RISK FACTORS FOR SEVERE POSTOPERATIVE COMPLICATIONS AFTER GASTRECTOMY FOR GASTRIC AND ESOPHAGOGASTRIC JUNCTION CANCERS

ABSTRACT Background: Gastrectomy is the main treatment for gastric and Siewert type II-III esophagogastric junction (EGJ) cancer. This surgery is associated with significant morbidity. Total morbidity rates vary across different studies and few have evaluated postoperative morbidity according to complication severity. Aim: To identify the predictors of severe postoperative morbidity. Methods: This was a retrospective cohort study from a prospective database. We included patients treated with gastrectomy for gastric or EGJ cancers between January 2012 and December 2016 at a single center. Severe morbidity was defined as Clavien-Dindo score ≥3. A multivariate analysis was performed to identify predictors of severe morbidity. Results: Two hundred and eighty-nine gastrectomies were performed (67% males, median age: 65 years). Tumor location was EGJ in 14%, upper third of the stomach in 30%, middle third in 26%, and lower third in 28%. In 196 (67%), a total gastrectomy was performed with a D2 lymph node dissection in 85%. Two hundred and eleven patients (79%) underwent an open gastrectomy. T status was T1 in 23% and T3/T4 in 68%. Postoperative mortality was 2.4% and morbidity rate was 41%. Severe morbidity was 11% and was mainly represented by esophagojejunostomy leak (2.4%), duodenal stump leak (2.1%), and respiratory complications (2%). On multivariate analysis, EGJ location and T3/T4 tumors were associated with a higher rate of severe postoperative morbidity. Conclusion: Severe postoperative morbidity after gastrectomy was 11%. Esophagogastric junction tumor location and T3/T4 status are risk factors for severe postoperative morbidity.


INTRODUCTION
G astric cancer (GC) is the fifth most common cancer 9 , with more than 900,000 new cases every year, and the third leading cause of cancerrelated death worldwide 29 . Surgery and adjuvant treatment are the main treatment modalities for GC. The gastrectomy is the approach universally agreed upon for gastric and Siewert type III esophagogastric junction (EGJ) cancer 15,24 . Although controversies exists respect the treatment for Siewert II tumors, the extended total gastrectomy also appear as an appropriate surgical option 5,10 . Postoperative morbidity rates after gastrectomy vary across different studies, but total morbidity is more than 20-30% in most studies 3,12,13,18,22,23, . In a previous study by our group, morbidity was present in 31% of 1066 gastrectomies 21 .

1/6
Data on postoperative morbidity predictors are heterogeneous. Patient (age, comorbidity, body mass index, serum albumin), tumor (local invasion and location), and surgery (open approach, total gastrectomy, lymph node dissection, and multi-organ resection) variables are described as potential factors for higher morbidity 3,22,23 .
In our previous study, we did not have data on complication severity because this type of score did not exist at the beginning of that study 21 . In the past decade, complication severity has gained great importance, and use of the Clavien-Dindo classification has been widely adopted 8 . However, only a few studies have evaluated postoperative morbidity predictors according to complication severity for gastrectomy 13,18 .
The aim of this study was to identify predictors of severe postoperative morbidity after gastrectomy for gastric (GC) and esophagogastric junction (EGJ) cancer.

METHODS
The local ethics committee approved this study. Informed consent of patients was waived because of the retrospective nature. This study was registered in ClinicalTrial. gov NCT03909997.
This was a retrospective cohort, including data from a prospective, institutional, single-center database. The database collected patients' demographics, tumor and surgery characteristics, and postoperative morbidity. All consecutive patients treated with a gastrectomy for GC or EGJ cancer between January 2012 and December 2016 were included. Only patients with stomach or EGJ adenocarcinoma were selected, and patients with other histology were excluded.

Preoperative assessment
The preoperative assessment consisted of upper gastrointestinal endoscopy, biopsy, complete blood count, liver function tests, electrocardiogram, and nutritional evaluation. Patients with diabetic, coronary heart disease and COPD were assessed additionally with a glycated hemoglobin test, echocardiogram and spirometry respectively. Preoperative imaging was a thorax-abdomen-pelvis computed tomography (CT) scan.

Operative procedure
Epidural analgesia was routinely employed in open surgery. Depending on the tumor's location, a total or subtotal gastrectomy was indicated. Surgery included omentectomy with bursectomy and D2 lymph node dissection, according to the Japanese classifications in patients with curative gastrectomy 25 . Multi-organ resection, including spleen, pancreas, colon, and liver, was performed in cases of direct tumor invasion. Partial distal esophagectomy with a transhiatal approach and mediastinal anastomosis was employed for Siewert types II and III cancers, with frozen section intraoperative biopsy to confirm an R0 resection. Partial distal esophagectomy was considered a multi-organ resection when more than 2 cm of the esophagus was resected. Routine cholecystectomy was performed in curative cases and was not considered a multi-organ resection. A reconstruction, using Roux-en-Y, was performed after a total gastrectomy; Roux-en-Y or Billroth II was used for subtotal gastrectomy. Esophagojejunal anastomosis was performed with a circular stapler and a second layer of running monofilament suture. One or two prophylactic drains were used routinely 21 . A laparoscopic approach was employed in patients with clinical early GC who were not candidates for endoscopic resection and patients with advanced GC without clinical invasion of adjacent structures and with lymph node metastases only in the perigastric area 20 .

