Analysis of drug-induced pulmonary embolism risk based on the Food and Drug Administration Adverse Event Reporting System database

Introduction

Pulmonary embolism (PE), one of the most severe clinical manifestations of venous thromboembolism (VTE), accounts for 5% to 10% of unexpected in-hospital deaths (1). While traditional risk factors, such as surgery and prolonged immobilization, are widely recognized, the risk of drug-induced PE has not received adequate attention (2,3). In recent years, the widespread application of tumor-targeting drugs, immune checkpoint inhibitors, and novel biological agents has brought the risk of drug-induced PE to the forefront of drug safety concerns. Studies indicate that the incidence of VTE in cancer patients undergoing chemotherapy with agents like bevacizumab can be as high as 12.6% (4). However, existing research is primarily limited to individual drugs or small-sample observational studies, and there remains a significant gap in systematic analyses regarding the epidemiological characteristics, temporal distribution patterns, and risk stratification of drug-induced PE. This gap results in an insufficient basis for developing precise clinical prevention and monitoring strategies.

The U.S. Food and Drug Administration Adverse Event (AE) Reporting System (FAERS), recognized as the world’s largest drug safety database, has documented over 20 million AE reports submitted by healthcare professionals, consumers, and pharmaceutical companies (5). Previous studies have validated the sensitivity of FAERS in detecting the thromboembolic risks associated with Janus kinase (JAK) inhibitors and the PE risks linked to antipsychotic medications (6,7). This study aims to comprehensively identify potential drug-induced causes of PE using the FAERS database, thereby providing valuable insights for clinicians. We present this article in accordance with the STROBE reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0005/rc).

Methods

Source of data

The data for this study are derived from AE reports in the FAERS open-source database, primarily covering the period from the first quarter of 2004 to the fourth quarter of 2024. All AE reports relevant to the study objectives were extracted using “PE” as the preferred term (PT). The dataset mainly comprises seven tables: demographic information (DEMO), drug information (DRUG), drug therapy duration (THER), indications for use (INDI), AE records (REAC), sources of AEs (RPSR), and patient outcomes (OUTC) (5). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Data processing

In the process of data mining, we adhered to the FDA-recommended deduplication guidelines and eliminated problematic records. We extracted the CASEID, FDA_DT, and PRIMARY ID fields from the DEMO table and organized them in a specific order. When multiple reports share the same CASEID, the report with the maximum FDA_DT value is retained to ensure it contains the most recent and comprehensive information. In cases where reports have identical CASEID and FDA_DT values, the report with the largest PRIMARYID value is kept to maintain data integrity. During the analytical process, we employed a precision definition strategy by exclusively selecting reports that used the Medical Dictionary for Regulatory Activities (MedDRA Version 27.1) PT “PE” as the primary suspected event (8). This precise PT was chosen due to its highest clinical diagnostic specificity, which maximally reduces the inclusion of non-PE thrombotic events such as deep vein thrombosis and non-specific embolisms. Consequently, this approach ensures that the core analytical results possess a high degree of specificity. It is also noteworthy that the FAERS database encounters issues with low standardization in drug nomenclature during the data collection phase. This challenge primarily arises from the diversity of report sources, which may result in inconsistent representations of the same drug during the submission process, such as generic names, brand names, or drug codes. To enhance data accuracy, this study implemented a standardized processing method for all drugs. Additionally, we adopted the World Health Organization’s anatomical therapeutic chemical (ATC) classification system to systematically categorize various types of drugs.

To pinpoint drug-related PE risks with significant potential public health implications, this study employed a stepwise, focused approach. Initially, we conducted a proportional imbalance analysis on all drugs in the FAERS database that were associated with at least one PE report. Subsequently, recognizing that the primary objective of spontaneous reporting systems is to identify and characterize safety signals with the highest reporting burden and the broadest potential impact, we selected the top 50 drugs with the highest PE report frequencies for in-depth analysis. This prioritization is justified because drugs with higher reporting frequencies typically correspond to a broader clinical user base and a greater absolute number of associated cases, making them more pertinent from a public health surveillance perspective. The substantial number of reports for these drugs provides a more stable statistical foundation for reliably estimating signal strength, effectively mitigating the volatility caused by insufficient reporting data.

Time-to-onset (TTO) analysis

To assess the temporal relationship between drug exposure and PE events, we analyzed cases that reported both the medication start date and the event date. The time interval from drug initiation to PE onset was calculated and described using median values along with the first (Q1) and third (Q3) quartiles to capture both central tendency and dispersion. Based on the distribution characteristics of TTO, we preliminarily categorized the risk patterns into two groups: (I) early/acute risk (TTO ≤90 days) and (II) delayed/chronic risk (TTO >90 days). This stratification aims to explore potential mechanistic differences underlying the risk under various treatment modalities.

