Abstract
Background/Objective:
In this study, we investigated a long-term trajectory of brain aging (from the 20 s to over-80) in cognitively normal (CN) individuals. We further determined whether differences in sex, education years, and apolipoprotein E ε 4 status affect age-related cortical thinning.
Methods:
A total of 2,944 CN individuals who underwent high-resolution (3.0-Tesla) magnetic resonance imaging were included in this study. Cortical thickness was measured using a surface-based method. Multiple linear regression analyses were performed to evaluate age-related cortical thinning and related factors.
Results:
Compared to those in their 20 s/30 s, participants in their 40 s showed thinning primarily in the medial and lateral frontal and inferior parietal regions, and cortical thinning occurred across most of the cortices with increasing age. Notably, the precuneus, inferior temporal and lateral occipital regions were relatively spared until later in life. Male and lower education years were associated with greater cortical thinning with distinct regional specificity.
Conclusion:
Our findings provide an important clue to understanding the mechanism of age-related cognitive decline and new strategies for preventing the acceleration of pathological brain aging.
INTRODUCTION
Cognitive aging is defined as a process of gradual, ongoing, yet highly variable changes in cognitive function that occur as individuals age and in the absence of frank neurodegenerative disease [1]. Specifically, declines in memory, conceptual reasoning, and processing speed are commonly observed in the elderly. Furthermore, these cognitive functions decrease as aging progresses. In contrast, some cognitive abilities, such as vocabulary, seem to be resilient to aging [2, 3]. There are several known factors that affect cognitive aging. Previous studies have shown that age-related cognitive function tends to be maintained more in women [4, 5] and apolipoprotein E (APOE) ε 4 non-carriers [6, 7], whereas the benefit of a high education level is controversial [8–10].
The measurement of cortical thickness is a useful surrogate marker of physiological brain aging. Neuroimaging studies have suggested that aging prominently affects cortical thickness even in cognitively normal (CN) individuals, which is referred to as age-related cortical thinning [11, 12]. In addition, several neuroimaging studies have reported that all regions of the cortex were not equally affected by aging, with prominent atrophy of the prefrontal cortex and relative sparing of the temporal and parahippocampal cortices [12–16]. However, these studies had small sample sizes [12, 13] and used less sophisticated imaging methods such as relatively low-field magnetic resonance imaging (MRI) [14, 15]. In addition, findings were inconsistent among these previous studies [14, 16].
Previous studies have shown that several factors affect brain structures even in CN individuals. For example, women had greater cortical thickness in the temporal and parietal regions than men, [17] high level of education was correlated with increased mean cortical thickness throughout the entire cortex, [18] and APOE ε 4 carriers showed prominent atrophy in the hippocampus and amygdala compared to non-carriers [19]. However, the relationships between these factors and age-related cortical thinning in the brain have not yet been fully established. Therefore, we hypothesized that the above-mentioned factors could have diverse effects on age-related cortical thinning with distinct regional specificity.
Despite growing concerns about brain aging, core questions on brain structures that are vulnerable to or preserved against aging and brain aging-related factors remain a matter of debate because of inconsistent results based on small sample sized participants. Furthermore, the majority of previous studies have conducted volumetric analyses to investigate brain aging. However, cortical thickness measurement using cortical surface has been proposed in studies of gray matter (GM) morphometry as a strategy for overcoming the limitation of volumetric analyses [20, 21]. With the strengths that include the large sample size (2,944 CN individuals) and sophisticated measurements using the same type of scanner with the same scan parameters, we investigated a long-term trajectory of brain aging in people ranging from the 20 s to older than 80. We hypothesized there would be specific regions for physiological brain aging. We further determined whether sex, education years, and APOE ε 4 status affect age-related cortical thinning.
