- Email: [email protected]
III Adenocarcinoma
Mediastinal Nodal Involvement After Neoadjuvant Chemoradiation for Siewert II/III Adenocarcinoma
Department of Thoracic and Cardiovascular Surgery, Department of Surgical Oncology, Department of Pathology, Department of Diagnostic Radiology, Department of Gastroenterology Hepatology and Nutrition, Department of Radiation Oncology, and Department of Gastrointestinal Medical Oncology, University of Texas MD Anderson Cancer Center, Houston, Texas
Background. Adenocarcinoma of the gastroesophageal junction (AEG) poses a management challenge, as preoperative prediction of occult mediastinal nodal metastasis is difficult. We sought to identify factors predictive of mediastinal involvement among patients undergoing trimodality therapy. Methods. Patients undergoing trimodality therapy for Siewert II and III AEG at a single institution between 2000 and 2015 were identified. Mediastinal involvement was defined as pathologic nodal involvement after neoadjuvant chemoradiation (ypND) in mediastinal stations or mediastinal recurrence 2 years or less after resection. Maximal c2 analysis and Youden’s J index were used to identify the pretreatment proximal tumor extent that best discriminated mediastinal involvement. Results. In all, 204 patients (151 [74%] AEG II, 53 [26%] AEG III) were included, of whom 47 (23%) had clinical evidence of thoracic nodal disease. Thirty-one of the 204 patients (15%) met criteria for mediastinal involvement (24 of 31 ypND, 10 of 31 mediastinal recurrence). Patients with mediastinal involvement had
greater proximal tumor extent (median 2 cm [interquartile range, 1.0 to 3.0 cm] vs 1.4 cm [interquartile range, 0.7 to 3.0 cm], P [ .030), were more frequently Siewert II lesions (27 of 31 [87.1%] vs 124 of 173 [71.7%], P [ .071), and were more often observed to have clinical thoracic nodal metastasis (cN) evidence (13 of 31 [42%] vs 34 of 173 [20%], P [ .007) than patients who did not. On multivariable analysis of patients with intrathoracic cN0, esophageal extent of 1.5 cm or greater was independently predictive of mediastinal involvement (odds ratio 5.46, P [ .011), whereas Siewert classification was not (Siewert II odds ratio 3.48, P [ .116). Conclusions. Pretreatment proximal tumor extent, rather than Siewert classification, is an independent predictor of mediastinal involvement among AEG II/III patients without clinical evidence of mediastinal metastasis and should be considered during treatment planning.
A
analyses of patterns of lymphatic drainage of the distal esophagus and proximal gastric cardia have largely corroborated this classification system.2-5 As a consequence, anatomic tumor location is frequently used to
denocarcinoma of the esophagogastric junction (EGJ) poses a distinct management challenge. The schema currently used to classify these tumors, which defines lesions by the relation of their epicenter to the true EGJ, was initially proposed after pathologic studies performed by Siewert and colleagues.1 Subsequent
(Ann Thorac Surg 2019;108:845-51) Ó 2019 by The Society of Thoracic Surgeons
Dr Das discloses a financial relationship with Adlai Nortye. Accepted for publication Apr 7, 2019. Presented at the Sixty-fifth Annual Meeting of the Southern Thoracic Surgical Association, Amelia Island, FL, Nov 7-10, 2018. Address correspondence to Dr Hofstetter, Department of Thoracic and Cardiovascular Surgery, University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd, Houston, TX 77030; email: whofstetter@ mdanderson.org.
Ó 2019 by The Society of Thoracic Surgeons Published by Elsevier Inc.
The Supplemental Table and Supplemental Figure can be viewed in the online version of this article [https:// doi.org/10.1016/j.athoracsur.2019.04.024] on http:// www.annalsthoracicsurgery.org.
0003-4975/$36.00 https://doi.org/10.1016/j.athoracsur.2019.04.024
GENERAL THORACIC
Kyle G. Mitchell, MD, Naruhiko Ikoma, MD, David B. Nelson, MD, MS, Dipen M. Maru, MD, Jeremy J. Erasmus, MD, Brian R. Weston, MD, Ara A. Vaporciyan, MD, Mara B. Antonoff, MD, Reza J. Mehran, MD, David C. Rice, MD, Jack A. Roth, MD, Stephen G. Swisher, MD, Boris Sepesi, MD, Garrett L. Walsh, MD, Arlene M. Correa, PhD, Prajnan Das, MD, MS, Mariela A. Blum, MD, Brian D. Badgwell, MD, and Wayne L. Hofstetter, MD
GENERAL THORACIC
846
MITCHELL ET AL SIEWERT II/III ADENOCARCINOMA
inform therapy decisions and the location of the tumor epicenter has been incorporated into the American Joint Committee on Cancer (AJCC) staging guidelines.6,7 Yet controversy persists regarding the anatomic location that best delineates esophageal and gastric cancers. Whereas type I tumors are typically managed according to esophageal cancer paradigms, the optimal approach to Siewert II and III lesions is less defined.3,8-10 Indeed, the seventh and eighth editions of the AJCC staging system are discordant in their classification of Siewert III tumors, with the more recent version classifying them as gastric cancers.6,7 Making a decision regarding the need for a total abdominal approach or an intrathoracic approach with esophagectomy requires preoperative identification of patients at risk for mediastinal nodal metastasis. In appropriate clinical scenarios, an adequate mediastinal lymphadenectomy will clear involved or at-risk nodal stations essential for the purposes of accurate staging and optimizing locoregional control. However, avoidance of unnecessary thoracotomy and its attendant morbidity is preferred when it is unlikely to provide a therapeutic benefit. Previous studies have suggested that proximal tumor extent can predict patterns of nodal involvement and could be used for preoperative risk stratification.4,5,11-13 However, these studies have largely examined patients undergoing surgical therapy without neoadjuvant therapy. As trimodality therapy has become the standard Western approach to the management of locoregionally advanced esophageal cancer and is being increasingly used in patients with gastric carcinoma, it is unclear whether location of the tumor epicenter or proximal tumor extent is the more discriminatory predictor of involvement of mediastinal nodal stations.14-16 As an ability to accurately predict mediastinal nodal metastasis could guide the extent of lymphadenectomy, we therefore sought to identify clinicopathologic characteristics predictive of mediastinal nodal disease among patients undergoing trimodality therapy for Siewert II and III adenocarcinoma. We hypothesized that pretreatment tumor extent proximal to the true EGJ would independently predict mediastinal involvement after neoadjuvant chemoradiotherapy.
Patients and Methods Patient Selection This study was approved by the University of Texas MD Anderson Cancer Center’s Institutional Review Board with a waiver of individual patient consent. Patients with EGJ adenocarcinoma classified as Siewert II (epicenter within 1 cm proximal and 2 cm distal to the true EGJ) or III (a gastric cardia tumor involving the EGJ with epicenter greater than 2 cm but not more than 5 cm distal to the EGJ) who received neoadjuvant chemoradiotherapy followed by curative-intent resection at the University of Texas MD Anderson Cancer Center (Houston, Texas) between 2000 and 2015 were eligible for inclusion. Patients were identified by querying
Ann Thorac Surg 2019;108:845-51
prospectively maintained clinical databases for the departments of surgical oncology and thoracic and cardiovascular surgery. Patients with distant metastases at time of presentation, patients with positive peritoneal cytology, patients undergoing salvage resection, and patients with synchronous malignancy other than indolent renal or thyroid carcinoma were excluded from analysis. Patients with pathologic nodal involvement after neoadjuvant chemoradiation (ypNþ) but without clear pathologic notation of the site of nodal involvement and who did not otherwise meet criteria for mediastinal involvement were also excluded.
Patient Management All patients underwent a thorough staging workup that included clinical examination, upper endoscopy, and axial imaging. Initial staging regularly included positron emission tomography and endoscopic ultrasonography with fine-needle aspiration in accordance with guidelines in effect at the time of treatment. Patients considered for trimodality therapy were evaluated and discussed in a multidisciplinary setting. Concurrent chemoradiotherapy regimens included a fluoropyrimidine (5-fluorouracil or capecitabine) with a platinum compound or taxane with 1.8 Gy fractions of radiotherapy delivered to a total dose of 45 to 50.4 Gy.17,18 The treatment volume included gross extent of tumor, a proximal and distal mucosal margin, and clinically involved or at-risk nodal basins.18 The choice of surgical approach was made at the treating surgeon’s discretion and was based on tumor location and clinical assessment of the extent of nodal involvement before and after neoadjuvant therapy. Postoperative surveillance protocols included clinical examination and axial imaging every 3 to 6 months for the first 2 years and then annually thereafter; upper endoscopy was performed when clinically indicated.
