Head-to-Head Comparison of Amyloid-PET and FDG-PET Accuracy in Predicting Alzheimer’s Disease Pathology


Orit Lesman Segev 1 Viktoriya Bourakova 1 Renaud La Joie 1 Nagehan Ayakta 1 Howard Rosen 1 Iryna Lobach 2 Sang Won Seo 1 James O’Neil 3 Mustafa Janabi 3 Rik Ossenkoppele 1 Manja Lehmann 1 Bruce Reed 1 John Olichney 4 Adam Boxer 1 Lee-Way Jin 1 Eric Huang 1 Marilu Gorno-Tempini 1 Charles DeCarli 3 Mackenzie Hepker 1 Bruce Miller 1 Lea Grinberg 1 William Seeley 1 William Jagust 3,4 Gil Rabinovici 1,3
1Memory and Aging Center, University of California San Francisco, USA
2Epidemiology & Biostatistics department, University of California San Francisco, USA
3Lawrence Berkeley National Laboratory, Life Sciences Division, USA
4Helen Wills Neuroscience Institute, University of California Berkeley, USA

BACKGROUND: Identifying the etiology underlying cognitive impairment and dementia during life is challenging given imperfect clinical-pathological correspondence. Diagnostic accuracy may be improved by introducing biomarkers, specifically imaging, to the diagnostic workup. Fluorodeoxyglucose-PET (FDG-PET) can help identify distinct topographic patterns of hypometabolism while amyloid-beta (Aβ) radiotracers, like Pittsburgh compound B (PIB), can detect Aβ plaques, a hallmark of Alzheimer’s disease (AD) pathology. In this work, we directly compared the diagnostic accuracy of PIB and FDG-PET in a large, clinical population with an autopsy-proven diagnosis. In addition, we assessed the added value of combining both modalities together.
METHODS: We included 101 individuals with available autopsy data (32 with pure AD pathology, 13- mixed AD and non-AD, and 56 non-AD pathology only, see Table 1). Pre-mortem FDG and PIB scans were visually read by three raters blinded to clinical information. PIB and FDG scans were read separately. PIB scans were rated as positive or negative for cortical retention while FDG scans were read as showing an AD or non-AD hypometabolism pattern (Figure1). Majority (2/3 raters in agreement) or consensus (full-agreement between raters) reads were compared to the pathological diagnosis of AD (i.e. intermediate or high levels of AD neuropathological changes).
RESULTS: PIB showed higher sensitivity in detecting AD pathology (96% vs 80% for FDG, p=0.02) and higher negative predictive value (96% vs 84%, p=0.01), but equivalent specificity (86% vs 84%, p=0.8) (Table2). PIB’s advantage was mainly in early disease stages (clinical dementia rating scale (CDR)≤0.5) and in young patients (<67 years). In ApoE4 gene carriers, however, PIB’s specificity decreased to 70% [CI:42-95%]. 84/101 cases had consensus PIB reads, while 72/101 cases had consensus FDG reads. Inter-rater agreement was higher for PIB (Fleiss’ Kappa=0.77) than FDG (Kappa=0.61), P=0.028. Sensitivity to detect AD pathology was 97% (CI:98-100%) and specificity 98% (CI:93-100%) for individuals that had congruent PIB and FDG (77/101 patients, example cases 1&2 in Figure1). When PIB and FDG were incongruent (n=24, example case 3 in Figure1): PIB was correct in 8 cases, FDG was correct in 7 cases and both were correct in 9 cases in which there was co-occurrence of AD and non-AD pathologies.
CONCLUSIONS: In our cohort of 101 patients with an autopsy proven diagnosis, PIB-PET had higher sensitivity in predicting AD pathology compared to FDG-PET, with similar specificity. When PIB and FDG are congruent, sensitivity and specificity approaches 100%. Mixed pathology should be considered when PIB and FDG are incongruent.

orit Lesman Segev
orit Lesman Segev