Comparison of reference regions for improved detection of change in florbetapir PET from phase 3 solanezumab trials


      Amyloid-PET is commonly utilized in Alzheimer’s disease(AD) anti-amyloid therapy trials for eligibility requirements and longitudinal evidence of target engagement. Improving methodologies to detect longitudinal change in amyloid-PET is of particular importance. Although cerebellum is a widely utilized reference region for cross-sectional quantitative evaluations, it may not be optimal for longitudinal assessments. We compared use of atlas based cortical white matter and whole-cerebellum reference regions for accuracy and power for detecting longitudinal change in Phase 3 solanezumab trial data.


      Florbetapir-PET scans were analyzed from 140 participants (72 placebo, 68 solanezumab) with mild dementia due to AD from the 18-month, randomized, placebo-controlled EXPEDITION 1&2 solanezumab studies. Participants had florbetapir positive scans at baseline, determined by mean cortical-to-whole-cerebellar standard uptake value ratio (SUVRs) ≥1.1, measured from a standard space template consisting of 6 cortical regions. For comparison to a whole-cerebellar reference region, a second normalization was performed on the longitudinal data using a centrum semiovale region as a correction factor (Figure). This method has been shown to improve signal to noise, while preserving the ability to use cerebellar SUVRs at baseline1. Analysis-of-Covariance models adjusted by baseline, study, treatment and age were used to assess baseline-to-endpoint change between treatment and placebo groups. For sample size estimations, 80% power and α=0.05 were used to detect magnitude of observed 18 month changes from baseline in the placebo group.


      Using a whole-cerebellar reference region at baseline and endpoint, least squares mean SUVRs for the placebo group increased 0.004±0.0129 (0.49%±0.91), and for the active treatment group decreased 0.006±0.0137 (0.19%±0.96)(p=0.62). Power analysis revealed a sample size of n=4056 to detect a difference between 18 month placebo group change and baseline SUVRs (i.e. no change in treatment group from baseline). White matter adjustments resulted in a mean increase of 0.011±0.0075 (0.79%±0.54) in the placebo group and mean decrease of 0.008±0.008 (-0.6%±0.57) in the active treatment group (p=0.08); the calculated sample size fell to n=421.


      Adjusting longitudinal SUVRs with a white matter reference region in these phase 3 anti-amyloid treatment trials increased mean change detection and decreased variance. This method resulted in a substantial improvement in statistical power to detect change. Reference: Abhinay Joshi, Michael Pontecorvo, Michael A. Navitsky, Ian A. Kennedy, Mark Mintun, Michael D. Devous. Measuring change in beta–amylod burden over time using florbetapir-PET and a subcortical white matter reference region. Alzheimer’s Dement. 2014;10(4):902.
      Figure thumbnail fx1
      FigureSecondary normalization for white matter reference region
      Figure: White matter normalization – the original baseline (Visit 1) cerebellar SUVR is used, but a new endpoint (Visit 2) accounts for the baseline to endpoint change observed in the WM.
      WM = White Matter; Vl = Visit 1; V2 = Visit 2; Cereb = whole cerebellum