Variable and Outcome Definitions Pretreatment staging studies were retrospectively reviewed and the extent of tumor proximal to the true EGJ was typically identified using endoscopic measurements. Exception was made for patients without endoscopic measurements available (18 of 204, 8.8%), in which case tumor extent was estimated using endoscopic images and pathology measurements of the tumor bed. Clinical involvement of mediastinal nodal stations (cNChestþ) was defined as cytopathologically proven nodal disease (by fine-needle aspiration), ultrasonographic findings meeting validated criteria for nodal involvement, nodal hypermetabolism, or enlargement on axial imaging.15 Pathologic nodal involvement (ypNChestþ) was defined as nodal metastasis (viable tumor cells) identified on pathology review of the resected specimen. To account for the lack of pathology examination of mediastinal stations among patients who did not undergo mediastinal lymphadenectomy, a composite outcome of mediastinal involvement was defined as (1) pathologically positive intrathoracic nodes (ypNChestþ) or (2) mediastinal recurrence within 2 years of resection.13 Overall survival was defined as the time from surgery to
Ann Thorac Surg 2019;108:845-51
Table 1.
MITCHELL ET AL SIEWERT II/III ADENOCARCINOMA
847
Cohort Characteristics
Variable
No Mediastinal Involvement (n ¼ 173)
59 (51-66)
59 (15.5)
Mediastinal Involvement (n ¼ 31) 57 (14)
38 (18.6) 166 (81.4)
34 (19.7) 139 (80.3)
4 (12.9) 27 (87.1)
151 53 5 2
124 49 5 1.4
27 4 5 2
P Value .661 .374
.071 (74) (26) (3-6.8) (1-3)
(71.7) (28.3) (3-7) (0.7-3)
(87.1) (12.9) (4-6) (1-3)
73 (35.8) 131 (64.2)
63 (36.4) 110 (63.6)
10 (32.3) 21 (67.7)
151 (74) 53 (26)
127 (73.4) 46 (26.6)
24 (77.4) 7 (22.6)
25 (12.3) 179 (87.7)
22 (12.7) 151 (87.3)
3 (9.7) 28 (90.3)
79 (38.7) 125 (61.3) 8.5 (5.4-14.3)
72 (41.6) 101 (58.4) 8.3 (5.3-14)
7 (22.6) 24 (77.4) 9 (5.9-15.1)
4 80 120 120
(2) (39.2) (58.8) (58.9)
4 76 93 102
(2.3) (43.9) (53.8) (59)
0 4 27 18
(0) (12.9) (87.1) (58.1)
5 54 8 3 129 5
(2.5) (26.5) (3.9) (1.5) (63.2) (2.5)
4 54 8 2 102 3
(2.3) (31.2) (4.6) (1.2) (59) (1.7)
1 0 0 1 27 2
(3.2) (0) (0) (3.2) (87.1) (6.5)
.693 .030 .656
.639
.774
.045
.542 .002
.926 <.001
.651 40 (19.6) 164 (80.4)
33 (19.1) 140 (80.9)
7 (22.6) 24 (77.4) .360
57 27 26 91 3
(27.9) (13.2) (12.7) (44.6) (1.5)
51 23 24 72 3
(29.5) (13.3) (13.9) (41.6) (1.7)
6 4 2 19 0
(19.4) (12.9) (6.5) (61.3) (0)
129 48 20 7
(63.2) (23.5) (9.8) (3.4)
125 33 23 3
(72.3) (19.1) (6.9) (1.7)
4 15 8 4
(12.9) (48.4) (25.8) (12.9)
<.001
.348 151 (74.5) 52 (25.5)
131 (75.7) 42 (24.3)
21 (67.7) 10 (32.3) (Continued)
GENERAL THORACIC
Age, y Sex Female Male Siewert II III Pretreatment tumor length, cm Esophageal extent, cm Differentiation Well/moderate Poor Signet ring cell No Yes Clinical tumor cT1-2 cT3-4 Clinical node cN0 cNþ Pretreatment SUVmaxa XRT dose, Gy <45 45-50.4 50.4 Induction chemotherapy Approach Transhiatal esophagectomy Total gastrectomy Proximal gastrectomy Three-field esophagectomy TTTA esophagectomy Total esophagogastrectomy Abdominal lymphadenectomy D1 D1þ/D2 Yp tumor ypT0 ypT1 ypT2 ypT3 ypT4 Yp node ypN0 ypN1 ypN2 ypN3 Lymphovascular invasion No Yes
All Patients (n ¼ 204)
848
MITCHELL ET AL SIEWERT II/III ADENOCARCINOMA
Table 1.
Ann Thorac Surg 2019;108:845-51
Continued
Variable
All Patients (n ¼ 204)
Margin R0 R1/2
196 (96.1) 8 (3.9)
No Mediastinal Involvement (n ¼ 173)
Mediastinal Involvement (n ¼ 31)
P Value .105
168 (97.1) 5 (2.9)
28 (90.3) 3 (9.7)
a
Available in 164 (80.4%).
Values are median (interquartile range) or n (%).
GENERAL THORACIC
SUVmax, maximum standardized uptake value; TTTA, transthoracic transabdominal; yp, stage after systemic therapy and surgery.
death from any cause; patients alive at the end of the study period were censored at the date of last contact. Disease-free survival (DFS) was defined as time from surgery to disease recurrence or death; patients without a DFS event were similarly censored at the time of last follow-up. All tumors were staged according to the seventh edition of the AJCC staging system.7
Statistical Analysis Categoric variables were analyzed using Pearson’s c2 and Fisher’s exact test, and continuous data were analyzed using Kruskal-Wallis and Wilcoxon rank sum tests. To define the pretreatment proximal extent of tumor that best discriminated the outcome of interest, two methods were used to independently identify the optimal cutpoint. Maximal c2 analysis, which determines an optimal dichotomization of a continuous variable by identifying the cutpoint with the greatest c2 statistic, was performed.19 In addition, Youden’s J index (J ¼ Specificity þ Sensitivity 1) was calculated for each point on a receiver-operating characteristic curve, and the cutpoint with the greatest J index was selected as optimal.19 Logistic regression was used to identify factors predictive of mediastinal involvement. Variables with P less than .20 on univariable analysis and those identified a priori as being clinically relevant were included in the multivariable analysis; backward selection was performed until all covariates reached a significance threshold of P less than .10 or were of clinical interest. Given the substantial selection biases inherent in choice of radiation dose and operative approach, these variables were deliberately excluded from analysis of the composite outcome. Predictive performance of the final model was quantified using the area under the receiveroperating characteristic curve. Overall survival and DFS were estimated using the Kaplan-Meier method, and differences in time-to-event outcomes between groups were analyzed using the log rank test. For all analyses, statistical significance was defined as two-tailed P less than .05. All analyses were performed using IBM SPSS Statistics 24 (IBM Corp, Armonk, NY) and STATA 14.2 (StataCorp, College Station, TX).
Results Baseline Patient and Treatment Characteristics In all, 204 patients met inclusion criteria, of whom 81.4% (166 of 204) were men (Table 1; Supplemental Figure 1).
Most tumors were large (median 5 cm; interquartile range [IQR], 3 to 6.8 cm) and centered at the true EGJ (Siewert II, 151 of 204 [74%]). Only 4 patients (2%) did not complete neoadjuvant chemoradiation (<45.0 Gy). Patients with Siewert II tumors preferentially underwent transthoracic esophagectomy (119 of 151 [78.8%] vs total gastrectomy 20 [13.2%], transhiatal esophagectomy 4 [2.6%], proximal gastrectomy 4 [2.6%], three-field esophagectomy 3 [2%], and total esophagogastrectomy 1 [0.7%]). Most patients with Siewert II tumors received 50.4 Gy radiation (105 of 151 [69.5%]). In contrast, patients with Siewert III tumors most frequently underwent total gastrectomy (34 of 53 [64.2%] vs transthoracic esophagectomy 10 [18.9%], proximal gastrectomy 4 [7.5%], total esophagogastrectomy 4 [7.5%], and transhiatal esophagectomy 1 [1.9%]). Radiation dose was typically 45.0 Gy in this group (36 of 53 [67.9%]). No difference in the number of examined lymph nodes was observed between patients undergoing transthoracic operations (median 22.0; IQR, 15 to 27.8) and transabdominal operations (median 21.5; IQR, 17 to 28.8; P ¼ .415).
