Genetic polymorphisms associated with therapeutic response and adverse effects to methotrexate in Taiwanese patients with rheumatoid arthritis: a real-world, hospital-based study Authors: Ta-Wei Yu, I-Chieh Chen, Guan-Cheng Lin, Chung-Mao Kao, Yen-Ju Chen, Yi-Ming Chen Published: 2026 ABSTRACT Although methotrexate (MTX) is extensively utilized in management of rheumatoid arthritis (RA), its pharmacogenomics are still unclear. This study explores the genetic determinants impacting MTX’s effectiveness and adverse effects in Taiwanese patients with RA, aiming to enhance understanding of RA treatment and pharmacogenetics. A retrospective analysis of 1,948 RA patients receiving MTX treatment from January 2009 to January 2022, using data from the Taiwan Precision Medicine Initiative (TPMI) at Taichung Veterans General Hospital. Among these patients, 872 who required treatment with biologic disease modifying anti-rheumatic drugs (bDMARDs) or targeted synthetic DMARDs (tsDMARDs) were classified as MTX ‘non-response’, while the 1,076 patients treated without these agents were categorized as MTX ‘response’. Chi-square analysis was employed to compare the carrier ratio of 82 selected MTX-associated SNPs from previous literature. Logistic regression models were utilized to explore associations between genetic variants and the risk of non-response and adverse effects, with adjustments made for potential confounding factors. Significant genetic markers influencing MTX response included ABCC1 rs212090, which predicted better MTX efficacy (p = 0.016), and ABCC3 rs2277624 and rs4148416, linked to lower MTX response rates (p = 0.004 and p = 0.023). SNPs in ADORA3 rs2298191 and ABCC4 rs7317112 were linked to a higher risk of thrombocytopenia (p = 0.045 and p = 0.049). Our study identifies genetic variations ABCC1 and ABCC3 related to MTX efficacy in treating RA, while SNPs in ADORA3 and ABCC4 are associated with the side effect of thrombocytopenia. These insights offer a foundation for tailoring individual treatment plans based on genetic profiles.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-54386-3. Introduction Rheumatoid arthritis (RA) is considered a chronic systemic autoimmune disease, predominantly marked by inflammation of synovial joint lining, resulting in bone and cartilage degradation, progression of deformities, and culminating in disability1,2. The global prevalence of RA exhibits relative consistency, with estimates ranging between 0.5% and 1.1%3. Prevalence rates of RA within the United States and Northern European nations are generally equivalent, often reported within the range of 0.5 to 1%4. However, the estimated prevalence of RA on the African continent varies widely, with figures in individual countries ranging from 0.06% to 3.4%5. A higher prevalence rate of RA among the Pima Indians has been documented at 5.3%. In Taiwan, the period prevalence of RA has been recorded at 0.96%6. These findings lead support to the hypothesis that hereditary elements may contribute to the susceptibility to the disease7. Therapeutic management in RA focuses on reducing synovitis and articular damage. The strategy for treatment is designed to secure remission or maintain low levels of disease activity. This objective is pursued through the prompt administration of disease-modifying antirheumatic drugs (DMARDs) at the onset of the disease. Nonetheless, patients exhibit varying levels of treatment response or disease progression8. Among the array of DMARDs, methotrexate (MTX) is generally regarded as the most commonly prescribed agent for the management of RA9,10. Furthermore, MTX is recognized for its cost-effectiveness and favorable tolerance profile in most of RA patients11. Notwithstanding, the therapeutic efficacy of MTX, as evaluated by the American College of Rheumatology (ACR) criteria, exhibits variability (46 to 65%), with approximately one-third of individuals exhibiting non-responsiveness to MTX, attributed to suboptimal therapeutic effectiveness or side effects12. Besides, the most commonly observed adverse drug reactions (ADRs) linked with MTX primarily include bone marrow suppression, leading to hematological disorders such as anemia, leukopenia, thrombocytopenia and even hepatotoxicity13. Currently, there remains no reliable clinical biomarkers that can effectively predict the efficacy of MTX or its potential toxic effects14,15. Hence, the need to develop innovative screening methodologies that can reliably forecast the effectiveness and ADRs prior to their prescribed in RA has intensified. In this endeavor, the field of pharmacogenetics (PGx) is recognized as fundamentally instrumental. Pharmacogenomics explores how genetic factors influence individual responses to medications, focusing on the effects of both hereditary and acquired genetic variations on drug efficacy and safety. This field aims to link gene expression patterns and single-nucleotide polymorphisms (SNPs) with pharmacokinetic properties to understand their impact on drug action12. Recently, multiple PGx investigations have been initiated to unearth the relationships between the particular SNPs that code for enzymes within cellular pathways modulated by MTX and the corresponding outcomes of the drug’s effectiveness and potential side effects9,16. Recent research has indicated that variations in transporter genes, specifically SLC19A1 and ABCC2, may serve as indicators of how well Chinese RA patients respond to MTX treatment, as