Abstract
Abstract
INTRODUCTION
Parkinson disease (PD) is a neurodegenerative disorder associated with a loss of dopamine in the Basal Ganglia (BG) and is characterized by motor signs such as tremor, rigidity, bradykinesia, and postural instability [1]. In addition to the commonly recognized motor signs, prior experimental studies have suggested that there may be focused deficits in motor learning, specifically implicit motor sequence learning, in persons with PD relative to age-matched controls [2–4]. Individuals frequently must learn new skills or relearn old skills during rehabilitation and often those skills occur in a sequence, such as the sequence of steps to safely and effectively learn a transfer. Improved function through rehabilitation relies, in part on implicit motor learning. Therefore, identification of a deficit in implicit motor sequence learning may aid rehabilitation specialists in determining methods to optimize learning of functional tasks.
Implicit motor sequence learning (IMSL) refers to the integration of simple or complex isolated serial movements into a single unit of movement without awareness of the serial pattern that has been learned [5]. IMSL has frequently been studied experimentally by the serial reaction time task (SRTT) [6]. This task embeds a repeating motor sequence within the random motor sequences during practice without informing the participants about the presence of the embedded repeating sequence, thus with practice participants come to learn the repeating sequence without awareness. At the end of practice, evidence of IMSL occurs if there is greater improvement (as measured by a decrease in reaction time) on the repeating sequence compared to the random sequence. In the SRTT, this difference is defined as sequence-specific learning (SSL).
A prior meta-analysis was performed by Siegert, Taylor, Weatherall, & Abernethy [7] which assessed implicit sequence learning that included both verbal and motor sequence training in individuals with PD and summarized six studies from 1987–2005. They concluded that individuals with PD were impaired on implicit sequence learning (standardized mean difference, 0.73). However, this prior meta-analysis failed to describe the motor, cognitive and disease characteristics of the individuals with PD that participated in the studies. In addition by examining both verbal and motor sequence learning, the study by Siegert, et al. [7] included skill beyond functional sequential motor tasks necessary and we sought to identify only the influence of motor sequence learning because of its importance to safe mobility.
Therefore, to address these limitations we utilized a meta-analysis with systematic review to address the following objectives: 1) Determine the extent to which IMSL is impaired in individuals with PD compared to healthy age-matched controls (HC), and 2) Summarize the design features and participants utilized in the included studies.
METHODS
A systematic review and meta-analysis of the existing literature were performed. Specific criteria were utilized to ensure that the review was comprehensive and that there was no bias in the decision to include or exclude studies. This was followed by a quantitative analysis of the data from the resultant research articles using Cochrane Review Manager for determination of the standardized mean difference [8].
Systematic review; Literature search
The following electronic databases were searched: Medline, CINAHL, Biomedical Reference Collection, Sport Discus, The Cochrane Library, PsycArticles, Psychology and Behavioral Health Collection, Psycinfo, and Google Scholar. The search was limited to studies in English, human subjects, published between January 1995 to end of September 2014. Authors (HH, NH) read titles and abstracts in separate databases to identify potential studies. This initial search assessed published studies containing the following keywords and their variants; Parkinson disease and implicit motor task, skill or learning.
The second step of the search was performed by two of the authors (HH, NH) and required further screening of the identified articles to ensure that initial criteria were met. Articles that met our criteria were read to identify if: 1) individuals studied had PD, 2) a motor task was performed (verbal tasks were excluded), 3) individuals were not informed of a repeating sequence, thus an implicit task was performed, and 4) SSL could be determined from the articles, by providing performance data for repeated and random segments at the end of the experiment. Discrepancies were discussed and two authors (HH, NH) came to a consensus decision.
