PDF
Citation
Print
INTRODUCTION
Graft-versus-host disease (GVHD) is a severe complication for transplant patients, particularly those undergoing allogeneic hematopoietic stem cell transplantation [1]. To mitigate this risk, pharmacological prophylaxis is routinely employed. In clinical settings, a wide range of drugs is administered in various combinations for GVHD prophylaxis based on each patient’s condition and institutional protocols. These drugs include cyclosporine [2], mycophenolate mofetil (including mycophenolic acid) [2,3], methotrexate [3], tacrolimus [3], and others.
Previous studies of GVHD prophylaxis have primarily focused on monotherapies [4] or a limited number of drug combinations [2,3,5]. However, the extensive range of pharmacological regimens used in clinical practice far exceeds the scope of these investigations. Although it is possible to aggregate research findings for each drug, such aggregation may not accurately reflect the outcomes achieved with their combinations. This is because the effects of drug combinations can be classified into three categories [6,7]: additive effects, where the impacts of each drug are simply combined; synergism, where the combined drugs enhance the overall effect; and antagonism, where the combination reduces the overall effect. Understanding these interactions is crucial for clinical decision-making. Identifying synergistic combinations may allow clinicians to achieve greater efficacy with lower doses, whereas recognizing antagonistic combinations could help avoid ineffective or potentially harmful regimens. Moreover, knowledge of additive effects can inform decisions on whether to add a drug to an existing regimen, balancing potential benefits against increased complexity and cost. Thus, we hypothesized that the efficacy of a particular GVHD prophylaxis drug may vary depending on other prophylactic medications administered concurrently. The purpose of this study was to test this hypothesis and provide guidance on combinatorial treatment strategies for GVHD prophylaxis. These insights could directly inform treatment decisions, such as selecting the most effective drug combinations for specific patient profiles or adjusting dosages based on combination effects. Investigating a wide array of drug combination patterns requires a large sample size, making it impractical to achieve this objective solely through clinical trials.
To examine a large-scale sample, we utilized data from the U.S. Food and Drug Administration Adverse Event Reporting System (FAERS) [8], which encompasses adverse event reports from around the world. The FAERS database provides essential information for transplantation studies, including the purpose of drug administration and the types of adverse events reported. It has been widely used in studies evaluating drug safety based on the frequency of adverse events [9,10]. For prophylactic drugs, this approach can be extended to evaluate efficacy by examining whether the targeted adverse events—those the drugs are intended to prevent—occur following administration [11,12]. Therefore, in this study of GVHD prophylaxis drugs, we were able to assess both safety and efficacy. Advantageously, the FAERS data also include a wide array of drug combinations from its large-scale sample. This breadth and depth of analysis are particularly valuable for guiding complex treatment regimens that may not be addressed in more limited clinical trials or smaller-scale studies. However, it is crucial to approach research using the FAERS database with an awareness of its limitations, including potential biases from voluntary reporting and the presence of incomplete data [13]. Despite these constraints, FAERS remains a powerful tool for postmarketing surveillance and drug safety analysis when used judiciously.
METHODS
Because FAERS is an unlinked, anonymized database that is publicly available, Institutional Review Board approval was not required.
Data Source
The FAERS database is an unlinked, anonymized resource made publicly accessible on a quarterly basis since the first quarter of 2004 (2004Q1). Between 2004Q1 and 2012Q3, it was provided as the Adverse Event Reporting System (AERS). In 2012Q4, it was renamed the FAERS database and updated to encompass a more comprehensive dataset. On May 4, 2024, we retrieved AERS (aers_ascii_yyyyQq.zip) and FAERS (faers_ascii_yyyyQq.zip) data files from the official FAERS website, where “yyyy” and “q” denote the year and quarter, respectively. Several differences existed between the AERS and FAERS databases, including variations in variable names and the types of variables provided. To ensure consistency, the variable names were mapped between the AERS and FAERS data using the instructions provided with each dataset. Thus, for the remainder of this manuscript, AERS data are included in references to the FAERS dataset. The FAERS dataset comprises seven files, of which five were utilized in this study: DEMOyyQq.txt (patient demographic and administrative information), DRUGyyQq.txt (drug information), INDIyyQq.txt (indications for use), REACyyQq.txt (adverse event information), and THERyyQq.txt (drug therapy start and end dates), where “yy” denotes the last two digits of the year. New information was appended to existing data in FAERS by incrementing the safety report version number {caseversion} rather than overwriting existing entries. Therefore, only the data with the highest numerical value of {caseversion} for each patient were utilized. Throughout this manuscript, braces indicate variable names used in FAERS. Although the AERS data did not include a {caseversion} variable, similar discrimination was achieved using the two variables {ISR} and {CASE}.
