Energetics of walking in individuals with cerebral palsy and typical development, across severity and age: A systematic review and meta-analysis

BACKGROUND
Individuals with cerebral palsy (CP) report physical fatigue as a main cause of limitation, deterioration and eventually cessation of their walking ability. A consequence of higher level of fatigue in individuals with CP leads to a less efficient and long-distance walking ability.


RESEARCH QUESTION
This systematic review investigates the difference in 1) walking energy expenditure between individuals with CP and age-matched typically developing (TD) individuals; and 2) energetics of walking across Gross Motor Function Classification System (GMFCS) levels and age.


METHODS
Five electronic databases (PubMed, Web of Science, CINAHL, ScienceDirect and Scopus) were searched using search terms related to CP and energetics of walking.


RESULTS
Forty-one studies met inclusion criteria. Thirty-one studies compared energy expenditure between CP and age-matched controls. Twelve studies correlated energy expenditure and oxygen cost across GMFCS levels. Three studies investigated the walking efficiency across different ages or over a time period. A significant increase of energy expenditure and oxygen cost was found in individuals with CP compared to TD age-matched individuals, with a strong relationship across GMFCS levels.


SIGNIFICANCE
Despite significant differences between individuals with CP compared to TD peers, variability in methods and testing protocols may play a confounding role. Analysis suggests oxygen cost being the preferred/unbiased physiological parameter to assess walking efficacy in CP. To date, there is a knowledge gap on age-related changes of walking efficiency across GMFCS levels and wider span of age ranges. Further systematic research looking at longitudinal age-related changes of energetics of walking in this population is warranted.


Abstract Background
Individuals with cerebral palsy (CP) report physical fatigue as a main cause of limitation, deterioration and eventually cessation of their walking ability. A consequence of higher level of fatigue in individuals with CP leads to a less efficient and long-distance walking ability.

Research question
This systematic review investigates the difference in 1) walking energy expenditure between individuals with CP and age-matched typically developing (TD) individuals; and 2) energetics of walking across Gross Motor Function Classification System (GMFCS) levels and age.

Methods
Five electronic databases (PubMed, Web of Science, CINAHL, ScienceDirect and Scopus) were searched using search terms related to CP and energetics of walking.

Results
Forty-one studies met inclusion criteria. Thirty-one studies compared energy expenditure between CP and age-matched controls. Twelve studies correlated energy expenditure and oxygen cost across GMFCS levels. Three studies investigated the walking efficiency across different ages or over a time period. A significant increase of energy expenditure and oxygen cost was found in individuals with CP compared to TD age-matched individuals, with a strong relationship across GMFCS levels.
Based on the Gross Motor Function Classification System (GMFCS), children with CP can be independently ambulatory (I, II), ambulatory with assistive devices (III), minimally ambulatory (IV) or predominantly use wheelchairs for mobility (V) [3]. Spastic CP is the most common subtype (~ 80% of cases) [4], characterized by spasticity, muscle weakness and impaired selective motor control, all of which affecting the gait pattern and walking ability [5,6]. Muscles in children with CP also have a significant increase in extracellular matrix (ECM), measured via collagen content [7][8][9], histologically [10] and transcriptionally [11][12][13]. Reduced levels of daily physical activity in CP are associated with higher perceived fatigue [5]. Individuals with CP report physical fatigue as one of the main causes of limitation, deterioration and eventually cessation of their walking ability [5,14].
Studies involving functional taskssuch as walking and bimanual movementsshowed increased selective muscle fatigue [15,16] and augmented external mechanical work [17]. Importantly, ambulatory children with CP, show reduced daily walking activity levels compared to typically developing (TD) children [18,19].
Walking capacity in individuals with CP often emerges as a multifaceted interplay of impairments at the neuromuscular, cardiorespiratory, and musculoskeletal systems [6,20,21]. Gross motor function in children with CP measured by the Gross Motor Function Measure (GMFM) [22,23] show changes in functional mobility over time, especially for the more severely impaired [24,25].
While GMFCS is normally stable for ambulatory children, those who are dependent on the use of assistive devices show a more pronounced loss of their ambulatory capacity [24,25]. Adults with CP have a reduction in walking capacity, with many studies reporting decline starting in their 20s and 30s [26]. In addition, age-related physiological changes occur earlier in adults with CP, and the 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62 63 64 65 prevalence of secondary manifestations (e.g., pain, osteoporosis, and musculoskeletal problems) is higher in adults with CP compared to age-matched healthy adults [27][28][29]. Age-related changes in motor function and in gait proficiency is therefore an important clinical marker to monitor physical capabilities of the individual with CP over the years from childhood to adulthood. However, a broad assessment of the relationship between walking efficiency, functional levels, and the natural progression of ambulatory ability across the lifespan for individuals with CP is still poorly understood.
Assessment of oxygen uptake (V O2) during submaximal exercise is a convenient and objective measure to determine walking efficiency in CP and evaluate changes after therapeutic interventions [30][31][32][33]. In addition, the measurement of energy expenditure while walking can provide a quantitative measure to evaluate differences between individuals with CP and age-matched TD individuals as well as to determine effective therapeutic interventions. Currently, a systematic approach primarily focused on studies that investigated energy expenditure by measuring V O2 during walking in population with spastic CP is lacking. Therefore, a comprehensive examination would further the understanding of the degree of impairment in walking energetics in individuals with spastic CP compared to age-matched TD individuals. Given the strong relationship between walking ability and GMFCS levels, it is also clinically relevant to summarize to what extent walking energy expenditure is related to GMFCS levels, and define the magnitude of change of energy expenditure across GMFCS levels.
In summary, the aims of this systematic review were to identify, appraise and synthesize the evidence describing 1) the difference in energy expenditure in walking between individuals with CP and age-matched TD; 2) the relationship between V O2 and GMFCS levels; and 3) age-related changes in walking energy expenditure over time in individuals with CP. Findings will provide further insights into the extent of walking ability and its proficiency in individuals with spastic CP, and outline key evidence gaps for development of future research.  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63 64 65 Methods This systematic review was completed following the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) 2020 statement [34] (S1). The protocol was registered on Prospero (ID: CRD42020146657) [35], where a complete description of the methodology and the complete list of searched terms, the searching process, and data extraction method is available.

