Anti-TNF immunogenicity in a Middle Eastern inflammatory bowel disease cohort: prevalence, predictors and context-specific drug level thresholds Authors: Shaima Wasim Khan, Noorah Al Hosani, Thaer Khaleel Swaid, Mazin M Edan, Mohammed Nabil Quraishi Published: 2026 ABSTRACT AbstractObjectiveAnti-tumour necrosis factor (anti-TNF) immunogenicity remains a major barrier to treatment persistence in inflammatory bowel disease, yet data from Middle Eastern populations are lacking. We characterised immunogenicity rates, predictors and loss of response mechanisms in a real-world cohort.MethodsThis retrospective cohort study included 314 anti-TNF treatment courses (212 infliximab, 102 adalimumab) in 248 patients at a tertiary centre in the United Arab Emirates. Immunogenicity was defined by detectable anti-drug antibodies on a drug-tolerant electrochemiluminescent bridging immunoassay with acid dissociation. Drug levels, predictors and loss of response mechanisms were analysed.ResultsImmunogenicity developed in 28.3% (89/314) of courses over median follow-up of 24.0 months (median time to detection 15.0 months, IQR 8.0–36.0). Infliximab and adalimumab had comparable rates (26.9% vs 31.4%; p=0.425). Immunogenicity-mediated failure was the leading cause of discontinuation (48.4%). Pre-event trough levels were significantly lower in immunogenic courses (median 3.0 vs 14.0 mcg/mL; p<0.001). Exploratory receiver operating characteristic thresholds of 5.3 mcg/mL for infliximab (area under the curve (AUC) 0.880) and 6.4 mcg/mL for adalimumab (AUC 0.861) predicted immunogenicity with high discrimination. On shared frailty Cox regression, prior surgery was the strongest predictor (HR 1.85, 95% CI 0.94 to 3.64; p=0.074). A sensitivity analysis excluding rescued patients strengthened the surgery signal (HR 2.80, 95% CI 1.41 to 5.56; p=0.003). Subcutaneous infliximab showed lower immunogenicity than intravenous infliximab (new starters 1/25 (4.0%), switchers 1/33 (3.0%) vs IV 57/212 (26.9%)); after propensity score matching on five covariates, the difference remained significant (3.8% vs 18.3%; p=0.013), though a new starters-only analysis did not reach significance (p=0.062).ConclusionImmunogenicity patterns in this Middle Eastern cohort are consistent with Western data, supporting pharmacokinetic rather than population-specific determinants. These findings suggest a potential benefit of proactive therapeutic drug monitoring and that subcutaneous infliximab may offer an immunogenic advantage warranting prospective validation.Keywords: IBD, INFLIXIMAB, IMMUNOGENETICS, INFLAMMATORY BOWEL DISEASE WHAT IS ALREADY KNOWN ON THIS TOPIC Anti-tumour necrosis factor (anti-TNF) immunogenicity drives loss of response in inflammatory bowel disease (IBD), but data from Middle Eastern populations are lacking, and context-specific drug level thresholds remain poorly defined. Anti-tumour necrosis factor (anti-TNF) immunogenicity drives loss of response in inflammatory bowel disease (IBD), but data from Middle Eastern populations are lacking, and context-specific drug level thresholds remain poorly defined. WHAT THIS STUDY ADDS In 314 anti-TNF courses across 248 IBD patients in the UAE, the overall immunogenicity rate was 28.3%, consistent with Western data and suggesting pharmacokinetic rather than population-specific determinants. Trough drug levels showed high discriminative accuracy for immunogenicity prediction, whereas concomitant immunomodulator use did not demonstrate a clear protective effect. In 314 anti-TNF courses across 248 IBD patients in the UAE, the overall immunogenicity rate was 28.3%, consistent with Western data and suggesting pharmacokinetic rather than population-specific determinants. Trough drug levels showed high discriminative accuracy for immunogenicity prediction, whereas concomitant immunomodulator use did not demonstrate a clear protective effect. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY These findings support proactive therapeutic drug monitoring (TDM) as the primary strategy for mitigating anti-TNF immunogenicity and suggest that adequate drug exposure through TDM-guided optimisation may be more important than immunomodulator coprescription alone. These findings support proactive therapeutic drug monitoring (TDM) as the primary strategy for mitigating anti-TNF immunogenicity and suggest that adequate drug exposure through TDM-guided optimisation may be more important than immunomodulator coprescription alone. Introduction Inflammatory bowel disease (IBD), encompassing Crohn’s disease (CD) and ulcerative colitis (UC), is a chronic immune-mediated disorder of the gastrointestinal tract with accelerating global incidence.1 Anti-tumour necrosis factor (anti-TNF) biologic therapies, principally infliximab (IFX) and adalimumab (ADL), remain a cornerstone of moderate-to-severe IBD management, with combination anti-TNF and immunomodulator therapy superior to monotherapy.2,5 Despite newer biologics and small molecules, network meta-analyses confirm the continued pre-eminence of anti-TNF agents.6,8 However, the long-term effectiveness of anti-TNF therapy is substantially limited by immunogenicity, the development of anti-drug antibodies (ADA). ADA formation has been reported in up to 61% of patients receiving episodic infliximab, with high ADA concentrations associated with shorter duration of response and increased infusion reactions.9 The UK Personalised Anti-TNF therapy in Crohn's disease (PANTS) study subsequently identified immunogenicity as a key predictor of anti-TNF treatment failure at both 1 and 3 years of follow-up, with a clear inverse relationship between drug trough