Objectives

Age-related cognitive changes can be influenced by both brain maintenance (BM), which refers to the relative absence over time of changes in neural resources or neuropathologic changes, and cognitive reserve (CR), which encompasses brain processes that allow for better-than-expected behavioral performance given the degree of life-course related brain changes. This study evaluated the effects of age, BM, and CR on longitudinal changes over two visits, 5 years apart, in three cognitive abilities that capture most of age-related variability.

Method

Participants included 254 healthy adults aged 20–80 years at recruitment. Potential BM was estimated using whole brain cortical thickness and white matter mean diffusivity at both visits. Education and IQ (estimated with AMNART) were tested as moderating factors for cognitive changes in the three cognitive abilities.

Results

Consistent with BM—after accounting for age, sex, and baseline performance—individual differences in the preservation of mean diffusivity and cortical thickness were independently associated with relative preservation in the three abilities. Consistent with CR—after accounting for age, sex, baseline performance, and structural brain changes—higher IQ, but not education, was associated with reduced 5-year decline in Reasoning (?=0.387, p=0.002), and education was associated with reduced decline in Speed (?=0.237, p=0.039).

Discussion

These results demonstrate that both CR and BM can moderate cognitive changes in healthy aging and that the two mechanisms can make differential contributions to preserved cognition.

Participants. The participants were drawn from our ongoing studies at Columbia University Irving Medical Center: The Reference Ability Neural Network (RANN) study and the Cognitive Reserve (CR) study (Stern et al., 2018; Stern et al., 2014). Subjects were recruited primarily by randomized market mailing. An initial telephone screening determined whether participants met basic inclusion criteria (i.e., right‐handed, English speaking, no psychiatric or neurological disorders, and normal or corrected‐to‐normal vision). Potentially eligible participants were further screened in person with structured medical and neuropsychological evaluations to ensure that they had no neurological or psychiatric conditions, cognitive impairment, or contraindication for MRI scanning. Global cognitive functioning was assessed with the Mattis Dementia Rating Scale (Lucas et al., 1998), on which a minimum score of 130 was required for retention in the study. In addition, participants who met diagnostic criteria for mild cognitive impairment were excluded. The studies were approved by the Internal Review Board of the College of Physicians and Surgeons of Columbia University. Additional details about procedures can be found in previous reports (Habeck, Gazes, et al., 2016; Stern et al., 2014). This study is currently in the process of completing a five-year follow-up on all participants using the same procedures. The current analysis included 254 participants who were assessed at both baseline and follow-up and had data from at least one of the 23 cognitive tasks that comprise the reference abilities at either time point, in which 209 participants also had pre-post testing on cognitive tasks during fMRI studies (fMRI data not examined in this manuscript).

Procedure. Out-of-scanner tasks were administered in one visit while in-scanner tasks were administered over the course of two 2-hour scanning sessions, with six activation tasks administered in each session. Prior to each scan session, computerized training was administered for the six tasks to be administered during that session. At the completion of training for each task, participants had the option of repeating the training. For all tasks except Picture Naming, responses were differential button presses. During training, responses were on the computer keyboard. During scans, they were made on the LUMItouch response system (Photon Control Company). 

Tasks administered out of scanner. As described in Salthouse (2015), twelve measures were selected from a battery of neuropsychological tests to assess cognitive functioning. Reasoning was assessed with scores on three different tests: WAIS III Block design task, WAIS III Letter–Number Sequencing test, and WAIS III Matrix Reasoning test. For processing speed, the Digit Symbol subtest from the Wechsler Adult Intelligence Scale-Revised (Wechsler, 1981), Part A of the Trail making test, and the Color naming component of the Stroop test were chosen. Three memory measures were based on sub-scores of the Selective Reminding Task (SRT) (Buschke & Fuld, 2011): the long-term storage sub-score, continuous long-term retrieval, and the number of words recalled on the last trial. Vocabulary (not included in analyses reported in this manuscript) was assessed with scores on the vocabulary subtest from the WAIS III, the Wechsler Test of Adult Reading (WTAR), and the American National Adult Reading Test (AMNART) (Grober & Sliwinski, 1991).

