Implementing the first program of minimally invasive esophagectomy for cancer in Israel: shifting the paradigm in a high-volume center – a cohort study

Abstract

Background:

Esophagectomy is associated with significant perioperative morbidity. Limited data are available on the process of implementation of minimally invasive techniques in esophagectomy and related outcomes. The authors sought to describe implementation processes and outcomes following the implementation of the first minimally invasive esophagectomy (MIE) program at a high-volume center in Israel under the mentorship of American early adopters.

Methods:

Patients who underwent esophagectomy for esophageal carcinoma 2011–2022 were included. Early and late cohorts were created for learning curve analysis. Secondary analysis included patients who underwent open esophagectomy 1997–2011.

Results:

Overall, 300 patients underwent MIE: three-field MIE (3F-MIE) was performed in 242 (80.7%) patients, two-field MIE (2F-MIE) in 58 (19.3%) patients. Following program implementation in 2012, the number of MIE performed increased during the first 3 years (n=33, 86.8% in 2015). Among 3F-MIE patients, a higher number of retrieved lymph nodes was reported during later cases (median, IQR1–3 17, 12–23 vs. 12, 8–12, P<0.001) while surgeries required a longer time (median, IQR1–3 300 min, 261–355 vs. 262.5, 239–300, P<0.001). Among 2F-MIE patients, the late cohort had lower rates of prolonged ICU admissions than earlier counterparts (n=2, 6.9% vs. n=9, 31%, P=0.041), overall and severe 30-day complications (n=12, 41.4% vs. n=23, 79.3%, P<0.001 and n=7, 24.1% vs. n=23, 79.3%, P=0.003).

Conclusions:

MIE was safely implemented. Nodal yield was higher among MIE patients than open esophagectomy. During the study years, open approach was gradually abandoned in favor of 3F-MIE procedures, while 2F-MIE increased over the course of the last years.

Introduction

More than one hundred years after the first subtotal transthoracic esophageal resection for esophageal carcinoma was successfully performed by Torek¹, esophagectomy remains among the most challenging and complex surgical procedures. A decade after Torek’s operation, Ohsawa in Japan performed the first intrathoracic anastomosis, and briefly thereafter Turner described the transhiatal approach².

While the adoption of multimodality treatment to esophageal and gastroesophageal junction carcinoma led to significant improvements in oncologic outcomes³, improving rates of postoperative complications is still challenging. In 2002, a retrospective analysis from a nationwide registry showed that 46% of patients suffered from major postoperative complications⁴. More recently, significant improvements in short-term outcomes have been demonstrated following the implementation of minimally invasive access techniques in esophageal resection⁵. However, available data on the subject are still scarce and highly variable in terms of both surgical techniques and results⁶.

In 2012, a team led by one of the authors (R.B.), one of the earliest adopters and developers of minimally invasive techniques in esophageal surgery^7,8, visited our institution and performed the first cases of esophagectomy with a minimally invasive approach. Prior to this visit, most of our patients were treated with transhiatal esophagectomy⁹. This mentorship resulted in the development and implementation of a minimally invasive esophagectomy (MIE) program at our hospital.

Aim

We sought to outline our framework for the implementation of a MIE program at a single, high-volume institution (tertiary academic medical center). Additionally, we sought to report changing trends in surgical techniques during the study years and differences in surgical and oncologic outcomes across an early and late cohort.

Materials and methods

Study design and patient population

This work has been reported in line with the STROCSS (Strengthening The Reporting Of Cohort Studies in Surgery) criteria¹⁰ (Supplemental Digital Content 1, http://links.lww.com/JS9/A923). A retrospective analysis of a prospective cohort of patients who underwent MIE between January 2011 and May 2022 (Institutional Review Board protocol 0772-19) was performed. Inclusion criteria were: patients age 18 years and older; primary cancer of the esophagus and esophagogastric junction (EGJ). Exclusion criteria were: benign disease; urgent esophagectomy; recurrent cancer; open esophagectomy.

Patients who underwent three-field MIE (3F-MIE) and two-field MIE (2F-MIE) were each divided into two cohorts based on early (first half of cases) and late (second half) experience for learning curve analysis.

Secondary analysis included patients who underwent open esophagectomy January 1997–January 2011.

Preoperative staging and treatment

Preoperative workup included clinical examination, biopsies, endoscopic ultrasonography (EUS), and computerized tomography (CT) or 18F-fluorodeoxyglucose positron emission tomography (PET-CT). Clinical staging was reported according to the 8th TNM classification¹¹.