concentrations and ADA risk.10 11 Immunogenicity is the principal driver of secondary loss of response (LOR) and treatment discontinuation, directly undermining the persistence that is essential to realise the long-term benefits of these highly effective agents. Yet systematic characterisation of immunogenicity and its impact on drug persistence in real-world cohorts remain limited. Despite this wealth of evidence from Western cohorts, immunogenicity rates, trough level thresholds and predictors of ADA formation have not been characterised in Middle Eastern IBD populations. No published study has reported comprehensive anti-TNF immunogenicity outcomes from the Gulf region.12 This study aimed to characterise the prevalence and temporal patterns of anti-TNF immunogenicity, identify clinical and treatment-related predictors of antibody development and determine drug-specific trough level thresholds predictive of immunogenicity in a real-world Middle Eastern IBD cohort. The study also sought to classify LOR by underlying mechanism, evaluate clinical outcomes following same-agent rechallenge and sequential anti-TNF courses, and compare immunogenicity rates between subcutaneous and intravenous infliximab formulations. Methods Study design and population This was a retrospective cohort study conducted at a tertiary care IBD facility in Abu Dhabi, United Arab Emirates that serves a predominantly Arab population. Data were extracted from the prospectively maintained SSMC IBD registry (up to February 2026) and supplemented by electronic health record review. Eligible patients were adults (≥18 years) with confirmed IBD who had received at least one course of anti-TNF therapy (IFX or ADL) with standard induction dosing. Both originator and biosimilar preparations were included. Patients were excluded if anti-TNF was administered solely within a clinical trial or if immunogenicity monitoring had not been performed. The primary unit of analysis was the treatment course, defined as a discrete period of exposure to a specific anti-TNF agent; patients could contribute multiple courses. A separate patient-level analysis accounted for within-patient correlation. This study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (online supplemental material). Data collection Variables extracted included patient demographics, disease classification (IBD subtype, Montreal Classification), clinical characteristics at treatment start (body mass index (BMI), serum albumin, C reactive protein (CRP), smoking status, prior surgery), treatment details (drug, line of therapy, dates), concomitant immunomodulator use, therapeutic drug monitoring (TDM) data (trough drug levels, ADA levels with detection dates) and clinical outcomes including LOR classification. TDM was usually performed proactively post-induction for infliximab-treated patients; however, hybrid approaches were used during the study period, with reactive monitoring also performed when patients presented with evidence of active disease supported by biochemical (faecal calprotectin, CRP) or endoscopic parameters, rather than clinical symptoms alone. This evolved over time as the centre developed its TDM programme. Drug levels were measured by selective reaction monitoring liquid chromatography tandem mass spectrometry (LC-MS/MS) and ADA by electrochemiluminescent bridging immunoassay with acid dissociation (Mayo Clinic Laboratories). This methodology is drug-tolerant, allowing ADA detection even in the presence of circulating drug. Infliximab samples were collected as preinfusion troughs; adalimumab and subcutaneous infliximab (SC IFX) samples were collected at routine clinic visits reflecting steady-state levels. Definitions and classifications Immunogenicity was defined as detection of ADA by routine TDM during a treatment course, irrespective of clinical consequence. Therapeutic trough thresholds were ≥3 mcg/mL for intravenous infliximab and ≥5 mcg/mL for adalimumab.13 14 Courses with subthreshold trough levels were classified as subtherapeutic. LOR was defined as clinical, biochemical or endoscopic deterioration following initial response. Among discontinued courses, LOR was classified as immunogenicity-mediated failure (ADA-positive), mechanistic failure (therapeutic levels, ADA-negative), non-immunogenicity-mediated pharmacokinetic (PK) failure (subtherapeutic levels, ADA-negative) or primary non-response. Rechallenge was defined as restarting the same anti-TNF after a break exceeding 3 months. Strict objective criteria were applied to classify LOR and primary non-response; patients not meeting these criteria were excluded from the analysis. LOR was defined based on objective evidence of active inflammatory disease: a faecal calprotectin elevation above 150 mcg/g with a rise compared with the prior reading or rise in CRP above upper limit of normal or worsening of endoscopic disease activity. Primary non-response was defined as no improvement in clinical or biochemical parameters during the postinduction period, with endoscopic assessment performed when this was equivocal or unclear. Any detectable ADA level above the assay lower limit of detection was classified as ADA-positive; no minimum titre threshold was applied. A subgroup of patients who received SC IFX was analysed separately, comprising new starters (SC IFX after IV induction) and switchers from established IV IFX. Propensity score matching (nearest-neighbour, 1:3, calliper 0.25) on baseline serum albumin and immunomodulator use was performed to compare SC and IV formulations. In a sensitivity analysis, the propensity score model was expanded to include five covariates: baseline serum