Tasks administered in the scanner. As described in Habeck et al. (2016) and Stern et al. (2014) twelve tasks were administered in the scanner and their behavioral performance measures were computed. Reasoning was assessed with the proportion of correct trials from Paper Folding (Ekstrom et al., 1979), Matrix Reasoning (Raven, 1962) and Letter Sets (Ekstrom et al., 1979). For processing speed mean reaction times on accurate trials for Digit Symbol, Letter Comparison, and Pattern Comparison tasks (Salthouse et al., 1991) were used. Memory scores were measured as the proportion of correctly answered questions from Logical Memory, Word Order Recognition, and Paired Associates. Three vocabulary measures (not included in the current analyses) were the proportion of correct responses for Synonyms, Antonyms (Salthouse, 1993), and Picture Naming (Salthouse, 1998). 

Individual difference factors. Years of education was assessed with the Education Questionnaire (Manly et al., 2002)and an IQ score was estimated from the American National Adult Reading Test (Grober & Sliwinski, 1991).

Image acquisition and processing. There were two 2-hour MR imaging sessions to accommodate the twelve fMRI tasks as well as the additional imaging modalities. Relevant to the current study, T1-weighted MPRAGE scan was acquired to determine cortical thickness, with a TE/TR of 3/6.5 ​ms and Flip Angle of 8°, in-plane resolution of 256 x 256, field of view of 25.4 ​× ​25.4 ​cm, and 165–180 slices in axial direction with slice-thickness/gap of 1/0 ​mm. Two diffusion MRI were acquired in 56 directions using the following parameters: b = 800 s/mm2, TE = 69 ms, TR = 11,032 ms, flip angle = 90 degrees, in-plane resolution 112 x 112 voxels, acquisition time = 12 min 56 s, slice thickness = 2 mm (no gap), and 75 slices. In addition, BOLD fMRI for twelve tasks, FLAIR, ASL and a 7-minute resting BOLD scan were acquired but not reported in the current study. A neuroradiologist reviewed each subject's scans. Any significant findings were conveyed to the subject's primary care physician.

Each subject's structural T1 scans were reconstructed using FreeSurfer v5.1 (http://surfer.nmr.mgh.harvard.edu/). This older version was used to maintain consistency across data for the parent study but with visual inspection already performed for every image, the quality of the parcellations was assured. The accuracy of FreeSurfer's subcortical segmentation and cortical parcellation (Fischl et al., 2004) has been reported to be comparable to manual labeling. Each subject's white and gray matter boundaries as well as gray matter and cerebral spinal fluid boundaries were visually inspected slice by slice, manual control points were added when any visible discrepancy was found, and reconstruction was repeated until we reached satisfactory results within every subject. The subcortical structure borders were plotted by TkMedit visualization tools and compared against the actual brain regions. In case of discrepancy, they were corrected manually. Finally, we computed the mean cortical thickness for each participant to be used in group-level analyses.

Diffusion data were analyzed with version 3.0.1 of the software, MRtrix3 (www.mrtrix.org), starting with a set of preprocessing steps to improve data robustness: (1) denoising (Veraart et al., 2016), (2) Gibbs ring correction (Kellner et al., 2016), (3) corrections for motion and eddy currents (FSL eddy) (Andersson & Sotiropoulos, 2016) and (4) bias field correction (Tustison et al., 2010). Diffusion tensor models were estimated for the preprocessed data from which mean diffusivity (MD) was calculated for each participant. To calculate the mean MD for the white matter across the whole brain, each participant’s T1 structural scan was registered to the mean of the non-diffusion weighted images, on which FreeSurfer parcellation was again performed. The white matter mask derived from FreeSurfer was then used to quantify each participant’s mean MD across all white matter in the brain, resulting in one MD value per participant and used in LCSM described below.

The data were analyzed with R.