Following multidisciplinary discussion, neoadjuvant treatment was usually administered to patients with locally advanced disease. Chemoradiation typically involved induction chemotherapy with cisplatin (90  ${\mathrm{mg}/\mathrm{m}}^2$ on day 1), and 5-fluorouracil (1000  ${\mathrm{mg}/\mathrm{m}}^2$ between days 1 and 5) followed by cisplatin (75  ${\mathrm{mg}/\mathrm{m}}^2$ ) and 5-fluorouracil (1000  ${\mathrm{mg}/\mathrm{m}}^2$ ) every 28 days with concomitant radiotherapy (total 50 Gy), or carboplatin (area under the curve = 2 mg/ml/min) and paclitaxel (50  ${\mathrm{mg}/\mathrm{m}}^2$ ) weekly for 5 consecutive weeks with concomitant radiotherapy (total 41 Gy) per CROSS (ChemoRadiotherapy for Oesophageal cancer followed by Surgery Study) protocol¹². Included in the chemotherapy regimen were the FLOT protocol, a combination of 5-fluorouracil infusion (2600  ${\mathrm{mg}/\mathrm{m}}^2$ for 24 h), oxaliplatin (85  ${\mathrm{mg}/\mathrm{m}}^2)$ and docetaxel (50  ${\mathrm{mg}/\mathrm{m}}^2$ ) every 2 weeks for a total of four courses¹³, and the ECF protocol, including epirubicin (50  ${\mathrm{mg}/\mathrm{m}}^2$ ) and cisplatin (60  ${\mathrm{mg}/\mathrm{m}}^2$ ) every 3 weeks, with continuous infusion of 5-fluorouracil (200  ${\mathrm{mg}/\mathrm{m}}^2$ ) per day¹⁴. Patients receiving modifications of the above protocols or experimental treatment were also included.

Surgical techniques

During the study period, the decision to adopt a minimally invasive versus an open approach was left to the surgeon’s discretion. A minimally invasive surgical approach was defined according to a recent consensus statement by Cerfolio et al.⁶.

A minimally invasive three-field esophagectomy (3F-MIE) was performed in the right chest cavity with the patient initially positioned on left lateral decubitus. The thoracic esophagus was mobilized thoracoscopically while dissecting paraesophageal and mediastinal lymph nodes with a Harmonic scalpel (Ethicon Endo-Surgery Inc., Cincinnati, Ohio, USA). The azygos vein was divided using an EndoGIA linear vascular stapler (Medtronic, Dublin, Ireland). The patient was then placed in a supine position. The stomach was mobilized laparoscopically with special care given to the preservation of the right gastroepiploic arcade. Short gastric vessels were divided with an ultrasonic scalpel, and the left coronary vein and left gastric artery were ligated with clips or linear stapler. The left neck was incised medially to the sternocleidomastoid muscle. Following the division of its cervical portion, the esophagus was delivered through a 5 cm midline incision. A gastric conduit was created with multiple EndoGIA firings. The gastric tube was subsequently mobilized to the neck through the posterior mediastinum. An esophagogastric anastomosis was performed either mechanically in an end-to-side fashion using a circular stapler (EEA 25, Medtronic, Dublin, Ireland; or CDH 25, Ethicon Endo-Surgery Inc., Cincinnati, Ohio, USA) or manually in an end-to-end fashion.

Minimally invasive two-field esophagectomy (2F-MIE) entailed laparoscopic mobilization of the stomach in the same manner as above, followed by thoracoscopic mobilization of the esophagus and mediastinal nodal dissection through a right thoracoscopic approach. An end-to-side stapled esophagogastrostomy was usually performed either thoracoscopically or through limited extension of one of the thoracic port incisions (incision length: 4–8 cm). The specimen was delivered through this same incision. We routinely added a feeding jejunostomy tube.

Postoperative management

Starting 2021, our program adopted an internally revised version of the Enhanced Recovery After Surgery (ERAS) recommendations¹⁵ (Supplemental Digital Content 2, http://links.lww.com/JS9/A924). Patients underwent early extubation, either in the operating room (OR) or in the post-anesthesia care unit (PACU). Following overnight stay in the PACU, patients were admitted to the surgical floor on postoperative day (POD) 1. The need for admission to the surgical ICU or step-down unit was evaluated on a case-by-case basis according to each patient’s status during and immediately after surgery. All patients underwent early respiratory rehabilitation and mobilization, starting in the PACU and continuing on the surgical floor afterward. Feeding was introduced enterally (via a jejunostomy) starting POD 1. The nasogastric tube was removed on POD 2 and oral clear fluids were reinstituted. Following 3F-MIE, the chest drain was usually removed on POD 2 and the cervical drain on POD 6 or prior to discharge. Following 2F-MIE, the chest drain was removed on POD 6 or prior to discharge. The jejunostomy was removed 4 weeks after surgery.