Before conducting statistical analyses on the four background variables ({sex}, {age} at adverse event occurrence, {weight}, and country of the reporter {reporter_country}), it was necessary to standardize the units of measurement and address unexpected inputs. The methodologies for these data preprocessing techniques are detailed in Supplementary Material 1. Additionally, because the AERS database was occasionally missing required line breaks, we manually inserted line breaks at line 322,967 of DRUG11Q2.txt, line 247,896 of DRUG11Q3.txt, and line 446,738 of DRUG11Q4.txt before conducting statistical analyses.
Study Design
The inclusion criterion was patients documented in the FAERS database as having received GVHD prophylaxis drugs between 2004Q1 and 2024Q1. Patients were identified based on the preferred term for the indication for use {indi_pt}, recorded as “prophylaxis against graft-versus-host disease.”
The exclusion criterion was patients who had not received GVHD prophylaxis drugs prior to the occurrence of adverse events—that is, those who started administration of these drugs following the occurrence of adverse events. This determination was based on the recorded dates regarding when therapy was started (or restarted) for the drug {start_dt}, when the adverse event occurred or began {event_dt}, and when therapy was stopped for the drug {end_dt}. However, in some cases, {start_dt}, {event_dt}, and {end_dt} were recorded as only the year, as only the year and month, or were missing. As excluding cases with missing data for these three dates would have reduced the sample size, only those cases in which it could be reliably determined that GVHD prophylaxis drugs were started after the occurrence of adverse events were excluded. Details of this judgment are provided in Supplementary Material 2. Notably, FAERS reports typically represent a single adverse event occurrence per report, without longitudinal follow-up of the same patient. Given this structure, as well as the fact that drug administration generally precedes adverse event reporting in FAERS, we reasonably assumed that drug administration occurred before the adverse event in cases with partial or missing date information. This approach maximized the inclusion of potentially relevant data while maintaining the integrity of our analysis.
The primary endpoint was the occurrence of GVHD, characterized by the preferred term for adverse events in the Medical Dictionary for Regulatory Activities, {pt}. The adverse event names considered indicative of GVHD occurrence are listed in Supplementary Table 1. If multiple adverse events were reported for a single patient, and at least one of these events was related to GVHD, the patient was classified as having experienced GVHD.
Between 2004Q1 and 2014Q2, the FAERS database provided only the variable name of a medical product {drugname}, which was listed as either the generic or brand name of the drug. Between 2014Q3 and 2024Q1, the database provided the product active ingredient {prod_ai}, listed by the generic name of the drug. However, in some cases, {prod_ai} was missing and only {drugname} was provided. This study utilized the names listed in Supplementary Table 2 to convert all entries to their generic names. Each drug combination was represented by a nine-digit binary number, in which a binary 1 indicates administration and a binary 0 indicates nonadministration. The order was as follows (from left to right): (1) cyclosporine, (2) mycophenolate mofetil, (3) methotrexate, (4) tacrolimus, (5) cyclophosphamide, (6) prednisone, (7) lapine T-lymphocyte immune globulin, (8) sirolimus, and (9) alemtuzumab. For example, the binary number 100000000 indicates that patients were administered cyclosporine without any of the other eight drugs. To ensure robust statistical analyses and meaningful interpretation, this study focused on drug combinations with reports from at least 20 patients.
Statistical Analyses
Continuous variables were summarized as medians with their first and third quartiles, while categorical variables were summarized as frequency and reporting proportion (RP) [14], where RP = (number of patients with the characteristic of interest) / (total number of patients reported to FAERS for GVHD prophylaxis) × 100. Regarding the distinction between incidence and RP, while they share a numerator, their denominators differ markedly. The formula for incidence is: (number of patients with the characteristic of interest) / (total number of patients exposed to GVHD prophylaxis) × 100. However, because the FAERS data include only adverse event occurrences (with no reports of zero adverse events) and lack information on the total exposed population, incidence cannot be calculated; instead, RP is used as a proxy measure. To distinguish our approach from traditional statistical methods, we follow the convention used in previous studies by adding the prefix “reporting” to the names of our statistical analysis methods. The reporting odds ratio (ROR) [15] and its 95% confidence interval (CI) were calculated using univariate binomial logistic regression analysis. Controlling for differences in patient background, the adjusted ROR (aROR) and its 95% CI were computed using multivariate binomial logistic regression analysis. The variables adjusted for when calculating the aROR were selected based on their P-values in the univariate analysis. Statistical analyses were performed using R ver. 4.2.2 (The R Foundation) [16]. Sample R code to reproduce the results in Tables 1−3 and Fig. 1 is provided in Supplementary Material 3.
