|Home | About | Journals | Submit | Contact Us | Français|
Assess the relationship between alcoholic etiology, tobacco use and severe acute pancreatitis (SAP).
Smoking and alcohol exposure were recorded upon admission in a cohort of acute pancreatitis patients within the United States (U.S.). Patients with first, “sentinel” attack of acute pancreatitis (AP) were identified for analysis.
Associations between alcohol, smoking and SAP were validated in an independent cohort of patients from Spain.
U.S. cohort (n=222): Thirty-five% developed organ failure (OF), 35% Pancreatic Necrosis (PNec), and 7% died. OF (54% vs. 33%, p=0.03), PNec (62% vs. 31%, p=0.006), intensive care (ICU) admission (58% vs. 36%, p=0.03) and length of stay (LOS) (20 vs. 8 days, p= 0.007) were greater in alcoholic when compared to other etiologies.
Spanish cohort (n=366): Similar differences in outcomes were also found with between alcoholic and non-alcoholic etiologies: OF (24% vs. 8%, p=0.001), PNec (38% vs. 14%, p<0.001), ICU admission (20% vs. 3%, p<0.001), and LOS (17 vs. 11 days, p=0.04).
Multivariable analysis confirmed alcoholic etiology to be independently associated with OF and PNec in both cohorts.
Alcoholic etiology is independently associated with OF and PNec in patients with sentinel AP and is important when evaluating risk for severe disease in AP.
Acute pancreatitis (AP) is the most common gastrointestinal diagnosis for inpatient hospitalization in the United States. 1 While majority of patients have an uncomplicated course and overall mortality is less than 5%; 15-20% of patients suffer a “severe” attack. Severe acute pancreatitis (SAP) is defined by the presence of organ failure (OF) and/or pancreatic necrosis (PNec); with mortality in SAP approaching up to 30%.2-5 Determination of risk factors for SAP is an area of intense research.
Alcohol is a known cause of AP, an independent risk factor for recurrent acute pancreatitis (RAP) and is associated with progression to chronic pancreatitis (CP).6,7 The results of an earlier prospective study by our group demonstrated that patients with heavy alcohol consumption were more likely to develop PNec in the setting of AP by any etiology.8 These results suggested that alcohol abuse may play a role in the natural history of SAP. However, the impact of alcohol consumption as a unique risk factor for SAP remains incompletely described in further retrospective studies.15,16
Cigarette smoking is also associated with a higher risk of developing AP, recurrent AP, and progression to CP in clinical studies.9-12 Studies have demonstrated smoking to be associated with elevated levels of pancreatic inflammatory cytokines, trypsinogen and the development of fibrotic parenchymal changes.13-16 While it is an exposure often paired with alcohol, smoking as a risk for acute and chronic pancreatic disease is now demonstrated to have a role that is independent of alcohol consumption. 6,17,18 However, the impact of smoking as an environmental risk factor for severity in the context of AP also remains poorly studied.
The aims of this study are to explore the association of alcoholic etiology and smoking with organ failure (OF) and pancreatic necrosis (PNec) using a tertiary training cohort of patients with sentinel AP (training cohort). Subsequently, the above findings are validated in an independent cohort from a community hospital (validating cohort).
The Severe Acute Pancreatitis Study (SAPS) at the University of Pittsburgh Medical Center (UPMC, U.S.A.) is an ongoing prospective observational study that has enrolled 344 patients with AP in four, one-year enrollment periods between June 2003 and May 2013. The diagnosis of AP was based on presence of at least two of the following three features: abdominal pain characteristic of AP, serum amylase and/or lipase at least three times the upper limit of normal, and characteristic findings of AP on abdominal computed tomography (CT) scan. All patients are enrolled within 24 hours from the time of admission or transfer to our institution(s).19 The SAPS Study protocol has been approved by the Institutional Review Board of the University of Pittsburgh. Informed consent is obtained from all patients or appropriate surrogates before study enrollment. For the present study, we included SAPS patients who were admitted with their first attack of AP (sentinel) and excluded patients who had a prior history of AP.
