There is no defined biomarker for BRONJ diagnosis with satisfactory performance in clinic. In this study, we established the BRONJ model and selected 7 microRNAs as candidate for BRONJ diagnosis from microRNA microarray reported by other research. Dysregulated microRNAs during BRONJ were detected and validated in two independent animal experiments using serum samples. In the first part, serum miR-21, miR-23a and miR-145 were significantly altered in between BRONJ and control group. And an Indice was constructed as-0.032+(0.154× miR-21)+(0.145× miR-23a)+(-0.700× miR-145) using logistic regression model to improve diagnostic performance. The performance of Indice to differentiate BRONJ subjects from control group was analyzed as AUC of 0.82 (95% CI, 0.72-0.92) or 0.85 (95% CI, 0.73-0.97) in the first or second part. Moreover, the predictive performance of Indice to discriminate BRONJ-1w and BRONJ-4w from control group was displayed as AUC of 0.65 (95% CI, 0.47-0.84) or 0.75 (95% CI, 0.60-0.91), which was better than individual circulating microRNAs. In addition, the expressions of candidate microRNAs were validated in human samples. Consequently, we investigated a combined Indice constructed with circulating microRNAs for BRONJ diagnosis and prediction.