The relationship between spasticity and muscle strength in persons with cerebral palsy (CP) has been discussed widely but remains inconclusive. Ross and Engsberg1 summarized several conflicting hypotheses on this issue. The first hypothesis is that spasticity within a single muscle group (eg, hamstrings) causes weakness in the opposing muscle group (eg, quadriceps). The second hypothesis is that spasticity within a single muscle group (eg, quadriceps) interferes with the force production of that muscle group, either in a spastic-strong or a spastic-weak relationship. The third hypothesis is that no relationship exists between spasticity and strength, either within a single muscle group or between agonist and antagonist.1
Many clinicians who work with persons with CP regard spasticity as a major factor that contributes to functional limitations and abnormal movement patterns. This assumption has lead to the development of some specific interventions, such as serial casting and selective posterior rhizotomy, aimed at decreasing spasticity. While the relationship between spasticity and gross motor function has been debated, it has not been investigated extensively.2 Clinicians and researchers hold different opinions on this issue. Bobath,3 for example, suggested that spasticity is the major cause of movement and postural abnormality and that the elimination of spasticity would increase an individual’s motor function. Damiano et al4 found that knee extensor Ashworth scores were moderately related to functional scores in subjects with spastic CP (r = −0.48 to −0.68). In addition, Maruishi and colleagues5 reported a negative relationship between muscle tone and the Barthel Index in 256 adults with CP. Guiliani,6 on the other hand, suggested that spasticity alone was insufficient to explain functional limitations. In a relatively large-scale study, Abel et al7 also noted that spasticity only weakly related to motor function in ambulatory children and adolescents with CP. Evidence is not yet sufficient to rule out the role of muscle spasticity in functional limitations.
Similarly, muscle weakness is a prominent impairment observed in persons with CP, but its relationship to functional limitations was not emphasized until recently. Several researchers have studied the effect of strength training in persons with CP and reported positive effects.8–11 Interestingly, all of the muscle strengthening programs in these studies produced gains in muscle strength, but not every one of them reported a positive changes in functional abilities.8,10,11 Thus, the relationship between strength and motor function in this population might be more complex than expected. Kramer et al12 showed that knee extensor isokinetic peak torque was moderately related to gross motor ability (r = 0.57 to 0.68), and to walking efficiency as measured by the Energy Expenditure Index (EEI; r = 0.65). Furthermore, Damiano et al9 demonstrated that isometric knee extensor strength was moderately related to walking velocity (r = 0.60) but no relationship was observed between knee flexor strength and gross motor function. In a subsequent study with 10 children with CP, Damiano et al13 further showed that isometric knee extensor strength had a positive relationship with gait velocity (r = 0.68) and the Gross Motor Function Measure (GMFM; r = 0.57). Previous studies either had too small of a sample size or only applied to adolescents with CP. To date, most studies have not demonstrated a relationship between hamstring strength and functional abilities. However, one study demonstrated that strength training of knee flexors improved gross motor function.14
Inconsistent results in the literature are likely attributable in part to the fact that no single study examined the relationships among strength and spasticity of both hamstrings and quadriceps, and functional abilities. Furthermore, the muscle imbalance across the joint may be a better predictor of function than the strength of a single muscle. No studies to date have looked at the relationship between muscle imbalance across the joint (quadriceps/hamstrings ratio) and functional abilities. Thus, further examination of these relationships is necessary before proceeding with experimental studies. While one cannot infer cause and effect from correlational methods, these methods do allow us to analyze how several variables, either singly or in combination, might be related to the phenomena of interest.15 Once the possible field of variables is narrowed, researchers can design experiments to determine cause and effect. The purpose of this study was twofold. First, we aimed to determine whether relationships exist among various measures of knee muscle spasticity, isometric knee muscle strength, knee muscle balance, gross motor function, and walking efficiency in children with spastic CP. Second, we aimed to evaluate the relative contributions of impairment measures (spasticity and muscle strength) to functional outcome measures (gross motor function and walking efficiency) in children with spastic CP.
