The thiopurine class of immunosuppressant includes azathioprine (AZA), 6-mercaptopurine (MP), and thioguanine [1]. These agents are widely used in clinical practice to manage immune-mediated responses in organ transplantation and autoimmune disorders [2]. The pharmacological mechanism of thiopurines is well established: AZA is converted to thioguanine nucleotides (TGNs) by several enzymes [3–5]. TGNs consist of thioguanine monophosphate, thioguanine diphosphate and thioguanine triphosphate [6]. These active metabolites exert cytotoxic and immunosuppressive effects through multiple pathways, ultimately leading to T-lymphocyte suppression and apoptosis [7]. Among the three agents, AZA has the longest history of regulatory approval and is indicated by the U.S. Food and Drug Administration (FDA) for preventing rejection in renal transplantation and for managing active rheumatoid arthritis by reducing signs and symptoms [8]. In addition, AZA and other thiopurines are frequently used off-label for chronic inflammatory diseases such as Crohn’s disease (CD), ulcerative colitis (UC) and systemic lupus erythematosus (SLE) [9, 10].

Despite its therapeutic benefits, AZA is associated with several adverse events (AEs), including hepatotoxicity, malignancy, cytopenias and other serious infections [8, 11]. Among the various AEs, intrahepatic cholestasis of pregnancy (ICP) has emerged as a clinically relevant safety concern in pregnant women [12]. ICP is a liver disorder characterized by impaired bile flow and elevated serum bile acid levels, which are associated with adverse maternal outcomes and increased risk of preterm delivery, fetal distress, and stillbirth [13]. Although the exact etiology of ICP remains unclear, environmental, genetic, immunological, and hormonal factors are implicated [14]. Clinical manifestations of ICP may include pruritus, jaundice, right upper quadrant pain, nausea, poor appetite, or sleep disturbance [15]. According to FDA prescribing information, administration of AZA is not recommended during pregnancy [8]. However, treatment guidelines from the American College of Rheumatology and the American Gastroenterological Association suggest that continuation of AZA or MP may be appropriate in selected cases to manage underlying disease activity during pregnancy [16–19]. Therefore, careful monitoring for AZA-associated ICP is clinically important. AZA has been reported to induce transient elevations in liver enzyme levels and biochemical cholestasis [20].

Recent FDA safety communications have highlighted the potential risk of ICP associated with thiopurines, including AZA, thereby raising concerns about its safety profile in pregnant women [16]. While the current FDA prescribing information for AZA acknowledges the possibility of ICP and advises discontinuation upon diagnosis, it does not provide quantitative evidence to establish the strength of this association [8]. Furthermore, there is a lack of pharmacovigilance studies that systematically evaluate the relationship between AZA exposure and ICP during pregnancy. To our knowledge, the present work is the first quantitative pharmacovigilance analysis to assess the association between AZA and ICP using the U.S. FDA Adverse Event Reporting System (FAERS) database. In addition, we compared the disproportionality of AZA with other drugs known to cause ICP, thereby contextualizing the potential risk in pregnant women.

This study suggests a significant association between AZA use and ICP, supporting the need for heightened clinical monitoring and risk management when prescribing AZA to pregnant patients. Our analysis revealed a markedly elevated ROR for ICP with AZA (ROR025 = 153.0; IC025 = 5.8; EBGM05 = 144.37), and pregnancy-specific subgroup analyses further identified significant signals, including ICP itself (ROR025 = 5.46; IC025 = 1.93; EBGM05 = 5.31) and hepatobiliary events such as drug-induced liver injury (ROR025 = 34.38; IC025 = 4.07; EBGM05 = 20.85). Additional subgroup analyses identified significant signals in patients with Crohn’s disease (ROR025 = 66.99; IC025 = 4.8; EBGM05 = 64.73), and Colitis ulcerative (ROR025 = 9.01; IC025 = 1.95; EBGM05 = 9.95). We also observed that the proportion of ICP reports for AZA (67 cases among 17,744 reports) was higher than that of other ICP-inducing drugs, approximately two-fold greater in relative percentage. To our knowledge, this is the first quantitative disproportionality analysis specifically evaluating AZA-associated ICP. By combining subgroup and comparative analyses, we attempted to strengthen the reliability of the signal and provide a more comprehensive understanding of this association. Importantly, pregnancy-specific interpretations in this study are based on the pregnancy-verified subset (n = 18), whereas the broader disproportionality signal reflects the full ICP-coded set (n = 67). This distinction should be considered when interpreting the findings for pregnant women.

We hypothesize that several factors may contribute to this phenomenon. One proposed mechanism involves pregnancy-related hormonal modulation of thiopurine metabolism. During pregnancy, changes in enzyme activity, including thiopurine S-methyltransferase, may alter the balance of AZA metabolism toward 6-mercaptopurine (6-MP), thereby increasing the generation of hepatotoxic metabolites [20, 27]. This metabolic diversion could amplify cholestatic effects in susceptible individuals. Additionally, AZA has been documented to cause transient elevations in liver enzyme levels and biological cholestasis [20]. Consistent with this, a French case series reported six women with IBD on AZA who developed atypical and more severe ICP earlier in pregnancy [20, 28]. In these patients, bile acid levels were markedly elevated despite normal liver enzymes, and abnormalities persisted despite UDCA therapy but resolved upon AZA discontinuation [28]. Such clinical evidence supports a potential causal role of AZA in exacerbating ICP.

