The British Journal of Radiology, 81 (2008), 545–548
Radiological detection of post-oesophagectomy anastomotic leak — a comparison between multidetector CT and fluoroscopy 1
S UPPONI, MPhil, FRCS, FRCR, 1A GANESHAN, MRCP, FRCR, 1H D’COSTA, FRCS, FRCR, 1M BETTS, 2 N MAYNARD, MS, FRCS, 1H BUNGAY, MRCP, FRCR and 1A SLATER, MRCP, FRCR
MRCP, FRCR,
Departments of 1Clinical Radiology and 2Surgery, John Radcliffe Hospital, Headley Way, Headington, Oxford OX3 9DZ, UK
ABSTRACT. The purpose of this study was to directly compare CT with fluoroscopy for the diagnosis of occult anastomotic leak following oesophagectomy. Patients undergoing oesophagectomy and gastric conduit formation for the treatment of oesophageal cancer were eligible for inclusion. Imaging was performed 6–8 days postoperatively. Patients underwent multislice CT examination of the chest and abdomen with a bolus of oral contrast, followed by fluoroscopic water-soluble contrast swallow (with subsequent use of barium if this was normal). The studies were reviewed by a consultant radiologist, who was blinded to the results of the other modality. Images were reported as showing ‘‘no leak’’, ‘‘possible leak’’ or ‘‘definite leak’’. The presence of mediastinal gas or fluid or extraluminal contrast at CT was recorded. The clinical outcome after reinstituition of oral intake was used as a reference standard. Patient preference for modality was recorded. 52 patients were recruited. Four were found to have leak on CT and fluoroscopy. 11 had possible leak at CT, but normal fluoroscopy: 2 of these had a leak confirmed later, whereas 9 had no leak. 37 had normal CT and fluoroscopy findings, and remained clinically well. The sensitivity, specificity, positive and negative predictive values were 100%, 80%, 40% and 100%, respectively, for CT, and 67%, 100%, 100% and 96%, respectively, for fluoroscopy. The positive predictive value of mediastinal air, air/fluid and extraluminal contrast were 25%, 75% and 50%, respectively. 35 patients found CT more tolerable. In conclusion, CT was better tolerated and more sensitive but less specific than fluoroscopy for detecting occult anastomotic leak.
The incidence of oesophageal cancer is rising predominantly because of the increasing incidence of adenocarcinoma [1, 2]. In operable cases, the most common procedure performed is that of oesophagectomy with formation of a gastric conduit. Both transthoracic and transhiatal approaches are described for oesophagectomy, and consequently the oesophagogastric anastomosis can be placed either in the neck (cervical anastomosis) or in the chest (thoracic anastomosis) according to surgical preference, site of tumour and fitness of patient. The incidence of leak following oesophagectomy for malignancy is between 4% and 8% [3, 4]. Symptomatic leaks can be associated with considerable morbidity and mortality [5–7]. At this time, there is little consensus as to if and when routine imaging should be undertaken in the post-operative period for the detection of occult leaks. The clinical significance of occult leaks is debatable. Currently, post-operative evaluation of occult leak following oesophagectomy is undertaken using fluoroscopic imaging with either water-soluble contrast alone or followed by barium sulphate [8]. Although highly specific, it has been shown that this technique is limited Address correspondence to: Dr A Slater, Department of Clinical Radiology, John Radcliffe Hospital, Headley Way, Headington, Oxford OX3 9DZ, UK. E-mail:
[email protected]
The British Journal of Radiology, July 2008
Received 28 February 2007 Revised 30 July 2007 Accepted 8 August 2007 DOI: 10.1259/bjr/30515892 ’ 2008 The British Institute of Radiology
by low sensitivity. This is particularly the case when only water-soluble contrast is used [9]. The addition of barium has been shown to increase sensitivity [10], although is not used routinely in many centres. It has been reported that 15% of leaks undetected at water-soluble contrast swallow can be detected by the addition of barium sulphate [11]. The use of high-density barium can further improve sensitivity [12]. In the early post-operative period, fluoroscopy is often ill tolerated by patients owing to general morbidity (predominantly respiratory) and difficulty with mobilization. The aim of this study was to evaluate prospectively the use of CT in the assessment of post-operative occult leak following oesophagectomy, and to compare this directly with fluoroscopy.
