Risk of Acute Pancreatitis in Users of Azathioprine:A Population-Based Case–Control StudyAndrea Floyd, M.D.,Lars Pedersen, Gunnar Lauge Nielsen, M.D., Ole Thorlacius-Ussing, Dr. Med. Sci., andHenrik Toft Sorensen, Dr. Med. Sci., Ph.D.Department of Clinical Epidemiology at Aalborg and Aarhus University Hospitals; Department ofGastrointestinal Surgery A, Aalborg University Hospital, Aalborg; Department of Medicine M, AalborgUniversity Hospital, Aalborg; and Department of Medicine V, Aarhus University Hospital, Aarhus, Denmark

OBJECTIVE: Azathioprine has been linked to subsequentacute pancreatitis in several case reports and small caseseries. We examined the risk for acute pancreatitis in usersof azathioprine in North Jutland County, Denmark, withabout 490,000 inhabitants.

METHODS: We identified patients with incident cases ofacute pancreatitis from the Hospital Discharge Registry ofthe county from 1991–2000, and selected 10 controls percase, matched by age and sex, from the Central PersonalRegistry using incidence density sampling technique. Allprescriptions of azathioprine within 90 days before admis-sion were likewise collected from the population-basedNorth Jutland Prescription Database. Data on potential con-founders were extracted from registries. We used condi-tional logistic regression to adjust for confounding.

RESULTS: A total of 1,388 patients and 13,836 controls wereincluded in the study. We found that 1,317 persons in theentire population redeemed a total of 15,811 prescriptions ofazathioprine in the county. The incidence rate for acutepancreatitis among all users of azathioprine was one per 659treatment year. The crude OR of having redeemed prescrip-tions for azathioprine within 90 days before admission foracute pancreatitis was 7.5 (95% CI� 2.6–21.6). Afteradjustment for gallstone disease, alcohol-related diseases,inflammatory bowel disease, and use of glucocorticoids, theOR increased to 8.4 (95% CI� 2.4–29.4). The population-attributable risk, which measures the proportion of all casesof pancreatitis that are attributable to the use of azathioprinein our study population, was 0.4%.

CONCLUSIONS: There was a substantially increased relativerisk of acute pancreatitis in users of azathioprine. (Am JGastroenterol 2003;98:1305–1308. © 2003 by Am. Coll. ofGastroenterology)

INTRODUCTION

Acute pancreatitis is a serious disease with a case fatalityrate of up to 15% (1). Established risk factors includegallstones, alcohol consumption, hypercalcemia, hyper-lipidemia, trauma, ERCP, and probably certain drugs (2, 3).Azathioprine is an effective drug in maintaining remission

for Crohn’s disease, with evidence for a steroid-sparingeffect (4). Furthermore, it has been used in renal transplantpatients, in the treatment of severe rheumatoid arthritis, andin chronic active hepatitis. Azathioprine has been linked toacute pancreatitis, but this alleged association has so farbeen based only on animal studies, case reports, and smallcase series (5–9). The mechanism of azathioprine-inducedpancreatitis remains speculative. It has been attributed to akind of hypersensitivity reaction (7) similar to that of 6-mer-captopurine-related pancreatitis (10). In an improved animalmodel, azathioprine causes deterioration of already compro-mized pancreatic microcirculation, thereby increasingischemia and acinar cell injury (11).

Given the morbidity and mortality associated with acutepancreatitis, estimates of its risk in patients using azathio-prine have important clinical and public health implications.We, therefore, conducted a population-based nested case–control study in Denmark to examine the risk of acutepancreatitis in patients who redeemed prescriptions for aza-thioprine, taking into account the use of other drugs and riskfactors associated with acute pancreatitis.

MATERIALS AND METHODS

Study Population and DesignWe conducted this register-based nested case–control studywithin the population of North Jutland County, Denmark,with about 490,000 inhabitants, representing 9% of theDanish population.

