Presented to a meeting of the American Pancreatic Association, Chicago, Illinois, USA, November 2010, and the International Research Workshop on Acute Pancreatitis, Szeged, Hungary, March 2011
Effect of type of alcoholic beverage in causing acute pancreatitis†
Article first published online: 3 AUG 2011
Copyright © 2011 British Journal of Surgery Society Ltd. Published by John Wiley & Sons, Ltd.
British Journal of Surgery
Volume 98, Issue 11, pages 1609–1616, November 2011
How to Cite
Sadr Azodi, O., Orsini, N., Andrén-Sandberg, Å. and Wolk, A. (2011), Effect of type of alcoholic beverage in causing acute pancreatitis. Br J Surg, 98: 1609–1616. doi: 10.1002/bjs.7632
- Issue published online: 3 OCT 2011
- Article first published online: 3 AUG 2011
- Manuscript Accepted: 23 MAY 2011
The effect of different alcoholic beverages and drinking behaviour on the risk of acute pancreatitis has rarely been studied. The aim of this study was to investigate the effect of different types of alcoholic beverage in causing acute pancreatitis.
A follow-up study was conducted, using the Swedish Mammography Cohort and Cohort of Swedish Men, to study the association between consumption of spirits, wine and beer and the risk of acute pancreatitis. No patient with a history of chronic pancreatitis was included and those who developed pancreatic cancer during follow-up were excluded. Multivariable Cox proportional hazards models were used to estimate rate ratios.
In total, 84 601 individuals, aged 46-84 years, were followed for a median of 10 years, of whom 513 developed acute pancreatitis. There was a dose–response association between the amount of spirits consumed on a single occasion and the risk of acute pancreatitis. After multivariable adjustments, there was a 52 per cent (risk ratio 1·52, 95 per cent confidence interval 1·12 to 2·06) increased risk of acute pancreatitis for every increment of five standard drinks of spirits consumed on a single occasion. The association weakened slightly when those with gallstone-related pancreatitis were excluded. There was no association between consumption of wine or beer, frequency of alcoholic beverage consumption including spirits, or average total monthly consumption of alcohol (ethanol) and the risk of acute pancreatitis.
The risk of acute pancreatitis was associated with the amount of spirits consumed on a single occasion but not with wine or beer consumption. Copyright © 2011 British Journal of Surgery Society Ltd. Published by John Wiley & Sons, Ltd.
Alongside gallstone disease, alcohol abuse is associated with an increased risk of acute pancreatitis1. The pathophysiological role of alcohol in the aetiology and occurrence of acute pancreatitis is complex, but increased oxidative stress2, 3, disruption of cytosolic calcium homeostasis4 and changes in gene expression5 in the pancreas seem to be involved. Yet, only 1–3 per cent of heavy alcohol drinkers (4–5 standard drinks of alcohol per day) develop acute pancreatitis after 10–20 years of follow-up6, 7.
For many years there has been an ongoing discussion on whether the type of alcoholic beverage might influence the occurrence of acute pancreatitis8. Indeed, a potential role for type of beverage was indicated by descriptive data from Stockholm County in Sweden showing a decline in the incidence of acute pancreatitis alongside a decline in the sales of spirits between 1971 and 1987, despite increased sales of beer and wine9. In Finland, there was also a decline in the incidence of alcoholic acute pancreatitis between 1989 and 200710. During the same period, the percentage of people drinking spirits each week dropped from 24 per cent in 1988 to 19 per cent in 200710. However, clinical studies have generally been too small to study the association between different alcoholic beverages and the risk of acute pancreatitis11–13. In a recent population-based study from Denmark the risk of pancreatitis was found to be associated with the amount of beer consumed6. However, the information on alcohol use was limited to the consumption of alcoholic beverages assessed as total number of drinks per week, not including the amount of alcohol consumed on a single occasion or overall drinking frequency.
To examine the association between types of alcoholic beverage and the risk of acute pancreatitis specifically, a large population-based cohort enrolled in a prospectively collected database, including information on alcohol consumption among middle-aged and elderly men and women, was used to investigate the possible association between type of alcoholic beverage, the amount consumed on a single occasion, frequency of drinking, total alcohol intake and the risk of developing acute pancreatitis.
