Case studies generated for members of this cohort include clinical data regarding onset and progression of cognitive problems, medical history and medications, family history of memory problems, neurological history and examination results, psychiatric history and examination results, neuropsychological examination results, and diagnostic imaging and laboratory results. In addition to demographic data, a clinical history, and a neurological history and examination, virtually all of the case summaries included a computed tomographic or magnetic resonance imaging scan, laboratory studies, at least a brief description of behavioral symptoms and/or signs, and a Mini-Mental State Examination and/or more detailed neuropsychological testing.

Diagnosticians followed the NINCDS-ADRDA criteria for the diagnosis of probable and possible AD, with two further specifications. First, a gradual progressive language deficit as the initial symptom would warrant a diagnosis of possible, not probable, AD. This change was instituted based on recent autopsy findings that such patients sometimes do not have AD.  Second, subjects with a prominent early behavioral disturbance should have a diagnosis of possible AD based on evidence that such patients do not invariably have AD. 

In phase 1 of the study, at least two clinicians at each of the two non-originating sites rated each patient as having probable AD, possible AD, or non-AD using the modified NINCDS-ADRDA criteria described above. All diagnosticians were “blinded” to the autopsy diagnosis and were not told which of the other two sites had provided the case. Where internal disagreements occurred, the clinicians at each site discussed the cases informally and agreed on a clinical diagnosis, which was tabulated as a pre-consensus diagnosis. In phase 2 of the study, consensus conferences were conducted for each case on which the non-originating sites disagreed, and post-consensus diagnoses (including the cases with agreement) were tabulated.