HPB SURGERY The role of magnetic resonance cholangiopancreatography in the management of acute gallstone pancreatitis

INTRODUCTION The aim of this study was to identify whether magnetic resonance cholangiopancreatography (MRCP) can be used selectively in patients with acute gallstone pancreatitis to detect choledocholithiasis, based on liver function tests (LFTs) and ultrasonography appearance. METHODS All patients admitted between January 2008 and January 2011 with gallstone pancreatitis (amylase >300u/l) who underwent MRCP were included in the study. LFTs and radiology reports were obtained from the respective computer systems. RESULTS Overall, 173 patients with acute gallstone pancreatitis underwent MRCP and 30% (52/173) showed choledocholithiasis. The mean bilirubin level was significantly higher in those with choledocholithiasis (46 ±5µmol/l vs 36 ±3µmol/l, p =0.0388) although there was no significant difference in alkaline phosphatase (276 ±25iu/l vs 229 ±16iu/l, p =0.1154). However, sensitivity of abnormal bilirubin (>21µmol/l) for choledocholithiasis was only 62% and specificity was 41%. Sensitivity

Acute gallstone pancreatitis is precipitated by obstruction of the distal common bile duct (CBD) by migrating gallstones. 1 This biliary obstruction is usually transient, with the stones passing spontaneously into the duodenum. 2 However, it is important to identify those patients with persistent stones obstructing the CBD as prompt intervention has been shown to reduce progression to severe disease and decrease overall morbidity. 3,4 Traditionally, investigation for choledocholithiasis involved endoscopic retrograde cholangiopancreatography (ERCP) but this is an invasive procedure with risks of perforation, pancreatitis and, in a small minority, death. 5,6 Over recent years, investigation has therefore moved increasingly towards magnetic resonance cholangiopancreatography (MRCP). This non-invasive technique has been shown to be comparable with ERCP in the detection of choledocholithiasis, both generally [7][8][9][10][11][12] and specifically in acute gallstone pancreatitis. [13][14][15][16][17] It has been shown to have a high negative predictive rate for choledocholithiasis when used unselectively. 18 Nevertheless, the accuracy may fall when MRCP is used selectively to evaluate patients with acute gallstone pancreatitis without strong clinical suspicion of choledocholithiasis. 19 The current evidence does not therefore make clear whether MRCP should be used routinely or selectively in the management of patients with acute gallstone pancreatitis. The British Society of Gastroenterology guidelines on the management of acute pancreatitis state that all patients BARLOW HAqq MCCORMACK METCALFE DENNISON GARCEA tHE RolE of mAGnEtic RESonAncE cHolAnGioPAncREAtoGRAPHY in tHE mAnAGEmEnt of AcUtE GAllStonE PAncREAtitiS with acute gallstone pancreatitis require imaging of the bile duct 20 but the optimal method of imaging has yet to be ascertained. One approach is to use MRCP selectively, based on ultrasonography appearances and deranged liver function tests (LFTs) to identify those patients at risk of choledocholithiasis. However, there is evidence that this approach is not sensitive enough to exclude choledocholithiasis reliably and we have recently adopted near routine use of MRCP in the management of acute gallstone pancreatitis. The aim of this study was to assess the validity of this approach and investigate whether the use of MRCP could be rationalised based on ultrasonography appearances or LFTs. methods All patients discharged with a diagnosis of pancreatitis between January 2008 and January 2011 were identified retrospectively from the hospital coding system. The hospital laboratory and radiology computer systems were then used to identify those patients with a diagnosis of acute gallstone pancreatitis as evidenced by serum amylase of >300u/l and the presence of gallstones on concurrent or previous radiology imaging. Those patients with acute gallstone pancreatitis who underwent MRCP as part of their initial investigation were included in the study. This was therefore not a series of consecutive patients admitted with gallstone pancreatitis but a consecutive series of patients who underwent MRCP for gallstone pancreatitis.
MRCP was performed according to an agreed protocol: T1 in/out phase, T2 HASTE, T2 HASTE thin slice, T2 true-FISP, T2 coronal thick slab, T2 HASTE thin slab. MRCP films were reported by one of a team of radiologists, who all had a special interest in hepatobiliary imaging. The median time to MRCP from admission was 4 days (interquartile range: 2.5-9.5 days).
