Allopurinol to Prevent Pancreatitis After Endoscopic Retrograde Cholangiopancreatography: A Randomized Placebo-Controlled Trial

Transcription

1 CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2008;6: Allopurinol to Prevent Pancreatitis After Endoscopic Retrograde Cholangiopancreatography: A Randomized Placebo-Controlled Trial JOSEPH ROMAGNUOLO,*, ROBERT HILSDEN, GURPAL S. SANDHA, MARTY COLE, SYD BASS, GARY MAY, JONATHAN LOVE, VINCENT G. BAIN, JOHN MCKAIGNEY, and RICHARD N. FEDORAK *Digestive Disease Center, Department of Medicine, and Department of Biostatistics, Bioinformatics and Epidemiology, Medical University of South Carolina, South Carolina; Department of Medicine, University of Calgary, Calgary, Alberta, Canada; Department of Medicine, University of Alberta, Edmonton, Alberta, Canada; and the Department of Medicine, University of Toronto, Ontario, Canada See CME exam on page 371. See Editorial on page 374. Background & Aims: Endoscopic retrograde cholangiopancreatography (ERCP) is associated with a risk of pancreatitis (PEP). Animal studies suggest that (single-dose) allopurinol (xanthine oxidase inhibitor with high oral bioavailability and long-lasting active metabolites) may reduce this risk; human study results are conflicting. The aim of this study was to determine if allopurinol decreases the rate of PEP. Methods: Patients referred for ERCP to 9 endoscopists at 2 tertiary centers were randomized to receive either allopurinol 300 mg or identical placebo orally 60 minutes before ERCP, stratified according to high-risk ERCP (manometry or pancreatic therapy). The primary outcome (PEP) was adjudicated blindly; pancreatitis was defined according to the Cotton consensus, and evaluated at 48 hours and 30 days. Secondary outcomes included severe PEP, length of stay, and mortality (nil). The trial was terminated after the blinded (midpoint) interim analysis, as recommended by the independent data and safety monitoring committee. Results: We randomized 586 subjects, 293 to each arm. The crude PEP rates were 5.5% (allopurinol) and 4.1% (placebo), (P.44; difference 1.4%; 95% confidence interval, 2.1% to 4.8%). The Mantel Haenszel combined risk ratio for PEP with allopurinol, considering stratification, was 1.37 (95% confidence interval, ). Subgroup analyses suggested nonsignificant trends toward possible benefit in the high-risk group, and possible harm for the remaining subjects. Logistic regression found pancreatic therapy, pancreatic injection, and prior PEP to be the only independent predictors of PEP. Conclusions: Allopurinol does not appear to reduce the overall risk of PEP; however, its potential benefit in the high-risk group (but potential harm for non high-risk patients) means further study is required. Biliopancreatic diseases are extremely common in North America and Europe and endoscopic retrograde cholangiopancreatography (ERCP) plays a pivotal role in their diagnosis confirmation and treatment. Pancreatitis is the most common complication of ERCP, with an overall incidence of 2% to 15%, and a related mortality rate of 0.1% to 0.5%. 1 5 High-risk procedures (eg, pancreatic sphincterotomy) and high-risk patients (eg, suspected sphincter of Oddi dysfunction [SOD]) can have post-ercp pancreatitis (PEP) rates as high as 20%. 1 4,6,7 Thus, each year ERCP is associated with an estimated 15,000 cases of pancreatitis and up to 300 procedure-related deaths per year in the United States. 1 4 Similarly, one Danish study reported a national insurance registry s findings of 7 deaths caused by ERCP-related pancreatitis over 4 years. 8 Recent studies have looked at various agents in attempts to try to prevent this complication. Simple regimens have included nonsteroidal anti-inflammatory drugs, steroids, and nitrates, 9 17 whereas others have been more complex (eg, prolonged continuous infusions), and/or used expensive therapies such as protease inhibitors (gabexate mesilate, ulinastatin), somatostatin, octreotide, interleukin-10, and N-acetylcysteine Pancreatic stents for high-risk patients appear to reduce the risk of PEP and be cost effective, but the approach also involves injection of additional pancreatic duct dye and wire manipulation, neither of which may be worth the risk for non high-risk patients. 32 In addition, there are 3 other barriers to routine prophylactic pancreatic stenting: (1) not all ERCP endoscopists are comfortable placing pancreatic stents; (2) placing a stent is not always possible; and (3) pancreatic stents can cause damage to the duct wall Oxygen radicals can mediate capillary endothelial injury, which may be an early step in the pathogenesis of acute pancreatitis Other anti-inflammatory drugs such as steroids have not been shown to provide significant reductions in PEP. 15 Interleukin-10 infusion has shown some promise; however, it is expensive and not widely available. 39 Several animal studies have suggested that allopurinol (a xanthine oxidase inhibitor and an effective anti-oxidant with anti-apoptotic effects, that historically has been effective in preventing attacks of acute gouty arthritis) might prevent the initial oxidative insult that can trigger the onset of PEP Allopurinol has high oral bioavailability (80% 90%), a rapid onset (peak, h), and a 70% hepatic transformation to a long-lasting active metabolite (oxypurinol half-life, 15 h). 46 These pharmacokinetic attributes mean a single oral dose of Abbreviations used in this paper: CI, confidence interval; DSM, data and safety monitoring; ERCP, endoscopic retrograde cholangiopancreatography; OR, odds ratio; PEP, post endoscopic retrograde cholangiopancreatography pancreatitis; SOD, sphincter of Oddi dysfunction by the AGA Institute /08/$34.00 doi: /j.cgh

