Interaction of age, cognitive function, and gait performance in 50–80-year-olds

First visit

The protocol involved two visits. During the initial visit, subjects provided written, informed consent to participate in the study, which was approved by the university’s institutional review board. Subjects also had to furnish a physician clearance form that demonstrated they were free from major disease and it was safe for them to participate in this walking study. To assess cognitive status, subjects completed both the Mini-Mental State Exam (MMSE) (Herman et al. 2010a; Folstein et al. 1975) and Trail Making Test (TMT) (Sanchez-Cubillo et al. 2009; Tombaugh 2004). The MMSE involves a series of questions and instructions given to the participants to assess their orientation, memory, attention, recall, and language abilities (Folstein et al. 1975). The TMT was used to assess central processing and executive function by testing task switching ability, visuomotor speed, and working memory (Sanchez-Cubillo et al. 2009). The TMT part A (TMT-A) requires participants to sequentially draw lines between numbered circles that are randomly arranged on a page as fast as possible and TMT part B (TMT-B) requires them to connect numbered and lettered circles in an alphanumeric sequence (1-A-2-B-3 etc.…). Cognitive tests were followed by the SPPB to evaluate functional mobility through standing balance, usual gait speed, and chair rise tests (Guralnik et al. 1994). Next, subjects completed a 10-min habituation walk at 0.8 m s⁻¹ on an instrumented gait analysis treadmill (Gaitway II, Kistler Instrument Corp., Amherst, NY, USA). At the end of the habituation walk, preferred walking speed was identified by asking participants to “choose the speed you would normally walk down the street when going to the store” while the investigators slowly increased the speed from 0.5 m s⁻¹.

Second visit

The second visit was scheduled within 1 week of the first and began by measuring height and body mass. After stepping onto the treadmill, an upper body harness was fit to the subjects’ torso and connected to an overhead support system to support body weight in the event of a fall. The line connecting the harness to the support system was kept slack and slightly posterior to the subjects such that it did not interfere with their normal movement patterns. Next, a 5-min warm-up walk was completed on the treadmill at each subject’s preferred speed determined during visit 1. Subjects were instructed to “look, walk, and talk straight ahead” during the test, but were not instructed on how to allocate attention between walking and cognitive tasks. Participants performed the following three tasks while walking: (1) single task, no cognitive load; (2) dual task, serial 1s—counting down from 100 by 1s; and (3) dual task, serial 3s—counting down from 100 by 3s (LaPointe et al. 2010). The serial 3s subtraction task has been shown to elicit the greatest number of calculation errors during dual-task walking in older adults and was the highest cognitive load condition in this study (Srygley et al. 2009). At the end of the warm-up, subjects continued to walk at their preferred speed and completed each of the three tasks in a counterbalanced, random order. Subjects walked for 10s while performing each task before 15s of gait data was collected that provided approximately ten strides (20 steps), similar to previous studies in our laboratory (LaRoche et al. 2012; LaRoche et al. 2011). Ten strides have been shown to be a sufficient number of strides to reliably measure gait variability (Brach et al. 2008; Lord et al. 2011), particularly during continuous walking (Galna et al. 2013), and allowed gait data to be collected in a single, nonfatiguing, uninterrupted trial. After gait data were collected, subjects continued walking for a period of 1 min to reestablish their normal gait pattern before completing the next task.

The treadmill’s deck contained two force plates that provided vertical ground reaction force (vGRF) and center of pressure measurements (COP) independently for each foot. Using these measures, spatial and temporal gait parameters were calculated by the software (Gaitway v. 2.0.8.50, Kistler Instrument Corp., Amherst, NY, USA) under each condition. While the software provides a multitude of gait variables, we identified six variables a priori that have been associated with mobility and fall risk in older adults and were most likely to be affected by cognitive function during dual-task conditions. These were three spatial gait variables including stride length, mediolateral foot strike location, and stride width (Brach et al. 2001) and three temporal gait variables including stride time, single-limb support time, and double-limb support time (Lajoie et al. 1993; Hausdorff et al. 1997). Both the mean of each gait parameter and the variability of each gait parameter were compared between age groups and across cognitive loading conditions.

