Prevalence of anastomotic leak and the impact of indocyanine green fluorescein imaging for evaluating blood flow in the gastric conduit following esophageal cancer surgery

Between January 2000 and December 2015, 133 patients underwent esophagectomy with lymphadenectomy for EC at the Department of Gastrointestinal and Pediatric Surgery of Mie University Graduate School of Medicine. Among the 133 patients, 13 patients underwent two-stage surgery or reconstruction with colonic or jejunal tissue. Hence, a total of 120 patients who underwent esophagectomy with esophagogastric anastomosis for EC were enrolled in the study. Patient characteristics were collected, including demographic data [sex, age, body mass index (BMI), and American Society of Anesthesiologists (ASA) classification], tumor-specific data (T classification, lymph node metastasis, pathological type, and lymphatic and venous invasion), preoperative biochemical variables (tumor markers and systematic inflammatory indicators), and surgery-related factors (estimated blood loss, duration of surgery, surgical approach, anastomosis location, route of reconstruction, and method of anastomosis). These factors were evaluated, along with ICG fluorescein imaging, to identify risk factors associated with AL. Univariate and multivariate analyses were performed to detect the demographic, tumor specific, laboratory, and surgical factors affecting AL. Peripheral blood samples were collected from patients prior to surgery and included neutrophil and lymphocyte counts, albumin (Alb) level, and C-reactive protein (CRP) level. The neutrophil and lymphocyte counts were used to calculate the neutrophil-to-lymphocyte ratio (NLR), and the CRP level was used to calculate the modified Glasgow prognosis score (mGPS). The cut-off values for carcinoembryonic antigen (CEA) and squamous cell carcinoma antigen (SCC) were 5 and 1.5 ng/ml, respectively, according to the normal range used at our institution. The cut-off values for Alb and CRP were 3.5 g/dl and 0.5 mg/dl, respectively, based on reference values of the mGPS [18]. The mGPS was designed to assess the nutritional status and prognosis of patients undergoing surgery for gastrointestinal malignancies, and patients were categorized according to a score of 0, 1, or 2, based on Alb and CRP values. NLR was designed as an index of systematic inflammation and its prognostic value has been studied in many types of cancer. Patients were divided into two groups using an NLR cut-off value of 2.5. Finally, patients were divided into two groups based on BMI, based on a cut-off value less than the 25th percentile (19.4 kg/m²).

A gastric conduit 4 cm wide was made by stapling the lesser curvature of the stomach. The right gastric artery, right gastroepiploic artery, and branches of the left gastroepiploic arteries were preserved and provided the vascular supply to the gastric conduit through an arcade of peripheral vessels. The omentum was freed from the transverse colon and divided at the edge adherent to the colon so vessel communications and a sufficient amount of omentum were fully preserved.

ICG fluorescein imaging was performed as follows: a 2.5 mg bolus injection of ICG dye (Diagnogreen; Daiichi-Sankyo Pharmaceutical, Tokyo, Japan) was administered after forming the gastric conduit. Vascular networks were assessed within the gastric wall and omentum about 15–60 s after ICG injection using an infrared ray imaging system (PDE; Hamamatsu Photonics K.K, Hamamatsu, Japan), and the data are recorded as a movie file [14]. Real-time visualization of the tissue perfusion enabled the operating surgeon to immediately calculate the perfusion index in the regions of interest using specially designed software (ROIs; Hamamatsu Photonics K.K, Hamamatsu, Japan), as detailed previously [19]. Based on the blood flow evaluation of the gastric conduit by ICG fluorescein imaging, the vascular territories with rapid, slow, and low perfusion areas were identified, and the border between each regions was marked (Fig. 1a, b). We defined rapid or slow (sufficient) perfusion areas as safely anastomosed, and low (insufficient) perfusion areas as unsafe for anastomosis [19] (Fig. 1b). In cases of sufficient perfusion at the level of the superior border of the sternum, cervical esophagogastric end-to-side anastomoses were generally performed using a circular stapler via a retrosternal route. This is one of the most common reconstruction methods in Japan, as reported in a comprehensive registry of esophageal cancer patients [20]. On the other hand, when blood flow to the gastric conduit was deemed insufficient for end-to-side anastomoses at the level of the superior border of the sternum, we constructed end-to-end anastomoses using a hand-sewn technique, adding vessels anastomosed between the short gastric vein and artery in the neck vessels. Alternatively, we used an adjunctive procedure whereby we decided good blood flow site of gastric conduit via sharpening part of the manubrium or by making a longitudinal incision over the manubrium and upper part of the sternum to make a secure anastomosis.