Statistical analysis

To identify noteworthy drug-PE safety signals from the extensive spontaneous reporting data, we employed a proportional imbalance analysis method widely utilized in international pharmacovigilance research. The core premise is as follows: if a drug is genuinely associated with PE, the proportion of PE reports mentioning that drug should be significantly higher than the proportion of other AE reports mentioning the same drug. In this study, we primarily utilized the reporting odds ratio (ROR) and the Bayesian confidence propagation neural network (BCPNN) methods for signal detection (9). This study adopted the disproportionality analysis method for signal mining of AEs associated with medications, which was fundamentally based on a 2×2 four-fold table statistical framework. Rows were classified into target drugs and other drugs, while columns were classified into target AEs and other AEs. The cells a, b, c, and d respectively represented the number of reports of target drug-target AE, target drug-other AE, other drug-target AE, and other drug-other AE. Combined with the marginal and grand totals (a+b, c+d, a+c, b+d, and a+b+c+d), the ROR and BCPNN were calculated. The specific algorithms of ROR and BCPNN were detailed in Table 1. The criteria for a positive ROR signal require that the number of reports (a) be ≥3, and the lower limit of the 95% confidence interval (CI) for the ROR must exceed 1. BCPNN signal detection primarily relies on the information component (IC). The IC is a Bayesian-derived metric that quantifies the strength of a drug-event combination relative to the background reporting rate. An IC value greater than 0 suggests a potential signal, and an IC lower limit (IC-2SD) above 0 is typically considered a statistically significant signal. These two methods identify signals that are significantly higher than expected by comparing the reporting proportion of specific drug-event combinations to the background proportion. While these signals indicate potential drug safety concerns, they may stem from genuine pharmacological risks or confounding factors. Therefore, all findings from this study should be regarded as tools for hypothesis generation rather than definitive evidence of causality. All data processing and statistical analyses were performed in R (Version 4.3.2), primarily utilizing the data.table, dplyr, ggplot2, and PhViD packages.

Results

In this study, we retrieved data from the FAERS database spanning 80 quarters, from the first quarter of 2004 to the fourth quarter of 2024. After excluding duplicate and questionable data, a total of 18,278,243 complete data reports were included. After conducting data mining, we discovered that there are 1,459 drugs associated with the AE of PE, affecting 86,810 patients. Among them, the top 50 drugs led to PE affecting a cumulative total of 50,513 patients. The data mining process is illustrated in Figure 1.

Figure 1 The data mining flow chart of this study. The comprehensive process of data cleaning and mining employed in this study. This research conducted a disproportionality analysis based on the top 50 drugs. DEMO, demographic information; DRUG, drug information; PS, primary suspect; REAC, adverse event records.

Baseline characteristics

We conducted a baseline analysis of all AE reports collected using “PE” as the keyword. The report, categorized by gender distribution and excluding reports with unspecified gender, indicates that female patients account for 57.1% (n=49,529), while male patients constitute only 34.7% (n=30,131) (Figure 2A). In terms of patient weight distribution, the highest proportion of patients had a weight between 50 and 100 kg, accounting for 29.3% (n=25,406) (Figure 2B). Regarding the identity of the reporter, the majority were physicians (n=28,876, 33.3%), followed by consumers (n=21,664, 25.0%) and other health professionals (n=13,770, 15.9%) (Figure 2C). When conducting data statistics, if a patient reports AEs related to PE with multiple outcomes, we prioritize recording the most severe consequence. If identical outcomes occur in different years or quarters, we consider them distinct event results. After analysis, the most common outcome was hospitalization—initial or prolonged (n=41,477, 47.8%), followed by other serious (important medical event) (n=20,423, 23.5%) and death (n=15,088, 17.4%) (Figure 2D). We conducted an analysis of the top 10 countries by the number of reports, with the United States accounting for the majority, representing 55.1% of all reported cases (n=47,869), followed by France (n=5,184, 6.0%), Germany (n=4,642, 5.4%), Canada (n=4,329, 5.0%), and the United Kingdom (n=4,231, 4.9%) (Figure 2E). The specific data for the above baseline analysis can be found in Table 2. Regarding age, after excluding missing data, we observed that the patients were predominantly between 51 and 80 years old, with the highest number of patients in the 61 to 70 age group, totaling 12,056 cases. Notably, there were also 6,118 patients in the age group of 21 to 30 (Figure 2F). Regarding the annual distribution of the reports, there has been a significant increase in the number of reports since 2011, particularly during the period from 2012 to 2014, when the annual number of reports consistently exceeded 5,500 (Figure 2G).

Figure 2 The baseline characteristics of the patients and reports of AEs included in this study. (A) Gender distribution of the included patients; (B) weight distribution of the included patients; (C) distribution of reporter occupations in the reports; (D) outcomes of the included patients; (E) number of reports from different countries; (F) age distribution of the included patients; (G) distribution of reported cases by year. AE, adverse event.

Disproportionality analysis

To assess the potential association between the drug and PE, we conducted a disproportionality analysis using the ROR and BCPNN, with all detailed data information available in https://cdn.amegroups.cn/static/public/10.21037jtd-2026-1-0005-1.xlsx. In the course of our data analysis, we identified the top 50 drugs associated with frequent AEs of PE (Table 3). Utilizing the ATC classification system, we categorized these 50 drugs into eight distinct groups. An analysis of the FAERS database revealed that the primary drug categories with strong reporting associations to PE included: antineoplastic and immunomodulating agents (ATC L, n=29, 58%), blood and blood-forming organs (ATC B, n=6, 12%), nervous system (ATC N, n=6, 12%), genitourinary system and sex hormones (ATC G, n=4, 8%), and musculoskeletal system (ATC M, n=2, 4%) (Figure 3). Furthermore, upon further classification of the most common antineoplastic and immunomodulating agents according to the third level of the ATC classification, we found that immunosuppressants (ATC L04A, n=9, 31.1%) and monoclonal antibodies and antibody-drug conjugates (ATC L01F, n=5, 17.3%) were the two primary drug categories. Additionally, we conducted a detailed classification of drugs for blood and blood-forming organs at the third level of the ATC classification, revealing that the category of drugs causing PE within this class of systemic medications consisted exclusively of antithrombotic agents (ATC B01A, n=6, 100%). We also classified the drugs for the nervous system according to the third level of the ATC classification, with antipsychotics (ATC N05A, n=5, 83.3%) emerging as the most predominant drug category.

