Global safety profile of PD-1/PD-L1 inhibitors in hepatic autoimmune disorders: A global disproportionality analysis

Abstract

The increasing use of programmed cell death protein-1 (PD-1) and programmed death-ligand 1 (PD-L1) inhibitors in cancer therapy has raised concerns regarding immune-related adverse events. However, global pharmacovigilance studies on autoimmune diseases remain scarce. This study aimed to evaluate the signal detection between PD-1/PD-L1 inhibitors and hepatic autoimmune disorders. A global pharmacovigilance study was conducted using data from individual case safety reports. The analysis centered on PD-1/PD-L1 inhibitors, classified by Anatomical Therapeutic Chemical codes (e.g., atezolizumab, avelumab, budigalimab, cemiplimab, dostarlimab, durvalumab, nivolumab, pembrolizumab, sintilimab, spartalizumab, and tislelizumab). Hepatic autoimmune disorders were defined based on MedDRA Version 26.0 (Maintenance and Support Services Organization, McLean). Disproportionality analyses were performed using reporting odds ratios with 95% confidence intervals and information components (ICs), with IC_0.25, to evaluate potential signal detection. A total of 1401 reports of hepatic autoimmune disorders associated with PD-1/PD-L1 inhibitors were identified. Nivolumab (59.89%) and pembrolizumab (27.62%) accounted for the highest reports. Notably, nivolumab (reporting odds ratio, 117.52 [95% confidence interval, 109.52–126.10]; IC, 6.67 [IC_0.25, 6.60]) and pembrolizumab (55.69 [50.31–61.64]; 5.65 [5.55]) showed the notable reporting signals, followed by cemiplimab, atezolizumab, durvalumab, avelumab, sintilimab, and tislelizumab. Subgroup analysis showed a stronger signal for hepatic autoimmune disorders in males than in females treated with PD-1/PD-L1 inhibitors. Additionally, the mean time to onset was 22.90 days (standard deviation: 81.66), although some reports presented with substantially delayed onset. Most PD-1/PD-L1 inhibitors showed pharmacovigilance signals for hepatic autoimmune disorders, particularly nivolumab and pembrolizumab. Although our findings do not permit causal inference, these findings underscore the necessity for sustained hepatic monitoring, risk stratification, and appropriate therapeutic management.

1. Introduction

Chronic immune-mediated hepatic disease is a severe liver disorder that affects both pediatric and adult populations worldwide.^[1] Recent studies have reported an increasing incidence and prevalence of this condition.^[2] Owing to its significant impact on the quality of life, and its association with liver-related morbidity and mortality, lifelong immunosuppressive therapy is often required.

Immune checkpoint inhibitors (ICIs), particularly those targeting programmed cell death protein-1 (PD-1) and programmed death-ligand 1 (PD-L1), have revolutionized cancer treatment by enhancing immune-mediated tumor suppression.^[3] However, as their clinical use continues to expand, concerns about immune-related adverse events resulting from unintended immune activation have grown.^[4] These adverse events impose a substantial financial burden on healthcare systems due to the need for ongoing treatment and monitoring, while also increasing the risk of hepatic failure and mortality in severe cases.^[5] Notably, individuals with underlying autoimmune diseases may be at a higher risk of developing hepatic dysfunction.

Recent studies indicate that PD-1/PD-L1 inhibitors can induce hepatic inflammation resembling autoimmune hepatitis, characterized by elevated aminotransferase levels, histopathological features of interface hepatitis, and T-cell infiltration.^[6] While previous studies have identified immune-related adverse events associated with PD-1/PD-L1 inhibitors, large-scale pharmacovigilance studies remain limited. Notably, no comprehensive analysis has been conducted to specifically assess the signal detections between PD-1/PD-L1 inhibitors and hepatic autoimmune disorders.

To address this gap, this study utilizes global pharmacovigilance to investigate the relationship between PD-1/PD-L1 inhibitors and hepatic autoimmune disorders. This analysis aims to advance our understanding of the hepatic safety profile of PD-1/PD-L1 inhibitors in patients with a predisposition to autoimmune diseases, providing valuable insights and practical implications for clinicians, regulatory authorities, and researchers.

