Ranges of Sensitivity and Specificity of Clinical Diagnoses Made With and Without the Use of PET, Compared with Neuropathologic Confirmation of Presence or Absence of AD
| Sensitivity | Specificity | Basis of AD diagnosis |
|---|---|---|
| 66% ± 17% | 77% ± 23% | Clinical evaluation, probable AD |
| 90.5% ± 5.5% | 55.5% ± 5.5% | Clinical evaluation, probable + possible AD |
| 91.5% ± 3.5% | 70% ± 3% | 18F-FDG PET, AD pattern |
In addition to those discussed in the text, another (class II) study reviewed by the AAN (83), which did not uniformly require diagnosis by NINCDS-ADRDA criteria, examined a large series of autopsy-confirmed cases pooled from multiple centers, with and without information about apolipoprotein E genotype. This study required no stratification between possible-AD and probable-AD patients, providing data based only on having pooled together all patients thought most likely to have AD. Nor did the study provide any baseline demographic data, so it was not possible to ascertain the level of severity of dementia in the subjects studied. Similarly to the studies summarized in Table 5, this study found that in achieving a sensitivity comparable to that obtainable with PET (93%), specificity was lower (55%) than that achieved with PET. When presence of apolipoprotein ε4 allele was used as the diagnostic criterion for AD, sensitivity and specificity were 65% and 68%, respectively. When the data from clinical and genotype evaluations were sequentially combined, sensitivity fell to an unacceptably low 61%, at a specificity of 84% and overall accuracy was 65%.