Analysis of the 2 x 2 crossover design with subsampling

Treadmill training with partial body weight support compared with conventional gait training for low-functioning children and adolescents with nonspastic cerebral palsy: a two-period crossover study

BACKGROUND

Partial body weight-supported treadmill training has been shown to be effective in gait training for patients with neurological disorders such as spinal cord injuries and stroke. Recent applications on children with cerebral palsy were reported, mostly on spastic cerebral palsy with single subject design. There is lack of evidence on the effectiveness of such training for nonspastic cerebral palsy, particularly those who are low functioning with limited intellectual capacity.

OBJECTIVES

This study evaluated the effectiveness of partial body weight-supported treadmill training for improving gross motor skills among these clients.

STUDY DESIGN

A two-period randomized crossover design with repeated measures.

METHODS

A crossover design following an A-B versus a B-A pattern was adopted. The two training periods consisted of 12-week partial body weight-supported treadmill training (Training A) and 12-week conventional gait training (Training B) with a 10-week washout in between. Ten school-age participants with nonspastic cerebral palsy and severe mental retardation were recruited. The Gross Motor Function Measure-66 was administered immediately before and after each training period.

RESULTS

Significant improvements in dimensions D and E of the Gross Motor Function Measure-66 and the Gross Motor Ability Estimator were obtained.

CONCLUSIONS

Our findings revealed that the partial body weight-supported treadmill training was effective in improving gross motor skills for low-functioning children and adolescents with nonspastic cerebral palsy. .

Use of the repeated cross-over designs in assessing bioequivalence

We consider applications of the repeated 2 x 2 cross-over design to evaluating bioequivalence between the two formulations. The repeated 2 x 2 cross-over design allows us not only to assess bioequivalence on average bioavailability and to examine the subject-by-formulation interaction but also to obtain independent unbiased estimates of intrasubject variability. One consequence of unequal intrasubject variabilities is that the sum of squares of intersubject residuals and the sum of squares of subject-by-formulation residuals are not independent. We also discuss the relative merits of this design as compared to the standard 2 x 2 cross-over design without repeated measurements in terms of precision and sample size with respect to the ratio of the number of subjects to the repeated measurements per subject. We investigate other uses of the 2 x 2 cross-over in examining the bioequivalence between the two different dosing regimens. Possible applications of other repeated cross-over designs to bioequivalence for more than two formulations are explored. A numerical example illustrates the proposed procedure.