Secondary Logo

Journal Logo

Research Articles

Interindividual Balance Adaptations in Response to Perturbation Treadmill Training in Persons With Parkinson Disease

Klamroth, Sarah MA; Gaßner, Heiko PhD; Winkler, Jürgen MD; Eskofier, Björn PhD; Klucken, Jochen MD; Pfeifer, Klaus PhD; Steib, Simon PhD

Author Information
Journal of Neurologic Physical Therapy: October 2019 - Volume 43 - Issue 4 - p 224-232
doi: 10.1097/NPT.0000000000000291



Balance dysfunction is highly prevalent in persons with Parkinson disease (PwPD), starting in the early stages and progressing during the disease course.1 It leads to impaired mobility, reduced quality of life,2 and contributes to an increased fall risk in Parkinson disease (PD)3,4 and is responsible for approximately one-third of falls in this population.5 Thus, effective intervention strategies for improving balance function and preventing falls in PwPD are urgently needed.

Balance dysfunction covers different dimensions of balance control in PwPD6 and is difficult to treat due to its complexity.7 There is good evidence showing that exercise,8 and particularly exercises addressing postural control,9 can improve balance function in PwPD, but has limited effects on fall incidence. One potential explanation discussed in the literature is a lack of task specificity and a call for exercise that involves practice of balance recovery mechanisms for fall prevention.10,11 Consequently, novel training concepts have emerged, which mimic unexpected falls by applying balance perturbations to participants while walking or standing. Among the different novel techniques for perturbation training (PBT), treadmill-based perturbations12–16 are one of the most common approaches. Treadmill-based PBT offers advantages since it provides a safe environment (harness and handrails), and is highly task-specific for fall prevention (eg, provoking slips or trips by sudden belt accelerations or changes of gait speed).

Two recently published reviews reported significant reductions in fall incidence (proportions of fallers or number of falls) after PBT among healthy elderly but also in some clinical populations, such as PD.17,18 Findings from motor learning studies19–21 observed improved postural responses after short-term PBT, where PwPD with poorer balance performance at baseline20,21 and older age19 showed most pronounced improvements. Further, studies investigating long-term effects of PBT showed a reduced fall rate and partially improved balance in PwPD.16,22,23 Although this preliminary evidence is promising, it also suggests that there are interindividual differences in response to PBT, which need further attention. Moreover, while PBT shows beneficial effects on fall incidence, little is known about the underlying mechanisms in balance adaptation, which are essential for further development of PBT as an effective approach for fall prevention in PwPD.

Recently, we evaluated the effects of an 8-week perturbation treadmill training (PTT) on gait and balance in PwPD,24 but our primary data analysis did not include a systematic evaluation of all domains of balance dysfunction.6 Further, we used group-based statistical approaches, which, first, are measures of central tendencies and do not allow evaluation of interindividual differences in treatment response, and secondly, do not necessarily indicate a meaningful change.25 On the contrary, other statistical methods like a responder analysis offer the opportunity to investigate individual effects on participants, and to calculate the proportion of participants showing a meaningful change (responder rates).26–28

The primary objective of this study was to evaluate interindividual differences in balance adaptations in PwPD in response to an 8-week program of PTT, compared with conventional treadmill training (CTT) without perturbations. To achieve this, the number of responders (responder rates) and the magnitude of response (individual change) were analyzed related to specific domains of balance dysfunction. Second, we aimed at identifying person-specific characteristics that may predict improved balance function in response to PTT.


Design and Participants

This study is a secondary data analysis of a single-blind randomized controlled trial (trial registration ID: NCT01856244), which provides a detailed description of the study design, participants, and intervention.24 Forty-three PwPD (age: 65,05 years; male/female: n = 30/13) were randomly assigned to either PTT or CTT, based on a computer-generated block randomization stratified by Hoehn & Yahr stage (H&Y 1-2 and H&Y 2.5-3.5). Both groups trained 40 minutes, twice per week, for 8 (to maximally 9) weeks. Recruitment of participants followed predefined inclusion and exclusion criteria, and group assignment was concealed.24 The study was approved by the ethics board of the Friedrich-Alexander-University Erlangen-Nürnberg (reference number: 181_12B) and participants gave written informed consent.


The experimental group (PTT) walked on a treadmill prototype, consisting of a standard medical treadmill (mercury, h/p/cosmos medical GmbH; Nußdorf, Germany) equipped with handrails and mounted on a tiltable platform construction (zebris Medical GmbH; Isny, Germany). The construction contains 3 pneumatic actuators (Airtec NXD Pneumatic; 4241N per actuator; 30-mm lifting capacity; Airtec Pneumatic GmbH; Kronberg, Germany) placed below the treadmill. Via a microcontroller (ATmega16 microcontroller), actuators are set to constantly varying positions, thereby inducing small 3-dimensional tilting movements of the treadmill (maximum 30 mm) and creating the characteristics of an uneven surface during walking (sudden changes of belt speed or belt direction were not included). Participants receiving CTT walked on the identical treadmill but without perturbations.

