The Simulation Analysis of SUV Normalization Methods in Clinical PET: A PRISMA-Compliant Evaluation with Bias Assessment

Simulation Analysis of SUV Normalization Methods in Clinical PET

Abstract

Background: Reliable standardized uptake value (SUV) normalization is essential for quantitative PET imaging and cross-center harmonization. Traditional body weight–based SUV (SUVbw) introduces variability, particularly in obese, pediatric, and cachectic patients. Alternative methods such as lean body mass–based SUV (SUVlbm) and liver-based SUV (SUL) may reduce bias, yet systematic comparisons across tracers, technologies, and populations remain limited.

Methods: A PRISMA 2020–compliant systematic review was conducted using PubMed, Scopus, and Web of Science (2000–2025). Inclusion required clinical PET studies reporting SUVs normalized by ≥2 methods. Risk of bias was assessed (ROBIS/QUADAS-2). Literature-derived mean ± SD values informed simulated datasets (n=500/method). Coefficient of variation (CV), intraclass correlation coefficient (ICC), Kolmogorov–Smirnov (KS) test, area under the curve (AUC), and Bland–Altman analyses were performed. Subgroups: tracer (FDG vs non-FDG), scanner (analog/digital/total-body), and patient cohorts (adult, pediatric, obese).

Results: From 17,432 records, 18 studies met eligibility. SUL showed the lowest variability (CV=9.3%) and highest reproducibility (ICC=0.94), outperforming SUVbw (CV=14.7%, ICC=0.87; p<0.01). SUVlbm reduced bias in obese and pediatric cohorts. ROC analysis confirmed superior classification for SUL (AUC=0.87) and SUVlbm (AUC=0.83) vs SUVbw (AUC=0.71). Subgroup analyses demonstrated SUL stability across tracers and scanners; SUVlbm benefitted body composition–sensitive groups.

Conclusion: SUL and SUVlbm provide more reliable, reproducible quantification than SUVbw. Adoption of these methods, supported by PRISMA methodology and bias assessment, enhances PET harmonization across populations and technologies.