Clinical Characteristics and Predictors of Biologic Failure in Inflammatory Bowel Disease: A Retrospective, Single-Center Study from Kuwait

This retrospective, single-center study provides novel insights into the clinical characteristics and predictors of biologic failure in a cohort of IBD patients from Kuwait. Our principal findings indicate that CD, longer disease duration, and Middle Eastern ethnicity were significantly associated with an increased risk of biologic failure. Furthermore, specific disease phenotypes, including ileocolonic (L3) CD and extensive ulcerative colitis (E3), were linked to higher rates of biologic failure, while complex CD behaviors (stricturing and penetrating) were strongly associated with the need for surgical intervention. These findings contribute crucial, region-specific data to the growing body of evidence on IBD treatment outcomes.

The landscape of IBD is rapidly evolving in the Middle East, necessitating a deeper understanding of treatment response in this unique patient population [3,4]. Our study identified several key predictors of biologic failure, some of which align with and others that diverge from the existing international literature. A primary finding was that patients with CD were significantly more likely to experience biologic failure than those with UC. This contrasts with a regional study from Saudi Arabia, which reported higher odds of anti-TNF failure in UC patients [17]. This discrepancy may reflect differences in cohort characteristics, prescribing patterns, our data showed a preference for anti-TNFs and ustekinumab in CD, or variations in the underlying disease biology between populations. However, our finding is consistent with several comparative-effectiveness datasets, particularly for vedolizumab. In a large U.S. cohort of older adults with IBD, vedolizumab was associated with a higher 1-year risk of treatment failure than anti-TNF therapy, and the excess risk was more pronounced in CD on subgroup analysis, supporting a CD>UC failure gradient for this agent [18]. Complementary evidence shows that in Crohn’s disease, ustekinumab outperforms vedolizumab across multiple effectiveness endpoints and persistence, indirectly indicating higher failure on vedolizumab in CD relative to alternatives; this contrast is less evident in UC where vedolizumab often performs comparatively well [19–21]. The reasonable biological explanation is that vedolizumab’s gut-selective α4β7-integrin blockade primarily limits lymphocyte trafficking to the mucosa, aligning better with UC’s superficial, continuous colitis, whereas CD’s transmural, patchy, and often fibrostenotic or penetrating phenotype engages deeper compartments and fibrotic pathways that are less responsive to trafficking blockade and slower-onset agents, mechanistic features that can translate to higher primary non-response or earlier loss of response in CD [22].

Our findings showed a significant association between longer disease duration (>5 years) and biologic failure. This likely reflects a more established and complex inflammatory process, a higher cumulative inflammatory burden, and potentially the development of fibrotic changes that are less responsive to anti-inflammatory agents [23], highlighting the importance of early and effective intervention to alter the natural history of the disease. Prior work on disease duration and biologic outcomes in IBD has been mixed. Abdelwahab et al. reported that longer duration in Egyptian patients with IBD was independently associated with a 2.5-fold higher risk of biologic failure risk of biological failure [24]. On the other hand, in post-hoc analyses of the ULTRA trials of adalimumab, Sandborn et al. found no difference in clinical remission between shorter (≤2 years) and longer (>2 years) duration, while a subsequent analysis suggested a lower colectomy risk with longer duration in adalimumab-treated UC [25,26]. In a retrospective cohort of infliximab-treated corticosteroid-dependent/refractory UC, Murthy et al. reported lower infliximab failure and colectomy with longer duration, though 35% had hospitalized acute severe UC, inherently at higher colectomy risk [27]. Conversely, Ma et al. showed that early anti-TNF use was not linked to higher surgery, hospitalization, or secondary loss of response despite greater baseline endoscopic severity [28]. For vedolizumab, the VICTORY consortium observed no association between disease duration and achieving clinical remission, steroid-free remission, or mucosal healing [29], whereas a validated prediction model from GEMINI 1 indicated that duration ≥2 years independently predicted steroidfree clinical remission by week 26 [30]. This apparent conflict likely stems from heterogeneity across studies, differences in biologic class and line of therapy, definitions and time horizons for “failure,” inclusion of incident vs prevalent users and acute severe UC, baseline disease severity and phenotype mix (CD vs UC), and varied adjustment for confounding by indication, all of which can shift the observed effect of disease duration.

We observed an exceptionally elevated rate of current smoking (41.75%) in our cohort, markedly higher than figures commonly reported from Western cohorts. Contrary to most studies where smoking is a well-established risk factor for biologic failure in CD [31,32], we did not find a statistically significant association. This could be due to our study being underpowered to detect such an effect, or its influence may have been overshadowed by other dominant factors like disease phenotype and duration. The inverse association we observed between smoking and surgical intervention is also perplexing and warrants further investigation, as it may represent a spurious correlation or unmeasured confounding.

Our analysis revealed that Middle Eastern ethnicity was a significant predictor of biologic failure. This novel finding points towards the potential influence of distinct genetic, environmental, or lifestyle factors within this population. For instance, certain genetic variants, such as the HLA-DQA1*05 allele, which are associated with immunogenicity to anti-TNF agents, have been reported to be more prevalent in some Middle Eastern populations [33]. This highlights the need for pharmacogenomic studies in the region to guide more personalized treatment strategies.