The study population consisted of participants in the Swedish Mammography Cohort (SMC)14 and the Cohort of Swedish Men (COSM)15, described in detail elsewhere. Briefly, the SMC was established between 1987 and 1990, when all 90 303 women born between 1914 and 1948 and residing in central Sweden (Västmanland and Uppsala counties) received a mailed questionnaire on diet, bodyweight and education. A total of 66 651 women (73·8 per cent) returned the questionnaire. In the autumn of 1997, all 56 030 surviving participants received a new expanded questionnaire that included approximately 350 items concerning diet, consumption of alcoholic beverages and other lifestyle factors, including smoking status and medical history of diabetes mellitus; additional information about history was obtained by computerized linkage of the study population with the Swedish Patient Register16. Some 38 988 women returned a completed questionnaire. The COSM was initiated in the autumn of 1997, when all 100 303 men born between 1918 and 1952, and residing in central Sweden (Västmanland and Örebro counties) received a mailed questionnaire that was identical (except for some sex-specific questions) to the SMC 1997 questionnaire; 48 645 men (48·5 per cent) answered the questionnaire.
The cohorts were representative of the Swedish population in terms of age distribution, relative bodyweight, educational level17, 18 and incidence of acute pancreatitis. For example, in 1998 the age-adjusted incidence (per 100 000 persons) in Sweden19 and in this cohort were 66 versus 64 for men, and 45 versus 46 for women, respectively. The study was approved by the Regional Ethics Board at Karolinska Institute.
Study population: patients with acute pancreatitis and follow-up
The SMC and COSM cohorts were followed up between 1 January 1998 and 31 December 2007. The Swedish Patient Register was used, through linkage of the cohorts, to identify all individual discharges after admission for acute pancreatitis (K85 in International Classification of Diseases, 10th revision). Gallstone-related pancreatitis was identified from surgical procedures for gallstone-related disease such as cholecystectomy (JKA20, JAK21), endoscopic retrograde cholangiography (JKE00, JKE02, JKE12, JKE18) or percutaneous transhepatic cholecystostomy (JKB30) within 3 months after the diagnosis of acute pancreatitis using the Swedish Classification of Operations and Major Procedures16. The majority of the remaining patients had K85.9 (unspecified acute pancreatitis) as the discharge diagnosis code, which made it difficult to assign them to one specific category of pancreatitis. However, because alcohol-related pancreatitis, alongside gallstone-related pancreatitis, constitutes by far the largest aetiological category of acute pancreatitis, it was felt reasonable to assume that the vast majority of this group comprised patients with alcohol-related disease. Therefore, all these other patients were classified as having acute alcohol-related or idiopathic pancreatitis. None of the individuals with acute pancreatitis had a history of chronic pancreatitis. Information on dates of death was obtained from the Swedish Death Register at Statistics Sweden, and information on cancer diagnoses in the cohorts from the Swedish Cancer Register.
Individuals with a previous diagnosis of cancer before 1 January 1998 (except non-melanoma skin cancer), and participants who returned an incomplete questionnaire or had an erroneous or missing National Registration Number were excluded from the databases, and were not followed up. In addition, those diagnosed with pancreatic cancer after inclusion in the cohorts were excluded.
Assessment of alcohol intake
The consumption of foods and beverages was assessed in 1997 as part of a 96-item food-frequency questionnaire. For each beverage, participants reported their average frequency of consumption over the past year. There were nine predefined response categories (from never to at least 3 times per day). The amount of alcohol consumed on a single occasion was expressed in centilitres (cl). This information was transformed to standard glasses of alcohol (12 g ethanol in each drink): 15 cl wine, 33 cl strong beer and 4 cl spirits. To estimate average alcohol intake from the questionnaire, the reported frequency of consumption of each alcoholic beverage was multiplied by self-reported quantities of the specific drinks (in centilitres; open question)20.
The validity of information on alcohol consumption has been described in detail elsewhere21. Briefly, 248 men, aged 40–74 years, were selected randomly from the Swedish Population Register. They were interviewed by telephone about once a month during a year about their diet, including alcohol drinking, using a 24-h recall method. Each person participated on average in 14 interviews. The Spearman correlation coefficient between the questionnaire-based estimates of alcohol intake and 24-h recall interviews was 0·81. In addition, in an unpublished validation study of the specific alcoholic beverages, 128 women were asked to record their dietary and alcohol intake during a week, four times in 1 year. The Spearman correlation coefficients between the questionnaire-based estimates of the consumption of different alcoholic beverages and the average of four 1-week records were 0·62 for strong beer, 0·89 for wine and 0·72 for spirits.
