Variable Selection and Model Development
Together with previous clinical researches, according to the results of the LASSO regression model and the multivariable logistic regression model, nine of the 22 variables were selected as the potential predictors of tophi formation (Figures 1A and 1B ). These variables were hospitalisation frequency, disease duration, number of joints involved in gouty arthritis, gout flares frequency, smoking, and whether combined with atherosclerosis, diabetes, hypertension and kidney dysfunction. The logistic regression analysis results of these variables are shown in Table 2 . A prediction model that contains the above predictors was then developed and represented as a nomogram (Figure 2 ).