Prospective Longitudinal Study of Gross Motor Function in Children With Cerebral Palsy

Participants 

Participants were recruited for a 2-year longitudinal study encompassing 3 annual measurements. Rehabilitation centers, special schools for physically and mentally disabled children, and outpatient clinics of departments of rehabilitation medicine in the northwest region of the Netherlands identified 244 children 9, 11, and 13 years of age with CP. These children and their parents received a letter with information about the study and a request that they participate. Of this group, 110 children and their parents returned the informed consent form and participated in the study. Reasons for nonparticipation could be determined in 20 cases: language problems (n=4), moved without a forwarding address (n=2), participation in other research (n=2), and family stress (n=12).

All the regional medical ethics committees approved the study protocol. This research was performed as part of the Pediatric Rehabilitation Research in the Netherlands (PERRIN) program (http://www.perrin.nl), which is a longitudinal study of children with CP.

Data Collection and Outcome Measure 

The children and their parents visited the Department of Rehabilitation Medicine at the VU University Medical Center in Amsterdam each year. During the first visit, 2 researchers asked standardized questions about diagnosis, epilepsy, and type of school; classified the children according to the GMFCS; carried out the physical examination; and assessed gross motor function with the Gross Motor Function Measure (GMFM). All measurements were repeated after 1 and 2 years. Both researchers were certified to administer the GMFM.

The GMFM is a standardized observational instrument that measures gross motor function in children with CP, based on their performance of 88 gross motor tasks upon instruction in a specific test situation.8, 9 The GMFM was analyzed with the Gross Motor Ability Estimator (GMAE) computer scoring program to obtain the GMFM-66 score.8, 10 The GMAE rescales the child’s abilities from an ordinal scale (GMFM-88) to an interval scale (GMFM-66), varying from 0 (poor motor function) to 100 (normal motor function for 5-year-old children).

Determinants 

The severity of the CP was classified according to the GMFCS, a 5-level classification system based on functional limitations, the need for assistive devices and, to a lesser extent, quality of movement.11, 12

We used the levels of impairments at the first measurement as potential determinants. These levels were limb distribution, selective motor control, muscle strength, limitations in hip and knee extension, spasticity, muscle tone, epilepsy, and type of education. Age was included as a personal characteristic.

Limb distribution was subdivided into 3 categories: hemiplegia (unilateral involvement), diplegia, and tetraplegia (both bilateral involvement). Tetraplegia was defined as the arms being affected as severely or more severely than the legs; diplegia was defined as the legs being more severely affected than the arms. To measure the selective motor control, the children were asked to extend the knee and dorsiflex the ankle of each leg in a short-sitting position without the support of the feet. Possible scores were: 0 (no selective, only synergistic movement), 1 (diminished selective movement [the first range of movement selective and later on, during the movement, no selective movement]), and 2 (full selective movement during extension of the knee and dorsiflexion of the ankle). The scores for the 2 sides together produced a total score varying from 0 to 8. The total scores were then subdivided into 3 categories: poor selective motor control (total scores 0, 1, or 2); moderate selective motor control (total scores 3, 4, or 5); and good selective motor control (total scores 6, 7, or 8).

To define muscle strength, the children were asked to stretch out from squat position 8 times (support for balance was allowed). They were subdivided into 3 categories: good strength if they could squat 8 or more times, moderate strength if they could squat fewer than 8 times or performed a part of the motion 8 times, and poor strength if they were not able to squat at all.

The range of motion (ROM) of hip and knee extension was measured, both in a supine position.13 We performed the Thomas test14 to detect the limitations in hip extension. To indicate the degree of limitations in ROM, the ROM scores were transformed according to the Spinal Alignment and Range of Motion Measure (SAROMM), to discriminate between no (1), mild (2), moderate (3), and severe (4) limitations in ROM (see appendix 1).15 To indicate the total extent of the limitations in extension in hips and knees we calculated an overall score as the mean of the SAROMM scores of the knee and hip of the right and left legs.

Spasticity was measured as an increase in muscle tone resulting in a catch during fast velocity stretching of the muscles, using standardized measurement procedures.16 Spasticity in the hamstrings, hip adductors, and the gastrocnemius muscle was measured in a supine position, and spasticity in the rectus femoris muscle was measured in a prone position. The scores were as follows: 0 if no spasticity was found in a muscle group, 1 if spasticity was found on one side, and 2 if the muscle group was found to be spastic on both sides. We calculated an overall score to indicate the total extent of the spasticity as the mean of the spasticity in the 4 muscle groups.

The muscle tone was measured as resistance in slow stretching during the ROM measurements and was defined as normal or abnormal if more than 50% of the muscles were hypotonic or hypertonic.

Children with more than 1 seizure during the previous 2 years were defined as having repeated seizures.

Type of education was based on the type of school: children with a “regular” education were those in a regular school or in a school providing education for physically disabled children, whereas children with “special” education were those enrolled in special schools for children with cognitive impairment (with or without physical disabilities), or in special day-care centers for severely handicapped children.

