Measuring neuroplasticity in cerebral palsy: What cohort sizes are needed for MR imaging?

Abstract

Background and Objective(s): Researchers are increasingly looking to supplement clinical trials of motor rehabilitation with neuroimaging in order to examine the relationship between behavioural training, brain changes, and clinical improvements. Randomised controlled trials calculate sample sizes to detect clinical improvements but rarely perform equivalent calculations for imaging measures of neuroplasticity. Study Design: Prospective Cohort study of Power analyses for a clinical trial. Study Participants & Setting: Participant numbers were calculated for a theoretical longitudinal study in which MR imaging took place immediately prior to, and immediately following, treatment in children with CP. Analyses utilised a paired t-test for differences in cortical thickness, or differences in FA of the corticospinal tract using ROI-seeded or functional MRI (fMRI) seeded diffusion tractography conducted on 3T MRI. Materials/Methods: A power analysis was conducted for two measures of brain changes that may be indexed in a trial of rehabilitative therapy for cerebral palsy: cortical thickness of the impaired primary sensorimotor cortex, and fractional anisotropy of the more-impaired corticomotor tract. Power for measuring fractional anisotropy was assessed for both regionof- interest-seeded and fMRI-seeded diffusion tractography. Power analyses were performed using 'pwr' package in R statistical software. Participant numbers for powers of 0.8, 0.9, and 0.95 were calculated for all analyses, assuming trials utilised a paired t-test. The longitudinal effect size was computed using the Cohen's d. Results: Taking into account practical limitations, as well as data loss due to behavioural and image-processing issues, estimated required participant numbers for a power of at 0.8 were 101 for cortical thickness, 128 for region-of-interest-based tractography, and 59 for fMRI-seeded tractography. These study numbers were not adjusted for study attrition. Conclusions/Significance: Although these sample sizes may be out of reach of many trials, several options are available to improve statistical power, including careful preparation of participants for scanning using mock simulators, careful consideration of image processing options, and enrolment of as homogeneous a cohort as possible. This work suggests that smaller and moderate sized studies give genuine consideration to harmonising scanning protocols between groups to allow the pooling of data.L6184697742017-09-29 <br />