Comparison of serum CYFRA 21-1 and ROMA in distinguishing ovarian cancer from benign pelvic masses.

Abstract

e15565 Background: CYFRA 21-1 is a known lung cancer biomarker. In subjects with ovarian cancer (OC) versus those with benign gynecological diseases, serum CYFRA 21-1 was reported sensitivities (SN) of 44% and 41% at specificity (SP) of 95% for OC detection using an ELISA assay by Hasholzner U (1994) and Tempfer C (1998), respectively. Nolen B (2010) reported a SN of 84.1% and 85.8% and a SP of 85% and 79.3% with a combination of CYFRA 21-1, CA 125 and HE4 using a multiplex bead-based immunoassay in a training and validation set, respectively. We carried out a cohort study to evaluate the serum CYFRA 21-1 as an ovarian cancer biomarker and compared it with ROMA (Risk of Ovarian Malignancy Algorithm), an FDA-cleared ovarian cancer algorithm. Methods: ARCHITECT CYFRA 21-1 assay was used to measure the levels of serum CYFRA 21-1. Single point serum samples from 170 female subjects with a pelvic mass were randomly selected from an archive cohort of female subjects. Among the 170 subjects, 47 had OC (42 with epithelial OC and 5 with low malignant potential tumor) and 123 with benign gynecological diseases. ROMA values were obtained with HE4 EIA and ARCHITECT CA 125 II. Results: For ARCHITECT CYFRA 21-1, the ROC-AUC, cut-off, SN, SP, PPV and NPV were 86%, 1.8 ng/mL, 70.2% (95% CI: 55.1% to 82.7%), 95.1% (95% CI: 89.7% to 98.2%), 84.6% and 89.3%, repectively. For ROMA algorithm, the ROC-AUC, cut-off, SN, SP, PPV and NPV were 92%, 2.44, 89.4% (95% CI: 76.9% to 96.5%), 84.6% (95% CI: 76.9% to 90.4%), 68.9% and 95.4%, repectively. Chi-square test showed an SN for ROMA was significantly higher than that of ARCHITECT CYFRA 21-1 (p < 0.05), and SP for ARCHITECT CYFRA 21-1 was significantly higher than that of ROMA (p < 0.01). Conclusions: In distinguishing OC from benign gynecological diseases, ROMA algorithm appears to be more sensitive than serum ARCHITECT CYFRA 21-1 (p < 0.05), and serum ARCHITECT CYFRA 21-1 appears to be more specific than (p < 0.01) in this cohort study.