Predictive model to identify patients requiring extended interferon therapy in chronic hepatitis B

Abstract

A contemporary study by Yan et al, published in the World Journal of Gastroenterology, demonstrates that baseline hepatitis B surface antigen and cirrhosis can be utilized in a prediction model to identify populations that may benefit from extended (≥ 48 weeks) interferon therapy. In this editorial, we summarize strengths of their study. First, the follow-up period was sufficiently long; second, the authors identified all major confounding factors; third, appropriate statistical methods were employed; fourth, the figures and tables are highly legible; and fifth, the training cohort exhibited stable discriminatory ability. The limitations of the study are as follows: First, the sample size in certain subgroups was too small; second, nearly all similar studies have been conducted in China, leaving readers with a limited understanding of global trends in this field; and third, the benefits of adding PEGylated interferon α-2b to nucleos(t)ide analog therapy were not addressed. Future research should investigate the promising role of combination therapies involving PEGylated interferon α-2b and nucleos(t)ide analogs, incorporate machine learning approaches, and consider the effects of concurrent nonalcoholic fatty liver disease.

Key Words: Chronic hepatitis B; Cirrhosis; Hepatitis B surface antigen; Nucleos(t)ide analog; Pegylated interferon α-2b

Core Tip: Yan et al published a study in the World Journal of Gastroenterology regarding interferon therapy for chronic hepatitis B clearance that warrants broad attention. In this editorial, we review their work and contextualize their findings within the scope of similar English-language publications in PubMed. A novel finding of their study is that stable discriminatory ability was noted for both the training cohort (area under the receiver operating characteristic curve = 0.83) and the validation cohort (area under the receiver operating characteristic curve = 0.81), with predictive efficacy unaffected by subgroup characteristics. We also discuss the limitations of their study and address its potential clinical implications.

This editorial refers to "Baseline hepatitis B surface antigen and cirrhosis predict extended interferon therapy in chronic hepatitis B: A retrospective study" by Yan et al, 2026; https://doi.org/10.3748/wjg.v32.i12.116287.

---

INTRODUCTION

In 2022, chronic hepatitis B virus (HBV) infection (CHB) was estimated to affect 257.5 million people worldwide and contributed to 820000 deaths annually due to cirrhosis and hepatocellular carcinoma[1]. In particular, 50% of all deaths due to hepatocellular carcinoma were reported to be linked to chronic HBV infection[1]. Therefore, HBV infection remains a significant global health challenge. The number of HBV infections was estimated to be 86 million cases in China, remaining a substantial burden to the health-care system[2]. In China, hepatitis B surface antigen (HBsAg) clearance through PEGylated interferon α-2b (PEG-IFNα-2b) therapy is recognized as the goal of contemporary hepatitis B treatment[2].

Although PEG-IFNα-2b therapy generally yields superior HBsAg loss rates compared with nucleos(t)ide analog (NA) therapy, its clinical efficacy exhibits significant variability. These associated disparities in therapeutic outcomes are primarily attributable to the heterogeneity of patient characteristics and the diversity of administered treatment regimens[1]. A contemporary study by Yan et al[2], published in the World Journal of Gastroenterology, addresses this variability by employing a predictive model to identify patients with CHB who may require extended interferon therapy. To contextualize this research, we reviewed similar articles that have been published on PubMed and wrote this editorial. Specifically, we reviewed published English-language full papers, excluding review articles and meta-analyses, and identified 37 similar articles published between January 1, 2021, and December 31, 2025. Excluding reviews and meta-analyses, we identified 37 relevant studies published in the 5 years[3-39] preceding the work by Yan et al[2].

STRENGTHS OF THE STUDY OF YAN ET AL

As indicated in Table 1, among the 39 articles, comprising 38 similar studies and the report by Yan et al[2], various predictors for PEG-IFN treatment response in CHB were identified. Notably, while baseline HBsAg, a key component of the predictive model developed by Yan et al[2], was cited in 20 publications, cirrhosis was utilized as a predictor in only two studies. In the model established by Yan et al[2], cirrhosis was identified as a critical parameter for predicting the necessity of extending therapy beyond 48 weeks. Specifically, Yan et al[2] demonstrated that the extended-course group exhibited significantly higher baseline HBsAg levels (2.75 ± 0.97 Log10 IU/mL vs 1.35 ± 1.20 Log10 IU/mL, P < 0.01) and a higher prevalence of cirrhosis (27.78% vs 10.00%, P < 0.05) compared to the standard-course group (Table 2 in[2]). Although the study by Tan et al[21] also incorporated cirrhosis, their research primarily focused on factors influencing HBsAg clearance at week 48. Consequently, the work of Yan et al[2] stands out as the only study among these 39 articles to utilize cirrhosis as a predictive parameter specifically for determining the extension of treatment duration.

Table 1 Comparison of predictors for PEGylated interferon treatment response in chronic hepatitis B between the study by Yan et al[2] and 37 similar studies published from 2021 to 2025.

Category	Predictors	n1	Ref.
Hepatitis B markers	HBsAg	20	[1,2,6,9,10,12,14,17-21,25,27,30,31,35,36,38,39]
	HBeAg	2	[1,13]
	Anti-HBs	2	[13,29]
	Anti-HBc	1	[17]
Biomarkers	Immunoglobulins	1	[4]
	Lipid	1	[15]
	Ferritin	1	[26]
	ALT	5	[27,30,31,36,39]
Patient characteristics	Cirrhosis	2	[2,21]
	Fatty liver	1	[12]
	Gender	1	[21]
	Age	3	[27,30,31]
mRNA	TRIM19/38	1	[8]
	LY6E/TRIM6	1	[23]
Immune-related markers	sPD-1/sPD-L1	1	[11]

¹The number includes the study of Yan et al[2].

