BACKGROUND: High-dose melphalan is of little benefit as a regimen for patients with relapsed or refractory myeloma undergoing an autologous stem-cell transplant (ASCT). The poor performance of single-agent melphalan in this setting prompted us to study a new high-dose combination of infused gemcitabine, busulfan, and melphalan. METHODS: We did a phase 2 trial at the University of Texas MD Anderson Cancer Center (Houston, TX, USA). We enrolled patients with primary refractory or relapsed myeloma who had received treatment with bortezomib, an immunomodulatory drug, or both, or who were receiving a salvage ASCT. Gemcitabine was infused at 1875 mg/m2 for 3 h for 2 days, followed by busulfan (target area under the curve 4000 μmol/L per min per day for 4 days) and melphalan (60 mg/m2 per day for 2 days). The primary endpoint of this trial was to establish the proportion of patients with measurable disease at ASCT receiving gemcitabine, busulfan, and melphalan who achieved stringent complete remission in accordance with the International Myeloma Working Group criteria. We then retrospectively compared the patients in this study with all other concurrent patients at the MD Anderson Cancer Center who were eligible for this trial but declined to participate or had no financial coverage for ASCT in a clinical trial and instead received melphalan at 200 mg/m2 intravenously over 30 min on 1 day, followed by ASCT (control group). To compare survival outcomes, we used a statistical algorithm to select a subset of patients from this control cohort who were matched in a 1-2:1 ratio with the patients in the gemcitabine, busulfan, and melphalan group by sex, age, disease status, refractory to both proteasome inhibitors and immunomodulatory imide drugs, time from diagnosis to ASCT, and cytogenetic risk. All analyses were per protocol. This is the final analysis of the clinical trial, which is registered at ClinicalTrials.gov, number NCT01237951. FINDINGS: Between Nov 30, 2010, and Dec 11, 2013, we enrolled 74 patients into the gemcitabine, busulfan, and melphalan trial. In these patients, median age was 58 years (IQR 51-62), median number of previous lines of therapy was two (2-5), 38 patients had high-risk cytogenetics, 17 were unresponsive to all previous treatments, and 32 were receiving a salvage ASCT. We identified 184 patients for the concurrent control cohort. The study patients and the concurrent controls received similar post-ASCT maintenance. Among patients with measurable disease at ASCT, 16 of 65 patients (24·6%, 95% CI 14·2-35·0) in the gemcitabine, busulfan, and melphalan group had stringent complete remission compared with 22 of 174 patients (12·6%, 10·1-15·1) in the concurrent control group (p=0·040). Median follow-up time was 36 months (IQR 30-46) in the patients receiving gemcitabine, busulfan, and melphalan and 34 months (25-53) in the matched control subset (n=111). With respect to the secondary survival endpoints, the gemcitabine, busulfan, and melphalan cohort had significantly longer median progression-free survival than the matched control cohort (15·1 months [95% CI 8·7-22·1] vs 9·3 months [8·0-10·7]) with a significantly reduced risk of progression or death (HR 0·55, 95% CI 0·38-0·81, log-rank p=0·030), as well as significantly longer median overall survival (37·5 months [26-not reached] vs 23·0 months [16·6-30·5]) and a lower risk of death (HR 0·60, 0·34-0·84, log-rank p=0·0092). For only the patients treated with gemcitabine, busulfan, and melphalan, grade 3 or worse adverse events included grade 3 mucositis (12 patients), grade 3 dermatitis (five patients), grade 3 aminotransferase elevation (seven patients), grade 3 diarrhoea (two patients), and three treatment-related deaths. One death was cardiac sudden death and two were due to sepsis. INTERPRETATION: Gemcitabine, busulfan, and melphalan is a comparatively safe and active regimen for ASCT in patients with refractory or relapsed myeloma. Better outcomes were achieved in patients who received this regimen than in a concurrent matched cohort receiving melphalan, although this will need to be confirmed in a prospective, randomised trial. FUNDING: Otsuka Pharmaceutical Development & Commercialization and US National Cancer Institute.