HAI Book 2025 - Flipbook - Page 73
Results: As of October 2024, we included 17 cohorts (n=30,409 participants) with heterogeneous pipelines,
radiotracers, and clinical and demographic characteristics (Table 1). In all cohorts, a two-Gaussian model best
fitted the distribution according to the integrated completed likelihood criteria. While the low Gaussian peaks
were near zero with mild variability across studies, the second peaks were more heterogeneous, with higher
values in cohorts that were more cognitively impaired (Figure-1A-B). The meta-analysis indicated a GMM-based
cutoff of 20 CL (Figure-1C) with no significant difference in cutoff by pipeline or center (p>0.1). When stratified by
tracer (Figure-1D), the meta-analysis showed significant (p