Using COP data and treadmill belt speed, the software determined stride length as the distance from a foot’s initial heel contact to the following heel contact of the same foot. Foot strike location was the average mediolateral position of the COP during foot-ground contact and was used to assess the lateral stability of the subject with respect to the center of the treadmill. Stride width was calculated by using the average distance between a foot’s COP and the opposing foot’s COP at the next foot strike. Stride time was the duration from the initial heel contact of one foot to heel contact of the same foot at the completion of a single gait cycle (e.g., time from right heel strike to right heel strike). Double-limb support time was calculated by comparing vGRF data between feet to determine the portion of stance when the body was supported by both limbs, and single-limb support time was the portion of stance when the body was supported by only one limb. Double support time and single support time were normalized as percent of the gait cycle (%GC) by dividing each variable by the stride time and multiplying by 100 %.

Association of cognitive tests and spatiotemporal gait performance

TMT-A time was significantly correlated with age (r = 0.58, P < 0.001), as was TMT-B time (r = 0.42, P = 0.003), whereas MMSE score was not correlated with age (r = −0.01, P = 0.486). Under the serial 3s cognitive loading condition, the time needed to complete the TMT-A test was positively correlated with stride time (r = 0.42, P = 0.003), double-limb support time (r = 0.30, P = 0.026), and foot strike location SD (r = 0.47, P = 0.001) and was inversely correlated with single-limb support time (r = −0.34, P = 0.013). The time to complete the TMT-B was not associated with any gait or gait variability measure during the serial 3s condition. Under the serial 3s condition, MMSE score was inversely correlated to single-limb stance time CV (r = −0.29, P = 0.033). Linear regression using age and TMT-A to predict stride length CV in the serial 3s condition revealed that age was significant in the model (partial correlation = 0.40, P = 0.01) and TMT-A was not (partial correlation = −0.07, P = 0.66). Collinearity diagnostics generated eigenvalues of 0.047 for age and 0.006 for TMT-A with condition indices of 7.91 and 22.53, respectively.

Spatial and temporal gait parameters

Results of the RMANOVA examining the effect of age and cognitive load on gait performance during dual-task walking are reported in Table 2. When averaged across age groups, stride time increased from 1.14 ± 0.02s in the single-task condition, to 1.16 ± 0.02s (P = 0.044) in the serial 1-s condition, to 1.17 ± 0.02s (P = 0.029) in the serial 3s condition (Fig. 2a). Similarly, stride length increased from 122.0 ± 2.7 cm, to 123.7 ± 2.8 cm (P = 0.08), to 125.8 ± 2.6 cm (P = 0.035) across cognitive loading conditions (Fig. 2d). Averaged across conditions, stride length was longer in the 50s (132.1 ± 16.5 cm) than in the 70s (116.2 ± 17.1 cm, P = 0.043) with no differences between the 60s (123.1 ± 16.01 cm, P > 0.05) and the other age groups. Likewise, double support time (P = 0.028) increased with age from 13.11 ± 0.61 %GC in the 50s to 15.55 ± 2.36 %GC in the 70s (P = 0.008), with no differences between the 60s and other age groups (14.15 ± 2.20 % GC, P > 0.05). An age group by cognitive load interaction occurred whereby subjects in their 70s had an increase in double support time during the serial 3s condition and younger age groups did not (Fig. 2c).

Open in a separate window

Fig. 2

Comparison of stride time (a), single support time (b), double support time (c), stride length (d), and stride width (e) between groups under single-task walking (single), low cognitive load dual-task walking (counting backwards by ones, Dual-1s), and higher cognitive load dual-task walking (counting backwards by threes, Dual-3s) conditions. %GC percent of gait cycle. Values are mean ± SD. ^* P < 0.05, significant difference between 50s and 70s; ^† P < 0.05, significant difference between 50s and 60s; ^§ P < 0.05, significant difference between 60s and 70s

Executive function and gait performance

The TMT was used to measure EF, task switching abilities, and visuomotor skills, and this measure was positively correlated to age and inversely related to gait performance. A review by Buckner (2004) shows that cognitive function changes minimally from 50 to 70 years of age and then significantly decreases from 70 to 90 years of age (Buckner 2004). The abrupt decline in cognitive function may occur as older adults retire from the workplace and move to an environment that includes less mental and physical stimulation (Ybarra et al. 2008). The mechanisms associated with reduced EF in the elderly, and the potential effects on gait, have been recently reviewed by Yogev-Seligmann et al. (2008) and include age-associated degeneration of frontal lobe structures including white matter lesions, loss of gray matter, reduced dendritic branching, and reduced dopaminergic activity.

