PEWATASDB

Ciltnos

csaiaa

ag(ost

CnbsJtmrhAt

A

PS

AsET

©P

GEN

ERA

LT

HO

RA

CIC

ORIGINAL ARTICLES: GENERAL THORACIC

roposed Modification of Nodal Status in AJCCsophageal Cancer Staging Systemayne Hofstetter, MD, Arlene M. Correa, PhD, Neby Bekele, PhD, Jaffer A. Ajani, MD,

lexandria Phan, MD, Ritsuko R. Komaki, MD, Zhongxing Liao, MD, Dipen Maru, MD,sung T. Wu, MD, Reza J. Mehran, MD, David C. Rice, MD, Jack A. Roth, MD,ra A. Vaporciyan, MD, Garrett L. Walsh, MD, Ashleigh Francis, BS,handa Blackmon, MD, and Stephen G. Swisher, MDepartments of Thoracic and Cardiovascular Surgery, Gastrointestinal Medical Oncology, Radiation Oncology, Pathology, and

iostatistics, University of Texas MD Anderson Cancer Center, Houston, Texas

papnrnn3(sAy0ns

cltt

Background. The current American Joint Committee onancer (AJCC) esophageal cancer staging for nodal status

s difficult to interpret and is based solely on lymph nodeocation relative to the primary tumor’s esophageal loca-ion. Recent reports suggest that the number of lymphodes involved is also an important factor. We reviewedur esophageal experience to propose an improved nodaltaging system.

Methods. In all, 1,027 patients with resected esophagealancer from 1970 to 2005 were reviewed. Lymph nodestations were assigned according to AJCC criteria. Over-ll survival was assessed by Kaplan-Meier analysis. Thempact of location, number of involved lymph nodes,nd use of preoperative chemotherapy or radiation ther-py, or both, was assessed.Results. Nonregional nodal involvement (n � 17) was

ssociated with decreased survival compared with re-ional (n � 441) or celiac nodal (n � 73) involvement3-year: 0% versus 24% and 23%; p < 0.001). The numberf involved lymph nodes was strongly associated withurvival (3-year: 0 nodes � 63%, 1 to 3 nodes � 31%, more

han 3 nodes � 13%; p < 0.001), and multivariable Cox

[ccpnloNms

ivctgcpp

sophageal Cancer Program, 1515 Holcombe Blvd, Box 445, Houston,X 77030; e-mail: whofstetter@mdanderson.org.

2007 by The Society of Thoracic Surgeonsublished by Elsevier Inc

roportional-hazards analysis suggested that the locationnd number of involved lymph nodes were independentredictors of survival (p < 0.001). We propose a modifiedodal staging system that designates celiac nodes asegional and includes number of involved nodes: pN0,o nodes (3 years � 63%, n � 496); pN1-regional, 1 to 3odes (3 years � 32%, n � 292); pN2-regional, more thannodes (3 years � 14%, n � 222); pN3-nonregional node

3 years � 0%, n � 17 [p < 0.0001]). This modified nodaltaging system better predicts survival than the currentJCC nodal staging system in which survival for pN1 (3ears � 24%) and pM1a (3 years � 23%) do not differ (p �.67). The use of induction before surgical resection didot alter the predictive effect of the new nodal stagingystem.

Conclusions. Modification of the AJCC nodal classifi-ation system to incorporate the number of involvedymph nodes with regional and nonregional node loca-ion simplifies and better predicts long-term survivalhan does the current AJCC nodal system.

(Ann Thorac Surg 2007;84:365–75)

© 2007 by The Society of Thoracic Surgeons

arcinoma of the esophagus and gastroesophagealjunction is an aggressive disease with a poor prog-

osis [1]. However, there are groups of patients whoenefit from therapy. Esophageal cancer staging systemsuch as the one devised and revised by the Americanoint Committee on Cancer (AJCC) [2] are used to stratifyhe prognosis of patients and are therefore vital for

aking decisions regarding various treatment algo-ithms. Therefore, a staging system that poorly stratifieseterogeneous groups of patients would require revision.s is the case, multiple authors have proposed modifica-

ions to the current TNM esophageal staging system

ccepted for publication Jan 29, 2007.

resented at the Fifty-third Annual Meeting of the Southern Thoracicurgical Association, Tucson, AZ, Nov 8–11, 2006.

ddress correspondence to Dr Hofstetter, University of Texas MD Ander-on Cancer Center, Department of Thoracic and Cardiovascular Surgery,

3–11]. In particular, the nodal, or N, portion of theurrent TNM staging system is cumbersome and omitsritical prognostic information while simultaneously anderhaps unnecessarily complicating the classification ofodal disease within the M category. The number of

ymph nodes involved, a factor shown to be important inther publications [9, 12] is not addressed in the current

staging whereas lymph node involvement and visceraletastasis are found grouped within the M staging

ystem.Because of poor outcomes with surgery alone, an

ncreasing number of patients with locoregionally ad-anced esophageal cancer are treated with preoperativehemoradiation therapy and surgery [13]. The impact ofhis treatment paradigm shift on the pathologic esopha-eal cancer staging system (pTNM) has not yet beenlearly defined. Because of pathologic downstaging, theossibility exists that chemoradiation therapy treated

atients may have different survivals than patients

0003-4975/07/$32.00doi:10.1016/j.athoracsur.2007.01.067

FCp

R

12

234

45678

8

911111112

366 HOFSTETTER ET AL Ann Thorac SurgPROPOSED MODIFICATION OF NODAL STATUS 2007;84:365–75

GEN

ERA

LT

HO

RA

CIC

ig 1. American Joint Commission on Cancer (AJCC) esophageal staging, nodal designation. Used with the permission of the American Jointommittee on Cancer (AJCC), Chicago, Illinois. The original source for this material is the AJCC Cancer Staging Manual, Sixth Edition (2002)ublished by Springer, New York (available at: www.springeronline.com).

egional Lymph Node Stations for Staging Esophageal Cancer, From Front (A) and Side (B)

