Risk of drug-induced pericardial effusion: a disproportionality analysis of the FAERS database

Abstract

Objective

By using the FAERS database, we aim to identify and assess risk signals of adverse drug events (ADEs) potentially causing pericardial effusion, to inform clinical drug management and promote rational drug use.

Methods

We obtained reports of pericardial effusion events from the FAERS database spanning from the first quarter of 2004 to the second quarter of 2024, and identified the top 50 drugs ranked by report frequency or signal strength. Four algorithms, namely the reported odds ratio (ROR), proportional reporting ratio (PRR), Bayesian confidence propagation neural network (BCPNN), and multi-item gamma Poisson shrinker (MGPS), were employed for signal detection of these drugs. Furthermore, for drugs with positive signals, we conducted sensitivity analyses and employed the Weibull shape parameter test to perform a time to onset (TTO) analysis.

Results

We identified 20,057 ADEs related to pericardial effusion, involving 19,693 patients for analysis. The patient population comprised 10,187 males (51.7%) and 7,939 females (40.3%). Adults aged 18–65 years were the largest group (7,798 cases, 39.6%). Regarding clinical outcomes, 9,924 patients (50.4%) experienced hospitalization, and 2,770 cases (14.1%) resulted in death. Ranked by the ROR risk signal strength, the top 3 drugs were hydralazine [ROR (95% CI): 27.11 (22.28–33)], dasatinib [ROR (95% CI): 15.62 (14.07–17.33)], and mesalazine [ROR (95% CI): 8.99 (6.84–11.8)]. We conducted a TTO analysis for the 26 drugs with positive signals. The median TTO and interquartile range (IQR) for the top 3 drugs causing the earliest pericardial effusion were: cytarabine 14 (7.5,38), selexipag 14.5 (4.25, 157.75), dabigatran etexilate 29 (9, 229). Most drugs exhibited an early failure type.

Conclusion

This study systematically compiled a list of drugs with potential risks of causing pericardial effusion. There is a significant association between pericardial effusion and the use of hydralazine, dasatinib, and mesalazine. Moreover, pericardial effusion is more common in patient groups receiving treatments with antineoplastic and immunomodulating agents.

Introduction

The pericardium is a conical, serous sac consisting of a visceral and a parietal layer. Pericardial effusion (PE) occurs when the fluid accumulation within the pericardial cavity exceeds 50 mL [1]. As a common clinicopathological phenomenon, the reported prevalence of PE is up to 9% when it is diagnosed by echocardiography [2]. The severity of PE is correlated with the effusion volume. Minor effusion generally has minimal impact on hemodynamics, while significant effusion can lead to cardiac tamponade, posing a serious threat to life [3]. The etiology of PE is multifaceted, often attributed to factors such as inflammation, autoimmune diseases, and tumors. However, drug-induced pericardial effusion (DPE) is frequently overlooked as a potential causative factor [4,5,6]. DPE often complicates the treatment of the underlying disease and may impede its progress. For example, minoxidil, a common treatment for refractory hypertension, can cause PE as a severe side effect, leading to dyspnea and chest pain [7]. Due to DPE’s significant latency, prompt diagnosis and intervention are crucial to prevent PE recurrence. Excessive fluid accumulation in the pericardial cavity severely impairs diastolic and systolic heart functions, potentially leading to serious adverse events such as hypotension and cardiac arrest [8]. Therefore, rapid and accurate identification of drug factors causing PE is crucial for optimizing treatment outcomes and enhancing patients’ clinical prognosis.

The U.S. Food and Drug Administration’s Adverse Event Reporting System (FAERS) is a publicly available, voluntary system, for post-marketing safety surveillance of FDA - approved drugs, facilitating the detection of potential links between drugs and adverse drug events (ADEs) [9]. Numerous scholars have uncovered the drug-associated factors underlying the onset of myocarditis [10], angioedema [11], and arrhythmia [12] by leveraging the FAERS database. This has effectively demonstrated the feasibility of drug-related study using the FAERS database. To date, various drugs, including immunosuppressants [13], antineoplastic agents [14], anticoagulants [15], and antihypertensive drugs [16], have been reported potentially to induce PE. However, the current related study has several limitations. On one hand, the data sources are predominantly case reports, lacking large-scale data. Consequently, the sample size is small and the representativeness is insufficient, making it difficult to comprehensively and accurately reflect the association between drugs and PE. On the other hand, there is a lack of systematic summarization of these drugs that induce PE. Current study fails to elaborate on the specific frequencies of drugs inducing PE and the strength of risk signals, thus being unable to provide adequate guidance for clinical medication. Compared with previous studies, this study relies on the advantages of large-scale and diverse data in the FAERS database, effectively overcoming the limitation of insufficient sample size. It also integrates the study results of predecessors to strengthen the evidence base. By adopting the disproportionality analysis method commonly used in the field of pharmacovigilance, this study accurately identifies the risk signals of drugs causing PE, provides references for clinical drug management and rational drug use, and reduces the risk of PE events in patients.

