Patients with unilateral cerebral palsy (CP) often have impaired movement coordination, reduced between-limb synchronization, and less weight bearing on the affected side, which can affect the maintenance of an upright weight-bearing position and gait. This study evaluated whether the different postural patterns of children with unilateral CP could be statistically recognized using cluster analysis. Forty-five outpatients with unilateral CP (mean age, 9 years and 5 months) and 51 able-bodied children with mild scoliosis (mean age, 9 years and 2 months) were included. One observer performed moiré topography (MT) examinations using a CQ Electronic System (Poland) device. A weight distribution analysis on the base of support (BOS) between the body sides was performed simultaneously. A force plate dynamographic platform (PDM), ZEBRIS (Germany), with FootPrint software was used for these measurements. Cluster analysis revealed three groups: Cluster 1 (
In children with cerebral palsy (CP), atypical body posture patterns (PPs) are observed [
A symmetric weight-bearing distribution between the legs during quiet standing provides optimal biomechanical stability, whereas weight shifts prevent the progressive build-up of fatigue in the legs [
Moiré topography (MT) is an imaging method for the body surface and is highly sensitive in detecting asymmetry [
Clustering analysis attempts to maximally separate subpopulations by exclusively assigning an instance to only one class. Colloquially, clustering attempts to identify groups of instances, so that the instances within a group are similar to each other while being dissimilar to those instances in all other groups. The most common approach is to use hierarchical cluster analysis and Ward’s method. K-means clustering is very different from the above, which are applied when there is no prior knowledge of how many clusters there may be or what they are characterized by. K-means clustering is used when hypotheses concerning the number of clusters in the cases or variables have already been made. K-means cluster analysis is thus a tool of discovery used to reveal associations and structure within data that, although not previously evident, are sensible and useful when discovered [
In our previous study, which presented a descriptive analysis of abnormal postural patterns in children with hemiplegic CP [
There are few studies on asymmetric weight bearing during standing [
The research protocol was approved by the Silesian Medical University Bioethics Committee in Katowice, Poland. The parents/guardians provided signed informed consent prior to the subjects’ enrollment in the study.
The study participants were 45 children (17 girls and 28 boys, mean age 9 years and 5 months, range 7 years and 4 months to 12 years and 2 months (SD = 2.11)) with unilateral CP. There were 29 patients with right-sided deficits and 16 patients with left-sided deficits. All participants were outpatients (75.5% Level I and 24.5% Level II by the Gross Motor Function Classification System) at local pediatric rehabilitation centers.
In the reference group, there were 51 able-bodied children with mild scoliosis (27 girls and 24 boys; range of lateral curvature, 11°–25°, mean, 18°; mean age, 9 years and 2 months, range, 7 years and 5 months to 12 years and 3 months (SD = 1,99)). All controls were outpatients at a local Center for Corrective Gymnastics. They were diagnosed by a physician as having idiopathic scoliosis. Twenty-two patients had recently undergone radiographic examination of the spine.
All subjects met the following criteria: (1) older than 7 years of age, (2) able to follow verbal directions, (3) scoliosis (angle of vertebral lateral curvature <20°), and (4) no previous surgical procedures. Additional criteria for subjects with CP were as follows: (1) the diagnosis of spastic hemiplegia, (2) the ability to stand without assistance, (3) not taking any pharmacological agents at the time of the study, and (4) no spasticity management 6 months before the evaluation.
The exclusion criteria were previous orthopedic surgery, severe asymmetrical fixed deformity or scoliosis (angle of vertebral lateral curvature >20°), and dislocation of the hip. Statistical analysis confirmed that the patient demographic characteristics were similar in both groups.