Supraclavicular nodes Above suprasternal notch and claviclesR Right upper paratracheal nodes Between intersection of caudal margin of innominate artery with trachea and the apex

of the lungL Left upper paratracheal nodes Between top of aortic arch and apex of the lungP Posterior mediastinal nodes Upper paraesophageal nodes, above tracheal bifurcationR Right lower paratracheal nodes Between intersection of caudal margin of innominate artery with trachea and cephalic

border of azygous veinL Left lower paratracheal nodes Between top of aortic arch and carinaAortopulmonary nodes Subaortic and para-aortic nodes lateral to the ligamentum arteriosumAnterior mediastinal nodes Anterior to ascending aorta or innominate arterySubcarinal nodes Caudal to the carina of the trachea

M Middle paraesophageal lymphnodes

From the tracheal bifurcation to the caudal margin of the inferior pulmonary vein

L Lower paraesophageal lymphnodes

From the caudal margin of the inferior pulmonary vein to the esophagogastricjunction

Pulmonary ligament nodes Within the inferior pulmonary ligament0R Right tracheobronchial nodes From cephalic border of azygous vein to origin of RUL bronchus0L Left tracheobronchial nodes Between carina and LUL bronchus5 Diaphragmatic nodes Lying on the dome of the diaphragm, and adjacent to or behind the crura6 Paracardial nodes Immediately adjacent to the gastroesophageal junction7 Left gastric nodes Along the course of the left gastric artery8 Common hepatic nodes Along the course of the common hepatic artery9 Splenic nodes Along the course of the splenic artery0 Celiac nodes At the base of the celiac artery

twrtlppwsofi

P

PTcctTawT13wn

ahsaa

PPosuttt5Rtmjrtadtwt

LLctLsicjmuuMneiecN

SCtomisOtst

T

D

AS

H

L

P

L

a

G

367Ann Thorac Surg HOFSTETTER ET AL2007;84:365–75 PROPOSED MODIFICATION OF NODAL STATUS

GEN

ERA

LT

HO

RA

CIC

reated with surgery alone, and it remains unclearhether pTNM status will predict survival after chemo-

adiation therapy. In this manuscript, we sought to assesshe impact of number of nodes and nodal location onong-term survival. With those factors in mind, we pro-ose a modified nodal staging system for the AJCCTNM esophageal cancer staging system. In our model,e assume that a modification to a staging system that

implifies the current system, better predicts long-termutcome, and appropriately predicts long-term survivalor patients treated with neoadjuvant therapy representsmprovement over the current system.

atients and Methods

atientshis study included 1,027 patients with histologicallyonfirmed invasive squamous cell carcinoma or adeno-arcinoma of the esophagus and gastroesophageal junc-ion who had a minimum of four resected lymph nodes.here were 595 patients who received neoadjuvant ther-py, and 432 were treated with surgery alone. All patientsere resected with curative intent at the University ofexas M. D. Anderson Cancer Center (MDACC) between970 and 2005. Patients who did not survive longer than0 days were not included in this analysis. Patients whoere operated on as salvage cases (recurrent cancer after

able 1. Patient Characteristics

emographics (n � 1,027)

ge (median, range) 61 (34–79)exMale 857 (83%)Female 170 (17%)istologyAdenocarcinoma 766 (75%)Squamous cell carcinoma 261 (25%)

ocationCervical/upper 54 (5%)Middle 153 (15%)Lower/GEJ 820 (80%)

athologic stage0a 136 (13%)I 127 (12%)IIA 223 (22%)IIB 118 (12%)III 324 (32%)IVA 60 (6%)IVB 39 (4%)

ymph nodes involved0 496 (49%)1–3 299 (29%)� 3 232 (22%)

No viable tumor in the esophagus after induction therapy (T0).

EJ � gastroesophageal junction.

onsurgical therapy with pathologic complete response s

nd observation) and patients with disease limited toigh-grade dysplasia were also not analyzed in thistudy. This retrospective review of patient data waspproved by the MDACC Institutional Review Board,nd the need for individual patient consent was waived.

reoperative Staging and Treatmentretreatment clinical staging involved available technol-gy at the time of diagnosis. Most recently, patients werecreened by computed tomography scans, endoscopicltrasonography, and positron emission tomography. Pa-

ients thought to have locally advanced disease werereated with neoadjuvant therapy. Preoperative chemo-herapy included three major chemotherapeutic agents:-fluorouracil-based, platinum-based, and Taxol-based.adiation treatment was given to 45 or 50.4 Gy over a 4-

o 6-week period and administered with concurrent che-otherapy. Four to 6 weeks after completion of neoad-

uvant therapy, patients underwent surgical esophagealesection. Surgical treatment (with or without inductionherapy) included Ivor Lewis (abdominal–right thoracicpproach), three-field (McKeown-type right thoracic, ab-ominal-cervical approach), or transhiatal esophagec-

omy (abdominal-cervical approach). Patients treatedith surgery alone received surgery without preopera-

ive chemotherapy or radiotherapy.

ymph Node Station Designationymph nodes were assigned a station designation ac-ording to the AJCC criteria (Fig 1). Nodal stations 2hrough 17 were considered regional (N1) lymph nodes.ymph node stations 18 and 19 and retroperitoneal nodesuch as retrocaval or retroaortic lymph nodes were des-gnated as nonregional (M1b) nodes. For esophagealancers of the distal esophagus or gastroesophagealunction, nodal station 20 (celiac node) was considered

etastatic (M1a) disease. Cases involving disease in thepper esophagus with involvement in the supraclavic-lar, station 1 lymph nodes were similarly designated as1a. For cases involving the upper or middle esophagus,

odal station 20 was designated nonregional (M1b) dis-ase. Finally, cases of middle to distal cancers with nodalnvolvement at station 1 were also considered M1b dis-ase. By definition, and for the purposes of our analysis,ases that were designated as M1a or M1b could be either0 or N1.

