Training dual-task walking in community-dwelling adults within 1 year of stroke: a protocol for a single-blind randomized controlled trial

The theoretical framework for the gait training intervention incorporates principles from Gentile’s taxonomy of tasks [27] and the Motor Relearning Program for Stroke [28]. Gentile’s taxonomy of tasks is a model for progressing patients from closed environments (stationary, predictable) to open environments (moving, unpredictable). The Motor Relearning Program for Stroke, from which the task-oriented training approach was derived, advocates 4 steps: (1) task analysis, (2) part-task practice, (3) whole-task practice, and (4) transfer of training to daily life. Using these frameworks, each session will comprise three specific components: (i) part-task practice, (ii) whole-task practice, and (iii) contextual practice to facilitate transfer to real-world environments.

Gait training activities will be performed initially in closed environments (quiet area with few distractions) and progressed as appropriate to an open environment (busy area with unpredictable and moving distractors). Part-task practice includes activities such as dynamic postural control and single-step training; whole-task practice includes activities that involve continuous walking; contextual transfer practice includes obstacle negotiation and outdoor walking, and is an extension of whole-task practice.

In each session, participants will complete 12 repetitions of two part-task and two whole-task practice activities, and 12 repetitions of one contextual transfer activity. To ensure the sessions are the same duration for all participants (30 minutes), after the warm-up, each component will be limited to 10 minutes. For participants with more severe impairments who are not able to complete 12 repetitions of two activities in 10 minutes, therapists will advance to the next component after 10 minutes to ensure that all components of the training program are practiced in each session. Conversely, participants who complete the activities in less than 30 minutes will perform a second set of repetitions of the contextual transfer task. Thus, the total number of repetitions may vary slightly based on individual ability levels, but the exposure time will be identical for all participants.

The gait training activities for each component (part-task, whole-task, contextual transfer) are standardized. Five levels of difficulty are defined for each standardized gait training activity. Participants will progress through the difficulty levels according to individual abilities and the progression guidelines detailed in the Intervention Manual of Procedures.

Participants in the DTGT group will practice identical gait activities as the STGT group. The critical difference is that the DTGT group will practice the gait activities while simultaneously performing a cognitive task for 75% of the repetitions. Specifically, in DTGT, the first 3 repetitions will be performed without any added cognitive task, and then 9 repetitions will be practiced under dual-task conditions. Participants in STGT perform all repetitions without any cognitive tasks. A variable-priority training strategy for the dual-task intervention will be used, such that participants will alternate their attentional focus between gait and the cognitive task within each session. Silsupadol et al. have reported the variable-priority approach to be more effective than fixed-priority practice (equal attention to both tasks at all times) and single-task training for improving dual-task gait speed [16], balance (body sway) [13] and cognitive performance (reaction time) [13] in older adults. Thus, we considered variable-priority to be the best available approach for dual-task training at this time.

Ten different cognitive tasks involving a range of cognitive functions (working memory, executive function, visuospatial processing, and language) will be used during DTGT. The sequence of cognitive tasks through the program is standardized to ensure that participants practice an array of cognitive tasks in varying combinations with the gait activities. Two different cognitive tasks will be used in each session. Progressions in difficulty within each cognitive task will be based on individual abilities and will be directed by progression guidelines detailed in the Intervention Manual of Procedures. Participants are not expected to reach the highest level of difficulty for each cognitive task; rather, the task progressions are provided for variation as the participants advance through the program.

The primary outcome measure will be the dual-task effect (DTE) on gait speed and cognition (reaction time and accuracy) during unobstructed walking at preferred walking speed and fastest comfortable walking speed. Dual-task effects will be calculated by dividing the difference between single and dual-task performance by single-task performance, expressed as a percentage [29]. Specifically, for variables in which a higher value indicates better performance (e.g., gait speed, cognitive task accuracy), DTE will be calculated as:

Conversely, for variables in which lower values indicate better performance (e.g., reaction time), DTE will be calculated as:

Thus, for all variables, positive DTE values indicate an improvement in performance in the dual-task condition relative to single-task (i.e., dual-task benefit) and negative DTE values indicate a decrement in performance in the dual-task condition (i.e., dual-task cost).

We will assess dual-task performance in two different dual-task combinations. The two cognitive distractor tasks that will be used in the dual-task conditions will be the auditory Stroop task [30], and the clock task [5]. The Stroop task involves executive function, which is critical in dual-task coordination. Participants will hear the words “high” and “low” spoken in either a high pitch (360 Hz) or a low pitch (180 Hz) and must indicate the pitch of the word they hear (ignoring the actual word presented) as accurately and as quickly as possible. The clock task is a visuospatial reaction time task. Participants will hear a time (e.g., “two-twenty-five”) and respond verbally (yes/no) based on whether the hands of clock for the given time are in the same half or not: “yes” if both hands are in the same half (left/right) of the clock, and “no” if they are not. Visuospatial abilities are highly relevant for navigating the environment. The participants will first practice the Stroop and clocks while sitting. Single-task performance of the cognitive task will then be recorded (sitting), followed by the dual-task conditions (walking at preferred and fastest comfortable speed). No specific instruction regarding task prioritization will be provided for the dual-task trials.

The gait tasks will involve continuous walking for one minute in the laboratory at Northeastern University. Single-task trials at self-selected and fastest comfortable speed will be performed before dual-task trials. Gait speed data will be acquired using a 12 camera Qualisys Motion Capture System. Thirty lightweight reflective markers (12mm) will be placed to anatomical positions of the subject’s lower extremities and pelvis in order to define the respective segments [31, 32]. Gait speed will be calculated from the Visual 3D software, and is computed using the actual stride length over the actual stride time.

