RESEARCH ARTICLES
Determinants of Dual-Task Training Effect Size
in Parkinson Disease: Who Will Benefit Most?
Carolien Strouwen, PhD, Esther A. L. M. Molenaar, MSc, Liesbeth Münks, MSc, Sanne Broeder, MSc,
Pieter Ginis, PhD, Bastiaan R. Bloem, PhD, Alice Nieuwboer, PhD, and Elke Heremans, PhD
Background and Purpose: Dual-task interventions show positive
effects in people with Parkinson disease (PD), but it remains unclear
which factors determine the size of these benefits. As a secondary
analysis of the DUALITY trial, the aim of this study was to assess
the determinants of the effect size after 2 types of dual-task practice.
Methods: We randomly allocated 121 participants with PD to re-
ceive either integrated or consecutive dual-task training. Dual-task
walking performance was assessed during (i) a backward digit span
task (digit), (ii) an auditory Stroop task (Stroop), and (iii) a func-
tional mobile phone task. Baseline descriptive, motor, and cognitive
variables were correlated with the change in dual-task gait velocity
after the intervention. Factors correlated with the change in dual-task
gait velocity postintervention (P < 0.20) were entered into a stepwise
forward multiple linear regression model.
Results: Lower dual-task gait velocity and higher cognitive capacity
(Scales for Outcomes in Parkinson’s Disease-Cognition [ScopaCog])
at baseline were related to larger improvements in dual-task gait
velocity after both integrated and consecutive dual-task training for
all 3 tasks (β[gait] = −0.45, β[ScopaCog] = 0.34, R2 = 0.23, P <
0.001, for digit; β[gait] = −0.52, β[ScopaCog] = 0.29, R2 = 0.26,
P < 0.001, for Stroop; and β[gait] = −0.40, β[ScopaCog] = 0.30,
R2 = 0.18, P < 0.001, for mobile phone task).
Discussion and Conclusions: Participants with PD who showed a
slow dual-task gait velocity and good cognitive functioning at baseline
Neuromotor Rehabilitation Research Group, Department of Rehabilitation
Sciences, KU Leuven, Belgium (C.S., L.M., S.B., P.G., A.N., E.H.); and
Department of Neurology, Nijmegen Centre for Evidence Based Prac-
tice (E.A.M.) and Department of Neurology, Donders Institute for Brain,
Cognition and Behaviour (B.R.B.), Radboud University Medical Center,
Nijmegen, the Netherlands.
This work was funded by the Jacques and Gloria Gossweiler Foundation and
the Malou-Malou funds of the King Baudouin foundation. Elke Heremans
is a postdoctoral researcher and Sanne Broeder a PhD researcher of the
Research Foundation Flanders. The funding sources were not involved in
study design or execution.
This study presents results from 121 participants included in the DUALITY
trial conducted in 2 clinical centers in Belgium and the Netherlands, regis-
tered on clinicaltrials.gov as NTC01375413.
The authors declare no conflict of interest.
Supplemental digital content is available for this article. Direct URL citation
appears in the printed text and is provided in the HTML and PDF versions
of this article on the journal’s Web site (www.jnpt.org).
Correspondence: Elke Heremans, PhD, Department of Rehabilitation Sci-
ences, KU Leuven, Tervuursevest 101 bus 1501, 3001 Leuven, Belgium
(Elke.Heremans@kuleuven.be)
Copyright C 2019 Academy of Neurologic Physical Therapy, APTA.
ISSN: 1557-0576/19/4301-0003
DOI: 10.1097/NPT.0000000000000247
JNPT r Volume 43, January 2019
Copyright © 2019 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
benefited most from the dual-task training, irrespective of the type of
training and type of dual-task outcome.
Video Abstract available for more insights from the authors (see
Video, Supplemental Digital Content 1, available at: http://links.lww.
com/JNPT/A242).
Key words: cognition, dual task, gait, human movement system,
prediction
(JNPT 2019;43: 3–11)
INTRODUCTION
T he DUALITY trial, a large-scale multicenter study on the
effects of dual-task (DT) training in people with Parkin-
son disease (PD), recently showed that 2 different training
programs had equally positive effects on DT motor and cog-
nitive performance without increasing fall risk.1 In this trial,
consecutive task training (CTT, training 2 tasks separately) was
contrasted with integrated task training (IDT, practicing 2 tasks
simultaneously). Both regimens were assumed to improve DT
outcomes via different mechanisms: (1) by increasing single-
task automaticity or fluidity (CTT) and 2) by improving actual
task integration (IDT). It was also expected that in IDT a better
retention of practice would be found at the expense of a higher
fall risk, a hypothesis that was not confirmed. These very sim-
ilar outcomes of training bring up a clinical dilemma, namely
how clinicians can determine the best training approach for
each individual. Therefore, this study aimed to inform clin-
ical decision-making by increasing insight into DT training
responses.
Thus far, a first exploratory analysis has been conducted
to compare the training response of subgroups based on di-
chotomous scoring of the presence of falls, gait freezing,
Hoehn and Yahr (H&Y) stage II or III, or having cognitive im-
pairment (a cutoff score at the MoCA of 27). Results showed
that no differences were apparent between these subgroups.1
Cross-sectional studies in participants with PD show that DT
gait velocity, irrespective of training, is predicted by several
motor, cognitive, and personal factors.2 Our own recent study
identified the combination of single-task (ST) gait velocity
and the score on the Alternating Intake Test, a test of ex-
ecutive function, as the key factors determining DT gait ve-
locity in PD.3 In persons with stroke, improvement in gait
velocity after training has been found to be associated with
age and gait velocity at the beginning of the intervention.4 In
older individuals, both ST and DT variabilities were important
3
Strouwen et al
in predicting DT practice effects.5 Furthermore, executive
function (inhibitory control, mental set shifting, and atten-
tional flexibility) appeared to be an important factor determin-
ing changes in ST mobility after 12 months of intervention.6 To
date, no studies have been performed in participants with PD to
identify factors that determine the size of DT training effects.
Cognitive dysfunction has previously been found to re-
late to learning deficits in people with PD.7-9 Individuals with
freezing of gait (FOG) were shown to suffer from more se-
