Chronic hepatitis B (CHB), affecting approximately 254 million people worldwide and accounting for an estimated 1.1 million deaths in 2022, remains a major global health burden [
1]. The American Association for the Study of Liver Diseases (AASLD) recommends entecavir (ETV), tenofovir disoproxil fumarate (TDF), and tenofovir alafenamide (TAF) as first-line therapies because of their antiviral efficacy [
2]. However, their potential associated with neuropsychiatric adverse events, including depression and suicidal tendencies, remains incompletely characterized. Pharmacovigilance reports in other therapeutic areas have highlighted the importance of evaluating possible psychiatric adverse effects of widely used medications [
3,
4]. In this study, we used the United States Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS) to explore whether disproportionality signals for depression and suicide/self-injury (DSSI) are present for first-line antiviral therapies used in CHB.
We conducted a pharmacovigilance study utilizing data from the FAERS database, encompassing reports from Q4 2012 to Q1 2023. Data were stored in a MySQL database (version 8.0) for efficient management. Information extraction, utilizing Python (version 3.10), focused on identifying reports related to ETV, TDF, and TAF, using “primary ID” as a key identifier. We performed deduplication following the commonly used FDA case-version approach for public FAERS quarterly files. Briefly, we first removed duplicate and follow-up versions at the CASEID level by retaining the most recent report, and then removed exact duplicate PRIMARYID-preferred term (PT) rows while preserving multiple distinct PTs within the same report. DSSI related events were identified using MedDRA version 26.0 terms (see
Supplementary Methods 1,
2) [
5].
Disproportionality analyses, including proportional reporting ratios (PRR) (95% confidence interval [CI]) and reporting odds ratios (ROR) (95% CI), were conducted using R 4.3.1. For each drug-event pair, a 2×2 contingency table was constructed, where a represented reports containing both the target drug and the target adverse event, b reports containing the target drug and all other adverse events, c reports containing comparator drugs and the target adverse event, and d reports containing comparator drugs and all other adverse events. PRR was calculated as [a/(a+b)]/[c/(c+d)], and ROR as (a×d)/(b×c). The 95% CI were derived using log-transformed Wald methods (see
Supplementary Methods 1,
2 for formulas). When a zero-cell occurred, an add-one continuity correction (Laplace smoothing) was applied by adding 1 to all four cells of the contingency table.
The primary analysis was restricted to primary suspect (PS) reports. Sensitivity analyses included analyses using primary or secondary suspect (PS+SS) reports and analyses excluding reports with missing age or sex. Exploratory analyses were conducted separately for secondary suspect (SS) and concomitant (C) reports.
Because both TDF and TAF are used for CHB and human immunodeficiency virus (HIV), and FAERS does not reliably capture treatment indication, we classified reports pragmatically as likely CHB when tenofovir was used as monotherapy and likely HIV when it appeared in combination therapy. This classification was supported by the observation that the accompanying drugs in combination-therapy reports were largely HIV antiretroviral agents, although some indication misclassification may still have occurred.
Exploratory time-to-onset analyses were performed among reports with non-missing onset dates. Time-to-onset categories were compared using Fisher’s exact test, and onset distributions were summarized as median (interquartile range [IQR]) and range and compared using the Kruskal–Wallis test. Missing onset-date information was reported separately.
We identified 90,989 FAERS reports involving ETV, TDF, or TAF. After deduplication, 76,229 reports remained, including 29,307 PS reports, 25,496 SS reports, 20,328 C reports, and 1,098 interacting reports. Expanded baseline tables for all reports and PS reports are provided in
Supplementary Tables 1 and
2, including age, sex, indication, time-to-onset characteristics, and reporter profession. The top 10 reporting countries are shown in
Supplementary Figures 1 and
2. In the PS analysis, baseline characteristics including, age, sex, and time-to-onset differed across the three drug groups (
Supplementary Table 2).
Within PS reports, ETV accounted for 3,869 reports, TDF for 19,232 reports (HIV, 12,674; CHB, 6,558), and TAF for 6,206 reports (HIV, 5,534; CHB, 672). Among these, 589 reports of DSSI were identified, including 370 reports for TDF (HIV, 304; CHB, 66), 169 for TAF (HIV, 155; CHB, 14), and 50 for ETV. Among DSSI reports, the median age was 40 years (IQR, 31.0–51.0), and 315/471 (66.9%) were male; sex was missing in 118 reports. There were nine reports of completed suicide, including five reported with TAF and four reported with TDF.
