We investigated the disease-free survival (DFS) of HER2-positive primary breast cancer patients treated with neoadjuvant chemotherapy plus trastuzumab, as well as predictive factors for DFS and pathologic response. Data from 829 female patients treated between 2001 and 2010 were collected from 38 institutions in Japan. Predictive factors were evaluated using multivariate analyses. The 3-year DFS rate was 87 % [95 % confidence interval (CI) 85-90]. The pathologic complete response (pCR: ypT0/is + ypN0) rate was 51 %. The pCR rate was higher in the ER/PgR-negative patients than in the ER/PgR-positive patients (64 vs. 36 %, P < 0.001). Patients with pCR showed a higher DFS rate than patients without pCR (93 vs. 82 %, P < 0.001). Multivariate analysis revealed three independent predictors for poorer DFS: advanced nodal stage [hazard ratio (HR) 2.63, 95 % CI 1.36-5.21, P = 0.004 for cN2-3 vs. cN0], histological/nuclear grade 3 (HR 1.81, 95 % CI 1.15-2.91, P = 0.011), and non-pCR (HR 1.98, 95 % CI 1.22-3.24, P = 0.005). In the ER/PgR-negative dataset, non-pCR (HR 2.63, 95 % CI 1.43-4.90, P = 0.002) and clinical tumor stage (HR 2.20, 95 % CI 1.16-4.20, P = 0.017 for cT3-4 vs. cT1-2) were independent predictors for DFS, and in the ER/PgR-positive dataset, histological grade of 3 (HR 3.09, 95 % CI 1.48-6.62, P = 0.003), clinical nodal stage (HR 4.26, 95 % CI 1.53-13.14, P = 0.005 for cN2-3 vs. cN0), and young age (HR 2.40, 95 % CI 1.12-4.94, P = 0.026 for ≤40 vs. >40) were negative predictors for DFS. Strict pCR (ypT0 + ypN0) was an independent predictor for DFS in both the ER/PgR-negative and -positive datasets (HR 2.66, 95 % CI 1.31-5.97, P = 0.006 and HR 3.86, 95 % CI 1.13-24.21, P = 0.029, respectively). These results may help assure a more accurate prognosis and personalized treatment for HER2-positive breast cancer patients.