For the MT examination, it was necessary to uncover the entire surface of the back and to mark some anatomical landmarks. These landmarks were the spinous process of C7 (2), spinous process of S1 (8), acromial angle of the shoulders (AAOS) (0, 4), superior angle of the scapula (SAOS) (1, 3), inferior angle of the scapula (IAOS) (5, 6), and the posterior superior iliac spine (PSIS) (7, 9), as suggested by the Society on Scoliosis Orthopedic and Rehabilitation Treatment (SOSORT) [
Surface topography parameter settings.
During the examination, the light was turned off, and the child stood quietly with his/her eyes open. The projection angle was 90°, which meant that the camera was placed perpendicularly to the measured surface. The 40 ms images of the back were captured with a CCD camera. An image recording sequence lasted from 5 to 15 seconds. The image most characteristic of the child was chosen for further analysis.
In the literature, many indices are computed in each of the three planes. The following indices were chosen (Figure
Indices measured on the coronal plane are as follows. Spinous process line (SP): the angle of inclination contained between two adjacent lines, a line situated within the sagittal plane and a line of spinous processes from C7 through S1 (Figure Shoulder line (SHL)*: bilateral SAOS (Figure Pelvic line (PL)*: bilateral PSIS (Figure Angle of the vertebral lateral curvature (ALC).
*The angle of inclination was contained between two adjacent lines: a line situated within the horizontal plane and a line connecting the SAOS (Figure
The coronal plane is the major plane for measuring back deformity because it is related to the Cobb angle definition (Figure
Indices measured on the transverse plane are as follows.
The angle of rotation was the major index used for the reference to this plane. Angle of trunk rotation (ATR)**. Angle of shoulder rotation (SHR)**. Angle of pelvic rotation (PR)**.
**The angle of surface rotation (
Indices measured on the sagittal plane are as follows.
These indices refer to the location and the magnitude of the maximum kyphosis and lordosis. The magnitude of the maximum kyphosis ( The magnitude of the maximum lordosis (
The MT examination was performed by one observer using a CQ Electronic System (Poland).
An analysis of the weight distribution between the right and left (in Ref) and between affected and unaffected (in SH) body sides was conducted simultaneously with an MT examination. A force plate PDM, ZEBRIS (Germany), with FootPrint software was applied for these types of PMs. Each measurement was recorded three times (3 trials, each lasted for 30 seconds with a 30-second pause between trials), and the most typical measurement of each trial was chosen as the mean weight value for the calculation of weight distribution on the right and left body sides in the reference subjects and on the unaffected/affected body sides in children with hemiplegia for further analysis.
Two experienced physical therapists selected both the moiré photographs and body weight distribution measurement. The image that was most characteristic of the child was chosen for further analysis. When the two experts agreed, the arithmetic mean of their assessments was recorded. When their assessments differed, the senior author (M. Domagalska-Szopa) chose the image that was analyzed. The accuracy of their evaluations was then analyzed.
Based on the index of asymmetry (IA) of weight distribution on the unaffected/affected body sides (>40%/60%), the hemiplegic children were divided into four subgroups (four postural patterns) based on the above criteria: LL—left side hemiplegic and the tendency to overload the affected body side ( RR—right side hemiplegic and the tendency to overload the affected body side ( LR—left side hemiplegic and the tendency to overload the unaffected body side ( RL—right side hemiplegic and the tendency to overload the unaffected body side ( NL—the tendency to overload the left body side ( NR—the tendency to overload the right body side (
Based on the same criteria, the children with scoliosis were divided into two subgroups:
The IA of weight distribution between the right and left body sides was calculated for the controls. The standard deviation was used as a criterion to define the asymmetry of weight distribution on the affected/unaffected body sides in children with SH (IA > 9.83%) to create four SH subgroups (LL, RR, RL, and LR) and two control subgroups (NR and NL).
