Surgery Today Jpn J Surg (1994) 24:24-29 @ Surr, mvTooAv

© Springer-Verlag 1994

Factors Affecting Leakage Following Esophageal Anastomosis YOUNG LEE, HIROMASA FUJITA, HIDEAKI YAMANA, and TERUO KAKEGAWA

First Department of Surgery, Kurume University School of Medicine, 67 Asahi-machi, Kurume, Fukuoka, 830 Japan

Abstract: Esophageal anastomotic leaks remain the most serious problem following extirpative procedures for esoph- ageal carcinoma. We conducted a retrospective analysis of 352 patients with carcinoma in the thoracic esophagus who had undergone esophageal anastomosis following esophagectomy at the Kurume University Hospital between 1981 and 1990. Of these, 94 patients (27%) developed anastomotic leaks, and out of this subgroup, 21 (6%) died as a direct result of the leak. A further 20 patients (6%) underwent repair of the leak, after which they were able to tolerate oral intake. The anastomotic leak healed spontaneously in the other 53 patients (15%). The risk factors predisposing to leaks from esophageal anas- tomoses were determined as: (1) the anastomosis being per- formed via a retrosternal or subcutaneous route as opposed to an intrathoracic route, (2) the use of colonic interposition as opposed to a gastric pedicle, (3) performing a manual anastomosis as opposed to a mechanical anastomosis, and (4) employing an end-to-end anastomosis, as opposed to an end-to-side anastomosis, using a mechanical method. By introducing an anastomotic stapling device, a microvascular technique, a staged operation based on the preoperative risk analysis, and improvement in pre- and postoperative man- agement, the incidence of anastomotic leakage could be de- creased from 35% to 14%, and that of consequent hospital mortality, from 9% to 2%.

Key Words: esophageal anastomosis, leak, mechanical suture, postoperative complication

Introduct ion

Subtotal esophagectomy with an esophagogastrostomy in the neck is the most common surgical treatment for carcinoma of the thoracic esophagus in Japan. This is firstly due to its radicalness, which allows for a wider

Reprint requests to: H. Fujita (Received for publication on Sept. 30, 1992; accepted on May 7, 1993)

resection of the esophagus and a more complete dis- section of the upper mediastinal lymph nodes; and secondly, because it is safe compared with intrathoracic anastomosis, which can be followed by severe com- plications associated with leakage. During the past ten years in Japan, esophageal cancer surgery has become increasingly radical through the introduction of aggressive cervical and mediastinal lymph node dis- section. ~ However, associated with the progress in radical lymphadenectomy are new postoperative com- plications, including pyothorax, sepsis, or multiple organ failure (MOF), subsequent to an anastomotic leak. To prevent mortality from anastomotic leakage, we have improved the modality of the initial treatment. We report herein on the risk factors predisposing to leakage from an esophageal anastomosis, including the preoperative evaluation of organ functions, and discuss the most appropriate procedure of esophageal recon- struction or anastomosis as well as the extent of cervical lymph node dissection.

Materials and Methods

During the 10 years from 1981 to 1990, 352 patients with squamous cell carcinoma in the thoracic esophagus underwent esophageal anastomosis following esoph- agectomy in our surgical department. There were 300 males and 52 females ranging in age from 38 to 86 years, with a mean age of 63 years. According to the Guidelines of the Japanese Society for Esophageal Diseases, 2 the tumor stage was: stage 0 in 35 cases (10%), stage I in 30 cases (8%), stage II in 14 cases (4%), stage III in 116 cases (33%), and stage IV in 157 cases (45%).

Following subtotal esophagectomy, esophageal anas- tomosis was generally performed in the neck, with reconstruction using the stomach, colon, or jejunum, and emplacement via a subcutaneous, retrosternal,

Y. Lee et al.: Esophageal Anas tomot ic Leaks

or posterior mediastinal route. Following lower esoph- agectomy, an anastomosis was performed in the thorax with reconstruction using the stomach or jejunum, and emplacement via an intrathoracic route. Until 1987, esophageal anastomosis was performed by hand with a single- or double-row suture in an end-to-end or end- to-side fashion. Since 1987, a circular stapling device (EEA or ILS) has been adopted for the end-to-side anastomosis.

