Abstract
Background Dual-task (DT) training is gaining ground as a physical therapy intervention in people with Parkinson disease (PD). Future studies evaluating the effect of such interventions need reliable outcome measures. To date, the test-retest reliability of DT measures in patients with PD remains largely unknown.
Objective The purpose of this study was to assess the reliability of DT outcome measures in patients with PD.
Design A repeated-measures design was used.
Methods Patients with PD (“on” medication, Mini-Mental State Examination score ≥24) performed 2 cognitive tasks (ie, backward digit span task and auditory Stroop task) and 1 functional task (ie, mobile phone task) in combination with walking. Tasks were assessed at 2 time points (same hour) with an interval of 6 weeks. Test-retest reliability was assessed for gait while performing each secondary task (DT gait) for both cognitive tasks while walking (DT cognitive) and for the functional task while walking (DT functional).
Results Sixty-two patients with PD (age=39–89 years, Hoehn and Yahr stages II–III) were included in the study. Intraclass correlation coefficients (ICCs) showed excellent reliability for DT gait measures, ranging between .86 and .95 when combined with the digit span task, between .86 and .95 when combined with the auditory Stroop task, and between .72 and .90 when combined with the mobile phone task. The standard error of measurements for DT gait speed varied between 0.06 and 0.08 m/s, leading to minimal detectable changes between 0.16 and 0.22 m/s. With regard to DT cognitive measures, reaction times showed good-to-excellent reliability (digit span task: ICC=.75; auditory Stroop task: ICC=.82).
Limitations The results cannot be generalized to patients with advanced disease or to other DT measures.
Conclusions In people with PD, DT measures proved to be reliable for use in clinical studies and look promising for use in clinical practice to assess improvements after DT training. Large effects, however, are needed to obtain meaningful effect sizes.
Dual tasking, defined as the simultaneous performance of 2 tasks with distinct goals,1 has been found to lead to decreased motor and cognitive performance in patients with Parkinson disease (PD).2–5 Both gait parameters and cognitive task performance deteriorate when walking while dual tasking in people with PD.2,3,6 Gait characteristics that represent postural control are especially vulnerable to dual-task (DT) interference.7 However, in people with early PD, it was shown that DT interference was very similar to that of controls,7 illustrating the complexity of DT behavior. Not only disease severity1,5 but also age-related changes in brain mass affect the degree of attention allocated to combine 2 or more tasks.8 Experience with and the difficulty level of the tasks, in combination with the environment in which the tasks are performed, further contribute to variable DT responses.5,9 In people with PD, the extent of the loss of movement automaticity5,10 and the effectiveness of medication11 play additional roles. Finally, the priority assigned to each of the tasks is of crucial importance.12 In general, people with PD use a “posture second” strategy in which they deprioritize gait or balance performance.9,13 However, they are able to change their focus when instructed to do so.14 Given the multitude of factors that influence DT performance and the consequential large variation in DT responses between and within people, it is plausible to assume that reliability of DT measures may be compromised and needs to be examined thoroughly before using them in clinical studies.
In PD, the reliability of DT gait outcomes is currently unknown, but that of single-task (ST) walking speed was demonstrated for comfortable and fast 10-m walks (intraclass correlation coefficient [ICC]=.98 and .99, respectively)15,16 within the same day, across multiple days, and in the home environment.17 In healthy elderly people, good reliability was found for DT walking in combination with a forward digit span task (ICC=.83),18 and almost perfect reliability was found for DT walking in combination with counting backward.19 Also, in people with frontotemporal degeneration,19 those prone to falling,20 and those with mild cognitive impairment,21 stroke,22 and multiple sclerosis,23 DT walking tests proved to be reliable.
Dual tasking will not only interfere with gait but also interfere with cognitive performance. In a study of healthy elderly people,18 reliable results were found on the DT digit span with a test interval of a maximum of 1 week. Other studies in healthy older adults20,24 demonstrated that speed and accuracy for various DT cognitive tests showed poor-to-good reliability, with ICC values ranging from −.16 to .83. The reliability of DT cognitive tests in PD is currently not known.
