API Generation Study 2 (GS2)

Generation Study 2 was one of the two Alzheimer's Prevention Initiative Generation Program trials run by Banner Alzheimer's Institute with Novartis and Amgen. It tested whether the oral BACE1 inhibitor umibecestat (CNP520) could delay symptom onset in cognitively unimpaired people aged 60 to 75 who carry one or two APOE4 alleles. Dosing started in August 2017 and the trial was stopped early after an interim analysis showed mild cognitive worsening under umibecestat; the registry lists it as terminated in March 2020. Screening and longitudinal clinical data and the trial images are requested through the LONI Image and Data Archive.

Composition

1145 participants were randomized 2:1:2 to umibecestat 50 mg (456), umibecestat 15 mg (233) or placebo (456). APOE4 homozygotes could join regardless of amyloid status, while heterozygotes needed elevated brain amyloid on PET or in CSF. Of the 1144 in the registry's baseline table, 720 are women; 201 were 64 or younger, 425 were 65 to 69 and 518 were 70 or older. Most participants are white, and 95 of the 194 listed sites are in the United States. The trial had planned about 2000 participants.

Acquisition

Brain MRI was done for safety (ARIA and cerebrovascular findings), volumetry and resting-state fMRI, every six months in the first year and yearly afterwards. Amyloid PET was voluntary at baseline unless it was used to establish amyloid status, and tau PET was planned at baseline, year 2 and year 5. Because of the early stop, the registry reports no follow-up amyloid or tau PET. CSF and blood samples were also collected. Scanner models, field strengths, PET tracers and the sequence list are not described in public sources.

Annotations

No image annotations are described. Central MRI reads and derived measures such as brain volumes and amyloid PET SUVR are part of the clinical data.

Known limitations

• Only 1145 of the planned 2000 participants were randomized, and treatment lasted far shorter than the planned five years.
• The structural MRI contrasts are not named in the public sources, so only the resting-state fMRI (BOLD) is listed.
• The Data Transfer Agreement signed after approval is not public, so its terms on commercial use, model training and publication are unknown.
• Generation Study 1 is a separate trial with its own LONI project.