The incidence of leakage from the esophageal an- astomosis was investigated with respect to preoperative and operative factors. The preoperative factors included age, organ functions, stage, and location of the tumor. With respect to organ functions, creatinine clearance (Ccr) was examined to assess renal function; trans- aminase, indocyanine green (ICG) secretion for 15 min (Ris), and liver biopsy specimens were examined to assess hepatic function; and the 75 g glucose tolerance test (750GTT) was used to assess whether there was any diabetes mellitus. The operative factors included the substituting organs, the route of esophageal recon- struction, and the procedure of anastomosis.

On the 9th or 10th postoperative day, a contrast esophagogram was routinely performed to check the esophageal anastomosis. Our definition of an anasto- motic leak included the following categories: (1) a minor leak detected only by radiological examination (a "radiological leak"), which usually healed spon- taneously, (2) a major leak with clinical manifestations such as a salival fistula, an approximate half of which healed spontaneously, (3) a major leak, another half of which required surgical repair, and (4) necrosis causing total disruption of the anastomosis through inadequate blood supply to the substituted organ, which always required surgical repair. Major leaks and necrosis were included in a "clinical leak".

The correlation between the occurrence of anasto- motic leak and the risk factors was investigated by statistical analysis using the chi-squared test.

Results

Incidence and Prognosis of Anastomotic Leakage

Of the 352 patients with carcinoma in the thoracic esophagus who underwent esophageal reconstruction and anastomosis following esophagectomy, 94 (27%) developed an anastomotic leak; out of this subgroup, 21 (6%) died without having tolerated any oral intake. Table 1 shows the clinical courses of these 21 patients. Before 1987, the most common procedure for esoph- ageal reconstruction was esophagogastrostomy by manual suturing via a retrosternal route. Since 1987, this has been changed to esophagogastrostomy by

25

Table 1. Incidence of anastomotic failures and hospital mortality due to leakage

No. of Anastomotic Hospital mortality Time period patients failures due to leakage

1980-86 211 74 (35%)* 18 (9%)** 1987-90 141 20 (14%)* 3 (2%)**

Total 352 94 (27%) 21 (6%)

* P < 0.01 **P < 0.05.

mechanical suturing via a subcutaneous route. Between 1980 and 1986, 74 (35%) of 211 patients who under- went esophageal reconstruction developed an anasto- motic leak. These included 16 leaks by necrosis, 37 major leaks, and 22 minor leaks, 18 (9%) of which resulted in hospital mortality. In the following period from 1987 to 1990, 20 (14%) of 141 patients who underwent esophageal reconstruction suffered from an anastomotic leak. These included 2 leaks by necrosis, 15 major leaks, and 3 minor leaks, 3 (2%) of which resulted in hospital mortality. There was a significant difference in the incidence of anastomotic leakage (P < 0.01), and also of hospital mortality (P < 0.05), be- tween the two periods.

As shown in Table 2, 18 (19%) of the 94 anastomotic leaks involved necrosis. Half of these patients under- went a second esophageal reconstruction using a free jejunal graft or a musculocutaneous flap and finally managed to tolerate oral intake; 3'4 however, the other half died before a second reconstruction of the esoph- agus could be performed. Fifty-two patients (55%) suffered from a major leak, which spontaneously healed in 29 patients, was closed by direct suturing or using a local skin flap in 11, and resulted in hospital mortality without any repair in the other 12. Twenty-four patients (26%) suffered from minor leaks, all of which spon- taneously healed. Overall, of the 94 patients with an anastomotic leak, 53 (56%) experienced spontaneous healing, 20 (21%) underwent a successful operative repair, and 21 (22%) became cases of hospital mortality before any repair could be done.

