Alcohol Consumption, Cigarette Smoking and the Risk of Recurrent Acute and Chronic Pancreatitis

Introduction:

Chronic Pancreatitis (CP) is an inflammatory syndrome of the pancreas characterized by progressive parenchymal fibrosis, maldigestion, diabetes mellitus and pain¹. Excessive alcohol consumption has been identified as the primary etiologic factor in numerous studies of adult CP^2–8, although fewer than 5% of heavy drinkers develop CP^{9, 10}. In 1978, a landmark South African study observed a dose-dependent risk of CP from alcohol consumption¹¹. However, threshold levels and drinking patterns were not evaluated and the study only included men. Other studies have suffered from similar limitations⁸. Cigarette smoking has been reported to be an independent risk factor for developing CP^{8, 12–14} and increases its rate of progression^{15, 16}. Recently, genetic factors have been identified which also increase the risk of pancreatitis, independent of alcohol consumption or cigarette smoking^17–21. No previous study has evaluated the role of alcohol and smoking in both men and women in recurrent acute pancreatitis (RAP) and CP. Additionally, the current prevalence of alcohol consumption and smoking and their associated risk with RAP and CP in the United States (US) is unknown.

We hypothesized that the etiology of RAP and CP is complex, and that the effects of alcohol and smoking may be limited to specific patient subsets. To further characterize the risk of RAP and CP in US men and women, we prospectively ascertained 1000 pancreatitis patients and 695 controls from 20 centers with expertise in pancreatic diseases and performed adjusted analyses of alcohol consumption and cigarette smoking.

Methods:

Statistics:

Descriptive analyses are presented as proportions for categorical data and as mean ± standard deviation (SD) or median and interquartile range (IQR) for continuous data. Comparisons of continuous variables between controls and pancreatitis patients were performed using one-way analysis of variance (ANOVA) and Kruskal-Wallis test where applicable. If significant results were observed, appropriate two group comparisons were performed using student t-test or Mann-Whitney U test. Categorical data were compared using the chi-squared test. The distribution of subjects in different drinking categories and amount of smoking was compared using the Cochran-Armitage test. The correlation between the drinking categories and TWEAK scores was tested using the Spearman correlation coefficient.

Multivariable logistic regression was used to assess associations of alcohol consumption and cigarette smoking with pancreatitis after controlling for potential confounding variables (age, gender, and current or maximum body mass index [BMI]). Data was evaluated for collinearity and interactions. For CP, we also performed stratified regression analyses for males and females; abstainers and ever drinkers; never and ever smokers; whites and blacks. Age was assessed as a continuous variable; all other variables in the models were categorical variables, including alcohol (drinking categories), smoking (never or ever; never, <1, ≥1 ppd; never, <12, 12–35, >35 pack years), BMI (≤25, >25-≤30, >30) and gender. We used abstainers and light drinkers as the reference category for alcohol consumption. Regression models were evaluated by the goodness of fit χ² test. Two-sided p-values <0.05 were considered significant. Data analysis was performed using the R Project software (www.r-project.org) and SPSS version 16 (SPSS Inc., Chicago, Ill., USA).

Results:

Overall, the mean age for the NAPS2 cohort was 49.7±15.4 yrs, 87.5% were White and 56.5% were female. Control and CP patients were older than RAP patients, and a greater proportion of controls were female (Table 1). Drinking behavior differed by gender and pancreatitis type. Heavy or very heavy drinking was reported by a much greater proportion of CP patients than controls or RAP patients. While approximately half of males with CP were heavy or very heavy drinkers, close to two-thirds of females with CP were self-reported abstainers or light drinkers. The demographics and distribution of drinking and smoking categories were generally similar across the top ten centers that recruited over 80% of all participants.

Figures 1a–1c illustrates the distribution of drinking categories for controls and pancreatitis patients. Alcohol exposure (average drinks/day [Figure 2a], duration and amount of alcohol consumed during the maximum lifetime drinking period), the proportion of lifetime drinking spent at higher drinking levels (i.e. levels 5 or 6) increased from light to very heavy drinkers for both controls and pancreatitis patients.

Very heavy drinkers formed a distinct subgroup of controls and pancreatitis patients. Alcohol exposure during the maximum lifetime drinking period in this subgroup was much higher than among the other drinking categories. These individuals also spent a higher proportion of their life engaging in heavier or binge drinking. An overlap for alcohol exposure with very heavy drinkers was observed for 10–20% of moderate and 15–30% of heavy drinkers.

Overall, the lifetime drinking duration for controls (median 29 years; IQR 17, 38) and CP patients (median 27; IQR 17, 37) was higher than RAP patients (median 22, IQR 9, 32). Binge drinking during any life period was reported by 4.6% of controls, 7.3% of RAP and 14.8% of CP patients who provided responses to questions on drinking categories. Among pancreatitis patients, the majority of binge drinkers were either heavy (30% of RAP, 8% of CP) or very heavy (65% of RAP, 88% of CP) drinkers. The proportion of time spent binge drinking by very heavy drinkers increased from controls to RAP and CP. At-risk drinking was more common for CP patients than RAP patients or controls, among both men and women (Table 1). A strong correlation was observed between TWEAK score and drinking categories for the entire cohort (r = 0.69, p<0.001), as well as among the individual groups (Figure 2b).