Postoperative management
Patients started early respiratory and physical therapy the day following surgery. An oral contrast study was performed on postoperative day 5-7 for total gastrectomy; after this, the patient started an oral diet and prophylactic drains were removed.
Esophagogastric junction (EGJ) cancer was classified according to Siewert classification 27 Only EGJ cancer Siewert types II and III were included, type I were excluded. Patients were staged using TNM Classification of Malignant Tumors, 7 th edition 19 .

Complication assessment
The primary outcome was severe 30-day or in-hospital morbidity, which was defined as a Clavien-Dindo score ≥3 8 . The complication data were prospectively collected by each attending surgeon and together at a monthly conference. Complications detected upon readmission were also included.
Postoperative bleeding was defined as any blood loss through abdominal drains or at reoperation. We considered an esophagojejunostomy leak as the appearance of contrast outside the anastomosis, using an oral contrast or CT scan, or by direct evaluation at reoperation. We considered a duodenal stump leak as the discharge of bile-containing liquid in drains or by direct evaluation at reoperation. A pancreatic fistula was considered as a drain output of any volume on or after postoperative day 3 with an amylase greater than three times the serum level. Intra-abdominal abscess was defined as septic fluid in the abdominal cavity on CT causing systemic inflammatory response syndrome. Postoperative pancreatitis was diagnosed with elevated levels of pancreatic enzymes and/or imaging finding 2 .

Sample size
Using previously published data and given a 0.05 alpha level, a percentage of unexposed outcome of 10.8%, and OR 4.28, a sample size of 116 patients would yield at least 80% statistical power.

Statistical analysis
Continuous variables were described by means and standard deviations or medians and interquartile range. Categorical variables were described with frequencies and percentage. The following factors were analyzed: age, gender, comorbidity, American Society of Anesthesiologists (ASA) physical status, tobacco and alcohol consumption, body mass index (BMI), hematocrit, serum albumin level, tumor location, the use of preoperative chemotherapy, laparoscopic or open surgery, total or subtotal gastrectomy, duodenal stump closure, multi-organ resection, lymphadenectomy, reconstruction method, T status, lymph node metastasis, and resection margin. We used a cut-off point of 65 years for statistical analysis 17 . The T stage was grouped by T1/T2 and T3/T4 for analysis. Univariate and multivariate analyses were performed to identify predictors of severe postoperative morbidity. All variables associated with severe morbidity with p<.05 in the univariate analysis were subsequently entered into a Cox multivariate regression model with backward elimination. Significance was set at two-sided p <.05. All analyses were performed using the statistical SPSS IBM Statistics software program, version 22.

RESULTS
Two hundred and eighty-nine gastrectomies were performed, 195 (67.5%) of whom were male, with a median age of 65 years (+/-11). The median BMI was 24.4 (21.8-26.9), and median albumin was 4.1 gr/dl (3.6-4.4). The patients' characteristics are summarized in Table 1. Eighty-three percent of patients OriginAl Article 2/6 had at least one comorbidity; the majority had ASA II (57%). Eight patients (2.8%) received neoadjuvant chemotherapy. The tumor was located in the EGJ in 14% of patients and in the stomach in 85%.
The open approach was employed in 231 (79.9%) patients.

