Three-Dimensional Gait Analysis in Children Undergoing Gastrocsoleus Lengthening for Equinus Secondary to Cerebral Palsy

Background and Objectives: Equinus is the most common deformity in children with cerebral palsy, and surgical lengthening of the gastrocsoleus muscle-tendon unit is the most commonly performed operation for children with cerebral palsy. Treatment outcomes of orthopaedic surgery can be measured objectively with three-dimensional gait analysis. This study examined the quality of evidence for gastrocsoleus lengthening surgery based on objective measures. Materials and Methods: A search was performed with Medline, Embase and PubMed from 1990 to 25 August 2020 using the keywords “cerebral palsy”, “equinus”, “surgery” and “gait analysis”. Only studies of gastrocsoleus lengthening surgery using three-dimensional gait analysis were included, yielding 34 studies. Results: Fourteen studies reported swing phase kinematics and all studies reported a significant improvement. Rates of recurrent equinus and calcaneus were reported in 21 studies and varied widely based on follow-up period and surgical technique. Conclusions: Poor study quality and marked variability in study samples and interventions made comparison difficult. Future studies should consider prospective design, controls or comparison groups and more detailed breakdowns of outcomes by cerebral palsy subtype, sagittal gait pattern, and equinus type in order to allow more rigorous treatment recommendations to be made.


Introduction
Cerebral palsy (CP) is caused by non-progressive brain injury during early development [1]. Equinus is the most common gait abnormality in children with CP and is characterised by excessive ankle plantarflexion during stance phase [2]. This is often accompanied by persistent plantarflexion during swing phase, commonly referred to as 'drop foot'. Equinus deformity is not only caused by contracture of the gastrocsoleus muscle, but can also involve abnormal activity of other muscles of the ankle joint, such as tibialis anterior and peroneus longus [3][4][5]. Dynamic equinus may be treated non-operatively with methods including physiotherapy, injections of Botulinum Neurotoxin A (BoNTA), serial casting or ankle-foot orthoses (AFOs). Over time, dynamic equinus usually progresses to fixed contracture, which is treated surgically by lengthening of the gastrocsoleus muscle-tendon unit (MTU) by a variety of techniques [6].
Clinical examination, observational gait analysis (OGA) and video recording of gait (VGA) can be utilised to measure treatment outcomes. These measures are largely subjective, with inherent problems of validity and reliability [7]. This makes interpretation of studies difficult, and synthesis of multiple studies impossible. Three-dimensional gait analysis (3DGA) is considered the "gold standard" for assessing gait and measuring outcomes. It is objective, reliable and allows detailed analysis of gait parameters. Global gait scores transform multiple kinematic variables from 3DGA into single objective measures of the deviation of a patient's overall gait pattern from typically developing controls (TDC) [8]. This is useful in defining the cumulative effects of either single level surgery or single-event multi-level surgery (SEMLS).
Surgical treatment for equinus has been shown to improve ankle kinematics during stance phase in children with both unilateral spastic cerebral palsy (USCP) and bilateral spastic cerebral palsy (BSCP) [9][10][11]. However, deviations from typically developing children's gait patterns may persist during swing phase, with many patients expressing dissatisfaction because of persistent drop foot, which is appreciated as a limp [12]. Understanding pre-operative predictors for post-operative outcomes is crucial to improve treatment planning.
This review provides an overview of the current evidence for surgical correction of equinus deformities in CP based on the use of 3DGA for both planning and assessing outcomes (please see Figures 1 and 2). It focuses on the outcomes of surgical management measurable using 3DGA, particularly swing phase dorsiflexion to investigate drop foot, gait scores to understand impact on overall gait pattern and rates of recurrent equinus and calcaneus to consider stability of correction.

