Recurrence risk after Ivor Lewis oesophagectomy for cancer

Objective The aim of this study was to analyze the profile of tumor recurrence for patients operated on for cancer of oesophagogastric junction or oesophagus by Ivor-Lewis oesophagectomy. Methods Patients undergoing potentially curative Ivor-Lewis oesophageal resection between January 1999 to December 2008 at a single center institution were retrospectively analyzed. Their clinical records, details of surgical procedure, postoperative course, pathological findings, recurrence and long term survival were reviewed retrospectively. Univariate and multivariate survival analyses were performed. Results One hundred and twenty patients were analyzed. Fifty three patients (44%) presented recurrence during median follow-up of 58 months. Five-year relapse free survival (RFS) rate was 51% (95%CI = [46; 65%]). On multivariate analysis, pT stage > 2 (HR = 2.42, 95%CI = [1.22; 4.79] p = 0.011), positive lymph node status (HR = 3.69; 95% CI = [1.53; 8.96] p = 0.004) and lymph node ratio > 0.2 (HR = 2.57; 95%CI = [1.38; 4.76] p = 0.003) were associated with a poorer RFS and their combination was correlated to relapse risk. Moreover, preoperative tumor stenosis was associated with an increased risk of local recurrence (HR = 3.46; 95% CI = [1.38; 8.70] p = 0.008) whereas poor or undifferentiated tumor was associated with an increased risk of distant recurrence (HR = 3.32; 95% CI = [1.03; 10.04] p = 0.044). Conclusion pT stage > 2, positive lymph node status and lymph node ratio > 0.2 are independent prognostic factors of recurrence after Ivor-Lewis surgery for cancer. Their combination is correlated with an increasing risk of recurrence that may argue favorably, in addition with preoperative tumor stenosis assessment, for adjuvant treatment or reinforced follow-up.


Background
Oesophageal cancer is a major public health concern as it is the fourth cause of cancer death after lung, colorectal and prostate cancers. Without contraindication of resectability or operability, surgery is the standard treatment of curative intent. However, in spite of optimal R0 resection, overall 5-year survival is poor, about 20 to 30%, because of frequent tumor recurrence [1]. Thus, exclusive chemoradiotherapy has become an alternative to surgical treatment, with a comparable overall survival in locally advanced squamous cell carcinoma [2,3]. However, ratio between adenocarcinoma and squamous cell carcinoma incidence is changing with an increasing incidence of adenocarcinoma developed on Barrett oesophagus in Western countries [4]. Regarding adenocarcinoma, combination of neo-adjuvant treatment seems to provide improved survival despite of its own morbi-mortality [5]. Study of tumor relapse could therefore allow adaptation and targeting perioperative treatment to patients with high risk of recurrence. The aim of this study was to analyze the profile of tumor recurrence in a homogenous group of patients operated on for cancer of oesophagogastric junction or oesophagus by Ivor-Lewis oesophagectomy.

Methods
From January 1999 to December 2008, 120 consecutive patients underwent an Ivor-Lewis oesophagectomy for cancer at a single institution. Their clinical records were reviewed retrospectively for age, sex, Body Mass Index (BMI), nutritional factors, American Society of Anesthesiologists (ASA) score, symptoms at diagnosis, preoperative treatment, details of the surgical procedure, pathological findings, postoperative course, recurrence and long term survival. The research was carried out in compliance with the Helsinki Declaration and approved by the Comité de Protection des Personnes (CPP) Sud Ouest (Toulouse, France).

