The Recurrence Pattern of Esophageal Carcinomaafter Transhiatal Resection
Jan BF Hulscher, MD, Johanna W van Sandick, MD, Jan GP Tijssen, PhD, Hugo Obertop, MD,J Jan B van Lanschot, MD
Background: There is much controversy about the op-timal resection for carcinoma of the esophagus. Little isknown about the pattern of recurrence after transhiatalresection for esophageal carcinoma.
Study Design: We retrospectively reviewed the chartsof 149 patients who underwent transhiatal esophagec-tomy for carcinoma of the mid or distal esophagus orgastroesophageal junction between June 1993 andJune 1997. Recurrence was classified as locoregional ordistant recurrence. Nine patients with macroscopicallyevident tumor left after resection and three patients(2.0%) who died in the hospital were excluded fromthe analysis. This left 137 patients; 105 men and 32women with a median age 65 years (range 37 to 84years).
Results: There were 95 adenocarcinomas (69.3%) and42 squamous cell carcinomas (30.7%). Overall the me-dian followup was 24.0 months (range 1.4 to 69.2months). For patients alive at the end of followup with-out recurrence, the median followup was 36.5 months(range 23.6 to 69.2 months). Seven patients died ofother causes. The median interval between operationand recurrence was 11 months (range 1.4 to 62.5months) for patients who had recurrence, with no sig-nificant difference in interval between locoregionaland systemic recurrence. Seventy-two of the 137 pa-tients (52.6%) developed recurrent disease. Thirty-two patients (23.4%) developed locoregional recur-rence only, 21 patients (15.3%) developed systemicrecurrence only, and 19 patients (13.9%) had a combi-nation of both. In only 8.0% of all patients was thererecurrence in the cervical lymph nodes. The most fre-
quent sites of distant recurrence were liver (37.5%),bone (25.0%), and lung (17.5%). Recurrence was re-lated to postoperative lymph node status (p
lymph node dissection for survival have never beenproved.1-5
We performed a retrospective analysis of thepattern of recurrence after transhiatal esophagec-tomy. Insight into the recurrence pattern might of-fer an indication of the possible advantages of amore extensive locoregional procedure.
We performed a retrospective chart review of allpatients who underwent a transhiatal resection foradeno- or squamous cell carcinoma of the mid ordistal esophagus or gastroesophageal junction be-tween June 1993 and June 1997. Patients with tu-mors of the esophagus proximal to the carina werepreferentially operated transthoracically, so theywere not included. Followup ended July 1999, en-suring a followup of at least 2 years for survivingpatients.
All patients were operated on with curative in-tent, that is, in the absence of distant metastasessigns of local irresectability, or both. Preoperativeanalysis included esophagogastroscopy with histo-logic biopsy, endosonography, external ultrasonog-raphy of the abdomen and neck, indirect laryngos-copy, sometimes combined with CT of the chestand abdomen. In selected patients, bronchoscopywas performed to rule out involvement of the tra-chea or bronchi. Staging was done according to thepathologic tumor node metastasis (pTNM) criteriafor carcinoma of the esophagus (Union Internatio-nale Centre le Cancer-American Joint Committeeon Cancer, 1987). This classification can also beapplied to carcinomas of the gastroesophageal junc-tion, because the biologic behavior of the tumor isconsidered to be comparable.4-6 Six patients withsquamous cell carcinoma received preoperative che-motherapy as part of a randomized trial.
Operations were performed by or under directsupervision of a senior staff member (JJBvL orHO). Transhiatal esophagectomy starts with a mid-line laparotomy. After careful exploration of the ab-dominal cavity, the diaphragm is incised to facilitateintrathoracic mobilization of the esophagus, in-cluding the tumor and the periesophageal tissue upto the carina. No formal lymphadenectomy is donein either abdomen or mediastinum, but when re-sectable positive lymph nodes around the celiac
trunk are encountered, these are removed whenpossible. When the esophagus has been mobilizedup to several centimeters proximal of the tumor, agastric tube is constructed. In the cervical phase ofthe operation the esophagus is mobilized circumfer-entially and transected low in the neck. A vein strip-per is then inserted distally and the normal esoph-agus proximal to the tumor is extracted from themediastinum by inversion. An anastomosis is per-formed cervically. No lymph node dissection is car-ried out in the neck.
A microscopically radical resection (histologi-cally negative margins) is called R0, a microscopi-cally irradical resection (histologically positive re-section margins but no macroscopic tumorremaining) is called R1, and a macroscopically ir-radical resection is called R2. Patients with an R2resection were excluded from the analysis.