2. Methods

2.1. Data sources

VigiBase, a global drug safety surveillance database maintained by the World Health Organization and the Uppsala Monitoring Centre, has been continuously updated since 1968 with reports submitted by member countries participating in the World Health Organization International Drug Monitoring Program.^[7] It serves as a comprehensive pharmacovigilance resource, integrating both spontaneous adverse drug reaction reports and clinical data. The global pharmacovigilance database contains over 35 million individual case safety reports from nearly 140 countries.^[8] Fundamentally, the global pharmacovigilance database does not contain personally identifiable information, ensuring compliance with research ethics while safeguarding patient confidentiality. Consequently, the requirement for informed consent was waived. This study was conducted using fully anonymized patient data in accordance with the ethical guidelines of the Institutional Review Board of Kyung Hee University.

2.2. Selection of reports

All reports from the inception of global pharmacovigilance through June 2024 were included in the analysis. Additionally, all preferred terms associated with hepatic autoimmune disorders were considered (Table S1, Supplemental Digital Content, https://links.lww.com/MD/Q64),^[9] as defined by the Medical Dictionary for Regulatory Activities (MedDRA), version 26.0 (Maintenance and Support Services Organization). Only drugs classified as “suspected” were analyzed, ensuring that hepatic autoimmune disorder reports identified using MedDRA codes reflect clinical significance.

To enhance the clinical reliability and consistency of drug categorization, the analysis was restricted to drugs classified under L01FF in the Anatomical Therapeutic Chemical classification system, and a separate analysis was performed. The “L01FF” category includes the following PD-1/PD-L1 inhibitors: atezolizumab, avelumab, budigalimab, cemiplimab, dostarlimab, durvalumab, nivolumab, pembrolizumab, sintilimab, spartalizumab, and tislelizumab.^[10] Time trend analyses were limited to those drugs for which the number of reports was considered adequate to ensure statistical robustness and analytical stability. In contrast, budigalimab, dostarlimab, and spartalizumab had only a single report each; therefore, while they were included in the basic descriptive analyses, they were excluded from the time trend analyses due to insufficient data for reliable signal detection.^[11]

Furthermore, trends in the information component (IC) and its 95% credibility interval lower endpoint (IC_0.25), along with the 95% confidence interval (95% CI) for autoimmune diseases associated with the 2 most frequently reported drugs, were analyzed. These findings provide insights for optimizing future prescribing practices and enhancing patient management strategies.

2.3. Data collection

Data were categorized by region, reporting period, reporter type (healthcare professional, non-healthcare professional, or unknown), sex, age group, time to onset (TTO), mortality outcome, and comprehensive details on adverse events.^[8] Additionally, TTO was determined in days and calculated as the difference between the treatment start date and the adverse event date (TTO = adverse event onset date – drug start date).

2.4. Statistical analysis

The disproportionality analysis conducted using the global pharmacovigilance database utilized ICs with IC_0.25 and reporting odds ratios (RORs) to evaluate signals. RORs and their corresponding 95% CIs, commonly used in case-control studies, were applied to assess the strength of signals by analyzing reports of hepatic autoimmune disorders within the global pharmacovigilance database.

Additionally, to further enhance the reliability and clarity of the signals between drugs and adverse drug reactions, the IC – a Bayesian disproportionality measure widely used in drug safety signal detection within global pharmacovigilance – was utilized. According to standard IC analysis, an event was considered disproportionately overreported if the lower bound of the 95% CI for the IC value was positive (IC_0.25 > 0.00).^[12]

To enhance the accuracy of signal detection and ensure consistency in interpretation, both the ROR and the IC were applied in parallel in this study. These 2 metrics offer complementary statistical perspectives, thereby addressing the limitations inherent in relying on a single analytical framework. This dual-metric strategy is particularly valuable in real-world pharmacovigilance settings, where data may be sparse or heterogeneous. Notably, this approach has been consistently employed in prior studies utilizing global pharmacovigilance databases.^[13]

All statistical analyses were performed using the SAS software (version 9.4; SAS Institute, Cary). A two-sided P-value < .05 was considered statistically significant.