Subjects and Tester
Twenty-seven children with spastic CP ages seven to 12 years were recruited in Taoyuan, Taiwan Republic of China, and Dallas, Texas. Inclusion criteria for this study were the ability to follow simple verbal commands, to walk independently for at least three minutes with or without assistive devices, and to have functional abilities at Level I, II, or III on the Gross Motor Function Classification System (GMFCS). Exclusion criteria were uncontrolled seizures, known cardiovascular disease, recent (less than six months before the study) selective dorsal rhizotomy, musculoskeletal surgery, or Botox® (Allergan, Irvine, Calif) injection in the lower extremity. Participants’ guardians signed the informed consent approved by one of the following institutional research boards: University of Texas at Southwestern Medical Center, Texas Woman’s University, or Lin Kou Chang Gung Memorial Hospital.
The primary author collected all data and was not blinded to participants’ raw scores. However, she was blinded to all derived scores because calculations were not performed until after all data collection was completed. Throughout the study, the tester examined intrarater reliability of muscle strength (measured 30 seconds apart) and spasticity (measured approximately one hour apart) during the same testing session on all participants. Intrarater reliability, expressed as interclass correlation coefficient (ICC), was excellent for all isometric strength measurements (ICC(3,1) = 0.88–0.95) and very good for spasticity measurements (ICC(3,1) = 0.86–0.90). Nine children were videotaped when performing the GMFM to assess its intrarater reliability in this study (ICC(3,1) = 0.97–0.99). Intrarater reliability of walking efficiency measures was not examined because of concerns of fatigue and excessive transportation. However, by following the standardized protocol established by Wiart and Darrah,16 good reliability was expected.
At first, participants’ body weight, lower leg length, and resting heart rate were measured and later used for normalization of strength and walking efficiency data. The tester used the Ashworth Scale (Table 1) to measure spasticity of hamstrings and quadriceps with the child in side lying and the tested leg on top to minimize the influence of tonic reflexes and to eliminate gravitational effects at the knee. The participant was asked to relax the top leg as much as possible throughout the procedure. To measure quadriceps spasticity, the tested hip was put in maximum extension but neutral in other movement planes while the knee was in maximum extension. The tester stood behind the subject with one hand supporting the thigh and the other hand supporting the ankle (Fig. 1). Then, the tester moved the lower leg from maximum extension to maximum flexion in one second. To measure the spasticity of the hamstrings, the tested hip and knee were put in maximum flexion but neutral in other movement planes. The tester stood in front of the subject, with one hand supporting the thigh and the other hand supporting the ankle (Fig. 2), and moved the lower leg from maximum flexion to maximum extension in one second. The same procedures were repeated on both sides of the body. This position and procedure has been shown to produce an 85% agreement in modified Ashworth scale scores for intrarater reliability of the quadriceps.17 In this study, intrarater reliability of methods tested one hour apart was good (ICC(3,1) >0.86). Our intrarater reliability of the Ashworth Scale was slightly higher than previously reported in children with CP (ICC(3,1) ranged from 0.67 to 0.80), which might have resulted from a more homogeneous sample and a shorter test-retest interval.18 For each child, spasticity data consisted of four scores (left and right quadriceps and hamstrings) repeated for a second trial to calculate intrarater reliability. The child’s best (lowest) scores during the first trial were used for data analysis as Blackburn et al. found that higher reliability was found for lower range scores.17
The tester measured maximal voluntary contraction of quadriceps and hamstrings with the Chatillon MSE 100 series digital dynamometer (NexGen Ergonomics Inc., Quebec, Canada). Strength measurements were taken with the knees positioned at 90 degrees, and then at 60 degrees of knee flexion. The 90-degree position was chosen for comparison with other studies.10,12 Sixty degrees was chosen because it was considered to be a position for optimal length–tension relationship to produce maximal force for both quadriceps and hamstrings.10 The order of strength testing started with right quadriceps, then left quadriceps, followed by right hamstrings, and ended with left hamstrings. The measuring sequence was repeated again on each muscle group to obtain second trial measurements to calculate intrarater reliability for each muscle. The order of muscle testing was chosen to minimize muscle fatigue and to allow at least a 30-second rest between two trials for each muscle group. During testing, the participant sat on the edge of a plinth, and his or her back was not supported. An assistant stood behind the participant to assure that the trunk remained in position throughout the testing. The tester performed a standard break test19 to measure quadriceps and hamstring strengths (Figs. 3 and 4). Each contraction lasted about six seconds, and standardized verbal encouragement was given during the contraction. The peak value (N) of two contractions (left and right) of each muscle group at each position was used to calculate the averaged values. The averaged strength values then were normalized by lower leg length and body weight and expressed in Newton-meters per kilogram to be used in data analysis.20,21 Furthermore, dividing normalized quadriceps strength by normalized hamstring strength derived a “Q/H ratio” at the 90-degree and 60-degree positions, which was used as an indicator of muscle balance at the knee. Because we were unsure of the utility of using the strength values measured at the 90-degree versus the 60-degree position, initial data analysis included both positions until a decision could be made to use one rather than the other in further analyses.
