Twenty-eight studies (1,527 participants, range 10 to 291) met the inclusion criteria; 26 studies evaluated PET, six evaluated PET/CT and 16 evaluated an alternative conventional imaging technology. Twelve studies recruited a representative patient spectrum. Twenty-two studies used an acceptable reference standard. Eleven studies avoided progression bias (delay less than one month) and 19 avoided partial verification bias. Nine studies reported blinding of interpreters of the index tests. Ten studies avoided clinical review bias. Only one study reported avoiding differential verification bias and none reported avoiding incorporation bias.
Compared with conventional imaging technologies (10 studies), PET had significantly higher sensitivity (89%, 95% CI 83% to 93% versus 79%, 95% CI 72% to 85%) and specificity (93%, 95% CI 83% to 97% versus 83%, 95% CI 67% to 92%). Indirect comparisons of conventional imaging technology (11 studies) and PET (25 studies) gave the same findings.
Compared with CT (four studies), PET/CT had significantly higher sensitivity (95%, 95% CI 88% to 98% versus 80%, 95% CI 65% to 90%), but not specificity. Indirect comparisons of conventional imaging technology (11 studies) and PET/CT (five studies) showed similar results. Direct comparisons between PET/CT and conventional imaging technologies where there was clearly less than one month between tests (three studies) showed no significance.
Compared with PET (four studies), PET/CT had significantly higher sensitivity (96%, 95% CI 90% to 98% versus 85%, 95% CI 77% to 91%), but not specificity. Results were similar for indirect comparison of PET/CT (five studies) and PET (25 studies).
There were no significant differences in the sensitivity or specificity of PET when compared with MRI. Results of subgroup and sensitivity analyses were reported. Results in the abstract were on a per patient basis; results on a per lesion basis were reported (nine studies).