; SUV standardized uptake value.) Fig 3. Kaplan-Meier survival analysis of esophageal cancer patients according to post-crt CT scan thickness at primary tumor site (p 0.20). (CRT concurrent chemotherapy and radiotherapy; CT computed tomographic scan.)

6 Ann Thorac Surg SWISHER ET AL 2004;78: PET, CT, AND EUS IN ESOPHAGEAL CANCER Table 4. Univariate Cox Regression of Survival a HR 95% CI p Value b Age Gender Male Female (Reference) Histology Squamous cell ACA (Reference) Tumor Location Lower/gastric/GEJ Cervical/upper/middle (Reference) Clinical stage Stage IV Stage III Stage II (Reference) SUV post primary c (Reference) SUV Post Primary Regional Distant c (Reference) EUS Post Mass Size (cm) c (Reference) CT post esophageal thickness (mm) c (Reference) a Survival analysis excludes 3 perioperative deaths (n 100); b p value 0.05 considered significant; c Diagnostic studies performed upon completion of CRT and prior to surgical resection. Table 5. Multivariable Cox Regression of Survival a 1157 HR 95% CI p Value b Histology Squamous cell ACA (Reference) SUV Post Primary c (Reference) EUS post mass size (cm) c (Reference) CT post esophageal thickness (mm) c (Reference) a Survival analysis excludes 3 perioperative deaths (n 100); b p value 0.05 considered significant; c Diagnostic studies performed upon completion of CRT and prior to surgical resection. Interestingly, both CT scans and EUS were also able to separate these groups to some extent when esophageal thickness (CT) and mucosal tumor size (EUS) were utilized. Several authors have described changes in size of the mass correlating with response in EUS but none have utilized mucosal mass length [18, 19]. As noted by others, assessment of TNM status by CT or EUS does not appear to correlate with pathologic response [14, 15]. This may be because these studies do not assess viability and cannot distinguish between fibrosis and viable cancer. Although FDG-PET scans can distinguish viable from nonviable tissue, false positive results can occur because metabolically active leukocytes and macrophages associated with post-crt inflammation can lead to an elevated SUV [20, 22]. Despite this limitation, FDG-PET imaging is more accurate at identifying pathologic nonresponders (Table 3) when compared with the other modalities. Several authors have suggested that two serial FDG- PET or EUS studies are needed to predict outcome because this allows the percent decrease relative to the baseline to be determined [11 13, 18, 19, 23, 24]. Our study suggests, however, that a single post-crt PET, EUS or CT scan may be sufficient to differentiate pathologic responders from nonresponders. This observation requires further analysis in a prospective study since the threshold values in our study were determined in a retrospective fashion based on analyzed data. Our findings also suggest that post-crt evaluation may be applicable to a broad range of patients treated with a variety of CRT regimens since our study evaluated patients treated both on and off-protocol with concurrent chemoradiation alone or induction chemotherapy followed by concurrent chemoradiation. In this regard, the timing of the imaging may be more important than the type of CRT so that the impact of post-crt inflammation and tumor size remains a constant. Another important observation was that only FDG- PET imaging was able to predict both the responses to preoperative CRT and poor prognosis (Figs 1 3, Tables 4 and 5). In our study patients who had a post-crt FDG-PET SUV of equal to or greater than 4 had significantly worse 2-year survival than did those with an SUV of less than 4 (34% vs 64%, p 0.01). The multivariable observation that CT scan esophageal thickness also correlated with survival (Table 5) may be a statistical abnormality since neither the univariate analysis nor the Kaplan-Meier evaluation revealed a CT scan esophageal thickness correlation with long-term outcome (Fig 3, Table 4). In fact, post-crt FDG-PET SUV was the only preoperative factor we evaluated that could predict survival except histology in a univariate fashion (Table 4). Other smaller studies in esophageal cancer have produced similar preliminary findings [11 13]. Because no other imaging modality can accurately predict prognosis before surgical resection, the importance of these observations cannot be emphasized enough. In this regard we have recently observed that many patients who receive induction chemotherapy before concurrent chemoradiation appear to have a higher pathologic response and a better prognosis [16]. Accordingly, if FDG-PET imaging could reliably identify CRT response then multimodality therapy could be tailored to individual patients to mini- GENERAL THORACIC