Materials and methods

Data source

The data for this study were sourced from the FAERS database, which has been publicly accessible and updated quarterly since 2004. The data were stored in ASCII or XML format (https://fis.fda.gov/extensions/FPD-QDE-FAERS/FPD-QDE-FAERS.html). We acquired ASCII data spanning 82 quarters, from the first quarter of 2004 to the second quarter of 2024. The dataset includes patient demographics (DEMO), indication (INDI), drug use records (DRUG), therapy duration (THER), adverse event records (REAC), and patient outcomes (OUTC).

Data processing

The 82 quarterly data files were imported into a MySQL 8.0 database for processing. ① SQL queries were employed to eliminate duplicate entries in the DEMO table. In accordance with FDA recommendations, only the data with the most recent report date (FDA_DT) were retained for each unique case number (CASEID). In cases where both CASEID and FDA_DT were identical, the report with the highest PRIMARYID was selected. Subsequently, the DEMO table was linked to other subfiles to create a summary table (Supplementary Material 1). Since the FAERS database allows for multiple ADEs to be recorded per patient, which may potentially result in multiple distinct clinical outcomes for each individual, we selected the most severe clinical outcome as the definitive outcome for this study. ② The drug names were normalized using Medex_UIMA_1.3.8 from Vanderbilt University. The ‘DRUG’ table was de-duplicated, each drug was numbered, and the result was saved as DRUGNAME.txt. Special characters such as “#”, “()”, “[]” were removed during the pre-processing step to prepare for Medex_UIMA_1.3.8 mapping. To ensure the software operates properly, one should install a Java environment and correctly configure the environment variables for Java command recognition. Then, unzip the downloaded Medex_UIMA_1.3.8 to a local disk, define it as MedEx_HOME, and save the pre-processed DRUGNAME.txt in its input folder as the identification source file. Run the Java command to start the software. It reads DRUGNAME.txt in the input folder for drug name recognition, it compares with the built-in standard drug dictionary, converts brand names and abbreviations to generic names, and outputs the results as an XML file to the output folder. ③ For drug classification, we adopted the Anatomical Therapeutic Chemical (ATC) classification system (https://www.who.int/tools/atc - ddd - toolkit/atc - classification), which is developed and regularly updated by the WHO Collaborating Centre for Drug Statistics Methodology.

By consulting the Medical Dictionary for Regulatory Activities (MedDRA) (https://www.meddra.org/), we designated “pericardial effusion” (code: 10034474) as the preferred term. Using this term, we searched for pertinent ADE reports. During the data processing phase, only reports with complete data were included in the analysis. Meanwhile, data points with an age greater than 120 years and a body weight exceeding 400 kg were defined as outliers and excluded. Given the wide-ranging sources of reports in the FAERS database, to effectively mitigate bias, the report sources were divided into two subsets: consumers and medical workers, for subsequent analysis. On this basis, a list of drugs potentially inducing PE was compiled. Depending on the magnitude of each drug’s impact on PE formation, the drugs were classified into three categories: primary suspect (PS), secondary suspect, and concomitant or interaction. To guarantee the accuracy of conclusions, only ADEs in which PS drugs induced PE were incorporated into the analysis. The PS drugs were subsequently ranked according to their frequency and signal strength. Finally, during the analysis of the time to onset (TTO) of ADEs, reports lacking either “START_DT” or “END_DT”, as well as those where “START_DT” was later than “EVENT_DT"’, were excluded to ensure the validity of the analysis.

