Calcification of Arteries Supplying the Gastric Tube: A New Risk Factor for Anastomotic Leakage after Esophageal Surgery

Introduction

Esophagectomy combined with neoadjuvant chemoradiation therapy is the mainstay therapy in patients with resectable esophageal cancer (1). Improvements in surgical techniques and perioperative care have led to a steady decrease in postoperative mortality over the years (2). However, the relatively high incidence (10%–30%) of anastomotic leakage after cervical esophagogastrostomy (1,3,4) results in an increased risk of stricture formation and higher perioperative mortality rates (5,6). Suggested risk factors for the development of anastomotic leakage include older age, ischemia of the gastric tube, malnutrition, hypotension, hypoxemia, neoadjuvant therapy, steroid use, smoking, comorbid conditions (eg, diabetes mellitus, chronic obstructive pulmonary disease, and cardiovascular disease), high body mass index, surgical approach, and low number of esophagectomy procedures performed in a hospital (2,4,5,7–11).

Relative ischemia of the gastric tube due to poor tissue perfusion is considered a major cause of anastomotic leakage (4,12). During esophagectomy, a major reduction of microvascular blood flow and concomitant compromised tissue oxygenation in the gastric tube has been shown in several studies (13–15). The most cranial part of the gastric tube is used to create the anastomosis. This part is supplied by a fine submucosal and mucosal microvascular network (16) and is fed exclusively by the right gastroepiploic artery after mobilization of the stomach during gastric tube reconstruction (Fig 1). Even in stable hemodynamic conditions, the blood flow in this upper part of the gastric tube is decreased when compared with that in the antrum and corpus (17,18). One study reported that all patients with tissue blood flow of less than 10 mL/min/100 g at the anastomotic site experienced leakage, indicating that blood flow at the anastomotic site may be an important predictive factor (19). Similarly, another study found a vascularization deficit of the reconstruction in 100% of patients with anastomotic leakage, as shown by selective arteriography (20).

Figure 1:

Download as PowerPointOpen in Image Viewer

Atherosclerosis is a known cause of ischemia and it has been speculated to compromise the vascular supply of the gastric tube and cervical anastomosis (21). A relatively recent study reported that vascular calcifications of the aortic wall and supraaortic arteries can be used as a measure for atherosclerosis to predict cardiovascular events (22). In this study, we present a similar pragmatic scoring system for the semiquantitative scoring of vascular calcifications of the supplying arteries of the gastric tube on routine preoperative computed tomographic (CT) images of the chest and abdomen in patients with esophageal cancer. The aim of this study was to evaluate the association between the amount and location of calcification of the supplying arteries of the gastric tube (as determined with a vascular calcification scoring system) and the occurrence of anastomotic leakage after esophagectomy with gastric tube reconstruction in patients with esophageal cancer.

Materials and Methods

Study Population

Institutional review board approval was obtained, and the informed consent requirement was waived for this retrospective study. Data from all patients who underwent elective esophagectomy at our tertiary referral center from October 2003 until August 2012 were prospectively collected in a database. Patients with an available preoperative CT study of the thorax and abdomen with a section thickness of 5.0 mm or less were selected for this study. Exclusion criteria were benign disease, premature discontinuation of surgery due to the discovery of T4b or M1 disease during surgery, a reconstruction other than gastric tube reconstruction, urgent nonelective surgery, and combined laryngeal resection and intrathoracic anastomosis (Fig 2).

Figure 2:

Download as PowerPointOpen in Image Viewer

All patients underwent gastric tube reconstruction with a cervical single-layer hand-sewn anastomosis. A linear stapling device was used for gastric tube reconstruction, and the staple line was manually oversewn in all subjects (Endo GIA Stapler; Covidien, Dublin, Ireland). During gastric tube reconstruction, the entire course of the right gastroepiploic artery was respected. Patient- and procedure-related characteristics (ie, sex, age, body mass index, American Society of Anesthesiologists score, chronic obstructive pulmonary disease, diabetes mellitus, cardiovascular comorbidity, smoking, chronic use of steroids, neoadjuvant treatment, surgical approach, type of anastomosis, duration of surgery, year of surgery, and supervising surgeon) were collected from a prospective database. In addition, intervals between CT examination and surgery and between surgery and occurrence of anastomotic leakage were recorded, if applicable.

