HLA-C*03:04:01 and HLA-B*15:18:01 but not HLA-DQA1*05 associated with anti- tumor necrosis factor antibody formation in Taiwanese inflammatory bowel disease patients

Abstract

BACKGROUND

Anti-drug antibodies (ADAs) can reduce the effectiveness of biologics. While human leukocyte antigen (HLA)-DQA1*05 allele is linked to ADA formation in European Crohn’s disease patients, its relevance in non-European populations remains unclear.

AIM

To investigate HLA genotypes associated with the development of ADAs in Taiwanese inflammatory bowel disease (IBD) patients treated with biologics.

METHODS

In this multicenter study, IBD patients treated with anti-tumor necrosis factor (TNF), anti-integrin, or anti-interleukin (IL)-12/23 therapies from April 2022 to June 2024 were enrolled. All participants underwent next-generation sequencing for HLA genotyping. ADA levels were measured via enzyme linked immunosorbent assay. HLA allele frequencies were compared between ADA-positive and ADA-negative groups, and against general Taiwanese population data.

RESULTS

Ninety-five IBD patients were included: 58 received anti-TNF therapy (38 infliximab, 20 adalimumab), 27 anti-integrin, and 10 anti-IL-12/23. ADAs occurred only in the anti-TNF group (n = 22): 19 infliximab (50%) and 3 adalimumab (15%). No ADAs developed in patients on anti-integrin or anti-IL-12/23 agents. HLA-C*03:04:01 was significantly associated with anti-infliximab ADAs (31.6% vs 0%, P = 0.02), and HLA-B*15:18:01 with anti-adalimumab ADAs (66.7% vs 0%, P = 0.016). HLA-DQA1*05 was not associated with ADA formation. Frequencies of HLA-C*03:04:01 (8.4% vs 10.5%) and HLA-B*15:18:01 (1.6% vs 0.6%) in IBD patients were comparable to those in the general population. ADA titers were inversely correlated with serum drug levels.

CONCLUSION

In Taiwanese IBD patients, HLA-C*03:04:01 and HLA-B*15:18:01 were significantly associated with ADA development to infliximab and adalimumab, respectively. HLA-DQA1*05 was not predictive, highlighting ethnic differences in genetic predisposition to immunogenicity.

Key Words: Inflammatory bowel disease; Crohn’s disease; Ulcerative colitis; Anti-tumor necrosis factor therapy; Anti-drug antibodies; Human leukocyte antigen genotype

Core Tip: The development of anti-drug antibodies (ADA) compromises the effectiveness of biologic therapies in treating inflammatory bowel disease. In European inflammatory bowel disease patients receiving anti-tumor necrosis factor therapy, the human leukocyte antigen (HLA)-DQA1*05 allele has been strongly associated with increased ADA formation. However, this association does not appear to hold in Taiwanese patients. Instead, distinct HLA variants have emerged: HLA-C*03:04:01 is linked to ADA formation in those treated with infliximab, while HLA-B*15:18:01 is associated with immunogenicity to adalimumab. Notably, in the Taiwanese cohort, HLA-DQA1*05 does not predict ADA development during anti-tumor necrosis factor treatment.

INTRODUCTION

Inflammatory bowel diseases (IBDs), which include Crohn’s disease (CD) and ulcerative colitis (UC), are chronic inflammatory disorders of the intestine. IBD incidence is increasing more quickly over time, as of 2020, in Taiwan and other newly industrialized countries[1]. In terms of cross-regional age-standardized rates, East Asia was reported to have the fastest average annual growth rate at 2.54% (95% confidence interval: 2.4-2.68)[2]. IBD progression causes organ damage, leading to disability and imposing a substantial economic and social burden[3,4].

