Biochemical model of logistic regression for early prediction of the etiology of acute pancreatitis.

OBJECTIVE

To define a simple model for the early prediction of the biliary or alcoholic etiology in acute pancreatitis, according to the results of several biochemical variables.

PATIENTS AND METHODS

Forty-five patients with acute pancreatitis were included in the study (33 of biliary and 12 of alcoholic etiology). Plasma levels of standard biochemical parameters (glucose, urea, albumin, calcium and C-reactive protein), liver function tests (glutamic oxalacetic transaminase, glutamic pyruvic transaminase, alkaline phosphatase, (gamma-glutamyl-transpeptidase, lactic dehydrogenase, bilirubin and bile acids) and pancreatic enzymes (lipase, amylase and p-amylase) were measured daily throughout the first three days of hospitalization. The lipase/amylase ratio was also calculated. Univariate and logistic regression analyses were performed.

RESULTS

Age, sex and plasma levels of C-reactive protein, glutamic oxalacetic transaminase, glutamic pyruvic transaminase, alkaline phosphatase, (gamma-glutamyl-transpeptidase, bilirubin and amylase were significantly different in the two groups. The lipase/amylase ratio was not useful. Logistic regression analysis based on four variables: sex, age, C-reactive protein(day) 1 and glutamic pyruvic transaminase(day) 2 allowed the correct classification in 44 of 45 cases (97.7%).

CONCLUSIONS

Biliary and alcoholic acute pancreatitis present differing biochemical profiles-Glutamic pyruvic transaminase(day)2 and C-reactive protein(day) 1 were the variables with the highest predictive value. Taking into account these two biochemical parameters plus age and sex, an accurate and early etiologic classification was possible in the vast majority of cases in the present study.