Abstract
Introduction:
This study aimed to characterize the association of cognitive decline starting in midlife with brain pathology in late-life in the absence of dementia.
Methods:
Non-demented Atherosclerosis Risk in Communities (ARIC) participants with brain imaging, all cognitive factor scores (CFS), and non-missing covariates were included. CFS were collected at three visits across 21 years (1990–2013) (short-term cognitive change [1990–1996], long-term cognitive change [1990–2013]), and brain magnetic resonance imaging (MRI) and florbetapir positron emission tomography (PET) imaging were collected in 2011–13 (PET subset n=327). Outcomes of interest were total and regional brain volumes (cm3), log2(white matter hyperintensity volume), white matter integrity (fractional anisotropy, mean diffusivity), ≥1 lacunar infarct (3–20 mm), and elevated brain β-amyloid (SUVR >1.2). Multivariable linear/logistic regression related outcomes to CFS slopes after adjusting for demographics and total intracranial volume.
Results:
At baseline, the 1734 participants had a mean (SD) age of 55 (5.2) years, were 60% female, and 26% Black. After adjustment, a 1-SD larger long-term decline in CFS was associated with a smaller relative total brain volume by 1.2% [95%CI: 1.0, 1.5], a smaller relative temporal lobe meta region volume by 1.9% [1.5, 2.3], a 13% [9, 17] larger volume of white matter hyperintensities, a 1.3-fold [1.2, 1.4] higher odds of having ≥1 lacune, and 1.7-fold [1.3, 2.2] higher odds of elevated brain β-amyloid deposition, and worse white matter integrity. Some long-term associations were also found for midlife short-term declines in CFS.
Conclusions:
This study provides evidence that starting in midlife, short-term and long-term declines in cognition are associated with multiple deleterious late-life differences in non-demented brains.
Keywords: brain, cognitive decline, imaging, cohort, longitudinal
Introduction
It is understood that dementia affects cognition as a result of long-term neuropathological processes in the brain.[1,2] However, there is scant literature elucidating the relationship between changes in cognitive function starting in midlife and differences in late-life brain pathology among those who have not yet developed dementia. Studies aiming to understand this association require a lengthy follow-up and large sample size since the pre-clinical period before overt dementia lasts decades. Unfortunately, much of the literature on this subject is limited by cross-sectional designs, small sample sizes, and–importantly–short follow-up prior to imaging.[3–6]
Cross-sectional study designs are useful because they mimic the information gathered in a typical clinical visit: a patient presents with concerning cognitive function, leading to a simultaneous assessment of cognition and brain imaging. A major limitation of previous studies using a cross-sectional evaluation of cognition is that they are confounded by many factors, including a patient’s education, ethnicity, culture, social, and baseline cognitive factors. Relating change in cognition to brain imaging–as we do–provides stronger, less confounded inferences about specific associations between declining cognition and brain pathology. Recent studies with this stronger design have nevertheless been limited by small sample sizes and short follow-up[6,7], or were limited by only examining late-life changes in cognition.[8–12] Thus an important question remains unanswered: How does a non-demented brain in late-life relate to long-term declines in cognition that began in midlife?
This study aimed to characterize the association of long-term cognitive decline starting in midlife with brain pathology in late-life in the absence of dementia. Specifically, which pathology and brain regions (e.g., gray matter volumes, white matter hyperintensity volume, white matter integrity, lacunar infarcts, and β-amyloid deposition) are most affected? We also aimed to assess whether short-term midlife cognitive decline was associated with late-life brain imaging pathology.
Methods
Study Design
This study used data collected during the Atherosclerosis Risk in Communities (ARIC) observational cohort study and ARIC Neurocognitive study (ARIC-NCS).[13] Briefly, the ARIC cohort study recruited 15,792 middle-to-late aged participants (44–66 years old) from four communities in the US (Washington County, Maryland; Forsyth County, North Carolina; Jackson, Mississippi; and selected Minneapolis suburbs, Minnesota). Demographic, clinical, and cognitive data were used from visits 2 (1990–1992), 4 (1996–1998) and 5 (2011–2013); imaging data were used from the visit 5 neurocognitive study (2011–2013).
Inclusion and Exclusion Criteria
General inclusion criteria were: 1) completed the visit 5 brain magnetic resonance imaging (MRI) study (n=1,979); 2) had cognitive factor scores for visits 2, 4, and 5 (n=1,847); 3) were not diagnosed with dementia at visit 5 (n=1750). The following exclusion criteria were applied due to small race-center sample sizes: 1) participants in Maryland or Minnesota who were not White; recruitment in Mississippi was limited to African Americans only; participants in North Carolina who were not White or African American (n=14 excluded); and 2) participants whose level of education was unknown (n=2 excluded).
