A total of 48 studies (41 diagnostic cohorts and 7 diagnostic case-controls) were included in the review.
Accuracy of SPECT in distinguishing AD from vascular dementia (13 studies, 768 participants).
The mean age of the study participants, where reported, ranged from 62.8 to 80.2 years. All studies scored + or ++ for both internal and external validity. There was no evidence of a threshold effect. The pooled estimates of sensitivity and specificity were 71.3% (95% CI: 67.5, 75.2) and 75.9% (95% CI: 70.8, 81.1), respectively. Regression analysis found no explanatory variables for the observed between-study heterogeneity in DOR. The funnel plot indicated an absence of publication bias.
Accuracy of SPECT in distinguishing AD from fronto-temporal dementia (7 studies, 424 participants).
The mean age of the study participants, where reported, ranged from 55.6 to 72.8 years. Six studies were graded ++ and one – for internal validity, whereas four were graded ++ and three + for external validity. There was no evidence of a threshold effect or significant between-study heterogeneity in DOR. The pooled estimates of sensitivity and specificity were 71.5% (95% CI: 66.3, 76.7) and 78.2% (95% CI: 71.2, 85.2), respectively. Regression analysis was not attempted. With only 7 studies, the funnel plot was inconclusive.
Accuracy of SPECT in distinguishing AD from non-dementia patients (13 studies, 1,082 participants).
The mean age of the study participants, where reported, ranged from 56.0 to 77.4 years. Control groups varied and included normal controls and/or patients with headaches and dizziness, memory problems, mild dementia, or psychiatric disorders. All studies scored + or ++ for internal validity and all but one for external validity. There was no evidence of a threshold effect. The pooled estimates of sensitivity and specificity were 65.7% (95% CI: 62.2, 69.3) and 79.1% (95% CI: 75.1, 83.1), respectively. Regression analysis found no explanatory variables for the observed between-study heterogeneity in DOR. The funnel plot indicated an absence of publication bias.