7 1158 SWISHER ET AL Ann Thorac Surg PET, CT, AND EUS IN ESOPHAGEAL CANCER 2004;78: mize treatment-related morbidity while maximizing therapeutic benefit. Recently it has been reported that PET imaging can predict pathologic response two weeks after initiating therapy [23 25]. This would be clinically useful as changes in induction therapy could be initiated before the completion of radiation therapy although this would have to be balanced by the increased toxicity of additional treatment. Additionally, if these observations are borne out in future trials then therapeutic strategies incorporating multiple regimens could be utilized based on early FDG-PET response. A third important observation of this study is that imaging with FDG-PET, CT, and EUS is unable to confirm the absence of residual viable disease in the primary tumor (Table 2, 3). Our study evaluated pathologic response and distinguished macroscopic residual disease ( 10%, responder) from microscopic residual and complete response (0% to 10% viable, nonresponder). We observed that the accuracy decreased dramatically when an attempt is made to distinguish microscopic residual disease (1% to 10% viable) from a complete (0% viable) pathologic response (data not shown). The clinical implication of these observations is that there is currently no imaging modality that can confirm a complete response after preoperative CRT. Consequently, patients with an apparent complete response after CRT can have residual disease and should not be precluded from additional therapeutic management. In summary, our study is a retrospective study but suggests that post-crt FDG-PET imaging may be useful for identifying a subset of esophageal cancer patients who have failed to respond to CRT and whose prognosis is poor. These results need to be confirmed in a prospective multiinstitutional setting but preliminary findings suggest that no modality not even FDG-PET can rule out the presence of residual microscopic disease after CRT. Therefore, esophagectomy should remain a therapeutic option even when the post-crt imaging modalities are normal. We thank Debbie Smith for help in preparation and review of the manuscript. Support for this study was obtained in part from the George Sweeney Esophageal Research Fund. References 1. Swisher SG, Hunt KK, Holmes EC, et al. Changes in the surgical management of esophageal cancer from 1970 to Am J Surg 1995;69: Muller JM, Erasmi H, Stelzner M, et al. Surgical therapy of oesophageal carcinoma. Br J Surg 1990;77: Hulscher JBF, van Sandick JW, de Boer AGEM, et al. Extended transthoracic resection compared with limited transhiatal resection for adenocarcinoma of the esophagus. N Engl J Med 2002;347: Roth JA, Pass HI, Flanagan MM, et al. Randomized clinical trial of preoperative and postoperative adjuvant chemotherapy with cisplatin, vindesine, and bleomycin for carcinoma of the esophagus. J Thorac Cardiovasc Surg 1988;96: Hagen JA, DeMeester SR, Peters JH, et al. Curative resection for esophageal adenocarcinoma: analysis of 100 en bloc esophagectomies. Ann Surg 2001;234: Hofstetter W, Swisher SG, Correa AM, et al. Treatment outcomes of resected esophageal cancer. Ann Surg 2002;236: Walsh TN, Noonan N, Hollywood D, et al. A comparison of multimodal therapy and surgery for esophageal adenocarcinoma. N Engl J Med 1996;335: Urba SG, Orringer MB, Turrisi A, et al. Randomized trial of preoperative chemoradiation versus surgery alone in patients with locoregional esophageal carcinoma. J Clin Oncol 2001;19: Burmeister BH, Smithers BM, Fitzgerald L, et al. A randomized phase III trial of preoperative chemoradiation followed by surgery (CR-S) versus surgery alone (S) for localized resectable cancer of the esophagus [abstract]. Proc Am Soc Clin Oncol 2002;21:130a. 10. Kelsen DP, Ginsberg R, Pajak TG, et al. Chemotherapy followed by surgery compared with surgery alone for localized esophageal cancer. N Engl J Med 1998;339: Downey RJ, Akhurst T, Ilson D, et al. Whole body 18FDG- PET, and the response of esophageal cancer to induction therapy. results of a prospective trial. J Clin Oncol 2003;21: Brucher BLDM, Weber W, Bauer M, et al. Neoadjuvant therapy of esophageal squamous cell carcinoma: response evaluation by positron emission tomography. Ann Surg 2001;233: Flamen P, Van Cutsem E, Lerut A, et al. Positron emission tomography for assessment of the response to induction radiochemotherapy in locally advanced oesophageal cancer. Ann Oncol 2002;13: Beseth BD, Bedford R, Isacoff WH, et al. Endoscopic ultrasound does not accurate assess pathologic stage of esophageal cancer after neoadjuvant chemoradiotherapy. Am Surg 2000;66: Jones DR, Parker LA, Detterbeck FC, et al. Inadequacy of computed tomography in assessing patients with esophageal carcinoma after induction chemoradiotherapy. Cancer 1999;85: Swisher SG, Ajani JA, Komaki R, et al. Long-term outcome of phase II trial evaluating chemotherapy, chemoradiotherapy, and surgery for locoregionally advanced esophageal cancer. Int J Radiat Oncol Biol Phys 2003;57: Swisher S, Erasmus J, Maish M, et al. PET predicts pathologic response and long-term survival following preoperative chemoradiation in esophageal cancer [abstract]. Proc Am Soc Clin Oncol 2004 (in press). 18. Chak A, Canto MI, Cooper GS, et al. Endosonographic assessment of multimodality therapy predicts survival of esophageal carcinoma patients. Cancer 2000;88: Bowrey DJ, Clark GWB, Roberts SA, et al. Serial endoscopic ultrasound in the assessment of response to chemoradiotherapy for carcinoma of the esophagus. J Gastrointest Surg 1999;3: Bakheet SM, Saleem M, Powe J, et al. F-18 fluorodeoxyglucose chest uptake in lung inflammation, and infection. Clin Nucl Med 2000;25: Bhargava P, Reich P, Alavi A, et al. Radiation-induced esophagitis on FDG PET imaging. Clin Nucl Med 2003;28: Gradinscak DJ, Fulham MJ, Mohamed A, et al. Skeletal muscle uptake detected on FDG PET 48 hours after exertion. Clin Nucl Med 2003;28: Weber WA, Ott K, Becker K, et al. Prediction of response to reoperative chemotherapy in adenocarcinomas of the esophagogastric junction by metabolic imaging. J Clin Oncol 2001;19: Ott K, Fink U, Becker K, et al. Prediction of response to preoperative chemotherapy in gastric carcinoma by metabolic imaging: results of a prospective trial. J Clin Oncol 2003;21: Weider HA, Brucher BL, Zimmermann F, et al. Time course of tumor metabolic activity during chemoradiotherapy of esophageal squamous cell carcinoma and response to treatment. J Clin Oncol 2004;22:900 8.

