A prospective assessment of gait kinematics and related clinical examination measures in cerebral palsy crouch gait

Background While prospectively assessed crouch gait in cerebral palsy (CP) does not necessarily progress, prospective changes in clinical examination measures have not been reported. This study prospectively examined the association between selected clinical examination variables and change in crouch gait in a cohort with bilateral CP. Methods Inclusion criteria were a diagnosis of ambulant bilateral CP, knee flexion at mid-stance >19 0 and a minimum of two-years between gait analyses. The change in kinematic variables was assessed using Statistical Parameter Mapping (SPM) and changes in clinical measures using appropriate paired tests. Linear regression examined the association between progression of crouch and clinical examination variables. Results There was no mean change in crouch in 27 participants over 3.29 years. However, there was significant variability within this group. Clinical hamstring tightness (60.00 0 to 70.48 0, p<0.01) and external knee rotation during stance (SPM analysis, p<0.001) increased but there was no association between changes in clinical examination variables and changes in crouch (p-values 0.06 - 0.89). Conclusions This prospective study found no association between the changes in clinical examination variables and changes in crouch highlighting the likely multi-factorial aetiology of this gait pattern and the need for larger prospective studies. The variability crouch gait progression among the 27 participants highlights the pitfall of group mean values in such a heterogeneous population.


v2 Abstract Background
While prospectively assessed crouch gait in cerebral palsy (CP) does not necessarily progress, prospective changes in clinical examination measures have not been reported. This study prospectively examined the association between selected clinical examination variables and change in crouch gait in a cohort with bilateral CP. Methods Inclusion criteria were a diagnosis of ambulant bilateral CP, knee flexion at mid-stance >19 0 and a minimum of two-years between gait analyses. The change in kinematic variables was assessed using Statistical Parameter Mapping (SPM) and changes in clinical measures using appropriate paired tests. Linear regression examined the association between progression of crouch and clinical examination variables.

Results
There was no mean change in crouch in 27 participants over 3.29 years. However, there was significant variability within this group. Clinical hamstring tightness (60.00 0 to 70.48 0 , p<0.01) and external knee rotation during stance (SPM analysis, p<0.001) increased but there was no association between changes in clinical examination variables and changes in crouch (p-values 0.06 -0.89).

Conclusions
This prospective study found no association between the changes in clinical examination variables and changes in crouch highlighting the likely multi-factorial aetiology of this gait pattern and the need for larger prospective studies. The variability crouch gait progression among the 27 participants highlights the pitfall of group mean values in such a heterogeneous population. This study prospectively examined the association between selected clinical examination variables and change in crouch gait in a cohort with bilateral CP.

Methods
Inclusion criteria were a diagnosis of ambulant bilateral CP, knee flexion at mid-stance >19 0 and a minimum of two-years between gait analyses. The change in kinematic variables was assessed using Statistical Parameter Mapping (SPM) and changes in clinical measures using appropriate paired tests. Linear regression examined the association between progression of crouch and clinical examination variables.

Results
There was no mean change in crouch in 27 participants over 3.29 years. However, there was significant variability within this group. Clinical hamstring tightness (60.00 0 to 70.48 0 , p<0.01) and external knee rotation during stance (SPM analysis, p<0.001) increased but there was no association between changes in clinical examination variables and changes in crouch (p-values 0.06 -0.89).

Conclusions
This prospective study found no association between the changes in clinical examination variables and changes in crouch highlighting the likely multi-factorial aetiology of this gait pattern and the need for larger prospective studies. The variability crouch gait progression among the 27 participants highlights the pitfall of group mean values in such a heterogeneous population.

