Scholarlink

Background and objective <p>Pegylated interferon (PEG-IFN) has been shown to significantly enhance the clinical cure rate in chronic hepatitis B (CHB) patients. This study aims to utilize machine learning to select optimal patient populations for PEG-IFN therapy and construct the nomogram model for predicting interferon response (IR).</p> Methods <p>Between April 2020 and October 2023, we collected data from CHB patients treated with PEG-INFα-2b at the Affiliated Hospital of Xuzhou Medical University. Based on HBeAg status, IR is defined as follows: For HBeAg-positive patients, an IR is defined as achieving HBeAg negativity, HBV DNA negativity, and HBsAg levels below 100&#xa0;IU/mL after 48&#xa0;weeks of treatment with pegylated interferon; For HBeAg-negative patients, an IR is achieved when HBV DNA becomes undetectable and HBsAg is negative after 48&#xa0;weeks of pegylated interferon treatment. Through a combination of LASSO COX regression, Random Survival Forests, and CoxBoost, we identified optimal patient populations. A multivariable COX regression was then employed to construct the nomogram model. The performance of the nomogram model was assessed using Receiver Operating Characteristic (ROC) curves, decision curves, and calibration curves.</p> Results <p>Utilizing machine learning, we pinpointed baseline HBsAg, △HBsAg, and △ALT as predictors of IR. Leveraging these key predictors, we developed a nomogram model that showcased robust accuracy in predicting IR at 48&#xa0;weeks across different datasets: the training cohort, validation cohort, and RNA cohort, with AUROC scores of 0.922 (95% CI: 0.88–0.96), 0.938 (95% CI: 0.87–1.00), and 0.933 (95% CI: 0.87–1.00), respectively. The results of the calibration and decision curves show the model has a good consistency and precision. Furthermore, Kaplan–Meier analysis identified that CHB patients with baseline HBsAg &lt; 2.43 lg IU/mL, △HBsAg &gt; 1.26 lg IU/mL, and △ALT &gt; 8.00 U/L were more likely to achieve IR.</p> Conclusion <p>The nomogram model can evaluate the efficacy of PEG-IFN in the treatment of CHB patients, and is helpful for clinical decision-making.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Leveraging machine learning to identify optimal patient populations for PEG-IFN therapy in CHB and constructing nomogram models of interferon response: a 48-week follow-up study

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月10日

查看全文
  • Deyang Xi,
  • Ke Xu,
  • Xuebing Yan,
  • Chunyang Li

摘要

Background and objective

Pegylated interferon (PEG-IFN) has been shown to significantly enhance the clinical cure rate in chronic hepatitis B (CHB) patients. This study aims to utilize machine learning to select optimal patient populations for PEG-IFN therapy and construct the nomogram model for predicting interferon response (IR).

Methods

Between April 2020 and October 2023, we collected data from CHB patients treated with PEG-INFα-2b at the Affiliated Hospital of Xuzhou Medical University. Based on HBeAg status, IR is defined as follows: For HBeAg-positive patients, an IR is defined as achieving HBeAg negativity, HBV DNA negativity, and HBsAg levels below 100 IU/mL after 48 weeks of treatment with pegylated interferon; For HBeAg-negative patients, an IR is achieved when HBV DNA becomes undetectable and HBsAg is negative after 48 weeks of pegylated interferon treatment. Through a combination of LASSO COX regression, Random Survival Forests, and CoxBoost, we identified optimal patient populations. A multivariable COX regression was then employed to construct the nomogram model. The performance of the nomogram model was assessed using Receiver Operating Characteristic (ROC) curves, decision curves, and calibration curves.

Results

Utilizing machine learning, we pinpointed baseline HBsAg, △HBsAg, and △ALT as predictors of IR. Leveraging these key predictors, we developed a nomogram model that showcased robust accuracy in predicting IR at 48 weeks across different datasets: the training cohort, validation cohort, and RNA cohort, with AUROC scores of 0.922 (95% CI: 0.88–0.96), 0.938 (95% CI: 0.87–1.00), and 0.933 (95% CI: 0.87–1.00), respectively. The results of the calibration and decision curves show the model has a good consistency and precision. Furthermore, Kaplan–Meier analysis identified that CHB patients with baseline HBsAg < 2.43 lg IU/mL, △HBsAg > 1.26 lg IU/mL, and △ALT > 8.00 U/L were more likely to achieve IR.

Conclusion

The nomogram model can evaluate the efficacy of PEG-IFN in the treatment of CHB patients, and is helpful for clinical decision-making.