Predictors of Mediastinal Involvement Of 204 patients, 31 (15.2%) met criteria for mediastinal involvement. Pathologic involvement of mediastinal nodes was seen in 11.8% (24 of 204 patients, 20 of 24 paraesophageal nodes, 3 of 24 station 7, and 1 of 24 thoracic paraaortic), whereas 4.9% (10 of 204) had mediastinal recurrence within 2 years of resection. Three patients had both ypNþ mediastinal involvement and subsequent mediastinal recurrence. An approximately linear relationship was noted between proximal tumor extent and mediastinal involvement (Figure 1). Patients meeting criteria for mediastinal involvement (31 of 204, 15.2%) had greater proximal tumor extent and were more likely to have Siewert II tumors than patients who did not (Table 1). No differences were observed regarding signet ring cell histology, tumor differentiation, or pretreatment 18 F-fluorodeoxyglucose avidity of the primary tumor. Pretreatment proximal tumor extent of 1.5 cm or greater was identified as the optimal discriminatory cutpoint for mediastinal involvement (Figure 2). Because there was an interaction between clinical intrathoracic nodal status (cNChestþ 47 of 204 [23%]) and proximal tumor extent (interaction term P ¼ .011), the cohort was stratified and further analysis was restricted to patients who were cNChest0. This is also a clinically relevant cohort
Ann Thorac Surg 2019;108:845-51
MITCHELL ET AL SIEWERT II/III ADENOCARCINOMA
849
1.5 cm or greater was not associated with prognosis (overall survival MST 102 months, 95% CI, 64.4 to 139.5 vs 105 months, 95% CI, 77.6 to 132.4, P ¼ .741; and MDFST 102 months, 95% CI, 37.4 to 166.6, vs 82.5 months, 95% CI, 32.2 to 132.8, P ¼ .967). In contrast, the presence and extent of pathologic nodal disease proved to have deleterious effects on postoperative prognosis in that patients with intrathoracic nodal disease performed more poorly (Figure 3).
Comment
for whom risk of occult mediastinal disease would alter treatment decisions. On multivariable analysis of patients who were cNChest0 (157 of 204 [77%]), proximal tumor extent 1.5 cm or greater (odds ratio 5.46, 95% CI, 1.49 to 20.10, P ¼ .011) remained independently predictive of mediastinal involvement when controlling for Siewert classification and clinically involved abdominal nodes (Table 2; Supplemental Table 1). The final model demonstrated good predictive ability with an area under the curve of 0.76 (95% CI, 0.66 to 0.86).
Oncologic Outcomes After a median follow-up duration of 42.5 months, there were 91 deaths (91 of 204 [44.6%]) and 100 disease-free survival events (100 of 204 [49%]). Rates of 30-day and 90-day mortality were 0% (0 of 204) and 1.5% (3 of 204), respectively. Median overall survival time (MST) and disease-free survival time (MDFST) for the entire cohort were 102 months (95% CI, 77.2 to 126.8) and 86.1 months (95% CI, 45.3 to 126.8), respectively. No differences between Siewert II and III tumors were observed in terms of overall survival (Siewert II MST 86.1 months, 95% CI, 53.9 to 118.2; Siewert III 114.9 months, 95% CI, 63.4 to 166.4, P ¼ .240) or DFS (Siewert II MDFST 86.1 months, 95% CI, 37.1 to 135.0; Siewert III 82.5 months, 95% CI, 22.3 to 142.7, P ¼ .710). Similarly, pretreatment proximal tumor extent
In this retrospective analysis of a single-institution large cohort undergoing trimodality therapy for Siewert II and III adenocarcinoma, we found that pretreatment tumor extent above the true EGJ is associated with mediastinal nodal involvement among patients without clinical evidence of intrathoracic nodal metastasis. This association remained after controlling for anatomic Siewert classification and the clinical status of intraabdominal nodes. The present study has identified proximal tumor extent as an independent predictor of mediastinal involvement after neoadjuvant chemoradiotherapy for Siewert II and III tumors. Our findings are supported by results of previous studies that have identified proximal tumor extent to be associated with increased frequency of metastasis to mediastinal nodal stations among patients undergoing surgery alone.4,5,12,13 In a large Western series of patients with Siewert I-III adenocarcinoma, Meier and colleagues5 observed that a cutoff value of proximal tumor extent similar to that identified in the present study proved to be discriminatory of rates of lower mediastinal nodal metastases. In contrast to these previous reports, which were largely restricted to patients who did not receive neoadjuvant therapy, we exclusively examined patients undergoing trimodality therapy, which represents a modern treatment paradigm for locoregionally advanced gastroesophageal cancer and brings into question the ability to sterilize mediastinal stations with chemoradiation to forego larger resection fields. Although we observed a difference in rates of mediastinal involvement between patients with Siewert II and III tumors, these differences did not persist after adjustment for proximal tumor extent. This observation suggests that the extent of tumor growth proximal to the true Figure 2. Discriminatory ability of cutoff values of pretreatment proximal tumor extent for mediastinal involvement as determined by (A) maximal c2 analysis and (B) Youden’s J statistic.
GENERAL THORACIC
Figure 1. Frequency of pathologic mediastinal involvement by pretreatment proximal tumor extent: mediastinal recurrence (orange bars); ypNþChest (green bars); and all mediastinal involvement (gray bars). (ypN, pathologic nodal stage after neoadjuvant chemoradiation.)
850
MITCHELL ET AL SIEWERT II/III ADENOCARCINOMA
Ann Thorac Surg 2019;108:845-51
Table 2. Univariable and Multivariable Analysis of Predictors of Mediastinal Involvement Among Patients Without Clinical Evidence of Intrathoracic Nodal Metastasis
GENERAL THORACIC
Variables Univariable analysis Age 65 y Sex, malea Siewert, IIa Pretreatment tumor length Esophageal extent 1.5 cma Signet ring cell Pretreatment SUVmax Clinical tumor (reference, cT1-2) Abdominal cNþa Differentiation (poor) Lymphovascular invasion Induction chemotherapy Multivariable analysisb Siewert (II) Esophageal extent 1.5 cm Abdominal cNþ a
Included in multivariable analysis;
OR
95% CI
P Value
0.66 4.10 3.83 1.06 6.39 0.53 1.00 1.50
0.21-2.12 0.52-32.16 0.84-17.38 0.88-1.28 1.77-23.09 0.15-1.94 0.94-1.07 0.32-7.01
.486 .180 .082 .551 .005 .339 .989 .603
1.69 1.75 1.10 0.76
0.62-4.61 0.59-5.20 0.37-3.30 0.28-2.03
.307 .310 .864 .759
3.48 5.46 1.59
0.74-16.44 1.49-20.10 0.55-4.60
.116 .011 .389
b
Area under the curve 0.76.
CI, confidence interval; cN, clinical node; cT, clinical tumor; OR, odds ratio; SUVmax, maximum standardized uptake value.