Methods and materials Local ethics committee approval was obtained. Patient inclusion criteria for the study were those who had undergone oesophagectomy using either a right or left thoracoabdominal approach or a transhiatal approach for malignancy, with construction of a gastric conduit only. Successive patients were recruited following surgery. Imaging was undertaken between days 6 and 8 after the operation. Patients were excluded from the study if 545
S Upponi, A Ganeshan, H D’Costa et al
source. Occult leak is only defined radiologically. Patient preference for modality of examination was also recorded.
they had undergone fluoroscopy or CT examination prior to the sixth post-operative day. CT evaluation was the first imaging study undertaken, immediately followed by fluoroscopic assessment at the same attendance. CT examinations were performed on a 4- or 16-slice helical GE lightspeed scanner (GE medical systems, Milwaukee, WI). 50 ml of diluted water-soluble oral contrast, consisting of 5 ml of iodinated contrast medium (Iopamidol, Niopam 300; Bracco UK Ltd, Bucks, UK) in 750 ml of water, was administered prior to the study. This was followed, if the patient was able to tolerate it, by a bolus of dilute oral contrast given immediately before the scan. The patient was imaged in the supine position only. No intravenous contrast was administered. CT images were acquired at 5 mm collimation and 1.5 pitch. Images were viewed in the axial plane and reconstructed in the coronal plane if necessary. CT studies were examined for the presence or absence of extra luminal gas, mediastinal fluid or extraluminal contrast. CT studies were then reported as ‘‘definite leak’’, ‘‘possible leak’’ or ‘‘no leak’’. Definite leak was defined as the presence of extraluminal contrast, mediastinal fluid, and gas collections or multiple locules of mediastinal gas. Possible leak was defined as small volumes of fluid and gas, or single locules of mediastinal gas. Contrast swallow examinations were performed by a consultant or specialist registrar in radiology. Fluoroscopy examination was undertaken according to our local protocol. The patient is positioned at a 30 ˚ caudal tilt on the table to prevent rapid passage of contrast across the anastomosis, while minimizing the risk of aspiration. The patient swallows a bolus of contrast in two positions: anteroposterior (AP) and lateral. If this is normal, the patient then drinks boluses of barium (100% weight/volume, Baritop; Bioglan Laboratories, Hitchin, UK) in four positions: AP, right lateral, left lateral and prone (if the patient is able). The purpose is to observe the primary anastomosis and the gastric staple line. In order to assess the gastric staple line fully, the patient is occasionally and cautiously tilted cranially to distend the gastric fundus. Fluoroscopic leak was defined as direct visualization of contrast leak beyond either the oesphagogastric staple line, gastric body staple line or gastrostomy line. The two studies were independently read by a consultant gastrointestinal radiologist. The reviewer was blinded to the results of the other modality. The reference standard for the presence of a leak was a clinically apparent leak at the time of, or subsequent to, imaging. Clinical leak was defined as a visible leak of saliva or fluid from a neck wound, or clinical signs of sepsis in a patient with no other demonstrable clinical