As previously described,20 data collection in the SAPS study included detailed information on demographic, clinical, laboratory and radiographic parameters. For patients transferred to our institution, pertinent information from outside records was obtained by contacting the transferring center.
Data regarding alcohol consumption and cigarette smoking was collected prospectively in the form of a detailed questionnaire administered to patients by the study investigators during their hospitalization. Patients were classified as never drinkers, prior drinkers (abstinent for the past six months), or active drinkers (use within the past six months). Level of active alcohol consumption was defined as “light” when patients consumed less than four drinks per month. Patients with heavy alcohol consumption were either daily users reporting more than two drinks per day for men and one drink per day for women or binge drinkers reporting multiple days per month of consuming greater than five drinks per day for men and three drinks per day for women. Finally “moderate” drinkers were those patients with daily and/or episodic patterns of use quantified to a level greater than “light”, but less than “heavy”21. Patients also identified the timing of their last drink prior to the attack of pancreatitis.
The etiology of AP was determined by the study physicians’ based on review of the prospectively collected questionnaires and the medical records. Biliary etiology was based on presence of gallstones and/or common bile duct stones on abdominal ultrasound, elevated serum alanine aminotransferase (ALT) concentration, especially more than threefold, and prior history of biliary colic. 22 Alcoholic etiology was based on active moderate or severe alcohol consumption in the absence of evidence for other common etiologies, i.e. biliary or hypertriglyceridemic.
Cigarette smoking was assessed with the same enrollment questionnaire. Patients were classified as never smokers, former smokers (abstinent for six months leading to enrollment), or active smokers. Smoking was quantified as greater than or less than 1 pack per day.
OF was defined by the Modified Marshal Score for cardiovascular, pulmonary, and/or renal dysfunction.5 OF was further sub-classified into transient (<48 hours) and persistent (≥48 hours). PNec was defined by a regional lack of parenchymal enhancement on contrast enhanced cross sectional imaging (CT or MRI scan). Other outcomes included ICU need, length of hospitalization, and in-hospital mortality. All the information on the clinical course during the index admission was collected prospectively.
To validate the association of alcoholic etiology and smoking with severe disease, we utilized an independent cohort of AP patients from Alicante, Spain. Demographic, clinical, laboratory and radiographic data for 366 consecutive patients with sentinel attack of AP admitted between 11/2008 and 11/2013 were collected prospectively.23 Definition of AP, etiologies, and outcomes were similar to the training cohort as described above. Data on alcoholic etiology and cigarette smoking were collected prospectively. Alcoholic etiology was assigned by study physician based on active moderate to severe alcohol consumption in the absence of other etiologies.
Descriptive data was represented as mean (standard deviation) or median (interquartile range; IQR) for continuous variables. Categorical variables were described using frequency (proportion). We used the chi-square or Fisher exact test for categorical data. Student's t-test or Wilcoxon rank-sum test was used for continuous variables as appropriate. The categorical outcomes for which odds ratios were calculated included any OF (transient or persistent) and PNec as assessed by contrast-enhanced imaging. Predictor variables included age (>60 years), sex, race, BMI≥30, alcoholic etiology and current smoking. Multivariable logistic regression was used to calculate adjusted odds ratios. We assessed for interactions and multi-colinearity. The Hosmer-Lemeshow goodness of fit test was used to assess whether there is evidence for lack of fit in the final logistic regression mode. All statistical analysis was conducted using STATA (StataCorp. 2011. Stata Statistical Software: Release 12. College Station, TX: StataCorp LP).