In addition to gait speed, we will also measure other spatiotemporal parameters of gait (e.g., stride length, double limb support duration) in the single and dual-task conditions described above. Other secondary outcome measures include: spatiotemporal and kinetic gait parameters during high and low obstacle crossing, spontaneous physical activity, executive function, lower extremity motor function, functional gait performance, balance self-efficacy, number of falls, and stroke-related disability.

Spatiotemporal data during obstacle crossing will be acquired as above for unobstructed gait. Additionally, kinetic data will be collected by two AMTI force platforms embedded in the floor of the gait laboratory. Signals from the markers will be digitized at a sampling rate of 100Hz, while raw analog data from the force platforms will be collected at a sampling frequency of 1000Hz. Obstacle crossing will be performed with obstacles at 5% and 15% of leg length. The obstacle will be placed between two force plates along the walkway, enabling force plate data to be acquired before and after the obstacle. We will collect data on step kinematics during obstacle approach to quantify hesitations incurred by the necessary modifications of strides towards the obstacle. In addition, we will measure vertical foot clearance over the obstacle, pre- and post-obstacle clearance distance, ground reaction forces and their variability measured by the force plates, as an indicator of the degree of control of stepping over the obstacle. Using motion capture, we will determine the time profile of the center of mass to examine smoothness of motion progression during obstacle approach and crossing. Since the pre-crossing phase of obstacle negotiation is particularly attention-demanding due to the planning required for step adaptation, this condition allows us to determine the effects of DTGT on a real-world dual-task that has the cognitive load inherently embedded. Participants will complete 4–5 trials in each obstacle condition. Since the obstacle negotiation task itself requires cognitive resources, we will not be assessing obstacle crossing during the Stroop or clock tasks. Rather, the obstacle condition enables direct investigation of a “natural” dual-task.

To measure spontaneous physical activity, participants will wear a physical activity monitor (PAMSys™, Biosensics LLC, MA, USA) embedded in a lightweight breathable T-shirt worn under clothing for two consecutive days after each evaluation. This device uses a combination of miniaturized kinematic sensors housed in a single portable sensor attached to the chest. It can detect body posture (e.g., sitting, lying) as well as provide an accurate assessment of periods of locomotion (e.g., walking, turning), including gait inter-cycle variability during activities of daily living. The validity of this approach has been established in three separate pilot studies and by benchmarking the results with independent analysis by an optical motion system [33‐35].

Executive function will be assessed using a computerized Stroop test [36]. DirectRT (Empirisoft, New York, NY) software will record the participants’ reaction times and responses. We will use color-word interference accuracy scores and reaction times to assess executive function at each assessment. Lower extremity motor function will be assessed using the Fugl-Meyer Motor Assessment scale [37]. Functional gait performance will be evaluated using the Timed Up and Go test [38], and the Activities-specific Balance Confidence Scale [39] will measure balance self-efficacy. The Stroke Impact Scale [40] will be used to assess stroke-specific disability. Participants will keep a falls diary between the post-intervention assessment and the 6-month follow-up assessment. Any participant reporting a fall will receive a phone call from the research team to obtain information about the fall and any related injuries.

Additional measures conducted at baseline to further characterize the study sample include: Digit Symbol Modalities Test (speed of processing) [41], Comprehensive Trail Making Test (inhibition of distraction) [42], and the Star Cancellation Test (unilateral spatial neglect) [43], National Institutes of Health Stroke Scale (stroke severity) [44], 6-minute walk test (walking endurance) [45], Melbourne Edge Test (contrast sensitivity), and lower extremity sensation and proprioception. Demographic data including age, time since stroke, medical history, employment status and living situation will also be collected at baseline.

The protocol for this study has been approved by the Northeastern University Institutional Review Board (IRB; #11-06-17) with IRB authorization agreement from New England Rehabilitation Hospital (Woburn, MA). Individuals who wish to participate in this study will provide written informed consent.

The study has been powered to detect a minimum difference of 10% in dual-task cost on gait speed between the DTGT and STGT groups. The difference of 10% is based on the established minimal clinically important difference for gait speed decline in older adults (including people with stroke). Perera et al. [46] determined that a decline in gait speed of 0.1 m/s was associated with substantial meaningful change in physical function. Based on the expected mean gait speeds throughout the phases of this study, a dual-task decline of 0.1 m/s corresponds to a 10-12% dual-task cost on gait speed. Assuming within-group standard deviation of 10.15, with 80% power at the 0.05 significance level, we need 18 subjects in each group. Allowing for 20% attrition at the 6-month follow up, we plan to randomize 22 subjects to each group (total 44 subjects).

DTGT and STGT groups will be compared on baseline characteristics using Student t-tests for continuous data and Fisher’s exact test for proportions. The hypotheses will be tested using an intention-to-treat analysis, applying a repeated measures design whereby the two groups will be compared on primary outcomes at: (i) baseline, (ii) after the completion of the intervention, and (iii) 6 months following completion of the intervention. For the primary outcomes of dual-task effects on gait and cognition, a 2 x 3 mixed model repeated measures ANOVA with group (DTGT, STGT) as between-subjects factor and time (pre-training, post-training, 6-month follow-up) as within-subjects factor will be applied to DTE for each gait task separately (preferred walking, fast walking) and for each cognitive task separately (Stroop, clock). A similar analysis will be performed for all secondary outcome measures. Tukey’s honestly significant difference (HSD) post-hoc tests will be applied to examine the significant differences for all ANOVAs. If the normality assumptions for ANOVA are not supported, nonparametric equivalent statistical methods including Mann–Whitney U-tests and Wilcoxon signed-ranks tests will be used.