Summary of Studies on Test–Retest Reproducibility of Untreated Tumors Without Interval Therapy
| Study | Pts/lesions | No. and time between PET scans | Imaging and reconstruction parameters | Variables and ROIs | Major findings |
|---|---|---|---|---|---|
| Minn 1995 | 10 pts; 10 lesions; primary lung cancer ≥ 2 cm | 2 scans; mean 1.8 ± 1.8 d | PET alone/68Ge AC; dynamic acquisition × 60 min; 3.4-mm slice thickness (n = 8); 6.75-mm slice thickness (n = 2); 128 × 128 matrices; FBP 0.3 Hanning filter; ∼8 mm FWHM; axial resolution not given | Maximal SUL 1.2 × 1.2 cm; ROI 4 × 4 pixels (“peak”) | Test–retest mean percentage difference between scans/correlation (SUL: 10% ± 7%/0.987; Ki: 10% ± 8%/0.969; SUL glucose correction: 6% ± 6%/0.995; K1: 24% ± 15%/0.812; k2: 42% ± 31%/0.0.765; k3: 24% ± 13%/0.953) |
| Weber 1999 | 16 pts; 50 lesions; various cancers; tumor volume 0.8–111 mL | 2 scans; mean 3 ± 3 d | PET alone/68Ge AC; dynamic acquisition × 70 min; 3.4-mm slice thickness; 128 × 128 matrices (4 × 4 mm); FBP 0.4 Hanning filter; ∼8 mm FWHM; axial resolution ∼5 mm | SUV bw in 50% threshold around maximal 18F-FDG ROI (mean diameter 32 ± 36 mm, range 12–60 mm) | Mean percentage difference in SUV for test–retest is ∼10%; 0.9 SUV unit required for significant change; greater variability in smaller lesions; glucose correction, no significant differences |
| Nakamoto 2002 | 10 pts; lung cancer | 2 scans; within 1 wk | Reassessment of Minn data; same parameters for image acquisition and reconstruction | Maximal SUL in 1 × 1 pixel anywhere in tumor; highest average SUL in 4 × 4 pixels in tumor; effective glycolytic volume (SUL × volume) | Mean percentage difference between scans (maximal SUL: 11.3% ± 8%; mean SUL: 10.1% ± 8.2%; effective glycolytic volume: 10.1% ± 8%; mean percentage differences slightly reduced with glucose correction) |
| Krak 2005 | 11 pts; 29 lesions; NSCLC; median volume ∼9 cm3* | 2 scans; 2 consecutive days | PET alone/68Ge AC; dynamic acquisition × 60 min; 2.5-mm slice thickness; 128 × 128 matrices; FBP 0.5 Hanning filter; OSEM (2 iterations,16 subsets); ∼7 mm FWHM; axial resolution not given | FBP vs. OSEM; SUL ROIs (manual; 15 mm fixed; 50%, 75% threshold; single pixel maximum) | Test–retest reproducibility similar for FBP vs. OSEM; mean percentage differences of SUV between 2 scans (8%−10% ± 7%−8% for manual and 15-mm fixed ROI; 12%−14% ± 11%−13% for threshold methods; 13% ± 11%−12% for single-pixel SUVmax); mean percent differences of ROI volume (23% ± 20% for 50% threshold; 55% ± 35% for 75% threshold); ICC highest for 15-mm fixed ROI (0.95); ICC for threshold/single-pixel SUVmax 0.89–0.91 |
| Nahmias 2008 | 26 pts; 26 lesions; various cancers; tumor size not given | 2 scans; mean 3 ± 2 d | PET/CT (CT AC); static acquisition; 90 min after tracer injection; 2.5-mm slice thickness; 256 × 256 matrices; OSEM (4 iterations,16 subsets); ∼8 mm FWHM; axial resolution ∼8 mm | Manual ROI definition in axial slice with most 18F-FDG uptake; mean SUV 9- to 17-mm circular ROI (30% of maximum guide); single-pixel SUVmax in 2.5 × 2.5 × 2 mm ROI | Mean SUV (large manual ROI) test–retest: high correlation (r = 0.99, 95% confidence interval (CI) 0.99–1.00); mean difference 0.01 ± 0.27 SUV (95% CI ± 0.53 SUV); absolute difference mean SUV < 0.5 SUV; SUVmax test–retest mean difference −0.05 ± 1.14 SUV (95% CI ± 2.2 SUV) (larger absolute difference in SUVmax as SUVmax increased, frequently more than 1.5 SUV units with SUVmax over 7.5) |
↵* PET metabolic tumor volume; CT size not given.
Pts = patients; AC = attenuation correction; FBP = filtered backprojection; FWHM = full width at half maximum; bw = body weight; NSCLC = non–small cell lung cancer; OSEM = ordered-subset expectation maximization; ICC = intraclass correlation coefficient.