Gross motor abilities were evaluated using the GMFM, which is composed of five dimensions: (1) lying and rolling, (2) sitting, (3) crawling and kneeling, (4) standing, and (5) walking, running, and jumping. The authors deemed dimensions D and E as most relevant to this study in terms of functional abilities. Clinically, the GMFM is usually scored in its entirety; therefore, all three scores (dimension D and E, and total score) were used for initial data analysis. Once the utility of using one functional measure derived from the GMFM over the other two could be determined, only one GMFM measure was used in further analyses.
Walking efficiency at both comfortable and fast speeds was measured by calculating the EEI. This index is calculated by obtaining the participant’s gait speed and heart rate (measured by the Polar heart rate monitor; heart rate monitor series A1, Polar Electro Inc., Lake Success, NY) at rest and during a three-minute walking test; EEI = (walking heart rate − resting heart rate)/walking speed.16 Participants walked on an oval track at a self-selected comfortable speed for three minutes under close supervision. Walking distance was recorded using pre-measured intervals (100 cm) on the track, later used to calculate walking speed. Walking heart rate was defined as the mean heart rate during the last 30 seconds of the walking session. After the comfortable speed walking test, there was a break of approximately five minutes to allow the heart rate to return to the resting level. The three-minute walk test was repeated with the instruction of “walk as fast as you can without running.” Self-selected fast walking distance and fast walking heart rate were recorded as described previously. This procedure was the same as described Wiart and Darrah16 and Kramer and MacPhail,12 who found test-retest reliability to be 0.94 and 0.81, respectively. Because we were unsure of the utility of examining walking efficiency measured either at a comfortable or fast walking speed, initial data analysis included both comfortable walking and fast walking EEI values until a decision could be made for using one over the other in further analyses.
Descriptive statistics were performed to describe the sample’s characteristics and performance on the variables of interest. Intrarater reliability of procedures using the Ashworth Scale, hand held dynamometer, and the GMFM were analyzed using ICC(3, 1). For the purpose of this study an ICC greater than 0.85 was considered acceptable. Pearson product moment correlation coefficients (r) were calculated to determine the inter-relationships among all the measures; alpha was set at the level of 0.05. The strength of the relationships was interpreted as follows: 0 to 0.25 indicated little to no relationship; 0.25 to 0.50 indicated a fair degree of relationship; values of 0.50 to 0.75 indicated a moderate relationship; and values greater than 0.75 indicated a strong relationship.15 The authors considered the utility of using (a) the strength values measured either at the 90-degree or 60-degree position; (b) total, dimension D, or dimension E GMFM scores; and (c) comfortable walking or fast walking EEI values, by considering the correlation coefficients among the variables, the intrarater reliability of the measures, and their professional judgment. Once a single variable was deemed most useful in each of the three aforementioned areas, they were used in further data analysis. A stepwise linear regression equation was calculated to determine any combinations of impairment variables that would be superior to any single variable for predicting gross motor function. Similarly, a second regression equation was calculated to determine which impairments best predicted walking efficiency.