Preoperative evaluation
Operability and resectability criteria were those usually used in oncologic guidelines [6]. All surgical indications were validated in pluridisciplinary meeting. Endoscopy, computed tomography and barium swallows were systematically performed. Ultrasonographic endoscopy (EUS) was performed systematically as soon as it was technically available and when there was no non-traversable strictures. Positron emission tomography (PET) was done depending on the suspicion of metastatic extension. Features of lymph nodes on EUS and computed tomography, including size 1 cm or more, rounded shape, well demarcated borders, and heterogeneous patterns, were used to describe the nodes as benign or malignant.
Preoperative tumor stenosis was defined as a difficulty or inability for the fiberscope (11.6 mm standard diameter) to pass through the lesion and/or at least a hemicircumferential narrowing light on barium swallow (both sides of the oesophagus narrowed by tumor noticed on at least one radiologic incidence).
Neo-adjuvant chemotherapy (CT) (Platin and 5 Fluorouracil) or chemoradiotherapy (CRT) (45 to 50 Gy and concomitant Platin/5 fluorouracil chemotherapy) was delivered in case of locally advanced cancer (T stage ≥ 3 and/or N stage ≥ 1). CRT was preferred for high volume tumors and/or for limited resection margins.
Clinical response was defined by regression of dysphagia and tumor size on post induction computed tomography. Endoscopic control, EUS nor PET were not routinely performed.
Immunonutrition was systematically delivered for seven preoperative days.

Surgical procedure
Conventional orotracheal intubation was performed without selective bronchial intubation. Gastrolysis was done through laparoscopy or midline laparotomy. Thoracotomy was done through a right posterolateral thoracotomy in the fifth intercostal space. Gastric transplant was created in the chest and anastomosed mechanically to the oesophagus. An extended upper abdominal lymphadenectomy was routinely performed comprising enbloc resection of the nodal tissue along the common hepatic and proximal splenic arteries together with that at the origins of the left gastric artery and celiac axis.
The lesser omentum was divided, encompassing the nodes along the lesser curve and an en-bloc hiatal dissection was performed removing the left and right paracardial stations and the respective crura. Within the thorax, standard lymphadenectomy was routinely performed that comprised middle and lower paraoesophageal nodes, paratracheal, carinal, left and right bronchial nodes and left recurrent laryngeal nerve chain. No cervical lymphadenectomy or three-field lymphadenectomy was undertaken even for tumors of middle third oesophagus. Patients were extubated in recovery room and then transferred to the ward.

Postoperative data
All tumors were staged post surgically by the TNM classification system of the American Joint Committee on Cancer 2002 for oesophageal or gastric cancer depending on tumor localization. The resection was designated R0 when it was thought that both macroscopic and microscopic clearance had been achieved; R1 when it was microscopically incomplete (margin inferior to one millimeter), and there was histologic evidence of invasion of the longitudinal or lateral margins; and R2 when it was macroscopically incomplete, with macroscopic residual tumor after surgery. Median lymph node ratio was defined by the ratio between number of positive lymph nodes and number of resected lymph nodes.
Operative mortality and morbidity was defined as any death or complication during the first 30 days after operation. A major complication was defined as any complication > 3 on Clavien Dindo classification.
Follow up data were obtained from patient charts, tumor registry and referring physicians. A clinical examination was carried out every three to six months for three years. Thoraco-abdominal computed tomography was performed at least twice a year for 3 years or according to symptoms. Endoscopy was done in case of dysphagia.
For recurrence study, final date of follow-up was scheduled for March 31, 2010. Histologic, cytologic, or unequivocal radiologic proof was required before recurrence was diagnosed. Recurrence was classified into three categories: local recurrence (occurring at the surgical resection site or on cervical site), distant recurrence (in case of distant metastasis) and disseminated (in case of local and distant recurrence). Operative deaths and deaths that were not related to cancer were included in the survival statistics.

Statistical analysis
Data were summarized by frequency and percentage for categorical variables and by median and range for continuous variables. All survival times were calculated from the date of surgery. Relapse free survival (RFS) and overall survival (OS) were estimated by the Kaplan-Meier methods using the following first-event definitions: loco-regional or distant recurrence for relapse free survival and death from any cause for overall survival. Univariate analysis was performed using the Log-Rank test. The following three step algorithm using Cox proportional hazard modeling was performed: Step 1: Influence of pre-operative variables with a p-value <0.05 in univariate analysis was evaluated.
Step 2: Influence of pathological variables with a p-value <0.05 in univariate analysis was evaluated.
Step 3: Influence of pre-operative and pathological variables statistically significative in step 1 and 2 was evaluated.
Competing risks multivariate analysis was conducted using the Fine & Gray model in order to evaluate the influence of different variables on the cumulative incidence in the presence of competing risks (i.e local recurrence vs distant relapse).
All p-values reported were two-sided. For all statistical tests, differences were considered significant at the 5% level. Statistical analyses were performed using the STATA 10.0 software.