3. Results

3.1. Overall analysis

Figure 1 presents the process of detecting potential safety signals of hepatic autoimmune disorders in reports involving PD-1/PD-L1 inhibitors. A total of 1401 related reports were identified, including 775 reports (55.32%) in males and 556 (39.69%) in females. The majority of reported reports originated from the European region (66.38%), followed by the Americas (21.20%) and the Western Pacific region (12.21%; Table 1).

Figure 1.

Flowchart illustrating the selection process for events of hepatic autoimmune disorders induced by PD-1/PD-L1 inhibitors based on global pharmacovigilance. PD-1/PD-L1 = programmed cell death protein 1/death ligand 1.

Table 1.

Baseline characteristics of reports on events of drug-induced hepatic autoimmune disorders.

Variables	Number (%)
Reporting region
African region	N/A
Region of the Americas	297 (21.20)
Southeast Asia region	N/A
European region	930 (66.38)
Eastern Mediterranean region	3 (0.21)
Western Pacific region	171 (12.21)
Reporting year
1968–1979	N/A
1980–1989	N/A
1990–1999	N/A
2000–2009	N/A
2010–2019	565 (40.33)
2020–2024	836 (59.67)
Reporter qualification
Health professional	613 (43.75)
Non-health professional	779 (55.60)
Unknown	9 (0.64)
Sex
Male	775 (55.32)
Female	556 (39.69)
Unknown	70 (5.00)
Age, yrs
1–17 yrs	6 (0.43)
18–44 yrs	98 (7.00)
44–64 yrs	409 (29.19)
65–74 yrs	275 (19.63)
≥75 yrs	197 (14.06)
Unknown	416 (29.69)
TTO, days
Mean days (SD)	22.90 (81.66)
Fatal outcomes
Recovered	738 (52.68)
Not recovered	158 (11.28)
Fatal	76 (5.42)
Death	N/A
Unknown	429 (30.62)
Single drug suspected	1401 (100.00)

SD = standard deviation, TTO = time to onset.

Figure 2 presents the frequency distribution of hepatic autoimmune disorder reports reported in relation to various PD-1/PD-L1 inhibitors. Among the reported reports, nivolumab (839 reports, 59.89%) and pembrolizumab (387 reports, 27.62%) were the most frequently implicated in hepatic autoimmune disorders, collectively accounting for approximately 87.51% of all reported reports.

Figure 2.

Frequency distribution of reports of hepatic autoimmune disorders induced by PD-1/PD-L1 inhibitors. PD-1/PD-L1 = programmed cell death protein 1/death ligand 1.

3.2. Disproportionality analysis of drug-induced hepatic autoimmune disorders

The disproportionality analysis of IC and ROR for various PD-1/PD-L1 inhibitors was performed to evaluate the strength of the signal with hepatic autoimmune disorders. A signal was observed between PD-1/PD-L1 inhibitors and the onset of hepatic autoimmune disorders (ROR, 85.39 [95% CI, 80.62–90.43]; IC, 6.19 [IC_0.25, 6.10]), with a stronger signal in males (101.88 [94.14–110.25]; 6.27 [6.15]) than in females (79.46 [72.87–86.66]; 6.10 [5.96]; Figure 3 and Table S2, Supplemental Digital Content, https://links.lww.com/MD/Q64).

Figure 3.

Analysis of subgroup disproportionality and association strength in drug-induced hepatic autoimmune disorders: IC_0.25 (A) and ROR (B). CI = confidence interval, IC = information component, PD-1/PD-L1 = programmed cell death protein 1/death ligand, ROR = reporting odds ratio.

