1. Introduction
During normal gait, the ipsilateral pelvis is internally rotated at the initial foot contact owing to forward positioning of the foot, and then there is a continuous external rotation of the pelvis to an extent of less than 5 degrees by the time of contralateral foot contact [
1]. This normal pelvic movement can be compromised in patients with underlying pathologic conditions. Pelvic retraction, excessive external rotation of the pelvis during gait, is a common finding seen in patients with cerebral palsy [
2]. It produces functional problems and cosmetic concerns due to an asymmetric gait.
Pelvic retraction has been considered a consequence of primary neurological deficit
per se, resulting from a lesion of the central nervous system [
3,
4]. However, it may be secondary to muscle spasticity or may be a coping response to the torsional deformity of the lower extremity. O’Sullivan et al. examined average and maximum values of gait parameters and found that ankle plantarflexor tightness, increased hip flexion and hip internal rotation were the significant features seen in patients with cerebral palsy and pelvic retraction [
2]. On the contrary, contractures of hip and knee flexors were reported to be factors associated with pelvic retraction in another study [
5]. Several surgical procedures have also been introduced to reduce pelvic retraction [
6,
7]; some improvements of pelvic retraction after derotational osteotomy of the femur were observed, and an increased hip internal rotation associated with an increased femoral anteversion was suggested to be the contributing factor for pelvic retraction in hemiplegia [
8,
9].
However, a review of the literature is difficult because heterogeneous groups of patients with hemiplegia and diplegia were included in most studies. To identify any factors related to pelvic retraction in cerebral palsy, the inclusion of a large patient cohort with homogeneous disease entity is the first requirement as patients with spastic hemiplegia and diplegia show different gait patterns. Furthermore, the definition of pelvic retraction was inconstant among studies and many authors simply used the average or maximum value of pelvic rotation in order to examine the degrees of external rotation of the pelvis and to determine pelvic retraction; however, those values cannot represent real persistent pelvic retraction during the entire phases of gait and patients with exaggerated pelvic motion may be misinterpreted as having pelvic retraction [
10,
11].
For better prediction of treatment outcomes, not only the three-dimensional gait parameters but also the clinical features related to abnormal gait pattern should be considered [
8,
12,
13,
14,
15]. Nonetheless, there have been more endeavours at the interpretation and biomechanical analysis of abnormal pelvic motions with small subgroups of patients enrolled in each study [
2,
5,
7,
16]. Also, it is necessary to investigate the effects of unilateral involvement of the entire limbs and the severity of clinical deficits of the ipsilateral limbs on the development of pelvic retraction. The purpose of this present study is to identify any clinical features and gait parameters that may be related to pelvic retraction in patients with spastic hemiplegic cerebral palsy (SHCP). To the best of our knowledge, this is the only study including a broader array of assessment domains of both clinical and gait parameters with a considerably large and homogenous population with hemiplegia.
2. Experimental Section
This was a retrospective study and was approved by our hospital’s institutional review board (IRB No. 4-2009-0035). Written informed consent was obtained from all participants’ parents. The research methods were carried out in accordance with the relevant institutional guidelines and regulations.
2.1. Subjects
Three hundred and twelve independent ambulatory patients with SHCP (Gross Motor Function Classification System, GMFCS I or II) who underwent three-dimensional instrumented gait analysis (VICON 370 Motion Analysis System, Oxford Metrics, Oxford, England) were recruited. First, we observed the pattern of pelvic rotation using each patient’s kinematic graphs, and the presence of a pelvic retraction was defined as when the consistently increased external rotation of the pelvis throughout the gait cycle (
Figure 1a) was more than two standard deviations (SDs) from our normative database (<−4.75°) [
2,
5]. Second, patients with increased internal rotation of the involved side compared to the normal contralateral side (
Figure 1b) were excluded as this represents primary external pelvic rotation on the contralateral side [
1,
11]. Also, patients with an excessive or exaggerated range of pelvic motion characterized by an increased internal rotation of the pelvis at early stance phase and then followed by a sinusoidal type of pelvic rotation (
Figure 1c) were excluded; this represents compensation for a reduced sagittal plane excursion of motion [
11]. The degree of pelvic rotation was calculated by averaging the maximum and the minimum values of pelvic rotation in all patients [
2].
In total, 212 patients were enrolled in the study (
Figure 2). There were 84 (39.6%) females and 128 (60.4%) males. The average age at the time of gait analysis was eight years and eight months (range, 38 months to 29 years). One hundred and twenty-seven patients (59.9%) had right hemiplegia, and 85 (40.1%) had left hemiplegia. The subjects were divided into two groups; Group I consisted of 113 patients (53.3%) who had pelvic retraction, and Group II comprised 99 patients (46.7%) who did not have pelvic retraction.