Level of evidence and quality of study
The final step was performed by authors (HH, NH) to determine the level of evidence and quality of each study using a scale described by the American Academy of Cerebral Palsy and Developmental Medicine (AACPDM; https://www.aacpdm.org/resources/outcomes/systematicReviewsMethodology.pdf). Level of evidence was rated on a five-category scale. Level I designs are well-controlled experiments requiring random allocation and manipulation of the intervention. Level II designs do not include randomization but are well-controlled experiments or prospective comparison studies. Level III designs are retrospective comparison studies. Level IV designs have no comparison group or condition. Level V designs are non-empirical evidence. For each study, a quality score, based on AACPDM guidelines, was calculated, with a point provided for positive responses on each of the following questions: 1) Were inclusion and exclusion criteria of the study population well described and followed?; 2) Was the intervention well described and was there adherence to the intervention assignment?; 3) Were the measures used clearly described?; 4) Was the outcome assessor unaware of the intervention status of the participants?; 5) Did the authors conduct and report appropriate statistical evaluation including power calculations?; 6) Were dropout/loss to follow-up reported and less than 20% ?; 7) Considering the study design, were appropriate methods for controlling confounding variables and limiting potential biases used? Quality scores were categorized as follows: strong (score of 7 or 6), moderate (score of 5) and weak (score of 4 or less).
Study inclusion criteria
Studies were included in this review if: 1) they were level I or II studies based on the AACPDM level of evidence criteria; 2) if they met a quality rating score of at least 4 according to the AACPDM scale; 3) if data were reported or could be extracted to determine SSL, such that a random sequence was provided and a repeated sequence was provided; and 4) if the comparison group were healthy, age-matched controls.
Data extraction methods
For the purposes of the meta-analysis, raw mean and standard deviation data for the final block of repeated and the final block of the random practice trials were obtained for each study according to the following methods: 1) full data were presented in the article; 2) for missing data, such as the standard deviation in the text, the authors of the studies were contacted to obtain missing data; 3) for missing data when the authors did not respond, the raw data was obtained by reading the data from the plots/figures provided. This graphical analysis required extracting the raw values (means and standard error or standard deviation) from the plots and the mean extraction values obtained by the authors (HH, NH) were utilized. If the article had been included in Siegert, et al. 2006, then this mean was also combined to obtain the value of SSL. When standard error was provided, the standard deviation was calculated by multiplying the standard error by the square root of the sample size.
In order to summarize the design features and participants included in the studies, the tasks utilized, the operational definitions of motor learning, assessment of explicit knowledge, and demographic/disease specific variables (age, gender, cognition, severity and duration of disease and medication status) for each included study were extracted.
Analysis
The mean between-group difference and the standard deviation of SSL were calculated for individuals with PD and healthy controls. Forest plots were used to depict the comparison of the groups by assessing the standardized mean difference [9]. Statisticalanalysis utilized random effect sizes with 95% confidence intervals (CI) and I2 value for overall heterogeneity. If heterogeneity was less than 74% , then fixed effect sizes were reported. An alpha value of less than 0.05 was considered statistically significant. All articles meeting the primary criteria were placed in the Forest plot to determine the overall effect of IMSL as described by the authors.
RESULTS
Systematic review
The initial literature search performed by the authors (HH & NH) identified 57 articles. After duplicates were removed, 46 articles were screened again to ensure they met initial criteria. Eighteen articles were excluded upon the second screening because they did not include a motor task or individuals were not identified as having PD or there was no age matched control. Thus, 28 articles met the original criteria and were assessed via full-text review. Of the 28 articles, 15 met the a priori inclusion criteria and were reviewed for meeting the AACPDM level of evidence and quality rating scale criteria. All 15 articles [2, 10–23] met the AACPDM criteria established (See Table 1). See Fig. 1 for an illustration of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram [24].
Meta-analysis
An initial overall assessment of SSL for all 15 studies was performed and standardized meandifferences, using a random effect size and 95% confidence intervals (CI) are summarized in Fig. 2 and Table 2. An I2 value of 87% was observed, indicating high heterogeneity. The overall random effect size was moderate, 0.83 (CI 95, 0.30 to 1.36; p < 0.01) favoring HC; suggesting that individuals with PD were impaired in motor SSL compared to HC.
