Transthoracic Versus Transhiatal Esophagectomy for Esophageal Cancer: A Nationwide Propensity Score-Matched Cohort Analysis

Data were obtained from the Dutch Upper GI Cancer Audit (DUCA). This audit was initiated in 2011 and is part of the Dutch Institute for Clinical Auditing (DICA). In the Netherlands, caregivers are obligated to register all patients with esophageal or gastric cancer with intended resection in the DUCA database. The main goal of this audit was to provide transparent information on the quality of care. Validation of completeness and accuracy of data registration is performed as has been described in earlier publications.7 Because the audit data are available anonymously, it is not possible to retrospectively retrieve missing data or include variables, such as surgical procedural data, hospital of treatment, 90-day mortality, or survival, outside the scope of the audit. This study was approved by the scientific committee of the DUCA. No informed consent or ethical approval was required under Dutch law. All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975 and later versions.

This study investigated the short-term outcomes of transthoracic and transhiatal esophagectomy for cancer in a nationwide propensity score matched analysis. The results show that a transthoracic approach provides a more thorough oncologic resection with a higher lymph node yield, at the cost of increased morbidity and short-term mortality. This is a population-based study, with all the variations in treatment this entails, giving a reflection of actual daily practice in the Netherlands. By utilizing a national database, we were able to study a much larger group of patients than would have otherwise been possible through a randomized, controlled trial. Furthermore, the present study reflects the results of the surgical treatment of esophageal cancer on a nationwide level compared with various publications, including only results from specialized tertiary centers. Our analysis therefore resembles real-world results more closely. However, this resemblance results in discrepancies compared with the guidelines: some patients with a mid-esophageal carcinoma were treated by THE, while the national guideline advises a transthoracic approach. Because this study also includes patients from smaller low-volume centers, the outcomes from our analysis also show, for example, higher anastomotic leakage and mortality rates and a lower lymph node yield compared with studies that only report outcomes of a single, specialized center or exclusively of high-volume (tertiary) centers.

The inclusion period of this study starts in 2011, which was the year of initiation of the DUCA. We know from earlier research that surgical care for esophageal cancer in the Netherlands has significantly evolved since 2011. The two largest changes in our country during the inclusion period of this study were the introduction of centralization of care with a minimum yearly hospital case-volume of 20 cases per year and the introduction of minimally invasive procedures.7^,11 The process of implementing the minimum volume per hospital is still in progress. In the Netherlands in 2016, 22 hospitals performed esophagectomies for esophageal cancer.12 Five of these performed less than 20 resections in that year, three performed 20–29 resections, five performed 30-39 resections, and the remaining nine performed 40 or more resections in that year. This means that in 2016, five hospitals did not meet the minimum volume set in the national guidelines. The number of cases per surgeon is not recorded in the audit. Through the matching process we aimed to correct this potential bias by including the surgical approach and year of surgery as covariates. This means that at the start of the cohort, most minimally invasive procedures could not be matched due to the smaller volume of minimally invasive procedures, whereas in the more recent years of the cohort, many open esophagectomies could not be matched due to the smaller volume of open resections.

TTE provided a superior lymph node yield over THE, accompanied by a higher (y)pN stage compared with patients who underwent a THE. The results also show that 50% of the transhiatal resections had a lymph node yield lower than 14. In the case of a transthoracic resection, only 25% of patients had a lymph node yield below the national guideline of 15. Even though the number of positive nodes was equal in both groups, the question arises whether positive nodes have been missed in the lymph node dissection during transhiatal surgery. Because the cN stage was comparable between groups, although borderline matched (SMD 0.100), a selection bias causing patients with a higher cN to be more likely to undergo a TTE seems an unlikely explanation. The surgical community is divided on the value of an extended lymph node dissection after neoadjuvant chemoradiation. Noordman et al. concluded from their study with the CROSS cohort, that nCRT for esophageal adenocarcinomas might reduce the need for an extended lymphadenectomy, as can be performed with a transthoracic resection.13 A transthoracic resection was independently associated with a more favorable prognosis in the surgery alone group, whereas a TTE with additional nCRT was not. In addition to this, Kurokawa et al. prospectively investigated the distribution of lymph node metastases in gastroesophageal junction tumors and found that a limited lymph node resection could be sufficient in patients with tumors involving less than 4 cm of the esophagus.14 Another study investigating whether a subgroup of patients could benefit from conservative management following neoadjuvant therapy is currently ongoing in the Netherlands.15 On the contrary, a recent study found an association between an extensive lymph node dissection during esophagectomy and prolonged survival.16 Furthermore, a recent study by Raja et al. on post-neoadjuvant esophageal resection found that resecting up to 25 lymph nodes in ypN0 tumors or resecting up to 30 lymph nodes in ypN + tumors resulted in increased survival.17 The presence of positive nodes after nCRT has been associated with survival, which makes lymph node dissection essential for determining the prognosis.18 Additionally, TTE has been linked to a higher 5-year survival compared with THE in the case of positive nodes in the resection specimen, making the risk of potentially missed positive nodes after THE even more relevant.5