2 466 ROMAGNUOLO ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 4 allopurinol before ERCP conceivably might prevent PEP because the drug targets those changes that contribute to the initial triggering of pancreatitis. 37,38 Allopurinol is also an inexpensive generic drug with an excellent safety record related to its use for prevention of acute gout. A pilot study of allopurinol treatment in 30 randomized human subjects, with the same dosing regimen used in the current study, showed very promising results with a 21% absolute risk difference in PEP and a significant reduction in 3-hour lipase levels. 47 In addition to this, 1 other positive-outcome randomized trial and 2 negativeoutcome randomized trials of allopurinol for PEP prevention have been published. 14,48,49 Given the promising pilot study data, the compelling animal data, and the practical benefits of allopurinol s potential use for PEP prevention, we designed a randomized blinded multicenter trial to compare the rates of PEP seen with treatment with either allopurinol or placebo. Methods Patients and Randomization After institutional review board approval at each site, the study screened for eligible adult patients presenting for an ERCP at 2 institutions (the University of Calgary, Peter Lougheed Center in Calgary, Alberta, Canada; and the University of Alberta Hospital in Edmonton, Alberta, Canada) between 2002 and At registration, eligible patients were randomized, after giving informed consent, to receive either allopurinol or an identical placebo. After the oral administration of the treatment drug or placebo, ERCP was performed by 1 of 9 ERCP endoscopists with comparable baseline complication rates (data not shown). An independent data and safety monitoring (DSM) committee oversaw the trial and was chaired from McMaster University in Hamilton, Canada. The study followed Consolidate Standards of Reporting Trials (CONSORT) guidelines. 50 Randomization was performed centrally at the University of Calgary by using computer-generated random numbers and allocation was concealed from subjects, coordinators, and investigators with coded packets of study drug delivered by the investigational pharmacy to the endoscopy unit after the patient was recruited and informed consent was obtained. Thus, the randomization assignments were known only to the investigational pharmacy. Patients were randomized in blocks of 4, stratified according to site and according to risk, so that the 2 arms of the trial contained equal numbers of high-risk ERCP cases (defined as planned pancreatic therapy or planned sphincter of Oddi manometry). Exclusion criteria included a hemoglobin level of less than 8 g/dl; a platelet count of less than /L; relative neutropenia (absolute neutrophil count /L); significant renal dysfunction (serum creatinine level, 200 mol/l); decompensated cirrhosis; a known allergy to allopurinol; a known or suspected pregnancy or lactation; current or recent (within 48 hours) use of allopurinol; current use of drugs with a known interaction with allopurinol, including cyclophosphamide, chlorpropamide, azathioprine/mercaptopurines, or probenecid; an inability to swallow or absorb oral medication; and recent (onset within 5 days of the ERCP) acute pancreatitis. The latter exclusion was mandated by the difficulty in diagnosing PEP when acute pancreatitis already is present (lipase often still is increased, pain often still is present) and imaging findings of pancreatic inflammation persist. After randomization, patients were given either allopurinol (300 mg) or the identical placebo orally with a sip of water at 60 minutes ( 20 min) before the start of ERCP. Baseline demographics, indication, and laboratory results were recorded. Procedural details were recorded, along with the final diagnosis. Follow-up evaluation, still blinded for both patients and research personnel, was performed either in the hospital (if the patient was admitted or remained an inpatient), or by telephone, at 48 hours and at 30 days. During this time, additional records and/or laboratory results were obtained when needed to clarify potential complications. PEP, unlike bleeding, perforation, or infection (all of which may be delayed in presentation), occurs within 24 hours of ERCP. Therefore, if no PEP was documented at either follow-up evaluation, the patients were considered not to have had PEP. Outcomes The primary outcome was the proportion of patients developing pancreatitis within 24 hours after ERCP (PEP). PEP was defined based on the Cotton consensus criteria. 