Effect of Musical Cues on Gait in Individuals with Parkinson Disease with Comorbid Dementia

Abstract

Background:

Individuals with Parkinson disease and comorbid dementia (PDD) demonstrate gait impairments, but little is known about how these individuals respond to interventions for gait dysfunction. Rhythmic auditory stimulation (RAS), which utilizes music or other auditory cues to alter gait, has been shown to be effective for improving gait in individuals with PD without dementia, but has not been explored in individuals with PDD.

Research question:

Can individuals with PDD modulate their gait in response to music and mental singing cues?

Methods:

This single center, cross-sectional, interventional study included 17 individuals with PDD. Participants received Music and Mental singing cues at tempos of 90, 100, 110, and 120% of their uncued walking cadence. Participants were instructed to walk to the beat of the song. Gait variables were collected using APDM Opal sensors. Data were analyzed using mixed effect models to explore the impact of tempo and cue type (Music vs Mental) on selected gait parameters of velocity, cadence, and stride length.

Results:

Mixed effects models showed a significant effect of tempo but not for cue type for velocity (F=11.51, p<.001), cadence (F=11.13, p <.001), and stride length (F=5.68, p=.002). When looking at the marginal means, velocity at a cue rate of 90% was significantly different from 100%, indicating participants walked slower with a cue rate of 90%. Participants did not significantly increase their velocity, cadence, or stride length with faster cue rates of 110% and 120%

Significance:

Individuals with PDD appear to be able to slow their velocity in response to slower cues, but do not appear to be able to increase their velocity, cadence, or stride length in response to faster cue tempos. This is different from what has been reported in individuals with PD without dementia. Further research is necessary to understand the underlying mechanism for these differences.

Introduction:

Parkinson disease (PD) is the second most common neurodegenerative disease, affecting nearly one million people in the United States.¹ While PD is defined by four cardinal motor symptoms (bradykinesia, rigidity, postural instability, and tremor²), individuals with PD can experience a variety of motor and non-motor symptoms, including gait dysfunction and cognitive impairment. Because symptoms vary from person to person, individualized treatment is important for optimizing function in individuals with PD.

Gait dysfunction is cited by individuals with PD as the most important symptom to address with treatment.³ Generally, gait dysfunction in PD can be characterized by short, shuffling steps and decreased speed.⁴ Many studies have been conducted to determine the best treatment for gait dysfunction in PD.^5,6 One area of focus for many years includes rhythmic auditory stimulation to improve gait.^7–9 Rhythmic auditory stimulation (RAS) can refer to various sounds, from metronome beats to actual music, and can result in improved gait speed, stride length, and cadence in individuals with PD.^7,10–13 While there are many forms of RAS, music is of particular interest as previous research suggests music cues increase gait speed more than metronome cues in healthy adults.¹⁴

Musical cues can be delivered as externally generated cues, such as playing a song on a speaker, or internally generated cues, such as singing aloud or in one’s head (i.e., mental singing). Previous work in our group demonstrated both types of cues are effective at improving gait parameters in individuals with PD, although internally generated cues have been shown to reduce gait variability whereas externally generated cues may increase gait variability.^15–17 These results are important as increased gait variability is linked to fall risk in individuals with PD¹⁸, but more research is needed to understand the underlying mechanisms behind these improvements as well as how other comorbidities may affect this response.

A complicating factor for gait dysfunction in PD is cognitive decline, which is often overlooked in research. While it appears gait dysfunction and cognitive decline are associated in PD^19–23, little is known about how cognitive impairment may impact the effectiveness of interventions to improve gait dysfunction. Individuals with cognitive impairment or dementia are often excluded from intervention studies in PD to reduce confounds.⁸ However, given that approximately 24–31% of people with PD are estimated to have comorbid dementia (PDD)²⁴, it is important to explore how individuals with PDD respond to interventions for gait dysfunction. In this study, we explored the effect of musical cues on gait parameters in individuals with PDD.

Methods:

Study Design

This study was a single center, cross-sectional, interventional study.

