The Impact of Paratracheal Lymphadenectomy on Lymph Node Yield and Short-Term Outcomes in Esophagectomy for Cancer: A Nation-Wide Propensity Score-Matched Analysis

Introduction: The balance between potential oncological merits and surgical risks is unclear for the additional step of performing paratracheal lymphadenectomy during esophagectomy for cancer. This study aimed to investigate the impact of paratracheal lymphadenectomy on lymph node yield and short-term outcomes in patients who underwent this procedure in the Netherlands. Methods: Patients who underwent neoadjuvant chemoradiotherapy followed by transthoracic esophagectomy were included from the Dutch Upper Gastrointestinal Cancer Audit (DUCA). After propensity score matching Ivor Lewis and McKeown approaches separately, lymph node yield and short-term outcomes were compared between patients who underwent paratracheal lymphadenectomy versus patients who did not. Results: Between 2011 and 2017, 2,128 patients were included. Some 770 patients (n = 385 vs. n = 385) and 516 patients (n = 258 vs. n = 258) were matched for the Ivor Lewis and McKeown approaches, respectively. Paratracheal lymphadenectomy was associated with a higher lymph node yield in Ivor Lewis (23 vs. 19 nodes, p < 0.001) and McKeown (21 vs. 19 nodes, p = 0.015) esophagectomy. There were no significant differences in complications or mortality. After Ivor Lewis esophagectomy, paratracheal lymphadenectomy was associated with longer length of stay (12 vs. 11 days, p < 0.048). After McKeown esophagectomy, paratracheal lymphadenectomy was associated with more re-interventions (30% vs. 18%, p = 0.002). Conclusions: Paratracheal lymphadenectomy resulted in a higher lymph node yield but also in longer length of stay after Ivor Lewis and more re-interventions following McKeown esophagectomy.


Introduction
Esophagectomy is the mainstay of curatively intended treatment for locally advanced esophageal cancer, with a 5-year survival rate of 40-50% when preceded by neoadjuvant therapy [1]. Lymphadenectomy is an integral part of an esophagectomy procedure, aiming to remove potential lymph node metastases and thereby increase the likelihood of curation [2]. The importance of lymphadenectomy during esophagectomy was emphasized by a recent study that demonstrated increased survival rates when at least 15 lymph nodes were retrieved in esophageal cancer patients [3]. Furthermore, a meta-analysis on this topic found that high lymph node yield is associated with improved overall and disease-free survival [4].
Aiming for an adequate lymphadenectomy during esophagectomy, most Western surgeons currently prefer a dissection of both abdominal and mediastinal lymph node stations (i.e., two-field lymphadenectomy) [5]. However, no consensus exists with regard to the exact mediastinal lymph node stations that should be routinely dissected [6]. Paratracheal lymphadenectomy is a particular subject of debate, as its potential oncological advantages must be weighed against the possible risk of iatrogenic damage to surrounding structures, especially the recurrent laryngeal nerves. In this context, some surgeons prefer to remove the paratracheal lymph nodes only in selected patients with proximal esophageal tumors or preoperatively identified lymph node metastases in the upper mediastinum [7]. Nonetheless, the entire peri-esophageal lymph node network is at risk for metastases, and several studies indicate that paratracheal lymph node metastases can be found in a substantial part of patients with distal esophageal tumors, even after neoadjuvant therapy [8,9]. Routine paratracheal lymphadenectomy, irrespective of the location of the primary tumor, is therefore desirable from an oncological perspective. On the contrary, efforts to avoid recurrent laryngeal nerve injury may be justified, as this complication increases the risk of aspiration pneumonia and frequently requires a re-intervention [10,11]. As insight is lacking with regard to the impact of paratracheal lymphadenectomy on oncological outcomes and postoperative morbidity, the aim of this study was to investigate the impact of paratracheal lymphadenectomy on lymph node yield and short-term outcomes following transthoracic esophagectomy for esophageal cancer.

Study Design
For this nation-wide population-based cohort study, patients were identified from the Dutch Upper Gastrointestinal Cancer Audit (DUCA) registry. Since 2011, Dutch hospitals are required to register all esophageal, gastroesophageal junction, and gastric cancer patients who undergo surgery with the intent of performing a resection in the DUCA registry. The scientific committee of the DUCA approved this study, and the need for written informed consent was waived.