In the analysis, nivolumab (117.52 [109.52–126.10]; 6.67 [6.60]) and pembrolizumab (55.69 [50.31–61.64]; 5.65 [5.55]) exhibited the highest IC_0.25 values, suggesting a potentially stronger signal with hepatic autoimmune disorders compared to other agents within the same class. Agents with significant pharmacovigilance signals included atezolizumab (40.11 [32.23–49.92]; 5.00 [4.79]), avelumab (40.36 [25.40–64.14]; 4.29 [3.82]), cemiplimab (110.74 [74.66–164.24]; 5.13 [4.74]), durvalumab (31.04 [22.75–42.36]; 4.50 [4.18]), sintilimab (114.66 [47.52–276.66]; 3.34 [2.46]), and tislelizumab (66.27 [21.30–206.15]; 2.68 [1.55]). In contrast, the IC values for budigalimab, dostarlimab, and spartalizumab were not significant.

3.3. TTO of PD-1/PD-L1 inhibitor-induced hepatic autoimmune disorders

The mean TTO of hepatic autoimmune disorders following PD-1/PD-L1 inhibitor administration was 22.90 days (standard deviation: 81.66; Table 1). As shown in Figure 4, most reports emerged within the 1st few weeks of treatment, while a few outliers exhibited delayed onset. A distinct clustering of reports was observed within the 1st month, indicating that hepatic autoimmune reactions predominantly develop early in the treatment course.

Figure 4.

Time to onset of events related to hepatic autoimmune disorders induced by PD-1/PD-L1 inhibitors. PD-1/PD-L1 = programmed cell death protein 1/death ligand 1 inhibitors.

3.4. Temporal trends in IC_0.25 values for nivolumab and pembrolizumab

Figure 5 and Table S3, Supplemental Digital Content, https://links.lww.com/MD/Q64 present the annual IC values for nivolumab and pembrolizumab, the most frequently reported PD-1/PD-L1 inhibitors, analyzed from 2014 to 2024 to evaluate temporal trends. For nivolumab, the number of reports peaked in 2020 and showed a consistent decline thereafter. In contrast, pembrolizumab reached its highest number of reports in 2019, followed by a decrease. However, it showed a sharp resurgence in 2023 that surpassed the 2019 peak, before declining again in 2024 and returning to the previous downward trend.

Figure 5.

Trends in nivolumab and pembrolizumab-induced hepatic autoimmune disorders (2014–2024). CI = confidence interval, IC = information component.

4. Discussion

4.1. Key findings

This disproportionality study presents a comprehensive pharmacovigilance assessment of hepatic autoimmune disorders associated with PD-1/PD-L1 inhibitors, utilizing a global pharmacovigilance database. Signals of hepatic autoimmune disorders were detected for most PD-1/PD-L1 inhibitors. Furthermore, nivolumab and pembrolizumab showed the strongest significant signals, followed by cemiplimab, atezolizumab, durvalumab, avelumab, sintilimab, and tislelizumab. In the subgroup analysis, signals of hepatic autoimmune disorders were more frequently detected in males than in females following PD-1/PD-L1 inhibitor use. Most reports occurred within the 1st month of treatment, while a few reports showed significantly delayed onset.

4.2. Comparison with previous studies

Previous studies have underscored the potential of PD-1/PD-L1 inhibitors to induce immune-related hepatic toxicities,^[14–17] and existing pharmacovigilance analyses have similarly observed hepatic autoimmune disorders in patients receiving these agents.^[18] However, the majority of prior studies have been observational in nature and have predominantly analyzed ICIs as a single category, with a limited number of investigations differentiating between individual PD-1/PD-L1 inhibitors.^[16] Moreover, existing research has largely relied on clinical trial data or case series, thereby restricting its generalizability. Additionally, publication bias and inter-study heterogeneity may have further compromised the reliability of these findings.^[19]

To address these limitations, we conducted a large-scale analysis utilizing global pharmacovigilance data to extend existing findings and applied rigorous disproportionality measures to enhance objectivity and validity. Through this approach, we provide a more extensive and representative evaluation based on pharmacovigilance data, specifically elucidating the signal detection between nivolumab and pembrolizumab and hepatic autoimmune disorders.