Finally, the children were subdivided into 3 age groups: children who were 9, 11, or 13 years of age at the first measurement.

Statistical Analyses 

We used random coefficient analyses, also known as multilevel analysis (MlwiN17,a), to analyze the changes in the GMFM over time and its determinants. This analysis method considers the dependency of repeated measures within the same person by allowing the regression coefficients to differ between subjects. In addition, the number of observations per person may vary (ie, subjects with missing values can be analyzed).18 The data were defined as follows: level 2 as patient and level 1 as measurement occasion. The GMFCS, limb distribution, selective motor control, muscle strength, ROM, spasticity, and age were analyzed as categorical variables using dummy variables.18 Type of education, epilepsy, and muscle tone were analyzed as dichotomous variables. Time was expressed as the measurement occasion in years.

Time, the GMFCS, and the interaction term GFMCS by time were added to the model to analyze the course of gross motor function, as described above. To analyze the relation between the course of gross motor function with impairments and age, each of the determinants and the interaction terms with time were entered into the model separately, corrected for GMFCS level. Subsequently, we made a multivariable model using a forward stepwise procedure, beginning with the most significant determinant. First, each determinant was added to the model as a single factor and then removed if not significant (P>.05). After this, the interaction action terms were added to the model one by one and then also excluded if not significant (P>.05). We used the chi-square test to determine whether determinants were significant (P<.05).

To determine the required sample size, we did a power analyses, using an α of .05, power of 0.8, smallest meaningful difference over a year of 2 points change on the GMFM, and a variance of 54. For the dependency of the repeated measures we assumed an intraclass correlation coefficient of .90. This resulted in a sample size of at least 11 subjects for each GMFCS group.

The Course of Gross Motor Function 

Overall, gross motor function remained stable over the 2 years. Table 2 shows the raw scores of the GMFM according to GMFCS levels. There was a significant difference between all 5 of the GMFCS levels, but there was no difference in the course of gross motor function over 2 years (table 3).

Table 2. Raw Data of the GMFM-66 per Year According to the GMFCS Level

	GMFM-66	Baseline		Time 1		Time 2
		n	Mean ± SD	n	Mean ± SD	n	Mean ± SD
	GMFCS I	50	89.84±7.22	49	90.86±7.91	47	91.70±7.83
	GMFCS II	16	76.01±6.85	14	76.88±7.17	13	77.24±9.17
	GMFCS III	13	58.91±7.89	10	60.91±7.64	10	59.81±7.41
	GMFCS IV	13	41.08±6.93	12	40.26±8.05	11	39.10±9.60
	GMFCS V	18	23.26±9.51	18	23.38±8.84	18	23.25±8.44

Abbreviations: Time 1, measurement after 1 year; Time 2, measurement after 2 years.

Table 3. Results of the Random Coefficient Analyses for the Course of Motor Functioning in Relation to the GMFCS

	GMFM-66		Regression Coefficient ± SE	Chi-Square (df)	P
	Intercept		89.875±1.072)
	Time		0.816±0.323)
	GMFCS	I	0 (ref)
		II	−13.980±2.179
		III	−30.867±2.363
		IV	−48.745±2.361
		V	−66.574±2.084
	GMFCS by time	I by time	0 (ref)	7.415 (4)	.116
		II by time	−0.251±0.682
		III by time	−1.484±0.775
		IV by time	−1.614±0.736
		V by time	−0.818±0.619

NOTE. GMFCS level I is set to zero because GMFCS levels II to V are analyzed versus GMFCS I (reference category). Abbreviations: ref, reference category; SE, standard error.

Determinants of the Course of Gross Motor Function 

Results of the random coefficient analyses, showing the relationship between the impairments and the course of the GMFM, corrected for GMFCS level, are shown in table 4. The regression coefficients of the interaction terms of the determinants with time indicate the change per year in GMFM item scores for each category, compared with the reference category (set to zero, see table 4). There was a significant difference (ie, significant interaction with time) in the course of the GMFM over 2 years for the following determinants: selective motor control, limb distribution, muscle strength, ROM, spasticity of the hamstrings, and type of education.

Table 4. Results of the Random Coefficient Analyses, Showing the Relationship Between Impairments and the Course of GMFM, Corrected for GMFCS Level