Data extraction
Two authors (MN and FR) extracted the pertinent data from the included articles using a customized data extraction form (S4). Study population characteristics (i.e., diagnosis, sample size, age, severity of CP) and details about the protocol used for the evaluationsuch as the type of test (over-ground walking or on a treadmill); testing modality (self-paced, constant speed or incremental test); protocol phases duration and study findingswere summarized and reported in Table II and   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63 64 65 Handbook [39]. Due to the heterogeneity in the primary outcomes among studies, and in some cases also the relatively small sample size (n<30), a random-effect model was fitted to the data for the calculation of the standardized effect size (Hedges' d) [40,41] and 95% confidence interval (C.I.) [42]. An effect size of 0.2 to 0.49 was interpreted as a small effect, 0.5 to 0.79 a medium effect, and over 0.8 a large effect size [43]. Publication bias was assessed by visual inspection of funnel plots, then the Egger's test was performed [44]. Heterogeneity was checked by means of the Higgins Inconsistency test (I 2 ). Values over 50 % were considered of high heterogeneity [45].

Summary of studies
After removing the duplicates, the initial search resulted in 1016 articles matching inclusion criteria. Eight hundred and nineteen articles were removed after title and abstract screening. The reasons for exclusion were: incorrect population (i.e., physical and mental disability other than CP), lack of metabolic data (e.g., biomechanics estimation or questionnaires score), and incorrect testing protocol (e.g., cycle-ergometer, isometric contractions). After the preliminary screening, 197 articles were kept for full-text review. Of these, 41 articles matched all the inclusion criteria ; reasons for exclusion of the additional 156 articles are listed in Figure 1.
All the included studies were screened with the RoBANS [38]. In most of the studies, CP and control groups were selected from comparable population group (80% of the studies). Concerning the confunding variables, these were adequately confirmed and considered during the design phase for 59% of the studies, while it was uncertain whether the confounding variables resulted in a high risk or a low risk of bias for the remaining studies (41%). The experiemental protocols were described, and the outcomes used were valid and reliable in all the studies (100%). Low risk of bias for inadequate blinding of outcome assessments and inadequate handling of incomplete outcome data A detailed summary of the characteristics and design of each study is provided in Table III, while details about protocol, energy expenditure outcomes, results and statistical significance for each study can be found in Table IV.  [49,51], and five other measures of energy cost of walking [49,51,63,81,82]. Of these, five studies reported more than one outcome [49,51,63,81,82].
Only three studies compared metabolic data in people with CP [66,68,82], either across different age ranges [68] or longitudinally on the same individuals [66,82]. One of these studies reported the net O2 cost of walking (ml/kg/m) [66]; two articles reported multiple outcomes [68,82]: both measured walking V O2, one the energy cost (J/kg/m), and one other measures of energy cost of walking (i.e., non-dimensional cost). Five studies addressed multiple comparisons and thus were included in more than one subgroup: four research papers compared CP to TD and also CP across GMFCS levels [49,51,62,82]; one compared CP to TD and also addressed the effect of age in CP [68]; while one study performed all three comparisons of interest [82]. For this reason, the sum of studies in each subgroup differs from the total of forty-one articles. For a more comprehensive overview and comparison of the study results, and for graphical purposes, we convertedwhen possiblethe originally reported values into V O2 (ml/kg/min) and O2 cost (ml/kg/m) using equations and formulae available from the literature (see details in [88]). Previously published V O2 and O2 cost values, and the converted or calculated ones, are shown in Figure 2 and 3, respectively.

