Optimization of serum miR-371a-3p for the detection of chemotherapy-naïve minimal residual germ cell tumor.

Abstract

423 Background: Currently used serum tumor markers exhibit middling sensitivity and specificity for the detection of germ cell tumors (GCTs). Circulating miRNAs, particularly miR-371a-3p, have exhibited strong performance characteristics in the pre-orchiectomy setting. However, the ability of miR-371a-3p to detect chemotherapy-naïve occult disease is less clear. We previously reported in a small cohort of patients that miR-371a-3p outperforms conventional markers in the context of chemotherapy-naïve minimal residual disease. Here, we expand upon that experience and further refine the assay, including the recommendation of an indeterminate range. Methods: We examined over 250 assays to compare performance between classification based on raw Cq, Cq normalized to miR-30b-5p (∆Cq), or relative quantification to normal serum (RQ). We prospectively collected pre-surgical serum samples and clinical information from GCT patients undergoing primary retroperitoneal lymph node dissection (RPLND) at our institution. Patients were classified based on histology as “Control” (pure teratoma or no GCT) or “GCT”. RNA was extracted from these samples, and miR-371a-3p (target) and miR-30b-5p (reference gene) were detected by qPCR. Performance was assessed by calculation of sensitivity, specificity, and area under the ROC curve (AUC). Results: Sensitivity, specificity, and AUC did not significantly change based on thresholding metric (Cq, ∆Cq, or RQ). We found a 25% chance that any given control sample would return a false positive result, and to combat this, estimated an indeterminate range of Cq 28-35. Repeating any indeterminate sample once improved performance. We then compared the performance of this updated process to a simple binary threshold in a cohort of patients with occult disease. 32 patients were enrolled in the study (15 Control, 17 GCT). 12 samples were indeterminate on first run (6 Control, 6 GCT). 8 remained indeterminate on a second run, all of which harbored teratoma or viable GCT. Inclusion of a second run prevented 3 false positives, improving specificity from 86% to 100% and yielding a sensitivity of 92% and AUC of 0.96 (95% CI: 0.89-1). Conclusions: The use of Cq instead of RQ to classify results for the miR-371a-3p test would eliminate the need for a normal sample in each run without injuring assay performance. In the context of chemotherapy-naïve minimal residual germ cell tumor, consideration of an indeterminate range may improve the performance of the miR-371a-3p test.