Psychometric considerations when measuring cognitive decline in Alzheimer's disease.

Measuring cognitive decline is important for both clinical and basic research purposes, but to do so is a complicated methodologic and statistical exercise. Some promising predictive measures have been identified, such as baseline severity of disease, early language deterioration, other early behavioral disturbance and extrapyramidal signs. Nevertheless, investigations of demographic, cognitive and biologic variables have not consistently identified factors affecting differences in the course or rate of decline. Moreover, contradictory results using similar measures are common. Such contradictory results may be attributed, in part, to differences among samples, cognitive tests selected, research design, and methods of statistical analysis. Large samples of patients with dementia examined repeatedly for long time periods are needed. However, tests developed for initial screening, diagnosing and categorizing Alzheimer's disease are not necessarily the most appropriate for longitudinal studies of disease course. New instruments with a broader range of item difficulty, and less susceptibility to floor and ceiling effects must be developed. Also, standardized ways of defining cognitive decline are needed which are more sophisticated than simple change scores. Standardization will improve the ability to compare investigations and perhaps reconcile apparent differences in results.