Cognitive Change
We used change in standardized (z-score) cognitive factor score (CFS) as previously defined by the ARIC study protocol.[14,15] Change in CFS was assessed over two time periods: 1) change from visit 2 to visit 5 (~21 years), and 2) change from visit 2 to visit 4 (~6 years). See the Supplemental Methods for detailed information on the derivation of cognitive change variables.
Brain Imaging
See the Supplemental Methods for information on acquisition of brain imaging. Brain characteristics for this study were assessed at ARIC-NCS visit 5 and were all imaging based. MRI measures were as follows: total gray matter brain volume (cm3), lobar cortical volume, subcortical gray matter volume, temporal lobe meta region gray matter volume, volumes of 23 other regions (Table, Supplemental Digital Content 1), volume of white matter hyperintensities (cm3), and presence of ≥1 lacunar infarct (3 to 20-mm). The temporal lobe meta region was defined as a cluster of the top performing brain regions best discriminating between amyloid PET positive and amyloid negative cognitively impaired individuals in the Mayo Clinic Study of Aging.[16] It is also a set of regions preferentially affected in Alzheimer’s and other less common neurodegenerative diseases.[17] Diffusion tensor imaging variables included fractional anisotropy and mean diffusivity. The PET measure for this study was elevated global cortical β-amyloid (florbetapir [18F] standardized uptake value ratio (SUVR) >1.2).[18]
Data Analysis
All analyses incorporated inverse probability sampling and non-response weights created by the ARIC coordinating center. Sampling weights were not available for the β-amyloid subpopulation. Multivariable linear regression estimated the mean difference in continuous brain measures associated with each 1-SD greater decline in the estimated CFS. Multivariable logistic regression estimated the adjusted odds of having ≥1 lacune or elevated β-amyloid. All models were adjusted for race-center, age, sex, and education. Models examining brain volumes, white matter hyperintensities, fractional anisotropy, mean diffusivity, and β-amyloid were also adjusted for estimated total intracranial volume (cm3). Analyses examining elevated global cortical β-amyloid were adjusted for race (White, Black) due to small race-center sample sizes in the PET subpopulation. See the Supplemental Methods for information on effect measure modification and sensitivity analyses. All statistical methods were agreed upon by multiple authors with an expertise in biostatistical methods. Hypothesis testing was two-sided with an alpha of 0.05, with the exception of the interaction analyses, which had alpha of 0.001. SAS 9.4 (Cary, NC) was used for all analyses.
Results
1734 (88%) participants were included in the final analytical sample. On average, participants at visit 2 were middle-aged (55.6 years), majority female (60%), majority White (74%), and a majority had more than a high school education (56%, Table 1). The average (SD) observed CFS at visits 2, 4, and 5 were 0.18 (0.78), 0.07 (0.74), and −0.72 (0.83). Scatterplots relating observed visit 2 CFS data to observed CFS at visits 4 and 5 are shown in Figures 1 and 2.
Table 1.
Baseline demographic differences by decline in estimated cognitive factor scores (V5-V2).
| n (%) | Overall | Decline in estimated CFS (V5-V2) | |
|---|---|---|---|
|
|
|
||
| (N=1734) | ≤ Median Decline (n=867) | > Median Decline (n=867) | |
|
| |||
| Change in eCFS, mean (SD) | −0.90 (0.30) | −0.67 (0.16) | −1.14 (0.21) |
| Age, yrs, mean (SD) | 55.6 (5.21) | 54.49 (4.99) | 56.75 (5.19) |
| Age, yrs | |||
| ≤55 | 902 (52%) | 525 (61%) | 377 (43%) |
| >55 | 832 (48%) | 342 (39%) | 490 (57%) |
| Male | 693 (40%) | 414 (48%) | 279 (32%) |
| Race-Center | |||
| Black, Forsyth | 30 (2%) | 16 (2%) | 14 (2%) |
| Black, Jackson | 426 (25%) | 205 (24%) | 221 (25%) |
| White, Forsyth | 414 (24%) | 203 (23%) | 211 (24%) |
| White, Minneapolis | 413 (24%) | 212 (24%) | 201 (23%) |
| White, Washington | 451 (26%) | 231 (27%) | 220 (25%) |
| Education | |||
| ≤ High school | 764 (44%) | 360 (42%) | 404 (47%) |
| > High school | 970 (56%) | 507 (58%) | 463 (53%) |
| V2 eCFS, mean (SD) | 0.18 (0.78) | 0.14 (0.68) | 0.22 (0.71) |
| V5 eCFS, mean (SD) | −0.72 (0.83) | −0.53 (0.69) | −0.91 (0.72) |
Abbreviations: eCFS, estimated cognitive factor score; SD, standard deviation.