Cases of Acute PancreatitisWe used computerized hospital discharge data from theHospital Discharge Registry in the county, from 1991–2000.This registry, established in 1977, includes civil registrynumbers, dates of admission and discharge, surgical proce-dures performed, and up to 20 discharge diagnoses coded bymedical doctors at discharge according to the Danish ver-sion of the International Classification of Diseases (ICD)(ICD-8 from 1977 to 1993 and ICD-10 from 1994; ICD-9was never used in Denmark) (12). The following ICD-8 andICD-10 codes were used to identify patients with acutepancreatitis: 577.00–577.09 and K85.9. Dates of death andemigration were obtained from the Civil Registration Sys-

THE AMERICAN JOURNAL OF GASTROENTEROLOGY Vol. 98, No. 6, 2003© 2003 by Am. Coll. of Gastroenterology ISSN 0002-9270/03/$30.00Published by Elsevier Inc. doi:10.1016/S0002-9270(03)00233-8

tem, which records all changes in vital status of all Danishcitizens, including change of address, date of emigration,and date of death since 1968 (13). We have recently vali-dated the diagnosis of incident acute pancreatitis in theHospital Discharge Registry by medical record review, from1981 to 2000. In this validation study, we accepted thediagnosis when the following criteria for acute pancreatitiswere met: a clinical presentation compatible with acutepancreatitis (abdominal pain) combined with at least one ofthe following criteria: a 2-fold increase in serum amylase,typical radiological signs by ultrasound or CT scan, typicalfindings at surgery, or typical pathological findings at au-topsy. The diagnosis was correctly recorded in 82% (95%CI � 72.8–88.9) (14). In the present study, we validated thediagnosis of acute pancreatitis of all cases exposed to aza-thioprine by using the same criteria, and validated the ex-posure by review of the hospital medical records. We ex-amined the severity of acute pancreatitis among casesassociated with use of azathioprine. In accordance with theconsensus from Atlanta, we considered a case of acutepancreatitis to be severe when it was associated with organfailure and/or local complications, such as necrosis, abscess,or pseudocyst, and to be mild when associated with minimalorgan dysfunction and an uneventful recovery (15).

ControlsFor each case with residence in North Jutland County, 10controls were randomly selected from the Central PersonalRegistry, in total 13,820 controls. The controls, matched byage and sex, were selected when their corresponding casewas diagnosed with acute pancreatitis. When using theincidence density sampling technique, the estimated expo-sure OR is an unbiased estimate of the relative risk (16).

Use of AzathioprineWe used the population-based North Jutland PrescriptionDatabase initiated on January 1, 1989, and with completecoverage from January 1, 1991 (17). This registry retainskey information on prescriptions for refundable drugs dis-pensed from all pharmacies in the county, including the civilregistry number of the patient, type of drug prescribedaccording to the anatomical therapeutical chemical (ATC)classification system (18), and date of prescription (date ofdispension of the drug). The National Health Service pro-vides tax-supported health care for all inhabitants, guaran-teeing free access to family doctors, hospitals, and publicclinics, and refunding a variable proportion (usually 50–75%) of the costs of medication prescribed by physicians.Using the civil registry number, which encodes gender anddate of birth, a complete prescription history can be estab-lished for each individual, and unambiguous linkages withother registries can be performed. A 90-day period beforethe date of hospital admission for acute pancreatitis of thecorresponding cases was chosen as the exposure window.We identified all prescriptions for azathioprine (ATC codeL04AX01) issued to cases and controls within this exposurewindow from the Prescription Database.

ConfoundersThe following discharge diagnoses were considered to bepotential confounders, and data were collected from theHospital Discharge Registry only if the diagnoses wererecorded before the date of admission for acute pancreatitis(ICD-8 and ICD-10 codes, respectively): gallstone disease(574.00, 574.09, K80.1-8), alcohol-related diseases (303.09,303.19, 303.20, 303.28, 303.29, 303.90, 303.91, 303.99,979, 980, F10.0-9, 571.09, 571.10, 570.0, 570.9, 571,K70.0-9, 573.00, 573.01, K71.1-2, 577.10, K86.0-9, Z72.1,R78.0, T51), and inflammatory bowel disease (563.01, K50,563.19, 569.04, K51.1-3). Furthermore, the concomitant useof glucocorticoids or cyclosporine may be an importantconfounding factor in patients exposed to azathioprine (19,20); they have both been associated with acute pancreatitis.We therefore extracted prescriptions for glucocorticoids andcyclosporine from the Prescription Database using the fol-lowing ATC codes: ATC H02AB09, H02AB06, H02AB07,H02AB01, H02AB02, and H02AB04 for glucocorticoids,and L04AA01 for cyclosporine. We likewise extracted allprescriptions issued to cases and controls within the 90 daysbefore admission for acute pancreatitis in the correspondingcases (exposure window).