RESULTS
Patient Background
A total of 18,400 patients who received at least one GVHD prophylaxis drug were selected between 2004Q1 and 2024Q1. Ultimately, 18,118 patients were included in the analysis set after excluding 282 patients who met the exclusion criterion. The breakdown of each drug combination in the analysis set is shown in Fig. 1. A total of 61 combinations had reports from 20 or more patients. Patient background and drug combinations are summarized in Table 1. For {reporter_country}, the top 15 countries with the highest patient counts were used, while those ranked 16th or lower were collectively categorized as “others.”
The study population exhibited a diverse demographic profile. Male patients outnumbered female patients in the cohort. The age distribution was broad, with median ages (first quartile–third quartile) of 28 years (10−52 years) for patients without GVHD and 25 years (10−50 years) for those with the condition. This wide age range indicates a cohort spanning pediatric to adult populations. Geographically, although the largest numbers of reports came from the United States and Japan, the study included data from various Western and Asian countries, providing a global perspective on GVHD prophylaxis. This diverse composition enables a comprehensive analysis across different age groups and geographical regions, potentially influencing the study results and their applicability to various patient populations.
Adverse Events
Tables 2 and Supplementary Table 3 summarize the ROR and aROR for the occurrence of GVHD. However, because 61 drug combinations were targeted in this study, 23 combinations with aROR values of P<0.05 are presented in the main text, while the remaining 38 combinations, with aROR values of P≥0.05, are provided in Supplementary Table 3.
In the patient background analysis, the ROR was significant (P<0.05) for age (0.996 [95% CI, 0.995−0.998]; P<0.001) and weight (0.990 [95% CI, 0.985−0.994]; P<0.001), as well as for reports from the United States (0.756 [95% CI, 0.701−0.815]; P<0.001), Japan (1.284 [95% CI, 1.187−1.390]; P<0.001), Korea (1.423 [95% CI, 1.058−1.915]; P=0.020), Turkey (1.336 [95% CI, 1.043−1.712]; P=0.004), France (0.766 [95% CI, 0.662−0.885]; P<0.001), the Netherlands (0.411 [95% CI, 0.289−0.584]; P<0.001), and the United Kingdom (0.678 [95% CI, 0.585−0.785]; P<0.001). Accordingly, the adjustment variables used in the aROR analysis were age, the United States, Japan, Korea, Turkey, France, the Netherlands, and the United Kingdom. Although the ROR for weight also exhibited a P-value of less than 0.05, weight was not included as an adjustment variable in the aROR due to the high rate of missing data (Table 1).
Drug combinations with aROR values significantly less than 1.000 included 011000000 (0.152 [95% CI, 0.066−0.351]; P<0.001), 001100001 (0.200 [95% CI, 0.046−0.859]; P=0.030), 000000100 (0.218 [95% CI, 0.129−0.366]; P<0.001), 001100010 (0.220 [95% CI, 0.067−0.723]; P=0.013), 000010000 (0.242 [95% CI, 0.144−0.409]; P<0.001), 000000010 (0.451 [95% CI, 0.219−0.930]; P=0.031), 001100100 (0.522 [95% CI, 0.321−0.849]; P=0.009), 101001000 (0.668 [95% CI, 0.471−0.945]; P=0.023), 010000000 (0.705 [95% CI, 0.555−0.895]; P=0.004), and 010100000 (0.781 [95% CI, 0.663−0.920]; P=0.003). Conversely, the drug combinations with aROR values significantly greater than 1.000 were 001100000 (1.207 [95% CI, 1.070−1.361]; P=0.002), 110000000 (1.233 [95% CI, 1.089−1.395]; P=0.001), 100001000 (1.391 [95% CI, 1.097−1.764]; P=0.007), 110010000 (1.457 [95% CI, 1.165−1.823]; P=0.001), 100000100 (2.094 [95% CI, 1.135−3.862]; P=0.018), 110000001 (2.350 [95% CI, 1.047−5.277]; P=0.038), 111000100 (2.599 [95% CI, 1.728−3.910]; P<0.001), 110000100 (2.756 [95% CI, 1.842−4.124]; P<0.001), 100101000 (2.851 [95% CI, 1.124−7.230]; P=0.027), 101000001 (2.874 [95% CI, 1.308−6.319]; P=0.009), 101101000 (4.561 [95% CI, 2.016−10.316]; P<0.001), 010110010 (9.156 [95% CI, 4.729−17.729]; P<0.001), and 000100001 (9.391 [95% CI, 3.822−23.074]; P<0.001).