Design features, demographic and disease-specific variables
Summaries of design features, demographic and disease-specific variables (age, gender, cognition, severity and duration of disease, and medication status) are provided in Table 3. All of the studies that met our final criteria utilized a variant of a serial reaction time task (SRTT) task with 14 of 15 studies examining skill acquisition after one day of practice. The number of individuals with confirmed PD was 299, compared to 244 age matched healthy controls. The mean age of the subjects; PD, M = 61.3 years [range 43.6–72.2] and HC, M = 57.6 years [range 38.7 –83.5] were similar across groups and number of subjects. Males (n = 274) and females (n = 211) were equally distributed. Eleven of the 15 studies utilized the Mini Mental State Exam (MMSE; max score of 30) as an assessment of cognitive function. On average, individuals with PD scored 28 compared to the HC scoring 29. In 5 studies, the Hoehn &Yahr (H&Y) scale scores were not reported. The remaining ten studies yielded an overall median H&Y score of 2.0 [range 1–3] and this included values for participants with PD on and off their dopamine medication. Four studies reported the Unified Parkinson Disease Rating Scale (UPDRS) motor score which averaged 26.0 [range 9 –43.9] and this included values on and off medications. Four studies reported a UPDRS total score, with an average of 37.4 [19.7–63.1]. Assessment of the H&Y and UPDRS scores based on medication status was unable to be determined because of variability in reporting. Ten of the studies reported duration of symptoms (mean 90.3 months [6–247]). Medication status during practice for the individuals with PD was reported in 14 of the 15 studies. Four of the studies had some or all of the individuals off their usual dosage of dopamine. Assessment of explicit knowledge found that three studies did not assess explicit knowledge of the repeating sequence, with one study reporting that individuals with PD did gain explicit knowledge. The remaining 11 studies reported that no greater than 50% of the individuals gained explicit knowledge of the repeatingsequence.
DISCUSSION
In the context of PD, IMSL is critical for successful rehabilitation to ensure that individuals can learn or re-learn successful movement strategies for the performance of functional tasks. Without IMSL, rehabilitation may not be successful. Recognition that individuals with PD are impaired in their implicit ability to learn motor sequences will allow rehabilitation specialists to identify alternative resources that may augment motor learning, such as providing explicit instruction, additional verbal cueing, or additional practice. In order to gain insight into the presence and consistency of a motor sequence learning deficit in persons with PD, we utilized a meta-analysis with systematic review to address the followingobjectives.
First we sought to determine the extent to which IMSL is impaired in individuals with PD compared to age matched HC. Our results revealed a moderate pooled effect size, which supports that individuals with PD demonstrate clear deficits in their ability to implicitly learn motor sequences compared to HC; similar to results assessing visuomotor and verbal sequence learning [7, 25–28]. Despite these deficits, individuals with PD do appear to improve performance as a result of practice, but to a lesser degreethan HC.
Second, we sought to summarize the designfeatures and participants utilized in the included studies. While our results suggest a deficit in IMSL in individuals with PD, our additional analysis of the articles finds multiple discrepancies that could account for the inconsistent results between studies [14]. We suggest that the following areas need to be considered in future studies to allow for clarification of the IMSL deficits observed in individuals with PD. These include: 1) the design features utilized, including the assessment of performance versus learning, the type of task, and the amount of practice utilized, and 2) PD-specific variables, including the role that medication status and cognition in persons with PD may play in impacting IMSL. By understanding how these components influence IMSL in individuals with PD we may be able to find a tool that may help ameliorate this deficit.
Design features used to assess IMSL
Our broad search criteria sought to find studies that assessed SSL outcomes using a motor task in individuals with PD. However, it is critical to note that all of the studies but one reported in this meta-analysis performed only one day of practice. Contemporary operational definitions of motor learning reserve the term learning for skills that have been acquired and are retained after a period of no practice. In contrast, motor performance is used to describe a change in skill observed during a single practice session [29]. When considered in the context of this definition of the motor performance-motor learning distinction, these studies have only examined changes in motor performance. Therefore, any assertion that these results reflect persistent changes in skill should be interpreted with caution.
Secondly, our broad search criteria focused on assessing motor tasks, and the only motor task that has been utilized assessing IMSL in individuals with PD is the SRTT. Other studies have assessed motor IMSL in other populations using a continuous tracking task (CTT), but no studies have assessed a CTT in individuals with PD [30]. Therefore, the results can only be applied to tasks of a discrete nature rather than a continuous nature. Future research is warranted to identify if an IMSL deficit persists during a CTT in individuals with PD.