measured by the European League Against Rheumatism (EULAR) criteria for good and moderate responses14. Moreover, a study in a Norwegian population with early RA who had not previously received DMARDs found that SNPs in the MTRR, AIF-1, and DHFR genes were significantly linked to continuous outcomes related to disease activity, such as disease activity, fatigue, and joint pain, after treatment with MTX alone17. Regarding the side effects caused by MTX, an analysis of Swedish patients with RA suggests RAVER2 rs72675408 were notably correlated with the highest level of Alanine aminotransferase (ALT) following MTX use18. Additionally, a cohort study in Mexican patients with RA identified polymorphisms in GGH rs1800909 and MTHFR rs1801131 as being associated with a range of adverse events, including nausea, vomiting, indigestion, hair loss, mouth sores, reduced white blood cell count, liver abnormalities, and lung toxicity19. Moreover, MTHFR rs1801133, rs1801131 and ADA rs244076 were linked to gastrointestinal side effects19. The aforementioned studies, as well as existing research, have predominantly focused on a limited number of SNPs and were primarily conducted in the Caucasian population. Moreover, the sample sizes in these studies were relatively small. Therefore, this research seeks to thoroughly explore how genetic variations relate to both the treatment effects and adverse effects in Taiwanese RA patients receiving MTX, while also incorporating a larger patient sample size for analysis. Methods Study population A sum of 63,000 patients undergoing treatment at Taichung Veterans General Hospital (TCVGH), were extended an invitation to take part in the Taiwan Precision Medicine Initiative (TPMI). This countrywide genetic research project is spearheaded by Academia Sinica in collaboration with partner hospitals. For each participant enrolled in the TPMI, blood samples were obtained from which DNA was isolated and subsequently analyzed for genotype. These genetic data were then integrated with each individual’s electronic health records at TCVGH, encompassing medical histories and biochemical test results. According to electronic medical records data from June 2019 to November 2022, individuals participating in TPMI at TCVGH who had been hospitalized at least once or attended outpatient clinics more than twice, and were diagnosed by physicians as meeting the RA criteria for ICD9 (714.0, 714.30–714.33) and ICD10 (M05.7-M05.9, M06.0, M06.2-M06.3, M06.8-M06.9, M08) codes, were included in this study. Excluding patients under the age of eighteen and those with insufficient medical records, the total remaining cohort comprised 3,090 individuals (Supplementary Fig. 1). The research methodology was approved by the Institutional Review Board at TCVGH, with protocol number SF19153A. All procedures adhered strictly to the ethical standards of the Declaration of Helsinki, including obtaining informed consent from every participant. Clinical assessment Under Taiwan’s strict National Health Insurance (NHI) regulations, approval for biologic or targeted synthetic DMARDs (bDMARDs/tsDMARDs) requires objective clinical evidence of conventional synthetic DMARDs (csDMARDs) treatment failure. Patients must meet ACR criteria for RA, fail a ≥ 6-month trial of at least two csDMARDs (including MTX), and exhibit persistently high disease activity (DAS28 > 5.1 on two consecutive evaluations ≥ 1 month apart). Consequently, NHI approval for these advanced therapies serves as a robust, real-world surrogate marker for MTX non-response. In this study, the “non-response” group was defined as patients requiring bDMARDs/tsDMARDs (Supplementary Table 1). Conversely, the “response” group comprised patients maintaining a DAS28 < 5.1 while receiving MTX alone or combined with other csDMARDs (sulfasalazine, hydroxychloroquine, cyclosporine, or leflunomide). Pertinent demographic and clinical parameters such as sex, age, body mass index (BMI), and smoking record, were systematically documented. Additionally, variables indicative of disease activity, such as Rheumatoid Factor (RF), anti-cyclic citrullinated peptide antibody (anti-CCP), erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), were analyzed using initial value. Regarding the laboratory testing methods, RF is quantified using immunoturbidimetric methods, while anti-CCP detection utilizes Fluorescence Enzyme Immunoassay (FEIA). According to the standards we employ, RF levels exceeding 14 IU/mL are considered positive. Similarly, for anti-CCP, any result greater than 10.0 U/mL is designated as positive20. Furthermore, the median and interquartile range (IQR) values of DAS-28 were included in the analysis. Among the various side effects caused by MTX, this research focuses on hematological and hepatic side effects after its usage. Hematological adverse reactions are defined by specific clinical parameters: anemia is indicated by hemoglobin levels falling below 10.0 g/dL21; leukopenia by a leukocyte count under 4000/µL19; and thrombocytopenia by a platelet count under 150,000/mL19.Finally, hepatic involvement is characterized as an increase in Alanine aminotransferase (ALT) levels to two times the normal upper limit19. SNP selection and genotyping In this research, genomic DNA was isolated from 6,3000 participants of the TPMI utilizing DNA isolation kits sourced from TIANGEN Biotech, Beijing, China. Subsequently, the concentration of DNA samples was determined utilizing a