Cognitive impairment is a key feature of Alzheimer's disease (AD). Cognitive performance will therefore be an important outcome measure in disease modifying drug trials. However, cognitive assessment also has several limitations such as a large inter and intra individual variability and floor and ceiling effects. To reduce the effect of these limitations on the accuracy with which drug-related changes can be detected, it is proposed to analyse cognitive change using a slope analysis. In addition, it is suggested to use as the main outcome measure for cognitive impairment a neuropsychological compound score that is based on a z-score transformation of a number of well-validated tests. The selection of these tests would depend on a number of study characteristics such as the inclusion criteria and the length of follow-up.