Finally, the quantities of practice provided by the studies in this meta-analysis varied extensively. It has been suggested that the deficit in IMSL for individuals with PD may be due to insufficient practice [31]. Since variations in practice dosages were not part of the design of individual studies, it is not known how much or if any improvement could be achieved if more practice was provided in individuals with PD using the SRTT. Current rehabilitation literature for multiple populations of individuals (Stroke, Spinal Cord Injury, and Multiple Sclerosis) have suggested that increased intensity, defined as increased practice, is required to improve motor function, perhaps via neuroplastic processes [32–34]. A greater understanding of the practice dose-response relationship is needed as increased motor skill practice may mitigate the IMSL deficit in individualswith PD.
Characteristics of participants with PD
The conceptual model which supports studying a potential deficit in IMSL in individuals with PD stems from the data suggesting that the Basal Ganglia (BG) is critical for implicit motor sequence learning [14, 35–38]. While PD is a neurodegenerative disorder associated with a loss of dopamine in the BG, broader neuronal damage is often observed and therefore complicates the use of PD as a model disorder to examine the role of the BG in IMSL [39]. To address the variability in neuronal damage, a more homogenous sample of individuals with PD could be captured or more clear characterization of the population being assessed could be utilized to aid in the understanding of this IMSL deficit. Recent research suggests that specific PD motor phenotypes such as freezing of gait may have greater IMSL deficits compared to non-freezers with PD and healthy controls [21]. Since none of the studies reviewed in this meta-analysis examined the influence of motor phenotype on the amount and rate of learning, future research exploring potential differences in sequence learning between varied motor phenotypes and other disease-specific differences in individuals with PD appears warranted.
Furthermore, given the ubiquitous presence of dopamine replacement medications in the pharmacologic treatment of PD, the role that medication may play on IMSL must be considered. Recent research has suggested that both endogenous dopamine and the treatment with dopamine replacement medications could influence IMSL [40–43]. In the studies included in this meta-analysis, there was broad variability of participant medication status. For this reason, there is insufficient data to determine the influence of medication status on IMSL. Future studies should describe if individuals were taking dopamine replacement medications and if so, summarize the amount using the levodopa equivalent daily dose calculation [44]. Additionally, future studies could manipulate dopamine status as an independent variable to gain insight into the impact of dopamine replacement on motor IMSL.
Finally, because cognitive decline and dementia occurs in 20–80% of patients with PD, a more thorough determination of cognitive status is warranted in all future studies [45]. Multiple neurobiological models have suggested that sequence learning requires intact dorsolateral prefrontal cortex and parietal regions, important regions for executive functions such as working memory and planning [37, 46]. There was inconsistency in the assessment of cognition in the studies described in this meta-analysis with the Mini-Mental State Exam (MMSE) being the most commonly used tool to assess cognitive status in the participants. Future research should examine more specific facets of cognition; such as attention, executive function and psychomotor speed, set shifting, and working memory since these areas are observed to be impaired in individuals with PD [4, 47–49]. Finally, differentiation of cognitive decline versus motor decline is warranted to determine the possible influence of these differing components on SSL in individuals with PD [50].
Limitations
This meta-analysis examined in-depth information from the multiple studies that met the inclusion criteria. Data were obtained via author, text, or extraction. Although attempts were made to directly obtain raw data from study authors, responses were limited. In some cases, the extraction method required estimation of data means and standard deviations. Efforts were made to maximize accuracy, yet in some cases actual data were not available.
Overall, the I-squared values in the meta-analysis were high and indicate the high variability observed in these studies. In an attempt to minimize these effects, a random effect size was utilized to account for the lack of homogeneity of the studies; however, the calculation of fixed effects found no difference in effect sizes or I-squared values.
CONCLUSIONS
The results of this meta-analysis suggest that during one day of practice on a SRTT, individuals with PD are impaired in IMSL compared to HC. However, more research is warranted to determine the impact of this deficit as a performance or learning deficit, the impact that the type of task may have on this deficit, and the impact that the amount and type of practice may play on IMSL in individuals with PD. Additionally, further clarification is warranted in determining the impact that medication and cognition may play on IMSL. Given that successful motor rehabilitation of functional tasks in persons with PD is highly dependent on IMSL, improved clarification of the influence of these variables is critical.
RELEVANT CONFLICTS OF INTEREST/FINANCIAL DISCLOSURES
Nothing to report.