Abstract Purpose: To investigate the potential of circulating-miRNAs (ct-miRNA) as noninvasive biomarkers to predict the efficacy of single/dual HER2-targeted therapy in the NeoALTTO study. Experimental Design: Patients with plasma samples at baseline (T0) and/or after 2 weeks (T1) of treatment were randomized into training (n = 183) and testing (n = 246) sets. RT-PCR–based high-throughput miRNA profiling was employed in the training set. After normalization, ct-miRNAs associated with pathologic complete response (pCR) were identified by univariate analysis. Multivariate logistic regression models were implemented to generate treatment-specific signatures at T0 and T1, which were evaluated by RT-PCR in the testing set. Event-free survival (EFS) according to ct-miRNA signatures was estimated by Kaplan–Meier method and Cox regression model. Results: In the training set, starting from 51 ct-miRNAs associated with pCR, six signatures with statistically significant predictive capability in terms of area under the ROC curve (AUC) were identified. Four signatures were confirmed in the testing set: lapatinib at T0 and T1 [AUC 0.86; 95% confidence interval (CI), 0.73–0.98 and 0.71 (0.55–0.86)], respectively; trastuzumab at T1 (0.81; 0.70–0.92); lapatinib + trastuzumab at T1 (0.67; 0.51–0.83). These signatures were confirmed predictive after adjusting for known variables, including estrogen receptor status. ct-miRNA signatures failed to correlate with EFS. However, the levels of ct-miR-140-5p, included in the trastuzumab signature, were associated with EFS (HR 0.43; 95% CI, 0.22–0.84). Conclusions: ct-miRNAs discriminate patients with and without pCR after neoadjuvant lapatinib- and/or trastuzumab-based therapy. ct-miRNAs at week two could be valuable to identify patients responsive to trastuzumab, to avoid unnecessary combination with other anti-HER2 agents, and finally to assist deescalating treatment strategies.