Materials and Methods
A search using keywords was performed with Medline and Embase using Ovid, with a date range from 1990 to 25 August 2020. PubMed was searched using keywords only to retrieve E-pubs and items not indexed in Medline. The Medline search strategy was adapted for use in other databases; the search histories are listed in the supplementary information. One additional study was obtained by direct contact with the author. Results were limited to English language only.
Duplicates were removed and titles and abstracts were screened. Studies with fewer than 10 participants were excluded, as well as diagnoses other than CP, and studies without original data. Studies of children aged zero to 18 were included. Full text articles were then screened to include studies using 3DGA. The original search strategy was used to find studies covering both non-operative and operative treatment options. However, for the purposes of this review, only studies of calf lengthening surgery were included. Sample demographics, interventions, and outcome measures for ankle kinematics were extracted from included studies.
Cohort studies were evaluated for methodological quality by the first and senior author using the Methodological Index for Non-Randomized Studies (MINORS) tool. MI-NORS is a valid instrument for assessing the quality of non-randomised surgical studies, based on criteria such as methods of patient recruitment, data collection, presence of a control group or comparison group, and the duration and completeness of follow-up [13].

Results
The search identified 103 articles in Medline, 176 articles in Embase, 37 articles in PubMed, and one article from direct contact with the author. After removal of duplicates, 193 articles remained, of which 34 met the inclusion criteria. Exclusion criteria and number of studies in each step of the search are described in a PRISMA diagram ( Figure 3).

Materials and Methods
A search using keywords was performed with Medline and Embase using Ovid, with a date range from 1990 to 25 August 2020. PubMed was searched using keywords only to retrieve E-pubs and items not indexed in Medline. The Medline search strategy was adapted for use in other databases; the search histories are listed in the Supplementary Information. One additional study was obtained by direct contact with the author. Results were limited to English language only.
Duplicates were removed and titles and abstracts were screened. Studies with fewer than 10 participants were excluded, as well as diagnoses other than CP, and studies without original data. Studies of children aged zero to 18 were included. Full text articles were then screened to include studies using 3DGA. The original search strategy was used to find studies covering both non-operative and operative treatment options. However, for the purposes of this review, only studies of calf lengthening surgery were included. Sample demographics, interventions, and outcome measures for ankle kinematics were extracted from included studies.
Cohort studies were evaluated for methodological quality by the first and senior author using the Methodological Index for Non-Randomized Studies (MINORS) tool. MINORS is a valid instrument for assessing the quality of non-randomised surgical studies, based on criteria such as methods of patient recruitment, data collection, presence of a control group or comparison group, and the duration and completeness of follow-up [13].

Swing Phase Dorsiflexion
Fourteen studies reported swing phase kinematics [4,9,10,12,14,18,21,[23][24][25][26]33,35,36]. All studies reported significant increases in maximum or mean swing phase ankle dorsiflexion. However, effect size and power were not reported in any study. Improvement in maximum dorsiflexion in swing in patients with BSCP ranged from 8.0 • to 17.7 • , whereas improvement in USCP ranged from 12.0 • to 20.9 • . These improvements are clinically meaningful but the Minimal Clinically Important Difference (MCID) has only been established for composite gait indices, such as the Gait Profile Score (GPS).

Gait Scores
Studies using gait scores are shown in Table 2. Four studies reported Gait Deviation Index (GDI) [11,29,32,37]. GDI combines 15 gait features to measure overall gait pathology. An improvement in 10 points of GDI corresponds to one standard deviation towards the mean for typically developing children [38]. Improvement in GDI ranged from 12.5 to 24.3 in patients with BSCP. Improvement in GDI of 10.6 for USCP was reported by one study [32].
Three studies reported Gillette Gait Index (GGI) [9,32,39]. GGI combines 16 kinematic parameters to measure overall gait, with higher scores indicating greater deviation from able-bodied individuals [40]. Improvement in mean GGI for USCP ranged from 157 to 200, while improvement in BSCP ranged from 184 to 337.
Seven studies [16,19,20,23,30,32,33] reported ankle Gait Variable Score (GVS). GVS measures the difference between patients and people with no gait pathology for a single gait variable [8]. In patients with USCP, reported improvement of mean GVS across the studies' samples ranged from 8.3 to 12.3, while the improvement in patients with BSCP ranged from 7.3 to 15.6. In USCP, improvement in GVS ranged from 83% to 100%, while in BSCP, improvement ranged from 81% to 93%.
Eight studies [16,19,20,23,29,30,32,33] reported Gait Profile Scores (GPS). The GPS measures the difference between patients and people with no gait pathology across nine kinematic variables [8]. All but two studies reported an improvement in GPS greater than the minimal clinically important difference (MCID) of 1.6 • [41]. Pilloni et al. [19] reported a 1.5 • deterioration in mean GPS in true equinus compared to 6.4 • improvement in jump gait. Kim et al. [29] investigated endoscopic surgery, reporting an improvement in GPS of 1.4 • . Excluding these studies, the improvement of mean GPS in patients with USCP ranged from 1.8 • to 3.8 • , while improvement in BSCP ranged from 2.2 • to 6.6 • . Post-operatively, improvement in GPS was noted in 70 to 100% of patients with USCP and 86% of patients with BSCP.

