given by the MCMC sampling and cross-validation with clear sep- aration between GM and CBF markers. Within modalities some po- sitional variance existed, especially within the GM markers where the Gaussian fit was not robust over cross-validation samples. Con- clusions: This EBM analysis suggests that ASL-MRI-based perfu- sion changes occur later in the AD disease course than GM volume changes. We should be careful in generalizing these find- ings which might arise from ASL-MRI being less sensitive than T1-weighted MRI. Our findings therefore only hold for the current regional measures, cohort, and ASL-MRI technique. Our future work concerns evaluation of other potentially more sensitive ASL-MRI-based markers. P1-328 INSULIN RESISTANCE AND LONGITUDINAL ASSOCIATIONS WITH TEMPORAL ATROPHY ACROSS THE ALZHEIMER’S DISEASE SPECTRUM Brandon Skylar Klinedinst 1 , Joseph L. Webb 1 , Jonathan Cerna 1 , Auriel A. Willette 1,2 , 1 Iowa State University, Ames, IA, USA; 2 National Institute on Aging/National Institutes of Health (NIA/NIH), Baltimore, MD, USA. Contact e-mail: brandon.klinedinst@gmail.com Background: Alzheimer’s disease (AD) is characterized by progres- sive medial temporal atrophy. Peripheral insulin resistance (IR) is associated with increased risk for AD. Previous reports indicate a consistent association between higher IR and temporal atrophy and hypometabolism, as well as cognitive impairment in late mid- dle-aged adults. To date, however, it is unknown to what degree IR predicts longitudinal temporal atrophy across the AD spectrum. Methods: In the Alzheimer’s Disease Neuroimaging Initiative (ADNI), we assessed associations between baseline IR and longitu- dinal atrophy in temporal areas across and between 57 Cognitive Normal (CN), 327 Mild Cognitive Impairment (MCI), and 94 AD participants. Freesurfer 5.1 was used to derive bilateral mean vol- ume for hippocampus and mean cortical thickness (CT) for entorhi- nal cortex and medial temporal lobe (MTL). Timepoints were baseline, month 6, month 12, month 24, and month 36. Linear repeated measures mixed models regressed log transformed IR against outcomes of interest. Predictors of interest were IR * Time, Diagnosis * Time, and IR * Diagnosis * Time interactions. Covariates included age, sex, APOE status, and the random effect of Subject. For hippocampus, intracranial volume was also a covar- iate. Results: There was a reliable step-wise decrease from CN to AD in temporal volume and CT over time. IR across diagnostic groups was not associated with temporal atrophy. However, IR * Time * Diagnosis interactions were significant for MTL CT [F¼5.006, p<.001], entorhinal CT [F¼3.030, p¼.004], and hippo- campus volume [F¼2.602, p¼.011]. Figure 1 illustrates the pattern of association for entorhinal CT. From baseline to month 36, higher IR predicted hypertrophy (p<.001) in MCI. For AD, higher IR did Table 1 Participant characteristics Data set Diagnosis # of participants (male/female) mean Age [years] (std) mean Mini-Mental State Examination score (std) VU University Medical Center Amsterdam/NL AD 64 (30/34) 65.9 (7.0) 20.6 (4.4) MCI 22 (12/10) 64.2 (6.5) 26.1 (1.7) Controls (cognitive complaints) 91 (56/35) 59.5 (8.9) 28.0 (1.7) Erasmus MC Rotterdam/NL AD 23 (14/9) 66.5 (7.7) 24.1 (3.9) Controls (cognitively healthy) 34 (22/12) 64.3 (6.4) 28.7 (1.3) Poster Presentations: Sunday, July 24, 2016 P553