vere cognitive dysfunction10-12 and consequently to experi-
ence greater deficits in learning than nonfreezers.11,13,14 The
impaired learning profile of persons with FOG as well as the
fact that the frequency of FOG is known to increase under DT
conditions15 predicts that IDT may be less effective in this sub-
group of individuals. In addition, IDT may be contraindicated
for individuals in the later disease stages and with a higher fall
risk to guarantee safety during training.12,16-18
The main aim of this study was to identify which factors
determine the change in DT gait velocity after DT training.
Based on previous studies in other groups, gains in DT per-
formance were hypothesized to be lower in those with initial
slower gait velocity, more FOG, more advanced disease stage
and age, and a higher prevalence of falling. The second aim of
this study was to assess whether treatment benefits depended
on the type of DT training strategy that was used. As atten-
tional resources were challenged to a greater extent in IDT,
it was expected that participants with cognitive problems, in-
cluding those with FOG, and participants with an increased
risk of falling would benefit less from this type of training.
METHODS
Participants
This study presents results from 121 participants in-
cluded in the DUALITY trial conducted in 2 clinical centers in
Belgium and the Netherlands (registered on clinicaltrials.gov
as NTC01375413).19 Inclusion criteria were (a) diagnosis of
PD according to the UK Brain Bank criteria20 ; (b) H&Y stage
3 or less in the subjective best “on” phase of the medication
cycle21 ; (c) being able to walk for 10 minutes continuously; (d)
the presence of DT interference as established by a structured
checklist19 ; (e) a score 24 or more on the MMSE22 ; (f) stable
medication over the past 3 months or stable DBS settings over
the past year; and (g) no hearing or visual problems that could
cause interference with testing or training. The study was eth-
ically approved in both centers (Belgium: CME KU Leuven,
B322201213165/S53419, and the Netherlands: CMO Regio
Arnhem-Nijmegen, NL39530.091.12) and participants signed
an informed consent form before study participation.
Predictors and Outcome Measures
Testing was performed as described in the previously
published protocol of the DUALITY trial.19 All indepen-
dent variables were derived from the first baseline assessment
and included disease descriptors, cognitive scores, and motor
scores. In all participants, ST and DT gait velocity (cm/sec) at
preferred speed was assessed before and after the intervention
by means of the GAITRite Walkway System,23 whereby walk-
ing started and ended 1 m before/after the walkway. Two cogni-
4 
Copyright © 2019 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
JNPT r Volume 43, January 2019
tive tasks (ie, a backward digit span task [digit] and an auditory
Stroop task [Stroop]) and 1 functional task (mobile phone task,
requiring typing in the date of the test on a mobile phone) were
used to assess DT performance while walking,19 with the in-
struction of focusing equally on both tasks. The backward digit
span task was trained during both types of dual-task training,
whereas both other tasks were untrained to capture the transfer
effects of training. Reaction times for these tasks were assessed
both under ST and DT conditions. Descriptors included (i) age,
(ii) disease duration, (iii) history of recurrent falls in the previ-
ous year (yes/no), (iv) levodopa-equivalent dose (LED),24 and
(v) presence of a deep brain stimulator (DBS) (yes/no). Cog-
nitive scores consisted of (i) Mini-Mental State Examination
(MMSE),22 (ii) Montreal Cognitive Assessment (MoCA),25
(iii) Frontal Assessment Battery (FAB),26 (iv) Scales for Out-
comes in Parkinson’s Disease-Cognition (ScopaCog),27 (v) Al-
ternating Names Test,28 (vi) Alternating Intakes Test,28 and
(vii) ST and DT reaction times on the cognitive tasks at base-
line. Motor scores included (i) H&Y stage,21 (ii) Movement
Disorders Society Unified Parkinson’s Disease Rating Scale—
part 3 (MDS-UPDRS-III),29 (iii) New Freezing of Gait Ques-
tionnaire (NFOG-Q),30 (iv) Activities-specific Balance Con-
fidence scale (ABC),31 and (v) ST and DT gait velocity at
baseline.
Testing was repeated 4 times over a period of 6 months
of follow-up and was performed in the gait laboratories of the
Neuromotor Research Group at KU Leuven or of the Rad-
boud University Medical Center. Baseline performance was
assessed at 2 different times, separated by a control period of
6 weeks. After the second assessment, all participants under-
went 6 weeks of DT intervention. Immediately afterward, the
training effect was assessed. After 12 weeks of follow-up, a fi-
nal assessment investigated potential retention effects. Change
in gait velocity after training was calculated by subtracting the
average DT gait performance during the 2 baseline tests from
the DT gait performance during the first postintervention test.
Intervention
As described in the DUALITY trial protocol,19 for all
participants training consisted of 3 identical components: (1)
gait practice, (2) auditory cognitive exercise, and (3) functional
training. Gait practice involved at-home gait exercises, aimed
to improve gait quality. Exercises depended on the clinical
need of the individual and included, for example, exercises
such as focusing on the taking big steps, focusing on heel
strike, turning, and other key parameters of gait. Difficulty
level was gradually increased over the training program. Cog-
nitive exercises addressed verbal fluency, discrimination and
decision-making, working memory, mental tracking, and reac-
tion time. The intervention included four 30-minute practice
sessions per week, of which 2 were performed in the pres-
ence of a physical therapist who was specifically trained and
experienced in treating individuals with PD. Participants were
randomly assigned to either CTT or IDT.1 During CTT, 15
minutes of gait exercises, and 15 minutes of cognitive training
exercises, were provided while participants were seated on a
chair. In contrast, IDT consisted of motor-cognitive tasks dur-
ing the entire session. All training sessions were performed
while participants where “on” medication.
C 2019 Academy of Neurologic Physical Therapy, APTA
JNPT r Volume 43, January 2019 Determinants of Dual-Task Training Effect Size in Parkinson Disease