Median change in estimated cognitive factor score = −0.88
Fig 1.
Relationship between observed cognitive factor scores at visit 2 and 4
Fig. 2.
Relationship between observed cognitive factor scores at visit 2 and 5
Long-term Decline in Cognitive Factor Scores
On average (SD), participants showed a 21-year decline of nearly 1-SD unit in estimated CFS from visit 2 to visit 5 (−0.90 (0.30)). When the demographic characteristics of the study population were compared between those with less than the median change in estimated CFS from visit 2 to visit 5, to those with at least the median value, the only notable differences were in age and sex. Participants with less than the median change in estimated CFS (i.e., more decline) were more often female and of older average age (Table 1).
The mean (SD) brain volume at visit 5 was 1015 (108) cm3 (Table 2). According to the linear regression model, after adjusting for age, sex, race, education, and intracranial volume, for every 1-SD greater decline in estimated CFS, on average total brain volumes were smaller by 12.5 cm3 (Table 2); this association was found for all lobes and the subcortical gray region. As a percentage of the mean regional brain volume, the largest difference in brain volume per 1-SD decline in estimated CFS was observed in the temporal lobe (1.77%) and temporal lobe meta region (1.90%). Brain regions associated with memory showed the largest differences in brain volume, relative to their mean volume (see Table, Supplemental Digital Content 1).
Table 2.
Differences in outcomes per standard deviation greater long-term decline in estimated cognitive factor score (V5-V2, 20 yrs)
| Linear Regression | Overall | Adjusteda | |
|---|---|---|---|
|
| |||
| Mean (SD) | Mean Difference (95% CI) | Mean Difference / Mean, (%) | |
|
| |||
| Total brain volume (cm3) | 1015.31 (108.16) | −12.49 (−14.85, −10.13) | −1.23% |
| Frontal lobe | 150.50 (15.93) | −1.89 (−2.34, −1.44) | −1.26% |
| Parietal lobe | 106.15 (12.50) | −1.33 (−1.68, −0.98) | −1.25% |
| Temporal lobe | 102.19 (11.40) | −1.81 (−2.15, −1.47) | −1.77% |
| Occipital lobe | 40.42 (5.53) | −0.58 (−0.77, −0.39) | −1.43% |
| Subcortical gray | 42.72 (4.24) | −0.27 (−0.41, −0.12) | −0.62% |
| Temporal lobe meta region, volume (cm3) b |
68.13 (8.33) | −1.30 (−1.55, −1.04) | −1.90% |
| log2(WMH volume, cm3) | 3.55 (1.28) | 0.17 (0.12, 0.23) c | − |
| White matter fractional anisotropy | |||
| Overall | 0.36 (0.03) | −0.0040 (−0.0054, −0.0027) | − |
| Frontal lobe | 0.33 (0.03) | −0.0036 (−0.0052, −0.0021) | − |
| Parietal lobe | 0.42 (0.04) | −0.0031 (−0.0049, −0.0013) | − |
| Temporal lobe | 0.36 (0.03) | −0.0051 (−0.0067, −0.0035) | − |
| Occipital lobe | 0.30 (0.04) | −0.0052 (−0.0072, −0.0032) | − |
| log2(white matter mean diffusivity) | |||
| Overall | −10.04 (0.12) | 0.0235 (0.0187, 0.0284) | − |
| Frontal lobe | −10.09 (0.14) | 0.0255 (0.0199, 0.0310) | − |
| Parietal lobe | −10.01 (0.14) | 0.0219 (0.0163, 0.0276) | − |
| Temporal lobe | −9.98 (0.14) | 0.0243 (0.0181, 0.0305) | − |
| Occipital lobe | −10.07 (0.13) | 0.0184 (0.0128, 0.0241) | − |
| log2(gray matter mean diffusivity) | |||
| Overall | −9.81 (0.12) | 0.0191 (0.0130, 0.0252) | − |
| Frontal lobe | −9.82 (0.19) | 0.0191 (0.0098, 0.0283) | − |
| Parietal lobe | −9.76 (0.16) | 0.0156 (0.0070, 0.0243) | − |
| Temporal lobe | −9.80 (0.13) | 0.0201 (0.0145, 0.0257) | − |
| Occipital lobe | −9.81 (0.14) | 0.0187 (0.0125, 0.0248) | − |
|
| |||
| Logistic Regression | n (%) | AOR (95%CI) d | |
|
| |||
| ≥1 lacunar infarct | 307 (18%) | 1.30 (1.18, 1.43) | |
| Elevated β-amyloid (>1.2 SUVR) | 165/326 (51%) | 1.68 (1.29, 2.18) | |
Abbreviations SD, standard deviation; WMH, white matter hyperintensity; CI, confidence interval; AOR, adjusted odds ratio. All volumes measured in cubic centimeters.