albumin, immunomodulator use, age, previous surgery and disease subtype. Immunogenicity rates were reported separately for SC new starters and IV-to-SC switchers, with the switcher subgroup identified as subjected to survivorship bias. Outcomes The primary outcome was the course-level immunogenicity rate, reported with Wilson 95% CIs. Secondary outcomes included LOR classification, time to immunogenicity (Kaplan-Meier), clinical predictors (Cox and logistic regression), trough level dose–response relationships with receiver operating characteristic (ROC) analysis, patient-level immunogenicity including rechallenge outcomes and SC versus IV infliximab immunogenicity comparison. Statistical analysis All statistical analyses were performed using R V.4.4.2.15 Continuous variables were expressed as median with IQR and compared using the Wilcoxon rank-sum test. Categorical variables were compared using Fisher’s exact test or χ2 test. Proportions with 95% CIs were calculated using the Wilson method. Time-to-immunogenicity was analysed using the Kaplan-Meier method with log-rank tests for between-group comparisons. Cox proportional hazards regression identified course-level predictors of immunogenicity; results are reported as HRs with 95% CIs. Multivariable models included clinically relevant variables and those with p<0.1 on univariable analysis. Cox proportional hazards models incorporated a shared gamma frailty term (patient-level random effect) to account for within-patient clustering of treatment courses; the proportional hazards assumption was tested using Schoenfeld residuals. Logistic regression models used cluster-robust SEs (sandwich variance estimator, clustered by patient). A sensitivity analysis using clinically meaningful immunogenicity (excluding rescued patients who developed ADA but were successfully managed and remained on therapy) as the endpoint was performed to assess whether removing these patients sharpened the signal for clinically relevant predictors. ROC analysis identified optimal trough cut-offs using Youden’s index; AUC with 95% CI is reported. ROC analyses were restricted to intravenous infliximab trough levels and adalimumab steady-state levels; SC IFX was excluded as SC steady-state concentrations are not pharmacokinetically comparable to IV preinfusion troughs. A two-sided p<0.05 was considered statistically significant.16 Results Study population A total of 314 anti-TNF treatment courses from 248 unique patients with IBD were included in the primary analysis. The median age at treatment commencement was 33.0 years (IQR 24.0–39.0), with a male predominance (198/314 courses, 63.1%). The cohort was predominantly of Arab ethnicity (232/248 patients, 93.5%), comprising 202 Emirati and 30 non-Emirati Arab individuals, with South Asian (13, 5.2%), Caucasian (2, 0.8%) and other (1, 0.4%) ethnicity also represented (table 1). CD accounted for 190 patients (76.6%) and UC for 58 (23.4%). Infliximab comprised 212 courses (67.5%) and adalimumab 102 (32.5%). Two hundred and sixteen courses (68.8%) were first-line anti-TNF therapy and 98 (31.2%) were later-line. Of 248 patients, 185 (74.6%) received a single anti-TNF agent, 60 (24.2%) received two sequential anti-TNF courses with different agents and 3 (1.2%) received a third course representing rechallenge with a previously used agent. Concomitant immunomodulator use was documented in 168 of 314 courses (53.5%). Table 1Baseline characteristics of 314 treatment courses, stratified by anti-TNF agentCharacteristicOverall (n=314)Infliximab (n=212)Adalimumab (n=102)P valueDemographics Age, years, median (IQR)33.0 (24.0–39.0)32.0 (24.0–38.2)33.0 (26.0–41.0)0.269 Male sex, n (%)198 (63.1)132 (62.3)66 (64.7)0.768 Montreal age at diagnosis, n (%)0.477 A1 (≤16 years)5 (1.6)3 (1.4)2 (2.0) A2 (17–40 years)238 (75.8)165 (77.8)73 (71.6) A3 (>40 years)71 (22.6)44 (20.8)27 (26.5)Disease characteristics Disease subtype, n (%)0.024 Crohn’s disease249 (79.3)160 (75.5)89 (87.3) Ulcerative colitis65 (20.7)52 (24.5)13 (12.7) Disease duration, years, median (IQR)8.0 (5.0–14.0)8.0 (4.0–13.0)11.0 (6.2–16.0)0.002 Previous IBD-related surgery, n (%)67 (21.3)39 (18.4)28 (27.5)0.092 Perianal disease, n (%)*114 (36.3)75 (35.4)39 (38.2)0.049Clinical parameters at treatment start BMI, kg/m², median (IQR)23.2 (19.0–27.7)22.9 (19.1–27.8)23.6 (18.6–27.5)0.792 Serum albumin, g/L, median (IQR)33.0 (29.0–37.5)32.5 (29.0–37.0)34.0 (29.0–38.0)0.747 CRP, mg/L, median (IQR)10.9 (3.2–36.9)10.9 (3.9–33.0)11.7 (2.6–42.4)0.454 Smoking status, n (%)0.038 Non-smoker264 (84.1)186 (87.7)78 (76.5) Current smoker36 (11.5)19 (9.0)17 (16.7) Ex-smoker14 (4.5)7 (3.3)7 (6.9)Treatment characteristics Line of therapy, n (%)0.027 First-line216 (68.8)151 (71.2)65 (63.7) Second-line70 (22.3)38 (17.9)32 (31.4) Third-line or later28 (8.9)23 (10.8)5 (4.9) Concomitant immunomodulator, n (%)168 (53.5)115 (54.2)53 (52.0)0.795 IMM type (n=168), n (%)0.207 Azathioprine148 (88.1)103 (89.6)45 (84.9) Methotrexate11 (6.5)5 (4.3)6 (11.3) Mercaptopurine9 (5.4)7 (6.1)2 (3.8) Treatment duration, months, median (IQR)24.0 (8.2–48.0)21.0 (7.9–37.4)29.9 (11.7–58.1)0.006*Perianal disease classification applies to Crohn’s disease patients; 65 ulcerative colitis courses are recorded as N/A. Continuous variables are presented as median (IQR) for skewed distributions. Categorical variables are presented as n (%). P values compare infliximab vs adalimumab using Fisher’s exact test (2×2 tables), χ2 test (larger tables) or Wilcoxon rank-sum test (continuous variables).BMI, body mass index; CRP, C reactive protein; IBD, inflammatory bowel disease; IMM, immunomodulator; TNF, tumour necrosis factor. Baseline characteristics