tatistical Analysisross-tabulation was employed on our raw data to assess

he minimum number of sampled nodes required tobtain a consistent percentage of lymph node involve-ent. Survival probability analyses were performed us-

ng the Kaplan-Meier method calculated from the date ofurgery to the date of death or most recent follow-up.perative mortality was excluded from survival analyses

o allow determination of long-term outcome rather thanhort-term morbidity. The prognostic significance of po-ential parameters was determined by univariate analy-

is. Cox proportional hazards models were fitted for

mclws

Nvswtdt

R

PTsfMmggm

gnpds

F

T

R

C

N

M

M

M

a

C

368 HOFSTETTER ET AL Ann Thorac SurgPROPOSED MODIFICATION OF NODAL STATUS 2007;84:365–75

GEN

ERA

LT

HO

RA

CIC

ultivariable analysis. Differences between groups wereonsidered statistically significant if the p values wereess than 0.05 in a two-tailed test. The statistical analysisas performed with SPSS Software for Windows (Ver-

ion 11.5.2.1; SPSS, Chicago, Illinois).The data were supplied to a second statistician (author.B.) who was blinded to the staging systems, current

ersus modified. Analyses were performed to detectuperiority (Appendix). Each of the two staging systemsas quantified for effectiveness using the log-rank trend

est. To assess if one staging system was more effective atifferentiation between stages than the other, a permu-

ation test was employed.

esults

atient and Survival Characteristicshe study population included all patients who hadquamous cell carcinoma or adenocarcinoma treatedrom 1970 to 2005 who underwent esophageal resection at

DACC. Mean potential follow-up for the group was 145onths (range, 7.2 to 442). Demographics of the sample

roup are depicted in (Table 1). As expected for thisroup of North American patients, the vast majority were

able 2. Univariate Analysis and Survival of Resected Patien

isk Factor N HR (CI)

urrent pTNMN0 496 1.00 (ref)N1 441 2.72 (2.30, 3.21)M1a 73 2.71 (2.04, 3.59)M1ba 17 6.94 (4.22, 11.41)umber of nodes involved0 496 1.00 (ref)1–3 299 2.17 (1.81, 2.61)�3 232 4.12 (3.40, 4.98)odified N stagingN0 496 1.00 (ref)N1 292 2.14 (1.78, 2.57)N2 222 4.01 (3.31, 4.87)N3 17 7.16 (4.35, 11.78)odified N staging, surgery onlyN0 174 1.00 (ref)N1 131 2.85 (2.13, 3.83)N2 116 5.67 (4.20, 7.66)N3 11 10.26 (5.38, 19.58)odified N staging, induction

therapy � surgeryN0 322 1.00 (ref)N1 161 1.82 (1.43, 2.32)N2 106 3.18 (2.45, 4.13)N3 6 5.37 (2.37, 12.20)

M1b by nonregional nodal disease only.

I � confidence interval; HR � hazard ratio; MS � median survi

ale with adenocarcinoma of the distal esophagus or J

astroesophageal junction. Pathologic downstaging aftereoadjuvant therapy resulted in a significant number ofatients achieving a pTNM stage 0. Unsuspected M1bisease discovered at the time of surgery resulted in atage IVb designation in 39 patients; 17 of these cases

ig 2. Kaplan-Meier survival, patients grouped by current American

cording to Nodal Disease

S (mos) 3-Year Survival 5-Year Survival p Value

72.4 63% 52% � 0.00116.7 24% 14%13.9 23% 19%9.2 0% 0%

72.4 63% 52% � 0.00120.4 31% 20%11.4 13% 7%

72.4 63% 52% � 0.00120.4 32% 20%12 14% 7%9.1 0% 0%

92.3 73% 61% � 0.00119.6 30% 19%11.1 13% 6%9.1 0% 0%

52.6 58% 46% � 0.00123 33% 21%12.5 15% 8%8.4 0% 0%

ts Ac

M

val.

oint Commission on Cancer (AJCC) nodal staging.

woprrth

ogvn2napirosMio2asib

vn3bfsT3

Kci

pltimst

Fi

TS

C

R

369Ann Thorac Surg HOFSTETTER ET AL2007;84:365–75 PROPOSED MODIFICATION OF NODAL STATUS

GEN

ERA

LT

HO

RA

CIC

ere due to nonregional nodal metastasis. The majorityf patients (66%, 665 of 1,027) were stage II to III. Mostatients (56%, 571 of 1,024) were resected by transtho-acic approach, 24% (241 of 1,027) had a transhiatalesection, and 13% (132 of 1,027) had three-field resec-ions. The remaining 7% were minimally invasive (1%) orad missing data unaccounted for.We looked at the significance of lymph node location

n long-term survival in our cohort of patients as cate-orized within the current pTNM staging system. Sur-ival at 3 years for patients with N0, N1, M1a, or M1b (byonregional nodal involvement) disease was 64%, 24%,3%, and 0%, respectively. Univariate analysis shows thatonregional lymph node metastasis (n � 17) was associ-ted with a significantly decreased overall survival com-ared with regional (n � 441) or celiac (n � 73) nodal

nvolvement (p � 0.001; Table 2). However, the hazardatios and confidence intervals of the N1 and M1a groupsverlap completely (Table 2). Whereas the Kaplan-Meierurvival curves (Fig 2) were distinct for nodal stages N0 and

1b, the N1 and M1a survival curves cross and are virtuallynterchangeable. This is reflected in the 3- and 5-yearverall survivals for the N1 and M1a groups of 24% and3%, and 14% and 19%, respectively. When the data arenalyzed in subgroups according to patients who receivedurgery only or induction therapy, the graphs are nearlydentical to that of the whole cohort, and the overlapetween the N1 and M1a groups persists (data not shown).To assess the impact of number of lymph nodes in-

olved on long-term survival, metastasis to any lymphode was separated into the following categories: 0, 1 to, and more than 3 involved nodes. These groupings wereased on previous publications [4, 5, 7, 8, 11, 12]. We

ound that the number of lymph nodes involved wastrongly associated with long-term survival (p � 0.001;able 2). Three-year survival for 0, 1 to 3, and more than

ig 3. Kaplan-Meier survival, patients grouped by number ofnvolved lymph nodes.

lymph nodes was 63%, 31%, and 13%, respectively. G

aplan-Meier survival curves show ordered, distincturves when analyzing survival by number of nodesnvolved (Fig 3).

