Results of the present study revealed a number of important considerations for FMRI studies of aging. First, a significant reduction in resting state perfusion in older adults relative to their younger counterparts was observed. This finding is consistent with several ASL, PET, and SPECT studies demonstrating age-related reductions in resting state CBF and cerebral metabolic rate of glucose (e.g., Hagstadius & Risberg, 1989
; Parkes et al., 2004
; Petit-Taboue et al., 1998
), although rates of perfusion decreases have varied. One particularly relevant study of 34 healthy adults (aged 20–67 years) using continuous arterial spin labeling (CASL) reported gray-matter perfusion reduction of 0.45% each year (Parkes et al. 2004
). The current study corroborates this and other previous findings of age-related differences in CBF and demonstrates consistency of the ASL method with the broader literature for the quantitative study of perfusion in aging.
Past findings regarding the impact of baseline CBF on BOLD response during functional tasks have been somewhat more equivocal (Stefanovic et al., 2006
). For instance, increased baseline CBF produced by carbon dioxide (CO2
) inhalation, acetazolamide administration, hyperventilation, or caffeine use has been associated with both decreased (Bandettini and Wong, 1997
; Brown et al., 2003
; Cohen et al., 2002
) and increased BOLD response (Cohen et al., 2002
; Mulderink et al., 2002
). In addition, vasoconstrictive indomethacin administration has been accompanied by a decreased BOLD response (Bruhn et al., 2001
). Despite these inconsistencies in the literature, studies demonstrate that baseline physiology impacts the magnitude of BOLD activation. Age-related baseline CBF and cerebrovascular alterations have been reported and the present findings of baseline hemodynamic differences across age may contribute, in part, to findings of age-related differences in BOLD contrast analyses. Therefore, resting state CBF is an important consideration when interpreting functional neuroimaging results of the BOLD response.
In addition, older adults demonstrated a significantly larger percent change in CBF response while viewing novel pictures as compared to the young participants, although a concomitant increase in the BOLD response was not observed. These differences in the CBF response were seen in the absence of discrepant performance on primary measures of learning and memory. This finding suggests that perfusion imaging has the potential to detect age-related alterations in MTL response to learning that go undetected when examined with the BOLD signal. Our findings further suggest that CBF and BOLD data yield different information and, therefore, the combination of both techniques may better elucidate brain activation patterns than either technique in isolation. However, it should be noted that greater percent change CBF during novel relative to familiar image presentation could reflect a lower level of absolute CBF response during the familiar image condition. As our recent work (Restom et al., 2007
) suggests, the greater percent change CBF response during novel picture encoding in the older participants may reflect a lower level of CBF response during the familiar image condition, which would be consistent with the observed age-related reduction in resting-state CBF.
Regarding cognitive performance, across both age groups, better performance on word-list learning was associated with a larger percent change in left MTL CBF response. These associations between CBF response and episodic memory suggest that participants with better performance on certain memory measures may have been invoking compensatory upsurges in CBF response to support successful memory operations. In contrast, percent change in BOLD signal during novel picture encoding was not significantly associated with any measures of cognition. Thus, in the present study it appears that CBF rather than BOLD response was better related to memory performance in ostensibly healthy and nondemented older adults.
In the older participant group, consistent with several previous reports, we also found higher blood pressure to be associated with poorer performance on cognitive measures (e.g., Elias et al., 1993
; Kilander et al., 1998
; Knopman et al., 2001
; Swan et al., 1998b
; Tzourio et al., 1999
) and smaller brain volumes (e.g., Akiyama et al., 1997
; Swan et al., 1998a
). Systolic blood pressure but not stroke risk showed a significant association with cognition. One distinction to note is that the FSRP is comprised of a combination of current and cumulative risk factors, whereas systolic blood pressure is a current index of the cerebrovascular system taken on the day of scanning and, as such, may represent a better index of cerebrovascular integrity than those factors assessing lifetime risk. As Knopman and colleagues (2001)
stated, the associations among individual risk factors is very complex. It is possible that the effects of hypertension vary from other risk factors (e.g., diabetes) included in our stroke risk assessment in terms of the locus or mechanisms of brain injury.
In recent years, interest has increased in examining the relative contributions of vascular risk factors (e.g. hypertension, diabetes, coronary artery disease) to the pathogenesis or exacerbation of age-related degenerative diseases like Alzheimer’s disease (AD), and recent studies have shown white matter (WM) lesions to be important predictors of subsequent cognitive decline in mild AD (Honig et al., 2004). Adak et al. (2004)
also reported that greater abnormal WM volumes were associated with shorter times to convert either from normal to ‘questionable dementia’ or from ‘questionable’ to mild dementia. Knopman et al. (2005)
demonstrated a link between diabetes and cerebral atrophy in their sample of normally-aging older adults and suggested that reduced cerebral perfusion and subtle microinfarction may be linked to the observed atrophy and concomitant cognitive decline. Even older adults in relatively good health experience ischemic brain changes before the development of frank, clinically diagnosed cerebrovascular disease (McPherson and Cummings, 1997
; Raz et al., 1998
). Although speculative, some of the determinants of abnormal WM likely include those captured by the stroke risk assessment used in the present study. Of course, such risk factors may also interact with one another or with still other risk factors such as advancing age. Thus, interest in the determinants of WM lesions in aging and their impact on cognition and risk for conversion to diseases such as AD has risen, especially because many of these risk factors may prove to be modifiable by medications, diet, or other lifestyle variables.
