Amyloid-β PET in Alzheimer's disease: A systematic review and Bayesian meta-analysis

Abstract

Background: In recent years, longitudinal studies of Alzheimer's disease (AD) have been successively concluded. Our aim is to determine the efficacy of amyloid-β (Aβ) PET in diagnosing AD and early prediction of mild cognitive impairment (MCI) converting to AD. By pooling studies from different centers to explore in-depth whether diagnostic performance varies by population type, radiotracer type, and diagnostic approach, thus providing a more comprehensive theoretical basis for the subsequent widespread application of Aβ PET in the clinical setting.

Methods: Relevant studies were searched through PubMed. The pooled sensitivities, specificities, DOR, and the summary ROC curve were obtained based on a Bayesian random-effects model.

Results: Forty-eight studies, including 5967 patients, were included. Overall, the pooled sensitivity, specificity, DOR, and AUC of Aβ PET for diagnosing AD were 0.90, 0.80, 35.68, and 0.91, respectively. Subgroup analysis showed that Aβ PET had high sensitivity (0.91) and specificity (0.81) for differentiating AD from normal controls but very poor specificity (0.49) for determining AD from MCI. The pooled sensitivity and specificity were 0.84 and 0.62, respectively, for predicting the conversion of MCI to AD. The differences in diagnostic efficacy between visual assessment and quantitative analysis and between ¹¹ C-PIB PET and ¹⁸ F-florbetapir PET were insignificant.

Conclusions: The overall performance of Aβ PET in diagnosing AD is favorable, but the differentiation between MCI and AD patients should consider that some MCI may be at risk of conversion to AD and may be misdiagnosed. A multimodal diagnostic approach and machine learning analysis may be effective in improving diagnostic accuracy.

Keywords: 11C-PIB; 18F-florbetapir; Alzheimer's disease; MCI converting to AD; amyloid-β PET.

FIGURE 1

Flowchart of the study screening process following the Preferred Reporting Items for Systematic Reviews and Meta‐analysis (PRISMA) 2020 statement

FIGURE 2

The typical brain images of Alzheimer's disease (AD) and normal controls are shown by amyloid‐β (Aβ) PET imaging. Part (a) shows axial ¹⁸F‐flutemetamol (¹⁸F‐FMM) PET images (upper row for AD patients, lower row for normal controls). Part (b) shows axial ¹¹C‐PIB PET images (upper row for AD patients, lower row for normal controls). Part (c) shows ¹⁸F‐AV45 PET images in transaxial, sagittal, and coronal positions (upper row for AD patients, lower row for normal controls). Part (d) shows axial ¹⁸F‐florbetaben PET images with aligned fused MRI images (upper row for AD patients, lower row for normal controls). Source: From Camus et al. (, Hatashita et al. (2014)), and Villemagne et al. (2011) with modifications

FIGURE 3

Overall assessment of the methodological quality of the 48 studies, including evaluation of risk bias and evaluation of applicability concerns

FIGURE 4

Forest plots of pooled sensitivity, specificity, and DOR for the included studies

FIGURE 5

Summary ROC curve for the overall performance assessment of amyloid‐β (Aβ) PET for the diagnosis of Alzheimer's disease (AD)

FIGURE 6

Funnel plot for assessing publication bias and heterogeneity