EGJ, rather than the location of the tumor epicenter, may have greater clinical relevance. Previous studies have identified disparities in the risk of mediastinal nodal metastasis between Siewert types and have suggested that the benefit of mediastinal dissection varies based on the anatomic classification of the tumor.4,9,20 However, though most investigators have agreed that dissection of lesser curvature, paracardial, and left gastric artery nodes is warranted, observed rates of mediastinal metastasis for Siewert II and III lesions have varied widely.1,3-5,9,20,21 Some of these discrepancies may be due to differences in rates of mediastinal lymphadenectomy, although it is worth noting that some of the highest rates of mediastinal
Figure 3. (A) Overall survival and (B) disease-free survival by pathologic nodal status: ypN0 (green line); ypN1 (red line); ypN2 (blue line); and ypN3 (purple line). (ypN, pathologic nodal stage after neoadjuvant chemoradiation.)
nodal involvement in type II/III lesions were observed in Western series.5,21 Given our findings herein, we submit that the proximal extent of tumor may be a more discriminatory guide for therapeutic decisions than location of the tumor epicenter, as two tumors with widely disparate proximal and distal tumor growth may still be classified similarly under the Siewert system. Critically, it must be noted that the association between proximal tumor growth and mediastinal involvement appears to be best interpreted as a continuum or spectrum of risk rather than as a dichotomized or trichotomized relationship. As such, a definitive account of which tumors are best treated by transthoracic or exclusively transabdominal operations cannot be provided in the absence of study in the prospective setting; we therefore propose that proximal tumor extent be considered in the context of other relevant patient and tumor characteristics during the course of therapeutic planning. The results of the present study highlight a critical distinction between patients who are noted to have preoperative evidence of clinical nodal involvement and those who do not. The observation that the extent of proximal tumor growth was only associated with mediastinal involvement among patients who were cNChest0 is clinically salient, as a mediastinal lymphadenectomy is warranted in the setting of suspected mediastinal metastasis. Identification of a relationship between proximal tumor growth and risk of mediastinal nodal disease among this group may permit better preoperative stratification of the risk of clinically occult mediastinal nodal disease. The present study is limited by its retrospective nature and the inherent biases in patient selection and treatment allocation. That the majority of patients meeting criteria for mediastinal involvement received esophageal dose radiation and underwent a transthoracic dissection precluded an ability to examine whether variations in radiation protocols or surgical approach are associated with improved locoregional control. However, these findings confirm the value of multidisciplinary review for identifying appropriate treatment algorithms for patients with complex lesions; we believe that these multidisciplinary discussions may be better informed by the observations
Ann Thorac Surg 2019;108:845-51
5.
6. 7. 8.
9.
10.
11.
12.
13. 14. 15.
16.
This research has been supported by generous philanthropic donations from the Mason family and anonymous donors.
17.
References
18.
1. Rudiger Siewert J, Feith M, Werner M, Stein HJ. Adenocarcinoma of the esophagogastric junction: results of surgical therapy based on anatomical/topographic classification in 1, 002 consecutive patients. Ann Surg. 2000;232:353-361. 2. Fujitani K, Miyashiro I, Mikata S, et al. Pattern of abdominal nodal spread and optimal abdominal lymphadenectomy for advanced Siewert type II adenocarcinoma of the cardia: results of a multicenter study. Gastric Cancer. 2013;16:301-308. 3. Goto H, Tokunaga M, Miki Y, et al. The optimal extent of lymph node dissection for adenocarcinoma of the esophagogastric junction differs between Siewert type II and Siewert type III patients. Gastric Cancer. 2015;18:375-381. 4. Hosokawa Y, Kinoshita T, Konishi M, et al. Clinicopathological features and prognostic factors of adenocarcinoma of
19. 20.
21.
851
the esophagogastric junction according to Siewert classification: experiences at a single institution in Japan. Ann Surg Oncol. 2012;19:677-683. Meier I, Merkel S, Papadopoulos T, Sauer R, Hohenberger W, Brunner TB. Adenocarcinoma of the esophagogastric junction: the pattern of metastatic lymph node dissemination as a rationale for elective lymphatic target volume definition. Int J Radiat Oncol Biol Phys. 2008;70:1408-1417. Amin MB, Edge S, Greene F, et al., eds. AJCC Cancer Staging Manual. 8th ed. New York, NY: Springer; 2017. Edge SB, Byrd DR, Compton CC, Fritz AG, Greene FL, Trotti A, eds. AJCC Cancer Staging Handbook. 7th ed. New York, NY: Springer; 2010. Gertler R, Stein HJ, Loos M, Langer R, Friess H, Feith M. How to classify adenocarcinomas of the esophagogastric junction: as esophageal or gastric cancer? Am J Surg Pathol. 2011;35:1512-1522. Hosoda K, Yamashita K, Moriya H, Mieno H, Watanabe M. Optimal treatment for Siewert type II and III adenocarcinoma of the esophagogastric junction: a retrospective cohort study with long-term follow-up. World J Gastroenterol. 2017;23:2723-2730. Kulig P, Sierzega M, Pach R, Kolodziejczyk P, Kulig J. Differences in prognosis of Siewert II and III oesophagogastric junction cancers are determined by the baseline tumour staging but not its anatomical location. Eur J Surg Oncol. 2016;42:1215-1221. Koyanagi K, Kato F, Kanamori J, Daiko H, Ozawa S, Tachimori Y. Clinical significance of esophageal invasion length for the prediction of mediastinal lymph node metastasis in Siewert type II adenocarcinoma: a retrospective single-institution study. Ann Gastroenterol Surg. 2018;2:187196. Nunobe S, Ohyama S, Sonoo H, et al. Benefit of mediastinal and para-aortic lymph-node dissection for advanced gastric cancer with esophageal invasion. J Surg Oncol. 2008;97:392395. Kurokawa Y, Hiki N, Yoshikawa T, et al. Mediastinal lymph node metastasis and recurrence in adenocarcinoma of the esophagogastric junction. Surgery. 2015;157:551-555. van Hagen P, Hulshof MCCM, van Lanschot JJB, et al. Preoperative chemoradiotherapy for esophageal or junctional cancer. N Engl J Med. 2012;366:2074-2084. National Comprehensive Cancer Network: esophageal and esophagogastric junction cancers (version 4.2017). Available at: https://www.nccn.org/professionals/physician_gls/pdf/ esophageal.pdf. Accessed October 15, 2017. Ikoma N, Das P, Hofstetter W, et al. Preoperative chemoradiation therapy induces primary-tumor complete response more frequently than chemotherapy alone in gastric cancer: analyses of the National Cancer Database 2006–2014 using propensity score matching. Gastric Cancer. 2018;21:1004-1013. Sudo K, Taketa T, Correa AM, et al. Locoregional failure rate after preoperative chemoradiation of esophageal adenocarcinoma and the outcomes of salvage strategies. J Clin Oncol. 2013;31:4306-4310. Chakravarty T, Crane CH, Ajani JA, et al. Intensity-modulated radiation therapy with concurrent chemotherapy as preoperative treatment for localized gastric adenocarcinoma. Int J Radiat Oncol Biol Phys. 2012;83:581-586. Unal I. Defining an optimal cut-point value in roc analysis: an alternative approach. Comput Math Methods Med. 2017;2017:14. Hasegawa S, Yoshikawa T, Rino Y, et al. Priority of lymph node dissection for Siewert type II/III adenocarcinoma of the esophagogastric junction. Ann Surg Oncol. 2013;20:4252-4259. Leers JM, DeMeester SR, Chan N, et al. Clinical characteristics, biologic behavior, and survival after esophagectomy are similar for adenocarcinoma of the gastroesophageal junction and the distal esophagus. J Thorac Cardiovasc Surg. 2009;138:594-602 [discussion 601-602].
GENERAL THORACIC
presented herein. Further investigation is warranted to validate our findings in an independent cohort and to clarify the ability of different neoadjuvant radiotherapy protocols to effectively sterilize the mediastinum, ideally in a prospective fashion. In addition, the findings of the present report highlight a distinction between patients who have clinical evidence of mediastinal nodal metastasis and those who do not, and the rate of clinically occult pretreatment mediastinal nodal disease that was subsequently sterilized with neoadjuvant therapy is unclear. The authors acknowledge temporal changes in positron emission tomography and endoscopic ultrasonography technology used at their institutions, particularly during the early part of the study period, and the potential effects of these changes on the accuracy of preoperative staging. Finally, although comprehensive pathologic examination of mediastinal nodes was not performed in patients who did not undergo a formal mediastinal lymphadenectomy, we applied a previously utilized methodologic approach in an effort to circumvent this constraint.13 In summary, we identified pretreatment proximal tumor extent as an independent predictor of mediastinal involvement among patients undergoing trimodality therapy for Siewert II/III adenocarcinoma who did not have preoperative evidence of mediastinal nodal metastasis. In light of the approximately linear relationship between proximal tumor growth and the observed frequency of mediastinal involvement, we propose that tumor extent above the EGJ be considered as a continuum rather than in a dichotomized fashion. Proximal extent of 1.5 cm or greater marks a distance in which significantly more cases involved the mediastinum in our analyses, but even in patients with less proximal extent there was risk. Decisions as to what cutoff represents optimum should be taken on a case by case basis. The results presented herein suggest that clinical evidence of intrathoracic nodal metastasis and proximal tumor extent should be used to guide operative planning when evaluating patients with Siewert II/III adenocarcinoma, and that tumors with limited proximal extent may be amenable to transabdominal resection irrespective of the location of their epicenter.
