Introduction

Surgical resection remains a critical component of potentially curative treatment of cancer of the esophagus, despite advances in radiotherapy and chemotherapy. Successful healing of the esophageal anastomosis is an important factor in postoperative outcomes after esophagectomy, as anastomotic leak is associated with significant morbidity and mortality, and the subsequent healing of a leak can lead to an anastomotic stricture, which negatively impacts quality of life. A major factor in anastomotic healing is adequate blood supply to the esophagus and the esophageal replacement. Transposition of the stomach from abdomen to chest during esophagectomy requires division of the left gastric, left gastroepiploic, and short gastric vessels, which compromises blood supply to the anastomotic site.

Evaluation of the blood supply to the gastric conduit has traditionally been made on clinical grounds by inspection of the gastric serosa. In an effort to reduce the risk of anastomotic leak, we employed the use of indocyanine green (ICG) fluorescence angiography and Doppler examination of the gastric conduit’s vascular pattern to evaluate blood supply and assist in construction of the conduit. The aim of this study was to determine the impact of intraoperative evaluation of the conduit vascular pattern on anastomotic failure after esophagectomy. We hypothesized that addition of intraoperative evaluation of the conduit vascular pattern would lead to a lower rate of anastomotic failure.

Methods

We have developed a single surgical team to focus on minimally invasive esophagectomy at Carolinas Medical Center consisting of a thoracic surgeon and gastrointestinal surgical oncologist. Approval for a retrospective study was obtained from the Institutional Review Board at Carolinas Medical Center (Charlotte, NC). From September 2007 to February 2013, 90 unselected patients underwent esophageal resection, reconstruction with a gastric conduit, and intrathoracic anastomosis. A majority of patients underwent some combination of a laparoscopic and thoracoscopic approach as described.1 The abdominal phase of the procedure was performed laparoscopically with an epigastric hand port and three 5-mm ports. Dissection began by entering the lesser sac via the gastrocolic omentum, taking care to preserve the right gastroepiploic artery. The left gastric artery was dissected of nodal tissue and divided with a vascular stapler. The right gastric artery was preserved. The lower mediastinum was dissected with bipolar cautery to clear nodal tissue from the pleurae, aorta, and pericardium for removal with the specimen.

The gastric conduit was fashioned using an endoscopic GIA stapler to divide the lesser curvature of the stomach between the right and left gastric arteries. Gastric drainage was facilitated with pyloroplasty, pyloromyotomy, or botulinim toxin injection.

The thoracic phase was then initiated, with the patient positioned in the left lateral decubitus position. The pleura was incised and dissected to the level of the azygous vein, which was divided. After division of the esophagus, the gastric conduit was transposed into the right chest and an anastomosis created with a circular stapler using a flip-top stapler anvil affixed to a nasogastric tube (OrVil, Covidien, North Haven CT, USA). The anvil was then brought through a fenestration in the esophageal stable line. The superior end of the lesser-curve gastric staple line was then opened, and a 25- or 21-mm DST XL EEA stapler (Covidien, North Haven, CT, USA) was selected and inserted through the lesser curvature with its spike brought out through the greater curvature of the gastric conduit. The circular stapler components were mated and fired. After firing of the stapler, the opening in the lesser curvature was closed with a GIA stapler.

In January of 2012, after 60 patients had undergone resection, review of our data showed an anastomotic leak rate of 20 % and a change in surgical practice made in an attempt to reduce the rate of anastomotic leak. Beginning in January of 2012 with patient #61, intraoperative examination of the vascular supply of the conduit was performed using intraoperative Doppler and ICG fluorescence imaging. The gastroepiploic arcade was evaluated with Doppler and a marking suture placed in the most proximal area in which where there was a strong biphasic arterial signal. This was designated as the preferred area for the anastomosis and was generally placed at the most cephalad arcade of the right gastroepiploic artery. A second suture was placed in the most cephalad area where there was any detectable pulse by Doppler examination, generally in the first or second of the left gastroepiploic arcades. ICG fluorescence angiography was performed using the SPY Elite System (LifeCell, Bridgewater, NJ, USA) using 5 mg of ICG injected intravenously. Images were taken at 60 s after injection. Relative fluorescence of the conduit was evaluated using a reference point midway between the greater and lesser curvature in the proximal (inferior) area of the conduit approximately 10 cm above the pylorus. An arbitrary cutoff of 75 % was selected and a running silk suture placed in the serosa at the border of the 75 % perfusion line. Representative images taken with conventional imaging (Fig. 1a) and with ICG fluorescene imaging (Fig. 1b) are shown.

Fig. 1
figure 1

Intraoperative photographs of a gastric conduit prior to transposition into the right chest. Normal photograph (a) and color-encoded perfusion map of relative indocyanine-green fluorescence (b). The lesser curvature staple line is to the bottom of the image, and the cephalad (distal) portion of the conduit is to the left of the image

Full size image

After evaluation and marking, the conduit was then passed into the right chest. An attempt was made to position the anastomosis between the two marking sutures on the greater curvature and inferior to the silk suture marking the 75 % relative perfusion line. It was possible to position the anastomosis inferior to the 75 % relative perfusion line in all cases. The anastomosis was constructed in conventional fashion.