Beyond individual case descriptions, the documented patterns of AZA-related hepatotoxicity offer mechanistic support for its potential to contribute to cholestatic conditions such as ICP. Early in therapy, elevations in aminotransferases have been linked to higher levels of methyl-mercaptopurine, a metabolite formed during AZA biotransformation and known to exert direct toxic effects on hepatocytes [29, 30]. AZA can also induce an acute cholestatic injury, typically presenting within 2–12 months and characterized histologically by intrahepatic cholestasis with focal hepatocellular necrosis and scant inflammation. This “bland cholestasis” pattern described as similar to the cholestasis observed with estrogens suggests impaired bile flow rather than immune-mediated hepatitis. Chronic thiopurine exposure has also been associated with sinusoidal dilation, portal venopathy, and nodular regenerative hyperplasia, reflecting structural disturbances that may interfere with biliary drainage. Rarely, prolonged cholestasis or vanishing bile duct syndrome has been described, and long-term thiopurine use has been associated with hepatocellular carcinoma and hepatosplenic T-cell lymphoma [31]. Considering these mechanisms and disease features together offers a clear biological explanation for the strong cholestasis risk observed. This strongly suggests that AZA could make cholestasis worse, especially during pregnancy.

ICP itself is a multifactorial disorder in which several converging risk factors impair bile acid homeostasis [28, 32]. Genetic variants in hepatobiliary transporters and nuclear receptors (e.g., ABCB4/MDR3, ABCB11/BSEP, ABCC2, NR1H4/FXR) predispose individuals to defective bile secretion and intrahepatic bile acid accumulation [32, 33]. Pregnancy-specific hormonal changes, particularly elevated sulfated progesterone metabolites (e.g., epiallopregnanolone sulfate) and estradiol, can downregulate BSEP expression and attenuate FXR signaling, thereby reducing bile acid clearance. In parallel, an imbalance in maternal immune responses, including increased IL-6, IL-12, IL-17, and TNF-α and reduced IL-4, may exacerbate hepatocellular injury [32, 34]. The convergence of these genetic, hormonal, and immunological influences with AZA’s hepatotoxic potential offers a plausible biological basis for the disproportionately high ROR for ICP observed in our analysis.

Beyond these biological mechanisms, disease-specific clinical and therapeutic factors may further explain the gradient in signal intensity observed across CD, UC, and SLE. Thiopurine exposure patterns during pregnancy differ across these conditions: CD carries a higher risk of flare, leading most patients to maintain thiopurine therapy throughout gestation, whereas UC is more often managed with lower or intermittent exposure. Real-world pregnancy registry data reinforce this difference. In the large prospective PIANO cohort, 41% of women with UC received no thiopurine or biologic therapy during pregnancy, compared with only 16% of those with CD [35]. Pregnant women with CD also demonstrated higher use of biologic-thiopurine combination therapy (18% vs. 11% in UC) and greater reliance on sustained immunomodulator treatment, resulting in markedly higher cumulative thiopurine exposure. In contrast, treatment strategies for SLE pregnancies differ fundamentally from those used in IBD. Management generally centers on hydroxychloroquine and low-dose corticosteroids, and when azathioprine is prescribed, it is typically administered at lower doses consistent with rheumatology guidelines recommending ≤2 mg/kg/day [18, 36, 37]. As a result, cumulative thiopurine exposure in SLE is considerably lower than in CD or UC. It is also important to acknowledge that the SLE subgroup in our dataset comprised only three ICP cases, raising the possibility that the weaker signal observed (ROR025 = 2.16; IC025 = −0.19; EBGM05 = 2.58) may reflect the small number of available reports, rather than indicating that no association exists.

This study has several limitations. First, inherent bias of disproportionality analyses and spontaneous reporting systems such as FAERS apply to this study. Underreporting and reporting biases are unavoidable, and although extensive deduplication was performed, some residual duplicates may remain. Moreover, the association between azathioprine and ICP observed in this analysis cannot be interpreted as causal. Disproportionality analyses do not allow for estimation of incidence, prevalence, or comparative risk due to the absence of reliable drug utilization data and background exposure rates in the FAERS database [38]. Given the observational design and reliance on voluntary spontaneous reports, the findings should be considered hypothesis-generating, and not interpreted as evidence of causality or as a quantification of absolute or comparative risk. Second, temporal bias may also have affected the observed disproportionality signal, as reporting patterns related to azathioprine or intrahepatic cholestasis of pregnancy may vary over time in response to regulatory alerts, increased clinical awareness, or media coverage. Third, potential confounding factors such as differences in disease severity, pregnancy status, and concomitant immunosuppressive therapies may have influenced the reporting patterns. In addition, indication bias and cumulative exposure should be considered, as azathioprine has been widely used for decades in chronic autoimmune conditions like CD, UC and SLE, which themselves may increase the risk of ICP. These factors make it difficult to determine whether the observed signal is attributable to azathioprine, the underlying disease, or their interaction. Finally, the mechanistic explanation presented in this study is based on a synthesis of existing literature and should be regarded as hypothesis-generating rather than conclusive, as this study did not directly assess biological mechanisms. Despite these limitations, we mitigated bias by conducting subgroup and comparative analyses, which consistently demonstrated elevated disproportionality signals for AZA-associated ICP.

In conclusion, this pharmacovigilance analysis identifies a novel disproportionality signal suggesting a possible association between AZA exposure and reports of intrahepatic cholestasis of pregnancy. Given the clinical importance of both disease control and maternal–fetal safety, clinicians should weigh the benefits of AZA therapy against the potential risk of ICP and consider enhanced monitoring strategies, especially in women of reproductive age with autoimmune diseases. Future prospective studies and mechanistic investigations are warranted to validate these findings and to clarify the underlying biological pathways.