Open Peer Review
Approval Status

Introduction
In cerebral palsy (CP), crouch gait and flexed knee gait are largely synonymous terms referring to excessive knee flexion in stance phase 1 and are among the most common pathological gait patterns in CP 2 . A systematic review on the progression of crouch gait in CP suggested that, in the absence of surgical intervention, knee flexion progresses over time 3 . However, our recent prospective follow-up of crouch gait in CP 4 found that crouch gait does not necessarily progress over time and on assessment every six-months, participants demonstrated episodes of both increasing and decreasing crouch. These contrasting findings appeared to be a function of prospective versus retrospective study designs. The majority of the studies included in the systematic review were retrospective cohort studies 5-8 ) likely based on data collected from those referred for clinical gait analysis. As referral for clinical gait analysis is often as a result of a deterioration in function, pain or to aid with planning of required intervention, it is not surprising, therefore, that retrospective cohort studies based on such data demonstrate deterioration in gait between repeated analyses. The need for clinically based prospective studies of walking in CP has previously been highlighted 9 and the findings of our recent prospective study of crouch gait 4 supports this.
That prospective study focussed only on the changes in knee flexion at mid-stance every six months and did not report on contemporaneous changes in other relevant gait kinematic variables or clinical examination measures. This was largely because the clinical examination variables in particular were unlikely to change sufficiently over six-monthly assessments to allow meaningful analysis, particularly when considering measurement error. By examining the first and last assessment only, the current study aims to examine the longer-term changes in crouch gait and the associated changes in relevant clinical examination and kinematic variables. Establishing which clinical examination variables are associated with progression of crouch gait has important implications in terms of treatment, and potential prevention, of crouch gait.
The most relevant kinematic and clinical examination variables are those potentially implicated in the causation of couch gait which is acknowledged to be multifactorial and due to a combination of muscular, neurologic, and/or bony pathologic processes 10 . Tightness in the knee flexors has been implicated, and crouch is often treated with surgical release of the hamstrings 11,12 . However, muscle-length modelling has suggested that most subjects with crouch gait have hamstrings of normal length or longer 13 . While it has previously been suggested that hamstring tighten secondary to crouch gait rather than causing this gait pattern 3 , this was based on retrospectively analyzed gait data and a prospective examination of the relationship between crouch gait and hamstring tightness has not previously been reported. In addition to hamstring and knee flexor tightness, studies have also demonstrated that those in crouch are likely to have psoas lengths shorter than normal during gait 13 and that ankle plantarflexor contracture may contribute to excessive knee flexion during gait 14 . Dynamically, during gait, external tibial torsion has been shown to reduce the capacity of muscles to extend both the hip and knee 15 . Therefore, lever arm dysfunction (LAD), which refers to faulty skeletal alignment 16 , is thought to contribute to crouch gait. Excessive external tibial torsion with excessive internal hip rotation due to femoral anteversion is the classic form of LAD in CP 17 . Therefore, based on existing literature the clinical examination variables examined in this study were tightness of the knee flexors (knee flexion contracture and hamstring tightness), tightness of the hip flexors (hip flexion contracture), tightness of the ankle plantarflexors (gastrocnemius tightness, soleus tightness) and LAD (femoral torsion, tibial torsion).
The aim of this study was to examine the prospective progression of crouch gait over a minimum of two years along with the contemporaneous changes in clinical examination measures of muscle tightness and estimates of bony torsions in a cohort of ambulant participants with CP crouch gait and to assess the association between progression in crouch gait and these variables.

Methods
Approval for the study was obtained from the Research Ethics Committee of the Central Remedial Clinic and written, informed consent was obtained from all parents/guardians. Participant recruitment and kinematic data collection methodology were as previously described and has been previously published 4 . Inclusion criteria were a diagnosis of bilateral, spastic CP, GMFCS level I-III, age 4-17 years at first analysis and crouch gait. Crouch gait was defined as knee flexion greater or equal to two standard deviations above laboratory reference values at mid-stance in at least one limb. This equated to mid-stance knee flexion greater or equal to 19°. Participants were excluded if they had surgery within one year prior to the first gait analysis. Sex was as recorded in the participants' medical file.
Kinematic data were captured using a four camera Codamotion cx1 active marker system (Charnwood Dynamics, Leicestershire, UK) at a rate of 200 Hz using a modified Helen Hayes marker protocol 18 . A dedicated, multisegment foot model was not used in this study and the foot was modelled as a straight-line segment from the base of the fifth metatarsal to the lateral aspect of the calcaneus. Therefore, only sagittal plane foot data (dorsiflexion/plantarflexion) were analysed. All participants walked in barefoot at a self-selected speed. Participants walked independently where possible. If independent gait was not possible data were collected while the participant walked with the assistance of two hands held in front by a physiotherapist as is standard practice in the gait laboratory. A minimum of