All patients underwent CT esophagram prior to oral intake. Anastomotic leak was defined as extravasation of contrast on postoperative CT esophagram or presence of empyema on chest CT or elevated drain amylase level with clinical evidence of leak. Trends in the rate of anastomotic leak over time were displayed using a CUMulative SUM (CUSUM) chart. CUSUM charts are used to plot normalized cumulative failure rates (cumulative scores) over time where the score at the first time point is zero. Each success results in an upward deflection of the plot and each failure results in a downward deflection of the plot.2 CUSUM charts were constructed with SAS version 9.4 (SAS Systems, Cary, NC). Comparisons between group were performed using R Statistics version 3.02 (R Foundation for Statistical Computing, Vienna, Austria). Categorical variables were compared using Fisher’s exact test. Continuous variables were compared between groups with Wilcoxon rank sum test. Differences were considered significant at the alpha = 0.05 significance level.

Results

For all 90 patients, data were collected and compiled, as summarized in Table 1. Sixty patients underwent operation prior to the change in practice in January 2012 (group 1), and 30 patients underwent operation subsequently (group 2). The mean age was 62 years old (range 22 to 81), 82 % were males, 80 patients were White, and 10 were African-American. Indications for surgery included adenocarcinoma (75 patients), squamous cell carcinoma (8 patients), neuroendocrine tumors (2 patients), and benign disease (5 patients, which included 3 esophageal strictures and 2 leiomyomata of the esophagus). Neoadjuvant chemotherapy along with radiation therapy was administered in 54 patients, and 6 patients received neoadjuvant chemotherapy alone. Adjuvant therapy was administered postoperatively in 18, 1 of whom received radiation only, 12 received chemotherapy, and 5 received radiation and chemotherapy. In the 85 patients undergoing surgery for malignancy, 81 R0 resections (95 %) and 4 R1 resections were obtained. Median lymph node harvest was 12 nodes.

Table 1 Patient characteristics (n = 90)
Full size table

In our series, for the abdominal portion, one case was converted from laparoscopy to laparotomy due to dense intestinal adhesions, and the remainder of cases were completed with laparoscopy. Similarly for the thoracic portion, 82 cases were performed thoracoscopically, 4 were converted from thoracoscopy to thoracotomy, and 4 were treated with initial thoracotomy (Table 2).

Table 2 Operative management
Full size table

Patient, operative, and tumor characteristics between the two groups (before and after change in practice) were similar with the exception that more patients in Group 2 were treated with preoperative chemotherapy and radiation therapy compared with Group 1 (90 vs 50 %). This difference in treatment was due to a higher proportion of patients in Group 1 with early stage disease (cT1N0 and cT2N0). In addition, the treatment of the pylorus differed between the two groups, with more patients in Group 1 treated with pyloromyotomy (22 vs 0 %) and more patients in Group 2 treated with botulinim toxin (53 vs 13 %).

Median hospital stay was 11 days (range 8 to 86 days) and did not differ between the two groups. Thirty-day mortality was zero, and 90-day mortality was 2.2 %, which did not differ between the two groups (see Table 3).

Table 3 Perioperative complications
Full size table

In the first 60 patients (Group 1), the rate of anastomotic leak was 20 %, prompting a change in surgical technique. Patient #61 in the series was evaluated with Doppler examination and ICG fluorescence angiography; the subsequent 9 patients were evaluated with Doppler examination alone, and the final 20 patients in this series were evaluated with both modalities.

The overall leak rate in the series was 13.3 % (12 of the 90 patients). After the practice change, there were no leaks in the succeeding 30 cases. This difference is significant by Fisher’s exact test (p = 0.007). The CUSUM chart is shown in Fig. 2. This chart shows a clear turning point in the CUSUM score at patient #61, coincident with the change in surgical practice. Prior to patient #61, there is a clear upward trend in CUSUM scores, due to an increase in the rate of anastomotic leak. After patient #61, there is a decline in CUSUM scores, due to improving performance (and no anastomotic leaks).

Fig. 2
figure 2

CUSUM graph of anastomotic leak rate over time. A change in surgical practice occurred with case #61. Until case #61, the cumulative sum (CUSUM) of failure was increasing and began to decline at case #61, coincident with a change in practice

Full size image

Discussion

Healing of the esophagogastric anastomosis is essential for optimizing outcomes after esophagectomy. Anastomotic leak following esophagectomy is one of the most serious complications and occurs in up to 30 % of cases in some published series.3 Such leaks are associated with significant morbidity and mortality.4 Perioperative anastomotic leaks have been shown to adversely affect long-term survival after gastroesophageal surgery, including cancer-specific survival.5