Each session comprised 30 minutes of training in the allocated condition, with an additional 5-minute warm-up and cool-down period prior to and following the training period, respectively. Based on participants' self-perceived exertion (Borg scale 6-20; target range 12-15) and difficulty of the task (Likert scale 1 “not difficult” to 7 “extremely difficult”; target range ≤5), treadmill speed was individually progressed during the intervention period (starting with 70% of individual comfortable overground walking speed). Participants were encouraged to walk without using handrails, but 7 participants (PTT = 3, CTT = 4) needed handrail support in 1 to 8 sessions (out of 16), and another 5 participants (PTT = 3, CTT = 2) for more than 8 sessions. Exercise sessions were supervised by trained physical therapists providing only standardized and rigorously restricted instructions.24

Data Collection and Analysis

Data were collected at baseline and following intervention during the “on” state of antiparkinson medication. All tests were conducted by an experienced assessor in a human movement laboratory, except for one measure (Pull-Test), which was conducted in parallel, by an experienced neurologist as part of a regular medical assessment at the local university hospital. Both assessors were blinded to the randomization procedure. To improve the quality of assessing individual change29,30 in our sample, we used a broad set of balance measures: Mini Balance Evaluation Systems Test (Mini-BESTest) (4 subscores),31 Timed Up and Go (TUG),32 dual-task TUG (counting backward in 3s from a randomly chosen number),32 Pull-Test (as part of Unified Parkinson's Disease Rating Scale-III [UPDRS-III]),33 and postural sway (center-of-pressure root mean square [COP RMS] assessed during 30 seconds of quiet stance with eyes closed with a pressure sensor matrix [108.4 × 47.4 cm, 7168 pressure/force sensors, FDM-T, zebris Medical GmbH, Isny, Germany]). Additionally, we assigned these balance assessments to specific domains of postural control, to provide a more differentiated picture of balance adaptations. For this classification, we used an accepted framework,6 which illustrates that balance dysfunction in PD is composed of balance during quiet stance, reactive postural adjustments, anticipatory postural adjustments, and dynamic postural control. The following tests were used to quantify changes in each of these domains:

  1. Balance during quiet stance: Mini-BESTest subscore sensory orientation, Postural sway
  2. Reactive postural adjustments: Mini-BESTest subscore reactive postural control, Pull-Test
  3. Anticipatory postural adjustments Mini-BESTest subscore anticipatory postural control
  4. Dynamic postural control: Mini-BESTest subscore dynamic gait, TUG, dual-task TUG

For the evaluation of interindividual differences in balance adaptations, we used the responder analysis as an innovative methodological approach, which classifies participants into responders (people with a meaningful improvement after therapy)26 and nonresponders (people with no meaningful improvement) based on a defined threshold value in the outcome of interest.28 To date there are a variety of methods, classified into anchor-based and distribution-based methods, which provide a dichotomous classification of responders and nonresponders. In anchor-based methods, change in the outcome of interest is compared with an external criterion (anchor), which can be based on either participants' perception or clinical experts.34 Distribution-based methods compare the observed change to measures of variability, and estimate thresholds based on distribution of data in the outcome of interest in a given sample.35 In the present analysis, we examined multiple estimates from anchor-based and distribution-based methods, and merged them into a single threshold value. This combination of approaches from a clinical and a statistical perspective is also called “triangulation of methods,”35,36 and is recommended to improve meaningfulness of threshold values.28,34,37,38

In total, we performed 8 responder analyses, 1 for each balance measure, respectively. As described earlier, we calculated 2 anchor-based thresholds (methods A and B) and 2 distribution-based thresholds (methods C and D) for every measure. The remaining 4 estimates were converged to a single threshold value, by calculation of the mean value. For anchor-based thresholds, we used a patient-based anchor by evaluating therapy success based on participants' perspective. Assessors who were blinded to group assignment asked participants after intervention whether they think the training was successful for them (yes/no) and why. If participants confirmed therapy success with “yes” and justified their decision with improvements in balance function, they were defined as anchor-based responders. For calculation of anchor-based thresholds, within-individuals change scores (method A: mean change in the outcome of interest within anchor-based responders) and between-individuals change scores (method B: mean changes in the outcome of interest between anchor-based responders and nonresponders) were calculated.34 For distribution-based threshold values, the standard error of measurement (method C: SEM) ()39 and effect size (method D: ES) ()40 were calculated. The final threshold value of every outcome measure represents the mean of anchor-based and distribution-based threshold values.

Statistical Analysis

All statistical analyses were performed with commercial statistics software (IBM SPSS version 25.0 for Windows, Armonk, New York). Initially all data were explored with descriptive data analysis and tested for normality using the Shapiro-Wilk test. Equality of variances was assessed with Levene's test. Depending on measurement scales and distribution of variables, differences at baseline between groups (PTT vs CTT) were tested using independent-samples t tests (interval variables and normally distributed), χ2 test (categorical variables), and Mann-Whitney U test (ordinal variables, nonnormal distribution). Responder rates (proportion of responders) were evaluated in the PTT and CTT groups, respectively. Relative risk ratios (RRs) with a 95% confidence interval (CI) were calculated to compare responder rates between groups. For the comparison of person-specific characteristics between responders and nonresponders, between-group differences were analyzed with independent-samples t tests (age, disease duration, and motor impairment) and Mann-Whitney U tests (cognition, disease stage, and balance confidence). Additionally, we descriptively analyzed magnitude of change in responders and nonresponders by plotting individual change scores (pre-post) for all balance measures.


