AZURE-1: antiviral response, liver injury and the limits of a 24-week endpoint

September 28, 2026 / A separate completed readout

Mirum announced that both brelovitug dose arms met the Phase 3 AZURE-1 primary endpoint at Week 24. This History event records the actual clinical disclosure, including the separate 48-week update from the initial Phase 2b cohort. It does not replace the pending AZURE-4 event or the AZURE-2/3 program. A planned first-half 2027 BLA is not a submitted or accepted application, and the results are not a marketing approval. [1] [4]

The evidence is stronger than an uncontrolled viral-load reduction because the pivotal comparison was randomized against delayed treatment. However, the endpoint combines laboratory measures rather than cirrhosis progression, liver transplantation, cancer or survival. A meaningful effect on viral replication and liver inflammation supports further evaluation; it does not quantify the eventual reduction in those clinical outcomes.

Biology / HBsAg-dependent viral disease

Hepatitis delta virus depends on hepatitis B surface antigen for its viral envelope. Brelovitug is a fully human IgG1 antibody targeting HBsAg, intended to neutralize and remove HBV/HDV virions and HBsAg-containing subviral particles. The clinical hypothesis is that reducing circulating infectious material can suppress HDV and decrease hepatocyte injury. ALT normalization provides a complementary injury marker, but a normal ALT value does not establish viral eradication or reversal of fibrosis. [1] [2] [3]

The program includes participants with advanced but compensated liver disease. The registry excludes clinical decompensation, including ascites, encephalopathy or variceal hemorrhage, and other specified liver diseases. Results should not be generalized to those excluded conditions. The statement that some participants had cirrhosis or portal hypertension is not evidence of efficacy or safety in decompensated cirrhosis.

Molecule / Dose and PK–PD interpretation

The two subcutaneous regimens are 300 mg once weekly and 900 mg every four weeks. They differ in dose intensity as well as interval; the trial is not a simple comparison of identical exposure delivered at different frequencies. The current releases do not provide concentration-time curves, numerical half-life, receptor occupancy or a validated exposure-response model. It would therefore be inappropriate to infer equivalent trough concentrations or a preferred regimen from the response percentages alone. [1] [2]

Biological activity must be distinguished from complete viral suppression. A decrease of at least 2 log10 HDV RNA can satisfy virologic response without reaching an undetectable level. Results below the 10 IU/mL quantification limit and target-not-detected results are therefore shown separately. Persistent response during continued dosing also does not establish durable control after treatment withdrawal.

April 27 version / Phase 2b randomized evidence

The initial cohort randomized 53 participants: 21 to weekly dosing, 20 to four-weekly dosing and 12 to delayed treatment. The efficacy analysis required at least one post-baseline assessment and used 20/20/12, while safety used 21/20/12. The later disclosure explains that one weekly recipient stopped after the first dose without on-treatment results. The earlier 778-001 Phase 2 study was cited as showing antiviral activity at Week 48, but its complete cohort-specific numerical results were not recovered here and are not treated as an additional randomized replication. [2] [6]

The April results are preserved below as their own version. Reported p-values used a stratified Cochran–Mantel–Haenszel test against delayed treatment. Full multiplicity allocation, confidence intervals and the complete SAP were not available in the extracted materials. The smaller cohort provided an early signal; the 153-person Phase 3 analysis below is a separate cohort and should not be pooled with it without a prespecified plan.

Phase 3 / Design, populations and estimand limits

AZURE-1 (NCT06907290) is an open-label, randomized, multicenter Phase 2b/3 study. The Phase 3 portion enrolled 153 treatment-naïve participants with 2:2:1 allocation. Adults required chronic HDV, screening HDV RNA >500 IU/mL, abnormal ALT and willingness to take or existing use of HBV nucleos(t)ide therapy. That background therapy is relevant: brelovitug was not studied as a substitute for all HBV care. Detailed rescue rules and background-treatment balance were not extracted. [1] [3] [4]

The Phase 3 full analysis set contains randomized active recipients who received at least one dose, plus randomized delayed-treatment recipients: 59/65/29. Delayed participants begin weekly brelovitug after Week 24, limiting longer-term untreated comparisons. The March registry snapshot lists estimated enrollment 150 and estimated completion in 2029; neither is the actual September disclosure date or the full program enrollment. Sponsor reports separately identify 53 initial and 153 pivotal participants.