Comparison of people with CP and typically developing peers
Thirty-one out of 41 studies compared people with CP and TD. There were discrepancies across studies on group matching modality: some controlled for age, sex and body sizes, others only for age. In several studies the sample size was relatively small (n<30) [90], while in others the enrolled controls were fewer than the experimental CP group.

Age-related differences
Only three studies considered age as potential confounder of V O2 and/or O2 cost during walking [66,68,82]. Two longitudinal studies [66,82] [82]. In the work of Marconi et al. [68] it was found that walking V O2 of individuals with diplegic CP was significantly higher than TD at each age, whereas in individuals with hemiplegic CP it was significantly higher only for the first age group (4-7 years); however, they did not run statistical analysis to determine differences across age groups of the same CP subtype [68].

Meta-analysis TD vs CP
Comparison between TD and CP was feasible for both our main outcomes. Nineteen studies were included for the energy consumption analysis [48,49, [56,61,84]. It is worth to note that the results are consistent among studies when O2 cost, which takes into account walking speed, is considered ( Figure 3 and Figure 6). The study by Rigby et al. [75] stands out for the considerable estimate of Hedges' g (>3). The sample size of the study was relatively small (8 TD, 8 CP) and the CP group was characterized by a medium-to-high severity involvement (4 spastic quadriplegia, 2 spastic diplegia) [75] which may have exacerbated the between-group differences. Visual inspection of both funnel plots did not reveal asymmetry There was no evidence of publication bias for the walking energy expenditure and the energy cost of walking comparisons (p = 0.446 and p = 0.628, respectively).

Across GMFCS
Seven studies were included in the analysis for the V O2 [49,51,63,79,[81][82][83]. The pooled r resulting from all the studies, based on the random-effects model was 0.965 (95% CI: 0.875-0.991; p < 0.001) and exhibited a notable heterogeneity (I 2 = 98.07%). In particular, two studies [79,81] presented estimates much lower than the main bulk of results. In both studies, participants classified as GMFCS III had a lower V O2 during walking than participants classified as GMFCS II. This result could be explained by the reduced number of participants in the most impaired group (n=6) and the characteristics of the protocol (i.e., self-paced walking for 3 minutes) in the study by Slaman et al. [79]. In the study by Thomas 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65 it an outlier was apparent to the present authors. Forest plots for both analyses are displayed as supplementary material (S7). Visual inspection of the funnel plot did not reveal asymmetry for the studies included for the V O2 correlation analysis. however, the number of studies included in the meta-analysis was too low to use tests for funnel plot asymmetry, as suggested in the Cochrane's Handbook (<10 studies) [39]. Visual inspection of the funnel plot did not reveal asymmetry for the studies included for the O2 cost correlation analysis. Egger's test indicated that there was no obvious publication bias (p = 0.268).

