Randomized Controlled Trial of the Behavioral Intervention for Physical Activity in Multiple Sclerosis Project: Response Heterogeneity and Predictors of Change

Background We reported that a social cognitive theory-based (SCT), Internet-delivered behavioral intervention increased device-measured minutes/day of moderate-to-vigorous physical activity (MVPA) over a 6-month period among persons with multiple sclerosis (MS). This paper examined the pattern and predictors of heterogeneity in change for MVPA. Based on previous research, we hypothesized that mild MS disability, fewer MS symptoms, lower baseline MVPA, and positive SCT characteristics (e.g., high exercise self-efficacy) would be associated with greater change in MVPA. Method Persons with MS ( N = 318) were randomized into behavioral intervention ( n = 159) or attention/social contact control ( n = 159) conditions that were administered via Internet websites and supported with behavioral coaching. Demographic, clinical, symptom, behavioral, and SCT data were from before the 6-month period of delivering the conditions, and MVPA data were from before and after the 6-month period. We examined heterogeneity based on waterfall plots, box plots, and the Levene statistic. We identified predictors of MVPA change using bivariate correlation and multiple, linear regression analyses per condition. Results The Levene statistic indicated statistically significant heterogeneity of variances for MVPA change between conditions ( p = .003), and the waterfall plots and box plots indicated greater heterogeneity in MVPA change for the behavioral intervention. MVPA change score was correlated with baseline MVPA ( r = − .33 and r = − .34, p = .0004 and p = .0001) in both conditions and walking impairment ( r = − .188, p = .047) and race ( r = .233, p = .014) in the behavioral intervention condition. The regression analysis indicated that baseline MVPA (Standardized B = − .449, p = .000002), self-reported walking impairment (Standardized B = − .310, p = .0008), and race (Standardized B = .215, p = .012) explained 25.6% of variance in MVPA change for the behavioral intervention condition. Conclusion We provide evidence for walking impairment, baseline MVPA, and race as predictors of the heterogeneity in the pattern of MVPA change with a behavioral intervention.