Werner M, Steinberg W M, Pauley C
Department of Pathology, George Washington University Medical Center, Washington, DC 20037.
Clin Chem. 1989 Jun;35(6):967-71.
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).
在67例疑似急性胰腺炎的急诊病例中，评估了以酶测定结果为指标诊断急性胰腺炎的最佳策略。我们在入院时以及住院期间多次测量了以活性浓度（U/L）、定时排泄量（U/h）和淀粉酶/肌酐清除率表示的尿淀粉酶，以及血清淀粉酶、弹性蛋白酶、脂肪酶和胰蛋白酶原。采用受试者工作特征函数来评估各变量在初始检查结果以及回顾性确定的最高个体检查结果中的诊断鉴别能力。我们对血清指标的六种可能组合和四种可能三联进行了多次单变量鉴别，并采用相同的处理方法。结果表明，此类尿检测应被摒弃，所有单个血清检测结合起来约有0.9的敏感性和0.9的特异性，两种检测的组合并不能明显提高鉴别能力，而三联检测可能会降低鉴别能力。敏感性和特异性之间的权衡不仅取决于所选的决策阈值，还取决于采样策略（初始值与最高值）和解释规则（布尔“与”策略与“或”策略）。
