Background: The multicenter phase III TRAIN-2 study showed high pathological complete response (pCR) rates after neoadjuvant chemotherapy with and without anthracylines plus dual HER2-blockade in stage II-III HER2-positive breast cancer patients (67% vs 68%, p = 0.95) (NCT01996267). Here we report 3-year efficacy and safety outcomes. Methods: Patients were randomly assigned (1:1) to receive 3 cycles 5-fluoruoracil (500mg/m²), epirubicin (90mg/m²), and cyclophosphamide (500mg/m²) followed by 6 cycles paclitaxel (80mg/m² day 1 and 8) and carboplatin (AUC = 6mg/ml·min) (FEC-PC) or 9 cycles paclitaxel and carboplatin (PC). Both regimens were combined with trastuzumab (T; 6mg/kg, loading dose 8mg/kg) and pertuzumab (Ptz; 420mg, loading dose 840mg). Patients completed one year of trastuzumab, radiotherapy and endocrine therapy as indicated. The primary endpoint pCR was previously reported. Secondary efficacy endpoints reported here are event-free-survival (EFS) defined as time from randomization to first event (disease progression resulting in inoperability, recurrence [contralateral DCIS excluded], secondary primary malignancies, or death) and overall survival (OS), both in the intention-to-treat population. Safety endpoints are reported for patients treated with at least one dose of study medication. Results: 438 patients were randomized (219/arm) and evaluable for long-term efficacy endpoints. After a median follow-up of 48.8 months 23 EFS events occurred in the FECT-PTC-Ptz-arm and 21 in the PTC-Ptz-arm (HR 0.90; 95% CI 0.50-1.63). Three-year EFS estimates were 92.7% (95% CI: 89.3-96.2) for FECT-PTC-Ptz and 93.6% (95% CI: 90.4-96.9) for PTC-Ptz. Three-year OS estimates were 97.7% (95% CI: 95.7-99.7) for FECT-PTC-Ptz and 98.2% (95% CI: 96.4-100) for PTC-Ptz. These results were irrespective of hormone receptor and nodal status. LVEF decline ≥10% from baseline and < 50% was more common in patients who received anthracyclines than in the PTC-Ptz arm (8.6% vs. 3.2%, p = 0.021). Two patients in the FECT-PTC-Ptz arm developed acute leukemia. No other new safety concerns were seen. Conclusions: The 3-year follow-up of the TRAIN-2 study confirms the results of the primary outcome that anthracylines do not improve efficacy and are associated with clinically relevant toxicity. A neoadjuvant carboplatin-taxane based regimen with dual HER2-blockade can be considered in all stage II-III breast cancer patients, regardless of hormone receptor and nodal status. Clinical trial information: NCT01996267.