Abstract
PURPOSE: Peginterferon (Peg-IFN) is a common treatment for chronic hepatitis B (CHB); however, some patients developing myelosuppression as a side-effect. In this study, we identified risk factors associated with increased myelosuppression, and incorporated them into a predictive nomogram. PATIENTS AND METHODS: This study is designed as a case-control study. A total of 312 CHB patients treated with Peg-IFN from two medical centers were retrospectively enrolled between December 2019 and December 2022. Patients from the First Affiliated Hospital of Nanchang University were randomly divided into a training cohort (n=153) and a test cohort (n=55) at a 3:1 ratio. Patients from the Jiangxi Provincial People's Hospital composed the validation cohort (n= 104). In the training cohort, based on the blood routine results of patients 1 week after Peg-IFN treatment, patients were further divided into Normal (myelosuppression grades 0-I) and Myelosuppression (grades II-IV) groups. Then uni- and multivariate logistic regression analyses were carried out to identify myelosuppression risk factors, which were subsequently incorporated into a predictive nomogram. The capability of the predictive nomogram was validated using an area under the curve (AUC) of the receiver operating characteristic (ROC) curve. The Hosmer-Lemeshow test, calibration curves, and decision curve analysis (DCA) were used to evaluate the nomogram. Finally, the developed predictive nomogram was validated both internally and externally using separate test and validation cohorts. RESULTS: Body mass index (BMI; odds ratio [OR]=0.841, 95% confidence interval [CI] 0.738-0.959, P=0.010), white blood cell counts (WBC; OR=0.657, 95% CI 0.497-0.868, P=0.003), globulin (GLB; OR=0.796, 95% CI 0.713-0.889, P<0.001) and serum creatinine levels (SCR; OR=1.029, 95% CI 1.002-1.058, P=0.038) are independent risk factors for myelosuppression in Peg-IFN-treated CHB patients. A predictive nomogram was constructed by incorporating the above independent risk factors, and its performance was assessed across the training, test, and validation cohorts. The model demonstrated AUC values of 0.824 (95% CI 0.757-0.891), 0.812 (95% CI 0.701-0.923), and 0.870 (95% CI 0.802-0.940), respectively, highlighting its good predictive accuracy. As for Hosmer-Lemeshow, it was P=0.351, (χ2= 8.898) for training, P=0.514 (χ2=6.226) for the test, and P=0.442 (χ2=7.918) for the validation cohort. The results of the calibration curves and DCA demonstrated good concordance between predicted probabilities and observed outcomes, with the model showing higher clinical net benefit. CONCLUSION: Lower BMI, WBC counts, GLB, and higher SCR levels are independent risk factors for myelosuppression among Peg-IFN-treated CHB patients. The predictive nomogram, based on those factors, is able to identify high-risk individuals for myelosuppression, thereby aiding in early alleviation of this side-effect.