Participants with bilateral CP were recruited through the Shirley Ryan AbilityLab and the Cerebral Palsy Research Registry (Hurley et al., 2011). Inclusion criteria for these individuals were (1) between the age of 5 and 19 years, (2) a medical diagnosis of bilateral CP affecting the lower limbs, (3) Gross Motor Function Classification System (GMFCS) level I-III, and (4) independent ambulatory function with ability to step up with or without assistive devices. Exclusion criteria were (1) botulinum toxin injections to lower limb muscles in the past 6 months, (2) surgeries affecting lower limb function in the past year, and (3) serious comorbidities or cognitive dysfunction that would affect ability to participate. Age and sex-matched participants without bilateral CP (typical development or TD) were recruited through word-of-mouth and flyers. This study was approved by Northwestern University’s Institutional Review Board. Participants under the age of 18 provided assent in addition to informed consent from their parent/guardian, while participants 18 and older provided informed consent themselves.

Participants performed multiple stepping trials on a 2 × 2 cluster of force plates (AMTI, Watertown, MA, USA) (Figure 1). Two 10.2-cm wooden platforms, each approximately the size of a single force plate, were placed on two side-by-side force plates to simulate a step. A previous study found that a 10.2-cm step height is both challenging and achievable for clinical populations with lower limb impairments (Goyal et al., 2023a). A 10-camera motion capture system (Qualisys, Göteborg, Sweden) recorded lower limb kinematics from retro-reflective markers placed on the trunk (sternum, C7 vertebrae, T10 vertebrae), pelvis (sacrum, posterior superior iliac spines, greater trochanters), and lower extremities (lateral femoral epicondyles, lateral malleoli, calcanei, the second and fifth metatarsals, and thigh and shank four-marker clusters). Ground reaction forces were captured at 1000 Hz, while kinematics were captured at 100 Hz. Participants also wore a harness that was attached to a passive overhead trolley to minimize the risk of falling.

Participants started the experiment with their feet on two independent force plates posterior to the two platforms. This ensured that all ground reaction forces for the left and right lower limbs were recorded separately. Participants were instructed to step up onto the platform at their typical walking speed (i.e., at the pace if they were walking with a friend, not rushing to catch a bus and not slowly moving in a big crowd). After a short pause on the step, participants were then instructed to step down and backward onto the starting force plates. These step-up trials were repeated 5–15 times per leading foot, depending on participant fatigue and comfort. Once the typical speed was established, we monitored that trials were consistently being completed within 10% of this speed. We offered breaks and asked for participant report of their fatigue. A licensed physical therapist guarded and offered support as needed for safety to participants who were GMFCS level III; we did not offer these participants a handrail or other consistent upper extremity support and they did not use their assistive devices. The trial did not begin until they were standing independently. These participants used a hand touch on the physical therapist’s hand if they felt they needed it. If they attempted to use anything more than support for balance (e.g., holding hands or pulling with arms to ascend), they were cued to reduce this type of support and the trial was not used for analysis. In addition to the step-up trials, all participants completed timed single-limb stance tests (Newton, 1989) and the Waterloo Footedness survey (Elias et al., 1998) to identify the dominant (TD) or less affected (bilateral CP) lower limb. Participants with bilateral CP also completed the Selective Control Assessment of the Lower Extremity (SCALE) (Fowler et al., 2009) and the locomotion ability assessment for kids (ABILOCO-Kids) (Caty et al., 2008).

Qualisys Track Manager (Qualisys, Göteborg, Sweden) recorded marker and ground reaction force data. Marker data were visually inspected to ensure that they were correctly labeled. All data were then imported to Visual 3D (C-Motion, Germantown, MD, USA). First, a static trial was used to estimate a biomechanical model using the following markers: trunk (sternum, C7 vertebrae, T10 vertebrae, acromions), pelvis (sacrum, posterior superior iliac spines, greater trochanters, ischial tuberosities), thigh (four-marker clusters, greater trochanters, lateral and medial femoral epicondyles), shank (four-marker clusters, lateral and femoral epicondyles, lateral and medial malleoli), and foot (lateral and medial malleoli, calcanei, first, second and fifth metatarsals). Dempster’s regression equations were used to define segment weights (Dempster, 1995) and segment initial properties were estimated based on segment weight and geometry (Hanavan, 1964). Bell’s regression equations were used to define hip joint centers (Bell et al., 1989), while knee and ankle joint centers were estimated based on the static markers defined above. At least three markers from each segment were used to track segment movement. The marker and ground reaction force data were first interpolated to fill in small gaps and passed through a 4th-order low-pass Butterworth filter with a 6 Hz cutoff frequency to filter out high-frequency oscillations. These data were then used in combination to perform inverse dynamics calculations of hip moments in the frontal plane and hip, knee, and ankle moments in the sagittal plane. The joint angular convention used was the Cardan sequence—global axes were used to define overall metrics and local axes were used to define joint angles. To account for the 10.2-cm step height, the height of the two corresponding force plates was adjusted virtually. Important gait events, including lift-off and initial contact for both lower limbs, were identified by the software using a 5 N ground reaction force threshold and manually inspected for accuracy.