of 314 treatment courses, stratified by anti-TNF agent Perianal disease classification applies to Crohn’s disease patients; 65 ulcerative colitis courses are recorded as N/A. Continuous variables are presented as median (IQR) for skewed distributions. Categorical variables are presented as n (%). P values compare infliximab vs adalimumab using Fisher’s exact test (2×2 tables), χ2 test (larger tables) or Wilcoxon rank-sum test (continuous variables). BMI, body mass index; CRP, C reactive protein; IBD, inflammatory bowel disease; IMM, immunomodulator; TNF, tumour necrosis factor. LOR classification At the time of data extraction, 115 courses were actively receiving therapy without immunogenicity and 9 had developed immunogenicity but were successfully rescued and remained on therapy. Of the 190 discontinued courses, 29 were stopped for non-LOR reasons (adverse events, remission or patient choice) and were excluded from the mechanistic classification, leaving 161 classifiable courses. Among these, immunogenicity-mediated failure was the most common category (78/161, 48.4%), followed by mechanistic failure (58/161, 36.0%), non-immunogenicity-mediated PK failure (14/161, 8.7%) and primary non-response (11/161, 6.8%) (table 2, online supplemental figure 2). Median treatment duration differed significantly across categories (Kruskal-Wallis, p=0.003), with primary non-response having the shortest duration (6.7 months, IQR 3.1–8.0) and mechanistic failure the longest (25.0 months, IQR 14.0–58.1) (online supplemental figure 2). A participant flow diagram is presented in figure 1. Table 2Loss of response classification (n=161 discontinued courses)Mechanismn%Immunogenicity-mediated failure7848.4Mechanistic failure5836.0Non-immunogenicity-mediated PK failure148.7Primary non-response116.8LOR classification restricted to patients who discontinued treatment: those who discontinued with immunogenicity (n = 78), those with non-immunogenic loss of response (n = 72) and primary non-responders (n = 11).LOR, loss of response; PK, Pharmacokinetic. Loss of response classification (n=161 discontinued courses) LOR classification restricted to patients who discontinued treatment: those who discontinued with immunogenicity (n = 78), those with non-immunogenic loss of response (n = 72) and primary non-responders (n = 11). LOR, loss of response; PK, Pharmacokinetic. Participant flow diagram. Flow of 314 anti-TNF treatment courses (248 patients) through the study, showing active courses (n=124), discontinued courses (n=190) and loss of response classification (n=161). LOR, loss of response; TNF, tumour necrosis factor; PK, pharmacokinetic. Primary outcome—overall immunogenicity rate ADA were detected in 89 of 314 treatment courses, yielding an overall course-level immunogenicity rate of 28.3% (95% CI 23.4% to 33.7%) (table 3). This comprised 78 courses in which ADA detection led to drug discontinuation, 9 courses that were rescued and remained on therapy, and 2 courses among those classified as primary non-responders in whom ADA were incidentally detected. Among the nine rescued courses, seven (78%) had detectable drug levels at the time of ADA detection (median 3.0 mcg/mL), compared with 28 of 66 (42%) non-rescued immunogenic courses with available drug level data (median 0.0 mcg/mL), suggesting that rescue was feasible when antibody burden had not yet fully neutralised circulating drug. At the patient level, 75 of 248 patients (30.2%, 95% CI 24.9% to 36.2%) developed ADA at any point during follow-up. The median time to immunogenicity detection was 15.0 months (IQR 8.0–36.0), with similar durations observed for infliximab (median 15.0, IQR 8.5–36.0 months) and adalimumab (median 14.8, IQR 8.0–42.8 months). The overall Kaplan-Meier immunogenicity-free survival curve is presented in figure 2. Median immunogenicity-free survival was estimated at 86 months. Table 3Immunogenicity by clinical subgroupSubgroupNEvents, n (%)95% CIP valueOverall 314 89 (28.3) 23.4 to 33.7% Anti-TNF agent 0.425 Infliximab (intravenous)21257 (26.9)21.3 to 33.2% Adalimumab (subcutaneous)10232 (31.4)23.0 to 41.1%Disease subtype 0.537 Crohn’s disease24973 (29.3)24.0 to 35.3% Ulcerative colitis6516 (24.6)15.8 to 36.3%Concomitant immunomodulator 0.901 No14642 (28.8)22.0 to 36.6% Yes16847 (28.0)21.7 to 35.2%Serum albumin 0.018 Low (<35 g/L)13545 (33.3)25.9 to 41.6% Normal (≥35 g/L)10019 (19.0)12.5 to 27.8%Disease duration 0.020 0–2 years292 (6.9)1.9 to 22.0% 3–5 years7620 (26.3)17.7 to 37.2% >5 years20967 (32.1)26.1 to 38.7%Previous surgery 0.001 No24759 (23.9)19.0 to 29.6% Yes6730 (44.8)33.5 to 56.6%Intravenous IFX trough level* <0.001 ≥5.3 mcg/mL18636 (19.4)14.3 to 25.6% <5.3 mcg/mL6543 (66.2)54.0 to 76.5%*Cut-off 5.3 mcg/mL derived from ROC analysis (AUC 0.880). For adalimumab, cut-off 6.4 mcg/mL (AUC 0.861).AUC, area under the curve; IFX, infliximab; ROC, receiver operating characteristic; TNF, tumour necrosis factor. Immunogenicity by clinical subgroup Cut-off 5.3 mcg/mL derived from ROC analysis (AUC 0.880). For adalimumab, cut-off 6.4 mcg/mL (AUC 0.861). AUC, area under the curve; IFX, infliximab; ROC, receiver operating characteristic; TNF, tumour necrosis factor. Kaplan-Meier immunogenicity-free survival curves (A) Overall cohort. (B) Stratified by drug (infliximab vs adalimumab). (C) Stratified by immunomodulator cotherapy. Log-rank p values are shown. Shaded areas represent 95% CIs. IMM, immunomodulator; TNF, tumour necrosis factor. The immunogenicity rate was 26.9% (57/212, 95% CI 21.3% to 33.2%) for infliximab and 31.4% (32/102, 95% CI 23.0% to 41.1%) for adalimumab; this difference was not statistically significant (p=0.425). Immunogenicity