Stroke is the leading cause of chronic disability in the world (Virani et al., 2021). Reduced speed and diminished balance control are primary deficits in post-stroke gait (Ng et al., 2017), which can lead to reduced community ambulation, restricted societal participation, and poor quality of life (Grau-Pellicer et al., 2019). Consequently, re-learning gait is a primary goal of stroke rehabilitation. Forward locomotor training (FLT), the most common form of post-stroke gait training, has been successful at improving gait speed (e.g., Bowden et al., 2013, Combs et al., 2012, Duncan, P.W., Sullivan, K.J., Behrman, A.L., Azen, S.P., Wu, S.S., Nadeau, S.E., Dobkin, B.H., Rose, D.K., Tilson, J.K., Cen, S., Hayden, S.K., LEAPS Investigative Team, 2011), balance, balance confidence and walking endurance (Bowden et al., 2013). On the other hand, backward locomotor training (BLT) is emerging as a more potent gait training approach in individuals with stroke (Chen et al., 2020, Rose et al., 2018). Several studies have reported BLT’s effectiveness in improving forward gait speed (e.g., Moon and Bae, 2019, Rose et al., 2018, Yang et al., 2005), balance (Chang et al., 2021, Rose et al., 2018, Yang et al., 2005), balance confidence (Moon and Bae, 2022, Rose et al., 2018) as well as walking endurance (Chang et al., 2021, Rose et al., 2020). Further, a recent meta-analysis of 10 studies bolsters these benefits of BLT as compared to FLT or control interventions post-stroke (Chen et al., 2020). Although these studies report considerable benefits of BLT over FLT based on improved clinical outcomes, there is yet no empirical evidence to elucidate the underlying mechanisms for improved gait speed and balance control post-BLT, and whether the gains were achieved through recovery of the paretic limb or compensatory mechanisms from the nonparetic limb (Allen et al., 2014, Balasubramanian et al., 2007, Bowden et al., 2006).

The main mechanism responsible for gait speed modulation is forward propulsion, also referred to as the propulsive impulses generated by each limb, which collectively propel the body center-of-mass forward. An increase in gait speed requires increased net propulsive impulses, which can be generated by one or both limbs. Post-stroke gait is frequently characterized by reduced propulsive impulses from the paretic limb and increased compensatory propulsive impulses from the nonparetic limb (Bowden et al., 2006, Roelker et al., 2019). The reduction in propulsion output from the paretic limb is the hallmark of an energy-deficient hemiparetic gait (Stoquart et al., 2012), and a strong predictor of diminished long-distance walking ability (Awad et al., 2015) as well as reduced total daily walking activity post-stroke (Awad et al., 2020b). Consequently, in the past decade, post-stroke gait rehabilitation has increasingly focused on improving propulsive force output from the paretic limb and reducing the compensatory reliance on the nonparetic limb (Alingh et al., 2020). Although FLT has been successful at improving gait speed, there are mixed reports if improved propulsion from the paretic limb was achieved (Bowden et al., 2013, Combs et al., 2012), unless FLT is accompanied by adjuvants such as backward resistive force to the pelvis (Lewek et al., 2018), functional electrical stimulation (Awad et al., 2014) or biofeedback (Genthe et al., 2018). Backward walking requires increased reliance on sensory and proprioceptive systems as well as increased cortical drive due to reduced visual feedback compared to forward walking (Kurz et al., 2012). With repetitive practice of backward walking during BLT, this increased cortical activation may more efficiently engage the damaged neuronal circuits, further promoting neuroplasticity (Kleim and Jones, 2008). However, it is still unknown whether BLT would promote improved propulsion output from the paretic limb.

Another important goal of post-stroke rehabilitation is to enhance dynamic balance control. While clinical measures provide a broad evaluation of balance control, a mechanistic approach to assess dynamic balance occurs through the analysis of whole-body angular-momentum (Neptune and Vistamehr, 2019). In order to maintain dynamic balance during walking, whole-body angular-momentum needs to be regulated (e.g., Herr and Popovic, 2008). Further, individuals post-stroke have deficits in mediolateral balance control, manifested as an increased range of frontal-plane angular-momentum, which is correlated with poor clinical balance outcomes (Nott et al., 2014, Vistamehr et al., 2014). Although our prior work has shown that FLT was effective for improving mediolateral balance control post-stroke in individuals that walked at higher speeds (≥ 0.8 m/s) pre-training (Vistamehr et al., 2019), it is unknown if BLT can improve regulation of whole-body angular-momentum.

Thus, the purpose of this study was two-fold: 1) to assess the influence of direction-specific locomotor training (FLT or BLT) on forward gait speed and dynamic balance control and identify if changes within each training group were achieved via recovery of the paretic limb or compensation from the nonparetic limb and 2) to quantify if the gains resulting from one intervention were superior to the other intervention. We hypothesized that both interventions would induce improvements in forward gait speed and balance control and that improvements would be achieved through recovery of the paretic limb. Further, we hypothesized that the gains post-BLT would be larger than those post-FLT.