Celiac disease (CD) is an immune-mediated disorder that. Risk of Pancreatitis in 14,000 Individuals With Celiac Disease

Transcription

1 CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2007;5: Risk of Pancreatitis in 14,000 Individuals With Celiac Disease JONAS F. LUDVIGSSON,*, SCOTT M. MONTGOMERY,,, and ANDERS EKBOM,, *Department of Pediatrics, Örebro University Hospital, Linkoping, Clinical Epidemiology Unit, Department of Medicine, Karolinska Institutet, Karolinska University Hospital, Stockholm, and Clinical Research Centre, Örebro University Hospital, Linkoping, Sweden; Department of Primary Care and Social Medicine, Imperial College, London, United Kingdom; and Harvard School of Public Health, Boston, Massachusetts Background & Aims: The aim of this study was to examine the risk of pancreatitis in patients with celiac disease (CD) from a general population cohort. Methods: By using Swedish national registers, we identified 14,239 individuals with a diagnosis of CD ( ) and 69,381 reference individuals matched for age, sex, calendar year, and county of residence at the time of diagnosis. Cox regression estimated the hazard ratios (HRs) for a subsequent diagnosis of pancreatitis. We restricted analyses to individuals with more than 1 year of follow-up and no diagnosis of pancreatitis before or within 1 year after study entry. Conditional logistic regression estimated the association of pancreatitis with subsequent CD. Results: CD was associated with an increased risk of subsequent pancreatitis of any type (HR, 3.3; 95% confidence interval [CI], ; P <.001; on the basis of 95 positive events in individuals with CD vs 163 positive events in reference individuals) and chronic pancreatitis (HR, 19.8; 95% CI, ; P <.001; on the basis of 37 and 13 positive events, respectively). Adjustment for socioeconomic index, diabetes mellitus, alcohol-related disorders, or gallstone disease had no notable effect on the risk estimates. The risk increase for pancreatitis was only found among individuals with CD diagnosed in adulthood. Pancreatitis of any type (odds ratio, 3.2; 95% CI, ; P <.001) and chronic pancreatitis (odds ratio, 7.3; 95% CI, ; P <.001) were associated with subsequent CD. Conclusions: This study suggests that individuals with CD are at increased risk of pancreatitis. Celiac disease (CD) is an immune-mediated disorder that occurs in some 1% of the Western population. 1 On being exposed to gluten, individuals with CD develop inflammation with crypt hyperplasia and villous atrophy of the small intestine. CD therefore requires a lifelong gluten-free diet. Although this is principally a disease of the small intestine, extraintestinal complications are common. 2 4 Each year 80,000 Americans have acute pancreatitis. 5 Typical symptoms include severe upper abdominal pain radiating through the back, vomiting, and nausea. Pancreatitis is commonly associated with excessive consumption of alcohol or the presence of gallstones, 6 but other causes have also been suggested. A number of studies have reported an association between CD and pancreatic disease, including exocrine pancreatic insufficiency, 7,8 pancreatic cancer (standardized incidence ratio, 1.9; 95% confidence interval [CI], ), 9 and death from pancreatic disorders (standardized mortality ratio, 2.8; 95% CI, ). 10 CD should also be suspected in patients with an unexplained increase of pancreatic enzymes. 11 Other than case reports, 12,13 we are only aware of 1 study evaluating the association of CD with pancreatitis. 14 Patel et al 14 screened 169 patients with possible sphincter of Oddi dysfunction for gliadin and endomysial autoantibodies and found 12 patents with CD, of whom 10 had a history of recurrent pancreatitis. Most of these patients had undiagnosed CD and improved on a gluten-free diet. Although the study by Patel et al supports a positive association between CD and pancreatic disease, it was based on only 10 patients with both diseases and used no comparison group. We examined the relationship between CD and pancreatitis by using data from Swedish national registers. Materials and Methods The Swedish National Board of Health and Welfare used the Swedish national inpatient register (IPR) to identify all individuals with CD diagnosed between The IPR became nationwide in CD, our outcome measures (any or chronic pancreatitis), as well as potential confounding factors (diabetes mellitus [DM], gallstone disease, and use of alcohol) were defined according to relevant International Classification of Disease (ICD) codes in the IPR (Appendix Table; see supplementary material online at The IPR does not distinguish between type 1 and type 2 DM. We used cholecystectomy as a measure of gallstone disease and hospital admission as a result of alcohol-related disorders as a measure of high alcohol consumption (Appendix). Statistics Sweden, the government agency responsible for producing official statistics on Sweden, matched each individual with CD with up to 