Adjusted for race-center, sex, age, educational level, and intracranial volume.
Temporal lobe meta region is preferentially affected in Alzheimer’s and other less common neurodegenerative diseases.
20.17 = 1.13; For each standard deviation greater decline in estimated cognitive factor score, there was a 13% larger volume of WMH.
Adjusted for race-center, sex, education level, and age. β-amyloid model adjusted for race (Black, White) not race-center and for estimated total intracranial volume.
According to the adjusted linear regression model for white matter hyperintensities, a 1-SD greater decline in estimated CFS was associated with a 13% larger volume of white matter hyperintensities (95%CI: 8.7%, 17.3%). Based on adjusted models for white matter, larger decreases in estimated CFS from visit 2 to visit 5 were associated with smaller values of white matter fractional anisotropy and larger values of mean diffusivity; this was found for the overall brain and individual lobes (Table 2). Gray matter mean diffusivity values for the overall brain and lobes were also larger for each 1-SD greater decline in estimated CFS.
At visit 5, 18% of participants presented with at least one lacune measuring 3–20 mm (Table 2). When examining the odds of presenting with at least one lacune or with elevated global cortical β-amyloid at visit 5, multivariable logistic regression suggested that for each 1-SD greater decline in estimated CFS the odds were higher.
Table, Supplemental Digital Content 2 displays the results of the interaction analysis for the change in estimated CFS from visit 2 to visit 5. These analyses suggested the relationships between change in estimated CFS and the primary brain measures did not significantly differ by age, education, race, sex, or for visit 2 CFS. Table, Supplemental Digital Content 3 shows that the results of the main analyses were robust to various exclusion criteria and modelling strategies including adjustment for vascular and genetic risk factors for dementia and Alzheimer’s disease.
Early Short-term Decline in Cognitive Factor Scores
The study sample average (SD) change in observed CFS from visit 2 to visit 4 was −0.11 (0.42). Across the six-years from visit 2 to visit 4, each 1-SD greater decline in observed CFS was associated with smaller total brain volume; the largest difference in volume, as a percentage of the average brain region volume, was observed in the temporal lobe meta region (Table 3). There was no association between change in CFS and subcortical gray or parietal lobe volumes. A greater visit 2 to visit 4 decline in CFS was not associated with a larger volume of white matter hyperintensities or with larger values of fractional anisotropy. White and gray matter mean diffusivity values were significantly larger for overall brain, and all lobes, for each 1-SD greater early decline in CFS, with the exception of the frontal lobe. Finally, a greater decline in CFS from visit 2 to visit 4 was not associated with significant differences in the odds of having at least one 3–20-mm lacune or with having elevated β-amyloid at visit 5.
Table 3.
Differences in outcomes per standard deviation greater short-term decline in estimated cognitive factor score (V4-V2, 6 yrs)
| Linear Regression | Overall | Adjusteda | |
|---|---|---|---|
|
| |||
| Mean (SD) | Mean Difference (95% CI) | Mean Difference / Mean, (%) | |
|
| |||
| Total brain volume (cm3) | 1015.31 (108.16) | −4.26 (−7.02, −1.50) | −0.42% |
| Frontal lobe | 150.50 (15.93) | −0.68 (−1.21, −0.16) | −0.45% |
| Parietal lobe | 106.15 (12.50) | −0.40 (−0.80, 0.01) | −0.38% |
| Temporal lobe | 102.19 (11.40) | −0.67 (−1.06, −0.27) | −0.65% |
| Occipital lobe | 40.42 (5.53) | −0.23 (−0.44, −0.01) | −0.57% |
| Subcortical gray | 42.72 (4.24) | 0.00 (−0.17, 0.16) | 0.00% |
| Temporal lobe meta region, volume (cm3) b | 68.13 (8.33) | −0.51 (−0.81, −0.21) | −0.75% |
| log2(WMH volume, cm3) | 3.55 (1.28) | 0.04 (−0.02, 0.11) | − |
| White matter fractional anisotropy | |||
| Overall | 0.36 (0.03) | −0.0001 (−0.0017, 0.0015) | − |
| Frontal lobe | 0.33 (0.03) | −0.0009 (−0.0027, 0.0009) | − |
| Parietal lobe | 0.42 (0.04) | 0.0010 (−0.0010, 0.0030) | − |
| Temporal lobe | 0.36 (0.03) | −0.0002 (−0.0021, 0.0016) | − |
| Occipital lobe | 0.30 (0.04) | 0.0002 (−0.0021, 0.0025) | − |
| log2(white matter mean diffusivity) | |||
| Overall | −10.04 (0.12) | 0.0078 (0.0021, 0.0134) | − |
| Frontal lobe | −10.09 (0.14) | 0.0063 (−0.0002, 0.0128) | − |
| Parietal lobe | −10.01 (0.14) | 0.0067 (0.0001, 0.0132) | − |
| Temporal lobe | −9.98 (0.14) | 0.0123 (0.0052, 0.0195) | − |
| Occipital lobe | −10.07 (0.13) | 0.0065 (0.0000, 0.0131) | − |
| log2(gray matter mean diffusivity) | |||
| Overall | −9.81 (0.12) | 0.0093 (0.0022, 0.0163) | − |
