Strategic use of individual and combined enzyme indicators for acute pancreatitis analyzed by receiver-operator characteristics

Abstract

The optimal strategy for the diagnosis of acute pancreatitis with enzyme assay results as indicators was evaluated in 67 emergency cases in whom this condition was suspected. We measured urine amylase expressed as activity concentration (U/L), timed excretion (U/h), and amylase/creatinine clearance ratio, and also serum amylase, elastase, lipase, and trypsinogen, at admission and repeatedly during hospitalization. The receiver-operator characteristic function was used to evaluate the diagnostic discrimination of each variable among initial findings and among the highest individual findings established retrospectively. We applied the same treatment to multiple univariate discrimination, using the six possible pairs and the four possible triplets of serum indicators. The results suggest that such urine assays should be abandoned, that all individual serum assays combine about 0.9 sensitivity with 0.9 specificity, that pairing of two assays does not clearly enhance discrimination, and that triplets of tests may degrade discrimination. The trade-off between sensitivity and specificity is a function not only of the chosen decision threshold but also of the sampling strategy (initial vs highest values) and of the interpretation rule (Boolean "and" vs "or" strategy).