Data from 37 participants (PTT: n = 18; CTT: n = 19) were available for this secondary data analysis. Demographic characteristics and baseline values of balance measures are provided in Table 1. There were no significant differences between intervention groups at baseline. Both groups increased their treadmill walking speed from T0 to T8, without a significant between-group difference (change PTT: +0.19 ± 0.11 m/s; CTT: +0.17 ± 0.10 m/s). The calculated threshold values (see the Table, Supplemental Digital Content 2, which provides details for calculation of thresholds, available at:, which were used to classify responders in every balance measure, are provided in Table 2.

Table 1. - Sample Characteristics
Perturbation Treadmill Training Conventional Treadmill Training
PTT (n = 18) CTT (n = 19)
Mean SD Mean SD P
Age, y 67.6 8.2 62.9 7.8 0.082
Gender (male/female) 11/7 15/4 0.235a
Height, cm 174.6 8.8 175.2 8.4 0.832
BMI, kg/m2 24.6 1.9 25.7 2.3 0.133
Disease duration, y 7.9 4.0 7.4 4.5 0.713
LEDD, mg/d 630.4 331.1 618.3 259.7 0.902
Hoehn & Yahr 2.6 0.5 2.5 0.6 0.468b
1-2, n 3 6
2.5-3.5, n 15 13
UPDRS-III (score) 17.7 6.1 20.0 8.2 0.337
MoCA (score 0-30) 25.9 3.8 25.5 4.1 0.951b
ABC Scale (score 0-100) 79.1 19.8 83.8 8.5 0.891b
COP RMS, mm 11.3 4.4 12.6 4.3 0.318b
Pull-Test (score 0-4) 1.0 0.5 0.9 0.5 0.498b
Mini-BESTest (score 0-28) 24.1 3.3 24.9 2.6 0.481b
Subscore sensory orientation (0-6) 5.6 0.6 5.7 0.6 0.433b
Subscore reactive postural control (0-6) 4.8 1.1 5.1 1.1 0.360b
Subscore anticipatory postural control (0-6) 5.2 1.0 5.5 0.6 0.485b
Subscore dynamic gait (0-10) 8.4 1.3 8.5 1.5 0.617b
TUG, s 9.9 2.6 9.2 1.4 0.296
Dual-task TUG, s 11.3 3.7 10.5 2.9 0.323
Abbreviations: ABC Scale, Activities-specific Balance Confidence Scale; BMI, body mass index; COP RMS, center-of-pressure root mean square; CTT, conventional treadmill training; LEDD, levodopa equivalent daily dose; Mini-BESTest, Mini Balance Evaluation Systems Test; MoCA, Montreal Cognitive Assessment; PTT, perturbation treadmill training; TUG, Timed Up and Go; UPDRS-III, Unified Parkinson's Disease Rating Scale-III.
aχ2 test.
bMann-Whitney U test.

Table 2. - Responder and Nonresponder Rates in PTT and CTT, With Relative Risk Ratios (With 95% CI)
Domain of Balance Dysfunction Outcome Measure Threshold Valuea Responder and Nonresponder Rates Difference in Response Rates Between Groups Relative Risk Ratio RR (95% CI)
PTT (n = 18) CTT (n = 19)
% n % n
Quiet stance COP RMSb, mm −1.42 Responder 23 4 16 3 1.49 (0.39-5.73)
Nonresponder 77 13 84 16
Sensory orientationc (0-6 points) 0.38 Responder 22 4 16 3 1.41 (0.36-5.44)
Nonresponder 78 14 84 16
Reactive postural adjustments Reactive postural controlc (0-6 points) 0.35 Responder 44 8 10 2 4.22d (1.03-17.28)
Nonresponder 56 10 90 17
Pull-Test (0-4 points) −0.50 Responder 39 7 26 5 1.48 (0.58-3.82)
Nonresponder 61 11 74 14
Anticipatory postural adjustments Anticipatory postural controlc (0-6 points) 0.45 Responder 22 4 21 4 1.06 (0.31-3.60)
Nonresponder 78 14 79 15
Dynamic postural control Dynamic gaitc (0-10 points) 0.30 Responder 39 7 10 2 3.69 (0.88-15.49)
Nonresponder 61 11 90 17
TUG, s −0.90 Responder 33 6 10 2 4.22 (0.52-34.29)
Nonresponder 67 12 90 17
Dual-task TUG, s −1.25 Responder 22 4 5 1 3.17 (0.73-13.7)
Nonresponder 78 14 95 18
Abbreviations: CI, confidence interval; COP RMS, center-of-pressure root mean square; CTT, conventional treadmill training; Mini-BESTest, Mini Balance Evaluation Systems Test; PTT, perturbation treadmill training; RR, risk ratio; TUG, Timed Up and Go.
a Threshold values: based on triangulation of distribution-based (standard error of measurement, effect size) and anchor-based methods (within- and between-subject change scores), for thresholds of COP RMS, Pull-Test, TUG, and dual-task TUG negative change indicate improvements.
bData from n = 17 available.
cMini-BESTest subscore.
dP value significant at level ≤ 0.05.

