Chapter 17. Cancer of the Esophagus

One of the earliest descriptions of esophageal cancer was in the second century ad, when Galen described a fleshy obstructing growth in the esophagus, which was responsible for the inability to swallow and led to emaciation and death. In early Chinese literature, a patient who had esophageal cancer was described as “one suffers in autumn, and does not live to see the coming summer.” Improvement in treatment strategies has resulted in better outcome. However, most patients are still diagnosed at an advanced disease stage, with consequent poor prognosis. In 1877, Czerny was the first to successfully resect a cervical esophageal cancer and the patient lived for 15 months. Torek in 1913 performed the first successful transthoracic resection.1 A 67-year-old woman had a squamous cell cancer of the midesophagus. Through a left thoracotomy, the esophagus was resected. The proximal cervical esophagus was brought out through an incision anterior to the sternocleidomastoid muscle and tunneled subcutaneously along the anterior chest wall, where a cutaneous esophagostomy was fashioned. The patient was fed via a rubber tube connecting the esophagostomy with a gastrostomy. The patient lived for 17 years.

The first successful resection of a thoracic esophageal cancer with reconstruction using the stomach was performed by Ohsawa, a Japanese surgeon in Kyoto, who reported the technique in 18 patients in 1933.2 In 1946, Lewis described esophageal resection using a two-phase approach via a right thoracotomy and laparotomy.3 Tanner independently also described the procedure in 1947.4

Although surgical resection has remained the mainstay treatment for esophageal cancer, recent years have seen a proliferation of treatment options especially with regards to different combinations of chemotherapeutic agents, radiotherapy and surgery. There has also been a divergence in the epidemiological pattern between Western and Eastern countries, which has made a major impact on the management of this disease.

Esophageal cancer is the eighth most common cancer worldwide and the sixth most common cause of death from cancer.5 There is marked geographic variation in the incidence of cancer of the esophagus and, to some extent, among different ethnic groups within a common area. The disease is especially common in countries of the so-called “Asian esophageal cancer belt,” which stretches from eastern Turkey and east of the Caspian Sea through northern Iran, northern Afghanistan, and southern areas of the former Soviet Union, such as Turkmenistan, Uzbekistan, and Tajikistan, to northern China and India. High incidences are also found, in the Transkei province of South Africa and Kenya. In high-incidence areas, the occurrence of esophageal cancer is 50- to 100-fold higher than that in the rest of the world. It is the fourth most common cancer in China.6 The age-standardized incidence rate of esophageal cancer in China is 27.4 per 100,000, compared to 10 in Japan, 7.9 in northern Europe and 7.6 in western Europe, 5.8 in North America, and 5.5 in Australia/New Zealand.5 The provinces of Henan, Hebei, Shanxi in central/northern China, and areas within, such as Linxian and Cixian, have particularly high incidences.7,8 The crude age-adjusted mortality is up to 140 per 100,000 and is the most common cause of cancer death.8 Esophageal cancer most commonly presents in the sixth and seventh decades of life. In most countries it is a male-predominant disease, although in high-incidence areas, the male-to-female ratio approaches unity.

The most striking change in epidemiological pattern for esophageal cancer in the past three decades has been the shift from squamous cell cancers to adenocarcinomas of the lower esophagus and cardia in the Caucasian populations in Western countries. In the United States, squamous cell cancers predominate in African Americans, but the incidence of this cancer has seen a decline since the mid-1980s, while adenocarcinoma has been rising in incidence rapidly in the white population. The incidence of adenocarcinoma has surpassed squamous cell cancers since 1990.9 Similar changes have been observed in Europe and Australia. In Asia, however, esophageal cancers remain predominantly squamous cell in type and are mostly located in the midesophagus.10

Apart from squamous cell cancers and adenocarcinomas, other tumor types less commonly encountered include mucoepidermoid cancer,11 adenosquamous cancer, small cell cancer,12 basaloid squamous tumor,13 sarcomatoid carcinoma, lymphoma, melanoma,14 and various subtypes of stromal tumors.15

Various factors associated with the development of esophageal cancer are shown in (Table 17-1). Smoking and drinking as independent contributing factors are shown by prospective studies of patients who drink but do not smoke and, conversely, of patients who smoke but do not drink.16

Genetic predisposition may be important in the pathogenesis of esophageal squamous cell cancer. Case-controlled studies have identified familial aggregation; suggesting that the cancer may be heritable.17 Mitochondrial studies have proved historical population migrations from central/northern to south-eastern China, where another high-incidence area is found, again suggesting that hereditary factors may play a part.18,19 Genetic polymorphism is important in individuals with chronic alcohol consumption.20 Approximately 36% of East Asians show a physiologic response to drinking that includes facial flushing, nausea, and tachycardia. This facial flushing response is predominantly related to an inherited deficiency in the enzyme aldehyde dehydrogenase 2 (ALDH2). Alcohol is metabolized to acetaldehyde by alcohol dehydrogenase and the acetaldehyde is in turn metabolized by ALDH2 to acetate. Two main variants for ALDH2 exist, resulting from the replacement of glutamate with lysine at position 487. Only individuals homozygous with the glutamate allele have normal catalytic activity. Homozygotes with the lysine alleles have no detectable activity, while heterozygotes with Glu/Lys alleles have much reduced ALDH2 activity. The inability to fully metabolize acetaldehyde results in its accumulation in the body leading to the facial flushing and unpleasant side effects. Lys/Lys homozygotes could not tolerate much alcohol because of the intensity of the side effects, and so paradoxically they do not have increased risk because they simply would not consume significant amount of alcohol. Individuals who are glu/lys heterozygotes may become habitual drinkers because they could become tolerant to the side effects of alcohol and yet they had suboptimal catalytic activity and thus the acetaldehyde accumulates. These are the individuals most susceptible to the carcinogenic effects of alcohol consumption, which is related to acetaldehyde causing DNA damage and other cancer-promoting effects.21 A simple questionnaire that elicits the history of a flushing response was shown to be useful in identifying at-risk individuals. They could be advised against drinking or to undergo screening endoscopy. The risk of developing cancer may be reduced or earlier diagnosis possible.22,23

For squamous cell cancer, in addition to drinking and smoking, dietary and environmental factors are important, especially in Asian countries. Nitrosamines and their precursors (nitrate, nitrite, and secondary amines), such as pickled vegetables, are incriminated.24 Nutritional depletion of certain micronutrients, particularly vitamins A, C, E, niacin, riboflavin, molybdenum, manganese, zinc, magnesium, selenium, as well as fresh fruits and vegetables, together with an inadequate protein intake, predisposes the esophageal epithelium to neoplastic transformation.25 Change in specific dietary habits, such as replacing traditional methods of food preservation and storage with refrigeration, together with consumption of vitamin-rich food, may have produced a drop in incidence rates in certain areas of China, especially in urban cities such as Shanghai.26 Other dietary risk factors include consumption of hot beverages, opium smoking, chewing betel nuts, and maté drinking in South American countries.

The human papillomaviruses27 and certain fungi belonging to the genera Fusarium, Alternaria, Geotrichum, Aspergillus, Cladosporium, and Penicillium are infective agents variably found to be associated with esophageal cancer.

Patients with other aerodigestive malignancies have a particularly high risk of developing squamous cell carcinoma (SCC) of the esophagus, presumably because of exposure to similar environmental carcinogens and “field cancerization.” Using esophageal cancer as the index tumor, multiple primary cancers were found in 9.5% of patients, of whom 70% were in the aerodigestive tract.28 The overall incidence of synchronous or metachronous esophageal cancer in patients with primary head and neck cancer is estimated to be 3%.29

Diseases that are known to predispose to esophageal cancer are few. The risk from achalasia is estimated to be 7- to 33-fold, but symptoms of achalasia are present for an average of 15–20 years before the emergence of cancer.30 Other diseases include lye corrosive strictures, Plummer-Vinson syndrome, tylosis, and celiac disease.

The reasons accounting for the dramatic rise in incidence of adenocarcinoma in Caucasian population is widely attributed to obesity, gastroesophageal reflux disease, and Barrett's esophagus,31–33 which are uncommon in Asian populations.34 Gastroesophageal reflux disease affects up to 44% of the general population in the United States, and approximately 5–8% will develop Barrett's esophagus,35 with an estimated annual rate of neoplastic transformation of 0.5%.36 Epidemiological data suggest a protective role of Helicobacter pylori against reflux. The high prevalence of H. pylori infection in Eastern populations may guard against reflux and Barrett's esophagus, and may account for the differences in cancer cell type.37 However, this association remains controversial.

Screening, Surveillance, and Prevention for Early Cancer

Squamous Cell Dysplasia and Cancer

Diagnosing esophageal cancer at its asymptomatic or early stage is crucial in improving prognosis, although at present this is only possible in the minority of patients. In high-incidence areas such as in China, abrasive cytology has been used for population screening. Two principal types of samplers have been used: an inflatable balloon developed in China38 and an encapsulated sponge sampler developed in Japan.39 When early-stage cancers are diagnosed by this method, excellent long-term results with 5-year survival rate approaching 90% and 25-year survival rate of 50% can be achieved, comparable to those of the normal population.40

Primary endoscopic screening with chromoendoscopy using Lugol's iodine as a useful adjunct is carried out in high-incidence areas in China (Fig. 17-1). It has been shown that dysplastic lesions seen in the esophagus have a quantifiable risk of malignant transformation.41 Long-term endoscopic screening studies are ongoing, integrating with early treatment and chemoprevention programs.42

Nutritional intervention trials were undertaken in Linxian in China for the general population in the 1980s as a form of chemoprevention. The trial was tested in 29,584 participants. At the end of 5-year intervention, the group receiving selenium, β-carotene and vitamin E was found to have a statistically significant reduction in all causes of mortality and cancer death. However, mortality reduction in combined esophageal/gastric cardia cancer was 10%, not reaching statistical significance.43 To date, no conclusive evidence is available for chemopreventive strategies for squamous cell esophageal cancer.

Barrett's Esophagus and Adenocarcinoma

For cancer due to Barrett's esophagus, screening and surveillance for early cancers have been controversial. Gastroesophageal reflux is prevalent; approximately 20% of adults have heartburn at least once per week, 5% of whom have Barrett's esophagus; thus a very substantial number of patients will require screening. However, the absolute risk of adenocarcinoma is low even in subgroups of patients with severe reflux symptoms. Moreover, 40% or more of patients with esophageal adenocarcinoma have no prior reflux symptoms and therefore would not be detected through screening programs targeted to those with such reflux symptoms.32 Most patients with Barrett's esophagus also die from unrelated causes,44 and the presence of Barrett's esophagus does not change life expectancy or overall survival.45,46 These arguments, together with the high cost of endoscopy, mitigate against general population screening. Although retrospective studies have demonstrated survival benefits in patients with Barrett's esophagus undergoing surveillance,47,48 these studies may have been biased because of selection, lead time, and length bias.49 There is currently no confirmed evidence proving that screening or surveillance will lead to improved survival in patients with Barrett's esophagus.50 Screening for Barrett's esophagus in the general population is not recommended. The use in selective populations at higher risk remains to be established.48

Despite the lack of clear evidence, individuals who are identified to have Barrett's esophagus should enter surveillance programs. Systemic four-quadrant, 2-cm biopsy protocol using large biopsy forceps is recommended.48 Dysplasia is so far the only reliable indicator of risk development of invasive cancer. The recommendation given by the American College of Gastroenterology with regards to endoscopy interval and treatment is shown in Table 17-2. Endoscopy is performed every 3 years for those with no dysplasia and yearly for low-grade dysplasia. Diagnosis of high-grade dysplasia implies the need for intervention (by surgery or endoscopic means), or intensive surveillance at 3-month intervals. If the latter is preferred, a four-quadrant, 1-cm protocol is required for diagnosis of early invasive cancer.