The presence of choledocholithiasis on MRCP was correlated with LFTs measured at and during admission as well as with ultrasonography appearances. Biliary dilatation was defined as a CBD of ≥8mm. The sensitivity and specificity of abnormal LFTs and biliary dilatation at ultrasonography for predicting choledocholithiasis was calculated using receiver operating characteristic (ROC) curves. All statistical analysis was performed using Prism ® version 5.0 (GraphPad Software, La Jolla, CA, US).

Results
Overall, 265 patients were admitted with acute gallstone pancreatitis between January 2008 and January 2011, of whom 173 (65%, 102 women and 71 men) underwent MRCP for choledocholithiasis. The median age of the study population was 59 years (range: 42-74.5 years). A fifth (21%) of patients had severe pancreatitis, defined as a Glasgow score on admission of ≥3. Sixty-nine per cent (120/173) of patients had abnormally raised bilirubin (>17µmol/l) on admission and sixty-two per cent (107/173) had abnormally raised alkaline phosphatase (ALP) (>140iu/l) while forty-three per cent (75/173) of patients had abnormally raised bilirubin and ALP. A fifth (20%, 34/173) of patients had neither abnormal bilirubin nor ALP at admission.
Choledocholithiasis was detected on MRCP in 30% (52/173) of patients. The characteristics of those patients with and without choledocholithiasis are shown in Table 1. Those patients with choledocholithiasis were significantly older and had a significantly higher bilirubin level at admission although this difference was small and probably not clinically relevant. The median Glasgow score was also higher in those patients with choledocholithiasis. However, this just failed to reach statistical significance.
Using ROC analysis (Fig 1), the sensitivity and specificity of raised bilirubin (>21µmol/l) for predicting choledocholithiasis on MRCP were 71% (95% CI: 57-83%) and 31% (95% CI: 23-40%) respectively. The sensitivity and specificity of raised ALP (>130iu/l) were 73% (95% CI: 59-84%) and The trend in LFTs over the course of the patients' admission was also investigated for any relationship with choledocholithiasis on MRCP. Of those patients with abnormal LFTs (n=139) at admission (raised bilirubin, ALP or both), 66% (92/139) remained persistently deranged and 25% (35/139) returned to normal. (Twelve patients did not have LFTs rechecked following admission.) However, there was no difference in the incidence of choledocholithiasis on MRCP for those patients with persistently deranged LFTs versus those whose LFTs returned to normal (relative risk: 1.07, 95% CI: 0.61-1.89, p=1.00). The resolution of abnormal LFTs cannot therefore be taken to demonstrate passage of a previously retained CBD stone.
Overall, there were five patients (10% of those with choledocholithiasis) with choledocholithiasis on MRCP who had entirely normal LFTs on admission and no biliary dilatation or choledocholithiasis on ultrasonography. Conversely, nine patients with abnormal LFTs on admission and biliary dilatation on ultrasonography did not have choledocholithiasis on MRCP.
Of the 52 patients with choledocholithiasis on MRCP, 38 went on to have ERCP. Choledocholithiasis was confirmed in 89% (34/38) of these patients. ERCP was unsuccessful in two patients. Of the remaining twelve patients with choledochiolithiasis on MRCP, four underwent a laparoscopic cholecystectomy and on-table cholangiography (OTC) demonstrating a clear CBD, three underwent CBD exploration, one underwent biliary bypass and four were deemed unsuitable for further treatment due to co-morbidities.
Of those patients undergoing MRCP for acute gallstone pancreatitis, 40 also underwent computed tomography (CT). This demonstrated choledocholithiasis in 4 out of a total of 12 patients with choledocholithiasis on MRCP who also underwent CT, giving an accuracy of 33%. No patients had choledocholithiasis demonstrated on CT but not on MRCP.

discussion
To our knowledge, this study is the largest to date to evaluate the role of MRCP in acute gallstone pancreatitis. It was performed with the specific aim of determining whether MRCP was necessary in all patients presenting with acute gallstone pancreatitis or whether it could be used selectively based on abnormal LFTs and or ultrasonography findings.