2.2. Clinical and Gait Parameters
A thorough physical examination was performed in all patients: modified Ashworth scale [
17], passive range of motion in the hip, knee, and ankle joints; popliteal angle for hamstring tightness, Duncan-Ely test for rectus femoris spasticity or tightness, and ankle dorsiflexion angle during the Silfverskiöld test for the detection of gastrocnemius/Achilles tendon tightness. The degree of ankle dorsiflexion was measured with the knee flexed and extended. Limb length discrepancy was assessed by measuring the distance from the anterior superior iliac spine to the medial malleolus in both sides in all patients. A scanogram was also checked in 175 (82.5%) patients (93 patients in Group I and 82 in Group II). Femoral anteversion was measured by the trochanteric palpation method [
18] and tibial torsion was measured using the bimalleolar axis in all patients [
19]. One hundred and seventy-two patients (81.1%) also underwent a computed tomographic (CT) scan to measure both femoral anteversion and tibial torsion.
Gait analysis was performed using a VICON 370 Motion Analysis System (Oxford Metrics, Oxford, England) with six infrared cameras and a Helen Hayes marker set. Data on ground–reaction forces were gathered from multiple force platforms (Advanced Mechanical Technology, Watertown, MA, USA). All subjects were asked to walk barefoot at a self-selected speed along a 15-m walkway with markers in place. We selected values at the initial contact, maximum, minimum, and average values for each kinematic parameter at each phase of gait. Several variables that have been regarded as clinically unimportant were excluded; furthermore, those variables have been known to be interrelated with each other in a complex way, and may act as confounding factors in the interpretation of gait data and in understanding abnormal movement [
2,
5,
9,
14,
20,
21,
22].
Two pediatric orthopaedic surgeons reviewed all the videotaping for visual gait analysis, three-dimensional gait analysis data, dynamic foot-pressure measurements (pedobarographs), gross clinical photos, and standing plain radiographs of the lower extremity in all patients. The presence of an equinovarus foot deformity was defined as when inversion and plantar flexion of the hindfoot were present in standing position and a significant pressure exerted on the lateral forefoot and midfoot compared to the non-involved were confirmed on the dynamic pedobarographs (Tekscan, South Boston, MA, USA). The determination of knee recurvatum gait and the type according to the classification by Winters et al. [
23] were done using gait data and visual analysis. We referred recurvatum gait as a clinical static variable as it is a descriptive type of qualitative classification. We modified the original Winters classification as follows; a patient with a normal range of knee motion but having increased knee flexion at initial contact and terminal stance phase of gait was categorized into type III (
Figure 3). Consequently, patients with the original type III with increased knee flexion and type IV with increased hip flexion, hip internal rotation, and pelvic retraction were then re-classified as type IV and type V, respectively. The presence of an asymmetrical posturing of the upper extremity was defined as when the patient has more than 30 degrees of elbow flexion contractures and typical posturing becomes apparent when the patient walks (
Figure 4) [
24,
25].
2.3. Statistical Analyses
Statistical analyses were performed using IBM
®SPSS
® software version 23 (IBM Corporation, Armonk, NY, USA). The level of significance was set at
p < 0.05. An independent t-test was used for initial comparison between the groups. Twenty-one gait parameters were selected for the clustering technique analysis: range of pelvic motion in the sagittal plane, average anterior pelvic tilt, range of pelvic motion in the coronal plane, average pelvic obliquity, minimum hip sagittal angle, average hip sagittal angle, maximum hip coronal angle, average hip coronal angle, maximum hip transverse angle, average hip transverse angle, minimum knee sagittal angle, range of knee motion in sagittal plane, knee sagittal angle at initial contact, average knee transverse angle, maximum ankle sagittal angle, minimum ankle sagittal angle, average ankle sagittal angle, ankle sagittal angle at initial contact, average ankle transverse angle, maximum foot progression angle, and average foot progression angle. With six temporospatial parameters, a total of 27 gait parameters were clustered as gait factors, with the consideration of the differences in patients’ ages and the correlation between the gait parameters [
2,
13,
14,
22,
27].
Finally, eight gait factors (ankle dorsiflexion, temporospatial parameter (walking speed, stride & step length), temporospatial parameter (cadence, stride & step time), internal rotation of foot and ankle, knee extension, pelvic obliquity and hip abduction, anterior pelvic tilt, hip internal rotation) and five clinical factors (modified Winters classification, Achilles tendon tightness, gastrocnemius tightness, pes equinovarus, asymmetrical posturing of the upper extremity) were included for the multivariate logistic regression analysis.
4. Discussion
Previous studies suggested several associated factors related to pelvic retraction during gait in cerebral palsy. However, the results of single event multi-level surgery performed in patients with cerebral palsy were not consistent in terms of the improvement of pelvic retraction [
6,
7,
20,
21,
30]. Furthermore, pelvic retraction observed to be improved at the short-term follow-up was not noted to be maintained at the long-term follow-up [
2,
9,
16,
20]. These findings suggest that other factors rather than those addressed in previous studies must have existed. In the present study, we have found that pelvic retraction in spastic hemiplegia is related to the following parameters: increased anterior pelvic tilt and hip internal rotation; decreased ankle dorsiflexion; type II according to the classification by Winters et al.; and asymmetrical posturing of the affected upper extremity during gait. If any factors related to pelvic retraction could be clarified before the operation, the surgical results may be predicted more reliably.
