Impact of Subcortical Hyperintensities on Dual-tasking in Alzheimer’s Disease and Aging

Participants

Participants between ages of 60 and 80 years who were enrolled in the Sunnybrook Dementia Study, a longitudinal study on cognitive impairment including AD and other dementias, were screened for inclusion in this sub-study. Participants had to be able to walk independently for 15 minutes without stopping. Patients with AD had to meet NINDS-ADRDA criteria for probable AD³⁷. Normal controls (NC) were community volunteers who performed within normal limits on all cognitive tests, were functionally independent in all activities of daily living, had no history of neurological or psychological disorder and were in a stable healthy condition. All participants gave informed consent to the protocol which was approved by the Research Ethics Board. Additionally, participants were only screened if they had a brain magnetic resonance imaging (MRI) within a six-month window of the gait assessments, as little change in SH volume was expected over this time period in this particular population.

Patients with a Mini-mental Status Exam (MMSE) score of ≤ 20, history of other neurological disorders or coexistent neurodegenerative conditions such as Lewy Body disease or Parkinson’s Disease, major depression, hip-fractures, significant arthritis, clinically significant joint deformity, recent hip/knee replacement, sedative medication use, dependence on alcohol and/or neuroleptics drugs, use of assistive devices such as cane/walker and presence of significant sensory and/or motor neuropathy on neurological examination were considered exclusionary. MMSE³⁸ and the Mattis Dementia Rating Scale (MDRS) ³⁹were also used to characterize the stage of AD. All participants underwent training on the working memory tasks and those with an accuracy of <80% after 5 trials of the 2-back working memory task were excluded.

Assessments

At the time of gait assessments, data pertaining to previous falls, concomitant medical conditions, cardiovascular risk factors, exercise history, current medications were obtained followed by a physical and neurological examination that included measurement of body-mass index, leg-length, mid-calf girth, blood pressure and resting heart rate. Timed-up-and-go (TUG) test ⁴⁰ is a validated bedside measure of mobility that measures the time required to arise from a chair, walk 3 meters, turn, and return to the seated position. Unified Parkinson’s Disease Rating Scale (UPDRS)⁴¹ is a standardized scale that includes rating severity of parkinsonism. In this study, the TUG, motor sub-scale of the UPDRS and the Tinetti gait scale ⁴² were also scored. Additionally, neuropsychological and neuroimaging data were available to ascertain cognitive, functional and other disease characteristics.

SH severity rating

SH was rated on the Age-related White Matter Change (ARWMC) scale, a four-point scale that assesses SH severity by aggregating scores in five bilateral regions as delineated by Wahlund et al.⁴³. The rating scale was applied to T2-weighted MRI scans obtained within six-months of the gait assessments. All brain images were acquired using a 1.5 T Signa MR imager (GE Medical systems, Milwaukee, WI). Three image sets were acquired in the same imaging session: T1-weighted (axial 3D SPGR with 5ms TE, 35ms TR, 35° flip angle, 1 NEX, 22 × 16.5 cm FOV, 0.859 × 0.859mm in-plane resolution, and 1.2 to 1.4mm slice thickness), proton-density (PD) and T2-weighted images (interleaved axial spin echo, with TEs of 30 and 80 ms, 3s TR, 0.5 NEX, 20 × 20cm FOV, 0.781 × 0.781mm in-plane resolution, and 3mm slice thickness).

Gait assessment

Cadence was measured using footswitches (B&L Engineering) placed in the insoles of participant’s shoes as they walked on a motorized treadmill (Biodex™ RTM400, Biodex Medical Systems, Inc., NY). Foot-switch data was digitized at the rate of 500 samples per second through an analog-to-digital converter. Digitized signals were processed (Labview®, National Instruments, Austin, TX) to measure the cadence.

All participants were instructed to wear comfortable walking shoes for the experiment. Footswitches made of pliable material were placed in both shoes such that the shoes once worn would fit comfortably. A non-weight bearing safety-harness (The Biodex Unweighing System®, Biodex Medical Systems, Inc., NY) was strapped across the trunk for safety reasons after which participants were asked to alight on the treadmill. The target self-selected gait speed on the treadmill was determined as follows. While holding on to the hand rails participants were instructed to start walking slowly and gradually increase their walking speed to reach a pace that they would normally walk at while on a leisure stroll. As the participant started walking, the treadmill speed was gradually increased to attain the most comfortable speed as dictated by each participant. When this was attained, participants were instructed to let go of the hand rails, one at a time, and continue walking at this speed with no incline on the treadmill until they felt familiarized to it. This speed was used as the target gait speed for subsequent walking conditions of the experiment. To account for the need to endure this target self-paced gait speed for the duration of the experiment, we ensured that the self-selected gait speed on the treadmill was at least 25% less than their over-ground gait speed measured separately on an automated walkway. After acclimatization on the treadmill, the speed was gradually reduced until the participant came to a halt in order to provide adequate rest prior to data recording. For every subsequent experimental condition, the protocol for gradually increasing the treadmill speed to attain the target self-selected gait speed was adhered to. Data recording began with walking only condition followed by walking while performing the working memory tasks described below. Prior to start of the dual-task condition, participants were instructed and encouraged to perform the cognitive task to the best of their ability both in terms of accuracy and speed of responses. For each condition, three trials each lasting 65 seconds were captured at target gait speed. Between trials, the participants were provided adequate rest periods, the duration of which was dictated by the participants.