Image Acquisition

All contrast material–enhanced routine preoperative CT protocols were considered suitable as long as the field of view included the full length of the thoracic aorta, celiac axis, and common hepatic, gastroduodenal, and splenic arteries. Chest and abdominal CT examinations were typically performed with 16- or 64-section CT scanners from a variety of manufacturers at referring centers or our own center. Images were typically acquired with 16 × 0.75 or 64 × 0.625 mm section collimation, a gantry rotation time of 400–750 msec, a tube potential of 120 or 130 kV, an effective tube current of 32–215 mAs (mean, 122 mAs), and a field of view of 300 (1% of patients), 320 (2% of patients), 371 (1% of patients), 400 (7% of patients), or 500 (89% of patients) mm. An iodinated 90-mL 300-mg/mL contrast material bolus with a saline solution chaser was administered intravenously at a rate of 3 or 4 mL/sec in all patients. In the standard protocol of both referring centers and our own center, a region of interest was placed in the aortic arch or the descending thoracic aorta, and image acquisition was automatically initiated once a selected threshold was reached within this region of interest with bolus tracking. Subsequently, chest and abdominal CT images were typically acquired during the arterial phase and the portal venous phase, respectively.

Image Evaluation

In patients with more than one CT examination during the preoperative work-up, the findings from the first diagnostic examination were considered, reflecting the moment of clinical decision-making. All preoperative CT studies were independently reviewed and scored for calcifications by one reader (P.S.N.v.R.). In addition, a randomly sampled subgroup (n = 50) was also scored independently by a second reader (L.H.). Both readers were clinical research physicians with 1 and 2 years of experience, respectively, in the field of surgical oncology. The readers were trained and supervised by a dedicated radiologist with more than 15 years of experience in gastrointestinal radiology (M.S.v.L.). Images were analyzed in the transverse plane. Readers were blinded to patient- and operation-related characteristics and clinical outcome in terms of anastomotic leakage.

A visual calcification grading system was modified from a previously reported score for grading aortic wall abnormalities in the prediction of cardiovascular events (22,23). This grading system was developed specifically for the supplying arteries of the gastric tube. The definitions used for visual grading are provided in Table 1, and examples of imaging characteristics are presented in Figure 3. The selection of studied vessels was based on anatomic studies that clearly showed that the right gastroepiploic artery exclusively supplies the gastric tube and originates from the aorta via the celiac axis, common hepatic artery, and gastroduodenal artery (16,21). Additionally, calcifications in the splenic and left gastroepiploic arteries were scored because a small part of the gastric tube length is suggested to benefit from the left gastroepiploic arteries (16). Aortic wall calcifications were scored on all transverse images from the descending thoracic aorta, starting right after the origination of the left subclavian artery, down to the level of the celiac axis origination (score 0–2). Furthermore, calcifications of the celiac axis (score 0–2), right postceliac arteries (common hepatic, gastroduodenal, and right gastroepiploic arteries; score 0–1), and left postceliac arteries (splenic and left gastroepiploic arteries, score 0–1) were scored. The right and left postceliac arteries were scored with a binary scale only because visible vascular calcifications in these smaller vessels were expected to occur relatively infrequent. Artificially scoring more than two categories with small cell counts could easily result in model overfitting, with imprecise estimates describing random error rather than a true association in this case.

Table 1 Definitions Used to Grade Calcification of the Supplying Arteries of the Gastric Tube Seen on Preoperative CT Images

Note.—Definitions may be considered as guidelines to distinguish absence or minor or major presence of calcifications. MCSD = maximum cross-sectional diameter.

*Aorta defined as descending part of thoracic aorta and abdominal part of aorta above celiac level.

^†Right postceliac arteries defined as common hepatic artery, gastroduodenal artery, and right gastroepiploic artery.

^‡Left postceliac arteries defined as splenic artery and left gastroepiploic artery.

