Power analysis was performed for the repeated measures analysis of variance (ANOVA) with a between factor and calculation was performed by the software GPower (version 3.1) (19), and the following inputs were used: (a) Effect size f (V) = 0.65; effect size was estimated by partial eta squared of 0.3; (b) alpha = 0.05; (c) Power = 0.8; (d) Number of groups = 3 (paretic x non-paretic x controls); and (e) Number of measurements (lateral x forward step) 2. A total sample size of 21 participants was required to achieve significance. Twenty community-dwelling adults with hemiparesis and ten controls were recruited for this study. For the inclusion criteria, participants were included if they were >6 months post-stroke, ≥ 50 years of age, could stand unsupported for 5 min, were able to walk 10 m with or without an assistive device, and did not have a medical condition that significantly impacted their ability to walk beyond the effects of the stroke. For the control group, participants were included if they had no self-reported history of a neurological injury or condition. All subjects provided written informed consent. Experimental procedures were approved by the University of Maryland, Baltimore Institutional Review Board.

Participants attended one testing session and performed a stepping assessment and a clinical assessment of balance and motor recovery.

The data focused on the WT phase, defined as the interval from the onset of the lateral WT to the instant of the first foot off (14) (Figure 1). Customized Matlab scripts were used to extract the variables during WT: onset time, duration, average mediolateral center of pressure (ML COP) velocity, ML COP displacement, and hip abduction torque (Figure 2), as previously performed elsewhere (6, 18). The WT onset time was defined as 3 SD above the baseline ground reaction force for at least 100 ms relative to the light cue onset time (6). The first foot off was defined as the time when the ground reaction force of the stepping leg was <10 N: The WT duration was the time from the WT onset until the first foot off (6). The ML COP displacement was normalized to the base of support (BOS) width. BOS width was the mediolateral distance between the medial ankle marker between both feet determined prior to the onset of the light cue. The average ML COP velocity was calculated by dividing the ML COP displacement by the WT duration. Inverse dynamics were used to estimate the hip abductor torque. Using the Newton-Euler equations for motion, calculations were performed using the bottom-up approach. The positions of segmental mass centers of gravity, and segmental masses and their moments of inertia were estimated from the anthropometric data of each subject (23). The torque estimates were normalized to the height and weight of the participant.

The values presented are means and standard deviations. The normality of the data was confirmed by using a Shapiro-Wilk test. An independent t-test was used to compare demographics and clinical outcome measures between groups (individuals with stroke vs. controls). Non-parametric statistics (Kruskal-Wallis test) were used to compare the cutaneous sensation. The right and left leg of controls were not significantly different, so the values from the right and left leg were averaged, and the average was used for the analyses. A repeated measures analysis of variance (ANOVA) with a between factor (leg) was conducted to assess the effects of step direction (lateral vs. forward) on leg (paretic, non-paretic, and control) for WT characteristics (WT onset time, duration, average ML COP velocity, and ML COP displacement) and hip abductor torque. When a significant main effect was found, post hoc comparisons were evaluated using Bonferroni (within comparisons) or Tukey (between comparisons) post hoc. Statistical analyses were performed using SPSS v26 (IBM Corp, Armonk, NY), and the significance level was set at p ≤ 0.05.

Interestingly, regardless of the leg used (paretic, non-paretic, control), the weight transfer before a lateral step was initiated faster, had a shorter weight transfer duration, resulted in a greater ML COP displacement, and generated a greater hip abduction torque of the stepping leg than a forward step. Before a forward step, a greater hip abduction torque was produced in the stance leg compared to the lateral step. This information reveals an important characteristic of stepping. Although there are differences between groups (individuals with stroke vs. controls), the WT characteristics between the different directions has a similar overall pattern across populations. Thus, weight transfer preceding a lateral step has intrinsic differences compared to forward stepping. The quick weight transfer before the lateral step may present problems for those individuals with stroke since their movements tend to be slower with reduced activation of the lower extremity muscles (24–26). The findings in this study indicate the importance and difficulty of weight transfer before the lateral step given the inherent differences from the forward step. Thus, training to increase the movement speed in individuals with stroke may be necessary for weight transfer control when stepping laterally.

The WT differences between step directions presented here may be explained by different factors. For example, independent of step direction, a step will be preceded by a lateral transfer of bodyweight toward the leg that will support the body during the stance. Thus, the body's center of mass moves toward the stance leg while the stepping leg is free to move. Considering the stepping leg may require greater hip abduction during the lateral step to push the foot off the ground and project the leg laterally, it is reasonable to expect a higher hip abduction torque during a lateral step than the forward step. Conversely, the stance leg had a greater hip abduction torque during the forward step. The longer weight transfer duration before the forward step may allow a greater time for generating torque of the stance leg and may be important for progressing the center of mass forward. Additionally, potentially moving into a smaller base of support for the forward compared to the lateral step may require more preparation. This might happen due to the necessity to move the stepping leg medially before moving forward, while for the lateral step, the leg mainly moves laterally before the step occurs. Therefore, when stepping forward into a smaller base of support greater hip abduction torque may be required to stabilize the stance leg before the step occurs. Since our aim was not to explain the possible differences in hip torque we found here, future studies should further investigate which kinematic variables may influence torque production before the step. This information might provide a further understanding of the factors that influence the ability to step in different directions.