51 Thus, PEP diagnosis required the presence of typical pancreatic pain (epigastric pain often radiating into the back and associated with nausea and/or vomiting) requiring medical attention, in association with a serum lipase or amylase level greater than 2 times the upper limit of normal. The pain had to have begun within 24 hours of ERCP and must have lasted longer than 24 hours (defined as residual pain 5 of 10 on an analogue scale). Pancreatic pain requiring hospitalization for more than 24 hours was considered sufficient to diagnose PEP in the case of patients with chronic pancreatitis because enzyme level increases may not occur (because of the potential lack of enzyme reserve in those patients). The main secondary outcomes were the proportion of patients with local complications of pancreatitis (necrotizing pancreatitis, pseudocyst, or abscess) or the need for surgery. The definition and grading of other post-ercp complications (bleeding, sepsis, perforation, other infection) was performed according to published guidelines. 51 Cardiorespiratory events also were recorded, as were other unusual unplanned events (such as basket impaction, aspiration pneumonia, and medication reactions). Thirty-day mortality rates attributable to complications also were recorded. The study used an intention-totreat analysis strategy. Statistical Analysis Blinded statistical analysis was completed with Stata (v. 7.0, StataCorp, College Station, TX). The analysis was stratified (Mantel Haenszel weights) in accordance with the stratified randomization approach. Logistic regression also was used to control for the possible confounding effects of chance imbalances in randomization. Model building was performed using likelihood ratio tests of nested models of clinically plausible factors. Proportions were compared using the chi-square or the Fisher exact test where applicable, and t tests were used for continuous variables. Nonparametric tests were used for nonnormal continuous variable comparisons. Ninety-five percent confidence intervals (CIs) were calculated for ratios and differences. Planned subgroup analyses of interactions (differential

3 April 2008 ALLOPURINOL FOR ERCP-RELATED PANCREATITIS 467 effect of allopurinol therapy among subgroups) included highrisk versus non high-risk ERCP and prophylactic pancreatic stent insertion versus no stent. The sample size calculation was based on an estimated baseline pancreatitis rate of 7%, a relative risk reduction of 0.5, a power of 80%, and a 2-sided of.05, yielding a target sample size of 1266 patients. A blinded interim analysis according to the primary outcome was planned at 2 years, using O Brien-Fleming spending function thresholds. 52 After this blinded review, the independent DSM committee recommended a second interim analysis at the patient recruitment midpoint (533 of 1266 patients). This second blinded analysis, along with conditional power and stochastic curtailment analyses with realistic and optimistic scenarios, then was presented to the DSM; in the meantime, additional patients were recruited. On the basis of these data, however, the DSM decided the trial should cease recruitment, primarily owing to futility in achieving significance with respect to the overall primary outcome. Thereafter, the previously recruited patients (n Table 1. Demographics, Indications, and Procedure Details for Allopurinol and Placebo Arms Variable Allopurinol (n 293) Placebo (n 293) P value Demographics/baseline data Age, y 53.9 (17.6) 55.5 (16.7) Male sex 127 (43.3%) 115 (39.2%) Body mass index 26.9 (5.3) 27.2 (5.8) Aspartate aminotransferase level, U/L (198.0) (173.1) Bilirubin level, mol/l 61.3 (89.0) 54.2 (70.8) Malignancy clinically suspected 32 (10.9%) 35 (11.9%) Biliary dilation suspected 84 (28.7%) 89 (30.4%) History of chronic pancreatitis 8 (2.7%) 11 (3.8%) Pre-ERCP abdominal pain 68 (23.2%) 67 (22.9%) Previous PEP 17 (5.8%) 11 (3.8%) University of Calgary site a 263 (89.8%) 264 (90.1%) Indications Acute pancreatitis 8 (2.7%) 7 (2.4%) Chronic pancreatitis 9 (3.1%) 8 (2.7%) Recurrent pancreatitis 13 (4.4%) 18 (6.1%) Suspected malignant obstruction 25 (8.5%) 25 (8.5%) Suspected bile duct stone 174 (59.4%) 169 (57.7%) Suspected sphincter dysfunction (SOD) 10 (3.4%) 5 (1.7%) Suspected bile leak 11 (3.8%) 8 (2.7%) Suspected postoperative biliary stricture 5 (1.7%) 12 (4.1%) Suspected primary sclerosing cholangitis 5 (1.7%) 6 (2.0%) Other 28 (9.6%) 31 (10.6%) Procedure details Time delay from study drug to endoscope insertion, min b 60.0 ( ) 55.0 ( ).01 Pre-ERCP antibiotics 81 (27.6%) 77 (26.3%).71 Total procedure time, min b 25.0 ( ) 25.0 ( ).52 Cannulation time, min b 4.0 ( ) 3.5 (1 11.0) Pancreatic duct injection 129 (44.0%) 102 (34.8%).02 Number of pancreatic injections b 2.0 ( ) 1.0 ( ).01 Acinarization with injection 3 (1.0%) 2 (0.68%).65 Invasive diagnostics Cytology 25 (8.5%) 16 (5.5%).15 Bile sample 2 (0.68%) 3 (1.0%).65 Intrabiliary biopsy 13 (4.4%) 11 (3.8%).68 Sphincter manometry (biliary) 6 (2.0%) 7 (2.4%).78 Sphincter manometry (pancreatic) 3 (1.0%) 8 (2.7%).13 Therapeutics (selected) Any therapeutics 252 (86.0%) 252 (86.0%) 1.00 Biliary stent insertion 54 (18.4%) 50 (17.1%).67 Stone extraction 68 (23.2%) 73 (24.9%).63 Mechanical lithotripsy 8 (2.7%) 14 (4.8%).19 Biliary sphincterotomy (EBS) 220 (75.1%) 213 (72.7%).51 Needle-knife/precut EBS 31 (10.6%) 30 (10.2%).89 Pancreatic stent 11 (3.8%) 17 (5.8%).25 ERCP cancelled 1 (0.34%) 4 (1.4%).18 High-risk ERCP (see text for definition) 32 (10.9%) 34 (11.6%).79 NOTE. Categoric/dichotomous variables are summarized as number (%), and continuous variables as mean (SD). P values were calculated only for postrandomization factors. EBS, Endoscopic biliary sphincterotomy. a Peter Lougheed Center, Calgary, Alberta. b Median (interquartile range).