Our analysis showed that patients treated by a transthoracic esophagectomy experienced more chyle leakages, pulmonary, and cardiac complications. Additionally, they needed more reinterventions and had a longer ICU and hospital stay. The number of recurrent nerve lesions of any severity was comparable between TTE versus THE. The THE group received tube feeding more frequently, also after propensity score matching. This difference could be explained by differences in treatment protocols between hospitals. Unfortunately, information on hospital of treatment and treatment protocol is lacking in the DUCA database, so this theory cannot be confirmed by data in this study. The short-term mortality (in-hospital/30-day mortality) was significantly higher in the transthoracic group. The authors suspect that despite a comparable anastomotic leakage rate, the lower mortality in the THE group may be explained by less severe manifestations of leakage. As a result of reduced surgical pleural dissection, especially in the upper mediastinum, any leakage will likely result in limited mediastinal manifestations. This hypothesis is supported by earlier research showing a reduced incidence of intrathoracic complications of anastomotic leakage after a THE (27%) compared with a TTE with cervical anastomosis (44%) with similar incidence of anastomotic leakage between groups.19

Many publications regarding this subject suffer from bias since patients undergoing a THE generally differ significantly from those undergoing a TTE. Although some centers prefer one of these approaches for all their patients, most studies show evidence of selective allocation to the procedures based on preoperative condition, comorbidities of the patient, and cTNM staging. The current study shows that the known increased morbidity and mortality after a transthoracic approach for esophagectomy can be expected even in patients matched on baseline characteristics. A randomized trial comparing total gastrectomy via an abdominal-transhiatal approach versus a left thoracoabdominal approach found results similar to this study: more complications in the group with the extended approach.20 In addition, survival did not improve with the more aggressive approach in this randomized, controlled trial. Results however, cannot directly be extrapolated to the patients in this study, because this randomized, controlled trial compares gastrectomy for cardia and subcardia cancer and not esophagectomy for esophageal and gastroesophageal junction cancer.

The generalization of our results is reduced by the fact that we excluded hybrid procedures and nongastric-tube reconstructions after esophagectomy. Performing a subgroup analysis after propensity score matching is statistically unfeasible.21 We have explored the possibility of a 3-arm propensity score matching analysis (TTE with intrathoracic anastomosis, TTE with cervical anastomosis, and THE). However, because propensity score matching discarded nonmatched subjects, this led to very small groups no longer representative of the original cohort. In addition, because only the transthoracic group could be divided in an intrathoracic and cervical anastomosis group, correcting for this confounder was not possible. Because the audit does not disclose the hospital of origin of patients, we were unable to compare the two groups regarding the relation between hospital volume and complications, mortality, and pathology outcome. Because the results of this study may be influenced by the results of individual hospitals or by hospital volume, this is a limitation of the current study. In addition, evaluation of large cohorts of patients could lead to an inherent selection bias: the choice of the surgery type can be made based on experience, principle but also by necessity. Apart from this possible selection bias, propensity score matching does not correct for unknown confounders, and as such residual bias may be present. As stated in the Methods section, DUCA only registers outcomes during the first 30 postoperative days. Because of this, long-term outcomes are not available for analysis. Data regarding lymph node yield per region was only recently added to the audit, and subject to change, and therefore not included. Additionally, we do not have any information on location of suspicious lymph nodes, which could have led to a TTE for oncologic reasons, nor on information regarding surgical preference. Survival data are not part of the DUCA registration: this would have increased the value of the mortality analysis. The current study is nonetheless of great value because of the large number of patients included and correction for the often-reported selection bias through propensity score matching on baseline characteristics.

Our analysis showed that, even after correction for baseline characteristics, a transthoracic approach provides a higher lymph node yield, at the cost of increased morbidity and short-term mortality. The lower lymph node yield after a transhiatal resection could indicate positive lymph nodes left in situ. Although results in high-volume centers and RCTs often are superior, these data reflect the national performance. We believe future research should investigate further whether long-term survival differs between a transthoracic and transhiatal resection in the era of (neo)adjuvant therapy, minimally invasive surgery, and increasingly centralized care.