Multivariable Cox proportional-hazards analysis waserformed including T stage, M stage, and number of

ymph nodes involved. All three are independent predic-ive factors of long-term survival (p � 0.001). Owing tossues of colinearity, a separate proportional-hazards

odel was constructed including nodal location and Ttage, showing that both factors are independent predic-ors of survival (p � 0.001).

able 3. Current and Proposed Modified AJCC Esophagealtaging

urrent American Joint Committee on Cancer staging system(pTNM)

Primary tumor (T)Tx Primary tumor cannot be assessedT0 No evidence of primary tumorTis Carcinoma in situT1 Tumor invades lamina propria or submucosaT2 Tumor invades muscularis propriaT3 Tumor invades adventitiaT4 Tumor invades adjacent structures

Regional lymph nodes (N)Nx Regional nodes cannot be assessedN0 No regional node metastasisN1 Regional node metastasis

Distant metastasis (M)Mx Presence of distant metastasis cannot be assessedM0 No distant metastasisM1a Celiac node metastasis for tumors of the lower

esophagus/GEJ, supraclavicular node metastasis fortumors of the upper esophagus

M1b Nonregional nodal metastasis or distant metastasisevised staging system (pTNM)Primary tumor (T)

Tx Primary tumor cannot be assessedT0 No evidence of primary tumorTis Carcinoma in situT1 Tumor invades lamina propria or submucosaT2 Tumor invades muscularis propriaT3 Tumor invades adventitiaT4 Tumor invades adjacent structures

Regional lymph nodes (N)Nx Regional nodes cannot be assessedN0 0 regional node metastasisN1 1–3 regional lymph node metastases, including nodes

previously labeled as M1aN2 � 3 regional lymph node metastases, including nodes

previously labeled as M1aN3 Nonregional lymph node metastasis

Distant metastasis (M)Mx Presence of metastasis cannot be assessedM0 No distant metastasisM1 Distant metastasis

EJ � gastroesophageal junction.

MSBAPplnngailctN(adKdscs

mmicnlm

CSTsieesaof(

C

TbMsi

4Fmsd

370 HOFSTETTER ET AL Ann Thorac SurgPROPOSED MODIFICATION OF NODAL STATUS 2007;84:365–75

GEN

ERA

LT

HO

RA

CIC

odified Nodal-Status for pTNM Esophageal Cancertaging Systemased on the above findings, modifications to the currentJCC nodal staging system were derived (Table 3).atients were stratified into four different nodal groups:N0, no nodal metastasis; pN1, one to three regional

ymph node metastasis, inclusive of previously desig-ated M1a nodes; pN2, more than three regional lymphode metastasis, inclusive of M1a nodes; pN3, nonre-ional nodal involvement (previously M1b). Univariatend multivariable analysis were performed to assess thempact of the modified staging on the prognosis ofong-term survival. Kaplan-Meier survival curves wereonstructed using this proposed staging modification tohe nodal status. With the patients grouped by modified

status, 3-year survival was significantly differentiatedp � 0.001) between all of the groups at 63%, 32%, 14%,nd 0%, respectively (Table 2). Hazard ratios and confi-ence intervals were distinct between the groups, andaplan-Meier analysis of long-term survival shows or-ered, distinct survival curves (Fig 4A). Subgroup analy-is of this modified grouping using the surgery-onlyohort or induction-therapy cohort yielded similar re-ults (Table 2, Fig 4B, C).

Multivariable analysis by Cox proportional hazardsodel was performed to investigate the contribution of theodified nodal status to long-term survival (Table 4). We

ncluded in the model the modified nodal groups and theurrent pT and pM factors and found that the modifiedodal system was a strong independent predictive factor for

ong-term survival (p � 0.001). The pT and pM factors re-ain significant independent prognostic elements as well.

omparison of Modified to Current Nodal Stagingystemshe data were supplied to a statistician blinded to whichtaging group was the current versus the proposed mod-fications with the queries of (1) are the staging systemsffective (do they maintain an ordered differentiation andffectively differentiate the stage groups); and (2) whichystem is the more effective. Using log-rank trend testnd permutation analysis, both systems were found to berdered and effective; however, the modified system wasound to be significantly better than the existing systemp � 0.005; see Appendix).

omment

he current AJCC esophageal cancer staging system isased on a T (tumor depth), N (regional nodal statues), and

(nonregional nodes or systemic metastases) staging clas-ification [2]. Although the staging system currently in places good at predicting long-term survival, as indicated by our

™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™ig 4. (A) Kaplan-Meier survival, patients grouped by proposedodification to the American Joint Commission on Cancer nodal

taging. (B) Analysis of surgery only subgroup. (C) Analysis of in-

uction therapy subgroup.

ddcRcdenniaMie

nljaerdcpoaJWrmrKtm

enftviiknimtg

atsgfsamcppadasmt

Mrto3psn5sspllntMtcblratpsb

TN

R

p

p

M

a

C

371Ann Thorac Surg HOFSTETTER ET AL2007;84:365–75 PROPOSED MODIFICATION OF NODAL STATUS

GEN

ERA

LT

HO

RA

CIC

ata, modifications to the system results in more orderedifferentiation of stages and eliminates some of the compli-ated features of the N and M portions of the TNM system.evision of the esophageal staging system is a process ofontinuous evaluation. More data, changing histology, andifferent treatment algorithms all affect the validity of anxisting system. The current proposal to modification isovel as it expands on previous appeals to include theumber of positive lymph nodes and also effectively elim-

nates the cumbersome M1a classification while includingll lymph node disease within the N category, reserving the

category for visceral metastatic disease. Furthermore, itncorporates patients treated with induction therapy andffectively predicts survival in that subgroup of patients.

Our data confirm the findings of other authors that theumber of involved nodes is an independent predictor of

ong-term survival in esophageal and gastroesophagealunction tumors. In separate publications, Hagen andssociates [9] and Lerut and associates [12] showed thatn bloc resection resulted in a significant number ofesected lymph nodes and that prognosis significantlyeclined as the number of involved lymph nodes in-reases. We modeled the numbered cut-offs for the ourroposed modifications to be similar to those used inther publications by Korst and associates [4] and Ricend associates [5], and they are also similar to theapanese nodal staging system for esophageal cancer [8].

e found that when our cohort of patients were catego-ized by number of involved lymph nodes that there is aonotone relationship to the designations, the hazard

atios and confidence intervals are distinct, and theaplan Meier survival curves are ordered and differen-

iated. These are criteria that an effective staging system

able 4. Multivariable Cox Regression Analysis of Modifiedodal Stage and pTNM Factors

isk Factor Frequency HR (CI) p Valuea

T � 0.001T0 (reference) 135 1.00 (ref)Tis 23 0.88 (0.45, 1.74) 0.72T1 165 0.87 (0.60, 1.25) 0.46T2 161 1.35 (0.96, 1.90) 0.09T3 523 2.18 (1.61, 2.96) 0.000T4 20 2.23 (1.26, 3.94) 0.006

M � 0.001M0 (reference) 928 1.00 (ref)M1a 60 1.04 (0.77, 1.40) 0.80M1b 39 4.19 (2.70, 6.49) 0.000

odified N stage � 0.001pN0 (reference) 496 1.00 (ref)pN1 292 1.57 (1.29, 1.92) 0.000pN2 222 2.56 (2.05, 3.18) 0.000pN3 17 1.27 (0.67, 2.41) 0.468

p value � 0.05 accepted as significant.