Statistic analysis

The disproportionality analysis and Bayesian approaches are pivotal analytical tools in pharmacovigilance. Within the framework of a 2 × 2 table (Table 1), the disproportionality analysis employs two algorithms: the reported odds ratio (ROR) and the proportional reported odds ratio (PRR) [17]. The Bayesian approaches primarily encompass two representative algorithms: the Bayesian confidence propagation neural network (BCPNN) [18]and the multi-item gamma Poisson shrinker (MGPS) [19]. The formulas for these four algorithms are provided in Supplementary Material 2. To enhance the reliability of the drug-ADE correlation analysis, an ADE signal is generated only when criteria from all four algorithms are concurrently met. The ROR algorithm is simple and can intuitively present the association between drugs and ADEs. In contrast, the PRR algorithm highlights differences by calculating ratios, enabling the efficient screening of drugs with abnormal risk signals. The BCPNN, based on Bayesian principles, integrates diverse data and supports cross validation, thereby enhancing signal robustness. The MGPS effectively controls biases caused by data sparsity or over reporting [20]. Furthermore, the FAERS database compiles spontaneous ADEs from diverse populations. Given the notable disparities in medical expertise between consumers and medical workers, this discrepancy may introduce false positive signals into the dataset. To assess and quantify this potential effect, we conducted a sensitivity analysis.

Table 1 Four-fold table of disproportionality analysis

We used the median and interquartile range (IQR) as key statistics to describe the TTO. To further explore the distribution characteristics of the TTO, we applied the Weibull shape parameter (WSP) test [21] to reveal its characteristics. The WSP test serves as an early - warning indicator by detecting ADEs in patients during specific periods, which enables prompt preventive measures or adjustments to treatment plans. All statistical analyses were conducted using R software.

Outcomes

Retrieval process

We retrieved a total of 53,769,390 ADEs from the FAERS database, and 20,057 of them were associated with PE. Subsequently, after data processing, 19,693 patients were included in the baseline analysis. The detailed workflow is presented in Fig. 1.

Fig. 1

Flowchart for identifying pericardial effusion reports. ADE: adverse drug event; PS: primary suspect

Basic characteristics of ADE reports related to pericardial effusion

Regarding gender distribution, males accounted for a higher proportion (51.7%) than females (40.3%). Across different age groups, adults aged 18–65 years made up the largest proportion (39.6%), followed by elderly patients (27.5%). The ADE reports mostly came from medical workers. Basic information is shown in Table 2.

Table 2 Clinical characteristics of pericardial effusion report

From 2004 to 2024, the number of ADE reports related to PE showed an overall upward trend (Fig. 2A). Significantly, since 2012, this number has surged and maintained a high level. The peak occurred in 2023, with 1757 cases reported. The reduced number of ADE reports in 2024 is attributable to the inclusion of data from only two quarters.

Fig. 2

(A) Annual number of reported adverse drug events related to pericardial effusion. (B) Proportion of clinical outcomes. CA: congenital anomaly, DE: death, DS: disability, HO: hospitalization, LT: life-threatening, OT: other, NA: not available. (C) Top 8 reporting countries. CA: Canada, CN: China, DE: Germany, FR: France, IT: Italy, JP: Japan, UK: The United Kingdom, US: The United States

The majority of ADE reports came from European and American countries. The top 8 countries with the most ADE reports were the United States (9380 cases), Japan (1186 cases), Canada (1136 cases), Germany (981 cases), France (977 cases), the United Kingdom (774 cases), Italy (388 cases), and China (379 cases). For details, see Fig. 2B.

As depicted in Fig. 2C, the distribution of patient outcomes is presented as follows. Hospitalization was the most frequent clinical outcome, with 9924 cases (50.4%). Moreover, 2770 cases (14.1%) resulted in death. The top 5 drugs that led to hospitalization and death outcomes are listed in Table 3.

Table 3 Top 5 drugs with case outcome of hospitalization and death

Top 50 drugs based on report frequency

By analyzing ADE reports, we identified the top 50 PS drugs based on reporting frequency. These drugs include: adalimumab (436 cases, 2.21%), ibrutinib (383 cases, 1.94%), treprostinil (375 cases, 1.90%), clozapine (374 cases, 1.90%), dasatinib (363 cases, 1.84%), nivolumab (339 cases, 1.72%), ambrisentan (325 cases, 1.67%), sacubitril/valsartan (299 cases, 1.52%), macitentan (297 cases, 1.51%), dabigatran etexilate (286 cases, 1.45%), lenalidomide (281 cases, 1.43%), rivaroxaban (264 cases, 1.34%), infliximab (243 cases, 1.23%), bosentan (239 cases, 1.21%), rosiglitazone (234 cases, 1.19%), among others (see Table 4 for details). It is noteworthy that the relatively high ranking of certain drugs among those causing PE may be due to their high usage volume. This does not necessarily mean that there is an inevitable link between these drugs and PE. According to the Anatomical Therapeutic Chemical (ATC) classification, the top 50 drugs are categorized as follows: antineoplastic and immunomodulating agents (31 drugs), cardiovascular system drugs (11 drugs), blood system drugs (3 drugs), alimentary tract and metabolism drugs (3 drugs), musculo - skeletal system drug (1 drug), and nervous system drug (1 drug).