4.3. Plausible mechanisms of PD-1/PD-L1 inhibitor-induced hepatic autoimmune disorders

PD-1/PD-L1 inhibitors interfere with the regulatory mechanisms of T-cell activation, leading to a loss of immune tolerance and an exaggerated immune response against hepatocytes.^[20] This dysregulation is further amplified by increased levels of proinflammatory cytokines, including IFN-γ, TNF-α, and IL-17, which drive hepatic inflammation and immune cell infiltration.^[20,21] Ultimately, this immune dysregulation may underlie the pharmacovigilance signal observed between PD-1/PD-L1 inhibitors and hepatic autoimmune disorders.^[21]

Among PD-1/PD-L1 inhibitors, nivolumab and pembrolizumab, which showed the strongest signals with hepatic autoimmune disorders, are PD-1 inhibitors. In contrast, PD-L1 inhibitors such as atezolizumab and durvalumab selectively block PD-L1 while preserving residual PD-1/PD-L1 interactions.^[22] This mechanism may help maintain immune homeostasis and reduce the risk of autoimmune reactions.^[22] The absence of this protective effect in nivolumab and pembrolizumab may have contributed to the detection of signals related to hepatic autoimmune disorders. Additionally, nivolumab and pembrolizumab are among the most widely prescribed PD-1/PD-L1 inhibitors across various cancer types and were among the 1st agents approved within this class.^[23] This may partially explain their higher reporting frequency in pharmacovigilance databases.

Our analysis showed that the disproportionality signals for hepatic autoimmune disorders were higher in males than in females after PD-1/PD-L1 inhibitor therapy. Previous studies have reported sex-based differences in immune-related adverse events associated with ICIs, suggesting that sex may be a potential determinant of immune toxicity.^[24] In certain cancers, estrogen is known to downregulate PD-L1 expression.^[25] This effect is probably mediated through mechanisms such as transcriptional repression of PD-L1, enhanced proteasomal degradation, and interleukin-17 downregulation.^[26] This hormonal influence on PD-L1 expression might partly contribute to observed sex-based variations in tumor immune responses and ICI treatment outcomes.^[25] Considering the reported positive association between immune-related adverse event occurrence and treatment efficacy, these findings may indirectly support our observation.^[27]

The delayed onset of hepatic autoimmune disorders following PD-1/PD-L1 inhibition suggests prolonged immune dysregulation. This may be driven by the persistence of memory T-cells with autoreactive potential and epitope spreading that induces cross-reactivity with hepatic self-antigens.^[20,21,28] Additionally, prolonged exposure to PD-1 inhibitors may induce lasting immune reprogramming, preventing the reestablishment of immune tolerance and increasing the risk of delayed-onset autoimmune reactions.^[29]

The number of reports of hepatic autoimmune disorders associated with nivolumab and pembrolizumab increased initially and subsequently declined. Although these patterns did not identically align with the Weber effect, which suggests a peak in adverse event reporting during the 2nd year after drug approval, the observed trend of increase followed by decline with nivolumab may still reflect a potential influence of this phenomenon.^[30]

4.4. Clinical and pharmacovigilance implications

This study highlights the necessity of a well-structured monitoring protocol to ensure treatment safety. First, the significant pharmacovigilance signals observed with nivolumab and pembrolizumab emphasize the need for rigorous risk stratification, particularly considering hepatic function. This finding underscores the importance of continuous hepatic function assessment and close monitoring for hepatic toxicity throughout the treatment course. Second, age-related declines in hepatic function may increase the risk of drug-induced toxicity in older patients, while age-associated immunological changes may influence responses to immune-mediated therapies.^[31] Accordingly, a more refined clinical framework and enhanced pharmacovigilance strategies are warranted for patients with underlying hepatic dysfunction and older populations. From a clinical perspective, early identification of high-risk patients and routine hepatic function assessments within the 1st month of treatment initiation are essential. Furthermore, integrating long-term surveillance measures into standard clinical practice is crucial for the timely detection and management of delayed-onset hepatotoxicity.^[32]

Finally, in certain regions, adverse events associated with PD-1/PD-L1 inhibitors remain underreported, posing a significant challenge to comprehensive global safety assessment.^[33] Therefore, the development of more robust adverse event reporting mechanisms and the enhancement of international collaboration are essential for establishing an integrated and globally coordinated drug safety surveillance system.