I � confidence interval; HR � hazard ratio.

ust meet. With these factors applied to modify the i

xisting staging system, we found that our proposedodal staging system was an independent prognostic

actor on long-term survival. Furthermore, comparison ofhe current nodal staging to the proposed system re-ealed the modified system to be a statistically significantmprovement over the current system. Whether threenvolved nodes is the correct inflection point is notnown. However, we agree that determining the optimalumber of positive nodes to stratify the staging system is

mportant and is a task that should be completed using auch greater sample size than is available to us here, as

hat would include more patients within each stagingroup.An increasing number of patients with locoregionally

dvanced esophageal cancer (stage II to IVA) are beingreated with chemotherapy or chemoradiotherapy beforeurgery because of poor long-term outcomes with sur-ery alone [1]. The impact of pathologic downstagingrom this preoperative treatment on the pTNM stagingystem has not been fully evaluated. This manuscriptttempts to address the important question of whether aodified pTNM staging is accurate after preoperative

hemoradiation. As our results would suggest, the pro-osed modified nodal staging system is appropriate foratients treated with induction therapy. Kaplan-Meiernalysis shows survival curves to be differentiated andistinct, very similar to those patients treated by surgerylone. When analyzing the cohort as a whole, includingurgery-only and induction-therapy patients together,ultivariable analysis shows the modified nodal system

o be a strong predictor of long-term survival.The proposed modifications eliminate the need for an1a classification and would designate level 20 nodes as

egional for lower esophageal or gastroesophageal junc-ion tumors. Analysis of our data shows similar survivalf patients with M1a and N1 disease (24% versus 23%-year survival). This finding agrees with previouslyublished data by Hagen and associates [12] who reportimilar survival for patients with celiac or regional lymphode disease after en-bloc esophagectomy (actuarial-year survival 28% versus 37%). Both of these manu-cripts, ours and that of Hagen and associates [12], woulduggest that intermediate survival can be achieved inatients with M1a disease similar to that of regional

ymph node metastasis, and that the number of involvedymph nodes is a stronger predictor of survival than M1aodal location. However, these findings disagree with

hose of Rice and associates [5] in which the survival of1a patients is described as equivalent to M1b, and

herefore their recommendation is to eliminate M1aategory and reclassify these patients as M1b. That maye explained by difficulty in the classification of an M1a

ymph node, especially in the celiac position whichepresents all of the M1a disease in our data. Malaisriend associates [14] reported data from our own institu-ion describing an intermediate to poor prognosis foratients identified as having celiac adenopathy by endo-copic ultrasonography. Similar findings were reportedy Eloubeidi and associates [15] in another single-

nstitution experience with endoscopic ultrasonography

incnstsals

sdamtofsMsnftacwsaTrdb

cltt

SaF

R

1

1

1

1

1

1

1

1

1

1

2

2

2

2

372 HOFSTETTER ET AL Ann Thorac SurgPROPOSED MODIFICATION OF NODAL STATUS 2007;84:365–75

GEN

ERA

LT

HO

RA

CIC

dentified celiac adenopathy that resulted in poor prog-osis. Given a lack of consensus on how to define theeliac node, it is possible that the cases that have a celiacode defined by the surgeon could differ slightly butignificantly in position and prognosis compared withhose whose disease is identified by endoscopic ultra-onography. The conflicting data underscore the need tochieve a consensus on the designation of the celiacymph node and eliminate the M1a distinction within thetaging system.

There are several notable weaknesses to this manu-cript. The current AJCC system accounts for tumorepth, nodal disease, and metastatic disease but does notccount for other tumor features such as length, micro-etastasis, molecular markers, or various histologic fea-

ures. Our proposed modifications pertain to nodal statusnly, and as is true for any staging system, numerousactors must be analyzed to create an entire stagingystem. Interactions between other factors such as T and

need to be taken into consideration. Our data set is toomall to refine the entire staging system; there simply areot enough patients within each stage subgroup to allow

or adequate evaluation. This will require evaluation ofhe T and M and minor classifications (ie, histology)nd the use of a larger data set obtained from multipleollaborating institutions, and we urge participationith cooperations currently under way. The relevant

trength of the paper lies within the independentnd blinded evaluation of our data by a statistician.hrough log-rank trend test and permutation test, theevised staging system was found to have better or-ered differentiation and more effective differentiationetween stages.In conclusion, modification of the AJCC nodal classifi-

ation system to incorporate the number of involvedymph nodes with regional and nonregional node loca-ion simplifies and better predicts long-term survivalhan does the current AJCC nodal system.

upported by generous grants from the George Sweeney Esoph-geal Research Fund and the Carl and Ginnie Edwards Researchund.

eferences

1. Hofstetter W, Swisher SG, Correa AM, et al. Treatmentoutcomes of resected esophageal cancer. Ann Surg 2002;236:376–85.

2. American Joint Committee on Cancer. AJCC cancer stagingmanual. 6th ed. Philadelphia: Lippincott-Raven, 2002:91–8.

3. Steup WH, De Leyn P, Deneffe G, et al. Tumors of theesophagogastric junction. Long term survival in relation tothe pattern of lymph node metastasis and a critical analysisof the accuracy or inaccuracy of pTNM classification. J Tho-rac Cardiovasc Surg 1996;111:85–95.

4. Korst RJ, Rusch VW, Venkatraman E, et al. Proposed revi-sion of the staging classification for esophageal cancer.

J Thorac Cardiovasc Surg 1998;115:660–70.

5. Rice TW, Blackstone EH, Rybicki LA, et al. Refining esoph-ageal cancer staging. J Thorac Cardiovasc Surg 2003;125:1103–13.