Responder Rates and Magnitude of Response

Effects in responder rates between groups (PTT and CTT) are presented in Table 2 and Figure 1. The largest differences in responder rates existed for the balance domain reactive postural adjustments (RR = 1.48-4.22), with a significant between-group effect in Mini-BESTest reactive postural control (PTT = 44%; CTT = 10%; RR = 4.22, CI 1.03-17.28). While were no significant between-groups effects for measures of dynamic postural control, based on evaluation of descriptive data, PTT showed a higher proportion of responders (PTT: 22%-39%) compared with CTT (CTT: 5%-10%) in all measures of dynamic postural control (Mini-BESTest dynamic gait, TUG, TUG dual-task; RR = 3.17-4.22). In the domains of balance during quiet stance (COP RMS, Mini-BESTest sensory orientation) and anticipatory postural adjustments (Mini-BESTest anticipatory postural control), small nonsignificant differences in responder rates (RR = 1.06-1.49) were observed in favor of PTT.

Figure 1.
Figure 1.:
Responder and nonresponder rates (percent) in both groups for all balance measures (assigned to 4 domains of balance dysfunction in Parkinson disease). COP, center-of-pressure root mean square; CTT, conventional treadmill training; PTT, perturbation treadmill training; TUG, Timed Up and Go.

The additional descriptive analysis of individual change of responders and nonresponders is presented in Figure 2 (domain reactive postural adjustments) and in Supplemental Digital Content 3 (see the Figure, Supplemental Digital Content 3, which illustrates individual change in all other domains, available at: for PTT and CTT, respectively. Overall, the magnitude of change in responders appeared not to be affected by group assignment. Among nonresponders, the majority (>50%) showed no or minimal change, and in both groups only very few individuals decreased their balance performance.

Figure 2.
Figure 2.:
Individual change scores of responders and nonresponders in PTT (perturbation treadmill training) and CTT (conventional treadmill training) for the domain reactive postural adjustments.

Characteristics of PTT Responders

Baseline characteristics of responders and nonresponders from the PTT group are presented in Table 3. Responders in the PTT group showed larger deficits in balance performance at baseline, with significant between-group differences in all Mini-BESTest subscores (P ≤ 0.01). Further, PTT responders presented larger cognitive impairments at baseline (PTT responders: Montreal Cognitive Assessment (MoCA) range = 23-26; PTT nonresponders: MoCA range = 26-28), showing significant differences in 3 measures (COP RMS: P = 0.05; Mini-BESTest reactive postural control: P = 0.05; Mini-BESTest dynamic gait: P = 0.03). Descriptive comparisons showed that PTT responders in most of the balance measures were older (PTT responders: age range = 69-73 years; PTT nonresponders: age range = 65-67 years) and higher in H&Y staging (PTT responders: H&Y range = 2.5-2.9; PTT nonresponders: H&Y range = 2.5-2.6), but a significant between-group difference existed only for age in COP RMS (P = 0.02). Regarding disease duration, motor impairment (UPDRS-III), and balance confidence (ABC Scale [Activities-specific Balance Confidence Scale]), no systematic differences between PTT responders and nonresponders were observed.

Table 3. - Characteristics of Responders and Nonresponders Exercising With Perturbation Treadmill Training
Outcome Measure Initial Balance Age Cognition Disease Duration Disease Stage Motor Impairment Balance Confidence
Mean Pa y P MoCA Pa y P H&Y Pa UPDRS-III P ABC Scale Pa
COP RMSb, mm Responder (n = 4) 16.29 0.13 73 0.02c 23 0.05c 9 0.62 2.8 0.40 20 0.46 81 0.61
Nonresponder (n = 13) 9.76 65 27 8 2.5 17 79
Sensory orientationd (0-6 points) Responder (n = 4) 4.75 0.00c 73 0.14 24 0.31 10 0.30 2.9 0.09 18 0.84 66 0.04c
Nonresponder (n = 14) 5.86 66 27 7 2.5 18 83
Reactive postural controld (0-6 points) Responder (n = 8) 4.00 0.00c 69 0.61 24 0.05c 7 0.64 2.6 0.92 22 0.01c 79 0.66
Nonresponder (n = 10) 5.50 67 28 8 2.5 14 80
Pull-Test (0-4 points) Responder (n = 7) 1.14 0.32 70 0.26 24 0.07 8 0.89 2.6 0.56 18 0.69 87 0.50
Nonresponder (n = 11) 0.91 66 27 8 2.5 17 74
Anticipatory postural controld (0-6 points) Responder (n = 4) 4.00 0.01c 73 0.16 23 0.22 10 0.18 2.8 0.36 19 0.77 84 0.87
Nonresponder (n = 14) 5.57 66 27 7 2.5 17 78
Dynamic gaitd (0-10 points) Responder (n = 7) 7.14 0.00c 71 0.15 24 0.03c 10 0.05c 2.7 0.30 17 0.90 78 0.79
Nonresponder (n = 11) 9.18 65 27 6 2.5 18 80
TUG, s Responder (n = 6) 10.97 0.25 71 0.30 26 0.89 8 0.97 2.6 0.96 17 0.59 71 0.30
Nonresponder (n = 12) 9.42 66 26 8 2.5 18 83
Dual-task TUG, s Responder (n = 4) 13.22 0.14 70 0.37 25 0.42 8 0.84 2.5 0.61 21 0.31 79 0.83
Nonresponder (n = 14) 10.77 67 26 8 2.6 17 79
Abbreviations: ABC Scale, Activities-specific Balance Confidence Scale; COP RMS, center-of-pressure root mean square; H&Y, Hoehn and Yahr disease stage; Mini-BESTest, Mini Balance Evaluation Systems Test; MoCA, Montreal Cognitive Assessment; TUG, Timed Up and Go; UPDRS-III, Unified Parkinson's Disease Rating Scale-III.
aMann-Whitney U test.
bData from n = 17 available.
cSignificant difference between groups (P ≤ 0.05).
dMini-BESTest subscore.