Supplementary Material 4 list the reported adverse events for each drug combination in descending order of frequency. Analysis of country-specific data revealed variations in the RPs of GVHD. For patients receiving cyclosporine without methotrexate (as shown in Supplementary Table 4), Italy, Korea, and Turkey exhibited comparatively high RPs of GVHD, with values of 47.2, 46.9, and 46.2, respectively. In contrast, for patients receiving methotrexate without cyclosporine (Supplementary Table 5), Russia and Japan reported higher RPs of GVHD occurrences, at 50.0 and 45.5, respectively. These findings highlight the complex interplay between drug combinations, patient characteristics, and geographical factors in the occurrence of GVHD, underscoring the need for tailored prophylactic strategies.
DISCUSSION
GVHD prophylaxis drugs demonstrate diverse mechanisms of action, targeting various components of the immune system implicated in GVHD pathogenesis. The complexity of these mechanisms has led previous studies to provide detailed visual explanations [17,18]. Calcineurin inhibitors such as cyclosporine and tacrolimus impede T cell activation by blocking the dephosphorylation and nuclear translocation of nuclear factor of activated T cells, which has been shown to reduce the incidence of acute GVHD by up to 50% in clinical trials. Antimetabolites like mycophenolate mofetil and methotrexate exert their effects by inhibiting key enzymes in nucleotide synthesis and folate metabolism, respectively, thereby attenuating T cell function. This mechanism has been associated with significant reductions in both acute and chronic GVHD when these drugs are used in combination with calcineurin inhibitors. When administered posttransplant, the alkylating agent cyclophosphamide selectively eliminates alloreactive T cells and may induce regulatory T cells, leading to marked decreases in severe acute GVHD and chronic GVHD requiring systemic therapy. Corticosteroids, including prednisone, prednisolone, and methylprednisolone, exhibit broad anti-inflammatory and immunosuppressive effects by targeting multiple cytokine pathways, making them particularly effective in treating established GVHD. Antibody-based therapies such as lapine T-lymphocyte immune globulin and alemtuzumab act by depleting T cells or modulating their function, significantly reducing chronic GVHD without compromising overall survival. Sirolimus, a mammalian target of rapamycin inhibitor, affects T cell proliferation and activation; this drug has shown promise in lowering the incidence of both acute and chronic GVHD when combined with tacrolimus. Given the diverse mechanisms of action and target sites of GVHD prophylaxis drugs, they are frequently administered in various combinations. For instance, the combination of cyclosporine, methotrexate, and prednisone has demonstrated superior efficacy in preventing acute GVHD compared to the dual therapy of cyclosporine and prednisone alone [19]. By analyzing FAERS data, this study characterized these extensive GVHD prophylaxis drug combinations used in clinical settings, providing insights into their relative effectiveness and potential synergistic effects.
The standard approach for GVHD prophylaxis typically involves a combinatorial regimen based on calcineurin inhibitors [20], usually administered in conjunction with antimetabolites. Despite the widespread implementation of these conventional preventive strategies, the incidence of GVHD remains substantial, affecting approximately 40% to 60% of patients who have undergone allogeneic hematopoietic stem cell transplantation. One potential reason for this high incidence is the suboptimal dosage of cyclosporine used. Previous studies have indicated that higher doses of cyclosporine may enhance the efficacy of GVHD prophylaxis [21,22]. However, higher doses of cyclosporine could increase the risk of adverse events, potentially limiting its use in certain patient populations. Alternatively, prior studies have investigated treatments that do not involve cyclosporine, with reports indicating that these alternative therapies demonstrate superior efficacy in GVHD prophylaxis compared to cyclosporine-based regimens [23,24]. Among the 13 drug combinations with aROR significantly exceeding 1.000 in this study, 10 included cyclosporine, five of which contained both cyclosporine and mycophenolate mofetil. These findings corroborate previous studies suggesting the limited efficacy of cyclosporine, as well as the combination of cyclosporine and mycophenolate mofetil, in GVHD prophylaxis. While avoiding cyclosporine may appear beneficial, this approach is not always feasible in clinical settings because treatment plans must be tailored to each patient. In this study, among the cyclosporine-containing combinations, the regimen of cyclosporine, methotrexate, and prednisone displayed an aROR significantly less than 1.000 and thus demonstrated high efficacy. This finding aligns with previous studies, which have reported that this combination exhibits superior GVHD prophylactic efficacy compared to other cyclosporine-containing regimens [19,25].