Patients were routinely seen in the outpatientclinic at 3-month intervals for 2 years and at6-month intervals thereafter. Apart from a carefulhistory and physical examination, additional diag-nostic procedures were only performed if indicated.During followup, locoregional and distant recur-rence was scored. Perianastomotic recurrence (tu-mor recurrence in the gastric tube near the anasto-mosis, as identified by endoscopy or endoscopicultrasonography and confirmed by histologic bi-opsy) was considered locoregional recurrence. Re-current disease was most often diagnosed radiolog-ically by external ultrasonography, CT scan, chestx-ray, or bone scan. Whenever possible, histologicconfirmation was obtained. In some cases, however,the diagnosis was established solely on clinicalgrounds. At followup cervical and celiac axis nodeswere scored as indicating locoregional recurrence;they were scored as indicating distant metastasis forthe pTNM classification of the primary tumor.
Statistical analysis was performed using the Sta-tistical Package for Social Sciences (SPSS Inc., Chi-cago, IL). The chi-square or Fishers exact test andStudents t-test were used when appropriate.Disease-free survival curves were calculated by theKaplan-Meier method, and curves were comparedstatistically using the log-rank test. Patients whodied from nonesophageal cancer-related causes areincluded in the disease-free survival analysis and aretreated as censored when they died. The Cox pro-portional hazards model was used to examine the
144 Hulscher et al Esophageal Cancer after Transhiatal Resection J Am Coll Surg
impact of covariates (differentiation grade, presenceof lymph node metastases [pTNM-stages I and IIaversus IIb, III, and IV] and radicality of the opera-tion) on recurrence. Covariates were dichotomizedin two categories: (a) an index group for which therisk was assumed to be increased was assigned avalue of 1, versus (b) the reference group that con-stituted all other cases, which was attributed a value0. The ratio of risk of the index group over thereference group is the hazard ratio.
One hundred forty-nine patients were identified.There were three early postoperative deaths (2.0%).These patients were excluded from the analysis.Also excluded were nine patients with a macroscop-ically irradical (R2) resection.
Followup was complete in the remaining 137patients. At operation the median age was 65 years(range 37 to 84 years). There were 105 men and 32women. Ninety-five patients (69.3%) had adeno-carcinoma and 42 (30.7%) suffered from squamouscell carcinoma. The main characteristics of the pa-tients and tumors are depicted in Table 1.
The median period of followup was 24.0months (range 1.4 to 69.2 months) for all patients;in patients still alive without signs of tumor recur-
rence, followup was 36.5 months (range 23.6 to69.2 months). During the followup period, 72 of137 patients (52.6%) developed tumor recurrence,52 of 95 patients (54.7%) with an adenocarcinomaand 20 of 42 patients (47.6%) with a squamous cellcarcinoma (p50.46, Fishers exact test). The me-dian interval between operation and recurrence was11 months (range 1.4 to 62.5 months) for patientsdeveloping recurrence. There was no significant dif-ference between time to recurrence in adenocarci-noma and squamous cell carcinoma, with a mediandisease-free interval of 9.1 months (range 1.4 to62.5 months) versus 17.2 months (range 2.7 to40.1 months; p50.43) respectively, as estimated bythe Kaplan-Meier curves.
The cumulative disease-free survival curve is de-picted in Figure 1. Seven patients died of othercauses without signs of recurrent disease. Causes ofdeath in these patients were: cardiac arrest (4), pro-gressive Parkinsons disease (1), cerebrovascular ac-cident (1), and primary lung carcinoma (1).
In 32 patients (23.4% of all 137 patients or44.4% of the 72 patients with recurrence) there waslocoregional recurrence only; only systemic diseasedeveloped in 21 patients (15.3% of all patients or29.2% of the patients with recurrence). Nineteenpatients (13.9% of all patients or 26.4% of the pa-tients with recurrence) developed a combination ofboth locoregional and distant recurrent disease (Ta-bles 2 and 3).
The recurrence rate was significantly associatedwith the pTNM-stage: there was only one recur-rence in the stage I group (5.9%). In stages IIa, IIb,III, and IV the recurrence rates were 42.4%, 63.6%,69.8%, and 76.9%. Lymph node status (corre-sponding with stages I and IIa versus stage IIb, III,and IV) was an independent prognostic factor inmultivariate analysis (p,0.001, Cox proportionalhazards model, see Table 4). The same holds forradicality: of the 109 patients in whom an R0 resec-tion could be achieved, 50 (45.9%) developed re-currence, significantly less than the 78.6% recur-rence rate in the 28 patients in whom microscopictumor remained (R1) (p,