Background Patients with large B-cell lymphoma (LBCL) primary refractory to or relapsed within 12 months of first-line therapy are at high risk for poor outcomes with current standard of care, platinum-based salvage immunochemotherapy and autologous haematopoietic stem cell transplantation (HSCT). Lisocabtagene maraleucel (liso-cel), an autologous, CD19-directed chimeric antigen receptor (CAR) T-cell therapy, has previously demonstrated efficacy and manageable safety in third-line or later LBCL. In this Article, we report a prespecified interim analysis of liso-cel versus standard of care as second-line treatment for primary refractory or early relapsed (within 12 months after response to initial therapy) LBCL. Methods TRANSFORM is a global, phase 3 study, conducted in 47 sites in the USA, Europe, and Japan, comparing liso-cel with standard of care as second-line therapy in patients with primary refractory or early (≤12 months) relapsed LBCL. Adults aged 18–75 years, Eastern Cooperative Oncology Group performance status score of 1 or less, adequate organ function, PET–positive disease per Lugano 2014 criteria, and candidates for autologous HSCT were randomly assigned (1:1), by use of interactive response technology, to liso-cel (100 × 106 CAR+ T cells intravenously) or standard of care. Standard of care consisted of three cycles of salvage immunochemotherapy delivered intravenously—R-DHAP (rituximab 375 mg/m2 on day 1, dexamethasone 40 mg on days 1–4, two infusions of cytarabine 2000 mg/m2 on day 2, and cisplatin 100 mg/m2 on day 1), R-ICE (rituximab 375 mg/m2 on day 1, ifosfamide 5000 mg/m2 on day 2, etoposide 100 mg/m2 on days 1–3, and carboplatin area under the curve 5 [maximum dose of 800 mg] on day 2), or R-GDP (rituximab 375 mg/m2 on day 1, dexamethasone 40 mg on days 1–4, gemcitabine 1000 mg/m2 on days 1 and 8, and cisplatin 75 mg/m2 on day 1)—followed by high-dose chemotherapy and autologous HSCT in responders. Primary endpoint was event-free survival, with response assessments by an independent review committee per Lugano 2014 criteria. Efficacy was assessed per intention-to-treat (ie, all randomly assigned patients) and safety in patients who received any treatment. This trial is registered with ClinicalTrials.gov, NCT03575351, and is ongoing. Findings Between Oct 23, 2018, and Dec 8, 2020, 232 patients were screened and 184 were assigned to the liso-cel (n=92) or standard of care (n=92) groups. At the data cutoff for this interim analysis, March 8, 2021, the median follow-up was 6·2 months (IQR 4·4–11·5). Median event-free survival was significantly improved in the liso-cel group (10·1 months [95% CI 6·1–not reached]) compared with the standard-of-care group (2·3 months [2·2–4·3]; stratified hazard ratio 0·35; 95% CI 0·23–0·53; stratified Cox proportional hazards model one-sided p<0·0001). The most common grade 3 or worse adverse events were neutropenia (74 [80%] of 92 patients in the liso-cel group vs 46 [51%] of 91 patients in the standard-of-care group), anaemia (45 [49%] vs 45 [49%]), thrombocytopenia (45 [49%] vs 58 [64%]), and prolonged cytopenia (40 [43%] vs three [3%]). Grade 3 cytokine release syndrome and neurological events, which are associated with CAR T-cell therapy, occurred in one (1%) and four (4%) of 92 patients in the liso-cel group, respectively (no grade 4 or 5 events). Serious treatment-emergent adverse events were reported in 44 (48%) patients in the liso-cel group and 44 (48%) in the standard-of-care group. No new liso-cel safety concerns were identified in the second-line setting. There were no treatment-related deaths in the liso-cel group and one treatment-related death due to sepsis in the standard-of-care group. Interpretation These results support liso-cel as a new second-line treatment recommendation in patients with early relapsed or refractory LBCL. Funding Celgene, a Bristol-Myers Squibb Company.