Recently, some preliminary training studies assessed the potential benefit of DT training in PD.25–27 In addition, a systematic review of DT postural training in older adults suggested positive effects, although firm conclusions were hampered by the heterogeneity of the outcome measures.28 Reliable outcome measures for DT interventions, therefore, are needed for people with PD and for older adults to address this lacuna. Training studies often include interventions that last several weeks. It is plausible to assume that the factors that influence DT performance will become less stable and less well controlled with increasing time intervals. Hence, the reliability over longer time spans needs to be established to be able to interpret the actual significance of training effects. Given this overview, the aim of the present study was to assess the reliability of DT gait and cognitive measures in people with PD and calculate minimal detectable change (MDC) scores over a period of 6 weeks, a common duration of training in PD studies.
Gait requires dynamic balance and coordinated movements, and its multidimensional nature currently can be captured in different outcomes.29 As various gait outcomes previously showed different DT interference,7 we addressed DT reliability separately for 5 different gait domains: pace, rhythm, postural control, asymmetry, and variability.30 These domains also may have a different contribution to various clinical phenomena. For instance, variability outcomes were found to be more sensitive than other domains for predicting falling.31 However, unlike spatiotemporal measures,32 the reliability of variability measures ranged from poor to excellent (ICC=.04–.86) and improved only by increasing the number of steps analyzed.15,18,19
In this study, cognition was assessed with 3 different tasks: an auditory Stroop task, a backward digit span task, and a mobile phone task. The auditory Stroop task addresses set shifting ability and incongruent response inhibition.33,34 The backward digit span task tests the working memory component of executive function.34 Given the importance of inducing transfer of training to daily life dual tasking, we also studied the reliability of a functional DT. For this purpose, we devised a mobile phone task consisting of fine motor, memory, and executive components.25 We hypothesized that the reliability of DT measures would be more compromised than that of ST measures, as dual tasking introduces both motor and cognitive sources of variability. We also expected that interference measures would be less reliable than absolute DT measures and that the variability domain of gait would be the least stable over time. Furthermore, we predicted that reliability would be worse in the mobile phone task, given the complexity of this task.
Method
Participants
Participants were recruited as part of the DUALITY study.25 The Belgian subsample of this study, consisting of 62 people with PD, was used to assess test-retest reliability. Inclusion criteria for recruitment were: (1) diagnosis of PD according to the UK Brain Bank criteria,35 (2) Hoehn and Yahr (H&Y) stage ≤3 in the subjective best “on” phase of the medication cycle,36 (3) able to walk for 10 minutes continuously, (4) presence of DT interference as established by a structured checklist,25 (5) score of ≥24 on the Mini-Mental State Examination (MMSE),37 (6) stable medication over the previous 3 months, (7) no hearing or visual problems that interfere with testing or training, and (8) stable deep brain stimulator (DBS) settings over the previous year.
Testing
The first 2 baseline assessments, as described in the protocol of the DUALITY trial,25 were used to assess test-retest reliability. A clinical test battery was performed, assessing several descriptive, disease, and cognitive characteristics of the participants.25 Assessments were performed in the “on” phase at a standardized moment after medication intake. In both test sessions, the same person (C.S.), who has a physical therapy background, conducted the tests. The DT assessments consisted of 3 different tests: (1) walking while performing the backward digit span task, (2) walking while performing the auditory Stroop task, and (3) walking while performing the mobile phone task. These tasks also were assessed in a sitting position. In addition, walking was assessed without these tasks (ST gait).
