Linear Regression of Postevacuation Serum Human Chorionic Gonadotropin Concentrations Predicts Postmolar Gestational Trophoblastic Neoplasia

Abstract

Currently, human chorionic gonadotropin (hCG) follow-up after evacuation of hydatidiform moles is essential to identify patients requiring chemotherapeutic treatment for gestational trophoblastic neoplasia (GTN). We propose a model based on linear regression of postevacuation serum hCG concentrations for the prediction of GTN. One hundred thirteen patients with at least 3 serum samples from days 7 to 28 after evacuation were selected from the Dutch Central Registry for Hydatidiform Moles (1994-2009). The slopes of the linear regression lines of the first 3 log-transformed serum hCG and free β-hCG values were calculated. Receiver operating characteristic curves were constructed to calculate areas under curve (AUCs). The slope of the hCG regression line showed an AUC of 0.906 (95% confidence interval, 0.845-0.967). Gestational trophoblastic neoplasia could be predicted in 52% of patients with GTN at 97.5% specificity (cutoff, -0.020). Twenty-one percent of patients with GTN could be predicted before diagnosis according to the International Federation of Gynecology and Obstetrics 2000 criteria. The slope of free β-hCG showed an AUC of 0.844 (95% confidence interval, 0.752-0.935), 69% sensitivity at 97.5% specificity, and 38% of patients with GTN could be predicted before diagnosis according to the International Federation of Gynecology and Obstetrics criteria. The slope of the linear regression line of hCG proved to be a good test to discriminate between patients who will achieve spontaneous disease remission and patients developing GTN. The slope of free β-hCG seems to be a better predictor for GTN than the slope of hCG. Although this model needs further validation for different assays, it seems a promising way to predict the more aggressive cases of GTN.