6. de Manzoni G, Pedrazzani C, Verlato G, et al. Comparison ofold and new TNM systems for nodal staging in adenocarci-noma of the gastro-oesophageal junction. Br J Surg 2004;91:296–303.

7. Ellis FH, Heatley GJ, Krasna MJ, Williamson WA, Balogh K.Esophagogastrectomy for carcinoma of the esophagus andcardia: a comparison of findings and results after standardresection in three consecutive eight-year intervals with im-proved staging criteria. J Thorac Cardiovasc Surg 1997;113:836–48.

8. Kunisaki C, Akiyama H, Nomura M, et al. Developing anappropriate staging system for esophageal carcinoma. J AmColl Surg 2005;884–90.

9. Lerut T, Nafteux P, Moons J, et al. Three-field lymphadenec-tomy for carcinoma of the esophagus and gastroesophagealjunction in 174 Ro resections: impact on staging, disease-freesurvival, and outcome. A plea for adaptation of TNM clas-sification in upper-half esophageal carcinoma. Ann Surg2004;240:962–74.

0. Swisher SG, Hofstetter WL, Wu T-T, et al. Proposed revisionof the esophageal cancer staging system to accommodatepathologic response (pP. following preoperative chemora-diation (CRT). Ann Surg 2005;241:810–20.

1. Roder JD, Busch R, Stein HJ, Fink U, Siewert JR. Ratio ofinvaded to removed lymph nodes as a predictor of survivalin squamous cell carcinoma of the oesophagus. Br J Surg1994;81:410–3.

2. Hagen JA, DeMeester SR, Peters JH, Chandrasoma P, De-Meester TR. Curative resection for esophageal adenocarci-noma. Analysis of 100 en bloc esophagectomies. Ann Surg2001;234:520–31.

3. Chirieac LR, Swisher SG, Ajani JA, et al. Post-therapypathologic stage predicts survival in patients with esopha-geal carcinoma receiving preoperative chemoradiation. Can-cer 2005;103:1347–55.

4. Malaisrie SC, Hofstetter WL, Correa AM, et al. Endoscopicultrasonography-identified celiac adenopathy remains apoor prognostic factor despite preoperative chemoradio-therapy in esophageal adenocarcinoma. J Thorac CardiovascSurg 2006;131:65–72.

5. Eloubeidi MA, Wallace MB, Hoffman BJ, et al. Predictors ofsurvival for esophageal cancer patients with and withoutceliac axis lymphadenopathy: impact of staging endosonog-raphy. Ann Thorac Surg 2001;72:212–20.

6. Vauthey J-N, Lauwers GY, Esnaola NF, et al. Simplifiedstaging for hepatocellular carcinoma. J Clin Oncol 2002;20:1527–36.

7. Smith DD, Schwarz RR, Schwarz RE. Impact of total lymphnode count on staging and survival after gastrectomy forgastric cancer: data from a large US-population database.J Clin Oncol 2005;23:7114–24.

8. Balch CM, Soong SJ, Gershenwald JE, et al. Prognosticfactors analysis of 17,600 melanoma patients: validation ofthe American Joint Committee on Cancer melanoma stagingsystem. J Clin Oncol 2001;19:3622–34.

9. Bouvier A-M, Haas O, Piard F, Roignot P, Bonithon-Kopp C,Faivre J. How many nodes must be examined to accuratelystage gastric carcinomas? Cancer 2002;94:2862–6.

0. Lagarde SM, ten Kate FJW, Reitsma JB, Busch ORC, vanLancschot JJB. Prognostic Oncol 2006;24:4347–55.

1. Efron B, Tibshirani R. An introduction to the bootstrap.London: Chapman & Hall, 1993.

2. Klein JP, Moeschberger ML. Survival analysis: techniquesfor censored and truncated data. Berlin: Springer-Verlag,1997.

3. Cantor A. Extending SAS survival analysis techniques for

medical research. Cary, NC: SAS Institute, 1997.

A

S(

TsAfitsec

s(t(dsccamcgogsmbmt(rt[stcw

sabasbtdphcrt(ddhtduc

a

rNhsnsdtctstssessdp

(

(

(

sdcblii

A1

2

FT

373Ann Thorac Surg HOFSTETTER ET AL2007;84:365–75 PROPOSED MODIFICATION OF NODAL STATUS

GEN

ERA

LT

HO

RA

CIC

ppendix

tatistical Comparison of AJCC to Modified N-StagingNeby Bekele, PhD)

he main purpose of this discussion is to provide a formaltatistical test comparing two staging systems (the standardJCC nodal staging system for esophageal cancer and a modi-ed AJCC nodal staging system). Before we can compare these

wo staging systems, we must first define what makes a stagingystem effective. We also must define how to quantify thisffectiveness, and lastly we must have a way of statisticallyomparing the effectiveness of the two staging systems.

The characteristic that defines the effectiveness of any stagingystem is its ability to differentiate between low-stage patientsthose patients who survive a long time), mid-stage patients (pa-ients who moderate amount of time, and high-stage patientspatients who survive a relatively short amount of time). Thisifferentiation is quantified graphically by a monotone relation-hip between stage of disease and area-under-the survivalurve. That is, a good staging system will result in patientsategorized has having low-stage disease to also have the mostrea under the survival curve; patients categorized as havingid-stage disease will have less area under the survival curve

ompared with the low-stage patients; and lastly, patients cate-orized as having high-stage disease will have the least amountf area-under-the-survival-curve relative to the other tworoups. Moreover, this type of monotone relationship betweentage and area-under-the-survival-curve may be quantified nu-erically by a log-rank trend test [1]. This statistic is used to

ecause it quantifies the effectiveness of a staging system: Theore effective a staging system, the larger will be the value of

he log-rank trend test. Any test statistic that increases only asordered) differentiation between stages increases (as the log-ank trend test does) would be adequate for our purposes. Ashe log-rank trend test is available and can easily be calculated2], we chose this statistic. Other test statistics such as thetandard log-rank test would not be appropriate because thisest statistic increases for nonordered differences in survivalurves (which for assessing effectiveness of a staging systemould not be appropriate).Once we have quantified the effectiveness of each of the two

taging systems using the log-rank trend test, we need to alsossess if one staging system is more effective at differentiatingetween stages than the other. The complicating factor inssessing the difference in effectiveness of the two stagingystems is that the same set of patients are categorized under theoth systems, inducing correlation between the two log-rank

rend statistics. We address this complicating factor by assessingifferences in the effectiveness of the two staging systems by aermutation (randomization) test. A permutation test is a type ofypothesis test in which the null distribution is obtained byalculating possible values of the test statistic under randomearrangements (ie, permutations) of the original labels assignedo the observed data [3]. By repeating this process many timeseg, 10,000 times), a null distribution is created. Creation of nullistributions in this way only differ from null distributionserived from statistical theory (eg, standard normal, �2, F) inow the null distribution is obtained but do not differ in how

hey are used or interpreted. An added benefit of using nullistributions derived from permutation tests is that they can besed in situations in which the null distribution is difficult toonstruct analytically (as in this case).