Figure 3a:

Download as PowerPointOpen in Image Viewer

Figure 3b:

Download as PowerPointOpen in Image Viewer

Figure 3c:

Download as PowerPointOpen in Image Viewer

Figure 3d:

Download as PowerPointOpen in Image Viewer

Statistical Analysis

The primary outcome measure of this study was anastomotic leakage, defined by either extravasation of water-soluble contrast material during a contrast material swallow study or CT scan, visualization of anastomotic dehiscence or fistulae during endoscopy, or visible loss of saliva through the cervical wound (24). Diagnostic tests to identify anastomotic leakage were conducted on indication only; no routine contrast material swallow examination was performed (25). As clinically apparent cervical leakages generally manifest between 2 and 10 days after esophagectomy and because no leakage is expected to occur after long-term observation (4), the follow-up time of subjects was uniformly truncated to 30 days. A formal check of our prospective database enabled us to confirm the absence of anastomotic leakage occurring after 30 days.

The association between clinical characteristics and calcification scores and anastomotic leakage was studied univariately. We used the χ² test for categorical variables, the Fisher exact test for categorical variables with small cell counts, and the T test or Mann-Whitney U test for normally or skewed distributed continuous variables. We used multivariate logistic regression analysis to study whether calcifications were independently and significantly associated with anastomotic leakage. Variables associated with anastomotic leakage at P < .25 at univariate analysis were entered into the multivariate model. In a stepwise matter, backward elimination of the least significant variables associated with leakage was performed based on the log likelihood ratio. The likelihood ratio test was used to confirm that the excluded covariates did not significantly change the model fit. Odds ratios (ORs) with 95% confidence intervals (CIs) were estimated. Statistical analysis was performed by using statistical software (SPSS 20.0; SPSS, Chicago, Ill). P < .05 was considered indicative of a significant difference.

Interobserver reproducibility was assessed in a random sample consisting of 50 patients from this study scored by two readers (P.S.N.v.R., L.H.). Intraobserver reproducibility was assessed in the same sample scored twice by one reader (P.S.N.v.R.) after a 2-month interval between readings. Overall percentages of agreement were calculated to determine the inter- and intraobserver agreement in grading the scoring model. Inter- and intraobserver reliability was assessed by using κ statistics. We used a weighted κ statistic to calculate the inter- and intraobserver reliability for grading the calcification scores of the aorta (score 0–2) and celiac axis (score 0–2) with linear weighting between successive ordinal categories set at 1 (23). The weighted κ statistic can be interpreted as follows (26): κ of 0.81 to 1.00 indicates excellent reliability; κ of 0.61 to 0.80, good reliability; κ of 0.41 to 0.60, moderate reliability; κ of 0.21 to 0.40, fair reliability; and κ of less than 0.20, poor reliability.

Results

Of the 309 patients who underwent esophagectomy in the study period, 63 were excluded because of insufficient quality of the preoperative CT study (n = 31), benign disease (n = 9), preoperative discontinuation of the procedure because of unsuspected T4b or M1 disease (n = 8), use of a colon interposition graft (n = 5), creation of an intrathoracic anastomosis (n = 2), urgent nonelective resection (n = 3), and combination with laryngeal resection (n = 4). One patient experienced a major acute myocardial infarction within 3 days after esophagectomy and was excluded from further analysis.

Fifty-eight (24%) of the remaining 246 patients experienced anastomotic leakage after esophagectomy. Median time between CT examination and esophagectomy was 86 days (range, 1–174 days). All but one patient survived the first 30 days after esophagectomy, with no loss to follow-up in this period. In addition, censoring was not an issue, as the one patient who died within the 30-day follow-up period experienced anastomotic leakage before death. Anastomotic leakage occurred after a median time of 6 days (range, 1–26 days) after esophagectomy. Patients in the anastomotic leakage group had a mean age of 65 years and 79% (46 of 58) were male, whereas patients without leakage had a mean age of 64 years and 71% (134 of 188) were male (Table 2). Chronic use of steroids (n = 4) was rare, and all these patients developed anastomotic leakage. Other patient- and procedure-related factors were not significantly associated with anastomotic leakage.

Table 2 Baseline Characteristics

Note.—Data are numbers of patients, with row-based percentages in parentheses.

*Data are mean ± standard deviation.

^†Significant difference between the anastomotic leakage group and the no anastomotic leakage group (P < .05).