Intraclass correlation coefficient (ICC) with 95% confidence interval was used to measure the overall intraobserver and interobserver agreement. Interobserver agreement was calculated separately for each of the MT and PT parameters, based on two examinations performed by the same two researchers in each group (SH and controls) of 10 subjects (40 examinations in total). Interobserver agreement was calculated (for the same subjects) for two of the reviewers. For the analysis, mean ICC values of 0.80 and above reflected excellent reliability, those between 0.70 and 0.79 indicated good reliability, and those below 0.70 reflected poor to moderate reliability. Because of the high dimensionality of the postural analysis data and the correlations between the parameters, a data reduction technique (specifically, factor analysis with six factors extracted) was used as an input for nonhierarchical
Using a data reduction technique, five grouping variables were extracted: SP, PL, SHL, ALC, and
Parameter descriptions.
MT parameter | Cluster | Mean |
|
Std deviation | Minimum | Maximum |
---|---|---|---|---|---|---|
SP (°) | 1 | 1.5 | 71 | 4.0 | −4.2 | 13.4 |
2 | −3.4 | 8 | 5.3 | −11.9 | 3.6 | |
3 | −5.4 | 17 | 5.0 | −11.3 | 3.5 | |
Total | −0.1 | 96 | 5.1 | −11.9 | 13.4 | |
| ||||||
PL (°) | 1 | −1.7 | 71 | 5.1 | −13.9 | 10.6 |
2 | 7.6 | 8 | 3.9 | 2.1 | 13.1 | |
3 | 10.2 | 17 | 3.3 | 2.4 | 14.3 | |
Total | 1.2 | 96 | 6.8 | −13.9 | 14.3 | |
| ||||||
SHR (°) | 1 | 1.1 | 71 | 9.1 | −20.9 | 18.0 |
2 | 1.3 | 8 | 11.8 | −16.6 | 15.9 | |
3 | −11.6 | 17 | 10.9 | −31.0 | 2.2 | |
Total | −1.1 | 96 | 10.8 | −31.0 | 18.0 | |
| ||||||
|
1 | −9.5 | 71 | 5.9 | −17.6 | 11.2 |
2 | 8.9 | 8 | 8.1 | −1.5 | 18.5 | |
3 | 7.9 | 17 | 4.3 | 2.8 | 15.1 | |
Total | −1.9 | 96 | 8.4 | −17.6 | 18.5 | |
| ||||||
ALC (°) | 1N | 162.9 | 71 | 4.4 | 168.4 | 156.0 |
2PP | 166.4 | 8 | 3.2 | 180.0 | 170.5 | |
3AP | 176.2 | 17 | 3.7 | 180.0 | 164.5 | |
Total | 176.6 | 96 | 5.5 | 180.0 | 156.0 |
MT: moiré topography; SP: spinous process line; PL: pelvic line; SHR: angle of shoulder rotation;
Results of analysis of variance (ANOVA). Differences between the means of various clusters for MT parameters.
MT parameter | Groups | Sum of squares | df | Mean square |
|
|
---|---|---|---|---|---|---|
SP | Between | 743.90 | 2 | 18.61 | 19.99 | 0.00000 |
Within | 371.95 | 93 | ||||
Total | 1730.62 | 95 | ||||
| ||||||
PL | Between | 2318.34 | 2 | 22.80 | 50.84 | 0.00000 |
Within | 1159.17 | 93 | ||||
Total | 2120.50 | 95 | ||||
| ||||||
SHR | Between | 2270.67 | 2 | 93.64 | 12.12 | 0.00002 |
Within | 1135.33 | 93 | ||||
Total | 8708.94 | 95 | ||||
| ||||||
|
Between | 3500.70 | 2 | 34.56 | 50.64 | 0.00000 |
Within | 1750.30 | 93 | ||||
Total | 3214.34 | 95 | ||||
| ||||||
ALC | Between | 1684.60 | 2 | 13.34 | 63.15 | 0.00000 |
Within | 842.30 | 93 | ||||
Total | 1240.44 | 95 |
MT: moiré topography; SP: spinous process line; PL: pelvic line; SHR: angle of shoulder rotation;
Tukey’s posthoc test revealed that five of the MT parameters (excluding ALC) reliably differentiated Cluster 1 and both Clusters 2 and 3 through their cluster means. Three of the MT parameters (PL, SHR, and ALC) demonstrated significant differentiation between Clusters 2 and 3 (Table
Nonhierarchical
Subgroup | Cluster 1 |
Cluster 2 |
Cluster 3 |
Total | |
---|---|---|---|---|---|
NR | ( |
22 | 1 | 0 | 23 |
(%) | 30.99 | 12.50 | 0.00 | 23.96 | |
NL | ( |
28 | 0 | 0 | 28 |
(%) | 39.44 | 0.00 | 0.00 | 29.17 | |
RR |
( |
13 | 0 | 0 | 13 |
(%) | 18.31 | 0.00 | 0.00 | 13.54 | |
RL |
( |
2 | 2 | 12 | 16 |
