Introduction: Although novel therapies have significantly improved the prognosis of multiple myeloma (MM) patients, there are still a subgroup of MM patients experienced early relapse or disease progression. To early identify the patients with compromised prognosis in our center, the study aimed to establish a predictive model for ultra-high risk (UHR) in MM. Methods: A study from a tertiary hospital in western China analyzed the patients with newly diagnosed MM (NDMM) receiving bortezomib and/or lenalidomide-included regimens from January 2015 to June 2021. The last follow-up date was December 31, 2022. The patients were randomly divided (2:1) into training and validation cohorts, and the patients with progression-free survival (PFS) less than 12 months were defined as UHR. Univariate and multivariate Cox regression analyses were performed for clinical feature selection. Independent baseline factors were incorporated and scored to develop the predictive model. The UHR group was then defined beyond the optimal cut-off value of total scores. Results: A total of 314 MM patients were enrolled and 102 (32.5%) patients underwent autologous hemopoietic stem cell transplantation (ASCT). The median time of follow-up was 36.6 months. Disease progression and death occurred in 178 (56.7%) and 69 (22.0%) patients. The median PFS (mPFS) was 36.5 months, and the median overall survival (mOS) was not reached. The mPFS of patients stratified into R-ISS III, R2-ISS III, R2-ISS IV, IMWG high-risk and mSMART double/triple hit was 23.8, 25.6, 10.0, 17.2 and 10.5 months, respectively. Among 54 patients at UHR, less than half of them could be identified by the aboved stratification systems. In the training cohort, elevated LDH, hypercalcemia, 1q21 gain/amplification (1q+) and the presence of extramedullary disease (EMD) showed significant association with UHR in NDMM. Referring to the results, the UHR predictive model was developed as 5 points for elevated LDH, 6 points for hypercalcemia, 7 points for 1q+ and 10 points for EMD. Patients with total scores of more than 12 points can be predicated as UHR. This model identified UHR and non-UHR groups with significantly different mPFS (9.8 v.s. 37.5 months, HR = 3.24, 95% CI:2.09-5.03, P < 0.001) and 5-year OS (63.3% vs. 83.8%, HR = 2.92, 95% CI: 1.47-5.81, P = 0.002). The area under the curve (AUC) of the receiver operating characteristic (ROC) curve was 0.77, and in the validation cohort, the UHR model kept good discrimination (AUC: 0.79) at the latest follow-up. Conclusions: In this study, elevated LDH, hypercalcemia, 1q+ and EMD were selected and scored to develop the UHR model for NDMM patients. At present, the extension of follow-up time and expansion of sample size in the validation cohort are still ongoing. External validity is also designed to verify the generalizability of the model. The research was funded by: Natural Science Foundation of Sichuan Province, China, Grant/Award Number: 2022NSFSC1299 Keywords: Risk Models, Multiple Myeloma No conflicts of interests pertinent to the abstract.