Calcifications of the aorta were highly prevalent; 102 (41%) and 37 (15%) of 246 patients were assigned aortic calcification scores of 1 and 2, respectively (Table 3). At univariate analysis, the aortic calcification score was significantly associated with the occurrence of anastomotic leakage (27% [28 of 102] and 35% [13 of 37] leakage in groups with minor and major calcifications, respectively, vs 16% [17 of 107] in the group without calcifications; P = .029). No significant association was found between celiac axis calcifications, which occurred in 105 (43%) of 246 patients, and anastomotic leakage (27% [20 of 74] and 35% [11 of 31] in groups with minor and major calcifications, respectively, vs 19% [27 of 141] in the group without calcifications; P = .107). Although calcifications in the right postceliac arteries were uncommon (n = 11 [4%]), the incidence of leakage was significantly higher in this group (55% [six of 11] vs 22% [52 of 235], P = .013). The common hepatic artery showed at least one calcification in all 11 patients; the gastroduodenal and right gastroepiploic arteries were additionally but not exclusively involved in five and two of these patients, respectively. The presence of left postceliac artery calcifications was not significantly associated with a higher incidence of leakage (27% [16 of 59] vs 22% [42 of 187], P = .462). Left postceliac artery calcifications included splenic artery involvement in all 59 patients, with visible additional involvement of the left gastroepiploic artery in only three of these patients.

Table 3 Distribution and Univariate Comparison of Calcification Score per Trajectory among Patients with Anastomotic Leakage versus Patients without Anastomotic Leakage

Note.—Data are numbers of patients, with percentages in parentheses. P values were determined with the χ² test.

*Significant difference between the anastomotic leakage and no anastomotic leakage groups (P < .05).

The calcification scores of the aorta, celiac axis, and right postceliac arteries were entered into the multivariate logistic regression model, along with the terms gender, chronic obstructive pulmonary disease, and year of operation. Chronic use of steroids was not considered suitable for regression analysis because of the very low prevalence in our sample. Duration of surgery was excluded from this model because this covariate would not be useful in preoperative risk assessment. After stepwise backward elimination, sex, chronic obstructive pulmonary disease, year of surgery, and calcification of the celiac axis were eliminated, and the vascular calcification scores of the aorta and right postceliac arteries remained independently and significantly associated with a higher risk of anastomotic leakage (Table 4). When compared with patients without aortic calcifications, patients with an aortic calcification score of 1 or 2 had an increased adjusted risk for leakage (OR, 2.00; 95% CI: 1.02, 3.94 and OR, 2.87; 95% CI: 1.22, 6.72, respectively). Similarly, a significant association with leakage was found for major calcifications of the right postceliac arteries (OR, 4.22; 95% CI: 1.24, 14.4).

Table 4 Results of Multivariate Logistic Regression Analysis with Stepwise Backward Elimination

Note.—As an example, an OR of 4.22 shows that the presence of calcifications in the right postceliac arteries (score, 1 vs 0) leads to a 322% increase in the odds of experiencing anastomotic leakage. Adjustment variables included sex, chronic obstructive pulmonary disease, and year of surgery. Data in parentheses are 95% CI.

*Marked parameters are each independently associated with anastomotic leakage.

The 50 randomly selected patients used to determine observer agreement appeared to be representative of the whole study group, with a mean age of 64.1 years ± 8.4 (standard deviation) and male sex in 74% (37 of 50) of cases. Anastomotic leakage occurred in 22% (11 of 50) patients. The aorta calcification score showed excellent inter- and intraobserver overall agreement (94% [47 of 50] and 94% [47 of 50], respectively) and excellent inter- and intraobserver reliability with weighted κ values of 0.93 (95% CI: 0.84, 1.00) and 0.93 (95% CI: 0.84, 1.00), respectively. Also, excellent inter- and intraobserver overall agreement was found in scoring celiac axis calcifications (88% [44 of 50] and 90% [45 of 50], respectively) with excellent inter- and intraobserver reliability (weighted κ of 0.81 [95% CI: 0.67, 0.95] and 0.84 [95% CI: 0.72, 0.97], respectively). Calcifications in the right postceliac arteries showed excellent inter- and intraobserver overall agreement (94% [47 of 50] and 98% [49 of 50], respectively) with good inter- and intraobserver reliability (κ of 0.64 [95% CI: 0.27, 1.00] and 0.79 [95% CI: 0.39, 1.00], respectively). Calcifications in the left postceliac arteries showed excellent inter- and intraobserver overall agreement (84% [42 of 50] and 88% [44 of 50], respectively) with moderate to good inter- and intraobserver reliability (κ of 0.59 [95% CI: 0.33, 0.84] and 0.69 [95% CI: 0.46, 0.92], respectively).