rates did not differ between CD (73/249, 29.3%) and UC (16/65, 24.6%) (p=0.537), nor between first-line (61/216, 28.2%) and later-line courses (28/98, 28.6%) (p=1.000). Kaplan-Meier analysis confirmed no significant differences in immunogenicity-free survival by drug type (p=0.707), disease subtype (p=0.859) or treatment line (p=0.392). Predictors of immunogenicity Immunogenicity rates were assessed across baseline clinical characteristics (table 3). On univariable Cox regression, previous surgery (HR 1.82, 95% CI 1.17 to 2.83, p=0.008), lower serum albumin (HR 0.96 per g/L, 95% CI 0.93 to 1.00, p=0.024), higher CRP (HR 1.00 per mg/L, 95% CI 1.00 to 1.01, p=0.020) and younger age (HR 0.97 per year, 95% CI 0.95 to 1.00, p=0.016) were significantly associated with immunogenicity. Drug type, concomitant immunomodulator use, sex and BMI were not significant (online supplemental table 1, figure 1). On multivariable shared frailty Cox regression (n=211, 54 events), incorporating a patient-level random effect to account for within-patient clustering (frailty variance 0.40), previous surgery showed the strongest association with immunogenicity (HR 1.85, 95% CI 0.94 to 3.64, p=0.074), with CRP showing a trend (HR 1.00 per mg/L, 95% CI 1.00 to 1.01, p=0.079). Drug type, immunomodulator use, albumin and age were not independently significant. The proportional hazards assumption was met (global Schoenfeld test p=0.49). On multivariable logistic regression with cluster-robust SEs, previous surgery was independently associated with immunogenicity (adjusted OR (aOR) 2.23, 95% CI 1.06 to 4.71, p=0.035). Treatment duration was excluded from aetiologic predictor models owing to reverse causation (patients developing immunogenicity early necessarily have shorter treatment duration). Model concordance was 0.785 (online supplemental table 1). A sensitivity analysis using clinically meaningful immunogenicity as the endpoint was performed by excluding the nine rescued patients who developed ADA but were successfully managed and remained on therapy, yielding 78 of 314 (24.8%) courses that met this definition. On shared frailty Cox regression, previous surgery was more strongly associated with immunogenicity (HR 2.80, 95% CI 1.41 to 5.56, p=0.003), and concomitant immunomodulator use showed a trend towards protection (HR 0.52, 95% CI 0.26 to 1.02, p=0.057). Drug level analyses Pre-event trough levels in immunogenic courses were compared with the most recent trough level (within 6 months) in non-immunogenic courses (online supplemental table 2). Among 77 immunogenic courses with available pre-event data, the median pre-event drug level was 3.0 mcg/mL (IQR 2.0–4.2). By drug, the median was 3.0 mcg/mL (IQR 1.4–4.1) for infliximab trough levels (n=50) and 3.0 mcg/mL (IQR 2.4–5.0) for adalimumab steady-state levels (n=27). Among 174 non-immunogenic courses with drug levels recorded within 6 months, the median was 14.0 mcg/mL (IQR 8.7–21.0); infliximab 15.8 mcg/mL (IQR 8.9–21.5, n=128) and adalimumab 12.0 mcg/mL (IQR 8.2–17.4, n=46). The difference between pre-event and non-immunogenic drug levels was highly significant (p<0.001), and this held for each drug individually (both p<0.001) (figure 3). Among 64 courses discontinued following immunogenicity with a drug level recorded at the time of the event, 37 (57.8%) had an undetectable level and the median was at the lower limit of quantification (0.0 mcg/mL, IQR 0.0–0.5), consistent with near-complete drug neutralisation at the point of clinically apparent immunogenicity. Trough level dose–response and ROC analysis for immunogenicity prediction (A) Infliximab immunogenicity rate by trough level category. (B) Infliximab ROC curve (AUC 0.880). (C) Adalimumab immunogenicity rate by trough level category. (D) Adalimumab ROC curve (AUC 0.861). Trough level categories in mcg/mL. AUC, area under the curve; ROC, receiver operating characteristic. A clear inverse dose–response relationship was observed between drug level measured within 6 months before the event and immunogenicity for both infliximab and adalimumab (figure 3, online supplemental table 3). For infliximab, immunogenicity rates declined from 63.2% at trough levels <1 mcg/mL to 3.1% at levels >10 mcg/mL (n=178 courses). ROC analysis yielded an AUC of 0.880 (95% CI 0.824 to 0.936) with an optimal cut-off of 5.3 mcg/mL (Youden’s index), corresponding to a sensitivity of 94.1% and specificity of 85.0% (figure 3A,B). For adalimumab, immunogenicity rates declined from 66.7% at steady-state levels <2 mcg/mL to 7.7% at levels >12 mcg/mL (n=73 courses). ROC analysis yielded an AUC of 0.861 (95% CI 0.769 to 0.952) with an optimal cut-off of 6.4 mcg/mL, corresponding to a sensitivity of 92.9% and specificity of 80.0% (figure 3C,D). A sensitivity analysis comparing drug levels in actively treated non-immunogenic patients (median 17.0 mcg/mL, IQR 12.0–22.0, n=110) with those in patients who were subsequently discontinued for non-immunogenic reasons (median 11.0 mcg/mL, IQR 6.0–16.5, n=59) confirmed higher levels in patients who maintained therapy. These ROC-derived cut-offs are cohort-derived exploratory thresholds derived and tested in the same dataset; independent external validation is required before they can be applied in clinical practice. Patient-level immunogenicity and rescued courses At the patient level (n=248), 75 patients (30.2%, 95% CI 24.9% to 36.2%) developed ADA during at least one treatment course. Sixty-three patients (25.4%) received more than one anti-TNF course (60 received two courses and 3 received three courses). Among patients with multiple anti-TNF courses, 42 of 63 (66.7%, 95% CI 54.4% to 