Thirty children were recruited to participate in this study. Data from 27 children were used in analysis, and data from three children were excluded: one participant was a statistical outlier, and two had incomplete data. Participants ranged in age from seven to 12 years (mean = nine years) and included both girls (n = 10) and boys (n = 17). Twelve children functioned at Level I, 10 at Level II, and five at Level III the GMFCS. Two children used a walker for ambulation, three used crutches, and the remaining 22 participants used no assistive device. All the children were receiving regular physical therapy treatments either in the school system or health care system. Descriptive statistics for the entire sample’s strength, spasticity, gross motor function, and walking parameters are reported in Table 2.
Tables 3 and 4 show the relationships among the various measures of knee muscle strengths. There were positive and strong relationships among the normalized muscle strengths in the two test positions (90 degrees and 60 degrees), suggesting that measurements taken at these two different joint positions may be assessing the same component of performance. Even though the measure of muscle balance at the knee (Q/H ratio) was calculated from the normalized muscle strength of the two muscle groups, with the exception of hamstring strength at 90 degrees, the relationships between the Q/H ratio and the individual normalized muscle strengths within a given test position ranged from none to moderate. The relationship between muscle balance at 90 degrees and 60 degrees was moderate at best.
Table 5 presents the results of the correlations between spasticity, normalized strength, and muscle balance at the two test positions. The relationships between spasticity and the strength measures ranged from none to moderate. The authors considered the correlation coefficients among these variables, the intrarater reliability of the strength measures, and decided to use the strength measurements at the 90 degree test position for all subsequent analyses.
The correlations between impairments and functional outcome measures are presented in Table 6. With the exception of hamstring spasticity, which had weak to fair relationships with gross motor function, all other impairment measures had very strong relationships with the total GMFM and with the two dimensions. Walking efficiency in the two test conditions had similar fair to moderate relationships with all the impairment measures. The authors considered the correlation coefficients among these variables, the intrarater reliability of the gross motor function measures, and decided to use the GMFM total score and the EEI in the comfortable walking speed test for further analyses.
The stepwise linear regression models for the gross motor function and walking efficiency at a comfortable speed are shown in Table 7. Five independent variables were considered: hamstring spasticity, quadriceps spasticity, normalized quadriceps strength at 90 degrees, normalized hamstring strength at 90 degrees, and muscle balance Q/H ratio at 90 degrees. Normalized hamstring strength contributed 68.7% of variance to the total GMFM scores, and the quadriceps Ashworth added another 12.8%. Together, these two measures explained 81.5% of the variance in the total GMFM. An analysis of the multicollinearity of the total GMFM regression model showed very little correlation between the two independent variables. In the second model, quadriceps spasticity was the only significant factor explaining the variance in the EEI at a comfortable walking speed, explaining 56.7% of the variance.
Generally, we found that children with more spasticity of the knee muscles tended to have less knee muscle strength, lower GMFM scores, and less efficient gait. Children who had more balanced strength around the knee tended to have better gross motor function and walked more efficiently.
The results in Table 5 suggest that children with more spasticity also were weaker. These findings are consistent with results reported by Damiano and colleagues,4 but contrary to Ross and Engsberg, who found no relationship between these two variables.1 In our study, hamstring spasticity was moderately related to both normalized hamstring and quadriceps strength, and quadriceps spasticity was related to both normalized quadriceps and hamstring strength. These correlations were all statistically significant. Thus, we are confident that the strength of the relationships are not caused by type I error.15 Therefore, quadriceps spasticity explains 33% of the variance in quadriceps strength and 40% of the variance in hamstring strength. Likewise, hamstring spasticity explains 18% of the variance in hamstring strength and 13% of the variance in quadriceps strength. The remainder of the variances among the variables must be explained by one or more unmeasured variables, such as muscle flexibility, availability of stored muscle energy, etc. Conflicting results in the literature about the relationship between spasticity and strength may be due in part to these uncontrolled, unmeasured variables.
