Respiratory failure is a major problem after esophagectomy. Several studies have reported that about half of the inhospital deaths after esophagectomy is due to pneumonia, which is the most frequent general complication after surgery [89]. Preventive measures include preoperative respiratory training, cessation of smoking, and continuous postoperative pain control by epidural analgesia in order to avoid restrictive respiration and insufficient coughing. Micro-aspiration as a consequence of impaired swallowing coordination because of a cervical anastomosis also plays a role in the pathophysiology of bronchopneumonias. Another reason for postoperative respiratory impairment is a large pleural effusion, which should be drained if provoking extended atelectasis. Avoiding the need for a combined thoracotomy and laparotomy may potentially reduce postoperative pain, ventilator dependence, and cardiopulmonary complications [90]. In a study comparing thoracoscopic resection with a historical cohort, the overall incidence of pulmonary complications was reduced from 33 to 20 % [91]. Probably cardiopulmonary complications do not depend on the incision size only. The benefit of smaller port sites that are needed during minimally invasive surgery may be offset by the lengthened time of operation and single-lung ventilation. The use of a prone position also plays a role but will be discussed elsewhere.

More recurrent laryngeal nerve injuries when using thoracoscopy have been reported, which might be attributed to the use of diathermia. Others claim that the use of minimally invasive techniques has lowered the incidence of hoarseness because of the magnified view [87].

Lack of standardization of definitions is a problem when reporting on complications. In a recent meta-analysis, anastomotic leakage was reported in most of the publications, but it was defined in only a minority with 22 differing definitions [84]. Early disruption of the esophagogastric anastomosis is the result of a technical problem and immediate reexploration is frequently indicated for correction. Many different suturing and (semi-)mechanical techniques have been described. The semimechanical side-by-side technique claims a lower leakage rate compared to a hand-sewn anastomosis, but has not been tested in a randomized trial [92, 93]. Leakage is more frequent in the neck than in the chest, but the associated mortality might be lower, especially after a transhiatal approach [94]. If a transmural necrosis of the gastric conduit is suspected, this can be diagnosed by endoscopy and when present is also an indication for surgery with formation of a cervical esophagostomy, resection of the gastric tube, and placement of a feeding jejunostomy. After rehabilitation of the patient, a colonic interposition can be performed at a secondary stage. Late disruptions become manifest generally between postoperative day 5 and 10 and are most frequently due to ischemia. They can be managed nonoperatively in most cases with aggressive drainage using radiologically guided drains or endoluminal vacuum therapy [95]. Self-expandable stents can be inserted in these situations but can have the disadvantage of migration or further necrosis due to tissue compression ultimately leading to, e.g., neoesophago-tracheal fistula formation.

The incidence of accidental thoracic duct leakage can be diminished by intraoperative identification and ligation of the duct. Reported incidence of chylothorax varies between 3 and 10 % and is seen more often in patients who undergo transthoracic esophagectomy and in patients who have more positive nodes. Patients with chyle leakage have more pulmonary complications. Conservative therapy (initial parenteral feeding and subsequent enteral diet with medium-chain triglycerides (MCT)) is often successful, but operative therapy should be seriously considered in patients with a persistently high daily output of more than 2 L after 2 days of optimal conservative therapy [96].

Cardiac arrhythmias are not uncommon in the postoperative phase. Atrial fibrillation (AF) is seen in 15–20 % of patients and requires further investigation because it can be an early manifestation of, e.g., mediastinitis due to intrathoracic anastomotic leakage. AF can also be associated with hypervolemia, preexistent pulmonary or cardiac disease, and dilation of the gastric conduit.

There is an increasing interest in comparing institutional performance. For surgical procedures postoperative mortality rate is generally used, because it is a relatively objective measure and reflects the summation of the most severe postoperative complications. Currently it is unclear which definition of postoperative mortality best reflects surgical quality of care. The 30-day operative mortality (30DM) and the inhospital mortality (IHM) after esophageal resection are well documented and vary from 4 % for specialized centers to > 10 % for nationwide registries [97]. Few studies report on mortality beyond 30 days. Damhuis et al. however showed in the Dutch Cancer Registry that 43 % of inhospital deaths after surgery for esophageal cancer occurred 30 days or more after the operation [98]. Therefore, 90-day mortality (90DM) might be preferred as a performance indicator. Using a longer time period after the operation for defining postoperative mortality may thus provide a better definition of quality of surgery [99]. Extending the mortality period beyond 30 days and beyond inhospital stay has the advantage that patients who die because of surgery-related complications outside the hospital are included as well.

Not only short-term outcomes but also long-term survival should be part of the benchmark as both aspects are relevant for comparing surgical performance. Both surgery-related deaths and cancer recurrence-related deaths are reflections of surgical quality of care. Less radical surgical resections will generally result in lower postoperative morbidity and mortality but will generally give less favorable oncological outcomes.

Even after agreement on a uniform definition of postoperative mortality, direct comparison of crude mortality rates between hospitals can be misleading as they do not take into account the case-mix difference, i.e., the differences in physiological condition and tumor stages of patients. Sophisticated models have been developed for prediction of 30DM and IHM [8, 14, 67, 100–104] after esophageal surgery, but models for 90DM have been mostly based on large multi-institutional databases with only few parameters available [105].

Over the past decades, better long-term survival results have been presented, evolving from 18 % 5-year survival in the era from 1980 to 1990 to 48 % in the most recently published RCT (Table 4.1) [17, 65, 99, 106, 107]. It is suggested that many factors are responsible for this positive effect, including large hospital volume, early tumor detection, improved patient selection based on novel staging modalities, increased use of neoadjuvant therapy, better surgical and anesthesiological techniques, and improved standardized perioperative clinical pathways [18, 108]. In many countries around the world, it has been decided that high-risk surgical procedures such as esophagectomy should be restricted to facilities with a yearly minimum volume [109, 110]. It has been demonstrated that the incidence of postoperative complications is similar across hospitals but that the associated mortality rates are lowest in high-volume centers, which generally show a lower “failure to rescue” [86, 111]. Centralization is currently implemented widely. Also auditing has been implemented as a way of improvement of care. Of course this results in an additional registration burden for the surgeon, but comparing individual or institutional results with the benchmark has proven valuable in other types of cancer surgery, such as for rectal cancer [112, 113]. For esophageal cancer, variables of interest are, for example, hospital mortality, radicality (R-status), extent of lymph node dissection, length of hospital stay, application of neoadjuvant therapy, availability of PET-CT, and the presence of a well-structured MDT. The quality indicators can be divided in structural, process, and outcome measures, respectively (Table 4.2) [114]. Heterogeneity and lack of standardized definitions of the outcome of interest are a problem here as well. In a review of esophagectomy outcomes from 164 NSQIP (National Surgical Quality Improvement Project) hospitals, it was demonstrated that even following case mix adjustment, results between centers varied by 161 % for 30-day mortality and 84 % for serious morbidity [67].

Finally, comparing the quality of infrequent operations such as esophagectomies is difficult, besides issues of definition and case-mix correction, because of another complex element in comparing surgical performance, i.e., the problem of sample size [115].