| Frontal lobe | −9.82 (0.19) | 0.0046 (−0.0061, 0.0152) | − |
| Parietal lobe | −9.76 (0.16) | 0.0107 (0.0009, 0.0205) | − |
| Temporal lobe | −9.80 (0.13) | 0.0112 (0.0047, 0.0177) | − |
| Occipital lobe | −9.81 (0.14) | 0.0146 (0.0075, 0.0216) | − |
|
| |||
| Logistic Regression | n (%) | AOR (95%CI) c | |
|
| |||
| ≥1 lacunar infarct | 307 (18%) | 1.08 (0.97, 1.21) | |
| Elevated β-amyloid (>1.2 SUVR) | 165/326 (51%) | 1.30 (0.97, 1.75) | |
SD, standard deviation; WMH, white matter hyperintensity; AOR, adjusted odds ratio; CI, confidence interval; SUVR, standardized uptake value ratio. All volumes measured in cubic centimeters.
Adjusted for race-center, sex, education level, age, and intracranial volume.
Temporal lobe meta region is preferentially affected in Alzheimer’s and other less common neurodegenerative diseases.
Adjusted for race-center, sex, education level, and age. β-amyloid model adjusted for race (Black, White) and estimated total intracranial volume.
Discussion
By presenting the association of 20-year pre-dementia cognitive decline with a broad range of late-life brain morphology measures in a large population-based prospective four-community cohort study of older adults of Black and White race in the United States, this study provides a comprehensive description of the patterns of cognitively relevant brain changes expected in populations without dementia. The changes found are suggestive of both Alzheimer’s and vascular pathology. Importantly, the observed patterns did not significantly differ by age, race, sex, or baseline level of cognitive performance. Additionally, short-term cognitive decline in midlife was also significantly associated with brain morphology nearly 15 years later.
The temporal lobe and temporal lobe meta region were the regions with the largest percent difference from average in brain volume related to cognitive decline, suggesting that differences in these regions contribute more to the overall cognitive loss in the general population than other cortical regions.[3,19–21] Leong and colleagues showed similar findings when combining longitudinal brain imaging with longitudinal cognitive assessments in an East Asian cohort study over eight years.[22] Annual percent decrease in brain volume was largest for the temporal lobe and hippocampus, and they also found that greater memory decline was significantly associated with greater loss of hippocampal volume; greater loss of temporal lobe volume was only associated with greater decline in global cognition, not memory. Even in a smaller study (n=297) examining 2-year declines in cognition and brain volumes, the most significant positive associations were found in the temporal lobe and the subcortical structures of the thalamus, corpus callosum, and posterior cingulate gyrus.[8] These studies all indicate a clear and sensitive relationship between the temporal lobe volume and cognition.
White matter hyperintensities, lacunes, and reduced subcortical gray matter volume are cognitively relevant characteristics that may be more sensitive to vascular pathology (i.e., cerebral small vessel disease and arteriolosclerosis), than are the cortical lobar volumes.[23–26] The current study suggests that greater 21-year cognitive decline was associated with significantly greater WMH volume and more lacunes, but only a small difference in subcortical gray matter volume. If WMH volume and lacunes were primarily the result of worsening brain microvasculature, subcortical gray matter would have also been expected to show differences in relative volumes, however, it showed one of the smallest relative volume differences among all regions examined. The reason for this is unknown. Studies examining WMH and lacunes have shown inconsistent relationships with decreased cognition, with some showing significant associations[25,27,28] but others showing none.[23,29,30] This incongruence is likely attributable to varying study designs with different temporal relationships between the measurements of WHM and cognition.
Measurement of elevated β-amyloid deposition in the brain identifies individuals with neurological disease who may not yet exhibit cognitive symptoms.[31] The current study observed that sharper long-term declines in cognition were associated with significantly higher odds of having elevated β-amyloid levels in later life, consistent with other studies of cognitive decline and risk factors for dementia and Alzheimer’s disease.[32–34] It is clear that the late-life state of the brain vis-à-vis gray matter volumes, WMH volume, white matter integrity, lacunes, and β-amyloid deposition are all preceded by long-term declines in cognition, with sharper declines in cognition relating to significantly worse levels of each of these measures.