METHODS
Study participants
CN individuals were recruited from the Health Promotion Center of Samsung Medical Center (Seoul, Korea). The study population was comprised of men and women 40 years of age or older who underwent a comprehensive health screening exam from January 1, 2009 to December 31, 2014. There were 3,370 eligible participants who attended a preventative medical check-up, which included an assessment of cognitive function and dementia status. All study participants underwent high-resolution 3.0-Tesla brain MRI, including three-dimensional (3D) volume images, as a part of their dementia assessment. The assessment procedure used for the participants has been described in detail elsewhere [22]. We excluded participants who had any of the following conditions: 202 participants with missing data on education years or Mini–Mental State Examination (MMSE) score; 178 participants with significant cognitive impairment defined by MMSE scores below the 16th percentile in age-, sex-, and education-matched norms or through an interview conducted by a qualified neurologist; and 136 participants with unreliable analyses of cortical thickness due to head motion, blurring of the MRI, inadequate registration to a standardized stereotaxic space, misclassification of tissue type, or inexact surface extraction. We also included 42 participants (27 men and 15 women) who ranged in age from the 20 s to 30 s and had no past medical history and normal cognition as determined by MMSE score. They underwent the same kind of high-resolution 3.0-tesla brain MRI, including 3D volume images. Therefore, the final sample size was 2,944 participants (1,481 men and 1,463 women). Participants were excluded if they had a cerebral, cerebellar, or brainstem infarction; hemorrhage; brain tumor; hydrocephalus; severe cerebral white matter hyperintensities (deep white matter ≥ 2.5 cm and caps or band ≥ 1.0 cm); or severe head trauma by personal history.
Standard protocol approvals, registrations, and patient consent
This study was approved by the Institutional Review Board at Samsung Medical Center. In addition, all methods were carried out in accordance with the approved guidelines. Written informed consent was obtained from all participants prior to study participation.
Brain MRI scans
All study participants underwent neurological and neuropsychological examination, MMSE, and a 3D volumetric brain MRI scan. An Achieva 3.0-Tesla MRI scanner (Philips, Best, the Netherlands) was used to acquire 3D T1 Turbo Field Echo (TFE) MRI data using the following imaging parameters: sagittal slice thickness, 1.0 mm with 50% overlap; no gap; repetition time of 9.9 ms; echo time of 4.6 ms; flip angle of 8° and matrix size of 240×240 pixels reconstructed to 480×480 over a field view of 240 mm.
Cortical thickness measurements
T1-weighted MR images were automatically processed using the standard Montreal Neurological Institute image processing software (CIVET) to measure cortical thickness. This software has been well-validated and extensively described elsewhere including in aging/atrophied brain studies [23, 24]. In summary, native MR images were first registered into a standardized stereotaxic space using an affine transformation [25]. Non-uniformity artifacts were corrected using the N3 algorithm, and the registered and corrected volumes were classified as GM, white matter (WM), cerebrospinal fluid (CSF), and background using an artificial neural net classifier [26]. The surfaces of the inner and outer cortices were automatically extracted by deforming a spherical mesh onto the gray/white boundary of each hemisphere using the Constrained Laplacian-Based Automated Segmentation with Proximities algorithm, which has also been well-validated and extensively described elsewhere [27].
Cortical thickness was calculated as the Euclidean distance between the linked vertices of the inner and outer surfaces after applying an inverse transformation matrix to cortical surfaces and reconstructing them in the native space [20, 27]. To control for brain size, we computed intracranial volume (ICV) using classified tissue information and a skull mask acquired from the T1-weighted image [28]. ICV was defined as the total volume of GM, WM, and CSF, with consideration of the voxel dimension. Classified GM, WM, CSF, and background within the mask were transformed back into the individual native space.
To compare the thicknesses of corresponding regions among the participants, the thicknesses were spatially registered on an unbiased iterative group template by matching the sulcal folding pattern using surface-based registration involving sphere-to-sphere warping [29]. For global and lobar regional analyses, we used the lobe-parcellated group template that had been previously divided into frontal, temporal, parietal, and occipital lobes using SUMA (http://afni.nimh.nih.gov) [20]. Average thickness values of the whole vertex in each hemisphere and lobar region were used for global analysis.
