Abstract: Objective To investigate the use of various carfilzomib-based combination therapies for relapsed/refractory multiple myeloma (RRMM) to explore the optimal combination regimens, treatment efficacy and safety in different subgroups of RRMM patients, and to construct an efficacy evaluation model. Methods Retrospective analysis was performed on the clinical data of 22 RRMM patients who received carfilzomib-based chemotherapy therapy at Department of Hematology, Xuzhou Central Hospital, from March 2023 to August 2024. Their therapeutic efficacy, survival and adverse reactions were observed. The efficacy and safety of different carfilzomib combination regimens and clinical outcomes in different subgroups were evaluated, and an efficacy assessment model was constructed. Results Among the 22 patients, 21 survived at the end of follow-up. The overall response rate (ORR) was 68.2%, and 45.5% of patients achieved complete remission (CR) or very good partial remission (≥VGPR). The ≥VGPR rate in patients receiving carfilzomib-dexamethasone-pomalidomide (KPD) regimen was higher than in those with carfilzomib-dexamethasone-lenalidomide (KRD) (P=0.029). Patients with ≤3 prior lines of treatment had a higher ≥VGPR rate than those with >3 prior lines of treatment (P=0.03). The ≥VGPR rate and ORR in the low neutrophil-to-lymphocyte ratio (NLR) group increased, compared to the high NLR group (P=0.043, P=0.02). The ≥VGPR rate and ORR in patients with DSP=1.00, P=0.616). The ≥VGPR rate and ORR in patients with low-risk cytogenetics were higher than in those with high-risk cytogenetics (P=0.126, P=0.302). The ≥VGPR rate and ORR in patients without extramedullary disease were higher than in those with extramedullary disease (P=0.594, P=0.565). The efficacy assessment models (Model 1) based on prior treatment lines, NLR, lymphocyte-to-monocyte ratio (LMR), DS staging, cytogenetics, and extramedullary disease, as well as Model 2 (excluding cytogenetic indicators), both demonstrated predictive value for treatment efficacy in RRMM patients. Model 2, which excluded cytogenetic data, showed better efficacy in predicting remission (AUC=92.5) than Model 1 (AUC=90). The adverse reaction rate was 59.1%, and all patients had elevated N-terminal pro B-type natriuretic peptide (NT-proBNP) levels, with 11 patients showing NT-proBNP levels above the normal range. After symptomatic treatment, all adverse reactions were controlled. Conclusions Carfilzomib -based combination therapy is an effective and safe treatment option for RRMM patients. The KPD regimen is superior to KRD in achieving ≥VGPR. Early use of carfilzomib leads to deeper remission. High NLR negatively impacts the treatment efficacy in RRMM patients. For patients with DS III stage, high-risk cytogenetic abnormalities, or extramedullary disease, the carfilzomib -based combination therapy is a reliable treatment choice. The efficacy assessment models based on prior treatment lines, NLR, LMR, DS stage, and extramedullary disease provide valuable predictive insights into treatment efficacy for RRMM patients.