Effect of Treadmill Perturbation-Based Balance Training on Fall Rates in Community-Dwelling Older Adults

This randomized clinical trial evaluates the effect of a perturbation-based balance training treadmill intervention vs standard treadmill walking on 12-month fall rates among community-dwelling older adults.


Introduction
6][7] Nevertheless, some of these falls are preventable, and physical exercise is considered to be one of the most efficient preventive approaches. 6,8A promising task-specific exercise modality is perturbation-based balance training (PBT). 9,10During PBT, individuals are exposed to repeated external disturbances in a safe environment that are intended to improve reactive responses after slips and trips (an overview of the principles and mechanisms has been described elsewhere [9][10][11] ).0][21] Nonetheless, the dose for optimal retention and generalization of adaptations still needs further evaluation. 10[24][25][26][27][28][29][30][31][32][33] The recently published world guidelines for fall prevention highlight PBT as a future research priority. 4Therefore, this randomized clinical trial (RCT) aimed to evaluate the effects of a 4-session treadmill PBT intervention compared with regular treadmill walking on daily-life fall rates among community-dwelling older adults.

Study Design, Setting, and Participants
This assessor-blinded, parallel-group (1:1 ratio) RCT was conducted between March 2021 and December 2022 at a laboratory in Aalborg University in Denmark.The trial protocol was approved by the North Denmark Region Committee on Health Research Ethics and the Danish Data Protection Agency.All participants provided written informed consent.The trial protocol (Supplement 1) and statistical analysis plan were preregistered at ClinicalTrials.gov and published elsewhere. 34There were no important deviations from the protocol.We followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline. 35dividuals were recruited through radio and television advertisements and snowball sampling from March through November 2021.Eligible individuals were 65 years or older, community dwelling, and able to walk without a walking aid.Individuals were excluded if they (1) had any of the following self-reported conditions: orthopedic surgery within the past 12 months, osteoporosis or osteoporosis-related fractures (low-impact hip, spine, or wrist fracture), or progressive neurological disease (eg, Parkinson or multiple sclerosis); (2) had an unstable medical condition that prevented safe participation; (3) had a severe cognitive impairment (score <8 in the Short Orientation-Memory-Concentration Test); or (4) were currently participating in another fall prevention trial.

Randomization and Interventions
After the pretraining assessments, participants were randomized in a 1:1 ratio to either the PBT (intervention) group or control group (Figure 1).A permuted block randomization sequence was were performed on the same day immediately after the pretraining assessments.A week later, the third training session was completed, whereas the fourth was conducted after 6 months (Figure 2A).The training sessions of the intervention and control groups were conducted in the same laboratory setting.

JAMA Network Open | Geriatrics
Before the first session began, participants' preferred treadmill walking speeds were determined by increasing and decreasing the belt speed to establish the upper and lower limits for comfortable walking.
The preferred walking speed was then defined as the mean of the upper and lower boundaries. 36 2A).The side (right or left leg) and timing (intervals of 10 to 50 steps between) of the perturbations were randomized to increase unpredictability.The slip perturbations were induced by a quick forward acceleration at heel strike (0% of the gait cycle), causing a reversal in the treadmill movement direction (Figure 2B).The trip perturbations were induced by a sudden deacceleration followed by a greater backward acceleration at midswing (approximately 80% of the gait cycle) (Figure 2C).The perturbation intensity was adjusted based on participants' preferred walking speed and was divided into 5 levels with progressively longer durations (slip) or greater accelerations (trip).The intensity was increased after every fourth perturbation if (1) the combined rating of perceived anxiety and difficulty was 4 or less, (2) the participant avoided falling during the previous 4 perturbations, and (3) the participant was comfortable increasing the intensity.However, if any criterion was not met, the intensity remained unchanged (for protocol tailoring, see eAppendix 1 in Supplement 2).

Treadmill Walking Training (Control)
In each of the 4 treadmill walking training sessions, the control participants performed 20 minutes of treadmill walking at their preferred walking speed.The duration of these walks matched the PBT group's time on the treadmill.

Outcome Assessment
This RCT's primary outcome was the daily-life fall rates that were collected from fall calendars for the 12 months after the third training session.p1619) Participants made daily recordings in fall calendars and returned these calendars to the research team at the end of each month using preaddressed envelopes.All fall calendars were reviewed by a research assistant who was blinded to the group randomization.If a participant recorded a fall, the blinded research assistant interviewed the participant over the telephone to validate the fall and retrieve additional information about the circumstances and consequences of that fall.
Secondary outcomes included the proportion of participants experiencing at least 1 fall and recurrent falls, time to first fall, fall-related fractures, fall-related injuries, fall-related health care contacts, and daily-life slip and trip falls.A fall-related fracture was a radiologically confirmed fracture from a fall. 37A fall-related injury was any minor (eg, bruises, lacerations, or soreness) or major (eg, fractures or head injuries) injury from a fall.A fall-related health care contact was defined as any contact with the health care system (eg, general practitioner, emergency department, or hospital admission) associated with a fall.
At the pretraining and posttraining assessments and the 6-month and 12-month follow-up, laboratory falls were evaluated by exposing participants to level-1 slip and trip perturbations.These perturbations were recorded in slow motion (100 fps [frames per second]) with a camera (Canon M200; Canon Inc) that was positioned 2 m away, perpendicular to the center of the treadmill.The video recordings were reviewed by a blinded research assistant to determine whether the participants experienced a fall.An attempt was categorized as a fall when the participant's body posture was in a falling mode that was clearly and unambiguously stopped by the safety harness.The laboratory fall rate (number of laboratory falls per exposed perturbation) was used for analysis.The personnel responsible for the training sessions (J.E.N.) registered all adverse events related to the intervention, including any injury or unpleasantness.