Table 4 Basic information and classification of the top 50 drugs based on report frequency

Disproportionality analysis and sensitivity analysis

Signal detection was conducted on the top 50 PS drugs, resulting in the identification of 26 drugs as positive signals (Fig. 3). The results showed that: hydralazine [ROR (95% CI): 27.11 (22.28-33)], dasatinib [ROR (95% CI): 15.62 (14.07–17.33)], mesalazine [ROR (95% CI): 8.99 (6.84–11.8)], epoprostenol [ROR (95% CI): 8.97 (4.79–10.74)], bosutinib [ROR (95% CI): 8.34 (6.45–10.78)], crizotinib [ROR (95% CI): 8.25 (6.73–10.11)], macitentan [ROR (95% CI): 7.44 (6.63–8.35)], gemcitabine [ROR (95% CI): 6.54 (5.65–7.57)], dabigatran etexilate [ROR (95% CI): 6.46 (5.79–7.21)], bosentan [ROR (95% CI): 5.53 (4.69–6.05)], nivolumab [ROR (95% CI): 5.29 (4.75–5.89)], atezolizumab [ROR (95% CI): 5.19 (4.27–6.3)], ibrutinib [ROR (95% CI): 5.06 (4.58–5.6)], selexipag [ROR (95% CI): 4.97 (4.16–5.94)], nilotinib [ROR (95% CI): 4.82 (4.13–5.64)], trastuzumab [ROR (95% CI): 4.71 (4.14–5.37)], osimertinib [ROR (95% CI): 4.6 (3.64–5.81)], cytarabine [ROR (95% CI): 4.4 (3.52–5.5)], pembrolizumab [ROR (95% CI): 4.29 (3.76–4.9)], ambrisentan [ROR (95% CI): 4.22 (3.78–4.71)], imatinib [ROR (95% CI): 4.22 (3.71–4.79)], clozapine [ROR (95% CI): 4.16 (3.76–4.61)], everolimus [ROR (95% CI): 3.9 (3.37–4.51)], rosiglitazone [ROR (95% CI): 3.64 (3.2–4.14)], treprostinil [ROR (95% CI): 3.57 (3.23–3.96)]. Among these 26 drugs, 7 drugs (dasatinib, mesalazine, bosutinib, nilotinib, trastuzumab, pembrolizumab, imatinib) have documented expected adverse reactions of PE in FDA - approved drug labels, while the remaining 19 drugs do not mention the risk of inducing PE. Physicians should intensify the monitoring of PE symptoms in patients using these drugs. This includes regular echocardiographic examinations and close observation of whether patients exhibit symptoms such as dyspnea, chest pain, and palpitations, enabling the early detection and timely management of potential adverse reactions.

Fig. 3

Signal detection results of the top 50 drugs based on report frequency. ROR: the reporting odds ratio; PRR: the proportional reporting ratio; EBGM: the empirical Bayes geometric mean; EBGM05: the lower limit of 95% CI, of EBGM; IC: the information component; IC025: the lower limit of 95% CI, of the IC; Y: yes, N: no

The data in the FAERS database come from diverse sources, including spontaneous reports submitted by both consumers (patients and other related individuals) and medical workers (physicians, pharmacists, and other health specialists). There may be significant differences between these two types of reporters in terms of professional knowledge, reporting motivations, and report content. By conducting sensitivity analyses, we can reduce biases arising from these varied data sources, thereby enhancing the precision and reliability of our analyses. We used the ROR method for a disproportionality analysis of ADEs reported by consumers and medical workers for the 26 positive drugs. The results are shown in Fig. 4. For some drugs, such as ambrisentan [ROR (95% CI): 0.74 (0.57–0.83)] and digoxin [ROR (95% CI): 0.37 (0.16–0.87)], ADEs were more frequently reported by consumers, and the results are not reliable. However, ADEs for the majority of drugs were predominantly reported by medical workers, lending stability to the results.