The overall aim of this study was to explore interindividual differences in balance adaptations in response to an 8-week PTT in PwPD. In comparison to CTT, PTT showed a significantly higher proportion of participants with meaningful improvements (responders) in the Mini-BESTest subscore reactive postural control. Additionally, descriptive analysis revealed larger responder rates in PTT for dynamic postural control, without significant between-group effects. Anticipatory postural control and balance during quiet stance revealed the lowest responder rates in both groups. Further, PTT responders showed a significantly lower initial balance performance (4/8 measures) and cognitive function (3/8 measures), and based on descriptive evaluation appeared to be older and in a higher disease stage, compared with nonresponders.

Responder Rates and Magnitude of Response

Based on previous research we hypothesized that PTT, in comparison to CTT, would reveal larger responder rates, and particularly in measures of reactive postural control. This hypothesis was only partially confirmed. Although descriptive analysis suggests more responders in PTT for the majority of balance measures, only in Mini-BESTest reactive postural control effects were significantly different between groups. These findings are not surprising since PTT is a highly task-specific motor training, which requires reactive and dynamic balance control to maintain balance during perturbed walking. Results from our study are in line with a previously published meta-analysis9 showing that exercises specifically addressing components of balance dysfunction are most beneficial for postural stability in PwPD. Further, it has been shown that PBT leads to improved neuromuscular adaptations in balance control in young healthy adults41 as well as in PwPD,42 which also could explain the improved balance function after PTT. However, considering the magnitude of individual change in both groups, the majority of participants showed no or minimal change in balance function. Thus, one could argue that most of the participants did not benefit from the treatment, and PTT may only be effective for a subset of PwPD. This interpretation would highlight the importance of analyzing specific factors that can predict treatment response. However, considering the progressive nature of the disease with its decline in postural control,1 a stabilization of balance function may also be interpreted as a positive effect. Epidemiologic studies showed that disease progression is not linear during the course of PD, and is more rapid in persons with late onset and with the PIGD (postural instability gait difficultly) form of PD.43 Thus, the course of PD is very individual, and since we evaluated short-term effects only, the interpretation of a stable balance function over 8 weeks remains speculative.

The largest proportion of responders was found in PTT in reactive postural adjustments, with 44% responders in the Mini-BESTest reactive postural control and 39% in the Pull-Test. But for the Pull-Test, results were not significantly different between groups (CTT: 26%). Considering the specific tasks in both tests, these findings may indicate that both groups improved reactive balance in posterior direction, and that only PTT additionally enhanced forward and lateral direction. Perturbations from our prototypic treadmill provoke balance adjustments in both, the sagittal and frontal planes, and thereby may have improved participants' reactive responses in multiple directions. These results are interesting with regard to fall prevention, since it has been proposed that exercise should be specific to fall direction.10

In contrast to enhanced reactive balance, anticipatory postural control improved in considerably fewer individuals (PTT: 22%). Similar results have been reported for persons with stroke, where PBT enhanced reactive balance yet no significant effects were seen in anticipatory balance and fall rate.44 Motor learning studies, investigating adaptations to short-term PBT (repeated slips during walking), observed adaptive changes in gait patterns (eg, reduced step length, a flat foot, and flexed knee at heel strike) and reduced fall incidence. These adaptations were attributed to enhanced feedforward and reactive control mechanisms, both leading to effective fall prevention.45,46 These previous findings suggest that effective fall prevention requires reactive as well as feedforward control, which occurs before or in anticipation of a perturbation. Since feedforward control mainly relies on previously acquired knowledge,46 applying various perturbations with multiple directions and amplitudes during dynamic and static motor tasks may help to address all relevant balance domains.17

Responders' Characteristics

Since responder rates were generally limited, we investigated specific characteristics of PTT responders to identify potential predictors for treatment response. The importance of such predictive factors in exercise has been constantly highlighted, particularly in PD.29,47–51 Based on previous research,19–21,52 we hypothesized that especially those with poorer balance and motor function at baseline would show balance response to PTT. Our results confirmed this hypothesis, at least for the Mini-BESTest subscores, suggesting that lower balance performance at baseline can predict improved balance function after PTT in PwPD. This enhances previous findings from short-term PBT, indicating that PwPD with worse balance function benefit most from short-term as well as from long-term PBT. However, participants in the present study had a relatively high level of balance function (Pull-Test: n = 34 scored ≤ 1; Mini-BESTest: n = 17 scored ≥ 26) and potential ceiling effects may have limited the chance to see improvements.

