Amyloid PET Meta-analysis Identifies Centiloid Thresholds and an Intermediate Caution Range

Amyloid PET is increasingly used to assess the etiology of cognitive decline and eligibility for amyloid-targeting therapies, but interpretation of standardized Centiloid measurements varies by method and threshold. This JAMA individual participant data meta-analysis pooled cross-sectional scans using five radiotracers from 49,227 participants in 53 studies across 15 countries. Mean age was 71 years, 54% were female, and 62% were cognitively impaired. Investigators applied a unified statistical pipeline and random-effects meta-analysis to estimate positivity thresholds and agreement with visual reads.

Gaussian mixture modeling identified bimodal distributions in 51 studies involving 48,786 participants. The pooled single positivity cutoff was 18 Centiloids (95% CI, 16-19). A double-cutoff approach indicated high-confidence negative classification below 11 Centiloids and positive classification above 26. Among 35,045 participants in 36 studies with visual reads, the cutoff maximizing agreement was 27 Centiloids (95% CI, 24-30), with Cohen κ of 0.86 (95% CI, 0.83-0.89).

These findings support context-dependent interpretation of amyloid PET quantitation, particularly within the 11-26 Centiloid range. Substantial between-study heterogeneity accompanied the estimates, with I² values of 80%-97%. The cross-sectional analysis assessed statistical classification and visual-read agreement, not treatment outcomes; the abstract alone does not establish a universal threshold for therapy eligibility.