Working memory task

The working memory task was based on a n-back letter paradigm consisting of three tasks, X-task (control), 1-back and 2-back, that were administered to participants while standing and while walking to capture their single-task and dual-task performance respectively⁴⁴. The working memory tasks included a display of a continuous stream of letters on the screen placed in front of the treadmill. Participants registered their responses by pressing a button held in their preferred hand. Three working memory tasks, ‘X’ [simple working memory], ‘1-back’ and ‘2-back’ [active working memory load tasks], were presented in random order. The display duration was 1500ms and the inter-stimulus interval (ISI) was 2000ms. All the three tasks required participants to maintain information in memory for immediate retrieval, continuously update this information, and register responses on the button-press while keeping track of each letter displayed on the monitor.

During the X-condition, participants were instructed to press the button whenever they saw a letter ‘X’ in the continuous stream of letters. For the 1-back condition, participants were instructed to press the button whenever a letter was the same as the one that came just before it in the sequence [for example, M-T-T or W-B-B]. For the 2-back task, participants were instructed to press the button whenever the letter was the same as the one that appeared two stimuli prior in the sequence of letters [for example U-T-U or B-Q-B]. The letter ‘X’ did not appear in any of the two active working memory tasks. Three trials for each working memory task, each trial lasting approximately 1 minute, with at least one potential response every 5-6 seconds were administered. Reaction times and task accuracy data were recorded from signals obtained by button-presses.

Statistical analysis

Demographic variables between the two groups were compared using Student’s t test and chi-square where applicable. To relate the effect of working memory performance on gait parameters to the SH load in the two groups (AD and NC), the median distribution of total SH score was determined for each group and was used as the cut-off. Those that were above the cut-off were denoted as AD+ and NC+ and those that equaled or fell below the cut-off were referred to as AD− and NC− respectively. ANOVA was used to compare the demographic variables in the four groups.

Single-task cadence, i.e., walking with gaze fixed on the screen (single-task walking), was compared with cadence while dual tasking across three dual-task conditions: the X-task, 1-back task and the 2-back task. To determine the costs on cadence while dual-tasking, a percentage change in the dual-task costs (DTC) of cadence on dual-tasking was calculated relative to participant’s single-task cadence as done in other such studies^{45, 46}:

Performance on working memory tasks was determined by combining accuracy and reaction time (RT) into one performance score, the speed-accuracy trade-off (SAT), calculated by the formula: (Accuracy/RT) × 100. This method has been used in other studies to obtain a composite score incorporating both speed and accuracy of performance on cognitive tasks ²⁴. SAT was calculated for each task during both the standing and walking conditions. Dual-task costs on SAT were calculated as a percentage of change relative to participant’s single-task SAT as above using the formula:

To permit a measure of overall demands on dual-tasking and compare groups we combined the dual-task costs on cadence and working memory task performance, i.e., SAT, similar to Logie et al²¹. Therefore, to obtain an overall DTC measure we obtained the mean of DTC on cadence and SAT for each of the three working memory tasks.

Baseline characteristics

Patients with AD (n=24) were not significantly different from the NC (n=20) on most baseline characteristics (age, gender, BMI, waist circumference, leg-length, mid-calf girth, blood-pressure and heart rate) but differed as expected on the MMSE (p<0.01) and DRS (p<0.01) scores. The AD group showed a trend towards a higher UPDRS score (p=0.06) and significantly longer time to complete the TUG test (p<0.01) (Table 1).

The total ARWMC score in the AD group (range: 1-24, median: 7) was significantly greater than that of the NC group (range: 1-15, median: 5) (p<0.01). Therefore, based on the median of the distribution on the total ARWMC score for each group, those above the median were compared with those equal to or below the median. Thus from the two groups, four groups were derived: AD+ and AD− (with cut-off of 7 on the total ARWMC score for the AD group) and NC+ and NC− (with the cut-off of 5 on the total ARWMC score for the NC group).