Statistical Analysis
Data analysis was performed with IBM SPSS, version
22 (IBM Corp, Armonk, New York).32,33 Missing data were
due to sporadic dropout or technical issues and were not in-
cluded in the analysis. A regression analysis was performed
to examine factors that determine the effect of DT training
on the change in gait velocity while performing (1) digit, (2)
Stroop, and (3) mobile phone tasks. Following the main re-
gression analysis, an additional subanalysis was performed to
identify determinants in both types of training separately. Cor-
relations between determinants and outcome variables were
assessed using simple linear regression models.32,33 For the
main analysis, as well as for the secondary subanalysis, fac-
tors were included into a stepwise forward multivariate linear
regression model when they were correlated with the outcome
variable, with a P value less than 0.20. Variables with a mu-
tual correlation higher than r = 0.70 were not entered into the
model to avoid multicollinearity.
RESULTS
Descriptive Data
Descriptive data are presented in Table 1. No differences
were found between the 2 training groups concerning descrip-
tive, cognitive, and motor predictors. As previously reported,
the change in DT performance after training did not differ
between the groups.1
Change in DT Gait Velocity—Digit in All
Participants
Table 2 presents the factors that exhibit univariate cor-
relations at P < 0.20 with the outcome variable, and were

Table 1. Descriptive Data for Independent and Dependent Variables of the Total Group, the Consecutive Task Training
Group, and the Integrated Task Training Group
Total Group CTT IDT CTT vs IDT:
N Mean (SD) N Mean (SD) N Mean (SD) P Value
Descriptive
Age, y 121 65.93 (9.22) 65 66.05 (9.30) 56 65.80 (9.19) 0.886
Disease duration, y 121 8.67 (5.83) 65 8.89 (6.30) 56 8.41 (5.29) 0.653
Falling in previous year, yes/no 121 40/81 65 22/43 56 18/38 0.843
LEDa , mg/24 h 121 610.00 [387.50-865.80] 65 710 [427.50-931.67] 56 612.99 (396.10) 0.060
DBSa , n 121 0 [0-0] 65 0 [0-0] 56 0 [0-0] 0.733
Cognitive
MMSE (0-30) 121 27.94 (1.59) 65 27.88 (1.65) 56 28.02 (1.53) 0.630
MoCAa (0-30) 121 26 [24-28] 65 25.71 (2.82) 56 25.93 (2.76) 0.665
FAB (0-18) 121 15.76 (1.99) 65 15.77 (1.97) 56 15.75 (2.03) 0.958
ScopaCog (0 – 43) 121 27.02 (5.67) 65 27.26 (5.36) 56 26.73 (6.05) 0.611
ANTa , s 120 24.19 [17.12-35.35] 65 24.00 [18.30-36.78] 55 24.38 [16.14-34.60] 0.202
AITa , s 120 35.01 [25.86-46.50] 65 39.00 [27.35-50.00] 55 32.50 [23.20-43.10] 0.255
ST reaction time, digit span 118 874.50 [559.17-1234.69] 63 852.52 [566.46-1204.32] 55 896.49 [525.53-1271.37] 0.694
taska , s
ST reaction time, Stroop task, s 119 1174.26 (201.98) 63 1177.80 (181.84) 56 1170.29 (224.10) 0.841
DT reaction time, digit span 117 795.48 [556.23-1168.84] 63 762.82 [520.31-1116.81] 54 940.84 (488.45) 0.444
taska , s
DT reaction time, Stroop task, s 118 1194.56 (199.24) 63 1191.05 (146.86) 55 1198.59 (247.41) 0.839
Motor
H&Ya (0-5) 120 2 [2-3] 65 2 [2-3] 55 2.00 [2-3] 0.691
MDS-UPDRS-III (0-132) 120 32.39 (12.79) 65 32.51 (13.71) 55 32.25 (11.73) 0.915
NFOG-Qa (0-28) 121 0.00 [0.00-13.00] 65 0.00 [0.00-13.00] 56 2.00 [0.00-14.00] 0.567
ABCa (0-100) 121 78.75 [66.88-92.50] 65 78.25 (17.50) 56 78.13 [61.72-92.34] 0.391
ST gait velocity, cm/s 119 106.70 (20.07) 64 107.38 (19.68) 55 105.91 (20.66) 0.692
DT gait velocity during digit 120 90.57 (20.56) 65 90.32 (19.14) 55 90.87 (22.30) 0.885
span task, cm/s
DT gait velocity during Stroop 121 92.07 (20.83) 65 92.63 (19.95) 56 91.43 (21.96) 0.752
task, cm/s
DT gait velocity during mobile 121 79.41 (22.08) 65 79.59 (22.17) 56 79.21 (22.18) 0.924
phone task, cm/s
Outcomes
Change in DT gait velocity 105 10.28 (14.55) 55 11.20 (14.34) 50 9.28 (14.86) 0.503
during digit span task, cm/s
Change in DT gait velocity 114 11.30 (13.51) 61 10.34 (13.56) 53 12.41 (13.50) 0.417
during Stroop task, cm/s
Change in DT gait velocity 114 7.96 (13.50) 61 9.30 (13.64) 53 6.41 (13.30) 0.256
during mobile phone task,
cm/s
Abbreviations: ABC, Activities-specific Balance Confidence scale; AIT, Alternating Intake Test; ANT, Alternating Names Test; CTT, consecutive task training; DBS, deep brain
stimulator; DT, dual task; FAB, Frontal Assessment Battery; H&Y, Hoehn and Yahr stage; IDT, integrated task training; LED, levodopa-equivalent dose; MDS-UPDRS-III, Unified
Parkinson’s Disease Rating Scale—motor part; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment; NFOG-Q, New Freezing of Gait Questionnaire;
ScopaCog, Scales for Outcomes in Parkinson’s Disease-Cognition; SD, standard deviation; ST, single task.
a
In the event of abnormally distributed data, the median [interquartile range] is provided.