The first biologic introduced for IBD management was tumor necrosis factor (TNF) inhibitor[5]. However, repeated administration often triggers the formation of anti-drug antibodies (ADAs) due to its immunogenicity, resulting in treatment failure[6]. Approximately 40% to 60% of patients initially responsive to TNF inhibitors lose responsiveness over time[7]. Additionally, among initial responders, the annual risk of response loss is approximately 13% per patient/year[8]. Both patient- and treatment-related factors influence the immunogenicity of TNF inhibitors. Patient-related risk factors include genetic predisposition, longer disease duration, higher baseline disease activity, and insufficient drug levels[9,10]. Treatment-related factors involve the structure and composition of the monoclonal antibody, dosage, administration route, and concurrent medication use[11]. Immunogenicity is more commonly observed with infliximab, a murine-human chimeric monoclonal antibody, than with adalimumab, a fully human monoclonal antibody. Despite the availability of newer biologics, such as anti-integrin monoclonal antibodies, anti-interleukin (IL)-12/IL-23 antibodies, and anti-IL-23 agents, TNF inhibitors remain the most well-supported treatment option, particularly for patients with complicated, severe, perianal, or post-operative IBD. Therefore, the optimization of therapeutic strategies, such as therapeutic drug monitoring (TDM) and immunogenicity management, remains a critical priority.

A recent genome-wide association study identified the human leukocyte antigen (HLA)-DQA1*05 variant as a key genetic factor that doubles a patient’s risk of developing antibodies against infliximab and adalimumab, regardless of concomitant immunomodulator (IM) use[12]. Among carriers of HLA-DQA1*05 who were administered infliximab monotherapy, the immunogenicity rate reached 92% at 1 year, whereas noncarriers receiving adalimumab combination therapy had the lowest immunogenicity rate at 1 year of 10%. However, the study was conducted exclusively in European patients, only 40% of whom carried the risk allele, limiting its generalizability to other populations. Currently, no diagnostic tools can predict which patients will develop ADAs and lose therapeutic responsiveness. In this study, we investigated the association between specific HLA alleles and immunogenicity to biologics, focusing on whether HLA-DQA1*05 is linked to ADA formation against TNF inhibitors in Taiwanese patients with IBD and its effect on drug persistence.

MATERIALS AND METHODS

Patient population

This multicenter study enrolled patients with IBD who received anti-TNF, anti-integrin, or anti-IL-12/23 therapy between April 1, 2022, and June 30, 2024. The study was conducted across three medical centers: National Taiwan University Hospital Main Hospital, National Taiwan University Hospital Hsin-Chu Branch, and MacKay Memorial Hospital. The institutional review board of National Taiwan University Hospital approved the study protocol, No. 202112095RIND. The study was conducted in accordance with the principles of the Declaration of Helsinki. All participants provided written informed consent before their enrollment.

CD or UC diagnoses were confirmed through standard clinical, endoscopic, radiological, and histological assessments. The inclusion criteria required participants to be patients with IBD undergoing anti-TNF, anti-integrin, or anti-IL-12/23 treatment. Patients who declined to participate or were human immunodeficiency virus-positive were excluded. Data on sociodemographic characteristics, disease duration, medication history, specific disease phenotypes, and clinical outcomes were gathered. Infliximab and vedolizumab were administered at weeks 0, 2, 6, 14, 22, 30, 38, and 46; adalimumab was administered every 2 weeks; and ustekinumab was administered as an initial IV dose at week 0 followed by subcutaneous doses at week 8 and every 12 weeks thereafter. Thus, week 46 was selected as a representative endpoint for assessment of trough drug levels and ADA formation at the conclusion of the standard 1-year treatment period for all biologics. TDM was conducted before drug administration at weeks 0, 6, 14, 22, 30, 38, 46, and 54 and subsequently every 4 months or during symptom flare-ups while on biologic therapy.

Drug persistence was defined as the duration from treatment initiation to discontinuation. An event was recorded when patients developed ADAs or switched to an alternative biologic agent due to treatment failure. Censoring was defined as event absence, loss to follow-up, or discontinuation due to the drug holiday specified under Taiwan’s National Health Insurance reimbursement criteria. Insurance budgetary constraints mandate a drug holiday after 1 year of biologic therapy, unless patients elect to continue treatment at their own expense. Blood samples for HLA genotyping were collected at week 0. HLA genotype data from individuals in the control group were obtained from the Taiwan Biobank (https://www.biobank.org.tw/).