For functional neuroimaging studies with older adults, if stroke risk is not assessed, cognitive decline may be attributed to aging when it may be at least partially due to cerebrovascular alterations. In the present study, all older adults were healthy and thus there was little variability in our stroke risk profiles (i.e., the stroke risk score only ranged from 5 to 15 points out of a possible 48 points), which resulted in a truncated range for our correlative analyses. Despite these limitations, we nonetheless observed some important relationships between measures of stroke risk and imaging and cognition. A larger sample size and increased variability in stroke risk profile may reveal additional relationships between CBF and BOLD responses and cognitive variables.
Despite the CBF and BOLD changes noted, neither provides a direct quantitative measure of the neural activity within this region. Rather, they provide a qualitative estimation at best. Nevertheless, our data demonstrate that the two measures are not redundant and provide complementary information. Ultimately, combination of ASL with BOLD is a method that, when combined with a vasodilatory stimulus like carbon dioxide inhalation, will allow for a direct quantitative estimate of metabolic change (e.g., CMRO2
; [Davis et al., 1998
; Hoge et al., 2005
]). The estimation of CMRO2
is a well-defined physiological quantity that is tightly linked to neural activity, and the opportunity to non-invasively estimate a quantitative measure linked to neural activity will be particularly valuable for clinical studies where factors such as disease, medication, and age can greatly impact the cerebrovascular system and cause changes in CBF and BOLD measures that are unrelated to changes in neural activity.
Our preliminary work in this area (for reviews, see Brown et al., 2007
; Wierenga and Bondi, 2007
) suggests that aging results in differences in baseline cerebral blood volume or differences in baseline 02
extraction fraction. According to the deoxyhemoglobin dilution model (see Brown et al., 2007
, for discussion), BOLD contrast can give misleading information about the underlying metabolic activity of neurons when differences in baseline CBV or CMRO2
are present. There is an inherent ambiguity of interpreting BOLD response in isolation, especially in the presence of influencing factors such as age, disease, or disease risk. Combined ASL/BOLD imaging thus offers a considerably richer set of indices by which to characterize the hemodynamic response to cognition.
As an applied MRI technique, ASL has only recently begun to be used with clinical populations such as Mild Cognitive Impairment and Alzheimer’s disease (Johnson et al., 2005
). Perfusion serves as an index to assess brain function and integrity in certain susceptible regions (e.g., MTL, posterior cingulate). Thus, ASL has important implications for diagnosis, treatment monitoring, and outcome prediction in dementia, stroke, and other neurological diseases (Warach et al., 1996
). Initial studies assessing both BOLD and CBF in the MTL during encoding have demonstrated the feasibility of this technique in older adults (Restom et al., 2007
). The main challenge is the relatively low contrast-to-noise ratio of the CBF measures. Technological advances (e.g., higher field strength) will further improve the temporal and spatial resolution of ASL and facilitate the use of combined BOLD and ASL data (Fernandez-Seara et al., 2007
). The combination of CBF data with BOLD responses to yield CMRO2
information may be particularly useful in interpretation of FMRI study results and has the potential to help reconcile many discrepancies that currently exist from BOLD studies.
Despite our intriguing findings with ASL/BOLD imaging, the present study has several caveats. In imaging-based comparative aging studies, one must account for between-group differences in degree of atrophy by age. In an effort to reduce these potential differences, we used a partial volume correction on each data set and native space ROI tracings to avoid errors that may occur when normalizing brains with varying degrees of atrophy to the same standard space template. Such steps, though helpful, may not completely eliminate atrophic differences between our groups. In correcting for partial voluming, we assumed that the CBF in gray matter is 2.5 times larger than that in white matter. It is important to acknowledge that this ratio may change with age. However, Restom and colleagues (2007)
performed 3 sets of CBF estimates: one estimate in which they assumed that gray matter CBF is 2.5 times larger than white matter CBF, a second in which they assumed that CBF reflected gray matter only, and a third estimate in which they made no partial volume correction. Restom and colleagues (2007)
reported that significant between-group differences remained the same regardless of which estimate was used to perform the partial volume correction. Corroborating that finding, Johnson and colleagues (2005)
also reported no significant differences in their results when they used the partial volume correction compared to when they did not.
Another consideration is that older adult groups may include individuals with preclinical disease. In order to ensure that our older participants were free from cognitive, neurological, or other disease, they were carefully assessed for the presence of cognitive impairment, significant stroke risk factors, and medical and psychiatric conditions known to affect cognition. Our older group was found to be free of medical or psychiatric disease, cognitive impairment, or significant stroke risk. Moreover, 4 of our 15 older adult participants (or 26%) were carriers of the APOE ε4 allele—a genetic susceptibility factor for Alzheimer’s disease—a percentage that roughly corresponds to its 20–25% frequency of occurrence in the population (Corder et al., 1993
; Ghebranious et al., 2005
). Thus, our sample was not disproportionately at increased risk for preclinical disease. Future efforts targeting at-risk groups (e.g., mild cognitive impairment, genetic risk, etc.) using functional ASL/BOLD imaging will help clarify the complex interactions underlying neurovascular coupling and ultimately provide a better understanding of neural substrates leading to cognitive dysfunction.
In summary, we have demonstrated differences between CBF and BOLD response to learning between ostensibly healthy young and older adults that are suggestive of age-related compensatory mechanisms. However, whereas the lack of difference in the BOLD responses could be interpreted as evidence for no change in neural activity with age, the significant difference in the percent CBF responses could be interpreted as evidence for an age-related compensatory increase in task-related neural activity. However, a definitive interpretation is lacking in the absence of a direct quantitative estimate of neural activity such as from measurement of CMRO2
response during learning. Future studies may be able to obtain such an estimate through simple breath-hold or CO2
inhalation challenges (see Davis et al. 1998