Of 344 patients prospectively enrolled in the SAPS trial, 222 patients were identified as having a sentinel (first attack) episode of AP. Their median age was 56 years (IQR 41-70), 50% were female, and 90% were Caucasian. 62% were transfers from referring institutions. The most common etiologies were biliary (48%), idiopathic (14%), and alcoholic (12%). Contrast enhanced imaging was performed in 158 patients. OF developed in 35% (78/222); PNec was identified in 35% (56/158), and death in 7% of patients (16/222; Table 1).
Forty eight % of patients were active drinkers; 30% of them reporting heavy consumption. Twenty two % of patients were active smokers, 60% of which were smoking over 1 PPD. Overall, 9% of patients reported combined, active heavy drinking and smoking. Patients reporting never having consumed alcohol and cigarettes comprised 34% and 49% of our cohort, respectively.
Among patients with alcoholic etiology (n=26, 12%), 85% classified as heavy alcohol drinkers and the remaining patients as moderate drinkers at the time of admission. Sixty-five % of these patients were also smokers. Conversely, among patients without alcoholic etiology, only 5% reported heavy alcohol use. In all these patients, there was clear evidence supporting a different etiology of AP, i.e. presence of common bile duct stones on abdominal ultrasound.
Alcoholic etiology was significantly associated with the development of OF (54% vs. 33%, p=0.03). Greater proportion of patients with alcoholic etiology required ICU admission (58% vs. 36%, p=0.03) and had a longer median hospitalization (20 vs. 8 days, p=0.007) when compared to patients with non-alcoholic etiologies (Table 1). Alcoholic etiology was also significantly associated with the development of PNec (62% vs. 31%, p=0.006). Mortality was similar between the two groups (7% vs. 7%, p=0.92). Active smoking was neither associated with OF (42% vs. 33%, p=0.3) nor with PNec (42% vs. 33%, p=0.3).
Age (>60 years), gender, race (Caucasians vs. non-Caucasians), obesity (BMI≥30 vs. BMI<30), etiology (alcohol vs. non-alcohol), and current smoking were then assessed as potential risk factors for severe disease. On univariate analysis, BMI ≥30 [odds ratio (OR) 2.3, confidence interval (CI) 1.3-4.1; p=0.003], male gender (OR 2.7, CI 1.5-4.8; p<0.001), and alcoholic etiology (OR 2.4, CI 1.1-5.5; p=0.03) were significantly associated with OF. For PNec, analyzing only patients who underwent contrast enhanced imaging (n=158), male gender (OR 2.0, CI 1.0-4.0; p=0.03), and alcoholic etiology (OR 3.6, CI 1.4-9.2; p=0.009) were found to be significant risk factors (Table 2).
Multivariate regression analysis including age > 60 years, sex, BMI≥30, and alcoholic etiology as predictors demonstrated BMI ≥ 30 (OR 2.5, CI 1.4-4.5; p=0.003), male sex (OR 2.6, CI 1.4-4.7; p=0.002), and alcoholic etiology (OR 2.6, CI 1.0-6.3; p=0.04) to be independently associated with OF. Only alcoholic etiology (OR 4.1, CI 1.5-11.3; p=0.007) was independently associated with PNec.
The median age of the 366 consecutive patients with sentinel AP in the validation cohort was 62 years (IQR 45-75) and 51% were female. The most common etiologies included biliary (69%), alcoholic (11%), and idiopathic (9%). Contrast enhanced imaging was available in 265 (72%) patients. OF was identified in 10% (36/366), PNec in 17% (46/265), and death in 3% (11/366; Table 3). Twenty-eight % of the patients were cigarette smokers at the time of presentation.
Alcoholic etiology (n=41, 11%) was significantly associated with the development of OF (24% vs. 8%, p=0.001). Greater proportion of patients with alcoholic etiology required ICU admission (20% vs. 3%, p<0.001) and had a longer median hospitalization (17 vs. 11 days, p=0.04) when compared to patients with non-alcoholic etiologies (Table 3). Alcoholic etiology was also significantly associated with the development of PNec (38% vs. 14%, p<0.001). Mortality did not reach statistical significance between the two groups (8% vs. 2%, p=0.08). Cigarette smoking was not associated with OF (14% vs. 7%, p=0.0643), but was associated with PNec (29% vs. 13%, p=0.004).