In the primary PS analysis, tenofovir (TDF and TAF) showed a disproportionality signal for reports of DSSI compared with ETV (
Fig. 1A). The PRR was 1.640 (95% CI, 1.223–2.242) for DSSI and 1.144 (95% CI, 0.848–1.572) for depression alone. For suicide/self-injury (SSI), the PRR was 24.955 (95% CI, 3.495–178.169). In the CHB-restricted PS analysis, tenofovir also showed a disproportionality signal for SSI compared with ETV (PRR, 8.029; 95% CI, 1.061–60.761), whereas no clear signal was observed for DSSI and depression alone (
Fig. 1B). SSI estimates were highly imprecise because of sparse counts and zero cells in some strata, resulting in wide confidence intervals; these estimates should therefore be interpreted cautiously.
Sensitivity analyses using PS+SS reports showed generally consistent directions of association compared with the primary PS analysis (
Fig. 1C,
D). In the PS+SS analysis, the PRR was 2.184 (95% CI, 1.724–2.765) for DSSI, 1.817 (95% CI, 1.415–2.334) for depression alone, and 5.638 (95% CI, 2.662–11.941) for SSI. In the CHB subgroup of PS+SS reports, the corresponding PRRs were 1.520 (95% CI, 1.181–1.957) for DSSI, 1.442 (95% CI, 1.104–1.884) for depression alone, and 2.258 (95% CI, 1.020–4.998) for SSI. ROR estimates were consistent with PRR results.
Restriction analyses excluding reports with missing sex or missing age also showed broadly similar directions of association to the primary PS and PS+SS analysis (
Supplementary Tables 3,
4). In addition, exploratory analyses of SS reports and C reports are presented in the
Supplementary Table 5 and were interpreted cautiously because of their greater susceptibility to confounding and reporter attribution bias.
In direct comparisons within CHB reports, TAF showed disproportionality signals for both DSSI (PRR, 2.070; 95% CI, 1.068–3.748) and depression alone (PRR, 2.393; 95% CI, 1.191–4.479) compared with TDF (
Supplementary Fig. 3B). Similar patterns were observed in analyses of tenofovir overall and tenofovir in HIV reports (
Supplementary Fig. 3A,
C). We also added TAF versus TDF comparisons in the PS+SS dataset (
Supplementary Table 6), as well as separate TAF versus TDF exploratory analyses in the SS and C datasets (
Supplementary Table 7).
Our pharmacovigilance analysis identified a disproportionality signal for reports of SSI with tenofovir compared with ETV, whereas the signal for depression alone was less consistent. Importantly, the SSI estimates were highly imprecise because of sparse counts, including zero cells in some strata, resulting in extremely wide confidence intervals. These findings should therefore be interpreted cautiously as hypothesis-generating associations rather than evidence of causality.
Any biological explanation for these signals remains speculative. Prior studies have suggested that tenofovir may affect mitochondrial function, including mitochondrial DNA polymerase-γ inhibition and downstream oxidative stress pathways, which could plausibly contribute to neuropsychiatric vulnerability [
6]. In contrast, ETV displays no mitochondrial toxicities or DNA polymerase-γ inhibition, potentially accounting for the lack of reported SSI events associated with its use [
7]. However, FAERS data cannot establish mechanistic links, and the apparent differences between TAF and TDF may also reflect pharmacokinetic differences, reporting patterns, or confounding rather than true biological divergence. Further mechanistic and clinical studies are needed to clarify these possibilities. Although there is emerging evidence linking certain antiretroviral drugs to adverse neuropsychiatric outcomes [
8], current extensive clinical research comparing TAF and TDF in the treatment of CHB primarily emphasizes safety endpoints like hepatocellular carcinoma, bone, and renal events [
9,
10]. Significantly, these studies lack direct examination regarding potential psychological side effects, especially in terms of DSSI tendencies.