5 reference individuals without CD. Matching variables were age, sex, calendar year, and county of residence at the time of diagnosis. To minimize the risk of ascertainment bias, follow-up time began 1 year after the diagnosis of CD was made (this is equivalent to study entry in individuals with CD and in matched reference individuals). It ended on the date of first discharge diagnosis of pancreatitis, emigration, death, or the end of the study period (December 31, 2003), whichever occurred first. In subanalyses in which the outcome measure was having at least 2 diagnoses of pancreatitis, the end of follow-up was the date of first diagnosis of pancreatitis in those with at least 2 IPR diagnoses of pancreatitis (or death, emigration, or Abbreviations used in this paper: CD, celiac disease; CI, confidence interval; DM, diabetes mellitus; HR, hazard ratio; ICD, International Classification of Disease [codes]; IPR, Swedish national inpatient register; OR, odds ratio; SEI, socioeconomic index by the AGA Institute /07/$32.00 doi: /j.cgh

2 1348 LUDVIGSSON ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 11 December 31, whichever occurred first). In those with only 1 diagnosis of pancreatitis and those without such diagnosis, the end of follow-up in this separate analysis was the date of emigration, death, or December 31, 2003, whichever occurred first. From the original 15,533 individuals with CD, 94 were excluded because of data irregularities, such as death recorded as occurring before diagnosis of CD. We then excluded 114 individuals with pancreatitis occurring before study entry, an additional 71 individuals with pancreatitis within the first year of follow-up, and 1015 individuals with a shorter follow-up than 1 year. Similar exclusion criteria were applied to the reference individuals. The main analyses of this study were therefore based on 14,239 individuals with CD and 69,381 reference individuals without a diagnosis of CD. In a subset of individuals we had data on socioeconomic index (SEI) based on occupation (CD: n 8752; reference individuals: n 36,422). Some 6500 of these individuals were children at final follow-up who had been assigned the socioeconomic code of their mother. Data Analysis Celiac disease and subsequent pancreatitis. Cox regression estimated the hazard ratio (HR) for subsequent pancreatitis (any, chronic). We used a matched approach in which each individual with CD was only compared with his/her ageand sex-matched reference individuals. Hence, comparisons were made within each risk-set. We evaluated the proportional hazards assumption through log minus log plots. In separate analyses, we stratified by sex and age at first recorded hospital discharge diagnosis of CD ( 15 years, 16 years). We evaluated the influence of SEI on our risk estimates through calculating both crude and adjusted HRs for pancreatitis in a subset of individuals with data on SEI. To increase the specificity for our outcome measures, we estimated the risk of having (1) a diagnosis of pancreatitis recorded in departments in which pancreatitis is most often investigated and diagnosed, and in which diagnostic specificity should be highest (surgery, medicine, gastroenterology, and pediatrics). (2) We examined CD and the risk of having at least 2 diagnoses of pancreatitis. (3) We also limited the definition of pancreatitis to where it was listed as the main diagnosis. In separate analyses, we also included the first year after study entry in the follow-up to assess how this would affect the risk estimates. We performed adjustment for DM, alcohol use, and gallstone disease, because they are potential confounding factors. In a separate analysis we excluded individuals with pancreatic cancer because CD is associated with this type of cancer. 9 Data on pancreatic cancer were obtained from the IPR (Appendix). All individuals with CD had been hospital inpatients at study entry. To minimize the risk of differential ascertainment bias as a result of hospitalization in individuals with CD, we carried out post hoc analyses restricting reference individuals to those with a hospitalization episode within 1 year before or after the diagnosis of CD in the index individual with CD. This restriction resulted in fewer reference individuals, and several strata thereby consisted only of 1 individual with CD. Instead of using an internally matched approach that would have resulted in low power, we included all individuals fulfilling the criteria of the post hoc analyses and then adjusted for age, sex, and calendar year. To maximize power we included the first year of follow-up. This was deemed reasonable because all individuals in the post hoc analyses were inpatients. Prior pancreatitis and celiac disease. We used conditional logistic regression to estimate the odds ratios (ORs) for subsequent CD in individuals with prior pancreatitis (any or chronic). In these analyses, CD was our dependent variable, and end of follow-up was defined as