Variability of Average SUVs Among Institutions for Particular Categories of 18F-FDG PET Studies
| Category | n | Average SUV | Average log10SUV ± SD |
|---|---|---|---|
| Non-Hodgkin’s lymphoma | 21 | 8.0 | 0.81 ± 0.29 |
| 22 | 9.2 | 0.89 ± 0.29 | |
| 22 | 12.5 | 1.02 ± 0.27 | |
| Breast cancer | 41 | 3.5 | 0.49 ± 0.20 |
| 24 | 4.5 | 0.57 ± 0.27 | |
| 36 | 5.1 | 0.63 ± 0.23 | |
| 26 | 12.8 | 1.02 ± 0.29 | |
| Pancreatic cancer | 42 | 3.2 | 0.45 ± 0.24 |
| 34 | 4.4 | 0.60 ± 0.18 | |
| 23 | 6.5 | 0.77 ± 0.18 | |
| Head and neck squamous cell cancer | 48 | 3.2 | 0.49 ± 0.11 |
| 22 | 6.3 | 0.74 ± 0.24 | |
| 37 | 9.4 | 0.94 ± 0.19 | |
| Normal liver | 82 | 1.7 | 0.22 ± 0.13 |
| 24 | 2.5 | 0.40 ± 0.07 | |
| 37 | 2.7 | 0.43 ± 0.09 |
In meta-analysis (22) within each category, using individual patient log10SUV values in Kruskal–Wallis ANOVA, the P value for at least 2 institutions differing in means is always found to be <0.0001. Reasons can be a combination of factors in Table 1 along with variations in populations and pathologies chosen for study. The higher SD values of the logarithms (which are approximately [1/ln 10] × coefficients of variation of SUVs here) for cancers compared with those of normal liver presumably show variability stemming from extents of disease.