Joint moment data were further analyzed in MATLAB (MathWorks, Inc., Natick, MA, USA). All joint moments were normalized to participant body weight for comparison in statistical analyses. Hip, knee, and ankle sagittal plane moments were considered separately and together, summed to calculate an overall extensor support moment (Novak and Brouwer, 2011). All data were evaluated for each individual trial during two stance phases of the step-up. The push-off phase was defined when the trailing limb was in single-limb stance, between leading limb lift-off and initial contact with the step. The pull-up phase was defined when the leading limb was in single-limb stance, between trailing limb lift-off and initial contact with the step. For each individual participant, any step-up trial with a length outside of two standard deviations from the average were not considered in statistical analysis.

All statistical analyses were performed in Stata IC 14.1 (StataCorp LLC, College Station, TX, USA) and the threshold for significance was set to p < 0.05. Independent two-sample t-tests were used to compare participant-specific metrics such as age, weight, and height. An ANOVA was used to compare single limb stance times between the limbs in each group [CP more affected (MA), CP less affected (LA), typical development dominant (TD)]. Paired t-tests were also used to compare SCALE scores between the limbs of participants with bilateral CP. All outcome measures were considered independently for each stance phase: (1) peak hip abduction moments, (2) peak support moments, (3) individual hip, knee, and ankle percent contributions to peak support moments, (4) time duration of stance phase, and (5) time to peak moments. Individual joint percent contributions were calculated by dividing hip, knee, and ankle contributions to peak support moment by the overall peak support moment. Timing of peak moments was identified as a percentage of the corresponding stance phase. Our analysis determined that the effect of limb dominance was not significant for children with TD (Goyal et al., 2023b); as such, only data from their dominant limb was considered in subsequent analyses. All outcome measures were statistically compared using linear mixed effects models with one fixed effect of limb (MA, LA, TD) and a random effect of participant. The distribution of residuals for each mixed effects model was visually inspected using histograms to confirm normality of the data. All trials for each participant were input individually into the statistical models to increase the effective sample size. Significance of multiple pairwise comparisons was adjusted using Bonferroni’s corrections. Finally, the effect size of each statistical analysis was calculated using G*Power (Faul et al., 2009).

We recruited 7 participants in each group (Table 1). There were no significant differences in age, weight, and height between the groups. Five participants with bilateral CP were GMFCS level II and two participants were GMFCS level III. No participants with CP had notable joint contractures in the lower extremities that influenced their ability to perform the step up task. Single limb stance times between the limbs were significantly different (p = 0.002), where the TD limb had a larger stance time than the LA (p = 0.007) and MA limbs (p = 0.004). Average SCALE scores were significantly different between the limbs of the CP group (p = 0.020).

There were no significant differences in peak hip abduction or support moments between the limbs in the push-off stance phase (Table 2). There were significant differences in hip (p < 0.001, effect size = 0.943) and ankle (p = 0.011, effect size = 0.492) percent contributions to peak support moments between the limbs (Figures 2, 3). The MA limb had a significantly higher hip percent contribution compared to the LA and TD limbs, while the LA limb also had a significantly higher hip percent contribution compared to the TD limb (p < 0.001 for all). In contrast, both limbs of participants with CP had lower ankle percent contributions compared to the TD limb (less affected: p = 0.016; more affected: p = 0.003). Joint moments in Nm/kg for the hip, knee, and ankle at the time of peak support moments are provided in Table 2, though statistics were not run on these values.

There were significant differences in average stance time of the push-off phase (p = 0.005, effect size 0.322), where the LA limb spent more time in the stance phase than the MA limb (p = 0.001) (Table 2). There were also significant differences in timing of peak moments between the limbs (both p < 0.001, effect size = 36.5 for peak hip abduction moment and 0.580 for peak support moment). Both limbs of participants with CP reached a peak hip abduction moment later compared to the control limb (LA: p < 0.001; MA: p = 0.004) and reached a peak support moment earlier compared to the TD limb (LA: p < 0.001; MA: p < 0.001).

There were significant differences in peak support moments (p < 0.001, effect size = 0.640) between the limbs in the pull-up stance phase (Table 2). Both the limbs of participants with CP generated higher peak support moments compared to the TD limb (LA: p = 0.016; MA: p = 0.001). There were also significant differences in hip and knee percent contributions to peak support moments between the limbs (both p < 0.001, effect size = 0.719 for hip and 0.942 for knee) (Figures 2, 4). The MA limb had a significantly higher hip percent contribution compared to the LA and TD limbs (both p < 0.001). In contrast, the MA limb had a significantly lower knee percent contribution compared to the LA and TD limbs (both p < 0.001), while the LA limb also had a significantly lower knee percent contribution compared to the TD limb (p = 0.003). Joint moments in Nm/kg for the hip, knee, and ankle at the time of peak support moments are provided in Table 2, though statistics were not run on these values.

There were no significant differences in the average stance time of the pull-up phase between the limbs, though there was a significant difference in time of peak support moment (p = 0.042, effect size = 0.419) (Table 2). The MA limb reached a peak support moment later compared to the LA (p = 0.020) and TD limbs (p = 0.012).