The time needed to complete TMT-A was associated with increased stride time, greater double-limb support time, shorter single-limb support time, and greater mediolateral instability during walking, which are characteristics of elderly gait (Winter et al. 1990). The findings of the present study suggest that a small reduction in EF in individuals in their 70s, even within the normative range (Tombaugh 2004), may be enough to reduce ability to maintain a normal, steady gait when cognitively challenged. Other researchers have shown that TMT time is associated with the ability to allocate attention, with age-related reductions in walking speed, increased gait variability, and poorer cognitive performance during dual-task conditions (Siu et al. 2008; Srygley et al. 2009; Montero-Odasso et al. 2009; Coppin et al. 2006; Yogev-Seligmann et al. 2010). The TMT may therefore be a simple screening tool that is sensitive to both declining cognitive and physical function in older adults.

It is not known if increased age and decreased EF independently influence movement quality, or if they are tightly coupled. To attempt to answer this question, we repeated the RMANOVA for the gait variability measures with TMT-A (our measure of EF) as a covariate. The significant main effects of cognitive loading condition and the age group by cognitive load interaction were eliminated, which demonstrated that EF was partly responsible for the age-associated decline in gait performance seen during dual-task walking. Controlling for EF enhanced the age group effects indicating that differences in EF within age groups contributed to the between-subject variability in gait performance. When age and TMT-A time were used as predictor variables in linear regression, the partial correlations revealed that age was the primary factor predicting stride length variability under a dual-task condition. This indicates that other deficits of aging (strength, balance, coordination, vision, somatosensory, etc.) may have contributed more to increased gait variability during dual-task walking than EF. However, collinearity diagnostics in the regression analysis (eigenvalues near zero along with high condition indices) indicated that age and TMT-A were collinear. Taken together, these post hoc analyses suggest that while both age and EF are independently related to poor gait quality, it may not be possible to separate their effects as increasing age and declining executive function occur concurrently.

Cognitive loading and stride kinematics

Results of the current study show that performing a cognitive task while walking alters gait of healthy, community-dwelling, older adults, and probably more so in the oldest individuals. The data indicate that the 70s group increased stride length, increased stride time, and increased the time spent in double-limb support in an effort to maintain their stability and orientation during the most challenging dual-task condition. We speculate that these subjects did not have the level of EF needed to simultaneously perform both the walking and cognitive tasks, and they prioritized the subtraction task over the maintenance of gait stability. This theory is supported by the small number of subtraction errors recorded in this condition. If a subconscious prioritization of cognitive tasks over stability and gait occurs in old age, it may predispose the oldest subjects to a loss of balance when distracted.

With age, decreased EF may lessen the attentional resources available to process both the motor planning and cognitive aspects of gait (van Iersel et al. 2007). It has been suggested that older adults adopt slower gait to minimize motor planning and allow their mind to utilize their diminished cognitive attention, with consequent increases in stride-to-stride variability and risk for falls (Montero-Odasso et al. 2009). It has been shown that walking on a treadmill requires some conscious attention and that individuals will reduce walking velocity while performing a second task (Montero-Odasso et al. 2009; Regnaux et al. 2006). The subjects in their 70s were the only group to choose slower treadmill speeds compared to their overground speeds, possibly as a compensatory mechanism to deal with the added challenge of treadmill walking.