Preoperative results
One hundred and twenty patients were included. Median age at surgery was 62 years [27-79 years] (Table 1). Sex ratio male: female was 4: 1. Dysphagia was the main symptom (68%) and 16% of patients were asymptomatic among them 11.6% were diagnosed for Barrett oesophagus follow-up and 5% (n = 6) on an examination prescribed for other indication. Severe weight loss (weight loss exceeding 10% of usual weight in one month or 15% in six months) was present in 20% of patients as 32% were overweight (BMI > 25 Kg/m 2 ) and 6% were obese (BMI > 30 Kg/m 2 ). Median BMI was 23.5 Kg/m 2 [12.5-43.3 Kg/m 2 ]. Most lesions were localized on the lower third of oesophagus (49%). EUS was performed on 62 patients and lymph node invasion was suspected in 62% of cases (n = 38). PET was performed in 20 patients (16%).

Discussion
Several studies have identified different risk factors of tumor recurrence after surgery of oesophageal cancer. Positive lymph node status is the most frequent factor identified [7][8][9]. It is also highlighted as an independent risk factor in our study. However, the impact of lymph  node involvement can be refined by studying the prognostic role of lymph node ratio ie ratio between positive lymph nodes and resected lymph nodes. Ratios varying from 0.1 to 0.5 have already been discussed [8,10]. In this study, ratio was set at 0.2, low enough to be discriminating. It turns out to be an independent major risk factor for relapse free survival that can be illustrated by a 5-year RFS of 67% with a ratio ≤ 0.2 versus 0% with a ratio > 0.2. An interesting finding of our study is the independency of lymph node ratio as a risk factor, particularly independent from lymph node positivity what could have been a confusing factor. Other authors have also identified the number of involved nodes as a prognostic factor [9,11]. These studies show the importance of node involvement as a quantitative and not only qualitative factor and are behind the recent changes to the 2009 UICC TNM classification. This last distinguishes nowadays different stages of nodal involvement for oesophageal cancer, superimposed on the oesophagogastric junction cancer classification, whereas it included only a single stage of nodal involvement in the previous version [12].
Unfortunately, information on the presence of this major prognostic factor is only obtained retrospectively on specimen examination. It does not allow optimal adaptation of neo-adjuvant treatment. Current diagnostic methods combining computed tomography, EUS and endoscopy permit an appropriate staging in 56% of cases and adjunction of PET does not provide better sensitivity for the preoperative lymph node diagnosis compared to EUS or computed tomography (sensitivity of 0.57 vs 0.8 and 0.5 resp.) [13,14].
Indication of adjuvant treatment after oesophageal surgery is not recommended even in case of lymph node involvement but may be discussed if the patient is healthy, demanding, and informed. However, lymph node invasion seems to be a very accurate predictor of relapse free survival and may be used to identify patients requiring adjuvant chemotherapy or chemoradiotherapy to treat systemic metastases developing after primary resection [15]. Tumor depth, defined with pT stage, was also found as an independent prognostic factor in our study. Depth invasion has also been described in several other studies as a prognostic factor independent of nodal status [16,17]. Similarly, the importance of tumor length or its development location have been described [18].
Preoperative tumor stenosis was found in our study to be an independent risk factor of local relapse that, to our knowledge, has never been previously reported. This is the main new factor identified in our study. The interest of this factor, simply defined on endoscopic and/or barium swallow criteria, is its preoperative availability. Despite its clear clinical correlation with a high pT stage, quality of resection but also with nutritional status, it appears, statistically, to be an independent risk for local relapse and not for distant relapse. It could represent an additional argument in the discussion of either preoperative chemoradiotherapy or in the indication of transthoracic vs transhiatal approach to get an optimal local disease control. Indeed, there is evidence that preoperative chemoradiotherapy increases the rate of complete resection i.e. local control for patients with locally advanced disease even if this was not always translated into a survival benefit in individual studies [19,20]. Similarly, transthoracic approach allows greater lymphadenectomy (eight more lymph node retrieved compared with transhiatal approach) [21] that might provide increased local control