This article is commented on by Lin on pages 78 of this issue.
Causal relation between spasticity, strength, gross motor function, and functional outcome in children with cerebral palsy: a path analysis
Article first published online: 3 DEC 2010
© The Authors. Journal compilation © Mac Keith Press 2010
Developmental Medicine & Child Neurology
Volume 53, Issue 1, pages 68–73, January 2011
How to Cite
KIM, W. H. and PARK, E. Y. (2011), Causal relation between spasticity, strength, gross motor function, and functional outcome in children with cerebral palsy: a path analysis. Developmental Medicine & Child Neurology, 53: 68–73. doi: 10.1111/j.1469-8749.2010.03777.x
- Issue published online: 3 DEC 2010
- Article first published online: 3 DEC 2010
- Accepted for publication 7th July 2010.
Aim This study examined the causal relation between spasticity, weakness, gross motor function, and functional outcome (expressed as activity limitation) in children with cerebral palsy (CP) and tested models of functional outcome mediated by gross motor function.
Method Eighty-one children (50 males, 31 females) with CP were recruited for this cross-sectional study. Their mean age was 10 years 4 months (SD 1y 9mo). Strength was assessed using the Manual Muscle Test. Spasticity was assessed by the Modified Ashworth Scale. The Gross Motor Function Measure assessed gross motor function. The Functional Skills domain of the Pediatric Evaluation of Disability Inventory assessed functional outcome. Twenty-eight children (34.6%) had quadriplegia, 44 children (54.3%) had diplegia, and nine children (11.1%) had hemiplegia. Children were classified using the Gross Motor Function Classification System with 14 (17.3%) in level I, 9 (11.1%) in level II, 13 (16.0%) in level III, 5 (6.2%) in level IV, and 40 (49.4%) in level V.
Results The proposed path model showed good fit indices. The direct effects were significant between spasticity and gross motor function (β=−0.339), between strength and gross motor function (β=0.447), and between gross motor function and functional outcome (β=0.708). Spasticity had a significant negative indirect effect (β=−0.240) and strength had a significant positive indirect effect (β=0.317) on functional outcome through effects on gross motor function.
Interpretation Activity-based rather than impairment-based intervention is more important for reducing activity limitation in children with CP. The study established a base from which researchers can further develop a causal model between motor impairments and functional outcome.
Pediatric Evaluation of Disability Inventory
What this paper adds
- • The first path analysis of prediction of functional outcome in children with CP to be published.
- • Our data confirm that spasticity and weakness are causal factors related to gross motor function.
- • Gross motor function is a causal factor in functional outcome.
Children with cerebral palsy (CP) have impairments that interfere with motor function, activity, and participation.1 These neuromuscular and musculoskeletal impairments include spasticity, dystonia, contractures, abnormal bone growth, poor balance, loss of selective motor control,2 and weakness.3 The focus of physical therapy intervention for children with CP is to improve functional outcome by reducing neurological impairments, improving strength, and preventing the development of secondary impairments.1 However, a causal relation between motor impairment and functional outcome has yet to be established.
Spasticity, one of the most common problems in children with CP and a component of upper motoneuron syndrome, is a motor disorder characterized by a velocity-dependent increase in tonic stretch reflexes with exaggerated tendon jerks, resulting from hyper-excitability of the stretch reflex.4 Evidence pointing to a relation between spasticity and function in children with CP is equivocal. For example, Tuzson et al.5 reported that the Ashworth Scale score correlated significantly with the Gross Motor Function Measure (GMFM; r=0.83) in 18 participants with CP. In contrast, Damiano et al.6 reported no significant correlation between spasticity measures and gait parameters in 25 children with CP. Østensjø et al.7 reported a significant correlation between the Modified Ashworth Scale and the GMFM-66 (r=0.64). Although the relation between spasticity and motor ability has not been conclusively established, many intervention strategies still focus on the reduction of spasticity.8 Therefore, it is important that evidence of a relation between motor impairment and the functional outcome of activity limitation in children with CP is established using causal relation analysis.