It is clear from these results that that neither abnormal LFTs nor biliary dilatation on ultrasonography predict choledocholithiasis on MRCP with enough sensitivity or specificity to allow selective use of MRCP. Furthermore, those patients with resolution of abnormal LFTs during admission are at no less risk of choledocholithiasis than those with persistently deranged LFTs. Consequently, the authors feel that all patients with acute gallstone pancreatitis should undergo MRCP, regardless of LFTs and ultrasonography findings. Although deranged bilirubin, ALT and ALP, and biliary dilatation on ultrasonography was 90% specific for choledocholithiasis on MRCP, it is important to note that 10% of patients with choledocholithiasis had no abnormalities on LFTs or ultrasonography. By excluding those patients with normal LFTs and ultrasonography from undergoing MRCP, up to 10% of bile duct stones may be missed.
These findings concur with those of other studies. Makary et al showed a sensitivity for raised bilirubin levels to predict choledocholithiasis on MRCP and ERCP of 65% and for biliary dilatation on ultrasonography of 55%. 21 Telem et al incorporated CBD size on ultrasonography, gammaglutamyltransferase, ALP, total bilirubin and direct bilirubin into a predictive tool for choledocholithiasis. 22 The presence of four or five abnormal variables correlated significantly with choledocholithiasis. Importantly, however, 18% of the patients with choledocholithiasis had no abnormality in any of the five variables.
The large number of patients involved in this study improves the reliability of the results. Despite this, as it was a retrospective study, data were not complete for all patients. This mainly affected repeat measurement of LFTs after admission. Furthermore, the retrospective nature meant there was no standard reporting system for either ultrasonog- The authors also accept that a median time of four days from admission to MRCP is short. This potentially reduces the opportunity for spontaneous passage of bile duct stones and could have led to unnecessary intervention in patients who would otherwise have passed stones uneventfully. The primary driving force for early MRCP was to facilitate a cholecystectomy during the same hospital admission, and to prevent delays in surgery and subsequent discharge while exclusion of choledocholithiasis was awaited.
A further source of bias is that not all patients admitted with acute gallstone pancreatitis during the study period underwent MRCP although two-thirds did. However, those that did not are most likely to be those with normal LFTs and ultrasonography appearance of the biliary tree. As these patients are less likely to have choledocholithiasis, their relative exclusion from the study is a confounding factor.
An alternative approach would be to perform OTC at the time of laparoscopic cholecystectomy in those patients deemed low risk for choledocholithiasis. Nevertheless, it is clear from the above findings that it is not easy to differentiate which patients are low risk based on LFTs and ultrasonography appearances. Performing OTC has logistical implications, with delays to theatre lists while waiting for a radiographer and increased operating time. Furthermore, if CBD stones are discovered, the question is whether to perform a CBD exploration with further disruption to the operating list or to perform postoperative ERCP with an increased risk of precipitating a bile leak.
OTC is also not necessarily more cost effective; MRCP costs in the region of £200 and, while the additional tariff for OTC over a standard laparoscopic cholecystectomy is £90, this does not factor in lost theatre time. Overall, OTC is unlikely to be significantly cheaper. Consequently, OTC is not the preferred option in this centre unless MRCP is not technically feasible owing to patient size or metallic implants.
Despite this, the authors believe that all patients with acute gallstone pancreatitis should undergo MRCP as part of their initial investigation. This study has shown there is no reliable method for selecting patients for MRCP based on laboratory or ultrasonography findings. Indeed, 10% of patients in this study with choledocholithiasis had no abnormal laboratory or ultrasonography findings. Although this approach of routine MRCP does entail some patients without choledocholithiasis having unnecessary imaging, it ensures all patients with choledocholithiasis are identified. As MRCP is a non-invasive procedure with minimal risk, and as delay in identification and treatment of choledocholithiasis increases the risk of haemorrhage, sepsis and necrotising pancreatitis, this seems an appropriate strategy. 2,23,24 conclusions All patients with acute gallstone pancreatitis should undergo specific imaging, preferably MRCP, to exclude choledocholithiasis as LFTs and ultrasonography appearances are inaccurate in predicting the presence of CBD stones.