Identification of persistent trophoblastic diseases based on a human chorionic gonadotropin regression curve by means of a stepwise piecewise linear regression analysis after the evacuation of uneventful moles.

Among the 191 patients with complete hydatidiform moles who were diagnosed and treated at Kyushu University Hospital from 1982 until 1996, 167 patients were diagnosed with uneventful moles retrospectively. The serial beta human chorionic gonadotropin (hCG) values in the 167 patients with uneventful moles were analyzed by a stepwise piecewise linear regression analysis in order to establish a normal regression curve of a human chorionic gonadotropin after a molar pregnancy. This normal regression curve is considered to be excellent regarding sensitivity (24/24-100%) and to be equivalent to the identification based on a plateau or a rise regarding specificity (156/167-93.4%). To distinguish patients with persistent trophoblastic disease (PTD) from uneventful moles, this normal curve is thus considered to be accurate since the accuracy was 180/191 (94. 2%). The weeks exceeding the normal regression curve in 24 PTD patients were 5.04 +/- 3.85 weeks and were also earlier than the weeks based on a plateau or a rise (P = 0.01). Within 7 weeks after evacuation, in 21/24 (87.5%) of the PTD cases, the beta-hCG values exceeded the normal range, while in only 14/24 (58.3%) the beta-hCG showed a change in the shape of a plateau or a rise. In addition, in 19/24 (79.2%) of the PTD patients, the time exceeding the normal range was shorter than the time exhibiting a plateau or a rise in the beta-hCG change. The above findings thus led us to conclude that this normal regression curve was useful for discriminating PTD from uneventful moles more precisely and more quickly than by identification based on a plateau or a rise.