Cigarette smoking was more prevalent among CP patients compared to controls and RAP patients (Table 1). The prevalence and amount of smoking (Figure 3a, b) increased linearly with the level of drinking in controls and pancreatitis patients. Almost two-thirds of very heavy drinkers and 51.6% heavy drinkers with CP reported smoking ≥1 ppd compared to 13.9% abstainers and 34.3% light drinkers. The amount of smoking was significantly higher in CP patients (26.6 pack years; IQR 12, 46) compared to RAP (19.5; IQR 7.9, 36.1, p=0.001) and controls (16.2; IQR 6, 32.7). CP patients also had significantly longer smoking duration (median 30.5 years; IQR 19.7, 39) than controls (21.9; IQR 11.4, 33.1; p<0.001) and RAP patients (22.7; IQR 12, 32; p<0.001). Duration of smoking for CP patients was significantly longer than controls for all drinking levels except abstainers and light drinkers, and from RAP patients who were light or heavy drinkers.

Discussion:

In the largest U.S. study of risk factors for RAP and CP to date, we observed a smaller than expected relationship of alcohol consumption and pancreatitis. We confirmed the results of past studies that have noted a significant association of very heavy alcohol consumption and cigarette smoking with CP. Our study is the first to demonstrate that a significant increase in the risk of CP occurs only above a threshold of ≥5 alcohol drinks per day. Contrary to previous studies^{2, 4–8}, only 38.4% of males and 11% of females with CP were very heavy drinkers. Although heavy smokers tended to be heavy drinkers, smoking itself was a significant risk factor for pancreatitis.

Accurate quantification of alcohol consumption and its relationship to disease risk is a challenge in any observational study. We used drinking category definitions similar to the US general population survey (NHIS)²³ and the recommendations of other government agencies²⁶, although we divided the NHIS heavier drinking category into heavy and very heavy to perform more precise comparisons. Self reported alcohol intake during the maximum lifetime drinking period served as primary data to create drinking categories. This criterion was chosen for subject classification, rather than total lifetime alcohol consumption, to reflect the time of highest alcohol-associated risk. Furthermore, use of lifetime alcohol consumption criteria would classify older individuals with small amounts of daily alcohol consumption over many years in the same risk category as younger individuals who consumed alcohol more heavily for a few years (e.g. 100 kg alcohol can be consumed in 11.6 years at 2 drinks/day or in only 4.6 years at 5 drinks/day).

During the maximum lifetime drinking period, only about one-fourth of CP patients reported alcohol consumption in amounts typically associated with CP. To avoid underestimation of “heavy drinking”, we analyzed our data using several cut-offs. Including subjects in the moderate and heavy drinking categories whose alcohol exposure during the maximum lifetime drinking period overlaps with very heavy drinkers increases the proportion of very heavy drinkers to 9.8% of controls, 13.8% of RAP and 31.7% of CP patients. Using a lower threshold for heavy drinking (i.e. >14 drinks/week for males and >7 drinks/week for females), the proportion of heavy or very heavy drinkers among CP patients would have been 37.9% (50.5% males, 23.9% females). Using presence of either an overlap of alcohol exposure during the maximum lifetime drinking period or At-Risk or Binge drinking, the proportion of very heavy drinkers increases to 17.4% among controls, 21.7% among RAP patients and 40.2% (58.3% males, 20.2% females) among participants with CP. Using only average number of drinks on drinking days and omitting frequency increases the proportion of subjects drinking ≥5 drinks on a drinking day to 41.6% (57.2% males, 19.7% females) for CP patients. However, such criteria will classify an infrequent drinker who consumes ≥5 drinks on a drinking day also as a “very heavy drinker”, and classify a large proportion of controls (24.1% overall, 39.1% males, 15.5% females) as very heavy drinkers. Regardless of classification, the proportion of pancreatitis that can be attributed to heavy alcohol intake in our study is much lower than reported in many studies of CP^{2, 4–8}.

Our observation of a lower prevalence of heavy drinking among CP patients is not completely unexpected. Most of the previous large studies of CP evaluated subjects between 1960s-80s (compared to 2000–2006 for NAPS2) when alcohol was the predominate recognized etiological factor for CP^{2, 5–7}. There is recent, growing knowledge of other causes of CP, including genetic mutations in PRSS1, CFTR, SPINK1 and chymotrypsin C genes^17–21. Therefore, subjects without an obvious etiology might be more likely to be referred to expert centers for evaluation (referral bias). Our overall lower rates of heavy drinking may also have been attributed to our study’s greater proportion of females compared to other studies^{2, 4–8}. Lastly, NAPS2 used imaging evidence on ERCP or CT as the primary enrollment criteria, while many previous studies of CP were conducted prior to routine, clinical use of CT or MRI, when diagnosis relied on less sensitive methodology, such as the presence of steatorrhea, diabetes mellitus and pancreatic calcifications on abdominal x-ray - which are commonly observed in alcoholic men who are heavy smokers.