Each of the staging systems under consideration in this

nalysis has four staging categories. We call these four catego-

3

ies N0, N1, N2, and N3 in the modified staging system and N0,1, M1a, and M1b in the current AJCC system. Under the nullypothesis for our permutation test, we assume that the twotaging systems are exchangeable. This means that under theull hypothesis we assume that there is no difference in the twotaging systems. The alternative hypothesis is that there isifference in the ability of the two staging systems to differen-

iate among these four stages. To assess these hypotheses, weonstruct our test statistic, which is the difference in the tworend tests calculated under each staging system. For the ob-erved data, the difference in the two staging system log-rankrend test statistics is �49.0758 (177.21 for the standard AJCCtaging system and 226.287 for the modified AJCC stagingystem). These log-rank trend tests tell us that both methods areffective staging systems, but it appears that the modified AJCCtaging system is better. To assess whether the modified AJCCtaging system is in fact statistically better, we construct our nullistribution for which this observed test statistic will be com-ared by performing the following steps:

1) For each patient with 50% probability, randomly rearrange(ie, permute) the staging system labels originally assignedto that patient. That is, for a given patient, the stageassigned under the standard staging system is switched,and the stage assignment is now considered to have beenassigned under the modified system and vice versa.

2) Once all patients have been permuted, we calculate thelog-rank trend test statistic for the two staging systems andrecord the permuted difference in log-rank trend tests.

3) We repeat steps one and two 10,000 times.

The null distribution we construct using the method de-cribed above is given in Appendix Figure 1. As shown, theifferences in the trend statistics under the null distribution isentered around 0, as one would expect if there is no differenceetween staging systems in their ability to differentiate between

ow-, middle-, and high-stage patients. Moreover, the probabil-ty of observing a differences in trend statistics as rare as �49.08s only 0.005 under the null hypothesis.

ppendix References. Klein JP, Moeschberger ML. Survival analysis: techniques for

censored and truncated data. Berlin: Springer-Verlag, 1997.. Cantor A. Extending SAS survival analysis techniques for

medical research. Cary, NC: SAS Institute, 1997.

ig A1. Distribution of differences under the null distribution. (Obs.est Stat. � observed test statistic.)

. Efron B, Tibshirani R. An introduction to the bootstrap.London: Chapman & Hall, 1993.

D

DbsHb

iemf

tand

Tlbwt

Dtas

dranahfwilfimdtntvrntwt

acetbdi

popa

374 HOFSTETTER ET AL Ann Thorac SurgPROPOSED MODIFICATION OF NODAL STATUS 2007;84:365–75

GEN

ERA

LT

HO

RA

CIC

ISCUSSION

twsTdddon

DWOfictttah

appnanshig

DqwwabIms

pMoasafi

dttspptts

R PAUL H. SCHIPPER (Portland, OR): Recently there haseen an upswelling of discontent with the esophageal cancertaging system, both in T and N staging, and I congratulate Drofstetter and his colleagues for tackling this, and it is a largeody of work and thank them for doing that.In light of your data, I would say that N staging is at least an

mportant prognostic indicator and more so than just the pres-nce or absence of disease in an old bed. More importantly,aybe how many nodes are involved? So I have two questions

or you.First question. What do you feel is an adequate lymphadenec-

omy? Should we be removing everything between certainnatomic borders? Should we be aiming for a certain number ofodes removed? What should that number be? Should we beissecting out the celiac nodes?Second question. Is there a chance that patients with a lesserstage in your study had fewer nodes resected because of the

ength of time of this study and because it was retrospective, andecause of fewer nodes being resected, that some nodes thatere positive were not identified? In other words, how consis-

ent do you think your denominator of lymph nodes resected is?

R HOFSTETTER: I would like to thank Dr Schipper for takinghe time to review our paper. Contributions from our reviewersnd questions from the audience are appreciated and willtrengthen our publication.

Regarding the first question, what is an adequate lymph nodeissection, our data really does not address that question di-ectly and we did not statistically look at what would be andequate lymph node dissection. We did set a cut-off for theumber of lymph nodes resected in the patients that we lookedt. In other words, we wouldn’t use patients in this analysis whoad less than four lymph nodes evaluated. The cut was set at

our because when we looked at our data, we found that patientsho had very few lymph nodes resected had a very low

ncidence of lymph node involvement, perhaps inappropriatelyow. At four or greater resected lymph nodes, the incidence ofnding positive nodes remained consistent; meaning that thereay be a minimum number of nodes that need to be resected to

ecrease a false-negative result, and if you remove less than, say,hree, four lymph nodes that you are not going to find lymphode involvement that is there. There have been other publica-

ions that allude to a minimum number of lymph nodes in-olved. In our situation, over 60% of our patients had transtho-acic dissections, and our median number of reported lymphodes in the specimen was 10. So I would say that the extent of

he lymph node dissection is important. If you get to the pointhere you are not dissecting lymph nodes out, you are not going

o detect whether the patient has lymph node involvement.Should we be dissecting the celiac lymph node out? I think

bsolutely. My lymph node dissection comes down through theeliac axis. I clean out the area through the porta hepatis, andxtend this left-lateral to the hilum of the spleen. Do I know thathat is going to make a difference in long-term survival? I do not,ut I believe that it makes a difference in prognosis and there isata to suggest that patients that have limited lymph node

nvolvement do benefit from an extended lymphadenectomy.The second question, does a lesser T stage portend a different

rognosis and does that mean that we were doing a lesserperation in terms of a lymphadenectomy? Our practice at thisoint is that any patient we see who has local regionally

dvanced disease undergoes an extensive lymph node dissec- c

ion, and that is independent of T stage. Evidence to that is thate know that patients who have T1b disease in a modified

taging system have upwards of 35 to 50% of nodal involvement.herefore, we are still pushing for an extended lymph nodeissection. The only situation where a formal lymph nodeissection might be avoided is in a patient with high-gradeysplasia or intramucosal carcinoma. However, for the purposesf this study, we limited our data-set to patients who had lymphodes removed.