In addition to 21-year cognitive decline, the current study also assessed whether 6-year midlife cognitive decline was associated with differences in brain morphology 15 years later. Despite the variability and small magnitude, more rapid short-term declines in midlife cognition were associated with smaller brains and higher mean diffusivity values 15 years later; a testament to the integrative and sensitive nature of the cognitive performance test scores used. Of particular interest was how short-term midlife cognitive decline was associated with the temporal lobe meta region, which showed the largest difference in volume, relative to average volume, among all brain regions examined. Not unlike the long-term findings, subcortical gray volumes were not significantly associated with steeper short-term declines in midlife cognition. The current study was underpowered to detect a significant difference in the odds of having an elevated β-amyloid for every 1-SD larger decline in 6-year cognition; yet the odds ratio estimate remained considerably above one.
This study is not without limitations. First, all brain features were measured at single time, precluding us from relating changes in cognitive performance to changes in brain morphology. However, we can reasonably presume that deleterious brain features (e.g., WMH and lacunes) found in late life were uncommon or not present 20 years prior to brain imaging. Second, measuring CFS has error as evidenced by the variability in change in CFS over just 6 years from visit 2 to visit 4. To address the variability and possible measurement error, a linear mixed effects model was used to estimate individual cognitive performance trajectories based on a combination on individual trajectories and study sample average trajectories. This methodology decreased variance at the expense of bias, as individual cognitive trajectories with larger variance were biased towards the average trajectory. Thirdly, β-amyloid data were limited to a subsample of 326 non-demented MRI participants; because no analytical weights were available for the β-amyloid analysis, caution should be used when generalizing these specific findings to the overall cohort. Finally, we must acknowledge the possibility for unmeasured or residual confounding to affect our results. The main regression models only adjusted for demographic factors (age, sex, race/center, and education) because the study objective was to describe the population’s overall associations of cognitive decline with a variety of MRI differences. We opine that adjustment for additional comorbid factors (e.g., hypertension, smoking) likely causally related to the true biologically-relevant exposure variables (all examined brain variables), is not appropriate. Notwithstanding these limitations, the strengths of the current study were its two-decade-long longitudinal design, with a large sample size where each participant was their own control, robust and validated data collection, and strong statistical analyses.
In conclusion, this large, multi-site community-based prospective cohort study of non-demented Black and White participants showed that sharper long-term decreases in midlife cognition were significantly associated with worse imaging features across all eight brain outcomes examined. Importantly, these long-term relationships were foreshadowed by short-term declines in midlife cognition, and were highly robust to a variety of exclusions and modelling strategies. Declining cognitive functioning in midlife and late-life is a clear surrogate marker for worsening brain pathology, even in those who have not developed dementia.
Data Availability Statement
Data used for this study are publicly available by request at https://biolincc.nhlbi.nih.gov/studies/aric/
Supplementary Material
Acknowledgement
The authors thank the staff and participants of the ARIC study for their important contributions.
Funding Sources
The Atherosclerosis Risk in Communities study has been funded in whole or in part with Federal funds from the National Heart, Lung, and Blood Institute, National Institutes of Health, Department of Health and Human Services, under Contract nos. (HHSN268201700001I, HHSN268201700002I, HHSN268201700003I, HHSN268201700005I, HHSN268201700004I). ARIC Neurocognitive (ARIC-NCS) The Atherosclerosis Risk in Communities Study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute contracts (HHSN268201700001I, HHSN268201700002I, HHSN268201700003I, HHSN268201700005I, HHSN268201700004I). Neurocognitive data is collected by U01 2U01HL096812, 2U01HL096814, 2U01HL096899, 2U01HL096902, 2U01HL096917 from the NIH (NHLBI, NINDS, NIA and NIDCD), and with previous brain MRI examinations funded by R01-HL70825 from the NHLBI. The ARIC-PET study is funded by the National Institute on Aging (R01AG040282). Avid Radiopharmaceuticals provided the florbetapir isotope for the study, but had no role in the study design or interpretation of results. Author AO was supported by a T32 Ruth L. Kirschstein National Research Service Award (NRSA) Institutional Research Training Grant.
Footnotes
Statement of Ethics
This study was approved by the Johns Hopkins School of Public Health Institutional Review Board (IRB number: 12998). Written informed consent was obtained from all participants at each visit.
Conflict of Interest
All authors have no conflicts of interest to declare.
Author disclaimer: This article was partially prepared while Dr. Rebecca Gottesman was employed at the Johns Hopkins University School of Medicine. The opinions expressed in this article are the author’s own and do not reflect the view of the National Institutes of Health, the Department of Health and Human Services, or the United States Government.