Amendments from Version 1
This version has addressed the points raised by the reviewers.
Clarifications have been added around a number of points and more detail added to teh methodology as required. The conclusions both in the abstract and main text now better reflect the study findings. Figure 1 now contains all kinematic graphs.
Any further responses from the reviewers can be found at the end of the article REVISED four walking trials was recorded for each participant and, as per clinical practice in the laboratory, one representative walking trial was chosen based on a qualitative evaluation by the examining clinician.
Additional to the previously described barefoot kinematic data collection procedures, relevant clinical examination measures of muscle tightness/contracture and bony torsions were collected, and all data collection (kinematic and clinical) was by the same examiner on each occasion. Popliteal angle measure of clinical hamstring tightness was recorded in supine with the contralateral hip and knee flexed to align the pelvis to neutral, the recorded measure was degrees from full extension. Hip flexion contracture was similarly assessed in supine with the pelvis in neutral. The hip was passively extended to the limit of motion and the recorded value was lack of extension to neutral. Knee flexion contracture was also recorded in supine as lack of full extension. Gastrocnemius tightness was assessed with the knee extended and recorded as the angle between the longitudinal line of the tendo-achilles and the longitudinal axis of the calcaneus. Soleus tightness was similarly measured with the knee flexed. Femoral torsion was estimated in prone lying using the trochanteric prominence angle test 19 . Tibial torsion was estimated by measuring the transmalleolar angle corresponding to the angle between the line connecting the medial and lateral malleoli and a line perpendicular to the long axis of the thigh 20,21 .
The previously described cohort were assessed at 6-monthly intervals 4 . For this study, only the first and last assessments were analysed in those who had a minimum of two years between their first and last assessments. The longer-term changes in relevant kinematic and clinical examination variables between those two assessments were examined. A number of participants progressed to surgical intervention during the course of the study and so did not have two years between their first and last pre-operative assessments. This group were not included in the present analysis but their data are included in the Extended Data for completeness (Figure S.1; Table S.1).
To avoid dependence between sides, only the side which was most flexed during gait at first assessment was chosen for subsequent analysis 12,22 . All clinical examination variables reported were first tested for normality using a Shapiro-Wilks test. Normally distributed data were summarised using means and standard deviations. Otherwise, medians and inter-quartile ranges were reported. Differences in clinical examination variables were assessed using paired t-tests for normally distributed data. Otherwise, Wilcoxon Signed-Ranks tests were used. The presence of a hip/knee flexion contracture was defined as a clinically measured contracture ≥ 5° and changes in the prevalence of flexion contractures were assessed using Pearson's chi-squared test. Significance was set at p<0.05 in all cases.
Gait kinematic data were analysed using Statistical Parametric Mapping (SPM) (SPM1d version 0.4, available for download at http://www.spm1d.org/) in MATLAB (The Mathworks Inc., Natick, M.A., 2015). Normality of kinematic data was checked using a built-in SPM function in MATLAB. Differences between groups for lower limb kinematics were assessed using a two-sample parametric two-tailed t-test. Significance level for all analyses was set at p<0.05.
Relevant participant, clinical and kinematic variables were further summarised separately for those who demonstrated increased, decreased, and unchanged knee flexion at mid-stance over time. A threshold of 5° was chosen to represent a true change in the value of knee flexion at mid-stance 23 . As group numbers within these categories were relatively small, formal statistical tests were not used. The association between the progression of crouch gait and relevant clinical examination variables was assessed using repeated liner regression using each relevant demographic or clinical examination variable as a predictor of change in knee flexion at mid-stance.

Results
A total of 48 participants were initially recruited as previously described 4 and of those, 27 participants had a minimum of two years (mean 3.29±0.49 years) between their first and last assessment and were included in the current analysis.
As per the inclusion/exclusion criteria, none of the participants had orthopaedic surgery within one year of initial analysis. Six participants had previous single-event multi-level surgery (defined as having at least two orthopaedic surgeries at a single surgical event 24 ) at a mean age of 10.03±2.60 years and a mean of 3.97±2.64 years prior to initial assessment. Thirteen participants progressed to surgical intervention without having the minimum of two years between first and last analysis while the remaining eight participants were lost to follow up or did not complete the required assessments.
Among the 27 included participants, there was no significant difference between the mean knee flexion at mid-stance at baseline (23.46°±5.04°) and final assessment (23.71°±13.01°) (p=0.91). The participant characteristics for the 27 participants are summarised in Table 1 below. There was no significant change in mean knee flexion at mid-stance in the 3.29 years between baseline and final assessment (23.46°±5.04° and 23.71°±13.01° respectively; p=0.91) and the changes in kinematic graphs between the first and last analysis are shown in Figure 1.
There was no significant change in any sagittal plane kinematics, including knee flexion, between the first and last assessment. The only significant change was in the transverse Table 1. Characteristics of the included participants (n=27).