During the auditory Stroop task, patients were asked to verbally respond to the words “high” and “low” pronounced with congruent and incongruent high and low tones delivered via headphones. Stimuli were presented with a variable interval (1.5–2 seconds) to control for cueing effects. During the digit span task, participants were asked to repeat an array of numbers in reverse order, the length of which was adapted to the level of the individual determined in a sitting position.7 The optimal sequence length for each individual was determined with the subscale item “digit span backward” as part of the Scales for Outcomes in Parkinson's Disease–Cognition (SCOPA-Cog),38 of which the arrays were not included in the actual ST and DT assessments. Cognitive task performance during the digit span and auditory Stroop tasks was recorded via a wireless headset system (t-bone DS16T transmitter and t-bone IEM100R receiver, Beyerdynamic Inc, Farmingdale, New York). The mobile phone task made use of a large-button mobile phone (Emporia Talk Premium, Linz, Austria). For this task, participants had to type the test date (8 numbers) using the phone's keyboard. Participants were instructed to hold the phone in one hand and to type with the other hand. They were free to choose which hand they used for typing. For the digit span and auditory Stroop tasks, there were 3 different versions that were repeated twice in the ST and DT conditions. For the digit span task, in each version, different numbers were presented to the participant. For the auditory Stroop task, in each version, different stimuli were presented. The mobile phone task consisted of one version that was repeated twice in both conditions.
Gait data were collected with the GAITRite Walkway System (GAITRite Platinum software version 4.0, CIR Systems Inc, Peekskill, New York), which has excellent reliability.32,39 Data were sampled at a frequency of 120 Hz over an active length of 792 cm. Participants walked at their preferred walking speed and started and ended walking 1 m before and after the mat to exclude acceleration and deceleration. The ST and DT conditions were randomized, but the order of testing was kept the same for each individual across test sessions. Participants were asked to concentrate on both tasks during dual tasking and were instructed as follows: “Walk over the GAITRite mat in the same way as you would normally walk while performing a concurrent task.” Instructions were given by the same tester before the start of the assessment and repeated with each new task. Specific instructions for the backward digit span task were: “You will hear an array of numbers through the headphones. After this array has finished, please repeat the numbers in a reverse order, starting with the last number you have heard.” Specific instructions for the auditory Stroop task were: “Through the headphones, you will hear the words ‘high’ and ‘low,’ pronounced in a high or low pitch. Please indicate whether the pitch that you've heard was high or low, regardless of the word you have heard.” For the mobile phone task, the following instructions were provided to the patient: “We will now give you this mobile phone. Please type the date of today into the device.” Before actual assessment, the tester checked whether participants had understood these instructions. In the rare case of freezing or voluntary stops, the trial was stopped and repeated to obtain valid data.
Descriptors
We collected information on age, height, weight, disease duration, H&Y stage,36 medication intake, physical therapy, freezing of gait, falls, MMSE,37 Montreal Cognitive Assessment (MoCA),40 Frontal Assessment Battery (FAB),41 and Movement Disorder Society–Unified Parkinson's Disease Rating Scale (MDS-UPDRS) total score.42 Tests that were performed in both sessions to document disease and cognitive status over time were: Unified Parkinson's Disease Rating Scale–part III (UPDRS-III)42; Freezing of Gait Questionnaire43; Activities-specific Balance Confidence scale44; and, as a general cognitive scale, the SCOPA-Cog.38 We also included 2 executive function tests, the Alternating Names Test and the Alternating Intakes Test,45 which required the verbal listing of alternating boys' and girls' names and food and drink products, respectively.
ST and DT Gait Outcomes
We analyzed reliability in the 5 domains of gait30 as follows: (1) gait speed and stride length (pace); (2) cadence, stride time, and swing percentage (rhythm); (3) stride width (postural control); (4) step length asymmetry (asymmetry); and (5) within-patient standard deviations of stride length and stride time (variability). We also calculated the proportional DT interference using the following formula: (ST performance − DT performance)/ST performance. Dual-task interference was determined for gait speed, cadence, stride length, stride time, and stride width.