NanoDrop 2000 Spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). Participants were genotyped using the Axiom Genome-Wide TWB 2.0 Array Plate (Affymetrix, Santa Clara, CA, USA), encompassing 714,431 SNPs and specifically tailored for Taiwan’s Han Chinese population22.This approach encompassed the analysis of 114,000 risk variants within 2,831 disease-associated genes known to be rare, as collated from CLIN Var, ACMG, GWAS Catalog, HGMD, locus-specific databases, and the scientific publications23. Genotyping quality control procedures were implemented, which involved the exclusion of SNPs displaying a sole allele occurrence within the study cohort, as well as those with a total call rate below 95% or a total minor allele frequency (MAF) below 0.01. Furthermore, SNPs exhibiting notable deviations from the Hardy-Weinberg equilibrium (P < 1 × 10− 5) were also removed. The PLINK 1.9 software package was used to perform statistical genetic analyses and to ensure the quality control of genetic data. Following implementing quality control measures, 591,048 SNPs were kept for analysis. We systematically searched PubMed, Embase, and the Cochrane Library from 2020 to 2024, focusing on comparative studies investigating SNPs associated with MTX responses or SNPs that could lead to MTX-related side effects. The search was conducted without language or geographical restrictions. The following keywords were utilized: “Single Nucleotide Polymorphisms,” " SNPs,” “Pharmacogenetics,” “Pharmacogenomics,” “Methotrexate,” “MTX,” “Therapeutic Response,” “Efficacy,” “Drug Response,” “Resistance,” “Toxicity,” “Adverse Effects,” and “Side Effects.” After excluding duplicate SNPs from the analysis and incorporating SNPs available within our institution’s dataset, we identified a total of 82 relevant SNPs, related to MTX response or its side effect. Information about the chosen SNPs is provided in Supplementary Tables 2 and 3. Statistical analysis The Statistical Package for the Social Sciences (SPSS version 22.0, Chicago, IL, USA) software for Windows was used to analyze the data. Demographic information was statistically summarized through frequencies and percentages. Parametric data were reported as means and standard deviations. The chi-square test was utilized to determine the statistical significance of categorical variables. Univariate logistic regression analyses were performed to assess the association between ABCC3 rs2277624, ABCC3 rs4148416, and ABCC1 rs212090 variants and the risks of treatment response and adverse effects. These models were adjusted for potential confounders, including age, sex, treatment era (before 2014 or not when Treat-to Target approach was fully implemented), rheumatoid factor (RF), and anti-CCP status, with odds ratios (OR) and 95% confidence intervals (95% CI) calculated. Thrombocytopenia-free survival probability was evaluated using the Kaplan-Meier method, and survival curves were compared using the log-rank test. To control for Type I error from multiple testing across 82 SNPs, Bonferroni correction was applied (Supplementary Table 4). All tests were 2 tailed. Besides, p value less than 0.05 was regarded as statistically significant. Result Baseline characteristics of study population This retrospective cohort study analyzed 1,948 Taiwanese RA patients treated with MTX, comparing those with (872 patients, non-response) and without (1,076 patients, response) concomitant of bDMARDs or tsDMARDs (Table 1). Distribution of sex showed resemblance between two groups, with a majority being female (78.30% vs. 75.80%; p = 0.214), and there was no notable disparity in age (p = 0.834).The non-response group presented a significantly younger RA onset age (mean 47 vs. 50 years; p < 0.001) but showed no statistical significance in BMI categories or smoking status (p = 0.065 and p = 0.990, respectively). Immunological markers differed significantly; the non-response group had higher percentages of positive RF and anti-CCP (69.97% and 77.39% vs. 55.93% and 58.87%, respectively; p < 0.001), elevated ESR and CRP levels, and the higher DAS28 (median/IQR) scores, all with p-values of < 0.001, indicating more severe disease activity. Medication patterns also show similar trends; In the non-response group, higher usage of hydroxychloroquine (p = 0.012), sulfasalazine, cyclosporin, and leflunomide was noted, with the latter three demonstrating statistically meaningful variances (p < 0.001). Furthermore, there were notable disparities in glucocorticoid use (98.28% non-response vs. 92.01% response; p < 0.001). Table 1Basic characteristics of study populations.RA patients (n = 1948)P valueNon response group (n = 872)Response group (n = 1076)N%N% Sex (n, %) Female68378.3081675.80 Male18921.7026024.200.214 Age (Mean, SD) 58.881458.77140.834 RA onset age (Mean, SD) 47.331449.5514< 0.001 BMI (n, %) < 2442153.7755558.3>=2436246.2339741.70.065 Smoking No73884.6391284.76Yes13415.3716415.240.990 Laboratory test RF Ab positivity > 14 (n, %)60869.9759455.93< 0.001Anti-CCP Ab positivity > 10 (n,%)59277.3956158.87< 0.001ESR (mm/hr)43.8631.3431.3527.34< 0.001CRP (mg/L)2.174.761.483.47< 0.001DAS28 (median/IQR)2.033.731.793.12< 0.001 Medication Hydroxychloroquine82494.598491.45 0.012Sulfasalazine44150.5745642.38< 0.001Cyclosporin35841.0621720.17< 0.001Leflunomide35040.1425823.98< 0.001 Glucocorticoids Yes85798.2899092.01< 0.001 Table 1Basic characteristics of study populations.RA patients (n = 1948)P valueNon response