date of first CD diagnosis (corresponding date in reference individuals). Individuals with 1 year between the date of first diagnosis of pancreatitis and CD diagnosis (corresponding date in reference individuals) were excluded. Ninety-five percent CIs for HRs not including 1.00 were considered statistically significant. Statistics were calculated with SPSS 11.0 (SPSS Inc, Chicago, IL). Power Calculation At a significance level of 5%, we had an 80% power to detect a 1.5-fold increased risk of any pancreatitis in individuals with CD and a 3.2-fold increased risk of chronic pancreatitis. Ethics The Research Ethics Committee of Karolinska Institutet approved this study. None of the participants were contacted. Patient information was anonymized before the analyses, and individual patient charts could not be traced. Results Characteristics of Study Participants The majority of patients were female (Table 1). Although most individuals were children at study entry, this study also included more than 4800 individuals with CD diagnosed in adulthood. Some 74% of study participants (10,573 individuals with CD and 51,283 reference individuals) were 16 years or older at end of follow-up. The median age at study entry was 2 years (range, 0 94 years; mean, 19.6 years) in individuals with CD and 3 years (range, 0 94 years; mean, 20.1 years) in reference individuals. The median age and median duration from study entry until first diagnosis of any or chronic pancreatitis are given in Table 1. Celiac Disease and Subsequent Pancreatitis of Any Type CD was associated with a 3-fold increased risk of pancreatitis (Table 2, Appendix Figure [see supplementary material online at The risk increase for any form of pancreatitis was only seen in individuals with CD diagnosed in adulthood in whom the risk estimate was 4.4 (95% CI, ). The absolute excess risk for pancreatitis in individuals with CD diagnosed in adulthood was 142/100,000 person-years (Table 3). There was insufficient follow-up time to evaluate effectively the risk of pancreatitis associated with CD diagnosed in childhood (HR, 1.2; 95% CI, ). The difference in risk estimates between males and females was not statistically significant (formal interaction test: P.181). Adjustment for DM, alcohol-related disorders, gallstone disease, or SEI in a subset of individuals did not affect the risk estimates (Table 2). When we restricted our analysis to ICD codes for acute pancreatitis (Appendix), the HR was 2.5 (95% CI, ; P.001; on the basis of 67 individuals with CD and later acute pancreatitis vs 158 reference individuals). When we included the first year of follow-up, there was a 5-fold increased risk of subsequent pancreatitis of any type in

3 November 2007 CELIAC DISEASE AND PANCREATITIS 1349 Table 1. Characteristics of Participants With at Least 1 Year of Follow-up Characteristics Reference (no CD) (%) CD (%) Total 69,381 14,239 Age at first recorded diagnosis of CD (y) ,368 (65.8) 16 4,871 (34.2) Sex Male 28,449 (41.0) 5,858 (41.1) Female 40,932 (59.0) 8,381 (58.9) Calendar period a ,385 (3.4) 491 (3.4) ,047 (27.5) 3,894 (27.3) ,578 (44.1) 6,255 (43.9) ,371 (25.0) 3,599 (25.3) SEI I 7,218 (20.4) 1,491 (10.5) II 9,197 (13.3) 2,142 (15.0) III 20,007 (28.8) 5,119 (40.0) Missing data 32,959 (47.5) 5,487 (38.5) Comorbidity DM 1,944 (2.8) 915 (6.4) Diagnosis indicating alcohol 1,094 (1.6) 319 (2.2) use b Gallbladder operation 1,339 (1.9) 368 (2.6) Median age at pancreatitis (range) c (y) Any pancreatitis 62, , 2 87 Chronic pancreatitis 73, , Median duration until pancreatitis (range) c (y) Any pancreatitis 9, ; 1 27 Chronic pancreatitis 9, , 1 22 NOTE. Individuals with at least 1 year of follow-up after CD diagnosis or corresponding date in matched individuals. SEI: I is highest category. For reference individuals we have given the number of individuals who constituted the basis for the Cox regression. We actually had data on SEI in another 6440 reference individuals, but these individuals were not part of the internally stratified calculations because of missing values on SEI in the matched individual with CD. Adding the 6,440 reference individuals to those presented above, the proportion of missing values was similar among individuals with CD and reference individuals. a The IPR became nationwide (complete coverage of all counties in Sweden) in b The proportion of individuals entering the study in adulthood with a diagnosis of alcohol-related disorder was CD, 4.2% and reference, 2.7%. c Age and duration in years. Median duration indicates the number of years from study entry (and diagnosis of CD) until first diagnosis of any/chronic pancreatitis. Please note that values are based on individuals with pancreatitis after study entry; individuals with pancreatitis recorded before or within the first year of study entry were excluded from the main analyses. The median duration between study entry and pancreatitis might seem short, considering that the median age at study entry was 2 3 