In AURA3 (NCT02151981), osimertinib, a third-generation epidermal growth factor receptor (EGFR)-tyrosine kinase inhibitor (TKI), significantly prolonged progression-free survival and improved response in patients with EGFR T790M advanced non-small-cell lung cancer (NSCLC) and progression on prior EGFR-TKI treatment. We report the final AURA3 overall survival (OS) analysis. Adult patients were randomized 2 : 1 to osimertinib (80 mg orally, once daily) or pemetrexed plus carboplatin/cisplatin (platinum–pemetrexed) intravenously, every 3 weeks (≤6 cycles). Patients could crossover to osimertinib on progression confirmed by blinded independent central review. OS and safety were secondary end points. A total of 279 patients were randomly assigned to receive osimertinib and 140 to platinum–pemetrexed (136 received treatment). At data cut-off (DCO; 15 March 2019), 188 patients (67%) receiving osimertinib versus 93 (66%) receiving platinum–pemetrexed had died. The hazard ratio (HR) for OS was 0.87 95% confidence interval (CI) 0.67–1.12; P = 0.277; the median OS was 26.8 months (95% CI 23.5–31.5) versus 22.5 months (95% CI 20.2–28.8) for osimertinib and platinum–pemetrexed, respectively. The estimated 24- and 36-month survival was 55% versus 43% and 37% versus 30%, respectively. After crossover adjustment, there was an HR of 0.54 (95% CI 0.18–1.6). Time to first subsequent therapy or death showed a clinically meaningful advantage toward osimertinib (HR 0.21, 95% CI 0.16–0.28; P < 0.001). At DCO, 99/136 (73%) patients in the platinum–pemetrexed arm had crossed over to osimertinib, 66/99 (67%) of whom had died. The most common adverse events possibly related to study treatment were diarrhea (32%; grade ≥3, 1%) and rash (grouped term; 32%; grade ≥3, <1%) in the osimertinib arm, versus nausea (47%; grade ≥3, 3%) in the platinum–pemetrexed arm. In patients with T790M advanced NSCLC, no statistically significant benefit in OS was observed for osimertinib versus platinum–pemetrexed, which possibly reflects the high crossover rate of patients from platinum–pemetrexed to osimertinib. ClinicalTrials.gov NCT02151981; https://clinicaltrials.gov/ct2/show/NCT02151981. •Median OS with osimertinib was 26.8 months versus 22.5 months with platinum–pemetrexed (HR 0.87, 95% CI 0.67–1.12; P = 0.277).•The lack of a significant survival benefit could reflect high percentage (73%) of platinum–pemetrexed to osimertinib crossover.•Analysis of OS adjusted for crossover showed an HR of 0.54 (95% CI 0.18–1.60).•Among patients receiving subsequent anticancer therapy, platinum chemotherapy was the most common after osimertinib (65%).•Grade ≥3 (possibly treatment-related) adverse events were observed less frequently with osimertinib (9% versus 34% with platinum–pemetrexed).