Variable
Age at first assessment (years) 10.43  plane knee graph which highlights a significant increase in external knee rotation during stance phase (p<0.001). While both the mean value of knee flexion at mid-stance and SPM kinematic curves highlights that there was no overall change in knee flexion between initial and final assessments, the relatively large standard deviation associated with knee flexion at mid-stance at final assessment (23.71°±13.01°) highlights that some participants increased in crouch during the study window. Therefore, relevant participant, clinical and kinematic variables are further summarised in Table 3 for those who demonstrated increased, decreased and unchanged knee flexion at mid-stance over time. A threshold of 5° was chosen to represent a true change in the value of knee flexion at mid-stance 23 . Table 3 confirms that the largest group of participants (48%, n=13) did not show a significant change in knee flexion at

Increased crouch (>5°); n=7
Unchanged crouch (±5°); n=13 Decreased crouch (<5°); n=7 mid-stance while smaller, equal groups (26%, n=7) demonstrated increase and decrease in crouch over time. While group sizes are small, this data might suggest that those who increased in crouch were older at initial assessment, had tighter hamstrings at initial assessment and had a larger increase in Tibial Torsion compared to the groups who either demonstrated a decrease in crouch gait or did not change. However, the association between change in crouch gait and these variables is summarised in Table 4. Regression analysis using relevant clinical examination variables as predictors of change in knee flexion at mid-stance found no significant associations.

Discussion
The aim of this study was to prospectively examine the progression of CP crouch gait along with the contemporaneous changes in clinical examination measures of muscle tightness and bony torsions and to examine any association between the change in crouch gait and these variables. Despite no overall change in the mean value of crouch gait in the 3.29 years between initial and follow-up analyses, there was a significant increase in mean clinical examination measure of hamstring tightness and knee eternal rotation during gait suggesting that neither of these variables contribute to crouch progression. Further regression analysis confirmed this and found that clinical examination variables were not associated with change in crouch gait. This prospectively confirms that tightening of the hamstrings does not contribute to crouch progression, is part of the natural history of children/adolescents with CP and occurs over time.
SPM analysis found no change in the majority of gait kinematic graphs, including knee flexion/extension, over two analyses 3.29 years apart. This was apart from an increase in knee rotation. Increasing external knee rotation during gait is often assumed to be part of the natural history of gait in CP and while a recent retrospective review of hamstring surgery in CP 25 reported similar in a small non-surgical control group (n=15), this is the first study to prospectively document this. Known issues in reliability of transverse plane kinematic data must be considered when interpreting these results. However, our laboratory has demonstrated good levels of reliability of knee transverse plane kinematics (~2.8°) 26 and we are confident in reporting an increase over time.
External rotation of the tibia is thought to be associated with LAD which in turn contributes to crouch. Despite the change in dynamic knee rotation evident on the kinematic graphs, the present results did not demonstrate a similar increase in mean clinical measure of tibial torsion over the same time period. Longer term prospective study might ascertain if more reliable radiological assessment of bony torsion also increases in time and as above, a more detailed foot model might establish the potential role of foot posture contributing to the reported increased dynamic knee rotation.
Perhaps as significantly, our results highlight that reporting mean data only does not capture the significant variability in CP gait data and can hide potentially important findings. There was no change in either the mean value of knee flexion at mid-stance or the overall mean knee flexion/extension kinematic curves but there was significant variability within the overall group. While the majority of the 27 participants demonstrated unchanged knee flexion at mid-stance (n=13; 48%), equal proportions demonstrated both significant increase (n=7; 26%) and decrease (n=7; 26%) in this kinematic variable. This highlights the need to examine, and report, data in detail in studies on gait in CP to reflect the heterogeneity of this cohort and suggests that statistical analysis based on mean changes only may not sufficiently capture change. Table 3 provides a further summary of these groups. The data suggest that those who demonstrated an increase in crouch gait had slightly older age and tighter hamstrings at baseline along with a larger change in clinical estimation of tibial torsion compared to those who demonstrated unchanged or decreased crouch. Each group demonstrated a similar increase in tightness in the hamstrings over time regardless of the changes in knee flexion at mid-stance. Likewise, the initial value, or severity, of crouch at baseline analysis did not appear to be different among the groups. However, it must be highlighted that the numbers in these groups are small (n=7, 7 and 13) and more robust regression analysis on the whole study cohort found no association between progression in knee crouch and hamstring tightness or age.
It is notable that the mean value of knee flexion at mid-stance was in excess of 20° at baseline. This value has been suggested as a risk factor for accelerated progression of crouch and as a potential indicator for surgery 3 . Based on these prospective findings, a value of 20° knee flexion at mid-stance cannot necessarily be taken as an indicator for likely progression of crouch gait and the variability of crouch progression seen in our data confirms this is a more nuanced issue.
There are some limitations and suggestions for further study that should be considered when interpreting the results of these results. This was a prospective assessment of a cohort who were already in crouch at baseline. Therefore, we cannot comment of how crouch gait initially developed in this population and likewise do not know if clinical hamstring tightness and knee external rotation during gait increase over time in a non-crouch population. The study participants were primarily GMFCS levels I and II and only two participants were GMFCS level III. This means the results of this study can only be generalised to the more functional CP population (GMFCS I and II). The time period between the first and last assessments (mean 3.3 years) is still relatively short though as our results suggest, participants with ambulant, bilateral CP will likely undergo single event multi-level surgery or other surgical intervention during the course of a prospective study making long term prospective follow up of natural progression difficult.
As an example, Church et al. 27 examined the change in flexed-knee gait over 12 years and reported improvements in mean knee flexion. However, 75% of participants had orthopaedic surgery between analyses and those who did not have surgery were not reported separately. Likewise, Kanashvili et al. 28 reported a decrease in knee flexion during gait over a 13-year time period in a cohort in crouch on initial presentation but again, all participants had intervening study so no conclusions can be drawn on natural progression. It is noted that comparison of the present results with other studies can be limited due to differing definitions and criteria used to define crouch gait. We used a definition of knee flexion greater or equal to two standard deviations above laboratory reference values at mid-stance and have previously highlighted the need for consistency in definitions and criteria in such studies 1 . Our results show that while the mean value of crouch did not change, there was variability within the group which may be lost when examining group means only. Gait analysis was carried out in barefoot as is standard clinical practice. However, many of the study cohort would have been supplied with orthoses which would be worn routinely for at least part of the day. The influence or impact of such orthoses on the progression of crouch gait was not considered here. Likewise, as is the case in many studies on CP gait, non-surgical interventions (botulinum toxin, orthoses, therapy interventions) was not considered here and it may be that these interventions play a role in the variability of gait progression and should be carefully considered in future studies. Likewise, the potential role of foot posture in the progression of crouch gait 29 was not fully examined in the current study and our data collection protocol did not include a multi-segment foot model. Our data suggest that those who demonstrated an increase in crouch had a larger change in clinical estimation of tibial torsion and future studies might also include regular, more reliable radiological measures of relevant bony torsions.

