Secondary mediation and regression analyses of the PTClinResNet database: Determining causal relationships among the International Classification of Functioning, disability and health levels for four physical therapy intervention trials

Background. Each of the 4 randomized clinical trials (RCTs) hosted by the Physical Therapy Clinical Research Network (PTClinResNet) targeted a different disability group (low back disorder in the Muscle-Specific Strength Training Effectiveness After Lumbar Microdiskectomy [MUSSEL] trial, chronic spinal cord injury in the Strengthening and Optimal Movements for Painful Shoulders in Chronic Spinal Cord Injury [STOMPS] trial, adult stroke in the Strength Training Effectiveness Post-Stroke [STEPS] trial, and pediatric cerebral palsy in the Pediatric Endurance and Limb Strengthening [PEDALS] trial for children with spastic diplegic cerebral palsy) and tested the effectiveness of a muscle- specific or functional activity-based intervention on primary outcomes that captured pain (STOMPS, MUSSEL) or locomotor function (STEPS, PEDALS). Objective. The focus of these secondary analyses was to determine causal relation- ships among outcomes across levels of the International Classification of Functioning, Disability and Health (ICF) framework for the 4 RCTs. Method and Design. With the database from PTClinResNet, we used 2 separate secondary statistical approaches-mediation analysis for the MUSSEL and STOMPS trials and regression analysis for the STEPS and PEDALS trials-to test relationships among muscle performance, primary outcomes (pain related and locomotor related), activity and participation measures, and overall quality of life. Results. Predictive models were stronger for the 2 studies with pain-related primary outcomes. Change in muscle performance mediated or predicted reductions in pain for the MUSSEL and STOMPS trials and, to some extent, walking speed for the STEPS trial. Changes in primary outcome variables were significantly related to changes in activity and participation variables for all 4 trials. Improvement in activity and participation outcomes mediated or predicted increases in overall quality of life for the 3 trials with adult populations. Limitations. Variables included in the statistical models were limited to those mea- sured in the 4 RCTs. It is possible that other variables also mediated or predicted the changes in outcomes. The relatively small sample size in the PEDALS trial limited statistical power for those analyses. Conclusions. Evaluating the mediators or predictors of change between each ICF level and for 2 fundamentally different outcome variables (pain versus walking) provided insights into the complexities inherent across 4 prevalent disability groups.