TDM

Venous blood samples were harvested in serum tubes immediately before biologic administration to ensure that the drugs were at the lowest possible blood concentration (trough level) in all patients. The collection tubes were centrifuged at 3000 rpm for 10 minutes at room temperature. The serum was then transferred into cryotubes and stored at -20 °C until analysis. Serum levels of infliximab, adalimumab, vedolizumab, and ustekinumab were assessed with IDKmonitor infliximab, adalimumab, vedolizumab, and ustekinumab drug level enzyme linked immunosorbent assay (ELISA) kits (Immundiagnostik AG, Bensheim, Germany). Anti-infliximab, anti-adalimumab, antivedolizumab, and antiustekinumab antibodies were measured with anti-infliximab TM09003 ELISA, adalimumab-antibody duo ELISA (Immuno-Biological Laboratories, MN, United States), IDKmonitor vedolizumab free ADA ELISA, and IDKmonitor ustekinumab free ADA ELISA (Immundiagnostik AG, Bensheim, Germany) kits, respectively. All assays were performed according to protocols provided by the kit manufacturers.

DNA preparation

Whole blood (2 mL) was collected from all participants with IBD. A Puregene DNA Isolation Kit (Gentra Systems, Minneapolis, MN, United States) was used for DNA extraction. A Qubit 2.0 fluorometer (Thermo Fisher Scientific, Waltham, MA, United States) was used to quantify and assess the quality of 1000 ng of input genomic DNA prior to library preparation. We fragmented the genomic DNA using Covaris (Covaris, Woburn, MA, United States), targeting a peak length of 800 bp, which was assessed using an Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, United States). We then added adapters and barcodes onto the ends of the fragmented DNA to generate indexed sequencing libraries using a TruSeq Library Preparation Kit (Illumina, San Diego, CA, United States).

HLA typing

The procedures for DNA fragmentation, library preparation, probe design for HLA genes, capture-based target enrichment, and next generation sequencing with an Illumina system followed those outlined in a previous paper[13]. The panel comprised HLA-A, HLA-B, HLA-C, HLA-DRB1, HLA-DPA1, HLA-DPB1, HLA-DQA1, HLA-DQB1, HLA-DRB3, HLA-DRB4, and HLA-DRB5.

Statistical analysis

Data are reported in terms of the median and interquartile range or in terms of the frequency and percentage. χ² or Fisher’s exact tests were performed as appropriate to compare categorical variables. Kaplan-Meier and Kruskal-Wallis tests were performed to assess drug persistence across medications. Immunogenicity rates were estimated according to the Kaplan-Meier method; patients who did not develop immunogenicity were censored at their last recorded visit. All statistical tests were two-tailed, with a significance threshold set at P < 0.05.

RESULTS

Patient characteristic

A total of 95 patients with IBD were enrolled consequently in the study, along with HLA genotype data from 1097 control individuals obtained from the Taiwan Biobank. Among these IBD patients, 20 had UC, and 75 had CD. In terms of biologic therapy, 38 patients received infliximab, 20 were treated with adalimumab, 27 received anti-integrin therapy (vedolizumab), and 10 were given anti-IL-12/23 therapy (ustekinumab). A total of 79 patients (83.2%) received combination therapy with biologics and IMs. All patients receiving anti-TNF therapy were also on concomitant immunosuppressive treatment, except for one individual who discontinued IM due to adverse effects. By contrast, approximately 40% of those on vedolizumab or ustekinumab were not prescribed additional IMs. Drug monitoring was conducted for all participants. IM selection remained consistent throughout the course of biologic therapy. The immunotherapy regimen remained unmodified throughout treatment. The median age was 36.1, 33.0, 43.8, and 40.9 years in the infliximab, adalimumab, vedolizumab, and ustekinumab groups, respectively. Median drug trough levels were 4.2, 9.4, 12.7, and 1.9 μg/mL for infliximab, adalimumab, vedolizumab, and ustekinumab, respectively. Among patients treated with TNF inhibitors, 50% of those receiving infliximab developed ADAs, compared with 15% in the adalimumab group. By contrast, no ADA development was observed in patients treated with anti-integrin or anti-IL-12/23 therapies (Table 1). We analyzed the drug persistence associated with individual biologic agents, which differed significantly between groups (P = 0.001). The median treatment durations were as follows: Infliximab, 104 weeks; adalimumab, 316 weeks; vedolizumab, 120 weeks; and ustekinumab, 157 weeks (Table 1). Within the overall IBD cohort, adalimumab was associated with a significantly longer treatment duration compared with both infliximab (P < 0.001) and vedolizumab (P = 0.005) but not ustekinumab (Figure 1A). After stratification by disease subtype, in patients with CD, adalimumab maintained its longer treatment duration compared with infliximab (P < 0.001) and vedolizumab (P = 0.002) and did not significantly differ from the treatment duration of ustekinumab (Figure 1B). Due to the limited number of UC cases, treatment duration was not compared between biologics within this subgroup. Kaplan-Meier survival analysis further supported these findings, indicating greater treatment persistence with adalimumab and a higher rate of treatment failure with infliximab (Figure 1C). Adverse events were infrequent: One patient experienced arthralgia following infliximab administration, and another developed skin itching and rash after adalimumab administration. No patient discontinued advanced therapy due to adverse events. No patient in the study cohort died during the observation period.