It has been a widely held opinion that weakness is not a major problem in children with CP. Consequently, muscle strengthening has not been recommended for children with CP because it was believed that it would lead to increased spasticity and that children with CP would not benefit from resistance training.9 However, the Research Committee of the Section on Pediatrics of the American Physical Therapy Association recently reported that weaker children with CP have less endurance and more limited physical capability than stronger children.10 On the basis of this report, the American Physical Therapy Association recommended physical fitness training and muscle strengthening for these children.10 Professionals now accept weakness as another clinical feature of CP, and many researchers have reported a relation between muscle strength and motor function in children with CP.11,12 Kramer and MacPhail13 reported that knee extensor strength was moderately positively related to the GMFM (r=0.57–0.69) in 17 adolescents with CP. Damiano et al.14 reported a significant positive correlation between extensor strength and the GMFM (r=0.57). However, a direct relation between muscle strength and activity, as evaluated by validated measures such as the GMFM, has not been established.3
Several studies have reported that gross motor function is positively related to functional outcomes such as activities of daily living. Østensjø et al.7 reported that the GMFM-66 showed a positive correlation with the Pediatric Evaluation of Disability Inventory (PEDI) Functional Skills scale (r=0.76–0.94). Smits et al.15 examined the positive relation between gross motor capacity and daily functional mobility in children with CP. They reported that scores on the GMFM-66 explained 90% and 84% of the variation in scores on the PEDI Functional Skills and Caregiver Assistance scales respectively.
The activity and participation domain of the International Classification of Functioning Disability and Health (ICF)16 is now a well-established concept in rehabilitation. A major goal of therapy for children with CP is that they master the tasks and activities of daily living. However, few studies have addressed the nature and significance of the restrictions faced by children with CP in participation and performance of daily activities.1 Instead, most studies have reported a correlation between motor impairment and functional outcome,3,7 using regression analysis to establish the relation.12 However, analysis methods such as correlation cannot determine causal relations between variables.
This study aimed to establish a causal relation between spasticity, weakness, gross motor function, and the functional outcome of activity limitation in children with CP. Using path analysis and model fit indices, it used a cross-sectional design to test models of functional outcome mediated by gross motor function. Path analysis is a useful technique for testing the relations between variables, whereas model fit statistics can illuminate whether variables are best construed as exogenous or endogenous.
Eighty-one children (50 males, 31 females) were recruited for this study. They had been diagnosed with spastic CP by a physician, and attended an elementary school for physical disabilities or received hospital-based rehabilitation therapy in Korea. Their mean age (SD) was 10 years 4 months (1y 9mo). Twenty-eight children (34.6%) had quadriplegia, 44 children (54.3%) had diplegia, and nine children (11.1%) had hemiplegia. Children were classified using the Gross Motor Function Classification System (GMFCS) with 14 (17.3%) in level I, nine (11.1%) in level II, 13 (16.0%) in level III, five (6.2%) in level IV, and 40 (49.4%) in level V. In an elementary school for physical disabilities, the proportion of children with severe impairments was high because children with mild CP are enrolled in mainstream schools. This contributed to the skewed number in GMFCS level V.
The GMFCS level, distribution of motor impairment, and comorbidities of the participants are summarized in Table I. Kline17 recommends using 10 times the number of participants as parameters, and ideally 20 times as many participants as parameters, for significance testing of model effects. Although for this study an adequate sample size was considered to be 40 participants because four variables were measured, we measured 81 children with spastic CP to increase statistical power.
|Characteristic||Number of participants||Percentage of participants|
|Type of cerebral palsy|
Consent to participate in the study was obtained from the parents of all children. Approval for this study was granted by the ethics committee of Jeonju University.
Spasticity was measured bilaterally, and a spasticity rating was assigned using the Modified Ashworth Scale. Strength was measured using the Manual Muscle Test, and the GMFM and PEDI Functional Skills scale were used to measure gross motor function and functional outcome respectively.
Spasticity in the flexors and extensors of the shoulder, elbow, and wrist was measured in both upper extremities. In the lower extremities, spasticity in the hip flexors and extensors, knee extensors, and ankle plantar flexors and dorsiflexors was also measured bilaterally. Tests were performed at moderate speed (180°/s) using standardized procedures. Because it was difficult to distinguish spasticity from dystonia or rigidity. We applied the recommendations of Sanger et al.18 Specifically, we selected children in whom hypertonia varied with the speed of externally imposed movement. Severity of tone was then measured according to the Modified Ashworth Scale. Mutlu et al.19 reported values of intraclass correlation coefficients between 0.61 and 0.87.