Discussion

Calcifications of the arterial supply of the gastric tube detected on routine preoperative CT images are associated with the occurrence of cervical anastomotic leakage after esophagectomy for cancer. Preoperative calcifications of the aorta (score, 1–2 vs 0) and right postceliac arteries (score, 1 vs 0) were independently associated with postoperative anastomotic leakage. A visual grading system with excellent inter- and intraobserver reproducibility was developed.

The finding that no standard patient- or operation-related factors (except for chronic steroid use) were significantly associated with anastomotic leakage in this reasonably large series of 246 patients emphasizes the current inability to accurately predict anastomotic leakage. In this context, it must be noted that the cause of anastomotic leakage is multifactorial and is not solely based on calcifications. More risk factors need to be identified to allow for individualized prediction. The calcification scoring method, which involves use of image characteristics detected at routine scanning, can be used as an important factor in future prediction models to identify patients at high risk for leakage.

The calcification grading system in this study was largely based on a previously described and validated visual grading system used to score aortic abnormalities seen on routine diagnostic CT images that was shown to be useful in the prediction of cardiovascular events (22,23). Besides calcifications, scores for irregularity, plaques, and elongation of the aortic wall were analyzed in these studies; however, the addition of these parameters did not lead to substantial improvement of the prediction model. A prediction model based solely on calcifications was shown to be most appropriate for clinical use, with good to excellent inter- and intraobserver reproducibility (22,23). Similar to these findings, the scoring model in our study also showed good to excellent inter- and intraobserver reproducibility (weighted κ range, 0.64–0.93). The model is based on simple definitions and can be used in standard diagnostic CT protocols. In contrast, others used calcium scoring techniques that require special semiautomatic calcium scoring software or special CT protocols that have limited reproducibility and substantial practical difficulties (27,28).

The final model resulting from multivariate analysis contains a plausible pathophysiologic rationale, since the right gastroepiploic artery originating from this trajectory mainly supplies the gastric tube and anastomosis (16,21). The finding that the calcification score in the left postceliac arteries (ie, splenic and left gastroepiploic arteries) did not significantly differ in patients with anastomotic leakage from those without such leakage adds to this hypothesis since these arteries barely contribute to the blood supply of the gastric tube.

Ischemia as an underlying mechanism for anastomotic leakage is likely to be moderated by a combination of generalized vascular disease (marked by aortic calcifications) and compromised local perfusion (marked by right postceliac arteries calcifications). Furthermore, anastomotic leakage has been related to congestion due to insufficient venous drainage at the anastomotic site, the method of anastomosis construction, the width of the gastric tube, mechanical tension, and poor nutrition (3,4,12). With regard to these suggested causes, different attempts to optimize the conditions of the anastomosis have been reported. The proposed calcification scoring system of this article can help in selecting patients who might benefit from one of these interventions in the future.

The strength that lies in the simple and straightforward assessment of calcifications in our model might be considered a potential limitation of this study. Typical chunky calcifications were not differentiated from long thin calcifications. Also, variant vascular anomalies of the gastric and hepatic arteries and the celiac axis were not assessed separately. Finally, calcifications in the arteries are at most a surrogate for atherosclerosis and impaired perfusion, and there might be more specific ways to determine the extent of vascular disease and local perfusion, such as use of dedicated CT angiography of the abdomen or laser Doppler flowmetry.

Our study showed that vascular calcifications of the arteries that supply the gastric tube, defined by a visual grading system with good to excellent inter- and intraobserver reproducibility, are independently associated with anastomotic leakage of the esophagogastrostomy. Future research should aim to include this new parameter in the development of a risk prediction model for anastomotic leakage.

Author Contributions

Author contributions: Guarantors of integrity of entire study, P.S.N.v.R., L.H., R.v.H.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, P.S.N.v.R., L.H., M.S.v.L., J.P.R.; clinical studies, P.S.N.v.R., L.H., M.S.v.L., R.v.H., J.P.R.; statistical analysis, P.S.N.v.R., H.M.V.; and manuscript editing, all authors