Design
We completed a 6-week, single-arm pilot study of a sideways walking intervention with baseline, post-intervention, and retention measurements of risk-of-falling related outcomes. Reporting followed the Consolidated Standards of Reporting Trials (CONSORT) statements for randomized pilot and feasibility trials [50]. To ensure completeness of reporting, and replication of the intervention we followed the Template for Intervention Description and Replication (TiDieR) guidelines [51], which is recommended as extension on the CONSORT guidelines [50].
Participants
Fifteen older adults enrolled in the study. Inclusion criteria were (i) ≥ 65 years, (ii) be independently residing in the community, and (iii) ability to walk independently. Participants were not eligible if they (i) had a neurological disorder or progressive neurologic condition, (ii) had a musculoskeletal disorder or injury that could affect gait, (iii) had a surgery within the past 6 months, (iv) had a history of a cardiovascular event, and (v) were participating in any other studies that involves walking, balance, or exercise intervention.
Participants were recruited from 3 sources: (i) local retirement community, (ii) employees of the University of Nebraska at Omaha, and (iii) a sample of 190 older adults enrolled at the Mind & Brain Health Labs (MBHL) of the University of Nebraska Medical Center (UNMC). Between September 2017 and March 2018, one of the authors (AS) went to local retirement homes, fitness classes, and libraries to talk about the research and to distribute approved flyers. Moreover, an invitation to participate to the study was sent by email to all the members of the MBHL. A notice seeking volunteers was also announced to the university’s employees through campus-wide email posts linked to text on university’s website news page. Interested older adults conducted us by email or telephone and a screening visit at the Biomechanics Research Building was scheduled. Eligible older adults were identified by one of the authors (AS), and they asked whether they would like to participate in the study. All participants were asked to read, understand, and sign an informed consent form approved by the Institutional Review Board of the UNMC prior to participating in the study.
Sideways walking intervention
Before the first session, the participants were given a visual demonstration of sideways walking. Precise instructions were as such: (i) ‘keep the head up while stepping laterally’, (ii) ‘do not cross feet at any point’, (iii) ‘feet and legs are to be pointed in the same direction as the body’, and (iv) ‘at no point can both feet be off the ground’. Every session began with ‘warm-up’ exercises, which included 200–300 m walking at self-selected speed and stretching exercises at the comfort level of the participants. After the ‘warm-up,’ the participants started the sideways walking training. The 6-week intervention was performed at the indoor walking track (circumference of about 200 m) of the Recreation Building of the University of Nebraska at Omaha, where the 10 m walkway was located. All training sessions had a single participant and were supervised by one of the authors (AS).
The load of the training was based on the American College of Sports Medicine guidelines for older adults [52], which recommend 20–30 min on 2–3 days∙week− 1 for neuromotor exercises (balance, agility, coordination and gait). Thus, each participant was trained 3 days∙week− 1 for 6 weeks, resulting in a total of 18 sessions. Each participant performed 6 trials∙session− 1 that were alternated with 1–3 min of rest. Each trial consisted of 3 min sideways walking across a 10 m walkway changing body direction at the ends, thus alternating lead and lag limbs. Each training session lasted 30–45 min. The first session was at the participant’s self-selected sideways walking pace. The participants were given instructions that they should strive to increase their pace if they can, as they progress through the 18 sessions. The participants were informed that they could increase their pace at the start of each trial but may not decrease it at the next session. The time to cover the 10 m sections in each trial was manually recorded and charts were used to monitor participants’ adherence to the intervention protocol. The session was rescheduled when a participant reported a level of muscle soreness or joint pain that prevented them for maintaining the previous walking speed. When this became evident after a session had already commenced, the session was curtailed but not rescheduled. For safety reasons, sideways walking was performed next to a horizontal handrail to grasp if required. A staff was standing nearby to help as an added safety measure.
Feasibility outcomes
To evaluate the eligibility criteria, we used the ratio of included participants to those who did not meet the eligibility criteria. Participant and instructor fidelity at the intervention protocol was assessed by monitoring the walking pace at all trials per session. The feasibility and suitability of intervention outcomes was assessed by measuring the extent of the missing data, and ceiling or floor effects. Feasibility was measured by the ability to recruit and retain older adults until complete the follow-up (i.e., 6 weeks after the completion of the intervention). The study was considered feasible if we were able to recruit 3 participants∙week− 1, and if ≥ 80% of the sample was able to complete the follow-up.
Intervention outcomes
Intervention outcomes were walking speed, gait variability (variability of step width, step length, stride time, and stance time), the Timed Up and Go test (TUG), the Berg Balance Scale (BBS), and the Falls Efficacy Scale-International (FES-I). Walking speed is a predictor of fall risk (when is less than 1 m·sec− 1), and of disability, mortality, and adverse events in older adults [53–57]. Change in walking speed near 0.05 m·sec− 1 is small but meaningful and change near 0.10 m·sec− 1 is substantial [58]. Increased variability of spatial (step length and width) and temporal (stride and stance time) gait characteristics compromises gait performance and increase the tendency of older adults to fall [59, 60]. Stance time variability is an indicator of preclinical disability mobility (when stance time variability ≥ 0.034 sec) [61], while step width variability ≥ 2.5 cm is considered excessive [34]. Meaningful changes are 0.25 cm for step length variability, and 0.01 sec for stance time variability [62]. The TUG test was designed for assessing mobility in older adults [63], and has been used for predicting fall risk (when > 12 sec) [53], as well as for screening for frailty in older adults [64]. The BBS (14 items, max score: 56) is a valid and reliable test to measure the functional balance in older adults and predicts fall risk (when < 50 points) [53, 65, 66]. The FES-I is a valid and reliable questionnaire (16 items, max score: 64) to assess confidence in the performance of activities relevant to daily life and is used as a subjective measure of fall risk (when > 24 points) [53, 67, 68]. Additional measurements included: the Mini-Mental State Examination score (MMSE) (12 items, max score: 30) to measure cognitive impairment [69]; the short form of Geriatric Depression Scale (GDS) (15 items, max score: 15) to assess older adults for depression [70]; the short form of Brief Pain Inventory score (BPI) (4 Pain severity items, max score: 40; 7 Pain interference items, max score: 70) to measure the impact of pain on daily functions [71]. Furthermore, participants have been asked if they had sustained 2 or more falls in the past year. A fall was defined as an event that caused participants to rest on the floor.