Coactivation of the antagonist muscle has been reported to contribute to agonist weakness,4,22 which is further supported by our study, having found a negative and modest relationship between antagonist spasticity and agonist strength. Agonist spasticity also correlated fairly and negatively with agonist strength, a finding similar to that of Damiano and colleagues,4 suggesting that a spastic muscle is also a weak muscle. This relationship may be explained by biomechanical changes in chronic spastic muscles. Researchers found that spastic muscle cells had significantly shorter sacromere length and higher stiffness than non-spastic muscle cells.23 According to the tension-length relationship theory, a shortened sacromere is in a less optimal position to produce active force because less actin-myosin linkages are in reserve. Rose and associates found that most children with CP had type I fiber predominance.24 The authors hypothesized that these changes were due to chronic, prolonged low frequency motor unit firing rates.25 Type I muscle fibers are slow twitch in nature, which produces less instant force and takes longer time to reach peak tension. Because spastic muscles contain a greater ratio of type I fibers, it is not surprising to see less force production in the six-second isometric contraction performed in this study. Another simple explanation of this relationship could be that children with more spasticity tend to be less active, which, over time, results in muscle atrophy and weakness.
This study also showed that quadriceps spasticity had a stronger relationship with muscle strength than hamstring spasticity, a result similar to what Damiano et al. found.4 These findings differ from the clinical expectation that both quadriceps and hamstring spasticity would be equally important in weakening muscle force production.1 More evidence is needed to draw a conclusion about this relationship.
Despite the relationships found between spasticity and strength, it does not mean that elimination of spasticity would be accompanied by increased strength or that strengthening muscles would reduce spasticity. The amount of variance that could be explained is small, ranging from 16% to 30%. Voluntary muscle force production is accounted for by many factors, such as recruitment rate of motor neurons, muscle size, elasticity of muscle fibers, spasticity,22,26 and motivation. Therefore, treating spasticity or strength alone may not cause much improvement in the other.
The idea of examining muscle strength in terms of muscle balance around the joint rather than individual muscles is a relatively new idea. Even though the measure of muscle balance at the knee (Q/H ratio) was calculated from the normalized muscle strength of the two muscle groups, with the exception of hamstring strength at 90 degrees, there were only moderate-to-no relationships between the Q/H ratio and the individual normalized muscle strengths within a given test position (Table 4), implying that muscle balance is a different construct than muscle strength. In addition, muscle balance at 90 degrees explains only 31% of the variance in muscle balance at 60 degrees (Table 3), suggesting that muscle balance may be different at various points in the range of motion. In this sample, all children (n = 27) had Q/H ratios at 90 degrees that were greater than one, while at 60 degrees only 14 children had Q/H ratios greater than one. Spasticity of the quadriceps and hamstrings had similar correlations (positive and moderate) with muscle balance at 90 degrees of knee flexion, but only spasticity of hamstrings was correlated with Q/H at 60 degrees (Table 5). When the knee joint was positioned at 60 degrees of flexion, the hamstrings were stretched, a situation in which spasticity is more likely to be evoked. Based on this finding, it would be more appropriate to choose 90 degrees of knee flexion when one wants to measure muscle balance without the influence of spasticity.
It appeared that children with stronger knee muscles, muscle balance, and less spasticity in their quadriceps tended to have better performance on the GMFM and walked more efficiently as measured by the EEI. This finding in part matched the results reported by Damiano and associates, who also found that knee muscle isokinetic strength was highly correlated with the GMFM in children with CP.4 Kramer and MacPhail also found a similar-but-weaker relationship between knee extensor strength and the GMFM, but no significant relationships between knee flexor strength and the GMFM.12 In contrast, Berry and colleagues showed that only knee flexor isometric strength correlated with the GMFM scores.27 Tuzson et al. observed that children with more spasticity have worse performance in the GMFM.28 Other researchers4,26 also found quadriceps spasticity to have a stronger relationship with the GMFM and walking parameters than hamstring spasticity. Although spasticity is considered a limiting factor of gross motor function, some scholars believe that some children use spasticity in extensor muscles to maintain an anti-gravity position.6,26 Regardless, muscle spasticity may cause the high energy consumption observed in children with CP. Dahlback and Norlin reported that children with CP became exhausted while walking at 50% to 60% of maximal oxygen uptake, possibly due to local muscle factors such as spasticity.29 The discrepancy between our findings and the findings of the other researchers12,27 may be due to differences in the sample (eg, age, diagnosis, functional level) and differences in measuring methods (eg, isokinetic versus isometric, angular acceleration versus Ashworth Scale). The same could be said for walking efficiency. Many factors may affect heart rate changes during gait, such as mood, fitness level, and medication. All these issues could result in a weaker relationship observed between the impairment measures and the EEI.