Discussion
In the present systematic review, we aimed to 1) appraise and synthetize the difference in energy expenditure in individuals with CP compared to their age-matched peers; 2) identify the rate of V O2 across different severity levels, based on GMFCS; and 3) define the age-related changes of walking energy expenditure in individuals with CP. Forty-one studies met the inclusion criteria.
Specifically, 31 studies compared V O2 during walking between individuals with CP and age-matched healthy subjects, 13 studies reported the changes of metabolic data across GMFCS and only 3 studies were found to investigate the variation of walking energy expenditure related to different ages. It is worth noting that several studies were considered for more than one aim of the current review. in individuals with CP was twofold than the age-matched controls with a range difference from 0.04 to 0.62 ml/kg/m, which resulted with an increased O2 cost ranged from 1.3 to 3.5 times than controls.
The remarkable greater cost of walking in CP respect to TD peers confirms the perceived effort and fatigue reported by children and adults with CP [5,91]. The O2 cost is considered a physiological marker describing the degree of locomotion impairment in pathological conditions such as multiple sclerosis [92], stroke [93], and Parkinson disease [94], and reflecting either an increase in the rate of V O2 during normal walking speed or an abnormal rate of V O2 respect to a reduced walking speed. It is worthwhile to consider that there is a U-shaped relationship between O2 cost and gait speeds, which indicates that there is a particular gait speed minimizing the O2 cost in each individual [95,96]. For this reason, when comparing CP to TD, the selection of walking speed needs to be carefully considered. A similar U-shaped relationship has been hypothesized in individuals with CP, however, to date it has not been confirmed by experimental results [68]. Recently, Schwartz et al.  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63 64 65 [97,98] suggested an interesting way to compare data by means of a non-dimensional measure of energy expenditure, however the use of a non-dimensional outcome may not translate to clinical practice. Therefore, based on our results, we suggest O2 cost as the parameter to physiologically characterize the walking efficiency in people with CP.
The ability to sustain a walking task for long period of time, maintaining an adequate force production with the lower possible O2 cost, is dependent on the integration of the neuromuscular and cardiorespiratory systems. Several factors play a role on one or more of these physiological systems that can lead to insurgence of fatigue and increase the energy cost of a functional task like ambulation.
Among these factors specific have been recognized as the main ones affecting the duration of walking in CP. Neural-driven weakness, defined as a loss of excitatory motor signals descending in the cortico-spinal tract resulting in reduced muscle activation and reduced muscle size [99][100][101], seems to be a major limiting factor in this population. Rose and McGill [102] demonstrated an equivalent ratio between recruitment and firing rate modulation at submaximal contractions between subject with CP and controls. However, they showed that submaximal contractions required more voluntary effort for subjects with CP, such that the neuromuscular activation level corresponded about 50% of the maximum voluntary contraction (MVC) in individuals with CP compared to control whose neuromuscular activation level was related to about 20% of the MVC [102]. This means that a person with CP might require full voluntary effort compared to a submaximal effort for the healthy control.
This reduced force-generating capacity of the muscle in individuals with CP might result with high relative demand of lower limb skeletal muscles during walking, making these individuals more prone to fatigue. These results are consistent with findings showing lower muscle endurance in quadriceps muscles of CP undergoing a submaximal repetition-to-fatigue protocol compared to control subjects [103]. In addition, it has been recently shown that the decrease of EMG median frequency and increase of EMG amplitude in lower leg muscles were larger in children with CP compared to typically developing peers during overground walking at self-selected speed [15].  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65 Greater energy cost during walking in CP has been associated with higher external mechanical work [86,87], which has been associated with greater potential, vertical and later kinetic mechanical work [17]. The higher mechanical work was proposed to be related to the equinus gait pattern commonly seen in children with CP due to a less effective exchange between potential and kinetic energy by the legs to lift and redirect the center of mass [17, [104][105][106]. Furthermore, it has been recently shown that reduced knee and hip joint extension are associated with gait inefficiency in children and adolescents with CP [107]. However, these results are in contrast with the work of Steele et al. [80] who found the crouch gait severity correlated poorly with elevated V O2 in in children with CP. Furthermore, it has also been recently reported that long-term reduction of spasticity by selective dorsal rhizotomy does not lead to reduced oxygen consumption [108].
Abnormal muscle activation patterns of agonist-antagonist, higher levels of co-contraction and impaired selective motor control are typical clinical signs of damaged corticospinal projections in CP [2,109,110], which have been shown to additionally contribute to gait abnormalities in this population [6]. It has been suggested that these impaired neural mechanisms of muscle activation might further contribute to the early manifestation of fatigue in CP [111]. Unnithan et al. [84] found that co-contraction of the quadriceps and hamstrings explained 51% of the variance in gross V O2 among children with CP walking on a treadmill [84]. However, Damiano et al. [112] found the opposite relationship, with greater co-contraction between the quadriceps and hamstrings related to a lower energy expenditure index [112], and Steele et al. [80] showed that co-contraction of the rectus femoris and biceps femoris only explained 2-3% of the variance in V O2 during gait in children with bilateral CP. Moreover, despite the fact that the impaired selective motor control has showed a strong correlation with severity scales and gait abnormalities in CP [113][114][115][116], it is still uncertain whether decreased selective motor control is correlated with higher oxygen consumption during gait.
Interestingly, the respiratory exchange ratio has been found equivalent between individuals with CP and healthy controls, demonstrating similar cardiorespiratory responses in both group during submaximal exercise [76,117]. Though, it has been speculated that spastic muscles could cause local obstruction of venous return and, therefore, result in inhibition of muscle lactate clearance leading to increased acidity and local muscular fatigue [61,118]. However, no evidence has been provided to support this hypothesis, as well as it has yet been investigated the mechanism underlying skeletal muscle oxidative capacity in CP and its contribution to exercise intolerance and fatigue.
Muscle metabolic factors might also reflect the increased O2 cost of movement in children with CP. Force generation is highly energetic and requires the constant replenishment of adenosine  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65 level and the O2 expenditure compared to V O2. The increase of O2 cost was more accentuated between GMFCS II vs III (+56 %) compared to GMFCS I vs II (+38 %). Only three studies reported the O2 cost for GMFCS IV, therefore they were not considered in the regression analysis with the other GMFCS levels. However, the results for these studies reported a trend of remarkable increased of O2 cost in comparison of GMFCS III. Moreover, it is worth noting that the least difference (+22%) on cost was found between GMFCS I and TD age-matched controls considering both the originally measured and estimated values. This is not surprising since according to GMFCS individuals classified as level I can walk and run without any particular limitations that could impact their participation to daily activities [3].
As far as the age-related changes of walking energy expenditure in CP, we identified three studies that satisfied the search inclusion criteria [66,68,82]. However, only two studies analyzed and compared the cost of walking at different ages or over time [66,82]. Both studies reported an average increase of cost of walking in children with CP over a time period of 12 [82] or 31 months [66]. In comparison of the healthy control peers, Thomas et al. [82] found that all the GMFCS levels (I, II and III) had an increase in O2 cost over one year. Nevertheless, a lack of statistical difference was found in the magnitude of O2 cost increment by the GMFCS levels, which could had been influenced, as stated by the authors, by the short time period considered and the small sample size recruited for GMFCS levels I and II. The study by Kerr 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65 levels. There is still a knowledge gap on the changes of O2 expenditure during walking in individuals at the early and late middle age with CP compared to age-matched unimpaired population. Further cross-sectional and longitudinal studies with larger sample size are needed to assess the trend of energy expenditure of walking with a broader range of ages classified at different GMFCS levels.
Additionally, the wide range in energy expenditure and cost of walking values, and sometime inconsistent results, points out the need for more standardized protocols in clinical and experimental settings, as well as encouraging for large multicenter studies. This would provide a more comprehensive understanding of the progress of walking efficiency in individuals with CP at different severity levels from childhood to the adulthood.

