Evaluation of midkine and anterior gradient 2 in a multimarker panel for the detection of ovarian cancer

Abstract

The aims of this study were: to characterise and compare plasma concentrations of midkine (MDK) in normal healthy women with concentrations observed in women with ovarian cancer; and to establish and compare the performance of MDK with that of anterior gradient 2 protein (AGR2) and CA125 in the development of multi-analyte classification algorithms for ovarian cancer. Median plasma concentrations of immunoreactive MDK, AGR2 and CA125 were significantly greater in the case cohort (909 pg/ml, 765 pg/ml and 502 U/ml, respectively n = 46) than in the control cohort (383 pg/ml, 188 pg/ml and 13 U/ml, respectively n = 61) (p < 0.001). The area under the receiver operator characteristic curve (AUC) for MDK and AGR2 was not significantly different (0.734 0.046 and 0.784 0.049, respectively, mean SE) but were both significantly less than the AUC for CA125 (0.934 0.030, p < 0.003). When subjected to stochastic gradient boosted logistic regression modelling, the AUC of the multi-analyte panel (MDK, AGR2 and CA125, 0.988 0.010) was significantly greater than that of CA125 alone (0.934 0.030, p = 0.035). The sensitivity and specificity of the multi-analyte algorithm were 95.2 and 97.7%, respectively. Within the study cohort, CA125 displayed a sensitivity and specificity of 87.0 and 94.6%, respectively. The data obtained in this study confirm that both MDK and AGR2 individually display utility as biomarkers for ovarian cancer and that in a multi-analyte panel significantly improve the diagnostic utility of CA125 in symptomatic women. © 2010 Rice et al; licensee BioMed Central Ltd.