Use of an FDG-PET derived hypometabolic convergence index enrichment strategy to reduce sample sizes in Alzheimer's disease clinical trials: findings from the Alzheimer's Disease Neuroimaging Initiative (ADNI)

Abstract

While the Alzheimer's disease Assessment Scale-Cognitive Subscale (ADAS-Cog) is commonly used to measure cognitive decline in probable Alzheimer's disease (pAD) and mild cognitive impairment (MCI) patients, rates of decline vary. We previously introduced a single index, Hypometabolic Convergence Index (HCI), to characterize the extent to which the pattern and magnitude of cerebral hypometabolism in a person's fluorodeoxyglucose-positron emission tomography (FDG-PET) image corresponds to that in pAD patients. Here, we used data from ADNI to examine the extent to which 1) a person's baseline HCI predicts cognitive decline using the ADAS-Cog, Mini-Mental State Exam (MMSE), Clinical Dementia Rating-sum of boxes (CDR-SB) and Auditory-Verbal Learning Test (AVLT-Total), and 2) the HCI could be used to enrich clinical trials for clinical decline and reduce the number of patients needed to detect a treatment's clinical effects. Baseline HCIs were computed for 120 MCI and 54 mild AD patients who had up to 24-month data. We first characterized the extent to which HCIs correlated with 12-month and 24-month clinical declines. We then estimated the sample sizes needed to detect an AD-slowing treatment's effects on ADAS-Cog before/after enrichment for those patients with HCIs greater than the predetermined threshold of 13.82 for pAD and 8.19 for MCI (Chen et al., 2011, Neuroimage) with 80% power, 0.05 type-I error and a 25% treatment effect. In pAD, HCIs correlated (p < = 0.05) with subsequent ADAS-Cog, MMSE, CDR-SB and AVLT-Total 12-month(24-month) decline (r = 0.42(0.47), -0.35(-0.55), 0.28(0.29), -0.40(-0.43), respectively). In MCI, HCI also correlated with subsequent decline at 12-month(24-month) (r = 0.33(0.32), -0.43(-0.50), 0.29(0.37), -0.17(-0.32), respectively). HCI enrichment is estimated to reduce the number of pAD patients needed per treatment arm to detect an AD-slowing treatment's effect on ADAS-Cog from 406 to 243 in a 12-month trial and from 168 to 137 in a 24-month trial. Similarly, it is estimated to reduce the number of MCI patients from 1,718 to 749 in a 12-month trial and from 926 to 374 in a 24-month trial. Baseline HCIs could be used to predict subsequent clinical declines in pAD and MCI patients and to reduce the number of patients needed to detect an AD-slowing treatment's effects in randomized clinical trials.