Abstract: Objective To investigate the efficacy of Ruxolitinib in the treatment of myelofibrosis and its predictive indices. Methods A total of 64 patients with myelofibrosis, treated with Ruxolitinib in the Department of Hematology, Zhongshan Hospital, Fudan University from November 2017 to December 2020, were collected as the study subjects. The endpoint was the reduction of spleen volume ≥35% from baseline at 12 weeks of treatment. LASSO regression was used to screen the predictive indices of the therapeutic effect of Ruxolitinib in the treatment of myelofibrosis. The regression model was established by the binary logistic regression, and the prediction efficiency of the model was evaluated by the areas under the curve (AUC) of receiver operating characteristic (ROC). Results Among the 64 patients with myelofibrosis, DIPSS scores were 10 low risk patients, 22 intermediate risk-1 patients, 14 intermediate risk-2 patients, and 18 high risk patients; 45 cases of JAK2V617F gene mutation, 3 cases of MPL W515L/K gene mutation, 8 cases of CALR gene mutation, and 8 cases of triple negative. The results of LASSO regression showed that 13 variables were related to the efficacy of Ruxolitinib. Using MPV, BLAST, PLT and RDW-SD as predictors, the predictive model of efficacy was successfully established based on Logistic regression, with AUC of 0.774, sensitivity of 81.0% and the specificity of 69.6%. There were 23 patients (35.94%) with myelofibrosis whose spleen volume decreased by ≥35% compared with baseline; 15 patients with the initial sufficient dose of Ruxolitinib were selected strictly for subgroup analysis according to the platelet count, and the result showed that the spleen volume of 7 patients (46.67%) was reduced by ≥35% compared with the baseline, which was statistically significant compared with all patients (P<0.001). Conclusions Early treatment with sufficient dose of Ruxolitinib shows good efficacy in myelofibrosis patients, and the Logistic regression model including MPV, BLAST, PLT and RDW-SD as the predictive factors has good prediction efficacy.

Key words: Myelofibrosis, Ruxolitinib, LASSO regression, Logistic regression, Therapeutic effect