Abstract
BACKGROUND: [(18)F]FDG-PET brain scan remains widely used in the evaluation of cognitive decline worldwide, however data on its diagnostic performance against gold-standard CSF Alzheimer's disease (AD) biomarkers are scarce. We aimed to assess the agreement between [(18)F]FDG-PET findings and CSF AD biomarkers in a real-world tertiary memory clinic setting. METHODS: Cross-sectional study of Mass General Brigham patients with cognitive concerns and available [(18)F]FDG-PET imaging and CSF AD biomarkers between 01/01/2013 and 06/30/2025. [(18)F]FDG-PET brain scan findings were categorized as "Normal," "Abnormal Inconclusive," "Abnormal Not AD-like," or "Abnormal AD-like," based on the narrative report. The CSF AD biomarker panel was classified as "Not AD," "Equivocal," or "Consistent with AD" following the lab report. [(18)F]FDG-PET was compared with gold-standard CSF AD biomarkers using kappa agreement test and regression models. FINDINGS: Among 360 eligible individuals, 151 had a CSF profile "Consistent with AD," 136 "Equivocal," and 73 "Not Consistent with AD." The [(18)F]FDG-PET showed an AD-like pattern in 73/151 (48.3%) of subjects with CSF "Consistent with AD" and was normal in 30/73 (41.1%) of those with CSF "Not Consistent with AD." However, 19/151 (12.6%) of individuals with a CSF profile "Consistent with AD" had normal [(18)F]FDG-PET scans (false negatives) whereas 8/73 (11.0%) of those with a CSF profile "Not Consistent with AD" had an AD-like [(18)F]FDG-PET pattern (false positives), resulting in 0.48 sensitivity, 0.84 specificity, and 0.66 AUC of [(18)F]FDG-PET report vs. gold-standard CSF AD biomarkers, and a fair agreement between both tests (κ = 0.334). An AD-like [(18)F]FDG-PET pattern was strongly associated with a CSF "Consistent with AD" (OR = 4.81, p < 0.0001) and a lower Amyloid-Tau Index (ATI; β = -0.43, p < 0.0001). By region, posterior cingulate gyrus glucose hypometabolism predicted both an AD-like [(18)F]FDG-PET result (OR = 6.41, p < 0.0001) and a CSF profile "Consistent with AD" (OR = 2.48, p = 0.0003), whereas frontal hypometabolism predicted a Not AD-like [(18)F]FDG-PET result (OR = 5.90, p < 0.0001) but also lower odds of a CSF "Not Consistent with AD" (OR = 0.41, p = 0.0016). INTERPRETATION: [(18)F]FDG-PET imaging demonstrated high specificity but limited sensitivity to identify AD as defined by CSF biomarker criteria. Although a report of a typical AD-like [(18)F]FDG-PET pattern of glucose hypometabolism predicted a positive CSF AD biomarker panel, the agreement between [(18)F]FDG-PET report and CSF AD biomarker results was only fair. FUNDING: NR-L was supported by a Research Fellowship from the Fundación Ramón Areces, Madrid (Spain). JC, BCD, SEA, PK, and AS-P were supported by the Massachusetts Alzheimer's Disease Research Center (NIH/NIA P30AG062421).