The primary composite requires virologic response and ALT normalization at Week 24. The pivotal p-values use a stratum-adjusted Miettinen–Nurminen test. The releases do not establish the full estimand, handling of every intercurrent event, exact responder numerators, adjusted difference confidence intervals or all multiplicity rules. Rounded percentages are kept as reported rather than reverse-engineering patient counts. The open-label design is especially relevant to subjective adverse-event reporting even though the primary measures are laboratory-based.

September 28 / Phase 3 Week 24 efficacy

The combined endpoint occurred in 56% and 45% of the two active groups versus 0% with delayed treatment, both p<0.0001. The descriptive differences from the rounded source rates are 56 and 45 percentage points; these are not substitutes for adjusted estimates and confidence intervals. There is no reported head-to-head test establishing that weekly dosing is superior to four-weekly dosing. Secondary response rates should not be promoted to independently confirmed claims without the hierarchy. [1] [4]

September update / Phase 2b extension and source discrepancies

At Week 48, combined response reached 55% in both earlier active groups. This is follow-up of the same initial cohort, not a third independent replication. One weekly participant with Week 24 virologic response moved away and was missing at Week 48, counted as a nonresponder. One four-weekly participant responded at Week 12 but was subsequently lost and counted as a nonresponder at Weeks 24 and 48. These disclosed rules should not be assumed to describe every missing-data situation in the pivotal cohort. [1] [2] [4]

The weekly arm’s Week 24 TND rate was 30% in April, then 35% in September because a late result arrived after the original analysis cutoff. Both versions are retained. A second discrepancy remains unresolved: the September IR release and April original report give p=0.024 for the four-weekly Phase 2b composite, while the September 8-K prints 0.0024. The table retains 0.024 as the original version and does not silently correct the filing. Neither value is a Week 48 comparative p-value.

Safety / Pivotal cohort through Week 24

Any adverse event was reported in 27/59 weekly recipients, 40/65 four-weekly recipients and 8/29 delayed recipients. No serious event or adverse-event discontinuation was reported in the active arms; the delayed arm had one serious, Grade 4 myocardial infarction. Deaths were not separately tabulated, so they are not assigned a zero. Absence of active-arm serious events in 124 participants over this period cannot exclude uncommon harms with chronic exposure. [1] [4]

Injection reactions and flu-like symptoms were more frequent in the active groups. The numerical difference between regimens is descriptive, with no supplied comparative safety test. The extension was described as having no new safety signal, but a complete Week 48 arm-level safety table was not provided. That qualitative statement does not replace longer-term counts or exposure-adjusted rates, particularly in patients with limited hepatic reserve.

Capital structure / Research execution and next evidence

Mirum is the direct developer and holds worldwide rights to brelovitug. Its August 5 financial report disclosed $561.3 million of unrestricted cash, equivalents and investments at June 30, versus $330.728 million of cash and equivalents alone. Convertible notes were carried at $750.079 million net. These are dated balance-sheet measures, not a current fully diluted capitalization or proof that every development obligation is funded. CEO Chris Peetz and clinical development leadership oversee a program that must sustain multinational follow-up, manufacturing and regulatory preparation. [1] [4] [5]

AZURE-4 remains a separate fourth-quarter 2026 readout. The September SEC filing now guides AZURE-2/3 data to the second half of 2027; the separate timeline record now reflects this updated window and retains the older guidance as historical evidence. The next clinical questions are the second pivotal replication, complete analysis plans, immunogenicity, long-term safety and whether laboratory response translates into reduced hepatic complications. None is resolved solely by the present positive primary endpoint.

Sources / References

1. Mirum · Phase 3 AZURE-1 and Phase 2b Week 48 results · September 28, 2026
2. Mirum · Phase 2b AZURE-1 Week 24 original release · April 27, 2026
3. ClinicalTrials.gov · NCT06907290 · posted March 2, 2026; retrieved September 30
4. Mirum · SEC 8-K covering September 25 and September 28, 2026 actions
5. Mirum · Second-quarter results · August 5, 2026
6. EASL Congress 2026 abstract book · AZURE-1 early population and Week 12 results