Accumulating evidence has suggested that concentrations of blood-based circulating micro-ribonucleic acids (microRNAs, miRNAs) in breast tumor patients are significantly higher/lower than that in normal individuals, indicating that circulating miRNAs may serve as novel blood-based biomarkers for breast tumor. However, the results of previous studies on this issue have been inconclusive. Therefore, we perform a meta-analysis to determine whether aberrant miRNA expression can be used as molecular markers in blood for the diagnosis of breast tumor. PubMed and other databases were searched to identify eligible studies. The sensitivity and specificity were used to plot the summary receiver operator characteristic curve and calculate the area under the curve (AUC). Finally, 15 articles with a total of 1,428 breast tumor patients and 952 healthy individuals were involved. The summary estimates revealed that the pooled sensitivity was 76 % with 95 % confidence interval (CI) of 67-83 %; the specificity was 87 % with 95 % CI of 77-93 %; the PLR was 5.9 with 95 % CI of 3.3-10.4; the NLR was 0.28 with 95 % CI of 0.20-0.39; the DOR was 21 with 95 % CI of 10-44; and the AUC was 0.88 with 95 % CI of 0.84-0.90. The most noteworthy is that multiple-miRNA assay displayed a better diagnostic performance than single-miRNA assay. In summary, the results of the present meta-analysis suggested that blood-based miRNAs may serve as novel molecular biomarkers for breast tumor, with a relative high level of accuracy, especially based on multiple-miRNA assay. Further large-scale prospective studies are necessary to validate their potential applicability for breast tumor prognosis, treatment, and surveillance.