4 468 ROMAGNUOLO ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 4 Table 2. Outcomes for Allopurinol and Placebo Arms Variable Allopurinol (n 293) Placebo (n 293) P value Procedure-related complications Overall complications 25 (8.5%) 18 (6.1%).34 Overall fatal complications 0 (0%) 0 (0%) 1.0 Pancreatitis (PEP) 16 (5.5%) 12 (4.1%).44 Mild a 8 (2.7%) 4 (1.4%).24 Moderate 6 (2.0%) 6 (2.0%) 1.0 Severe 2 (0.7%) 2 (0.7%) 1.0 Fatal 0 (0%) 0 (0%) 1.0 Bleeding 0 (0%) 0 (0%) 1.0 Perforation 0 (0%) 0 (0%) 1.0 Infection (cholecystitis/cholangitis) 6 (2.0%) 1 (0.3%).12 Other (including unexplained pain) 4 (1.4%) b 5 (1.7%) 1.0 Disease-related adverse events Pancreatitis 0 (0%) 0 (0%) 1.0 Bleeding 0 (0%) 0 (0%) 1.0 Infection 4 (1.4%) 0 (0%) 1.0 Other 2 (0.7%) 3 (1.0%) 1.0 Resource use Critical care unit length of stay, d 0.02 c 0.32 Hospital length of stay for complication or adverse event, d d 6.0 ( ) 5.0 ( ).40 Hospital length of stay for post-ercp pancreatitis, d d 3.5 ( ) 5.5 ( ).50 NOTE. Categoric/dichotomous variables are summarized as number (%), and continuous variables as mean (SD). a All patients admitted except 1. b One patient also had pancreatitis. c One patient had a 5-day critical care admission. d Median (interquartile range) shown. 586) completed their follow-up evaluations, and the formal blinded statistical analysis was completed. Results We enrolled 586 subjects, randomizing 293 patients each to the allopurinol and the placebo arms. The baseline characteristics and procedural details are summarized in Table 1. The average subject age was 54.7 years ( SD 17 y) and 42.8% of the subjects were male. Overall, 29.1% of patients were randomized, based on screen-fail data from the first 1000 patients screened. Based on detailed data from 100 consecutive screening failures, the average age was slightly older at 61.9 years ( SD 20.6 y) and 41.4% were male. The most common reasons for exclusion were as follows: an inability to obtain consent (dementia, inability to speak English) or a refusal to consent (48%); issues related to current allopurinol use or contraindication (9%); onset of acute pancreatitis within 5 days (15%); insufficient time window (to allow the minimum time between drug and procedure) (11%); and pregnancy or lactation (6%). The frequencies of all other exclusion criteria were less than 5%. The most common indications for ERCP among screen-fails were similar to those of the study population. The most common indication among amon patients failing screening still was suspected bile duct stone (51.4%). The next two most common were bile leaks and suspected malignant obstruction (9.7% and 8.3%, respectively); the former was significantly more common than it was in the study population, perhaps because patients with bile leaks are generally in the immediate postoperative state, in considerable pain, and are less likely to elect to participate in a clinical trial. The majority of the ERCPs performed in randomized subjects were therapeutic (86.0%), and just over half were performed for a suspected bile duct stone. A small proportion (n 66) of the participants was classified as being in a high-risk category, as defined earlier. Most patients (n 527) were recruited at the University of Calgary, and the remaining were recruited at the University of Alberta. There were no clinically significant chance imbalances in baseline or procedural characteristics. Five patients (0.85%; 4 in the placebo group and 1 in the allopurinol group) had their ERCP cancelled after randomization (while the study still was blinded), but they were included in their original group assignment in the primary (intention-to-treat) analysis. There were no significant differences in the overall rates of PEP in the 2 groups (Table 2). The raw PEP rates were 5.5% (allopurinol) and 4.1% (placebo), (P.44; difference, 1.4%; 95% CI, 2.1% to 4.8%). The Mantel Haenszel combined risk ratio, considering stratification on center and high-risk status, for PEP with allopurinol was 1.37 (95% CI, ). The stochastic curtailment analysis showed that despite recruiting the additional planned patients, using an optimistic view of assuming that the relative risk ratio of 0.5 (the risk ratio used in the sample size calculation) would be seen in the subsequent recruits, using the baseline PEP rate observed in the trial so far (lower PEP rate of the 2 groups), the rates at the end of the trial would be 5.5% (group with higher event rate) and 3.4% (group with lower event rate) (difference 95% CI, 1.7% to 4.4%, P.07). The DSM believed further recruitment was futile. There were, however, differences between the treatment effects seen for the high-risk and non high-risk subgroups; the corresponding interaction term (P.01) was significant. In the non high-risk group (n 520), the crude PEP rates were 5.4% (allopurinol) and 1.5% (placebo) (P.017, favoring placebo, indicating harm associated with allopurinol), whereas in the