Interestingly, we found that hamstring spasticity explained only 12% to 25% of the variance in any of the outcome measures, while the other impairment measures explained up to 75% of the variance in gross motor function and up to 58% of the variance in walking efficiency (Table 6). This finding is not consistent with clinical expectations and experience in which hamstring spasticity is considered to have more influence on function than quadriceps spasticity.28,30
These results imply important linkages between impairments and functional outcome measures. The correlation matrixes discussed thus far only explore simple relationships between two variables at a time. Before we can suggest that strengthening knee muscles, seeking muscle balance, or reducing spasticity may result in improvements in gross motor performance or walking efficiency in this population, or suggest experimental studies needed to establish cause-effect between variables, we must evaluate the relative contribution of the impairment measures as a whole to each functional outcome measure in our study with stepwise linear regression. Therefore, we considered five independent variables for the regression equation. The implications for physical therapy practice are that interventions in this population should aim to improve hamstring strength and decrease spasticity in the quadriceps. These findings provide some evidence to support the clinical observation that both strength and spasticity have a role in determining functional performance, and provide clarification as to which muscle groups may be targets for interventions. Readers should be cautioned that these results provide only an overview of the possible influencing variables. Longitudinal research is needed to determine whether hamstring strength and quadriceps spasticity can be manipulated to improve performance in children with CP, and experimental designs are necessary to determine effective interventions.
There were several limitations inherent in this study. In terms of generalizabiltiy, only school-aged children with spastic diplegia were recruited and all had CP that was relatively mild to moderate in severity. Second, other joints (hip, ankle, and trunk) and other factors (cognition, balance) should be considered. Third, the reliability of the Ashworth Scale remains problematic. Fourth, the same tester recorded scores for all measurements and was not blinded to the dependent and independent variables directly measured. Non-randomized measurement order may have introduced bias; however, the order of measurements was necessary to ensure adequate rest periods between tasks and to ensure that strength measures did not create fatigue before other performance measures. Initial data analyses narrowed the field of potentially important variables from eight to five independent variables and from five to two dependent variables. Finally, Zar31 encourages investigators to include at least five participants (ideally 10 to 15) for every independent variable included in regression analyses. With five independent variables, we met the minimum requirement of at least 25 participants, but 50 to 75 participants would have been ideal.
Normalized hamstring strength at 90 degrees explained much of the variance in the total GMFM in our group of children with mild to moderate spastic CP. Quadriceps spasticity as measured by the Ashworth scale also explained some of this variance. For comfortable walking EEI, only quadriceps spasticity significantly contributed to the score. Although this study did not demonstrate cause and effect relationships between impairments and functional limitations, it provides information to help clinicians in decision making and setting treatment goals. Strategies to improve functional outcomes in this population should address both impairments. Further research is necessary to develop concrete strategies to increase function by focusing treatment on hamstrings strengthening and reducing quadriceps spasticity in patients with CP.
1. Ross SA, Engsberg JR. Relation between spasticity and strength in individuals with spastic diplegic cerebral palsy. Dev Med Child Neurol.
2. Landau WM, and Hunt CC. Dorsal rhizotomy, a treatment of unproven efficacy. J Child Neurol.
3. Bobath B. Abnormal Postural Reflex Activity Caused by Brain Lesion
. Rockville, Md: Aspen Publishing, 1985.
4. Damiano DL, et al. Spasticity versus strength in cerebral palsy: relationships among involuntary resistance, voluntary torque, and motor function. Eur J Neurol
5. Maruishi M, et al. Cerebral palsy in adults: Independent effects of muscle strength and muscle tone. Arch Phys Med Rehabil.