Results 52 patients were recruited to the study: 38 male patients and 14 female patients. The mean age was 63 years, with a median age of 66 years (range 35– 80 years). There were 32 mediastinal and 20 cervical anastomoses. Five patients had a clinically apparent leak either at the time of imaging or following radiological assessment. One patient had an occult leak. 37 patients showed no evidence of leak at CT and remained clinically well. Four patients had definite leak identified at CT and clinically defined at the time of imaging. 11 possible leaks were reported on CT. In one patient, a leak became clinically apparent following imaging and, in a further patient, an occult leak was subsequently defined on fluoroscopy. The remaining nine patients did not show evidence of clinical leak. CT therefore demonstrated 100% sensitivity and 80% specificity (Table 1). 46 patients had normal fluoroscopy and normal clinical outcome. In four cases, a leak was identified at fluoroscopy as well as clinically. In one patient with normal fluoroscopy, there was a subsequent clinical leak defined, as well as an occult leak in a further patient. Fluoroscopy showed 67% sensitivity and 100% specificity (Table 1). Of the four patients with definite leak defined at CT, two had mediastinal fluid and air (Figures 1 and 2), one had mediastinal air alone, and one had direct leak of contrast into the pleural space. Of the 11 patients with possible leak at CT, two had small volumes of mediastinal fluid and air reported on CT. One of these patients developed clinical symptoms relating to a leak, which was confirmed on a subsequent CT scan (Figures 3 and 4). At the time of initial imaging, there were no clinical features to suggest a leak. Following this case, all patients with possible leak on CT and normal fluoroscopy underwent repeat fluoroscopy the following day to establish the presence of a leak. A further patient with fluid and air present on CT remained clinically well following oral intake. 7 of these 11 patients had air locules defined only on CT. One of these had an occult leak described only on repeat fluoroscopy, but remained clinically well following oral intake. Of the two remaining patients, one had unexplained free intraperitoneal air but remained well; the other had high-attenuation fluid in the pleural space, which was thought initially to be contrast but, on subsequent review, was likely to represent haemorrhage.
Table 1. Comparison of CT, fluoroscopy and reference standard Leak
Clinically apparent leak/ 6 occult leak CT-defined leak 15 Fluoroscopy-defined leak 4
No leak
TP
FP
TN
FN
Sensitivity
Specificity
PPV
NPV
6 4
9 0
37 46
0 2
100 67
80 100
40 100
100 96
46 37 48
PPV, positive predictive value; NPV, negative predictive value; TP, true positive; FP, false positive; TN, true negative; FN, false negative.
546
The British Journal of Radiology, July 2008
Radiological detection of post-oesophagectomy anastomotic leak Table 2. The positive predictive values of the CT signs of leak
Leak No leak PPV
Air locules alone
Air/fluid
Leak of ‘‘contrast’’
2 6 25
3 1 75
1 1 50
Includes 15 patients with leak or possible leak reported on CT. A single patient with free intraperitoneal air is excluded. PPV, positive predictive value.
Figure 2. CT of the same patient described in Figure 1 showing a large air locule (arrow), indicating leak, and fluid adjacent to the gastric pull-through (arrowheads).
Figure 3. CT showing a large air locule (short arrow) and small volume of fluid adjacent to the gastric pull-through (long arrow).
Figure 4. CT of the patient described in Figure 3 showing Figure 1. Leak of water-soluble contrast (arrow) from the oesophagogastric anastomosis into the superior mediastinum at fluoroscopy.
The British Journal of Radiology, July 2008
progression of leak on subsequent imaging. The long arrow indicates a mediastinal collection, and the short arrow a small pleural effusion.
547
S Upponi, A Ganeshan, H D’Costa et al
The positive predictive values of the CT signs used to assess the presence of leak are given in Table 2. Patients were asked to record their preference for imaging modality. 35 patients found CT more tolerable; 7 patients preferred fluoroscopy; and 8 had no preference. In two cases, preference was not recorded.