C 2019 Academy of Neurologic Physical Therapy, APTA 5

Copyright © 2019 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
Strouwen et al JNPT r Volume 43, January 2019

Abbreviations: ABC, Activities-specific Balance Confidence scale; AIT, Alternating Intakes Test; ANT, Alternating Names Test; DBS, deep brain stimulator; DT, dual task; FAB, Frontal Assessment Battery; H&Y, Hoehn and
Yahr stage; LED, levodopa-equivalent dose; MDS-UPDRS-III, Unified Parkinson’s Disease Rating Scale—motor part; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment; NFOG-Q, New Freezing of
included into a stepwise forward multivariate linear regres-

r = −0.37

r = −0.36

r = −0.27
P < 0.001

P < 0.001

P = 0.004
DT Gait
Velocity
sion model. The change in DT gait velocity during the digit
span task after training was correlated with the MMSE, MoCA,
FAB score, ScopaCog, ST and DT reaction time on the digit
r = −0.20

r = −0.18
P = 0.04

P = 0.06
Velocity
ST Gait

span task, MDS-UPDRS-III, NFOG-Q, and ST and DT gait

–
velocity during the digit span task at baseline. Stepwise for-
ward linear regression (Table 3) revealed that lower DT gait
velocity during the digit span task at baseline (β = −0.45;
ABC
–

–
Table 2. Total Group Results for Pearson Correlation Coefficients Resulting From the Univariate Analysis Significant at Level P < 0.20
Motor

P < 0.001) and higher performance on the ScopaCog, reflect-

ing better cognitive functioning (β = 0.34; P < 0.001), related
r = −0.14

r = −0.17
NFOG-Q

P = 0.15

P = 0.07

significantly to the change in DT performance after training

–

(R2 = 0.23; P < 0.001). The differences between the observed

and predicted values based on the multiple linear regression
UPDRS III
r = −0.16

r = −0.15
P = 0.11

P = 0.12
MDS-

analysis are shown in the Figure for the digit span (top), Stroop
–

(center), and mobile phone tasks (bottom).

H&Y

Change in DT Gait Velocity—Stroop in All

–

Participants
DT Reaction

P = 0.06
r = 0.19

Change in DT gait velocity during the Stroop task was

Time

correlated with the MMSE, MoCA, FAB score, ScopaCog,

MDS-UPDRS-III, NFOG-Q, and ST and DT gait velocity dur-
ST Reaction

ing the Stroop task at baseline (Table 2). After forward linear
P = 0.01
r = 0.27
Time

regression (Table 3), lower DT gait velocity during the Stroop

–

task at baseline (β = −0.52; P < 0.001), higher ScopaCog

performance (β = 0.29; P = 0.002), and lower MDS-UPDRS-
AIT

III scores, reflecting better motor function (β = −0.25; P =

–

0.005), remained significant determinants for change in DT

Gait Questionnaire; r, regression coefficient; ScopaCog, Scales for Outcomes in Parkinson’s Disease-Cognition; ST, single task.

gait velocity during the Stroop task after training (R2 = 0.26;
ANT
–

–
Cognitive

P < 0.001). The difference between the observed and pre-

dicted values based on the multiple linear regression analysis
ScopaCog

P = 0.03

P = 0.05

P = 0.08
r = 0.22

r = 0.18

r = 0.16

is shown in the Figure.

Change in DT Gait Velocity—Mobile Phone

P = 0.19

P = 0.08
r = 0.13

r = 0.17
FAB

in All Participants
–

Univariate analysis (Table 2) showed that the LED,

P = 0.03

P = 0.03

P = 0.03

MoCA, ScopaCog, and DT gait velocity during the mobile

r = 0.22

r = 0.21

r = 0.20
MoCA

phone task at baseline were correlated with change in DT gait

velocity during the mobile phone task after training. Forward
P = 0.10

P = 0.06
r = 0.16

r = 0.18
MMSE

linear regression (Table 3) revealed that lower DT gait veloc-

–

ity during the mobile phone task at baseline (β = −0.40; P <

0.001), higher ScopaCog performance (β = 0.30; P = 0.002),
and higher LED (β = 0.19; P = 0.03) significantly deter-
DBS
–

mined change in DT gait velocity during the mobile phone task

(R2 = 0.18; P < 0.001). The difference between the observed
P = 0.01
r = 0.23
LED

and predicted values based on the multiple linear regression

–

analysis is shown in the Figure.

Descriptive

Falls
–

Consecutive Versus Integrated Task Training

Group
Duration
Disease

Pearson correlation coefficients at P < 0.20 resulting

–

from the univariate analysis in consecutive and integrated task

training groups are shown in the Appendix. In the CTT group,
Age
–

significant determinants of change in DT gait velocity during

the trained digit span task (Table 4) were lower baseline DT gait
velocity during Stroop
velocity during digit

velocity during the digit span task (β = −0.37; P = 0.007) and

span task (n =103)

mobile phone task

Change in DT gait

Change in DT gait

Change in DT gait

higher performance on the ScopaCog (β = 0.31; P = 0.021)

velocity during
task (n = 112)

(R2 = 0.14; P = 0.007). Results for gait velocity were sim-

(n = 114)

ilar in the IDT group (R2 = 0.38; p < 0.001) (β = −0.51;

P < 0.001). However, in this group higher MoCA scores

6 
C 2019 Academy of Neurologic Physical Therapy, APTA

Copyright © 2019 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
JNPT r Volume 43, January 2019 Determinants of Dual-Task Training Effect Size in Parkinson Disease

