Comparison of the modified Collard and hand‐sewn anastomosis for cervical esophagogastric anastomosis after esophagectomy in esophageal cancer patients: A propensity score‐matched analysis

Abstract Background Several studies have reported that modified Collard anastomosis is useful for cervical anastomosis after esophagectomy for thoracic esophageal cancer. However, no large‐scale study has confirmed the efficacy of the modified Collard anastomosis. Methods Between 2008 and 2016, 398 consecutive esophageal cancer patients who underwent esophagectomy and cervical anastomosis were enrolled in this study. Patients with a short remnant cervical esophagus were excluded. We investigated the utility of the modified Collard anastomosis by comparing the results of postoperative complications using a propensity score‐matched analysis between the hand‐sewn method (HS) and the modified Collard anastomosis (MC) for esophagogastric anastomosis of the neck after esophagectomy in thoracic esophageal cancer patients. Results Of the 398 patients, 127 were included in the MC group and 127 were included in the HS group after propensity score matching. Clinical characteristics did not differ between the two groups. Frequency of anastomotic leakage tended to be lower in the MC group than in the HS group (3% vs. 7%, P = 0.127). Frequency of anastomotic stenosis was significantly lower in the MC group than in the HS group (13% vs. 59%, P < 0.001). Multivariate logic analysis showed that anastomotic technique (HS) and performance status were independent factors associated with anastomotic stenosis (odds ratio, 12.24 and 2.52; P‐value <0.001 and 0.047, respectively). Conclusion In cervical esophagogastric anastomosis after esophagectomy, the modified Collard anastomosis is more suitable than hand‐sewn anastomosis in terms of reducing the frequency of anastomotic stenosis.


| INTRODUC TI ON
Radical esophagectomy provides the best cure for patients with resectable esophageal cancer. However, this represents the most invasive procedure, and the frequency of postoperative complications is reported to range from 45% to 80%. [1][2][3] Of the postoperative complications, anastomotic complications, such as anastomotic leakage and stenosis, are among the major concerns.
The anastomotic technique after esophagectomy has been investigated extensively in terms of location (intrathoracic or neck), suturing (hand-sewn or mechanical) and type (end-to-end, side-toside, or end-to-side). 4,5 However, the optimal anastomotic method has not yet been established. Hand-sewn anastomosis is widely used, and leakage rate and stenosis rate for this approach are 0%-33% and 2%-89%, respectively. 4,5 Side-to-side anastomosis using a linear stapler was reported by Collard et al 6 and modified by Orringer et al. 7 This modified Collard method is considered to have made progress in reducing anastomotic complications. According to published reports, esophagogastric anastomosis using the modified Collard method has low rates of anastomotic leakage and stenosis. 8,9 In our hospital, in principle, hand-sewn anastomosis was carried out for cervical esophagogastric anastomosis after esophagectomy until 2011. After 2012, the modified Collard anastomosis has been used. Few reports have compared these methods with other anastomotic methods, and such reports included small numbers of cases.
In the present large-scale study, we investigated the utility of the modified Collard anastomosis by comparing data regarding postoperative complications between the hand-sewn method and the modified Collard anastomosis in the esophagogastric anastomosis of the neck after esophagectomy in thoracic esophageal cancer patients.

| Patients
Between January 2008 and December 2016, 582 consecutive patients with esophageal cancer underwent esophagectomy with radical lymph node dissection at the Osaka International Cancer Institute in Japan. Of these patients, 531 with thoracic esophageal cancer underwent subtotal esophagectomy and cervical esophagogastric anastomosis. Of those 531 patients, 25 underwent reconstruction using the jejunum or colon, 26 underwent two-stage reconstruction, 80 underwent esophagogastric anastomosis using a circular stapling technique and two underwent hand-sewn anastomosis because the remnant cervical esophagus was too short to undergo the modified Collard method. After excluding the above 133 patients, 398 patients were enrolled in the present study ( Figure 1).
The 7th edition of the Union for International Cancer Control TNM staging system was used. Details of preoperative treatment in our institution were described priviously. 10,11 The Human Ethics Review Committee of the Osaka International Cancer Institute approved the protocol of this retrospective cohort study (No. 1609089101).

