Serum mirnas predict response to CAR-T cell therapy in large B-cell lymphoma

Abstract Background: Chimeric antigen receptor T-cell (CAR-T) therapy has revolutionized the treatment of relapsed or refractory diffuse large B-cell lymphoma (R/R LBCL), improving patients' progression-free survival (PFS) and overall survival (OS). However, CAR-T treatment failure remains a significant challenge in LBCL, with over 50% of patients experiencing disease relapse or progression within the first 6 months post-CAR-T therapy. There is an urgent clinical need for tools to identify patients at high risk of CAR-T failure prior to treatment initiation. Previously, we identified four serum miRNAs (miR-21, miR-28, miR-130b and miR-155) in patients with diffuse large B-cell lymphoma (DLBCL) predictive of disease outcome. Herein, we evaluated the clinical value of these four miRNAs in predicting CAR-T efficacy and prognosis using retrospective serum samples from patients with R/R LBCL. Methods: Here was a retrospective single-center study involving 22 patients with R/R LBCL, including 19 cases of DLBCL and 3 cases of primary mediastinal large B-cell lymphoma (PMBCL). All patients received CAR-T therapy, with 13 (59.1%) treated with CD28 CAR-T and 9 (40.9%) with 4-1BB CAR-T. Efficacy was assessed via PET/CT at 4 weeks, 3 months, 6 months, and 12 months (D28, M3, M6, and M12) post-CAR-T infusion, using criteria of complete response (CR), partial response (PR), stable disease (SD), and progressive disease (PD). The median follow-up duration for relapse was 9.1 months (range: 3.2–36.5 months). PFS was defined as the time from CAR-T infusion to the first PD or death. We assayed the expression levels of the four miRNAs and an external reference in pretreatment serum samples by digital PCR (dPCR). The patients were stratified into CR vs. non-CR (PR+SD+PD) groups or response (CR+PR) vs. non-response (SD+PD) groups at each time point. We then optimized the predictive models through logistic regression and logit cutoff values for single or combined miRNAs and calculated the area under the receiver operating characteristic curve (AUC). Kaplan-Meier (KM) curves for PFS were generated to compare differences between groups stratified by miRNA expression, with hazard ratios (HR) calculated. A p-value < 0.05 was considered statistically significant. Results: Among the 22 patients before receiving CAR-T therapy, 63.6% had refractory disease, 36.4% had relapsed disease, and 36.4% had received ≥3 lines of prior therapy. Patient demographics and clinical characteristics included: 64.4% aged ≤60 years, 63.6% with Ann Arbor stage III–IV disease, 72.7% with elevated serum lactate dehydrogenase (LDH), 59.1% with maximum tumor size ≥5 cm, 45.4% with double-expressor/double-hit/triple-hit lymphoma and 47.4% with TP53 mutations. For the CR group vs. non-CR group, the AUC values of the 4-miRNA predictive model at D28, M3, M6 and M12 were 0.823 (p = 0.0016), 0.692 (p = 0.1034), 0.696 (p = 0.1208) and 0.769 (p = 0.0243), respectively. For the response (CR+PR) group vs. non-response group, the AUC values of the predictive model at M3, M6 and M12 were 0.924 (p < 0.0001), 0.883 (p < 0.0001), and 0.783 (p = 0.0111). Notably, the predictive model at M6 based solely on miR-28 also achieved an AUC of 0.792 (p = 0.0056). Kaplan-Meier curves stratified by the optimal logit cutoff values of the 4-miRNA model (for response vs. non-response at M3 or M6) showed significant differences in median PFS between groups. For the M3 model, median PFS was 843 days vs. 189 days (p < 0.0001, HR = 0.05254, 95%CI: 0.0135 – 0.2045); for the M6 model, median PFS was 850 days vs. 314 days (p = 0.0064, HR = 0.2083, 95%CI: 0.06745 – 0.6430). Conclusion This study demonstrates that a combined panel of 4 serum miRNAs (miR-21, miR-28, miR-130b, and miR-155) serves as a robust and clinically feasible predictive tool for assessing the efficacy and prognosis of CAR-T therapy in patients with R/R LBCL. Their integration into clinical practice has the potential to advance personalized medicine in lymphoma treatment, guiding pretreatment decision-making and ultimately improving the success rate of CAR-T therapy.