Table 3. Total Group Results From the Multiple Regression Analysis With Stepwise Forward Regression
Significance of
β [95%CI] Standardized β t Value P Value VIF R2 Model
Change in DT gait velocity during
digit span task (n =103)
Constant 14.08 [0.21 to 27.94] 2.02 0.05 0.23 <0.001
DT gait velocity during digit span −0.297 [−0.414 to −0.180] − 0.45 − 5.04 <0.001 1.07
task
ScopaCog 0.837 [0.404 to 1.270] 0.34 3.83 <0.001 1.07
Change in DT gait velocity during
Stroop task (n = 112)
Constant 32.01 [15.84 to 48.17] 3.93 <0.001 0.26 <0.001
DT gait velocity during Stroop task −0.33 [−0.44 to −0.22] − 0.52 − 5.83 <0.001 1.18
ScopaCog 0.68 [0.26 to 1.09] 0.29 3.22 0.002 1.19
UPDRS-III −0.26 [−0.44 to −0.08] − 0.25 − 2.85 0.005 1.15
Change in DT gait velocity during
mobile phone task (n = 114)
Constant 3.97 [−8.08 to 16.01] 0.65 0.52 0.18 <0.001
DT gait velocity during mobile −0.25 [−0.37 to −0.13] − 0.40 − 4.28 <0.001 1.22
phone task
ScopaCog 0.74 [0.28 to 1.19] 0.30 3.20 0.002 1.25
LED 0.006 [0.001 to 0.011] 0.19 2.24 0.03 1.03
Abbreviations: CI, confidence interval; DT, dual task; LED, levodopa-equivalent dose; MDS-UPDRS-III, Unified Parkinson’s Disease Rating Scale—motor part; ScopaCog =
Scales for Outcomes in Parkinson’s Disease-Cognition; VIF, variance inflation factor.

(β = 0.45; P < 0.001) instead of performance on the Sco-
paCog determined better gains in DT gait velocity during the
digit span task.
For the CTT group, change in DT gait velocity during
the untrained Stroop task (Table 4) was solely related to a
lower DT gait velocity during Stroop at baseline (β = −0.27;
P = 0.04) (R2 = 0.05; P = 0.04). In contrast, in the IDT
group change in DT gait velocity was related to lower DT gait
velocity during the Stroop at baseline (β = −0.62; P < 0.001)
as well as to higher performance on the ScopaCog (β = 0.35;
P = 0.004) and lower MDS-UPDRS-III score (β = −0.26;
P = 0.03) (R2 = 0.39; P < 0.001).
In the CTT group, only a higher LED (β = 0.28; P =
0.03) determined change in DT gait velocity during the mo-
bile phone task significantly (R2 = 0.06; P = 0.03). In the IDT
group on the other hand, lower DT gait velocity during the mo-
bile phone task at baseline (β = −0.37; P = 0.006) and higher
MoCA score (β = 0.34; P = 0.01) were both significant deter-
minants (R2 = 0.17; P = 0.003) (Table 4). Overall, cognitive
factors played an important part in predicting improvements
in the IDT group. It is of note that a higher variance was ex-
plained by the prediction models in the IDT group (R2 ranged
between 0.17 and 0.39 compared with between 0.05 and 0.14
in the CCT group).
DISCUSSION
The main result of the current study was that both motor
and cognitive baseline factors contributed to the prediction of
DT training gains. More specifically, people with a lower DT
gait velocity at baseline and better cognitive function were
more likely to experience a greater benefit. In contrast to our
hypothesis, this finding was irrespective of the type of training
(ie, IDT vs CTT) and type of dual-task outcome.
The predictive effect of the motor and cognitive base-
line factors is in line with our hypotheses and previous stud-
ies in older individuals and persons with stroke.4,6 As well,
these results are supported by the relationship between cog-
nition and DT gait velocity and between dual-task deficits
and reduced movement automaticity in older individuals and
in persons with PD.3,34-36 As a result, participants with PD
showed a larger increase in frontostriatal networks than healthy
controls.37 Participants with PD were previously shown not to
be able to activate cerebellar areas involved in dual tasking to
the same extent as their healthy counterparts.38 Participants
with limited cognitive reserve, in particular, may no longer
have the cognitive resources to optimally gain from DT train-
ing. As well, cognitive dysfunction has been linked to learn-
ing deficits in people with PD,7-9 particularly in participants
with FOG.10-14 These individuals have specific alterations in
functional connectivity in dual-task-related regions involving
the precuneus and the striatum,39 and this may partly explain
why cueing is of value for reducing FOG.40 In the current
study, FOG severity correlated univariately with the outcome
variables, but in contrast to our hypotheses, it was not main-
tained as a primary factor determining the gains found after
DT training. Instead, cognitive function, irrespective of FOG,
was a consistent and significant determinant. The fact that the
cognitive load was different between the 3 tasks may explain
why different factors did not contribute equally to changes in
each of the tasks.
Beyond cognition, DT training benefits were also deter-
mined by motor factors, including DT gait velocity at baseline
and MDS-UPDRS-III scores. Contrary to our hypothesis,
people with PD with a low initial gait velocity showed the
largest improvements after training. This can be explained by
their larger potential to improve compared with people who
started at higher levels and by the fact that the training level
was adapted specifically to the participants’ gait velocity,
which was the primary outcome measure of the DUALITY
trial. It is also in line with previous work by Strouwen et al3