5 April 2008 ALLOPURINOL FOR ERCP-RELATED PANCREATITIS 469 Table 3. Summary of Randomized Trials of Allopurinol in PEP Prevention Study (y), SC vs MC, country n Dose, mg Allopurinol vs placebo PEP rates Percentage high risk a Comment Romagnuolo 2007, MC, Canada (current study) % vs 4.1% 16 vs 12 Mosler et al, 49 (2005) MC, USA b 13.0% vs 12.1% 46 vs 42 Katsinelos et al, 48 (2005) SC, c 3.2% vs 17.8% Greece 4vs21 Budzynska et al, 14 (2001) SC, d 12.1% vs 7.9% Poland 12 vs 8 Raw pooled 1837 (923 vs 914) 8.5% vs 9.1% 78 vs % Harm in average risk; benefit in high risk 70.2% 4% absolute benefit in high risk; 4% absolute harm in average risk 0% 2 patients with suspected SOD 0% 3-arm study, with third arm (n 100) given prednisone 0.6% difference (95% CI, 3.2% to 2.0%) MC, multicenter; SC, single center. a As defined in this protocol, namely sphincter manometry and/or pancreatic therapy. Other higher-risk cases (eg, precut sphincterotomy, suspected SOD) were not considered. b 600 mg 4 hours before, 300 mg 1 hour before. c 600 mg 15 hours before, 600 mg 3 hours before. d 200 mg 15 hours before, 200 mg 3 hours before. high-risk group (n 66), the PEP rates were 6.3% (allopurinol) and 23.5% (placebo) (P.050, favoring allopurinol). By using logistic regression, correcting for the blocked strata (center and high-risk group), age, and sex, the only independent predictors of PEP were found to be previous PEP (odds ratio [OR], 15.8; 95% CI, ; P.001), pancreatic injection (OR, 4.8; 95% CI, ), and pancreatic therapy (OR, 5.9; 95% CI, ; P.02). Correcting for these factors, the adjusted OR for PEP with allopurinol was 1.0 (95% CI, ; P.99). An older age also showed a trend toward a lower PEP rate (OR, 0.8; 95% CI, per decade; P.13). Other factors, including sex, body mass index, time from drug ingestion to procedure, indication (pancreatitis, stone/malignancy, suspected SOD, other), procedure, cannulation success and/or duration of cannulation attempt, number of pancreatic injections, biliary sphincterotomy (regular or precut), and pancreatic stent placement were not significant predictors. As stated earlier, there was an interaction between the allocation (treatment vs placebo) and the risk group (P.01), but not between the allocation and pancreatic stent insertion, or between the allocation and prior PEP history, age, sex, or recruitment site. Although pancreatic injection was performed, by chance, more frequently in the allopurinol arm, there was no significant confounding in the model in the non high-risk group. Furthermore, correcting for this factor, in a logistic model stratified by high-risk group, allopurinol still had a nonsignificant trend toward harm in the non high-risk group (OR, 3.1; 95% CI, ; P.17), albeit with a wide CI. Only 14 patients had confirmed SOD, with manometry followed by pancreatic and/or biliary sphincterotomy. There was a trend toward a higher rate of PEP in this group (P.09). Fifty-seven percent of these persons received a pancreatic stent. In the confirmed SOD subgroup, no differences were seen with respect to pancreatic stenting and treatment allocation arm; however, the statistical power of this subgroup analysis was very limited. In terms of the secondary outcomes, there were no differences in length of stay or critical care stay attributed to PEP (Table 2). There were also no differences in the spectrums of PEP severity; severe PEP was rare in both groups and there were no fatal complications. Discussion Despite a great deal of scientific interest in the area, endoscopists and their patients are still in great need of a simple, widely available, and inexpensive intervention that can reduce the rate of PEP. This complication is the main driving force in algorithms that lead toward noninvasive imaging of patients with suspected biliopancreatic disease for whom therapy or invasive diagnostics likely are not needed. However, a large proportion of patients still need ERCP, for both diagnostic and therapeutic reasons, and pancreatitis remains an important limitation of this procedure. Three randomized trials have been published in full to date: a negative study from Poland 14 (n 300), a positive study from Greece 48 (n 243), and a negative study from the United States 49 (n 701) (Table 3). There was variability in the doses used in the studies, and in the baseline rates of PEP in the control (placebo) groups (some of which are out of the usual range reported), but these differences do not appear to completely explain the heterogeneity in the results. There remains a possibility for a threshold effect or a minimally effective dose for allopurinol, given that the positive study 48 used the highest dose (1200 mg); however, there does not seem to be a clear dose-response relationship with a larger negative study 49 having used a fairly high dose (900 mg) (Table 3). In fact, our study found significant differences in treatment effects for different subgroups (ie, significant interactions). The 3 prior studies all formally or informally checked for interactions, presenting