Cross-sectional versus longitudinal estimates of cognitive change in nondemented older people: A CERAD study

OBJECTIVES: To investigate the effect of increasing age on cognition in nondemented older people, DESIGN: A cross-sectional and longitudinal analysis. PARTICIPANTS: A total of 454 control subjects for Alzheimer's cases from the cohort assembled by the Consortium to Establish a Registry for Alzheimer's Disease (CERAD). MEASUREMENT: The Mini-Mental State Examination (MMSE) to assess cognitive function. RESULTS: Cross-sectional estimates were derived by generalized linear models and longitudinal estimates by generalized estimating equations. The cross-sectional model indicated a small but significant decline in MMSE of -.4 points per 10 years. The longitudinal model indicated a small but significant increase in MMSE of about +.6 points per 10 years. Evidence of an early learning effect and nonrandom dropout exists. CONCLUSIONS: The question of "normal" aging can be approached by considering cross-sectional information and, usually separately, longitudinal information. This study does both using recently developed statistical methods. We conclude that there is a small but significant decline in scores on the MMSE with increasing population age. The effect can be masked in longitudinal cohorts by a learning effect (especially early in follow-up) and other factors associated with repeated testing.