ST and DT Cognitive Outcomes
To assess ST and DT cognitive outcomes, we calculated reaction times (for the digit span and auditory Stroop tasks), response times (for the digit span task), and errors (for the digit span, auditory Stroop, and mobile phone tasks). Verbal responses were recorded and analyzed with the Audacity 1.3 Beta program (Free Software Foundation, Boston, Massachusetts) and Matlab (R2011b) (The Mathworks Inc, Natick, Massachusetts). For the digit span task, reaction times were determined as the time between the end of the sound fragment and the beginning of the response of the patient. Response time was defined as the time it took to reproduce the span divided by the total duration of the span recorded on tape to account for differences in span levels among the participants. For the auditory Stroop task, reaction time was calculated as the time between the beginning of the stimulus and the beginning of the response of the patient. Interference was calculated for DT cognitive measures in the same manner as described for gait.
Data Analysis
To meet the minimum number of steps needed to calculate reliable DT gait outcomes with the GAITRite system,15,21,46 we averaged the results of all trials in each test session. Therefore, calculations were based on between 65 and 125 steps for each patient for the digit span and auditory Stroop tasks and between 21 and 77 steps for the mobile phone task. Data were analyzed using IBM SPSS Statistics for Windows version 22 (IBM Corp, Armonk, New York).47 Normality of the data was assessed with the Kolmogorov-Smirnov test, QQ-plots, and histograms. Mean and standard deviation are provided for normally distributed data. Median and interquartile range are given for non-normally distributed data. We used t tests and the calculation of Cohen d effect sizes to look for systematic differences between tests 1 and 2. To assess reliability, ICC values were calculated with a 2-way random-effects model with absolute agreement.48,49 We used ICC values for averaged measurements and applied Shrout and Fleiss' cutoffs of <.40 as poor, .40 to .75 as fair to good, and >.75 as excellent.48 For data that were normally distributed, we calculated the standard error of measurement (SEM=SDpooled × √[1−ICC]) and the MDC (MDC=SEM × 1.96 × √[2]).49 The SEM provides an absolute index for reliability, largely independent from the population from which it was determined.49 The MDC is used to determine whether a change in score can be considered without measurement error.49 As SEM and MDC could not be calculated for abnormally distributed data, we used Wilcoxon signed rank tests to assess systematic differences between test sessions and Spearman correlation coefficients to indicate reliability, using Landis and Koch's50 cutoffs of >.80 as almost perfect and between .61 and .80 as substantial. We used Bland and Altman plots51 to visualize the mean difference between the 2 tests for DT gait speed.
Role of the Funding Source
This work was funded by the Jacques and Gloria Gossweiler Foundation.
Results
Descriptive Data
Descriptive data of the participants are presented in Table 1. Of the 62 patients with PD included in this study, 71.0% were male and 29.0% female. The majority of participants were in H&Y stage 2 (61.0%) or 3 (36.0%). More than half (56.5%) of the participants indicated that they had fallen in the year previous to the first assessment, and 54.8% had freezing of gait. Although all participants had an MMSE score ≥24, 16 participants scored lower than 24 on the MoCA. Four participants had an FAB score lower than 13, the cutoff for frontal cognitive impairment.52 The average number of digits that participants could repeat backwards was 4.21 (SD=1.16).
Participants' Descriptive Data (N=62)a
Table 2 summarizes participants' evolution over the 6-week follow-up period for a number of disease characteristics, showing that they remained largely stable. During the second test, one participant changed from H&Y stage 3 to stage 2. All other participants were ranked in the same way. Medication dose changed for 8 participants in the period between test 1 and test 2, but average dose did not differ in the 6-week period. Two participants had a higher levodopa equivalent dose in test 2, 2 participants had a lower levodopa equivalent dose in test 2, and 4 participants had a change in DBS parameters. The t tests revealed significantly better scores in test 2 for UPDRS-III and SCOPA-Cog. Effect sizes for these differences ranged from small to medium (0.17, −0.34, and −0.19 milliseconds, respectively). All other measures were constant.
