I've read several papers that show that in Alzheimer's patients Amplitude is smaller and Latency is longer and that it is a clinically useful test.
There are indeed multiple papers that note that amplitude is smaller and latency is longer in AD patients w.r.t. normal subjects. The issue I've been discussing is whether the P300 can be used to assess drug effects. Thus far they have only been able to show that for latency.
Given the known changes in P300 with AD its conceivable that a drug COULD have a amplitude effect. However, the latency effect should be an earlier/easier marker to identify. As such, the absence of an associated latency effect makes it far more likely that an amplitude change is a spurious finding.