We found the threshold for association between alcohol use and CP to be ≥5 drinks/day. The wide spectrum of reported smoking and drinking habits in this US population allowed us to evaluate the role of alcohol and smoking in stratified analyses. Although ours is the largest study of its kind to date, the current study was not adequately powered for analysis of all important subgroups (e.g.: small proportion of - very drinkers among females, very heavy drinkers among non-smokers, non-smokers among very heavy drinkers, small number of Blacks). However, our results provide important trends that should be addressed in future studies. Although there were a limited number of Blacks in the NAPS2 study, the previously reported association between alcohol and CP in Blacks was identified, suggesting racial differences in susceptibility to alcoholic pancreatitis^{14, 27}.

Our results confirm the association between smoking and CP^{8, 12, 16, 28}, and demonstrated that this effect is dose-dependent. A lack of significance for the interaction term between drinking and smoking was likely due to coexistent heavy drinking and smoking habits, and a low prevalence of very heavy drinking among non-smokers and heavy smoking at lower levels of drinking. However, a trend for synergist effect of increasing tobacco plus alcohol use observed in this study is expected based on several observations. In our study, the primary diagnostic criteria for CP was the presence of pancreatic fibrosis. Although the biology of fibrosis is complex, it is now clear that pancreatic stellate cell is responsible for deposition of most of the extracellular fibrotic proteins, including collagen I and II, fibronectin and other matrix proteins^29–31. Stellate cells are transformed from inactive to active states and proliferation in the context of injury, free radicals, proinflammatory cytokines, and other factors. The deposition of matrix proteins is driven by additional factors including the anti-inflammatory cytokine transforming growth factor ß1 (TGF-ß1) and various growth factors^29–32. Both alcohol and components of tobacco (e.g. carbon monoxide) are known to have significant effects on the immune system^{33, 34}, but the synergistic mechanisms that lead to fibrosis in CP in humans have not yet been determined. An inverse association was observed between BMI and CP. Plausible hypotheses for this association include weight loss from malabsorption or fear of eating secondary to pain^{5, 35}; additionally, CP patients were often smokers, and cigarette smoking is associated with lower weight and BMI³⁶.

Potential limitations for generalizability of our results could be the choice of controls, under-reporting of alcohol consumption by study subjects, misclassification of subjects into drinking categories, higher proportion of females, and participation of expert centers for patient enrollment. The choice of spouses, relatives or friends as controls who may have shared drinking and smoking habits with affected participants could have introduced a conservative bias and lowered the odds ratios for alcohol use and smoking with pancreatitis. The prevalence of heavy or very heavy drinkers among controls was higher than in the general population²³. However, this finding was likely due to our use of a different reference period for creation of drinking categories (“maximum lifetime drinking period” rather than “past 12 months”)²³. Analysis of current drinking in our controls demonstrated that the distribution of heavier drinking was generally similar to the US general population²³.

Self-reported alcohol consumption has been shown to be a reliable and valid measure of an individual’s alcohol consumption when compared to other measures (in-person interviews, collateral reports/sources) in research settings^{37, 38}. The validity of drinking category assignments in our study was confirmed by the significant correlation between the drinking categories with self-reported alcohol exposure, drinking patterns and answers to the TWEAK questionnaire^{24, 25}. The sensitivity of detecting At-Risk drinking in CP patients in our study (using very heavy drinking as the reference) was slightly lower (87%) and specificity higher (86%) compared to previously reported results of studies of known alcoholics²⁴.

Although our study included a higher proportion of females (47.6% of CP), 35% of patients in the largest US study on idiopathic and alcoholic CP were females⁵. We performed stratified analyses for the prevalence and associations of alcohol and smoking. Therefore, a lower prevalence of very heavy drinking in our cohort cannot be explained solely on the basis of a higher proportion of females in the NAPS2 cohort. Our study population was enrolled from secondary or tertiary centers with expertise in pancreatic diseases. A lower proportion of very heavy drinkers and a higher proportion of females in our study highlight the need for future studies to determine the current distribution of demographics and risk factors in the US at a population level.

While our study confirms very heavy alcohol consumption and cigarette smoking as independent predictors of CP, it also demonstrates that the great majority of CP subjects seen at US referral centers are not “alcoholics” and that many, in fact, are abstainers. Our results indicate that patients with CP can be stratified into three distinct categories: those with no or minimal alcohol intake where alcohol is unrelated to disease, very heavy drinkers where alcohol and smoking may play a dominant role, and moderate or heavy drinkers where alcohol could be an etiologic cofactor with smoking, other environmental or genetic variables.

In conclusion, only very heavy alcohol consumption and cigarette smoking are significant independent risk factors for CP. Risk for CP from alcohol consumption occurs above a threshold level while risk due to smoking is dose-dependent. Drinking levels in subjects with RAP are similar to controls. Only a minority of patients with RAP and CP patients currently seen at secondary or tertiary US centers could be categorized as very heavy drinkers.