Several alternative explanations and biases inherent to spontaneous reporting systems may have influenced our findings. Because TAF was introduced more recently than ETV, the observed disproportionality signals may partly reflect stimulated reporting (Weber effect) and differential surveillance rather than true differences in risk. In addition, our categorization of “for CHB” versus “for HIV” based on monotherapy versus combination therapy was a pragmatic proxy and may have introduced indication misclassification, potentially biasing subgroup estimates in either direction. Although this approach is imperfect, it was not arbitrary: in TAF combination-therapy reports, the accompanying drugs were predominantly antiretroviral agents typically used for HIV treatment, supporting the use of combination therapy as a proxy for likely HIV-related use. Nevertheless, this classification may not fully capture the true indication in all cases. Because multiple contrasts were evaluated, some findings may also represent chance signals; in pharmacovigilance, disproportionality analyses are primarily intended for signal detection rather than confirmatory inference.
Additional limitations should also be acknowledged. FAERS is subject to under-reporting, selective reporting, and differential reporting by geography and reporter type, and it cannot be used to estimate event incidence. Moreover, FAERS does not reliably capture liver disease severity, prior psychiatric history, socioeconomic factors, treatment duration, or other important clinical determinants; therefore, residual confounding by indication and disease burden is likely. These factors may independently increase the likelihood of depression or SSI and may also influence reporting behavior. Onset-date fields were missing in a substantial proportion of reports, limiting the interpretability of time-to-onset analyses (
Supplementary Table 8). To improve transparency and assess robustness, we provided expanded baseline reporting characteristics and performed sensitivity analyses using alternative suspect-role definitions and restriction analyses excluding reports with missing age or sex. Overall, these analyses showed generally consistent directions of association, but they do not eliminate the possibility of residual bias.
Clinically, our findings highlight the importance of vigilant monitoring for mental health symptoms in patients undergoing antiviral therapy for hepatitis B. They also suggest a need for integrated care approaches that address both physical and mental health needs. Further research is imperative to understand the mechanistic pathways and to validate our findings in sizable, more diverse cohorts in prospective design.
In conclusion, this FAERS analysis identified a disproportionality signal for reports of DSSI with tenofovir compared with ETV, with particularly imprecise estimates for SSI. These findings are hypothesis-generating and should not be interpreted as evidence of causality. Clinician awareness and appropriate monitoring/reporting of neuropsychiatric adverse events may be considered, while confirmation in well-designed epidemiologic studies is needed.
FOOTNOTES
-
Authors’ contribution
Dr. You Deng had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Wen Xie, Chao Sun.
Acquisition, analysis, or interpretation of data: You Deng, Wenya Chen, and Weina Lu.
Drafting of the manuscript: You Deng, Wenya Chen, and Weina Lu.
Critical review of the manuscript for important intellectual content: Wen Xie, Chao Sun.
Statistical analysis: You Deng, Wenya Chen.
Obtained funding: Wen Xie.
Supervision: Chao Sun, Wen Xie.
-
Acknowledgements
The study was funded by Beijing Hospitals Authority Clinical Medicine Development of special funding support (ZLRK202334) and National Natural Science Foundation of China (No.82500714). The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
-
Conflicts of Interest
The authors disclose no conflicts.
SUPPLEMENTARY MATERIAL
Supplementary material is available at Clinical and Molecular Hepatology website (
http://www.e-cmh.org).
Supplementary Table 3.
Disproportionality analyses of depression and suicide/self-injury (DSSI) reports for tenofovir (TDF & TAF) vs. ETV after excluding reports with missing sex
Supplementary Table 4.
Disproportionality analyses of depression and suicide/self-injury (DSSI) reports for tenofovir (TDF & TAF) vs. ETV after excluding reports with missing age
Supplementary Table 5.
Disproportionality analyses of depression and suicide/self-injury (DSSI) reports for tenofovir (TDF & TAF) vs. ETV based on SS reports and C reports
Supplementary Figure 1.
Distribution of reporting countries for all FAERS reports of ETV, TDF, and TAF (top 10). FAERS, United States Food and Drug Administration Adverse Event Reporting System; ETV, entecavir; TDF, tenofovir disoproxil fumarate; TAF, tenofovir alafenamide.
Supplementary Figure 2.
Distribution of reporting countries for primary suspect (PS) FAERS reports of ETV, TDF, and TAF (top 10) FAERS, United States Food and Drug Administration Adverse Event Reporting System; ETV, entecavir; TDF, tenofovir disoproxil fumarate; TAF, tenofovir alafenamide.
Supplementary Figure 3.