Participants

Participants were recruited through the Movement Disorders Clinic (MDC) at Washington University in St. Louis School of Medicine. Inclusion criteria for this study were as follows: a diagnosis of idiopathic, typical Parkinson disease according to the UK Brain Bank Criteria; Hoehn & Yahr stage 2–3 (mild to moderate disease severity)²⁵; stable on all PD medications for at least two months prior to study entry; normal or corrected to normal hearing; score of ≥ 1 on the Movement Disorders Society Unified Parkinson Disease Rating Scale – Part III – Motor Aspects (MDS-UPDRS-III)²⁶ item #10 indicating observable gait impairment; able to walk for ten continuous minutes independently; a score of one or less on item #7 on the New Freezing of Gait Questionnaire (nFOG-Q)²⁷, indicating freezing episodes are not moderately or significantly disturbing to daily walking; and dementia as defined by a Clinical Dementia Rating (CDR) score of 0.5–1.0 (very mild to mild dementia)²⁸. Seventeen individuals met these inclusion criteria and were enrolled in the study.

As this was a within-group cross-over study, there was no control group and no group randomization. This study was approved by the Institutional Review Board (IRB) of Washington University in St. Louis School of Medicine.

Protocol

This study consisted of one in-person lab visit. All testing was done with participants in their on-state for medication. During the lab visit, participants were first assessed using the MDS-UPDRS-III and nFOG-Q to confirm eligibility for the study. Participants were also asked if they had any previous musical training. After eligibility was confirmed, participants were fitted with six APDM Opal sensors²⁹ worn on the feet, wrists, lumbar spine, and sternum (Figure 1). These sensors sample at a frequency of 128 Hz and were used to collect various gait parameters, with a focus on velocity, cadence, and stride length. For all gait tasks, participants completed three 30-second trials of each cueing type and condition in a 100-foot hallway. All participants were followed by a trained physical therapist during walking trials to reduce fall risk during the intervention. No falls or adverse events occurred during data collection.

Figure 1:

APDM Opal Sensor placement.

All participants first completed three 30-second uncued walking trials to determine baseline gait parameters. The average cadence over these uncued trials was then used to calculate four cueing tempos for each participant: 90%, 100%, 110%, and 120% of uncued walking cadence. These cues were selected based upon previous work in our group looking at internal vs. external cues in people with PD without dementia.¹⁶ Cues were rounded to the nearest five beats per minute. While cueing tempos were being calculated, participants were presented with a list of songs to choose from for the cued walking tasks. The list included songs selected with the help of a music therapist that have a 4/4 beat, salient tempo, and are likely to be familiar to many individuals. The songs available were as follows: “I’ve Been Working on the Railroad”, “You Are My Sunshine”, “This Land is Your Land”, “Skip to My Lou”, “She’ll Be Coming Around the Mountain”, “Don’t Sit Under the Apple Tree”, “When the Saints Go Marching In”, “You Can’t Hurry Love”, “When Johnny Comes Marching Home Again”, and “Seventy-Six Trombones.” Participants were provided with the song lyrics prior to starting the cued tasks to ensure they knew the words. After reviewing the lyrics, participants were recorded singing the song out loud without reading the lyrics from the sheet to ensure they remembered the lyrics correctly. Additionally, they practiced walking and singing out loud to ensure they were able to perform the two tasks simultaneously.

After cueing tempos were determined, participants received two different cueing methods at each tempo. The two cueing methods were A) Music and B) Mental (Figure 2A and B). Music served as the externally generated cue, while Mental served as the internally generated cue. For all cued trials, one chorus of the selected song was played at the correct tempo while the participant was standing still as a reminder of the tempo. Cues were delivered via speakers and volume was adjusted so participants could hear throughout the entire trial. During the Music task, the cue kept playing and participants were instructed to walk to the beat, rather than the melody, of the song as they listened. During the Mental task, the cue turned off and participants were instructed to walk to the beat of the song while singing in their heads in silence. To ensure participants were doing the task, they were asked if they were singing in their heads after each trial. Participants were instructed not to move their mouths during the mental singing task, and this was monitored by the physical therapist who was walking with the participant for safety. Prior to the start of the cued trials, a member of the research team demonstrated walking on the beat to ensure all participants received the same instruction. To mitigate order effects, task order for each participant was randomized first by tempo, then Music vs Mental condition order was randomized within each tempo. This meant that both cueing types at a given tempo occurred one after the other, but the order of cue type presented at each tempo was random. An example randomization is provided in Figure 2C. In total, participants completed 27 walking trials of 30 seconds each.