The phenotypic analysis provided further clarity. The association of ileocolonic (L3) CD with both biologic failure and surgery aligns with its characterization as a more extensive and often aggressive disease form [34]. Similarly, the link between stricturing (B2) and penetrating (B3) behaviors and a high likelihood of surgery is a classic finding, confirming that despite biologic therapy, disease complications often necessitate surgical management [23]. For UC, the finding that extensive colitis (E3) was most associated with biologic failure is supported by data from the IBD-ME registry, which also identified E3 as a key characteristic in patients requiring biologics [35]. This suggests that a greater disease burden is a consistent predictor of treatment challenges across different IBD subtypes.

Our observation that concomitant 5-ASA use was independently protective against biologic failure (aOR 0.06; 95% CI 0.02–0.22) contrasts with much of the contemporary evidence, which generally finds no added benefit to continuing mesalamine once advanced therapy is initiated. In UC, two large database studies showed that stopping 5-ASA after starting anti-TNF did not worsen clinical outcomes, arguing against a pharmacologic synergy with biologics [36,37]. Similarly, concomitant 5-ASA did not improve clinical or endoscopic outcomes with vedolizumab, and an RCT found no advantage to adding mesalamine to systemic therapy during induction [38]. Guidelines also do not recommend 5-ASA for Crohn’s disease and provide limited rationale for its continuation alongside biologics in moderate–severe UC [39]. Thus, the protective association in our cohort likely reflects confounding by indication and disease severity (e.g., 5-ASA preferentially used in milder, predominantly colonic UC), channeling/survivor bias (patients who tolerate and remain on 5-ASA may have inherently more stable disease), and unmeasured treatment behaviors (adherence, earlier healthcare contact). While hypothesis-generating, this signal should be interpreted cautiously and validated in designs that account for time-varying exposure and confounding (e.g., marginal structural models) before informing practice.

Clinical Implication

The findings of this study have direct and practical implications for clinicians managing IBD in Kuwait and similar Gulf regions. The identification of CD, longer disease duration, and extensive phenotypes (L3 CD, E3 UC) as key risk factors allows for the early stratification of patients. Individuals presenting with this high-risk profile may benefit from more aggressive initial therapy, closer monitoring, and a lower threshold for treatment optimization. In a high-risk patient, clinicians might consider earlier initiation of biologics or selecting agents with a lower risk of immunogenicity. The data suggests that a ‘watch-and-wait’ approach in patients with extensive disease may lead to a higher probability of future treatment failure. Although not directly measured in our study, the high rate of biologic failure (24.9%) underscores the potential utility of TDM. In patients with risk factors for failure, proactive or reactive TDM can help differentiate between mechanistic failure and pharmacokinetic issues (e.g., low drug levels, anti-drug antibodies), thereby guiding more rational treatment decisions, such as dose escalation or switching to a new agent [40]. The high prevalence of smoking in our cohort, even if not statistically linked to failure in this analysis, remains a critical modifiable risk factor. Clinicians must continue to emphasize smoking cessation as a cornerstone of IBD management, particularly in CD.

Future Direction

This study opens several avenues for future research to address the remaining knowledge gaps. A large-scale, prospective, multicenter registry across Kuwait and the GCC region is imperative to validate our findings, mitigate the limitations of retrospective analysis, and create a more comprehensive picture of IBD outcomes. Future studies should integrate pharmacokinetic and pharmacogenomic analyses. Measuring drug trough levels, anti-drug antibodies, and genetic markers (e.g., HLA-DQA1*05) would provide crucial mechanistic insights into why patients of Middle Eastern ethnicity may be at higher risk for biologic failure. Leveraging multi-omics approaches (genomics, proteomics, microbiomics) can help identify novel, non-invasive biomarkers for predicting treatment response [41]. This aligns with the global push towards precision medicine and could lead to predictive algorithms tailored for the regional population. Future research should clearly differentiate between primary non-response and secondary loss of response. Understanding the distinct predictors for each type of failure is essential for developing targeted management strategies [42,43]. The integration of AI and machine learning models could help synthesize complex clinical, endoscopic, and biomarker data to create robust predictive tools for clinical practice, moving beyond traditional statistical analysis [44].

Limitations

We acknowledge several limitations inherent to this study’s design. First, its retrospective and single-center nature limits the generalizability of our findings. The practices and patient population at our hospital may not be representative of all IBD patients in Kuwait. Second, our study was under the target sample size, which, while sufficient for the primary model’s estimates, likely reduced power for subgroup analyses (e.g., the effect of smoking on biologic failure). Third, our reliance on electronic health records is subject to information bias and missing data. We could not consistently differentiate between primary and secondary biologic failure, nor could we capture granular data on disease activity indices or inflammatory biomarkers like fecal calprotectin at the time of failure. The counterintuitive finding regarding comorbidities may also be a result of under-reporting in the medical records. Finally, the observational design is susceptible to confounding by indication, and while we identified strong associations, we cannot establish definitive causality. These limitations underscore that our findings should be considered hypothesis-generating and require validation in prospective studies.