Statistical analysis
For cortical thickness analyses of MRI data, we used a MATLAB-based toolbox (available free online at the University of Chicago website: http://galton.uchicago.edu/faculty/InMemoriam/worsley/research/surfstat/). Diffusion smoothing with a full-width half-maximum of 20 mm was used to blur each cortical thickness map, leading to increased signal-to-noise ratio and statistical power [23]. We entered age (continuous or categorized) as a predictor and vertex-by-vertex cortical thickness as an outcome to analyze the relationship between cortical thickness and age in the surface model. A linear regression analysis was then performed after controlling for sex, education years (continuous), ICV, hypertension, and diabetes mellitus (DM). To assess the significance of age and group (according to sex, education level, or APOE ε 4 status) interactive effects on cortical thickness, a multiple linear regression analysis was also applied after controlling for group (sex, education level, or APOE ε 4 status), ICV, hypertension, and DM. Education level was divided into higher (≥12 years) and lower (<12 years) groups. In this analysis, 42 participants from the 20 s/30 s group were excluded because they were all higher-educated, and there was no information on their APOE genotype (Supplementary Table 1). The cortical surface model contained 81,924 vertices; thus, the resulting statistical maps were thresholded using a false discovery rate (FDR) [30] with a Q-value of 0.05 after pooling the p-values from regression analyses. In addition, we modeled age as a continuous variable using restricted cubic splines with knots at the 5th, 35th, 65th, and 95th percentiles of the sample distributions to provide a flexible estimate of the dose-response relationship between age and cortical thinning on representative regions of interest (ROIs). Statistical analyses were performed using SPSS version 20.0 (SPSS Inc., Chicago, IL, USA).
RESULTS
Characteristics of the study participants
Table 1 summarizes the demographic and clinical characteristics of the study participants. A total of 2,944 CN individuals were included, with a mean (SD) age of 63.2 (8.4) years, ranging from 22 to 91 years, and a mean (SD) education years of 12.8 (4.3) years. The number of participants in the 20 s/30 s/40 s/50 s/60 s/70 s/80 s/90 s was 25/17/79/694/1,509/572/47/1, respectively. Since the numbers of participants in their 20 s and 30 s or 80 s and 90 s were relatively small, we combined these two age groups into the 20 s/30 s or over-80 groups.
Demographic and clinical characteristics of the study participants (N = 2,944)
Values are mean (SD) or N (%). Education level was divided into higher (≥12 years) and lower (<12 years) groups.
APOE genotyping was performed in 673 (22.9%) of the 2,944 CN individuals.
Scores in each row are significantly different in pairwise comparisons (Student’s t- or Chi-square test). N, number; SD, standard deviation; CN, cognitively normal; APOE, apolipoprotein E; MMSE, Mini-Mental State Examination; DM, diabetes mellitus; IHD, ischemic heart disease; ICV, intracranial volume.
Topographical differences in cortical thickness of each age group
We classified the participants into six groups according to age (20 s/30 s, 40 s, 50 s, 60 s, 70 s, and over-80). The mean and SD of cortical thickness for each age group based on sex, education, APOE ε 4 status, hypertension, and DM were presented in supplementary data (Supplementary Tables 2–6). Vertex-wise mean cortical thickness images for each age group were presented in Supplementary Figure 1. Actual differences in cortical thickness images of each age group from the 40 s to over-80 compared to the 20 s/30 s are presented in Fig. 1A. Compared to those in their 20 s/30 s, participants in their 40 s showed cortical thinning primarily in the frontal and inferior parietal regions (Fig. 1B). The topography of cortical thinning was widespread and severe based on age. However, the precuneus, inferior temporal, and lateral occipital regions were relatively preserved against age effects until later in life. This result was consistent with that of the analysis conducted with reference to the over-80 group (Supplementary Figure 2). Results of cortical thickness analysis in right hemisphere were presented in Supplementary Figure 3. In addition, statistical maps for t-value and β-coefficient were presented in Supplementary Figures 4 and 5, respectively.
Topographical differences in cortical thinning of each age group from the 40s to over-80 compared to the 20s/30s in CN individuals. A) Actual difference in cortical thickness. B) FDR-corrected p-value. FDR-corrected (p < 0.05) results were adjusted for sex, education years, ICV, hypertension, and DM. C) Reduction ratios in mean cortical thickness of CN individuals in the seven representative ROIs by age. The curves were estimated using restricted cubic splines for age with knots at the 5th, 25th, 75th, and 95th percentiles of the age sample distributions. The reference value (diamond dot) was set at the 1st percentile. Reduction ratios were expressed as a percentage (%). CN, cognitively normal; FDR, False discovery rate; ICV, intracranial volume; DM, diabetes mellitus; ROIs, regions of interest.
