Research ArticleClinical Investigations

Repeatability of Quantitative 18F-NaF PET: A Multicenter Study

Christie Lin, Tyler Bradshaw, Timothy Perk, Stephanie Harmon, Jens Eickhoff, Ngoneh Jallow, Peter L. Choyke, William L. Dahut, Steven Larson, John Laurence Humm, Scott Perlman, Andrea B. Apolo, Michael J. Morris, Glenn Liu and Robert Jeraj
Journal of Nuclear Medicine December 2016, 57 (12) 1872-1879; DOI: https://doi.org/10.2967/jnumed.116.177295

Article Figures & Data

Figures

  • FIGURE 1.
    FIGURE 1.

    Whole-body paired baseline 18F-NaF PET/CT scans of men with metastatic castration-resistant prostate cancer: a 74-y-old imaged 3 d apart at UWCCC (A), a 57-y-old imaged 2 d apart at MSKCC (B), and a 69-y-old imaged 1 d apart at NCI (C).

  • FIGURE 2.
    FIGURE 2.

    Box plots of relative differences in each SUV metric (log-transformed) for lesion-level ROIs (left; 411 lesions) and patient-level ROIs (right; 35 patients). Whiskers extend from minimum to maximum values.

  • FIGURE 3.
    FIGURE 3.

    Bland–Altman plots of SUV metrics for all lesion-level ROIs (411 lesions): SUVmax (A), SUVmean (B), and SUVtotal (C). Different sites are indicated by different symbols (▪ = UWCCC, ● = MSKCC, and ▲ = NCI). Solid line denotes mean difference, and dotted lines denote upper and lower 95% LOA. Both mean and difference uptake values have been log-transformed.

  • FIGURE 4.
    FIGURE 4.

    Bland–Altman plots of SUV metrics for all patient-level ROIs (35 patients): SUVmax (A), SUVmean (B), and SUVtotal (C). Different sites are indicated by different symbols (▪ = UWCCC, ● = MSKCC, and ▲ = NCI). Solid line denotes mean difference, and dotted lines denote upper and lower 95% LOA. Both mean and difference values have been log-transformed.

  • FIGURE 5.
    FIGURE 5.

    Overall ICC plotted against overall coefficient of variation of lesion-level (black) and patient-level (red) SUV metrics.

  • FIGURE 6.
    FIGURE 6.

    Bland–Altman plots of lesion SUVmax by site: UWCCC (265 lesions) (A), MSKCC (78 lesions) (B), and NCI (68 lesions) (C). Each point represents a lesion, and different subjects are indicated by different colors. Solid lines denote site-specific mean difference, and dotted lines denote site-specific upper and lower 95% LOA. Both mean and difference uptake values have been log-transformed.

Tables

  • TABLE 1

    Patient Demographics

    DemographicUWCCCMSKCCNCI
    Patients (n)18116
    Age (y)
     Median72.575.068
     Range47–8757–8157–83
    Height (cm)
     Median178177171
     Range166–191162–191161–189
    Weight (kg)
     Median92.394.084.6
     Range70.7–145.073.0–119.075.4–91.6
    PSA
     Median71.28.185.9
     Range1.6–310.02.5–246.832.0–460.7
    Gleason score (n)
     61 (6%)2 (18%)1 (17%)
     77 (39%)5 (45%)2 (33%)
     84 (22%)1 (9%)2 (33%)
     93 (17%)3 (27%)1 (17%)
    LDH (U/L)
     Median200219264
     Range139–470157–251119–903
    Hemoglobin (g/dL)
     Median12.813.811.8
     Range7.7–14.911.3–15.39.0–13.9
    Lesions (n)
     ≤56 (33%)5 (45%)2 (33%)
     6–100 (0%)4 (36%)1 (17%)
     11–2010 (56%)2 (18%)2 (33%)
     202 (11%)0 (0%)1 (17%)

    • PSA = prostate-specific antigen; LDH = lactic acid dehydrogenase.

  • TABLE 2

    Scanner Harmonization Parameters

    ParameterUWCCCMSKCCNCI
    ScannerDiscovery VCTDiscovery VCTGemini
    Reconstruction3D OSEM3D OSEM3D OSEM
    Grid size256 × 256256 × 256144 × 144
    Subset141433
    Iteration222
    Postprocessing filter4 mm4 mm—

    • 3D OSEM = 3-dimensional ordered-subsets expectation maximization.