The patients with a “relative” internal rotation of the pelvis on the affected side caused by primary external pelvic rotation on the contralateral side, and the patients with an excessive range of pelvic motion were not included in the present study. Only the patients with an excessive external rotation of the pelvis outside two standard deviations throughout the gait cycle were defined as having pelvic retraction. The determination of pelvic retraction should not be made by a simple measurement of the average values of the external rotation of the pelvis [
1,
10]. Not only the average values of the parameters but also the “patterns” of pelvic rotation throughout the gait cycle should be considered, as described earlier in
Section 2.1.
The transverse motion of the hip joint during gait is affected by the torsional deformity of the femur and/or tibia as well as by the spastic muscles. Another concern is whether the correction of an increased internal rotation of the hip may also treat pelvic retraction. Improvements in pelvic rotation after femoral derotational osteotomy have been reported, however there was no differentiation between the hemiplegic and diplegic patients in their studies [
7,
20]. Furthermore, the number of each subgroup of the patients enrolled was too small to draw a definitive conclusion [
8,
27]. Rutz et al. observed an improvement in pelvic rotation and a correction of hip internal rotation only in patients with Winters classification type IV; they had had soft tissue releases of the hip flexors and adductors in addition to femoral derotational osteotomy [
31]. It is our opinion that to determine the effects of derotation osteotomy on the improvement of pelvic retraction, those factors found in our study as well as any abnormal movement occurring in the non-involved side of limb should be considered as well.
Decreased ankle dorsiflexion found in patients with Group I was due to tightness of the gastrocnemius and Achilles tendon, and this was confirmed by comparing the physical examination findings between the groups. Normal forward progression of the tibia over the supporting foot during the stance phase of gait is prevented by tight calf muscles [
2,
5,
7,
30]. Pelvic retraction might occur as a consequence of insufficient forward progression of the body during the stance phase, and the lengthening of shortened ankle plantarflexors would improve pelvic retraction [
6]. However, we could not appreciate any other effects of tight gastrocnemius and Achilles tendon, as these clinical features were not statistically significant in the multivariate logistic regression analysis.
On the contrary to the previous observations [
2], our results showed that there is no relationship between pelvic retraction and tightness/spasticity of the rectus femoris. Despite no difference in the rectus tightness, patients in Group I had lower degrees of hip extension and had decreased maximum knee flexion during gait compared to Group II. The rectus femoris tightness has been thought to be secondary to longstanding inappropriate hip extension, and this diminished range of hip extension is related to ankle equinus [
2]. Decreased ankle dorsiflexion may cause reduced hip extension and may also elicit decreased knee flexion. On the other hand, reduced hip and knee motions in the sagittal plane have been known to be associated with an excessive range of pelvic motion [
1]. It is our opinion that there may be differences in terms of the degree of spasticity of the rectus femoris or the amount of reduced range of motion, compared to the patients with excessive pelvic motions.
Some previous studies introduced modified Winters classifications [
26,
32] in order to allocate subsets of patients that cannot be classified with the original system, and to make up the limitations in classifying types I, II, and III which were determined only by abnormalities in the sagittal plane. We also adapted a modified Winters classification in the study because 20% of our patients had increased knee flexion at initial and terminal stance phasse of gait, as well as having a normal range of knee motion in the sagittal plane; they were categorized into modified type III, and patients with original type III with increased knee flexion were classified as modified type IV. The characteristic finding of modified types III and IV was knee flexion compared to type II which has only ankle equinus. In our logistic regression analysis, allocation into type II was found to be related to pelvic retraction. Increased knee flexion is the characteristic finding in modified types III and IV; this increased knee flexion may act as a compensation for forward progression of the tibia. We think that pelvic retraction may decrease with forward progression of the knee. Patients with modified type V in our series showed less pelvic retraction than patients with type I. However, as only six patients (2.8%) were modified type V compared to 47 patients with type I (22.2%), it may be difficult to conclude that patients with modified type V have less risk of pelvic retraction.
No previous studies mentioned an association between asymmetrical posturing of the upper extremity during walking and pelvic retraction. Asymmetrical posturing was more frequent in the group with pelvic retraction, and was also found to be a related factor to pelvic retraction in the logistic regression analysis. Nevertheless, it is not certain that any surgeries on the upper extremity may improve pelvic retraction as the asymmetrical posturing itself is related not only to the trunk balance during gait but also to the severity of a primary neurologic impairment per se [
33].
Our study has several limitations: we selected only five clinical factors, however those clinical factors have traditionally been frequent in patients with hemiplegia and have also been considered as significant factors related to the abnormal gait; scanogram and CT scan were not performed in about 20% of our series as they refused radiological exposure. Nonetheless, the trochanteric palpation method is a useful and reliable method to measure femoral anteversion in patients with cerebral palsy [
18], and the use of the bimalleolar axis for tibial torsion shows a high correlation with CT measurement [
19]. Furthermore, there were no significant differences in both physical examination and radiographic evaluation between the groups; and as the clinically perceived pelvic retraction may be a movement that may result from complex movements occurring in the sagittal, coronal, or transverse plane, the clinician should be aware that there may be differences between the real gait and three-dimensional gait analysis.