Endoscopy and systemic biopsies remains the gold standard for diagnosis of Barrett's esophagus, dysplasia, and early cancer. Other modalities such as cytology with or without fluorescence in situ hybridization (FISH), autofluorescence imaging, narrow band imaging, optical coherence tomography, and confocal laser endomicroscopy are investigational techniques aimed at enhancing diagnostic capabilities.51

Chemoprevention can potentially prevent Barrett's esophagus from developing into invasive cancer. Proton pump inhibitors (PPIs) and nonsteroidal anti-inflammatory drugs (NSAIDs) have drawn the most attention in recent years. Currently there are no data that directly support the use of PPIs to prevent cancer, although retrospective studies have shown decrease in development of dysplasia in long-term users.52 For NSAIDs, a meta-analysis of pooled studies found a protective association between aspirin/NSAIDs and esophageal cancer of both histologic types.53 However, a randomized controlled trial showed that celecoxib, a COX-2 inhibitor, was not more effective than placebo in patients with Barrett's esophagus and dysplasia in changing the proportion of biopsies with dysplasia.54 An ongoing phase III randomized trial in the United Kingdom (AspECT [Aspirin Esomeprazole Chemoprevention Trial] trail) aims at assessing whether intervention with aspirin results in decreased mortality or conversion rate from Barrett's metaplasia to adenocarcinoma or high-grade dysplasia.55

Advanced Cancer

For symptomatic patients, the spectrum of symptoms varies depending on the extent of disease. Elderly patients who complain of dysphagia must be assumed to have esophageal cancer until proven otherwise, especially in high-risk areas. Patients with chronic reflux symptoms who develop dysphagia must have the diagnosis of tumor entertained in addition to a reflux stricture. In advanced cases the most common presenting symptom is dysphagia (80–95%), which is progressive in severity. However, many patients delay seeking medical attention until severe dysphagia and weight loss have occurred. Regurgitation is common. In high-grade obstruction, this symptom may be worse at night when the patient lies supine. Fluid regurgitation can lead to bouts of coughing, aspiration, and even chest infection. Odynophagia (retrosternal pain associated with swallowing) is not uncommon. Hoarseness is the result of recurrent laryngeal nerve compression, either by the primary tumor or by metastatic lymph nodes.

The demographics of patients who suffer from squamous cell cancers and adenocarcinomas are different.56 Patients with adenocarcinomas tend to be of higher socioeconomic class and have obesity-related chronic disease such as ischemic heart disease (Table 17-3). Examination of these patients therefore rarely reveals a wasted individual. Patients with squamous cell cancers are blue-collar workers, and general examination may show evidence of weight loss and muscle wasting. Chronic smoking and alcohol consumption leads to a higher prevalence of chronic lung disease and liver cirrhosis. The more proximal tumor location more easily predisposes to pneumonia from aspiration or the development of a tracheoesophageal fistula. Lymph nodes in the supraclavicular regions should be sought in all patients.

Staging System

Accurate staging serves to provide information for stage-directed therapies and is important for quality control for clinical trials.

The clinical staging system follows the American Joint Committee on Cancer (AJCC) staging or the International Union Against Cancer (UICC) TNM (tumor-node-metastasis) system, which are recently modified.57 The definitions of TNM, tumor grade, level of tumors and nodal stations are shown in Tables 17-4,Tables 17-5, Tables 17-6,Tables 17-7 and Figs. 17-2 and 17-3. This recently modified TNM system differs from the previous versions mainly on (1) the regional nodes encompass areas from the neck, through the mediastinum to the upper abdomen, including the celiac nodes; previously used M1a and M1b categories are deleted; (2) the separation of N1 to N3 depends on the number of nodes involved; (3) squamous cell cancers are stage-grouped differently to adenocarcinoma; and (4) location of tumor and grade of differentiation are also taken into consideration. Previously, it has been uncertain whether adenocarcinoma of the cardia should be staged as gastric or esophageal cancer. In this new edition, tumors whose epicenter is in the lower thoracic esophagus, gastroesophageal junction (GEJ), or within the proximal 5 cm of the stomach (cardia) that extend into the GEJ or esophagus are stage-grouped similar to adenocarcinoma of the esophagus and not that of the stomach. Cancers with their epicenter in the stomach greater than 5 cm distal to the GEJ, or those within 5 cm of the GEJ but not extending into the GEJ or esophagus, are stage-grouped using the gastric cancer staging system.

Siewert's classification aims at classifying tumors that are located 5 cm proximal and distal to the GEJ into types I to III (esophageal, cardiac, and subcardiac), depending on the relative extent of involvement of either the esophagus or stomach (Fig. 17-4). The three types of cancers are different in patient demographics, possible etiology, histopathologic features, and prognosis.58 This classification is useful clinically but is not considered in the new staging system.

The Japan Esophageal Society further classifies T1a/T1b tumors into finer categories; as there are important therapeutic implications (Table 17-8). This is discussed in later sections.

Apart from physical examination and simple chest radiograph, specific methods in clinical staging include barium contrast studies, bronchoscopy, computed tomography (CT) scan, percutaneous ultrasound of cervical lymph nodes ± fine-needle aspiration (FNA) cytology, endoscopic ultrasound (EUS) ± FNA, 2-[18F] fluoro-2-deoxy-d-glocose (FDG) positron emission tomography (PET) scan, and laparoscopy and/or thoracoscopy.

Barium Contrast Studies

Typical features on a contrast barium study include mucosal irregularity and shouldering, narrowing of the lumen and proximal dilation of the esophagus (Fig. 17-5). Tortuosity, angulation, axis deviation from the midline, sinus formation, and fistulization to the bronchial tree are signs indicative of advanced tumor that has traversed the adventitia and involved the neighboring fixed organs.59 With the availability of other staging modalities, barium studies are becoming less essential.

Bronchoscopy

Use of the fiberoptic endoscope allows histologic confirmation of the cancer by biopsy or brush cytology. Flexible bronchoscopy is performed to assess tumor involvement of the tracheobronchial tree, especially for tumors in the middle and upper esophagus. Signs of involvement include a widened carina, external compression, tumor infiltration, and fistulization. The last two signs contraindicate resection.60 Gross macroscopic bronchoscopic appearance may not be accurate, and biopsy and brush cytology is recommended.61

Computer Tomography Scan

The main value of CT scan in staging of esophageal cancer lies with its ability to detect distant disease, such as that in liver, lungs, bone, and kidneys. When metastasis to the liver is more than 2 cm, sensitivity is 70–80% although it drops to approximately 50% when it is less than 1 cm.62 Solitary lung metastases are rare in patients presenting with esophageal carcinoma63 and thus, when seen on CT, are more likely to be primary lung cancers or benign nodules and should be investigated as such.

In the diagnosis of T4 disease, obliteration of the fat plane between the esophagus and the aorta, trachea and bronchi, and pericardium is suggestive of invasion, but the paucity of fat in cachectic patients makes this criterion unreliable. When the area of contact between the esophagus and the aorta extends for more than 90 degrees of the circumference, an 80% accuracy of infiltration was reported,64 but this is by no means absolute and the accuracy is inferior to that of EUS.

The sensitivity of detecting mediastinal and abdominal nodal involvement is suboptimal with CT scans because only size alone can be used as diagnostic criterion. However, normal-sized lymph nodes may contain metastatic deposits and enlargement of lymph nodes may be due to reactive and inflammatory hyperplasia. Studies using high-resolution helical CT scanning have demonstrated sensitivities of 11–77% as well as specificities of 71–95% for detection of regional nodal disease.65,66 CT scanning is nowadays commonly performed together with PET scanning; a composite picture is created in the same setting to correlate more accurate anatomy with metabolic uptake (Fig. 17-6). Experience with magnetic resonance imaging (MRI) has shown limitations similar to those of CT.67

Endoscopic Ultrasound and Percutaneous Ultrasound

Endoscopic ultrasound (EUS) is the only imaging modality able to distinguish the various layers of the esophageal wall, usually seen as five alternating hyper- and hypoechoic layers (Fig. 17-7). The accuracy of EUS for tumor and nodal staging averages 85 and 75%, respectively, compared to 58 and 54% for CT scanning.68 One problem with EUS is the nontraversable tumor stricture, which occurs in about one-third of patients.69,70 Early studies showed that predilation may result in up to a 25% chance of perforation.71,72 More recent results suggest that predilation is safe, and the success rate of complete examination depends on the size of dilation—36% for 11–12.8 mm, and 87% for 14–16 mm.73 An alternative is to use miniaturized ultrasound catheter probes passed through the working channel of a conventional endoscope, which can achieve comparable accuracy to conventional EUS.74

Echo features of lymph nodes that suggest malignant involvement include echo-poor (hypoechoic) structure, sharply demarcated borders, rounded contour, and size greater than 10 mm, in increasing order of importance.75 A collective review showed that the overall accuracy of staging nodal disease was 77%.68 The accuracy of EUS may differ for different lymph node locations and is related to the depth of penetration of EUS (about 3 cm). It is best for detecting paraesophageal nodes, and sensitivity varies inversely with the axial distance of the nodes from the esophageal axis.76 The ability to perform EUS-guided FNA cytology of suspicious nodes (such as celiac nodes) is another factor that makes EUS superior to CT scanning.77

Percutaneous ultrasound is particularly useful for obtaining FNA biopsies of cervical lymph nodes. In one large study in 519 patients, cervical lymph node metastasis was detected in 30.8% of patients (160/519). The sensitivity, specificity, and accuracy of US diagnosis in patients who underwent subsequent cervical lymphadenectomy were 74.5%, 94.1%, and 87.6%, respectively. In those who did not undergo neck dissection, the chance of cervical nodal recurrence was low, at less than 5%.78

Information gained by combining preoperative cervical ultrasound and EUS can be highly prognostic. In one study, when the number of metastatic nodes was stratified into subdivisions of 0, 1–3, 4–7, and 8 or more, the number of involved lymph nodes was prognostically similar to the eventual subdivisions as determined by histological diagnosis.79 However, both percutaneous and EUS are highly operator-dependent, and their meticulous application is required to produce these results.

Fdg-PET Scans

PET scan is gaining popularity in esophageal cancer staging (see Fig. 17-6).80,81 For the detection of the primary tumor, the sensitivity of PET ranges from 78 to 95% with most false-negative tests occurring in patients with T1 or small T2 tumors.65,82 Adenocarcinomas of the GEJ and proximal stomach sometimes show limited or absent FDG accumulation regardless of tumor volume (FDG nonavidity). Some investigators observed this phenomenon in as many as 20% of these patients, which seems to be related to the diffusely growing subtype and poorly differentiated tumors.83

PET does not provide definition of the esophageal wall and thus has no value in T stage. For locoregional nodal metastases, its spatial resolution is also insufficient to separate the primary tumor with juxtatumoral lymph nodes because of interference from the primary tumor, and thus most studies demonstrated poor sensitivity.82,84 This is especially true for nodes in the middle and lower mediastinum, where most primary tumors are found. In one study, the sensitivities of PET for detecting cervical, upper thoracic, and abdominal nodes were 78%, 82%, and 60%, respectively, but was only 38 and 0% respectively for the mid- and lower mediastinum.65 Specificity of PET in detecting regional nodes is usually much better, reaching 95–100% in some studies.82,84 The low rate of false-positive findings is important in preoperative staging.