Test-Retest Scores for General Disease Characteristics During the Study Perioda
Reliability of ST and DT Gait Parameters
Bland and Altman plots are presented in the Figure. Points were equally distributed around zero, showing no bias in the results and a similar pattern for ST and DT gait speed. In all DT conditions, overall scores fell within the limits of agreement, around 0.20 m/s for the digit span and auditory Stroop tasks. The limits of agreement were larger for the mobile phone task (around 0.30 m/s), indicating more variability than in the digit span and auditory Stroop gait measures. The ST gait parameters showed excellent reliability, with ICC values ranging from .77 to .94 (eTab. 1).
Bland and Altman plots for single-task (ST) and dual-task (DT) gait speed: (A) ST gait speed, (B) DT digit span task gait speed, (C) DT auditory Stroop task gait speed, and (D) DT mobile phone task gait speed. The plots represent the difference between test 1 and test 2 (y-axis) plotted against the average of test 1 and test 2 (x-axis). The average difference between test 1 and test 2 is presented as a horizontal line on the plot (middle line), and the upper and lower lines represent the 95% upper and lower limits of these differences between the 2 tests.
The ICC values for DT digit span task gait ranged from .87 to .95, also indicating excellent reliability (Tab. 3). Spearman correlation coefficients for variability and asymmetry parameters, which did not allow calculation of ICCs, were moderate to substantial (.44–.74). When analyzing the results in more detail, Table 3 reveals that a significantly better performance in test 2 was found in the gait domains of rhythm (cadence and stride time), postural control (stride width), and variability (stride length and stride time). Effect sizes, however, were small (−0.19 to 0.28). In addition, these changes were not correlated with changes in MDS-UPDRS scores (cadence: r=−.16, P=.23) or SCOPA-Cog scores (cadence: r=−.11, P=.41). The ICC values for the digit span interference measures, calculated as a percentage of the ST measure, ranged from .47 to .72, indicating fair-to-good reliability (eTab. 2). Here, a systematically lower DT interference in test 2 was shown for speed, cadence, and stride time. The SEM for DT digit span task gait speed was 0.06 m/s, leading to an MDC score of 0.16 m/s.
Test-Retest Scores for Dual-Task Gait Parametersa
A similar pattern of results emerged from walking while performing the auditory Stroop task (Tab. 3). Absolute values for DT auditory Stroop task gait showed excellent reliability (.89–.94), whereas interference measures showed poor-to-good reliability (.19–.81). Parameters belonging to the variability and asymmetry domains showed moderate-to-substantial correlations (.52–.77). Interestingly, no systematic differences were found between test 1 and test 2 for the DT auditory Stroop task gait outcomes. The SEM for DT auditory Stroop task gait speed was 0.06 m/s, and the MDC score was 0.16 m/s.
Good-to-excellent reliability (ICC=.72–.89) was found for performing the mobile phone task during gait. For the auditory Stroop and digit span tasks, a moderate correlation was seen for the variability and asymmetry domains (rs=.48–.58). Reliability of the DT interference measures for the mobile phone task was good (ICCs=.64–.74). However, interference measures for speed, cadence, and stride time were systematically lower during test 2, albeit with small effect sizes (−0.29 to 0.27). The SEM for gait speed during the mobile phone task was 0.08 m/s, and the MDC score was 0.22 m/s.
Reliability of ST and DT Cognitive Parameters
Table 4 illustrates that DT cognitive task outcomes were less stable than DT gait outcomes. Reliability measures were fair to good for the DT backward digit span task (reaction time: ICC=.75; response time: ICC=.69), but not for interference measures (interference reaction time: ICC=.41; interference response time: rs=.09) (eTab. 3). Significantly fewer errors were made on the digit span task during test 2 compared with test 1. The SEM values were 266.31 milliseconds for DT digit span task reaction time and 184.12 milliseconds for DT digit span task response time, leading to an MDC score of 738.16 milliseconds and 510.35 milliseconds, respectively.