Preoperative Factors

There was no difference in the incidence of anastomotic leakage between males and females, being 27% (81/300) in males, and 25% (13/52) in females, or with respect to age, occurring in 26% (69/268) of the patients aged less than 70 years and in 30% (25/84) of those aged 70 years or older. No correlation was seen between the stage of the tumor and the incidence of anastomotic leakage, being 23% (18/79) for stages 0-I I , 28% (32/116) for stage III, and 28% (44/157) for stage IV. Moreover, no difference was found in the incidence of anastomotic

26 Y. Lee et al. : Esophageal Anastomotic Leaks

Table 2. Types of anastomotic failure and course of the affected patients

Course

Type of Spontaneously Died without anastomotic failure healed Repaired repair Total

Necrosis 0 9 a 9 18 (19%) Major leakage 29 11 b 12 52 (55%) Minor leakage 24 0 0 24 (26%)

Total 53 (56%) 20 (21%) 21 (22%) 94 (100%)

aThe esophageal defect was repaired using a free jejunal graft or musculocutaneous flap. b The esophageal fistula was closed by direct suturing or using a small skin flap.

leakage according to the location of the upper margin of the tumor, being seen in 30% (15/50) for the cervical or upper thoracic esophagus, 29% (58/201) for the middle thoracic esophagus, and 21% (21/101) for the lower thoracic esophagus.

The incidence of anastomotic leakage was 38% (20/52) in patients with liver fibrosis or cirrhosis, and 25% (75/300) in those without. Thus, patients with liver fibrosis or cirrhosis showed a greater tendency toward having an anastomotic leak than did those without, although the difference was not statistically significant (Table 3).

The incidence of anastomotic leakage was 64% (7/11) in patients with linear type diabetes mellitus and 26% (88/343) in those without. While the former showed a higher incidence of anastomotic leakage than the latter, there was no statistical difference between them (Table 4).5

Preoperat ive renal function showed no correlation with the incidence of anastomotic leakage, which de- veloped in 27% (74/274) of patients with good renal function (Ccr ~ 60ml/min), 28% (18/65) of those with intermediate renal function (30 _-> Ccr < 60ml/min), and 15% (2/13) of those with poor renal function (Ccr < 30 ml/min).

Operative Procedures

Table 5 shows the correlation between the incidence of anastomotic leakage and the procedure of esophageal reconstruction. With respect to the route of esophageal reconstruction using the stomach, the incidence of an- astomotic leakage was 30% (27/91) in patients who underwent reconstruction via a subcutaneous route, 24% (37/152) in those who underwent reconstruction via a retrosternal route, 22% (8/36) in those who underwent reconstruction via a posterior mediastinal route, and 4% (1/25) in those who underwent recon- struction via an intrathoracic route. The incidence of anastomotic leakage in those who underwent recon- struction via a subcutaneous or retrosternal route was

Table 3. Incidence of anastomotic failure with respect to liver function

Liver function No. of patients No. of failures

Normal a 204 56 (27%) Abnormal 148 38 (26%)

Chronic hepatitis b 96 18 (19%) Liver fibrosis c 18 8 (44%) Liver cirrhosis c 34 12 (35%)

a Normal transaminases and normal ICG secretion test. b Abnormal transaminases and normal ICG secretion test.

Abnormal ICG secretion test and histological diagnosis.

Table 4. Incidence of anastomotic failure with respect to diabetes mellitus

No. of No. of Diabetes patients failures

N o #1 221 58 (26%) Impaired Glucose Tolerance #2 74 16 (22%) Yes #3 57 20 (35%)

Parabolic type #1 46 13 (28%) Linear type #2 11 7 (64%)

#Cri ter ia of the National Diabetes Data Group using 750GTT (venous whole blood) 5.

#1: (1) Fasting value < 100mg/dl, (2) 2-H OGTT value < 120mg/dl, and (3) OGTF value of 1/2-h, lh, or 11/2-h < 180 mg/dl.

#2: (1) Fasting value < 120 mg/dl, (2) OGTT value of 1/2-h l-h, or 11/2-h _-> 180mg/dl, and (3) 120mg/dl < 2-h OGTT < 180 mg/dl.

