Independent validation of a novel noninvasive 4-microRNA diagnostic model for multicancer early detection.

3065 Background: Cancer early detection is critical to reduce mortality as treating early stage cancers is more likely to have better outcomes. We previously developed a diagnostic model based on 4 serum cell-free microRNAs (miRNAs) capable of detecting 12 cancer types with high accuracy (AACR 2022, Poster 2890; Manuscript submitted). In the current study, we aimed to validate this diagnostic model using independent serum microRNA microarray datasets. Methods: Four microarray datasets assessing the expression of 2588 serum miRNAs in patients with esophageal squamous cell carcinoma, gastric cancer, prostate cancer and glioma, as well as non-cancer healthy controls with a standardized platform were identified from Gene Expression Omnibus (GEO). The datasets were combined, cases that were redundant among them or with the cases used in our previous study (correlation > 0.99) were excluded. The 4-miRNA model was applied to the final combined dataset to calculate a diagnostic index and make prediction of cancer vs. no-cancer using previously determined algorithm and cut-point. The performance of the diagnostic model was assessed using Receiver Operating Characteristic (ROC) analysis, sensitivity and specificity. Results: After excluding redundant cases, the final combined dataset consisted of 3877 subjects, including 447 esophageal, 1267 gastric, 769 prostate and 196 glioma cancer patients, as well as 1198 healthy controls. The 4-miRNA model demonstrated an area under the curve (AUC) of 0.986, 0.995, 0.992 and 0.990 in the ROC analysis, with a sensitivity of 85.0%, 99.6%, 90.6% and 86.7% for esophageal, gastric, prostate and glioma cancers, respectively, while achieving a 99.1% specificity. These performance metrics were highly consistent to those reported in our previous study (Table). Conclusions: The study provided an independent validation of the previously developed 4-miRNA model, further demonstrating that the diagnostic model we developed has the potential to be developed into a simple, inexpensive and noninvasive blood test for multi-cancer early detection with high accuracy. [Table: see text]