Age-standardized baseline characteristics of the cohort were presented as tertiles of beer, wine, spirits and total alcohol (ethanol) consumption. Person-years were calculated from 1 January 1998 until the date of death, moving out of the study area, admission to hospital with a discharge diagnosis code of acute pancreatitis or until 31 December 2007, whichever occurred first. A Cox proportional hazard regression model was used to estimate the incidence rate ratio (RR) and 95 per cent confidence interval (c.i.) of the association between the amount of strong beer (continuous, cl), wine (continuous, cl) and spirits (continuous, cl) consumed on a single occasion, the total amount of alcohol (ethanol) intake per month (continuous, cl) and the risk of acute pancreatitis. The choice of co-variables included in the model was based on clinically relevant factors and was decided on before conducting the analysis. The multivariable model was adjusted for age (continuous), sex, educational level (primary school, high school, university), diabetes, smoking (never, previous smoker, current smoker), and fruit (continuous, items per day) and vegetable (continuous, items per day) consumption. The estimates for all alcohol variables were adjusted mutually. To account for drinking behaviour, the results were further adjusted for frequency of consumption of different alcoholic beverages. To investigate potential departures from linearity for the alcohol-related variables, restricted cubic splines (3 splines corresponding to 4 knots at fixed percentiles 5, 35, 65 and 95 of the distribution) were used. A P value for non-linearity was obtained by a Wald test type of the null hypothesis that the coefficients of the second and third spline variables were simultaneously equal to zero22. The analyses were also performed by excluding patients with gallstone-related pancreatitis to evaluate any change in the trend of association between different alcoholic beverages and acute pancreatitis.
Whether the proportional hazard assumption was reasonable in the multivariable models was also checked. Scaled Schoenfeld's residuals were regressed against survival time. There was no evidence of departure from the assumption.
The role of sex and age (less than 65 years, at least 65 years) was further examined as a potential effect modifier of the relationship between alcohol-related variables and acute pancreatitis by testing the statistical significance of the interaction terms by means of the Wald test. As drinking behaviour at baseline could have been secondary to preclinical or chronic illness or pancreatic disease, sensitivity analyses were also performed by excluding the first 3 years of follow-up.
The proportion of missing values for beer, wine, spirits and total alcohol were 33·7, 30·9, 36·7 and 19·1 per cent respectively. A total of 48·1 per cent of the participants had complete data on alcohol variables, 20·1 per cent had one missing value, 11·5 per cent had two missing values, and the remaining 20·3 per cent had three to four missing values. The proportion of incomplete data was less than 5 per cent for the remaining co-variable data. A complete-subjects approach reduced the analytical cohort to 38 923 individuals and 193 incidence patients. Therefore, to evaluate the impact of missing values on the observed RRs, multivariable imputation by chained equations (MICE) was used to obtain five imputed data sets of the analytical cohort23. The RRs of the imputed data sets from the MICE procedure were pooled together by using Rubin's rule in order to obtain valid statistical inferences24.
The final cohort consisted of 84 601 participants (38 820 women and 45 781 men), aged 46–84 years in 1998, after exclusion of 293 individuals with an incident diagnosis of pancreatic cancer (Fig.1). During a median follow-up of 10 years (789 937 person-years), 513 patients with acute pancreatitis were identified. Two hundred and eighty-five (55·6 per cent) had alcohol-related or idiopathic pancreatitis, of whom 187 (65·6 per cent) were men and 98 (34·4 per cent) women. The median time to acute pancreatitis was 5·1 years. Of the remaining 228 patients (44·4 per cent) with gallstone-related pancreatitis, 110 (48·2 per cent) were men and 118 (51·8 per cent) women. The mean(s.d.) age of those with acute pancreatitis was 64·0(9·4) years compared with 61·1(9·6) years for those who did not develop the disease.
The age-standardized distribution of baseline characteristics according to the number of standard drinks of wine, beer and spirits consumed on a single occasion, and the total alcohol intake per month is shown in Table1. Those with the highest consumption of alcohol on a single occasion, irrespective of the type of alcoholic beverage, were more likely to be men and were younger. Diabetes was most common among those with the highest consumption of spirits on a single occasion (8·8 per cent versus 5·5 per cent of those with the lowest intake). Never-smokers were less likely to drink the largest quantities of beer and spirits than current and previous smokers. The intake of beer and hard liquor was inversely associated with the level of education, and also with the consumption of fruit and vegetables.