Figure 3 Classification of top 50 drugs associated with AE of PE. Figure 3 displays 50 drugs associated with PE, subdivided into eight main categories according to ATC classification rules, and clearly illustrates the proportion of each category. AE, adverse event; ATC, anatomical therapeutic chemical; PE, pulmonary embolism.

Among the 50 drugs analyzed, the overall statistical frequency of reported cases ranged from 307 to 9,246, while the ROR varied from 0.27 to 68.64, IC varied from −1.84 to 5.58 (Table 3). The five drugs with the highest number of reported cases are drospirenone/ethinylestradiol (n=9,246), rivaroxaban (n=4,166), ethinylestradiol/etonogestrel (n=3,890), lenalidomide (n=3,124), and testosterone (n=2,257). Among the 50 drugs statistically analyzed, 42 exhibited positive signals associated with cough (Figure 4). The top five drugs with the highest ROR values are drospirenone/ethinylestradiol (ROR: 68.64, IC: 5.58), ethinylestradiol/etonogestrel (ROR: 56.51, IC:5.43), ethinylestradiol/norelgestromin (ROR: 14.28, IC: 3.73), rofecoxib (ROR: 13.61, IC: 3.65), and testosterone (ROR: 13.58, IC: 3.65). Furthermore, we found that the RORs of the four drugs classified under the genitourinary system and sex hormones (ATC G, n=4, 8%) category were all greater than 10.

Figure 4 The ROR for each of the top 50 drug-induced PE reports. Figure 4 presents the forest plot of the top 50 drugs examined in this study, of which 42 drugs demonstrate positive signals associated with PE (drugs marked with * in the forest plot indicate those with positive signals). ATC, anatomical therapeutic chemical; CI, confidence interval; PE, pulmonary embolism; ROR, reporting odds ratio.

Temporal analysis of drug exposure to pe onset

Among the cases with definitive temporal information (n=31,832), the interval between drug exposure and PE onset exhibited a remarkably broad distribution. The median TTO was 84 days (Q1 =19 days, Q3 =355 days), with an interquartile range (IQR) of 336 days.

Discussion

In this study, we utilized the FAERS database from 2004 to 2024 to conduct a comprehensive analysis of drug-related PE AEs. We examined the top 50 drugs that exhibited the highest frequency of drug-related PE and found that 42 of these drugs demonstrated a positive association with PE. The following discussion aims to analyze the patterns of data associations identified in this study and to explore the potential clinical and biological hypotheses. All inferences drawn herein require validation through subsequent research.

During the baseline analysis, we observed that the number of female patients exceeded that of male patients. This is consistent with previous research findings (10). In females, estrogen contributes to a hypercoagulable state by elevating the levels of coagulation factors (such as VII and X) and inhibiting the activity of antithrombin III (11). Women of childbearing age who utilize oral contraceptives or hormone replacement therapy may also face an increased risk of drug-induced PE (12). This finding aligns with our study’s observations regarding the age distribution of patients, where the incidence of PE is generally higher among younger individuals (21–30 years old). Recent years have seen a rise in immune diseases, such as Crohn’s disease, among younger patients. The use of JAK inhibitors (e.g., tofacitinib) for these conditions is associated with a risk of thrombosis, potentially contributing to the trend of younger patients developing PE (13,14). Tumor types more prevalent in women, such as breast cancer, are frequently treated with medications that may heighten the risk of thrombosis, including bevacizumab (15). In contrast, potent prothrombotic drugs are less commonly employed in the treatment regimens for tumors more frequently found in men, such as prostate cancer. Additionally, female patients are often more proactive in reporting adverse reactions, whereas male patients may underestimate symptoms such as chest pain and dyspnea, leading to underreporting. In this study, we observed that the proportion of patients with a body weight ranging from 50 to 100 kg was 29.3%. Many medications necessitate dose adjustments based on body weight (e.g., bevacizumab at 5–10 mg/kg). For overweight patients, this can lead to reduced efficacy due to insufficient dosing or an increased risk of toxicity from excessive dosing. However, our findings indicate that the risk was not significantly elevated in the group exceeding 100 kg (only 7.8%), suggesting that drug exposure is not the sole determinant of risk. In cases of drug-induced PE, hospitalization is primarily utilized, underscoring the complexity of acute phase management that necessitates multidisciplinary collaboration (including anticoagulation, thrombolysis, and interventional thrombectomy). Among hospitalized cases, approximately 20% of patients experience prolonged hospital stays due to bleeding complications, highlighting the need for individualized anticoagulation strategies (16).