years, and most individuals had a diagnosis of pancreatitis in adulthood. The short median duration until pancreatitis is explained by the fact that duration until pancreatitis was only calculated in individuals with pancreatitis. These patients were most often adults at study entry. individuals with CD (166 positive events in 15,325 individuals with CD vs 189 positive events in 75,340 reference individuals) (Table 2). Of the 166 individuals with CD and pancreatitis, 9 patients died within 30 days (5.4%) as opposed to 14 of 189 (7.4%) reference individuals. Although the death causes are not recorded in the IPR, the last main inpatient diagnoses before death in the 9 individuals with CD were any pancreatitis (3), ischemic heart disease (2), CD (1), malnutrition (1), chronic alcoholism (1), and chronic renal failure (1). Restricting our outcome to pancreatitis listed as the main diagnosis; pancreatitis diagnosed in departments of surgery, medicine, gastroenterology, and pediatrics; or having at least 2 diagnoses of pancreatitis had only a marginal effect on the risk estimates (Table 2). When we excluded those with 5 years of follow-up (and those with a diagnosis of pancreatitis within the first 5 years after study entry), the HR for subsequent pancreatitis was 2.7 (95% CI, ; P.001). Celiac Disease and Subsequent Chronic Pancreatitis CD was associated with a 19-fold increased risk of subsequent chronic pancreatitis (Table 2; Appendix Figure) and an even higher risk when we included the first year of follow-up (Table 2). Two individuals with both CD and pancreatitis (2.5%) and 2 reference individuals with pancreatitis (12.5%) died within 30 days of first diagnosis of chronic pancreatitis. Sex did not modify the association between CD and chronic pancreatitis as indicated by formal interaction testing (P.404). All individuals with a subsequent diagnosis of chronic pancreatitis entered the study in adulthood. The absolute excess risk for chronic pancreatitis in individuals with CD diagnosed in adulthood was 71/100,000 person-years (Table 3). Adjustment for DM, alcohol-related disorders, gallstone disease, or SEI in a subset of individuals did not affect the risk estimates (Table 2). This study found a highly increased risk of chronic pancreatitis when it was listed as the main diagnosis; when diagnosed in departments of surgery, medicine, gastroenterology, and pediatrics; or where the diagnosis was recorded at least twice (Table 2). Excluding the first 5 years of follow-up altered the HR for subsequent chronic pancreatitis to 9.6 (95% CI, ; P.001). Incidence of Pancreatitis The incidence of all forms of pancreatitis (per 100,000 person-years) was 50 for individuals with CD and 17 among reference individuals (Table 3). The incidence of chronic pancreatitis was 19 per 100,000 person-years among patients with CD and 1 per 100,000 person-years among the reference population (Table 3). Inpatient Reference Individuals Post hoc analyses found that CD was positively associated with subsequent pancreatitis (adjusted HR, 2.6; 95% CI, ; P.001) and subsequent chronic pancreatitis (adjusted HR, 7.6; 95% CI, ; P.001) when reference individuals were restricted to inpatients. This means that differential ascertainment bias as a result of hospital admission of all individuals with CD at study entry is unlikely to explain the positive association between CD and pancreatitis.

4 1350 LUDVIGSSON ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 11 Table 2. CD and Risk of Subsequent Pancreatitis Any pancreatitis Chronic pancreatitis Outcome HR, 95% CI P value HR, 95% CI P value Pancreatitis 3.3, , Adjusting for DM 3.4, , Adjusting for alcohol a 3.1, , Adjusting for gallstone disease b 3.2, , Adjusting for DM, alcohol, and gallstone disease ab 3.1, , Subgroups Excluding DM 3.6, , Excluding high alcohol consumption a 3.1, , Excluding gallstone disease a 3.7, , Excluding pancreatic cancer 3.3, , Individuals with data on SEI (crude) 3.8, , Individuals with data on SEI (adjusted) 3.9, , Male 4.2, , Female 2.7, , Modified outcome measure Main diagnosis 2.9, , Core departments c 3.3, , diagnoses of pancreatitis 5.0, , First year included Pancreatitis 5.6, , NOTE. First year of follow-up excluded except when otherwise specified. a Diagnoses indicating high alcohol consumption. b Gallstone disease defined as cholecystectomy before end of follow-up (or in individuals with pancreatitis within 1 year after the first diagnosis of pancreatitis). When we extended our concept of gallstone disease and also included individuals with gallstone disease but without cholecystectomy, the HR for any pancreatitis was 4.7 (95% CI, ; P.001) (chronic pancreatitis: HR, 28.5; 95% CI, ; P.001). c Departments of surgery, medicine, gastroenterology, and pediatrics. Prior Pancreatitis and Celiac Disease Conditional logistic regression found that both prior pancreatitis of any type (OR, 3.2; 95% CI, ; P.001) and prior chronic pancreatitis (OR, 7.3; 95% CI, ; P.001) were