the active treatment and placebo PEP rates in different subgroups. None found significant interactions between diagnostic and therapeutic procedures. The most detailed analysis of this type in the 3 studies was found in the study by Mosler et al. 49 If one were to group that study s manometry and pancreatic therapy subjects together into a high-risk group as we did in this study, one would see the same (although in this case, nonsignificant) trends: reduction of PEP by allopurinol from 27% to 23% in the high-risk group, and harm by allopurinol (8% vs 12% PEP) in the non high-risk group. Budzynska et al 14 also included primarily non high-risk patients and showed a higher rate of PEP with allopurinol. In contrast, the patients in the study by Katsinelos et al 48 also

6 470 ROMAGNUOLO ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 4 primarily were non high-risk patients, and yet the study showed a significant benefit for allopurinol. If allopurinol is truly harmful for non high-risk patients undergoing ERCP (the adjusted subgroup OR was not significant), the mechanism responsible for this harm is unclear. It could be the result of an idiosyncratic reaction to the medicine itself; one study does suggest that medications with a history of inducing pancreatitis could increase the risk of PEP. 53 However, allopurinol is not particularly known to cause drug-induced pancreatitis. Other anti-inflammatory medicines (ie, steroids) appear to either have no effect or increase the risk of pancreatitis in randomized trials. 10,14 17 Again, the mechanism responsible for this increase also is unknown, although for hydrocortisone, drug-related pancreatitis is a well-recognized phenomenon. Nonsteroidal anti-inflammatory drugs might have a more consistent benefit. 9,12 We considered that the association between allopurinol and PEP could be owing to the fact that the percentage of patients with pancreatic injections was significantly higher in the allopurinol group (P.02), as was the median number of injections; this could have confounded that subgroup analysis. However, confounding was not confirmed statistically, and correcting for pancreatic injection in a stratified model still showed a nonsignificant trend toward harm for allopurinol in the non high-risk subgroup (OR, 3.1; P.17). Pancreatic therapy, pancreatic injection, and a history of prior PEP were the only significant independent predictors of PEP in our study, with a trend being seen toward higher PEP rates in younger patients. Several other studies have commented on PEP predictors. 4 Multivariate analyses have shown suspected SOD, 4,6 female sex, 5,6 history of post-ercp pancreatitis, 3,4,6 younger age, 2,4,5 normal bilirubin level, 6 normal bile duct size, 2 absence of chronic pancreatitis, 6 difficult cannulation, 5 pancreatic sphincterotomy, 6 and (multiple) pancreatic injection 2,4,6 to be independent risk factors, but the factors are not consistently predictive among the studies. For example, female sex, difficult cannulation, and pancreatic sphincterotomy were not confirmed by Cheng et al 4 and lower case volume was not confirmed by Freeman et al. 6 Biliary sphincterotomy does not appear to be an independent predictor of PEP. 2,4,6 The factors found in our study are among those often cited and reproduced. In summary, it appears that allopurinol does not decrease the overall rate of post-ercp pancreatitis; it potentially could be harmful for non high-risk patients (we were underpowered for that specific question, and the CI is wide). Its effect on the highest-risk subgroup remains unclear, and although there is some indication of benefit, the drug-procedure interaction needs to be tested specifically in a randomized trial that enrolls only that subgroup. Unfortunately, interventions that are effective in high-risk subgroups, such as prophylactic pancreatic stenting, although helpful for referral ERCP centers, are not especially useful for community ERCP centers. Allopurinol potentially might be used as rescue therapy for high-risk patients in whom pancreatic stent placement has failed, but the protocols studied to date have involved providing the allopurinol before the procedure, rather than afterward. Our analysis does strengthen the suggestion that subgroup analysis by risk category is desirable, and perhaps even necessary, in PEP prevention trials, because results may not be generalizable across risk groups. References 1. Freeman ML. Complications of endoscopic biliary sphincterotomy: a review. Endoscopy 1997;29: Loperfido S, Angelini G, Benedetti G, et al. Major early complications from diagnostic and therapeutic ERCP: a prospective multicenter study. Gastrointest Endosc 1998;48: Ho KY, Montes H, Sossenheimer MJ, et al. Features that may predict hospital admission following outpatient therapeutic ERCP. Gastrointest Endosc Clin N Am 1999;49: Cheng CL, Sherman S, Watkins JL, et al. Risk factors for post- ERCP pancreatitis: a prospective multicenter study. Am