Data regarding major postoperative complications were collected according to the esophagectomy complication consensus group (ECCG)¹⁶ and categorized according to the Clavien–Dindo classification¹⁷. Leakage was diagnosed on clinical suspicion and further confirmed radiologically or surgically. Among clinically stable 3F-MIE patients, confirmation of anastomotic leakage included bedside exploration of the cervical wound with findings of purulent discharge. Dysphagia was assessed clinically both during the inpatient and outpatient phases of care. Patients who developed dysphagia subsequently underwent barium studies and subsequent endoscopic examination in case of abnormal findings or persistent symptoms.

Pathological assessment and long-term follow-up

Pathologic staging was reported according to the 8th TNM classification¹⁸. Long-term follow-up included physical examination, CT or PET-CT scans at a 3-month or 6-month interval at least during the first 2 years, followed by at least 3 more years of yearly visits and imaging.

Statistical analysis

Data were analyzed with the Statistical Program of Social Sciences (SPSS version 22, Chicago, IL). Student t-test and Mann–Whitney test were used to compare continuous variables, and χ ² test or Fisher’s exact test was used to compare categorical variables. Categorical data were expressed as percentages and continuous data were expressed as means. Overall (OS) and disease-free survival (DFS) were estimated using Kaplan–Meier estimators, and comparisons among the different curves were achieved using the logrank test. Survival endpoints included date of death (OS) and date of disease recurrence or death (DFS). Whenever missing, data among demographic and preoperative variables [ASA (American Society of Anesthesiology) scores, preoperative histology, location at endoscopy, clinical staging] were not imputed and were thus omitted from the analysis.

Results

Study population and perioperative characteristics

Overall, 565 patients underwent esophagectomy during the study years. Twenty patients (3.5%) had benign diagnoses. Among 545 patients diagnosed with esophageal carcinoma, 245 (43.4%) underwent open esophagectomy, while 300 (55%) underwent MIE and were included in the study (Fig. 1). Median age in the study population was 66 (IQR1-360-72), and most patients were males (n=122, 70.7%). Most tumors were adenocarcinomas (n=218, 72.7%) located either at the gastroesophageal junction or distal esophagus (n=235, 78.4%). Most patients presented locally advanced disease, including 216 (72%) patients with clinical stage III or IVA. Neoadjuvant treatment was administered to 240 (80%) patients, mostly according to the CROSS protocol (n=145, 48.3%) (Table 1).

Figure 1.

Flowchart of the study.

Table 1.

Demographic and preoperative characteristics.

Variables	Overall, n=300
Age, median, years (IQR)	66 (60–72)
Gender, Female	88 (29.3)
ASA
1	44 (14.7)
2	135 (45)
3	91 (30.3)
4	6 (2)
Missing	24 (8)
Histology
AC	218 (72.7)
SCC	80 (26.7)
Missing	2 (0.7)
Location
GEJ/distal esophagus	235 (78.4)
Mid-esophagus	45 (15)
Proximal esophagus	3 (1)
Missing	17 (5.7)
Clinical stage
I	36 (12)
II	42 (14)
III	213 (71)
IVA	3 (1)
Missing	6 (2)
CEA (mg/dl), median (IQR)	2.1 (1.4–5.7)
CA 19-9 (mg/dl), median (IQR)	13.9 (7.3–32.1)
Neoadjuvant treatment
CROSS	145 (48.3)
Cisplatin, 5-FU, RTx	21 (7)
ECF/ECX	5 (1.7)
FLOT	34 (11.3)
Other/Clinical trial	35 (11.7)
Neoadjuvant radiation	197 (65.7)

All data are n (%) unless otherwise specified.

AC, adenocarcinoma; ASA, American Society of Anesthesiology; CA 19-9, carbohydrate antigen 19-9; CEA, carcinoembryonic antigen; GEJ, gastroesophageal junction; IQR, interquartile range; SCC, squamous cell carcinoma.

Perioperative variables are described in Table 2. Three-field MIE was performed in 242 (80.7%) patients. Two-field MIE was performed in 58 (19.3%) patients. Overall, 17 (5.7%) patients required conversion to open procedure.

Table 2.

Intraoperative and perioperative characteristics.