REFERENCES
- 1.Bateman RJ, Xiong C, Benzinger TLS, Fagan AM, Goate A, Fox NC, et al. Clinical and biomarker changes in dominantly inherited Alzheimer’s disease. N Engl J Med. 2012. Aug;367(9):795–804. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Rajan KB, Wilson RS, Weuve J, Barnes LL, Evans DA. Cognitive impairment 18 years before clinical diagnosis of Alzheimer disease dementia. Neurology. 2015. Sep;85(10):898–904. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Dawe RJ, Yu L, Arfanakis K, Schneider JA, Bennett DA, Boyle PA. Late-life cognitive decline is associated with hippocampal volume, above and beyond its associations with traditional neuropathologic indices. Alzheimer’s Dement. 2020. Jan;16(1):209–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Hedman AM, van Haren NEM, Schnack HG, Kahn RS, Hulshoff Pol HE. Human brain changes across the life span: A review of 56 longitudinal magnetic resonance imaging studies. Hum Brain Mapp. 2012;33(8):1987–2002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Sowell ER, Thompson PM, Toga AW. Mapping changes in the human cortex throughout the span of life. Neuroscientist. 2004;10(4):372–92. [DOI] [PubMed] [Google Scholar]
- 6.Oschwald J, Guye S, Liem F, Rast P, Willis S, Röcke C, et al. Brain structure and cognitive ability in healthy aging: a review on longitudinal correlated change. Rev Neurosci. 2019;31(1):1–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Charlton RA, Schiavone F, Barrick TR, Morris RG, Markus HS. Diffusion tensor imaging detects age related white matter change over a 2 year follow-up which is associated with working memory decline. J Neurol Neurosurg Psychiatry. 2010;81(1):13–9. [DOI] [PubMed] [Google Scholar]
- 8.Fletcher E, Gavett B, Harvey D, Farias ST, Olichney J, Beckett L, et al. Brain volume change and cognitive trajectories in aging. Neuropsychology. 2018;32(4):436–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Valech N, Sánchez-Benavides G, Tort-Merino A, Coll-Padrós N, Olives J, León M, et al. Associations Between the Subjective Cognitive Decline-Questionnaire’s Scores, Gray Matter Volume, and Amyloid-β Levels. J Alzheimer’s Dis. 2019. Dec;72(4):1287–302. [DOI] [PubMed] [Google Scholar]
- 10.Caillaud M, Hudon C, Boller B, Brambati S, Duchesne S, Lorrain D, et al. Evidence of a Relation Between Hippocampal Volume, White Matter Hyperintensities, and Cognition in Subjective Cognitive Decline and Mild Cognitive Impairment. J Gerontol B Psychol Sci Soc Sci. 2020. Oct;75(7):1382–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Schneider ALC, Senjem ML, Wu A, Gross A, Knopman DS, Gunter JL, et al. Neural correlates of domain-specific cognitive decline. Neurology. 2019. Mar;92(10):e1051–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Chen X, Farrell ME, Rundle MM, Chan MY, Moore W, Wig GS, et al. The relationship of functional hippocampal activity, amyloid deposition, and longitudinal memory decline to memory complaints in cognitively healthy older adults. Neurobiol Aging. 2021;105:318–26. [DOI] [PubMed] [Google Scholar]
- 13.The ARIC Investigators. The Atherosclerosis Risk In Communities (ARIC) Sudy: Design and Objectives. Am J Epidemiol. 1989. [PubMed] [Google Scholar]
- 14.Rawlings AM, Bandeen-Roche K, Gross AL, Gottesman RF, Coker LH, Penman AD, et al. Factor structure of the ARIC-NCS neuropsychological battery: An evaluation of invariance across vascular factors and demographic characteristics. Psychol Assess. 2016. DOI: 10.1037/pas0000293 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Gross AL, Power MC, Albert MS, Deal JA, Gottesman RF, Griswold M, et al. Application of Latent Variable Methods to the Study of Cognitive Decline When Tests Change over Time. Epidemiology. 2015. Nov;26(6):878–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Schwarz CG, Gunter JL, Wiste HJ, Przybelski SA, Weigand SD, Ward CP, et al. A large-scale comparison of cortical thickness and volume methods for measuring Alzheimer’s disease severity. NeuroImage Clin. 2016. DOI: 10.1016/j.nicl.2016.05.017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Nelson PT, Dickson DW, Trojanowski JQ, Jack CR, Boyle PA, Arfanakis K, et al. Limbic-predominant age-related TDP-43 encephalopathy (LATE): Consensus working group report. Brain. 