To investigate the pattern of age-related cortical thinning, we selected seven representative ROIs (dorsolateral prefrontal cortex [DLPFC], precuneus, superior parietal lobule, inferior parietal lobule [IPL], medial temporal, inferior temporal, and lateral occipital regions) based on the results of topographical differences in cortical thickness. In spline regression models, increasing age was associated with higher reduction in the ratio of mean cortical thickness overall and in all regions showing distinct patterns of cortical thinning (Fig. 1C). The slopes of the lines on the plots showed a steep reduction in the ratio of mean cortical thickness in the DLPFC and IPL regions with age, even in the 40 s. However, declining slopes were relatively gradual in the precuneus, inferior temporal, and lateral occipital regions. In addition, the slope of the medial temporal region showed a gentle decline until the early 60 s, after which it demonstrated a steep, accelerated decline.
Increased age was associated with decreased MMSE scores on a multiple linear regression analysis after controlling sex, education level, ICV, hypertension, and DM (β= –0.042, standard error = 0.004, p-value <0.001). MMSE was negatively correlated with cortical thickness in the bilateral frontal, parietal, medial surface of occipital, and temporal regions including hippocampal gyri (Supplementary Figure 6).
The interactive effects between age and related factors on cortical thickness
To assess the interactive effects between age and related factors (sex, education level, and APOE ε 4 status) on cortical thickness, multiple regression analyses were performed after controlling for group (sex, education years, or APOE ε 4 status), ICV, hypertension, and DM. As a result, significantly less age-related cortical thinning was observed in the women group in the bilateral DLPFC, medial frontal, lateral temporal, IPL, and posterior cingulate/ precuneus regions (Fig. 2A). We also found significantly less age-related cortical thinning in the higher education group in the bilateral premotor area, left insular, and right medial frontal regions (Fig. 2B). However, there was no interaction between age and APOE ε 4 status. Vertex-wise mean cortical thickness images for each age group based on sex, education level, and APOE ε 4 status were presented in Supplementary Figures 7–9.
The topography of age*group interactive effects on cortical thickness. A) Age* sex interaction after controlling for education years, ICV, hypertension, and DM. B) Age*education level interaction after controlling for sex, ICV, hypertension, and DM. FDR-corrected (p < 0.05). ICV, intracranial volume; DM, diabetes mellitus; FDR, False discovery rate.
DISCUSSION
We assessed age-related cortical thinning in a large population of 2,944 participants, ranging in age from the 20 s to over-80 and determined whether sex, education, and APOE ε 4 status affected age-related cortical thinning. The major findings of the present study were as follows. First, cortical thickness in the DLPFC and IPL was affected by aging earlier in life, but cortical thickness was relatively preserved in the precuneus, inferior temporal, and lateral occipital cortices until later in life. Second, there were interactive effects between age and related factors (sex and education) on age-related cortical thinning with distinct regional specificity. Taken together, our findings provide an important clue to understanding the mechanism of age-related cognitive decline and new strategies for preventing the acceleration of pathological brain aging with development of programs that reinforce vulnerable regions in each subgroup.
Our first major finding was that there were selectively vulnerable brain regions for age-related cortical thinning in CN individuals. Previous studies have reported that aging affected cortical thinning diffusely [12–15] or prominently in the frontal regions, suggesting that cortical thinning may be due to increased cerebrovascular burden [31, 32]. However, we found that age-related cortical thinning was initiated and accelerated in the inferior parietal region as well as in the medial and lateral frontal regions earlier in life (40 s). Our findings are consistent with a well-known theory age-related cognitive decline in which aging is associated with declines in executive processing tasks such as working memory, executive functions, and processing speed [33, 34]. Previous studies demonstrated that inferior parietal regions play important roles in spatial cognition, vocabulary learning, and episodic memory retrieval [35–38]. Therefore, our findings might explain the cognitive deficits in these domains that are commonly observed in the elderly. Our finding that the frontal and inferior parietal regions are vulnerable to aging could be explained by the “last in, first out” hypothesis [13, 39]. That is, late-maturing regions of the brain, such as the heteromodal association cortices, are preferentially vulnerable to age-related loss of structural integrity. Therefore, early thinning in these regions provides a clue to understanding the pathomechanism of age-related cognitive decline.