Variables	Overall, n=300
Type of surgery
Three-field	242 (80.7)
Two-field	58 (19.3)
Conversion	17 (5.7)
Type of anastomosis
Stapled	256 (88.7)
Handsewn	34 (11.3)
Operative time (min), median (IQR)	292 (246.5–345)
Length of stay (days), median (IQR)	9.5 (8–18)
Intensive care admission	57 (19)
Overall, any 30-day complications	173 (57.6)
30-day ECCG classification
Gastrointestinal	87 (29)
Pulmonary	62 (20.6)
Cardiac	25 (8.3)
Thrombotic	6 (2)
Urologic	43 (14.3)
Infective	24 (8)
Neurologic/psychiatric	11 (3.6)
Wound/diaphragmatic	27 (9)
Other	27 (9)
Clavien–Dindo
1	16 (5.3)
2	40 (13.3)
3	64 (21.3)
4	36 (12)
5	12 (4)
Leak	55 (18.3)
Ischemic conduit	6 (2)
Chylothorax	1 (5.7)
Reoperation	28 (9.3)
Recurrent laryngeal nerve paralysis	10 (3.3)
Any dysphagia at follow-up	72 (24)
Index admission dysphagia	5 (1.7)

All data are n (%) unless otherwise specified.

ECCG, esophageal complication consensus group; IQR, interquartile range.

The number of 3F-MIE increased during the first 3 study years: n=6 (23.1%) 3F-MIE, n=0 2F-MIE, n=20 (76.9%) open in year 2012; n=33 (86.8%) 3F-MIE, n=5 (13.2%) 2F-MIE, n=0 open in year 2015. The number of 2F-MIE increased in the following years, from n=5 (13.2%) in 2015 to n=11 (32.4%) in 2017–2018 and n=15 (34.9%) in 2021–2022. During the first 6 months of 2022, more patients underwent 2F-MIE than 3F-MIE (n=8, 75%) (Fig. 2). Out of n=35 handsewn anastomoses, n=24 (68.6%) were performed within the first 3 study years.

Figure 2.

Trends in MIE, 2011–2022.

Postoperatively, the median length of stay was 9.5 days (IQR 8–18), and 57 (19%) patients required prolonged ICU admission. Major complications occurred in 121 (40.3%) patients. Twelve (4%) patients died during the admission. Leak was detected in 55 (18.3%) patients, and ischemia of the gastric conduit in 6 (2%). Twenty-eight (9.3%) patients required reoperation. Rates of anastomotic leak were comparable among patients who received or did not receive neoadjuvant radiation (n=37, 18% vs. n=18, 18.5%, respectively, P=0.947). At follow-up, dysphagia was reported by 72 (24%) patients, and five (1.7%) patients had dysphagia during index admission requiring endoscopic dilation.

Learning curve analysis

On a learning curve analysis including patients who underwent 3F-MIE, patients from the late cohort had longer operative time (median minutes, IQR1–3 300, 261–355 vs. 262.5, 239–300, P<0.001) and higher lymph node yield than those from the early cohort (median, IQR1–3 n=17, 12–23 vs. n=11, 8–20, P<0.001). Overall and severe complication rates were comparable among cohorts (n=75, 62.5% vs. n=63, 52.9%, P=0.151 and n=42, 43% vs. n=44, 36.4%, P=0.358, respectively). Recurrent laryngeal nerve paralysis was more frequent in the late cohort (n=8, 7% vs. n=1, 0.8%, P=0.011) (Table 3). Among patients who underwent 2F-MIE, the late cohort had a significantly lower frequency of prolonged ICU admissions (n=2, 6.9% vs. n=9, 31%, P=0.041). Rates of overall and severe complications were lower among patients from the late cohort (n=12, 41.4% vs. n=23, 79.3%, P<0.001 and n=7, 24.1% vs. n=18, 62.1%, P=0.003) (Table 4).

Table 3.

Learning curve variables, early versus late cohort, 3F-MIE (n=242).

Variables	Early cohort, n=121	Late cohort, n=121	P
Operative time (minutes), median (IQR)	262.5 (239–300)	300 (261–355)	<0.001
Length of stay (days), median (IQR)	9 (7–15)	11 (8–20)	0.101
Lymph node yield	12 (8–12)	17 (12–23)	<0.001
Prolonged ICU admission	20 (16.5)	26 (21.5)	0.413
30-day complications	63 (52.9)	75 (62.5)	0.151
Severe complications (Clavien–Dindo >2)	44 (36.4)	52 (43)	0.358
Pulmonary complications	28 (23.1)	23 (19)	0.529
Leak	18 (14.8)	24 (19.8)	0.239
Recurrent laryngeal nerve paralysis	1 (0.8)	8 (7)	0.011
30-day mortality	4 (3.3)	4 (3.3)	1

Table 4.

Learning curve variables, early versus late cohort, 2F-MIE (n=58).