This study conducted a systematic analysis of drugs associated with PE using the FAERS database and the ATC classification system. We found that antineoplastic and immunomodulating agents (ATC L) were the primary category linked to the risk of PE. Within the three-tier classification of ATC L, immunosuppressants (ATC L04A) and monoclonal antibodies and antibody-drug conjugates (ATC L01F) accounted for the highest proportions. In the ATC L04A category, lenalidomide (ATC L04AX04, n=3,124) exhibited the highest frequency of PE occurrence, while in the ATC L01F category, bevacizumab (ATC L01FG01, n=1,010) also showed a significant frequency, both indicating positive signals. Immunosuppressants, particularly lenalidomide, can target the ubiquitin-proteasome pathway, reducing the degradation of apoptotic factors. Concurrently, they can activate the NF-κB pathway, leading to the upregulation of pro-inflammatory factors [such as interleukin (IL)-6 and IL-8] and pro-coagulant factors [such as tissue factor (TF) and von Willebrand factor (VWF)], thereby diminishing downstream signaling and contributing to the vicious cycle of inflammation and coagulation (17). Lenalidomide can also directly inhibit endothelial cell migration and angiogenesis, resulting in decreased vascular repair capacity (18). Furthermore, when lenalidomide is used in conjunction with dexamethasone, it further inhibits the fibrinolytic system through the upregulation of plasminogen activator inhibitor-1 (PAI-1) induced by glucocorticoids, thereby synergistically increasing the risk of thrombosis (19). Research indicates that the risk of this type of thrombosis can be as high as 25% (20). PE induced by bevacizumab is primarily associated with vascular endothelial growth factor (VEGF), which is essential for maintaining vascular integrity. Bevacizumab neutralizes VEGF-A, inhibiting the proliferation and survival of endothelial cells, which leads to the exposure of subendothelial collagen and triggers platelet aggregation (21). Bevacizumab can also disrupt the structure of tumor blood vessels, reducing oxygen supply and resulting in hypoxia within the tumor. Hypoxic conditions activate hypoxia-inducible factor-1α (HIF-1α), which directly binds to the PAI-1 promoter, promoting its expression and leading to the formation of a regional hypercoagulable state (22). Chemotherapeutic agents, such as paclitaxel, inhibit the proliferation and migration of endothelial cells by affecting microtubule function, thereby exerting anti-angiogenic effects and synergistically disrupting vascular stability with bevacizumab (23). This aligns with our findings in this study, where paclitaxel (ATC L01CD01, n=387) was ranked among the top 50 drugs associated with PE and showed a positive correlation (ROR =2.34). Additionally, it is noteworthy that lenalidomide, as a first-line treatment for multiple myeloma, is in high demand globally, which may amplify the absolute number of reported adverse reactions to this drug. Similarly, the indications for bevacizumab include colorectal cancer and ovarian cancer, among others. The treatment of these diseases typically involves combination chemotherapy regimens (such as the FOLFOX regimen), and chemotherapeutic agents (such as oxaliplatin) may damage endothelial cells, thereby exacerbating the risk of thrombosis (24).