Figure 1 Drug persistence of biologic agents. A: Overall inflammatory bowel disease cohort; B: Crohn’s disease cohort. Both determined according to a Kruskal-Wallis test; C: Kaplan-Meier analysis with log-rank test comparing drug persistence among biologic agents in the inflammatory bowel disease cohort.

Table 1 Clinical characteristics of the inflammatory bowel disease patients, n (%)/median (interquartile range).

	Infliximab (n = 38)	Adalimumab (n = 20)	Vedolizumab (n = 27)	Ustekinumab (n = 10)
Age (years)	36.1 (28.5-46.6)	33.0 (28.6-42.2)	43.8 (34.8-54.8)	40.9 (32.2-51.5)
Gender (men)	29 (76.3)	14 (45.2)	17 (63.0)	8 (80)
Phenotype
UC	5 (13.2)	2 (10)	13 (48.1)	0 (0)
CD	33 (86.8)	18 (90)	14 (51.9)	10 (100)
Disease duration (years)	8.3 (3.8-13.3)	11.4 (8.3-12.3)	10 (6.1-11.1)	8.2 (3.58-14.5)
Concomitant IM
Thiopurine	33 (86.8)	15 (70)	16 (59.3)	5 (50)
Methotrexate	4 (10.5)	5 (25)	0 (0)	0 (0)
Tacrolimus	0 (0)	0 (0)	0 (0)	1 (10)
No	1 (2.6)	0 (0)	11 (40.7)	4 (40)
Drug duration (weeks)	104 (33.5-221)	316 (189-489)	120 (55-214.5)	157 (110-240)
Drug trough level (μg/mL)	4.2 (2-6.8)	9.4 (4.5-14.9)	12.7 (5.4-16.7)	1.9 (1.1-4.2)
ADA formation	19 (50)	3 (15)	0 (0)	0 (0)

UC: Ulcerative colitis; CD: Crohn’s disease; IM: Immunomodulator; ADA: Anti-drug antibody.

Association between HLA alleles and immunogenicity

A comparison of patients with and without ADAs revealed a significant association between the HLA-C*03:04:01 allele and the development of anti-infliximab antibodies (31.6% vs 0%, P = 0.02). Similarly, the HLA-B*15:18:01 allele was significantly associated with the development of anti-adalimumab antibodies (66.7% vs 0%, P = 0.016). By contrast, the HLA-DQA1*05 allele was not associated with the development of anti-infliximab or anti-adalimumab antibodies (Table 2). A further analysis of HLA-DQA1*05 subtypes likewise revealed no significant association of any agent with immunogenicity (Supplementary Tables 1 and 2).

Table 2 Association between human leukocyte antigen alleles and immunogenicity, %.

HLA	Anti-infliximab antibody (-) (n = 19)	Anti-infliximab antibody (+) (n = 19)	P value
HLA-C*03:04:01	0	31.6	0.02
HLA-DQA1*05	36.8	52.6	0.328
HLA	Anti-adalimumab antibody (-) (n = 17)	Anti-adalimumab antibody (+) (n = 3)
HLA-B*15:18:01	0	66.7	0.016
HLA-DQA1*05	52.9	66.7	1.000

HLA: Human leukocyte antigen.