Figure 2:

Conditions and Randomization. (A) The Music condition is an external cue. and requires the participant to listen to the song while attempting to match their footfalls to the beat. (B l The Mental condition is an internal cue. requiring the participant to sing the song in their head and attempt to match their footfalls to the beat. (C) The tasks were randomized first by tempo (90. 100, 110. or 120%) and then by condition (Music vs Mental). The numbers 2–9 indicate the order in which the tasks would be performed, with uncued always being task 1.

Statistical Analysis

All data were analyzed using R statistical software.³⁰ Gait parameters including velocity, cadence, and stride length were averaged across the three trials for a given condition (i.e., Mental 110%). Mixed-effect models were used to predict velocity, cadence, and stride length as a function of cue type and tempo (as fixed effects) and to account for the within-subject nature of the design (as random effects) using the lmer4 package in R.³¹ These models tested the main effects of Cue Type (Music v. Mental), Tempo (90, 100, 110, and 120%), and their interaction as categorical factors. Random effects of participant, the interaction of participant:cue type, and the interaction of participant:tempo were included to account for the within-subject nature of the manipulation.^32,33 To determine the statistical significance of these effects, we used F-tests with Satterthwaite’s approximation for the degrees of freedom.³⁴ The Type I error rate for all tests was set at α=.017 to account for multiple comparisons.

As an initial step, we also conducted t-tests to determine if velocity, cadence, or stride length were significantly different between uncued walking and the 100% cue for Music and Mental conditions. Statistically significant differences here would suggest a substantial effect of cuing on the parameter of interest (velocity, cadence, or stride length), so a lack of significance is interpreted as minimal disturbance due to cuing, although possible effects of cuing cannot be completely ruled out (i.e., we cannot prove there was truly zero dual task cost). Since we were cuing cadence, we also looked at the percent change of cadence from Uncued to Music 100% and Uncued to Mental 100% to further support a non-substantial effect of cueing. We expected the percent change from Uncued to the cued 100% conditions to be within +/− 5% due to the rounding to the nearest five beats per minute of the song.

Results:

Seventeen individuals with PDD were enrolled and completed the study protocol. One participant only completed two trials of each condition due to fatigue, but all others completed all trials. Demographic information is presented in Table 1. The majority of participants (16 of 17) were Hoehn and Yahr stage 2. Mean velocity for uncued walking was 1.02 m/s (SD = 0.04m/s). Other relevant gait parameters are presented in Table 2. T tests comparing uncued walking to the 100% cued conditions were not significant for any variables (Table 2), indicating any potential dual task costs of the cues were likely minimal. The percent change of cadence for Uncued to Music 100% was −0.18% (95% CI = [−1.70, 1.34]%) and Uncued to Mental 100% was 1.68% (95% CI = [−0.82, 4.19]%), also indicating any potential dual task costs of the cues were likely minimal and the 100% cued conditions are a suitable comparison. Song choice distribution is presented in Figure 3. Five songs (“This Land is Your Land”, “Don’t Sit Under the Apple Tree”, “You Can’t Hurry Love”, “When Johnny Comes Marching Home Again”, and “Seventy-Six Trombones”) were not chosen by any of the participants.