Conclusions
This prospective study found no association between the changes in relevant clinical examination variables and changes in crouch gait highlighting the likely multi-factorial aetiology of this gait pattern and the need for larger prospective studies.
Our results further highlight the need for larger, prospective studies of gait development in CP. We would suggest that a future prospective study beginning at the onset of gait with routinely scheduled follow-up assessments not dependent on clinical referral would be extremely useful and worthwhile. A study beginning at a young baseline age would allow the development of many pathological gait patterns and associated clinical examination variables to be examined while regularly scheduled research assessments would remove any bias in terms of clinical gait analysis potentially being primarily in response to a deterioration or clinical need.

Introduction
The main point in your introduction is that sampling bias likely led to misinterpretation as to the rate of crouch gait progression in the CP population. Please describe, in this manuscript, how the prospective sample was recruited (i.e. were these individuals drawn from a population that was already referred to the gait lab). If they were drawn from a population already referred to the gait lab, how is this prospective sample any different from those evaluated via -as per your introduction -'inferior' retrospective study designs (i.e. "referral for clinical gait analysis is often as a result of a deterioration in function, pain or to aid with planning of required intervention"). If these individuals were NOT recruited from those already referred to the gait lab, can you be certain that collection/capture of gait analysis data did not alter future treatment decisions regarding crouch gait (i.e. treatment plan prior to gait lab might have changed to include different interventions based on the additional gait lab data collected as part of this study, which otherwise would not have been performed)? Even if these individuals were recruited from individuals within the hospital system, doesn't that also suffer from the same pitfalls as you previously described for retrospective analyses (i.e. patients are often at the hospital for as a "result of a deterioration in function, pain or to aid with planning of required intervention"). Finally, if the 'prospective' part of the study was simply to follow-up with all individuals who were initially referred to the gait lab, the applicability of these results becomes much narrower (e.g. change/no change in crouch gait applies only to those individuals undergoing standard clinical practice following referral to the gait lab, and don't necessarily reflect any other CP population).

Methods
You only utilize a single 'representative' trial despite collecting four for each individual? Why not utilize all 4 trials to give a more robust assessment of variability at each timepoint? A rhetorical question would be: If you are only going to use one trial, why do you even bother collecting four? The significant methodological issue when using a single trial is that you don't have the ability to quantify any inter-session variability. It's especially problematic in a population that commonly has large stride-to-stride variability. As such, how is it possible to draw robust conclusions about improvement/worsening of crouch on an individual basis (like you go on to do) when only utilizing a single trial per visit?