Test-Retest Scores for DT Cognitive Task Measuresa
For the auditory Stroop task, excellent values were found for absolute measures of reaction time (ICC=.82), but values were lower for interference measures (rs=.59). Errors during dual tasking did not appear to correlate well between the 2 test sessions (rs=.41), although no systematic difference was seen (median=0). The SEM for DT Stroop reaction time was 86.31 milliseconds, resulting in an MDC score of 239.24 milliseconds. Dual-task errors while performing the mobile phone task were not correlated between tests (rs=.21).
Discussion
In line with previous studies in the elderly population and in people with neurological disorders,18–23 we found that DT gait measures in people with PD showed excellent reliability over a period of 6 weeks using standard tests of executive function (auditory Stroop and digit span tasks) as DT measures. The functional task, the manipulation of a mobile phone during gait, showed slightly worse reliability, probably as a result of greater task complexity. Overall, DT gait reliability was comparable to that of ST gait. Furthermore, DT cognitive performance, expressed as absolute values of reaction times, showed good-to-excellent reliability. These results support the use of DT gait and DT cognitive outcomes to assess patients' evolution over time in longitudinal studies and in clinical practice.
Proportional interference values, expressed as a percentage of ST performance, had lower reliability than absolute values for both gait and cognitive performance. We expected interference measures to have more compromised reliability, as variability inherent to ST and DT performance may play a role. In addition, when assessing effects of DT training, there may be an effect on both ST and DT performance. Therefore, interference measures may be less suitable as primary outcomes in studies with multiple assessments in the same population. However, they can be relevant when comparing 2 different populations in a single test session or when comparing the DT cost of different secondary tasks.
Repeatability in different domains of gait30 was assessed to obtain a better understanding of DT effects in PD. Dual-task gait measures representing pace, rhythm, and postural control had excellent test-retest reliability over a period of 6 weeks. In contrast, variability and asymmetry showed moderate-to-substantial reliability. Variability reflects the stride-to-stride fluctuations within one individual. It was found to be higher in patients with PD and interpreted as reflecting decreased automaticity and higher input of cognition required for motor control.53 In this sample, variability and asymmetry were found to have a skewed distribution, showing greater variation in people with PD than in healthy elderly people. This finding is in line with earlier work,15,46 which showed that more than 30 to 100 steps are needed to assess variability during ST in a reliable way. The number of steps per task in this study varied between 65 and 125 for the digit span and auditory Stroop tasks and between 21 and 77 for the mobile phone task. Although these numbers are considered sufficient for ST gait,46 they might have been insufficient to assess DT gait. For other gait domains, however, the number of steps included in this study seemed adequate. Future study is needed to address the number of steps that are minimally needed to achieve stable DT variability measurements.
This is the first study, to our knowledge, to evaluate the reliability of DT cognitive measures in PD. We found that cognitive performance showed lower reliability than gait in ST and DT conditions (eTab. 3 and eTab. 4). Several factors, such as a good night's sleep prior to testing, attention, and emotional state, may influence cognitive performance to a greater extent than gait performance.54 The degree of cognitive challenge and the high level of executive dysfunction55 also might have affected test-retest stability. Sixteen participants (25.8%) had scores lower than 24 on the MoCA, and 4 (6.5%) had low FAB scores. Also, learning effects from one test to the other may be greater for cognitive tasks than for gait tasks, as gait is habitual behavior and cognitive tests require more goal-directed control.56 To counteract this possible learning effect, a washout period may be recommended when assessing cognitive tasks repeatedly. We found that the backward digit span task in particular was subject to learning effects, despite the fact that we individualized the length of the digit span.7 We found no correlation between DT digit span task improvements and better UPDRS-III and SCOPA-Cog scores at test 2 compared with test 1. Therefore, the learning effects are unlikely due to patients' generally better disease and cognitive state at test 2. In contrast to the Stroop task, which was performed without any errors in test 1, the backward digit span task did show baseline errors. This finding points to a possible ceiling effect for the auditory Stroop task, which might have masked learning effects. Finally, the nature of the digit span task might have stimulated patients to practice, whereas for the auditory Stroop task, this was not possible. Cognitive performance was measured by reaction times and error scores. As the error scores showed very low reliability, we recommend that error scores not be used as a DT cognitive outcome measure.