Strength in the shoulder flexor and extensors, elbow flexors and extensors, and wrist flexors and extensors was measured in both upper extremities. Strength in the hip flexors and extensors, knee extensors, and ankle plantar flexors and dosiflexors was measured in both lower extremities. Several methods may be used to evaluate muscle strength. The Manual Muscle Test, the simplest and most widely used method, was used to determine muscle strength in children with CP in this study. Klingels et al.20 reported moderate high to very high (0.60–0.91) interrater reliability and high (>0.78) test–retest reliability in children with hemiplegic CP.
GMFM data were collected for measurement of gross motor function. The GMFM is a standard criterion-referenced test designed to assess changes in gross motor function in children with CP.21 The 88-item test assesses activities in five dimensions: lying and rolling; sitting; crawling and kneeling; standing; and walking, running, and jumping. Each item is rated according to a four-point Likert scale. Reliability of the GMFM scores is reported to be good. Nordmark et al.22 found the interrater and intrarater reliabilities were between 0.88 and 0.68 respectively. In this study, Cronbach’s alpha of the GMFM was 0.994.
The PEDI includes three scales: Functional Skills, Caregiver Assistance, and Modifications. The Functional Skills scale consists of 197 questions in three domains: self-care, mobility, and social function. Each question is scored as (1) positive or (0) negative.23 We measured a child’s functional outcome using the PEDI Function Skills scales, the English version of which was translated into Korean by Jung,24 and a bilingual physical therapist. After the initial translation, three American-educated bilingual users verified the accuracy of the translation and item compatibility was examined by 38 paediatric allied health professionals.
Cronbach’s alpha of the original PEDI Functional Skills scale was reported as 0.98 to 0.99, and the intraclass correlation coefficient was 0.88 to 0.98 for clinical samples.23 In this study, Cronbach’s alpha of the PEDI Functional Skills scale was 0.986 to −0.987.
Measurements were administered by eight therapists (six physical therapists, two occupational therapists), each with at least 3 years’ experience of providing therapy for children with CP.
A path model was used to investigate the causal relations between spasticity, strength, gross motor function, and functional outcomes in children with spastic CP. As shown in Fig. 1, the model for testing procedures in the path analysis was a recursive one in which no variable in the model had an effect on itself. The values of spasticity and strength were quantified respectively by the sum of the Modified Ashworth Scale score and the sum of the Manual Muscle Test score at each joint to indicate the extent or magnitude of impairment.
Path analysis is a statistical technique that uses bivariate and multiple linear regression techniques to test causal relations between the variables of a specialized model.25 The variables included in the path model and the presumed directions of causation were based on previous research and general logic. The correlation of variables in this study was analysed before performing the path analysis. Path analysis is an extended form of the regression model, used to test the fit of a correlation matrix against two or more causal models being compared. The task of path analysis is to predict the regression weight, which is compared with the observed correlation matrix. Regression analysis with univariate or multivariate dependent variables is a standard procedure for modelling relations among observed variables. Path analysis also allows the simultaneous modelling of several related regression relations. The goodness-of-fit test is performed to show that the model is the best possible fit.
Path coefficients were computed through a series of multiple regression analyses based on the hypothesized model. Path coefficients describe numerically to what extent a change in a predictor variable affects a dependent variable. We used the AMOS 7.0 statistical program (SPSS Inc, Chicago, IL. USA) to analyse the path models, to obtain maximum-likelihood estimates of model parameters, and to provide goodness-of-fit indices.
Inter-correlations of variables computed using Pearson’s correlation coefficients are shown in Table II. Four variables were identified as significant (p<0.001) for this study and were found to correlate with the expected directions. Multi-colinearity was not detected, as bivariate correlations did not exceed 0.80 (Jobson26).
|Gross motor function||p||Functional outcome||p||Spasticity||p|
Preliminary path analysis
Preliminary path analyses were performed to determine significant variables for model identification. The direct effects of spasticity and strength on functional outcome were not significant (Table III). The proposed path model used for this study is shown in Fig. 2.
|Path||βa||Standard error||Critical ratio||p|
|Spasticity Gross motor function||−0.339||4.560||−3.481||<0.01|
|Strength Gross motor function||0.447||0.341||4.595||<0.01|
|Gross motor function Functional outcome||0.734||0.148||6.648||<0.01|
|Spasticity Functional outcome||0.027||6.467||0.259||0.796|
|Strength Functional outcome||−0.016||0.507||−0.149||0.881|
Evaluation of the model fit was based on multiple criteria, including the theoretical meaningfulness of the model, absolute-fit indices and incremental fit measures, and model cross-validation. Absolute-fit indices measure how well a model fits the data without comparison with a baseline model. Incremental fit measures determine how well a model fits compared with a baseline model. In this study, absolute fit indices included the χ2 and root mean square error of approximation. Incremental fit measures included the normed fit index and the comparative fit index. Although χ2 was reported for the analysis of model fit, it was not used to judge model fit because of sensitivity to sample size. Therefore, χ2 was not used for fit statistics in this study.