Responders' characteristics further indicate that cognitive impairments, higher age, and advanced disease stages may also predict improved balance function after PTT. Results for cognitive function are in line with a recently published study,53 showing that PwPD with cognitive impairments are most likely to benefit from a highly challenging balance training. Interestingly, these potential predictive variables, together with balance impairments, are also known to be associated with an increased fall risk in PwPD.3,4,54 Acute adaptations to a single session of PTT point in a similar direction, indicating that PwPD with a fall history exhibited more improved gait adaptations, compared with nonfallers.52 This suggests that PTT may be especially effective for individuals at high fall risk, supporting the idea of an individual-based approach. Notably, we analyzed a small set of potential outcome predictors, but there are many more factors potentially impacting balance response, particularly freezing of gait (FoG). Current literature focusing on the relationship between FoG and postural instability suggests that both can influence each other.55–57 Thus, future studies should consider FoG as an additional factor potentially associated with response to balance training.


There are some limitations that should be considered when interpreting the results of this study. First, our sample included mildly to moderately affected PwPD with relatively high functional and cognitive levels,58 which limits generalizability of the results. Second, our anchor-based methods are based on a dichotomous question, which may not be sufficiently differentiated to completely reflect participants' perspective. Third, our analysis focused on balance adaptations immediately after training and did not provide evidence related to sustainability of effects. Further, our findings are based on exploratory analyses in a relatively small sample, which lowers statistical power and limits the strength of conclusions. Finally, direct transfer into practice is hindered given the prototypic nature of the perturbation-based treadmill used in this study. Thus, although highly valuable for research purposes in the field of postural control and fall prevention, such systems need further development before widespread application in rehabilitation practice.


Based on responder analyses, this study provides first insights into balance adaptations in response to an 8-week PTT in PwPD. Findings from the present study suggest that PTT is particularly beneficial to improve reactive balance in PwPD, and there were some interesting findings in dynamic postural control that need further attention. In contrast, only very few PwPD improved in anticipatory balance, which, as well as reactive balance, plays a crucial role in fall prevention. Thus, future studies on PBT should evaluate whether integration of different perturbation types, during dynamic and static motor tasks, can improve multiple balance recovery strategies and thereby reduce fall incidence in this population. Another important finding is that PTT appeared to be effective only in a part of PwPD, especially in those with lower balance and cognitive function. These preliminary results are interesting, but more research is needed to better understand adaptations toward PBT and to develop effective fall prevention strategies in individuals with PD.


We would like to thank Marie-Kristin Dunker, Lyusyena Novokreshchenova, Surendar Devan, and Hasan Tariq for their support in supervising the training sessions. Furthermore, we thank Dr Zacharias Kohl, Dr Franz Marxreiter, Dr Johannes Schlachetzki, and Dr Martin Regensburger for their support in recruitment of participants, Dr Sebastian Krinner for providing orthopedic advice during the study, and Florence Theil for her support in language editing.