Univariate analysis was performed including age (>60 years), gender, BMI≥30, alcoholic etiology, and current smoking as predictors and OF and PNec as outcomes. BMI≥30 (OR 2.1, CI 1.1-4.3; p=0.03) and alcoholic etiology (OR 3.7, CI 1.6-8.4; p=0.002) were significantly associated with OF. For PNec, male sex (OR 2.8, CI 1.4-5.7; p=0.004), alcoholic etiology (OR 3.9, CI 1.8-8.2 p<0.001), and cigarette smoking (OR 2.7, CI 1.4-5.4; P=0.005) were found to be significant risks (Table 4).
Multivariate regression analysis using age (>60 years), sex, BMI≥30, alcoholic etiology and cigarette smoking as predictors demonstrated BMI ≥ 30 (OR 2.6; CI 1.1-5.9; p=0.03) and alcoholic etiology (OR 5.6, CI 1.7-18.3; p=0.005) to have an independent association with OF. Male sex (OR 2.5, CI 1.1-5.8;p=0.03) and alcoholic etiology (OR 3.0, CI 1.1-7.6;p=0.03) were independently associated with PNec.
In this study, we describe prospectively recorded patterns of alcohol consumption and cigarette smoking in a large cohort of patients from the U.S. with their first attack of AP. Around 50% of patients were active drinkers, with one third of them being heavy drinkers. 6 In respect to smoking, around one quarter of patients were active smokers with the majority smoking more than 1 PPD. Furthermore, we explored possible associations between alcoholic etiology and smoking with severe disease as defined by the current revised classifications, specifically by OF and PNec.5,24 We demonstrated that the odds of developing OF and PNec are significantly higher for patients with alcoholic etiology, independently of other clinical risk factors such as BMI and gender. This is a novel and important finding in disease severity stratification.
Our early, previously published work described an association of chronic alcohol use with development of PNec.8 However, this previous study was of smaller sample size and was heterogeneous from the standpoint of including patients with prior pancreatic disease. There is recent data reporting that severity of AP diminishes with subsequent episodes of AP.25,26 Therefore, by excluding patients with recurrent attacks in the current study, we identified a sentinel cohort with minimal parenchymal fibrosis, which we believe was early in the natural history of pancreatic disease. Hence by focusing on sentinel attacks, we aimed to eliminate bias and/or confounding of outcomes by preexisting pancreatic disease.
The mechanism of pancreatic injury by alcohol remains an area of scientific debate. Theories implicating alcohol in AP include direct mitochondrial injury by metabolic end products of fatty acid ethanol esters, intracellular activation of pancreatic enzymes, pancreatic duct obstruction by alterations in the lithogenicity and viscosity of pancreatic secretions, sensitization of the exocrine pancreas to secretaogues, oxidative stress within pancreatic tissue and destabilization of zymogen membranes.9,27,28 Emerging evidence indicates that genetic polymorphisms play a role in susceptibility to alcoholic AP.28-30 Of note, this research focuses on mechanisms of direct acinar toxicity.
The strength of association between alcoholic etiology and OF suggests the role of alcohol in SAP is beyond direct pancreatic toxicity. A consequent, severe systemic inflammatory response in the setting of alcoholic etiology appears to influence outcomes in AP. This is supported in the critical care literature with alcohol identified as an independent risk factor for the acquisition of ICU associated bacterial infections and ICU mortality.31,32 There is evidence that alcohol abuse may translate to increased systemic exposure to endotoxins and lipopolysaccarhides by elevating gut permeability.9,33 The extent to which alcohol withdrawal, poor nutritional status, and concomitant end-organ damage (i.e. liver, or heart) also play a role are unclear. What is evident is that sentinel AP secondary to alcoholic etiology is associated with severity, and is likely due to a robust systemic insult. This translates to higher rates of hospitalization, greater ICU need, and a prolonged length of stay. Based on our results, we believe that these patients should be singled out as an at risk group at the time of presentation for worse outcomes within the context of AP. Future basic research studies are needed to further identify pathogenic mechanisms of OF in alcoholic AP.