R MARK J. KRASNA (Towson, MD): I want to congratulateayne and the group from Anderson on an excellent review.ne quick historical comment, Wayne. This idea was proposedrst by Dave Skinner in 1988 with his so-called WNM classifi-ation. My mentor, Bunkey Ellis, then rekindled this in a paperhat we wrote in 1990 that suggested that lymph node prognos-ication was the most important independent prognosticator andhat using the number of nodes should be incorporated. Ipplaud you because it wasn’t included in that last AJCC andopefully we will get them to change it the next time.I will just ask one specific question. Although I enjoyed your

nalysis, I would suggest that another separate analysis, anderhaps it was done in the manuscript, be done excluding all theatients who received neoadjuvant therapy. Although yourumbers are very high, therefore it is an excellent database tonalyze, you actually may be underestimating the degree ofodal extent and therefore underestimating the impact onurvival since your neoadjuvant patients may have appeared toave fewer nodes involved at the time of resection. So I wonder

f you would comment on whether you have analyzed the overallroup without the neoadjuvant patients.

R HOFSTETTER: Thanks, Dr Krasna, for your comments anduestions. First of all, thank you for reviewing the historical data,hich I didn’t have time for in our presentation. Nonetheless,e all recognize your important contributions to the field and

ppreciate your comments. The concept of including the num-er of nodes involved into the staging system is clearly not new.t just hasn’t been picked up in esophageal staging, even in the

ost recent AJCC revision in 2002. Many of us are interested ineeing this incorporated into a new revision.

We did perform a subgroup analysis including only theatients who did not receive neoadjuvant therapy. The Kaplan-eier survival curves of the patients who underwent surgery

nly versus induction therapy are included in the manuscriptnd these show that the relationship of nodal involvement tourvival was consistent, whether or not the patient had surgerylone or neoadjuvant therapy. We believe that these proposalsor modification of the nodal staging system would include bothnduction patients and surgery-alone patients.

Now, this data doesn’t specifically address the importance ofownstaging in neoadjuvant therapy patients, but we do think

hat it is appropriate that downstaged patients be categorized inhe same staging system as patients who have undergoneurgery alone. Separate data from our group has shown thatatients who are downstaged perform to the level of theirost-treatment stage, and this manuscript lends some support to

his as well. However, this is not the focus of our manuscript. Soo answer your question specifically, yes, we did evaluate it inurgery-only patients and the relationship seems to be

onsistent.

Dpihcostfict

DqbsrodUwcciwmtstt

Dfctoaodsbtfwc

Dtadttrfoadsdd

375Ann Thorac Surg HOFSTETTER ET AL2007;84:365–75 PROPOSED MODIFICATION OF NODAL STATUS

GEN

ERA

LT

HO

RA

CIC

R CAROLYN E. REED (Charleston, SC): Wayne, a greatresentation again from the MD Anderson group. I am dissat-

sfied, as many of us are, with the present staging system, but weave to be careful that the staging system needs to be usedlinically as well as pathologically, because we need to talk tour patients clinically ahead of time, we need to use the stagingystem to pick the appropriate therapy, as well as to talk abouthe prognosis. So my question is really how do you use what youound, what you have discovered clinically so that we canncorporate it into a new staging system? How is this going tohange how you would look at a patient preoperatively ratherhan after the fact? Thank you.

R HOFSTETTER: Dr Reed, thank you for your comments anduestions. Specifically we did deal with pathologic stagingecause we wanted to get a start on revising the nodal stagingystem and the overall staging system. This is just a first step inevision of the system. I agree that the staging system shouldffer prognostic information that is available to clinicians toiscuss with their patients prior to making treatment decisions.ltimately as the entire staging system comes under review, weould like to be able to take that information with us into the

linics. In the manuscript, we discuss the importance of EUS inlinical staging of these patients and how we go on to use thisnformation in terms of offering neoadjuvant therapy. Patientsho appear clinically to have minimal to no lymph node involve-ent may be candidates for a less aggressive approach to induction

herapy. However, based on the prognoses seen in this modifiedtaging system, patients who on clinical staging have greaterhan three lymph nodes involved may be more appropriately

reated with an aggressive approach to induction therapy. w

R MARK B. ORRINGER (Ann Arbor, MI): My complimentsor a nice paper that highlights some of the problems with oururrent system. Wayne, I would like to ask you if you don’t thinkhat part of the reason that your M1a and M1b survival groupsverlap is that all M1a patients aren’t the same? And isn’t theredifference when an M1a, a celiac node, is involved right at the

rigin of the left gastric artery from the celiac trunk and can beivided flush with the celiac axis and resected en bloc with apecimen versus an independent metastasis, still a celiac node,ut going up the portal system heading toward the liver along

he hepatic? And so I would ask you if there is not a need tourther subdivide what a celiac metastasis is, because that ishere I think the big shortcoming of the current system is. All

eliac metastases are not the same.

R HOFSTETTER: Dr Orringer, thank you for that comment. Ihink you have hit on the crux of the discussion for M1a disease,nd that is the distinction of how a celiac lymph node isistinguished by the surgeon and other clinicians involved in

he work-up of esophageal cancer patients. To my knowledge,here is no standard agreement within the staging systemegarding the definition as to what a celiac lymph node is. Inact, from data that we have previously presented, EUS detectionf celiac disease portends an intermediate to poor prognosis,nd this may be different from outcomes seen when the celiacisease was defined by the surgeon. The lymph node that theurgeon calls a celiac lymph node may be different than thatescribed by the endoscopist. So there is a need to reallyecipher and pin down what a celiac lymph node is. Perhaps

ith our collaborative group we can address this question.