Clinical evaluation of an algorithm for the interpretation of hyperamylasemia.

Total amylase concentration in serum continues to be widely determined in the diagnosis of acute pancreatic disease. Accumulated experience has made clear, however, that this determination has distinct limitations. Consequently, the knowledge of the origin of hyperamylasemia may have an important influence on treatment, hospitalization, and extent of clinical investigations. We undertook a logical and systematic approach to the interpretation of hyperamylasemia through the use of an algorithm that can be applied in clinical situations without the need for the integration of radiologic procedures or clinical data. The proposed algorithm was tested for effectiveness in 97 consecutive hospitalized patients with hyperamylasemia (amylase level greater than twice the upper reference limit) for a 2-year period. The majority (52.5%) of these patients had acute pancreatitis. The algorithm assigned the correct diagnostic categories in 95.8% of cases, with a disagreement between patient diagnosis and algorithm-generated diagnosis in only four cases. These four patients (two with acute biliary disease, one with bacterial peritonitis, and one with chronic renal failure) had pancreatic lipase values greater than five times the upper reference limit, so that the algorithm classified their condition as acute pancreatitis. The clinical trial indicated that the proposed decision tree, which requires only knowledge of biochemical data that are readily available, is useful in the evaluation of elevated amylase activity and facilitates arrival at a definitive diagnosis.