group (n = 872)Response group (n = 1076)N%N% Sex (n, %) Female68378.3081675.80 Male18921.7026024.200.214 Age (Mean, SD) 58.881458.77140.834 RA onset age (Mean, SD) 47.331449.5514< 0.001 BMI (n, %) < 2442153.7755558.3>=2436246.2339741.70.065 Smoking No73884.6391284.76Yes13415.3716415.240.990 Laboratory test RF Ab positivity > 14 (n, %)60869.9759455.93< 0.001Anti-CCP Ab positivity > 10 (n,%)59277.3956158.87< 0.001ESR (mm/hr)43.8631.3431.3527.34< 0.001CRP (mg/L)2.174.761.483.47< 0.001DAS28 (median/IQR)2.033.731.793.12< 0.001 Medication Hydroxychloroquine82494.598491.45 0.012Sulfasalazine44150.5745642.38< 0.001Cyclosporin35841.0621720.17< 0.001Leflunomide35040.1425823.98< 0.001 Glucocorticoids Yes85798.2899092.01< 0.001 Basic characteristics of study populations. Analysis of the influence of genetic variants on distribution differences and curative effect risk, adjusted for demographic and clinical variables In our analysis of candidate SNPs (Supplementary Table 4), ABCC3 rs2277624, ABCC3 rs4148416, and ABCC1 rs212090 demonstrated nominally significant differences in genotype frequencies between effective MTX responders and non-responders (nominal p = 0.0043, 0.0232, and 0.0160, respectively). Specifically, ABCC3 variants were associated with an increased risk of treatment failure, whereas ABCC1 rs212090 was linked to a favorable therapeutic response (Supplementary Table 4). To address potential clinical confounders, we analyzed genotype distributions between the non-response and response groups using multivariable logistic regression, adjusting for age, sex, treatment era (pre/post-2014), RF, and anti-CCP (Table 2). The ABCC3 rs2277624 CT genotype significantly increased the risk of treatment failure compared to the CC genotype (OR: 1.27, 95% CI: 1.03–1.56, p = 0.027), whereas its TT genotype showed no effect (OR: 0.89, 95% CI: 0.59–1.32, p = 0.559). Conversely, the ABCC1 rs212090 TA genotype conferred a significant protective effect (OR: 0.77, 95% CI: 0.61–0.98, p = 0.035), with the AA genotype showing no association (OR: 1.01, 95% CI: 0.50–2.07, p = 0.971). Furthermore, the ABCC3 rs4148416 CT genotype presented a marginally significant risk trend (OR: 1.25, 95% CI: 0.99–1.57, p = 0.052), while the TT genotype remained non-significant (OR: 0.65, 95% CI: 0.31–1.35, p = 0.248). Table 2Genetic variations associated with methotrexate therapeutic efficacy in patients with RA.Non-response group (n = 872)Response group (n = 1076)p value#OR (95% CI)p value## n % n %ABCC3 rs2277624CC43650.1160956.61-CT37743.3338435.691.27 (1.03–1.56)0.027TT576.55837.710.0030.89 (0.59–1.32)0.559ABCC3 rs4148416CC60669.6679774.071-CT25128.8525223.421.25 (0.99–1.57)0.052TT151.72262.420.0190.65 (0.31–1.35)0.248ABCC1 rs212090TT66776.6777071.561-TA17820.4627425.460.77 (0.61–0.98)0.035AA202.3211.950.0281.01 (0.50–2.07)0.971* Logistic regression analyses were performed, adjusting for age, sex, year of treatment (pre-2014 vs. post-2014), RF and anti-CCP.# test whether there is a significant difference in the distribution of genotypes between the non-response group and the response group.## assess the significance of the odds ratio, determining whether certain genotypes have a significantly different impact on drug response.SNP, single nucleotide polymorphism; OR, odds ratio;95% CI,95% confidence interval. Table 2Genetic variations associated with methotrexate therapeutic efficacy in patients with RA.Non-response group (n = 872)Response group (n = 1076)p value#OR (95% CI)p value## n % n %ABCC3 rs2277624CC43650.1160956.61-CT37743.3338435.691.27 (1.03–1.56)0.027TT576.55837.710.0030.89 (0.59–1.32)0.559ABCC3 rs4148416CC60669.6679774.071-CT25128.8525223.421.25 (0.99–1.57)0.052TT151.72262.420.0190.65 (0.31–1.35)0.248ABCC1 rs212090TT66776.6777071.561-TA17820.4627425.460.77 (0.61–0.98)0.035AA202.3211.950.0281.01 (0.50–2.07)0.971* Logistic regression analyses were performed, adjusting for age, sex, year of treatment (pre-2014 vs. post-2014), RF and anti-CCP.# test whether there is a significant difference in the distribution of genotypes between the non-response group and the response group.## assess the significance of the odds ratio, determining whether certain genotypes have a significantly different impact on drug response.SNP, single nucleotide polymorphism; OR, odds ratio;95% CI,95% confidence interval. Genetic variations associated with methotrexate therapeutic efficacy in patients with RA. * Logistic regression analyses were performed, adjusting for age, sex, year of treatment (pre-2014 vs. post-2014), RF and anti-CCP. # test whether there is a significant difference in the distribution of genotypes between the non-response group and the response group. ## assess the significance of the odds ratio, determining whether certain genotypes have a significantly different impact on drug response. SNP, single nucleotide polymorphism; OR, odds ratio;95% CI,95% confidence interval. To further validate these findings, a sensitivity analysis was conducted by excluding patients who switched to biologics within 6 months, yielding a final cohort of 821 non-responders and 1,076 responders (Supplementary Table 7). The results remained largely consistent with the primary analysis. For ABCC3 rs2277624, the CT genotype continued to demonstrate a significant association with non-response (OR: 1.24, 95% CI: 1.003–1.54, p = 0.047), while the TT genotype remained non-significant (OR: 0.89, 95% CI: 0.59–1.34, p = 0.589). A significant difference in genotype