Brazikumab worked against IL-23 in Crohn's disease, then lost a portfolio decision

What was tried

Brazikumab is a human monoclonal antibody that binds the p19 subunit of interleukin-23, the subunit encoded by IL23A, and blocks IL-23 signaling without touching IL-12 (Verstockt et al., Nature Reviews Gastroenterology and Hepatology, 2023). AstraZeneca ran NCT03759288 as an operationally seamless Phase 2b/3 study in moderate to severe Crohn's disease. Patients were randomized to low-dose brazikumab, high-dose brazikumab, the anti-TNF antibody adalimumab as an active comparator, or placebo, with CDAI remission at week 12 as the primary endpoint (ClinicalTrials.gov, NCT03759288). The study started in December 2018 and was terminated with an actual primary completion date of 18 October 2023. The posted reason was a strategic decision to discontinue the development of brazikumab in inflammatory bowel disease (NCT03759288). Total enrollment reached 89 patients, split across the four arms as 35, 31, 2, and 18 (NCT03759288 participant flow).

The biological hypothesis

IL-23 sits upstream of the IL-17 and IL-22 effector axis that sustains chronic intestinal inflammation, and human genetics place the pathway near the center of Crohn's disease. Coding variants in IL23R, the receptor for IL-23, are among the strongest protective and risk signals in inflammatory bowel disease genetics, which is why p19 blockade became a validated strategy rather than a guess. In Open Targets, IL23A associates with Crohn's disease at an overall score of 0.62, carried largely by clinical evidence at 0.99 and animal model evidence at 0.40 (Open Targets Platform, IL23A, ENSG00000110944). The therapeutic logic was direct. Neutralize IL-23 through its unique subunit, spare IL-12, and reduce downstream effector cytokine output in the gut.

What actually happened

Brazikumab had already produced positive controlled efficacy data before this trial ran. In the Phase 2a MEDI2070 study (NCT01714726) in 119 anti-TNF-failed Crohn's patients, clinical response at week 8 reached 49.2 percent on brazikumab versus 26.7 percent on placebo, an absolute difference of 22.5 percent with a 95 percent confidence interval of 5.6 to 39.5 and p = 0.010 (Sands et al., Gastroenterology, 2017). The confirmatory Phase 2b/3 then stopped before it could read out. The active-comparator adalimumab arm enrolled 2 patients and the placebo arm enrolled 18, numbers far below any prespecified analysis population (NCT03759288). The registry lists results as posted, but no interpretable between-arm efficacy estimate exists, because the trial was halted during early enrollment. The end came as a portfolio decision, not a signal of failed biology or unacceptable safety.

Failure mechanism, best guess

The failure here was commercial and strategic rather than mechanistic. By the time the confirmatory trial was enrolling, the IL-23 p19 class had already crowded. Risankizumab, another p19 inhibitor, met all co-primary endpoints in two Phase 3 Crohn's induction trials, with CDAI clinical remission of 45 percent versus 25 percent on placebo in the ADVANCE study (D'Haens et al., Lancet, 2022), and went on to approval. Guselkumab and mirikizumab advanced on parallel tracks (Verstockt et al., 2023). A late entrant facing approved or near-approved competitors with the same mechanism has a narrow path to differentiation and pricing power. The most likely driver of discontinuation is that brazikumab sat behind in a saturated class, and the expected commercial return no longer justified the cost of a global Phase 3 against an active comparator. The biology worked. The timing did not.

How to prevent this next time

Two quantitative tools would have surfaced this risk before the seamless Phase 2b/3 was launched. The first is a competitive landscape red-team that scores time-to-market against named competitors. With risankizumab reporting Phase 3 Crohn's induction data across 2020 to 2022 while brazikumab was still enrolling its confirmatory study, a lead-time model would have shown brazikumab reaching the market years behind a same-mechanism rival, which compresses net present value regardless of trial success. The second is a Bayesian gating analysis run alongside the clinical one. The predictive probability of program success is

PP(success)=∫θ​P(final success∣θ,interim data)⋅π(θ∣interim data)dθ

where the value-weighted version multiplies that probability by a market-share term that decays with each competitor that launches first. A program with high clinical predictive probability but low share-adjusted value is exactly the configuration that gets stopped mid-trial, and pricing that decay at the design stage would have set an explicit go or no-go threshold before 89 patients were enrolled and an active-comparator arm was opened. A third lever, historical base-rate adjustment, would have compared brazikumab's launch position against prior late entrants in crowded biologic classes, where uptake runs consistently low. The single highest leverage change would have been gating the confirmatory Phase 2b/3 on a prespecified share-adjusted value threshold, so the program either committed fully or stopped before it consumed a global trial's worth of enrollment.

What this means for similar programs

Brazikumab is a case where a genetically validated mechanism and real efficacy were not enough. The IL-23 field shows that being right about biology is necessary but not sufficient when several companies are right at the same time. For any program entering a class with approved or imminent competitors, the decisive question shifts from whether the drug works to whether it can arrive in time to matter. Seamless late-stage designs are expensive commitments, and they belong with assets that hold a defensible position, not assets racing to catch a field that has already moved.

Open questions

Will AstraZeneca or a licensee release the partial INTREPID data, including any pharmacodynamic or biomarker readouts from the patients who were dosed. Could a biomarker-enriched approach, for example selecting patients by baseline IL-22 as the MEDI2070 analysis suggested, have created enough differentiation to justify continuation. And how many other validated mechanisms are being shelved not because they fail, but because they finish second in a race.

Sources

1. • ClinicalTrials.gov, NCT03759288, full record and participant flow (enrollment 89, arms 35, 31, 2 and 18, terminated, primary completion 18 October 2023). https://clinicaltrials.gov/study/NCT03759288 - Sands BE, et al. Efficacy and Safety of MEDI2070, an Antibody Against Interleukin 23, in Patients With Moderate to Severe Crohn's Disease: A Phase 2a Study. Gastroenterology, 2017 (week 8 response 49.2 percent vs 26.7 percent, difference 22.5 percent, p = 0.010). https://- D'Haens G, et al. Risankizumab as induction therapy for Crohn's disease: ADVANCE and MOTIVATE. Lancet, 2022 (CDAI remission 45 percent vs 25 percent in ADVANCE). https://- Verstockt B, et al. IL-12 and IL-23 pathway inhibition in inflammatory bowel disease. Nature Reviews Gastroenterology and Hepatology, 2023 (p19 inhibitor landscape). https://- Deepak P, Sandborn WJ. Ustekinumab and Anti-Interleukin-23 Agents in Crohn's Disease. Gastroenterology Clinics of North America, 2017. https://- Open Targets Platform, IL23A (ENSG00000110944), Crohn's disease association 0.62. https://platform.opentargets.org/target/ENSG00000110944 - ClinicalTrials.gov, NCT01714726, Phase 2a MEDI2070 source record. https://clinicaltrials.gov/study/NCT01714726.