Statistical Analysis
An a priori sample size calculation was performed using a Poisson regression model in G*Power, version 3.1.9.4 (University of Kiel).To detect a significant (2-sided α = .05)difference between groups of 50% from a base fall rate of 0.85 with 80% power and an expected 20% dropout, a total of 140 participants, with 70 in each group, was needed.9][40] The base fall rate of 0.85 was extracted from the control groups of the studies on physical exercise and fall prevention that were included in a systematic review. 8e statistical analyses were conducted in collaboration with an external statistician according to a preregistered statistical analysis plan using Stata, version 17.0 (StataCorp LLC).The analyses were based on the intention-to-treat principle.Two-sided P < .05 was considered to be significant.
Between-group differences in count outcomes were analyzed using a negative binomial regression with person-years of follow-up as an offset variable.We applied the negative binomial regression as the data fit better with this model than with the preregistered Poisson regression with bootstrapping.A sensitivity analysis using the preregistered approach was conducted.Binary daily-life fall outcomes were evaluated using Poisson regression with robust error variance. 41The between-group difference in time to first fall was analyzed with a Cox proportional hazards regression model (the eFigure in Supplement 2 provides the survival plot).Data for participants with no falls were censored at the end of follow-up.The proportional hazards assumption was tested statistically, and data met this assumption (χ 2 = 1.55;P = .21). 42The Stata codes used for analyses are available in eAppendix 2 in Supplement 2.
Preregistered sensitivity analyses were conducted for each daily-life fall outcome, adjusting for age, sex, and previous falls and following the per-protocol principle, among participants who completed at least 75% of the training sessions.Sensitivity analyses that were decided after data collection included fall rates after stratifying participants based on their fall history and laboratory fall rates after multiple imputations.

Enrollment and Adherence
Of the 199 individuals screened for eligibility, 140 were included (mean [SD] age, 72 [5] years; 79 females [56%] and 61 males [44%]) and randomized to either the PBT or control group (Figure 1).Fifty-seven participants (41%) had a fall in the past 12 months before study enrollment.The baseline characteristics of the participants were comparable between groups (Table 1).
The PBT group completed 90% and the control group completed 93% of the assigned training sessions.At least 75% (per-protocol criterion) of the training sessions were completed by 90% of the PBT group and 97% of the control group.and 25 participants (36%) in the control group had missing laboratory fall data.Data were treated as missing at random because the reasons for missing data and the demographic characteristics of participants with missing data were similar between the 2 groups (eTables 1 and 2 in Supplement 2).

Daily-Life and Laboratory Falls
For the primary outcome, 62 falls in the PBT group (rate of falls per person-year of follow-up: 0.90; 95% CI, 0.58-1.21)and 78 in the control group (rate of falls per person-year of follow-up: 1.14; 95% CI, 0.76-1.52)were reported, resulting in a nonsignificant 22% relative between-group difference in fall rates (incidence rate ratio [IRR], 0.78; 95% CI, 0.48-1.27)(Figure 3 and Table 2).No significant between-group differences in any secondary daily-life fall outcomes were found (Table 2).

Sensitivity Analyses
The Poisson regression with bootstrapping yielded results that were comparable to those of the negative binomial regression for all count outcomes (eAppendix 3 in Supplement 2).Moreover, both preregistered sensitivity analyses yielded no difference in results in daily-life fall outcomes and primary analyses (eTables 4 and 5 in Supplement 2).The analysis of participants with a history of falls showed similar fall rates between the PBT (n = 28) and control (n = 29) groups (IRR, 0.91; 95% CI, 0.50-1.65).However, 45% fewer falls (IRR, 0.55; 95% CI, 0.28-1.08)were registered in the PBT group (n = 42) compared with the control group (n = 41) among participants without a history of falling (eTable 6 in Supplement 2).Multiple imputations of the laboratory fall outcomes did not significantly change the estimates (eAppendix 4 in Supplement 2).