DUCA Data Collection
The data that are collected by DUCA include patient and treatment characteristics, as well as the postoperative outcomes during postoperative hospitalization and until 30 days after surgery. The type of mediastinal lymphadenectomy is captured in the DUCA by registering whether or not a lymph node dissection was performed high in the mediastinum (i.e., paratracheal, along the recurrent laryngeal nerves) and low in the mediastinum (i.e., subcarinal, para-esophageal).

Patient Population
Esophageal cancer patients (cT1-4aN0-3M0) who received neoadjuvant chemoradiotherapy followed by elective transthoracic esophagectomy with two-field lymphadenectomy and gastric conduit reconstruction between 2011 and 2017 were included. Neoadjuvant chemoradiotherapy was typically provided according to the CROSS regimen [1]. Patients who did not undergo at least a dissection of the low mediastinal lymph node stations during esophagectomy were excluded. Patients were also excluded when insufficient data regarding the type of lymphadenectomy were available.

Outcome Measures
The primary endpoints included lymph node yield (i.e., total number and number of tumor-positive lymph nodes based on pathological examination), pathological lymph node staging after neoadjuvant therapy (i.e., ypN stage), radicality of the resection (i.e., R0, R1-2), recurrent laryngeal nerve injury (i.e., diagnosis based on either clinical symptoms or laryngoscopy before publication of the Esophagectomy Complications Consensus Group (ECCG) definitions in 2015, based on direct laryngoscopic examination in the later years), overall postoperative complication rate, and mortality (i.e., in-hospital or within 30 days after surgery). Other outcome measures included anastomotic leakage (i.e., clinically or radiologically diagnosed), chylothorax, pulmonary complications (i.e., pneumonia, pleural effusion, pneumothorax, atelectasis requiring bronchoscopy, acute aspiration, respiratory failure requiring re-intubation, acute respiratory distress syndrome, tracheobronchial injury), re-interventions (i.e., radiological re-intervention including the placement of drains under ultrasound, endoscopic re-interventions including drain placement or stenting, or re-operations), length of hospital stay, and re-admissions (within 30 days after hospital discharge).
Complications and re-interventions were registered during the postoperative hospitalization and until 30 days after esophagectomy in the DUCA registry.

Statistical Analyses
The study population was analyzed in two separate groups: (1) patients who underwent Ivor Lewis esophagectomy (i.e., transthoracic esophagectomy with an intrathoracic anastomosis) and (2) patients who underwent McKeown esophagectomy (i.e., transthoracic esophagectomy with a cervical anastomosis). Missing values were encountered in 10 variables, with a maximum of missing values per variable limited to 5.1%. Missing data could not be recovered because no patient and hospital identifiers are provided with data from the DUCA registry. Missing data were therefore considered at random and handled using multiple imputation via chained equations (20 iterations) [12,13].
After multiple imputation, propensity score matching was performed separately in both groups to balance the differences in patient and treatment characteristics between patients who underwent paratracheal lymphadenectomy and patients who did not. Propensity scores were calculated based on a logistic regression model that included the patient and treatment characteristics that are demonstrated in Tables 1 and 3. One-to-one propensity score matching was performed with a nearest-neighbor algorithm, allowing a maximum tolerated difference between propensity scores (caliper width) of no larger than 20% of the standard deviation of the estimated propensity score [14]. The balance in patient and treatment characteristics was assessed by using standardized mean differences (SMD) both before and after propensity score matching. An SMD between <0.1 was considered to indicate a proper balance [15].
After propensity score matching, the primary and secondary endpoints were compared between both groups by means of χ 2 tests (for categorical data), the Student's t test (for continuous data with a normal distribution), or Mann-Whitney U test (for continuous data with a nonnormal distribution), as appropriate. All statistical analyses were performed using SPSS version 23.0 (IBM Corp., Armonk, NY, USA) and R 3.1.2 open-source software (http://www.R-project.org: "mice," "matchit," and "optmatch" packages). A two-sided p value <0.05 was considered statistically significant.

Patient Population
A total of 2,896 patients who underwent neoadjuvant chemoradiotherapy followed by elective transthoracic esophagectomy with two-field lymphadenectomy and gastric conduit reconstruction were identified from the DUCA registry. Patients who underwent esophagectomy in the year 2015 were excluded (n = 573) because the extent of lymphadenectomy was not registered in that year. Furthermore, patients who did not receive at least subcarinal and para-esophageal lymphadenectomy (n = 68) or with insufficient registered data regarding their lymphadenectomy (n = 127) were excluded. Consequently, 2128 patients were included for propensity score matching (Fig. 1).

