Estimating sample size for longitudinal studies of age-related cognitive decline

Cross-sectional age-group norms for cognitive tests reveal a progressive age-related decline. There is current interest in the possibility of developing interventions that forestall the normal cognitive decline in elderly adults. This raises new research design issues, one of which is the estimation of appropriate sample sizes for longitudinal studies when only cross-sectional data are presently available. Formulae for estimating change parameters from cross-sectional data are presented in this article. Based on age-related changes in cross-sectional norms for the Wechsler Adult Intelligence Scale (WAIS), the sample sizes required for controlled intervention studies of 1 to 3 years duration are very large. If cognitive decline within individuals is not as great as the decline evident in cross-sectional norms, the required sample sizes will be even larger.