Safety
No serious adverse events were reported in the present RCT.Nevertheless, 2 injuries occurred during PBT: a knee injury and a muscle strain in the thigh.The participant with the knee injury had 1 appointment with the general practitioner, but none of the injuries required any further health care treatment.An overview of the adverse events is available in eTable 7 in Supplement 2.
the current study was estimated based on an expected 50% decrease; hence, the trial was underpowered to detect a 22% between-group difference.It is noteworthy that the 22% decrease in daily-life fall rates was similar to the 23% reduction reported in a large systematic review of studies on physical exercise. 8In contrast to 80 minutes of total training in the PBT protocol, most studies in that systematic review applied multiple weekly sessions for 12 or more weeks. 8Thus, the findings of the present trial indicate that PBT is more time efficient than general exercise approaches.However, the results must be validated in larger trials with sufficient sample sizes to detect approximately 20% fall rate reductions.
Of the 4 RCTs with a primary aim of investigating the effects of PBT on daily-life falls, 23,28,32,33 3 used treadmill PBT and 1 used an overground walkway.These 3 RCTs found no beneficial effects on daily-life fall rates (IRR, 0.93-1.19), 28,32,33and 2 of these trials included only 1 type of perturbation (either slips or trips), 28,33 diminishing the transferability of these interventions since unidirectional perturbations are not associated with improved reactive balance control after perturbations in other directions. 10,44,45However, RCT investigated adding multidirectional treadmill PBT to usual balance training but showed no effects on daily-life fall rates (IRR, 1.11; 95% CI, 0.98-1.24). 32Its participants were categorized as having a high risk of falling, 70% of whom had experienced a fall in the preceding year. 32However, it has been shown that older adults with a history of falls may be less responsive to PBT than participants without a previous fall. 46In this trial, the sensitivity analysis found larger effects among participants without falls (IRR, 0.55; 95% CI, 0.28-1.08)than participants with a previous fall (IRR, 0.91; 95% CI, 0.50-1.65),suggesting that PBT may be most effective in primary prevention.However, the results join the current evidence, which has not yet demonstrated that treadmill PBT can prevent daily-life falls among community-dwelling older adults. 28,32,33ile treadmill PBT did not show significant decreases in daily-life falls, overground PBT may be more effective. 23Pai et al 23 showed that overground PBT reduced the fall rate by 49% and any falls by 51%.This finding suggested that overground walkway PBT offers superior adaptation transferability to daily life compared with treadmill PBT.A possible explanation is that overground walkway perturbations more closely resemble daily-life situations than perturbations induced by treadmill accelerations. 10For instance, recent evidence showed that repeated trips induced by treadmill accelerations did not improve reactive balance control in a subsequent obstacle-induced trip. 47In daily life, trips occur by obstructing the swinging leg, demanding elevation or lowering of the leg before a large forward step is taken to regain balance. 48In contrast, the response following a trip from a treadmill acceleration does not necessitate such an elevation or lowering of the leg before the compensatory step is taken, leading to biomechanical differences. 47These differences may have resulted in suboptimal adaptation transferability. 47,49Moreover, while treadmill slip PBT has been associated with better reactive balance control following a subsequent overground slip perturbation, greater improvements and retention were found after practicing on an overground walkway. 50Although treadmills may be less task specific, they are less space consuming, cheaper, and easier to operate than overground setups, making them more clinically feasible. 45Thus, developing more task-specific treadmill perturbations (eg, obstacle-induced trips) may facilitate clinically feasible and more effective PBT.

Limitations
This trial has several limitations.First, participants were unblinded to group randomization due to the nature of training interventions.Given that the initial fall reporting was a participant-reported outcome, the lack of blinding may have contributed to fewer falls reported in the PBT group.However, daily-life falls were collected according to the recommendations of the Prevention of Falls Network Europe. 37cond, there was no established dose-response association between PBT and daily-life falls.Thus, the applied intensity and dose of the PBT were selected based on previous studies showing the sustained effectiveness of PBT in laboratory falls.

Conclusions
In this RCT, participants who received an 80-minute PBT intervention experienced a statistically nonsignificant 22% reduction in daily-life fall rates.There was no significant effect on other daily-life fall-related metrics; however, a statistically significant decrease in falls was found in the laboratory setting.Further studies are warranted to corroborate this effect.

Figure 3 .
Figure 3.Comparison of the Number of Falls in the Intervention and Control Groups 90 Effect of Treadmill Perturbation-Based Balance Training on Fall Rates in Older Adults StataCorp LLC) and uploaded in REDCap, version 9.5.6 (Vanderbilt University) by a research member who was not involved in any training or testing activities.Allocation sequence concealment was ensured by a random block size of 2, 4, 6, or 8.All participants in the PBT group were assigned to four 20-minute PBT sessions, and participants in the control group were assigned to 4 treadmill walking training sessions; the initial 2 training sessions JAMA Network Open.2023;6(4):e238422.doi:10.1001/jamanetworkopen.2023.8422(Reprinted) April 20, 2023 2/12 Downloaded From: https://jamanetwork.com/ on 09/27/2023 generated in Stata (

Table 1 .
Baseline Characteristics of Participants a Tilburg Frailty Indicator: score range of 0 to 15, with a lower score indicating less frailty.b Vulnerable Elders-13 Survey: score range of 0 to 10, with a lower score indicating less vulnerable.c Short Falls Efficacy Scale: score range of 7 to 28, with a lower score indicating less concern about falling.d Short Orientation-Memory-Concentration Test: score range of 0 to 28, with a higher score indicating better performance.e Short Physical Performance Battery: score range of 0 to 12, with a higher score indicating better performance.f Time to complete the Trail Making Test Part A was subtracted from Part B time, with less time indicating better executive function.