Anastomotic failure is thought to be related to inadequate blood supply to the anastomosis.6 The blood supply of the gastric tube appears to depend entirely on the right gastroepiploic artery with no appreciable contribution from the right gastric artery or from branches of the left gastroepiploic arteries.7 The fact that a constructed gastric tube is supplied by this single artery makes it of paramount importance that the distal tube is well perfused. The dissection of the gastrocolic ligament and the transformation of the stomach into a gastric tube for esophageal replacement have been shown to reduce perfusion at the distal portion of the stomach.8 , 9 Decreased conduit perfusion, as measured by laser Doppler spectroscopy or intramucosal gastric pH, has been shown to predict increased risk of anastomotic complications.6 , 10 , 11

Beyond measuring perfusion, augmentation of blood flow to the distal aspect of the conduit may represent a strategy for improving anastomotic healing. One such method is ischemic preconditioning by partial devascularization of the stomach through ligation of the left gastric artery or short gastric vessels prior to surgery in an effort to facilitate the formation of collateral vessels that would improve distal flow. Despite the attractiveness of this strategy, several studies, including a randomized trial, have failed to demonstrate an impact of ischemic preconditioning on the rate of anastomotic leak.12 – 15

Without successful strategies for increasing the blood supply to the gastric conduit, analysis of the gastric conduit perfusion may be helpful in construction of the conduit and choice of anastomotic site. Traditionally, perfusion to the gastric conduit and anastomosis has been evaluated based on clinical inspection. Many methods have been employed to evaluate perfusion of the conduit. Focal measurements of tissue perfusion such as laser flowmetry and spectrophotometry6 , 8 , 11 or intramucosal pH or mucosal CO2 10 have been shown to predict anastomotic failure but do not provide a geographic image of perfusion within the conduit.

Fluorescent imaging using indocyanine green is readily available in the operative room, gives real-time feedback, and can display a geographic view of tissue perfusion. ICG fluorescence angiography has been employed during breast reconstruction to predict skin flap necrosis.16 , 17 Preliminary published reports suggest the value of this technology in esophageal cancer surgery.18 A surgical series of 11 patients who underwent esophagectomy using ICG fluorescent imaging suggests that the perfusion pattern did predict a leak in one patient in this series. Ten patients were judged to have good perfusion of their gastric conduit, and one, in retrospect, was judged to have poor perfusion of the conduit. One of two leaks in the series occurred in the patients whose conduit had poor perfusion at the cephalad end.19

We report our experience with the use of evaluation of the vascular supply of the gastric conduit using a combination of ICG fluorescence imaging and Doppler. We have analyzed outcomes with the cumulative sum (CUSUM) chart, which allows sequential evaluation of processes with binary outcomes. This test is ideally suited to detect small changes in outcomes over time, such as the effects of cumulated experience (“learning curve”) or the impact of a change in process. Originally used for industrial processes, the CUSUM test has recently been applied to surgical outcomes.20 – 22 In our series, the CUSUM statistic demonstrated that immediately prior to the change in practice with patients #57 and #60, the rate of anastomotic failure had increased to 20 %. With the change in practice that begun with patient #61, the anastomotic leak rate declined for the remainder of the study period.

A limitation of our study is its retrospective nature. The reduction in anastomotic leak rate may have been influenced by other factors related to the experience of the surgical team, known as the learning curve. To look for this effect, we compared the first 30 patients with the second 30 patients within Group 1. The rate of anastomotic leak was slightly higher in the second 30 patients (chi-square p = 0.333, data not shown), perhaps due to the tendency to operate on higher-risk patients as our team gained experience. We did not see a difference in operative time between the first 30 patients and second 30 patients within Group 1 (Wilcoxon p = 0.6735, data not shown).

In comparing Group 1 with Group 2, we noted that operative time was significantly reduced in Group 2, despite the introduction of additional evaluation of the gastric conduit. This reduction in operative time in Group 2 may be reflective of the experience of the operative team, but our subjective experience is that over time, vascular evaluation of the conduit decreased operative time by imparting confidence in the adequacy of the conduit. In any case, the introduction of intraoperative vascular examination of the conduit was associated with a dramatic decrease in anastomotic failure.

Conclusion

Despite advances in nonoperative management of esophageal cancer, esophagectomy remains the mainstay of potentially curative treatment for adenocarcinoma of the esophagus. Anastomotic leak remains a significant source of morbidity and reduces cancer-specific survival. We report our early results after employing the use of ICG fluorescence imaging and Doppler evaluation of the gastric conduit used to evaluate blood supply during esophagectomy reconstruction. This evaluation has helped determine the optimal site for anastomosis based on perfusion to the distal conduit. Despite the limited numbers, we have experienced a statistically significant and clinically relevant decrease in leak rate after this change in practice. Evaluation of conduit perfusion allows for enhanced discretion in choosing the optimal site of anastomosis based on vascular pattern. Further investigation via a large-scale multicenter randomized controlled trial is needed to establish the utility of this novel technology for widespread use and for standardization of techniques during esophagectomy or other gastrointestinal anastomosis.