#3 : (1) Fasting value _->120mg/dl, and (2) 2-h OGTT value and some other sample _->180 mg/dl.

* Our original criteria for a subtype of diabetes mellitus using 750GTT.

*1 : l-h OGTT value >2-h OGTT value. *2 : 1-h OGTT value < 2-h OGTT value.

significantly higher than in those who underwent recon- struction via an intrathoracic route (P < 0.05).

With respect to the substituting organ for esophageal reconstruction, the incidence of anastomotic leakage was 24% (73/308) in patients who underwent recon- sruction using the stomach, 60% (15/25) in those who

Y. Lee et al.: Esophageal Anastomotic Leaks

Table 5. Incidence of anatomotic failure with respect to the reconstructive route and substituted organ

27

Location of anastomosis

Organ

Reconstructive route Stomach Colon Jejunum Total

Neck

Thorax

Total

Subcutaneous 27/91 (30%) .1 9/18 (50%) 1/8 (13%) #I Retrosternal 37/156 (24%) *2 6/7 (86%) - Posterior mediastinal 8/36 (22%) -

Subtotal 72/283 (25%) 15/25 (60%) 1/8 (13%)

Intrathoracic 1 /25 ( 4 % ) *3 - 5/11 (45%) #2

73/308 (24%) *.1 15/25 (60%) **2 6/19 (32%)

37/117 (32%) 43/163 (26%)

8/36 (22%)

88/316 (28%)

6/36 (17%)

94/352 (27%)

#1 : Free jejunal graft #2 : Pedicled jejunum. "1-'3: P < 0.05, *2-*3: P < 0.01, *'1-*'2: P < 0.01.

underwent reconstruction using the colon, and 32% (6/19) in those who underwent reconstruction using the jejunum. There was a significant difference in the incidence of anastomotic leakage between patients who underwent reconstruction using the stomach and those who underwent reconstruction using the colon (P < 0.01).

Table 6 shows the correlation between the anasto- motic technique and leakage. The incidence of an- astomotic leakage was 33% (75/225) in patients who underwent a manual anastomosis, resulting in necrosis in 15 patients, a major leak in 38, and a minor leak in 22, while it was 15% (19/127) in those who underwent a mechanical anastomosis, resulting in necrosis in 3 patients, a major leak in 14, and a minor leak in 2. The incidence of anastomotic leakage in patients who underwent manual anastomosis was significantly higher than in those who underwent mechanical anastomosis (P < 0.01).

There was no difference in the incidence of anasto- motic leakage among the various kinds of anastomotic techniques such as the Albert-Lembert double layer anastomosis, the layer-to-layer or double-layer anasto- mosis, and the Olsen single-layer anastomosis. On the other hand, there was a significant difference in the

Table 6. Incidence of anastomotic failure according to the various anastomotic techniques

No. of No. of Technique patients failures

Manual 225 75 (33%)** Albert-Lembert (double layers) 16 5 (31%) Layer-to-layer (double layers) 178 60 (34%) Olsen (single layer) 28 8 (29%) Gambee (single layer) 3 2 (67%)

Mechanical 127 19 (15%)** End-to-end 7 4 (57O/o) * End-to-side 120 15 (13%)*

*P < 0.05, **P < 0.01.

incidence of anastomotic leakage between patients who underwent a mechanical anastomosis using an end-to- end technique and an end-to-side technique, being 57% v s 13% respectively, (P < 0.05).

Table 7 shows the correlation between the incidence of anastomotic leakage and the procedure of cervical lymph node dissection. There were no differences among the incidences of anastomotic leakage when there was no dissection, left cervical node dissection, right cervical node dissection, or bilateral cervical node dissection,

Discussion

Following extirpative surgery for esophageal car- cinoma, leakage from the esophageal anastomosis remains a serious problem. Its incidence generally de- pends on the surgical technique and procedure, the preoperative conditions, and the incidence of post- operative complications.

