Risk factors analysis and prediction model establishment of malignant tumor in patients with polymyositis/dermatomyositis

Abstract

Objective: To analyze the risk factors of polymyositis/dermatomyositis (PM/DM) complicated with malignant tumor and to construct clinical prediction model. Methods: A total of 427 PM/DM patients, who were admitted to Rheumatism Immunity Branch, the Second Affiliated Hospital, Air Force Medical University from January 1, 2015 to January 1, 2021, were enrolled in the study, including 129 males and 298 females. The mean age was (51.4±12.2) years. The patients were divided into control group (without malignant tumor, n=379) and case group (with malignant tumor, n=48) according to whether they were complicated with malignant tumors. In the two groups, 70% of the patients' clinical data were randomly selected as the training set data, and the remaining 30% were used as the validation set data. The clinical parameters were retrospectively collected, and risk factors of PM/DM complicated with malignant tumor were analyzed by binary logistic regression. R software was used to construct a clinical prediction model for malignant tumors in PM/DM patients using training set data. Validation set data were used to assess the feasibility of the model. The area under the receiver operating characteristic (ROC) curve (AUC), calibration curve and decision curve analysis (DCA) were used to evaluate the predictive ability, accuracy and clinical applicability of the nomogram model. Results: The age of the control group was (50.4±11.8) years, and males accounted for 26.9%(102/379); the age of the case group was (59.1±12.7) years, and the proportion of males was 56.3% (27/48). The proportion of male, age, the positive rate of anti-transcription mediator 1-γ (TIF1-γ) antibody, glucocorticoid therapy resistance, and levels of creatine kinase (CK), carbohydrate antigen 125(CA125) and carbohydrate antigen 199 (CA199) were all higher in the case group than those in control group, while incidence of interstitial lung disease (ILD), arthralgia, Raynaud's phenomenon, serum albumin (ALB) level and lymphocyte (LYM) count were all lower than those in control group (all P<0.05). Binary logistic regression analysis showed that male (OR=2.931, 95%CI: 1.356-6.335), glucocorticoid therapy resistance (OR=5.261, 95%CI: 2.212-12.513), older age (OR=1.056, 95%CI: 1.022-1.091), elevated CA125 (OR=8.327, 95%CI: 2.448-28.319) and positive anti-TIF1-γ antibody (OR=7.529, 95%CI: 2.436-23.270) were risk factors of malignancy in PM/DM patients (all P<0.05); and complicated with ILD (OR=0.261, 95%CI: 0.099-0.689), complicated with arthralgia (OR=0.238,95%CI:0.073-0.779), elevated LYM count (OR=0.267, 95%CI: 0.103-0.691) were protective factors of malignancy in PM/DM patients (all P<0.05). The AUC of ROC curve predicting malignancy in PM/DM patients with the training concentrated prediction model was 0.887 (95%CI: 0.852-0.922), with a sensitivity of 77.9% and a specificity of 86.3%; it was 0.925 (95%CI: 0.890-0.960), 86.5% and 88.0% in the validated centralized prediction model, respectively. The correction curves of the training set and the validation set indicated that the predictive model had good calibration ability. Both the DCA curves of the training set and the validation set showed that the proposed predictive model had good clinical applicability. Conclusions: Older age, male, glucocorticoid therapy resistance, not complicated with ILD and arthralgia, elevated CA125, positive anti-TIF1-γ antibody, decreased LYM count are risk factors for malignancy in PM/DM patients, and the established nomogram model shows good predictive ability.