|Consumption on each occasion (cl)|
|Wine*||Beer†||Spirits‡||Total alcohol (g/month)§|
|< 21||21–35||> 35||< 34||34–50||> 51||< 7||7–10||> 10||< 126||126–352||> 352|
|No. of participants (×103)||20·3||21·7||16·4||32·3||16·2||7·6||22·1||16·1||15·4||22·8||22·8||22·8|
|Median no. of standard drinks on a single occasion¶||1·0||2·0||2·7||1·0||1·5||3·0||1·2||2·5||5·0||61#||227#||786#|
|Mean age (years)||63||59||56||62||57||53||60||60||58||60||60||58|
|Sex (% men)||48·9||51·0||65·7||49·0||81·6||85·5||44·3||68·4||87·5||41·4||56·6||79·7|
|Fruit (per day)||1·8||1·8||1·6||1·8||1·5||1·4||1·9||1·6||1·4||1·8||1·8||1·6|
|Vegetables (per day)||3·0||3·0||2·9||3·0||2·7||2·5||3·2||2·8||2·4||2·9||2·9||2·8|
|Tobacco smoker (%)|
|Educational level (%)|
The multivariable adjusted RRs for the risk of acute pancreatitis with respect to wine, beer and spirit consumption on a single occasion, and total monthly alcohol consumption are shown in Table2. When mutually adjusted for the number of standard drinks of beer and wine, and monthly consumption of alcohol (in grams) and other confounding factors, there was a 52 per cent (RR 1·52, 95 per cent c.i. 1·12 to 2·06) significantly increased risk of acute pancreatitis for every increase by five standard drinks in the amount of spirits consumed on a single occasion (Fig.2). There was no association between the risk of acute pancreatitis and the number of standard drinks of wine or beer consumed on each occasion. Moreover, the average amount of alcohol consumed per month did not affect this risk.
|All patients||Excluding gallstone-related pancreatitis|
|Age and sex*||Multivariable model 1†||Multivariable model 2‡||Multivariable model 1†|
|Wine||0·85 (0·42, 1·74)||0·86 (0·42, 1·75)||0·88 (0·40, 1·93)||0·88 (0·36, 2·15)|
|Beer||0·93 (0·48, 1·80)||0·92 (0·47, 1·79)||0·92 (0·47, 1·81)||1·15 (0·56, 2·39)|
|Spirits||1·54 (1·15, 2·07)||1·52 (1·12, 2·06)||1·59 (1·16, 2·19)||1·39 (0·95, 2·05)|
The estimates for wine, beer or average amount of alcohol consumed per month did not change when patients with gallstone disease were excluded. The association between spirit intake on a single occasion and the risk of acute pancreatitis weakened only slightly (RR 1·39, 0·95 to 2·05) (Fig.2). Further adjustment for the frequency of consumption of different beverages was performed, but none of the frequency variables was significant (all P > 0·400) and they had negligible influence on the estimates for amount of alcohol consumed. For instance, every increment of five standard drinks of spirits conferred a RR of 1·59 (1·16 to 2·19). There was no evidence of departure from linearity in any of the regression models (all P > 0·150).
There was no evidence of non-linearity throughout the analyses for variables analysed as continuous. Furthermore, there was no evidence of significant interactions between alcohol-related variables and sex (P for interaction = 0·930 for beer; P for interaction = 0·471 for wine; P for interaction = 0·878 for spirits; P for interaction = 0·402 for total alcohol) and age (P for interaction = 0·756 for beer; P for interaction = 0·716 for wine; P for interaction = 0·126 for spirits; P for interaction = 0·285 for total alcohol) in predicting acute pancreatitis. Excluding the first 3 years of follow-up had a negligible influence on the observed associations. For instance, the RR for acute pancreatitis associated with every increment of five standard drinks of spirits consumed on a single occasion was 1·43 (1·01 to 2·02).
The magnitude and direction of the RRs based on the complete data and multiple imputation analyses were basically identical. The risk of acute pancreatitis increased by 49 per cent (RR 1·49, 1·19 to 1·88) for every increase by five standard drinks in the consumption of spirits after adjustment for confounding factors, and the corresponding risk for acute pancreatitis excluding patients with gallstone disease was increased by 58 per cent (RR 1·58, 1·19 to 2·11).
As extreme hypertriglyceridaemia is a risk factor for the development of acute pancreatitis, adjustments were made for body mass index, a proxy for high blood triglyceride levels. The results were, however, not substantially changed. For instance, for every increment of 5 units in the consumption of spirits on each occasion, the adjusted risk was increased by 49 per cent (RR 1·49, 1·10 to 2·03).
To determine whether baseline drinking behaviour had changed over follow up, the results were stratified by median follow-up period. However, the risk estimates for spirits did not change when the data were analysed according to the timing of acute pancreatitis; for every increment of five standard drinks during the first 5 years of follow-up the RR was 1·45 (0·89 to 2·37), and during the second part of follow-up (least 5 years or more) it was 1·55 (1·05 to 2·28).