The incidence of anastomotic leakage has been re- ported to be as low as 2% ,6 and as high as 36% ,7 but these rates clearly depend on the definition of an anastomotic leak. We report its incidence in this study to be 27%, by a definition which included four kinds of anastomotic leaks: (1) necrosis, (2) a major leak with clinical manifestations requiring surgical repair, (3) a major leak which spontaneously healed, and (4) a minor leak (a "radiological leak"), as cited above. The

Table 7. Incidence of anastomotic failure with respect to neck dissection

Neck dissection No. of patients No. of failures

No 234 59 (25%) Yes 118 35 (30%)

Bilateral 72 21 (29%) Right 5 2 (40%) Left 41 12 (29%)

28 Y. Lee et al.: Esophageal Anastomotic Leaks

incidence of a clinical leak, which included all except radiological leaks, was 20%, while the incidence of a leak which needed a surgical repair was 12%.

it is well known that leakage from a cervical an- astomosis often closes spontaneously after adequate drainage, whereas leakage from an intrathoracic an- astomosis is associated with a very high mortality rate, as reported by Lu et al. s Until 1987, we routinely peformed a cervical esophagogastrostomy via a retro- sternal route; however, since the introduction of a more aggressive dissection of the cervical and upper mediastinal lymph nodes, being the 3-field dissection, we have experienced several patients who developed pyothorax following leakage from the cervical an- astomosis which resulted in mortNity. In order to improve the postoperative course, we have routinely performed esophageal reconstruction since 1987 via a subcutaneous route for patients undergoing the 3-field dissection. Since then, we have had no-6xperience of pyothorax as a result of leakage, and the overall rate of hospital mortality following esophagectomy for esophageal carcinoma has markedly improved from 26% (55/211) to 6% (9/141) (P < 0.05).

Many authors have recommended mechanical an- astomosis offering a lower incidence of leakage than manual anastomosis, as observed in our series. 9 Intro- ducing mechanical anastomosis, we performed end-to- end anastomosis in seven patients using a stapler, four of whom (57%) developed a leak. Therefore, we modified the mechanical technique to perform an end-to-side anastomosis, after which the incidence of leakage decreased to 13%.

There remains some controversy regarding the optimum shape of the gastric pedicle. Some authors reported that a thin gastric tube using the greater curvature improved blood circulation, 1°'11 while others reported that a thick gastric tube using the subtotal stomach resulted in better circulation. ~2'13 The thin gastric tube is longer than the thick gastric tube and more suitable for end-to-end anastomosis. In contrast, the thick gastric tube is suitable for end-to-side anasto- mosis using a circular stapler. We commonly produce a thick gastric tube for mechanical anastomosis, unless the stomach is too small to pull up in the neck.

Hermreck and Crawford 14 reported that leakage oc- curred in 5 (42%) of 12 patients who underwent an esophagocolostomy, and in only 5 (18%) of 28 patients after an esophagogastrostomy. In fact, many authors have reported a high incidence of anastomtic leakage following esophagocolostomy. 15 In our series, the in- cidence of leakage after esophagocolostomy was very high because the colon was used only for patients who had undergone a previous gastrectomy or for those in whom an esophagogastrectomy was needed due to double primary carcinomas in the esophagus and

stomach. Thus, since 1989, we have added vascular anastomoses immediately after esophagocolostomy using microvascular techniques between the vessels in the neck or chest wall and in the tip of the colonic pedicle, a6 and no anastomotic leakage has occurred in any of the five patients on whom we have used this technique.

Peracchia et al. 9 reported that leakage from an esophageal anastomosis was not correlated with dia- betes mellitus, renal disease, or hepatic disease, whe- ther singly or in combination, although in our series, an anastomotic leak was more frequently observed in patients with liver fibrosis or cirrhosis, or linear-type diabetes mellitus, but without a statistically significant difference. Thus, for patients with those diseases, we have recently begun to use a two-stage operation consisting of esophagectomy and esophageal recon- struction with cervical esophagostomy in the first stage, and esophageal anastomosis alone in the second stage, performed around 1 month after the first stage. In the patients who have so far undergone the two-stage operation, there has been no experience of necrosis or a major leak, or of hospital mortality.