Reliability was assessed for 3 different DT measures, each representing different domains of executive function. This study demonstrated that the digit span and auditory Stroop tasks had comparable reliability levels, whereas the ICC values for the mobile phone task were somewhat lower. Both the digit span and auditory Stroop tasks are purely cognitive tasks relying on executive function, albeit in different domains. The mobile phone task, however, represents a functional task, relying on visual and motor control of the display functions and remembering the date. The greater complexity of this task and the fact that it induced more interference may explain the somewhat lower reliability. These findings are in contrast to those of Muhaidat et al,20 who compared 8 different tests but did not find lower reliability for the more complex tasks. However, as reliability of the mobile phone task could still be considered as very good, this task may be particularly suitable to assess DT problems in clinical practice. Using a mobile phone is not only relevant for daily life but also user-friendly, as most patients and clinicians have this tool at their disposal.
The MDC values found in this study, indicating clinically relevant DT changes, were comparable to previous findings on test-retest reliability of ST measures. Reference MDC values of ST gait speed16,57 were found to range between 0.09 and 0.18 m/s. Monticone et al23 reported MDC values for gait speed of 0.13 m/s for a motor-motor DT measure and of 0.19 m/s for a motor-cognitive DT measure in people with multiple sclerosis. In the current study, MDC values for DT gait speed were 0.16 m/s for the digit span task, 0.16 m/s for the auditory Stroop task, and 0.22 m/s for the mobile phone task. These findings imply that in order to achieve an effect of DT training, which is beyond measurement error, patients with PD should improve their gait speed with large effect sizes. Calculated as a percentage of the current baseline DT gait speed, this would amount to 17% for the digit span task, 17% for the auditory Stroop task, and 26% for the mobile phone task.
The current study assessed DT reliability in a large group of people with PD, giving sufficient power to the results. Limitations of the study are that MDC values could not be provided for all parameters assessed, due to skewed distributions. The results cannot be generalized to tasks other than those used in this study or to late-stage PD. We also recommend that future DT reliability studies address other possible sources of variability, such as age, cognitive impairment, and the presence of freezing of gait. Eight patients in this study had a change in prescribed medication dose. As this is a common problem in clinical studies, future sensitivity analyses comparing groups with and without changes in medication could be conducted to determine how this factor affects reliability of DT outcomes.
In conclusion, this study showed that DT gait can be measured reliably in people with PD when repeating assessments over a 6-week period. However, large effect sizes are needed to induce clinically relevant changes, extending beyond the MDC values. We found that absolute values of DT performance are more reliable than proportional values and that cognitive DT performance, expressed as reaction time but not as error rates, can be ascertained reliably. Although a higher variation was found in the mobile phone task, this task proved to be a reliable functional dual-task test, which is suitable for use in clinical practice.
Footnotes
Dr Keus and Professor Nieuwboer provided concept/idea/research design. Dr Strouwen and Professor Nieuwboer undertook the writing. Dr Strouwen, Mrs Molenaar, Dr Keus, Mrs Münks, and Professor Bloem were responsible for data collection. Dr Strouwen and Professor Nieuwboer performed data analysis. Professor Nieuwboer, Mrs Molenaar, Dr Keus, and Professor Bloem provided project management. Mrs Molenaar, Mrs Münks, Dr Keus, and Professor Bloem provided consultation (including review of manuscript before submission). The authors thank all patients who participated in the study.
Ethical approval for the study was obtained (CME KU Leuven–B322201213165/S53419).
This work was funded by the Jacques and Gloria Gossweiler Foundation.
The DUALITY trial was registered on ClinicalTrials.gov (NCT01375413).
- Received May 19, 2015.
- Accepted January 25, 2016.
- © 2016 American Physical Therapy Association
References
Issue highlights