Among patients treated with infliximab who developed ADAs, the median time to antibody development was 37 weeks (interquartile range, 44.0-241.0). Kaplan-Meier analysis revealed a higher incidence of ADA development in infliximab-treated patients carrying HLA-C*03:04:01, although this difference was nonsignificant (P = 0.067; Figure 2). Patients treated with adalimumab had a median ADA development time of 32 weeks, and a Kaplan-Meier analysis revealed a significantly higher rate of ADA development in those carrying HLA-B*15:18:01 (P < 0.001; Figure 2). No significant association was observed between the HLA-DQA1*05 allele and the development of anti-infliximab or anti-adalimumab antibodies (Figure 2).

Figure 2 Kaplan-Meier estimation of the rate of anti-drug antibodies development. A: Infliximab-treated group, stratified according to human leukocyte antigen (HLA)-C*03:04:01 carrier status; B: Infliximab-treated group, stratified according to HLA-DQA1*05 carrier status; C: Adalimumab-treated group, stratified according to HLA-B*15:18:01 carrier status; D: Adalimumab-treated group, stratified according to HLA-DQA1*05 carrier status. HLA: Human leukocyte antigen.

Negative correlation of serum ADA titer with serum drug trough levels

We further investigated the association between serum ADA titers and serum drug trough levels. Immunogenicity was defined as an ADA titer exceeding 10 AU/m. Serum analysis was conducted at the time of immunogenicity development or at week 46 for patients who did not develop ADAs. A negative correlation was observed between ADA titers and serum drug trough levels in both the infliximab-treated group (r = -0.446, P = 0.006) and the adalimumab-treated group (r = -0.455, P = 0.044), suggesting that higher ADA levels were associated with lower drug trough (Figure 3).

Figure 3 Spearman’s correlation between serum anti-drug antibodies titers and serum drug trough levels. A: Infliximab-treated group; B: Adalimumab-treated group. ADA: Anti-drug antibody.

Comparison of HLA allele frequencies between Taiwanese patients with IBD and the general population

To further explore genetic predisposition toward ADA development, we compared the frequency of HLA alleles in Taiwanese patients with IBD with those in the general population. The frequency of HLA-C*03:04:01 (8.4% vs 10.5%) and HLA-B*15:18:01 (1.6% vs 0.6%) was comparable between IBD patients and the general population. Likewise, the prevalence of the HLA-DQA1*05 allele did not differ significantly between patients with IBD and the general population (25.3% vs 19.8%; Table 3). The frequency of several other HLA alleles differed significantly between patients with IBD and the general population, suggesting potential genetic associations with disease susceptibility or resistance (Supplementary Table 3). These findings provide further insight into the immunogenetic landscape of IBD in Taiwanese patients.

Table 3 Frequency of anti-drug antibody associated human leukocyte antigen alleles in Taiwanese inflammatory bowel disease patients and the general population, %.

HLA	IBD (n = 190)	Control (n = 2194)	P value
HLA-C*03:04:01	8.4	10.5	0.3701
HLA-B*15:18:01	1.6	0.6	0.1484
HLA-DQA1*05	25.3	19.8	0.0749

IBD: Inflammatory bowel disease; HLA: Human leukocyte antigen.

DISCUSSION

Taiwanese patients with IBD in our cohort who carried the HLA-C*03:04:01 allele had a higher prevalence of anti-infliximab antibodies compared with those without this allele. Similarly, the HLA-B*15:18:01 allele was significantly associated with the development of anti-adalimumab antibodies. ADA titers and serum drug trough levels were negatively correlated. HLA-C*03:04:01 is a specific allele of the HLA-C gene, which belongs to the major histocompatibility complex class I system responsible for presenting peptide antigens to cytotoxic T lymphocytes. The HLA-C*03:04:01 allele has been associated with acetylcholine receptor antibody-positive adult-onset myasthenia gravis in Chinese patients[14] and identified as a risk variant for sarcoidosis, an autoimmune condition characterized by granulomatous inflammation[15]. HLA-B*15:18, an allele of the HLA-B gene within the major histocompatibility complex class I family, also plays a role in antigen presentation and has been linked to improved responsiveness to immunosuppressive therapy in patients with severe aplastic anemia[16] and an increased risk of amoxicillin-clavulanate-induced liver injury[17]. These associations collectively suggest that both alleles participate in immunologically relevant pathways and may influence immune tolerance or susceptibility to aberrant immune activation. Despite these observations, the precise molecular and structural mechanisms through which these alleles might contribute to ADA formation remain undefined. This knowledge gap is not unexpected, as detailed mechanistic explanations are lacking for most well-established HLA-disease or HLA-phenotype associations. It is plausible that specific peptide-binding properties or differential interactions with killer cell immunoglobulin-like receptors could alter immune regulation in carriers of these alleles, but such hypotheses require empirical validation. Consequently, replication of our findings in independent cohorts and the implementation of functional studies, such as peptide-binding assays, structural modeling, and T-cell activation analyses, will be essential to clarify these relationships. Understanding the mechanistic underpinnings of HLA-associated ADA formation would not only strengthen the biological plausibility of our results but could also inform the development of predictive tools for immunogenicity risk stratification and guide personalized therapeutic strategies. We view our current findings as an important first step in this direction, providing a foundation for future studies aimed at bridging the gap between genetic association and functional immunology.