Our findings indicate that blood and blood-forming organs (ATC B) represent the second largest category of drugs associated with PE surpassed only by antineoplastic and immunomodulating agents. After conducting a statistical analysis of the top 50 drugs, we found that all of these drugs belong to the category of antithrombotic agents (ATC B01A). Antithrombotic drugs are commonly employed for the prevention or treatment of thrombotic events. However, in this study, the reporting frequency and risk signal intensity of these drugs associated with PE were significantly elevated, particularly for rivaroxaban (ATC B01AF01), which exhibited the most pronounced results (n=4,166, ROR: 7.81, 95% CI: 7.57–8.06). Other antithrombotic drugs ranked within the top 50, such as apixaban and enoxaparin, also demonstrated positive risk signals (ROR >1). Taking rivaroxaban as an example, this drug exerts its anticoagulant effect by inhibiting factor Xa; however, its efficacy is influenced by renal function (approximately 36% is excreted via the kidneys) and drug interactions, particularly when co-administered with potent CYP3A4/P-gp inhibitors (25). Rivaroxaban reduces thrombin generation by inhibiting factor Xa but may compensatorily activate platelet membrane receptors, promoting platelet aggregation (26). Long-term use of factor Xa inhibitors can also lead to increased expression of coagulation factors VII and XI, potentially resulting in a rebound hypercoagulable state after discontinuation (27). When rivaroxaban is co-administered with CYP3A4/P-gp inhibitors, such as amiodarone, the pharmacokinetics of rivaroxaban may be affected, leading to an increase in its plasma concentration and thereby increasing the risk of bleeding (28). Following a bleeding event, blood transfusion or the implementation of hemostatic measures (e.g., the use of tranexamic acid) may indirectly promote thrombosis formation. In clinical practice, the use of low-dose rivaroxaban for prophylactic treatment (e.g., 10 mg/day) may not effectively inhibit thrombin generation, particularly in high-risk patients (e.g., those with cancer or obesity). Poor medication compliance among patients, characterized by missed doses or arbitrary dosage reductions, can lead to fluctuations in drug concentration and result in an intermittent hypercoagulable state. In comparison to apixaban (ATC B01AF02), another factor Xa inhibitor, rivaroxaban presents a significantly higher risk of PE, with an ROR of 7.81 compared to Apixaban’s ROR of 3.36. This discrepancy may be attributed to insufficient drug concentrations resulting from rivaroxaban’s shorter half-life of 5–9 hours versus apixaban’s 12 hours (29). Previous randomized controlled trials, such as the EINSTEIN-PE study, have demonstrated the efficacy and safety of rivaroxaban in treating PE; however, this study underscores its potential thrombosis risk (30). This risk is primarily due to the stringent inclusion criteria of the EINSTEIN-PE study, which excluded patients with a creatinine clearance rate below 30 mL/min, whereas real-world patients often present with complex comorbidities. While the EINSTEIN-PE study focused primarily on symptomatic recurrent VTE, the FAERS database utilized in this study encompasses all reported PE cases, including both new and recurrent instances. Additionally, the follow-up period in the EINSTEIN-PE study was limited to a maximum of 12 months, while the FAERS database includes delayed thrombotic events in patients on long-term medication, facilitating a more comprehensive evaluation of drug safety. Moreover, we observed significantly strong signals from heparin-based preparations, such as enoxaparin (ATC B01AB05, ROR =7.94) and heparin (ATC B01AB01, ROR =7.04). The pronounced signal associated with heparin strongly underscores the necessity of focusing on heparin-induced thrombocytopenia (HIT) (31,32). HIT is a severe, antibody-mediated complication triggered by medication, with venous thrombosis being its hallmark manifestation (33). When heparin binds to platelet factor 4 (PF4) to form an antigenic complex, it stimulates the production of IgG antibodies. These antibodies subsequently activate platelets via Fc receptors, leading to the release of procoagulant microparticles and TF, ultimately culminating in thrombosis (34). Notably, despite enoxaparin being a low molecular weight heparin with a lower risk of typical HIT compared to unfractionated heparin, it exhibits the highest ROR (35). This discrepancy may reflect the broader clinical application spectrum of enoxaparin, which is frequently employed for perioperative thromboprophylaxis, cancer-associated thrombosis treatment, and anticoagulation in patients with renal insufficiency. Consequently, in this context, PE events may represent “treatment failure” or the natural progression of underlying disease rather than a direct prothrombotic effect of the drug.

In addition to the aforementioned two major drug categories, the nervous system (ATC N) constitutes the third largest risk category for drug-induced PE, with antipsychotics (ATC N05A) being the predominant contributors. Olanzapine (ATC N05AH03), a second-generation antipsychotic, demonstrates a significantly higher reporting frequency and signal strength (ROR =3.54) compared to its counterparts, such as quetiapine (ROR =1.82) and risperidone (ROR =1.18). Olanzapine primarily modulates psychiatric symptoms through multi-receptor actions, including 5-HT2A receptor antagonism; however, this medication carries a certain risk of PE. Notably, olanzapine can lead to substantial weight gain, which is one of its most common side effects. Obesity is an independent risk factor for PE and may indirectly increase the risk of thrombosis by promoting inflammatory states, insulin resistance, and a hypercoagulable state (36). Olanzapine can enhance adenosine diphosphate (ADP)-induced platelet aggregation by inhibiting the 5-HT2A receptor, thus reducing cyclic adenosine monophosphate (cAMP) levels in platelets (37). Olanzapine can also antagonize α1 receptors, leading to venous vasodilation and decreased blood flow velocity, thereby contributing to the ‘venous stasis’ factor in Virchow’s triad (38). Our findings indicate that the risk of PE induced by olanzapine (ROR =3.54) is significantly higher than that of other similar drugs, such as aripiprazole (ROR =1.3), possibly due to the more severe metabolic disorders caused by olanzapine’s potent H1 receptor antagonism (39). It is also crucial to highlight that patients using antipsychotic medications, particularly those with schizophrenia or bipolar disorder, often exhibit unhealthy lifestyle habits, including smoking, sedentary behavior, and irregular eating patterns. These factors, independent of medication effects, already place them at a high risk for PE. In the specific process of medication, olanzapine is frequently combined with mood stabilizers, such as sodium valproate, or antidepressants, like selective serotonin reuptake inhibitors (SSRIs). Consequently, psychiatrists must increase their awareness of the need to screen for thrombotic events.

