Seventeen studies met the inclusion criteria (40 datasets; N=2,940; range 29 to 1,000 patients). Seven studies evaluated FDG-PET–CT, seven evaluated FDG-PET, four evaluated MRI and 13 evaluated bone scintigraphy. Five studies were prospective, eight retrospective and the direction of data collection was unclear in four. Patient recruitment was consecutive in eight studies and not documented in nine. All studies scored over 8 out of 14 on the QUADAS tool and were considered moderate quality.
On a per-patient basis, pooled sensitivities for the detection of bone metastasis in lung cancer were 92% (95% CI, 88 to 95; seven studies) for FDG-PET–CT , 0.87% (95% CI, 81 to 92; five studies) for FDG-PET, 77% (95% CI, 65 to 87; three studies) for MRI and 86% (95% CI, 82 to 89; 12 studies) for bone scintigraphy. Pooled specificities were 98% (95% CI, 97 to 98) for FDG-PET–CT, 94% (95% CI, 92 to 96) for FDG-PET, 92% (95% CI, 88 to 95) for MRI and 88% (95% CI, 86 to 89) for bone scintigraphy.
Pooled DORs were 449.17 for FDG-PET–CT, 118.25 for FDG-PET, 38.27 for MRI and 63.37 for bone scintigraphy; the DOR for FDG PET-CT was statistically significantly higher than the other modalities.
Results were also presented for each diagnostic modality on a per-lesion basis. There was no evidence of publication bias. Results of meta-regression that explored the impact of study design, year of publication, sample size, histopathology analysis, prevalence of bone metastasis and study quality were also reported. The presence of differential verification bias resulted in significantly lower sensitivity and higher specificity.