Reliability of physical-activity measures over six months in adults with multiple sclerosis: implications for designing behavioral interventions.

Reliability estimates are important parameters for power analyses and sample-size calculations in behavioral interventions for increasing physical activity among persons with multiple sclerosis (MS). This study involved a secondary data analysis for providing reliability estimates for measures of physical activity over a six-month period in MS. On two occasions separated by six months, participants completed the Godin Leisure-Time Exercise Questionnaire (GLTEQ) and wore an ActiGraph model 7164 accelerometer on a belt around the waist for a seven-day period. The outcomes from the accelerometer were total activity counts and minutes of moderate-to-vigorous physical activity (MVPA) per day. The main analyses involved two-way, random effects intraclass correlation coefficients (ICCs). The ICCs for GLTEQ scores, total activity counts, and minutes of MVPA were .66, .84, and .84, respectively. Additionally, the sample was less physically active than adults without MS, and correlation analyses supported the validity of the measures, but within a substantially larger sample. Overall, the measures had acceptable, but different, reliability over six months in this sample of mostly women with relapsing-remitting MS. Power analyses using the reliability estimates indicated vastly different sample sizes for the measures of physical activity and when compared with the default parameter in software packages.