A meta-analysis of 12 publications on PET scanning in esophageal cancer showed that the pooled sensitivity and specificity for the detection of locoregional metastases were 0.51 (95% CI, 0.34–0.69) and 0.84 (95% CI, 0.76–0.91), respectively. For distant metastases, the corresponding figures were 0.67 and 0.97. When two studies (out of 11) that had particularly low sensitivities for detection of distant metastases were excluded (probably because they included more early tumors), the pooled sensitivity improved to 0.72 and specificity to 0.95.85 This study highlights again that the accuracy of PET in locoregional nodes is only moderate. EUS-FNA is superior in this regard. PET is more useful for picking up distant metastases.

A multi-institutional trial with a primary objective to evaluate whether PET could detect metastatic disease that would preclude esophagectomy was recently published. Patients who had operable disease after conventional staging (including CT scan) were evaluated with PET scan. Of 189 patients, only 9 (4.8%) had M1b disease found and confirmed as true positives and were excluded from surgery. An additional 3.7% had unconfirmed M1b disease. However, apparent M1 findings by PET were also found in at least 3.7% of patients.86 The true value of PET scan may therefore be limited and cost-effectiveness should be evaluated further.

Thoracoscopy and Laparoscopy

Thoracoscopy and laparoscopy have their advocates. Thoracoscopic staging usually involves a right-sided approach, with opening of the mediastinal pleura from below the subclavian vessels to the inferior pulmonary vein with lymph node sampling. Laparoscopic staging can include celiac lymph node biopsy and the use of laparoscopic ultrasound for detecting liver metastases. One multi-institutional study (CALGB 9380) reported results in 113 patients, and the strategy was feasible in 73% of patients. Thoracoscopy and laparoscopy identified nodes or metastatic disease missed by CT scan in 50% of patients, by MRI in 40%, and by EUS in 30%. Although no deaths or major complications occurred, it did involve a general anesthesia, one-lung anesthesia, a median operating time of 210 minutes, and a hospital stay of 3 days.87 Laparoscopy could be used in diagnosing metastases (especially peritoneal spread) or identifying unsuspected cirrhosis, which may contraindicate resection, and it could be performed as a preliminary procedure during the time of planned esophagogastrectomy. Its main contribution would be in lower esophageal and cardial adenocarcinoma, while its value is expected to be minimal for more proximally located tumors.88 Given their invasiveness, thoracoscopy and laparoscopy should be reserved for cases in whom positive confirmation of metastatic disease not otherwise obtainable is essential in deciding on treatment.

Stage-Directed Therapy

In the past, esophageal cancer was treated by surgical resection alone, radiotherapy, or use of a plastic stent for palliation. Increasing choices and combinations of therapeutic options have made staging important; the treatment for early and advanced cancers should be individualized.

Treatment for Early Squamous Cell Cancers

Early tumors include T1a-EP, LMP, MM and T1b-SM1, SM2, and SM3 lesions as defined in Table 17-8. The distinction is important because of the risk of nodal metastases. The incidence of lymph node involvement in T1a-EP, T1a-LMP, and T1a-MM tumors are 0%, 3.3%, and 12.2%, respectively. For T1b-SM1, SM2, and SM3 lesions, the respective incidences of lymph node involvement are 26.5, 35.8, and 45.9%, respectively.89 For mucosal cancers 5-year survival rates are 80–100% and for submucosal cancers 50–65%.

T1a-EP or LMP tumors are amenable to endoscopic resection because they carry a very small risk of nodal metastases and endoscopic resection is a sufficiently radical treatment. Because circumferential resection is likely to be associated with cicatricial stenosis, this procedure is indicated for lesions not exceeding two-thirds of the circumference. Lesions reaching T1a-MM or T1b-SM1 (200 μm deep from the muscularis mucosa) may be associated with nodal metastases, but endoscopic mucosal resection (EMR) is feasible for patients without clinical evidence of lymph node metastasis (relative indication). Lesions showing deep invasion (T1b-SM2 or SM3) are associated with metastasis at a frequency of about 30–50% and are treated in the same manner as advanced cancers. Recommendations by the Japan Esophageal Society with regards to endoscopic resection are shown in Fig. 17-8.90 Other unfavorable features for endoscopic resection in addition to depth of infiltration and extent of involvement include poorly differentiated tumor and findings of lymphovascular infiltration in the resected specimen.

There are many techniques of EMR but the most commonly performed is perhaps EMR-cap EMR-C. Using a cap-fitted forward-viewing endoscope, saline is injected into the submucosal layer in order to raise the lesion from the deeper wall layer. The lesion is sucked into the cap and a snare wire that has been prelooped is used to snare the lesion. The strangulated mucosa is cut by blend-current electrocautery. In a series of 250 patients, 72% had absolute indications when EMR was performed for T1a-EP or LPM lesions. In these patients, no local or distant metastases occurred during follow-up. The 5-year survival rate was 95%. All those who died within 5 years died of non–cancer-related causes.91

Endoscopic submucosal dissection (ESD) techniques are now preferred by many endoscopists. In this method, the lesion's border is first marked, and like EMR, submucosal injection is carried out. Various types of fluid have been used for injection to delay dispersion, for example glycerol, hyaluronic acid, hypertonic saline, and mannitol. Through-the-scope “knives” such as hook, needle, flex, or insulated tip (with ceramic) knives are used to cut out the lesions. This technique is thus not limited by the size of the cap in EMR-C and can be used to remove large lesions of substantial length in one piece, thus achieving the aim of en bloc removal. The depth of resection can also be deeper and controlled, often revealing the underlying muscularis propria. Positive margins are less likely, and an en bloc specimen is more suitable for more complete pathological examination. The skill to perform ESD, however, is much more difficult to acquire compared to EMR. Complications common to both EMR and ESD include bleeding (which is usually minor), perforation (which can be prevented by adequate submucosal saline injection and can be treated sometimes with hemoclip), and stenosis (which tends to occur when the lesion is large).

Treatment for High-Grade Dysplasia and Early Adenocarcinoma

Barrett's high-grade dysplasia, synonymous with intraepithelial cancer, is the last preinvasive stage in the metaplasia-dysplasia-cancer sequence. Options of treatment include intensive surveillance, mucosal ablation, and esophagectomy.

Intensive Surveillance

Proponents of endoscopic surveillance claim that such a strategy can diagnose invasive cancer at an early stage and treatment can be delayed until then without compromising prognosis. The high morbidity and mortality rate of esophagectomy is also thought by some to be a deterrent to immediate surgical resection. Opponents of surveillance observe that most patients with high-grade dysplasia will have an invasive adenocarcinoma identified during the following 5 to 10 years, in approximately 25% of patients at 1.5 years, 50% at 3 years, and up to 80% 8 years later.31 High-grade dysplasia is currently the only reliable marker in preinvasive cancer detection, but interobserver concordance is suboptimal in distinguishing invasive and noninvasive lesions.92 When esophagectomy is carried out in patients who have high-grade dysplasia, invasive cancer is identified in the surgical specimen in up to 42% of patients, even when patients have been recruited in surveillance programs.93 More recent evidence, however, suggests that this figure is an overestimation; a meta-analysis of histologic findings after esophagectomy for high-grade dysplasia revealed invasive adenocarcinoma (at least submucosal cancer) in 12.7% and most of these had visible lesions at endoscopy, a known risk for invasive cancer.94 In the absence of visible lesions, this figure is as low as 6.7%.95 Most would regard the finding of high-grade dysplasia as a threshold for intervention. In patients who have visible lesions, such as raised nodules and not just a flat Barrett's mucosa, endoscopic resection is recommended to ensure no invasive cancer is present.48 If surveillance is to be carried out, the American College of Gastroenterology recommends three-monthly surveillance.48 The intensity that is required in surveillance of patients with high-grade dysplasia does make this an unattractive option.

Endoscopic Therapies

The rationale of endoscopic mucosal treatments is that the incidence of nodal metastases is low in high-grade dysplasia or T1a (intramucosal) cancers, and therefore treating the mucosal disease alone will result in cure. In T1a lesions, the rate of nodal metastases is low, reported as 0–6%. Once the submucosa is invaded (T1b lesions), this figure rises to around 20%.96,97 EMR methods can be used to locally resect visible lesions in Barrett's esophagus. The largest series on the use of localized EMR was reported by Ell and colleagues. One hundred patients were treated; all had mucosal lesions of a diameter up to 20 mm, without lymphovascular invasion proven by histology of the resected specimen and histologic grades G1 and G2 arising in Barrett's metaplasia. Complete local remission was achieved in 99% of patients, 11% developed recurrence (6% locally and 5% at different locations), but successful repeated treatments were possible in all. Calculated 5-year survival rate was 98%.98

One problem about Barrett's metaplasia is multifocality of dysplasia and potential malignant transformation. Thus, in addition to localized resection of suspicious lesions, ablation of the whole Barrett's mucosa is desirable. Mucosal ablative therapies consist of various methods for ablating the metaplastic mucosa combined with high-dose acid-suppressive therapy so that normal squamous mucosa will replace the ablated metaplastic mucosa in a pH-neutral environment. Methods include circumferential EMR, bipolar electrocautery, argon beam coagulation, photodynamic therapy (PDT), and radiofrequency ablation.

If circumferential EMR is possible, more than one procedure is usually necessary to lessen the chance of stricture formation. In a series of 41 Barrett's patients who had high-grade dysplasia or early adenocarcinoma, circumferential EMR achieved complete removal of the Barrett's epithelium in 76% after a median follow-up of 32 months. Recurrent or metachronous early cancer was found in 12% of patients.99 Recurrence is thus a significant problem with this type of EMR; this is not unexpected as resection is piecemeal and not en bloc. Barrett's epithelium could be missed and grow again. The advantage of EMR compared to other forms of mucosal ablative therapy is that specimens are available for histopathologic examination.

PDT has been demonstrated in a randomized trial to reduce the cancer risk in Barrett's esophagus. In this study, 208 patients with high-grade dysplasia were randomized comparing PDT using porfimer sodium plus a PPI against PPI only. High-grade dysplasia was eliminated in 77% of the PDT group, although 39% in the PPI group also lost high-grade dysplasia on subsequent biopsies. Barrett's epithelium elimination was achieved in 52% in the PDT compared to 7% in the PPI group. Adenocarcinoma developed in 15% of the PDT group compared with 29% in the PPI group, with a longer time to progression to cancer favoring PDT.100 The problems with PDT treatment include the need for repeated sessions, photosensitivity, stricture formation (6% in the series just described), and the phenomenon of buried glands or pseudoregression; residual metaplastic mucosa beneath the regenerated squamous mucosa can be present, which makes continual surveillance necessary. This incidence can be as high as 51%.101 Because PDT does not treat nodal disease and there is not specimen histological examination, accurate pretherapy diagnosis of noninvasiveness is necessary.

Recently a more promising ablative method is shown to be effective in treating both nondysplastic and dysplastic Barrett's esophagus. It is a balloon-based circumferential endoscopic radiofrequency device (HALO360); 60 tightly spaced bipolar electrodes that deliver radiofrequency are wrapped around the balloon. A sizing balloon is first introduced into the esophagus; an appropriately sized radiofrequency frequency balloon is then used to ablate the mucosa. Ablation is based on frictional heating of cellular water molecules. The advantages of the system are that it is easy to use, and, because of its controlled depth of injury up to 500–1000 μm (to the muscularis mucosae), stricture formation is uncommon. A more focal device (HALO90) mounted on the tip of a gastroscope is also available. The upper surface of the device is a 20-mm-long × 13-mm-wide articulated platform with an electrode array identical in pattern to the circumferential device. It is best used for ablating residual Barrett's mucosa after initial HALO360 treatment.