77.1%) developed immunogenicity during at least one course, compared with a substantially lower rate in single-course patients. The number of anti-TNF courses was the strongest patient-level predictor of ever developing immunogenicity (OR 7.95 per additional course, 95% CI 4.31 to 15.18, p<0.001), with previous surgery also significant (OR 2.46, 95% CI 1.27 to 4.73, p=0.007). Nineteen treatment courses were identified as rechallenges (restarting the same anti-TNF agent after a break exceeding 3 months), of which 12 of 19 (63.2%, 95% CI 41.0% to 80.9%) developed immunogenicity, confirming a high rate of recurrent ADA formation on re-exposure. Among 61 patients who received both infliximab and adalimumab sequentially, the rate of immunogenicity to the second anti-TNF was numerically higher in patients who were immunogenic to the first agent (13/32, 40.6%) compared with those who were not (8/29, 27.6%), although this association did not reach statistical significance (OR 1.78, 95% CI 0.54 to 6.15; p=0.42). An exploratory interaction analysis found no significant difference in the effect of immunomodulator cotreatment on immunogenicity by anti-TNF drug type (OR 1.13, 95% CI 0.40 to 3.19, p=0.817). Of the 87 courses with clinically relevant immunogenicity, 9 were successfully rescued with dose intensification and continued on the same anti-TNF agent. Among rescued patients, 7 of 9 (77.8%) were receiving concomitant immunomodulator therapy, and the median trough drug level following rescue was 3.0 mcg/mL (IQR 2.0–5.0). All nine rescued patients were managed with reinduction at higher doses (10 mg/kg or more), with intensified frequency where indicated, combined with the addition of an immunomodulator (thiopurine or methotrexate). Early proactive TDM was performed before the second or third reinduction dose to confirm effective salvage of drug levels and suppression of ADA. Immunogenicity analysis of subcutaneous versus intravenous infliximab A subgroup of 58 patients who received SC IFX was analysed, comprising 25 new starters (initiated on SC IFX 120 mg Q2W following IV induction) and 33 switchers (transitioned from established IV IFX). Fifty-seven of 58 (98.3%) had a corresponding IV IFX course in the primary dataset. Median follow-up was shorter for SC IFX (9.6 months, IQR 6.6–13.6) than IV IFX (21.0 months, IQR 7.9–37.4; p<0.001), reflecting the more recent introduction of SC IFX at our centre. Drug levels are reported descriptively and separately by route, as SC steady-state levels and IV preinfusion troughs reflect different pharmacokinetic parameters and cannot be directly compared; IV IFX trough levels were a median of 12.0 mcg/mL (IQR 3.1–18.8, n=178) and SC IFX steady-state levels a median of 25.0 mcg/mL (IQR 16.0–42.0, n=53). Among SC patients, steady-state levels were higher in switchers (median 36.5 mcg/mL, IQR 23.0–45.0) than new starters (median 16.0 mcg/mL, IQR 12.0–27.0), consistent with longer exposure and potentially higher drug accumulation. Immunogenicity was lower with SC IFX than IV IFX: 2/58 (3.4%, 95% CI 0.9% to 11.7%) versus 57/212 (26.9%, 95% CI 21.3% to 33.2%) for IV IFX (p<0.001; OR 0.10, 95% CI 0.01 to 0.39). Rates were similar in new starters (1/25, 4.0%, 95% CI 0.7% to 19.5%) and switchers (1/33, 3.0%, 95% CI 0.5% to 15.3%); however, the switcher subgroup is subjected to survivorship bias, as these patients had already tolerated IV IFX without developing clinically significant immunogenicity. To account for potential confounders, propensity score matching was performed using an enhanced model with five covariates (baseline serum albumin, immunomodulator use, age, previous surgery, disease subtype; nearest-neighbour, 1:3, calliper 0.25), yielding 52 SC and 109 IV patients with good covariate balance (mean albumin: SC 35.7 vs IV 34.2 g/L; immunomodulator use: SC 54% vs IV 52%; mean age: SC 30.7 vs IV 32.2 years; prior surgery: SC 10% vs IV 10%; CD subtype: SC 79% vs IV 75%). The immunogenicity difference remained significant: SC 3.8% versus IV 18.3% (Fisher exact p=0.013; OR 0.18, 95% CI 0.02 to 0.79). Median follow-up in the matched cohort was 9.6 months for SC versus 19.6 months for IV (p<0.001). In a sensitivity analysis restricted to SC new starters versus IV IFX, the unmatched comparison was significant (Fisher exact p=0.018; OR 0.13, 95% CI 0.00 to 0.85), but after propensity matching (24 SC vs 70 IV), statistical significance was not reached (p=0.062; OR 0.16, 95% CI 0.00 to 1.17), reflecting the limited sample size. These exploratory findings suggest that SC IFX may be associated with lower immunogenicity, but prospective validation is required. Discussion This study provides the first comprehensive characterisation of anti-TNF immunogenicity in a real-world Middle Eastern IBD cohort, addressing a critical evidence gap. The overall course-level immunogenicity rate of 28.3% is broadly consistent with Western data. The UK PANTS Extension study reported cumulative immunogenicity rates of 44% for infliximab and 20% for adalimumab at 3 years.11 Our comparable findings suggest that ADA formation is fundamentally pharmacokinetic rather than population-specific. The only other Gulf-region study, a cross-sectional Saudi cohort, provided point-prevalence data but lacked longitudinal analysis.17 Among classifiable discontinued courses, immunogenicity-mediated failure accounted for 48.4% of treatment discontinuations. The context-aware trough level analysis, comparing pre-event levels in immunogenic courses with the most recent TDM results in non-immunogenic courses, revealed a strong inverse dose–response relationship. This comparison is analytically asymmetric: