As the only FDA-approved therapeutic agent for TGCT, pexidartinib warrants comprehensive post-marketing surveillance of its safety profile. This study provides a comprehensive pharmacovigilance analysis of pexidartinib through the FAERS database, addressing critical gaps in real-world safety evidence for this CSF1R inhibitor. Unlike prior clinical studies, our research utilized five-year real-world data. It described the demographic information of pexidartinib—related adverse events and listed and analyzed these events across subgroups based on gender and age differences. We also carried out a more thorough analysis of both common and uncommon adverse events, including common systemic reactions like fatigue and some rare hepatic injury responses. It was noted that neurological damage, rarely mentioned in previous studies, deserves attention, particularly the association between pexidartinib and memory impairment. In total, these findings provide viable data for the clinical use of pexidartinib and risk management strategies for TGCT drug therapy.
In pharmacovigilance studies utilizing the FAERS database, a combination of four disproportionality analyses (ROR, PRR, EBGM, and BCPNN) are commonly applied18. The ROR leverages an intuitive odds ratio framework for rapid signal screening, yet its neglect of prior distributions increases susceptibility to false positives for rare events with low reporting frequencies. The PRR, while offering enhanced stability in cross-drug comparisons through proportional rate analysis, risks overlooking weakly associated signals due to empirical threshold constraints (e.g., PRR ≥ 2)19. The EBGM employs a gamma-Poisson shrinkage model to mitigate random variability in sparse data, though its reliance on large-sample prior parameter estimation limits sensitivity for newly approved medications or rare AEs. The BCPNN quantifies signal uncertainty via information component (IC) metrics, demonstrating utility in multifactorial analyses, yet its probabilistic outputs require advanced statistical interpretation, and computational complexity hinders real-time clinical implementation20. Notably, the methodological limitations are exacerbated by the spontaneous reporting nature of FAERS, with consumer-submitted reports accounting for 50.2% of cases and a significant geographic bias in reports related to pexidartinib. In addition, ROR/PRR are vulnerable to selective reporting biases, while EBGM/BCPNN performance is highly dependent on data completeness. Consequently, the convergence of multiple algorithms (e.g., 84 PTs showing positive signals across all methods) combined with sensitivity analyses (e.g., exclusion of concomitant medication confounders) enhances signal reliability.
Within FAERS data, female patients accounted for a significantly higher proportion of pexidartinib-related reports compared to males. Among age-documented cases, the 18–64.9 age cohort predominated (84.6%), potentially reflecting TGCT’s epidemiological predilection for young adults and female populations2. Gender-stratified disproportionality analysis revealed positive signals for constipation, nausea, dysgeusia, and alopecia, suggesting potential sex-dependent variations in AE susceptibility. Notably, while only 844 pexidartinib-related reports were documented in FAERS, they encompassed 7311 PTs, indicating a high incidence of multi-AE co-reporting per patient (mean 8.66 PTs/report)-a phenomenon potentially attributable to pexidartinib’s AE-prone pharmacological profile21. Outcome analysis demonstrated that 78.80% of reports were classified as non-serious, with most AEs being manageable, possibly reflecting effective risk mitigation through the mandated Risk Evaluation and Mitigation Strategy (REMS) program22,23.
Our investigation identified multiple hepatic injury-related positive signals under the Investigations SOC, including Aspartate Aminotransferase Increased (ROR 28.34, 95% CI 24.76–32.43), Alanine Aminotransferase Increased (ROR 24.17, 95% CI 20.89–27.97), Gamma-Glutamyltransferase Increased (ROR 18.92, 95% CI 15.23–23.49), and Hepatic Enzyme Increased (ROR 15.64, 95% CI 12.89–18.97). Clinical trial data corroborate these findings, with 95% (133/140) of pexidartinib-treated patients experiencing hepatic AEs, predominantly reversible, low-grade dose-dependent transaminase elevations (91% of cases). However, 4% (5 cases) developed severe mixed/cholestatic liver injury requiring treatment discontinuation21,24. Critical cases necessitating liver transplantation have been documented25, prompting FDA-mandated REMS implementation to ensure appropriate risk management26. The mechanisms underlying pexidartinib-associated hepatotoxicity are not fully elucidated but likely involve metabolic and immune-mediated pathways. Pexidartinib is primarily metabolized by cytochrome P450 (CYP) 3A4 and uridine glucuronosyltransferase (UGT) 1A4, generating reactive intermediates that may directly damage hepatocytes or trigger immune responses27. Competitive inhibition of these enzymes by concomitant medications (e.g., CYP3A4 inducers or inhibitors) may alter pexidartinib’s pharmacokinetics and exacerbate hepatotoxicity28. Preclinical and clinical data suggest that CSF-1R inhibition itself may disrupt hepatic macrophage homeostasis, impairing Kupffer cell function and promoting cholestasis or inflammation29,30. Despite significant hepatotoxicity risks, pexidartinib remains the sole pharmacotherapeutic option for inoperable or recurrent TGCT patients31. Strict adherence to REMS protocols, including baseline/frequent hepatic monitoring and provider/patient education, enables effective risk–benefit balance in clinical practice23.