Results
This study is framed as a 'natural progression of crouch', but it is not clear that this is the case. Please include all details of the study group relevant to this study rather than referring back to your previous publication. How were these individuals recruited (gait lab referral, hospital population, general population), what general procedures were performed prior to study participation, what procedures were performed during the study period, how many of these individuals have AFOs, etc.
Related to the previous comment, in is unclear (and somewhat confusing) to me based on what is written in the first paragraph of the results section which participants were included and what types of intervention were performed at what point in time before and during the experimental time period. For example, regarding the statement "Thirteen participants progressed to surgical intervention without having the minimum of two years between first and last analysis while the remaining eight participants were lost to follow up or did not complete the required assessments.". Are these 13 included in the 27 or are they part of the 48 that weren't included in this study? It's also important to know what procedures (in bulk) the 27 individuals had during the course of the study.
Were changes in knee rotation towards or away from typically developing norms? Regardless of direction, do you think changes in transverse plane rotations (both kinematic changes and linear regression modeling) provide strong enough evidence to question the belief that lever arm dysfunction has a significant effect on crouch gait?
You mention large variability in knee flexion suggests differing group responses (worsened, same, improved). How can you be certain that this variability is true and not a consequence of only using one trial per visit ("one representative walking trial was chosen for further analysis"). In other words, these results could simply be caused by trial-to-trial gait variability (e.g. picked a single trial that happened to be the high end relative to the others for visit 1 and a trial that happened to be on the low relative to the others for visit 2, while overall mean values of all trials for both visits are the same).

Discussion
Related to previous comment, the "natural history" statement in "This prospectively confirms that tightening of the hamstrings does not contribute to crouch progression, is part of the natural history of children/adolescents with CP and occurs over time." implies that there were no treatments provided to any of these individuals during the study period. A purist might even say that any treatment provided at any time to these individuals (even prior to recruitment to this study) should prohibit the use of 'natural history' since any intervention *might* have the potential to alter body development. Please explicitly list out what general interventions this sample population underwent prior to and during the study period. This would help readers better understand the context in which to evaluate these results and conclusions.
You mention 20 deg is a commonly utilized 'tipping point' for worsening of crouch, and you concluded from the results of this study that it doesn't necessarily mean that this is the case. I understand that you only studied individuals with significant crouch, but I wonder if future studies wouldn't benefit from looking at crouch in CP more broadly, especially considering the high rate of crouch gait in CP, in general. Instead of 20 deg of crouch being a tipping point where PE and kinematic changes start to occur, maybe physiological changes begin to occur much earlier, say from 0-10 deg of crouch, and then don't really change after that. My point is, why not focus on the bigger picture, what causes crouch or what changes occur during early onset of crouch rather than hyper-focusing on what causes bad crouch to get worse.
You state that "our results highlight that reporting mean data only does not capture the significant variability in CP gait data and can hide potentially important findings." But isn't this also true for single trial data (i.e. 'Reporting single trial data does not capture the significant variability in CP gait data and can hide potentially important findings.')? I don't argue with the potential that subgroups can exist that have different responses than the overall group, but it seems naïve to pretend that comparing single gait trials at two time points gives a robust picture of gait kinematics, especially considering the large stride-to-stride variability often observed in individuals diagnosed with CP.

If applicable, is the statistical analysis and its interpretation appropriate?
Please specify which SPM tests you used for your analyses.
Results: Figure 1: I know the additional figures are in the supplementaty material but I believe it would be better to report pelvic, hip and knee angles in all three planes in the main manuscript. Figure 1: Hypothesis test waveforms take a lot of space and could be included in the kinematic waveform subplot, e.g. horizontal bars indicating significant differences below the kinematic waveforms. From my point of view it would make the figure nicer. However, this is my own preference and I leave it up to the authors to decide how they will present the SPM results in this figure. Table 2: Why is hamstring tightness from the first assessment reported as median (interquartile range) while the last assessment is presented as mean (standard deviation)? Please use the same method to make it easier for the reader to compare both assessments. Table 3: Label at the top is wrong. '>50' should be '>5°', etc. Furthermore, having the same numbers in brackets for increased and decreased crouch was a bit confusing for me. Might be better to report the numbers in brackets as >5°, +/-5° and <5°.
Furthermore, the authors could consider to perform statistical test between these 3 groups.
In general, I believe the additional analyses are interesting. I, however, suggest to mention the additional analyses in the method section. Discussion: Third paragraph: Further thought from my side which might be worthwhile mentioning: Clinical measures of bony torsion are not very reliable. I would recommend long term prospective studies based on medical images, which enable a more reliable and accurate estimation of bony torsion.
Fourth paragraph: Very interesting findings. Change in tibia torsion was larger in the increased crouch gait group which might be worthwhile to mention/discuss in this paragraph.
The lack of medical images to quantify bony torsion is another limitation of this study.