pre-event levels in immunogenic courses were sampled at different clinical timepoints from the most recent monitoring levels in non-immunogenic courses, and because TDM was performed reactively during the maintenance phase rather than under a prespecified sampling schedule, informative sampling bias cannot be excluded. Our findings are nonetheless consistent with established TDM targets and international Delphi consensus.13 Meta-analytic and network meta-analytic data confirmed that proactive TDM reduces treatment failure and ranks highest for maintaining remission.18 19 These data reinforce the rationale for routine proactive drug-level monitoring, although our study did not formally compare proactive versus reactive TDM strategies.20 The therapeutic trough thresholds applied in our study (infliximab ≥3 mcg/mL, adalimumab ≥5 mcg/mL) were the consensus treat-to-target levels at the time our study cohort was established, based on a 2019 international Delphi consensus and the TAXIT trial.13 14 As this retrospective cohort spans several years, we applied the thresholds in clinical use during the study period. We acknowledge that the AGA guideline on TDM recommends higher target thresholds of ≥5 mcg/mL for infliximab and ≥7.5 mcg/mL for adalimumab, supported by dose–response data demonstrating improved outcomes at higher levels.21,26 The ROC-derived cut-offs for immunogenicity prediction reported here are analytically distinct from these treat-to-target thresholds and should be regarded as cohort-derived exploratory values requiring external validation; nonetheless, they identified similar ranges. Patients with low-titre antibodies and subtherapeutic levels may benefit from dose intensification and/or immunomodulator addition before being classified as treatment failures. Nine patients who developed immunogenicity were successfully rescued with intensified dosing (10 mg/kg or more) and concomitant immunomodulator addition, with early proactive TDM confirming restoration of therapeutic drug levels. The rescued patients were characterised by detectable drug levels at the time of ADA detection, consistent with low-titre antibodies and an early pharmacokinetic window amenable to dose intensification. This suggests that early detection of immunogenicity through proactive TDM may create a window of opportunity for intervention before antibodies become clinically irreversible. A sensitivity analysis excluding these nine patients yielded concordant but sharper results, confirming the robustness of the primary analysis. The relatively low rate of concomitant immunomodulator use (53.5%) reflects the evolution of practice at our centre over the study period. In earlier years, monotherapy was more commonly used, with immunomodulators added reactively when subtherapeutic drug levels or early immunogenicity signals were detected. Current practice has evolved to favour combination therapy from initiation, aligning with guideline recommendations. The lack of a significant immunomodulator-immunogenicity association in our data warrants careful interpretation, and it may partly reflect this heterogeneity in how immunomodulator exposure was classified, potentially diluting any protective signal from true concomitant use. Nevertheless, this contrasts with SONIC, SUCCESS and PANTS, in which combination therapy or immunomodulator cotreatment was protective.4 5 10 Several factors may explain this, including confounding by indication and variability in immunomodulator dosing and adherence. Mechanistically, immunomodulators may maintain higher trough levels; thiopurine metabolite concentrations directly correlate with infliximab trough levels.27 Concurrent immunomodulator use is similarly associated with higher adalimumab levels.28 Higher drug concentrations may themselves suppress the ADA response.29 The relationship may also be genotype-dependent through HLA-DQA1*05, although applicability to non-European populations remains uncertain.30,32 Importantly, proactive TDM revokes the DQA1*05 association, and immunomodulator therapy can promote reversal of established ADA-mediated LOR.32 33 These observations suggest that adequate drug exposure through TDM-guided optimisation may be more important than immunomodulator coprescription or genotype-guided prescribing alone. Prior IBD-related surgery was independently associated with immunogenicity in Cox regression, possibly reflecting greater systemic inflammatory burden. Longer disease duration was associated with increased risk, consistent with the highest ADA hazard occurring in the first year of treatment. Treatment duration was not included in the aetiologic predictor models owing to reverse causation (patients who develop immunogenicity early necessarily have shorter treatment exposure).11 Low baseline albumin was also independently associated with immunogenicity. Population pharmacokinetic studies show that serum albumin is inversely related to infliximab clearance, with low levels corresponding to up to fourfold increased clearance.34 35 Albumin and IgG-based monoclonal antibodies share a common neonatal Fc receptor (FcRn) salvage pathway; hypoalbuminaemia reflects reduced recycling and accelerated catabolism. Furthermore, inflamed mucosa permits direct enteric loss of therapeutic antibodies. Low albumin may thus serve as a surrogate for impaired FcRn salvage and protein-losing enteropathy, contributing to subtherapeutic exposure.35 The persistence of immunogenicity across treatment courses has important clinical implications. Same-agent rechallenge carried substantial risk of recurrent antibody formation, and immunogenicity to a second anti-TNF was numerically