STOCKHOLM — Predictors of primary nonresponse and loss of response to anti-tumor necrosis factor (anti-TNF) therapy in inflammatory bowel disease (IBD), including demographic, biochemical, and treatment-related factors, have been identified by a large, population-based registry study from Denmark.

The findings support the recent shift from a bottom-up to a top-down approach in treating IBD, said Mads Lie, MD, PhD fellow, from the Center for Molecular Prediction of Inflammatory Bowel Disease (PREDICT), Aalborg University Hospital and University of Copenhagen, Denmark, who presented the study here at this year’s European Crohn’s and Colitis Organisation (ECCO) Congress 2026. 

“We established a registry-based phenotype for primary nonresponse and loss of response to anti-TNF therapy in IBD. This approach allowed us to analyze a cohort far larger than any clinical trial , reinforcing many previously reported findings while also identifying novel associations that require further validation,” said Lie. 

“While this means we cannot fully predict non response before treatment initiation, it does suggest that early identification during induction may be possible, potentially allowing clinicians to adjust treatment strategy before irreversible treatment failure occurs,” he added. 

Moving Beyond Trial Subset Populations

Anti-TNF agents were introduced about 25 years ago and have dramatically reduced symptoms, hospitalizations, corticosteroid use, and surgery for many patients with IBD; however, about 20% of patients experience primary nonresponse, and a further 40%-50% eventually develop loss of response, Lie noted. 

Earlier identification of nonresponders would help determine who is most likely to benefit from the anti-TNF therapy, he said. “Biologic therapies have transformed IBD care, but the reality is that the response is far from universal. A substantial proportion of patients are exposed to ineffective treatment, risks, and costs before we know it hasn’t worked.”

Most previous efforts to predict treatment failure have relied on clinical trials, typically enrolling a few hundred participants. “Those studies are invaluable, but they inevitably represent a selected subset of patients,” said Lie. 

In the current study, the researchers drew on Denmark’s national health registries, which offered an opportunity to examine treatment response across an entire population. “Denmark has been described as one large patient cohort, with extensive national health registries capturing detailed clinical information,” Lie explained.

Using linked registry data, the investigators aimed to establish registry-based phenotypes for primary nonresponse and time to secondary loss of response, and to analyze demographic, temporal, pharmaceutical, biochemical, and genetic associations with treatment outcomes.

Defining Response Using Registry Data

One of the key challenges was defining clinical outcomes without direct access to physician assessments. To address this, the investigators used proxy definitions closely aligned with those used in clinical trials.

Primary nonresponse was defined as predefined clinical events occurring between weeks 10 and 18 after treatment initiation, including need for systemic corticosteroids, hospitalization longer than 2 days, major IBD surgery, lack of C-reactive protein (CRP) reduction, treatment intensification, or switching to another biologic. Loss of response was defined using similar criteria occurring later in the treatment course.

Using these definitions, the team extracted data on blood tests, medication use, disease severity, IBD subtype, and demographics for each patient. Genetic data were available for a subset of patients.

The analysis included more than 11,000 patients with IBD — roughly evenly split between ulcerative colitis (UC) and Crohn’s disease — who were treated with infliximab as their first biologic therapy. 

Overall, 19.7% experienced primary nonresponse , 45.7% developed loss of response , and 34.6% maintained response. “These figures closely mirror what we see in clinical practice,” Lie noted.

Differences Between UC and Crohn’s

Regarding primary nonresponse a cross both diseases, Lie reported that concomitant immunomodulator therapy at the time of anti-TNF initiation was protective , consistent with findings from randomized clinical trials. 

Several predictors of primary nonresponse emerged, with important differences between UC and Crohn’s. In patients with UC, no baseline biochemical markers were significantly associated with subsequent primary nonresponse. In contrast, among patients with Crohn’s, multiple baseline laboratory measures were linked to primary nonresponse, including CRP, hemoglobin, leukocyte count, albumin, and — more unexpectedly — baseline sodium levels.

Using demographic data and routinely available biochemical markers, the researchers trained a prediction model with moderate accuracy for predicting primary nonresponse and found that predictive performance improved when laboratory values obtained during the early induction period (weeks 1-10) were included (UC area under the curve [AUC]: 0.72; Crohn’s AUC, 0.74).

They also examined predictors of loss of response and found a higher risk for loss of response associated with longer disease duration before initiation of biologic therapy, which supports an earlier, top-down treatment strategy, Lie said.

In Crohn’s, male sex was associated with a lower risk for loss of response, a finding previously reported in clinical trials. 

The investigators found no genetic variants that meaningfully predicted primary nonresponse or loss of response, added Lie. 

Distinguishing Mechanisms for Treatment Failure

Commenting on the findings, Konstantinos Papamichail, MD, PhD, physician-scientist and director of the Inflammatory Bowel Disease Center at Beth Israel Deaconess Medical Center in Boston, said the study adds to a growing body of work exploring biomarkers and precision medicine in IBD, but highlighted an important limitation.

“To move precision medicine forward, it’s critical to distinguish between truly refractory disease, so primary nonresponse or loss of response and treatment failure driven by pharmacokinetic issues vs mechanistic failure [of the disease],” Papamichail said. “Around 70% of treatment failure with anti-TNF agents is related to pharmacokinetics, such as underdosing, rather than lack of mechanistic efficacy.”

Without therapeutic drug monitoring (TDM) data, he added, it is difficult to determine which patients fall into each category. “They probably didn’t have sufficient TDM data to do that analysis here,” he said.

Still, Papamichail viewed the work as a valuable incremental step. “There are many studies in the biomarker field, and it’s important to build cumulative data,” he said. “The challenge is that, so far, there is no single biomarker that can reliably be used in routine clinical practice.”

Ultimately, he said, meaningful progress will require integrating multiple data streams. 

“We need multiomics, so proteomics, genomics, microbiota, alongside therapeutic drug monitoring,” he explained. “Treatment failure isn’t driven by one mechanism alone.”

Lie and Papamichael reported no relevant financial relationships.