distribution between groups was observed for this variant (p = 0.005), supporting its overall relevance in MTX response. For ABCC3 rs4148416, although the CT genotype did not reach statistical significance (OR: 1.22, 95% CI: 0.96–1.53, p = 0.099), a nominally significant difference in genotype distribution was detected (p = 0.030), with the TT genotype remaining non-significant (OR: 0.62, 95% CI: 0.29–1.31, p = 0.209). Regarding ABCC1 rs212090, the TA genotype retained its protective effect (OR: 0.78, 95% CI: 0.61–0.99, p = 0.048), and a significant genotype distribution difference was again observed (p = 0.031), whereas the AA genotype showed no significant association (OR: 1.08, 95% CI: 0.53–2.22, p = 0.830). Collectively, these sensitivity analyses corroborate the robustness of the primary findings, suggesting that the identified genetic associations are not substantially confounded by early biologic switching. We subsequently performed a stratified analysis using a 2014 cutoff to evaluate the impact of the Treat-to-Target (T2T) strategy (Supplementary Tables 5 and 6). In the pre-2014 cohort (n = 1,140; 610 in the non-response group vs. 530 in the response group), the ABCC3 rs2277624 CT genotype demonstrated a significant risk effect (OR: 1.32, 95% CI: 1.004–1.74, p = 0.047), and ABCC3 rs4148416 showed a similar trend (OR: 1.32, 95% CI: 0.98–1.78, p = 0.068). In the post-2014 cohort (n = 808), improved therapeutic outcomes resulted in fewer patients in the non-response group (546 in the response group vs. 262 in the non-response group). Although this reduced sample size decreased statistical power and yielded non-significant p-values, the OR directionality remained highly consistent with the primary model. Specifically, post-2014 ABCC3 variants maintained a risk trend (rs2277624 OR = 1.20; rs4148416 OR = 1.15), while ABCC1 rs212090 sustained a protective trend (OR: 0.72, 95% CI: 0.49–1.06, p = 0.094). These results suggest that the underlying genetic influences on MTX efficacy remain stable despite evolving treatment strategies. In a secondary sensitivity analysis stratified by the 2014 cutoff—which excluded patients switching to bDMARDs/tsDMARDs within 6 months of MTX initiation (Supplementary Tables 8 and 9), the reduced sample size decreased statistical power, shifting some p-values to marginal significance (e.g., pre-2014 ABCC3 rs2277624 CT: OR = 1.31, 95% CI: 0.99–1.72, p = 0.059; post-2014 ABCC1 rs212090 TA: OR = 0.69, 95% CI: 0.46–1.03, p = 0.072). However, the consistent directionality and magnitude of the Odds Ratios across both strata suggest a stable underlying trend for these genetic effects. Association between SNPs and the occurrence of adverse effects in individuals with RA receiving MTX treatment In a study assessing the link between SNP variants and the occurrence of side effects in RA patients undergoing MTX therapy, most SNPs showed no statistically significant correlation with hematological and hepatic side effects, as indicated by p-values above 0.05 (Supplementary Table 10). However, two SNPs demonstrated a notable correlation with thrombocytopenia: ADORA3 rs2298191 (p = 0.045) and ABCC4 rs7317112 (p = 0.049), suggesting increased susceptibility to this side effect among carriers undergoing MTX treatment (Table 3). Table 3Comparison of ADORA3 rs2298191 and ABCC4 rs7317112 allele frequencies with methotrexate adverse events in RA patients.TotalADORA3 rs2298191P-valueABCC4 rs7317112P valuecarriernon-carriercarriernon-carriern (%)n (%)n (%)n (%) Hematological adverse events Anemia396235(21.38)157(18.62)0.149226(20.55)169(20)0.811Leukopenia483271(24.66)211(25.03)0.893275(25)207(24.5)0.840Thrombocytopenia293182(16.56)111(13.17)0.045181(16.45)111(13.14)0.049 Hepatitis Male : ALT > 100 UI/L6536(14.34)29(14.72)1.00040(15.56)25(13.09)0.548Female : ALT > 70 UI/L336197(23.23)139(21.52)0.469180(21.35)155(23.7)0.308ALT, Alanine aminotransferase; n (%), number and percentage. Table 3Comparison of ADORA3 rs2298191 and ABCC4 rs7317112 allele frequencies with methotrexate adverse events in RA patients.TotalADORA3 rs2298191P-valueABCC4 rs7317112P valuecarriernon-carriercarriernon-carriern (%)n (%)n (%)n (%) Hematological adverse events Anemia396235(21.38)157(18.62)0.149226(20.55)169(20)0.811Leukopenia483271(24.66)211(25.03)0.893275(25)207(24.5)0.840Thrombocytopenia293182(16.56)111(13.17)0.045181(16.45)111(13.14)0.049 Hepatitis Male : ALT > 100 UI/L6536(14.34)29(14.72)1.00040(15.56)25(13.09)0.548Female : ALT > 70 UI/L336197(23.23)139(21.52)0.469180(21.35)155(23.7)0.308ALT, Alanine aminotransferase; n (%), number and percentage. Comparison of ADORA3 rs2298191 and ABCC4 rs7317112 allele frequencies with methotrexate adverse events in RA patients. ALT, Alanine aminotransferase; n (%), number and percentage. Distribution of genotypes and allele frequencies, and their correlation with the risk of side effects Upon further analysis of SNPs ADORA3 rs2298191 and ABCC4 rs7317112, and their genotypic effects on hematological side effects in patients, distinct associations were observed. For ADORA3 rs2298191, while no significant link was found with anemia (p = 0.323) or leukocytosis (p = 0.983), the TC genotype presented a significant association with thrombocytopenia compared to the TT genotype, with an Odds Ratio (OR) of 1.38 (95% CI 1.06–1.80, p = 0.017). In contrast, ABCC4 rs7317112 showed no significant association with anemia (p = 0.942), leukocytosis (p = 0.087), or thrombocytopenia (p = 