Notably, many studies have demonstrated that HLA-disease associations are often shaped by ethnic background. For example, HLA-B*15:02 is strongly associated with carbamazepine-induced Stevens-Johnson syndrome in Han Chinese[18], whereas HLA-A*30:01 has been identified as the primary risk allele in Japanese and Caucasian populations[19]. Likewise, HLA-B*38:02 is the key risk allele for antithyroid drug-induced agranulocytosis in Taiwanese patients[20], while HLA-B*27:05 is implicated in antithyroid drug-induced agranulocytosis among White Europeans[21]. These findings suggest that while HLA-associated genetic susceptibility to immunogenicity exists, the relevant loci may differ between Western and Asian populations.

In contrast to previous studies, we observed no significant association between HLA-DQA1*05 and ADA development in patients treated with TNF inhibitors. In the PANTS (Personalised Anti-TNF Therapy in CD) study, Sazonovs et al[12] identified HLA-DQA1*05 as a risk allele for increased immunogenicity to infliximab and adalimumab. Similar associations were reported by Reppell et al[22] in the SERENE UC and CD trials, and a meta-analysis of 10 studies involving 2984 patients confirmed that HLA-DQA1*05 carriers had a higher risk of developing immunogenicity to TNF inhibitors[23]. A subsequent meta-analysis of 13 studies involving 3756 patients with various immune-mediated inflammatory diseases reinforced this association[24].

The proposed mechanism involves HLA class II molecules, heterodimers consisting of DQA (alpha) and DQB (beta) chains, which present epitopes to T-helper cells. HLA-DQA1 is expressed on antigen-presenting cells and plays a key role in forming the antigen-binding site. Immunogenicity predictions often rely on well-characterized interactions between HLA molecules and specific protein epitopes[25]. Several studies suggest that the risk of immunogenicity may be influenced by multiple HLA class II alleles, not just a single genetic variant[26].

Consistent with our findings, several studies have not demonstrated a significant association between HLA-DQA1*05 and anti-TNF immunogenicity[27,28]. Several factors may explain this discrepancy. First, patient ethnicity may contribute to the variation in study findings. Second, nearly all patients in our cohort (57/58) received combination therapy with TNF inhibitors and IMs. Concomitant IMs reduce ADA development and increase the efficacy of TNF inhibitors, possibly attenuating the effects of genetic predisposition on immunogenicity[12,29]. We acknowledge that concomitant IM use could modulate ADA development and, consequently, affect the observed associations. However, the high proportion of combination therapy in our study enabled us to investigate genetic predisposition within a clinically relevant context, where immunomodulatory therapy is considered the standard of care.

We observed significant differences in drug persistence between biologic agents. Adalimumab had the longest treatment duration, which was nonsignificantly longer than that of ustekinumab and significantly longer than the duration of both vedolizumab and infliximab treatment. This variation may be partly explained by the reimbursement timeline for biologics in Taiwan: Adalimumab was approved in 2011, followed by infliximab in 2015, vedolizumab in 2017, and ustekinumab in 2020. Under the consensus guidelines of the Taiwan Society of IBD, infliximab, adalimumab, vedolizumab, and ustekinumab are all recommended induction and maintenance therapy options for patients with moderate-to-severe active CD or refractory to conventional treatment for patients with UC[30]. Additionally, all patients administered adalimumab in our cohort were treated with concomitant IMs, which may have contributed to its superior persistence. Thiopurines are known to increase the durability of anti-TNF therapy in both CD and UC, and methotrexate has demonstrated similar benefits in CD[31].

