Predictors and Outcomes of Discordance Between Initial Clinical Diagnosis and Amyloid‐PET Results in the Imaging Dementia—Evidence for Amyloid Scanning (IDEAS) Study

Abstract

AbstractBackgroundDiscordance between the clinical diagnosis of AD by experts after a comprehensive evaluation and AD pathology on autopsy remains common. We sought to identify clinical situations where amyloid‐PET would be most valuable as a diagnostic tool by identifying predictors associated with discordance between initial clinical diagnosis and amyloid‐PET results.MethodIDEAS was a real‐world study that assessed the clinical utility of amyloid‐PET in 11,829 participants (age 76±6, 51% female, MMSE 24.5±5) across 595 US sites. Participants met amyloid‐PET Appropriate Use Criteria and completed pre/post‐PET visits where specialists documented clinical diagnosis. Diagnosis‐PET discordance was defined as either an initial AD diagnosis pre‐PET with a negative amyloid‐PET, or non‐AD diagnosis with positive amyloid‐PET. Univariate logistic regressions were performed to identify predictors and outcomes of discordance in the total sample and in subgroups stratified by pre‐PET diagnosis (AD/non‐AD).ResultDiagnosis‐PET discordance occurred in 40% of total participants, including 36%(n = 3323/9115) with an initial AD diagnosis and negative amyloid‐PET and 52%(n = 1409/2714) with an initial non‐AD diagnosis and positive amyloid‐PET. In the total sample (Figure 1), predictors associated with greater discordance included non‐White race (Black/African American: OR 1.26; 95%CI [1.04‐1.53]; Asian: (1.62[1.23‐2.13]) and more medical comorbidities (each comorbidity: 1.07[1.05‐1.07]). Predictors associated with less discordance included older age (5‐year increase: 0.91[0.89‐0.94]), female sex (0.91[0.85‐0.98]), dementia severity (vs MCI: 0.70[0.65‐0.75]), greater disease duration (≥3yr: 0.83[0.77‐0.90]), greater physician diagnostic confidence (0.89[0.86‐0.89]), AD medication use (0.64[0.59‐0.69]), and family history of dementia (0.89[0.81‐0.97]). Practice features associated with less discordance included using image quantification (0.90[0.81‐0.99]), using multimodal imaging (0.90[0.82‐0.97]), and greater physician time(≥50%) in dementia care (0.92[0.86‐0.99]). Subgroup analyses (Figure 2) identified additional predictors associated with greater discordance in those initially diagnosed with AD included Hispanic ethnicity (1.35[1.10‐1.66]), living alone (1.16[1.04‐1.30]), and non‐English primary language (1.22[1.01‐1.47]).ConclusionThe high frequency of discordance supports the clinical utility of AD biomarkers. Amyloid‐PET may be most valuable to support AD diagnosis in participants who are early in the disease course, younger, from underrepresented racial/ethnic groups, live alone, have multiple medical comorbidities, and lack a suggestive family history. The risk of discordance may be reduced by using image quantification, multimodal imaging, and highly specialized providers.