Fig. 4

Forest plot of sensitivity analysis of reporting odds ratios for the 26 positive drugs causing pericardial effusion. MW: medical worker; CN: consumer

Time-to-onset analysis

When analyzing the TTO of the 26 drugs that generated positive signals (Table 5), we used the WSP test and focused specifically on the top 10 drugs that induced PE in the shortest time. The median TTO and IQR of these drugs are presented as follows: cytarabine 14 (7.5, 38); selexipag 14.5 (4.25, 157.75); dabigatran etexilate 29 (9, 229); pembrolizumab 29 [11, 68]; osimertinib 32 (11.75, 107); clozapine 34 (16, 1623); crizotinib 40 (26, 108.5); atezolizumab 55 (18, 128); nivolumab 59.5 (26.75, 138.5); everolimus 60 (21, 195). The results of the WSP analysis showed that the shape parameter β and the upper limit of its 95% CI for most drugs were less than 1, indicating that the probability of these drugs inducing PE tends to decrease over time.

Table 5 Analysis of time-to-onset of 26 positive signal drugs causing pericardial effusion

Top 50 drugs based on signal strength

Using reporting frequency as the inclusion criterion for drugs may introduce potential biases. To address this shortcoming, we used disproportionality analysis to rank the top 50 drugs according to signal strength. This strategy can help identify potential safety issues independent of drug use frequency and enables focused monitoring of drugs with high-risk signals. The results are shown in Fig. 5. Specifically, the top 10 drugs are as follows: pergolide [ROR (95% CI): 35.97(20.2–64.05)], arsenic trioxide [ROR (95% CI): 28.53(20.47–39.76)], loncastuximab tesirine [ROR (95% CI): 28.27(12.54–63.71)], hydralazine [ROR (95% CI): 27.11(22.28–33)], dinutuximab [ROR (95% CI): 22.71(11.26–45.81)], plerixafor [ROR (95% CI): 18.89(10.1–35.33)], ceritinib [ROR (95% CI): 17.44(13.06–23.28)], denileukin diftitox [ROR (95% CI): 15.85(5.06–49.62)], dasatinib [ROR (95% CI): 15.62(14.07–17.33)], porfimer [ROR (95% CI): 14.69(6.07–35.56)].We found that the number of antineoplastic and immunomodulating agents reached 34, accounting for a significant proportion (68%) among the top 50 drugs. These types of drugs remain important risk factors for the occurrence of PE.

Fig. 5

Forest plot of the top 50 drugs based on signal strength

Discussion

This is the first comprehensive signal-detection study using the FAERS database to investigate high-risk drugs potentially inducing PE. Consistent with previous studies [22], our results show that DPE events are predominantly concentrated in adult patients, especially males. Regarding clinical outcomes, 50.4% of patients were hospitalized, and 14.1% of patients even died. The data, mainly from medical workers’ reports, are credible to some extent. Clinicians should be aware of the PE risk when prescribing drugs to adult males. The annual number of reported DPE events has been increasing, peaking at 1,757 cases. We performed signal detection on the top 50 drugs associated with PE using the ROR and PRR methods, and 26 drugs with positive signals were identified. Remarkably, only 7 of these drugs listed PE as an expected ADE in FDA - approved drug labels. Due to the complexity of drug development and clinical trials, ADE information in drug labels may not be updated promptly, which may lead to inadequate warnings and preventive measures for patients during drug use.

Adalimumab, a commonly used tumor necrosis factor-α blocker for rheumatoid arthritis and ankylosing spondylitis [23], is the drug most frequently associated with PE events in the FAERS database. Many studies [24,25,26] have documented PE events in patients undergoing adalimumab therapy. Despite being identified as a negative drug in pharmacovigilance signal detection, close monitoring for PE in patients using adalimumab remains essential, considering the patient’s overall health and medication risks. Ibrutinib is linked to the highest number of ADEs of PE among drugs with positive signals. Ibrutinib, a novel targeted anticancer agent approved by FDA, is routinely employed in the treatment of mantle cell lymphoma and chronic lymphocytic leukemia [27]. Several clinical studies [28,29,30] have shown that ibrutinib can induce hemorrhagic PE. The underlying mechanism is postulated to involve its interference with platelet aggregation through inhibition of Bruton’s tyrosine kinase and tyrosine kinase expressed in hepatocellular carcinoma; when administered concomitantly with antiplatelet and anticoagulant medications, the risk of hemorrhagic PE is substantially elevated [31, 32]. Utilizing the WSP test, this study determined that the median TTO for ibrutinib-induced PE is 80 days, and notably, the drug’s package insert omits warning information regarding PE risk. This underscores the necessity for heightened vigilance and close monitoring of PE risk when initiating ibrutinib therapy in clinical practice [33].