Conclusion
The results of this review demonstrate a meaningful higher energy expenditure and energy cost during walking in individuals with CP despite a variability in the experimental protocols and testing type. A strong association between walking inefficiency and gross motor function was found across studies with a noticeable increase of cost of walking for GMFCS levels II and III. The analysis of the studies suggests a preference for using the O2 cost as a physiological parameter to assess walking efficiency in CP. Due to a limited number of studies, partially with small sample sizes, the impact of age-related changes on walking efficiency with different functional severity remains still undetermined, as well as the trend of these longitudinal changes across the lifespan for individuals with CP.   14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64 65

Conflict of interest statement
The authors have no conflict of interest.

Acknowledgement
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Recalculated values
Walking energy consumption

Study selection
16a Describe the results of the search and selection process, from the number of records identified in the search to the number of studies included in the review, ideally using a flow diagram.
Page 9-11; Figure 1 16b Cite studies that might appear to meet the inclusion criteria, but which were excluded, and explain why they were excluded.

Search strategy in Scopus:
((cerebral palsy) AND (fatigue OR "energy metabolism" OR "metabolic cost" OR "energy cost" OR VO2 OR endurance OR "energy expenditure" OR "aerobic capacity" OR "oxygen consumption"))

Search strategy in Web of Science:
((cerebral palsy) AND (fatigue OR "energy metabolism" OR "metabolic cost" OR "energy cost" OR VO2 OR endurance OR "energy expenditure" OR "aerobic capacity" OR "oxygen consumption"))

Search strategy in ScienceDirect:
((cerebral palsy) AND (fatigue OR "energy metabolism" OR "metabolic cost" OR "energy cost" OR VO2 OR endurance OR "energy expenditure" OR "aerobic capacity" OR "oxygen consumption")) S3: Scores for methodological quality assessment for included studies (questions reported below). S7a: Forest plot with pooled correlation coefficients (r) and C.I. for walking energy consumption across GMFCS levels.