1. Kim SD, Allen NE, Canning CG, Fung VSC. Postural instability in patients with Parkinson's disease. Epidemiology, pathophysiology and management. CNS Drugs. 2013;27:97–112.
2. Soh S-E, Morris ME, McGinley JL. Determinants of health-related quality of life in Parkinson's disease: a systematic review. Parkinsonism Related Disord. 2011;17:1–9.
3. Allen NE, Schwarzel AK, Canning CG. Recurrent falls in Parkinson's disease: a systematic review. Parkinsons Dis. 2013;2013:906274.
4. Canning CG, Paul SS, Nieuwboer A. Prevention of falls in Parkinson's disease: a review of fall risk factors and the role of physical interventions. Neurodegener Dis Manag. 2014;4:203–221.
5. Michalowska M, Fiszer U, Krygowska-Wajs A, Owczarek K. Falls in Parkinson's disease. Causes and impact on patients' quality of life. Funct Neurol. 2005;20:163–168.
6. Schoneburg B, Mancini M, Horak F, Nutt JG. Framework for understanding balance dysfunction in Parkinson's disease. Mov Disord. 2013;28:1474–1482.
7. van der Kolk NM, King LA. Effects of exercise on mobility in people with Parkinson's disease. Mov Disord. 2013;28:1587–1596.
8. Tomlinson CL, Herd CP, Clarke CE, et al. Physiotherapy for Parkinson's disease: a comparison of techniques. Cochrane Database Syst Rev. 2014;6:CD002815.
9. Klamroth S, Steib S, Devan S, Pfeifer K. Effects of exercise therapy on postural instability in Parkinson disease: a meta-analysis. J Neurol Phys Ther. 2016;40:3–14.
10. Grabiner MD, Crenshaw JR, Hurt CP, Rosenblatt NJ, Troy KL. Exercise-based fall prevention: can you be a bit more specific? Exerc Sport Sci Rev. 2014;42:161–168.
11. Oddsson LIE, Boissy P, Melzer I. How to improve gait and balance function in elderly individuals—compliance with principles of training. Eur Rev Aging Phys Act. 2007;4:15–23.
12. Lurie JD, Zagaria AB, Pidgeon DM, Forman JL, Spratt KF. Pilot comparative effectiveness study of surface perturbation treadmill training to prevent falls in older adults. BMC Geriatr. 2013;13:49.
13. Hak L, Houdijk H, Steenbrink F, et al. Stepping strategies for regulating gait adaptability and stability. J Biomech. 2013;46:905–911.
14. Fasano A, Schlenstedt C, Herzog J, et al. Split-belt locomotion in Parkinson's disease links asymmetry, dyscoordination and sequence effect. Gait Posture. 2016;48:6–12.
15. Sessoms PH, Wyatt M, Grabiner M, et al. Method for evoking a trip-like response using a treadmill-based perturbation during locomotion. J Biomech. 2014;47:277–280.
16. Protas EJ, Mitchell K, Williams A, Qureshy H, Caroline K, Lai EC. Gait and step training to reduce falls in Parkinson's disease. NeuroRehabilitation. 2005;20:183–190.
17. Gerards MHG, McCrum C, Mansfield A, Meijer K. Perturbation-based balance training for falls reduction among older adults: current evidence and implications for clinical practice. Geriatr Gerontol Int. 2017;17:2294–2303.
18. Mansfield A, Wong JS, Bryce J, Knorr S, Patterson KK. Does perturbation-based balance training prevent falls? Systematic review and meta-analysis of preliminary randomized controlled trials. Phys Ther. 2015;95:700–709.
19. Barajas JS, Peterson DS. First-trial protective step performance before and after short-term perturbation practice in people with Parkinson's disease. J Neurol. 2018;265:1138–1144.
20. Peterson DS, Dijkstra BW, Horak FB. Postural motor learning in people with Parkinson's disease. J Neurol. 2016;263:1518–1529.
21. van Ooteghem K, Frank JS, Horak FB. Postural motor learning in Parkinson's disease: the effect of practice on continuous compensatory postural regulation. Gait Posture. 2017;57:299–304.
22. Shen X, Mak MKY. Technology-assisted balance and gait training reduces falls in patients with Parkinson's disease: a randomized controlled trial with 12-month follow-up. Neurorehabil Neural Repair. 2015;29:103–111.
23. Smania N, Corato E, Tinazzi M, et al. Effect of balance training on postural instability in patients with idiopathic Parkinson's disease. Neurorehabil Neural Repair. 2010;24:826–834.
24. Steib S, Klamroth S, Gaßner H, et al. Perturbation during treadmill training improves dynamic balance and gait in Parkinson's disease: a single-blind randomized controlled pilot trial. Neurorehabil Neural Repair. 2017;31:758–768.
25. Turk DC. Statistical significance and clinical significance are not synonyms!. Clin J Pain. 2000;16:185–187.
26. Haley SM, Fragala-Pinkham MA. Interpreting change scores of tests and measures used in physical therapy. Phys Ther. 2006;86:735–743.
27. Bolton JE. Sensitivity and specificity of outcome measures in patients with neck pain: detecting clinically significant improvement. Spine. 2004;29:2410–2417; discussion 2418.
28. Snapinn SM, Jiang Q. Responder analyses and the assessment of a clinically relevant treatment effect. Trials. 2007;8:31.
29. Hecksteden A, Kraushaar J, Scharhag-Rosenberger F, Theisen D, Senn S, Meyer T. Individual response to exercise training—a statistical perspective. J Appl Physiol 1985. 2015;118:1450–1459.
30. Scharhag-Rosenberger F, Walitzek S, Kindermann W, Meyer T. Differences in adaptations to 1 year of aerobic endurance training: individual patterns of nonresponse. Scand Med Sci Sports. 2012;22:113–118.
