Longitudinal MRI analysis from our large cohort of normal controls, individuals with MCI, and individuals with AD provides strong evidence that neocortical atrophy rates are not uniform across the cortical mantle, nor are changes in atrophy rates constant in many regions with increased level of clinical impairment. We demonstrate a pattern of atrophy in MCI–CDR-SB 0.5–1 that begins in medial and inferolateral temporal, inferior parietal, and posterior cingulate and spreads to encompass superior parietal, widespread prefrontal, and lateral occipital cortex with progression to MCI–CDR-SB 1.5–2.5. In patients with mild AD, atrophy rates are greater in all cortical regions with the exception of primary visual and auditory cortex relative to age-matched controls. These data support our primary hypothesis and are consistent with previous findings that atrophy in MCI and AD progresses along a posterior-to-anterior gradient with relative sparing of sensory regions.4,23,24
However, regional atrophy rates do not always progress in a linear manner. Whereas increases in atrophy rates were greatest between controls and those with MCI–CDR-SB 0.5–1 within most medial temporal regions and plateaued slightly between those with MCI–CDR-SB 1.5–2.5 and early AD, increases in atrophy rates within the hippocampus and lateral temporal neocortex did not differ across levels of impairment. Conversely, increases in atrophy rates within most prefrontal, parietal, and anterior cingulate regions were modest between controls and patients with MCI–CDR-SB 0.5–1, but the degree of increase was greater as disease severity increased. Therefore, although atrophy rates increased throughout the neocortex with disease severity, the rate of increase differed across regions and level of clinical impairment. A few longitudinal studies have shown acceleration in whole brain, ventricular, and hippocampal atrophy rates with disease progression.5–7
Our cross-sectional findings are in line with other studies suggesting that hippocampal atrophy rates increase linearly with level of disease severity.25
Such discrepancies in the literature may reflect methodologic differences in hippocampal measurements, variability in MRI scanning intervals, and/or differences in study design and patient cohorts (i.e., acceleration of hippocampal atrophy rates has been shown in familial AD).7
Nevertheless, our findings extend the literature by revealing that observed increases in atrophy rates vary by neocortical region and level of clinical impairment.
Understanding the different spatial patterns and regional trajectories of neocortical loss that accompany various stages of disease can provide critical information for early detection of disease, as well as response to treatment. Whereas atrophy rates within medial temporal lobe regions seem to be the earliest indicator of mild disease, atrophy rates within anterior and posterior association cortex may help to determine treatment efficacy with respect to slowing disease progression and cognitive decline. Such longitudinal data will be particularly important for studies of disease-modifying treatments because each subject serves as his or her own control, allowing disease progression to be assessed directly through repeat evaluations. When monitoring response to treatment, our data suggest that the cortical areas targeted will depend on where patients reside along the disease spectrum. Slowing of entorhinal atrophy rates in patients with mild impairment may represent a true treatment effect, whereas later in the disease it may reflect the natural course of disease progression. Similarly, a steady rate of decline in prefrontal regions in patients with greater impairment may reflect a strong response to treatment because increases in atrophy rates would be expected with conversion to early AD. These patterns should be carefully considered in studies that select MRI biomarkers as primary or secondary outcome measures because overlooking the natural history of disease progression could result in overestimation or underestimation of treatment effects. High-throughput image analysis procedures, such as the ones used in this study, would be particularly useful and cost-effective in large-scale clinical trials designed to measure treatment effects in hundreds of individuals, within multiple regions of interest, and over several time points.
Despite support for our hypothesis of the spatial distribution of changes observed in our patient groups, the pattern of atrophy in our control group did not support our hypothesis of greater atrophy rates in prefrontal and parietal regions relative to posterior association cortex. Rather, we found 1-year atrophy rates to be relatively small and of similar magnitude across frontal, parietal, and temporal lobe regions. Although these data initially appear at odds with existing literature, they are commensurate with other studies showing minimal, diffuse cortical atrophy over a 1- to 2-year period in healthy individuals aged 59–85 years.4,26,27
Therefore, it is possible that with a longer time interval, the predicted anterior-to-posterior gradient of atrophy frequently observed in normal aging would have emerged. Additional insights into patterns of normal age-related atrophy will be critical to the interpretation of pathologic patterns of aging.
Despite the potential clinical value of our findings, there are limitations of this study that should be noted. First, longitudinal follow-up was limited to 1 year. Therefore, our longitudinal data were supplemented with cross-sectional analysis to investigate the change in atrophy rates across levels of clinical impairment. Longitudinal studies that follow the same individuals for 5 to 10 years will provide a more definitive analysis of the pattern of atrophy (i.e., regional acceleration) associated with disease progression. Long-term follow-up is particularly important for individuals diagnosed with MCI because of the heterogeneity within MCI and the uncertainty as to what this group truly represents. Second, we do not currently have histologic verification of disease. Therefore, it is possible that some subjects have disorders other than AD, or have comorbid pathology that was not detected with our in vivo measures. Taken together, these data provide additional insight into the complex patterns and rates of neocortical atrophy that characterize different stages of clinical impairment. This information may aid in the selection of biomarkers for future clinical trials designed to evaluate the effect of disease-modifying treatments.