MITCHELL ET AL SIEWERT II/III ADENOCARCINOMA
Department of Thoracic and Cardiovascular Surgery, Department of Surgical Oncology, Department of Pathology, Department of Diagnostic Radiology, Department of Gastroenterology Hepatology and Nutrition, Department of Radiation Oncology, and Department of Gastrointestinal Medical Oncology, University of Texas MD Anderson Cancer Center, Houston, Texas
Background. Adenocarcinoma of the gastroesophageal junction (AEG) poses a management challenge, as preoperative prediction of occult mediastinal nodal metastasis is difficult. We sought to identify factors predictive of mediastinal involvement among patients undergoing trimodality therapy. Methods. Patients undergoing trimodality therapy for Siewert II and III AEG at a single institution between 2000 and 2015 were identified. Mediastinal involvement was defined as pathologic nodal involvement after neoadjuvant chemoradiation (ypND) in mediastinal stations or mediastinal recurrence 2 years or less after resection. Maximal c2 analysis and Youden’s J index were used to identify the pretreatment proximal tumor extent that best discriminated mediastinal involvement. Results. In all, 204 patients (151 [74%] AEG II, 53 [26%] AEG III) were included, of whom 47 (23%) had clinical evidence of thoracic nodal disease. Thirty-one of the 204 patients (15%) met criteria for mediastinal involvement (24 of 31 ypND, 10 of 31 mediastinal recurrence). Patients with mediastinal involvement had
greater proximal tumor extent (median 2 cm [interquartile range, 1.0 to 3.0 cm] vs 1.4 cm [interquartile range, 0.7 to 3.0 cm], P [ .030), were more frequently Siewert II lesions (27 of 31 [87.1%] vs 124 of 173 [71.7%], P [ .071), and were more often observed to have clinical thoracic nodal metastasis (cN) evidence (13 of 31 [42%] vs 34 of 173 [20%], P [ .007) than patients who did not. On multivariable analysis of patients with intrathoracic cN0, esophageal extent of 1.5 cm or greater was independently predictive of mediastinal involvement (odds ratio 5.46, P [ .011), whereas Siewert classification was not (Siewert II odds ratio 3.48, P [ .116). Conclusions. Pretreatment proximal tumor extent, rather than Siewert classification, is an independent predictor of mediastinal involvement among AEG II/III patients without clinical evidence of mediastinal metastasis and should be considered during treatment planning.
A
analyses of patterns of lymphatic drainage of the distal esophagus and proximal gastric cardia have largely corroborated this classification system.2-5 As a consequence, anatomic tumor location is frequently used to
denocarcinoma of the esophagogastric junction (EGJ) poses a distinct management challenge. The schema currently used to classify these tumors, which defines lesions by the relation of their epicenter to the true EGJ, was initially proposed after pathologic studies performed by Siewert and colleagues.1 Subsequent
(Ann Thorac Surg 2019;108:845-51) Ó 2019 by The Society of Thoracic Surgeons
Dr Das discloses a financial relationship with Adlai Nortye. Accepted for publication Apr 7, 2019. Presented at the Sixty-fifth Annual Meeting of the Southern Thoracic Surgical Association, Amelia Island, FL, Nov 7-10, 2018. Address correspondence to Dr Hofstetter, Department of Thoracic and Cardiovascular Surgery, University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd, Houston, TX 77030; email: whofstetter@ mdanderson.org.
Ó 2019 by The Society of Thoracic Surgeons Published by Elsevier Inc.
The Supplemental Table and Supplemental Figure can be viewed in the online version of this article [https:// doi.org/10.1016/j.athoracsur.2019.04.024] on http:// www.annalsthoracicsurgery.org.
0003-4975/$36.00 https://doi.org/10.1016/j.athoracsur.2019.04.024
GENERAL THORACIC
Kyle G. Mitchell, MD, Naruhiko Ikoma, MD, David B. Nelson, MD, MS, Dipen M. Maru, MD, Jeremy J. Erasmus, MD, Brian R. Weston, MD, Ara A. Vaporciyan, MD, Mara B. Antonoff, MD, Reza J. Mehran, MD, David C. Rice, MD, Jack A. Roth, MD, Stephen G. Swisher, MD, Boris Sepesi, MD, Garrett L. Walsh, MD, Arlene M. Correa, PhD, Prajnan Das, MD, MS, Mariela A. Blum, MD, Brian D. Badgwell, MD, and Wayne L. Hofstetter, MD
GENERAL THORACIC
846
MITCHELL ET AL SIEWERT II/III ADENOCARCINOMA
inform therapy decisions and the location of the tumor epicenter has been incorporated into the American Joint Committee on Cancer (AJCC) staging guidelines.6,7 Yet controversy persists regarding the anatomic location that best delineates esophageal and gastric cancers. Whereas type I tumors are typically managed according to esophageal cancer paradigms, the optimal approach to Siewert II and III lesions is less defined.3,8-10 Indeed, the seventh and eighth editions of the AJCC staging system are discordant in their classification of Siewert III tumors, with the more recent version classifying them as gastric cancers.6,7 Making a decision regarding the need for a total abdominal approach or an intrathoracic approach with esophagectomy requires preoperative identification of patients at risk for mediastinal nodal metastasis. In appropriate clinical scenarios, an adequate mediastinal lymphadenectomy will clear involved or at-risk nodal stations essential for the purposes of accurate staging and optimizing locoregional control. However, avoidance of unnecessary thoracotomy and its attendant morbidity is preferred when it is unlikely to provide a therapeutic benefit. Previous studies have suggested that proximal tumor extent can predict patterns of nodal involvement and could be used for preoperative risk stratification.4,5,11-13 However, these studies have largely examined patients undergoing surgical therapy without neoadjuvant therapy. As trimodality therapy has become the standard Western approach to the management of locoregionally advanced esophageal cancer and is being increasingly used in patients with gastric carcinoma, it is unclear whether location of the tumor epicenter or proximal tumor extent is the more discriminatory predictor of involvement of mediastinal nodal stations.14-16 As an ability to accurately predict mediastinal nodal metastasis could guide the extent of lymphadenectomy, we therefore sought to identify clinicopathologic characteristics predictive of mediastinal nodal disease among patients undergoing trimodality therapy for Siewert II and III adenocarcinoma. We hypothesized that pretreatment tumor extent proximal to the true EGJ would independently predict mediastinal involvement after neoadjuvant chemoradiotherapy.
Patients and Methods Patient Selection This study was approved by the University of Texas MD Anderson Cancer Center’s Institutional Review Board with a waiver of individual patient consent. Patients with EGJ adenocarcinoma classified as Siewert II (epicenter within 1 cm proximal and 2 cm distal to the true EGJ) or III (a gastric cardia tumor involving the EGJ with epicenter greater than 2 cm but not more than 5 cm distal to the EGJ) who received neoadjuvant chemoradiotherapy followed by curative-intent resection at the University of Texas MD Anderson Cancer Center (Houston, Texas) between 2000 and 2015 were eligible for inclusion. Patients were identified by querying
Ann Thorac Surg 2019;108:845-51
prospectively maintained clinical databases for the departments of surgical oncology and thoracic and cardiovascular surgery. Patients with distant metastases at time of presentation, patients with positive peritoneal cytology, patients undergoing salvage resection, and patients with synchronous malignancy other than indolent renal or thyroid carcinoma were excluded from analysis. Patients with pathologic nodal involvement after neoadjuvant chemoradiation (ypNþ) but without clear pathologic notation of the site of nodal involvement and who did not otherwise meet criteria for mediastinal involvement were also excluded.
Patient Management All patients underwent a thorough staging workup that included clinical examination, upper endoscopy, and axial imaging. Initial staging regularly included positron emission tomography and endoscopic ultrasonography with fine-needle aspiration in accordance with guidelines in effect at the time of treatment. Patients considered for trimodality therapy were evaluated and discussed in a multidisciplinary setting. Concurrent chemoradiotherapy regimens included a fluoropyrimidine (5-fluorouracil or capecitabine) with a platinum compound or taxane with 1.8 Gy fractions of radiotherapy delivered to a total dose of 45 to 50.4 Gy.17,18 The treatment volume included gross extent of tumor, a proximal and distal mucosal margin, and clinically involved or at-risk nodal basins.18 The choice of surgical approach was made at the treating surgeon’s discretion and was based on tumor location and clinical assessment of the extent of nodal involvement before and after neoadjuvant therapy. Postoperative surveillance protocols included clinical examination and axial imaging every 3 to 6 months for the first 2 years and then annually thereafter; upper endoscopy was performed when clinically indicated.
