Evaluation of early pregnancy by serial chorionic gonadotropin determinations: a comparison of methods by receiver operating characteristic curve analysis

Two methods of evaluating early pregnancies by serial human chorionic gonadotropin (hCG) determinations were evaluated by receiver operating characteristic (ROC) curve analysis. The first method used normal values (mean and upper confidence limits) of the hCG doubling time (DT) determined by a linear regression analysis of hCG concentrations from 35 normal pregnancies during the first 42 postmenstrual days. In the second method, normal values were determined at multiple sampling ranges of hCG concentrations from 49 normal pregnancies. In the evaluation of 91 pregnancies of which 58 progressed to viability, 25 spontaneously aborted, and 8 were ectopic gestations, the ROC curves demonstrated that for any given false-positive rate, the use of multiple normal values of DT was more sensitive than the use of a single mean DT determined by linear regression analysis. The study indicates that a more precise definition of the normal values of the hCG-DT results in improved early detection of abnormal pregnancies.