BACKGROUND AND AIMS: Plasma cell-free DNA (cfDNA) has been used to monitor gene mutations and diagnose tumors. Discriminating parathyroid carcinoma (PC) from parathyroid adenoma (PA) before surgery is difficult because of the overlap in clinical features between parathyroid neoplasms. We aimed to detect cfDNA mutations in plasma samples from PC and PA patients before surgery to predict the CDC73 status in tumor tissue and help in the differential diagnosis of parathyroid neoplasms. MATERIALS AND METHODS: Eighteen PC patients and 13 PA patients were enrolled. Plasma cfDNA was detected using next-generation sequencing, with DNA from matched peripheral white blood cells used as a control. CDC73 gene mutations were detected via whole-exome sequencing or parafibromin staining via immunohistochemistry of tumor tissues. Logistic regression was used to evaluate the ability of cfDNA mutations to predict the CDC73 status in tumor tissue and for differential diagnosis. CDC73 gene mutation or parafibromin staining loss were defined as CDC73 abnormalities. RESULTS: One PC patient was not tested for CDC73 abnormalities due to the absence of tumor specimen. CDC73 abnormalities were not detected in all 13 PA patients, whereas 10 PC patients harboured CDC73 abnormalities in tumor specimens (P = 0.001). Among the 10 patients, CDC73 mutations were identified in the cfDNA of 8 patients. In another 20 patients without CDC73 abnormalities in tumors, CDC73 mutation was detected in the cfDNA of 4 patients. Using the CDC73 status in cfDNA, the area under the receiver operating characteristic curve (AUC) for predicting CDC73 abnormalities in tumor tissue was 0.80 (95% CI: 0.622-0.978), and the AUC for predicting malignancy was 0.795 (95% CI 0.632-0.958). CONCLUSION: This study is the first attempt to evaluate the gene mutation status of parathyroid neoplasms through the deep sequencing of plasma cfDNA, which could also help to identify PC prior to surgery.