As shown in Table IV, the proposed model showed excellent fit indices. Root mean square error of approximation scores lower than 0.05 indicated a good model fit. Scores for the normed and comparative fit indices were more than 0.9, which also indicated a good model fit.
|Fit index||χ2||Root mean square error of approximation||Normed fit index||Comparative fit index|
Direct and indirect effect of variables
Path analysis was used to predict that the variables of spasticity, strength, and gross motor function influence functional outcome and that functional outcome mediated gross motor function in children with spastic CP (Table V).
|Path||βa||Standard error||Critical ratio||p|
|Spasticity Gross motor function||−0.339||4.560||−3.481||<0.01|
|Strength Gross motor function||0.447||0.341||4.595||<0.01|
|Gross motor function Functional outcome||0.708||0.106||6.648||<0.01|
The dependent variable was functional outcome based on the PEDI Functional Skills scale. Exogenous independent variables were spasticity and strength. Gross motor function was an endogenous independent variable used as a mediating variable. As shown in Table VI, results for the direct paths were as follows: −0.339 (p=0.001) from spasticity to gross motor function, 0.447 (p=0.003) from strength to gross motor function, and 0.708 (p=0.001) from gross motor function to functional outcome. Results for the indirect paths from spasticity to functional outcome were −0.240 (p=0.001) and 0.317 (p=0.002) from strength to functional outcome.
|Predictor variables||Independent variables||Total effect||p||Direct effect||p||Indirect effect||p||R2|
|Spasticity||Gross motor function||−0.339||0.001||−0.339||0.001||0.488|
|Gross motor function||0.708||0.001||0.708||0.001||0.001|
Functional outcome is concerned with the impact of disability on an individual’s level of independence in daily life. The ICF suggests that our understanding of health and disability should be viewed in terms of the social aspects of disability.16 Almost all therapy aimed at decreasing impairments in children with CP is based on the hypothesis that decreased motor impairment leads to increased activity and participation in everyday life; however, the relation between motor impairment and functional outcome has seldom been demonstrated empirically.12,27 The factors contributing to functional outcome are important for improving functional ability in children with CP. Although motor impairments such as spasticity, weakness, restricted range of motion, and loss of selective motor control have been reported to affect functional gross motor and daily activities outcomes,3,7,12 to our knowledge, no path analysis of prediction of functional outcome in children with CP has been published.
This study used a path analysis method to determine the causal relations between spasticity, strength, gross motor function, and functional outcome in children with CP. The strength of a path analysis is that it provides an explicit theory of the relations between variables rather than simply testing a set of data from a linear relation. In addition, a path analysis produces clear results of the strength of the inherent mathematical relations. Although regression analysis provides information about the correlation of mathematical relations, path analysis looks explicitly at cause. Path analysis also is superior to regression analysis because it provides an explanation both of causal relations and the relative importance of alternative paths of influence. Depiction of a path analysis involves the use of circles and arrows that indicate causation. In path analysis, a variable that is dependent in one relation and independent in another is referred to as a mediating variable. Path analysis provides methodological advantage in that multiple measures can be used both as independent and dependent variables.25
The model proposed in this study was designed to determine the contribution of the variables of spasticity, strength, and gross motor function to functional outcome. This model was shown to be the best fit for these variables. Our data confirm and extend previous findings indicating that spasticity and weakness are causal factors in gross motor function, and that gross motor function is a causal factor in functional outcome. The path coefficients show that spasticity and strength had the largest direct effect and explain most of the variance on gross motor function. An increase of 1SD in spasticity produces a decrease of 0.339SDs in gross motor function. These findings are supported by previous studies showing significant correlation between spasticity and gross motor function.5–7 When strength goes up by 1SD, gross motor function goes up by 0.447SDs. This is consistent with previous studies in which muscle strength had a relation with motor function.11,12 An increase of 1SD in gross motor function produces an increase 0.708SDs in functional outcome level. It has previously been reported that gross motor function correlates with activities of daily living.1,15
The exogenous variables of spasticity and strength accounted for substantial proportions of variance (48.8%) in explaining gross motor function. The explanatory variables of spasticity, strength, and gross motor function accounted for substantial proportions of variance (50.1%) in explaining functional outcome. The indirect effects of spasticity and strength on functional outcome were confirmed in this study. Indirect effects include those that an exogenous variable such as spasticity or strength can have on the endogenous variable of functional outcome through another variable (gross motor function). The path model proposed here shows that spasticity had a significant negative indirect effect and that strength had a significant positive indirect effect on functional outcome through their effects on gross motor function.