Variable and Outcome Definitions Pretreatment staging studies were retrospectively reviewed and the extent of tumor proximal to the true EGJ was typically identified using endoscopic measurements. Exception was made for patients without endoscopic measurements available (18 of 204, 8.8%), in which case tumor extent was estimated using endoscopic images and pathology measurements of the tumor bed. Clinical involvement of mediastinal nodal stations (cNChestþ) was defined as cytopathologically proven nodal disease (by fine-needle aspiration), ultrasonographic findings meeting validated criteria for nodal involvement, nodal hypermetabolism, or enlargement on axial imaging.15 Pathologic nodal involvement (ypNChestþ) was defined as nodal metastasis (viable tumor cells) identified on pathology review of the resected specimen. To account for the lack of pathology examination of mediastinal stations among patients who did not undergo mediastinal lymphadenectomy, a composite outcome of mediastinal involvement was defined as (1) pathologically positive intrathoracic nodes (ypNChestþ) or (2) mediastinal recurrence within 2 years of resection.13 Overall survival was defined as the time from surgery to
Ann Thorac Surg 2019;108:845-51
Table 1.
MITCHELL ET AL SIEWERT II/III ADENOCARCINOMA
847
Cohort Characteristics
Variable
No Mediastinal Involvement (n ¼ 173)
59 (51-66)
59 (15.5)
Mediastinal Involvement (n ¼ 31) 57 (14)
38 (18.6) 166 (81.4)
34 (19.7) 139 (80.3)
4 (12.9) 27 (87.1)
151 53 5 2
124 49 5 1.4
27 4 5 2
P Value .661 .374
.071 (74) (26) (3-6.8) (1-3)
(71.7) (28.3) (3-7) (0.7-3)
(87.1) (12.9) (4-6) (1-3)
73 (35.8) 131 (64.2)
63 (36.4) 110 (63.6)
10 (32.3) 21 (67.7)
151 (74) 53 (26)
127 (73.4) 46 (26.6)
24 (77.4) 7 (22.6)
25 (12.3) 179 (87.7)
22 (12.7) 151 (87.3)
3 (9.7) 28 (90.3)
79 (38.7) 125 (61.3) 8.5 (5.4-14.3)
72 (41.6) 101 (58.4) 8.3 (5.3-14)
7 (22.6) 24 (77.4) 9 (5.9-15.1)
4 80 120 120
(2) (39.2) (58.8) (58.9)
4 76 93 102
(2.3) (43.9) (53.8) (59)
0 4 27 18
(0) (12.9) (87.1) (58.1)
5 54 8 3 129 5
(2.5) (26.5) (3.9) (1.5) (63.2) (2.5)
4 54 8 2 102 3
(2.3) (31.2) (4.6) (1.2) (59) (1.7)
1 0 0 1 27 2
(3.2) (0) (0) (3.2) (87.1) (6.5)
.693 .030 .656
.639
.774
.045
.542 .002
.926 <.001
.651 40 (19.6) 164 (80.4)
33 (19.1) 140 (80.9)
7 (22.6) 24 (77.4) .360
57 27 26 91 3
(27.9) (13.2) (12.7) (44.6) (1.5)
51 23 24 72 3
(29.5) (13.3) (13.9) (41.6) (1.7)
6 4 2 19 0
(19.4) (12.9) (6.5) (61.3) (0)
129 48 20 7
(63.2) (23.5) (9.8) (3.4)
125 33 23 3
(72.3) (19.1) (6.9) (1.7)
4 15 8 4
(12.9) (48.4) (25.8) (12.9)
<.001
.348 151 (74.5) 52 (25.5)
131 (75.7) 42 (24.3)
21 (67.7) 10 (32.3) (Continued)
GENERAL THORACIC
Age, y Sex Female Male Siewert II III Pretreatment tumor length, cm Esophageal extent, cm Differentiation Well/moderate Poor Signet ring cell No Yes Clinical tumor cT1-2 cT3-4 Clinical node cN0 cNþ Pretreatment SUVmaxa XRT dose, Gy <45 45-50.4 50.4 Induction chemotherapy Approach Transhiatal esophagectomy Total gastrectomy Proximal gastrectomy Three-field esophagectomy TTTA esophagectomy Total esophagogastrectomy Abdominal lymphadenectomy D1 D1þ/D2 Yp tumor ypT0 ypT1 ypT2 ypT3 ypT4 Yp node ypN0 ypN1 ypN2 ypN3 Lymphovascular invasion No Yes
All Patients (n ¼ 204)
848
MITCHELL ET AL SIEWERT II/III ADENOCARCINOMA
Table 1.
Ann Thorac Surg 2019;108:845-51
Continued
Variable
All Patients (n ¼ 204)
Margin R0 R1/2
196 (96.1) 8 (3.9)
No Mediastinal Involvement (n ¼ 173)
Mediastinal Involvement (n ¼ 31)
P Value .105
168 (97.1) 5 (2.9)
28 (90.3) 3 (9.7)
a
Available in 164 (80.4%).
Values are median (interquartile range) or n (%).
GENERAL THORACIC
SUVmax, maximum standardized uptake value; TTTA, transthoracic transabdominal; yp, stage after systemic therapy and surgery.
death from any cause; patients alive at the end of the study period were censored at the date of last contact. Disease-free survival (DFS) was defined as time from surgery to disease recurrence or death; patients without a DFS event were similarly censored at the time of last follow-up. All tumors were staged according to the seventh edition of the AJCC staging system.7
Statistical Analysis Categoric variables were analyzed using Pearson’s c2 and Fisher’s exact test, and continuous data were analyzed using Kruskal-Wallis and Wilcoxon rank sum tests. To define the pretreatment proximal extent of tumor that best discriminated the outcome of interest, two methods were used to independently identify the optimal cutpoint. Maximal c2 analysis, which determines an optimal dichotomization of a continuous variable by identifying the cutpoint with the greatest c2 statistic, was performed.19 In addition, Youden’s J index (J ¼ Specificity þ Sensitivity 1) was calculated for each point on a receiver-operating characteristic curve, and the cutpoint with the greatest J index was selected as optimal.19 Logistic regression was used to identify factors predictive of mediastinal involvement. Variables with P less than .20 on univariable analysis and those identified a priori as being clinically relevant were included in the multivariable analysis; backward selection was performed until all covariates reached a significance threshold of P less than .10 or were of clinical interest. Given the substantial selection biases inherent in choice of radiation dose and operative approach, these variables were deliberately excluded from analysis of the composite outcome. Predictive performance of the final model was quantified using the area under the receiveroperating characteristic curve. Overall survival and DFS were estimated using the Kaplan-Meier method, and differences in time-to-event outcomes between groups were analyzed using the log rank test. For all analyses, statistical significance was defined as two-tailed P less than .05. All analyses were performed using IBM SPSS Statistics 24 (IBM Corp, Armonk, NY) and STATA 14.2 (StataCorp, College Station, TX).
Results Baseline Patient and Treatment Characteristics In all, 204 patients met inclusion criteria, of whom 81.4% (166 of 204) were men (Table 1; Supplemental Figure 1).
Most tumors were large (median 5 cm; interquartile range [IQR], 3 to 6.8 cm) and centered at the true EGJ (Siewert II, 151 of 204 [74%]). Only 4 patients (2%) did not complete neoadjuvant chemoradiation (<45.0 Gy). Patients with Siewert II tumors preferentially underwent transthoracic esophagectomy (119 of 151 [78.8%] vs total gastrectomy 20 [13.2%], transhiatal esophagectomy 4 [2.6%], proximal gastrectomy 4 [2.6%], three-field esophagectomy 3 [2%], and total esophagogastrectomy 1 [0.7%]). Most patients with Siewert II tumors received 50.4 Gy radiation (105 of 151 [69.5%]). In contrast, patients with Siewert III tumors most frequently underwent total gastrectomy (34 of 53 [64.2%] vs transthoracic esophagectomy 10 [18.9%], proximal gastrectomy 4 [7.5%], total esophagogastrectomy 4 [7.5%], and transhiatal esophagectomy 1 [1.9%]). Radiation dose was typically 45.0 Gy in this group (36 of 53 [67.9%]). No difference in the number of examined lymph nodes was observed between patients undergoing transthoracic operations (median 22.0; IQR, 15 to 27.8) and transabdominal operations (median 21.5; IQR, 17 to 28.8; P ¼ .415).