  • TABLE 3

    Repeatability of Lesion 18F-NaF PET SUV Metrics

    MetricRCICC*CV (%)CPD (%)B†
    UWCCC (265 lesions)
     SUVmax0.230.980 (0.974, 0.984)11.737.51.00 (0.79, 1.25)
     SUVmean0.100.983 (0.979, 0.987)5.515.91.00 (0.90, 1.11)
     SUVtotal0.400.990 (0.987, 0.992)20.775.91.04 (0.69, 1.56)
    MSKCC (78 lesions)
     SUVmax0.310.958 (0.935, 0.973)16.854.31.04 (0.75, 1.45)
     SUVmean0.140.970 (0.953, 0.981)7.822.21.03 (0.88, 1.19)
     SUVtotal0.600.990 (0.985, 0.994)32.7133.61.08 (0.57, 2.06)
    NCI (68 lesions)
     SUVmax0.370.865 (0.791, 0.915)20.669.20.97 (0.65, 1.46)
     SUVmean0.160.876 (0.807, 0.922)9.226.20.98 (0.82, 1.17)
     SUVtotal0.650.993 (0.989, 0.996)36.6151.41.00 (0.49, 2.06)
    All sites (411 lesions)
     SUVmax0.270.969 (0.963, 0.975)14.147.21.00 (0.76, 1.32)
     SUVmean0.130.975 (0.970, 0.980)6.619.61.00 (0.88, 1.14)
     SUVtotal0.490.990 (0.988, 0.992)25.5100.41.04 (0.63, 1.71)

    • ↵* Data in parentheses are 95% confidence intervals.

    • ↵† Data in parentheses are 95% LOA.

    • RC = repeatability coefficient for α = 0.05 (log-transformed SUV); CV = log-transformed coefficient of variation; CPD = critical percentage difference; B = ratio of test-to-retest bias.

    • B and 95% LOA have been back-transformed to original units.

  • TABLE 4

    Repeatability of Patient 18F-NaF PET SUV Metrics

    MetricRCICC*CV (%)CPD (%)B†
    UWCCC (18 patients)
     SUVmax0.170.984 (0.959, 0.994)8.827.61.00 (0.84, 1.19)
     SUVmean0.080.990 (0.974, 0.996)4.212.31.01 (0.93, 1.09)
     SUVtotal0.200.993 (0.981, 0.999)10.132.21.05 (0.86, 1.28)
    MSKCC (11 patients‡)
     SUVmax0.300.965 (0.874, 0.990)15.553.80.96 (0.71, 1.32)
     SUVmean0.130.920 (0.731, 0.978)6.319.00.99 (0.87, 1.11)
     SUVtotal0.450.950 (0.825, 0.986)23.189.90.96 (0.61, 1.51)
    NCI (6 patients)
     SUVmax0.280.921 (0.548, 0.989)14.449.21.03 (0.77, 1.36)
     SUVmean0.130.826 (0.190, 0.974)6.720.20.97 (0.85, 1.11)
     SUVtotal0.540.985 (0.895, 0.999)27.6115.00.95 (0.55, 1.63)
    All sites (35 patients)
     SUVmax0.240.974 (0.949, 0.987)12.039.51.00 (0.79, 1.26)
     SUVmean0.100.981 (0.962, 0.990)5.316.00.99 (0.89, 1.10)
     SUVtotal0.360.989 (0.978, 0.994)18.567.11.00 (0.70, 1.44)

    • ↵* Data in parentheses are 95% confidence intervals.

    • ↵† Data in parentheses are 95% LOA.

    • ↵‡ Two patients underwent partial whole-body scans.

    • RC = repeatability coefficient for α = 0.05 (log-transformed SUV); CV = log-transformed coefficient of variation; CPD = critical percentage difference; B = ratio of test-to-retest bias.

    • B and 95% LOA have been back-transformed to original units.

Additional Files

Back to top

In this issue

Print
Download PDF
Article Alerts
Email Article
Citation Tools
Bookmark this article

Jump to section

Keywords