In addition to HLA-associated immunogenicity, we evaluated several nongenetic factors that have been implicated in ADA development, including age, sex, and IM use. None of these variables were significantly associated with ADA development (Supplementary Table 4). Because biologic therapy in Taiwan is reimbursed only for patients with severe disease activity (composite dietary antioxidant index > 300 for CD and Mayo score > 9 for UC) and only at standard dosing, baseline disease severity and treatment dose were likely consistent across our cohort, minimizing their potential as confounders. Regarding treatment interruptions, we noted that the median time to ADA development for anti-TNF agents was approximately 32-37 weeks, suggesting that only a limited number of patients would have experienced treatment holidays during this critical window of immunogenicity. Future studies with larger sample sizes and detailed medication records are warranted to clarify the differential effects of various treatments on ADA suppression.

We observed a consistent inverse relationship between ADA titers and serum drug trough levels in both the infliximab and adalimumab groups, reflecting the findings of previous studies[32,33]. The primary reasons for discontinuation of biologic therapy in our cohort were subtherapeutic drug levels, which often result from ADA development, and financial constraints, given that Taiwan’s National Health Insurance currently covers only a 1-year course of biologic treatment. Patients wishing to continue therapy beyond this period must cover the cost out of pocket. TDM, the measurement of drug levels and ADA titers, is a crucial tool for maintaining therapeutic drug exposure, assessing immunogenicity, and optimizing treatment[34]. TDM is grounded in a well-established exposure-response relationship, wherein adequate serum drug levels are associated with better clinical outcomes, such as remission and mucosal healing[35,36]. The formation of neutralizing ADAs can hasten drug clearance, often leading to secondary loss of response. In such cases, dose intensification, involving either an increased dosage or shortened dosing interval, is commonly employed to restore drug levels. TDM also informs decisions regarding combination therapy with IMs. For patients at high risk of ADA development (e.g., those carrying specific HLA alleles), TDM can aid providers in selecting the most appropriate biologic agent and determining whether prophylactic combination therapy is necessary. TDM is based on both pharmacokinetics and immunogenicity and supports a personalized and proactive approach to IBD management.

Another important observation in our study was the timing of ADA detection: The median time to ADA development for anti-TNF agents was approximately 32-37 weeks, even in patients receiving IMs. Although current guidelines recommend the administration of combination therapy with infliximab and thiopurines for at least 6-12 months to reduce immunogenicity[37], our findings suggest that even sustained combination therapy may only delay rather than fully prevent ADA development. Thus, trough levels should be continuously monitored to ensure adequate drug exposure and long-term treatment persistence.

This study has several limitations. First, the sample size was relatively small. Larger cohorts, particularly from Asian populations, may be needed to detect subtle genetic effects. Second, most of the patients received combination therapy, regardless of their HLA status. As immunosuppressive co-therapy reduces ADA formation, this may have masked any genetic predisposition. Third, we could detect only unbound antibodies and not those complexed with the drug through the free ADA ELISA. Consequently, certain ADAs may have gone undetected, potentially causing us to underestimate the true immunogenicity rates. Forth, the small number of ADA-positive cases, especially in the adalimumab subgroup, limits statistical power, potentially increasing the likelihood of false-positive associations. Despite these limitations, this study has some notable strengths. First, this study had a long follow-up period for its cohort, allowing for meaningful longitudinal assessment. Second, our use of TDM and pharmacokinetic data allowed us to evaluate drug exposure and immune response patterns in detail. Third, and crucially, high-resolution HLA genotyping was performed through next generation sequencing, yielding comprehensive and accurate genetic data for analysis.

CONCLUSION

We identified HLA-C*03:04:01 and HLA-B*15:18:01 as risk alleles for anti-infliximab and anti-adalimumab antibody development, respectively, in a Taiwanese IBD cohort. No significant association was observed between HLA-DQA1*05 carrier status and ADA formation. Further research involving larger, ethnically diverse cohorts treated with individual biologics is warranted to validate these findings and clarify the role of specific HLA alleles in immunogenicity risk.

ACKNOWLEDGEMENTS

We thank the 2^nd Core Lab of the National Taiwan University Hospital main hospital for the facility support.