The Ablation Intestinal Metaplasia-II (AIM-II) trial examined the use of the HALO system in ablating nondyplastic Barrett's esophagus up to 6 cm in length. HALO360 treatment was performed at baseline and repeated at 4 months if there was residual intestinal metaplasia. Focal ablation with HALO90 was carried out after 12 months if needed. At 12 months complete remission of metaplasia was achieved in 48 of 69 patients (70%) and at 30 months 60 of 61 patients (98%). No stricture or buried glands were found.102

Another trial recently published examined the use of HALO system in ablating dysplastic Barrett's esophagus; 127 patients were randomly assigned in a 2:1 ratio to radiofrequency ablation or to a sham procedure. Randomization was stratified according to the grade of dysplasia and the length of Barrett's esophagus. Primary outcomes at 12 months included eradication rates of dysplasia and intestinal metaplasia. In the intention-to-treat analyses, among patients with low-grade dysplasia, complete eradication of dysplasia occurred in 90.5% of those in the ablation group as compared with 22.7% of those in the control group. Among patients with high-grade dysplasia, the respective figures were 81 and 19%. Overall, 77.4% of patients in the ablation group had complete eradication of intestinal metaplasia, compared with 2.3% in the control group. Patients in the ablation group had less disease progression (3.6 vs 16.3%) and fewer cancers (1.2 vs 9.3%). Stricture only developed in 6% of ablated patients.103

Esophagectomy

Surgical resection is the only method to ensure complete eradication of the dysplastic mucosa, and the frequently undetected invasive cancer. Surgical resection was considered a standard treatment because of the high frequency of invasive cancers found in surgical specimens when resection was performed for high-grade dysplasia (up to 42%), though more recent evidence suggests that this figure is much lower at 13%.94 The supposedly high morbidity and mortality rates of esophagectomy are also deterrents against surgical resection. However, in specialized centers, the mortality rate from esophagectomy, especially in this group of patients, is minimal. Minimally invasive surgical methods, including thoracoscopy, laparoscopy, or esophageal stripping, further reduce the trauma of surgical access. Excellent long-term survival with good quality of life is reported.104,105

Vagal-sparing esophagectomy leaves the vagi intact, is another approach aimed at preserving quality of life, and has been shown to result in much fewer postvagotomy symptoms.106 In the Merendino procedure, limited surgical resection of the distal esophagus and GEJ, together with lymphadenectomy of the lower mediastinum and upper abdominal compartment, has also been advocated. An isoperistaltic jejunal interposition graft is used to restore intestinal continuity. This method combines the adequacy of nodal dissection and improved quality of life, as the jejunal loop prevents gastroesophageal reflux.107

In summary, in patients with high-grade dysplasia or early intramucosal cancer, there is a definite risk of progression to invasive cancer, treatment needs to be individualized. The choice between intensive surveillance, mucosal ablative therapies, and esophagectomy needs to be considered based on available expertise and patient's preference.

Surgical Resection for Esophageal Cancer

Surgical resection remains the mainstay treatment for patients with localized esophageal cancer. In dedicated high-volume centers, mortality rate from surgery of 2–3% can be achieved.56,108–112 A volume-outcome relationship is evident.111,113 Centralization of service to high-volume hospitals also improves outcome.114

Important aspects to enhance better outcome after esophagectomy are (1) selecting appropriate patients for resection, (2) choice of surgical techniques and their execution, and (3) enhancing perioperative care.

Patient Selection for Esophagectomy.

How stringent one selects patients for esophagectomy will influence the resection rate. Selection depends on many factors, including (1) the referral pattern of individual centers, (2) the prevailing treatment philosophy, (3) the availability of alternative therapies, and (4) the possible mortality that the surgeon and patient are prepared to accept. Reported resection rates range from 21 to 70–80%.110,115 This wide variation suggests probable prereferral bias or a high prevalence of early cancers in those with high resection rates.

In studies that report on improvement of surgical results over time, more stringent patient selection often comes into play, either by excluding high-risk patients or by treating advanced disease by nonoperative means.116 Resection with a clear aim for palliation is becoming uncommon, and most would only operate on patients for potential cure.

Factors often cited as being predictive of morbidity and mortality after esophagectomy include advanced age,112 poor performance status,116 nutritional depletion117 and weight loss, more proximally located tumor,112 poor pulmonary function,118 cirrhosis,119 and abnormal cardiac evaluation.116 Patients suffering from adenocarcinoma and squamous cell cancers also have different risk profiles. Patients with squamous cell cancers are more likely to be malnourished, have high alcohol intake, are smokers, and have more impairment of pulmonary and hepatic function. Patients with adenocarcinomas on the other hand are more likely to be overweight and are more at risk from cardiovascular diseases.120

Assessing a patient's fitness is often based on the surgeons' experience and intuition and is not an exact science. Objective scores can help assess operative risk and patient selection.116,118,121 In one series of studies using a scoring system based on compromised general status and poor cardiac, hepatic, and respiratory function as independent predictors of postoperative death, 30% of patients with otherwise resectable tumors were excluded from surgery. When this was applied in prospective patient selection, it led to decrease in postoperative mortality rates from 9.4 to 1.6%116

It is uncertain if patient selection based on a strict mathematical scoring system is better than one based on surgeon and anesthesiologist assessments alone. They are more likely to be complementary to each other.

Choice of Surgical Approaches.

There are many important variables in esophagectomy, such as surgical access, the extent of resection and lymphadenectomy, the type and the method of preparation of the esophageal substitute, the route of reconstruction, and the technique of esophageal anastomosis. Many of these variables are interrelated and could affect immediate morbidity and mortality rates, long-term quality of life, and survival. Tumor location and stage, patient's risk profile, and surgeon's preference and experience are important variables in deciding the surgical procedure. The surgeon should be versatile and well versed with the many different techniques to adapt to different clinical situations.

Cervical Esophageal Cancer.

In 1960, Ong and Lee first described the procedure of pharyngolaryngoesophagectomy (PLE) as a one-stage, three-phase operation that involved cervical and abdominal incisions and a thoracotomy.122 Tumors involving the hypopharyngeal and upper cervical esophageal region were resected together with the whole esophagus, and the stomach was delivered via the posterior mediastinum to the neck for pharyngogastric anastomosis. A terminal tracheostome was constructed. The thoracotomy was later replaced by transhiatal esophageal mobilization. Thoracoscopic esophageal mobilization has become another and our preferred alternative.123 PLE is associated with significant morbidity and mortality, partly related to the fact that the procedure is often performed as a last resort for salvage when no other means of palliation exists.123 So despite improvements in surgical care, results remain worse compared to patients with intrathoracic cancers. At the authors' institution, of 317 PLE performed from 1966 to 1995, mortality rate decreased from 31 to 9%.124

For tumors confined to the proximal portion of the cervical esophagus with sufficient distal margin, free jejunal interposition graft or deltopectoral or pectoralis major myocutaneous flaps are options for reconstruction after resection. The use of a free jejunal graft is advantageous because it avoids mediastinal dissection, though expertise in performing microvascular anastomosis is essential. Graft necrosis, fistula formation, and late graft strictures are specific problems. When compared with gastric pull-up, graft survival and leak rates are similar. Stricture was the most common late complication for free jejunal transfers, whereas reflux was most common in gastric pull-ups, both occurring in approximately 20% of patients.125 Functional study showed a satisfactory swallowing mechanism in all patients.126 The jejunal graft is also tolerant to postoperative radiotherapy.127 The need to sacrifice the larynx does make surgical resection an unattractive option, and chemoradiation has been used up-front in many series, with surgery reserved for salvage.128

Intrathoracic Esophageal Cancer.

For tumors in the upper thoracic esophagus, obtaining a sufficient proximal resection margin dictates an anastomosis placed in the neck. For this reason, resection is best carried out by a three-phase esophagectomy or the McKeown approach.129 In this procedure a right thoracotomy is first carried out to mobilize the thoracic esophagus together with lymphadenectomy; this is followed by abdominal and neck incisions for the mobilization of the esophageal substitute, placing the anastomosis in the neck. The split-sternum approach is an alternative, especially for tumors close to the thoracic inlet.130,131

The majority of intrathoracic cancers are squamous esophageal cancers located in the middle and lower esophagus, and Barrett's adenocarcinomas in the lower esophagus. The most widely used approach was that described independently by Lewis3 and Tanner.4 The operation begins with an abdominal phase, in which the stomach is prepared, followed by a right thoracotomy and resection of the tumor together with lymphadenectomy. The stomach is then brought up into the chest for anastomosis with the proximal esophagus at the apex of the pleural cavity.

An alternative approach involves a single left thoracotomy incision. Through a left thoracotomy and incision in the diaphragm, both the esophagus and stomach could be mobilized and resection carried out, and stomach delivered into the chest for anastomosis, either below or above the aortic arch. Proximally the aortic arch does hinder surgical access, making mobilization of the proximal esophagus and subsequent anastomosis difficult. The approach is therefore more suitable for cancer of the cardia or the distal esophagus where an adequate resection margin is obtained below the aortic arch.

A transhiatal approach, whereby the thoracic part of the esophagus is mobilized by blunt and often blind dissection through the enlarged esophageal hiatus, and the mobilized stomach is then delivered to the neck and anastomosed to the cervical esophagus This is advocated especially for distal esophageal tumor or early-stage tumors of other parts of the esophagus.

Abdominal Esophagus and Gastric Cardia Tumors.

For cancers that are limited to the abdominal esophagus or gastric cardia cancers, an abdominal-right thoracic approach as in a Lewis-Tanner esophagectomy is one option, with the proximal stomach also resected in order to gain an adequate distal resection margin. A left thoracoabdominal incision through the seventh or eighth rib space also gives excellent exposure of the low mediastinum and upper abdomen. A single left thoracotomy with opening up of the diaphragm is also an option. This gives reasonable exposure of the upper abdomen. However, lymphadenectomy toward the hepatoduodenal ligament is hampered. When a thoracotomy is not desired, opening the hiatus widely by splitting the crura laterally and the diaphragm anteriorly can gain access to the low posterior mediastinum, and distal esophagectomy can be performed with the anastomosis performed from the abdomen without the need for a thoracic incision. The anastomosis is made easier with a mechanical stapler. Recently a stapler designed with a transoral placement of the anvil into the distal esophagus makes construction of a lower mediastinal anastomosis easier. When the proximal stomach is involved by tumor, a total gastrectomy with Roux-en-Y reconstruction is preferred by many.

Transthoracic versus Transhiatal Resection.

This continues to be controversial. Proponents of transhiatal resection believe that surgical resection for esophageal cancer is mostly palliative and a cure is a chance phenomenon for only those with very early tumors. More thorough lymphadenectomy through a thoracotomy merely improves staging but does not affect prognosis. The operating time is also shorter and postoperative morbidity is less with the transhiatal approach.132 Conversely, surgeons who practice transthoracic esophagectomy (TTE) consider the open approach to be safer, with dissection under direct vision.133 A more thorough lymphadenectomy leads to better staging and survival.