and disease free survival [22]. Several other risk factors of recurrence, very different, have been described as circumferential margin involvement [23,24], involvement of subcarinal and lower paraoesophageal lymph nodes [25], tumor-stroma ratio [26], endoscopic tumor length [27], good performance status of the patient [28]. Even if tumor stage and lymph node ratio are probably the most relevant risk factors of recurrence, this kind of studies, as ours, are necessary to refine oesophagus cancer prognosis and to help in the decision for adjuvant therapy. Locoregional recurrence has been described with longer survival than distant or disseminated recurrence particularly in case of cervical recurrence [29,30]. On the other hand, Bhansali et al. [31] reported no difference in different types of recurrence. In our report, survival in case of local recurrence was surprisingly 3 months shorter than for distant recurrence. This is probably due to differences in treatment protocol after relapse.
Half of our patients presented with pT < 3 or negative preoperative lymph node status. That partly explains our low rate of patient who received preoperative treatment (30%). Moreover, the length of our study, begun in 1998, is another explanation, first main studies of neo-adjuvant treatment still not published at that time. Finally, low rate of 50% of preoperative EUS performed may have led to a preoperative downstaging and also explains the rate of preoperative treatment.
A same approach was performed with overall survival and brought similar results. As our study focused on recurrence risk, these results are not detailed here. Five year overall survival estimated by Kaplan Meier method was 49% in our survey. This result, higher than OS reported in literature [1], may be due to our high rate of low stage tumors and drastic patient selection for surgical treatment.
Adenocarcinoma and squamous cell carcinoma were both included in this study that may be a bias. Indeed, Siewert et al. [32] reported two different diseases with different pathogenesis, epidemiology, tumor biology and prognosis requiring different therapeutic strategies. However, focusing on relapse and survival, several studies comparing adenocarcinoma and squamous cell carcinoma have reported worse, similar or better survival according to histopathology [33][34][35]. In our study, histopathology doesn't appear to be correlated to RFS and OS and multivariate analysis allowed us to statistically study all of our patients in the same cohort. Signet ring cell adenocarcinoma is well known to have poorer prognosis [36]. Our number of patient and particularly with signet ring cell was too small to study it as a risk factor.
This study might initiate thinking for patients who received neoadjuvant treatment as indicated in guidelines, if they could benefit, in presence of these risks factors, of post-operative treatment or of a more intensive clinical or radiological follow-up. Moreover, recurrence risk is directly correlated with the number of positive risk factors (pN stage, pT stage and lymph node ratio), creating a risk score of recurrence, what may ponder discussion according to this score. Our score has to be confirmed by others studies but strongly argue for adjuvant therapy in case of cumulated risks, particularly in case of tumor regression after preoperative treatment.
Limitations of this study are numerous. The main is it retrospective type, while conducted on a homogeneous group of patients, that has led to heterogeneity in preoperative administrated treatments. These results have to be confirmed in a prospective one with standardized protocols. Then, monocentric aspect of the study allowed us to analyze patients with standardized practices but represents also a bias as the results must be confirmed in multicentric studies. Besides, our department is not a high volume center for oesophageal surgery, due to a low incidence of oesophageal cancer in our county, but belongs to the university hospital and is experienced in oesophagogastric surgery.

Conclusion
Finally, our study demonstrates that preoperative tumor stenosis is the only prognostic factor associated with an increased risk of local relapse. Concerning relapse free survival, pT stage > 2, positive lymph node status and lymph node ratio > 0.2 are independent prognostic factors of recurrence after Ivor-Lewis surgery for cancer. Furthermore, addition of these risk factors is correlated with an increasing risk of recurrence. Thus, a routine postoperative treatment or reinforced follow-up should probably be considered in patients combining these risk factors.

Consents
Written informed consent was obtained from the patients included or their family for the publication of this study.