2019. DOI: 10.1093/brain/awz099 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Gottesman RF, Schneider ALC, Zhou Y, Chen X, Green E, Gupta N, et al. The ARIC-PET amyloid imaging study: Brain amyloid differences by age, race, sex, and APOE. Neurology. 2016. Aug;87(5):473–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Pelletier A, Bernard C, Dilharreguy B, Helmer C, Le Goff M, Chanraud S, et al. Patterns of brain atrophy associated with episodic memory and semantic fluency decline in aging. Aging (Albany NY). 2017. DOI: 10.18632/aging.101186 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Zhang H, Sachdev PS, Wen W, Kochan NA, Crawford JD, Brodaty H, et al. Gray matter atrophy patterns of mild cognitive impairment subtypes. J Neurol Sci. 2012. DOI: 10.1016/j.jns.2011.12.011 [DOI] [PubMed] [Google Scholar]
- 21.Chen KHM, Chuah LYM, Sim SKY, Chee MWL. Hippocampal region-specific contributions to memory performance in normal elderly. Brain Cogn. 2010. DOI: 10.1016/j.bandc.2009.11.007 [DOI] [PubMed] [Google Scholar]
- 22.Leong RLF, Lo JC, Sim SKY, Zheng H, Tandi J, Zhou J, et al. Longitudinal brain structure and cognitive changes over 8 years in an East Asian cohort. Neuroimage. 2017;147(August 2016):852–60. [DOI] [PubMed] [Google Scholar]
- 23.Moon SY, de Souto Barreto P, Chupin M, Mangin JF, Bouyahia A, Fillon L, et al. Associations between white matter hyperintensities and cognitive decline over three years in non-dementia older adults with memory complaints. J Neurol Sci. 2017;379:266–70. [DOI] [PubMed] [Google Scholar]
- 24.Van Der Flier WM, Van Straaten ECW, Barkhof F, Verdelho A, Madureira S, Pantoni L, et al. Small vessel disease and general cognitive function in nondisabled elderly: The LADIS study. Stroke. 2005. DOI: 10.1161/01.STR.0000179092.59909.42 [DOI] [PubMed] [Google Scholar]
- 25.Bilello M, Doshi J, Nabavizadeh SA, Toledo JB, Erus G, Xie SX, et al. Correlating Cognitive Decline with White Matter Lesion and Brain Atrophy Magnetic Resonance Imaging Measurements in Alzheimer’s Disease. J Alzheimer’s Dis. 2015. Oct;48(4):987–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Debette S, Markus HS. The clinical importance of white matter hyperintensities on brain magnetic resonance imaging: systematic review and meta-analysis. BMJ. 2010. Jul;341:c3666. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Kloppenborg RP, Nederkoorn PJ, Geerlings MI, van den Berg E. Presence and progression of white matter hyperintensities and cognition: A meta-analysis. Neurology. 2014. Jun;82(23):2127–38. [DOI] [PubMed] [Google Scholar]
- 28.Debette S, Markus HS. The clinical importance of white matter hyperintensities on brain magnetic resonance imaging: Systematic review and meta-analysis. BMJ. 2010;341(7767):288. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Vibha D, Tiemeier H, Mirza SS, Adams HHH, Niessen WJ, Hofman A, et al. Brain Volumes and Longitudinal Cognitive Change. Alzheimer Dis Assoc Disord. 2018;32(1):43–9. [DOI] [PubMed] [Google Scholar]
- 30.Blom K, Koek HL, Zwartbol MHT, van der Graaf Y, Kesseler L, Biessels GJ, et al. Subjective cognitive decline, brain imaging biomarkers, and cognitive functioning in patients with a history of vascular disease: the SMART-Medea study. Neurobiol Aging. 2019;84:33–40. [DOI] [PubMed] [Google Scholar]
- 31.Sperling RA, Aisen PS, Beckett LA, Bennett DA, Craft S, Fagan AM, et al. Toward defining the preclinical stages of Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimer’s Dement. 2011. DOI: 10.1016/j.jalz.2011.03.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Rabin JS, Schultz AP, Hedden T, Viswanathan A, Marshall GA, Kilpatrick E, et al. Interactive associations of vascular risk and β-amyloid burden with cognitive decline in clinically normal elderly individuals findings from the Harvard Aging Brain Study. JAMA Neurol. 2018;75(9):1124–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Betthauser TJ, Koscik RL, Jonaitis EM, Allison SL, Cody KA, Erickson CM, et al. Amyloid and tau imaging biomarkers explain cognitive decline from late middle-age. Brain. 2020;143(1):320–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Lopez OL, Becker JT, Chang Y, Klunk WE, Mathis C, Price J, et al. Amyloid deposition and brain structure as long-term predictors of MCI, dementia, and mortality. Neurology. 2018;90(21):e1920–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data used for this study are publicly available by request at https://biolincc.nhlbi.nih.gov/studies/aric/