< Back to previous page

Project

Consensus-based classification of gait in children with cerebral palsy

The most common physical disability in children is cerebral palsy (CP). The prevalence of this neuromotor disorder is estimated at 1.7 per 1000 live births. CP is caused by non-progressive brain lesions that occur during the antenatal, perinatal, or postnatal period, at a time when the brain and spinal cord are not yet fully developed. As brain lesions are not curable, treatment mainly focusses on symptom management. Primary motor symptoms in CP are spasticity, weakness, impaired balance, and loss of selective motor control. These symptoms often have a destructive influence on a patient’s ability to walk, which is one of the most crucial functional activities of daily life. In CP, about 70% of children are able to walk, albeit with major or minor gait deviations and with or without the use of walking aids. The clinical presentation of gait in CP is ever changing due to the complex interplay of those primary motor symptoms with a maturing brain, growth, and treatment. As a result, secondary symptoms such as muscle contractures and lever arm dysfunctions will eventually occur, for which invasive treatments such as orthopedic surgery are required. To guide treatment planning, gait is typically evaluated through instrumented, three-dimensional motion analysis (3DGA), which provides a highly detailed assessment of joint angles, joint moments, and power during walking. The challenge with using this comprehensive biomechanical measurement of gait is the clinical interpretation of the vast amount of multidimensional data that it generates. The benefits of 3DGA as opposed to observational gait analysis are therefore dependent on the expertise of the clinical professional who is analyzing the data. To this date, there is no standardized method to qualitatively interpret 3DGA data, and there is a lack of effective and robust tools that capture the full complexity of gait reliably and validly, with widespread clinical acceptance and applicability. In routine clinical and research practice, the amount of 3DGA data is reduced before it is analyzed and interpreted. Two approaches for reduction are commonly applied. The first approach analyses gait features, which are specific points extracted from the kinematic and kinetic waveforms. Within the scope of this PhD research, important steps were undertaken in the search for an alternative, standardized method to extract and analyze the clinically relevant information from 3DGA data (study 1). The second approach to reduce 3DGA data is to define gait patterns, which allocate multiple gait features, either within a joint or across multiple joints, into groups. The principal goal of the PhD was to develop a clinically relevant, reliable, and valid classification for pathological movement patterns during gait in children with CP (study 2, 3, 4, 5).

In the first study, a literature review found that approximately 220 papers have reported on the outcome of treatment in CP by evaluating children pre- and post-treatment using 3DGA features. Focusing on the studies that evaluated the effect of Botulinum Toxin type A (a common treatment intervention to manage spasticity), this first study shows that there is no consensus regarding the selection of gait features that are expected to be sensitive to change after treatment in CP. Feature analysis may fail to provide a full understanding of the effect of treatment. Clinically relevant information could be missed, as the selection of features is most likely based on the clinical expertise of the medical team that decides on the treatment plan. In a subsequent retrospective intervention study, statistical parametric mapping (SPM) was identified as a valid, unbiased statistical alternative. SPM allowed kinematic and kinetic waveforms to be evaluated as a whole. This statistical approach eliminates the need for a priori data reduction, and keeps the probability of making a Type I or Type II error stable by considering the interdependency of all points of the waveform.

Regarding gait classifications in CP, a literature review showed that several gait patterns and classifications based on kinematic and kinetic data have been previously reported in literature. However, their clinical applicability is limited because psychometric properties of reliability and validity are often not yet established. In the second study, a Delphi consensus project was organized to ensure that the new classification was clinically relevant. The Delphi approach is a semi-quantitative research method where an international expert panel was consulted via iterative surveys to provide their opinion on the problem of gait classification in CP. The study started with a first proposal of gait patterns that should be included in the classification, based on previous literature and on the expertise of the clinical and research team of University Hospitals Leuven. After three consecutive survey rounds, consensus was reached on 49 gait patterns across the pelvis, hip, knee, and ankle joints in the sagittal, coronal, and transverse plane.

After the development of the classification, a necessary next step was to ensure that the level of clinician agreement on the patterns of a patient was at a sufficiently high level so that the patterns could be used reliably in practice. In the third study, an international agreement study was conducted among 29 clinicians with varying levels of experience with regards to CP and 3DGA. Apart from a few individual patterns, the study demonstrated that, after a brief learning phase, clinicians could assign the consensus-based joint patterns with good consistency between and within raters. The amount of patient data that were found to be ‘unclassifiable’ by the clinicians was low compared to previously published classifications and formed an indirect confirmation of the content validity of the patterns.

As the patterns defined during the Delphi study were the result of an informed, yet subjective opinion of an expert panel, the classification might have still provided an incomplete picture on gait pathology in CP. Therefore, the fourth study examined the content validity of the classification. It was assessed whether objective patient data supported the existence of the patterns. To this end, SPM was used to analyze a large database of classified kinematic and kinetic trials. This was to confirm whether each of the 49 patterns differed from the gait pattern of typically developing children in the key areas of the gait cycle that were indicated in the pattern definitions by the experts. Even though this hypothesis was largely confirmed, some additional areas that were not included within the definitions of the Delphi patterns were highlighted by the SPM analysis.  

The fifth study examined the construct validity of the classification, measuring the extent to which the gait patterns of the classification system are able to distinguish between the categories of other validated scales that measure the same or a related construct. Therefore, the prevalence of the patterns in a large cohort of children with CP was evaluated. It was found that the distribution of these patterns was associated with the distribution of other relevant and validated scales in CP, such as topographical classification, gross motor function, and levels of spasticity and weakness.

In conclusion, this doctoral research has made important contributions to the analysis and standardized interpretation of 3DGA. The classification that was developed within this research presents clinicians and researchers with a comprehensive overview of clinically relevant gait features and gait patterns, which will hopefully serve as a basis for improved communication and a more uniform terminology regarding gait pathology in CP. Patterns and their definitions should be adapted when necessary, and future research should further demonstrate the validity, responsiveness, and clinical applicability of the classification. One of the main contributions of this PhD thesis is that the developed methodological framework, combining qualitative and quantitative research methods to build a clinically relevant, reliable, and valid classification, could also be generalized to any other medical condition that affects movement.

Date:15 Oct 2012 →  14 Oct 2016
Keywords:statistical parametric mapping, cerebral palsy, gait, classification
Project type:PhD project