A predictive nomogram for short-term outcomes of myasthenia gravis patients treated with low-dose rituximab

BackgroundThis study aims to establish and validate a predictive nomogram for the short-term clinical outcomes of myasthenia gravis (MG) patients treated with low-dose rituximab.MethodsWe retrospectively reviewed 108 patients who received rituximab of 600 mg every 6 months in Huashan Hospital and Tangdu Hospital. Of them, 76 patients from Huashan Hospital were included in the derivation cohort to develop the predictive nomogram, which was externally validated using 32 patients from Tangdu Hospital. The clinical response is defined as a >= 3 points decrease in QMG score within 6 months. Both clinical and genetic characteristics were included to screen predictors via multivariate logistic regression. Discrimination and calibration were measured by the area under the receiver operating characteristic curve (AUC-ROC) and Hosmer-Lemeshow test, respectively.ResultsDisease duration (OR = 0.987, p = 0.032), positive anti-muscle-specific tyrosine kinase antibodies (OR = 19.8, p = 0.007), and genotypes in FCGR2A rs1801274 (AG: OR = 0.131, p = 0.024;GG:OR = 0.037, p = 0.010) were independently associated with clinical response of post-rituximab patients. The nomogram identified MG patients with clinical response with an AUC-ROC (95% CI) of 0.875 (0.798-0.952) in the derivation cohort and 0.741(0.501-0.982) in the validation cohort. Hosmer-Lemeshow test showed a good calibration (derivation: Chi-square = 3.181, p = 0.923; validation: Chi-square = 8.098, p = 0.424).ConclusionsThe nomogram achieved an optimal prediction of short-term outcomes in patients treated with low-dose rituximab. Our study developed and validated the clinical response prediction nomogram for myasthenia gravis patients treated with low-dose rituximab. The nomogram can effectively identify patients with clinical response, help make treatment-related decisions, alleviate the financial burden on patients, and rationalize the allocation of healthcare resources.image