Data collection and analysis
Data were collected at the Biomechanics Research Building at baseline, post-intervention, and retention period (6 weeks following completion of the intervention). The building featured a 3D motion capture system with 17 high-speed Raptor cameras (Motion Analysis Corporation, Santa Rosa, CA, US) synchronized with an instrumented treadmill (AMTI, Watertown, MA, US). Upon arrival at the Biomechanics Research Building for the baseline assessment, the participants changed into a tight-fitting suit, then they performed the TUG and BBS, completed the clinical questionnaires (FES-I, BPI, GDS, and MMSE), and were asked for previous falls. Then, retroreflective markers were placed at anatomical locations to gather kinematics data during walking on the treadmill at sampling rate of 100 Hz.
After participants’ ‘warm-up’ for 3 min on a treadmill at a self-chosen pace, they asked to self-select a walking speed based on comfort level. The participants started walking at treadmill’s slowest speed and then incrementally (intervals of 0.17 m∙sec− 1) the speed was increasing until the participant stated that this was their preferred comfortable speed. The treadmill was then increased another increment, so that the participant could confirm that the previous speed was the preferred speed. This procedure was continued and repeated until successful confirmation that a comfortable speed was reached. The participants performed 3 trials of 3 min walking on a treadmill at their preferred speed, alternated with 2 min of rest. Participants wore a harness during all treadmill trials.
Following the 6-week sideways walking intervention, the participants performed a post-intervention assessment. The data collection mirrored the baseline assessment with additional trials on treadmill to enable possible comparisons to be made with speed both fixed across sessions (at baseline preferred speed) and free to reflect functional post-intervention differences (post-intervention session using newly determined self-selected preferred speed). Participants performed 6 trials of 3 min treadmill walking. Three trials were at the baseline speed and the other 3 trials were at the post-intervention preferred speed. When there was not post-intervention difference on the preferred speed, the participants performed only 3 trials at the baseline speed. Participants returned 6 weeks after the completion of the intervention for the retention assessment. The data collection at the retention mirrored the post-intervention assessment.
We determined gait events from the filtered (low-pass Butterworth, 6 Hz cut-off frequency) heel and toe markers trajectories using custom MATLAB code (v. R2019a, The MathWorks, Natick, MA, US). We used the standard deviation of step width (mediolateral distance between the locations of the sequential left and right heel strikes), step length (anteroposterior distance between the locations of the sequential left and right heel strikes), stride time (the time between 2 consecutive ipsilateral heel strikes), and stance time (the time elapse during the stance phase of a leg) to evaluate gait variability [34].
Sample size
The issue of statistical power was addressed via consideration of the magnitude of the treadmill walking speed difference at baseline and post-intervention. A preliminary sideways walking study that we performed showed an average increase in treadmill walking speed from 0.50 ± 0.26 m·sec− 1 to 0.97 ± 0.38 m·sec− 1 in older adults, and from 0.93 ± 0.18 m·sec− 1 to 1.19 ± 0.16 m·sec− 1 in young adults, which corresponds to a between-groups effect size of 1.12 (Cohen’s d statistic), and within-group effect size of d > 3 for both groups. Therefore, 12 participants provided 80% power to detect a within-subjects effect size of d = 1.3 at a α = 0.05 significance level with non-sphericity correction of ε = 0.5. We recruited 15 participants to allow for 20% attrition.
Statistical methods
The outcomes measures of walking speed, TUG, and gait variability (step width variability, step length variability, stance variability, and stride time variability) were analyzed separately using a repeated measures ANOVA with Time (baseline, post-intervention, and retention assessments) as within factor. When sphericity was violated, a Greenhouse-Geisser correction was applied. The average values at each Time level were computed for the statistical analysis. If the ANOVA revealed effects (p < 0.05), further univariate comparisons were performed using a planned (simple) contrast in which all conditions were compared with the baseline. For the ANOVA comparisons, the Cohen’s f effect size was reported (f < 0.10 negligible, f < 0.25 small, f < 0.40 moderate, otherwise large effect). Responsiveness of outcomes was reported using the ES statistic, this is the mean change between baseline and post-intervention divided by the standard deviation of the measurement at baseline [72, 73] (| ES | < 0.02 negligible, | ES | < 0.50 small and | ES | < 0.80 moderate, otherwise large effect). The ordinal outcomes of the clinical tests (BBS, FES-I) were analyzed separately using Friedman test, followed if needed by Wilcoxon singed-rank test. Kendall’s W coefficient of concordance was used to report effect size (W < 0.10 negligible, W < 0.25 small, W < 0.40 moderate, otherwise large effect). Significance level was set at α = 0.05. Statistical analyses were performed in R software Version 3.6.3 (R Foundation for statistical computing, Vienna, Austria) [74] using the afex [75], emmeans [76], sjstats [77], and rstatix [78] packages.