The ROR serves as an index to assess the strength of the association between a drug and an ADE. A higher ROR indicates a stronger correlation between the drug and the onset of PE [34]. This study identified hydralazine, dasatinib, and mesalazine as the top 3 drugs with higher risk signals. Hydralazine exerts its hypotensive effect by targeting and relaxing vascular smooth muscles, thereby reducing peripheral resistance and lowering blood pressure. Moreover, it enhances cardiac output and optimizes cardiac function, leading to its widespread use in the treatment of hypertension and heart failure [35]. Recent study indicates that the PE caused by this drug is not due to direct cell toxicity, but rather closely linked to the mechanism of drug-induced lupus-like syndrome [36,37,38]. Further investigation reveals that this pathological process frequently involves autoimmune reactions triggered by the inhibition of T-cell DNA methylation and the accumulation of autoantibodies, particularly in patients with slow acetylation following hydralazine treatment [39]. Clinicians must maintain high vigilance in monitoring for signs of hydralazine-induced lupus-like syndrome (HILS), particularly in patients receiving doses of 200 milligrams or more daily for over three months. Any nonspecific symptoms or unexplained cases of PE should be viewed with a high index of suspicion for HILS, warranting immediate drug withdrawal [40, 41].

Dasatinib, a potent tyrosine kinase inhibitor, is primarily administered for targeted therapy in chronic myelogenous leukemia (CML) [42]. With rapid absorption, dasatinib is primarily metabolized and eliminated through the cytochrome P450 (CYP) 3A4 enzyme system, with a half-life of about 3 to 4 h [43]. Studies indicate that dasatinib may induce severe PE in 1% of patients [6]. A clinical study of 102 CML patients treated with dasatinib revealed that 30.9% developed pericardial or pleural effusions [44]. Although the exact mechanism of dasatinib-induced PE remains unclear, studies suggest it may involve the drug’s inhibition of platelet-derived growth factor receptor-β (PDGFR-β) or SRC family kinases, resulting in increased vascular permeability [45]. Moreover, an analysis of the correlation between drug dosage and ADEs revealed that CML patients receiving dasatinib at doses of 140 mg, 100 mg, or 50 mg were all susceptible to developing PE. Further analysis indicated that the TTO for dasatinib-induced PE is 155 days. In light of this finding, regular chest X-ray examinations are recommended for patients undergoing long-term treatment with dasatinib to promptly detect and manage potential PE [46].

Mesalazine, commonly used to treat inflammatory bowel diseases like ulcerative colitis and Crohn’s disease, contains 5-aminosalicylic acid as its primary active ingredient. This ingredient inhibits intestinal mucosal inflammation by suppressing the synthesis of pro - inflammatory prostaglandins and the formation of inflammatory mediators such as leukotrienes [47, 48]. Despite its proven efficacy, the literature consistently reports rare but severe side effects of Mesalazine, including myocarditis, pericarditis, and PE, with symptoms typically resolving after drug discontinuation [49,50,51]. However, the exact pathological mechanism of mesalazine-induced PE remains unclear. Existing study indicates that IgE-mediated allergic reactions and direct cardiotoxic effects may be potential pathogenic mechanisms [52]. Specifically, a retrospective study of 52 patients with inflammatory bowel disease revealed that even within the recommended dosage range, most patients may still experience PE due to cardiotoxic reactions to mesalazine, with a median onset time of 14 days [53]. Therefore, clinicians must remain vigilant for this serious complication potentially caused by mesalazine and implement early monitoring and intervention strategies.