6. Giuliani CA. Dorsal rhizotomy for children with cerebral palsy: support for concepts of motor control. Phys Ther.
7. Abel MF, et al. Relationships among musculoskeletal impairments and functional health status in ambulatory cerebral palsy. J Pediatr Orthop.
8. Damiano DL, Abel MF. Functional outcomes of strength training in spastic cerebral palsy. Arch Phys Med Rehabil.
9. Damiano DL, Kelly LE, Vaughn CL. Effects of quadriceps femoris muscle strengthening on crouch gait in children with spastic diplegia
. Phys Ther.
1995;75:658–667; discussion 668–671.
10. Damiano DL, Vaughan CL, Abel MF. Muscle response to heavy resistance exercise in children with spastic cerebral palsy. Dev Med Child Neurol.
11. MacPhail HE, Kramer JF. Effect of isokinetic strength-training on functional ability and walking efficiency in adolescents with cerebral palsy. Dev Med Child Neurol.
12. Kramer J, MacPhail H. Relationships among measures of walking efficiency, gross motor ability, and isokinetic strength in adolescents with cerebral palsy. Pediatr Phys Ther.
13. Damiano DL, et al. Muscle force production and functional performance in spastic cerebral palsy: relationship of cocontraction. Arch Phys Med Rehabil.
14. Morton JF, Brownlee M, McFadyen AK. The effects of progressive resistance training for children with cerebral palsy. Clin Rehabil.
15. Portney L and Watkins M. Correlation, ed. Foundation of Clinical Research: Applications to Practice.
Stanford: Appleton & Lange; 2000:439–455.
16. Wiart L, Darrah J. Test-retest reliability of the energy expenditure index in adolescents with cerebral palsy. Dev Med Child Neurol.
17. Blackburn M, van Vliet P, Mockett SP. Reliability of measurements obtained with the modified Ashworth scale in the lower extremities of people with stroke. Phys Ther.
18. Clopton N, Dutton J, Featherston T, Grigsby A, Mobley J, Melvin J. Interrater and intrarater reliability of the Modified Ashworth Scale in children with hypertonia. Pediatr Phys Ther.
19. Bohannon RW. Test-retest reliability of hand-held dynamometry during a single session of strength assessment. Phys Ther.
20. Damiano DL, Dodd K, Taylor NF. Should we be testing and training muscle strength in cerebral palsy? Dev Med Child Neurol.
21. Backman E, et al. Isometric muscle strength and muscular endurance in normal persons aged between 17 and 70 years. Scand J Rehabil Med.
22. Elder GC, et al. Contributing factors to muscle weakness in children with cerebral palsy. Dev Med Child Neurol.
23. Friden J, Lieber RL. Spastic muscle cells are shorter and stiffer than normal cells. Muscle Nerve.
24. Rose J, Haskell WL, Gamble JG, Hamilton RL, Brow DA, Rinsky L. Muscle pathology and clinical measures of disability in children with cerebral palsy. J Orthop Res.
25. Rose J, McGill KC. The motor unit in cerebral palsy. Dev Med Child Neurol.
26. Damiano DL, Quinlivan JM, Owen BF, Payne P, Nelson KC, Abel MF. What does the Ashworth scale really measure and are instrumented measures more valid and precise? Dev Med Child Neurol.
27. Berry ET, Giuliani CA, Damiano DL. Intrasession and intersession reliability of handheld dynamometry in children with cerebral palsy. Pediatr Phys Ther.
28. Tuzson AE, Granata KP, Abel MF. Spastic velocity threshold constrains functional performance in cerebral palsy. Arch Phys Med Rehabil.
29. Dahlback GO, Norlin R. The effect of corrective surgery on energy expenditure during ambulation in children with cerebral palsy. Eur J Appl Physiol Occup Physiol.
30. Papadonikolakis AS, Vekris MD, Korompilias AV, Kostas JP, Ristanis SE, Soucacos PN. Botulinum A toxin for treatment of lower limb spasticity in cerebral palsy: gait analysis in 49 patients. Acta Orthop Scand.
31. Zar J. Biostastical Analysis
. 3rd ed. Upper Saddle River, NJ: Prentice Hall; 1996.