A systematic review of sample representativeness and homogeneity in exercise trials using group designs for people with cerebral palsy.

Background: Evidence quality from exercise trials which use group designs is partly dependent on whether study samples represent the population and acceptable sample homogeneity is achieved. This study aimed to review studies evaluating exercise training for people with cerebral palsy (CP) and appraise to what extent i) participants were representative of people with CP; and ii) internal validity was threatened by sample heterogeneity. Methods: A search of 5 major databases was conducted to identify exercise trials which use group designs for people with CP. Participant characteristics were extracted and used to divide participants into sub-populations. Indicative Participant Prevalence Ratio (iPPR) was calculated to assess representativeness. Sample homogeneity was appraised for each study. Results: Fifty-one studies evaluating 836 participants were appraised. Adults comprise 60 ​% of the CP population but were grossly underrepresented (iPPR<0.4). Older adults, people with dyskinetic and ataxic CP and wheelchair users were also grossly underrepresented. The number of studies that accounted for key prognostic variables was: age - 26 studies (51 ​%); sex - 0 studies (0 ​%); neurological subtype - 21 studies (41 ​%); functional effects - 14 studies (27 ​%) and comorbidities - 1 study (2 ​%). Conclusion: Adults with CP and people with high support needs are underrepresented, and future research should prioritise this subpopulation. Trials using group designs require researchers to reconcile two competing interests - adequate sample size and sample homogeneity and to date, sample homogeneity has received insufficient priority. This threatens internal validity and, consequently, the overall quality of evidence underpinning clinical exercise prescription is likely to be lower than previously thought.