Related Experiment Videos
A nomogram based on LASSO-logistic regression for predicting tocilizumab-associated hypofibrinogenemia in patients
Chunmei Dai1,2, Yiting Wang1,2, Yilei Chen3
1Department of Pharmacy, The Sixth Affiliated Hospital, Sun Yat-Sen University, Guangzhou, Guangdong, China.
Introduction:
Tocilizumab (TCZ), an IL-6 receptor inhibitor used in rheumatic diseases, can suppress fibrinogen synthesis and cause hypofibrinogenemia, yet its incidence, risk factors and predictive tools remain incompletely characterized. This study determined the incidence of TCZ-associated hypofibrinogenemia and developed a nomogram-based prediction model for individualized risk stratification.
Methods:
This retrospective study was conducted at the Sixth Affiliated Hospital of Sun Yat-sen University, enrolling 121 patients with rheumatic diseases who received TCZ between 1 January 2019 and 31 December 2025 and had at least one fibrinogen measurement. Hypofibrinogenemia was defined as a plasma fibrinogen concentration below 2.0 g/L at any point during treatment. Hypofibrinogenemia was defined as a plasma fibrinogen concentration below 2.0 g/L at any point during treatment. All candidate predictors were entered into Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression with 10-fold cross-validation under the λ.1se criterion, and the retained candidates were examined in multivariable logistic regression. The resulting coefficients were formulated into a nomogram, whose performance was assessed in terms of discrimination by the area under the receiver operating characteristic curve (AUC), calibration by the Hosmer-Lemeshow test and bias-corrected calibration curves, internal validity through 1000-replicate bootstrap resampling, clinical utility via decision curve analysis and clinical impact curves, and predictor-level interpretability through SHapley Additive exPlanations (SHAP) decomposition.
Results:
Hypofibrinogenemia occurred in 39 of 121 patients (32.23%). Independent predictors were higher BMI, lower IgA, lower platelet count, and treatment with Actemra® rather than Tofidence®. The nomogram showed acceptable discrimination (AUC = 0.795, 95% CI: 0.716-0.873; bootstrap-corrected AUC = 0.782, 95% CI: 0.751-0.795), good calibration (Hosmer-Lemeshow P = 0.835; Brier score = 0.170), and positive net benefit on decision curve analysis. SHAP analysis identified platelet count as the dominant contributor, followed by TCZ brand, IgA and BMI.
Conclusion:
Hypofibrinogenemia occurred in one-third of TCZ-treated rheumatic patients. Higher BMI, lower IgA, lower platelet count and use of the Actemra® were independently associated with increased risk. The nomogram offers a preliminary tool for risk stratification, but requires external validation before clinical use.