0.056) across genotypes (Table 4). Table 4Genetic variations associated with adverse events of methotrexate in patients with RA.Hgb < 10.0 g/dLHgb≥10.0 g/dLp-value*Risk of side effect(n = 396) (%)(n = 1552) (%)OR (95% CI)P value**ADORA3 rs2298191TT157 (40.05)686 (44.26)1TC186 (47.45)686 (44.26)1.19(0.93–1.50)0.1616CC49 (12.5)178 (11.48)0.3231.20(0.84–1.73)0.3155ABCC4 rs7317112GG169 (42.78)676 (43.61)1GA178 (45.06)693 (44.71)1.03(0.81–1.30)0.8220AA48 (12.15)181 (11.68)0.9421.06(0.74–1.52)0.7481WBC < 4000/µLWBC≥4000/µLp-value*Risk of side effect(n = 483) (%)(n = 1465) (%)OR (95% CI)P value**ADORA3 rs2298191TT211 (43.78)632 (43.29)1TC215 (44.61)657 (45.0)0.98(0.79–1.22)0.8579CC56 (11.62)171 (11.71)0.9830.98(0.70–1.38)0.9114ABCC4 rs7317112GG207 (42.95)638 (43.61)1GA205 (42.53)666 (45.52)0.95(0.76–1.18)0.6413AA70 (14.52)159 (10.87)0.0871.36(0.98–1.87)0.0631PLT < 150,000 mcLPLT≥150,000 mcLp-value*Risk of side effect(n = 293) (%)(n = 1655) (%)OR (95% CI)P value**ADORA3 rs2298191TT111 (37.88)732 (44.39)1TC151 (51.54)721 (43.72)1.38(1.06–1.80)0.017CC31 (10.58)196 (11.89)0.0461.04(0.68–1.60)0.847ABCC4 rs7317112GG111 (38.01)734 (44.4)1GA143 (48.97)728 (44.04)1.30(0.99–1.70)0.056AA38 (13.01)191 (11.55)0.1271.32(0.88–1.97)0.180* Chi-square test was employed to evaluate the statistical significance of categorical variables.** Logistic regression analyses were performed, adjusting for both sex and age.Abbreviation: Hgb, hemoglobin; PLT, platelet; n, number; OR, odds ratio; 95% CI, 95% confidence interval. Table 4Genetic variations associated with adverse events of methotrexate in patients with RA.Hgb < 10.0 g/dLHgb≥10.0 g/dLp-value*Risk of side effect(n = 396) (%)(n = 1552) (%)OR (95% CI)P value**ADORA3 rs2298191TT157 (40.05)686 (44.26)1TC186 (47.45)686 (44.26)1.19(0.93–1.50)0.1616CC49 (12.5)178 (11.48)0.3231.20(0.84–1.73)0.3155ABCC4 rs7317112GG169 (42.78)676 (43.61)1GA178 (45.06)693 (44.71)1.03(0.81–1.30)0.8220AA48 (12.15)181 (11.68)0.9421.06(0.74–1.52)0.7481WBC < 4000/µLWBC≥4000/µLp-value*Risk of side effect(n = 483) (%)(n = 1465) (%)OR (95% CI)P value**ADORA3 rs2298191TT211 (43.78)632 (43.29)1TC215 (44.61)657 (45.0)0.98(0.79–1.22)0.8579CC56 (11.62)171 (11.71)0.9830.98(0.70–1.38)0.9114ABCC4 rs7317112GG207 (42.95)638 (43.61)1GA205 (42.53)666 (45.52)0.95(0.76–1.18)0.6413AA70 (14.52)159 (10.87)0.0871.36(0.98–1.87)0.0631PLT < 150,000 mcLPLT≥150,000 mcLp-value*Risk of side effect(n = 293) (%)(n = 1655) (%)OR (95% CI)P value**ADORA3 rs2298191TT111 (37.88)732 (44.39)1TC151 (51.54)721 (43.72)1.38(1.06–1.80)0.017CC31 (10.58)196 (11.89)0.0461.04(0.68–1.60)0.847ABCC4 rs7317112GG111 (38.01)734 (44.4)1GA143 (48.97)728 (44.04)1.30(0.99–1.70)0.056AA38 (13.01)191 (11.55)0.1271.32(0.88–1.97)0.180* Chi-square test was employed to evaluate the statistical significance of categorical variables.** Logistic regression analyses were performed, adjusting for both sex and age.Abbreviation: Hgb, hemoglobin; PLT, platelet; n, number; OR, odds ratio; 95% CI, 95% confidence interval. Genetic variations associated with adverse events of methotrexate in patients with RA. * Chi-square test was employed to evaluate the statistical significance of categorical variables. ** Logistic regression analyses were performed, adjusting for both sex and age. Abbreviation: Hgb, hemoglobin; PLT, platelet; n, number; OR, odds ratio; 95% CI, 95% confidence interval. Impact of ADORA3 and ABCC4 genetic variants on thrombocytopenia risk in RA patients undergoing MTX treatment: a Kaplan-Meier analysis The Kaplan-Meier (KM) curves illustrate the time-to-event analysis for thrombocytopenia in RA patients treated with MTX. The analysis was focused on the genetic variants ADORA3 rs2298191 and ABCC4 rs7317112, which were identified to potentially cause thrombocytopenia (Fig. 1). The vertical axis of the KM curves represents the Probability of thrombocytopenia-free survival, while the horizontal axis tracks the passage of time in years. These curves depicted thrombocytopenia-free survival over time, revealing that carriers of the ADORA3 rs2298191 variant had a significantly lower probability of avoiding thrombocytopenia compared to non-carriers (p = 0.035), a trend similarly observed in carriers of the ABCC4 rs7317112 variant (p = 0.042).Continuing from the previous discussion, our subsequent analysis explored the KM curves for the ADORA3 rs2298191 and ABCC4 rs7317112 polymorphisms across three genetic subgroups: wild type, carriers of one variant, and carriers of both variants. The analysis revealed a statistically meaningful difference in the KM curves across three genetic subgroups (p = 0.026). The curve for wild-type individuals shows a higher probability of remaining free from thrombocytopenia over time, followed by those carrying one variant, with the lowest probability observed in individuals carrying both genetic variants. Fig. 1Kaplan-Meier curves of 5-year thrombocytopenia-free survival probability in MTX-treated RA patients: carriers (red) vs. non-carriers (blue). (A) Carrier of ADORA3 rs2298191 (B) Carrier of ABCC4 rs7317112 (C) Comparisons of RA patients carrying both variants (red), either one variant (blue), and neither (green) variants of ADORA3 rs2298191 and ABCC4 rs7317112. Kaplan-Meier curves of 5-year thrombocytopenia-free survival probability in MTX-treated RA patients: carriers (red) vs. non-carriers (blue). (A) Carrier of ADORA3 rs2298191 (B) Carrier of ABCC4 rs7317112 (C) Comparisons of RA patients carrying both variants (red), either one variant (blue), and neither (green) variants of ADORA3 rs2298191 and ABCC4 rs7317112. Discussion This comprehensive retrospective cohort study of 1,948 Taiwanese RA patients receiving MTX unveiled significant findings concerning the impact of specific genetic variants on therapeutic outcomes and side effects of MTX. Our findings indicate that the ABCC1 rs212090 variant was correlated with a better response to MTX, while the ABCC3 rs2277624 and rs4148416 variants were associated with a greater risk of non-response. Further exploration of the incidence of side effects revealed no significant association for most SNPs; however, ADORA3 rs2298191 and ABCC4 rs7317112 were significantly associated with thrombocytopenia. KM-curves confirmed the detrimental impact of these variants on thrombocytopenia-free survival, with a marked difference observed across genetic subgroups (wild type, one variant, both variants), indicating a gradation in risk. These findings underscore the importance of genetic screening in refining RA treatment with MTX, setting the stage for personalized medicine. Identifying SNPs that affect treatment response and side effects allows for customized treatment plans, improving outcomes and reducing adverse reactions. MTX metabolism involves numerous complex steps and is influenced by a variety of SNPs. Numerous studies have explored various SNPs connected to the effectiveness of MTX in individuals with RA, employing diverse methods to assess its effects (Supplementary Table 2). In this study, we identified three SNPs affecting the response to MTX treatment. Among these SNPs, ABCC3 rs2277624 and rs4148416 have also been implicated in previous research14. This study of 100 Chinese RA patients treated with MTX identified significant correlations with changes in DAS28-ESR (ΔDAS28-ESR), evaluated as the difference from the initial measurement to three months. Regarding the ABCC1 rs212090 variant, which our study found to be linked to a better response to MTX in RA patients, it was also discussed in the aforementioned article14. However, it did not observe a significant benefit in treatment efficacy. Few studies in similar topics have found a correlation with the efficacy of MTX in ABCC1 rs212090. Within the pharmacogenomic landscape, ABCC1 and ABCC3 are critical in mediating the cellular efflux of MTX, thereby influencing its intracellular levels and therapeutic outcomes24–26. The rs212090 polymorphism in ABCC1 is theorized to enhance the transporter’s efficiency, potentially augmenting MTX efflux and, consequently, increasing its concentration in the bloodstream. This may facilitate the administration of higher MTX dosages or improve the precision in targeting diseased areas. Conversely, mutations in ABCC3, such as rs2277624 and rs4148416, could impair transporter function, diminishing MTX efflux and, in turn, lowering the drug’s intracellular concentrations (Fig. 2A). As for the advantages of this paper, our study encompasses the latest research on genes potentially related to MTX, featuring not only a broader range and number of genes but also a larger cohort of participants (n = 1,948). Uniquely, we assessed therapeutic efficacy by evaluating the requirement for bDMARDs or tsDMARDs. In treating RA patients, treatment typically begins with MTX. And bDMARDs or tsDMARDs are considered if there is no significant benefit after six months. This method allows for a more direct evaluation of the impact of different SNPs on treatment efficacy, eliminating the need for excessive biochemical blood tests or complex scoring assessments. Fig. 2Impact of SNPs on Methotrexate efficacy and side effects in Rheumatoid Arthritis patients. (A) The influence of ABCC3 and ABCC1 polymorphisms on MTX metabolism in RA treatment. (B) The role of ADORA3 and ABCC4 polymorphisms in MTX-induced thrombocytopenia in RA patients. Impact of SNPs on Methotrexate efficacy and side effects in Rheumatoid Arthritis patients. (A) The influence of ABCC3 and ABCC1 polymorphisms on MTX metabolism in RA treatment. (B) The role of ADORA3 and ABCC4 polymorphisms in MTX-induced thrombocytopenia in RA patients. MTX is effective for most patients with RA, but it can cause side effects in a specific proportion of patients12. Additionally, the methods used to assess side effects vary across different studies (Supplementary Table 3). Our study reported that ADORA3 rs2298191 and ABCC4 rs7317112 were significantly associated the risk with thrombocytopenia. Although previous literature does not directly link ADORA3 rs2298191 with thrombocytopenia, a study of 127 Serbian patients with RA indicated that this SNP was associated with bone erosions27. Regarding ABCC4 rs7317112, a retrospective cohort study of 375 pediatric patients in the USA with acute lymphoblastic leukemia (ALL) treated with MTX found that variants in ABCC4 were related to an higher probability of delayed clearance28. In pharmacogenomic literature, initiation one of the MTX anti-inflammatory cascade is triggered by the activation of ADORA2A and ADORA329. However, stimulation of ADORA3 (rs2298191) through alterations in cytokine production indirectly impacts metalloproteinase generation, potentially affecting bone architecture30. This may ultimately lead to a decrease in bone marrow activity. Furthermore, ABCC4 is known to encode an anion export transporter protein that facilitates MTX clearance in renal cells31. The SNP rs7317112, a variant located within an intron of ABCC4 has been