C 2019 Academy of Neurologic Physical Therapy, APTA 7

Copyright © 2019 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
Strouwen et al
Figure. Differences between observed and predicted values
based on the multiple linear regression analysis for 5
dependent variables (total group): digit span task (top),
Stroop task (center), and mobile phone task (bottom).
showing that single-task gait velocity had the highest associ-
ation with dual-task gait velocity, suggesting that the loss of
motor function is the most important explanatory factor of
dual-task impairment in PD. In contrast, the higher MDS-
UPDRS-III score correlated negatively with the DT training
benefits, indicating that people with a greater disease severity
8 
Copyright © 2019 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
JNPT r Volume 43, January 2019
were less likely to improve, particularly when performing
the auditory Stroop task. This is in line with previous trials
investigating the effect of exercise on other outcomes, such
as balance and upper limb movements, which indicate that
learning effects may be hampered in subgroups with greater
disease severity.41,42 One potential explanation for this finding
could be that individuals with greater disease severity have
a higher fall risk, and may as such deliberately walk slower
and/or focus less on the cognitive task as they may prioritize
safety over improvement under dual-task conditions. Future
interventions, including subgroup analyses comparing the
effects of interventions on participants with lesser versus
greater disease severity, would be extremely valuable in the
field of PD rehabilitation. Other factors, such as disease stage,
age, and prevalence of falling, did not contribute to the effect
size. As for the mobile phone task, LED was an additionally
important factor in determining training benefit, as higher
medication intake was associated with larger improvements in
the CTT group. Both the Stroop and mobile phone tasks were
untrained tasks and thus training benefits could be considered
as transfer effects. As generalization of learning is specifically
impaired in some subgroups of PD,14 this may explain why
a higher MDS-UPDRS-III score was found to be related
with less DT gains in the untrained Stroop task. Dopamine
replacement therapy has been previously demonstrated
to reinforce learning in persons with PD,43,44 which was
confirmed by the current results on the mobile phone task.
The second aim of the current study was to investigate
whether the factors determining DT training gains depended
on the type of practice. We hypothesized that IDT would de-
mand higher attentional load than CTT.45 Therefore, it was
expected that participants with cognitive problems, falling,
more advanced disease stage, and FOG would benefit less
from integrated compared with consecutive dual-task train-
ing. This hypothesis was only partially confirmed. We indeed
found that cognitive outcomes were related to training im-
provements more frequently after CTT than IDT Participants
with a higher cognitive status, measured by either MoCA or
ScopaCog, showed larger effects after IDT in both trained and
untrained tasks compared with participants with lower cogni-
tive scores. In contrast, falling, disease stage, and FOG were
not significant in determining benefits of IDT or CTT.
A limitation of the current study is that the explained
variance was low, and predicted outcomes corresponded only
moderately to the observed outcomes. Importantly, the predic-
tion of the explained variance was higher in the IDT group
compared with the CTT group. This finding indicates that DT
training gains are also explained by other factors that were
not included in this study, particularly in the CTT group. In
the DUALITY trial by Strouwen et al,1 it was found that the
total number of falls was higher in the CTT than in the IDT
group. As such, it is likely that fear of falling could have been
a contributing factor explaining the differences in explained
variance between groups. Furthermore, a previous study found
that depression was important in predicting DT gait interfer-
ence in PD.2 In addition, other factors such as motivation,
activity level, training frequency, and training intensity should
be taken into account in future studies predicting DT training
effects.
C 2019 Academy of Neurologic Physical Therapy, APTA
JNPT r Volume 43, January 2019 Determinants of Dual-Task Training Effect Size in Parkinson Disease

Table 4. Consecutive Task Training Versus Integrated Task Training Group Results From the Multiple Regression
Analysis With Stepwise Forward Regression
Consecutive Task Training Group Integrated Task Training Group
Significance Significance of
Standardized β P Value R2 of Model Standardized β P Value R2 Model
Change in DT gait velocity during
digit span task (n = 55/48)
DT gait velocity during digit span − 0.368 0.007 0.14 0.007 − 0.509 <0.001 0.38 <0.001
task
ScopaCog 0.311 0.02 – –
MoCA – – 0.450 <0.001
Change in DT gait velocity during
Stroop task (n = 61/53)
DT gait velocity during Stroop task − 0.265 0.04 0.05 0.039 − 0.623 <0.001 0.39 <0.001
ScopaCog – – 0.347 0.004
MDS-UPDRS-III – – − 0.263 0.03
Change in DT gait velocity during
mobile phone task (n = 61/53)
LED 0.281 0.03 0.06 0.028 – – 0.17 0.003
DT gait velocity during mobile – – − 0.367 0.01
phone task
MoCA – – 0.336 0.01
Abbreviations: DT, dual task; LED, levodopa-equivalent dose; MDS-UPDRS-III, Unified Parkinson’s Disease Rating Scale—motor part; MoCA, Montreal Cognitive Assessment;
ScopaCog, Scales for Outcomes in Parkinson’s Disease-Cognition

Despite these limitations, the current study offers valu- REFERENCES

able information regarding the implementation of dual-task
training in the rehabilitation of individuals with PD. Lower
dual-task gait velocity at baseline and higher cognitive reserve
were found to be the most important predictors of the train-
ing effect for both IDT and CTT. As factors such as FOG,
falling, and disease duration did not contribute significantly to
improvements in gait velocity after training, dual-task training
should be offered to individuals with various clinical profiles,
including those with greatest disease severity. Disease severity
measured by the MDS-UPDRS-III did not predict improve-
ments on the trained task after either CTT or IDT and on both
trained and untrained tasks after CTT. It had a significant effect
only on the untrained Stroop task after IDT. Transfer to un-
trained tasks was compromised after IDT in patients with lower
cognitive functioning, reflecting the inherent greater cognitive
challenge related to IDT in comparison to CTT.
CONCLUSIONS
DT gait velocity at baseline and cognitive function, as
assessed by the MoCA and the ScopaCog scale for global
cognitive function, are the most important determinants of the
effect of DT training on DT gait velocity. Whereas individuals
with PD seem to benefit from consecutive dual-task train-
ing regardless of disease severity, their cognitive capacity and
UPDRS-III scores are important factors to determine whether
they would be eligible for integrated DT training.
ACKNOWLEDGMENTS
We would like to thank all patients who participated in
the study.
1. Strouwen C, Molenaar EALM, Münks L, et al. Training dual tasks together
or apart in Parkinson’s disease: results from the DUALITY trial. Mov
Disord. 2017;32(8):1201-1210.
2. Rochester L, Nieuwboer A, Baker K, et al. Walking speed during single
and dual tasks in Parkinson’s disease: which characteristics are important?
Mov Disord. 2008;23(16):2312-2318.
3. Strouwen C, Molenaar EA, Keus SH, et al. Are factors related to dual-task
performance in people with Parkinson’s disease dependent on the type of
dual task? Parkinsonism Relat Dis. 2016;23:23-30.
4. Goldie PA, Matyas TA, Kinsella GJ, Galea MP, Evans OM, Bach TM. Pre-
diction of gait velocity in ambulatory stroke patients during rehabilitation.
Arch Phys Med Rehabil. 1999;80(4):415-420.
5. Strobach T, Gerstorf D, Maquestiaux F, Schubert T. Does initial perfor-
mance variability predict dual-task optimization with practice in younger
and older adults? Exp Aging Res. 2015;41(1):57-88.
6. Gothe NP, Fanning J, Awick E, et al. Executive function processes predict
mobility outcomes in older adults. J Am Geriatr Soc. 2014;62(2):285-290.
7. Deroost N, Kerckhofs E, Coene M, Wijnants G, Soetens E. Learning
sequence movements in a homogenous sample of patients with Parkinson’s
disease. Neuropsychologia. 2006;44(10):1653-1662.
8. Vandenbossche J, Deroost N, Soetens E, Kerckhofs E. Does implicit learn-
ing in non-demented Parkinson’s disease depend on the level of cognitive
functioning? Brain Cogn. 2009;69(1):194-199.
9. Price A, Shin JC. The impact of Parkinson’s disease on sequence learn-
ing: perceptual pattern learning and executive function. Brain Cogn.
2009;69(2):252-261.
10. Amboni M, Cozzolino A, Longo K, Picillo M, Barone P. Freezing of gait
and executive functions in patients with Parkinson’s disease. Mov Disord.
2008;23(3):395-400.
11. Vandenbossche J, Deroost N, Soetens E, et al. Impaired implicit sequence
learning in Parkinson’s disease patients with freezing of gait. Neuropsy-
chology. 2013;27(1):28-36.
12. Heremans E, Nieuwboer A, Spildooren J, et al. Cognitive aspects of freez-
ing of gait in Parkinson’s disease: a challenge for rehabilitation. J Neural
Transm (Vienna). 2013;120(4):543-557.
13. Vandenbossche J, Deroost N, Soetens E, et al. Freezing of gait in
Parkinson’s disease: disturbances in automaticity and control. Front Hum
Neurosci. 2012;6:356.
14. Heremans E, Nackaerts E, Vervoort G, Broeder S, Swinnen SP, Nieuw-
boer A. Impaired retention of motor learning of writing skills in