We have identified that the genitourinary system and sex hormones (ATC G) represent a high-risk category for drug-induced PE (ROR >10). Among these, hormonal contraceptives for systemic use (ATC G03A) and topical contraceptives (ATC G02B) are the most prevalent. Analysis of specific medications reveals that drospirenone/ethinylestradiol (ATC G03AA12) has the highest frequency of PE reports (n=9,246) and signal strength (ROR =68.64, 95% CI: 67–70.32). The PE induced by drospirenone/ethinylestradiol primarily depends on the synergistic prothrombotic effects of its two components. This drug induces a hypercoagulable state in the body by activating estrogen receptors in the liver, which upregulates the levels of coagulation factors II, VII, VIII, X, and fibrinogen, while simultaneously inhibiting the synthesis of antithrombin III and protein S (11). Estrogen can also enhance the oxidative stress response of endothelial cells and expose subendothelial collagen by reducing the bioavailability of nitric oxide (40). Drospirenone, a synthetic progestin, enhances signal transduction mediated by the VWF receptor glycoprotein Ib-IX-V and promotes botrocetin/VWF-induced platelet aggregation (41). Compared to drospirenone/ethinylestradiol, ethinylestradiol/etonogestrel (ATC G02BB01), as a topical contraceptive, exhibits a lower systemic absorption rate, with the blood concentration of ethinylestradiol being approximately 10% of that from oral administration. However, its PE signal intensity is significantly increased (ROR =56.51). Although some local contraceptives utilize transdermal drug delivery, their active ingredients can still enter systemic circulation, leading to prolonged exposure to low levels of estrogen. Other local formulations, such as vaginal rings, may induce local inflammatory responses due to mechanical irritation or material allergies, promoting the release of pro-inflammatory cytokines [e.g., IL-6, tumor necrosis factor-alpha (TNF-α)] that activate the expression of TF in endothelial cells, thereby initiating the coagulation cascade (42). In addition to estrogen-induced PE, testosterone (ATC G03BA03, ROR =13.58) also poses a risk of promoting thrombus formation. Testosterone stimulates hematopoiesis in the bone marrow, increases hematocrit, and consequently raises blood viscosity, which can lead to venous stasis (43). Prolonged or high-dose testosterone use may also damage vascular endothelial cells through oxidative stress or inflammatory responses, promoting TF exposure and thrombin generation, thus activating the extrinsic coagulation pathway (44).

This study quantitatively delineated the temporal distribution of drug-related PE risk, revealing significant time heterogeneity. Approximately 25% of events occurred within 19 days of drug initiation, strongly suggesting that clinicians must maintain heightened vigilance during the first month of treatment—particularly for agents such as heparins and novel oral anticoagulants used for acute prevention or therapy—and closely monitor for rapid-onset complications, such as HIT. The longer median time and third quartile values indicate that for patients undergoing targeted cancer therapy, immunomodulatory treatment, or long-term hormonal therapy, the risk of PE persists as a continuous threat throughout the entire treatment course. This finding necessitates the establishment of dynamic, long-term monitoring strategies rather than assessments confined to the early phase of drug exposure. In summary, early risks are more likely directly associated with acute pharmacologic toxicities (e.g., endothelial injury, platelet activation), whereas delayed risks may be linked to cumulative endothelial dysfunction, metabolic alterations, chronic inflammatory states, or immune system remodeling induced by prolonged drug exposure.

In this study, we identified PE-related reports using the highly specific MedDRA PT “PE.” This approach was intended to precisely capture clinically confirmed PE events while minimizing misclassification bias that could arise from the inclusion of non-specific thrombotic events such as “pulmonary thrombosis” or “venous embolism”. This strategy ensures a high degree of relevance between the core risk signals and the target disease. However, this selection may not encompass cases that, although reported under other related terms, are clinically consistent with PE. Therefore, conducting a sensitivity analysis using broader standardized MedDRA queries (SMQs) or grouping by PTs would be an ideal method to verify the robustness of our findings. While this study focuses on the most definitive events to establish a highly specific signal, we acknowledge that the lack of such an analysis represents a limitation of our work. It is noteworthy that the primary drug categories identified as high-risk (e.g., antitumor agents, anticoagulants, sex hormones) are already supported by a substantial body of literature regarding their pro-thrombotic or high-risk pathophysiological mechanisms. Hence, we hypothesize that even with an expanded definition of events, the overall pattern of significant signals from these core categories is unlikely to undergo fundamental changes. Future studies employing the SMQ approach for a broader analysis will further confirm and refine these associations.

Limitations

This study has several limitations. First, spontaneous reporting data inherently contain biases, such as reporting bias and confounding by indication. Consequently, the observed associations cannot be directly equated to causal relationships, nor are they sufficient evidence to inform specific clinical management guidelines. Second, although our analytical framework—focusing on the top 50 drugs by report frequency—was grounded in considerations of public health impact and statistical stability, it may inadvertently downplay the potential risks associated with newer agents that have a narrower usage spectrum and fewer reports. This is an inherent trade-off in any study that prioritizes analysis based on report frequency. Finally, as a retrospective pharmacovigilance study, our research solely elucidates statistical associations between drugs and AEs, without establishing causality. These findings provide clear hypotheses that warrant priority in future epidemiological and pharmacological investigations.

Conclusions

This study leveraged the FAERS database to identify the 50 drugs most strongly associated with reports of PE. It was crucial to emphasize that research based on spontaneous reporting data is inherently exploratory and hypothesis-generating; consequently, the findings of this study should be considered preliminary risk signals. Our analysis revealed that 42 drugs were significantly positively associated with PE, with risk signals predominantly clustering within the categories of antitumor/immune-modulating agents, anticoagulants, and sex hormones. This pattern suggested that drug-related PE risk might have possessed a specific pharmacological basis and could be linked to patients’ underlying disease states. These findings underscore the necessity for clinicians to be vigilant about these potential risk signals, particularly when prescribing these drugs for long-term therapy in patients with pre-existing thrombotic risk. Prospective studies, large-scale cohort investigations, and mechanistic research are essential for confirming these associations, quantifying the true risk, and elucidating the underlying biological mechanisms.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0005/rc