This study also provides initial evidence for the mediating role of gross motor function as a variable in functional outcome and identifies impairment variables that influence gross motor function and functional outcome. These variables accounted for 48.8% of the variance in explaining gross motor function. The study also identifies the indirect effects of impairment variables and the mediating effect of gross motor function on functional outcome, accounting for 50.1% of the variance in explaining functional outcome. These results show that activity-based intervention to reduce activity limitation in children with CP is more effective than impairment-based intervention.
The GMFM is used to measure gross motor function based on a developmental model and is a measure of activity level in the ICF. In contrast, the PEDI measures capacity (a measure of activity limitation, rated by the functional skill) and performance (a measure of participation restrictions, rated by the level of caregiver assistance required). Participation restriction is a limitation at the social level, whereas activity limitation is a limitation at the personal level. It is thought that the GMFM score affects the PEDI score. Previous literature about the relation between the GMFM and PEDI has shown that the GMFM has high explained variance. Smits et al.15 used only the mobility domain of the PEDI; the participants in their study were mostly in GMFCS level I (47.3%). Østensjø et al.7 reported that the GMFM-66 score was a strong predictor of PEDI, accounting for 88% of the variation in mobility, 76% of the variation in self-care, and 57% of the variation in social function. An explanation of the lower explained variance in the current study may be the different participants and measurements used.
Although the results showed 50.1% of explained variance, the identification of the indirect effects of spasticity and strength on functional outcome is clinically meaningful. Also, 48.8% of the variance of spasticity and strength explained gross motor function, suggesting that 51.2% of the variance of gross motor function could be explained by other neurological impairments.
Further study is needed to provide a more detailed description of the nature of the relations between gross motor function and functional outcome. The amount of explained variance in this study suggests that other variables should be included to explain functional outcome more fully. In addition, we will use a path model to examine the discrepancy according to the domain of functional outcomes and GMFCS level with more participants.
There were three main limitations to this study. First, participants were skewed to GMFCS level V. Second, no data on potentially significant variables such as range of motion,7 postural control, and loss of selective motor control were collected.2 The explained variance could change according to the GMFCS level of the participants and exogenous variables. Third, we did not measure the interrater reliability between evaluators. The causal relation among neurological impairments, gross motor function, and functional outcome should be investigated in other groups in different GMFCS levels. We also recommend that research looks for other exogenous variables of functional outcomes.
- 2Treatment of neuromuscular and musculoskeletal problems in cerebral palsy. Pediatr Rehabil 2001; 4: 5–16..
- 13Relationships among measures of walking efficiency, gross motor ability and isokinetic strength in adolescents with cerebral palsy. Pediatr Phys Ther 1994; 6: 3–8., .
- 16World Health Organization. The International Classification of Functioning, Disability and Health (ICF). Geneva: World Health Organization, 2008.
- 17Principles and Practice of Structural Equation Modeling. 2nd edn. New York: Guilford Press, 2005..
- 19Reliability of Ashworth and modified Ashworth scales in children with spastic cerebral palsy. BMC Musculoskelet Disord 2008; 10: 9–44., , .
- 21Gross Motor Function Measure Manual. 2nd edn. Toronto: McMaster University, 1993., , , et al.
- 23Pediatric Evaluation of Disability Inventory (PEDI): Development, Standardization and Administration Manual. Boston, MA: New England Medical Center and PEDI Research Group, 1992., , , , .
- 24Adaptation of the pediatric evaluation of disability inventory in a Korean setting: preliminary study. [Masters thesis]. Seoul, Korea: Yonsei University, 2006..
- 25A User’s Guide to Path Analysis. Oxford: University Press of America, 2006..
- 26Applied Multivariate Data Analysis. Volume I. Regression and Experimental Design. New York: Springer, 1991..
- 27People with cerebral palsy: effects and perspective of therapy. Neural Plast 2001; 2: 51–9..