Construction of an easy‐to‐use predictive model for ultrasound‐detected tophi to improve the detection of hidden tophi

Abstract Background Musculoskeletal ultrasound is used in clinical practice to evaluate gout patients and is an effective imaging tool for the detection of tophi. The aim of this study was to analyze the factors associated with ultrasound‐detected tophi in gout patients and to construct a clinical model to predict its occurrence and improve the detection of hidden tophi. Methods Data of gout patients admitted to Beijing Jishuitan Hospital from January 2015 to December 2021 were collected. The complete and detailed information from gout cases with completed musculoskeletal ultrasound was included in the analysis. Univariate and multivariate analyses were used to identify independent factors associated with ultrasound‐detected tophi. A nomogram was used to visualize the clinical predictive models. Results Among 517 gout patients, rheumatologists found that 67 patients (13.0%) had subcutaneous tophi by visual observation, while musculoskeletal ultrasound revealed that 123 patients (23.8%) had ultrasound‐detected tophi with odds ratio [OR] (95% confidence intervals [CIs]) = 2.20 (1.81–2.67). Disease duration, upper limb joint flare (ULJF), persistent joint pain (PJP), uric acid, and homocysteine levels were independently associated with ultrasound‐detected tophi, and they had ORs (95% CIs) of 1.092 (1.050–1.136), 3.732 (2.312–6.025), 1.864 (1.086–3.200), 1.003 (1.001–1.004), and 1.015 (1.000–1.030), respectively. After balancing the complexity and accuracy of the model, Model 2 (incorporating disease duration, ULJF, PJP, and uric acid) was chosen to create a nomogram to predict the occurrence of ultrasound‐detected tophi. The nomogram had good discrimination (consistency index [C‐index] = 0.774) and excellent calibration, demonstrated by calibration curves. Conclusion Using easily available indicators, such as disease duration, the nature of the joint pain, and uric acid levels, we successfully developed an easy‐to‐use clinical model to improve the detection of hidden tophi.