C 2019 Academy of Neurologic Physical Therapy, APTA 9

Copyright © 2019 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
Strouwen et al
patients with Parkinson’s disease with freezing of gait. PLoS One.
2016;11(2):e0148933.
15. Spildooren J, Vercruysse S, Desloovere K, Vandenberghe W, Kerckhofs
E, Nieuwboer A. Freezing of gait in Parkinson’s disease: the impact of
dual-tasking and turning. Mov Disord. 2010;25(15):2563-2570.
16. Rascol O, Perez-Lloret S, Damier P, et al. Falls in ambulatory non-
demented patients with Parkinson’s disease. J Neural Transm (Vienna).
2015;122(10):1447-1455.
17. Bekkers EMJ, Dijkstra BW, Dockx K, et al. Clinical balance scales indi-
cate worse postural control in people with Parkinson’s disease who exhibit
freezing of gait compared to those who do not: a meta-analysis. Gait
Posture. 2017;56:134-140.
18. Heinzel S, Maechtel M, Hasmann SE, et al. Motor dual-tasking deficits
predict falls in Parkinson’s disease: a prospective study. Parkinsonism
Relat Disord. 2016;26:73-77.
19. Strouwen C, Molenaar EA, Keus SH, et al. Protocol for a randomized com-
parison of integrated versus consecutive dual task practice in Parkinson’s
disease: the DUALITY trial. BMC Neurol. 2014;14(1):61.
20. Hughes AJ, Daniel SE, Kilford L, Lees AJ. Accuracy of clinical diagno-
sis of idiopathic Parkinson’s disease: a clinico-pathological study of 100
cases. J Neurol Neurosurg Psychiatry. 1992;55(3):181-184.
21. Hoehn MM, Yahr MD. Parkinsonism: onset, progression and mortality.
Neurology. 1967;17(5):427-442.
22. Tombaugh TN, McIntyre NJ. The Mini-Mental State Examination: a com-
prehensive review. J Am Geriatr Soc. 1992;40(9):922-935.
23. Menz HB, Latt MD, Tiedemann A, Mun San Kwan M, Lord SR. Re-
liability of the GAITRite R walkway system for the quantification of
temporo-spatial parameters of gait in young and older people. Gait
Posture. 2004;20(1):20-25.
24. Deuschl G, Schade-Brittinger C, Krack P, et al. A randomized trial of deep-
brain stimulation for Parkinson’s disease. N Engl J Med. 2006;355(9):896-
908.
25. Dalrymple-Alford JC, MacAskill MR, Nakas CT, et al. The MoCA: well-
suited screen for cognitive impairment in Parkinson disease. Neurology.
2010;75(19):1717-1725.
26. Cohen OS, Vakil E, Tanne D, Molshatzki N, Nitsan Z, Hassin-Baer S.
The Frontal Assessment Battery as a tool for evaluation of frontal lobe
dysfunction in patients with Parkinson disease. J Getr Psychiatry Neurol.
2012;25(2):71-77.
27. Kulisevsky J, Pagonabarraga J. Cognitive impairment in Parkinson’s dis-
ease: tools for diagnosis and assessment. Mov Disord. 2009;24(8):1103-
1110.
28. Hyde T, Fritsch T. Assessing executive function in Parkinson disease: the
alternating names test. Part I. Reliability, validity, and normative data.
Parkinsonism Relat Disord. 2011;17(2):100-105.
29. Siderowf A, McDermott M, Kieburtz K, Blindauer K, Plumb S, Shoulson
I. Test-retest reliability of the Unified Parkinson’s Disease Rating Scale in
10
Copyright © 2019 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
JNPT r Volume 43, January 2019
patients with early Parkinson’s disease: results from a multicenter clinical
trial. Mov Disord. 2002;17(4):758-763.
30. Nieuwboer A, Rochester L, Herman T, et al. Reliability of the new freezing
of gait questionnaire: agreement between patients with Parkinson’s disease
and their carers. Gait Posture. 2009;30(4):459-463.
31. Dal Bello-Haas V, Klassen L, Sheppard MS, Metcalfe A. Psychometric
properties of activity, self-efficacy, and quality-of-life measures in indi-
viduals with Parkinson disease. Physiother Can. 2011;63(1):47-57.
32. Field A. Discovering Statistics Using IBM SPSS Statistics. 4th ed. London,
England: Sage Publications; 2013.
33. Portney L, Watkins M. Foundations of Clinical Research: Application to
Practice. 3rd ed. Upper Saddle River, NJ: Pearson-Prentice Hall; 2009.
34. Hausdorff JM, Schweiger A, Herman T, Yogev-Seligmann G, Giladi N.
Dual-task decrements in gait: contributing factors among healthy older
adults. J Gerontol A Biol Sci Med Sci. 2008;63(12):1335-1343.
35. Holtzer R, Verghese J, Xue X, Lipton RB. Cognitive processes related to
gait velocity: results from the Einstein Aging Study. Neuropsychology.
2006;20(2):215-223.
36. Wu T, Hallett M. A functional MRI study of automatic movements in
patients with Parkinson’s disease. Brain. 2005;128(10):2250-2259.
37. Gratwicke J, Jahanshahi M, Foltynie T. Parkinson’s disease dementia: a
neural networks perspective. Brain. 2015;138(6):1454-1476.
38. Gao L, Zhang J, Hou Y, Hallett M, Chan P, Wu T. The cerebellum in
dual-task performance in Parkinson’s disease. Sci Rep. 2017;7:45662.
39. Vervoort G, Heremans E, Bengevoord A, et al. Dual-task-related con-
nectivity changes in patients with Parkinson’s disease. Neuroscience.
2016;317:36-46.
40. Spildooren J, Vercruysse S, Heremans E, Galna B, Verheyden G, Vervoort
G, Nieuwboer A. Influence of cueing and an attentional strategy on freez-
ing of gait in Parkinson disease during turning. J Neurol Phys Ther.
2017;41(2):129-135
41. Stephan MA, Meier B, Zaugg SW, Kaelin-Lang A. Motor sequence learn-
ing performance in Parkinson’s disease patients depends on the stage of
disease. Brain Cogn. 2011;75(2):135-140.
42. Canning CG, Sherrington C, Lord SR, et al. Exercise for falls pre-
vention in Parkinson disease: a randomized controlled trial. Neurology.
2015;84(3):304-312.
43. McNeely ME, Duncan RP, Earhart GM. Medication improves bal-
ance and complex gait performance in Parkinson disease. Gait Posture.
2012;36(1):144-148.
44. Nackaerts E, Vervoort G, Heremans E, Smits-Engelsman BC, Swinnen
SP, Nieuwboer A. Relearning of writing skills in Parkinson’s disease: a
literature review on influential factors and optimal strategies. Neurosci
Biobehav Rev. 2013;37(3):349-357.
45. Strouwen C, Molenaar EA, Münks L, et al. Dual tasking in Parkin-
son’s disease: should we train hazardous behavior? Expert Rev Neurother.
2015;15(9):1031-1039.