Population data from the Surveillance, Epidemiology, and End Results-Medicare linked database including 868 patients from 1992 to 2002 who underwent either transhiatal or transthoracic approach were studied in one recent study; 225 underwent transhiatal and 643 received TTE. Lower operative mortality rate was observed after a transhiatal than transthoracic approach (6.7 vs 13.1%). Survival was not different after adjusting for tumor stage, patient, and provider factor.134 The largest randomized controlled trial comparing the two approaches studied 106 patients who underwent transhiatal esophagectomy and 114 patients who had the transthoracic approach for mid-lower third/cardia adenocarcinomas. Pulmonary complication rates were 27% in the former group compared to 57% in the later. Ventilation time, intensive care, and hospital stay were longer in the transthoracic group. There were, however, no significant differences in in-hospital mortality at 2 and 4%. Significantly more lymph nodes were dissected in the transthoracic group (16 vs 31). Overall 5-year survival was 34% (transhiatal) and 36% (transthoracic). Importantly, it showed that in individuals with limited nodal spread (one to eight positive lymph nodes), TTE imparted a survival advantage (64 vs 23%). Survival was not different in patients with no nodal metastases or in those with more nodal metastases.135

The location and stage of the primary tumor has bearing on which surgical approach is selected. From a purely safety point of view, transhiatal resection is not suitable for patients with advanced middle- or upper-third tumors, especially in patients with tumors closely related to the tracheobronchial tree and after neoadjuvant radiation therapy; tumor infiltration or fibrosis may obliterate tissue planes and make blind dissection unsafe. As such, its application is more suitable for lower esophageal tumors for which much of the mobilization can be performed under vision. From an oncological standpoint, the philosophy toward lymphadenectomy dictates the surgical approach.

Minimally Invasive Esophagectomy (Mie).

Various combinations of minimally invasive approaches including thoracoscopy, laparoscopy, mediastinoscopy, hand-assisted laparoscopy, and open laparotomy and thoracotomy have been explored.136 The myriad of surgical methods implies a lack of consensus on which is superior.

Large single-center series are few; some have experience of over 100 patients.137–140 Several reviews on MIE have been published136,141–144; none could conclusively show that MIE is better or worse than that of the open approach, and no randomized controlled trial has been undertaken. Conversion rate is approximately 5%, respiratory complications 13–22%, and a very low mortality rate of 3% is achieved.142,143 Biere and colleagues examined 10 comparative studies comparing MIE with open esophagectomy, comprising 1061 patients. Three comparative groups were created for meta-analysis: (1) total MIE versus open TTE; (2) thoracoscopy and laparotomy versus open TTE; (3) laparoscopy versus open transhiatal esophagectomy. There was a trend toward less mortality with MIE in groups 1 and 2, and fewer anastomotic leaks with MIE in group 2 were found. Again, definitive conclusions could not be reached because of selection bias and the variety of techniques used.144

Potentially serious intraoperative complications can occur with MIE, such as bleeding from the azygous vein145 and from intercostal vessels,146 injury to the aorta,147,148 tracheobronchial tree,149–151 and recurrent laryngeal nerve,152 but certainly they are not specific for these methods. The increased magnification and excellent visualization offered by thoracoscopy might help lessen complications.

Whether MIE could reduce morbidity and mortality rates remains controversial. This is partly because of the number of patients studied generally was too small to have enough statistical power to demonstrate a difference. There are also other reasons why benefits are difficult to confirm. With modern analgesic methods, such as epidural analgesia, postoperative pain control is less critical a problem.153 The genesis of cardiopulmonary complications is multifactorial and does not depend solely on the size of the incisions. Surgical trauma from mediastinal dissection is independent of the incision size. The benefit of smaller port sites compared with open thoracotomy may be offset by the lengthened time of single-lung anesthesia. A learning curve obviously exists for such complicated procedures.154,155 The duration of the thoracoscopic procedure, blood loss, and the incidence of postoperative pulmonary infection were all less, and the number of mediastinal nodes retrieved was more, in the latter half of a group of 80 patients who had thoracoscopic esophagectomy.154 Thus, for most series, the full technical potential may not have been realized. The number of procedures that need to be performed before the learning curve is overcome is uncertain.

Patient selection is evident in many series, and in some studies most subjects had early-stage disease or high-grade dysplasia in Barrett's esophagus.138,156 The most important test will be long-term survival by stage-by-stage comparison, but stage migration may be hard to eliminate. Most series do not report on survival data and, in those that do, there is no reported difference compared with historical controls. Existing data, however, do show that nodal harvesting can be equivalent to that of open surgery.136 The place of MIE thus remains controversial without a well-conducted randomized controlled trial.

Extent of Resection: Axial and Lateral Margin.

One of the most controversial aspects of treating gastrointestinal malignancies is the appropriate extent of resection, and this debate is exemplified by esophageal cancer.157

An R0 resection is consistently identified as the most important prognostic factor for long-term survival. An R0 resection results in total removal of the tumor mass (primary and lymph nodes) with clear proximal, distal, and lateral margins. The need to obtain clear axial and lateral margins is less controversial. The propensity of esophageal cancer to spread intramurally and to have multiple separate tumors in the esophagus is well recognized. The prevalence of intraepithelial, subepithelial, or intramural spread was as high as 46 and 54%,158,159 and multiplicity of tumor was found in around 30% of patients.159,160 The deeper the wall penetration of the primary tumor, the farther away such spread can take place.158 It is clear that the chance of a histologically positive margin declines with increasing distance at which the esophagus is transected away from the tumor edge, and that the frequency of anastomotic recurrence is a function of the length of proximal resection margin attained. Taking into account shrinkage of the specimen after resection, as a guide to surgery, an in situ margin of 10 cm (fresh contracted specimen of approximately 5 cm) should be aimed at, to allow a less than 5% chance of anastomotic recurrence.161 Intraoperative frozen section is one method to ensure a negative margin. However, a histologically involved resection margin does not necessarily lead to definite anastomotic recurrence, and a negative margin does not preclude anastomotic recurrence. The occurrence of skip lesions or submucosal spread can be missed even by a conscientious pathologist; hence margins may be falsely negative. Extramural recurrence with infiltration back to the anastomosis may also be indistinguishable from true anastomotic recurrence. Patients who have positive histologic margins are those likely to have more advanced disease, and early recurrences at more distant sites may make anastomotic recurrence less relevant. In our study, a positive histologic margin (diagnosed with definitive histology and not with frozen section) occurred in 7.5% of patients who had esophagectomy, which had an anastomotic recurrence rate of 10.3% compared to 4.9% in those with a negative margin. The difference, however, did not reach statistical significance.161

Microscopic involvement of the lateral margin (macroscopically clear) results in increased chance of local recurrence and worse survival.162 Obtaining a clear lateral margin is difficult with esophageal cancer because of its anatomical position and adjacent indispensable structures. Neoadjuvant therapy may help achieve this. Some Western centers advocate the concept of “en bloc” resection, which aims at removing the primary tumor together with the pericardium, thoracic duct, azygous vein, intercostal vessels, and bilateral pleurae overlying the primary tumor and a surrounding cuff of crura (where the primary tumor is abutting) to enhance lateral clearance.108,163 Obviously this type of resection is less suitable for upper esophageal cancers in close proximity to the trachea. The concept of “en bloc” resection is thus more applicable for Western patients, where most tumors are adenocarcinomas of the lower esophagus.

Extent of Lymphadenectomy: Squamous Cell Cancers.

As discussed previously, the ability to perform lymphadenectomy is closely related to the surgical approach utilized, and an open transthoracic or thoracoscopic approach is necessary, unless only a limited lower mediastinal dissection is planned. In countries where squamous cell cancers are prevalent, transhiatal resection is uncommonly performed based on safety concerns, as well as because the value of lymphadenectomy is less questioned.

Conventional lymph node dissection for esophageal cancer usually involves a “standard two-field” lymphadenectomy, which entails removing the nodes and periesophageal tissue below the level of the carina, and the lymph node stations around the celiac trifurcation. When superior mediastinal lymph node dissection is performed, it is sometimes known as “extended two-field lymphadenectomy.” “Three-field” lymphadenectomy involves additional bilateral cervical lymph node clearance (Figs. 17-9, Figs. 17-10, Figs. 17-11, Figs. 17-12, Figs. 17-13, Figs. 17-14). For intrathoracic squamous cell cancers, detailed lymph node mapping of metastatic disease in Japan shows that lymph nodes can spread to the neck, mediastinum, and upper abdomen around the celiac trifurcation. The overall rate of cervical lymph node metastases is approximately 30%. In relation to the level of primary tumor, cervical lymph nodes are involved in 60, 20, and 12.5% of upper-, middle-, and lower-third tumors, respectively. When nodes along the recurrent laryngeal nerves from the superior mediastinum are considered together with the cervical nodes as one entity, this “cervicothoracic” group nodes are involved in up to 63.4% of proximal-third, 45.2% of middle-third, and 42.0% of lower-third cancers.164 These data provide the rationale behind “three-field” lymphadenectomy, where the true value of extended lymphadenectomy does not lie with the addition of a cervical phase, but the completeness of the superior mediastinal dissection along the recurrent laryngeal nerves to the neck.

Three-field lymphadenectomy as practiced in Japan shows an overall hospital mortality rate of 4%. Although this very low mortality rate is achieved, most of these results come from experienced and specialized institutions and such extensive surgery is expected to carry with it a more unfavorable outcome if it were more widely and unselectively applied. In addition, morbidity rates are substantial; septic complications were the most common at 26.8%, followed by pulmonary ones at 21.3%.165 Recurrent laryngeal nerve injury can occur in more than 50% of patients, which predisposes to pulmonary complications and impairs long-term quality of life.166

Perhaps based on the realization that such an extensive operation carries with it substantial morbidity and that not all patients can benefit, the recent focus of research in this area is to further refine the indications for extended lymphadenectomy. A survival advantage was only evident for upper- and middle-third cancers in some studies.164,167,168 Other poor prognostic factors include (1) when all three fields have metastatic nodes; (2) when a lower-third tumor has positive cervical nodes; and (3) when five or more lymph nodes are involved.169 These situations suggest advanced metastatic disease and three-field lymphadenectomy may not be justified. Other suggested strategies include using intraoperative polymerase chain reaction to examine recurrent laryngeal nerve lymph nodes to predict the need for cervical dissection,170 similar to the concept of sentinel lymph node metastasis,171 and taking a two-stage operative approach to select patients suitable for cervical lymphadenectomy.172 Replacing three-field lymphadenectomy by neoadjuvant, adjuvant, or intraoperative radiotherapy173 are alternatives, but their roles remain controversial.

Another major criticism of three-field dissection is that the prognostic superiority over conventional resection is only a result of stage migration. While retrospective studies provide evidence for benefits of three-field dissection,174,175 the more robust evidence of a well-performed randomized controlled trial is lacking. Two small randomized trials could not demonstrate convincing survival advantage, and, in both, the patient groups appeared to be highly selected and not well-matched, and adjuvant therapies were not controlled for.176,177

Barrett's Adenocarcinoma and Gastric Cardia Cancers.