observed to exhibit a strong correlation with extended MTX clearance times, leading to elevated MTX levels in the blood stream28. Consequently, this could amplify the side effects associated with MTX’s suppression of bone marrow function (Fig. 2B). Finally, MTX is extensively used to treat a variety of diseases, not limited to RA. Consequently, we also explored genes associated with side effects of MTX across different conditions over the past five years. For our analysis of side effects, we utilized more quantifiable measures such as complete blood count and liver enzymes to objectively assess the impact of SNPs on adverse effects. The identified SNPs and their potential side effects have not been previously documented in the literature, possibly owing to variations in the populations studied or the definitions applied. The principal strengths of our study included: Firstly, the sample size of our current study was comparatively large (n = 1948), providing substantial statistical power to identify variants with strong effect sizes Taiwan Han Chinese. Secondly, we performed a comprehensive assessment of a total of 82 SNPs associated with MTX metabolism, which enabled us to thoroughly analyze its clinical efficacy and potential side effects. Thirdly, our approach included a rigorous statistical evaluation, accounting for common potential confounders such as age and sex. Additionally, we applied the Bonferroni correction for multiple testing to mitigate the risk of false positive results, further enhancing the reliability of our findings. These methodological strengths significantly contribute to the robustness and scientific validity of our research. Despite these significant strengths, our study is not without its limitations. Firstly, the retrospective design of our study inherently limits our ability to establish causality and may introduce selection bias. The lack of longitudinal follow-up data restricts our comprehension of the long-term impact of genetic markers on treatment efficacy and safety. Further research is needed to confirm these genetic associations. Secondly, despite a substantial sample size, this single-center study lacks external validation. Furthermore, our cohort comprised solely Han Chinese patients. While these findings may provide relevant insights for other East Asian populations, their generalizability to non-Asian cohorts remains uncertain due to ethnic-specific variations in genetic architectures and allele frequencies. Thirdly, our study determined MTX therapeutic effectiveness using subsequent bDMARD/tsDMARD initiation as a surrogate for treatment failure. While this approach is strongly anchored in strict Taiwan NHI regulations, which mandate objective evidence of high disease activity (DAS28 > 5.1) and thereby minimize biases from physician preference or biologic accessibility, the lack of continuous, longitudinal disease activity scores (e.g., EULAR response) in our database remains a limitation. Furthermore, to mitigate the potential misclassification of early treatment intolerance as inefficacy, we performed a sensitivity analysis excluding patients who escalated therapy within six months (Supplementary Table 7). The findings remained highly consistent with the primary analysis. Fourthly, the inability to confirm patient adherence to MTX therapy presents a significant limitation to our study. Nonadherence to prescribed medication regimens is a well-documented issue in clinical studies and can substantially influence treatment outcomes32,33. Fifthly, due to the retrospective nature of this database study, we lacked granular data on several dynamic clinical variables, including the precise MTX dosage, route of administration, compliance with folic acid supplementation, longitudinal changes in renal and liver function, exact disease duration, and specific concomitant medications. The absence of these variables may lead to residual confounding, acting as unmeasured factors that could independently influence MTX efficacy and the occurrence of adverse events. Future prospective studies comprehensively adjusting for these variables are warranted to validate our findings. Finally, to control Type I error when evaluating multiple SNPs, we applied the conservative Bonferroni correction (Supplementary Table 4). Although the associations for ABCC3 and ABCC1 variants did not survive this adjustment (Bonferroni p > 0.05) and are considered exploratory, their effect sizes remained directionally consistent across primary and sensitivity analyses. Specifically, ABCC3 variants consistently conferred an increased risk of MTX non-response (OR > 1), while ABCC1 rs212090 showed a protective trend (OR = 0.83), suggesting a plausible biological signal rather than random variation. These findings are biologically plausible given the roles of ABCC transporters in MTX efflux and intracellular drug availability, and should be considered hypothesis-generating, warranting replication in larger, prospective cohorts and functional validation. Conclusion In summary, our research delineates specific SNPs in ABCC1 and ABCC3 that correlate with the effectiveness of MTX in treating RA, while identifying SNPs in ADORA3 and ABCC4 that are linked to the side effect of thrombocytopenia. These findings provide a basis for customizing treatment plans for individual patients based on their genetic makeup. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2 Supplementary Material 3 Supplementary Material 3