higher among patients immunogenic to the first, consistent with the IMSAT study.36 The lack of statistical significance in our cross-anti-TNF analysis likely reflects limited power. These observations suggest immunogenicity may reflect an inherent patient-level propensity to mount ADA responses, reinforcing the rationale for proactive TDM when switching agents. In our exploratory comparison, SC IFX was associated with lower immunogenicity rates than intravenous infliximab (new starters 4.0%, IV-to-SC switchers 3.0%, vs IV 26.9%). SC CT-P13 produces higher, more stable drug concentrations with lower peak-to-trough variability, and real-world switch cohorts confirm threefold to fourfold trough increases.37 38 The median SC level in our cohort (25.0 mcg/mL) was well above published targets. The LIBERTY programme and REMSWITCH study confirmed SC infliximab efficacy regardless of immunosuppressant use.39,41 However, this comparison should be interpreted cautiously. First, immunogenicity rates were similarly low in SC new starters (1/25, 4.0%) and in IV-to-SC switchers (1/33, 3.0%), but the switcher subgroup is subjected to substantial survivorship bias: these patients had, by definition, already tolerated IV infliximab without developing clinically significant immunogenicity, and therefore represent a favourably selected subgroup. Second, although immunogenicity remained significantly lower with SC after enhanced propensity score matching on five covariates (3.8% vs 18.3%, p=0.013), a sensitivity analysis restricted to SC new starters versus propensity-matched IV controls did not reach statistical significance (p=0.062), reflecting the limited sample size. Third, median follow-up was substantially shorter in the SC group than in the IV group (9.6 vs 21.0 months), which is particularly relevant for an outcome that accumulates over time. These findings should, therefore, be regarded as exploratory and hypothesis-generating, and prospective studies with longer follow-up and head-to-head comparisons in treatment-naive patients will be required before SC infliximab can be confidently positioned as preferable for reducing immunogenicity. This study has several strengths. It represents the largest anti-TNF immunogenicity cohort from the Middle East, providing novel data from an ethnically diverse, under-represented population. The comprehensive TDM dataset enabled context-aware trough level analysis distinguishing between routine monitoring levels and pre-event levels, more accurately reflecting clinical scenarios. The propensity-matched SC versus IV infliximab comparison adds to a growing evidence base on formulation-related immunogenicity differences, although it is exploratory. The LOR classification moves beyond a binary framework to allow granular understanding of treatment failure mechanisms. Limitations include the retrospective single-centre design. ADA testing used a drug-tolerant platform that mitigates but does not eliminate the risk of underestimating low-level immunogenicity in the presence of circulating drug. The SC IFX subgroup was relatively small, with substantially shorter follow-up than the IV group. TDM was performed at clinical discretion rather than at protocol-mandated time points, introducing potential informative sampling bias. HLA-DQA1*05 genotyping was not performed. Our findings reflect practice from a tertiary IBD centre with established TDM expertise, and TDM awareness and utilisation vary across the wider Middle Eastern region.42 A direct comparison of proactive versus reactive TDM strategies was not possible. The multivariable regression models used complete-case analysis (n=211 of 314 courses), with 103 courses excluded due to missing baseline albumin or CRP data, which may introduce selection bias; missing data rates for key variables are reported in online supplemental table 4. In conclusion, this study provides the first comprehensive characterisation of anti-TNF immunogenicity in a Middle Eastern IBD cohort, demonstrating patterns broadly consistent with Western data and confirming that immunogenicity is driven by pharmacokinetic rather than population-specific determinants. The findings are consistent with a potential benefit of proactive TDM and support avoidance of same-agent rechallenge and closer surveillance of patients with prior surgery or hypoalbuminaemia. The lower immunogenicity observed with SC IFX is encouraging but requires confirmation in prospective studies with longer follow-up. Concomitant immunomodulator use did not show a clear protective effect in this cohort, possibly reflecting heterogeneity in the timing and duration of coprescription. These findings should be interpreted within the limitations of the retrospective design, and future studies with protocolised TDM sampling will help validate the exploratory thresholds identified here. Supplementary material Footnotes Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Ethics approval: This study was approved by the Sheikh Shakhbout Medical City Research Ethics Committee (SSMC-REC-585). Patient consent was waived due to the retrospective nature of the study. Data availability free text: The data that support the findings of this study are available from the corresponding author upon reasonable request. The data are not publicly available due to their sensitive nature, containing information that could compromise patient confidentiality and privacy. Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information. References Associated Data Supplementary Materials Data Availability Statement All data relevant to the study are included in the article or uploaded as supplementary information.