The primary entry criteria for the study was "Patients aged 60 to 85 years, inclusive, with a NIA-AA diagnosis of mild cognitive impairment (MCI) due to AD or early stage mild dementia due to AD" (see https://clinicaltrials.gov/study/NCT03790709#participation-criteria). This was not a presymptomatic (FDA Stage 1 and 2) where patients would be mostly CDR 0 but show amyloid pathology only (stage 1) or amyloid pathology and subtle neurocognitive findings not typically apparent on short screens like the MMSE (which had a range 20-28 in the study).
Doc is proving too much by arguing that a CDR-G score of 0.5 doesn't really fit within the criteria of the draft FDA guidance for stages 1 to 3, because it leaves no situation in which the CDR-G scale can be used by the draft FDA guidance. Why would the FDA even bother to propose to change course in endpoints if, in practice, it couldn't be reasonably applied?