Disproportionality analyses of depression and suicide/self-injury event reports for TAF vs. TDF based on PS report. (A) Disproportionality analyses using proportional reporting ratio (PRR) and reporting odds ratio (ROR) for TAF (total) vs. TDF (total) in PS reports. (B) Disproportionality analyses using PRR and ROR for TAF (for CHB) vs. TDF (for CHB) in PS reports. (C) Disproportionality analyses using PRR and ROR for TAF (for HIV) vs. TDF (for HIV) in PS reports. PS, primary suspect; DSSI, depression and suicide/ self-injury; TAF, tenofovir alafenamide; TDF, tenofovir disoproxil fumarate; CI, confidence interval; CHB, chronic hepatitis B; HIV, human immunodeficiency virus.
Figure 1.Disproportionality analyses of depression and suicide/self-injury (DSSI) event reports for tenofovir (TDF & TAF) vs. ETV. (A) Disproportionality analyses using proportional reporting ratio (PRR) and reporting odds ratio (ROR) for TDF & TAF (total) vs. ETV in primary suspect (PS) reports. (B) Disproportionality analyses using PRR and ROR for TDF & TAF (CHB) vs. ETV in PS reports. (C) Disproportionality analyses using PRR and ROR for TDF & TAF (total) vs. ETV in primary or secondary suspect (PS+SS) reports. (D) Disproportionality analyses using PRR and ROR for TDF & TAF (CHB) vs. ETV in PS+SS reports. TDF, tenofovir disoproxil fumarate; TAF, tenofovir alafenamide; ETV, entecavir; CI, confidence interval; CHB, chronic hepatitis B.
Abbreviations
The American Association for the Study of Liver Diseases
depression and suicide/ self-injury
FDA Adverse Event Reporting System
United States Food and Drug Administration
human immunodeficiency virus
proportional reporting ratios
primary or secondary suspect
tenofovir disoproxil fumarate
REFERENCES
- 1. World Health Organization. Hepatitis B. World Health Organization web site, <https://www.who.int/news-room/fact-sheets/detail/hepatitis-b>. Accessed 28 Feb 2026.
- 2. Ghany MG, Pan CQ, Lok AS, Feld JJ, Lim JK, Wang SH, et al. AASLD ISDA practice guideline on treatment of chronic hepatitis B. Hepatology 2026;83:974-997.
- 3. Zhou C, Peng S, Lin A, Jiang A, Peng Y, Gu T, et al. Psychiatric disorders associated with immune checkpoint inhibitors: a pharmacovigilance analysis of the FDA Adverse Event Reporting System (FAERS) database. EClinicalMedicine 2023;59:101967.
- 4. Wang W, Volkow ND, Berger NA, Davis PB, Kaelber DC, Xu R. Association of semaglutide with risk of suicidal ideation in a real-world cohort. Nat Med 2024;30:168-176.
- 5. MedDRA Maintenance and Support Services Organization. Introductory guide for standardised MedDRA queries (SMQs) version 26.0. International Council for Harmonisation website, <https://admin.new.meddra.org/sites/default/files/guidance/file/SMQ_intguide_26_0_English.pdf>. Accessed 28 Feb, 2026.
- 6. Apostolova N, Blas-García A, Esplugues JV. Mitochondrial interference by anti-HIV drugs: mechanisms beyond Pol-γ inhibition. Trends Pharmacol Sci 2011;32:715-725.
- 7. Mazzucco CE, Hamatake RK, Colonno RJ, Tenney DJ. Entecavir for treatment of hepatitis B virus displays no in vitro mitochondrial toxicity or DNA polymerase gamma inhibition. Antimicrob Agents Chemother 2008;52:598-605.
- 8. Ciccarelli N, Fabbiani M, Di Giambenedetto S, Fanti I, Baldonero E, Bracciale L, et al. Efavirenz associated with cognitive disorders in otherwise asymptomatic HIV-infected patients. Neurology 2011;76:1403-1409.
- 9. Agarwal K, Brunetto M, Seto WK, Lim YS, Fung S, Marcellin P, et al. 96 weeks treatment of tenofovir alafenamide vs. tenofovir disoproxil fumarate for hepatitis B virus infection. J Hepatol 2018;68:672-681.
- 10. Lim YS, Chan HLY, Ahn SH, Seto WK, Ning Q, Agarwal K, et al. Tenofovir alafenamide and tenofovir disoproxil fumarate reduce incidence of hepatocellular carcinoma in patients with chronic hepatitis B. JHEP Rep 2023;5:100847.