We found that some regions were relatively preserved against age effects until later in life. In other words, all age groups from the 40 s to over-80 did not show significant cortical thinning in the precuneus, inferior temporal, or lateral occipital regions compared with the younger subjects in the 20 s/30 s. The precuneus has a central role in highly integrated tasks, including episodic memory retrieval, visuospatial processing, and self-consciousness [40]. In addition, the inferior temporal and lateral occipital regions are particularly important for the ventral stream, as the lateral occipital complex, [41, 42] and have been implicated in object recognition and semantic processing [43, 44]. Therefore, our findings support previous studies showing that semantic memory or visual object recognition functions were relatively preserved in the elderly [45–47]. In fact, the precuneus is particularly vulnerable to the early deposition of amyloid [48] and seems to be affected even in the early-stages of AD [49–51]. We therefore suggest that the precuneus might be important regions to discriminate pathological brain aging from physiological brain aging.
Our second major finding was that there were interactive effects between age and related factors (sex and education) on age-related cortical thinning with distinct regional specificity. Women showed significantly less age-related cortical thinning in the bilateral DLPFC, medial frontal, lateral temporal, IPL, and posterior cingulate/precuneus regions. This finding was partially consistent with several prior studies that found that women had thicker cortices in the bilateral middle frontal, inferior parietal, posterior temporal, and cingulate regions [17, 52]. However, caution is needed in interpreting our findings since cognitive functions are the result of complex activities of several brain regions. Indeed, a systematic review described that elderly women showed better performance in tests of episodic memory, whereas elderly men had better visuospatial ability [53]. These sex differences in brain aging could be associated with a number of factors including lifestyle effects, genetic factors, cardiovascular risk factors, or possibly sex hormones [54–56]. Consistent with a previous study from our group, [18] the higher education group was predicted to undergo a better course of age-related cortical thinning in the bilateral premotor area, left insular, and right medial frontal regions. Previous studies reported that cortical volume or thickness in the insular and anterior cingulate regions was related to education [57, 58]. In addition, previous neuropsychological studies have shown that higher levels of education are associated with better performance on attention, verbal fluency, and working memory, which are known to be related to frontal region [59, 60]. Thus, our results support that the protective effects of education on cortical thickness might be mediated by increased resistance to structural loss from aging, rather than by simply providing a fixed advantage in the brain.
The strengths of this study include the large sample size and sophisticated measurements of cortical thickness using the same type of scanner with the same scan parameters across different waves of data collection. However, some limitations should be considered when interpreting the results. First, our study was designed to be cross-sectional, precluding claims of causality. The cross-sectional design does not taken account of individual differences in the process of aging. Second, our participants were recruited from individuals seeking a comprehensive preventive health exam not covered by national medical insurance, which might not be completely representative of the general population. Finally, we did not have molecular imaging or neuropathologic data from the participants. Further studies should be performed to investigate the effects of neurodegenerative pathologies on age-related cortical thinning.
In conclusion, our findings suggest that there are regions of the brain that are vulnerable or resistant to aging. Furthermore, our findings provide information on the trajectory of normal brain aging and identify factors that might affect the trajectory of normal brain aging.
Footnotes
ACKNOWLEDGMENTS
This research was supported by the Brain Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (2016M3C7A1913844), by the NRF grant funded by the Korean government (2015R1C1A2A01053281), by the NRF grant funded by the Korea government (2017R1A2B2005081), by the Korea Ministry of Environment as the Environmental Health Action Program (2014001360002), by a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute, funded by the Ministry of Health & Welfare, Republic of Korea (HI17C1915) and by a fund (2018-ER6203-00) by Research of Korea Centers for Disease Control and Prevention.
