TABLE 3.

Linear Mixed-Effects Models (Linear Regression for Liver)

Model	Fitted % difference in repeat scans	95% confidence interval
Model 1 (SUVmax)*
A. Same scanner (n = 10)	8%	6%–11%
B. Same institution,  different scanner (n = 2)	6%	3%–11%
C. Different institution  and scanner (n = 11)	18%	13%–24%
Model 2 (SUVmax)†
Same scanner or  institution	8%	6%–10%
Different institution and  scanner	18%	13%–24%
Model 3 (liver SUVmean)‡
Same scanner or  institution	5%	3%–10%
Different scanner and  institution	6%	3%–11%

^*C > A (P = 0.0015), C > B (P = 0.003), A and B not different on average (P = 0.66) (Tukey–Kramer adjustment for pairwise comparisons).

^†P < 0.001, Wald test.

^‡P = 0.85, Wald test (n = 23).

Group differences are back-transformed from log(absolute percentage difference + 1); n = 162 tumors in 23 patients.