associated with a subsequent diagnosis of CD. Discussion This study found statistically significant positive associations with CD for any form of pancreatitis and chronic pancreatitis. This relationship was independent of temporal sequence. Adjustment for a number of potential confounding factors, including socioeconomic status, markers of high alcohol consumption, gallstone disease, and DM did not affect our risk estimates. Our results are unlikely to be explained by ascertainment bias because we excluded the first year after diagnosis of CD in the main analysis. Even when we excluded the first 5 years after diagnosis of CD and when we restricted our reference individuals to inpatients, this study found statistically significant positive associations with CD for any form of pancreatitis and chronic pancreatitis. Our study examined hospital admissions for pancreatitis from 1964 to 2003, and incidence data were consistent with those provided by another Swedish study. 15 The characteristics of our patients are also consistent with the earlier study, indi- Table 3. Incidence Rates for Any and Chronic Pancreatitis Any pancreatitis Chronic pancreatitis Outcome CD Reference CD Reference First year excluded Any CD 95/190,122 (50) 163/962,561 (17) 37/190,506 (19) 13/964,103 (1) Children 8/141,534 (6) 33/696,761 (5) Adults 87/45,588 (191) 130/265,800 (49) 37/48,890 (76) 13/267,166 (5) Males 53/79,008 (67) 77/401,310 (19) 22/79,236 (28) 6/402,116 (1) Females 42/111,114 (38) 86/561,252 (15) 15/111,270 (13) 7/561,987 (1) First year included Any CD 166/204,969 (81) 189/1,062,726 (18) 81/205,536 (39) 16/1,064,586 (1) NOTE. Incidence figures per 100,000 person-years are given within parentheses. NB: Y-axis scales are not identical

5 November 2007 CELIAC DISEASE AND PANCREATITIS 1351 cating that they are representative of this population. In our study, the median age at first diagnosis of pancreatitis was 62 years in reference individuals, compared with 59 years in the study by Lindkvist et al. 15 In the study by Patel et al, 14 the mean age of individuals with both CD and sphincter of Oddi dysfunction (10/12 of these had recurrent pancreatitis) was 61 years. Insufficient follow-up time explains why this study was not able to evaluate the risk of subsequent pancreatitis among individuals with CD diagnosed in childhood. The low median age at study entry otherwise mirrors the early diagnosis of CD in Sweden during most of the study period. 16 Individuals with a diagnosis of CD in adulthood were at a 4-fold increased risk of any form of pancreatitis, whereas the risk of subsequent chronic pancreatitis was almost 20 times that of the reference population. Several factors that are not mutually exclusive might help to explain our findings. One of the key features of CD both before and after diagnosis is malnutrition. Whereas low energy intake has been linked to pancreatic dysfunction, 17,18 it might also influence the composition of the bile, inducing microlithiasis and impaired secretion of pancreatic enzymes, 19 thus increasing the risk of pancreatitis. A second explanation involves chronic inflammation. Patel et al 14 suggested that secretion of pancreatic enzymes might be impaired by papillary scarring caused by CD inflammation. It is noteworthy that inflammation in CD might persist for many years after the introduction of glutenfree diet, 20 which could explain the continuing high risk estimates for pancreatitis even 5 years after diagnosis of CD in this study. A third potential explanation for the association of CD with pancreatitis involves shared immunologic traits in both diseases. Although earlier studies report increased levels of both Th1 and Th2-associated cytokines in pancreatitis (CD is skewed toward Th1 responses 21 ), Th1-associated cytokines (interferongamma 22,23 and interleukin-18 24,25 ) are increased in both CD and pancreatitis. A tumor necrosis factor alpha polymorphism has repeatedly been suggested as a marker of CD susceptibility. 26 Data on tumor necrosis factor alpha 308 and pancreatitis are conflicting, but recent data indicate that this tumor necrosis factor alpha polymorphism might be more common in severe pancreatitis 27 and in pancreatitis leading to pancreatic cancer. 28 This study has some potential weaknesses including the definition of alcohol use and the lack of smoking data and dietary data. The IPR only contains data on alcohol-related disorders; 2.7% of reference individuals entering our study in adulthood had a diagnosis indicating high alcohol consumption. Smoking has a positive effect on the risk of certain forms of pancreatitis 29,30 but has, if any, 31 a negative effect on CD. 32 Therefore, our results cannot be explained by confounding through smoking. Unfortunately, the IPR contains no dietary data. Such data could otherwise have helped us understand the relationship between pancreatic insufficiency and consumption of gluten in individuals with CD. Several case reports have indicated that gluten-free diet might reverse pancreatic exocrine insufficiency. 