Cognitive loading and gait variability

Although the variability of stride length, foot strike location, single-limb support time, and stride time were higher in the 70s group, the variability of these temporal gait measures was only moderately influenced by the addition of a cognitive load in the dual-task walking conditions. While the mean stride length CV and stride time CV appear to increase with cognitive loading in the 70s group, the interaction was not statistically significant because of the high between-subject variability (see SD error bars in Fig. 3 for the 70s group). It was our observation that some subjects in the 70s group had a very difficult time maintaining a steady gait in the serial 3s condition, whereas others completed the trial with limited difficulty. The group possessed a range of physical and cognitive capacities and it is likely that this task differentially challenged the subjects with the poorest function.

The modest effect of dual-task loading on temporal gait variability may have also occurred because the assessment was performed while walking on a treadmill, at a fixed speed, and may have been different if the subjects had performed the dual-task trials overground where their speeds would not have been kept constant. For example, counting backwards from 100 by ones has been shown to reduce gait speed of older adults with mild cognitive impairment by 28 % (Montero-Odasso et al. 2009). In the absence of a dual task, Van Emmerik et al. (1999) demonstrated that stride time variability increases as speed decreases. Similarly, Dubost et al. (2006) showed that the increase in stride time variability under dual-task conditions was a product of both a verbal fluency task and the decreased walking speed that occurred in this dual-task condition. It is therefore plausible that much of the increased temporal gait variability usually seen during dual-task walking overground is a result of reduced speed.

An unexpected finding of this study was that single-limb support time variability decreased in all groups with an increase in cognitive loading. It is not clear what the source of the reduced variability is, but it may have occurred because of the increase in stride time that occurred in the serial 3s condition. Single support time variability was calculated using time in seconds (not normalized as percent of gait cycle), which increased concurrently with stride time in the serial 3s trial. It is possible that as the magnitude of single-limb support time increased, the reliability of its measurement between strides increased and contributed to the reduced variability seen in the serial 3s condition. Nevertheless, the 70s group demonstrated a higher level of single-limb support time variability in the serial 3s condition than other age groups.

Cognitive loading and balance control

Gabell and Nayak (1984) originally proposed that the variability of stride width and double support time is primarily determined by balance control. Two spatial gait variability parameters related to balance (i.e., foot strike location SD and stride length CV) were more affected by age and cognitive loading than the temporal aspects, even though people normally utilize a “posture first” strategy to prioritize balance ahead of other tasks (Yogev-Seligmann et al. 2008). We hypothesized that the dual-task conditions would challenge balance control resulting in a wider stride width, increased time spent in double-limb support, along with a more variable stride width and greater foot strike location variability (our marker of mediolateral instability). Double support time and foot strike location variability increased in the 70s group in the serial 3s condition, whereas there was no effect of cognitive load on stride width or stride width variability. In fact, the older subjects walked with a narrower stride width under the most challenging conditions, which was likely a suboptimal strategy. Guimaraes and Isaacs (1980) showed that elderly fallers walked more slowly, with a shorter step length, narrower stride width, and increased step length variability compared to their nonfalling peers, a pattern similar to this study’s oldest subjects. The greater mediolateral instability in the 70s group may partly be explained by a linear decline with age in the mechanical work performed at the hip and ankle during walking, which causes a decreased capacity to control body movement in the oldest individuals (Ko et al. 2010).

Fall risk in older adults has been studied from the perspective of muscle weakness, gait performance, balance, visual acuity, and cognitive function (AGS 2001). This study demonstrates that when older individuals are challenged by simultaneously performing a cognitive and physical task, walking performance and balance control are diminished. A comprehensive reading of the fall risk literature illustrates that it is impossible to separate the effects of the comorbidities of aging, despite researchers’ attempts to do so. With aging comes a generalized decline in sensory, central nervous, motor control, musculoskeletal, cardiopulmonary, and metabolic function, along with changes in body composition, that predispose older individuals to falling. It is likely that the elderly are aware of their declining function and slow and alter gait to operate within their diminished capacities; however, this awareness may be compromised in those with cognitive dysfunction. To reduce the risk of falls, it is therefore important to identify cognitive and physical weaknesses, attempt to improve these weaknesses, and counsel older adults about limiting the number of challenges while walking.

D.P. LaRoche was supported by the National Center for Advancing Translational Sciences via NIH Grant L30 AG038028-02.