Stability of Correction
Twenty-one studies (Table 1) reported the presence of recurrent equinus, calcaneal gait, or crouch gait following calf lengthening surgery. Definitions varied and were based on either stance or swing phase, with different distances from the mean of normal used. In children with USCP, recurrent equinus ranged from 0% to 38%, and calcaneal gait from 4% to 30%. In children with BSCP, recurrent equinus ranged from zero to 35%, and calcaneal gait from 0% to 40%. Recurrent equinus ranged from 0% to 18% in studies with less than four years of follow up, and 16% to 38% in studies with more than four years of follow up. For calcaneus, studies with less than four years of follow up reported a rate of 0% to 30%, while studies with more than four years follow up reported a rate of 3% to 40%.

Study Quality
Evaluation of the 33 cohort studies using MINORS showed marked variability in the quality of the studies (Figure 4). Scores for cohort studies ranged from six to 11 (maximum score 16), with no studies being blinded, none reporting any prospective calculation of study size and most studies being retrospective.

Study Quality
Evaluation of the 33 cohort studies using MINORS showed marked variability in the quality of the studies (Figure 4). Scores for cohort studies ranged from six to 11 (maximum score 16), with no studies being blinded, none reporting any prospective calculation of study size and most studies being retrospective.

Discussion
Introduction of 3DGA has assisted pre-operative planning, detailed analysis of preoperative gait pattern and objective assessment of outcomes following surgery for equinus gait, in children with CP. Correction of equinus is important to improve quality of life in patients, as well as prevent further musculoskeletal consequences such as midfoot break, which can result in pain, calluses, and brace intolerance making walking and wearing of supportive footwear difficult [47]. In 2020, Rutz et al. [45] reported in a Delphi study a process to achieve consensus between experienced surgeons on the identification of factors, including 3DGA that provide indications for GSL surgery. Pre-operative predictors

