DIAGNODE-3: Intralymphatic GAD-alum in HLA DR3-DQ2 recent-onset T1D

What this trial tested and why it mattered

DIAGNODE-3 was the pivotal test of antigen-specific immunotherapy in type 1 diabetes (T1D): not general-population prevention, but a genetically targeted attempt to preserve the insulin production that remains at diagnosis. The idea is to re-teach the immune system to tolerate GAD65 — one of the main proteins the immune system attacks in T1D — by injecting it (formulated in an aluminium adjuvant, as "GAD-alum") directly into a lymph node, where immune cells are trained, alongside vitamin D.¹

The trial's defining feature was genetic selection. Earlier GAD-alum studies had missed their endpoints overall, but the reported rationale was a benefit signal in people carrying the HLA DR3-DQ2 haplotype, so DIAGNODE-3 enrolled only DR3-DQ2 carriers — ages 12 to 28, diagnosed within the previous 6 months.¹ That made it a direct test of the precision-immunology bet: give the therapy to exactly the people it was supposed to work in.

What happened

It did not work. On 27 March 2026 Diamyd announced that the pre-specified interim analysis — 174 of the 321 participants, baseline to month 15 — did not demonstrate statistical significance on the primary C-peptide endpoint at that timepoint, and that the pre-specified criteria required to continue the trial were not met.² The company's follow-up detail was blunter: no treatment effect on C-peptide was observed, neither in the overall trial population nor in any pre-specified subgroup within it, including the group previously flagged as potential "super responders" (DR3-DQ2 positive, DR4-DQ8 negative). HbA1c and time-in-range showed no difference between the treatment and placebo arms.²

On 10 April 2026, after independent external statistical validation of the interim data, Diamyd confirmed the results met the pre-specified futility criteria and did not support continuation. The company discontinued the trial and began an orderly wind-down, and its Board opened a formal review of the organization and its strategic alternatives.³ No new safety concerns were identified in either announcement.²³

How to read this

Two honest caveats. First, the efficacy numbers still come from company press releases, not a peer-reviewed paper — the full C-peptide tables are not posted. Second, the registry has now caught up: as of the 26 August 2026 update, ClinicalTrials.gov lists DIAGNODE-3 as terminated for futility, with actual enrollment 321 and actual completion 24 June 2026, and still has no posted results.¹ Treat the stop as confirmed; treat the detailed efficacy tables as still unpublished.

What it means for the field

This is a genuinely important negative result. DIAGNODE-3 was the strongest form of the argument that antigen-specific immunotherapy works if you find the right patients — a 321-participant phase-3 trial enrolling only the genetically selected population in which the signal had supposedly appeared. Per the company's reported interim analysis, the signal did not replicate. That is the classic fate of subgroup findings pulled from trials that missed their primary endpoint, and it is a caution for every "it works in this haplotype" claim in the field.

It does not sink immune-modulating therapy in T1D more broadly: teplizumab, a CD3-directed antibody, is approved to delay stage-3 onset in people with stage-2 disease. But the specific hope that a single autoantigen (GAD65) delivered into a lymph node could durably preserve the body's own insulin production — in the very people most likely to respond — did not survive contact with a phase-3 trial, per the company's reported interim analysis.