For Barrett's adenocarcinomas of the lower esophagus and cancer of the gastric cardia, data suggest that nodal spread tends to occur later than for squamous cell cancers. Positive nodes are found in approximately 10% of patients with squamous cell cancers for T1a lesions, while in Barrett's cancer this is only 0–6%. In T1b cancers, the respective figures are 30–50% for squamous cell and 20% for adenocarcinomas. In addition, the pattern of lymphatic spread also differs; more than 85% of all positive nodes in early adenocarcinoma are located in close proximity to the primary tumor in contrast to fewer than 60% in squamous cell cancers.178 Nodes are not commonly found in the superior mediastinum and, when present, probably indicate very widespread disease.179 Thus lymphadenectomy is generally performed using a standard two-field approach. The advent of transhiatal esophagectomy came at a time when esophagectomy was a high-risk operation with high mortality rates, and this less invasive method probably contributed to reducing overall death rates. With improvement in surgical techniques and perioperative care, it seems that, in most experienced centers, when selected appropriately, both procedures can be carried out safely and the margin of benefit in reducing morbidity for most patients with the transhiatal operation is not overwhelming. There is also increasing evidence of the benefits of radical lymphadenectomy in recent years.

The concept of en bloc resection has been discussed in a previous section; this enhances lateral margin clearance, results in a complete lymphadenectomy within a facial envelop surrounding the primary tumor,108,163 and is especially advocated for adenocarcinoma of the lower esophagus. In dedicated centers, en bloc resection has a morbidity rate of 40%, a mortality rate of less than 5%, and a 5-year survival rate of 37–52%.163,180,181 It has been suggested that local recurrence can be reduced to an impressive 5% within the field of dissection,163,182–184 and nodal recurrences are mostly found outside the limits of dissection in the superior mediastinum or aortopulmonary window, in areas along the recurrent laryngeal nerves that are not routinely removed. Taking en bloc resection further, in selected centers in the United States and Europe, three-field lymphadenectomy has been tested and interestingly also yielded similar incidences of positive cervical lymph nodes of around 30%.109,184 This type of resection, however, is not commonly performed in the West.

For tumor of the gastric cardia (Siewert types II and III tumors), most surgeons would perform a total gastrectomy with a Roux-en-Y jejunal loop reconstruction, though some would prefer to preserve the distal stomach for anastomosis. An upper abdominal compartment nodal dissection around the celiac axis seems routine for all, but complete lower mediastinal nodal dissection is somewhat controversial. Some argue that thorough lower mediastinal dissection is needed and this is only possible with the addition of a thoracotomy; others believe that this is unnecessary, and mediastinal nodal involvement could indicate advanced disease for which survival is poor regardless of the extent of lymphadenectomy. The Japanese Oncology Group trial 9502 addressed this question. Patients whose tumors were Siewert II or III adenocarcinomas and which have infiltrated into the esophagus for less than 3 cm were randomly assigned to a transabdominal (n = 82) or left thoracoabdominal approach (n = 85). A more thorough mediastinal dissection was deemed only possible with the later approach. The trial was closed prematurely after the first interim analysis, when the predicted probability of left thoracoabdominal approach having a significantly better overall survival than transabdominal route at the final analysis was only 3.65%. The morbidity rate was worse after the left thoracoabdominal approach. Thus a transabdominal approach seems adequate, though the surgeon must be prepared to add a thoracotomy when frozen section indicates a positive proximal resection margin.

Regardless of tumor histologic type, increasing evidence is emerging to show that extended lymphadenectomy is related to survival, from single and multi-institutional studies,185,186 as well as from population data.187,188 The number of nodes removed correlates significantly with long-term survival. One international multicenter study showed that the number of nodes removed was an independent prognostic factor, in addition to age, gender, cell type, presence of nodal metastases, number of nodes involved, and depth of tumor invasion.186 The optimal number of nodes removed was identified as 23, though this number varies among studies. From a Worldwide Esophageal Cancer Collaboration including institutions from the United States, Europe, and Asia, the number of nodes that must be removed to maximize survival depends on the pT classification: for pT1, approximately 10 nodes must be resected; for pT2, 20 nodes; and for pT3 or pT4, 30 nodes or more.189 Thus, one should resect as many regional nodes as possible, balancing the extent of lymphadenectomy with morbidity.

Reconstruction after Esophagectomy

The reconstruction phase of an esophagectomy determines to a significant extent the postoperative morbidity and long-term quality of life. The most commonly used conduit is the gastric tube, and of the many configurations, a tailored isoperistaltic tube based on the greater curvature with preservation of the right gastric and right gastroepiploic vessels is most reliable. A 4-cm gastric tube on the greater curvature gives the best blood supply.190 The simplicity of preparation, adequate length, and robust blood supply makes it the first choice as the esophageal substitute (Fig. 17-15). Disadvantages of the gastric conduit include the fact that patients who have an intrathoracic stomach often experience postprandial discomfort and early satiety related to loss of normal gastric functions such as receptive relaxation. Patients can also suffer from acid reflux, possible gastric ulceration, and dysfunctional propulsion.191 In addition, Barrett's esophagus has been reported to develop in the esophageal remnant,192 although the clinical relevance of this finding is at present unknown. These are important considerations though, in our experience, serious problems are uncommon. The level of the esophagogastric anastomosis has a bearing on the severity of reflux. Patients who have a low intrathoracic anastomosis tend to have more severe reflux and esophagitis compared with the high intrathoracic or cervical anastomosis. Preserving a longer length of esophagus, on the other hand, theoretically may enhance swallowing function. Inadequate gastric emptying can be a problem. A pyloric drainage procedure is not universally practiced. In a randomized trial, 13% of patients who did not have a pyloroplasty had problems with gastric emptying.193 A meta-analysis suggested that a drainage procedure lessens the chance of early postoperative gastric stasis, but long-term function is not affected.194

Many other factors contribute to emptying of the intrathoracic gastric conduit. A smaller stomach enhances postoperative emptying.195 The straighter position of the stomach, when delivered to the neck via the posterior mediastinal or the retrosternal route, may make the stomach empty more efficiently compared to one placed in the right pleural cavity, where the angulation at the diaphragmatic hiatus as the stomach continues from the right paravertebral gutter into the abdomen may produce relative obstruction. Rotation of the stomach at the hiatus should be avoided. With a gastric conduit, diet modifications and the use of acid suppressive and prokinetic drugs such as erythromycin may be useful.196,197

There are instances when the stomach cannot be used, such as after previous gastric resection, and tumor involvement of a substantial part of the stomach dictating its removal. In these situations the use of the colon is preferred. For most, colonic interposition remains an infrequently performed procedure and has the potential for more complications.198 Mobilization of the colonic loop is more complex; its blood supply is less reliable than the gastric conduit; three anastomoses are required; when the colon becomes ischemic, the choice of alternative conduit is restricted. In our experience, use of a colon loop is associated with more blood loss, a longer operating time, and a higher anastomotic leak rate. Colon ischemia occurs in 1 of 42 patients (2.4%), which compares favorably to a rate of 3–10% reported in the literature.199

A colonic conduit provides good long-term swallowing function; it seems to have active peristalsis, and this is cited as an explanation for its superior function as an esophageal substitute when compared with a passive gastric conduit.200,201 Although peristalsis can be demonstrated immediately following surgery,202 long-term emptying likely relies on gravity.203 When the distal stomach is retained in the abdomen after a colon interposition with a cologastric anastomosis, the latter provides additional reservoir function.204

The jejunum is used most frequently after distal esophagectomy and total gastrectomy for cancer of the lower esophagus and gastric cardia. A Roux-en-Y configuration seems best, as it prevents bile reflux to the esophagus. A jejunal loop used in a modified Merendino procedure to interpose between the esophagus and proximal stomach after limited resection of the distal esophagus and GEJ has also been advocated.205 Excellent postoperative quality of life and function is claimed. A long jejunal loop is sometimes used to reach the neck, but preparation is tedious and the vasculature may not be reliable; a “supercharge” using a microvascular anastomosis to cervical vessels may be required.206 A free jejunal graft is used for reconstructing the defect after resection of the pharyngoesophageal segment in the neck.127

The method of reconstruction is in part related to the surgical approach for resection. When a cervical anastomosis is chosen, one must decide whether to place the conduit via the orthotopic, retrosternal, or subcutaneous route. The subcutaneous route is rarely used because it is cosmetically unsightly. The retrosternal route has variably been shown to be associated with increased or similar cardiopulmonary morbidity and mortality rates.207–209 The retrosternal route is 2–3 cm longer compared to the orthotopic route,210 but this is rarely of relevance because the esophageal replacement conduit is usually of sufficient length. Some suggest that the tight space at the thoracic inlet in the neck could cause potential constriction on the conduit and recommend partial manubrial, clavicular head and first rib resection211; we have found this unnecessary. Functionally, although it was shown that there is a higher rate of gastric retention when the retrosternal route is used, quality of life is not adversely affected.208,212

When palliative resection is carried out for advanced tumor, recurrent tumor could infiltrate into the conduit placed in the posterior mediastinum. In a retrospective study of 209 patients who had undergone curative resection and orthotopic reconstruction, or 73 patients (35%) who had locoregional tumor recurrence, 46 (22%) had secondary dysphagia as a result. The authors concluded that in 27 patients (13%) dysphagia would likely have been prevented by using a retrosternal reconstruction route.213 However, the site of the obstruction that produced dysphagia was not clearly stated. The stomach is usually spacious and tumor infiltration will not readily result in dysphagia. Only at the thoracic inlet and in the cervical region, where there is limited space, can tumor involvement lead to obstruction. Using the retrosternal route will eliminate tumor involvement in the posterior mediastinum, but infiltration from tumors in the neck cannot be avoided. The benefits of choosing the retrosternal route in reducing secondary dysphagia from recurrent tumor infiltration may be overemphasized. In our own study, only 4 out of 28 patients (14%) developed tumor infiltration into the gastric conduit in the posterior mediastinum. The main symptom was bleeding in two patients and none had dysphagia.214 It is our policy therefore to only use the retrosternal route for reconstruction when resection is palliative, especially when postoperative radiotherapy is planned, or when the reconstructive phase of the operation precedes tumor resection.

Perioperative Care and Postoperative Morbidity and Mortality

With adequate preoperative workup, serious cardiac events like myocardial infarction should be rare. Atrial arrhythmia is common, affecting about 20% of patients. In itself, atrial fibrillation is benign; rather it serves as a marker for more serious underlying pulmonary and septic surgical complications.215 Occurrence of atrial arrhythmia should prompt thorough search for a more ominous underlying cause.

Pulmonary complications remain the most common and serious postoperative morbidity. Major complications can affect 30% of patients; most series report a rate of about 20%.216 Pneumonia and respiratory failure occurred in 15.9% of our patients and were responsible for 55% of hospital deaths. Predictive factors include advanced age, supracarinal tumor location (in part related to recurrent laryngeal nerve injury), and lengthy operating time. Neoadjuvant therapy did not lead to increased morbidity.112 Measures to improve respiratory outcome include cessation of smoking preoperatively, chest physiotherapy, avoidance of recurrent laryngeal nerve injury, cautious fluid administration to avoid fluid overload, use of smaller chest tubes,217 early ambulation, regular bronchoscopy, and early tracheostomy for sputum retention.218 Epidural analgesia is invaluable in postoperative pain relief and has been shown to improve outcomes.153

The most common surgical complication after esophagectomy is still anastomotic leak and can reach 30%,219 although in experienced centers leak rates of below 5% can be achieved. Most leaks are probably related to technical errors,118,220 such as tension between the conduit and the esophageal stump, ischemia of the conduit because of rough handling and poor preparation, and suboptimal technique. The intrinsic vascular perfusion of the stomach can be enhanced by certain methods, such as “ischemic preconditioning,” whereby partial mobilization of the gastric conduit is followed by a second stage-anastomosis later. The perfusion of the stomach could be shown to improve in the interim period.221 Although an interesting concept and potentially useful, the existing wide range of reported leak rates (from 2–3 to 30%) suggests that much improvement is possible by other means, even without ischemic conditioning. It would be ideal if one could identify the right patients on whom to perform ischemic conditioning pre- or intraoperatively, so that such elaborate preparation can be selectively applied.