Statistical AnalysisThe number of person-year of exposure to azathioprine wasestimated for the patients who redeemed a prescription forazathioprine in North Jutland assuming an average exposurewindow of 90 days for each prescription. We then calculatedthe incidence rate of acute pancreatitis in all users of aza-thioprine in North Jutland County by dividing the number ofcases of acute pancreatitis by the person-year of exposure toazathioprine.

In the case–control analysis, we analyzed the data first byobtaining a contingency table for the main study variable, andcalculated the OR for the risk of acute pancreatitis in users ofazathioprine, with approximate 95% CIs. We then stratified thecontingency table according to each of the confounding vari-ables. We estimated the OR with conditional logistic regressionanalysis adjusting for possible confounders.

To examine the impact of azathioprine on the overall riskof acute pancreatitis, we calculated the population-attribut-able risk for the use of azathioprine, i.e., the proportion of allcases of acute pancreatitis that are attributable to azathio-prine. In case–control studies, the population-attributablerisk may be calculated by the following equation:

Population-attributable risk

� p � (OR � 1)/{[p � (OR � 1)] � 1}.

The proportion of exposed subjects in the reference popu-lation, p, can be estimated by the exposure prevalence incontrols in rare diseases, and when controls are representa-tive for all noncases in the population (16). Analyses weredone using SAS Version 8.02 (SAS Institute, Cary, NC).The study was approved by the Scientific Ethics Committee

1306 Floyd et al. AJG – Vol. 98, No. 6, 2003

of North Jutland and Viborg Counties and by the DanishRegistry Board (No. 2-16-4-5-96[96/97]).

RESULTS

A total of 1388 incident cases with the discharge diagnosisof acute pancreatitis were identified during the study period,six of which had reimbursed a prescription for azathioprineduring the preceeding 3 months. Further details about casesand controls and possible confounder variables are shown inTable 1. There were no prescriptions for cyclosporine incases and controls within the exposure window.

Incidence Rate Among All Users of Azathioprine inNorth Jutland, 1991–2000We found that 1,317 persons redeemed a total of 15,811prescriptions in the county (573 men, median age 46 yr; 744women, median age 47 yr), making a total of 3,953 per-son-yr of exposure to azathioprine. The six incident cases ofacute pancreatitis that had been exposed to azathioprinewithin 90 days (median 26.5 days) gave an incidence rate ofacute pancreatitis of one per 659 person-yr of exposure toazathioprine among all users of azathioprine.

Risk EstimatesThe crude OR of having reimbursed prescriptions for aza-thioprine within 90 days before admission for acute pancre-atitis was 7.5 (95% CI � 2.4–26.0); after adjustment, theOR increased to 8.4 (95% CI � 2.4–29.4) (Table 2).

Validation of the Diagnosis of Acute PancreatitisAmong Users of AzathioprineWe reviewed the hospital medical records of the six exposedpatients. Five of them fulfilled the above mentioned a prioricriteria for acute pancreatitis. In one case, the diagnosis wassuspected acute pancreatitis. This patient developed epigas-tric pain and tenderness 2 wk after the onset of treatmentwith azathioprine for Crohn’s disease in the colon, and wasadmitted with a moderately elevated serum amylase. Theultrasound scan was normal apart from an enlarged pan-creas. No other risk factors for acute pancreatitis werefound. After azathioprine treatment was stopped, the amy-lase level decreased, and the patient recovered uneventfully.At discharge, the patient was rechallenged with azathio-prine, and a minor increase of serum amylase followed.Subsequently, the patient stopped all treatment for Crohn’sdisease and had no recurrent symptoms for acute pancreati-tis during 8 yr of follow-up.

Validation of Exposure to Azathioprine Among Cases ofAcute PancreatitisAll six exposed patients were current users of azathioprinewhen they were admitted to hospital for acute pancreatitis.

Severity of Acute PancreatitisThe clinical course was mild in four cases, with an unevent-ful recovery. In one case, vena cava thrombosis was diag-nosed by ultrasound, treatment for thrombosis was started,and the patient recovered. One patient died, and hemor-raghic pancreatitis was found post mortem.

Population-Attributable RiskUnder the assumption that the association is causal, and anexposure prevalence in controls of 0.058%, the population-attributable risk (etiological fraction) was 0.4%.