The demographic differences between the four sub-groups were evaluated using ANOVA and are highlighted in Table 2. There were no statistically significant differences in baseline body characteristics such as leg-length, BMI and waist circumference. Statistically significant differences between groups were observed in age (p<0.01), MMSE (p<0.01), DRS (p<0.01), UPDRS score (p<0.01) and TUG (p<0.01). Post-hoc Tukey’s test revealed that the NC− group was significantly younger than NC+ (p<0.05) and AD+ and AD−(p<0.01) groups. The two AD groups had significantly lower MMSE and DRS scores than NC but there were no statistically significant differences between AD+ and AD− and between NC+ and NC− on these baseline cognitive measures. The AD+ group had significantly higher UPDRS score than the other three groups (p<0.01). The AD+ group also took significantly longer time to complete the TUG as compared to the NC− and NC+ groups. However, there were no statistically significant differences between AD+ and AD− and between NC+ and AD− groups on the TUG test.

There were no differences in the two groups on cadence, cycle time and double-support during the ‘walk only’ condition though the AD group had a significantly slower self-selected treadmill speed compared to the NC group (0.61 m/sec vs 0.77 m/sec, p=0.02).

Effect of working memory task performance on cadence

A positive dual task cost on cadence implies that the cadence increased while dual-tasking compared to their cadence during the walking without cognitive tasking condition. Similarly, a negative dual-task cost on cadence implies that the cadence actually decreased compared to the walking only condition. There was a significant effect of dual task costs of cadence between AD and NC groups (F(2,42)=11.6, p=0.001, η²_p=0.21). The dual task costs on cadence were significantly higher in the NC group compared to the AD group [(X task: 4.4±5.0 vs 0.75±5.8 (p=0.026); 1-back: 4.9±5.7 vs −0.82±5.2 (p=0.001), 2-back: 5±8.7 vs −0.28±4.5 (p=0.002)]. In fact, the AD group had negative dual task costs on cadence in the 1-back and 2-back conditions indicating that their cadence decreased while performing the 1-back and 2-back tasks.

Next, the dual-task costs on cadence were compared between four groups using repeated measures ANOVA. There was a significant effect of dual task costs on cadence in the four groups (F(3, 41)=5.3, p=0.003, η²_p = 0.28). Post hoc tests revealed that the AD+ group showed a significant decline in terms of their mean dual-task costs on cadence compared NC− (p=0.001), NC+ (p=0.004) and AD− (p=0.05) groups.

Effect of Treadmill Walking on Working Memory Task

A higher number on the SAT denotes better performance and lower numbers denote a poorer performance. The SAT scores were compared between AD and NC groups and then between the four groups subdivided on their SH load. Table 3a highlights the differences in SAT during standing (single-task) and walking (dual-task) conditions in the AD and NC groups. The performance of NC group on all working memory tasks, indicated by the respective SAT scores, on both single- and dual-task conditions was significantly better than the AD group (all p<0.01). When the four sub-groups were compared (table 3b), the NC− group alone performed significantly better on all three working memory tasks during both single- and dual-task conditions compared to the other three groups (NC+, AD+ and AD−).

When performance on the cognitive task was compared within-groups, no significant differences emerged between single-task and dual task conditions within the NC and AD groups. There were no significant within-group differences between single and dual-task conditions in the sub-groups (NC−, NC+, AD− and AD+).

We then compared the DTC of SAT for the three tasks (x-task, 1-bac, 2-back) during dual-tasking compared to standing condition. There was an overall effect of DTC on SAT between groups (F= 4, p=0.05, η²_p =0.1) but not within-groups. When the four SH subgroups were compared there were no statistically significant differences in the DTC on SAT (F=2.5, p=0.07, η²_p = 0.1)

Overall dual-task costs on the three working memory tasks

With the overall dual-task costs for the control task, 1-back and 2-back as dependent variables on the repeated measures ANOVA and diagnostic group (AD vs NC) as between-subjects factor, there was a statistically significant effect between-groups (F (1,43)=7, p= 0.012, η²_p =0.14) but not within-groups (p=0.1). Further analysis revealed that the differences in overall DTC between AD and NC were statistically significant only for active working memory tasks - the 1-back (t=−2.3, p=0.03) and 2-back tasks (t=−2.7, p=0.01) [Figure 1].

Repeated measures ANOVA for the four subgroups with the three overall DTC variables as dependent variables with SH subgroup as between-group factor revealed a significant effect of overall DTC between groups (F (3, 40) =4.2, p=0.01, η²_p =0.24) but not within groups (p=0.7). Post hoc tests revealed that the AD+ group differed significantly from their AD− counterparts (p=0.04), NC− (p=0.002) and NC+ (p=0.01) groups [Figure 2]. There were no statistically significant differences between the NC−, NC+ and AD− groups.