In conclusion, esophageal anastomotic leakage was more frequently observed: (1) after reconstruction via a retrosternal or subcutaneous route, (2) when recon- struction involved a colonic interposition, (3) when manual anastomosis was used, and (4) when end-to-end mechanical anastomosis was used. Our technical im- provements in esophageal reconstruction and anasto- mosis, using a two-stage operation when indicated by a preoperative risk analysis, and employing careful pre-, peri-, and postoperative management are the key factors to decreasing the incidence of anastomotic leakage and consequent hospital mortality.

References

1. Isono K, Sato H, Nakayama K (1991) Results of a nationwide study on the three-field lymph node dissection of esophageal cancer. Oncology 48:411-420

2. Japanese Society for Esophageal Diseases (1992) Guidelines for the clinical and pathologic studies on carcinoma of the esophagus. Kanehara, Tokyo

3. Fujita H, Kakegawa T, Tai Y, Yamana H, Shirohzu G, Yoshimura Y, Minami T, Negoto Y, Irie H, Shima I, Machi J (1988) Surgical treatment of cervical esophageal fistula following reconstruction. Dis Esophag 1:179-183

4. Fujita H, Inoue Y, Tanaka S, Kakegawa T, Yamana H, Shirohzu G, Minami T, Tai Y, Hirano M (1991) Atypical free-gut transfer for esophageal reconstruction (in Japanese with English abstract). Nippon Kikan Shyokudouka Gakkai Kaiho (J Jpn Broncho- esophagol Soc) 42:27-34

5. National Diabetes Data Group (1979) Classification and diagnosis of diabetes mellitus and other categories of glucose intolerance. Diabetes 28:1039-1057

6. Dark JF, Mousalli H, Vaughan R (1981) Surgical treatment of carcinoma of the esophagus. Thorax 36:891-895

Y. Lee et al.: Esophageal Anastomotic Leaks 29

7. Postlethwait RW (1983) Complications and deaths after operation for esophageal carcinoma. J Thorac Cardiovasc Surg 85:827-831

8. Lu YK, Li YM, Gu YZ (1987) Cancer of esophagus and esoph- agogastric junction: Analysis of results of 1025 resections after 5 to 20 years. Ann Thorac Surg 43:176-181

9. Peracchia A, Bardini R, Ruol A, Asolati M, Scibetta D (1988) Esophageal anastomotic leak: A prospective statistical study of predisposing factors. J Thorac Cardiovasc Surg 95:685-691

10. Sugimachi K, Inokuchi K, Okudaira Y, Ueo H, Natsuda Y, Ikeda M, Nakamura T, Yaita A (1979) A safer and more reliable operative technique for esophageal reconstruction using a gastric tube (in Japanese with English abstract). Nippon Geka Gakkai Zasshi (J Jpn Surg Soc) 80:1164-1167

11. Sugimachi K, Ikeda M, Ueo H, Kai M, Okudaira Y, Inokuchi K (1982) Clinical efficacy of the stapled anastomosis in esophageal

reconstruction. Ann Thorac Surg 33:374-378 12. Akiyama H, Miyazono H, Tsurumaru M, Kawamura T (1979)

Use of the stomach as a substitute of the esophagus (in Japanese). Rinshyo Geka (J Clin Surg) 34:695-702

13. Isono K, Onoda S, Okuyama K, Yamamoto Y, Asano T, Koide Y, Sato H (1985) Protection against suture insufficiency in ante- thoracic esophagogastrostomy. Jpn J Surg 15:43-48

14. Hermreck AS, Crawford DG (1976) The esophageal anastomotic leak. Am J Surg 132:794-798

15. EI-Domeiri A, Martini N, Beattie EJ (1970) Esophageal recon- struction by colon interposition. Arch Surg 100:358-362

16. Fujita H, Inoue Y, Kakegawa T, Yamana H, Shirohzu G, Minami T, Tai Y (1991) Esophageal reconstruction using micro- vascular anastomosis to the thoracoacromial artery and cephalic vein. Jpn J Surg 21:512-516