Ivor Lewis Esophagectomy
Esophagectomy was performed by an Ivor Lewis approach in 1,062 patients, and a paratracheal lymphadenectomy was performed in 431 of these patients (41%). After propensity score matching, 770 patients (385 in each group) remained to be evaluated. The distribution of patient and treatment characteristics before and after matching is demonstrated for both groups in Table 1.

Discussion
In this nation-wide propensity score-matched analysis of patients who underwent transthoracic esophagectomy, a higher total lymph node yield was achieved in patients who received a paratracheal lymphadenectomy when compared to patients who did not, while no significant differences were found regarding the rates of mortality and postoperative complications such as recurrent laryngeal nerve injury. Furthermore, more metastatic lymph nodes were harvested in patients who underwent paratracheal lymphadenectomy during McKeown esophagectomy, which led to more advanced pathological nodal staging in this group. However, paratracheal lymphadenectomy was also associated with more re-interventions after McKeown esophagectomy and a 1-day increase in the median length of hospital stay after Ivor Lewis esophagectomy.
Paratracheal lymphadenectomy requires a meticulous dissection along the recurrent laryngeal nerves, which can be technically challenging despite thorough knowledge of the anatomical area [16]. One might therefore expect that the procedure carries an increased risk of damaging these adjacent nervous structures. Nevertheless, in line with the findings of an Asian study that included patients who underwent esophagectomy by a McKeown approach [17], the current study did not find significantly higher incidences of recurrent laryngeal nerve injury or other postoperative complications when a paratracheal lymphadenectomy was performed. Based on the current results, paratracheal lymphadenectomy seems to increase the number of lymph nodes harvested without compromising the safety of the procedure, which are important findings that contribute to the ongoing debate on the necessity of performing a paratracheal lymphadenectomy. The more advanced pathological lymph node staging that was found in patients who underwent paratracheal lymphadenectomy is likely the result of selection bias, as some centers in the Netherlands only remove paratracheal lymph node stations in patients with PET-positive lymph nodes in the upper mediastinum. On the other hand, one could consider that a higher total lymph node yield may result in the detection of more tumor-positive lymph nodes and thereby higher pathological staging, a phenomenon that is known as stage migration [18]. Although the current study is unfortunately unable to provide a definite explanation for these findings due to a lack of data on long-term follow-up, survival, and recurrence, the key question to be answered next is whether routine removal of paratracheal lymph nodes can also improve disease-free and overall survival after esophagectomy. In a previous study that included patients who underwent esophagectomy for distal esophageal cancer with nodal involvement, clinical staging by means of (positron emission tomography) computed tomography (CT) or endoscopic ultrasound (EUS) identified positive paratracheal lymph nodes in 10% of patients [19]. However, the accuracy of clinical nodal staging is notoriously poor [20][21][22]. Although recent initiatives such as the TIGER trial (NCT03222895) will likely provide more insight into the distribution of lymph node metastases in esophageal cancer, accurate preoperative detection of paratracheal lymph node metastases is impossible at this moment. This is an argument for standardized paratracheal lymphadenectomy in patients with esophageal cancer that is located in any part of the esophagus and gastroesophageal junction, which  maximizes the likelihood of removing all present lymph node metastases. Studies from Asia suggest that paratracheal lymphadenectomy in fact has a high therapeutic value to optimize the survival perspectives after esophagectomy, although these findings might not be generalizable to the Western world due to considerable differences in patient demographics [8]. Future studies investigating potential survival benefits should consider that harvesting more lymph nodes may lead to stage migration [18], which might make stage-by-stage comparisons between different types of lymphadenectomy challenging when evaluating survival. In order to accurately investigate the impact of resecting specific lymph node stations, it would be necessary to know the exact locations of tumor-positive lymph nodes, and studies with such detailed data are warranted.
Although not associated with increased complication and mortality rates, paratracheal lymphadenectomy was associated with a significantly higher re-intervention rate in patients who underwent a McKeown esophagectomy in this study. One possible explanation for this phenomenon could be that cervical anastomotic leakage might reach the mediastinum more easily in case the paratracheal lymph nodes have been removed, leading to intrathoracic manifestations that are known to be associated with more intensive treatment, a longer length of hospital stay, and higher in-hospital mortality [23,24]. Based on additional analyses that were performed in the subgroup of patients who suffered anastomotic leakage after McKeown esophagectomy in this study, it appears that patients who underwent paratracheal lymphadenectomy might have had more re-interventions and a longer length of hospital stay when compared to patients in whom the paratracheal lymph nodes were left in place. However, proper statistical analyses could unfortunately not be performed due to uncertainty regarding potentially present confounding co-variables in this subgroup of patients. Furthermore, as center bias cannot be ruled out in this study, another possibility could be that centers that performed paratracheal lymphadenectomy were also more aggressive in their complication management. Follow-up studies on more in-depth data are required to investigate this hypothesis further.
The foremost strength of this study is the inclusion of a nation-wide cohort that is representative of daily practice in the Netherlands during the inclusion period. As this study aimed to evaluate short-term clinical (mainly recurrent laryngeal nerve injury) and oncological outcomes (mainly total lymph node yield), it was considered most important to first divide the cohort into patients undergoing Ivor Lewis and patients undergoing McKeown esophagectomy. Although factors such as tumor histology and location were expected to have less impact on our primary outcomes, propensity score matching was additionally used to further minimize bias arising from differences in confounding factors between groups. Yet, it should be noted that only known factors can be considered in propensity score matching. Potential unknown confounders that led to residual bias can therefore not be ruled out. As variation between centers might be present with regard to the perioperative outcomes, correction for center would have been desirable. However, DUCA data are anonymous and cannot be traced back to specific centers. Center bias in this study might therefore be a possible explanation for the increased length of hospital stay that was found in the group of patients who underwent paratracheal lymphadenectomy during Ivor Lewis esophagectomy, since the postoperative complication rates were comparable. In addition, the level of detail was limited for several outcome measures. For example, no information was available regarding as the exact side (i.e., right, left, or both) and resected stations (i.e., stations 2, 4, or both) in patients who underwent paratracheal lymphadenectomy, which may have caused some variation in this regard. Moreover, as the extent of lymphadenectomy was not registered in 2015, patients who underwent surgery in that year could not be included in this study. Also, only the total lymph node yield and the total number of positive lymph nodes are registered in the DUCA registry, which means that no information was available on the exact locations of (positive) resected lymph nodes. Lastly, the type of re-intervention was grouped into three categories (i.e., radiological reintervention, endoscopic re-intervention, or re-operation). As a result, no detailed analyses could be performed in an attempt to explain the higher reintervention rate that was observed in patients who underwent paratracheal lymphadenectomy during McKeown esophagectomy.

