MDVN and AD outcome studies:
I can categorically state that every AD trial that I ever participated in strictly required that patients be seen at the same time of day for each measurement visit, so this would limit the variance associated with diurnal rhythm and sundowning.
That said, there are points well made today about the subjective nature of AD outcome measures, as well as the reliance on caregivers' recall of patient behavior during the time in between the last visit. There are biases that we have yet to be able to control for that include misattribution biases, hindsight bias, and serial position effects, to name a few. I cannot recall any study that even starts to attempt to control these rater biases, which may wash out the significance of any dataset.
AD outcome trial aren't the only sphere that rely on "subjective" (wallstarb's term, not mine) measures for outcome. Pain studies, and virtually all psychiatric drug studies rely on similar outcome data.
JM2C
aj