Variables	Early cohort, n=29	Late cohort, n=29	P
Operative time (minutes), median (IQR)	347 (309–399.5)	366 (321–405)	0.588
Length of stay (days), median (IQR)	10 (8–16)	8 (7–10)	0.128
Lymph node yield, median (IQR)	15 (12–27)	19 (13.7–27)	0.161
Prolonged ICU admission	9 (31)	2 (6.9)	0.041
30-day complications	23 (79.3)	12 (41.4)	<0.001
Severe complications (Clavien–Dindo >2)	18 (62.1)	7 (24.1)	0.003
Pulmonary complications	7 (24.1)	4 (13.8)	0.504
Leak	8 (27.6)	5 (17.2)	0.349
Recurrent laryngeal nerve paralysis	0	1 (3.4)	1
30-day mortality	3 (5.2)	1 (3.4)	0.611

All data are n (%) unless otherwise specified. P values <0.05 are considered significant.

ICU, intensive care unit; IQR, interquartile range.

Survival analysis

At a median follow-up period of 20 months (interquartile range, IQR1–3 2–36 months), median OS was 40 months (95% confidence interval, CI 32.5–47.5 months) and OS at 1-year, 3-year, and 5-year follow-up was 78.1%, 54.6%, and 41.5%. Overall, 106 (35.3%) patients recurred, median DFS was 29 months (95% CI 20.3–37.7 months) and DFS at 1-year, 3-year, and 5-year follow-up was 68.5%, 45.7%, and 34.9%. No differences in OS or DFS were observed when comparing 2F-MIE and 3F-MIE (P=0.877 and P=0.485, respectively) (Figs 3A and 3B).

Figure 3.

Kaplan–Meier curves for overall survival (OS) (A) and disease-free survival (DFS) (B) comparing two-field (2F-MIE) and three-field (3F-MIE) minimally invasive esophagectomy.

Secondary analysis, open versus MIE

Overall, 245 patients underwent open esophagectomy from 1997 to 2011 and were compared to their 300 MIE counterparts, as shown in Table 5. The distribution of clinical staging and histology were comparable (P=0.340 and P=0.845, respectively). Patients who underwent MIE received neoadjuvant treatment more frequently (n=241, 80.3% vs. n=114, 46.5%, P<0.001). The frequency of cervical versus thoracic anastomosis was similar among MIE and open esophagectomy patients (cervical anastomosis, n=242, 80.7% vs. n=208, 84.9%, P=0.241). Leak rates were comparable among the open and MIE cohorts (n=55, 18.3% vs. n=32, 14.1%, P=0.192) and when compared across patients who received or did not receive neoadjuvant treatment (n=59, 17.3% vs. n=28, 15.6%, respectively, P=0.711). Incidence of leak rate, R status, and median nodal yield stratified by two-field versus three-field approaches and MIE versus open esophagectomy are reported in Supplementary Table 1 (Supplemental Digital Content 3, http://links.lww.com/JS9/A925). On univariate logistic regression, neoadjuvant treatment was not associated with higher rates of leak. Patients who underwent MIE had higher lymph node yield (median, IQR1–3 14.5, 10–21 vs. 12, 7–17, P<0.001). At a median follow-up of 22.5 months (IQR1–3 7.75–44 months), median OS was comparable between MIE and open esophagectomy patients (40 months 95% CI 32.5–47.5 vs. 29 months, 95% CI 20.2–37.8 months, P=0.198) (Supplementary Fig. 1, Supplemental Digital Content 4, http://links.lww.com/JS9/A926).

Table 5.

Secondary analysis: comparison of select outcomes, open versus minimally invasive esophagectomy.

Variables	Open, n=245	MIE, n=300	P
Gender, Female	69 (28.2)	91 (30.3)	0.637
Age, median (IQR)	66 (57–74)	66.4 (59.9–72.1)	0.831
Histology			0.845
AC	182 (74.3)	218 (73.2)
SCC	63 (25.7)	80 (26.8)
cTNM			0.340
I	32 (15.9)	36 (12.2)
II	29 (14.4)	42 (14.3)
III	140 (69.7)	213 (72.4)
IVA	0	3 (1)
Neoadjuvant treatment	114 (46.5)	241 (80.3)	<0.001
Type of anastomosis			0.241
Cervical	208 (84.9)	242 (80.7)
Thoracic	37 (15.1)	58 (19.3)
Leak	32 (14.1)	55 (18.3)	0.192
R status, negative	204 (91.5)	239 (92.6)	0.735
LN yield, median (IQR)	12 (7–17)	14.5 (10–21)	<0.001

All data are n (%) unless otherwise specified. P values <0.05 are considered significant.

AC, adenocarcinoma; IQR, interquartile range; LN, lymph node; MIE, minimally invasive esophagectomy; SCC, squamous cell carcinoma.