| Surgical procedure
Patients underwent esophagectomy and extensive mediastinal lymph node dissection through a right thoracotomy in the leftlateral decubitus position. Patients were subsequently repositioned in the supine position, and cervical and abdominal lymph node dissection were then carried out. Cervical lymph node dissection was not carried out in lower thoracic esophageal cancer patients without cervical or upper mediastinal lymph node metastasis. Abdominal lymph node dissection was carried out using either open laparotomy or hand-assisted laparoscopic surgery. A 4-cm-wide gastric conduit was created using a linear stapler along the greater curvature of the stomach. At the point of the reconstruction route, in our institution, the retrosternal route is routinely adopted. When the retrosternal route was unable to be used, the posterior mediastinal or percutaneous route was used. After the gastric conduit was pulled up to the neck, esophagogastric anastomosis was carried out on the left side of the neck. Regardless of any reconstruction route, anastomosis was done in the same way. In the present study, four surgeons carried out each method of anastomosis. We divided the four surgeons into two groups; two senior surgeons with over 20 years experience and two junior surgeons with under 20 years experience.
The hand-sewn anastomosis was carried out as follows. First, an appropriate site was selected on the anterior wall of the gastric tube away from the stapled line approximately 2 cm below the highest point of the gastric tube to ensure good blood flow. Then, interrupted posterior seromuscular sutures were made using 3-0 vicryl to approximate the esophagus and stomach. The stomach was then opened transversely approximately 3-5 mm away from the posterior seromuscular suture line. Interrupted stitches through the full thickness of the stomach and esophagus were placed to achieve Conclusion: In cervical esophagogastric anastomosis after esophagectomy, the modified Collard anastomosis is more suitable than hand-sewn anastomosis in terms of reducing the frequency of anastomotic stenosis.

K E Y W O R D S
anastomotic stenosis, cervical esophagogastric anastomosis, esophageal cancer, modified Collard anastomosis, propensity score matching mucosa-to-mucosa approximation. The anterior wall of the anastomosis was completed in a way similar to that of the posterior wall.
The modified Collard anastomosis was carried out as follows

| Perioperative management
Postoperative management was identical for both groups. After the operation, patients were admitted to the intensive care unit under anesthesia. On the day after surgery, the patients were extubated.
Jejunostomy feeding was started after extubation. A nasogastric tube was used routinely and was suctioned every 2 hours. The tube was removed on postoperative day (POD) 5. Videofluorography was carried out on POD 8, and we then evaluated swallowing function and the state of the anastomotic site. Oral intake was started on POD 9 when the above two points were satisfactory. After discharge from the hospital, the patients were observed at the outpatient clinic F I G U R E 1 Patient selection for the evaluation of cervical esophagogastric anastomosis after esophagectomy in patients with thoracic esophageal cancer. CDH, curved intraluminal stapler (Ethicon) every 2-3 weeks. Three months after the surgery, regular blood tests and diagnostic imaging were carried out every 3 months until at least 5 years or recurrence.

| Definition of perioperative complications
In the present study, anastomotic leakage was defined as the presence of saliva leaking through the cervical wound or the presence of extraluminal contrast as seen by videofluorography/computed tomography or visualization of dehiscence or fistula by endoscopy.
Anastomotic stricture was defined according to previous methods. 12 When the patient complained of dysphagia after surgery, endoscopic examination was carried out. The anastomotic site was observed using an XQ260 or XQ240 fiberscope with a front-edge size of 9.0 mm (Olympus, Tokyo, Japan). If we were not able to push through the anastomotic site, presence of anastomotic stricture was defined.
If the patient did not complain of dysphagia until 1 year after surgery, routine endoscopic examination was carried out 1 year after surgery, and the presence or absence of stricture was diagnosed.
Dilation frequency was defined as the number of times endoscopic balloon dilatation was carried out until an endoscopic dilatation-free state was achieved for at least 3 months after the last dilatation.
Reflux esophagitis was defined as greater than grade A according to the Los Angeles Classification System. Concerning other postoperative morbidities such as pneumonia and recurrent nerve paralysis, ≥grade 2 postoperative morbidities according to the Clavien-Dindo classification were defined as the appearance of complications.

| Propensity matched analysis
Propensity matched analysis was conducted using a logistic regression model and the following covariates: age, gender, performance status (PS), concomitant disease, tumor location, clinical T factor, cN factor, cM factor, c stage, preoperative chemotherapy, preoperative chemoradiotherapy, route of reconstruction, field of dissection, and approach taken in the thoracic procedure.

| Patients and characteristics
Of the 398 patients, 173 patients were included in the hand-sewn group (HS group), and the remaining 225 were included in the modified Collard group (MC group) before matching. After matching, 127 patients were included in the HS group, and 127 were included in the MC group ( Figure 1). Table 1 shows the clinical characteristics of the patients before and after matching. After matching, there was no significant difference in backgrounds between the two groups.