Discussion Leak following oesophagectomy is associated with considerable morbidity and mortality. The clinical relevance of occult leak is unknown; however, it remains common practice in many centres to undergo routine radiological assessment in the post-operative period prior to removal of the nasogastric tube placed at the time of surgery and instigation of oral intake. As reported, CT shows high sensitivity but is limited by poor specificity. Fluoroscopy is highly specific but lacks sensitivity, which is further reduced when detecting occult leak rather than clinically apparent leak. There are no studies to date reporting normal appearances of the mediastinum at CT following oesophagectomy, as imaging using this modality is not usually undertaken. Therefore, specificity may be low because the presence of any air was defined as possible leak for the purposes of this study. However, it is recognized that this is likely to be a normal post-operative finding at 6–8 days, as is the case for patients undergoing midline sternotomy [13, 14]. Air locules alone showed a poor positive predictive value on CT in defining a leak. In our study, all patients with symptoms attributable to an anastomotic leak had this defined by both fluoroscopy and CT. In one case in which a patient was suspected of having a leak at CT but remained clinically well, repeat fluoroscopy revealed an occult leak. This had not been identified on initial fluoroscopic examination. The clinical relevance of an entirely occult anastomotic leak is debatable. Given the poor sensitivity of fluoroscopy in the assessment of leak, its ‘‘routine’’ application in clinically well patients is limited. Imaging may, therefore, be more appropriate in cases where patients have clinical symptoms of leak. If ‘‘routine ‘‘ imaging is to continue, CT may be more appropriate owing to its high sensitivity and greater patient acceptability, with subsequent fluoroscopic examination being used in a subset of patients with suspicious, but not definitive, findings at CT.
548
References 1. Pera M, Manterola C, Vidal O, Grande L. Epidemiology of esophageal adenocarcinoma. Surg Oncol 2005;92:151–9. 2. Wei JT, Shaheen N. The changing epidemiology of esophageal adenocarcinoma. Semin Gastrointest Dis 2003;14:112–27. 3. Karl RC, Schreiber R, Boulware D, Baker S, Coppola D. Factors affecting morbidity, mortality and survival in patients undergoing Ivor Lewis esophagectomy. Ann Surg 2000;231:635–43. 4. Valverde A, Hay JM, Fingerhut A, Elhadad A. Manual versus mechanical esophagogastric anastomosis after resection for carcinoma: a controlled trial. French Association for Surgical Research. Surgery 1996;120:476–83. 5. Alanezi K, Urschel JD. Mortality secondary to anastomotic leak. Ann Thorac Cardiovasc Surg 2004;10:71–5. 6. Crestanello JA, Deschamps C, Cassivi SD, Nichols FC, Allen MS, Schleck C, et al. Selective management of intrathoracic leak after oesophagectomy. J Thorac Cardivasc Surg 2005;129:254–60. 7. Juneman-Ramirez M, Awan MY, Khan Zm, Rahamim JS. Anastomotic leakage post – esophagectomy for esophageal carcinoma: retrospective analysis of predictive factors, management and influence on longterm survival in a high volume centre. Eur J Cardiothorac Surg 2005;27:3–7. 8. Bruce J, Krukowski ZH, Al-Khairy G, Russell EM, Park KGM. Systematic review of the definition and measurement of anastomotic leak after gastrointestinal surgery. Br J Surg 2001;88:1157–68. 9. Tirnaksiz MB, Deschamps C, Allen MS, Johnson DC, Pairolero PC. Effectiveness of screening aqueous contrast swallow in detecting clinically significant anastomotic leaks after esophagectomy. Eur Surg Res 2005;37:123–8. 10. Gollub MJ, Bains MS. Barium sulfate: a new (old) contrast agent for diagnosis of postoperative esophageal leaks. Radiology 1997;202:360–2. 11. Tanomkiat W, Galassi W. Barium sulfate as contrast medium for evaluation of postoperative anastomotic leaks. Acta Radiol 2000;41:482–5. 12. Swanson JO, Levine MS, Redfern R, Rubesin SE. Usefulness of high-density barium for detection of leaks after esophagogastrectomy, total gastrectomy and total laryngectomy. AJR Am J Roentgenol 2003;181:415–20. 13. Kay HR, Goodman LR, Teplick SK, Mundth ED. Complications of median sternotomy: computed tomographic evaluation. AJR Am J Roentgenol 1983;141:225–30. 14. Kay HR. Use of computed tomography to assess mediastinal complications after median sternotomy. Ann Thorac Surg 1983;36:706–14.
The British Journal of Radiology, July 2008