(%) | 2.82 | 25.00 | 70.59 | 16.67 | |
LR |
( |
6 | 0 | 0 | 6 |
(%) | 8.45 | 0.00 | 0.00 | 6.25 | |
LL | ( |
0 | 5 | 5 | 10 |
(%) | 0.00 | 62.50 | 29.41 | 10.42 | |
Total | ( |
71 | 8 | 17 | 96 |
(%) | 73.96 | 8.33 | 17.71 | 100.00 |
Two subgroups of children with scoliosis. NL: the tendency to overload the left body side; NR: the tendency to overload the right body side and four subgroups of children with CP; RR: right side hemiplegic and the tendency to overload the affected body side; RL: right side hemiplegic and the tendency to overload the unaffected body side; LL: left side hemiplegic and the tendency to overload the affected body side; LR: left side hemiplegic and the tendency to overload the unaffected body side.
In this large cohort of children, asymmetrical body posture was recognized in all three clusters. Cluster 1 (
The average IA of weight distribution between the right/left body sides or the affected/unaffected body sides in children from this cluster indicated almost symmetrical weight bearing (Table
Summary of the index of asymmetry of weight distribution between right/left body sides in control subjects and the affected/unaffected body sides in children with hemiplegia in particular clusters.
Cluster | Index of asymmetry | ||||
---|---|---|---|---|---|
Mean (%) |
|
SD (%) | Minimum (%) | Maximum (%) | |
Cluster 1 | −1.45 | 51 | 9.83 | −18.00 | 22.00 |
Cluster 2 | 6.96 | 23 | 28.48 | −38.00 | 48.00 |
Cluster 3 | −12.25 | 22 | 26.75 | −46.00 | 46.00 |
Total | −1.91 | 96 | 20.99 | −46.00 | 48.00 |
Significantly higher ALC values (approximately 10°) were noted in subjects in Cluster 2 (Table
Based on the aforementioned relationships, three types of postural patterns in children with body posture asymmetry have been recognized: the asymmetrical postural pattern with almost symmetrical weight bearing (SS), the asymmetrical postural pattern with asymmetrical weight bearing and overloading of the affected body side (+AS), the asymmetrical postural pattern with asymmetrical weight bearing and underloading of the affected body side (−AS).
Every outcome from the MT and PT examinations demonstrated good to very high level of intraobserver agreement for both groups of subjects, with the ICC ranging from 0.72 to 0.96 for CP and from 0.79 to 0.99, in all variables for able-bodied subjects. ICC values indicated very high level of interobserver agreement among researchers, with the ICC ranging from 0.92 to 0.99 for both groups.
Children with asymmetrical body posture have a variety of postural patterns. Cluster analysis was used to recognize different postural patterns and to search for underlying pathological mechanisms that could explain the large intersubject variability in the body postures of children with asymmetry of body posture. Three asymmetrical postural patterns were described based on the weight bearing between body sides and MT parameters, which were extracted using a data reduction technique; these included one pattern with almost symmetrical weight bearing and two different patterns with asymmetrical weight bearing. The asymmetrical postural pattern with almost symmetrical weight bearing was characteristic of all children with moderate scoliosis and for hemiplegic subjects with right-sided hemiplegia and a tendency to overload the affected body side (RR). The cluster analyses also identified two asymmetrical postural patterns with asymmetrical weight bearing in hemiplegic children; one was overloading of the affected body side (+AS), and the other was underloading of this side (−AS).
