MicroRNAs have emerged as new diagnostic and therapeutic biomarkers for breast cancer. Herein, we analysed miR-99a-5p expression levels in primary tumours and plasma of breast cancer patients to evaluate its usefulness as a minimally invasive diagnostic biomarker. MiR-99a-5p expression levels were determined by quantitative real-time PCR in three independent cohorts of patients: (I) Discovery cohort: breast cancer tissues (n = 103) and healthy breast tissues (n = 26); (II) Testing cohort: plasma samples from 105 patients and 98 healthy donors; (III) Validation cohort: plasma samples from 89 patients and 85 healthy donors. Our results demonstrated that miR-99a-5p was significantly downregulated in breast cancer tissues compared to healthy breast tissues. Conversely, miR-99a-5p levels were significantly higher in breast cancer patients than in healthy controls in plasma samples from both testing and validation cohorts, and ROC curve analysis revealed that miR-99a-5p has good diagnostic potential even to detect early breast cancer. In conclusion, miR-99a-5p’s deregulated expression distinguished healthy patients from breast cancer patients in two different types of samples (tissues and plasma). Interestingly, expression levels in plasma were significantly lower in healthy controls than in early-stage breast cancer patients. Our findings suggest circulating miR-99a-5p as a novel promising non-invasive biomarker for breast cancer detection.