J Gastroenterol 2006;101: Williams EJ, Taylor S, Fairclough P, et al. Risk factors for complication following ERCP; results of a large-scale, prospective multicenter study. Endoscopy 2007;39: Freeman ML, DiSario JA, Nelson DB, et al. Risk factors for post- ERCP pancreatitis: a prospective, multicenter study. Gastrointest Endosc 2001;54: Andriulli A, Loperfido S, Napolitano G, et al. Incidence rates of post-ercp complications: a systematic survey of prospective studies. Am J Gastroenterol 2007;102: Trap R, Adamsen S, Hart-Hansen O, et al. Severe and fatal complications after diagnostic and therapeutic ERCP: a prospective series of claims to insurance covering public hospitals. Endoscopy 1999;31: Murray B, Carter R, Imrie C, et al. Diclofenac reduces the incidence of acute pancreatitis after endoscopic retrograde cholangiopancreatography. Gastroenterology 2003;124: Sherman S, Urszula B, Watkins JL, et al. Does prophylactic administration of corticosteroid reduce the risk and severity of post-ercp pancreatitis: a randomized, prospective, multicenter study. Gastrointest Endosc 2003;58: Kaffes AJ, Bourke MJ, Ding S, et al. A prospective, randomized, placebo-controlled trial of transdermal glyceryl trinitrate in ERCP: effects on technical success and post-ercp pancreatitis. Gastrointest Endosc 2006;64: Sotoudehmanesh R, Khatibian M, Kolahdoozan S, et al. Indomethacin may reduce the incidence and severity of acute pancreatitis after ERCP. Am J Gastroenterol 2007;102: Kwanngern K, Tiyapattanaputi P, Wanitpukdeedecha M, et al. Can a single dose corticosteroid reduce the incidence of post- ERCP pancreatitis? A randomized, prospective control study. J Med Assoc Thai 2005;88(Suppl 4):S42 S Budzynska A, Marek T, Nowak A, et al. A prospective, randomized, placebo-controlled trial of prednisone and allopurinol in the prevention of ERCP-induced pancreatitis. Endoscopy 2001;33: De Palma GD, Catanzano C. Use of corticosteriods in the prevention of post-ercp pancreatitis: results of a controlled prospective study. Am J Gastroenterol 1999;94: Dumot JA, Conwell DL, O Connor JB, et al. Pretreatment with methylprednisolone to prevent ERCP-induced pancreatitis: a randomized, multicenter, placebo-controlled clinical trial. Am J Gastroenterol 1998;93: Manolakopoulos S, Avgerinos A, Vlachogiannakos J, et al. Octreotide versus hydrocortisone versus placebo in the prevention of post-ercp pancreatitis: a multicenter randomized controlled trial. Gastrointest Endosc 2002;55: Milewski J, Rydzewska G, Degowska M, et al. N-acetylcysteine does not prevent post-endoscopic retrograde cholangiopancreatography hyperamylasemia and acute pancreatitis. World J Gastroenterol 2006;12: Li ZS, Pan X, Zhang WJ, et al. Effect of octreotide administration in the prophylaxis of post-ercp pancreatitis and hyperamylasemia: a multicenter, placebo-controlled, randomized clinical trial. Am J Gastroenterol 2007;102: Zheng M, Chen Y, Yang X, et al. Gabexate in the prophylaxis of

7 April 2008 ALLOPURINOL FOR ERCP-RELATED PANCREATITIS 471 post-ercp pancreatitis: a meta-analysis of randomized controlled trials. BMC Gastroenterol 2007;7: Deviere J, Le Moine O, Van Laethem JL, et al. Interleukin 10 reduces the incidence of pancreatitis after therapeutic endoscopic retrograde cholangiopancreatography. Gastroenterology 2001;120: Dumot JA, Conwell DL, Zuccaro G Jr, et al. A randomized, double blind study of interleukin 10 for the prevention of ERCP-induced pancreatitis. Am J Gastroenterol 2001;96: Andriulli A, Leandro G, Federici T, et al. Prophylactic administration of somatostatin or gabexate does not prevent pancreatitis after ERCP: an updated meta-analysis. Gastrointest Endosc 2007;65: Cavallini G, Tittobello A, Frulloni L, et al. Gabexate for the prevention of pancreatic damage related to endoscopic retrograde cholangiopancreatography. Gabexate in digestive endoscopy Italian Group. N Engl J Med 1996;335: Fujishiro H, Adachi K, Imaoka T, et al. Ulinastatin shows preventive effect on post-endoscopic retrograde cholangiopancreatography pancreatitis in a multicenter prospective randomized study. J Gastroenterol Hepatol 2006;21: Tsujino T, Komatsu Y, Isayama H, et al. Ulinastatin for pancreatitis after endoscopic retrograde cholangiopancreatography: a randomized, controlled trial. Clin Gastroenterol Hepatol 2005;3: Fazel A, Quadri A, Catalano MF, et al. Does a pancreatic duct stent prevent post-ercp pancreatitis? A prospective randomized study. Gastrointest Endosc 2003;57: Tarnasky PR, Palesch YY, Cunningham JT, et al. Pancreatic stenting prevents pancreatitis after biliary sphincterotomy in patients with sphincter of Oddi dysfunction. Gastroenterology 1998;115: Singh P, Das A, Isenberg G, et al. Does prophylactic pancreatic stent placement reduce the risk of post-ercp acute pancreatitis? A meta-analysis of controlled trials. Gastrointest Endosc 2004; 60: Das A, Singh P, Sivak MV Jr, et al. Pancreatic-stent placement for prevention of post-ercp pancreatitis: a cost-effectiveness