31. Franchignoni F, Horak F, Godi M, Nardone A, Giordano A. Using psychometric techniques to improve the Balance Evaluation Systems Test: the mini-BESTest. J Rehabil Med. 2010;42:323–331.
32. Shumway-Cook A, Brauer S, Woollacott M. Predicting the probability for falls in community-dwelling older adults using the Timed Up & Go Test. Phys Ther. 2000;80:896–903.
33. Movement Disorder Society Task Force on Rating Scales for Parkinson's Disease. The Unified Parkinson's Disease Rating Scale (UPDRS): status and recommendations. Mov Disord. 2003;18:738–750.
34. Copay AG, Subach BR, Glassman SD, Polly DW Jr, Schuler TC. Understanding the minimum clinically important difference: a review of concepts and methods. Spine J. 2007;7:541–546.
35. Revicki D, Hays RD, Cella D, Sloan J. Recommended methods for determining responsiveness and minimally important differences for patient-reported outcomes. J Clin Epidemiol. 2008;61:102–109.
36. Leidy NK, Wyrwich KW. Bridging the gap: using triangulation methodology to estimate minimal clinically important differences (MCIDs). COPD. 2005;2:157–165.
37. Cella D, Eton DT, Lai JS, Peterman AH, Merkel DE. Combining anchor and distribution-based methods to derive minimal clinically important differences on the Functional Assessment of Cancer Therapy (FACT) anemia and fatigue scales. J Pain Symptom Manag. 2002;24:547–561.
38. Yost KJ, Eton DT. Combining distribution- and anchor-based approaches to determine minimally important differences: the FACIT experience. Eval Health Prof. 2005;28:172–191.
39. Wyrwich KW, Tierney WM, Wolinsky FD. Further evidence supporting an SEM-based criterion for identifying meaningful intra-individual changes in health-related quality of life. J Clin Epidemiol. 1999;52:861–873.
40. Kazis LE, Anderson JJ, Meenan RF. Effect sizes for interpreting changes in health status. Med Care. 1989;27:S178–S189.
41. Krause A, Freyler K, Gollhofer A, et al. Neuromuscular and kinematic adaptation in response to reactive balance training—a randomized controlled study regarding fall prevention. Front Physiol. 2018;9:1075.
42. Pasluosta CF, Steib S, Klamroth S, et al. Acute neuromuscular adaptations in the postural control of patients with Parkinson's disease after perturbed walking. Front Aging Neurosci. 2017;9:638.
43. Jankovic J. Parkinson's disease: clinical features and diagnosis. J Neurol Neurosurg Psychiatry. 2008;79:368–376.
44. Mansfield A, Aqui A, Danells CJ, et al. Does perturbation-based balance training prevent falls among individuals with chronic stroke? A randomised controlled trial. BMJ Open. 2018;8(8).
45. Pai Y-C, Yang F, Bhatt T, Wang E. Learning from laboratory-induced falling: long-term motor retention among older adults. Age (Dordr). 2014;36:9640.
46. Bhatt T, Pai YC. Generalization of gait adaptation for fall prevention: from moveable platform to slippery floor. J Neurophysiol. 2008;101:948–957.
47. Abbruzzese G, Marchese R, Avanzino L, Pelosin E. Rehabilitation for Parkinson's disease: current outlook and future challenges. Parkinsonism Related Disord. 2016;22(suppl 1):S60–S64.
48. Ellis TD, Dibble LE, Peterson DS. Moving beyond effectiveness. J Neurol Phys Ther. 2019;43:1–2.
49. Buford TW, Pahor M. Making preventive medicine more personalized: implications for exercise-related research. Prev Med. 2012;55:34–36.
50. Landers MR, Lopker M, Newman M, Gourlie R, Sorensen S, Vong R. A cross-sectional analysis of the characteristics of individuals with Parkinson disease who avoid activities and participation due to fear of falling. J Neurol Phys Ther. 2017;41:31–42.
51. Whipple MO, Schorr EN, Talley KMC, Lindquist R, Bronas UG, Treat-Jacobson D. Variability in individual response to aerobic exercise interventions among older adults. J Aging Phys Act. 2018;26(4):655–670.
52. Klamroth S, Steib S, Gaßner H, et al. Immediate effects of perturbation treadmill training on gait and postural control in patients with Parkinson's disease. Gait Posture. 2016;50:102–108.
53. Lofgren N, Conradsson D, Joseph C, Leavy B, Hagstromer M, Franzen E. Factors associated with responsiveness to gait and balance training in people with Parkinson disease. J Neurol Phys Ther. 2019;43:42–49.
54. Contreras A, Grandas F. Risk of falls in Parkinson's disease: a cross-sectional study of 160 patients. Parkinsons Dis. 2011;2012:362572.
55. Plotnik M, Giladi N, Hausdorff JM. Is freezing of gait in Parkinson's disease a result of multiple gait impairments? Implications for treatment. Parkinsons Dis. 2012;2012:459321.
56. Schlenstedt C, Muthuraman M, Witt K, Weisser B, Fasano A, Deuschl G. Postural control and freezing of gait in Parkinson's disease. Parkinsonism Related Disord. 2016;24:107–112.
57. Bekkers EMJ, Dijkstra BW, Heremans E, Verschueren SMP, Bloem BR, Nieuwboer A. Balancing between the two: are freezing of gait and postural instability in Parkinson's disease connected? Neurosci Biobehav Rev. 2018;94:113–125.
58. Gill DJ, Freshman A, Blender JA, Ravina B. The Montreal cognitive assessment as a screening tool for cognitive impairment in Parkinson's disease. Mov Disord. 2008;23:1043–1046.

exercise; fall prevention; perturbation; responder; treadmill

Supplemental Digital Content

© 2019 Academy of Neurologic Physical Therapy, APTA.