Clear differences in the MT parameters were characterized by the spinal deformities and the fringe deviations in pelvic obliquity and the shoulder girdle rotation, which were observed between these three postural patterns. Interestingly, greater spinal deformity was more commonly observed in children with almost symmetrical weight bearing (SS), not in the groups with asymmetrical patterns of weight bearing. Conversely, the pelvic obliquity and shoulder girdle rotation were the most important pathogenic factors in hemiplegic children with asymmetrical weight bearing (AS). Interestingly, hemiplegic subjects with a tendency to overload the affected body side (+AS) were commonly found in the group of children with scoliosis. Most likely, the observed reclustering was due to the higher values of the angle of curvature for scoliosis and the symmetrical weight bearing between the affected and unaffected body sides, which were more typical for children with scoliosis. The postural pattern did not appear to be determined by the diagnosis but primarily by the symmetry/asymmetry distribution of body mass between body sides and the value of scoliosis as well as the spatial relationship between the shoulder rotation and pelvic obliquity. In all subjects exhibiting almost symmetrical weight bearing, scoliosis represented the most common pattern of deformity, whereas children with asymmetrical weight bearing presented with laterality and pelvis obliquity (up) and a large externally rotated shoulder girdle on the affected side. Therefore, the differences between the asymmetrical postural patterns of children with moderate scoliosis and children with hemiplegia were not as clear as expected.
Despite the fact that a group of children with spastic hemiplegia appears to be relatively homogeneous, the present study has shown that their postural patterns differ. Based on the MT and the PMs of the body mass distribution between the affected and unaffected body sides, two different postural patterns were recognized in children with unilateral CP, one with overloading of the affected body side (+AS) and the second with underloading of the affected body side (−AS). This finding suggests that the distribution of body mass between the affected and unaffected body sides determined the characteristic compensatory action, which was the spatial relationship between the shoulder rotation and pelvic obliquity and the type and value of scoliosis. The obtained results confirmed the hypothesis that children with unilateral CP are not homogeneous in terms of their body weight distribution and body posture patterns. However, the diversities of the PGPPs and AGPPs described in our previous study were not completely confirmed. This finding should be confirmed in other series of statistical analyses before hypotheses can be formulated regarding this difference.
To our knowledge, this study is the first to examine body posture using MT as an objective evaluation of body posture in children with CP. The data demonstrated a very high intertrial reliability for every variable calculated from the MT examination in both able-bodied children and children diagnosed with CP. Previously, Chowanska and coauthors reported good intraobserver repeatability using CQ surface topography to examine children with scoliosis [
An asymmetric alignment while standing is often characteristic of children with a unilateral neurological lesion, such as hemiplegia [
It is well known that it is not possible to achieve thoroughly correct postural patterns when treating children with CP. The entire rehabilitation process for these children is based on the steering of compensation and alleviating the brain lesion symptoms. Certain consequences of compensatory postural patterns will develop and exceed their natural abilities of acting against gravity; these consequences must be considered in every case [
The present study recognized and defined differences between the asymmetry of the body postures of children with mild scoliosis and children with unilateral CP. Additionally, this study demonstrated that despite apparent similarities in children with unilateral CP, their postural patterns differed.
This awareness may be essential in the decision making process regarding the management, facilitation, modification, or elimination of each compensatory sign. The results of this study were promising, and, therefore, these findings should be confirmed in another series of statistical analyses that will precisely define postural patterns in children with unilateral CP and demonstrate the basic differences between them.
The authors declare that they have no conflict of interests.