We showed for the first time, individuals with stroke are slower to initiate the transfer of weight compared to individuals. Overall (lateral and forward direction), individuals with stroke, paretic and non-paretic leg, were approximately 48% slower initiating weight transfer, took 89% longer to transfer the weight and 35% slower executing the weight transfer than controls. Along these lines, a study demonstrated that step initiation time took 93% longer in individuals with stroke before forward/backward voluntary steps than controls (27). It is not surprising since individuals with stroke have neural impairments that might affect motor behavior by decreasing the ability to perform a specific motor task (28, 29). Additionally, individuals with stroke also took more time to perform the weight transfer regardless of the step directions, which might be due to different motor planning. A previous study has shown that individuals with stroke may need more time to plan a voluntary step (30) which would explain the differences between individuals with stroke and controls. The slower response of the individuals with stroke compared to controls, and also between paretic and non-paretic leg, may be related to a reduced capacity to activate the muscles needed to transfer weight before the lateral step. Previous research showed that the rate of activation of the hip abductors and adductors muscles influence the time of the weight transfer preceding a voluntary lateral step in older adults (7); hence, individuals with stroke may present a reduced rate of activation of the hip abductors and adductors muscles compared to controls. Moreover, it is possible that a reduced motor cortex excitability (31) and loss of motor units (32, 33) may also affect the ability to prepare the body for stepping.

There were no significant differences in torque production (stance or stepping leg) between individuals with stroke and controls. In contrast, in a different context, other studies demonstrated that individuals with stroke have a lower capacity to produce hip abduction torque (measured at isokinetic dynamometer) than matched controls (9, 34). Thus, torque values measure in-task (e.g., stepping) may provide different outcomes compared to measurements performed outside the task (e.g., isometric test) when comparing individuals with stroke and controls. Although, differences were found between the paretic and non-paretic, the non-paretic leg had an overall hip abduction torque up to ~44% higher than the paretic leg. A study showed that the non-paretic leg may have a negative influence on the paretic leg by decreacreasing the performance of the paretic leg during walking (35), which could impact on step performance. In addition to that, previous research found individuals with stroke had reduced weight bearing on the paretic side (36), which could affect the ability to produce torque in the stance leg during a voluntary step (both directions). However, we controlled the weight bearing in the present study to assure equal weight distribution between paretic and non-paretic legs. Thus, it is possible the reduced velocity demonstrated by the paretic leg (Figure 2), may lead to a lower torque production, helping to explain the differences in torque between legs. Moreover, the ability to produce torque is dependent on the ability to activate the muscles involved in the task (37). For instance, previous research demonstrated impaired muscle activation in the muscles of the paretic leg, while the non-paretic leg had altered muscle activation as a result of compensatory strategies during a forward step (38), which may also help to explain the differences in torque we demonstrated here. Yet, as mentioned above, people individuals with stroke may have a reduced motor cortex excitability and loss of motor unis (31–33) which may lead to an overall motor impairment in the paretic leg (39) which would affect the ability of the paretic leg to perform a step.

Individuals with stroke were slower (COP velocity) during the weight transfer, which also helps to explain the above-mentioned differences between groups. As pointed out above, the neural impairment associated with stroke (28, 29, 31, 33) possibly affects the ability to react quickly and move the lower limbs during a motor task. Furthermore, a similar motor planning, regardless of the leg stepping, may also explain the differences between the individuals with stroke and controls we found in this study (30). Considering motor planning is the integration of sensory afferent information responsible for the position of the limbs (30, 40), this might directly affect the performance during the WT (weight transfer onset, duration, ML COP velocity and ML COP displacement). Indeed, researchers have shown that individuals with stroke may need more time to plan a voluntary step (30), which would explain the differences in WT characteristics between individuals with stroke and controls in the present study.

Nonetheless, a stroke can have an impact on brain function (41), which may reduce cortical excitability (42, 43) and impact balance control (44). Although we did not investigate here the differences in cortical excitability (individuals with stroke vs. controls), it is possible that the impairment of the individuals with stroke we presented might be connected with a decreased brain function. For example, a review pointed out that brain structure (e.g., cerebellum, brainstem) is associated with dynamic balance and leads to a balance disorder (44). Thus, exploring the motor cortex excitability in relation to the WT would be important for future studies.

The interpretation of this investigation should be made taking into consideration its limitations. The participants of the present study performed voluntary steps with no external perturbations, and half of all falls an external perturbation is present (e.g., tripping on an obstacle). Yet, understanding a voluntary step is of fundamental importance for individuals with stroke since they present an explicit motor limitation due to their stroke and often limit their ability to step. Moreover, although we investigated the differences between lateral and forward stepping, it would be important to investigate the backward step since many falls also occur in this direction (45). The present study has important implications for health care professionals for understanding directional vulnerability among individuals with stroke. Since our results showed that the transfer of weight preceding a lateral step appears to require a quicker movement than the forward step, the risk of falling may be greater in during lateral stepping. Our results also indicate that a reduced hip abduction torque capacity may place individuals with stroke at risk for falls. Future studies may want to explore how hip abduction neuromuscular activation would contribute to weight transfer across different step directions since individuals with stroke often present a neural impairment. Nonetheless, further understanding if a quicker weight transfer and stronger hip abduction torque can reduce falls among individuals with stroke, would advance training prescription to reduce falls in this population.