FIGURE 1 -Postoperative morbidity according to Clavien-Dindo classification
The esophagojejunal anastomosis leak rate was 4.2%. Five cases (1.7%) of esophagojejunal anastomosis leak were managed with nothing by mouth, antibiotics, and prophylactic drains, corresponding to a Clavien-Dindo score of II. Severe esophagojejunal anastomosis leak was diagnosed in seven cases (2.4%). Three patients underwent a reoperation due to clinical deterioration or abdominal abscess and later recovered. In one case of conservative treatment, the patient developed anastomotic stenosis and required an endoscopic dilation as the only intervention. Three patients who died due to esophagojejunal fistula had mediastinal and uni/bilateral pleural contamination; two of these patients underwent a reoperation, and one was treated conservatively, dying shortly after, due to multiple organ failure. Esophagojejunal anastomosis leak rate was 11.9% in patients with an EGJ tumor location.
Duodenal stump leak developed in 2.8% of patients. Three cases (1%) of duodenal stump fistula were managed with antibiotics and prophylactic drains or spontaneous drainage through the abdominal wound, corresponding to a Clavien-Dindo score of II. Five patients (1.7%) underwent reoperation due to duodenal stump leak, and in one of them a duodenostomy was performed. One patient underwent reoperation because the duodenal fistula, and developed respiratory and hemodynamic failure associated also with a pulmonary embolism, requiring intensive care treatment corresponding to Clavien IV. No postoperative mortality was associated with duodenal stump leak.
Six patients (2%) developed severe respiratory complications. Two with pneumonia required mechanical ventilation, and one was treated with a thoracocentesis due to parapneumonic pleural effusion. Thoracocentesis was also performed in two pleural effusions, and one patient required bi-level positive airway pressure (BiPAP) because of pulmonary atelectasis.
The frequency of severe complications was significantly higher for patients with a T3/T4 EGJ cancer, reaching 26% compared to the group of non-EGJ cases, and a T1/T2 status, with only 4% of severe complications (Figure 2).  In the T3/T4/EGJ group (n=38), the main severe complications were represented by two cases of severe esophagojejunal anastomosis leak (5%), two severe respiratory complications (5%), and three severe cardiac (8%) complications. Morbidity among EGJ cancer types according to the Siewert classification did not have a statistically significant difference (36% Siewert II and 18% Siewert III, p=0.37).
RR (Relative Risk) statistically significant with respect to T1-T2/Non EGJ group

DISCUSSION
Most studies describe and evaluate total postoperative morbidity 3,22,23 , and only more recent studies have evaluated postoperative morbidity according to complication severity 13,18 . Using the Clavien-Dindo classification, severe morbidity was present in 11% of patients after gastrectomy. We identified tumor location in the EGJ and gastric wall involvement beyond the muscular layer as predictors of severe postoperative morbidity.
The use of severity grading to evaluate postoperative complications offers several advantages. The Clavien system is easy to apply and has gained widespread use. The data collected from Clavien ≥3 cases allow us to focus on the complications with greater clinical significance and potentially life-threatening consequences and enables a more precise comparison between studies. Because levels I and II complications are often not fully documented across different centers, this is supported by the great variation of total morbidity description 3,13,18,22,23 , but a generally stable rate of severe complications between 9% and 12% 13,18 , was very similar to the rate described in this study (11%).
Previously reported risk factors for morbidity, such as age, preoperative co-morbidity, open surgical approach, multi-organ resection, splenectomy, or total gastrectomy observed in other series 3,11,12,13,14,18,22,23,26 , were not associated with severe morbidity in our study, probably because these factors were more often associated with overall morbidity. Laparoscopic approach was not associated with a significantly lower rate of severe morbidity. This result may be due in part to the less frequent use of laparoscopic gastrectomy in our study, particularly in the EGJ location. EGJ cancer is more prevalent in Western centers and represents a subgroup of esophagogastric malignancies that have special staging and treatment modalities, depending on patients' and tumors' factors and mainly tumor location as described by Siewert classification 10,27, . In our study, we included Siewert types II and III tumors treated with gastrectomy, most of them with transhiatal distal esophagectomy and mediastinal esophagojejunal anastomosis. In the group of patients with EGJ location and a T3/T4 status, significantly high severe morbidity reached 26%, associated with a higher rate of esophagojejunal anastomosis leak and cardiorespiratory complications. This might have been due to the technical difficulties in the resection phase of the operation because of the manipulation of the pericardium and both pleurae. Also the technical difficulties to perform an esophagojejunal anastomosis high in the mediastinum with restricted space or the peripheral inflammation associated with an EGJ and T3/T4 cancer may have facilitated second injuries, future bleeding, or leaks that eventually explained their risk association.
Generally, gastrectomy is viewed as a procedure with lower morbidity compared to esophagectomy 7 , but our data show that for T3/T4 and EGJ tumors this may not be the case. These morbidity data must be considered in the preoperative planning in patients who may be candidates to either esophagectomy or gastrectomy, according to local morbidity and mortality results from each procedure 4 . In patients with a higher risk of severe complications, the preoperative nutritional and physical condition in the weeks previous to surgery needs to be improved, the postoperative management should be optimized, and these conditions carefully monitored for possible complications.
Several studies of esophageal and colon cancers have shown that postoperative complications, by themselves, are associated with worse oncological survival 1,30, . For gastric cancer, the data are contradictory on how postoperative complications affect long-term survival 6,16,28, . The follow-up of our patients will allow us to define whether global, severe or infectious complications change survival.
This study has a limitation that corresponds to a retrospective cohort, and some factors, such as preoperative weight loss, intraoperative bleeding, pre-or intraoperative red blood cell transfusion, were not available to include in the analysis.

CONCLUSION
Eleven percent of patients present severe morbidity after gastrectomy. EGJ tumor location and gastric wall involvement beyond the muscular layer represent predictors of severe postoperative morbidity. This risk stratification allows a more precise decision-making process for patient selection, evaluation, and optimization as well as improved counseling about the risks of surgery.