Background
The Placebo Group Simulation Approach (PGSA) aims at partially replacing Randomized Placebo-Controlled Trials (RPCTs) using data from historical control groups in order to decrease the needed number of study participants exposed to lengthy placebo treatment. PGSA algorithms were originally derived from Mild Cognitive Impairment (MCI) data of the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. To produce more generalizable algorithms, we aimed to compile five different MCI databases in an heuristic manner to create a ‘standard control algorithm’ for use in future clinical trials.
Methods
We compared data from two North American cohort studies (n= 395 and 4,328, respectively) one international clinical drug trial (n= 831) and two convenience patient samples, one from Germany (n=726), and one from Switzerland (n=1,558).
Results
Despite differences between the five MCI samples regarding inclusion and exclusion criteria, their baseline demographic and cognitive performance data varied less than expected. However, the five samples differed markedly with regard to their subsequent cognitive performance and clinical development: (1) MCI patients from the drug trial did not deteriorate on verbal fluency over 3 years, whereas patients in the other samples did; (2) relatively few patients from the drug trial progressed from MCI to dementia (about 10% after 4 years), in contrast to the other four samples with progression rates over 30 percent.
Conclusion
Conventional MCI criteria were insufficient to allow for the creation of well-defined and internationally comparable samples of MCI patients. More recently published MCI criteria are unlikely to remedy this situation. The Alzheimer scientific community needs to agree on a standard set of neuropsychological tests including appropriate selection criteria to make MCI a scientifically more useful concept. Patient data from different sources would then be comparable and the scientific merits of algorithm-based study designs such as the PGSA could be properly assessed.