Comparison of imaging biomarkers in the Alzheimer Disease Neuroimaging Initiative and the Mayo Clinic Study of Aging

Abstract

Objective: To determine whether magnetic resonance imaging measurements observed in the Alzheimer Disease Neuroimaging Initiative (ADNI) convenience sample differ from those observed in the Mayo Clinic Study of Aging (MCSA) population-based sample.

Design: Comparison of 2 samples.

Setting: Fifty-nine recruiting sites for the ADNI in the United States and Canada and the MCSA, a population-based cohort in Olmsted County, Minnesota.

Patients: Cognitively normal subjects and amnestic subjects with mild cognitive impairment were selected from the ADNI convenience cohort and MCSA population-based cohort. A simple random sample of subjects from both cohorts in the same age range was selected, and a second sample applied matching for age, sex, educational level, apolipoprotein E genotype, and Mini-Mental State Examination score.

Main outcome measures: Baseline hippocampal volumes and annual percentage of decline in hippocampal volume.

Results: In the population-based sample, MCSA subjects were older, had less education, performed worse on the Mini-Mental State Examination, and had a family history of Alzheimer disease less often than did ADNI subjects. Baseline hippocampal volumes were larger in ADNI compared with MCSA cognitively normal subjects in the random sample, although no differences were observed after matching. Rates of decline in hippocampal volume were greater in the ADNI compared with the MCSA for cognitively normal subjects and those with amnestic mild cognitive impairment, even after matching.

Conclusions: Rates of decline in hippocampal volume suggest that ADNI subjects have a more aggressive brain pathologic process than MCSA subjects and hence may not be representative of the general population. These findings have implications for treatment trials that use ADNI-like recruitment mechanisms and for studies validating new diagnostic criteria for Alzheimer disease in its various stages.

Figure 1. A comparison of hippocampal volume measured from scans performed at 1.5T and 3.0T

The comparison was performed using 91 subjects from ADNI that had both a 1.5T and 3.0T scan at the same visit (32 CN, 39 aMCI and 20 AD). Scatter-plots show the 3.0T vs. 1.5T hippocampal volume (Panel A) and TIV (Panel B). Different colors are used to represent each diagnostic group (CN=black, aMCI=blue, AD=red) and the identity line indicating perfect agreement is shown as a solid black line. The Spearman correlation (Spearman rho) and Lins' concordance correlation coefficient (ccc), a measure of intra-class correlation, are shown at the top of each plot. The data demonstrates an excellent agreement between 1.5T and 3.0T hippocampal volumes and TIV.

Figure 2. Box-plots of hippocampal volumes in CN and aMCI subjects contrasting findings in the ADNI study with findings in the MCSA study

Panel A shows the results in two simple random samples. Panel B shows the results in two samples frequency matched by age, sex, education, APOE genotype, and MMSE score. The boxes indicate the median and interquartile range (IQR) of the distributions while the vertical lines extending from the boxes stop at the most extreme data points within 1.5 IQRs. Because all individual points are shown, the points have been shifted randomly in the horizontal direction to avoid overlap and improve the visual display. We summarize group-wise difference using the Probabilistic Index (PI) and Wilcoxon rank-sum p values (shown in brackets). A PI of 0.50 indicates no difference across groups, whereas a PI of 0.60 indicates that 60% of the time the hippocampal volume from a random subject in ADNI is higher than the corresponding value in a random subject from the MCSA.

Figure 3. Box-plots of annual percent decline in hippocampal volume in CN and aMCI subjects contrasting findings in the ADNI study with findings in the MCSA study

Panel A shows the results in two simple random samples. Panel B shows the results in two samples frequency matched by age, sex, education, APOE genotype, and MMSE score. Negative values represent a decline in hippocampal volume over time. The boxes indicate the median and interquartile range (IQR) of the distributions while the vertical lines extending from the boxes stop at the most extreme data points within 1.5 IQRs. Because all individual points are shown, the points have been shifted randomly in the horizontal direction to avoid overlap and improve the visual display. We summarize group-wise difference using the Probabilistic Index (PI) and Wilcoxon rank-sum p values (shown in brackets). A PI of 0.50 indicates no difference across groups, whereas a PI of 0.60 indicates that 60% of the time the annual percent decline in hippocampal volume from a random subject in ADNI is greater than the corresponding value in a random subject from the MCSA.