C 2019 Academy of Neurologic Physical Therapy, APTA
 
JNPT r Volume 43, January 2019

APPENDIX. Pearson Correlation Coefficients Resulting From the Univariate Analysis Significant at Level P < 0.20—Consecutive Versus Integrated Task
Training Group
Descriptive Cognitive Motor
Disease ST Reaction DT Reaction MDS-UPDRS ST Gait DT Gait
Age Duration Falls LED DBS MMSE MoCA FAB ScopaCog ANT AIT Time Time H&Y III NFOG-Q ABC Velocity Velocity

C 2019 Academy of Neurologic Physical Therapy, APTA

Consecutive task training group
Change in DT gait velocity during – – – r = 0.19a – – r = 0.19 – r = 0.22 – – r = 0.27 r = 0.25 – – – – – r = −0.29
digit span task (n = 55) P = 0.17 P = 0.17 P = 0.11 P = 0.05 P = 0.07 P = 0.03
Change in DT gait velocity during – – – r = 0.224 – – r = 0.19 – – – – – – – – – – – r = −0.27
Stroop task (n = 61) P = 0.08 P = 0.14 P = 0.04
Change in DT gait velocity during – – – r = 0.281 – – – – – – – – – – – – – – r = −0.24
mobile phone task (n = 63) P = 0.03 P = 0.06
Integrated task training group
Change in DT gait velocity during – – – – – r = 0.26 r = 0.25 – r = 0.21 – – r = 0.31 – – – r =−0.19 – r =−0.34 r = −0.46
digit span task (n = 48) P = 0.06 P = 0.08 P = 0.14 P = 0.03 P = 0.19 P = 0.02 P = 0.001
Change in DT gait velocity during – – – – – r = 0.28 r = 0.23 r = 0.19 r = 0.25 – – – – – r =−0.19 r =−0.20 – r =−0.36 r = −0.46
Stroop task (n = 53) P = 0.04 P = 0.10 P = 0.18 P = 0.07 P = 0.18 P = 0.16 P = 0.01 P = 0.001
Change in DT gait velocity during – – – – – – r = 0.27 – r = 0.22 – – – – – – – – – r = −0.31
mobile phone task (n = 53) P = 0.05 P = 0.11 P = 0.03
Abbreviations: ABC, Activities-specific Balance Confidence scale; AIT, Alternating Intakes Test; ANT, Alternating Names Test; DBS, deep brain stimulator; DT, dual task; FAB, Frontal Assessment Battery; H&Y, Hoehn and
Yahr stage; LED, levodopa-equivalent dose; MDS-UPDRS-III, Unified Parkinson’s Disease Rating Scale; MMSE, Mini-Mental State Examination; r, regression coefficient; MoCA, Montreal Cognitive Assessment; NFOG-Q, New
Freezing of Gait Questionnaire; ScopaCog, Scales for Outcomes in Parkinson’s Disease-Cognition; ST, single task.
a
Factors which were correlated in one group but not in the other are underlined.

Copyright © 2019 Academy of Neurologic Physical Therapy, APTA. Unauthorized reproduction of this article is prohibited.
11
Determinants of Dual-Task Training Effect Size in Parkinson Disease