Abstract: Objective: To explore the risk factors of early recurrence in multiple myeloma (MM) patients after initial complete response and establish a prediction model. Methods: A total of 122 MM patients were selected,of which 87 were used as the training group for model creation,and 35 were used as the validation group for model evaluation.The clinical characteristics,platelet count (PLT),hemoglobin (HGB),serum uric acid (SUA),serum creatinine (SCR),lactate dehydrogenase (LDH) and cytogenetic abnormalities of patients with early recurrence and those without recurrence after initial treatment were analyzed in the training group.The risk factors were screened by multivariate logistic regression analysis,and the regression equation was established.The receiver operating characteristic (ROC) curve was used for self-validation and K-fold cross-validation was used for out-of-group validation. Results: There were no significant differences in gender,age,M protein type,ISS stage,chemotherapy cycle and bone disease grade between the two groups (P>0.05).In the training group,patients with recurrence were older than 60 years old,ISS stage Ⅲ,mSMART3.0 high risk,double/triple strike ratio,SUA and SCR levels were higher than those without recurrence (P<0.05).Logistic regression analysis showed that age ≥60 years old,high ISS stage,mSMART3.0 high risk,double/triple blow,high expression of SUA and SCR were independent risk factors for early recurrence after initial treatment and complete remission of MM (P<0.05).The risk prediction model was established:logit (P) =-15.763+age×0.646+ISS stage×1.049+mSMART3.0 stratification×1.032+ double/triple strike×1.557+SUA×1.212+SCR×0.725.Self-verification:the model AUC was 0.859,the diagnostic sensitivity was 70.92%,and the specificity was 84.16%.K-fold cross verification:the training accuracy was 0.904±0.009,and the prediction accuracy was 0.881±0.049. Conclusions: The early recurrence of MM after initial complete remission is affected by age,ISS stage,mSMART3.0 stratification,double/triple blow,SUA and SCR level.Therefore,the logistic regression model can be built to predict the risk of early recurrence,which is conducive to early clinical decision-making.