Our results suggest that having a larger hippocampal and total brain volume sets cognitively intact individuals with a high burden of AD pathology apart from individuals with overt dementia and a similar amount of AD pathologic changes. This may be interpreted in several ways. First, larger brain volumes may indicate a greater preexisting brain reserve. Second, the cognitively intact group may have other forms of compensation or protection from the pathologic processes that underlie AD-related changes. Third, the brain volume loss in the AD group is not directly caused by the NFTs and NPs. In this case NFTs and NPs are makers of other pathologic processes that actually lead to brain atrophy in AD.
The relationship between preexisting brain reserve and risk of AD has been investigated in several studies. While some studies have shown an association between indirect measures of brain size such as head circumference or intracranial volume and risk of AD, others have failed to show such an association.23-27
In our study, there was no difference in intracranial volume between the two groups. If we assume that intracranial volume is a crude measure of maximum brain size attained during life, then this argues that the larger brain volume we observed does not reflect a preexisting reserve. We also did not observe a difference in education or SES between our cases and controls. Both education and SES have been suggested to be an index of cognitive reserve providing protection against dementia.28
The second interpretation of our results is that there are other physiologic or molecular mechanisms providing protection against the pathologic changes associated with AD. An example of such mechanisms that may promote or protect cognitive function are the number of synapses.29
A study investigated the number of synapses in individuals with AD, mild cognitive impairment, and no cognitive impairment.29
The authors reported that synapse loss was higher in the AD group compared to the mild cognitive impairment and cognitively intact groups, and was a structural correlate of cognitive decline. The authors also investigated an association between total Braak scores, NIA Reagan scores, and synapse numbers, and these did not show an association. Lack of such an association further supports the notion that features such as synapse number may contribute to whether cognitive decline proceeds in the presence of AD pathology. Studying protein expression in the brains of cases and controls like ours with equivalent AD pathology may also improve our understanding of other mechanisms playing a role in AD pathophysiology.30
Apoptotic pathways represent another mechanism that may play a role in resistance to dementia in these individuals. Apoptosis has been suggested to be one of the main causes for the cell loss accompanying neurodegenerative diseases such as AD.31
Brain volume loss in patients with AD may be secondary to activation of apoptotic pathways by the NFT and plaques.32
Thus differences in regulators of apoptosis may lead to resistance to neurodegeneration and associated brain volume loss. For example, the FAS
gene, which is a member of the tumor necrosis factor receptor superfamily, plays a role in apoptosis and has been associated with AD.33
It has also been shown to be associated with brain volumes obtained by MRI scans in patients with AD.34
Polymorphisms in genes such as FAS, which play a role in regulation of apoptosis, may mediate the relationship between plaques and tangles and the degree of neurodegeneration, brain volume loss, and presence of symptoms of AD.
Another interpretation of our finding is that the plaques and tangles do not directly cause loss of brain volume observed in patients with AD. Other mechanisms that are closely correlated with plaques and tangles may lead to brain volume loss in patients with AD. Several studies have shown a correlation between postmortem brain volumes and postmortem Braak stage, NFT measures, and neuron numbers.35-37
One study found that postmortem neocortical NFT and NP pathology correlated well with last ventricular volume prior to death and rate of ventricular volume increase in patients with AD while in cognitively intact individuals such a correlation did not exist.37
The authors also reported that the last hippocampal volume prior to death correlated well with hippocampal NFT pathology in patients with AD, while in the cognitively intact subjects the hippocampal NFT pathology did not correlate with antemortem hippocampal volume. Lack of an association between brain volumes and pathology in the cognitively intact subjects may mean that NFTs and NPs are markers of another process in AD, but do not lead to brain volume loss in the absence of these other processes related to AD. However, a correlation between brain volume and AD neuropathology in nondemented individuals has been reported in some other studies.35,36
This may be because the nondemented group in these studies may represent a more heterogeneous group with some having preclinical AD or some mild memory problems.
Our entry criteria and prospective clinical evaluations enhanced the likelihood that subjects in our study do not fall into the preclinical AD group and did not have subtle memory problems. When we compared several psychometric tests between the cognitively intact group with high AD pathology and another group of cognitively intact elderly who were found to have low AD neuropathology (Braak stage I or II and no or sparse NPs by CERAD criteria), we did not find significant differences between these two groups (table e-1 on the Neurology
® Web site at www.neurology.org
). Similarly, a study using Pittsburgh compound to image amyloid deposition reported no significant differences in cognitive performance between cognitively intact elderly with high vs low amyloid binding.38
Another study suggested that the relationship between postmortem brain volumes and cognitive function is more robust than the relationship between neuropathology and cognitive function.39
This further supports the notion that brain volume seems to play a major role whether cognitive decline occurs in the setting of AD neuropathology.
This study has several limitations. First, matching subjects with high AD pathologic lesion burden and selecting only those with antemortem MRI scans resulted in a restricted sample size. Nevertheless, we observed a statistically significant difference in the brain volumes between AD and cognitively intact subjects even with our relatively small sample size. Our results need replication in a larger sample. Second, since most patients with AD at the final stages of their disease become housebound, subjects in the AD group had a longer time between their MRI and death. This has an effect of minimizing the magnitude of difference in the brain volumes. Ideally one would prefer to have MRI scans within 1 year of death for both groups and we tried to correct for this confounder statistically. Finally, we matched the subjects based on widely used semiquantitative methods to assess the presence of tangles and plaques. Given the possibility of individual variation within the same Braak or CERAD scores, quantitative pathologic assessment methods may be needed in future studies to be able to match cases and controls more precisely.
Our results suggest that individuals with a high burden of AD pathologic lesions do not manifest overt cognitive impairment if they also have larger hippocampal and brain volumes. Identifying the mechanisms whereby larger hippocampal and brain volumes are protective, either by providing more brain reserve or as a result of other processes leading to resistance to neuronal loss traditionally attributed to NFT and NPs, warrants further investigation.