TRIM19/38: Triple motif protein 19/38 mRNA; LY6E: Lymphocyte antigen 6 family member E; TRIM6: Triple motif protein 6; sPD-1: Soluble programmed cell death-1; sPD-L1: Soluble programmed cell death ligand-1; HBsAg: Hepatitis B surface antigen; HBeAg: Hepatitis B e antigen; anti-HBs: Anti-hepatitis B surface antibody; anti-HBc: Hepatitis B core antibody; ALT: Alanine aminotransferase.

Table 2 Number of related studies published each year.

Publication year	2021	2022	2023	2024	2025
n1	1	3	2	10	22
Ref.	[3]	[4-6]	[7,8]	[9-18]	[2,19-39]

¹The number includes the study of Yan et al[2].

The study by Yan et al[2] has several other strengths, including: (1) An extensive follow-up period, spanning January 2018 to March 2024; (2) Thorough control of confounding factors through exclusion of individuals with coinfections (hepatitis A, C, D, or E), individuals who interrupted PEG-IFNα-2b treatment for more than 3 months, and individuals with incomplete data; (3) The use of appropriate statistical methods, such as the t-test, Kruskal-Wallis one-way analysis of variance, χ² test, and Fisher’s exact test, ensuring reliable results; (4) A clear presentation of data, including a well-designed patient enrollment flow diagram and accurately rendered receiver operating characteristic and calibration curves; and (5) Stable discriminatory performance in both training [area under the receiver operating characteristic curve (AUC) = 0.83] and validation (AUC = 0.81) cohorts. Furthermore, their findings indicated that the predictive efficacy of the model was not affected by subgroup characteristics: No significant interactions were observed regarding sex, age, baseline HBV DNA, or baseline alanine aminotransferase levels (P > 0.05; Figure 6 in[2]). This evidence of a stable, independent correlation was not reported in the other articles that used AUC analysis[6,10,11,14,15,19,21,23,25,27,29,30,33-36,39], indicating it is a novel finding.

LIMITATIONS OF THE STUDY OF YAN ET AL

The study by Yan et al[2] has four key limitations: (1) The sample size in certain subgroups is quite small. In the analysis of factors associated with PEG-IFNα-2b treatment, the observed statistical significance (P = 0.056) of 12-week HBsAg (log10 IU/mL) therapy (Table 3 in[2]) may not be accurate given the limited sample size; further verification in expanded studies is required. Similarly, the small number of patients with cirrhosis (fewer than 5) may affect the reliability of the findings regarding cirrhosis as a predictive factor for extending therapy (Figure 6 in[2]); (2) Potential selection bias and a lack of prospective validation may affect the results; (3) As presented in Table 2, the number of similar studies has increased from 1 in 2021 to 10 in 2024 and 22 in 2025. However, nearly all of these studies originated in China (37 of 38)[2-15,17-39] (Table 3); only one study was conducted in another country, South Korea[16], despite earlier contributions from Taiwanese and Korean researchers prior to 2010 and 2016, respectively[40,41]. Providing reasons for this geographic concentration would help readers better understand trends in the field; and (4) The benefits of adding PEG-IFNα-2b to NA therapies were not addressed: Before the manuscript was revised on December 25, 2025, at least three relevant articles regarding such predictive models had been published[42-44] (two from China[42,43] and one from the Netherlands[44]). However, Yan et al[2] did not incorporate this information into their discussion.

Table 3 Locations of relevant studies.

Study subject	Chinese	Taiwanese	Koreans	White men	Other ethnics
n1	371	0	1	0	0
Ref.	[2-15,17-39]		[16]

¹The number includes the study of Yan et al[2].

POTENTIAL CLINICAL IMPLICATIONS

(1) As mentioned earlier in this editorial, patients are likely to benefit from the addition of PEG-IFNα-2b to NA therapy. Consequently, continued research into combination therapies involving PEG-IFNα-2b and NAs remains a promising area of study[45,46]; (2) Regarding machine learning, researchers must construct and validate models that support early clinical decision-making. Developing user-friendly clinical tools will assist in providing personalized medicine and tailoring treatment decisions[19,27,34,36,39]; and (3) The prevalence of metabolic dysfunction-associated steatotic liver disease (MASLD) has been increasing rapidly, and MASLD has become a major contributor to chronic liver disease complications[28,47]. Because effective drug therapies for MASLD are currently lacking, clinicians must maximize HBsAg loss through anti-HBV therapy (PEG-IFNα-2b and antiviral drugs) and minimize liver damage to reduce the risk of end-stage liver disease in patients with concurrent CHB and MASLD.

CONCLUSION

Yan et al[2] demonstrates that baseline HBsAg levels and the presence of cirrhosis can be used in a prediction model to identify patients who may benefit from extended interferon therapy. However, several limitations of the study must be considered: The sample size in certain subgroups is notably small, and nearly all similar studies were performed in China. Furthermore, potential selection bias, a lack of prospective validation, and a failure to add PEG-IFNα-2b to NA therapy for CHB treatment may influence the overall findings.