Quasi-Experimental Study of Weight-Supported Treadmill Training for Myelopathy

In their exploratory, cohort report using body weight–supported treadmill training (BWSTT) in persons with nonprogressive spinal cord injury (SCI), Harkema et al state, “There is a need to evaluate the effectiveness of locomotor training by using a standardized protocol and outcomes with a large sample of participants with upper motor neuron, incomplete SCI.” For this rationale, the Reeve Foundation funded 7 sites (NeuroRecovery Network [NRN]) to treat 196 patients graded C and D on the American Spinal Injury Association Impairment Scale (AIS). The investigators did not use any of the design features needed to scientifically test the impact of their training. They acknowledge limitations but did nothing to mitigate them. The protocol did not randomize to a comparison intervention, specify number of treatments, obtain blinded outcomes, or assess outcomes at standard posttreatment times or examine their sustainability. No measures of home-based walking practice, lower limb strength, physical functioning, and quality of life were obtained. Instead, they conjectured about the possible impact of their findings based on unrelated studies. No statistical correction for multiple outcome tests was performed. Indeed, the cohort's high variability of baseline and change scores, which the authors decry as a problem of past studies, is a contraindication for their use of the Wilcoxon statistic for 2-group outcome studies. Lack of a standard protocol, with high baseline variability (eg, time from onset was 32d to 25y; participants trained for 20–251 sessions over 28–649d), made it unlikely, a priori, that they could generalize about impact or establish an effect size.