8 Ann Thorac Surg SWISHER ET AL 2004;78: PET, CT, AND EUS IN ESOPHAGEAL CANCER DISCUSSION DR MARK J. KRASNA (Baltimore, MD): I enjoyed the talk. I did want to make at least one comment and ask you a question. The comments regarding the comparison that you did on that one slide showing the comparisons between the sensitivity and specificity. There were really only 64 patients from the abstract and from your presentation who had all 3 modalities. You had some who had CTs, some who had EUS, some who had PET, but really the comparison that you re making for PET scan now and I would suggest for your manuscript that you revise this that should have the McNemar s test where you are comparing between the sensitivity and the specificity, and you have to have the same n in all of those that you re comparing. I would suggest for that comparison that you just take the 64 who had all of the tests, and see if it still holds true. My question for you is regarding your summary, but really what you haven t told us, which is your conclusion. Your institution recently has advocated doing only salvage esophagectomy for patients after chemoradiation. I wonder if you are now accepting that not only is path CR, complete path CR going to be the predictor for best long-term prognosis, but even basically partial response is a predictor for long-term prognosis. Does this mean now that you are going to go back and do more esophagectomies on patients postchemorad, or do you still believe in this concept of salvage esophagectomy? DR SWISHER: Thank you, Dr Krasna. You brought up some very interesting points. Our institution does not put forward the concept that we should proceed with salvage esophagectomy after chemoradiation as first line therapy. We still think that this is something that needs to be tested. At our institution, if patients are in good physiologic condition and have locoregionally advanced esophageal cancer, we typically treat them with preoperative chemoradiation and surgery. Having said that, we had hoped very much that the PET scan might be able to pick out those people who had a complete pathologic response without viable tumor who might be candidates to avoid surgery in. Unfortunately, the PET scan is not able to rule out microscopic residual and is not, therefore, going to be able to definitively tell you that there is no viable cancer without surgical resection. DR THOMAS A. D AMICO (Durham, NC): Dr Swisher, this is an excellent study. I have a couple of questions on the use of the PET. First of all, have you looked at the pretreatment PET regarding survival and the predictability of whether someone would be sterilized by chemoradiation? If you look at lung cancer, the pretreatment PET actually tells you all you a great deal about relative survival. It doesn t change our decision to operate or not operate, but you get a lot of information on that pretreatment PET. Have you analyzed that? Number two, with regard to sterilization, it has been shown that the most important aspect of sterilization is if the lymph nodes are sterilized. Now, you have carefully analyzed both the SUV (standardized uptake value) and microscopic versus complete sterilization of the primary, but do you think PET is sensitive or specific enough to look at the lymph node status only, because that appears to be more closely linked to survival. Lastly, are you advocating restaging PET for all patients after therapy with esophagus cancer? Currently that s fundable for lung cancer patients, but in esophagus cancer, it s one patient, one PET. Do you think everybody should be able to get two PETs? Thanks. That was a great presentation DR SWISHER: Thank you, Dr D Amico. For time purposes, I did not show you all the multiple analyses that we performed, but one of them included the PET SUV of the primary and another one combined the PET SUV of the primary with the regional lymph nodes. We found that there was no added benefit to adding the SUV of the regional lymph nodes. Additionally, the pretreatment PET did not predict responders to chemoradiation only the posttreatment PET after the chemoradiation had been delivered. We do not advocate restaging PET for all patients at this time until further data can confirm these results. In the future, it may be possible to add additional preoperative therapy if the PET SUV does not show response after initial chemoradiation. DR FRANK C. DETTERBECK (Chapel Hill, NC): I enjoyed your presentation. We have looked at the issue of restaging as well and found that neither CT nor endoscopy and biopsy predicted who was a complete responder, in other words, in whom we could avoid surgery because they had already responded. My interpretation of