| Operative outcome
Operative outcome for each group is summarized in Table 2. There was no significant difference in surgical experience of the surgeons between the two groups. Before matching, the operative time was significantly shorter in the MC group than in the HS group (523 minutes vs. 551 minutes, P = 0.003). The total blood loss was significantly less in the MC group than in the HS group (625 mL vs. 1074 mL, P < 0.001). After matching, the tendency was the same.
The frequency of blood transfusion was similar between the two groups (39.9% vs. 32.0%, P = 0.113). Table 3 shows the postoperative complications for the two groups.

| Postoperative outcome
Anastomotic leakage was less frequent in the MC group than in the HS group, but the difference did not reach statistical significance (3% vs. 8%, P = 0.063). Anastomotic stenosis was significantly less frequent in the MC group than in the HS group (15% vs. 59%, P < 0.001). Furthermore, the period between esophagectomy and the first dilatation was significantly shorter in the HS group than

| Multivariate logistic analysis of anastomotic stenosis
Finally, we carried out logistic analysis to identify the independent factors associated with anastomotic stenosis after esophagectomy (

| D ISCUSS I ON
In the present study, we evaluated the utility of the modified Collard anastomosis and compared it with that of handsewn anastomosis using propensity score-matched analysis in Results showed that the frequency of anastomotic leakage was similar between the two groups. The results also showed that anastomotic stenosis was significantly less frequent in the MC group than in the HS group. These results show that modified Collard anastomosis is effective in reducing the incidence of anastomotic stenosis. This is the first report comparing treatment outcomes between the modified Collard anastomosis and other methods of anastomosis in a large-scale study.
The results of the present study showed that anastomotic leakage was less frequent in the MC group than in the HS group, but this difference was not statistically significant. Compared with the 13.3% frequency in the Japanese nationwide database, anastomotic leakage was less frequent in both groups of our study. 13 According to Honda's review of 12 prospective randomized con- Collard anastomosis and noted that the incidence of anastomotic leakage was 21%. Compared to these previous reports, the incidence of leakage in our large-scale study was lower. According to our results, modified Collard anastomosis advantageously shows a lower rate of anastomotic leakage.
The results of the present study showed that anastomotic stenosis was significantly less frequent in the MC group than in the HS group. In this study, the stenosis rate in the HS group was 59%.   (Table S1). In the posterior mediastinal or subcutaneous route reconstruction, the modified Collard anastomosis might also be useful as well as in the retrosternal route. One explanation as to why anastomotic leakage happened only in the retrosternal route in the MC group is that the number of patients with posterior mediastinal or subcutaneous route reconstruction was small.
Another explanation is that the sternoclavicular joint may impinge on the retrosternal conduit. It is possible that gastric pull-up by the retrosternal route may cause more mechanical stress to the stomach than the other routes, resulting in reduced perfusion, and oxygen supply. It might eventually lead to anastomotic leakage in the retrosternal reconstruction.
In the present study, after matching, a significant difference still exists in total operative time and blood loss between the two groups. The cause of the difference between the two groups may be attributed to the difference in abdominal approach or learning curve associated with the surgical period. The difference in surgical invasiveness between the two groups may result in the different frequencies in anastomotic stenosis. Indeed, according to an analysis of postoperative anastomotic stenosis by Ahmed et al, 15 anastomotic stenosis was associated with intraoperative hypoperfusion.
In the current study, poor PS was an independent factor associated with anastomotic stenosis. We investigated whether poor PS was associated with specific concomitant diseases. Results showed that poor PS tended to be associated with the presence of cardiovascular disease (P = 0.097). According to the analysis of anastomotic stenosis by Ahmed et al, 15 ASA grade and cardiovascular disease were independent risk factors for anastomotic stenosis. Our results were consistent with their findings.
The present study had limitations. This was a retrospective consecutive cohort study. From now, we should prospectively investigate the efficacy of the modified Collard anastomosis.
In conclusion, in cervical esophagogastric anastomosis after esophagectomy, the modified Collard anastomosis is superior to hand-sewn anastomosis in terms of reducing the frequency of anastomotic stenosis. To validate the above results, prospective clinical trials should be carried out.

D I SCLOS U R E
Funding: The authors declare no sources of support.
Conflicts of interest: Authors declare no conflicts of interest for this article.