Cases that were not evaluated by ICG fluorescein imaging were included in the ICG (−) group, and cases that were evaluated by ICG fluorescein imaging were included in the ICG (+) group.

AL was defined as any esophagogastric anastomosis dehiscence that was clinically symptomatic (abscess, mediastinitis, externalized drainage of digestive fluid) or clinically asymptomatic but detected by contrast study within 30 days after esophagectomy, and included necrosis of the gastric conduit and anastomotic-bronchial fistulas. AL was assessed by a water-soluble, monomeric, ionic X-ray contrast medium (Gastrografin; Schering AG, Berlin, Germany) on postoperative day 6. In cases of uncertainty, the diagnosis was confirmed by an upper gastrointestinal endoscopy. Surgical site infections was defined as superficial pus expressed from the abdominal, thoracic, or drains incision sites, requiring surgical debridement and antibiotic treatment. Respiratory complications included bronchial circulatory disturbance, disorders of ventilation, atelectasis, pneumonia, respiratory failure, and acute respiratory distress syndrome. Recurrent laryngeal nerve paralysis was defined as a disturbance of vocal cord mobility with insufficient glottics closure as observed with flexible laryngoscopy, and was recorded according to the affected side.

The study group was comprised of 101 males and 19 females and the median age was 68 years old (25th percentile; 63 and 75th percentile; 74). Characteristics of the study population are shown in Table 1. According to the Japanese classification of EC (10th Ed.) [21], pathological stages included stage 0 (14 patients, 11.7%), stage I (31 patients, 25.8%), stage II (40 patients, 33.3%), stage III (24 patients, 20%), and stage VIa (11 patients, 9.2%). Fifty-nine (49.2%) patients underwent ICG fluorescein imaging to evaluate blood flow in the gastric conduit. Median estimated blood loss was 495.0 g (25th percentile; 305.5 g and 75th percentile; 719.8 g) and median duration of operation was 551.0 min (25th percentile; 420.5 min and 75th percentile; 632.0 min).

Among the 120 study patients, 10 (8.3%) developed AL. Of those 10 patients, 1 patient (0.83%) developed necrosis of the gastric conduit and another patient (0.83%) developed an anastomotic-bronchial fistula. In addition, a total of 24 (20.0%), 23 (19.2%), and 14 (11.7%) patients had surgical site infections, respiratory complications, or recurrent laryngeal nerve paralysis, respectively (Table 1).

A univariate analysis showed no significant associations between AL and demographic data (age, sex, comorbidities, preoperative treatment, and BMI). Furthermore, no significant associations between AL and clinicopathological factors were observed (Table 2). Preoperative tumor markers, Alb, CRP, and mGPS were not associated with the development of AL, but NLR tended to be higher in patients who developed AL [HR 3.8130, 95% confidence interval (CI) 0.9999–18.4563, p = 0.0500] (Table 2). No significant associations between AL and surgery-related factors were observed (estimated blood loss, duration of surgery, surgical approach, anastomosis location, reconstruction route, and method of anastomosis). However, the ICG (−) group had a significantly higher incidence of postoperative AL compared with the ICG (+) group [HR 10.0384, 95% confidence interval (CI) 1.7969–188.2576, p = 0.0057] (Table 2).

Using the significant factors identified by univariate analysis, a multivariate analysis was performed to identify factors independently associated with AL, as shown in Table 3. We found that the absence of ICG fluorescein imaging was an independent risk factor for AL following EC surgery [HR 9.0740, 95% confidence interval (CI) 1.5923–171.2574, p = 0.0098].