Discussion
Introduction of 3DGA has assisted pre-operative planning, detailed analysis of preoperative gait pattern and objective assessment of outcomes following surgery for equinus gait, in children with CP. Correction of equinus is important to improve quality of life in patients, as well as prevent further musculoskeletal consequences such as midfoot break, which can result in pain, calluses, and brace intolerance making walking and wearing of supportive footwear difficult [47]. In 2020, Rutz et al. [45] reported in a Delphi study a process to achieve consensus between experienced surgeons on the identification of factors, including 3DGA that provide indications for GSL surgery. Pre-operative predictors for surgical outcomes are important for informed decisions to be made about interventions. Improvement in ankle kinematics by calf lengthening surgery during stance phase is well described in children with USCP and BSCP [9][10][11]14,[24][25][26][27]33]. However, post-operative changes in swing phase are also important as insufficient foot clearance during swing can lead to tripping, and compensatory movements such as increased knee or hip flexion [18]. A drop foot in swing may require the use of an AFO and is perceived as a "limp" or abnormal gait pattern by patients, peers and parents/carers [1].
All studies in this review reported an improvement in equinus gait during stance phase and the improvements were clinically and statistically important in the majority of studies [10,11,14]. All studies in this review found improved ankle dorsiflexion in swing, in some patients. Lofterod et al. reported that 47.5% of their patients still exhibited drop foot during swing after calf lengthening [12]. This study suggested a relationship between increased pre-operative maximum plantarflexion in initial swing and post-operative drop foot. It also suggested that a pre-operative Selective Motor Control (SMC) score of four using the Boyd and Graham score [48] is equivalent to normal swing phase postoperatively. However, this was based on only four limbs, so care should be taken with making generalisations from these limited numbers.
Davids et al. [4] failed to identify any predictors for active ankle dorsiflexor function in swing, defined by a positive slope of the sagittal ankle kinematic graph during swing phase. Presence of active ankle dorsiflexion in swing increased post-operatively from 79% to 96% of patients, with 19% of patients improving from a grading of "absent" to "present", and 2% changing from "present" to "absent". Davids et al. also reported 38% of patients improving in SMC, with 40% of patients graded as having normal SMC pre-operatively and 53% of patients having normal SMC post-operatively. Dynamic electromyography (EMG) showed no clear improvement in gastrocnemius and tibialis anterior co-activation during swing, with 11% of patients having abnormal co-activation pre-operatively, and a different 11% having abnormal co-activation post-operatively. The absence of change in EMG activity post-operatively despite improvement in swing phase kinematics is supported by other studies [14,39]. While drop foot in swing phase was not an outcome measured in the study, these findings suggest that some patients are able to exhibit active ankle dorsiflexor function without normal SMC or normal phasic activation of the gastrocnemius and tibialis anterior. Further investigation of muscle activation patterns is required as excessive cocontraction of agonist and antagonist muscles, specifically the dorsiflexor and plantarflexor muscles in the case of equinus and drop foot, play a role in the mechanism of spastic paresis [5]. The role of peroneus longus (PL) in drop foot could be considered as premature onset of PL activity in swing phase in conjunction with premature gastrocnemius activity has been implicated in equinovalgus deformity [3].
Kay et al. [35] also found no significant improvement in dorsiflexor control, despite improvement in swing phase dorsiflexion. This casts further doubt on the role of SMC in predicting swing phase kinematics.
In 2020, Rajagopal et al. [44] reported greater improvement of ankle kinematics after GSL in patients with short gastrocnemius lengths in gait pre-operatively, compared to patients without a short gastrocnemius. However, the zone of GSL used for each patient was not reported and the patient sample was not classified by CP type. Additionally, the use of the ankle deviation index (ADI) as the outcome measure did not allow information about swing phase dorsiflexion to be extracted. Hence, further investigation is required for the use of gastrocnemius muscle length as a pre-operative predictor of GSL outcomes.
It is well accepted that surgical outcomes differ between CP topographical types, USCP and BSCP [45]. However, none of the aforementioned studies distinguished patients according to topographical distribution, USCP or BSCP. Hence, further studies investigating swing phase kinematics should consider analysis of outcomes based on CP subtype and movement disorder.
Gait scores provide an objective measure of the deviation of a patient's gait from that of a typically developing child, as well as averaging kinematics at multiple joints. This is important as SEMLS is the most widely used surgical intervention in children with BSCP [1]. The gait outcomes of SEMLS are the result of corrections of deformities in multiple locations and anatomic planes, not just GSL (Figure 1). Global gait scores reported by studies [32,39] suggest that patients with BSCP have a more abnormal global gait pattern than patients with USCP pre-operatively, allowing more room for improvement [1]. This is supported by greater post-operative improvement in gait found in these studies in patients with BSCP compared to those with USCP.