Table 1:

Participant Demographics (n=17)

Age (years, (SD))	72.7 (5.1)
Gender (% male)	94%
MDS-UPDRS-III (score, (SD))	48.1 (16.2)
H&Y = 2 (n)	16
H&Y = 3 (n)	1
CDR = 0.5 (n)	10
CDR = 1 (n)	7

MDS-UPDRS-III: Movement Disorders Society – Unified Parkinson Disease Rating Scale Part III; H&Y – Hoehn and Yahr; CDR – Clinical Dementia Rating

Table 2:

Gait Parameters for Uncued and 100% Cued Conditions (n=17)

	Uncued	Mental 100%	Music 100%	Uncued vs Mental 100%	Uncued vs Music 100%
Velocity (m/s)	1.02 (0.04)	1.04 (0.05)	1.01 (0.05)	t = 0.248 p = .806	t = −0.120 p = .905
Cadence (step/min)	110.00 (2.60)	112.04 (2.90)	109.90 (3.29)	t = 0.557 p = .582	t = −0.029 p = .977
Stride Length (m)	1.12 (0.03)	1.12 (0.05)	1.10 (0.04)	t = −0.004 p = .997	t = −0.158 p = .876

Mean (SD). t values and p values are presented for Uncued vs Mental 100% t tests and Uncued vs Music 100% t tests. No tests were significant (p<0.01).

Figure 3:

Song choice distribution.

Results of each mixed effect model are presented in Tables 3 and 4. The main effect of Tempo was significant for all models, while main effect of Cue Type and the Cue Type x Tempo interaction were not. This indicates velocity, cadence, and stride length all vary with regards to the Tempo, but Cue Type (Music vs Mental) and the Cue Type x Tempo interaction (i.e., Mental 110% vs Music 110%) were not substantially altering velocity, cadence, or stride length. Marginal means were calculated for the velocity, cadence, and stride length models (Table 4) for each cue percent.³⁵ Only velocity at a cue rate of 90% was significantly different from 100%, indicating participants walked slower with a cue rate of 90%, but did not significantly alter their cadence or stride length. Participants did not significantly increase their velocity, cadence, or stride length with faster cue rates of 110% and 120% (Table 4).

Table 3:

Results of mixed effect models, with main effect of type (music vs mental cue) and percent (90, 100, 110, and 120%) and the interaction of type and percent.

	Type		Tempo		Type x Tempo
Variable	F-value	p-value	F-value	p-value	F-value	p-value
Velocity (m/s)	0.79	.388	11.51	<.001	1.19	.323
Cadence (step/min)	3.04	.100	11.13	<.001	2.08	.115
Stride Length (m)	0.06	.806	5.68	.002	0.75	.526

^*Models included random intercepts of Participant, Participant:Cue Type, and Participant:Tempo to account for the within-subject nature of the design.

Table 4:

Marginal means for each cue percent.

Variable	90%	100%	110%	120%
Velocity (m/s)	0.97 [0.83, 1.11]*	1.03 [0.89, 1.16]	1.07 [0.93, 1.21]	1.08 [0.95, 1.22]
Cadence (step/min)	107 [101, 113]	111 [105, 117]	114 [108, 120]	115 [109, 121]
Stride Length (m)	1.09 [0.95, 1.22]	1.11 [0.98, 1.24]	1.13 [1.00, 1.27]	1.13 [0.99, 1.26]

Mean [95% confidence interval],

^*denotes significant difference from 100% cue.

Figure 4 shows individual changes in response to cues for each participant. While most of the lines are fairly flat, indicating little to no change in response to the cues, some lines have clear increases from 90% to 120% potentially indicating an ability to modulate gait (Figure 4). Due to the variation seen across participants, an exploratory analysis was performed to assess the relationship of previous musical training on response to musical cues. The models described above were re-run with music experience included as a main effect. Seven individuals indicated they had previous musical training. When adding this into the models, the main effect of music experience was not significant for velocity (F= 0.98, p=.338), cadence (F= 0.18, p=.680), or stride length (F= 1.57, p=.230).

Figure 4:

Gait parameters for each condition by participant.