19,33 The risk of observing spurious positive associations might be increased where the same material is used for several studies; however, there are several reasons why this is only of limited concern here. One potential reason for observing positive associations between diseases in material such as this is ascertainment bias; hospital patients are more likely to have multiple investigations and diagnoses. Evidence that this source of bias is unlikely to account for our results is from our previous studies, which have found several negative disease associations, 34 and in many studies we have not observed an association between diseases (eg, schizophrenia 35 and urinary tract infection 36 ). A second reason for observing false-positive associations is that multiple testing increases the risk of a false association arising by chance. This is an unlikely explanation for our findings because they were consistent when stratified by sex and temporal sequence. This consistency is indicative of a true association rather than a chance finding that is more likely to relate to a single subgroup. Nevertheless, we cannot rule out that ascertainment bias has influenced our risk estimates. Such bias will increase the risk of false-positive pancreatitis in individuals with CD. Meanwhile, sensitivity for pancreatitis might be low in reference individuals in our study. The IPR has previously been used to identify individuals with CD, 10 as well as those with chronic pancreatitis. 37 In a subset of individuals with both CD and concomitant lymphoma, the misclassification rate for CD in the IPR was below 15%. 38 This partly stems from the well-established practice in Sweden to perform a small-bowel biopsy before confirming a diagnosis of CD. This is important because the prevalence of CD might be overestimated in individuals with gastrointestinal symptoms, if the diagnosis of CD is based only on serologic findings. In a selected population of patients who underwent pancreaticobiliary manometry on the suspicion of sphincter of Oddi dysfunction, of 169 patients were positive for IgA/ IgG gliadin or endomysial antibodies, but only 12 had CD. The IPR is a hospital-based register. Many gastrointestinal investigative procedures, including small-bowel biopsy, used to be performed in an inpatient setting. This is still often the case in very young children. The increased use of endoscopy, often performed under general anesthesia, might also explain why CD was diagnosed in hospital. After initial investigations, patients are usually transferred to an outpatient setting. Outpatient care has increased over time and might explain the lower number of individuals with a hospital-based diagnosis of CD in as compared with (Table 1). Incomplete coverage of the IPR before 1987 and low degree of awareness of CD might explain the lower numbers of individuals with diagnosed CD during the first 2 calendar periods of this study. This study probably identified a large proportion of those with diagnosed CD. Although screening studies have indicated a prevalence of CD of some 1% in the Western world, 1 we expect to find a diagnosis of CD in about 0.1% of the population because an earlier Swedish study reported that only 2 of 1894 randomly selected adults had a known diagnosis of CD 39 (Sweden currently has 9 million inhabitants). Not every individual with CD will be identified through a hospital-based register. This, however, should not be a major drawback because we identified more than 14,000 individuals with CD. The existence of false-negative individuals with CD (now classified as reference individuals) is unlikely to affect the risk estimates because the overall prevalence of CD is around 1%. 1 Pancreatitis is often diagnosed on the basis of typical symptoms as well as laboratory and radiologic findings. Unfortunately, the IPR does not contain any data on laboratory findings such as amylase levels or investigative procedures such as endoscopic retrograde cholangiopancreatography/magnetic reso-

6 1352 LUDVIGSSON ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 11 nance cholangiopancreatography. The lack of validation is unfortunate because there is a risk that increased levels of pancreatic enzymes (occurring in a subset of individuals with CD) might be misinterpreted as pancreatitis. 