Conclusion:
I agree with the content of this paragraph, but I would have expected that the conclusion paragraph is more related to the findings/data from this study.
Are all the source data underlying the results available to ensure full reproducibility?
I answered 'partly' to this question because the raw data, i.e. c3d files from the motion capture data are not provided. However, this is common within the communuity and I do not have a problem with it.
small study population of 27 individuals?
Introduction: Please clarify why it's important to examine associated changes in clinical examination and crouch? Add references to how those variables are associated with crouch.

Methods:
You refer to your previous study for description of data collection. In that study-the participants had an examination every 6 months: was that the case also for this study? Please clarify this.
○ You used two SD or more as a definition of crouch. In your previous study -you did not consider ankle kinematics and therefore referred to "flexed knee gait" instead of crouch. As far as I can tell, you used the same participants in this study -So how do you explain the different use of "crouch" definition? ○ Results: I miss a summary of level of crouch at first assessment. ○ Did the participants receive any specific interventions to improve gait between assessments? If they were referred to a gait lab, is it likely that their coming interventions were based on the results of the analysis? Please discuss this and how it might affect the results.

Conclusion:
Does not answer to your aim to analyse the association between clinical variables and crouch. You found no association? ○ References: time in a group of individuals diagnosed with CP.

General Comments
It is a bit unclear to me what the actual history is of the study group. Details critical to understanding the implications of this work should be included in this manuscript and not referenced to previous work. Missing information on recruitment pool, pre-study interventions, interventions performed during the study period (surgical and non-surgical), etc.

Introduction
The main point in your introduction is that sampling bias likely led to misinterpretation as to the rate of crouch gait progression in the CP population. Please describe, in this manuscript, how the prospective sample was recruited (i.e. were these individuals drawn from a population that was already referred to the gait lab). If they were drawn from a population already referred to the gait lab, how is this prospective sample any different from those evaluated via -as per your introduction -'inferior' retrospective study designs (i.e. "referral for clinical gait analysis is often as a result of a deterioration in function, pain or to aid with planning of required intervention"). If these individuals were NOT recruited from those already referred to the gait lab, can you be certain that collection/capture of gait analysis data did not alter future treatment decisions regarding crouch gait (i.e. treatment plan prior to gait lab might have changed to include different interventions based on the additional gait lab data collected as part of this study, which otherwise would not have been performed)? Even if these individuals were recruited from individuals within the hospital system, doesn't that also suffer from the same pitfalls as you previously described for retrospective analyses (i.e. patients are often at the hospital for as a "result of a deterioration in function, pain or to aid with planning of required intervention"). Finally, if the 'prospective' part of the study was simply to follow-up with all individuals who were initially referred to the gait lab, the applicability of these results becomes much narrower (e.g. change/no change in crouch gait applies only to those individuals undergoing standard clinical practice following referral to the gait lab, and don't necessarily reflect any other CP population).

Methods
You only utilize a single 'representative' trial despite collecting four for each individual? Why not utilize all 4 trials to give a more robust assessment of variability at each timepoint? A rhetorical question would be: If you are only going to use one trial, why do you even bother collecting four? The significant methodological issue when using a single trial is that you don't have the ability to quantify any inter-session variability. It's especially problematic in a population that commonly has large stride-to-stride variability. As such, how is it possible to draw robust conclusions about improvement/worsening of crouch on an individual basis (like you go on to do) when only utilizing a single trial per visit?

Results
This study is framed as a 'natural progression of crouch', but it is not clear that this is the case. Please include all details of the study group relevant to this study rather than referring back to your previous publication. How were these individuals recruited (gait lab referral, hospital population, general population), what general procedures were performed prior to study participation, what procedures were performed during the study period, how many of these individuals have AFOs, etc.
Related to the previous comment, in is unclear (and somewhat confusing) to me based on what is written in the first paragraph of the results section which participants were included and what types of intervention were performed at what point in time before and during the experimental time period. For example, regarding the statement "Thirteen participants progressed to surgical intervention without having the minimum of two years between first and last analysis while the remaining eight participants were lost to follow up or did not complete the required assessments.". Are these 13 included in the 27 or are they part of the 48 that weren't included in this study? It's also important to know what procedures (in bulk) the 27 individuals had during the course of the study.
Were changes in knee rotation towards or away from typically developing norms? Regardless of direction, do you think changes in transverse plane rotations (both kinematic changes and linear regression modeling) provide strong enough evidence to question the belief that lever arm dysfunction has a significant effect on crouch gait?
You mention large variability in knee flexion suggests differing group responses (worsened, same, improved). How can you be certain that this variability is true and not a consequence of only using one trial per visit ("one representative walking trial was chosen for further analysis"). In other words, these results could simply be caused by trial-to-trial gait variability (e.g. picked a single trial that happened to be the high end relative to the others for visit 1 and a trial that happened to be on the low relative to the others for visit 2, while overall mean values of all trials for both visits are the same).