DISCUSSION

We found an 8-fold increase in the risk of acute pancreatitisin users of azathioprine, and four of 1000 admissions ofincident acute pancreatitis in our study population may beattributed to the prescription of azathioprine.

The main strengths of our register-based study are itslarge size, the uniformly organized health care system al-lowing a population-based design, and the use of appropri-ate population controls. Furthermore, the nested case–con-trol design with the incidence density sampling techniqueallowed us to estimate the relative risk for the association

Table 1. Descriptive Characteristics of Cases With Acute Pancre-atitis and Controls From the County of North Jutland, Denmark,1991–2000

Characteristics Cases (%) Controls (%)

Total 1,388 13,836Gender

Men 715 (51.5) 7,123 (51.5)Women 673 (48.5) 6,713 (48.5)

Age (yr)�40 338 (24.4) 3,379 (24.4)40–60 445 (32.1) 4,450 (32.2)�60 605 (43.6) 6,007 (43.4)

Gallstone disease 462 (33.3) 242 (1.7)Alcohol-related diseases 165 (11.9) 228 (1.6)Inflammatory bowel disease 14 (1.0) 50 (0.4)Prescription of

Glucocorticoids 58 (4.2) 334 (2.4)Cyclosporine 0 0

Table 2. Crude and Adjusted ORs of Acute Pancreatitis According to Prescription of Azathioprine Within 90 Days Before the Date ofHospitalization

Casesn

Controlsn

Crude OR(95% CI)

Adjusted OR*(95% CI)

Prescription of azathioprineNo 1,382 13,828 1 (reference) 1 (reference)Yes 6 8 7.5 (2.6–21.6) 8.6 (2.4–29.4)

* OR adjusted for gallstone disease, alcohol-related diseases, inflammatory bowel disease, and use of glucocorticoids.

1307AJG – June, 2003 Acute Pancreatitis and Azathioprine

between acute pancreatitis and the intake of azathioprine.The lack of specificity in coding of acute pancreatitis in thedischarge registry may have led us to underestimate the truerisk. By contrast with other studies, prescription and out-come data were collected independently, thereby avoidingrecall bias, which often hampers case–control studies based oninterviews or questionnaires, the most frequently used designin examining the risk factors for acute pancreatitis (21).

Azathioprine is usually used in serious complications ofchronic diseases, with regular laboratory surveillance of thepatients, and the compliance is therefore considered high.However, any noncompliance is probably nondifferentialbetween cases and controls, and this will lead to a conser-vative risk estimate.

Furthermore, clinicians caring for patients with underly-ing diseases, e.g., patients with gallstones or alcohol-relateddisease, may have increased their surveillance for acutepancreatitis, and vice versa. In cases of acute pancreatitis,one is more likely to discover and register possible con-founders, such as gallstone disease. A more complete reg-istration of confounders among cases would also lead to aconservative risk estimate. Inflammatory bowel disease maybe an independent risk factor for acute pancreatitis (22).However, controlling for inflammatory bowel disease in theanalyses did not alter the risk estimate.

Our study supports previous findings based on case re-ports and case series, reporting a possible association be-tween the intake of azathioprine and acute pancreatitis (19,20, 23, 24). In a recent Danish case series on spontaneousreports about drug-related acute pancreatitis, five of 47reported cases of acute pancreatitis in Denmark were relatedto azathioprine from 1968 to 1995 (9). There are limitationsto the use of spontaneous reporting systems when the pop-ulation treated with the drug in question is not known, and it isthought that only few cases are reported to the registry (23).

In conclusion, we found an increased risk of acute pan-creatitis in users of azathioprine in the population of NorthJutland. There may be important reasons for prescribing thedrug, which in some cases may be the only suitable treatment.

ACKNOWLEDGMENTS

This study has received financial support from the WesternDanish Research Forum for Health Sciences (“VestdanskForskningsforum” ) and by grants from the Medical Re-search Council of North Jutland (grant no. 2-16-4-0005-98)and the “Obelske Familiefond.”

Reprint requests and correspondence: Andrea K. Floyd, M.D.,Department of Clinical Epidemiology, Aalborg University Hospi-tal, Stengade 10, DK-9000 Aalborg, Denmark.

Received June 11, 2002; accepted Oct. 30, 2002.