In this study, we systematically classified 50 drugs and found that antineoplastic and immunomodulating agents are the most prevalent, followed by cardiovascular and blood system drugs. Notably, a clinical study of 2005 CML patients treated with BCR-ABL1 tyrosine kinase inhibitors (TKIs) clearly identified PE as one of the most closely related adverse reactions to TKIs treatment [54]. This may be due to the off-target effects of TKIs on the immune system, specifically their interaction with the PDGFR-β receptor, which could significantly contribute to PE occurrence [55]. Furthermore, in non-small cell lung cancer treatment, a meta-analysis revealed that PD-1 inhibitor monotherapy significantly increases the risk of PE and cardiac tamponade [56]. Multiple studies have further confirmed the association between PD-1 inhibitors and PE [57, 58], yet the exact mechanism remains elusive. It is hypothesized that this may be linked to immunotherapy-induced T-cell overactivation, resulting in myocardial damage and severe complications, including PE and even pericardial tamponade [59]. Cardiovascular drugs, particularly those for pulmonary arterial hypertension, account for a large proportion and warrant our utmost attention regarding safety issues. Specifically, a safety evaluation study of bosentan revealed a correlation between its use and the occurrence of PE [60]. However, the underlying mechanism remains to be further explored. Additionally, direct oral anticoagulants (DOACs) are widely used to prevent embolic events in atrial fibrillation patients, but recent study increasingly suggests an association between DOACs and hemorrhagic PE [15, 61, 62]. Possible reasons for this phenomenon may include drug accumulation due to hepatorenal insufficiency and drug-drug interactions [63].

To explore the temporal patterns of DPE and offer a scientific foundation for optimizing drug administration, we conducted a systematic analysis of the TTO of DPE using the FAERS database. Our findings show that the median TTO for DPE is the shortest with cytarabine, at 14 days, and the longest with digoxin, up to 1624 days. Remarkably, the median TTO for most drugs causing PE is within three months. Additionally, the results of our WSP test indicate that the incidence of DPE decreases over time. This finding suggests that the risk of DPE is highest in the initial treatment stages and progressively declines thereafter. Consequently, clinicians should closely monitor patients in the early stages of medication.

Among the drugs ranked according to signal strength, pergolide, arsenic trioxide, and loncastuximab tesirine are the top three drugs with the highest signal strength. Pergolide, a dopamine receptor agonist used to treat Parkinson’s disease, has been proven to cause valvular tissue fibrosis and thickening by interfering with the metabolism of 5 - hydroxytryptamine and the expression of genes related to its receptors and transporters, ultimately leading to cardiac valve regurgitation [64,65,66]. Recent studies have shown that the adverse effects of pergolide also extend to the pericardial tissue, promoting its fibrosis and potentially developing into constrictive pericarditis [67]. Both of these conditions can impair the heart’s pumping function, leading to congestion in the systemic and pulmonary circulations, elevation of venous pressure, and possible obstruction of pericardial lymphatic return, thus resulting in PE. Arsenic trioxide is commonly used in the treatment of acute promyelocytic leukemia. Multiple studies have shown that patients treated with arsenic trioxide may develop PE or pericarditis [68,69,70], and the underlying mechanism may be related to the cardiotoxicity induced by arsenic trioxide [71]. Loncastuximab tesirine is a novel antibody-drug conjugate. It is noteworthy that the FDA’s drug label has clearly indicated that the use of this drug may cause severe adverse reactions such as effusion and edema. Due to its short marketing time, there is a lack of sufficient clinical evidence. However, in clinical practice, attention still needs to be paid to the potential risk of inducing PE.

This study has several limitations: (1) The spontaneous reporting nature of the FAERS database leads to a proclivity for recording severe events with clear causal relationships, introducing an inherent bias. Concurrently, issues such as data gaps, including the absence of crucial information like patient age, gender, and drug duration, along with under-reporting, have a high likelihood of causing an underestimation or overestimation of risk signals. This, in turn, can impact the accuracy and reliability of conclusions. (2) The majority of reports originate from European and American countries, with limited representation from Asia, which may introduce racial bias. (3) The detected positive signals merely indicate a statistical correlation between drugs and adverse events; clinical trials are necessary to establish their clinical significance. Despite these limitations, our study marks the first exploratory analysis of the FAERS database to identify potential drugs that may induce PE.

Conclusion

In this study, the FAERS database was utilized to identify 26 drugs potentially associated with PE. Notably, 19 of these drugs did not list PE as an ADE in their drug labels. In future research, it is advisable to further integrate electronic health records (EHR) and insurance claim databases to enhance the accuracy and comprehensiveness of the dataset. In addition, logistic regression can be applied to re-evaluate the research findings. Such efforts will enable a deeper understanding of DPE, enhance the precision of drug safety assessment, and provide a more solid basis for decision-making in clinical practice.