analysis. Gastrointest Endosc 2007;65: Andriulli A, Forlano R, Napolitano G, et al. Pancreatic duct stents in the prophylaxis of pancreatic damage after endoscopic retrograde cholangiopancreatography: a systematic analysis of benefits and associated risks. Digestion 2007;75: Cheon YK, Cho KB, Watkins JL, et al. Frequency and severity of post-ercp pancreatitis correlated with extent of pancreatic ductal opacification. Gastrointest Endosc 2007;65: Smith MT, Sherman S, Ikenberry SO, et al. Alterations in pancreatic ductal morphology following polyethylene pancreatic stent therapy. Gastrointest Endosc 1996;44: Sherman S, Hawes RH, Savides TJ, et al. Stent-induced pancreatic ductal and parenchymal changes: correlation of endoscopic ultrasound with ERCP. Gastrointest Endosc 1996;44: Alvarez C, Robert M, Sherman S, et al. Histologic changes after stenting of the pancreatic duct. Arch Surg 1994;129: Sanfey H, Bulkley GB, Cameron JL. The role of oxygen-derived free radicals in the pathogenesis of acute pancreatitis. Ann Surg 1984;200: Sanfey H, Cameron JL. Increased capillary permeability: an early lesion in acute pancreatitis. Surgery 1984;96: Sanfey H, Bulkley GB, Cameron JL. The pathogenesis of acute pancreatitis. The source and role of oxygen-derived free radicals in three different experimental models. Ann Surg 1985;201: Deviere J, Le Moine O, Van Laethem JL, et al. Interleukin 10 reduces the incidence of pancreatitis after therapeutic endoscopic retrograde cholangiopancreatography. Gastroenterology 2001;120: Marks JM, Dunkin BJ, Shillingstad BL, et al. Pretreatment with allopurinol diminishes pancreatography-induced pancreatitis in a canine model. Gastrointest Endosc 1998;48: Hirano T, Manabe T, Steer M, et al. Protective effects of therapy with a protease and xanthine oxidase inhibitor in short form pancreatic biliary obstruction and ischemia in rats. Surg Gynecol Obstet 1993;176: Sanfey H, Sarr MG, Bulkley GB, et al. Oxygen-derived free radicals and acute pancreatitis: a review. Acta Physiol Scand Suppl 1986; 548: Schimpl G, Pesendorfer P, Steinwender G, et al. Allopurinol reduces bacterial translocation, intestinal mucosal lipid peroxidation, and neutrophil-derived myeloperoxidase activity in chronic portal hypertensive and common bile duct-ligated growing rats. Pediatr Res 1996;40: Schimpl G, Pesendorfer P, Steinwender G, et al. Allopurinol and glutamine attenuate bacterial translocation in chronic portal hypertensive and common bile duct ligated growing rats. Gut 1996; 39: Albuquerque RG, Sanson AJ, Malangoni MA. Allopurinol protects enterocytes from hypoxia-induced apoptosis in vivo. J Trauma 2002;53: Welbanks L, ed. Compendium of pharmaceuticals and specialties. 36th ed. Ottawa: Canadian Pharmacists Association, Romagnuolo J, Bailey RJ, Mortimer ST, et al. Allopurinol in the prevention of ERCP pancreatitis: preliminary results from a randomized double-blind placebo-controlled trial (abstr). Can J Gastroenterol 2001;15:43A. 48. Katsinelos P, Kountouras J, Chatzis J, et al. High-dose allopurinol for prevention of post-ercp pancreatitis: a prospective randomized double-blind controlled trial. Gastrointest Endosc 2005;61: Mosler P, Sherman S, Marks J, et al. Oral allopurinol does not prevent the frequency or the severity of post-ercp pancreatitis. Gastrointest Endosc 2005;62: Moher D, Schulz KF, Altman DG. The CONSORT statement: revised recommendations for improving the quality of reports of parallelgroup randomized trials. Ann Intern Med 2001;134: Cotton PB, Lehman G, Vennes J, et al. Endoscopic sphincterotomy complications and their management: an attempt at consensus. Gastrointest Endoscopy 1991;37: Piantadosi S. Data-dependent stopping. In: Piantadosi S, ed. Clinical trials: a methodologic perspective. Hoboken, NJ: John Wiley & Sons, Perney P, Berthier E, Pageaux GP, et al. Are drugs a risk factor of post-ercp pancreatitis? Gastrointest Endosc 2003;58: Address requests for reprints to: Joseph Romagnuolo, MD, FRCPC, MSc, Medical University of South Carolina, 96 Jonathon Lucas Street, CSB #210, PO Box , Charleston, South Carolina romagnuo@musc.edu; fax: (843) Previously presented in abstract form at Canadian Digestive Disease Week March, 2001, Banff, Canada; American College of Gastroenterology October, 2002, Seattle, WA; and Digestive Disease Week May, 2005, Chicago, IL. The authors would like to thank the Canadian Association for Gastroenterology (and the Canadian Digestive Health Foundation), the Canadian Institutes for Health Research, and the University of Calgary for joint funding of this trial s operating budget. In addition, the authors would like to thank the Alberta Heritage Foundation for Medical Research for funding our investigators (J.R., R.H.) with salary support during part of the trial. The authors also are grateful for the help of Scott Miller, MSc (Department of Biostatistics, Epidemiology and Bioinformatics, Medical University of South Carolina) with the interim analysis and conditional power calculation.