REFERENCES

1. Eland IA, Sturkenboom MJ, Wilson JH, et al. Incidence andmortality of acute pancreatitis between 1985 and 1995. ScandJ Gastroenterol 2000;35:1110–6.

2. Steinberg W, Tenner S. Acute pancreatitis. N Engl J Med1994;330:1198–210.

3. Gullo L, Migliori M, Olah A, et al. Acute pancreatitis in fiveEuropean countries: Etiology and mortality. Pancreas 2002;24:223–7.

4. Pearson DC, May GR, Fick G, Sutherland LR. Azathioprinefor maintaining remission of Crohn’s disease (Cochrane Re-view). In: The Cochrane Library Issue 2, 2003. Oxford: Up-date Software. Cochrane Database Syst Rev 2000;CD 000067.

5. Tragnone A, Bazzocchi G, Aversa G, et al. Acute pancreatitisafter azathioprine treatment for ulcerative colitis. Ital J Gas-troenterol 1996;28:102–4.

6. Siwach V, Bansal V, Kumar A, et al. Post-renal transplantazathioprine-induced pancreatitis. Nephrol Dial Transplant1999;14:2495–8.

7. Nogueira JR, Freedman MA. Acute pancreatitis as a compli-cation of Imuran therapy in regional enteritis. Gastroenterol-ogy 1972;62:1040–1.

8. Frick TW, Speiser DE, Bimmler D, et al. Drug-induced acutepancreatitis: Further criticism. Dig Dis 1993;11:113–32.

9. Andersen V, Sonne J, Andersen M. Spontaneous reports ondrug-induced pancreatitis in Denmark from 1968 to 1999. EurJ Clin Pharmacol 2001;57:517–21.

10. Bank L, Wright JP. 6-Mercaptopurine-related pancreatitis in 2patients with inflammatory bowel disease. Dig Dis Sci 1984;29:357–9.

11. Foitzik T, Forgacs B, Ryschich E, et al. Effect of differentimmunosuppressive agents on acute pancreatitis: A compara-tive study in an improved animal model. Transplantation1998;65:1030–6.

12. Andersen TF, Madsen M, Jorgensen J, et al. The DanishNational Hospital Register. A valuable source of data formodern health sciences. Dan Med Bull 1999;46:263–8.

13. Sørensen HT. Regional administrative health registries as a re-source in clinical epidemiology. Int J Risk Safety Med 1997;10:1–22.

14. Floyd AK, Pedersen L, Nielsen GL, et al. Secular trends inincidence and 30-day case fatality of acute pancreatitis. ScandJ Gastroenterol 2002;37:1461–3.

15. Bradley EL III. A clinically based classification system foracute pancreatitis. Summary of the International Symposiumon Acute Pancreatitis, Atlanta, September 11 through 13,1992. Arch Surg 1993;128:586–90.

16. Szklo M, Nieto NF. Epidemiology. Beyond the basics. Gaith-ersberg, Maryland: Aspen Publishers, Inc. 2000.

17. Gaist D, Sorensen HT, Hallas J. The Danish prescriptionregistries. Dan Med Bull 1997;44:445–8.

18. Capella D. Descriptive tools and analysis. WHO Reg Publ EurSer 1993;45:55–78.

19. Wilmink T. Drug-induced pancreatitis. Drug Saf 1996;14:406–23.

20. Underwood TW, Frye CB. Drug-induced pancreatitis. ClinPharm 1993;12:440–8.

21. Blomgren KB, Sundstrom A, Steineck G, et al. A Swedishcase-control network for studies of drug-induced morbidity—acute pancreatitis. Eur J Clin Pharmacol 2002;58:275–83.

22. Rasmussen HH, Fonager K, Sørensen HT, et al. Risk of acutepancreatitis in patients with chronic inflammatory bowel dis-ease. Scand J Gastroenterol 1999;34:199–201.

23. Eland IA, van Puijenbroek EP, Sturkenboom MJ, et al. Drug-associated acute pancreatitis. 21 years of spontaneous report-ing in the Netherlands. Am J Gastroenterol 1999;94:2417–22.

24. Mallory A, Kern F Jr. Drug-induced pancreatitis: A criticalreview. Gastroenterology 1980;78:813–20.

1308 Floyd et al. AJG – Vol. 98, No. 6, 2003