	GMFM-66	Time (y)		Determinants⁎		Determinants by Time⁎
		Intercept ± SE	Regression Coeff ± SE		Regression Coeff ± SE	Regression Coeff ± SE	Chi-Square (df)	P†
	Model 1: SMC	90.9±1.1	1.02±0.32	Good SMC	0 (ref)	0 (ref)	10.860(2)	.004
				Moderate SMC	−5.884±2.161	−0.704±0.606
				Poor SMC	−14.462±4.027	−1.657±0.490
	Model 2: Limb distribution	90.0±1.2	0.96±0.36	Hemiplegia	0 (ref)	0 (ref)	6.566(2)	.038
				Diplegia	−0.322±2.119	−0.804±0.501
				Tetraplegia	−3.743±3.373	−1.559±0.615
	Model 3: Strength	90.0±1.1	0.84±0.30	Good strength	0 (ref)	0 (ref)	8.272(2)	.016
				Moderate strength	−4.931±2.967	−0.428±0.735
				Poor strength	−4.952±4.019	−1.428±0.488
	Model 4: ROM	89.8±1.2	1.05±0.32	Full ROM	0 (ref)	0 (ref)	11.322(3)	.010
				Mild limitations ROM	0.499±1.957	−1.494±0.542
				Moderate limitations ROM	−0.027±3.049	−1.840±0.657
				Severe limitations ROM	2.058±3.814	−0.504±0.878
	Model 5: Tone	90.6±1.1	0.60±0.28	Normal tone	0 (ref)	0 (ref)	3.526(1)	.060
				Abnormal tone	−4.966±2.019	−0.924±0.488
	Model 6: Spasticity	91.5±1.4	0.55±0.40	Spas total_0 side	0 (ref)	0 (ref)	2.926(2)	.232
				Spas total_1 side	−2.751±1.747	0.147±0.573
				Spas total_2 sides	−1.786±1.791	−0.758±0.558
	Model 7: Hamstring spasticity	91.6±1.3	0.65±0.31	Spas ham_0 side	0 (ref)	0 (ref)	9.015(2)	.011
				Spas ham_1 side	−3.704±1.933	0.336±0.583
				Spas ham_2 sides	−2.249±1.730	−1.354±0.510
	Model 8 of education	90.1±1.1	0.67±0.26	Regular education	0 (ref)	0 (ref)	7.375(1)	.007
				Special education	−3.156±1.976	−1.401±0.506
	Model 9: Repeated seizures	90.2±1.1	0.43±0.24	No	0 (ref)	0 (ref)	2.526(1)	.112
				Yes	−1.368±2.379	−1.197±0.745
	Model 10: Age Group (y)	88.1±1.5	1.17±0.43	9	0 (ref)	0 (ref)	5.674(2)	.058
				11	2.405±1.809	−1.094±0.580
				13	3.889±1.793	−1.288±0.558

NOTE. Significant interactions between impairments and time (determinant by time) indicate an association with the course of GMFM. The SMC, limb distribution, strength, ROM, age group, and spasticity are analyzed as dummy variables. Abbreviations: coeff, coefficient; SMC, selective motor control; Spas ham, spasticity of the hamstrings; Spas total, average score of total spasticity.

⁎ Corrected for GMFCS level. † P of the model with the interaction term.

Children with poor selective motor control showed a 1.7 points greater decrease in GMFM score per year than children with good selective motor control (reference category) (see table 4). Children with tetraplegia showed a 1.6 points greater decrease per year than children with hemiplegia. Even so, for the determinants muscle strength, spasticity of the hamstrings, and type of education, changes in GMFM score were less favorable in the most severely affected category (ie, children with poor strength, spasticity in the hamstrings on both sides and having special education) compared with the least affected category. The ROM analyses showed that children with mild or moderate ROM limitations had a 1.5 and 1.8 points greater decrease in GMFM score per year, compared with children with a full ROM. The course of the motor function of children with severe ROM limitations, however, did not differ from that of children with full ROM.

Table 5 shows the results of the multivariable analysis. GMFCS, selective motor control, tone, and selective motor control by time were included in the final model, indicating that selective motor control is the most important determinant of the course of gross motor function.

Table 5. Results of the Random Coefficient Analyses, Showing the Multivariable Model

	GMFM-66		Regression coefficient ± SE	Chi Square (df)	P
	Intercept		91.264±1.093
	Time		1.019±0.323
	GMFCS	I	0 (ref)	99.006(4)	.000
		II	−12.028±2.127
		III	−20.440±3.348
		IV	−36.625±4.477
		V	−52.803±4.340
	Tone		−4.152±1.970	4.442(1)	.035
	SMC	Good SMC	0 (ref)
		Moderate SMC	−5.475±2.164
		Poor SMC	−11.073±4.265
	SMC by time	Good SMC by time	0 (ref)	10.851(2)	.004
		Moderate SMC by time	0.698±0.607
		Poor SMC by time	1.662±0.492

NOTE. The P value for main effects of selective motor control is not reported. Because the interaction for selective motor control and time is significant, the significance of these main effects is not relevant.

Study Limitations 

Apart from the study’s small sample size, the conclusions that can be drawn from it may be limited because 11 children and their parents did not complete the 3 measurements. Although the dropouts were equally divided over the GMFCS levels, it is not known whether they influenced the results, because reasons for dropping out were variable (repeated no-shows at appointments [n=5], serious psychologic problems of the child [n=1], relocation without leaving a new address or phone number [n=2], parents with serious health problems [n=2], refusal by the child [n=1]).