4.5. Limitations

Although this study highlights a specific signal detection between PD-1/PD-L1 inhibitors and hepatic autoimmune disorders, several important limitations should be considered. First, insufficient reporting numbers may inflate ROR values and widen the CI for IC values, potentially resulting in statistical distortion.^[34] This is particularly true for budigalimab, dostarlimab, and spartalizumab, which have only 1 report and limited clinical use.

Second, voluntary reporting databases, such as global pharmacovigilance systems, are inherently prone to reporting bias, which may result from underreporting, selective reporting, and regional disparities in drug safety surveillance practices.^[35] Additionally, the retrospective and observational nature of these data imposes inherent constraints on establishing definitive causal relationships. To address these limitations, we conducted a broader and more comprehensive analysis by integrating global pharmacovigilance data and applying 2 statistical methods to enhance the generalizability and reliability of our findings.

Third, certain clinical factors, such as the type and stage of cancer, underlying hepatic conditions, and concomitant medications, may have influenced the risk of hepatic autoimmune disorders. However, such information was not available in the database used for this study and therefore could not be considered in our analysis.

Fourth, further prospective studies and mechanistic investigations are required to elucidate the precise immunopathological mechanisms underlying PD-1/PD-L1 inhibitor-induced autoimmune hepatic disease. However, despite these uncertainties, our findings identify a statistically signal detection between PD-1/PD-L1 inhibitors and autoimmune hepatic disease.

Finally, the lack of reports of hepatic autoimmune disorders associated with PD-1/PD-L1 inhibitors in Africa and Southeast Asia underscores the need for enhanced surveillance and further research.^[36] Future studies should prioritize the identification of potential biomarkers for early detection and examine the influence of genetic predisposition and environmental factors on susceptibility to hepatic autoimmune disorders.

5. Conclusions

By utilizing a large global pharmacovigilance dataset, this study provides reliable evidence of a signals between PD-1/PD-L1 inhibitors and hepatic autoimmune disorders. Nivolumab and pembrolizumab exhibited the strongest signal detection, followed by cemiplimab, atezolizumab, durvalumab, avelumab, sintilimab, and tislelizumab. In the subgroup analysis, compared to females, males had a higher risk of hepatic autoimmune disorders showed elevated reporting signals with PD-1/PD-L1 inhibitors. Most cases occurred within the 1st month of treatment, while a few reports showed delayed onset. Although our findings do not permit causal inference, these results highlight the importance of early detection, regular monitoring, and risk stratification in clinical practice. As the global use of PD-1/PD-L1 inhibitors continues to increase, further research is needed to establish more effective safety monitoring strategies and improve patient outcomes.

Acknowledgments

All databases used in this study were licensed until February 7, 2025. The content presented herein does not represent the official views of the Uppsala Monitoring Centre or the World Health Organization (WHO). It should be noted that certain terms used in this manuscript, such as “association” or “induced,” may unintentionally imply clinical or causal relationships; however, no such interpretations are intended. The data were derived from a spontaneous reporting system, which inherently lacks the capacity to determine incidence or prevalence. Moreover, as this study employed a disproportionality analysis approach, it is not suitable for drawing causal inferences. All interpretations and discussions reflect the authors’ viewpoints and should not be considered direct representations of the database. Accordingly, these findings should be interpreted with caution and must not be used as the basis for clinical decision-making.

Author contributions

Formal analysis: Jeongseon Oh, Jaehyun Kong, Tae Hyeong Kim.

Investigation: Jeongseon Oh.

Methodology: Jeongseon Oh, Jaehyun Kong.

Supervision: Jiyoung Hwang, Tae Hyeong Kim, Dong Keon Yon.

Validation: Jaehyun Kong.

Visualization: Jeongseon Oh, Jaehyun Kong.

Writing – original draft: Jeongseon Oh, Jaehyun Kong, Jiyoung Hwang, Tae Hyeong Kim.

Writing – review & editing: Jeongseon Oh, Jaehyun Kong, Jiyoung Hwang, Tae Hyeon Kim, Jaeyu Park, Jaehyeong Cho, Tae Hyeong Kim.

Supplementary Material

medi-104-e44700-s001.docx ^{(19.3KB, docx)}