Predictors of Mediastinal Involvement Of 204 patients, 31 (15.2%) met criteria for mediastinal involvement. Pathologic involvement of mediastinal nodes was seen in 11.8% (24 of 204 patients, 20 of 24 paraesophageal nodes, 3 of 24 station 7, and 1 of 24 thoracic paraaortic), whereas 4.9% (10 of 204) had mediastinal recurrence within 2 years of resection. Three patients had both ypNþ mediastinal involvement and subsequent mediastinal recurrence. An approximately linear relationship was noted between proximal tumor extent and mediastinal involvement (Figure 1). Patients meeting criteria for mediastinal involvement (31 of 204, 15.2%) had greater proximal tumor extent and were more likely to have Siewert II tumors than patients who did not (Table 1). No differences were observed regarding signet ring cell histology, tumor differentiation, or pretreatment 18 F-fluorodeoxyglucose avidity of the primary tumor. Pretreatment proximal tumor extent of 1.5 cm or greater was identified as the optimal discriminatory cutpoint for mediastinal involvement (Figure 2). Because there was an interaction between clinical intrathoracic nodal status (cNChestþ 47 of 204 [23%]) and proximal tumor extent (interaction term P ¼ .011), the cohort was stratified and further analysis was restricted to patients who were cNChest0. This is also a clinically relevant cohort
Ann Thorac Surg 2019;108:845-51
MITCHELL ET AL SIEWERT II/III ADENOCARCINOMA
849
1.5 cm or greater was not associated with prognosis (overall survival MST 102 months, 95% CI, 64.4 to 139.5 vs 105 months, 95% CI, 77.6 to 132.4, P ¼ .741; and MDFST 102 months, 95% CI, 37.4 to 166.6, vs 82.5 months, 95% CI, 32.2 to 132.8, P ¼ .967). In contrast, the presence and extent of pathologic nodal disease proved to have deleterious effects on postoperative prognosis in that patients with intrathoracic nodal disease performed more poorly (Figure 3).
Comment
for whom risk of occult mediastinal disease would alter treatment decisions. On multivariable analysis of patients who were cNChest0 (157 of 204 [77%]), proximal tumor extent 1.5 cm or greater (odds ratio 5.46, 95% CI, 1.49 to 20.10, P ¼ .011) remained independently predictive of mediastinal involvement when controlling for Siewert classification and clinically involved abdominal nodes (Table 2; Supplemental Table 1). The final model demonstrated good predictive ability with an area under the curve of 0.76 (95% CI, 0.66 to 0.86).
Oncologic Outcomes After a median follow-up duration of 42.5 months, there were 91 deaths (91 of 204 [44.6%]) and 100 disease-free survival events (100 of 204 [49%]). Rates of 30-day and 90-day mortality were 0% (0 of 204) and 1.5% (3 of 204), respectively. Median overall survival time (MST) and disease-free survival time (MDFST) for the entire cohort were 102 months (95% CI, 77.2 to 126.8) and 86.1 months (95% CI, 45.3 to 126.8), respectively. No differences between Siewert II and III tumors were observed in terms of overall survival (Siewert II MST 86.1 months, 95% CI, 53.9 to 118.2; Siewert III 114.9 months, 95% CI, 63.4 to 166.4, P ¼ .240) or DFS (Siewert II MDFST 86.1 months, 95% CI, 37.1 to 135.0; Siewert III 82.5 months, 95% CI, 22.3 to 142.7, P ¼ .710). Similarly, pretreatment proximal tumor extent
In this retrospective analysis of a single-institution large cohort undergoing trimodality therapy for Siewert II and III adenocarcinoma, we found that pretreatment tumor extent above the true EGJ is associated with mediastinal nodal involvement among patients without clinical evidence of intrathoracic nodal metastasis. This association remained after controlling for anatomic Siewert classification and the clinical status of intraabdominal nodes. The present study has identified proximal tumor extent as an independent predictor of mediastinal involvement after neoadjuvant chemoradiotherapy for Siewert II and III tumors. Our findings are supported by results of previous studies that have identified proximal tumor extent to be associated with increased frequency of metastasis to mediastinal nodal stations among patients undergoing surgery alone.4,5,12,13 In a large Western series of patients with Siewert I-III adenocarcinoma, Meier and colleagues5 observed that a cutoff value of proximal tumor extent similar to that identified in the present study proved to be discriminatory of rates of lower mediastinal nodal metastases. In contrast to these previous reports, which were largely restricted to patients who did not receive neoadjuvant therapy, we exclusively examined patients undergoing trimodality therapy, which represents a modern treatment paradigm for locoregionally advanced gastroesophageal cancer and brings into question the ability to sterilize mediastinal stations with chemoradiation to forego larger resection fields. Although we observed a difference in rates of mediastinal involvement between patients with Siewert II and III tumors, these differences did not persist after adjustment for proximal tumor extent. This observation suggests that the extent of tumor growth proximal to the true Figure 2. Discriminatory ability of cutoff values of pretreatment proximal tumor extent for mediastinal involvement as determined by (A) maximal c2 analysis and (B) Youden’s J statistic.
GENERAL THORACIC
Figure 1. Frequency of pathologic mediastinal involvement by pretreatment proximal tumor extent: mediastinal recurrence (orange bars); ypNþChest (green bars); and all mediastinal involvement (gray bars). (ypN, pathologic nodal stage after neoadjuvant chemoradiation.)
850
MITCHELL ET AL SIEWERT II/III ADENOCARCINOMA
Ann Thorac Surg 2019;108:845-51
Table 2. Univariable and Multivariable Analysis of Predictors of Mediastinal Involvement Among Patients Without Clinical Evidence of Intrathoracic Nodal Metastasis
GENERAL THORACIC
Variables Univariable analysis Age 65 y Sex, malea Siewert, IIa Pretreatment tumor length Esophageal extent 1.5 cma Signet ring cell Pretreatment SUVmax Clinical tumor (reference, cT1-2) Abdominal cNþa Differentiation (poor) Lymphovascular invasion Induction chemotherapy Multivariable analysisb Siewert (II) Esophageal extent 1.5 cm Abdominal cNþ a
Included in multivariable analysis;
OR
95% CI
P Value
0.66 4.10 3.83 1.06 6.39 0.53 1.00 1.50
0.21-2.12 0.52-32.16 0.84-17.38 0.88-1.28 1.77-23.09 0.15-1.94 0.94-1.07 0.32-7.01
.486 .180 .082 .551 .005 .339 .989 .603
1.69 1.75 1.10 0.76
0.62-4.61 0.59-5.20 0.37-3.30 0.28-2.03
.307 .310 .864 .759
3.48 5.46 1.59
0.74-16.44 1.49-20.10 0.55-4.60
.116 .011 .389
b
Area under the curve 0.76.
CI, confidence interval; cN, clinical node; cT, clinical tumor; OR, odds ratio; SUVmax, maximum standardized uptake value.
EGJ, rather than the location of the tumor epicenter, may have greater clinical relevance. Previous studies have identified disparities in the risk of mediastinal nodal metastasis between Siewert types and have suggested that the benefit of mediastinal dissection varies based on the anatomic classification of the tumor.4,9,20 However, though most investigators have agreed that dissection of lesser curvature, paracardial, and left gastric artery nodes is warranted, observed rates of mediastinal metastasis for Siewert II and III lesions have varied widely.1,3-5,9,20,21 Some of these discrepancies may be due to differences in rates of mediastinal lymphadenectomy, although it is worth noting that some of the highest rates of mediastinal
Figure 3. (A) Overall survival and (B) disease-free survival by pathologic nodal status: ypN0 (green line); ypN1 (red line); ypN2 (blue line); and ypN3 (purple line). (ypN, pathologic nodal stage after neoadjuvant chemoradiation.)