40,41 Ideally, cases of pancreatitis should have been validated against patient charts or renewed investigation. This was not possible because all data were anonymized by the National Board of Health and Welfare to protect the integrity of the patients. Three to 6 months after a registry linkage is performed, the Board routinely deletes the identification key used for linkage purposes so that subject identity cannot be traced at a later stage. We know of no validation study of pancreatitis in the IPR. Instead, we increased the specificity of our outcome measures in several ways (pancreatitis listed as the main diagnosis; pancreatitis diagnosed in departments of surgery, medicine, gastroenterology, and pediatrics; 2 diagnoses of pancreatitis), and this did not affect our HRs more than marginally. In a validation study, Nilsson et al 42 evaluated 900 random inpatient episodes from In that study, the specificity for diagnoses listed as the main diagnoses was 86% 89% at the first decimal level of the ICD code (eg, and not 577.1). The lack of specificity in the study by Nilsson et al was often caused by coding errors. Coding errors are most often nondifferential by nature, and although nondifferential bias will lead to a reduction in precision, it should not alter the risk estimates substantially. However, because we were unable to specifically validate the registry diagnoses in this study, the specificity of CD and pancreatitis remains uncertain. Misclassification might therefore have influenced our risk estimates. In conclusion, this general population-based study found a positive association between CD and pancreatitis. This risk was seen both before and after diagnosis of CD. The positive association between CD and pancreatitis might be explained through shared immunologic traits or a combination of malnutrition and long-term inflammation caused by CD. However, we cannot rule out that our risk estimates were influenced by misclassification and ascertainment bias. In this registry-based study we were unable to validate individual diagnoses. Future studies should preferably apply strict diagnostic criteria to both CD and pancreatitis and validate these diagnoses against patient charts and laboratory data. Supplementary Data Note: To access the supplementary materials accompanying this article, visit the online version of Clinical Gastroenterology and Hepatology at References 1. Dube C, Rostom A, Sy R, et al. The prevalence of celiac disease in average-risk and at-risk Western European populations: a systematic review. Gastroenterology 2005;128:S57 S Holmes GK, Prior P, Lane MR, et al. Malignancy in coeliac disease: effect of a gluten free diet. Gut 1989;30: Collin P, Reunala T, Pukkala E, et al. Coeliac disease: associated disorders and survival. Gut 1994;35: Ludvigsson JF, Wahlstrom J, Grunewald J, et al. Coeliac disease and risk of tuberculosis: a population based cohort study. Thorax 2007;62: NIH. Pancreatitis. Available at: ddiseases/pubs/pancreatitis/index.htm#acute. National Digestive Diseases Information Clearinghouse (NDDIC), 2004; last Accessed July 4, Yadav D, Lowenfels AB. Trends in the epidemiology of the first attack of acute pancreatitis: a systematic review. Pancreas 2006;33: Carroccio A, Iacono G, Montalto G, et al. Exocrine pancreatic function in children with coeliac disease before and after a gluten free diet. Gut 1991;32: Leeds JS, Hopper AD, Hurlstone DP, et al. Is exocrine pancreatic insufficiency in adult coeliac disease a cause of persisting symptoms? Aliment Pharmacol Ther 2007;25: Askling J, Linet M, Gridley G, et al. Cancer incidence in a population-based cohort of individuals hospitalized with celiac disease or dermatitis herpetiformis. Gastroenterology 2002;123: Peters U, Askling J, Gridley G, et al. Causes of death in patients with celiac disease in a population-based Swedish cohort. Arch Intern Med 2003;163: Carroccio A, Di Prima L, Scalici C, et al. Unexplained elevated serum pancreatic enzymes: a reason to suspect celiac disease. Clin Gastroenterol Hepatol 2006;4: Fitzgerald O, Fitzgerald P, Fennelly J, et al. A clinical study of chronic pancreatitis. Gut 1963;4: Pitchumoni CS, Thomas E, Balthazar E, et al. Chronic calcific pancreatitis in association with celiac disease. Am J Gastroenterol 1977;68: Patel RS, Johlin FC Jr, Murray JA. Celiac disease and recurrent pancreatitis. Gastrointest Endosc 1999;50: Lindkvist B, Appelros S, Manjer J, et al. Trends in incidence of acute pancreatitis in a Swedish population: is there really an increase? Clin Gastroenterol Hepatol 2004;2: Ivarsson A, Persson LA, Nystrom L, et al. Epidemic of coeliac disease in Swedish children. Acta Paediatr 2000;89: Nordgren L, von Scheele C. Hepatic and pancreatic dysfunction in anorexia nervosa: a report of two cases. Biol Psychiatry 1977; 12: Cox KL, Cannon RA, Ament ME, et al. Biochemical and ultrasonic abnormalities of the pancreas in anorexia nervosa. Dig Dis Sci 1983;28: Regan PT, DiMagno EP. Exocrine pancreatic insufficiency in celiac sprue: a cause of treatment failure. Gastroenterology 1980;78: Lee SK, Lo W, Memeo L, et al. Duodenal histology in patients with celiac disease after treatment with a gluten-free diet. Gastrointest Endosc 2003;57: Stepniak D, Koning F. Celiac disease: sandwiched between innate and adaptive immunity. Hum Immunol 2006;67: Nilsen EM, Lundin KE, Krajci P, et al. Gluten specific, HLA-DQ restricted T cells from coeliac mucosa produce cytokines with Th1 or Th0 profile dominated by interferon gamma. Gut 1995; 37: Pietruczuk M, Dabrowska MI, Wereszczynska-Siemiatkowska