Under-correction after calf lengthening surgery can lead to recurrent equinus, while over-correction can lead to calcaneus or crouch gait. Shore et al. [49] highlighted that cerebral palsy type, surgical technique, and follow-up period can influence the rate of recurrent equinus and calcaneus. This review supports the finding that studies with longer follow-up periods had higher rates of recurrence or calcaneus, emphasising the need for longer follow-up periods, preferably to skeletal maturity and beyond, to determine the stability of correction more accurately. However, this review did not find any obvious difference between the rates of recurrent equinus and calcaneal gait when comparing USCP and BSCP. This may be due to the range of sample sizes, follow-up periods, surgeries and definitions used.
Type of surgery can impact the rates of calcaneus and recurrent equinus after surgery. Results from Fujita et al. [23] suggest that anterior transfer of flexor hallucis longus with gastrocsoleus lengthening can increase the rate of post-operative calcaneus. Additionally, Borton et al. [15] reported post-operative calcaneal gait in 40% of patients with BSCP. This is much higher than other studies of similar follow-up period that reported rates of 10% to 11% [9,11]. The higher rate of calcaneus reported by Borton et al. may be due to the use of isolated zone 2 and 3 calf lengthening surgeries and inclusion of patients as young as two years old ( Figure 2). Zone 3 surgeries are now considered contraindicated for patients with BSCP, and there is consensus that the preferred age for calf lengthening surgery is between six and ten years old [45]. Multiple features from the Delphi study have been suggested to optimise the indications for GSL, including age at surgery, CP subtype, physical examination, and kinematic features [45]. These criteria remain to be tested in prospective studies and clinical trials.
In 2011, Rutz et al. described a novel technique utilising concurrent tibialis anterior tendon shortening (TATS) with calf lengthening to improve dorsiflexor function [32]. The improvements in GDI and GVS fell within the range found by other studies, with the exception of a greater improvement in GPS in patients with BSCP [11,19,20,23,29,37]. This makes interpretation difficult as to whether TATS provides additional benefit to ankle kinematics, when added to GSL. However, using TATS, Klausler et al. reported 0% overcorrection at a mean six-year follow-up period, with 25% of patients with USCP and 0% of patients with BSCP having recurrent equinus [30]. This is lower than other reported recurrent equinus rates of 38% in USCP and 16-35% in BSCP at similar follow-up periods [9,11,15,16]. Overcorrection or calcaneal gait at long-term follow-up has been reported as 4% in children with USCP and 3-11% in children with BSCP [9,11,15,16]. However, it should be noted that Klausler et al. did not provide a definition for recurrent equinus, overcorrection or calcaneal gait, which have been variably defined across studies [15,28,35].
Limitations of this review include a single author screening studies for inclusion, extracting study characteristics and outcomes, and performing qualitative synthesis. This was addressed by using a systematic approach to the inclusion of articles (Figure 3). Additionally, the studies included in this review were of generally poor methodological quality ( Figure 4). Study samples and interventions were heterogenous and ranges reported in this review include outcomes of studies with variable sample characteristics and incomplete documentation of key variables. Improvement in study design is critical for future studies. Gait patterns involve multiple joints, and interventions are tailored to the individual, making rigorous study designs challenging to conduct. However, while cohort studies remain the dominant design for investigating surgical interventions, more rigorous evidence could be provided by the use of prospective studies and longer follow-up periods. Comparison of studies could be improved by agreed, universal definitions for equinus, recurrent equinus, calcaneus, and crouch gait based on physical examination and 3DGA parameters. Further breakdown of outcome measures based on topographic distribution of CP, Rodda and Graham equinus classifications [50] and calf lengthening zone could provide useful information in predicting outcomes. Finally, predictors of post-operative drop foot in swing, and recurrent equinus or calcaneus requires further exploration.

Conclusions
Various methods of calf lengthening surgeries are available to treatment fixed equinus deformities in children with cerebral palsy. Studies using 3DGA provide objective outcome measures but vary greatly in sample characteristics, with many studies pooling patients and interventions without considering the impact of features such as topographic distribution and type of equinus. Further prospective studies and randomised control trials to determine pre-operative predictors for swing phase kinematics could consider CP type, equinus type, type of surgery, active ankle dorsiflexor function, gastrocnemius muscle length, muscle activation patterns, and SMC.