Discussion

In ‘The Structure of Scientific Revolutions’ from 1962, Thomas Kuhn, physician and philosopher, described scientific progress as recurrent periods of stable, incremental evolution (normal science) followed by sudden, abrupt scientific revolutions (paradigm shifts)¹⁹. If we look at the history of general surgery in these terms, the development and swift adoption of minimally invasive techniques would be the single, most important paradigm shift in modern times. More than 30 years passed since the first laparoscopic cholecystectomy was reported²⁰. Over the course of the last decades, the advantages of minimally invasive techniques in treating different abdominal and thoracic malignancies have been demonstrated in a number of studies presenting Level I evidence^21–24. Esophageal surgery and, specifically, esophageal resections, however, lagged behind this trend, and the adoption of laparoscopy and thoracoscopy has been a somewhat slower process. Only at the beginning of the last decade did new evidence support the introduction of minimal access techniques in esophageal resections⁵. Advantages in short-term outcomes have been shown in multiple studies, and potential benefits on survival have been recently reported as well^25–27. However, data from the largest U.S. surgical database showed that between the years 2008 and 2017, more than 80% of esophagectomies were still being performed with an open approach²⁷. Therefore, data on the implementation of MIE programs and their outcomes provides evidence-based grounds to support the adoption of minimally invasive techniques as the preferred approach for esophageal resections while offering ‘real-world’ practical suggestions on the processes needed to achieve safe results. The series of patients presented in this paper describes the successful adoption of minimal access techniques in a single, high-volume center and provides an overview of our experience facing substantial changes in surgical practice. While reporting on the learning curve effect on short-term outcomes among MIE patients, we also compare our outcomes to a previous cohort of open esophagectomy patients.

Prior to program initiation, the increasing adoption of minimally invasive techniques in general surgery and encouraging results from emerging literature⁵ led to the decision to establish an MIE program at our institution. A team of physicians and nursing staff was created to head start the program, including two attending surgeons (H.K. and N.M.) who had performed more than one hundred transhiatal esophagectomies⁹, a dedicated anesthesiologist, and a scrub nurse. Our vision at that time was to perform most elective esophagectomies minimally invasively within the first 3 years. In 2012, a team led by one of the coauthors (R.B.) was invited to our institution with the purpose of initiating a MIE mentorship program. At that time, the mentoring team had more than 10 years of experience in performing MIE⁷. The visiting team included two attending surgeons, a dedicated anesthesiologist, an OR nurse-in-charge, and a surgical technologist. During their week-long visit, the team completed the first two MIE cases in Israel while providing progressive autonomy to the local surgical and nursing staff during the second case. The rising number of MIE performed during the first 3 years was positively perceived and generated confidence in the program. Indeed, during the first 6 months of 2022, more patients underwent 2F-MIE than 3F-MIE. With the aim of further standardizing processes and reinforcing implemented changes, several strategies were developed to improve outcomes and prevent stepbacks:

(1) A dedicated Foregut Unit was created in 2019, including four attending surgeons, one surgical intensivist, one junior resident, and one senior resident; (2) a protocolized, multidisciplinary approach to the perioperative management of patients following MIE was developed in 2020, based on current recommendations from the Enhanced Recovery After Surgery Society¹⁵ (Appendix, Supplemental Digital Content 4, http://links.lww.com/JS9/A964); and (3) a systematic approach to MIE training during surgical residency was developed locally, promoting early exposure to MIE, with one junior resident rotating in the foregut unit, while senior residents and junior attendings are gradually allowed to perform limited stages of the procedure including mobilization of the stomach, abdominal lymph node dissection, feeding jejunostomy insertion, mediastinal dissection and, eventually, anastomosis.

