Objective: To evaluate the accuracy of Single Photon Emission Computed Tomography (SPECT) scanning in the diagnosis of Alzheimer's Disease (AD) and its capacity to improve the diagnostic accuracy of conventional clinical evaluation.
Design: Comparison of SPECT scanning of AD and normal subjects with the criterion standard of clinical diagnosis confirmed by 1-year repeated evaluation.
Setting: A memory clinic in a tertiary care university hospital.
Patients: One hundred twenty patients were evaluated upon entering the Jewish General Hospital (McGill University) Memory Clinic. Fifty-eight patients were diagnosed as having AD and 17 as having vascular dementia. Twenty unmatched controls (recruited mainly through newspaper advertisements) were normal, and 25 had cognitive impairment without dementia (not included in the statistical analysis).
Main outcome measures: Comparison of visual inspection of SPECT, based on the system of classification developed by Holman et al., using B pattern alone as positive or B (bilateral posterior temporal and/or parietal cortex deficits) or C (bilateral posterior temporal and/or parietal deficits with additional defects) pattern and B or C or D (unilateral posterior temporal and/or parietal defects with or without additional defects) as positive compared with clinical diagnosis after repeated evaluations. Sensitivity and specificity, as well as positive predictive value (PPV) and negative predictive value (NPV) based on the prevalence of AD in a memory clinic setting of 30% or 50%, were calculated.
Results: With B pattern as positive, the sensitivity of SPECT was 21% whereas the specificity was 80%. With B or C as positive, the sensitivity was 29% and the specificity was 75%. With B or C or D as positive, the sensitivity was 55% and the specificity was 65%. With a 30% prevalence, the PPV with B pattern as positive was 31% whereas the NPV was 30%. The PPV with B or C as positive was 33% while the NPV was 29%, and the PPV with B or C or D as positive was 40% whereas the NPV was 23%. With a 50% prevalence, the PPV with B pattern as positive was 51% and the NPV 49.6%; the PPV with B or C as positive was 54% and the NPV 48.6%; the PPV with B or C or D as positive was 61% while NPV was 41%.
Conclusion: The sensitivity and specificity were too low for SPECT to be useful as a diagnostic test for AD. The poor positive and negative predictive values in our tertiary care clinic mean that SPECT is not useful in "ruling-in" or "ruling-out" AD in that setting. In fact, clinical evaluation is more accurate.