Conclusion
Paratracheal lymphadenectomy increases the number of lymph nodes harvested, which likely improves the accuracy of pathological nodal staging in esophagectomy for cancer. In Ivor Lewis esophagectomy, paratracheal lymphadenectomy was not associated with more complications. However, paratracheal lymphadenectomy was associated with a higher re-intervention rate in McKeown esophagectomy and with a slightly increased length of hospital stay in Ivor Lewis esophagectomy. Variation in complication management amongst the including centers might explain these findings, as the higher re-intervention rate and increased length of hospital stay could not be explained by complication rates that were actually similar for patients undergoing paratracheal lymphadenectomy and those who did not. On the other hand, the addition of paratracheal lymphadenectomy might lead to more overall morbidity following esophagectomy. Further studies looking into the impact of paratracheal lymphadenectomy on survival are required to investigate whether this operative step should be a routine part of an esophagectomy procedure for esophageal cancer.

Statement of Ethics
This study was performed with a fully anonymized national dataset that was provided by the DUCA. Ethical approval was granted by the Scientific Committee of the DUCA. Therefore, written informed consent from individual patients was not applicable for this study.

Conflicts of Interests Statement
R. van Hillegersberg and J.P. Ruurda are proctors for Intuitive Surgical Inc. There are no other potential conflicts of interests.

Funding Sources
No funding was received for this study.

Author Contributions
B.F. Kingma: design of the study, data analysis, writing of the manuscript, and submission. E.R.C Hagens: design of the study, data analysis, and writing of the manuscript. M.I. van Berge Henegouwen and J.P. Ruurda: expert input and co-writing of the manuscript. A.S. Borggreve: data analysis and co-writing of the manuscript. R. Van Hillegersberg: project leader, expert input, and co-writing of the manuscript.

Data Availability Statement
The data of this study are not publicly available. However, upon reasonable request, data might be shared.