Discussion:

We did not find reliable evidence that participants with PDD were able to increase gait velocity, cadence, or stride length in response to music or mental singing cues. Previous work including individuals with PD without dementia found velocity and cadence were significantly different from uncued walking at cue rates of 90% and 110% of uncued cadence for both Music and Mental conditions.¹⁶ This indicates slower velocity and decreased cadence with the 90% cue and faster velocity and increased cadence at 110% cue respectively.¹⁶ Stride length was also shorter at the 90% cue rate for both Music and Mental when compared to uncued.¹⁶ The 120% cue rate was not utilized in prior work. While participants with PDD did walk significantly slower at the 90% cue rate, there were no significant differences seen for cadence or stride length at 90% nor were there significant differences for velocity, cadence, nor stride length at the faster cue rates.

It is unclear why individuals with PDD respond differently to the musical cues than those with PD without dementia. One explanation might be a reduced ability of individuals with PD and cognitive impairment to adapt their gait. A recent study found individuals with PD and freezing of gait (FOG), who did not show adaptation during split belt treadmill training, had significantly lower scores on working memory and visuospatial processing assessments, indicating that cognitive impairment may hinder the ability to adapt.³⁷

Difficulty adapting to cues may be inflated by the differences in baseline walking characteristics. Compared to the previously referenced study¹⁶, the individuals in this sample walked slower (1.02±0.04 m/s vs. 1.24±0.19m/s) than individuals with PD without dementia. Additionally, individuals with PDD took shorter strides (1.12±0.03m vs. 1.34±0.18m). It is difficult to directly compare these two groups as the participants in the current sample are older (72.7±5.1 vs. 65.8±6.5) and have higher MDS-UPDRS-III scores (48.1±16.2 vs. 24.9±10.27) than the PD without dementia cohort, but the differences in gait noted may point to individuals with PDD having greater gait impairment generally, which may explain why cues were not able to be utilized in the same manner as by those with PD without dementia. It is important to note that the current PDD sample walks faster than PD patients in H&Y stage 2–2.5 (0.97m/s)³⁸, so the differences in baseline gait speed likely do not account for the difference in response to cues.

Another explanation for the lack of adaptation could be the manner in which participants were instructed to use the cues. Participants were instructed to walk to the beat of the song, but were not given further instruction on how to modify their gait. A previous study utilizing metronome cues in individuals with PD and mild cognitive impairment found gait speed and stride length were significantly increased when participants were instructed “as you walk try to take a big step in time to the beat” vs when instructed solely “as you walk try to step in time to the beat.”³⁹ This may indicate the addition of attentional cues with RAS may better modify gait in this population. As stated above, a research team member demonstrated how to walk on the beat in the present study, but perhaps an additional attentional cue may have improved performance.

The lack of significant changes to gait parameters in individuals with PDD may also be due to the variability among the participants included in the present study. As shown in Figure 4, some participants show clear increases in gait velocity and cadence with faster cue rates, which mirrors data from individuals with PD without dementia.¹⁶ This indicates there may be some individuals with PDD who are able to respond to the cues. There is evidence to suggest individuals with PD without dementia also have varying abilities to respond to auditory cues^40,41, so it may be worth further exploration of individuals who are able to respond to the cues to determine what factors predict responsiveness. Some factors that may be important include previous music training (i.e., playing an instrument or vocal training), initial gait speed, and motivation. As mentioned above, music experience did not appear to influence response to cues in this sample. We also did not see obvious trends for other variables of interest including UPDRS score and CDR rating. A larger sample may be useful to further explore various confounders.

Conclusions:

Neither music nor mental singing cues increase gait velocity, cadence, or stride length in individuals with PDD. Further research is necessary to understand the underlying mechanism of auditory cues for gait impairment in PD and specifically how that may differ with a diagnosis of comorbid dementia.

Highlights.

• People with Parkinson disease and dementia have walking issues

• Music cues can help people with Parkinson disease without dementia walk better

• We wanted to explore the effect of music cues and singing in one’s head in people with Parkinson disease and dementia

• Music and singing in one’s head were not effective at improving walking in people with Parkinson disease and dementia