Retrospective data comparing outcomes following MIE and open esophagectomy is growing. In terms of survival, we failed to detect a statistically significant difference in OS and DFS among the study groups. Given that two recently published analyses from nationwide databases did demonstrate that MIE was independently associated with improved OS, even when adjusting for ERAS protocol (Supplemental Digital Content 2, http://links.lww.com/JS9/A924) implementation and modern neoadjuvant chemoradiation protocols^25,26, our results might be due to an underpowered analysis. In this sense, the improved nodal retrieval reported in the MIE cohort appears to support a survival benefit in this group of patients. Regarding short-term outcomes, older studies demonstrated limited benefits in short-term outcomes with MIE²⁸. However, more recent data has since challenged those findings. Indeed, newly published data from a large national database reported that MIE was associated with lower levels of overall morbidity. Specifically, respiratory complications were seemingly affected to a greater extent by the adoption of minimal access surgery²⁷, a result that has been confirmed in additional data from a single-center retrospective analysis²⁹. While we were unable to perform a direct comparison on the incidence of pulmonary complications between MIE and open esophagectomies, the incidence of pulmonary complications following MIE in our analysis (20%) is similar to the rates reported in the aforementioned studies. Level I data on postoperative outcomes appears to confirm this trend. Recently published results from the MIRO trial demonstrated better early postoperative outcomes in patients undergoing hybrid two-field esophagectomy than those undergoing open esophagectomy³⁰. The TIME trial, published more than 8 years ago, is the only randomized controlled study reporting results from a direct comparison of MIE to open techniques³¹. In this benchmark trial, patients undergoing MIE had fewer pulmonary complications (9% vs. 29%), shorter length of stay (11 vs. 14 days), and better patient-reported outcome measures than patients undergoing open esophagectomy. Recently, investigators from the TIME trial reported their experience in mentoring different institutions abroad³². We acknowledge that the mentorship received from the mentoring team was of foremost importance in the initiation and development of our program. While two initial cases proved important in providing a baseline for a step-to-step approach to MIE, the impact of such a limited number of cases was unlikely to effectively shorten a potential learning curve. However, our institution featured the favorable conditions outlined in the abovementioned paper: (i) a high volume of patients referred for open esophageal resection and (ii) a team of two experienced surgeons fully dedicated to the implementation of the MIE program³². These factors likely contributed to the steep increase in the number of MIE performed one year after program initiation, with perioperative and long-term outcomes overall similar to historic controls^33,34. Moreover, the initial encounter between our two surgical teams provided benefits beyond technical aspects. We recognize that there is an important human factor in establishing mentorship. Although only for one week, this face-to-face, hands-on approach paved the way for further collaboration, ultimately leading to hosting years-long fellowship programs for Israeli foregut surgery trainees.

Importantly, this report reinforces the idea of surgery as an everchanging practice, showing the need to adapt approaches and techniques once new data challenges previous routines. At our institution, prior to the introduction of the MIE program, esophagectomy was performed with a transhiatal approach, preferred to Ivor Lewis esophagectomy due to concerns regarding the severe sequelae of intrathoracic anastomotic leak over the more limited morbidity of cervical anastomotic dehiscence³⁵. Therefore, with gradual implementation of minimally invasive access in our practice, we felt that cervical anastomoses still provided the safest complication profile, and 3F-MIE became a natural progression of our previous experience. However, recent evidence challenged our rationale. A recent analysis of a nationwide database reported cervical anastomosis as an independent predictor of anastomotic leakage³⁶. Moreover, the alleged increased morbidity associated with thoracic anastomotic leak has been disproved in randomized controlled trials^37,38. This growing evidence favoring intrathoracic anastomosis was the main driver for change in our practice, resulting in a tailored approach based on tumor location: 3F-MIE was performed for Siewert 1 and distal esophageal tumors, while 2F-MIE was usually reserved for Siewert 2 malignancies. We are now performing an increasing number of 2F-MIE for EGJ malignancies regardless of Siewert classification, as demonstrated by 2022 data shown in Figure 1. Anastomotic techniques also reflected the dynamic nature of the surgical practice, with most handsewn anastomoses being performed over the first 3 study years, with later years showing a more consistent adoption of stapled anastomosis. This is likely consistent with increased standardization of surgical techniques with growing experience.