The actual method of anastomosis is perhaps less important than its proper application. Stapled anastomosis is popular for intrathoracic anastomosis while the hand-sewn technique is preferred in the neck. There is no evidence from randomized trials that leak rates differ between stapled and hand-sewn anastomoses, but the circular stapler may give rise to more strictures.222 The linear stapler has also been advocated in the neck. One group reduced their cervical leak rate from 10 to 15% using a hand-sewn technique to 2.7% using linear staples with a side-to-side anastomosis.223 With experience, however, the hand-sewn method is as safe, if not more so, and certainly less expensive.

As mentioned already, technical variables play an important role in the genesis of postoperative complications. Anastomotic leaks (largely technical) and recurrent laryngeal nerve injury, for instance, are related to higher incidences of postoperative pulmonary morbidities. At the author's center, pulmonary complications occurred in 10% of patients without technical complications, and in 38% of patients who developed such morbidities, and mortality rates were 3.3 and 9.2%, respectively.224 Multivariate analyses also demonstrated that a long operating time was related to pulmonary complications, and increasing intraoperative blood loss was related to postoperative mortality.112 In sum, the meticulous and expeditious execution of an esophagectomy and its subsequent reconstruction are of paramount importance in lessening complication and mortality rates.

Vigilant and aggressive treatment of complications is important for good outcomes. Management of complications has improved with time. At the author's unit, anastomotic leak rate was 16% in the 1960s to 1970s, 61% of whom died, resulting in a leak-related mortality of 9.8%.225 In the 1980s the leak rate was 3.5%, of whom 35% died, a leak-related mortality of 1.2%,220 while in the late 1990s leak occurred in 3.2% of patients and none died as a result.226

Other surgical complications like chylothorax and herniation of bowel through the diaphragmatic hiatus are rare but should be recognized early; both are corrected by surgical reexploration.

Combined Multimodal Treatment Strategies

The past two decades have seen a proliferation of additional treatments for esophageal cancer. The rationale is based on the suboptimal long-term results of surgery or radiotherapy. Both the spatial and synergistic actions of chemotherapeutic agents and radiotherapy are explored in multimodality treatments. How surgical resection and these new combinations should be integrated into treatment programs is an active area of research.

Neoadjuvant Radiotherapy

Trials of neoadjuvant radiotherapy have failed to show increased resection rate or improved survival compared with surgery alone.227–232 The European Organization for Research and Treatment of Cancer (EORTC) study suggested improved local disease control but no better long-term outcome.229 One study, which also involved chemotherapy, suggested a survival advantage imparted by preoperative radiotherapy but only in the pooled groups of patients receiving radiotherapy.232 A Cochrane meta-analysis showed that if preoperative radiotherapy regimens do improve survival, the effect is likely to be modest with an absolute survival benefit of 3% at 2 years and 4% at 5 years that was not statistically significant (p = .062).233

Adjuvant Radiotherapy

Postoperative radiotherapy was studied in three randomized trials234–236; all three demonstrated improved local disease control. The largest study randomized 495 patients with intrathoracic squamous cell cancers. Postoperative radiotherapy of 50–60 Gy was given in 220 patients to the entire mediastinum and bilateral supraclavicular fossae. Per protocol, analysis showed no overall difference in 5-year survival at 31.7% for the surgery alone group and 41.3% for the radiotherapy group. A benefit in the radiotherapy group was observed in stage III patients; 5-year survival rates were 13.1 and 35.1%, respectively. In patients with node-positive disease, the difference in survival was of borderline significance. The chance of mediastinal, cervical lymph node and anastomotic recurrence was also reduced.236 Survival benefit was not demonstrated for the other trials. From these studies it seems reasonable to give postoperative radiotherapy to subgroups of patients, especially those who have palliative resections, to enhance local disease control.

Neoadjuvant Chemotherapy

Eleven randomized trials studied the role of preoperative chemotherapy.232,237–246 The two largest trials were the Intergroup (INT 0113) trial in the United States, and the Medical Research Council (MRC) trial in the United Kingdom (Table 17-9). The first study randomized patients to undergo surgery alone, or to have three cycles of cisplatin and 5-fluouracil before surgery, and in those who had stable or responsive disease, two additional postoperative courses.245 Of 440 eligible patients, 213 were assigned to the neoadjuvant group. The median survival was 14.9 months for the chemotherapy group compared with 16.1 months for the surgery group. Two-year survival rates were no different at 35 and 37%, respectively. The MRC trial (OE02) involved 802 patients and similar preoperative regimens with two courses of cisplatin and 5-fluouracil.246 Overall survival was better in the chemotherapy group. Median survival was 16.8 versus 13.3 months, and 2-year survival rates were 43 and 34%. Recently the long-term follow-up data were presented; with a median follow-up is of 6 years and 93% of patients followed to 5 years or death, 5-year survival rates were 23% in the chemotherapy group compared with 17% in surgery group. Benefits were evident for both squamous cell cancer and adenocarcinoma.247

Many differences between the two studies could explain the different outcomes, including the chemotherapy regimen, distribution of histologic cell types (66% adenocarcinoma in MRC and 54% in INT trials), number of patients who underwent resection, time to resection, type of surgery performed, and number of patients who also had radiotherapy. The larger sample size in the MRC trial also could have facilitated the detection of a small improvement with chemotherapy.

A Japanese study conducted by the Japanese Clinical Oncology Group (JCOG 9907) randomized 330 patients with stage II/III squamous cell cancers (excluding T4 disease) comparing two courses of preoperative cisplatin and 5-fluorouracil to a similar regimen given after esophagectomy. Overall 5-year survival was significantly better at 60% in the preoperative chemotherapy group compared to 38% in the postoperative group.248 Although this trial did not specifically compare preoperative chemotherapy to surgical resection alone, this has quickly become a standard-of-care treatment in Japan. In the United Kingdom, the MRC OE02 trial has also established preoperative chemotherapy as a widely practiced strategy. Another ongoing trial (OE05) compares the OE02 preoperative chemotherapy regimen with four courses of preoperative epirubicin, cisplatin, and capecitabine (ECX) in treating patients with adenocarcinoma of the esophagus and GEJ. Accrual was planned for 1300 patients.

The Medical Research Council Adjuvant Gastric Infusional Chemotherapy (MAGIC) trial, a randomized study, included 503 patients with adenocarcinoma of the stomach, GEJ, and the lower esophagus. Initially planned for gastric cancers, eligibility criteria were extended to include lower esophageal adenocarcinoma coinciding with termination of OE02. Thus 14% of patients had lower esophageal tumors, and another 12% had GEJ tumors. Three courses of epirubicin, cisplatin, and infused fluorouracil (ECF) were given to patients before surgery, and three courses were repeated afterward, comparing this to patients undergoing surgical resection alone. Both progressive-free and overall survival rates were improved in the chemotherapy group.249

A recent individual patient data-based meta-analysis on nine randomized trials (2102 patients) showed a statistically significant overall survival benefit in favor of preoperative chemotherapy translating into a 5-year absolute increase of 4% (from 16 to 20%). Disease-free survival and curative resection rates were also improved.250 Similar benefits were demonstrated by another meta-analysis, with a 2-year absolute survival benefit of 7%. However, adenocarcinomas may benefit more than squamous cell cancers.251

Adjuvant Chemotherapy

This is an area perhaps least well studied, and trials on pure postoperative chemotherapy are limited. JCOG 9907, mentioned in the previous section, was in fact a follow-up study on JCOG 9204, which randomized 242 patients comparing surgical resection with the addition of two courses of postoperative cisplatin and 5-fluouracil.252 The 5-year disease-free survival rate was significantly different at 45% with surgery alone and 55% with surgery plus chemotherapy. The overall 5-year survival rates were not significantly different at 52 and 61%, respectively. The effect was more marked in the subgroup with lymph nodes metastases.252 However, another small French study also using cisplatin and 5-fluouracil as adjuvant therapy did not show any advantage with chemotherapy.253

Neoadjuvant Chemoradiation

Several groups have explored chemoradiation as neoadjuvant therapy (Table 17-10).232,254–261 The radiation dose ranged from 20 to 45.6 Gy. In five trials, only squamous cell cancers were recruited232,254,255,257,258; three included mostly adenocarcinomas259–261 and one treated adenocarcinomas only.256 A survival advantage with neoadjuvant chemoradiation over surgery alone was demonstrated only in two trials.256,260 The trial reported by Walsh and colleagues on adenocarcinomas only has been criticized because of inadequate preoperative staging, unclear surgical procedures, and the large number of protocol violations, and survival from the surgery group was exceptionally poor (3-year survival rates were 32 and 6% for the preoperative treatment group compared to surgery alone).256 In CALGB 9781, 475 patients were planned, but the trial was terminated after 56 patients because of poor accrual. Nevertheless, a survival advantage was seen in the chemoradiation group; median survival was 4.5 versus 1.8 years and 5-year survival was 39 versus 16%.260 However, the statistical analyses of the trial were much criticized.262

The results from these studies are conflicting and thus inconclusive. Several meta-analyses have addressed the role of neoadjuvant chemoradiation.251,263–267 The latest published meta-analysis included the randomized trials comprehensively; 10 studies included 1209 patients. The hazard ratio for all-cause mortality with neoadjuvant chemoradiation versus surgery alone was 0.81 (95% CI 0.70–0.93; p = .002), corresponding to a 13% absolute difference in survival at 2 years, with similar results for different histological tumor types: 0.84 (0.71–0.99; p = .04) for SCC and 0.75 (0.59–0.95; p = .02) for adenocarcinoma.

Although it cannot be said conclusively that neoadjuvant chemoradiation therapy is superior to surgery alone in the treatment of localized esophageal cancer, it is widely practiced, especially in the United States. Neoadjuvant chemoradiation therapy does result in more pathological complete responses compared with chemotherapy (25–30% vs <10%). One recent trial compared preoperative chemotherapy with preoperative chemoradiation therapy in advanced adenocarcinoma of the lower esophagus and GEJ. More pathological complete responses were observed in the chemoradiation group (16 vs 2%), and more patients had negative nodal involvement (64 vs 38%). A trend toward improved median survival (32.8 vs 21.1 months) and 3-year survival (47.4 vs 27.7%) were also seen, though these did not reach statistical significance.268

Definitive Chemoradiation

The Radiation Therapy Oncology Group (RTOG 85-01) trial of chemoradiation versus radiotherapy provided convincing evidence of the superiority of chemoradiation.269. The 5-year survival rate reported for the combined therapy group was 26% compared to 0% following radiotherapy (median survival 14 vs 9 months). Data on recurrence patterns showed that both local and distant disease control were superior with combined treatment. Local persistence of disease and recurrence were 47% compared to 65%. Intensification of radiation dose to beyond 50.4 Gy, whether by external beam270 or by brachytherapy,271 did not yield further advantage but potentially added complications.

A Cochrane meta-analysis on 13 randomized trials that compared chemoradiation with radiation confirmed the superiority of chemoradiation. Concurrent chemoradiation provides a significant overall reduction in mortality at 1–2 years, an absolute reduction in death rate by 7%, and a reduction in local persistence/recurrence rate by 12%. The downside is a 17% increase in grades 3–4 toxicities. Sequential chemoradiation provides no benefit, perhaps demonstrating the need to maximize the radiosensitizing properties of chemotherapy.272

The Role of Surgery

The RTOG trial suggested that, in patients with T1-3 N0-1 M0 disease a 14–26% 5-year survival can be expected. It has been suggested that surgery may be of no additional value to chemoradiation and should be relegated to use as an adjuvant treatment.