what you ve said is that you agree with that and furthermore that PET is also not able to predict a complete response. The converse, though, is also of interest. You imply that a PET that is still positive predicts patients that still have tumor present, and that perhaps we should not operate on those patients or we should do something in addition. But the key issue in being able to apply this logic to an individual patient is definition of the false positive rate of a posttreatment PET scan. What is your false-positive rate for PET? In other words, if you have a positive PET, how many of those patients actually had no viable tumor at the time of resection? If this rate is low enough, then perhaps we can say that they need additional treatment, but, if not, then I m not so sure we can. I bring this question up because this is after radiation, and, of course, radiation causes PET to be positive for a while. In addition, could you comment on what interval after finishing the radiation treatment you think the PET should be done in order to avoid that? DR SWISHER: Thank you very much Dr Detterbeck. Those are very important points. We performed these PET scans 4 to 6 weeks after the completion of radiation therapy to try and keep consistency, but there clearly still is a false-positive from inflammation due to radiation that throws us off a little bit. The false-positive rate is about 34% when we look at this closely, and some of that is due to postchemoradiation changes such as esophagitis. These are clearly false-positives. It is interesting that there have been some recent reports from Europe in which they have performed PET scans 2 weeks after starting chemotherapy rather than after the completion of therapy and they seem to have a much lower false-positive rate. This early time point still appears to be predictive of long-term survival and pathologic response so this may be able to allow clinicians to change therapy if the patient s do not appear to be responding. The problem now is that there are few alternative treatments to try but in the future with the advent of the new biological therapies this may change. DR JEFFREY A. HAGEN (Los Angeles, CA): I would like to follow-up Dr Detterbeck s question with a suggestion and then pose a different question. GENERAL THORACIC

9 1160 SWISHER ET AL Ann Thorac Surg PET, CT, AND EUS IN ESOPHAGEAL CANCER 2004;78: If you have the chance to revise this manuscript, you might consider reporting the positive and negative predictive values rather than simply the sensitivity and specificity values. This would give the reader more useful information, providing a better estimate of the false-positive and false-negative rates. My question is, since most of the literature would suggest that the benefit, if any, of chemoradiation therapy is in the group of patients who are complete responders, I wonder if you could explain the rationale behind choosing to include both complete and partial responders as the endpoint to your study? Thanks for a nice presentation. DR SWISHER: Thank you, Dr Hagen. We have found in our experience that there tends to be three groups of patients. There are patients who have a complete pathologic response that do the best, and their long-term survival is about 60% to 70%. There is another group with macroscopic residual ( 50% viable tumor), and those do the least well and they have about a 20% long-term survival. Those patients, however, who have a microscopic residual disease ( 50% viable tumor) fall somewhere in between and have about a 40% long-term survival so we tend to group the microscopic residuals and the pathologic complete responders together as responders to chemoradiation. DR JAMES D. LUKETICH (Pittsburgh, PA): I enjoyed your study. Did you have trouble interpreting the EUS postchemoradiation? Is this why you just chose not to present the T status pretreatment and posttreatment? Since it is problematic to delineate the esophageal layers postchemoradiation, how were you measuring the lesion, the volume size? I notice you were presenting a lesion size. How were you certain the visible lesion was viable tumor and not scar tissue? And after you made those measurements, did you individually look at the pathology specimen and see how close that EUS assessment of viable tumor was indeed viable tumor? We have had a lot of trouble interpreting EUS postchemoradiation. DR SWISHER: Thank you Dr Luketich. You bring up some very good points. We looked at a variety of different things with the EUS, and as has been previously reported the posttreatment stage by EUS following chemoradiation is not very helpful because one cannot differentiate between fibrosis and residual viable tumor. The measurement that we used was mucosal mass size which was reproducible and could be very easily reported and followed. As I mentioned before, we had been hopeful that the PET scan would be able to get over some of the problems with EUS and CT in differentiating between fibrosis and viable cancer. Unfortunately, I do not think that the PET is sensitive enough to differentiate microscopic residual disease from fibrosis since both have low PET SUV uptake. The PET SUV does appear to pick up macroscopic residual disease and this is what is allowing a survival prediction since these patients tend to do poorly.