Discussion
Related to previous comment, the "natural history" statement in "This prospectively confirms that tightening of the hamstrings does not contribute to crouch progression, is part of the natural history of children/adolescents with CP and occurs over time." implies that there were no treatments provided to any of these individuals during the study period. A purist might even say that any treatment provided at any time to these individuals (even prior to recruitment to this study) should prohibit the use of 'natural history' since any intervention *might* have the potential to alter body development. Please explicitly list out what general interventions this sample population underwent prior to and during the study period. This would help readers better understand the context in which to evaluate these results and conclusions.
You mention 20 deg is a commonly utilized 'tipping point' for worsening of crouch, and you concluded from the results of this study that it doesn't necessarily mean that this is the case. I understand that you only studied individuals with significant crouch, but I wonder if future studies wouldn't benefit from looking at crouch in CP more broadly, especially considering the high rate of crouch gait in CP, in general. Instead of 20 deg of crouch being a tipping point where PE and kinematic changes start to occur, maybe physiological changes begin to occur much earlier, say from 0-10 deg of crouch, and then don't really change after that. My point is, why not focus on the bigger picture, what causes crouch or what changes occur during early onset of crouch rather than hyper-focusing on what causes bad crouch to get worse.
You state that "our results highlight that reporting mean data only does not capture the significant variability in CP gait data and can hide potentially important findings." But isn't this also true for single trial data (i.e. 'Reporting single trial data does not capture the significant variability in CP gait data and can hide potentially important findings.')? I don't argue with the potential that subgroups can exist that have different responses than the overall group, but it seems naïve to pretend that comparing single gait trials at two time points gives a robust picture of gait kinematics, especially considering the large stride-to-stride variability often observed in individuals diagnosed with CP.

If applicable, is the statistical analysis and its interpretation appropriate? Partly
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Yes © 2023 Kainz H. This is an open access peer review report distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. This is a well written paper with some interesting findings. Please find below my detailed feedback.

Hans Kainz
Introduction is well written and provides a good summary of previous research related to the topic of this paper.
Last sentence in the introduction: Please change the wording from 'measures of bony torsions' to 'estimates of bony torsions'. Otherwise the reader expects that you used medical images to quantify bony torsions.

Methods:
First paragraph in the method section: Please clearly state that data of the same participants has been published in your previous paper.
How did you choose the representative walking trial for your analysis? Was it based on a qualitative evaluation of the kinematic waveforms or did you use a quantitative measure?
Please specify which SPM tests you used for your analyses.
Results: Figure 1: I know the additional figures are in the supplementaty material but I believe it would be better to report pelvic, hip and knee angles in all three planes in the main manuscript. Figure 1: Hypothesis test waveforms take a lot of space and could be included in the kinematic waveform subplot, e.g. horizontal bars indicating significant differences below the kinematic waveforms. From my point of view it would make the figure nicer. However, this is my own preference and I leave it up to the authors to decide how they will present the SPM results in this figure. Table 2: Why is hamstring tightness from the first assessment reported as median (interquartile range) while the last assessment is presented as mean (standard deviation)? Please use the same method to make it easier for the reader to compare both assessments. Table 3: Label at the top is wrong. '>50' should be '>5°', etc. Furthermore, having the same numbers in brackets for increased and decreased crouch was a bit confusing for me. Might be better to report the numbers in brackets as >5°, +/-5° and <5°.
Furthermore, the authors could consider to perform statistical test between these 3 groups.
In general, I believe the additional analyses are interesting. I, however, suggest to mention the additional analyses in the method section. Discussion: Third paragraph: Further thought from my side which might be worthwhile mentioning: Clinical measures of bony torsion are not very reliable. I would recommend long term prospective studies based on medical images, which enable a more reliable and accurate estimation of bony torsion.
Fourth paragraph: Very interesting findings. Change in tibia torsion was larger in the increased crouch gait group which might be worthwhile to mention/discuss in this paragraph.
The lack of medical images to quantify bony torsion is another limitation of this study. Conclusion: I agree with the content of this paragraph, but I would have expected that the conclusion paragraph is more related to the findings/data from this study.
Are all the source data underlying the results available to ensure full reproducibility?
I answered 'partly' to this question because the raw data, i.e. c3d files from the motion capture data are not provided. However, this is common within the communuity and I do not have a problem with it.

Is the work clearly and accurately presented and does it cite the current literature? Yes
Is the study design appropriate and is the work technically sound? Yes

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Partly

Are the conclusions drawn adequately supported by the results? Yes
Competing Interests: No competing interests were disclosed.