There are different ways of validating a new cognitive test. The easiest trial of a new test is to compare the performance of patients with established Alzheimer’s disease with pre-screened healthy controls. A reasonable test should perform very well in such a trial. The specificities and sensitivities produced by such studies can be impressive and are sometimes used (erroneously) in review papers to compare tests. The problem is that this is too easy; the more advanced the dementia and the more selected the controls, the more impressive will be the sensitivity and specificity.

A second method is to use patients with mild disease and matched, unscreened controls. This is the model we used.

A third method of validation is to use the test in the clinic on all patients presenting with memory problems and then compare the results of patients diagnosed with Alzheimer’s disease with those not given a diagnosis of dementia. This has the advantage of having direct clinical application but leads to other problems. The major problem is that in memory clinics, not all patients on their first visit are divided into two groups: Alzheimer’s disease (or dementia) and normal. Many patients are in between. Some of these are regarded as having mild cognitive impairment (MCI). One form of MCI, amnestic MCI, is on a spectrum with AD [10]. Should these patients be regarded as having mild AD or as “not demented?” If they are treated as not demented, then a sensitive test which picks up their deficits may appear inferior to an easier test that fails to detect milder problems.

The original TYM test validation [4] was performed with patients, with predominantly mild AD, usually on their first visit to the Cambridge Memory Clinic at Addenbrooke’s Hospital. The controls were relatives of the patients attending the clinic. When we needed to extend the age range and number of controls, relatives of other patients attending Addenbrooke’s Hospital and the Queen Elizabeth and north Cambridgeshire hospitals were recruited. The memory clinic controls are likely to be of the same educational background as the patients and are the most useful group to compare to the patients.

In the study, 108 patients with a clinical diagnosis of Alzheimer’s disease or amnestic MCI were compared to age-matched controls. There is a problem deciding where amnestic MCI ends and where AD begins. The official discriminator, whether the cognitive problems affect lifestyle, is too subjective. The patients with a clinical diagnosis of amnestic MCI were divided into AD and amnestic MCI on the basis of their ACE-R score using the official cut-off of <83/100 [1]. Therefore, patients with a clinical diagnosis of amnestic MCI who scored 82 or less were included in the AD cohort. Patients with a clinical diagnosis of amnestic MCI who scored 83 or more on the ACE-R were treated separately as amnestic MCI.

The 94 patients in the AD cohort had an average age of 69 years. These patients had mild to moderate AD, scoring an average of 67/100 on the ACE-R and 23/30 on the MMSE. On the TYM test, they scored an average of 33/50. The age-matched controls scored 47/50 – so there was a clear difference between the patients and controls. This was highly significant and indeed all the subtest scores (except copying) showed significant differences between AD patients and controls. The data from this study and a second TYM validation study (performed using a similar protocol) are shown in Table 9.2. The second validation study excluded all patients with “moderate” AD, that is, patients scoring less than 20 on the MMSE, and this is reflected in higher TYM, ACE-R, and MMSE scores. The results from the two studies show an almost identical pattern.