In this study, the most frequently reported SOC was General Disorders and Administration Site Conditions, with the predominant PTs being Fatigue, Pain, and Feeling Abnormal. Notably, Fatigue emerged as the most commonly reported adverse event across all PTs and demonstrated a positive signal in disproportionality analysis (ROR: 4.1 with 95% CI 3.7–4.56). This finding aligns with observations from other CSF-1R antagonists, where fatigue is frequently documented32,33,34.
Eye Disorders also exhibited a positive ROR signal. Signal analysis revealed that pexidartinib induced multiple ocular adverse events, including periorbital swelling, with high reporting frequency and signal strength. Consistent with our results, these ocular adverse reactions were previously identified in clinical trials35,36. As a CSF-1R inhibitor, pexidartinib reduces macrophage recruitment and activity in peripheral tissues29,31. Given that CSF-1R is also expressed in ocular tissues, such as retinal microglia and choroidal macrophages37, which are crucial for maintaining ocular immune privilege and tissue homeostasis, disrupting CSF-1R signaling may impair these functions. Considering macrophages are implicated in lymphatic drainage and interstitial fluid homeostasis, inhibition of CSF-1R may disrupt these processes, leading to localized fluid retention and edema. A clinical study involving 41 subjects treated with the CSF-1R inhibitor PLX3397 reported periorbital edema, fatigue, hair color changes, nausea, and dysgeusia as common adverse events38. These findings were corroborated by both FAERS database reports and positive disproportionality signals, likely attributable to CSF-1R inhibition.
Nervous System Disorders also warrant attention. Beyond commonly reported and generally tolerable events such as headache, dizziness, and somnolence, pexidartinib was associated with taste disorder and memory impairment, both demonstrating positive disproportionality signals. Pexidartinib selectively inhibits CSF-1R, which predominantly targets microglia. Prolonged treatment results in the elimination of ~ 99% of brain-wide microglia, a phenomenon further validated in vitro39. Preclinical models suggest that CSF1R inhibition may alter microglial dynamics, potentially leading to neuroinflammation or synaptic remodeling40. While clinical trials in TGCT patients have not reported direct neurotoxicity, the drug’s ability to cross the blood–brain barrier warrants caution41. Intriguingly, studies have exploited pexidartinib’s microglial-depleting effects to restore hippocampal synaptic plasticity and ameliorate social isolation-induced emotional deficits42. However, microglia are critical for normal brain function, and their dysfunction is implicated in numerous neurological disorders43,44. Specifically, microglial synaptic pruning is essential for synaptic plasticity and memory consolidation45,46,47. Although microglial populations gradually recover after pexidartinib discontinuation, the long-term neurological implications—particularly with chronic use—remain to be fully elucidated.
This study has several inherent limitations due to the nature of the FAERS database and the methodologies employed. First, as a spontaneous reporting system, FAERS data suffer from underreporting, selective reporting, and incomplete information (e.g., missing demographic details such as age and weight). Second, disproportionality analyses (e.g., ROR, PRR, BCPNN) may generate false-positive signals due to confounding factors like concomitant medications or underlying diseases, which retrospective analyses cannot fully adjust for. Third, FAERS data lack clinical context such as medication dosage, treatment duration, and patient comorbidities, making causal inference challenging. Fourth, there exists geographical and ethnic bias, with pexidartinib adverse event reports predominantly originating from the United States, potentially misrepresenting AE profiles in Asian or other populations. Finally, it should be noted that while the FAERS database offers substantial data volume, it cannot establish causality between drugs and adverse events. Future studies should incorporate prospective designs or real-world data to validate these findings and address these limitations.