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0005/prf

Funding: This work was funded by Kunshan Key Laboratory of Chronic Cough (No. KZ202401), Key R&D Program of Jiangsu Province (No. BE2022761), Shandong Second Provincial General Hospital Scientific Research Fund (No. 2023MS09) and Suzhou Key Medical Discipline Funding Project (No. szzdxk1902).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0005/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Rajapakse S. Thromboprophylaxis: striking a balance between under-treatment and over-treatment. Int J Clin Pract 2015;69:4-5. [Crossref] [PubMed]
2. Freund Y, Cohen-Aubart F, Bloom B. Acute Pulmonary Embolism: A Review. JAMA 2022;328:1336-45. [Crossref] [PubMed]
3. Li J, Wang X, Mei L, et al. Integration of Surgical and Coagulation Risk Factors for Predicting Postoperative Pulmonary Embolism in Thoracic Surgery: A Multi-Center Retrospective Study. Int J Gen Med 2025;18:6821-32. [Crossref] [PubMed]
4. Khorana AA, Dalal M, Lin J, et al. Incidence and predictors of venous thromboembolism (VTE) among ambulatory high-risk cancer patients undergoing chemotherapy in the United States. Cancer 2013;119:648-55. [Crossref] [PubMed]
5. Sakaeda T, Tamon A, Kadoyama K, et al. Data mining of the public version of the FDA Adverse Event Reporting System. Int J Med Sci 2013;10:796-803. [Crossref] [PubMed]
6. Setyawan J, Azimi N, Strand V, et al. Reporting of Thromboembolic Events with JAK Inhibitors: Analysis of the FAERS Database 2010-2019. Drug Saf 2021;44:889-97. [Crossref] [PubMed]
7. Huang J, Zou F, Zhu J, et al. Association between antipsychotics and pulmonary embolism: a pharmacovigilance analysis. Expert Opin Drug Saf 2025;24:1259-64. [Crossref] [PubMed]
8. Tieu C, Breder CD. A Critical Evaluation of Safety Signal Analysis Using Algorithmic Standardised MedDRA Queries. Drug Saf 2018;41:1375-85. [Crossref] [PubMed]
9. Huang L, Guo T, Zalkikar JN, et al. A Review of Statistical Methods for Safety Surveillance. Ther Innov Regul Sci 2014;48:98-108. [Crossref] [PubMed]
10. Marinelli A, Dragonieri S, Portacci A, et al. Reconsidering Gender in Asthma: Is It All About Sex? A Perspective Review. J Clin Med 2025;14:2506. [Crossref] [PubMed]
11. Abou-Ismail MY, Citla Sridhar D, Nayak L. Estrogen and thrombosis: A bench to bedside review. Thromb Res 2020;192:40-51. [Crossref] [PubMed]
12. Douxfils J. Cost-effective screening strategy to prevent venous thromboembolism in combined oral contraceptive users. Front Endocrinol (Lausanne) 2025;16:1559162. [Crossref] [PubMed]
13. Rogler G. Efficacy of JAK inhibitors in Crohn's Disease. J Crohns Colitis 2020;14:S746-54. [Crossref] [PubMed]
14. Kotyla PJ, Engelmann M, Giemza-Stokłosa J, et al. Thromboembolic Adverse Drug Reactions in Janus Kinase (JAK) Inhibitors: Does the Inhibitor Specificity Play a Role? Int J Mol Sci 2021;22:2449. [Crossref] [PubMed]
15. Marchetti M, Gomez-Rosas P, Russo L, et al. Baseline Hemostatic Biomarker Assessment Identifies Breast Cancer Patients at High Risk for Venous Thromboembolism During Chemotherapy. Cancers (Basel) 2025;17:2712. [Crossref] [PubMed]
16. Pagkratis N, Matsagas M, Malli F, et al. Prevalence of Hemorrhagic Complications in Hospitalized Patients with Pulmonary Embolism. J Pers Med 2022;12:1133. [Crossref] [PubMed]
17. Comerford C, Glavey S, Quinn J, et al. The role of VWF/FVIII in thrombosis and cancer progression in multiple myeloma and other hematological malignancies. J Thromb Haemost 2022;20:1766-77. [Crossref] [PubMed]
18. Liu X, Zhang J, Yi T, et al. Decoding tumor angiogenesis: pathways, mechanisms, and future directions in anti-cancer strategies. Biomark Res 2025;13:62. [Crossref] [PubMed]
19. De Stefano V, Larocca A, Carpenedo M, et al. Thrombosis in multiple myeloma: risk stratification, antithrombotic prophylaxis, and management of acute events. A consensus-based position paper from an ad hoc expert panel. Haematologica 2022;107:2536-47. [Crossref] [PubMed]
20. De Stefano V, Za T, Rossi E. Venous thromboembolism in multiple myeloma. Semin Thromb Hemost 2014;40:338-47. [Crossref] [PubMed]
21. Patel SA, Nilsson MB, Le X, et al. Molecular Mechanisms and Future Implications of VEGF/VEGFR in Cancer Therapy. Clin Cancer Res 2023;29:30-9. [Crossref] [PubMed]