Two clinical trials attempted to examine whether surgical resection was necessary after chemoradiation. A French study (FFCD 9102) treated 444 patients with both squamous cell cancers and adenocarcinomas of stage T3-4 N0-1 M0 with two cycles of 5-fluouracil, cisplatin, and concurrent radiation (46 Gy at 2 Gy/d or split course 15 Gy weeks 1 and 3). Only 259 patients who had at least a partial response were randomized to undergo immediate surgery or to have three more cycles of chemotherapy with 20 Gy at 2 Gy/d or split course 15 Gy. The death rate within 3 months after starting induction treatment was 9% for surgery group compared with 1% in the chemoradiation group. Two-year survival rates were not different at 34 and 40%, so were median survival at 17.7 and 19.3 months for surgical and nonsurgical groups, respectively. Patients in the surgical arm, however, required stenting less often (13 vs 27%) or dilations (22 vs 32%).273 There was no difference in the long-term quality of life, but the surgery arm had transient deterioration in the immediate postoperative period.274

A German multicenter trial recruited 172 patients with squamous cell cancers (T3-4 N0-1 M0). Three cycles of 5-fluouracil/leucovorin/etoposide/cisplatin were given followed by chemoradiation (cisplatin/etoposide + 40 Gy). Resection was then performed. This was compared to a control group with the same chemotherapy, followed by definitive chemoradiation (cisplatin/etoposide + >60 Gy).275 Long-term data from this trial were presented recently.276 A nonsignificant trend toward better overall survival at 5 and 10 years was observed: 27.9 and 19.2% in the resection group, compared to 17.0 and 12.2% in the chemoradiation alone group. Local tumor control was significantly worse in the nonsurgical arm. Three-year survival rate was 35% in nonresponders undergoing complete tumor resection compared to 11% in nonresponders who did not undergo resection. Both the French and German studies concluded that surgical resection may not be necessary after chemoradiation therapy.

It may be premature to negate the value of surgical resection. First, chemoradiation is by no means harmless, and surgical resection may not be as morbid as described. Treatment duration of chemoradiation is often long and compliance is problematic. Only 68% of the patients in the RTOG-8501 trial could complete the planned treatment.269 In the control arm of INT 0123, acute grades 3 and 4 toxicity affected 43 and 26%, respectively, and long-term grades 3 and 4 toxicity affected 24 and 13% of patients, respectively.270 Treatment-related mortality was 5–9% as reported by the INT trials.270,277 In studies that showed a benefit for chemoradiation or questioned the value of surgical resection, the results of the surgical arm were often suboptimal. In the FFCD 9102 trial, death rate within 3 months in the surgical arm was 9% compared to 1% in the nonsurgical arm273; in the German trial again the mortality rates were 10 and 3.5%, respectively.276 The early surgical deaths likely biased the long-term survival results. Comparisons with nonoperative treatments will only be valid when better results from high-volume centers are integrated into clinical trials.

Second, local disease control with chemoradiation alone is less than satisfactory. It can be shown that with increasing extent of lymphadenectomy, better local control is achieved with surgery; by comparison, nonoperative chemoradiation has a much higher local persistence/recurrence rate of over 50%.270 The relief of dysphagia, the main symptom requiring palliation, is much more certain with surgical resection; the need to treat dysphagia with a stent was twice in the nonsurgical group in the FFCD 9102 trial.273

Third, residual disease exists for the majority of patients treated by chemoradiation. The pathological complete response rate for most trials is in the region of 25%. Thus it is logical to assume that surgical resection would enhance cure at least in the remaining 75%, who did not completely respond. In the German trial, the 3-year survival of nonresponding patients who underwent resection was 35% compared with 11% in those who did not.276 In the FFCD 9102 trial, 192 patients were not randomized primarily because of lack of objective response but also because of medical contraindications or patient refusal. Out of these, 112 patients had operations; among these 80 had R0 resection (42%). The median survival for the patients who underwent surgery was 17.3 versus 6.1 months for those who did not, and was comparable for those who were randomized. The data suggest that salvage surgery could benefit a subset of patients who do not respond to initial therapy.278 Conversely, the role of surgery is less obvious in those with a complete response. However, ascertaining true complete response is difficult, whether by endoscopy, EUS, or CT scanning.279,280 Recent studies using 18-FDG-PET scans show promise,82,281 but, while PET scan can more reliably distinguish responders and nonresponders, it is not accurate enough to pinpoint the complete pathological responders.282

Prediction of Response and Response-Directed Therapy

Reliable predictors for response to chemoradiation would be useful, because multimodality treatments are toxic, time consuming, and costly. Various markers have been explored, such as simple histology,283 proliferative cell nuclear antigen (PCNA), epithelial growth factor (EGFR), Ki-67, cyclin D1, thymidylate synthase, and microvessel density, both in tissue and serum. To date none have been proven to help clinical decision making.284

Metabolic imaging with PET scan has some promise. The degree of response detected by PET imaging has been shown by many studies to correlate with pathological response after chemotherapy or chemoradiation therapy (Fig. 17-16).82,281

The MUNICON (the Metabolic response evalUatioN for Individualization of neoadjuvant Chemotherapy in oesOphageal and oesophagogastric adeNocarcinoma) trial evaluated patients with locally advanced adenocarcinoma of the distal esophagus or type II cardia tumors undergoing neoadjuvant chemotherapy. Early metabolic response was defined as a reduction of 35% or more in the mean glucose standard uptake value (SUV) measured by serial PET scans at the beginning and at 2 weeks after commencement of treatment. Responders carried on chemotherapy for an additional 12 weeks before resection, while nonresponders went directly to immediate surgery. Out of 119 patients, 110 were evaluable for metabolic responses, of whom 54 (49%) were responders. Significantly improved R0 resection rate (96 vs 74%), major pathological response rate (defined as <10% residual tumor) (96 vs 0%), longer median event-free survival (29.7 vs 14.1 months), and median overall survival (median not reached versus 25.8 months) were found for metabolic responders versus nonresponders. More importantly, the outcomes for nonresponders were not different from previous results in such patients who completed 3 months of chemotherapy, indicating that such a strategy did not compromise these patients and could save them from suboptimal chemotherapy.285

The same investigators reported on their MUNICON-2 trial recently. Metabolic nonresponders as defined in MUNICON were switched to chemoradiotherapy (both chemotherapy and chemoradiotherapy were cisplatin-based). Out of 32 patients recruited, 13 (41%) were metabolic nonresponders. Subtotal histologic response (<10% residual tumor) following chemoradiotherapy was reported in three patients (23%), but no complete responses was observed. In contrast, complete histological response rate in metabolic responders were seen in 16%. Higher rates of R1/2 resections were also observed for nonresponders (31 vs 16%). One-year progressive-free survival was also inferior (46 vs 63%). The study suggested that merely adding radiotherapy to the same cisplatin-based chemotherapy in nonresponders was only marginally better.286 Another strategy may be to switch to alternative, non–cross-resistant chemotherapy during radiation.

It seems that cisplatin and 5-fluouracil–based chemoradiation therapy has reached its therapeutic limit in treating esophageal cancer. More novel chemotherapeutic agents are being explored, including paclitaxel, docetaxel, the topoisomerase I inhibitor irinotecan (CPT-11), vinorelbine, gemcitabine, Herceptin (trastuzumab), oxaliplatin, and biomodulators such as interferon or targeted therapies with bevacizumab or cetuximab. This remains a very active area of research. In addition, advances in techniques in radiation delivery, such as intensity-modulated radiotherapy, may further reduce radiation toxicity.287

Endoscopic Palliation

Endoscopic palliative treatments for more advanced tumors include placement of an esophageal prosthesis, laser therapy, intralesional injection of various substances, and PDT. The most commonly employed technique is perhaps insertion of a prosthesis, especially self-expanding metallic stents (SEMS) (Fig. 17-17). The smaller diameter of the delivery mechanism makes aggressive dilation of the tumor before insertion unnecessary. These stents are more flexible than conventional plastic prostheses; membrane-covered versions have been developed to seal esophagoairway fistulae and prevent tumor ingrowth. Three randomized trials were reported comparing the use of metallic stents with plastic prostheses. Perforation, pneumonia, bleeding, or migration rates were significantly less with metallic stents. Because of the lower morbidity, metallic stents were also more cost-effective despite their higher initial cost.288–290 The choice of various metallic stents depends on their individual characteristics, in terms of flexibility, tensile force, and degree of shortening on deployment in relation to the site of placement. Compared with more conventional methods of palliation such as laser therapy, patients with SEMS spent less time in the hospital and required less frequent reinterventions.291

The main problems with SEMS are stent migration, tumor ingrowth or overgrowth, and, if placed across the GEJ, they allow acid reflux. Placing uncovered stents across the cardia lessens the chance of migration, and stents have been developed with a one-way flap valve to prevent reflux.292 It has also been shown that “tumor” ingrowth is sometimes due to granulation tissue or hyperplastic reaction by the esophageal mucosa.293 Patency can be achieved again by laser, argon beam application, or sometimes placement of a second stent within the first. One recent randomized trial compared the use of the Ultraflex stent (Boston Scientific, MA) with the Polyflex stent (Boston Scientific, MA), and the Niti-S double stent (Taewoong Medical, Seoul, Korea). The Polyflex stent is a silicone device with an encapsulated monofilament braid made of polyester. The silicone and polyester material is designed to lessen nontumoral tissue overgrowth, a problem common with SEMS. The Niti-S stent has an inner polyurethane layer over its entire length, and an outer uncovered nitinol wire tube to allow the mesh to embed itself in the esophageal wall. Success rates were similar for all three stents, but recurrent dysphagia was more common with the Ultraflex stent, because of tissue ingrowth and overgrowth, and, to lesser degree, the Niti-S stent. Polyflex stent had a higher chance of migration, not surprisingly, because the stent is also designed to be removable in benign esophageal stenosis.294

Another problem of stent insertion is for placement near to the upper esophageal sphincter. Foreign body sensation, pain, odynophagia, and airway compression can be troublesome and demand accurate placement. This is illustrated in the situation when recurrent disease is found at the anastomosis or in the esophageal remnant after subtotal esophagectomy. Placement of SEMS is still possible and achieves good palliation.295

Advances have been made in the management of esophageal cancer; survival of patients has improved.296 The key is to select the most appropriate combination for individual patients. Surgeons play a central role in directing management treatment of this disease by advising on how best to integrate surgical resection with nonoperative programs. Surgeons should aim at improving their results further, so that low mortality rates for resections are used to compare with seemingly safer therapies. The technique and extent of surgical resection may change when more information is made available, and should vary with patients and disease stage. MIE will be more widely practiced; it should achieve the same radicality of operation with less morbidity. Chemoradiation therapy has made a real impact on current management strategies,296 but perhaps its overenthusiastic adoption and its presumed benefit have to be balanced against the lack of clear evidence of superiority over surgery.297 Distant failure remains a major problem, and search for more effective systemic drugs as well as our ability to predict responders with precision must be therapeutic targets. Management strategies are going to evolve further, with improvements in molecular techniques, imaging methods, and introduction of more novel tumoricidal agents. The challenge for the future is for us to critically test our strategies in a scientific, unbiased manner, and to explore other innovative treatments.