In this large prospective cohort study, there was a dose–response association between the amount of spirits consumed on a single occasion and the risk of acute pancreatitis. Neither the amount of wine nor the amount of beer consumed on a single occasion was associated with the risk of acute pancreatitis. Furthermore, there was no association between the risk of acute pancreatitis and the consumption frequency of the different alcoholic beverages or the average amount of alcohol consumed per month. It is important to note that the majority of the present cohort consisted of individuals with drinking behaviour within allowed ranges, that is a maximum intake of one to two standard glasses of alcoholic beverage per day25. Thus even within these ranges the consumption of spirits increased the risk of acute pancreatitis.
The metabolism of alcohol (ethanol) is known to induce oxidative stress, which in turn depletes cellular glutathione storage2, 26 and results in lipid peroxidation and damage to pancreatic tissue2, 26. However, in animal models it seems that ethanol alone is not enough to induce acute pancreatitis27, 28. Beer29 and wine30 include polyphenols with antioxidant capabilities. In experimental studies, other constituents in spirits such as long-chain alcohols have been shown to be more potent than ethanol in inducing oxidative stress31. Comparing the same amounts of alcohol, spirits deplete the antioxidant capacities more readily than beer or wine32. Thus, there might be other constituents in spirits that induce acute pancreatitis, in combination with ethanol or alone. Those drinking spirits might also have lower reserves of antioxidants at baseline, which could be depleted rapidly after intake of spirits33. The present analyses were adjusted for the level of fruit and vegetable consumption, which is a valid indicator of antioxidant status34, but the results were not changed.
The trend in the association between the consumption of spirits and acute pancreatitis did not change after excluding patients with gallstone disease. Patients with acute pancreatitis normally undergo abdominal ultrasonography as part of their clinical diagnostic evaluation. Thus, in some patients the detection of gallstones might mask an underlying unhealthy alcohol drinking behaviour.
A limitation of this study was that information on alcohol consumption was self-reported. There could have been some underreporting among those with high levels of alcohol consumption. However, such misclassification would have biased the results towards the null association. Another concern was the presence of missing data, in particular with respect to alcohol. It has been shown that complete data analysis in the proportional hazard regression model may lead to substantial bias estimates when missing co-variable data are dependent on the outcome of both the outcome and exposure35. Yet, there were small differences between complete data and multiple imputation analyses in this study, suggesting that any missing data were most likely to be missing at random.
There is always a possibility that the people who accept participation in a cohort study are more motivated and might have a healthier lifestyle. However, the cohorts were representative of the general population in terms of age, education and bodyweight. Moreover, non-respondents to the alcohol-related questions in this study might have had greater alcohol consumption than the respondents. Yet, among those with missing alcohol-related information, 0·2 per cent had gallstone-related acute pancreatitis and 0·3 per cent had alcohol-related or idiopathic pancreatitis. The corresponding figures for those with complete alcohol information were 0·3 and 0·4 per cent. Thus, it seems reasonable to assume that the results of this study could be generalized to a population with non-excessive alcohol consumption. The applicability of the findings to individuals with excessive alcohol consumption (more than 60 g alcohol per day) needs further investigation.
Finally, although the study controlled for known risk factors for acute pancreatitis, the possibility cannot be ruled out that the present findings were affected by residual confounding. For instance hypercalcaemia, extreme levels of serum triglycerides, cystic fibrosis and many others have been described as risk factors for acute pancreatitis36. However, the first two co-morbidities explain less than 1 per cent of all cases of acute pancreatitis, and cystic fibrosis should not influence the observed associations in this study population in view of the mean age. One proxy for high triglyceride levels may be obesity. However, the results did not change markedly after adjustment for this factor.
Although the exact mechanism by which spirits increase the risk of acute pancreatitis needs to be explored, more focus should be given to constituents other than ethanol. If the present findings are confirmed in future studies, the consumption of spirits in those susceptible to acute pancreatitis should be avoided.
This work was supported by research grants from the Swedish Cancer Foundation and the Swedish Research Council Committee for Infrastructure. O.S.A. was supported as a postdoctoral researcher by a grant from Olle Engkvist Byggmästare Foundation. The funding sources had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; or preparation, review or approval of the manuscript. The authors declare no conflict of interest.
- 25Implementing American Heart Association pediatric and adult nutrition guidelines: a scientific statement from the American Heart Association Nutrition Committee of the Council on Nutrition, Physical Activity and Metabolism, Council on Cardiovascular Disease in the Young, Council on Arteriosclerosis, Thrombosis and Vascular Biology, Council on Cardiovascular Nursing, Council on Epidemiology and Prevention, and Council for High Blood Pressure Research. Circulation 2009; 119: 1161–1175., , , , , et al.