One hundred and eight studies met the inclusion criteria. Of these, 104 reported on some aspect of diagnostic accuracy (n=5,935, range 15 to 517): 68 prospective studies and 36 retrospective. Unless specified, outcomes are for 18FDG PET. Of the 104 diagnostic accuracy studies, 43 recruited a representative patient spectrum, 100 avoided verification bias, 97 incorporation bias and 33 differential verification bias, and 86 reported some level of blinding (most commonly of the interpretation of the index test). Results for the impact on decision making and management were also reported.
Bladder cancer (two prospective and one retrospective studies):
For staging (two prospective, n=88), sensitivity was 53% (95% CI 27% to 79%) and 77% (95% CI 46% to 95%) and specificity 72% (95% CI 51% to 88%) and 94% (95% CI 81% to 99%). The pooled LR+ was 4.68 (95% CI 0.65 to 33.90) and LR- was 0.43 (95% CI 0.15 to 1.19); significant heterogeneity was observed.
Brain cancer (three prospective and two retrospective studies):
For staging (two prospective, n=42), sensitivity was 63% (95% CI 35% to 85%) and 63% (95% CI 38% to 84%). Specificity was reported in one study as 100% (95% CI 59% to 100%). Pooled likelihood ratios could not be calculated.
Cervical cancer (21 prospective and 12 retrospective studies):
For staging (five prospective, three retrospective, n=561), sensitivity ranged from 10% (95% CI 0% to 45%) to 100% (95% CI 77% to 100%) and specificity 76% (95% CI 53% to 92%) to 100% (95% CI 59% to 100%). From the prospective studies, pooled LR+ was 8.22 (95% CI 2.59 to 26.08) and LR- 0.38 (95% CI 0.12 to 1.20); significant heterogeneity was observed.
For detecting recurrence (three prospective and three retrospective, n=627), sensitivity ranged from 50% (95% CI 1% to 99%) to 100% (95% CI 16% to 100%) and specificity 76% (95% CI 70% to 82%) to 100% (95% CI 79% to 100%) depending on site of recurrence. From the prospective studies, pooled LR+ ranged from 15.24 (95% CI 5.63 to 41.27) to 45.89 (95% CI 14.09 to 149.49) and LR- from 0.09 (95% CI 0.02 to 0.40) to 0.37 (95% CI 0.22 to 0.60) depending on site of recurrence.
Kidney cancer (three prospective and five retrospective studies):
For staging (three retrospective, n=42), sensitivity ranged from 60% (95% CI 32% to 84%) to 90% (95% CI 55% to 100%) and specificity from 80% (95% CI 28% to 99%) to 100% (95% CI 3% to 100%). Pooled LR+ was 3.95 (95% CI 1.14 to 13.73) and LR- was 0.30 (95% CI 0.12 to 0.79); no significant heterogeneity was observed.
For primary diagnosis and staging (two retrospective, n=27), sensitivity was 60% (95% CI 32% to 84%) and 89% (95% CI 52% to 100%) and specificity was 100% (95% CI 3% to 100%). The pooled LR+ was 3.48 (95% CI 0.60 to 20.15) and LR- was 0.42 (95% CI 0.21 to 0.84); no significant heterogeneity was observed.
Ovarian cancer (14 prospective and six retrospective studies):
For detecting recurrences (six prospective and five retrospective), sensitivity for 18FDG PET/CT ranged from 73% (95% CI 39% to 94%) to 100% (95% CI 72% to 100%) and specificity 40% (95% CI 5% to 85%) to 100% (95% CI 29% to 100%). From four prospective studies using any reference standard, pooled LR+ was 6.97 (95% CI 1.94 to 25.0) and LR- 0.12 (95% CI 0.06 to 0.26).
Pancreatic cancer (14 prospective and three retrospective studies):
For primary diagnosis and staging (seven prospective, n=479), sensitivity ranged from 73% (95% CI 62% to 83%) to 97% (95% CI 91% to 99%) and specificity 41% (95% CI 18% to 67%) to 97% (95% CI 86% to 100%). Pooled LR+ was 2.77 (95% CI 1.62 to 4.73) and LR- 0.19 (95% CI 0.10 to 0.43).
Small cell lung cancer (six prospective and four retrospective studies):
For staging (three prospective and two retrospective, n=276), sensitivity was 100% (95% CI 48% to 100%). Specificity was reported in one study as 98% (95% CI 91% to 100%).
Testicular (three prospective and one retrospective studies) and prostate cancer (two prospective and two retrospective studies): No results were presented due to insufficient data. Results for other outcomes were reported, as were sROC curves.