22. Mohammad MA, Kumar N, Ghosh S, et al. Estrogen and obesity synergistically suppress protein S via HIF1α, enhancing thrombosis potential. J Clin Invest 2025;135:e193976. [Crossref] [PubMed]
23. Beretta GL, Cassinelli G, Rossi G, et al. Novel insights into taxane pharmacology: An update on drug resistance mechanisms, immunomodulation and drug delivery strategies. Drug Resist Updat 2025;81:101223. [Crossref] [PubMed]
24. Mohelnikova-Duchonova B, Melichar B, Soucek P. FOLFOX/FOLFIRI pharmacogenetics: the call for a personalized approach in colorectal cancer therapy. World J Gastroenterol 2014;20:10316-30. [Crossref] [PubMed]
25. Ashton V, Kerolus-Georgi S, Moore KT. The Pharmacology, Efficacy, and Safety of Rivaroxaban in Renally Impaired Patient Populations. J Clin Pharmacol 2021;61:1010-26. [Crossref] [PubMed]
26. Dobbing HC, Webb KL, Ariëns RAS. Novel and experimental anticoagulant strategies beyond current direct oral anticoagulants. J Thromb Haemost 2026;S1538-7836(25)00919-5.
27. Frydrych M, Janeczek M, Małyszek A, et al. Prothrombotic Rebound After Discontinuation of Direct Oral Anticoagulants Therapy: A Systematic Review. J Clin Med 2024;13:6606. [Crossref] [PubMed]
28. Ding H, Wang Z, Wang J, et al. Co-Administration of Amiodarone Increases Bleeding by Affecting Rivaroxaban Pharmacokinetics in Patients with Atrial Fibrillation. Pharmaceutics 2024;16:1006. [Crossref] [PubMed]
29. Trujillo T, Dobesh PP. Clinical use of rivaroxaban: pharmacokinetic and pharmacodynamic rationale for dosing regimens in different indications. Drugs 2014;74:1587-603. [Crossref] [PubMed]
30. EINSTEIN–PE Investigators. Oral rivaroxaban for the treatment of symptomatic pulmonary embolism. N Engl J Med 2012;366:1287-97. [Crossref] [PubMed]
31. Liu L, Zheng H, Chen S, et al. Heparin-induced thrombocytopenia associated with low-molecular-weight heparin: clinical feature analysis of cases and pharmacovigilance assessment of the FAERS database. Front Pharmacol 2023;14:1247253. [Crossref] [PubMed]
32. Kunyu L, Shuping S, Chang S, et al. An Updated Comprehensive Pharmacovigilance Study of Drug-Induced Thrombocytopenia Based on FDA Adverse Event Reporting System Data. J Clin Pharmacol 2024;64:478-89. [Crossref] [PubMed]
33. May J, Cuker A. Practical guide to the diagnosis and management of heparin-induced thrombocytopenia. Hematology Am Soc Hematol Educ Program 2024;2024:388-95. [Crossref] [PubMed]
34. Ng JY, D'Souza M, Hutani F, et al. Management of Heparin-Induced Thrombocytopenia: A Contemporary Review. J Clin Med 2024;13:4686. [Crossref] [PubMed]
35. Junqueira DR, Zorzela LM, Perini E. Unfractionated heparin versus low molecular weight heparins for avoiding heparin-induced thrombocytopenia in postoperative patients. Cochrane Database Syst Rev 2017;4:CD007557. [Crossref] [PubMed]
36. Rasras H, Beghi M, Samti M, et al. Rare and Underestimated Association of Pulmonary Embolism and Olanzapine Therapy; Report of Two Cases. Arch Acad Emerg Med 2021;9:e17. [Crossref] [PubMed]
37. Dietrich-Muszalska A, Rabe-Jablonska J, Nowak P, et al. The first- and second-generation antipsychotic drugs affect ADP-induced platelet aggregation. World J Biol Psychiatry 2010;11:268-75. [Crossref] [PubMed]
38. Masopust J, Bazantova V, Kuca K, et al. Venous Thromboembolism as an Adverse Effect During Treatment With Olanzapine: A Case Series. Front Psychiatry 2019;10:330. [Crossref] [PubMed]
39. Oruch R, Rajab HA, Elderbi MA, et al. Drawbacks of Olanzapine Therapy: An Emphasis on Its Metabolic Effects and Discontinuation. J Clin Med 2025;14:8125. [Crossref] [PubMed]
40. SenthilKumar G. Estrogen and the Vascular Endothelium: The Unanswered Questions. Endocrinology 2023;164:bqad079. [Crossref] [PubMed]
41. Fan X, Chen X, Wang C, et al. Drospirenone enhances GPIb-IX-V-mediated platelet activation. J Thromb Haemost 2015;13:1918-24. [Crossref] [PubMed]
42. Al-Haddad S, Branham KKR, Clare CA. Advances in contraception: vaginal contraceptive rings. Ther Adv Reprod Health 2023;17:26334941231186733. [Crossref] [PubMed]
43. Liu J, Chin-Yee B, Ho J, et al. Diagnosis, management, and outcomes of drug-induced erythrocytosis: a systematic review. Blood Adv 2025;9:2108-18. [Crossref] [PubMed]
44. Koukoulis GN, Filiponi M, Gougoura S, et al. Testosterone and dihydrotestosterone modulate the redox homeostasis of endothelium. Cell Biol Int 2022;46:660-70. [Crossref] [PubMed]