Among various limitations to the widespread adoption of MIE, a challenging learning curve stands at the forefront. Indeed, a recent review of available literature described a range of 20–175 cases as a learning curve for 3F-MIE and 40–54 for 2F-MIE³⁹. Claassenet al.⁴⁰ showed that more than 10% of 2F-MIE patients experience learning curve-associated anastomotic leakage. Moreover, results from a propensity score-matched analysis reported increased anastomotic leaks and the need for reintervention among patients undergoing 2F-MIE, which the authors attributed to the long learning curve⁴¹. Concerningly, higher rates of gastric conduit necrosis have been described during initial experience with MIE as well⁴². We failed to detect increased anastomotic leaks among early 3F-MIE cases, and no study to date demonstrated a significant learning curve effect in this population³⁹. In our series, the overall leak rate was 18%, the incidence of ischemic conduit was 2%, and perioperative mortality was 4%. While remaining higher than recent data from the ECCG¹⁶ and single-center studies⁴³, similar rates have been reported in multicentric databases⁴⁴ as well as prospective data¹². We, therefore, investigated whether our leak rate might have been secondary to the administration of neoadjuvant treatment and, more specifically, chemoradiation. Unfortunately, data on specific protocols adopted among the open esophagectomy patients’ cohort was missing. An analysis of the administration of neoadjuvant treatment revealed no significant association with leak rate in the open and MIE populations. Among MIE patients only, anastomotic leak rates were also comparable among patients who received or did not receive neoadjuvant chemoradiation. This further confirms the limited impact of neoadjuvant treatment on short-term postoperative outcomes⁴⁵. Notably, mortality rates decreased from 4.6 to 3.3% among patients from the later cohorts. Although the relatively high morbidity might be secondary to the high number of 3F-MIE performed and to a rather inclusive definition of the leak, these numbers should not be ignored nor downplayed. Low et al.⁴³ previously attributed their remarkably low leak rate to standardized clinical pathways. In this context, we have adopted the recently published ERAS guidelines¹⁵ and developed an Institutional protocol striving to improve standardization of care and surgical outcomes. Low rates of early dysphagia requiring intervention confirmed that early introduction of oral nutrition, a controversial item among the ERAS guidelines was overall well tolerated. Regarding 2F-MIE, our analysis was limited by a low number of patients and by the relatively recent implementation of this technique. While patients from the early cohort presented high rates of postoperative complications, our late cohort experienced improved perioperative outcomes, similar to previous reports^33,46. Lastly, 3F-MIE patients from the late cohort were more likely to experience recurrent laryngeal nerve injury. Unexpectedly, our rate of 7% is almost two-fold than the data included in the ECCG database¹⁶. Since a higher lymph node yield and longer operative times were found among 3F-MIE patients from the late cohort, this finding could represent a consequence of progressively more radical nodal dissection with growing experience. The high rate of recurrent laryngeal nerve injury might also be associated with extended dissection along the left main bronchus. Given higher rates found among 3F-MIE patients than their 2F-MIE counterparts, exposure to the left recurrent nerve during cervical dissection might have further contributed to this increase. We do recognize that a higher lymph node yield might be the result of changes in the pathologic examination of the specimens. Interestingly, OR back-table lymph node station dissection prior to submission of specimen to pathology has become a routine since 2019 and might have been a concurring factor in the increase in lymph node yield seen in the late cohort. We believe that reporting the experience from a single center, reducing bias resulting from different surgical techniques and postoperative management, may strengthen results from our analysis since all patients were operated on and followed by the same surgical team during the entire study period. However, different perioperative and neoadjuvant protocols were adopted during the study period, resulting in significant bias. Specifically, patients who underwent open procedures belonged to earlier years than MIE patients and, therefore, were less likely to receive neoadjuvant chemoradiation protocols, which might have exerted a significant bias on long-term outcomes. While potentially eliminating this confounder, the low number of open cases performed during the study years prevented us from performing a direct comparison to patients from the MIE cohort. The decision to adopt a minimally invasive or open approach was also left to each surgeon’s preference, generating additional selection bias. Moreover, prior to the adoption of the ERAS guidelines (Supplemental Digital Content 2, http://links.lww.com/JS9/A924), no institutional protocol was available, with perioperative management left to each attending’s discretion. To some degree, these factors might account for the differences in perioperative outcomes (e.g. need for ICU stay). Lastly, since this series includes data from a tertiary referral center, a large number of patients received adjuvant treatment at outside locations, and this variable was not included.

Conclusions

Minimally invasive techniques of esophagectomy were successfully and safely implemented in a high-volume center. The nodal yield was higher among MIE patients than their open esophagectomy counterparts. During the study years, the open approach was gradually abandoned in favor of 3F-MIE procedures, while an increase in the number of 2F-MIE is reported over the course of the last years, reflecting the dynamic nature of the surgical practice.

Ethical approval

Rabin Medical Center’s Institutional Review Board, Protocol no. 0772-19.

Sources of funding

No direct funding was obtained for this research work.

Author contribution

D.S.: study design, data collection, data analysis, and writing; E.S.: data collection and writing; N.M. and V.B.: writing; R.B. and H.K.: study design and writing.

Conflicts of interest disclosure

Dr Bueno receives grants from NCI, NIH, DOD, Merck, Roche, Genentech, Verastem, Siemens, North Pond, Momentaro, Novocure, Epizyme, and Bayer through the Institution. None of the other authors has interests to declare.

Research registration unique identifying code (UIN)

1. Name of the registry: www.researchregistry.com.

2. Unique identifying number or registration ID: researchregistry8501.

3. Hyperlink to your specific registration (must be publicly accessible and will be checked): https://www.researchregistry.com/browse-theregistry#home/registrationdetails/637bebde9d887c002289216d/.

Guarantor

Daniel Solomon and Hanoch Kashtan.

Data availability statement

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Provenance and peer review

Not commissioned, externally peer-reviewed.

Supplementary Material

js9-109-3467-s001.docx ^{(31.3KB, docx)}

js9-109-3467-s002.docx ^{(323.5KB, docx)}

js9-109-3467-s003.docx ^{(14KB, docx)}

js9-109-3467-s004.docx ^{(299.2KB, docx)}

js9-109-3467-s005.docx ^{(323.5KB, docx)}