nodal involvement in type II/III lesions were observed in Western series.5,21 Given our findings herein, we submit that the proximal extent of tumor may be a more discriminatory guide for therapeutic decisions than location of the tumor epicenter, as two tumors with widely disparate proximal and distal tumor growth may still be classified similarly under the Siewert system. Critically, it must be noted that the association between proximal tumor growth and mediastinal involvement appears to be best interpreted as a continuum or spectrum of risk rather than as a dichotomized or trichotomized relationship. As such, a definitive account of which tumors are best treated by transthoracic or exclusively transabdominal operations cannot be provided in the absence of study in the prospective setting; we therefore propose that proximal tumor extent be considered in the context of other relevant patient and tumor characteristics during the course of therapeutic planning. The results of the present study highlight a critical distinction between patients who are noted to have preoperative evidence of clinical nodal involvement and those who do not. The observation that the extent of proximal tumor growth was only associated with mediastinal involvement among patients who were cNChest0 is clinically salient, as a mediastinal lymphadenectomy is warranted in the setting of suspected mediastinal metastasis. Identification of a relationship between proximal tumor growth and risk of mediastinal nodal disease among this group may permit better preoperative stratification of the risk of clinically occult mediastinal nodal disease. The present study is limited by its retrospective nature and the inherent biases in patient selection and treatment allocation. That the majority of patients meeting criteria for mediastinal involvement received esophageal dose radiation and underwent a transthoracic dissection precluded an ability to examine whether variations in radiation protocols or surgical approach are associated with improved locoregional control. However, these findings confirm the value of multidisciplinary review for identifying appropriate treatment algorithms for patients with complex lesions; we believe that these multidisciplinary discussions may be better informed by the observations
Ann Thorac Surg 2019;108:845-51
5.
6. 7. 8.
9.
10.
11.
12.
13. 14. 15.
16.
This research has been supported by generous philanthropic donations from the Mason family and anonymous donors.
17.
References
18.
1. Rudiger Siewert J, Feith M, Werner M, Stein HJ. Adenocarcinoma of the esophagogastric junction: results of surgical therapy based on anatomical/topographic classification in 1, 002 consecutive patients. Ann Surg. 2000;232:353-361. 2. Fujitani K, Miyashiro I, Mikata S, et al. Pattern of abdominal nodal spread and optimal abdominal lymphadenectomy for advanced Siewert type II adenocarcinoma of the cardia: results of a multicenter study. Gastric Cancer. 2013;16:301-308. 3. Goto H, Tokunaga M, Miki Y, et al. The optimal extent of lymph node dissection for adenocarcinoma of the esophagogastric junction differs between Siewert type II and Siewert type III patients. Gastric Cancer. 2015;18:375-381. 4. Hosokawa Y, Kinoshita T, Konishi M, et al. Clinicopathological features and prognostic factors of adenocarcinoma of
19. 20.
21.
851
the esophagogastric junction according to Siewert classification: experiences at a single institution in Japan. Ann Surg Oncol. 2012;19:677-683. Meier I, Merkel S, Papadopoulos T, Sauer R, Hohenberger W, Brunner TB. Adenocarcinoma of the esophagogastric junction: the pattern of metastatic lymph node dissemination as a rationale for elective lymphatic target volume definition. Int J Radiat Oncol Biol Phys. 2008;70:1408-1417. Amin MB, Edge S, Greene F, et al., eds. AJCC Cancer Staging Manual. 8th ed. New York, NY: Springer; 2017. Edge SB, Byrd DR, Compton CC, Fritz AG, Greene FL, Trotti A, eds. AJCC Cancer Staging Handbook. 7th ed. New York, NY: Springer; 2010. Gertler R, Stein HJ, Loos M, Langer R, Friess H, Feith M. How to classify adenocarcinomas of the esophagogastric junction: as esophageal or gastric cancer? Am J Surg Pathol. 2011;35:1512-1522. Hosoda K, Yamashita K, Moriya H, Mieno H, Watanabe M. Optimal treatment for Siewert type II and III adenocarcinoma of the esophagogastric junction: a retrospective cohort study with long-term follow-up. World J Gastroenterol. 2017;23:2723-2730. Kulig P, Sierzega M, Pach R, Kolodziejczyk P, Kulig J. Differences in prognosis of Siewert II and III oesophagogastric junction cancers are determined by the baseline tumour staging but not its anatomical location. Eur J Surg Oncol. 2016;42:1215-1221. Koyanagi K, Kato F, Kanamori J, Daiko H, Ozawa S, Tachimori Y. Clinical significance of esophageal invasion length for the prediction of mediastinal lymph node metastasis in Siewert type II adenocarcinoma: a retrospective single-institution study. Ann Gastroenterol Surg. 2018;2:187196. Nunobe S, Ohyama S, Sonoo H, et al. Benefit of mediastinal and para-aortic lymph-node dissection for advanced gastric cancer with esophageal invasion. J Surg Oncol. 2008;97:392395. Kurokawa Y, Hiki N, Yoshikawa T, et al. Mediastinal lymph node metastasis and recurrence in adenocarcinoma of the esophagogastric junction. Surgery. 2015;157:551-555. van Hagen P, Hulshof MCCM, van Lanschot JJB, et al. Preoperative chemoradiotherapy for esophageal or junctional cancer. N Engl J Med. 2012;366:2074-2084. National Comprehensive Cancer Network: esophageal and esophagogastric junction cancers (version 4.2017). Available at: https://www.nccn.org/professionals/physician_gls/pdf/ esophageal.pdf. Accessed October 15, 2017. Ikoma N, Das P, Hofstetter W, et al. Preoperative chemoradiation therapy induces primary-tumor complete response more frequently than chemotherapy alone in gastric cancer: analyses of the National Cancer Database 2006–2014 using propensity score matching. Gastric Cancer. 2018;21:1004-1013. Sudo K, Taketa T, Correa AM, et al. Locoregional failure rate after preoperative chemoradiation of esophageal adenocarcinoma and the outcomes of salvage strategies. J Clin Oncol. 2013;31:4306-4310. Chakravarty T, Crane CH, Ajani JA, et al. Intensity-modulated radiation therapy with concurrent chemotherapy as preoperative treatment for localized gastric adenocarcinoma. Int J Radiat Oncol Biol Phys. 2012;83:581-586. Unal I. Defining an optimal cut-point value in roc analysis: an alternative approach. Comput Math Methods Med. 2017;2017:14. Hasegawa S, Yoshikawa T, Rino Y, et al. Priority of lymph node dissection for Siewert type II/III adenocarcinoma of the esophagogastric junction. Ann Surg Oncol. 2013;20:4252-4259. Leers JM, DeMeester SR, Chan N, et al. Clinical characteristics, biologic behavior, and survival after esophagectomy are similar for adenocarcinoma of the gastroesophageal junction and the distal esophagus. J Thorac Cardiovasc Surg. 2009;138:594-602 [discussion 601-602].
GENERAL THORACIC
presented herein. Further investigation is warranted to validate our findings in an independent cohort and to clarify the ability of different neoadjuvant radiotherapy protocols to effectively sterilize the mediastinum, ideally in a prospective fashion. In addition, the findings of the present report highlight a distinction between patients who have clinical evidence of mediastinal nodal metastasis and those who do not, and the rate of clinically occult pretreatment mediastinal nodal disease that was subsequently sterilized with neoadjuvant therapy is unclear. The authors acknowledge temporal changes in positron emission tomography and endoscopic ultrasonography technology used at their institutions, particularly during the early part of the study period, and the potential effects of these changes on the accuracy of preoperative staging. Finally, although comprehensive pathologic examination of mediastinal nodes was not performed in patients who did not undergo a formal mediastinal lymphadenectomy, we applied a previously utilized methodologic approach in an effort to circumvent this constraint.13 In summary, we identified pretreatment proximal tumor extent as an independent predictor of mediastinal involvement among patients undergoing trimodality therapy for Siewert II/III adenocarcinoma who did not have preoperative evidence of mediastinal nodal metastasis. In light of the approximately linear relationship between proximal tumor growth and the observed frequency of mediastinal involvement, we propose that tumor extent above the EGJ be considered as a continuum rather than in a dichotomized fashion. Proximal extent of 1.5 cm or greater marks a distance in which significantly more cases involved the mediastinum in our analyses, but even in patients with less proximal extent there was risk. Decisions as to what cutoff represents optimum should be taken on a case by case basis. The results presented herein suggest that clinical evidence of intrathoracic nodal metastasis and proximal tumor extent should be used to guide operative planning when evaluating patients with Siewert II/III adenocarcinoma, and that tumors with limited proximal extent may be amenable to transabdominal resection irrespective of the location of their epicenter.
MITCHELL ET AL SIEWERT II/III ADENOCARCINOMA