Another standard combination for GVHD prophylaxis is methotrexate and tacrolimus. In this study, the combination of these two drugs with either lapine T-lymphocyte immune globulin, sirolimus, or alemtuzumab demonstrated an aROR significantly less than 1.000. A previous study on the combination of tacrolimus, methotrexate, and sirolimus reported a reduction in the incidence of GVHD with the addition of sirolimus [26]. However, the group receiving sirolimus tended to display an elevated incidence of veno-occlusive disease and thrombotic microangiopathy. In contrast, no reports are yet available on the safety and efficacy of the three drug combinations involving methotrexate and tacrolimus with either lapine T-lymphocyte immune globulin or alemtuzumab. The effectiveness of these combinations should be validated in future clinical trials.
Based on the present study, the RP of GVHD occurrence varied significantly across countries, as shown in Table 3. The United States and European nations (France, the Netherlands, and the United Kingdom) demonstrated ROR values significantly less than 1.000, whereas Asian countries (Japan, Korea, and Turkey) exhibited RORs significantly greater than 1.000. These variations may be attributed to several factors. Differences in genetic backgrounds between populations could influence GVHD susceptibility [27]. Medical practices, including donor selection criteria, conditioning regimens, and GVHD prophylaxis protocols, may vary between regions due to differences in clinical guidelines or resource availability [28]. Regulatory environments also play a role, as variations in approved drugs and protocols can impact GVHD management strategies [29]. Moreover, cultural factors affecting patient-physician communication and reporting practices could influence the observed differences [30]. For example, variations in the willingness to report adverse events or differences in the interpretation of GVHD symptoms across cultures might contribute to these disparities. These findings underscore the importance of considering regional factors when developing and implementing GVHD prophylaxis strategies, highlighting the need for further research into the underlying causes of these geographical variations.
This study had several limitations. First, because the FAERS data did not include reports detailing the absence of adverse events, the incidence of GVHD could not be calculated. Second, the spontaneous nature of FAERS data reporting may have introduced bias. Third, the available information was limited to the reason for drug administration (namely, GVHD prophylaxis) without detailed transplantation circumstances. However, given that GVHD is highly likely to follow allogeneic hematopoietic stem cell transplantation, it is reasonable to assume that most of these data pertained to such patients. Fourth, the occurrence or nonoccurrence of GVHD may depend on adherence to the prescribed prophylaxis regimen, but the FAERS data provided no information on medication adherence. Lastly, {pt} was recorded as acute and chronic GVHD in some cases and simply as GVHD in others. Due to the small sample sizes that would result from considering specific variations, this study used a broad definition for the occurrence of GVHD.
Despite the inherent limitations of FAERS data, its global coverage represents a major advantage, offering a wealth of reports from diverse geographical locations. This study's methodology allowed for the investigation of drug combinations that are typically challenging to incorporate into clinical trials. The identification of several GVHD prophylaxis combinations in this study has potential clinical applications, even in countries where certain regimens may not be approved. Although subject to the limitations of the FAERS dataset, our findings provide a valuable starting point for future studies by identifying promising drug combinations for GVHD prophylaxis that warrant further investigation. This approach can potentially streamline future research efforts. To leverage our results, healthcare providers could adopt a stepwise strategy—initially incorporating one additional agent into their standard prophylaxis regimen based on our findings, while closely monitoring outcomes. This cautious implementation enables real-world evaluation of the identified drug combinations, potentially leading to improved GVHD prophylaxis strategies. Alternatively, researchers could conduct randomized controlled trials comparing the most promising drug combinations identified in our study against the current standard of care. These practical applications of our findings could contribute to advancing GVHD prophylaxis strategies.
ARTICLE INFORMATION
Conflict of Interest
No potential conflict of interest relevant to this article was reported.
Author Contributions
Conceptualization: all authors. Data curation: TO. Formal analysis: TO. Investigation: TO. Methodology: all authors. Project administration: TO. Visualization: TO. Writing–original draft: TO. Writing–review & editing: CS, AU. All authors read and approved the final manuscript.
XML Download