U, et al. Alteration of peripheral blood lymphocyte subsets in acute pancreatitis. World J Gastroenterol 2006;12: Salvati VM, MacDonald TT, Bajaj-Elliott M, et al. Interleukin 18 and associated markers of T helper cell type 1 activity in coeliac disease. Gut 2002;50: Ueda T, Takeyama Y, Yasuda T, et al. Significant elevation of serum interleukin-18 levels in patients with acute pancreatitis. J Gastroenterol 2006;41: McManus R, Wilson AG, Mansfield J, et al. TNF2, a polymorphism of the tumour necrosis-alpha gene promoter, is a component of the celiac disease major histocompatibility complex haplotype. Eur J Immunol 1996;26: Balog A, Gyulai Z, Boros LG, et al. Polymorphism of the TNFalpha, HSP70-2, and CD14 genes increases susceptibility to severe acute pancreatitis. Pancreas 2005;30:e46 e Duell EJ, Casella DP, Burk RD, et al. Inflammation, genetic polymorphisms in proinflammatory genes TNF-A, RANTES, and CCR5,

7 November 2007 CELIAC DISEASE AND PANCREATITIS 1353 and risk of pancreatic adenocarcinoma. Cancer Epidemiol Biomarkers Prev 2006;15: Morton C, Klatsky AL, Udaltsova N. Smoking, coffee, and pancreatitis. Am J Gastroenterol 2004;99: Maisonneuve P, Frulloni L, Mullhaupt B, et al. Impact of smoking on patients with idiopathic chronic pancreatitis. Pancreas 2006; 33: Ludvigsson JF, Montgomery SM, Ekbom A. Smoking and celiac disease: a population-based cohort study. Clin Gastroenterol Hepatol 2005;3: Snook JA, Dwyer L, Lee-Elliott C, et al. Adult coeliac disease and cigarette smoking. Gut 1996;39: DiMagno EP, Go WL, Summerskill WH. Impaired cholecystokininpancreozymin secretion, intraluminal dilution, and maldigestion of fat in sprue. Gastroenterology 1972;63: Ludvigsson JF, Askling J, Ekbom A, et al. Diagnosis underlying appendectomy and coeliac disease risk. Dig Liver Dis 2006;38: Ludvigsson JF, Osby U, Ekbom A, et al. Coeliac disease and risk of schizophrenia and other psychosis: a general population cohort study. Scand J Gastroenterol 2007;42: Olen O, Montgomery SM, Ekbom A, et al. Urinary tract infections in pregnant women with coeliac disease. Scand J Gastroenterol 2007;42: Ye W, Lagergren J, Weiderpass E, et al. Alcohol abuse and the risk of pancreatic cancer. Gut 2002;51: Smedby KE, Akerman M, Hildebrand H, et al. Malignant lymphomas in coeliac disease: evidence of increased risks for lymphoma types other than enteropathy-type T cell lymphoma. Gut 2005;54: Ivarsson A, Persson LA, Juto P, et al. High prevalence of undiagnosed coeliac disease in adults: a Swedish population-based study. J Intern Med 1999;245: Hodgson HJ, Whitaker KB, Cooper BT, et al. Malabsorption and macroamylasemia: response to gluten withdrawal. Am J Med 1980;69: Rajvanshi P, Chowdhury JR, Gupta S. Celiac sprue and macroamylasemia: potential clinical and pathophysiological implications case study. J Clin Gastroenterol 1995;20: Nilsson AC, Spetz CL, Carsjo K, et al. [Reliability of the hospital registry: the diagnostic data are better than their reputation]. Lakartidningen 1994;91:598, Address requests for reprints to: Jonas F. Ludvigsson, MD, Department of Pediatrics, Örebro University Hospital, Örebro, Sweden. jonasludvigsson@yahoo.com; fax: Supported by the Swedish Research Council, the Swedish Society of Medicine, the Örebro Society of Medicine, the Majblomman Foundation, the Sven Jerring Foundation, the Clas Groschinsky Foundation, the Samariten Foundation, the Karolinska Institutet, and the Swedish Celiac Society. J.F.L. was supported by the Swedish Research Council and the Örebro University Hospital while carrying out this study.

8 November 2007 CELIAC DISEASE AND PANCREATITIS 1353.e1 Appendix Table. ICD Codes of CD, Pancreatitis, DM, and Alcohol-Related Disorders ICD-7 ICD-8 ICD-9 CD-10 CD , A K90.0 Any pancreatitis a A-B K85, K Chronic pancreatitis a , , B K Acute pancreatitis a , , , , , 577A K , DM b E10-14 Alcohol use , , , , , 322, , , , 291, 303, , , 979, , , , , 303, 305A, 357F, 425F, 535D, 571A-D, 760W, 790D, 977D, 980A, 980X, V97B Pancreatic cancer C25 F10, G31.2, G62.1, G72.1, I42.6, K29.2, K70, K86.0, Q35.4, R78.9, T51.0, T51.8, T51.9, X65, Y15, Y57.3, Y90, Y91, Z50.2, Z71.4, Z71.2 NOTE. Cholecystectomy (gallstone disease) was defined according to the Swedish Classification of Operation and Major Procedures (data obtained from the inpatient register) with the following codes signifying cholecystectomy: JKA, 5300; 5302, 5304, 5306, 5309, 5340, 5341, 5350, 5351, 5352, 5356, 5357, 5359, and In additional analyses we used a more extended concept of gallstone disease that also included individuals without cholecystectomy but with the following ICD diagnoses: ICD-7: 584, ICD-8: 574, ICD-9: 574, ICD-10: K80, K81. a We did not include ICD codes representing pancreatitis caused by epidemic parotitis (mumps) or cytomegalovirus infection. b The IPR does not distinguish between type 1 and type 2 DM.

9 1353.e2 LUDVIGSSON ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 11 Appendix Figure. Cumulative risk of pancreatitis.