WASHINGTON – Hippocampal metabolic reductions seen on PET scans in people with no clinical signs of cognitive impairment may be able to predict who will go on to develop Alzheimer's disease and mild cognitive impairment, according to data presented at an international conference sponsored by the Alzheimer's Association.
Pending further investigation, the technique holds the potential to fast-forward the search for preventive measures–such as new medications–that could hold the disease at bay before it becomes clinically evident.
In a study of 53 people that spanned 24 years, those who eventually progressed from normal cognition at baseline to develop Alzheimer's disease (AD) showed 23% less hippocampal glucose metabolism on baseline fluorodeoxyglucose (FDG) PET scans, compared with a cognitively stable control group. Those who progressed from normal cognition to develop mild cognitive impairment (MCI) showed 11% less hippocampal glucose metabolism, compared with the control group, said Lisa Mosconi, Ph.D., of New York University.
Reduced hippocampal metabolism at baseline predicted progression to AD with 83% sensitivity and 86% specificity–a total accuracy of 85%. Progression to MCI was predicted with 74% sensitivity and 68% specificity–a total accuracy of 71%.
“These numbers are interesting when you consider that the signs of Alzheimer's disease were observed an average of 9 years after baseline, and the signs of MCI were seen 11 years after baseline,” said Dr. Mosconi.
Dr. Mosconi and her colleagues followed 53 normal elderly volunteers for 10–24 years. Participants were at least 50 years old at baseline, had at least 12 years of education, and scored at least 28 on the Mini-Mental State Examination (MMSE) and no greater than 2 on the Global Deterioration Scale.
Every 2 years, the volunteers underwent complete clinical, neuropsychologic evaluations. Throughout the study, all participants also underwent at least two PET scans–30 underwent three scans–at least 3 years apart. All images were acquired with the same scanner.
Over the course of the study, 28 participants remained cognitively normal, 19 eventually developed MCI, and 6 developed AD.
Diagnosis of MCI and AD was made by standard neuropsychologic evaluation.
The three groups (stable normal, normal to MCI, and normal to AD) were similar in terms of gender, education, and MMSE scores. However, the stable normal group was several years younger on average, so the researchers corrected the data for age.
The participants who eventually developed AD lost an average of 4% of their hippocampal glucose metabolism capacity per year, those who developed MCI lost 2%, and the control group lost 1%. The difference between those with AD and the control group was statistically significant, but the difference between those with MCI and the control group was not.
The diagnoses of two of the patients who progressed from normal cognition to AD were confirmed by postmortem examination–the most definitive means of diagnosing the disorder. At this examination, hippocampal and cortical volume reductions were observed that were consistent with AD pathology. At baseline, these two patients had 35% and 15% reduced hippocampal metabolism, respectively, at baseline, compared with the control group.
The PET scans were analyzed using HipMask, a software program developed at the university. To develop this program, the researchers drew hippocampal regions of interest on MRI scans of some study patients (ranging in cognition from normal to AD). The MRI scans (and regions of interest) for each patient were then spatially normalized to the shape of an anatomical brain reference image, which is a custom-made template. The resulting images were then overlaid to produce an image containing only areas of the hippocampus where all of the images overlapped–the HipMask.
“The HipMask is a hippocampal masking image that includes only those portions of the hippocampus where the overlap of the subjects is maximized after size normalization procedures,” said Dr. Mosconi. The HipMask was verified against individual MRI scans. On average, 96% of the HipMask represented true hippocampal tissue.
The HipMask was then applied to PET scans–performed within 3 months of the MRIs–to derive estimates of the hippocampal glucose metabolism–a measure of brain activity.
The results also were validated against the standard region-of-interest technique. There was very good correspondence between the HipMask results and the standard technique in all clinical groups, said Dr. Mosconi.