We also found that alcoholic etiology in AP is independently associated with development of PNec. This finding is consistent with our previous clinical study, as described above 8. Recent experimental studies provide mechanistic insight into alcohol induced necrotizing pancreatitis. Oxidative metabolites of alcohol sensitizes the pancreatic mitochondria, increasing mitochondrial permeability, and ultimately leading to mitochondrial failure and acinar cell necrosis.34 Necrotic cell death appears to activate the local immune response resulting in further pancreatic injury.
We did not find any association of alcoholic etiology with mortality in AP. This finding should, however, be interpreted with caution, owing to the small number of patients that died.
While patients with alcohol abuse tend to be heavy smokers, cigarette smoking was not found to be an independent predictor of severity. It is likely that smoking plays a larger role in mediating chronic inflammation, parenchymal fibrosis, and is synergistic with alcohol to progression to recurrent and chronic pancreatitis rather than influencing the severity of the sentinel attack.6,17,18
Limitations of this study include the fact that our training cohort was captured at a tertiary care center. As a referral center, a significant proportion of enrolled patients were transfers from other institutions. Transferred patients differ in their elevated complexity and acuity. This is reflected by the prevalence of OF and PNec (35% for both), which is higher than expected in the community setting, raising a concern for ascertainment bias. To investigate this potential limitation, we utilized a second, independent, prospective group of patients with sentinel AP as a validating cohort. It included patients from a community hospital, which serves a large geographic area in Spain, Europe. This is reflected in the lower prevalence of OF (10%) and PNec (17%). We were able to reproduce the significant findings between alcoholic etiology and OF and PNec in the validating cohort, which diminishes the concern for ascertainment bias. Furthermore, the large sample size of almost 600 patients with sentinel AP between the two cohorts is a major, unique strength of our study.
Another limitation was the absence of detailed information on the amount of alcohol consumption of patients in the Spanish cohort. However, both centers used the exact same definition for alcoholic etiology. The fact that similar percentages of alcoholic pancreatitis were found in both cohorts increases the internal validity of the study and minimized any potential cultural differences. Finally, as we report our findings within context of patients with sentinel attack of pancreatitis, we advise caution in generalizing our findings to patients with recurrent acute and/or chronic pancreatitis.
In summary, this is the first study to identify alcoholic etiology as a significant risk for severe disease. These findings were generated in a cohort derived from a tertiary care center in the U.S. and subsequently validated in an independent cohort from a community hospital in Spain, Europe. Early recognition of subjects at high risk for severe disease remains a challenging task for clinicians as they strive to realize optimal management of patients with AP. Herein, we demonstrate that alcoholic etiology represents another, if not one of the more important, considerations when evaluating these patients. Within the context of a first episode of AP, alcoholic etiology should be immediately explored for accurate early risk stratification of such patients. Consequently, this group of patients would benefit from early, aggressive medical management at presentation (adequate volume resuscitation), close clinical surveillance once admitted with a low threshold for managing these patients at a higher level of care or transfer to a tertiary center.
Grant Support: VA Merit Review Grant: Pro00000496
Disclosures: None relevant to this manuscript.
Author Contributions: Study Design (Easler, Papachristou, Yadav, Enrique de-Madaria), Data Collection (Easler, Nawaz, Moya-Hoyo, Koutroumpakis, Rey-Reveiro, Acevedo-Piedra, de-Madaria), Statistical Analysis (Easler, Nawaz, Papachristou, Yadav), Manuscript Drafting (Easler, Papachristou, Nawaz), Critical Editing (Whitcomb, Yadav, Singh, de-Madaria)